Evolution of wind-pollination in Fraxinus (Oleaceae) – an ecophylogenetic approach Eva Wallander

Botanical Institute Göteborg University Sweden, 2001

Göteborg University Faculty of Science 2001

Dissertation

Evolution of wind-pollination in Fraxinus (Oleaceae) – an ecophylogenetic approach

Eva Wallander

Systematic Botany Botanical Institute Göteborg University Box 461 SE 405 30 Göteborg SWEDEN

Akademisk avhandling för filosofie doktorsexamen i systematisk botanik (examinator: professor Lennart Andersson) som enligt beslut av naturvetenskapliga fakultetsnämnden kommer att försvaras offentligt fredagen den 7 december 2001 kl. 10.00 i föreläsningssalen, Botaniska institutionen, Carl Skottsbergs gata 22 B, Göteborg. Fakultetsopponent: W. Scott Armbruster, Norges TekniskNaturvitenskapelige Universitet, Trondheim.

Göteborg, November 2001. ISBN 91-88896-37-4 1

2

Dissertation for the Degree of Doctor of Philosophy in Systematic Botany, defended in public at the Botanical Institute, Göteborg University, Sweden, 7 December 2001. Wallander, E. 2001. Evolution of wind-pollination in Fraxinus (Oleaceae) – an ecophylogenetic approach. PhD thesis, Göteborg University, Sweden. ISBN 91-88896-37-4.

ABSTRACT Studies in Fraxinus (ash) and other genera of Oleaceae have revealed interesting trends and patterns in the evolutionary history of reproductive traits. They contribute empirical data for hypotheses under which conditions anemophily (wind-pollination) and dimorphic breeding systems, such as dioecy and the rare androdioecy, may evolve. The study includes a molecular phylogeny of all 25 genera of the family Oleaceae, based on two noncoding chloroplast loci (the intron of rps16 and the trnL-F region), and of 40 of the 43 recognised species of the genus Fraxinus, based on the internal transcribed spacers of the nuclear ribosomal DNA (ITS-1 and ITS-2, and the intervening 5.8S gene). The reproductive ecology of three species of Fraxinus, F. excelsior, F. ornus, and F. longicuspis, representing different combinations of pollination and breeding systems, were studied in the field in order to gain insight in their adaptations to wind and/or insect-pollination and compare different reproductive strategies. Based on estimates of phylogenetic relationships in combination with reproductive data, I have shown that anemophily has three separate origins in Fraxinus and that it is preceded by ambophily (both wind- and insect-pollination). Dioecy is correlated with anemophily and has evolved after the origin of anemophily in three instances. In one case, dioecy has originated via androdioecy, which in turn has a hermaphroditic origin. This is the first study that shows the evolution of dioecy from hermaphrodites via the rare androdioecious pathway. In the other windpollinated taxa, dioecy has evolved twice, independently, from polygamous ancestors. In the subtribe Oleinae, the sister group of Fraxinus, there are five genera with species that are adapted to wind-pollination. Here, anemophily is believed to have four separate origins. Many species have inconspicuous, whitish to yellowish-green flowers, which appear to attract unspecialised pollinators by offering pollen as a reward. This is believed to be an exaptation (a trait that facilitates or is a prerequisite) for ambophily, which in turn may have selected for anemophily. Many of the taxa of Oleinae are hermaphrodites, but dioecy and androdioecy are common breeding systems, especially among the anemophilous or ambophilous species. The interpretations of the results are that ambophily and a breeding system with unisexual males, either androdioecy or polygamy, were some of the exaptations for the evolution of anemophily. Dioecy, on the other hand, has followed in evolutionary sequence and is interpreted as an adaptation that evolved in response to anemophily and is adaptive for increasing the efficiency and function of that system. In wind-pollinated flowers there are different structural and productivity demands for optimal male and female function that are hard to compromise between in one flower. Thus, the selection for sexual specialisation leads to dioecy through total or partial reduction of the opposite sex in the resulting unisexual flowers. Key words: androdioecy, ambophily, anemophily, breeding system evolution, dioecy, floral evolution, Fraxinus, Oleaceae, pollination system evolution.

Eva Wallander, Botanical Institute, Göteborg University, Box 461, SE 405 30 Göteborg, SWEDEN. [email protected] Botanical Institute, Göteborg 2001. ISBN 91-88896-37-4 Printed in Sweden by Kompendiet, Göteborg 2001. 3

Svensk sammanfattning Avhandling for filosofie doktorsexamen i systematisk botanik, offentligt försvarad på Botaniska institutionen, Göteborgs universitet, 7 december 2001. Wallander, E. 2001. Evolution of wind-pollination in Fraxinus (Oleaceae) – an ecophylogenetic approach. Doktorsavhandling, Göteborgs universitet, ISBN 91-88896-37-4. Denna avhandling handlar främst om hur vindpollination har utvecklats inom ask-släktet (Fraxinus) i olivfamiljen (Oleaceae), men också hos några andra släkten i familjen. Först har jag tagit fram ett släktskapsträd över 40 av de 43 ask-arterna i världen samt ett över alla 24 nutida släkten (och ett utdött!) i olivfamiljen, baserat på DNA-sekvenser. Med hjälp av dessa har jag sedan spårat utvecklingen av de olika karaktärer som hör ihop med vindpollination. Jag har främst varit intresserad av att ta reda på vilka karaktärer som föregick uppkomsten av vindpollination, sk ”exaptationer”, och vilka som uppkom efteråt, dvs anpassningar (adaptationer) till detta pollinationssätt. Jag har också specialstuderat tre arter i fält för att se hur deras anpassningar till vindpollination eventuellt skiljer sig från varandra. Resultaten visar att vindpollination har uppkommit inte mindre än sju separata gånger i familjen, varav tre gånger i ask-släktet. Vid två av de tre tillfällena inom Fraxinus har det skett en utveckling från tvåkönade till enkönade blommor, antingen via androdioeci (han- och hermafrodit-blommor på olika individ i en population) eller via polygami (mångbyggare, dvs både enkönade och/eller tvåkönade blommor finns på samma eller olika individ i populationen). Flera karaktärer verkar ha varit exaptationer, dvs förutsättningar för eller underlätta för uppkomsten av vindpollination, t ex förekomst av många och små oansenliga blommor som besöks av pollensamlande insekter. Man kan skönja en trend i blomutvecklingen inom Oleaceae, från skyltande, väldoftande och nektarbelönande blommor (t ex jasmin, syren, liguster och Forsythia), till dessa flertaliga, små, öppna och kopiöst pollenproducerande blommor. I lämpliga miljöer sprider många av dessa en hel del av sitt pollen också med hjälp av vinden och är således både vind- och insektspollinerade. Ur dessa arter har sedan ”ren” vindpollination uppkommit genom ytterligare anpassningar som underlättar vindspridning av pollen. Sådana anpassningar är bl a reducering av krona och doft samt enkönade blommor. Förutom att kronbladen inte längre behövs för att locka till sig insekter så kan den tvärtom vara i vägen för både effektiv pollenspridning och infångande av pollen. Även ståndare och pistill kan vara i vägen för varandra och minska effektiviteten för de två könsfunktionerna. Det är också konkurrens om resurserna till pollenproduktion respektive fruktproduktion. Dessa motsättningar mellan han- och honkönet hos vindpollinerade blommor medför ett starkt selektionstryck för separation av könen och tolkas som en starkt bidragande orsak till varför många vindpollinerade växter har enkönade blommor, antingen på samma individ (sambyggare) eller på olika individ (tvåbyggare). Detta är en av de slutsatser som jag har dragit från denna studie. Jag har också visat att förekomsten av det mycket ovanliga reproduktionssystemet androdioeci, som är vanligt hos de arter som är både vind- och insektspollinerade i Oleaceae, kan förklaras med hjälp av denna modell. Nyckelord: androdioeci, dioeci, evolution, Fraxinus, Oleaceae, vindpollination

Eva Wallander, Botaniska institutionen, Göteborgs universitet, Box 461, 405 30 Göteborg. [email protected] Botaniska institutionen, Göteborg 2001. ISBN 91-88896-37-4 Tryckt av Kompendiet, Göteborg 2001.

4

Papers This thesis is based on the following four papers, which are referred to by their Roman numerals: Paper I: Wallander, Eva and Åslög Dahl. Studies on the reproductive ecology of F. excelsior, F. ornus and F. longicuspis, and a comparison with some other species of Fraxinus (Oleaceae). (Manuscript) Paper II: Wallander, Eva. Evolution of pollination and breeding systems in Fraxinus (Oleaceae), and a new classification. (Manuscript) Paper III: Wallander, Eva and Victor A. Albert. 2000. Phylogeny and classification of Oleaceae based on rps16 and trnL-F sequence data. American Journal of Botany 87(12): 1827–1841. Paper IV: Wallander, Eva. Evolution of wind-pollination and gender specialisation in Oleaceae – exaptations and adaptations. (Manuscript)

Paper I is co-authored by Åslög Dahl and based on initial ideas developed by her. I am responsible for carrying out the fieldwork, most of the analyses, and writing the ms. Paper III is co-authored by Victor A. Albert, then head of the Cullman Program for Molecular Systematics at the New York Botanical Garden where the molecular work was conducted. I am responsible for all molecular work, most of the cladistic analyses, and writing the ms.

5

6

INTRODUCTION This thesis is mainly about the evolution of pollination and breeding systems and related reproductive traits in the genus Fraxinus, the ashes, but also in a number of other related genera in the family Oleaceae. Hypotheses on the evolutionary scenario is based on phylogenetic estimates based on molecular data, coupled with detailed field studies of three species of Fraxinus, and literature and herbarium studies of the other taxa. First, I give a background to wind-pollinated plants and their characteristics, i.e. their ‘syndrome’. Then I explain the ‘ecophylogenetic’ method of reconstructing the evolutionary history of traits. I go on presenting the main study plants, the genus Fraxinus, and the characteristics that make them an ideal model taxon for this kind of study. I also briefly summarise characteristics of other genera of the family Oleaceae and explain why some of them were included in the work as well. Finally, I state the aims of this work and the questions I wished to address. The materials and methods section explains what I have done, where and how. For a more detailed exposition of methodology and taxon sampling, see Papers I-III. Authors of plant names are not given in the thesis, but are to be found in the appended papers. The results and discussion are combined into one section and is divided into three parts: reproductive ecology of Fraxinus, phylogenies and systematics, and evolution of wind-pollination. In the latter part I discuss possible exaptations and adaptations for wind-pollination, occurrences of different breeding systems in Fraxinus and Oleaceae and general trends in the evolution of reproductive systems in Oleaceae. I also summarise hypotheses on the evolution of dioecy in wind-pollinated plants and suggest why these may also explain the existence of androdioecy in Oleaceae and other windpollinated plants.

Wind-pollination – a background Plant reproductive ecology is a vital and highly productive field within modern biology (Lovett Doust and Lovett Doust 1988, Richards 1997, Geber et al. 1999). Most studies, however, have focused on herbs that are pollinated by animals. As to long-lived, woody species, of which the

majority are wind-pollinated at temperate latitudes, most of our knowledge emanates from cultivars of species of economical interest. This fact is notable, since about 20% of Scandinavian flowering plants species are wind-pollinated. Many of these are important constituents in all our terrestrial ecosystems, not least with regard to biomass. Biotic pollination (zoophily) is ancestral in angiosperms. Although other seed plant families are wind-pollinated, e.g. conifers, anemophily (wind-pollination) is a derived condition within the angiosperms and has evolved independently in several families (Whitehead 1968, Cox 1991, Linder 1998). Anemophily has rarely been experimentally verified in the species where it is assumed to occur (e.g. as by Goodwillie 1999, de Figueiredo and Sazima 2000). Instead anemophily has been inferred on the basis of a number of traits considered to be typical of anemophilous taxa, the so-called anemophilous syndrome (Whitehead 1968 and 1983, Faegri and van der Pijl 1979). The most obvious of those traits are flowers without (or with a very reduced) corolla, nectar or scent. The flowers are often dichogamous and/or unisexual and borne either on the same (monoecy) or different (dioecy) individuals. The anthers are usually large and well exposed on long slender filaments or in hanging catkins that can easily be moved by wind. The stigmata and/ or surrounding structures often have a morphology favouring efficient pollen trapping (Niklas 1985). Many are trees, but there are also shrubs and non-woody plants, e.g. the whole of Poaceae, that are wind-pollinated. In temperate areas, anemophilous trees usually flower earlier than other plants, when they and/or the surrounding plants are leafless, i.e., many are deciduous and occur in deciduous forests or open environments. Anemophily is especially common among temperate trees and there is a striking increase with latitude and decrease with species diversity in percentages of anemophilous trees (Regal 1982). Anemophily is rare in habitats like tropical rainforests (Bawa et al. 1985, Renner and Feil 1993, Linskens 1996, but see Williams and Adam 1999). High pollen/ovule (P/O) ratios are also a feature of wind-pollinated plants (Cruden 1977, Tormo et al. 1996). A number of plants assumed to be entomophilous (insect-pollinated) may in fact be ambophilous, i.e., pollinated by both insects 7

and wind. Some examples are Piper (de Figueiredo and Sazima 2000), Calluna (cited by Faegri and van der Pijl 1979), Salix (Vroege and Stelleman 1990, Tollsten and Knudsen 1992), Erica arborea (Aronne and Wilcock 1994), Luzula, Acer, Tilia, and other taxa reviewed by Cox (1991). Ambophilous species are generally characterised by numerous, simple flowers with small and open corollas (Linder 1998). Ambophily is common among neotropical dioecious trees (Bullock 1994). All the above traits typical of anemophiles have not evolved at the same time. There has probably been a gradual evolution of adaptations to wind-pollination from biotic pollination, where ambophily may have been an intermediate stage. Therefore, one may ask which characters generally precede the transition to anemophily, which characters are coincidental, and which evolve later as a consequence of that system (Linder 1998). The first type of trait has been termed exaptation by Gould and Vrba (1982) and is a character that facilitates or is a prerequisite for a transition. Traits that follow in evolutionary sequence have evolved in response to anemophily and are adaptive for increasing the efficiency and function of that system. Such traits are adaptations in the strict sense of Gould and Vrba (1982). Particularly, in this regard, it is unresolved whether unisexual flowers, which are strongly correlated with anemophily, evolve prior to or after the shift to anemophily (Charlesworth 1993). In a number of entirely or partly anemophilous genera there is a variation in sexuality. For example, monoecy as well as subdioecy occurs in Myrica (Lloyd 1981), and in Acer (de Jong 1976) and Fraxinus there are species with hermaphrodite flowers, as well as polygamous and dioecious ones. Although the pollination mode should strongly influence the way different types of flowers are distributed in time and space, the evolution of breeding systems have rarely been examined from the pollination point of view (Bawa 1980, Bawa and Beach 1981). The reasons for the correlations between pollination and breeding systems, especially dioecy, are still unclear in many cases. In general, there are few empirical studies in this area and, in particular, none of the pathways, other than the one via gynodioecy, is well understood (Webb 1999). A study of adaptations to anemophily may comprise several aspects of reproduction, apart 8

from the morphological characteristics that belong to the ”syndrome of anemophily”, and which are predictable on the basis of aerodynamic principles. Several features of the reproductive ecology are likely to be the consequence of other conditions coupled to this pollination method, e.g., strong dependence on weather conditions (suitable for pollen dispersal); an often extreme cross-pollen limitation resulting in a small chance of multiple cross-pollination and for sexual selection; and the apparent risk of self-pollination when the tree is surrounded by a cloud of its own pollen. The degree of synchrony in pollen release among individuals, among sexual phases within individuals, and among different ramets of individuals, may be even more important for the possibilities for multiple cross-pollination and self-pollination/ self interference, respectively, in anemophilic plants than in entomophilous ones. Studies of stigmatic receptivity schedules, the phenology of megagametophyte development and fertilization may also be informative.

Ecophylogenetics Studies of character evolution have frequently relied on correlation rather than on phylogeny and have thus been lacking a historical perspective. However, correlations do not estimate the number of times that a trait evolved, and they are insensitive to the direction or temporal sequence of character transformation (Donoghue 1989). In contrast, cladograms can provide this information. Phylogenies are fundamental to comparative biology and help to identify independent evolutionary events (Felsenstein 1985). Correlation between traits do not necessarily imply causation and careful use of phylogenetic information can help to distinguish between cause and effect (Harvey and Pagel 1991). They may also help to identify the context in which a feature evolved and to determine whether the trait is an adaptation or an exaptation for the function. Determining the order of evolutionary events may give insights in, for example, what may be the major selective force promoting the evolution of unisexual flowers. Armbruster (1992) reviewed the field of studying plant-animal interactions and used the term “ecophylogenetic” for those recent studies that combined phylogenetic information with ecological data to infer the evolutionary history of ecological relationships. He

pointed out that the stronger the selective association between two characters, the less likely we are to detect the order of their appearance in a phylogeny. Furthermore, if there is no cladogenesis (tree branching) between the origins of two traits, or if species in between go extinct, the two characters will appear simultaneously on the tree. The basis for any ecophylogenetic study is an estimate of the phylogeny of the group of organisms (Armbruster 1992). This estimate should ideally be complete for the group in question and corroborated by independent data, i.e., the topology needs to be robust and reliable. Of course, one also needs to know the states of all character to be mapped, preferably for all included taxa.

The study plants

5 mm

The olive family, Oleaceae comprises 24 extant genera and one extinct (Table 1). Well-known genera include the lilacs (Syringa), privets (Ligustrum), Forsythia, and jasmines (Jasminum). Many species of these genera are cultivated as ornamentals because their flowers are showy and fragrant. They are typical examples of insect-pollinated plants and the white or yellow, sometimes pink or purple, flowers usually offer nectar as a pollinator reward. However, the focus of this study was the genera with inconspicuous, often wind-pollinated flowers. The subtribe Oleinae comprises 12 very closely related genera, all characterised by drupes, and it is sister group to subtribe Fraxininae, consisting solely of the genus Fraxinus (Paper III). This clade of two subtribes contains numerous taxa with adaptations to wind-pollination and the main aim of this study was to find out how

they have become adapted to this type of pollination system and how many times it has happened. The genus Fraxinus, the ashes, comprises 43 species distributed in temperate to subtropical areas of the northern hemisphere (Table 2 and Paper II). The species are mainly trees, most of which are deciduous, but there are also some shrubby xerophytes (plants adapted to arid environments). About two thirds of the species of the genus are wind-pollinated and one third are insect-pollinated, and they exhibit a variety of breeding systems (Table 4 in Paper II). Nineteen species occur in the New World (from Canada south to Guatemala) and 24 in the Old World (Europe, N Africa, and Asia). There are three European species: F. excelsior, F. angustifolia and F. ornus. Most of the insect-pollinated species occur in eastern Asia, but the wind-pollinated species are more evenly distributed in temperate forests and arid regions of the northern hemisphere. Three species, representing different pollination and breeding systems, were selected for detailed and comparative studies. F. excelsior, the common or European ash, occurs over central and northern Europe and is the only native species of ash in Sweden. It is wind-pollinated and polygamous, i.e., male, female and hermaphrodite flowers occur on the same or different individuals in a population. The trees are large, up to 30 m, but a more normal height is around 20 m. Flowers are without both petals (apetalous) and calyx (asepalous). Male flowers consist solely of two stamens, hermaphrodite (bisexual) ones of two stamens and one pistil, and female flowers of only one pistil, with or without rudimentary stamens (Figure 1).

Figure 1. The five floral types of Fraxinus excelsior: (a) male, (b) male with rudimentary pistil, (c) hermaphrodite, (d) female with rudimentary stamens, and (e) female flower. Reproduced with permission from Binggeli and Power (1991).

9

Nyctanthes L.

Dimetra Kerr

Fontanesia Labill.

Abeliophyllum Nakai

Forsythia Vahl

Menodora Humb. & Bonpl. 24

Jasminum L.

Ligustrum L.

Syringa L.

Myxopyreae

Myxopyreae

Fontanesieae

Forsythieae

Forsythieae

Jasmineae

Jasmineae

Oleeae: Ligustrinae

Oleeae: Ligustrinae

Insect

Long-tubed & sweet-scented or shorttubed & inodorous, yellow or white

Fraxinus L.

Chionanthus L.

Forestiera Poir.

Haenianthus Griseb.

Hesperelaea A. Gray

Nestegis Rafin.

Noronhia Stadm. ex Thou.

Notelaea Vent.

Olea L.

Osmanthus Lour.

Phillyrea L.

Picconia DC.

Priogymnanthus P.S. Green 2

Oleeae: Fraxininae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Total number of species: 600+

1(-2)

2

30

40+

12

41

5

White to purple, nectariferous

Insect

S America (Bolivia, Brazil, Paraguay, Ecuador)

Macaronesia

Mediterranean region to W Asia

Subtropical parts of E Asia and N America (1 sp.)

Tropical and subtropical parts of the Old World

Australia and Tasmania

Madagascar

New Zealand and Hawaii (1 sp.)

Insect

Insect

Wind

Insect

Insect

Petals very early caducous

Small whitish corolla

Small whitish corolla

Small, white or yellowish corolla, extrafloral nectaries

Wind?

Insect

Wind

Insect

Small, whitish , lobes shorter than corolla Insect & tube, stamens exserted in some spp. wind

Small corolla

Small, fleshy & globular, nectar

Apetalous

Yellow corolla, four stamens

Small, white, deeply 4-lobed corolla

Subtropical N America, C America, and West Indies Apetalous West Indies

Wind

Insect

Small, white or yellow, corolla lobes divided to the base or with short tube

Tropical and subtropical parts of Africa, America, Asia, and Australia

Insect Insect & Wind

White or pinkish corolla Small, with 4 (or 2) free or fused white petals, some scented, or without petals

Insect

Tropical parts of Africa and India

White or pinkish corolla

Mainly temperate and subtropical regions of the Northern Hemisphere

1 (extinct) Mexico (was endemic to Guadalupe Island)

3

ca 15

ca 100

4 43

Oleeae: Schreberinae Schrebera Roxb.

Madagascar and the Comores

Mainly subtropical parts of Eurasia

3

20

Oleeae: Schreberinae Comoranthus Knobl.

Insect

Insect

Insect

Yellow, strongly scented, stamens not exserted

White or yellow, scented, nectariferous

Insect

Insect

Insect

Insect

Insect

Hermaphroditic

Hermaphroditic

Androdioecious

Hermaphroditic, androdioecious

Hermaphroditic, a few dioecious or andromonoecious

Hermaphroditic

Hermaphroditic

Dioecious

Hermaphroditic

Hermaphroditic

Dioecious, a few hermaphroditic

Hermaphroditic, androdioecious or dioecious

Hermaphroditic, androdioecious, polygamous, or dioecious

Hermaphroditic, heterostylous

Hermaphroditic

Hermaphroditic

Hermaphroditic

Hermaphroditic, some heterostylous

Hermaphroditic

Hermaphroditic

Hermaphroditic

Hermaphroditic

Hermaphroditic

Hermaphroditic

Hermaphroditic

Pollination Breeding system(s) system(s)

White or pinkish corolla

Small, deeply lobed corolla

Tubular, white with yellow mouth

Tubular with sessile anthers, white or orange, night-scented

Small yellow or pink corolla

Floral morphology

Temperate to tropical parts of the Old World, except Usually white, nectariferous Africa

Tropical and subtropical parts of the Old World

Subtropical N and S America and S Africa

E Asia and SE Europe (one sp.)

Korea

SW Asia (and Sicily) and China

Thailand

Tropical and subtropical SE Asia

Tropical SE Asia

Distribution

45

200+

11

1

1-2

1

2

4

Myxopyrum Blume

Myxopyreae

No. of species

Genus

Tribe: subtribe

Table 1. The 25 genera of Oleaceae, with tribal and subtribal assignments according to Paper III, the approximate number of species and their world distribution, short floral descriptions, and occurrences of different pollination and breeding systems.

Table 2. The 43 recognised species of Fraxinus and their classification according to Paper II, with information on distribution, pollination and breeding system. Taxa

Pollination system

Breeding system

Fraxinus L. sect. Dipetalae (Lingelsh.) E. Nikolaev F. anomala Torr. ex S. Wats. W USA F. dipetala Hook. et Arn. W USA F. quadrangulata Michx. C & E USA, C Canada

Distribution

Wind Insect (and wind?) Wind

Hermaphroditism Hermaphroditism Hermaphroditism

Fraxinus L. sect. Ornus (Boehm.) DC. F. apertisquamifera Hara F. bungeana DC. F. floribunda Wall. F. griffithii C. B. Clarke F. lanuginosa Koidz. F. malacophylla Hemsl. F. ornus L. F. paxiana Lingelsh. F. raibocarpa Regel F. sieboldiana Blume F. trifoliolata W. W. Smith F. baroniana Diels F. chinensis Roxb. F. japonica Blume ex K. Koch F. longicuspis Sieb. et Zucc. F. micrantha Lingelsh.

Japan China Himalayas, E Asia SE Asia Japan China, Thailand C & E Mediterranean Himalayas, China C Asia China, Japan, Korea China China E Asia Japan Japan Himalayas

Insect (and wind?) Insect (and wind?) Insect (and wind?) Insect (and wind?) Insect and wind Insect (and wind?) Insect and wind Insect (and wind?) Insect (and wind?) Insect (and wind?) Insect (and wind?) Wind Wind Wind Wind Wind

Androdioecy Androdioecy Androdioecy Hermaphroditism Androdioecy Hermaphroditism Androdioecy Androdioecy Hermaphroditism Androdioecy Androdioecy Dioecy Dioecy Dioecy Androdioecy Androdioecy

Fraxinus L. sect. Fraxinus subsect. Fraxinus F. angustifolia Vahl ssp. angustifolia F. angustifolia Vahl ssp. oxycarpa (Willd.) Franco & Rocha Afonso F. angustifolia Vahl ssp. syriaca (Boiss.) Yalt. F. excelsior L. F. mandshurica Rupr. F. nigra Marsh. F. platypoda Oliv.

SW Europe SE Europe

Wind Wind

Polygamy Polygamy

Middle East N. & C. Europe China, Japan, Korea, E Russia E USA, E Canada China

Wind Wind Wind Wind Wind

Polygamy Polygamy Dioecy Polygamodioecy Dioecy

subsect. Melioides (Endl.) Lingelsh. F. americana L. F. caroliniana Mill. F. latifolia Benth. F. papillosa Lingelsh. F. pennsylvanica Marsh. F. profunda (Bush) Bush F. texensis (Gray) Sarg. F. uhdei (Wenzig) Lingelsh. F. velutina Torr.

E USA & E Canada SE USA N & C USA, Mexico N & C USA, Mexico C & E USA, Canada E USA USA (Texas) C America SW USA, N Mexico

Wind Wind Wind Wind Wind Wind Wind Wind Wind

Dioecy Dioecy Dioecy Dioecy Dioecy Dioecy Dioecy Dioecy Dioecy

subsect. Pauciflorae Lingelsh. F. dubia (Willd. ex Schult. & Schult. f.) P. S. Green & M. Nee F. gooddingii Little F. greggii A. Gray F. purpusii Brandegee F. rufescens Lingelsh.

Mexico, Guatemala

Wind

Polygamy

SW USA, N Mexico SW USA, Mexico Mexico, Guatemala Mexico

Wind Wind Wind Wind

Polygamy Polygamy Polygamy Polygamy

subsect. Sciadanthus (Coss. et Dur.) Lingelsh. F. xanthoxyloides (G. Don) DC. F. hubeiensis S. Z. Qu, C. B. Shang & P. L. Su

N Africa to China China

Wind Wind

Polygamy Polygamy

Incertae sedis F. cuspidata Torr. F. chiisanensis Nakai F. spaethiana Lingelsh.

SW USA, Mexico Korea Japan

Insect (and wind?) Wind Wind

Hermaphroditism Polygamy Polygamy

Figure 2. Fraxinus ornus, Sicily 1996. (top) Inflorescences and (below) hermaphrodite and male flowers.

F. ornus (manna ash) occurs in the Mediterranean area and eastwards to Iraq. It has showy, white-flowered and fragrant panicles that attract numerous insects. The populations consist of separate male and hermaphrodite individuals, an extremely rare breeding system that is termed androdioecy. However, at the time of planning for the study the species was consistently described as hermaphroditic in the literature. Hermaphrodite flowers consist of a small calyx, four free petals, two stamens and one pistil. Male flowers resemble hermaphrodite ones except that they have a rudimentary pistil. The flowers are borne in large terminal inflorescences that emerge together with the leaves (Figure 2). F. longicuspis (Japanese ash) is endemic to Japan and occurs in deciduous forests at 1001100 m altitude in the central and southern parts (Honshu, Shikoku, and Kyushu Islands). It is a large deciduous canopy tree, up to 25 m tall, growing at low density among other broadleaved trees along rivers and on slopes. The species is androdioecious with male and hermaphrodite flowers on separate trees. The flowers emerge in many-flowered panicles together 12

Figure 3. Fraxinus longicuspis, Japan 1999. (top) Male inflorescences with unopened anthers, (left) male inflorescences with empty anthers as the leaves begin to unfold, and (right) hermaphrodite inflorescences at the same time. Compared to F. ornus (Fig. 2), flowering takes place before leafing in this wind-pollinated species.

with the leaves from terminal buds in early spring (Figure 3). In contrast to F. ornus, the leaves are still undeveloped while flowering takes place. The wind-pollinated flowers have a small cup-shaped calyx but no corolla, the hermaphrodite flowers with a pistil and two stamens and the male flowers with two stamens only. In the subtribe Oleinae, five of the twelve genera contain species that appear to be windpollinated, Phillyrea, Forestiera, Priogymnanthus, Nestegis and Olea (Table 1). The N and C American genus Forestiera consists of ten apetalous species, most of which are dioecious. It is related to the S American genus Priogymnanthus, whose two species have hermaphrodite flowers with ephemeral petals, and which flower when the trees are leafless during the dry season. Phillyrea contains two androdioecious species in the Mediterranean and west Asia. They are wind-pollinated shrubs with small whitish

flowers. Nestegis contains four apetalous species in New Zealand. They are either large canopy trees or smaller trees in windswept coastal areas, and are dioecious through reduction of the opposite sex organs in the unisexual flowers. The genus Olea consists mainly of tropical species, but many of the African species are mostly wind-pollinated. Olea europaea and its many cultivars are the most common wind-pollinated species of that group in Europe. Although it has small and dull whitish flowers that may be visited by unspecialised insects, the anthers contain large amounts of pollen that is mainly distributed by the wind. In fact, pollen from Olea europaea has been identified as one of the most important cause of seasonal respiratory allergy in the Mediterranean area (Liccardi et al. 1996). Many of the other seven genera of subtribe Oleinae have numerous small, whitish, open flowers that appear to be an exaptation for the evolution of anemophily or ambophily in this group. During the work it soon became clear that hypotheses on the evolution of anemophily in Fraxinus could not be seen in isolation from other similar events in the family. Although anemophily appeared to have developed independently once or more times in Fraxinus, the same evolutionary path might have been taken on several other independent occasions in the sister group. For that reason I expanded my work to include the Oleinae as well (Paper IV), although it has not been studied in such detail as Fraxinus.

Aims The main aim of this study was to develop a hypothesis for the course of evolution of windpollination in the genus Fraxinus. In order to achieve this I integrated data from comparisons of reproductive strategies among anemophilous and entomophilous species with estimates of phylogeny. The aims of this study were… • to estimate a phylogeny of the genus Fraxinus and its sister genera based on DNAsequences, • to study and compare the phenology of pollination, gender specialisation, floral morphology, and other adaptations to wind-pollination in selected species of Fraxinus that appeared to have evolved these adaptations separately, and

• to combine these data and use them as a basis for hypotheses on the occurrence and nature of shifts in reproductive ecology during the course of differentiation of the genus Fraxinus and its sister group. Based on results from the ecophylogenetic study, I wished to address questions such as: • Is entomophily ancestral in Oleaceae and in Fraxinus? • Have adaptations to wind-pollination originated only once in the family or have there been multiple, independent, transitions from entomophily to anemophily? • Are there even multiple origins within the genus Fraxinus? • If there are multiple origins, do the flowers in the different taxa have the same morphological adaptations? • Is ambophily an intermediate step, i.e., an exaptation? • Are there any other characters that preceded the evolution of anemophily, i.e. exaptations that made it easier for multiple transitions to occur? • Which traits originated after the transition, i.e. can be interpreted as adaptations for the function of anemophily? • Are there any patterns in the distribution of unisexual flowers?

MATERIALS AND METHODS Field studies Field studies were conducted on three species of Fraxinus in three countries: Fraxinus excelsior in Sweden, F. ornus on Sicily (Italy) and F. longicuspis in Japan (Paper I). The study of reproductive ecology of F. excelsior was conducted during seven years (1995-2001) in one population located in Göteborg, on the Swedish west coast. In the population, 30 trees were individually marked and six branches per tree were tagged and studied during the flowering season, from mid April to mid May, each year. Data on start and end of male and female anthesis and leafing phenology were recorded each year, as were estimates of flowering intensity and fruit set of individual trees. The phenotypic sex expression at the floral, inflorescence and tree level, as well as sex ratio in the population, were also recorded each year. Data on flowering intensity in the popu13

lation were compared to the total sum of airborne Fraxinus pollen in the Göteborg region in order to see if flowering was synchronised over a larger area. I determined the receptivity of stigmata at the onset of flowering through a hydrogen peroxide test in vitro, and verified through pollen tube growth tests in vivo. I counted the number of flowers per inflorescence in four types of inflorescences (male, hermaphrodite, female, and mixed) and number of pollen grains per anther in a few male and hermaphrodite flowers. The sizes of pollen grains were also compared among males and hermaphrodites, as well as the germination capacity in vitro. Self-compatibility was determined through a simple experiment where a few hermaphrodite inflorescences were bagged to exclude cross-pollen. The development of ovules, growth of pollen tubes, and fertilization event was studied in hermaphrodite flowers preserved in FPA on different days during the anthesis, with the aid of a Confocal Laser Scanning Microscope (CLSM), in order to determine the timing of reproductive events. The pollination ecology of F. ornus, at that time thought to be insect-pollinated and hermaphroditic, was studied in two populations in the Madonie mountains on Sicily. During the first two weeks of May 1996, I recorded the flowering phenology and sex expression for 51 trees (31 males and 20 hermaphrodites) and the sex ratio in three populations. I collected a few insect visitors on the inflorescences and estimated the extent to which pollen from the trees was dispersed into the air. Size of pollen grains and the number of pollen grains per anther were estimated in four to five flowers each from two male and two hermaphrodite trees. Stigmatic receptivity was tested with hydrogen peroxide on a few newly opened flowers from one hermaphrodite tree in Göteborg Botanical Garden (because it was too late in the season when I arrived to the study site on Sicily). A preliminary study on the reproductive ecology of F. longicuspis was conducted in collaboration with Dr. Tsutom Hiura (Tomakomai Experimental Forest, Hokkaido) and Dr. Kiyoshi Ishida (Forestry and Forest Products Research Institute, Kyoto) in Japan, May 1999. In a small population located in the Tokyo University Forest at Chichibu, consisting of 14 flowering trees, we studied the flowering phenology and sex expression and compared the number of flowers per inflorescence, stamen 14

size, pollen grain size, and pollen germination rate in vitro, among males and hermaphrodites. No pollination experiments could be conducted because flowering had already started on arrival to the study population.

Herbarium material and literature data For the molecular work and for studies on the morphology of Fraxinus and other genera of the family, loans of plant specimens were obtained from the following herbaria: BM, C, E, GB, K, MO, NY, S, and UPS (abbreviations from Index Herbariorum). In addition, I received a large number of samples of leaves and/or flowers from various people and institutions. Since only three species of Fraxinus were studied in the field, data on the reproductive biology of all other species of Fraxinus, plus many other wind-pollinated species in Oleaceae, were derived from literature or inferred from herbarium studies. In addition, a number of species were studied in botanical gardens and arboreta (Göteborg Botanical Garden and Arboretum, New York Botanical Garden, Missouri Botanical Garden and Shaw Arboretum, Palermo Botanical Garden, Kyoto Botanical Garden, and the Royal Botanic Gardens at Kew). In taxa where wind-pollination has not been experimentally verified by field studies, anemophily was inferred on the basis of possession of apetalous flowers in combination with a suitable habitat, such as open coastal habitats (Nestegis), seasonally deciduous and/ or open habitats (the two species of Priogymnanthus, some species of Forestiera), or temperate forests (many species of Fraxinus), and/or evidence for airborne pollen (e.g. Olea europaea, Phillyrea, and species of Fraxinus). Data on the breeding system in section Ornus, many of which seemed to be androdioecious, were not always readily gained from the literature. Therefore, the phenotypic sex expression was recorded on 179 flowering herbarium specimens of Fraxinus belonging to 13 of the 16 recognised species in section Ornus (Paper I). In order to estimate the male function in hermaphrodite flowers, I compared stamen size relative to other floral parts in the two gender morphs of each species, under the assumption that size and form of staminate parts reflect one aspect of the functionality. On 71 of these specimens, measurements of anther, fila-

ment, pistil (only for hermaphrodites), and petal length (only for petalous taxa) were taken on ten flowers per specimen and averaged separately for the gender morphs of each species. Stamen length was a posteriori calculated as anther plus filament length. For the phylogeny of Fraxinus, a thorough inventory of most of the taxa accepted in various recent treatments and their synonyms, were undertaken in order to come up with a reasonable representation. This inventory was based on original descriptions, floristic treatments, regional monographs, and monographs of the entire genus (see references in Paper II). The work resulted in a provisional list of about 50 taxa, most of which I eventually accepted (Table 2 and Paper II).

Molecular work In order to estimate the phylogenies of the genus Fraxinus and the whole family Oleaceae, I used molecular data. A major part of the molecular work was carried out at the Cullman Lab of Molecular Systematics, New York Botanical Garden. The rest of the work was completed at the Botanical Institute at Göteborg University. DNA was extracted from fresh, silica-gel dried or herbarium specimens using a modified CTAB buffer. PCR amplification and sequencing was performed as described in Papers II and III. Based on the list of accepted Fraxinus taxa, at least two representatives of each species, and in some cases subspecies, were chosen for DNA sequencing. In addition, some taxa of uncertain standing (synonyms) were chosen to evaluate their relationships. I sequenced the ITS region of the nuclear ribosomal DNA (the internal transcribed spacer 1, the 5.8S gene, and the internal transcribed spacer 2), approximately 700 base pairs. In addition, I used the two chloroplast sequences of the 10 Fraxinus species from the family phylogeny (Paper III), added five new ones (to represent all sections) and combined them with 15 ITS sequences from the same taxa. Vouchers for included taxa are listed in Paper II (Table 2). Outgroup taxa were chosen from the sister groups, subtribes Oleinae and Ligustrinae, as identified by the Oleaceae phylogeny (Paper III). For the phylogeny of Oleaceae I sequenced two noncoding chloroplast loci, the intron of rps16 and the trnL-F region (consisting of the trnL intron and the contiguous intergenic

spacer), about 800-900 base pairs each. At least two representatives from each genus in the family were sequenced, including Nyctanthes and Dimetra (previously Verbenaceae). Where possible I tried to use material from the type-bearing species of the genus. In the monotypic genera Abeliophyllum, Dimetra, and Picconia two different individuals of each species were sequenced. The genus Hesperelaea is also monotypic, but because it is extinct and known solely from the type collection, only one sample from this could be used. Vouchers for all sequenced taxa are listed in Paper III (Table 3) along with their GenBank accession numbers. As outgroup taxa I chose species of Verbenaceae and Myoporaceae (Lamiales) to test the position of Dimetra and Nyctanthes, and members of Rubiaceae, Loganiaceae, Strychnaceae, and Gelsemiaceae (Gentianales) were included to provide a root hypothesis for Oleaceae. The forward and reverse sequences were checked and edited using the Sequencher™ software version 3.1 (Gene Codes Corporation, Ann Arbor, Michigan, USA). Alignment was done using the assembly feature in Sequencher, and then manually adjusted using criteria described by Andersson and Rova (1999). Consensus sequences from each of the two chloroplast loci and the ITS region were aligned separately. Twenty-two indels in the rps16 matrix and 20 in the trnL-F were considered informative, and indel characters were added to the combined matrix (using A/T for present/ absent). A few insertions, which did not contain informative characters, were then deleted. Autapomorphic insertions were also removed.

Cladistic and statistical analyses The combined chloroplast data set consisted of 78 ingroup Oleaceae taxa and ten outgroup taxa and the ITS data set of 101 ingroup Fraxinus taxa and five outgroup taxa. Cladistic analyses were performed using PAUP* 4.0b8 (Swofford 2000). Heuristic searches were conducted with TBR branch-swapping in a number of random addition sequence replicates (10-100). The optimality criterion was set to maximum parsimony, all characters were given equal weight (=1), and gaps were treated as missing data. In the ITS data set multistates were interpreted as polymorphisms but in the combined chloroplast data set as uncertainties. Parsimony jackknifing was performed using XAC (James S. Farris, Swedish Museum of Natural History, 15

Stockholm), doing 1000 replicates, each with ten random addition sequence replicates and nonrotational branch-swapping. In Paper I, containing data on the reproduction of a number of species of Fraxinus, statistical analyses were performed using StatView® version 5.0.1 and SuperANOVA version 1.11 (Abacus Concepts, Inc., Berkeley, CA). Data that were not normally distributed and/or with non-homogenous variances were log-transformed before using parametric tests. Percentage data were arcsine square root transformed before testing. In some cases, the unpaired ttest (two-tailed) was used to test the null hypothesis of no difference in group mean between the two sex morphs. A nested ANOVA model (type III sums of square) was used to test the distribution of variance at different levels (among sex morphs, among trees within sex morphs, and among inflorescences within trees).

RESULTS AND DISCUSSION Reproductive ecology of Fraxinus I have studied the reproductive ecology of three species of Fraxinus (Paper I). Fraxinus excelsior, wind-pollinated and polygamous, was studied in detail in the field over seven years. Comparative studies were made of F. ornus, insect-pollinated and androdioecious, and F. longicuspis, wind-pollinated and androdioecious for one flowering season each. The results from the 7-year study on F. excelsior showed that sexuality in ash is quite complicated, with a continuum of phenotypic gender expressions. There are five types of flowers (Figure 1) that can be combined in various ways into four main types of inflorescences: male, female, hermaphrodite, and mixed male and hermaphrodite. One or more of these inflorescence types can occur on one tree, but it is mostly possible to classify a tree into one of three main categories of phenotypic gender expression, male, hermaphrodite, or female. These three sex morphs occur in approximately equal proportions in the population. The phenotypic gender expression of a tree can also vary between years, but it usually remains within either the male or female side of the continuum. Male and mixed inflorescences contain significantly more flowers than hermaphrodites, which in turn contain significantly more 16

flowers than female ones. Male anthers contain more pollen than hermaphrodite ones and male pollen also has higher germination rates. Most female flowers have rudimentary stamens, some of which dehisce, but contain pollen that is largely inviable. There are no differences in flowering phenology between the sex morphs. Hermaphrodite flowers are protogynous for about a week and the overlap between male and female phases is also about a week. At the time of pollination there are no mature ovules in the ovary and the megagametophytes do not develop until the receptive period of the stigmata is over. This makes sexual selection through pollen tube competition possible and if multiple cross-pollination has taken place, pollen with the fastest growing pollen tubes may be the ones that achieve fertilisation. Hermaphrodite flowers are self-compatible but it is believed that self-pollen is not successful in competition with cross-pollen. Likewise, pollen from other hermaphrodites may lose in competition with male pollen (because of the higher germination rate of the latter). This fact is important for the maintenance of entirely male individuals in the populations (developed further below, under hypotheses for evolution of dioecy). The field study on F. ornus confirmed that it is androdioecious (as independently found by Dommée et al. 1999) and I also showed that it is ambophilous. Both male and hermaphrodite flowers produce copious amounts of pollen and they are visited by many unspecialised pollenfeeding insects, mainly small beetles, but a lot of pollen is also spread by the wind. However, relative rates of successful wind- or insect-pollination are unknown. The pollen:ovule ratio is more than twice that of F. excelsior, and this high pollen production is thought to be significant for the ambophilous strategy and ultimately for the maintenance of androdioecy (explained further below). The male flowers have a pistil rudiment but there are no significant differences between males and hermaphrodites in any other traits. Currently there is no evidence for a higher male fitness as compared to that of the male function in hermaphrodites. The study of F. longicuspis in Japan was the first documentation of some parts of the reproductive biology of this species, and also the first for any species in the wind-pollinated group of section Ornus. It was shown to be functionally androdioecious, with males and hermaph-

rodites in equal proportions in the small population. The male flowers have no pistil rudiments. Apart from 2.2 times higher number of flowers in male inflorescences compared to hermaphrodites, and about 40% larger male pollen grains, there are no differences in any other variables measured. Pending pollination experiments to evaluate potential differences in siring success between males and hermaphrodites, I suggested that males may have the necessary doubled male pollen fertility compared to hermaphrodites (as required by ESS models [Charlesworth 1984], see Paper I, II, and IV), in order for them to be maintained in the population.

Phylogenies and systematics A reliable estimate of the phylogeny of the group of organisms in interest is the basis for any hypothesis concerning its evolutionary history. This thesis includes a phylogenetic estimate of the family Oleaceae (paper III) and in more detail of the genus Fraxinus (paper II). In addition, a phylogenetic estimate of the subtribe Oleinae (Wallander, Green, and Harris unpublished data), which is the sister group of Fraxinus and contains a number of other windpollinated genera, is used together with the Oleaceae phylogeny as a basis for hypotheses in Paper IV. The phylogeny of Oleaceae (Paper III) includes representatives of all 25 genera, including one extinct, and is based on two noncoding chloroplast regions, the intron of rps16 and the trnL-F region. Both regions gave congruent gene phylogenies and the combined phylogenetic estimate was evaluated against nonmolecular data (morphology, embryology, karyology, wood anatomy). The result was also corroborated through the distribution of iridoid glucosides in the family (Jensen, Franzyk and Wallander, in press). Based on the phylogeny, a revised tribal classification of the family was given (Table 1), consisting of five tribes and four subtribes. Because the resolution among the genera of subtribe Oleinae (where the five wind-pollinated genera belong) was very poor using the noncoding chloroplast DNA data, I sequenced the more variable nuclear ribosomal ITS region. This locus gave many more informative sites and the ITS phylogeny was nearly completely resolved (unpublished data). The taxon sample included several representatives from 10 of

the 12 genera and with a concentration in the anemophilous and dimorphic genera. The complete, but less resolved, chloroplast phylogeny was used together with the more resolved ITS phylogeny to infer the evolutionary history of pollination and breeding systems in Oleaceae (Paper IV). The phylogeny of the genus Fraxinus was estimated based on ITS data for 40 of the 43 recognised species in the genus, and in combination with chloroplast data and morphology gave a solid ground for evolutionary interpretations (paper II). The paper also includes a revised infrageneric classification (Table 2), consisting of three sections and four subsections.

Evolution of wind-pollination – exaptations and adaptations Insect-pollination is ancestral in the family and all taxa in genera outside subtribes Fraxininae and Oleinae have hermaphrodite flowers (Table 1 and Paper IV). These are usually quite showy and fragrant and many provide nectar as a pollinator reward. Examples are the flowers of Forsythia, Jasminum, Syringa and Ligustrum. Six of the 24 extant genera contain anemophilous species (Table 1) and adaptations to anemophily have evolved seven times, independently, in the family. The genus Fraxinus has been studied in detail and there are three independent origins of anemophily (Paper II), in each of the sections Dipetalae, Ornus and Fraxinus, which are preceded by ambophily. Dioecy is correlated with anemophily in Fraxinus and has evolved after the origin of anemophily in three instances. In section Ornus, dioecy has originated via androdioecy. The androdioecious and ambophilous species have a hermaphrodite origin and this is the first study that shows the evolution of dioecy from hermaphrodites via the rare androdioecious pathway. In the wind-pollinated section Fraxinus, dioecy has evolved twice, independently, from polygamous ancestors. The other five genera with wind-pollinated species belong to subtribe Oleinae and show four independent origins of adaptations to anemophily, in Forestiera/Priogymnanthus, Phillyrea, Nestegis and Olea. (Paper IV). In the Oleinae, dioecy has evolved independently in the two wind-pollinated genera Forestiera and Nestegis, and in Chionanthus virginicus and Osmanthus 17

section Leiolea (Paper IV). The pollination system of the latter two taxa is unknown. The rare androdioecious breeding system occurs in four clades in the Oleinae: Phillyrea, one species of Chionanthus, Osmanthus section Notosmanthus, and the three remaining sections of Osmanthus (the genus Osmanthus is polyphyletic as presently circumscribed [Wallander, Green, and Harris, unpublished data]). The prevalence of anemophily in Oleaceae and its correlation with different breeding systems is remarkable. However, since it has happened so many times it is possible to develop a hypothesis to explain why anemophily has evolved repeatedly in this family and what might be exaptations and adaptations for this abiotic pollination system. What are the common features of these taxa? To begin with, all species in subtribe Oleinae are characterised by drupes. All of them have four ovules but only one develops into a seed (the drupes are very rarely two-seeded). The drupes are more or less fleshy and mostly birddispersed. The samaras of Fraxinus are oneseeded and wind-dispersed. These types of fruit have bearing on the hypotheses for evolution of dioecy, as will be explained further on. Second, the flowers are generally small (smaller than the ones in the other tribes of the family), whitish to yellowish-green and appear to be visited by unspecialised insects. Many show evidence of being ambophilous, e.g. Fraxinus section Ornus and African species of Olea (Dyer 1991) and, indeed, Olea europaea is the main source of airborne Oleaceae pollen in the Mediterranean area (Liccardi et al. 1996). There is a trend from sympetalous and longtubed flowers to short-tubed and more open flowers or flowers with free or almost free petals. Many of the anemophilous taxa have no corollas at all, e.g. in Forestiera, Fraxinus, and Nestegis, and in some species of Fraxinus and Forestiera not even calyces. This floral pre-condition in Oleinae, in combination with habitats suitable for wind-pollination (or unsuitable for insect-pollination), may have selected for ambophily and anemophily. Thus, the small whitish flowers were an exaptation for anemophily, and the trend in reduction of corolla size is viewed as an adaptation for enhanced efficiency of wind-pollination. These results and interpretations are consistent with patterns found among other angiosperm families (Linder 1998). 18

Part of the exaptations is also the shift from nectar reward to pollen reward. This shift was probably selected for in habitats where generalist pollinators were more common or reliable than specialist pollinators. The pollenrewarding flowers were then an exaptation for ambophily, which in turn was an exaptation for anemophily. In Fraxinus, at least in section Ornus, there is a trend in increase in tree size that preceded the origin of anemophily. This, in combination with other traits, may also be interpreted as an exaptation for anemophily. Breeding systems are clearly correlated with pollination system in Oleaceae. All entomophilous taxa are hermaphrodites and dimorphic breeding systems with unisexual males (either androdioecy, dioecy, or polygamy) have evolved in most anemophilous taxa. Some ambophilous or anemophilous species of Olea, which morphologically appear to be hermaphrodites, may in fact be functionally andromonoecious (Dyer 1991). Dioecy has evolved after the origin of anemophily in three instances in Fraxinus and it is also present in the four New Zealand species of Nestegis and in eight of ten species of Forestiera. In the latter two genera it has not been possible to discern whether dioecy evolved after anemophily or not, since the two traits appear simultaneously on the tree. However, their sister taxa are hermaphroditic or androdioecious. In the case of Forestiera, there are two species that are hermaphroditic but they appear to be derived from dioecy (Wallander, Green, and Harris, unpublished data). Dioecy also occurs in Chionanthus virginicus and in six of eight species of Osmanthus section Leiolea, but it is not known how they are pollinated. Androdioecy occurs in ten ambophilous species of Fraxinus, two anemophilous species of Phillyrea, in two separate clades of Osmanthus and in one species of Chionanthus. The pollination system of these two genera is not known. Thus, dioecy and androdioecy are clearly correlated with some degree of anemophily in a number of independent instances and they are interpreted as adaptations to pollination by wind. Hypotheses on the evolution of dioecy and androdioecy are reviewed in Paper IV and below I summarise the hypotheses on dioecy and suggest why these may also explain the occurrence of androdioecy in Oleaceae, in particular, and other wind-pollinated plants in general.

Dioecy – why is it so common among wind-pollinated plants? Dioecy occurs in about 6 % of flowering plants and has been correlated with a number of characters such as fleshy fruits, woodiness, climbing habit, small and white to yellowish or greenish flowers, unspecialised insect pollinators, wind-pollination, etc (reviewed by Renner and Ricklefs 1995). Although the great majority of dioecious taxa are entomophilous, in temperate regions many dioecious species are anemophilous (Bawa 1980). The occurrence of dioecy in plants has generated much debate and a number of hypotheses have been put forward to explain the evolution of dioecy and the above associations (see references in Paper IV). The models for the evolution of dioecy can be divided into two groups, based on whether they invoke a genetic or ecological mechanism as favouring unisexual individuals over hermaphrodites (Givnish 1982). The ecological mechanisms invoked are quite diverse and include, for example, sexual selection, division of labour leading to optimal resource allocation, decreased intraspecific competition, pollinator attraction to massive pollen crops and frugivore attraction to massive fruit crops. Ever since Darwin’s (1877) comprehensive work, the existence of a variety of breeding systems in plants have traditionally been viewed mainly as consequences of selection for outbreeding (see discussion in Bawa 1980, Givnish 1982). Organisms with separate sexes do not face the problem of selfing and inbreeding depression (at least not in large populations). In cosexual plants, on the other hand, which may suffer from inbreeding depression after selfing, there is a substantial selective pressure to promote outcrossing. Bawa (1980) and Bawa and Beach (1981) have emphasised that viewing plant breeding systems as the result of regulation of genetic recombination is unlikely to account fully for the evolution of breeding systems. They argue that the key to understand them lies in considering patterns of sexuality as means of optimising male and female reproductive success in different ways within the constraints imposed by the pollination system. In my review of the different hypotheses that are based on ecological mechanisms (Paper IV), I began by asking why plants should be hermaphrodites. It is commonly the case that we

try to find explanations for the rarest conditions, thinking that the most common does not need an explanation. However, what is common in this case is relative. Most sexually reproducing organisms on Earth are unisexual and hermaphroditism is common only in angiosperms, where it is, in fact, the predominant condition (Richards 1997). Why is this? The obvious explanation is their dependence on external pollinators visiting their flowers and transferring their male gametes to other conspecific individuals. This fact imposes a strong selective pressure for combined sexes in zoophilous plants (sometimes viewed as a constraint), partly because costly floral displays and pollinator rewards can be shared between the two sexes (Lloyd 1982) and partly because one pollinator can then perform two services in one visit, both delivering and picking up pollen (Richards 1997). Another advantage is that a cosexual and self-compatible plant, growing at low population density, such as after a colonising event, or where pollinators are scarce or absent, may self-pollinate (Lloyd 1982). Although there are advantages of having cosexual flowers, these may also turn into disadvantages, such as selfing is if it confers inbreeding depression. Therefore, while maintaining the advantageous cosexual flowers in zoophilous species, selection has favoured other mechanisms that promote outcrossing, e.g. dichogamy, herkogamy, and pre- or post-zygotic self-incompatibility mechanisms. So, what happens if a plant is no longer dependent on animal pollen vectors? The abiotic vectors, wind and water, are ubiquitous (in most habitats) and there is no longer any selection pressure to maximise both sexual functions for one visit. Neither do they have costly floral displays that it would be advantageous to share. On the contrary, the spatial interference between the male and female structures in a wind-pollinated flower may be directly disadvantageous and lead to conflicts between optimal male and female functions (Faegri and van der Pijl 1979, Lloyd 1982). There are different structural requirements for optimal pollen dispersal and pollen capture in wind-pollinated flowers (Frankel and Galun 1977, Niklas 1985, Freeman et al. 1997). Furthermore, because wind-pollination is a rather inefficient (in the sense that it is wasteful or imprecise) means for plants to spread their male gametes, there is a strong selection pressure for maximising pol19

len production, which increases the likelihood for pollen to reach conspecific stigmas (Frankel and Galun 1977). Consequently, those plants that divert more resources to pollen production, either through more pollen per anther, more anthers per flower and/or more flowers per individual, may have higher paternal success. In fact, in several wind-pollinated species of Fraxinus it has been shown that male inflorescences have significantly more flowers than hermaphrodite and female ones (Paper I). Increased allocation to male function is in conflict with allocation to female function, which may be resource limited (Lloyd and Bawa 1984). The strongest selection pressures promoting unisexual flowers in wind-pollinated plants may be the above described factors. It is difficult for a wind-pollinated plant to optimise both male and female fitness in one flower, when there are such different demands, both structural and productional, imposed upon the two sexual functions (Faegri and van der Pijl 1979). Consequently, since they are no longer affected by selection pressures for optimising floral displays and reward and maximising the returns from insect visits to bisexual flowers, anemophilous plants may “freely” respond to the selection for separate sexes due to the different demands on optimal male and female function in a flower. Thus, selection for sexual specialisation leads to a dioecy through total or partial reduction of the opposite sex in the resulting unisexual flowers. This division of labour in unisexual plants was recognised by Darwin (1877) who suggested that it was a possible factor in the evolution of dioecy. Dioecy could also be the result of sexual selection (Willson 1979), which in short says that the reproductive success of males is limited by their access to the female gametes, while that of the females is limited by the resources available for egg production and parental care of the offspring. Thus males tend to optimise the quantity of matings while females tend to optimise the quality. Another factor, proposed by Burd and Allen (1988), considers the suggestion that paternal fitness in wind-pollinated plants might be linearly related to male reproductive investment, because wind will not be saturated as a pollen vector even at high levels of pollen production (Charnov 1979, Charlesworth and Charlesworth 1981). Since the allocation patterns found in anemophilous species do not follow these 20

predictions, Burd and Allen (1988) suggest that the relation between male fitness and male investment follows a pattern of saturation or diminishing marginal returns, which is largely determined by the height of a plant, because pollen dispersal distances (and thus access to mates) are dependent on the height of the pollen emitting source. A taller individual should experience reduced local mating competition and have a less saturating male-fitness curve due to wider dispersal of its pollen. Thus, larger wind-pollinated plants are expected to have a relatively greater male investment. In addition, models of plant mating systems (e.g. Lloyd and Bawa 1984) show that male allocation should increase as the potential mating-group size increases (based on the arguments of sexual selection). Selfing, on the other hand, is expected to shift allocation in favour of female function, because the reduced pool of outcrossing ovules restricts potential paternal reproductive success (Lemen 1980, Charlesworth and Charlesworth 1981, Schoen 1982). Taken together, these models predict that a large wind-pollinated plant, in a large outcrossing population, should devote much of its available resources to pollen production. An associated explanation for gender specialisation involves fruit type. The correlation between dioecy and fleshy fruits is strong (Bawa 1980, Renner and Ricklefs 1995). In general, many dioecious species are tropical trees with fleshy one- or few-seeded fruits. These fruits or seeds are generally distributed by birds, which are common in the tropics and effective long-distance dispersers. In Oleaceae, many genera have fleshy and bird-dispersed fruits, e.g. Ligustrum and all genera of subtribe Oleinae. Fraxinus have wind-dispersed samaras. How does fruit type and dispersal influence gender specialisation? If the range of seed dispersal increases, the degree of local resource competition among sibling seedlings will decrease, favouring investment in female function. Wind dispersal of fruits or seeds, such as the samaras of Fraxinus, may be enhanced by height for reasons similar to those affecting pollen dispersal and lead to selection for increased female allocation (Burd and Allen 1988). Frugivore dispersal of large crops of fleshy fruits, such as the drupes of Oleinae, could create disproportionate gains for female reproduction among those plants capable of more massive fruit set (Burd and Allen 1988). The con-

flicts between selection for increased male allocation in outcrossing wind-pollinated plants and selection for increased female allocation in plants with effective long-distance fruit dispersal (by wind or birds), favour separation of sexes. Thus, these opposing selective pressures on sexual allocation will lead to males specialising in pollen production and dispersal and females specialising in fruit production. According to Givnish (1980), dioecy should predominate only in those wind-pollinated taxa that have fleshy fruits dispersed by animals. As stated above, the general explanation for the presence of dioecy in plants has been selection for outcrossing. It is evident that dioecy confer obligate outcrossing, but the question is if it has been the main selective force? The presence of the monoecious breeding system among many wind-pollinated plants, e.g. in Betula, Quercus, Alnus, and Corylus, show that selection for outcrossing is not the main factor, since monoecy may still lead to selfing (Bertin 1993). In Fraxinus, all cosexual individuals are protogynous and not only does protogyny promote outcrossing, it also prevents self-pollen from clogging the stigma. Since these mechanisms were present before the origin of dioecy, it shows that selection for outcrossing cannot have been an important force in the evolution of unisexual flowers. It is therefore my suggestion that selection for sexual specialisation is the main factor that has repeatedly led to the evolution of unisexual flowers in Oleaceae, as well as in other windpollinated plants. Even when outcrossing is advantageous, is not the main selective force and has merely conferred a beneficial side effect.

Androdioecy – why so common in Oleaceae? Although not originally intended to be covered in this thesis, the remarkable incidence of androdioecy in Oleaceae deserves an explanation. The results obtained in this study have given me the opportunity to take a fresh look at the problem — the problem of how males can coexist with hermaphrodites. The conditions under which males could be established and maintained in an originally hermaphrodite population have been modelled by Lloyd (1975), Charlesworth and Charlesworth (1978), Charlesworth (1984), and Pannell

(2000). Basically the ESS models show that males can “invade” an outcrossing hermaphrodite population if the males have at least twice the male fertility compared to the male function in hermaphrodites. The conditions for males become more severe as the selfing rate increases (as fewer ovules are accessible for the males) and the models also show that androdioecy cannot evolve as a means to promote outcrossing. Since a more than doubled male fertility has been considered implausible (Charlesworth 1984), it has been used as explanation for the apparent rarity of androdioecy among angiosperms. Androdioecy is not so unusual in Oleaceae and no less than 37 species are morphologically androdioecious. Four of these cases have so far been proven: Phillyrea angustifolia (e.g. Vassiliadis 1999), Fraxinus ornus (Dommée et al. 1999, Paper I), F. lanuginosa (Ishida and Hiura 1998) and F. longicuspis (Paper I). Why, then, does the reality in Oleaceae obviously contradict the prediction of the ESS models? The answer may be that models have not taken into account the ecological factors, discussed above, that select for unisexual flowers in wind-pollinated plants. When this is taken into consideration, males may be selected for on less stringent conditions because of the advantages of being a specialist of pollen production and dispersal in ambophilous or anemophilous species. Or put differently, because anemophily or ambophily selects for specialisation in pollen production, males may become so good at it that they may coexist with hermaphrodites. All four functionally androdioecious species are anemophilous or ambophilous and males have in these cases, except perhaps F. ornus (Paper I), a more than doubled pollen production as well as in most cases also a higher pollen germination rate and fertilisation success than hermaphrodites have. Thus, in this way the males may satisfy the conditions specified by the models. Therefore, I suggest that the possible explanations for evolution and maintenance of androdioecy in partially or completely anemophilous species are much the same as for dioecy, with increased male investment selected for in large outcrossing populations. ‘Females’ may retain partially or fully functional stamens for two main reasons: either the stamens are selected for, or they have not been selected against. A simple argument for the latter explanation is that since all of the investigated 21

species are protogynous, stigma clogging and/ or self-pollination is not a problem. On the contrary, since at least some of them are self-compatible (F. ornus and F. lanuginosa), the hermaphrodites may self-pollinate if cross-pollination fails, or during episodes of colonisation. This may be one factor favouring retention of stamens as a means of reproductive assurance. Another selection pressure for keeping the stamens, in these nectarless speices, may be for pollinator attraction (Charlesworth 1984). But more importantly, if hermaphrodites actually do contribute to part of the progeny in the population, this may be a strong enough selection pressure for keeping the male function, even if it is inferior compared to the specialised males.

Conclusions The observed trends in floral evolution, viz. number of flowers per individual, their size, form, and type of pollinator reward, can be interpreted as selection for more generalised pollination systems, leading to ambophily in suitable habitats. The presence of ambophily and/ or androdioecy may be an exaptation for the evolution of anemophily and ultimately dioecy. The main conclusion of this study is that androdioecy and dioecy are adaptations to ambophily and anemophily, respectively, and that these dimorphic breeding systems with unisexual males are mainly the result of selection for male specialisation in pollen production.

ACKNOWLEDGEMENTS First, I must thank Åslög Dahl, my supervisor, the reason for why you are now reading this PhD thesis (or perhaps just this part, often viewed as the most important!). When I started in 1995, the initial aim seemed rather simple: “check out what the ashes are doing”! I think neither of us, at least not me, could foresee at that time what would come out of this work. Now I am very grateful for having been guided by you through the study of this ideal model taxon. Thank you Åslög! It is now finally prepared… I am indebted to Victor Albert for inviting me to the Cullman Lab of Molecular Systematics at the New York Botanical Garden and let me use the facilities there for four months. It 22

really made the difference for the large amount of sequences I got done for this project. Thank you also Lena Struwe and Tim Motley for all help at the lab, innumerable dinners, etc., and Mats Gustafsson for letting me stay at your apartment. And David Kizirian, thanks for showing me how to use the sequencing equipment at the American Museum of Natural History, and for making my time in New York so enjoyable (remember: “the people only want strawberry”!). Several people have in one way or another helped me during this study. I will name a few here and I’m sure I have forgot someone important. Peter S. Green, Kew Herbarium, have patiently answered all my questions about Oleaceae and helped me select material for sequencing, and Mark Chase provided many DNA extractions from Kew. Mari Källersjö and Steve Farris (Swedish Museum of Natural History) performed the jackknife analyses. I’ve had many valuable discussions on my research and other stuff with e.g. Pierre Binggeli, Tim Chumley, Wayne Harris, Søren R. Jensen, Woong-Ki Min, Jeff Ollerton, John Pannell, Jens G. Rohwer, J-F Veldkamp, and all friends and colleagues at the SCAPE (Scandinavian Society for Pollination Ecology) meetings. I am also grateful to all those people who sent me leaves or floral samples of Fraxinus. Thank you all! For assistance during fieldwork in Sweden I am grateful to Cilla Odenman and Karin H. Persson and for assistance in the field in Sicily I thank Kerstin Helgesson, for local support Giuseppe Venturella and Pietro Mazzola at the University of Palermo, and Gioacchino Genchi for help to find populations of F. ornus. For help with carrying out the field studies in Japan and for permission to include the preliminary results on F. longicuspis in paper I, I am grateful to Tsutom Hiura (Tomakomai Research Station, Hokkaido University Forests) and Kiyoshi Ishida (Forestry and Forest Products Research Institute, Kyoto). I also want to thank the staff at the Tokyo University Forest at Chichibu for all valuable help in the field, especially locating and climbing the trees! I am also grateful to Satoshi Nanami and Takashi Osono (Laboratory of Forest Ecology, Graduate School of Agriculture, Kyoto University) for taking care of me and Björn in Kyoto. I thank Vivian Aldén, Tudlik Bergqvist, Margit Fredrikson, and Jan Helgesson for as-

sistance in the lab. Of course, I also need to thank all past and present colleagues at the Botanical Institute, especially Juha Alatalo, Uno Eliasson, Roger Eriksson, Claes Gustafsson, Magnus Lidén, Karin Lindblad, Bengt Oxelman, Claes Persson, Magnus Popp, Johan Rova, Anna and Mikael Stenström, the staff at the library for all help, and all those who I have forgot to mention! Finally, but not least, I want to thank my husband Johan and my parents, Kerstin and Ingvar Helgesson, for supporting me in this work. Isn’t it amazing how much there is to write about some leaves?! Kerstin was also a very good assistant during the field work in Sicily (although she thought we would have some days off…!). This PhD research was funded by Knut & Alice Wallenbergs stiftelse, The Royal Swedish Academy of Sciences, Kungliga & Hvitfeldtska Stiftelsen (Överskottsfonden), Adlerbertska forskningsfonden, Paul & Marie Berghaus donationsfond, Wilhelm & Martina Lundgrens Vetenskapsfond, Helge Ax:son Johnsons Stiftelse, Stiftelsen Kulturfonden Botaniskas Vänner (Carl Skottbergs-stipendium), Kungliga Vetenskaps- & Vitterhets-Samhället in Göteborg, Uddenberg-Nordingska Stiftelsen, and Collianders stiftelse. Thank you all so much!

REFERENCES Andersson, L. & Rova, J. H. E. 1999. The rps16 intron and the phylogeny of the Rubioideae (Rubiaceae). Plant Systematics and Evolution 214: 161-186. Armbruster, W. S. 1992. Phylogeny and the evolution of plant-animal interactions. BioScience 42: 13-20. Aronne, G. & Wilcock, C. C. 1994. Reproductive characteristics and breeding system of shrubs of the Mediterranean region. Functional Ecology 8: 6976. Bawa, K. S. 1980. Evolution of dioecy in flowering plants. Annual Review of Ecology and Systematics 11: 15-39. Bawa, K. S. & Beach, J. H. 1981 Evolution of sexual systems in flowering plants. Annals of the Missouri Botanical Garden 68: 254-274. Bawa, K. S. & Bullock, S. H. & Perry, D. R. & Coville, R. E. & Grayum, M. H. 1985. Reproductive biology of tropical lowland rain forest trees. II. Pollination systems. American Journal of Botany 72: 346-356. Bertin, R. I. 1993. Incidence of monoecy and dichogamy in relation to self-fertilization in angio-

sperms. American Journal of Botany 80: 557-560. Binggeli, P. & Power, A. J. 1991. Gender variation in ash (Fraxinus excelsior L.) In: Proceedings of the Irish Botanist Meeting, p 42. University College Dublin, Dublin. (Abstract). Bullock, S. H. 1994. Wind pollination of neotropical dioecious trees. Biotropica 26: 172-179. Burd, M. & Allen, T. F. H. 1988. Sexual allocation strategy in wind-pollinated plants. Evolution 42: 403407. Charlesworth, B. & Charlesworth, D. 1978. A model for the evolution of dioecy and gynodioecy. American Naturalist 112: 975-997. Charlesworth, D. 1984. Androdioecy and the evolution of dioecy. Biological Journal of the Linnean Society 22: 333-348. Charlesworth, D. 1993. Why are unisexual flowers associated with wind pollination and unspecialized pollinators? American Naturalist 141: 481-490. Charlesworth, D. & Charlesworth, B. 1981. Allocations of resources to male and female function in hermaphrodites. Biological Journal of the Linnean Society 15: 57-74. Charnov, E. L. 1979. Simultaneous hermaphroditism and sexual selection. Proceedings of the National Academy of Science, USA 76: 2480-2484. Cox, P. A. 1991. Abiotic pollination: an evolutionary escape for animal-pollinated angiosperms. Philosophical Transactions of the Royal Society London B 333: 217-224. Cruden, R. W. 1977. Pollen-ovule ratios: a conservative indicator of breeding systems in flowering plants. Evolution 31: 32-46. Darwin, C. 1877. The different forms of flowers on plants of the same species. Appleton, New York. Dommée, B. & Geslot, A. & Thompson, J. D. & Reille, M. & Denelle, N. 1999. Androdioecy in the entomophilous tree Fraxinus ornus (Oleaceae). New Phytologist 143: 419-426. Donoghue, M. J. 1989. Phylogenies and the analysis of evolutionary sequences, with examples from seed plants. Evolution 43: 1137-1156. Dyer, C. 1991. A biosystematic study of the African species of Olea L. (Oleaceae). Dissertation, Univ. of the Witwatersrand, Johannesburg, South Africa. Faegri, K. & van der Pijl, L. 1979. Abiotic pollination. In: The principles of pollination ecology, pp. 3440. Pergamon Press. Felsenstein, J. 1985. Phylogenies and the compartive methods. American Naturalist 125: 1-15 de Figueiredo, R. A. & Sazima, M. 2000. Pollination biology of Piperaceae species in southeastern Brazil. Annals of Botany 85: 455-460. Frankel, R. & Galun, E. 1977. Pollination mechanisms, reproduction and plant breeding. In: Monographs on theoretical and applied genetics 2. SpringerVerlag. Freeman, D. C. & Lovett Doust, J. & El-Keblawy, A. & Migla, K. J. & McArthur, E. D. 1997. Sexual specialization and inbreeding avoidance in the evolution of dioecy. The Botanical Review 63: 65-92.

23

Geber, M. A. & Dawson, T. E. & Delph, L. F. (eds.) 1999. Gender and sexual dimorphism in flowering plants. Springer-Verlag, Berlin Heidelberg, Germany. Givnish, T. J. 1980. Ecological constraints on the evolution of breeding systems in seed plants: dioecy and dispersal in gymnosperms. Evolution 34: 959972. Givnish, T. J. 1982. Outcrossing versus ecological constraints in the evolution of dioecy. American Naturalist 119: 849-865. Goodwillie, C. 1999. Wind pollination and reproductive assurance in Linanthus parviflorus (Polemoniaceae), a self-incompatible annual. American Journal of Botany 86: 948-954. Gould, S. J. & Vrba, E. S. 1982. Exaptation - a missing term in the science of form. Paleobiology 8: 4-15. Harvey, P. H. & Pagel, M. D. 1991. The comparative method in evolutionary biology. Oxford University Press. Ishida, K. & Hiura, T. 1998. Pollen fertility and flowering phenology in an androdioecious tree, Fraxinus lanuginosa (Oleaceae), in Hokkaido, Japan. International Journal of Plant Science 159: 941-947. Jensen, S. R. & Franzyk, H. & Wallander, E. (submitted) Chemotaxonomy of the Oleaceae: Iridoids as taxonomic markers. de Jong, P. C. 1976. Flowering and sex expression in Acer L. – a biosystematic study. Mededelingen Landbouwhogeschool Wageningen Nederland 76: 1-201. Lemen, C. 1980. Allocation of reproductive effort to the male and female strategies in wind-pollinated plants. Oecologia 45: 156-159. Liccardi, G. & D’Mato, M. & D’Amato, G. 1996. Oleaceae pollinosis: A review. International Archives of Allergy and Immunology 111: 210-217. Linder, H. P. 1998. Morphology and the evolution of wind pollination. In: S. J. Owens and P. J. Rudall (eds.) Reproductive Biology, pp. 123-135. Royal Botanic Gardens, Kiew. Linskens, H. F. 1996. No airborne pollen within tropical rain forests. Proceedings of the Koninklijke Nederlandse Akademie van Wetenschappen 99: 175-180. Lloyd, D. G. 1975. The maintenance of gynodioecy and androdioecy in Angiosperms. Genetica 45: 325-339. Lloyd, D. G. 1981. The distribution of sex in Myrica gale. Plant Systematics and Evolution 138: 29-45. Lloyd, D. G. 1982. Selection of combined versus separate sexes in seed plants. American Naturalist 120: 571-585. Lloyd, D. G. & Bawa, K. J. 1984. Modification of the gender of seed plants in varying conditions. Evolutionary Biology 17: 255-338. Lovett Doust, J. & Lovett Doust, L. 1988. Plant reproductive ecology: patterns and strategies. Oxford University Press.

24

Niklas, K. J. 1985. The aerodynamics of wind pollination. The Botanical Review 51: 328-386. Pannell, J. 2000. A hypothesis for the evolution of androdioecy: the joint influence of reproductive assurance and local mate competition in a metapopulation. Evolutionary Ecology 14: 195211. Regal, P. J. 1982. Pollination by wind and animals: Ecology of geographic patterns. Annual Review of Ecology and Systematics 13: 497-524. Renner, S. S. & Feil, J. P. 1993. Pollinators of tropical dioecious angiosperms. American Journal of Botany 80: 1100-1107. Renner, S. S. & Ricklefs, R. E. 1995. Dioecy and its correlates in the flowering plants. American Journal of Botany 82: 596-606. Richards, A. J. 1997. Plant breeding systems. Chapman & Hall. Schoen, D. J. 1982. Male reproductive effort and breeding system in an hermaphroditic plant. Oecologia 53: 255-257. Swofford, D. L. 2000. PAUP*: phylogenetic analysis using parsimony (* and other methods), version 4. Sinauer, Sunderland, Massachusetts, USA. Tollsten, L. & Knudsen, J. 1992. Floral scent in dioecious Salix (Salicaceae) — a cue determining the pollination system? Plant Systematics and Evolution 182: 229-237. Tormo, M. R. & Munoz, R. A. & Silva, P. I. & Gallardo, L. F. 1996. Pollen production in anemophilous trees. Grana 35: 38-46. Vassiliadis, C. 1999. Evolution et maintien de l’androdioécie: Etude théorique et approaches expérimentales chez Phillyrea angustifolia L. PhD Dissertation, Université des Sciences at Technologies de Lille. Vroege, P. W. & Stelleman, P. 1990. Insect and wind pollination in Salix repens L. and Salix caprea L. Israel Journal of Botany 39: 125-132. Webb, C. J. 1999. Empirical studies: evolution and maintenance of dimorphic breeding systems. In: Geber, M. A. & Dawson, T. E. & Delph, L. F. (eds.) Gender and sexual dimorphism in flowering plants, pp. 61-95. Springer-Verlag, Berlin Heidelberg, Germany. Whitehead, D. R. 1968. Wind pollination in the angiosperms: evolutionary and environmental considerations. Evolution 23: 28-35. Whitehead, D. R. 1983. Wind pollination: Some ecological and evolutionary perspectives. In: Real, L. (ed.) Pollination Biology, pp. 97-108. Academic Press, London, U.K. Williams, G. & Adam, P. 1999. Pollen sculpture in subtropical rain forest plants: Is wind pollination more common than previously suspected? Biotropica 31: 520-524. Willson, M. F. 1979. Sexual selection in plants. American Naturalist 113: 777-790.

ISBN 91-88896-37-4

Paper I

Paper II

Paper III

Paper IV

Paper V

Paper VI

Studies on the reproductive ecology of F. excelsior, F. ornus and F. longicuspis, and a comparison with some other species of Fraxinus (Oleaceae)

Eva Wallander and Åslög Dahl

SUMMARY 1. The reproductive ecology of the wind-pollinated tree Fraxinus excelsior L. (Oleaceae) in Sweden is described in detail and compared with two other species of the genus, the wind-pollinated F. longicuspis Sieb. & Zucc. in Japan, and the insect-pollinated F. ornus L. in the Mediterranean. 2. F. excelsior is polygamous with a continuous phenotypic sex expression, from unisexual male via hermaphrodite to female. Five basic types of flowers in various combinations make up four main types of inflorescences, which in turn can be combined to form three classes of phenotypic gender expression: male, hermaphrodite, and female. 3. Over the seven-year field study, the population consisted of on average equal numbers of the three gender morphs. Sex expression varied within the continuum among years for some of the trees, but none of them changed in functional gender. 4. Flowering and fruiting in the study population varied over the years in an irregular pattern, which seems to have some sort of environmental component determining the apparent large-scale synchronicity among trees within years. 5. Hermaphrodite flowers are self-compatible but protogynous for about a week, thus avoiding self-pollination. They remain receptive for another week following pollen release and the overlap between male and female anthesis is about a week, depending on weather. At the time of pollination in F. excelsior, the ovules do not yet contain mature megagametophytes. 6. F. ornus is functionally androdioecious and was shown to be protogynous and ambophilous (both wind- and insect-pollinated). A review of the records on flower visitors of F. ornus is provided. 7. F. longicuspis is the first documented case of an androdioecious and wind-pollinated species within Fraxinus. The studied population consists of 50% males and there are no significant differences between males and hermaphrodites as to size of trees and stamens, pollen content per anther, and germination rate of pollen, i.e., no difference in potential male function between males and hermaphrodites. However, males produce 2.2 times more flowers per inflorescence and 40% larger pollen grains than hermaphrodites. Pending further evidence on pollen viability, it is suggested that F. longicuspis is functionally androdioecious. Key words: androdioecy, anemophily, breeding system, flowering phenology, polygamy 1

INTRODUCTION The genus Fraxinus, the ashes, is one of 24 extant genera in Oleaceae, the olive family (Wallander & Albert 2000). The genus consists 43 recognised tree and shrub species distributed in the Northern Hemisphere (Wallander ms. 1). About one third are entomophilous (insect-pollinated), and most of these occur in Asia (only two in N America). The remainder are anemophilous (wind-pollinated) and are distributed in both North America and Eurasia. The common ash in Europe, F. excelsior L., is a wind-pollinated and polygamous tree. The polygamous breeding system of this ash has been known for a long time (Schultz 1892) and the individuals in a population show a continuum of gender, from trees bearing only male (staminate) flowers, via hermaphrodites (with both stamens and pistil functional), to functionally female trees (pistillate flowers with, or without, infertile stamens). There also seems to be variation within trees, branches, and inflorescences and between years. Although the breeding system of F. excelsior has been characterised in morphological terms in a number of studies (e.g., Schultz 1892, Larsen 1945, Rohmeder 1952, Picard 1982, Binggeli & Power 1991), the functionality of male, female and hermaphrodite individuals, and their relative abundance in populations, is poorly understood. The literature review by Lamb & Boshier (1994), on the current state of knowledge on the reproductive biology of F. excelsior, identified several areas where knowledge was poor or contradictory: (1) There are few studies on periodicity of flowering and fruiting. One single long-term study of flowering (Hyde 1951), based on airborne pollen in Cardiff (Wales) during 1943-1949, indicated a pronounced three-year cycle. Other studies have been concentrating on the fruiting pattern, most of which suggest either a two-year, three-year, or an irregular cycle (Gardner 1977, Tapper 1992). (2) There are no studies on the flowering phenology of ash, an important piece of information as to get a complete picture in understanding the complicated reproductive system of ash. (3) In particular, the extent to which individual trees change sex, or sexual function, from year to year is still poorly known. (4) The temporal variation in sex ratio within a population and between populations is unknown. (5) Incompatibility mechanisms are unknown and (6) the 2

function of different sex morphs is also unknown. The aims of this study are to fill some of the gaps in the knowledge of the reproductive system of F. excelsior and to compare with species exhibiting other types of pollination and breeding systems. Within the genus Fraxinus there are three distinct groups (sections Dipetalae, Ornus, and Fraxinus) and within each of these groups there have been independent origins of adaptations to wind-pollination (Wallander ms. 1). F. excelsior belongs to section Fraxinus, a group of 23 wind-pollinated species, which are all characterised by having apetalous flowers in inflorescences that emerge from lateral buds before the leaves of the terminal bud begin to expand. Section Ornus consists of 16 Eurasian species and is characterised by combined floral and vegetative terminal buds. Eleven petalous species are pollinated by insects, or by both insects and wind (ambophily), and flowering begins after the leaves are fully developed. Within this section, five species (informal group ‘Ornaster’) have evolved anemophily through loss of petals and scent and delayed leafing. Finally, section Dipetalae consists of three American species, two of which have evolved anemophily in much the same way as in section Fraxinus. As a representative for the insect-pollinated species in section Ornus, F. ornus L., distributed in the Mediterranean area, was chosen. In floras, F. ornus is described as being hermaphrodite. Already at the first inspection, it turned out to be morphologically androdioecious (separate male and hermaphrodite individuals), a fact that was also independently found by Dommée et al. (1999) who showed that F. ornus is functionally androdioecious. In order to compare the reproductive biology of F. excelsior with a wind-pollinated species from the ‘Ornaster’ group, F. longicuspis Sieb. & Zucc. was selected for comparative studies. It is endemic to Japan and is described as having male and bisexual flowers on different trees (Sawada 1937, Yamazaki 1993), i.e., it is also androdioecious. In addition to the detailed studies of these three species, we describe and compare the floral morphology of all other species in section Ornus, most of which are androdioecious. Androdioecy is a very rare breeding system among plants (Charlesworth 1984) and hypotheses on its occurrence in Fraxinus will be discussed.

This study aims to increase the existing knowledge about reproductive systems of F. excelsior and F. ornus, two important tree species in Europe. Previously, nothing was known about the reproductive ecology of F. longicuspis and, although requiring a more complete study for a full understanding of the reproductive biology, the present preliminary study contributes to the understanding of potentially functional androdioecy in this anemophilous species.

11°56’E). The site is within the nemoral zone, and experiences a suboceanic temperate climate with 750 mm of precipitation annually and a mean annual temperature of 7°C. The soil cover is thin and most of the trees are smaller than average (up to 10 m). The population consists of about 50 trees. In 1995, 25 trees were individually marked and studied intensively over seven years. In 1996, three additional trees were included. All study trees have branches that can be reached from ground. Phenological studies

MATERIALS AND METHODS The reproductive biology of F. excelsior was studied in detail over seven years in one population on the Swedish west coast. Reported here are also results from short-term field studies of F. ornus and F. longicuspis, studied during one reproductive season in Italy and Japan, respectively. Field site and methods are described separately for each of the three studied species. Other taxa have only been studied using herbarium material.

Fraxinus excelsior in Sweden Study species The European ash, F. excelsior L. (the only native Fraxinus species in Sweden) is a wind-pollinated and polygamous species. It occurs over central and northern Europe and is relatively common on base and nutrient rich, moist soils in southern and central Sweden. The trees are large, up to 30 m, but normally around 20 m, and the flowers lack calyx and corolla. There are hermaphrodite flowers with a pistil and two stamens with well-developed anthers, male flowers with two functional stamens only, and female flowers with one pistil and two rudimentary stamens, or no stamens at all. Many flowers of one or more of these types are borne together in lateral inflorescences (panicles) that emerge before the leaves expand from the terminal leaf buds in the spring. There are two locules in the syncarpous ovary, with two ovules each, but the mature fruit is a one-seeded samara. The ovary and fruit characters apply to all species of Fraxinus. Population and field site characteristics The study population grows on a north-west facing slope in the centre of Göteborg (57°41’N,

Six branches on each tree were tagged in 1995. The same trees were studied over the seven years but some of the branches changed between years (due to loss of tags, dead or broken branches, etc.). There were up to ten inflorescences per tagged branch. The tagged branches were scored about every second day during the flowering period, usually from mid April to mid May. Data gathered per inflorescence included bud opening date = start of female anthesis (first day of stigma emergence), start of male anthesis (first day of pollen dispersal), end of female anthesis (stigmata all withered), and end of male anthesis (anthers empty). The overlap between the male and female phases was calculated as end day of male or female anthesis (whichever comes first) minus start day of male anthesis (female anthesis always begins before male). Each year the flowering intensity of the trees were classified into the following categories: 0 = no flowering, 1 = few inflorescences (scattered inflorescences, fewer than ten), 2 = medium flowering (few inflorescences on all branches or many inflorescences on few branches), 3 = mast flowering (many inflorescences on all branches). Mean flowering intensity per year was calculated only for the flowering trees (i.e. excluding zero values). Flowering rate (%) was calculated for each year as number of flowering trees per total number of trees. Flowering rate and mean flowering intensity were compared to the total sum of airborne Fraxinus pollen in Göteborg region over the years. The total pollen sum is expressed as the sum of average number per cubic meter of air for 24-h periods during the anthesis in a year. Vegetative phenology was scored through recording the first date for the following four stages: 1 = outer bud scales begin to separate, 2 = leaves visible between fully separated bud 3

scales, and 3 = leaves fully unfolded (but not yet fully grown). Recording stage 2 is important because a late frost can damage the leaves at this point, and we wished to know if there are any differences at this stage between flowering and non-flowering trees. Fruiting was recorded in September/October of 1997, 1998, and 1999, and trees were classified into the following categories (which follows the flowering categories): 0 = no fruits, 1 = few fruits, 2 = medium fruit set, and 3 = mast fruiting. Values of zero for males and non-flowering trees were not included in the mean fruit set for each year. A number of inflorescences were collected at different dates from some of the flowering trees in 1995-1997 and were preserved in FPA (formaldehyde 37%, propionic acid and ethanol 50% in a 5:5:90 ratio). They were later used to study the development of female gametophytes, growth of pollen tubes, and fertilization event, with the aid of a Confocal Laser Scanning Microscope (CLSM). Measurements A number of the collected inflorescences were also randomly selected and used for counting number of flowers per inflorescence. In total, 54 inflorescences from 12 trees, belonging to four different inflorescence categories (see Results), were counted. From three of the trees (one male, one hermaphrodite and one female), the number of flowers in the inflorescences from two flowering years with different flowering intensity were compared. From the above collected inflorescences we picked two male and two hermaphrodite ones from four different trees. The size of pollen grains (diameter) was measured on 20 grains from five random anthers on each inflorescence. The pollen was embedded in glycerol gelatine and measured at 400x magnification in a Zeiss™ microscope with an objective micrometer. Means per anther were used in the statistical analysis. Number of pollen grains per anther was estimated by smearing the contents of an anther onto a microscope slide and cover with glycerol gelatine and a cover slip. The number of pollen grains per 11 rows across the slide was counted and multiplied by a factor to make up the total area. There are two stamens and four ovules in all Fraxinus flowers and the pollen:ovule (P/O) ratio was calculated as number of pollen grains per anther times two 4

divided by four. As a simple test for determining if the hermaphrodites were self-compatible we put nylon bags (10 µm mesh size) around three twigs each of two hermaphrodite trees to exclude other pollen. There were 4-8 inflorescences in each bag, just about to open at the time of bagging. Determination of stigma receptivity Stigma receptivity was tested in vitro by dipping stigmata from newly opened buds in 3 % hydrogen peroxide (H2O2). Bubbling on the stigma indicate presence of peroxidase enzymes (Galen & Plowright 1987). Stigma receptivity was also tested in vivo through pollen germination tests. From trees with hermaphrodite and female flowers, shoots with newly opened buds were cut and placed in cups with water under a bell jar in a climate chamber where the temperature was 12°C. The stigmata were hand pollinated with a small brush using fresh pollen from male flowers. After 24 and 48 hours buds were collected and fixed in FPA for 24 hours and then transferred to 70 % ethanol for storage. The stigmata were later examined for germinated pollen in a drop of Aniline blue in a microscope. Pollen germination in vitro and in vivo To test if there is a difference in viability between pollen from anthers of male, hermaphrodite and ‘female’ flowers, pollen from 10 trees (3 males, 2 females and 5 hermaphrodites) were grown in vitro. Twigs with inflorescences ready to release pollen were collected from the study population and brought into the laboratory. Inflorescences were put on a Petri dish and after a couple of hours in room temperature the stamens had opened. Several pollen germination media were tested and the best one retained for further usage was a 15 % sucrose medium, modified from Brewbaker & Kwack (1963), consisting of 15 g sucrose, 30 mg calcium nitrate, 20 mg magnesium sulfide, 10 mg potassium nitrate, 10 mg boric acid, 300 mg yeast extract (0.3%) and distilled water added to 100 ml. When all had dissolved, 600 mg agarose was added to the mixture and heated to boiling temperature in a microwave oven. The pollen germination medium was then smeared on a microscope slide and after a couple of minutes, when the medium had cooled

and solidified, fresh pollen was dusted onto the slide. We used three replicates from each individual, which were placed together in a petri dish with a filter paper moistened with distilled water. After 24 hours, pollen germination was stopped by spraying the slides with FPA (this also killed the growing bacteria and fungi). A couple of hundred pollen (randomly chosen) was scored for presence of germination.

Fraxinus ornus in Italy Study species Fraxinus ornus, the manna ash, occurs in the central and eastern Mediterranean area and eastwards to Iraq, in mixed woods, along roadsides, and in rocky places. The trees may be up to 20 m tall, but are usually much smaller, around 5-10 m, and the flowers are borne in large (15-20 cm long) and showy terminal inflorescences together with the leaves. Each hermaphrodite flower consists of a small campanulate calyx, four white free petals, two stamens and a pistil. Male flowers resemble hermaphrodite flowers but have no pistil. At the time of planning for this study, there was no report in the literature about androdioecy in this species. Populations and field site characteristics The flowering phenology and pollination was studied in two populations at two different elevations about 5 km apart, in the Madonie Mountains (Parco Naturale Regionale delle Madonie), near Castelbuono (37°55’N, 14°05’E) on the island of Sicily, Italy. In the population at Montagnola, at about 400 m altitude along the road from Castelbuono to Pollina, 28 flowering trees were tagged (21 males and 7 hermaphrodites). This population consisted of many young (=small) trees, most of them males. In the population near Barraca, at around 800 m altitude along the road from Castelbuono to Milocca, 23 flowering trees (10 males and 13 hermaphrodites) were tagged for phenological studies. These trees were older (bigger) than those in the other population. The two populations plus a third were used for checking the gender ratio. Vouchers from all trees were collected and are deposited at Herbarium GB. Inflorescences, or parts thereof, were also collected from all trees and preserved in 70% ethanol.

Phenological observations Five inflorescences on each of the tagged trees in the two populations were checked every day from 25 April to 4 May 1996 and the progressing male and female anthesis noted. The stages of anthesis used were 0 = flowers unopened, 1 = opened, 2 = anthers shedding pollen, and 3 = stigma necrotised and/or anthers empty. As it turned out, the beginning of stage 1 was never recorded because all flowers had already opened on arrival to the study population. Pollination studies A few insects visiting the flowers during the phenological observations were caught and preserved in ethanol. This was not done in any systematic way and the species listed from this study in Table 6 only reflects the ones that happened to be seen and was possible to catch. To find out if or to which extent pollen was also released into the air, 14 microscope slides covered with a sticky Vaseline and glycerol paste were put out on different objects (stones, stone walls, fence posts, etc.) in the population surroundings. It was slightly windy and the directions varied. The slides were collected after 10 hours, covered with Safranin (red) dyed glycerol gelatin and a cover slip and later checked under a microscope. No total count was made, just an estimate if there was no, little or much Fraxinus pollen on the slides. One cannot distinguish between pollen from different Fraxinus species, but there was no other Fraxinus species flowering in the area, only different species of Quercus (Fagaceae), Erica arborescens (Ericaceae), Calycotome spp. (Fabaceae), and Cistus spp. (Cistaceae), which are easy to distinguish from Fraxinus pollen. Testing of stigma receptivity in vitro could not be made in the Sicily populations (because it was too late in the season). Instead, this was tested on one hermaphrodite tree of the species flowering in Göteborg Botanical Garden at the end of May 1997. As soon as the flowers had opened, a couple of inflorescences were brought into lab and stigmata from newly opened flowers were tested with 3 % hydrogen peroxide (H2O2). Bubbling on the stigma indicated the presence of peroxidase enzymes. Embryological studies were conducted on a few flowers from one hermaphrodite tree in the Sicily population, using the same methods and equipment as in F. excelsior. 5

Measurements Size of pollen grains and the number of pollen grains in anthers were estimated in the same way as for F. excelsior. The sample size was two male and two hermaphrodite inflorescences from four different trees. Five random anthers per inflorescence were used for the size measurements and two anthers for the pollen counts.

Fraxinus longicuspis in Japan Study species and site Fraxinus longicuspis Sieb. et Zucc. is endemic to Japan and occurs in deciduous forests at 1001100 m altitude in the central and southern parts (Honshu, Shikoku, and Kyushu Islands). It is a large deciduous canopy tree, up to 25 m tall, growing at low density among other broadleaved trees along rivers and on slopes. Flowers emerge in many-flowered panicles together with the leaves from terminal buds in April to May. They are described as androdioecious (Yamazaki 1993) with male and hermaphrodite flowers on separate trees. The flowers have a small cup-shaped calyx but no corolla. The hermaphrodite flowers have a pistil and two stamens, and the male flowers have two stamens only. Populations and field site characteristics A scattered population was found at Tokyo University Forest in Chichibu (lat. 35°55’N, long. 139°00’E), west of Tokyo in Saitama prefecture, central Japan. Fourteen trees were located growing at 500-800 m altitude in section Ohchigawa and at 700 m altitude in section Tochimoto and these were studied daily during 7-15 May 1999. Because of the low sample size of trees, difficulty of reaching the flowers in some tall trees, and because flowering had already started on arrival to the study site, a full field study of this species, including phenology and pollination experiments, could not be done as planned. Nevertheless, we wanted to include some of the data in this paper in order to compare the reproductive strategies of this species with the others. The results, however, should be regarded as preliminary until more thorough studies can be done. Observational studies and measurements The stage of male anthesis in male and hermaphrodite trees was noted on arrival to the 6

site and then almost every day for the following eight days. Measures of height and girth at breast height (GBH) were taken on all 14 trees. Number of flowers per inflorescence was counted in 5-8 fresh inflorescences per tree from five hermaphrodites and one male. Because there was only one male in the field that had not yet started to release pollen (and therefore its inflorescences still contained all flowers), additional counting of 1-6 inflorescences from each of five male voucher specimens taken from the study trees were later done at home. The latter male counts are very conservative as the stamens and parts of the inflorescence drop soon after the anthers are empty and in reality the number of male flowers are higher than reported. In total 22 male and 28 hermaphrodite inflorescences were counted. To quantify male floral characters, measurements of anther and filament length were made using a 10x magnifying Zeiss™ stereoscopic microscope with an objective micrometer. Ten flowers each in 1-3 inflorescences in each of four male and three hermaphrodite trees were measured. Stamen length was later calculated as anther plus filament length. Size of pollen grains and the number of pollen grains in anthers were estimated in the same way as in F. excelsior. The sample size was one male and two hermaphrodite inflorescences from three different trees. Five random anthers per inflorescence were used for the size measurements and two anthers per inflorescence for the pollen counts. Vouchers from all trees were collected and are deposited at GB. Inflorescences, or parts thereof, were also collected from all trees and preserved in 70% ethanol. Pollen germination in vitro To assess one aspect of potentially differential viability between pollen from male and hermaphrodite flowers, a pollen tube germination test was carried out in vitro in almost the same way as on F. excelsior. The same kind of medium was used and tested with 5%, 10%, and 15% sucrose. There was very little growth on the 10% and 15% media so only the 5% medium was finally used. After the growth medium was mixed, smeared onto a microscope slide and allowed to cool, pollen from newly opened anthers in two inflorescences each of three hermaphrodite trees (2 x 3 slides, i.e., only one replicate per inflorescence) was dusted onto the

slides. Since only one male tree with recently opened anthers was accessible, four inflorescences picked from these trees the day before were used. The four twigs were put in vases and kept in room temperature overnight. Pollen from anthers that opened during the night was dusted over each of the four slides. The air in the covered Petri dishes was kept moist with a wet filter paper. Pollen germination was stopped by ethanol after 13 hours and then counted immediately. Percentage germinated pollen was calculated as number of germinated pollen grains out of c. 200 grains in randomly selected rows on the microscope slide.

Other species in section Ornus

Figure 1. The five floral types of Fraxinus excelsior: (a) male, (b) male with rudimentary pistil, (c) hermaphrodite, (d) female with rudimentary stamens, and (e) female flower. Reproduced with permission from Binggeli & Power (1991).

All statistical analyses were performed using StatView® version 5.0.1 and SuperANOVA version 1.11 (Abacus Concepts, Inc., Berkeley, CA). Data that were not normally distributed and/ or with non-homogenous variances, were logtransformed before using parametric tests. Percentage data were arcsine square root transformed before testing. In some cases, the unpaired t-test (two-tailed) was used to test the null hypothesis of no difference in group mean between the two genders. A nested ANOVA model (type III sums of square) was used to test the distribution of variance at different levels (among genders, among trees within genders, and among inflorescences within trees). The error term of the main effect was set to the nested factor. In case of significant differences between groups, Fisher’s PLSD post-hoc test was used to evaluate comparisons between genders for a significance level set to 0.05. All means are given ± one standard deviation (SD), unless it is means of means which are reported ± one standard error (SE).

RESULTS Fraxinus excelsior Floral morphology Basically, there are five floral types (Fig. 1). These types can be combined in various ways in an inflorescence and here we distinguish four main types of inflorescences (Table 1). A male inflorescence consists only of male flowers (type a) or quite frequently with one or a few hermaphrodite flowers (type c) at the top of an otherwise male inflorescence. In the latter, male flowers with a rudimentary pistil (type b) may be present as well. Hermaphrodite inflorescences consist mostly of type c flowers, but sometimes these are mixed with type d, (i.e.,

5 mm

In this study, three species with different pollination and breeding systems were selected for detailed studies of the reproductive ecology. To compare with other species, data on the reproductive biology were derived or inferred from literature and herbarium studies. Data on the breeding system in section Ornus, many of which seemed to be androdioecious, were not always readily gained from the literature. Therefore, the phenotypic sex expression was recorded on 179 flowering herbarium specimens of Fraxinus belonging to 13 of the 16 recognised species in section Ornus. In order to estimate the male function in hermaphrodite flowers, we compared stamen size relative to other floral parts in the two gender morphs of each species, under the assumption that size and form of staminate parts reflect one aspect of the functionality. On 71 herbarium specimens, measurements of anther, filament, pistil (only for hermaphrodites), and petal length (only for petalous taxa) were taken on ten (dry) mature flowers per specimen and averaged separately for the genders of each species. Stamen length was a posteriori calculated as anther plus filament length. Equipment used was the same as for the measurements of F. longicuspis.

Statistical analyses

a

b

c

d

e 7

Table 1. Four main types of inflorescences of Fraxinus excelsior, consisting of different proportions of the five basic floral types (Fig. 1). + few or rare (80% hermaphrodite flowers were categorised as hermaphrodites. Number of flowers/inflorescence

male mixed male & hermaphrodite hermaphrodite female

Floral types

n=10

n=10

n=16

n=18

M

X

H

F

150 140 130 120 110 100 90 80 70 60 50

Figure 3. Mean number of flowers per inflorescence in the four different types of inflorescences of Fraxinus excelsior in 1997. Error bars represent ± 1 SE. M = male, X = mixed male and hermaphrodite, H = hermaphrodite, and F = female.

further on, we describe different variants which have been given Roman numerals. Male individuals are defined as those who never produce fruit (male variant I), or rarely produce a few fruits (male variant II). Hermaphrodite variant III consist of only ±25% hermaphrodite flowers, but since they do set a low fruit crop they are classified as hermaphrodites in this study. The ‘true’ hermaphrodites (variant I) carry only hermaphrodite inflorescences but some trees (variant II) have a low number of mixed male

Table 2. Categories of phenotypic gender of Fraxinus excelsior individuals, based on the types of inflorescences (Table 1) present on an individual in one flowering year. Different variants of a category are given Roman numerals within parentheses and are discussed in the text. + few or rare ( 0.05. Species

Mean pollen grain size (µm) male

F. excelsior F. longicuspis F. ornus F. lanuginosa F. sieboldiana

hermaphrodite

24.8±0.3 21.6±0.3 22.0±0.4 23.7±0.2 20.7±0.2

24.1±0.3 19.2±0.2 22.3±0.2 20.8±0.5 20.9±0.3

Mean no. pollen grains/anther

P

male

hermaphrodite

P

ns. 0.0001 ns. 0.0006 ns.

31126±9357 12404±870 17753±4122 15065±3301 6407±1640

10722±7140 15923±2938 26854±10730 12216±814 5264±2112

0.01 ns. ns. ns. ns.

P/O ratio 5361 7962 13427 6108 2632

Table 6. Visitors to Fraxinus ornus flowers found during this field study, and from the literature. Order

Family (subfamily)

Species

Reference

Heteroptera

Pentatomidae Miridae

Eurygaster maura sp.

Porsch (1958) this study

Hymenoptera Megachilidae Apidae Halictidae

Chelostoma sp. Apis mellifica var. ligustica Spin. Halictus sp

this study Scotti (1911) Scotti (1911)

Diptera

Tachinidae Psilidae

Lypha dubia Fallen Psila sp.

this study this study

Lepidoptera

Micropterigidae Eriocraniidae

Micropterix sp. Eriocrania sp.

this study this study

Coleoptera

Alleculidae Alleculidae Alleculidae Chrysomelidae Cryptophagidae Elateridae Elateridae Elateridae Elateridae Malachiidae Melyridae (Dasytinae) Melyridae (Dasytinae) Melyridae (Dasytinae) Melyridae (Dasytinae) Melyridae (Dasytinae) Scarabaeidae (Cetoniinae) Scarabaeidae (Cetoniinae) Scarabaeidae (Cetoniinae) Scarabaeidae (Melolonthinae) Scarabaeidae (Melolonthinae) Staphylinidae

Gonodera luperus Hbst. Isomira murina L. Isomira semiflava Kuest. Galeruca calmariensis L. Cryptophagus cylindricus Kies. Cardiophorus cinereus Hbst. Cardiophorus discicollis Hbst. Cardiophorus ebeninus Germ. Cardiophorus rufipes Goeze Attalus dalmatinus Er. Danacaea marginata a. thoracica Schils. Dasytes subaeneus Schönherr Haplocnemus sp. Danacaea (cf. denticollis Baudi) Danacaea (cf. ambigua Mulsant) Oxythyrea funesta Poda Oxythyrea stictica L. Cetonia aurata L. Hoplia farinosa L. Hoplia argentea Poda Omalium cinnamoneum Kr.

Novak (1952) Novak (1952) Novak (1952) Scotti (1911) Novak (1952) Novak (1952) Novak (1952) Novak (1952) Novak (1952) Novak (1952) Novak (1952) this study this study this study this study this study Scotti (1911) Scotti (1911) Scotti (1911) Hegi (1924) Novak (1952)

15

a

b

c

Figure 11. Fraxinus longicuspis, Japan, May 1999. (a) Male inflorescences with unopened anthers, (b) male inflorescences with empty anthers as the leaves begin to unfold, and (c) hermaphrodite inflorescences with unopened anthers, at the same time as in 11b.

Fraxinus longicuspis Morphology Photos of inflorescences are shown in Fig. 11. The results of measurements of male floral parts are reported in Table 7 and show that there are no significant size differences in anther, filament or stamen length between males and hermaphrodites. No rudimentary pistil was ever found in a male flower. Phenological observations Flowers of male trees opened their anthers before the hermaphrodites did (compare Figs. 11b and 11c). All but one of the seven male trees had already opened their anthers and shed their pollen at the time of arrival on the study site. The hermaphrodites are protogynous because their anthers dehisce some time (about a week) after the flowers have emerged. The question if the stigmata are receptive from the start of flower emergence could not be answered, since hermaphrodite flowers had already been open for some days at the time of arrival to the study population. Presumably, tough, the stigmata are receptive at the time of emergence as in other investigated Fraxinus species. Breeding system: androdioecy Sex ratio was 50:50 in the small population we studied (N = 14). As Table 7 shows, there are no significant differences in height and girth of trees (n = 14), or in lengths of filaments, anthers or stamens (n = 7) between males and hermaphrodites. On the other hand, there are 2.2 times more flowers (conservatively estimated) in male compared to hermaphrodite inflores16

cences (nested ANOVA, df = 1, F = 10.5, p = 0.010). The number of pollen grains per anther and pollen germination rate could only be measured in flowers from one male individual and 2-3 hermaphrodite individuals. Therefore these measurements were not subjected to statistical tests. However, there is an indication of a differential pollen quality between the two gender morphs, because five anthers from the investigated male had a 42% greater mean pollen grain volume (based on a 12% greater mean diameter) compared to the two hermaphrodites (Table 7). Although male anthers are insignificantly larger than hermaphrodite ones, since the larger male pollen grains occupy a larger space, the hermaphrodites have a slightly higher number of pollen grains per anther. Pollination system No insects were ever seen visiting the flowers and based on pollination syndrome characters, F. longicuspis shows all the typical signs of being an anemophilous species. Although the inflorescences emerge with the leaves in terminal panicles, the leaves are still quite small at the time of flowering (Fig. 11a). The non-fragrant flowers are without petals and are borne on long slender pedicels in lax, hanging inflorescences. The anthers of both male and hermaphrodite flowers contain a comparatively large amount of pollen (Table 5). The trees are tall, usually around 20 m, which is comparable to the height of many other typically wind-pollinated trees in Fraxinus (and other families). Also, being a canopy species it has a good chance of getting its pollen dispersed by the wind.

Table 7. Means and standard deviations (SD) for tree height, stamen length, pollen diameter and volume, pollen germination rate, and counts of pollen per anther and flowers per inflorescence, for males and hermaphrodites in a population of 14 trees of Fraxinus longicuspis in Japan. n = sample size (see explanations in table footnote), P = significance of difference, ns. = not significant at P < 0.05. Variables with sample size = 1 for the male group, were not tested for statistical significance. Variable

Tree height (m) Tree girth at breast height (cm) Number of flowers/inflorescence Stamen length (mm) Anther length (mm) Filament length (mm) Number of pollen grains/anther Pollen grain diameter (µm) Pollen grain volume (µm3) Pollen germination rate (%)

Male

Hermaphrodite

P

Mean ± SD

n

Mean ± SD

n

18.1 ± 3.2 73.7 ± 20.4 382 ± 115 4.46 ± 0.74 2.17 ± 0.31 2.29 ± 0.56 12404 ± 870 21.6 ± 0.8 5314 ± 551 17.8 ± 14.9

7 7 6a 4c 4c 4c 1e 1g 1g 1i

16.5 ± 3.7 72.4 ± 16.1 174 ± 75 4.58 ± 0.81 1.88 ± 0.28 2.71 ± 0.67 15923 ± 2938 19.2 ± 0.5 3732 ± 306 13.4 ± 10.0

7 7 5b 3d 3d 3d 2f 2h 2h 3j

ns. ns. 0.01 ns. ns. ns.

a

Five to eight inflorescences each from six male trees. One to eight inflorescences from five hermaphrodite trees. c Ten flowers each measured in ten inflorescences from four male trees. d Ten flowers each measured in seven inflorescences from three hermaphrodite trees. e Two anthers from one male tree. f Two anthers each from two hermaphrodite trees. g Five anthers from one male tree. h Five anthers each from two hermaphrodite trees. i Pollen from four different inflorescences in one male tree. j Pollen from two different inflorescences in each of three hermaphrodite trees. b

Other species The observations on phenotypic sex expression and measurements of floral parts on 13 of 16 taxa in section Ornus are reported in Table 8. Data on Fraxinus ornus and F. lanuginosa are not included in the table, since they have been extensively studied elsewhere (Dommée et al. 1999 and Ishida & Hiura 1998), and neither is F. baroniana, of which we had no material. Of the 13 species, we found specimens with either male or hermaphrodite flowers (never co-occurring) in ten species and only hermaphrodite flowers in three species. Although sample sizes for some of the species are small, this result is in accordance with literature data where only the first three are stated to be hermaphroditic and all the others either polygamous or androdioecious (Nakaike 1972, Yamazaki 1993, Wei & Green 1996). In section Ornus there are eleven petaliferous (entomophilous or ambophilous) spe-

cies and five apetalous (anemophilous) species. Among the petaliferous and androdioecious species, there are no significant differences between males and hermaphrodites in length of anthers, filaments and petals. But of the five apetalous species (of which F. baroniana was not studied), both F. chinensis and F. japonica, have much smaller and seemingly non-functional anthers in hermaphrodite flowers. One specimen of F. chinensis even had female flowers without visible remnants of stamens. Observations on a hermaphrodite tree of F. japonica, flowering in Göteborg Botanical Garden in 2000, showed that most of the small anthers never dehisced, but remained closed and soon fell to the ground. Rudimentary pistils were observed in male flowers of all taxa except F. longicuspis. In F. ornus the rudimentary pistil contains small, undeveloped ovules. 17

N

27

4

1

3

11

55

20

20

4

21

3

3

7

179

Species

griffithii

malacophylla

raibocarpa

apertisquamifera

bungeana

floribunda

paxiana

sieboldiana**

trifoliolata

chinensis

japonica

longicuspis

micrantha

Total

98

5

1

2

11

3

15

13

35

9

3

0

0

0

m

81

2

2

1

10*

1

5

7

20

2

0

1

4

27

h

A

A

A-D

A-D

A

A

A

A

A

A

H

H

H

B

71

2/2

5/4

3/1

5/4

2/1

1/4

7/4

9/4

3/1

2/0

-/1

-/1

-/5

n (m/h)

yes (only in some?)

yes, very small no

yes

yes, very small yes

yes

in some, very small in some, very small yes

Rudimentary pistil in males?

2.9±0.4

2.1±0.4

1.7±0.2

2.1±0.5

1.8±0.4

1.1±0.2

1.8±0.4

1.5±0.3

1.4±0.1

1.1±0.1

1.4±0.3

2.3±0.5

1.5±0.4

1.1±0.3

3.9±1.0

0.5±0.1

3.0±0.7

2.9±0.7

4.2±1.1

2.5±1.5

5.4±1.6

1.9±0.2

3.8±0.7

3.0±0.6

5.5±1.4

3.7±1.5

Male flowers Anther FilaPetal length ment length length

4.4±0.6

4.4±0.8

3.3±0.5

3.2±0.5

5.8±1.3

1.6±0.2

4.8±0.9

4.4±0.8

5.6±1.0

3.5±1.5

Stamen length

3.0±0.4

3.2±0.4

2.8±0.2

2.8±0.6

2.6±0.4

4.7±0.8

2.3±0.2

2.5±0.5

5.3±0.3

no data

2.6±0.3

1.2±0.3

1.4±0.3

Pistil length

2.8±0.2

1.9±0.2

1.0±0.2

1.7±0.9

1.9±0.3

1.1±0.1

1.6±0.3

1.7±0.2

1.6±0.2

no data

2.7±0.2

2.4±0.2

2.5±0.3

1.5±0.4

2.2±0.9

0.4±0.1

1.2±0.7

3.8±0.4

3.3±1.0

3.2±0.5

4.0±0.6

5.1±0.7

no data

4.2±0.3

1.2±0.3

1.3±0.3

4.3±0.5

5.6±0.8

3.7±0.6

3.9±0.5

9.8±0.9

no data

6.4±0.8

2.8±0.3

3.0±0.3

Hermaphrodite flowers Anther Filament Petal length length length

4.2±0.5

4.0±1.0

1.3±0.2

2.8±1.4

5.7±0.7

4.4±1.1

4.7±0.6

5.7±0.8

6.6±0.8

no data

6.9±0.3

3.5±0.3

3.8±0.5

Stamen length

Table 8. The phenotypic gender expression based on 179 flowering specimens of Fraxinus, belonging to 13 of 16 recognised species in section Ornus. Measurements of anther, filament, petal (in petalous species), and pistil length (only for hermaphrodites) were taken on 71 available specimens and averaged separately for the sexes of each species. Based on these observations and measurements, combined with data from the literature, ten species were considered to be morphologically androdioecious and 3 hermaphroditic. Mean lengths in mm ± SD. Stamen length is anther plus filament length. * denotes a few female flowers found. ** The flowers on the only measured male specimen of F. sieboldiana were somewhat immature. N = number of flowering specimens seen, m = male individual, h = hermaphrodite individual, B = phenotypic breeding system, n = number of measured specimens (number of males/number of hermaphrodites), A = androdioecy, A-D = androdioecious but functionally dioecious.

DISCUSSION Fraxinus excelsior – polygamous In accordance with other authors we have found that the breeding system of F. excelsior is quite complicated, and that the sex expression varies at the floral, inflorescence, and tree level, as well as between years. In an attempt to bring some order in the terminology regarding sex expression, we classified the different, but to some extent intergrading, inflorescence types into four categories depending on combinations of floral types. We also tried to assign a functional gender to an individual, but because trees can bear more than one type of inflorescence they did not always fall into neat categories. Gender (maleness or femaleness) is a quantitative phenomenon in plants (Lloyd 1980). Phenotypic gender describes male and female functions in the initial investment of parental resources, whereas functional gender (realised gender) of an individual estimates it relative success as either paternal or maternal parent. Since functional gender describes the proportion of an individual’s genes that are transmitted through pollen (its maleness) or through ovules (its femaleness), it cannot be estimated for an individual without reference to the interbreeding population that it is part of. Although we have estimates of differential pollen quality among male and hermaphrodite flowers, we cannot estimate the relative contributions of each sex morph to the next generation. Thus, we have no ideal way of describing functional gender and we have therefore used the term phenotypic gender expression to refer to predominantly male, hermaphrodite and female individuals (Table 2). The phenotypic gender of an individual varies to some extent among years. However, the functional gender may not change much since most of the variation in sex expression is within a narrow range of the continuum. It is mainly only males (variant II) that vary in their number of mixed inflorescences and/or the number of hermaphrodite flowers in those inflorescences (i.e. variation between male II and hermaphrodite III). Some female trees varied in the degree of rudimentary stamens (which are not very fertile anyway). Hermaphrodite individuals (variants I and II), whether they are functionally more male or more female, do not vary much. Modification of both phenotypic and

functional gender may take place as a result of an individual’s circumstances, which may be a function of both the external environment and a plant’s internal resource status (Lloyd and Bawa 1984). Lloyd and Bawa distinguish two patterns of gender modification, based on the extent of departure from the mean gender of a class of plants: (1) Gender adjustments, in which gender varies continuously about one modal value in response to environmental or status signals, or (2) phase choices, in which the distribution of gender is bimodal and individuals choose between discrete phases (male or female forms) according to conditions. In Fraxinus excelsior it appears that the pattern may be described as gender adjustments. As Larsen (1945) also noted, F. excelsior is protogynous, not protandrous as Wardle (1961) and people citing him report. The stigmata in hermaphrodite flowers are receptive for about a week before their own anthers open so, theoretically, there is ample time for cross-pollination. But since the duration of the pre-male phase is dependent on weather and temperature during flowering, there could be anywhere from total overlap of gender phases to up to two weeks delay of the male phase within a tree. However, the possibility of self-pollination in hermaphrodites does not necessarily imply self-fertilisation, as relative rates of self versus outcross pollen tube growth may differ. The hermaphrodites were shown to be self-compatible but in competition between self and outcross pollen, the latter may have higher fertilisation success if they have faster growing pollen tubes. We tested the germination capacity and found that it differed significantly between pollen from males, hermaphrodites and ‘females’, in decreasing order. Pollen from females with rudimentary anthers has very low viability and no competitive ability. Apart from more competitive pollen, males also produce more pollen (more pollen per anther and more flowers per inflorescence). Pollen from male trees may thus have an advantage over selfpollen in hermaphrodite flowers, both in terms of number and in terms of quality. Of course, pollen from other hermaphrodite trees may also have higher fertilisation success than self-pollen, but overall, pollen from males seem to have greater advantage. When the number of pollen grains deposited on a stigma is greater than the number of ovules in the ovary, competition for ovules can 19

result and the fastest growing pollen tubes may be the ones resulting in fertilisation (Lee 1988). An important aspect of this is the fact that there are no mature megagametophytes in the ovules until after the stigmata are no longer receptive. This makes sexual selection through pollen tube growth possible, even for those pollen grains that arrive late to the stigma (Dahl & Fredrikson 1996). An interesting comparison in this regard is the observation that in the ambophilous F. ornus, the megagametophytes are already developed while the stigmata are receptive. The significance of this condition may indicate that there is a selection pressure on fruiting individuals of F. excelsior (and perhaps other windpollinated species in this group) for delayed fertilisation, that increases the chances for multiple cross-pollination and selection among a greater number of male gametes. In entomophilous plants, that generally receive many pollen grains per pollen load, this selection pressure may not be strong since only one pollinator visit may bring more pollen than there are ovules. Wind-pollinated plants, on the other hand, are generally believed to be pollen limited and if this is the case, it would be advantageous to delay fertilisation. This is because the possibilities for sexual selection among male gametes are increased if a greater number of pollen grains are allowed to accumulate on the stigma or in the style, before any of them is allowed to access the ovules. The flowering and fruiting pattern over the seven years seem to indicate a 2 or 3-year cycle (Fig. 4). However, the synchronicity among trees within years in the study population (Table 3) and the correlation between the flowering rate and intensity in the study population and the regional airborne pollen count (Fig. 6) suggest that the flowering behaviour to a large extent may be dependent on climatic factors. Tapper (1992 and 1996), on the other hand, found an irregular fruiting cycle that was neither negatively correlated with fruiting the previous year nor positively related to climatically favourable years. Instead, date of leafing the previous year seemed to determine whether individual trees flowered or not.

Fraxinus ornus – ambophilous and functionally androdioecious The flowers of Fraxinus ornus are mostly regarded as entomophilous. The strongly sweet20

scented flowers produce copious amounts of pollen (Table 5) and during peak flowering they are visited by a number of different pollen-feeding insects, mainly small beetles (Table 6). Whether these visitors also act as pollinators is not known. However, species of Scarabaeidae are known to be effective pollinators over long distances, e.g. in Cistus (Cistaceae) (Bosch 1992) and Viburnum (Caprifoliaceae) (Englund 1993), and they may be important pollinators of F. ornus as well. The Hymenopteras are known to be pollen feeders and may also act as pollinators. Many of the other flower visitors belong to what is called the “rendez-vous guild”, i.e. insects gathering at attractive flowers or inflorescences to find mates (Thomas Appelqvist, Göteborg University, pers. comm.). These visitors may be contributing mostly to geitonogamous pollination. Dommée et al. (1999) state that F. ornus is nectariferous, but I have not seen nectar in any of the flowers, and related species, e.g. Fraxinus lanuginosa (Ishida & Hiura 1998), do not have nectar. Although the inflorescences attract a lot of different non-specialist pollinators, wind may still play a significant role in pollination because of the high pollen production (Table 5) and the fact that a lot of pollen was also found to be spread by wind. Mincigrucci et al. (1987) report high levels of F. ornus pollen in the air in Perugia, Italy, and Scotti (1911), although he mostly observed beetles on the flowers, believed the principal pollination of this species to be by wind. Ilardi & Raimondo (1999) in their review state that “the anthers stand out between the petals favouring both entomophilous and anemophilous pollen dispersion”. All this points to the fact that the flowers of F. ornus are ambophilous, i.e., rely on both wind and insects for pollination. The ambophilous strategy is also documented for a related species, F. lanuginosa (Ishida & Hiura 1994). Ambophily is also known in other families and genera, e.g., Salix (Vroege & Stelleman 1990, Tollsten & Knudsen 1992). During the studies, F. ornus was discovered to be morphologically androdioecious. This was a surprising finding, because up till then nothing was found in the literature that suggested this. Independently, this condition was also found by Dommée et al. (1999) who presented evidence showing that populations in France were functionally androdioecious. Their study included comparative measurements of

Table 9. Comparison of sex ratios and mean number of flowers per inflorescence (±SE) in male and hermaphrodite trees of three species in Fraxinus section Ornus. Data on number of flowers are not available from F. ornus. Species

Sex ratio

Male

14-66% males

490.4 ± 48.8 (n = 16) 160.7 ± 13.1 (n = 19) 401.3 ± 20.9 (n = 107) 153.8 ± 5.9 (n = 125)

Ishida & Hiura (1994) Hiura & Ishida (unpubl.)

F. longicuspis

50% males (n = 14)

382.3 ± 47.1 (n = 6)

174.4 ± 33.4 (n = 5)

this study

F. ornus

48-60% males (n = 319) no data 45-65% males (n = 85)

no data

Dommée et al. (1999) this study

F. lanuginosa

Hermaphrodite

Reference

traits related to male function (size of staminate floral structures and pollen grains), sizes of trees, sex ratios and controlled cross- and selfpollinations. They found no significant differences in size of trees and floral traits related to male function in the two genders. Although we never took any measurements, their data are in line with our field observations. Similarly, they found an average of 54% males in five populations (Table 9), and no evidence of gender change. Thus, our results are consistent with theirs and gender seems to be genetically determined. Charlesworth (1984) believes functional androdioecy to be extremely rare and suggests that hermaphrodites in most cases actually are functionally female with dysfunctional pollen. This is not the case in F. ornus, however, where hermaphrodites have been shown to produce viable pollen (Dommée et al. 1999). Dommée et al. also report of unpublished data on pollen germination tests showing that there is no difference in germination capacity of pollen from males and hermaphrodites. Pollen from hermaphrodites is able to produce seed when crossed with other hermaphrodites, and the trees are even self-compatible. No difference in siring success between pollen from males and hermaphrodites was detected. Thus it seems that hermaphrodites of F. ornus do have a male function.

on one male and two hermaphrodites, showed no large differences. This may indicate that hermaphrodites produce potentially functional pollen. However, pollen grains from the measured male had a 42% larger mean pollen grain volume. No crossing experiments were conducted, but if larger pollen grain size is correlated with faster growing pollen tubes (e.g., through more nutrients), male pollen could outcompete hermaphrodite pollen on a stigma or in the style. However, the most striking difference between the gender morphs is the fact that males have more than twice as many flowers per inflorescence as hermaphrodites. This, in combination with the observation that male trees seem to open their anthers earlier and that the hermaphrodite trees are protogynous (as in F. ornus and other species of Fraxinus), makes it possible for pollen from males to pollinate the stigmata and fertilise the ovules before the hermaphrodite’s own pollen is released. Only if no cross-pollination has taken place may selfing eventually occur. All this could contribute to higher fitness of males and explain their maintenance in the population. If hermaphrodites are self-compatible (which many species of Fraxinus are), the possession of anthers may be viewed as a reproductive assurance in this wind-pollinated tree.

Fraxinus longicuspis – functionally androdioecious?

Although androdioecy is considered a very rare breeding system (Charlesworth 1984, Richards 1997), there are quite a number of occurrences in the Oleaceae family. In Fraxinus, phenotypic androdioecy occurs in at least 10 species of section Ornus (Table 8) and in several other genera in Oleaceae as well (Wallander ms. 2). The measurements of floral parts in section Ornus did, for most of the taxa, not show any signifi-

The preliminary field study of F. longicuspis indicated no differences in size of individuals and male organs between the two gender morphs (Table 7). Although a very small sample size, the preliminary data on germination rate and number of pollen grains per anther, measured

Androdioecy in Fraxinus

21

Table 10. Comparison of possible fitness superiority in male function of males over hermaphrodites in three functionally androdioecious species of Fraxinus. Data on F. lanuginosa from Ishida & Hiura 1998 and Hiura & Ishida 1994, and on F. ornus from Dommée et al. 1999. Species

More male flowers/

More male pollen/anther? inflorescence?

Larger male pollen grains?

Greater male pollen germinability?

Greater male siring success?

F. ornus F. lanuginosa F. longicuspis

? yes yes

no no no

no maybe yes

no yes no

no yes ?

cant differences between males and hermaphrodites regarding length of stamens, anthers, filaments, pistils, and petals. This speaks for potentially functional male organs in the hermaphrodite flowers. But in order to conclude whether the pollen of hermaphrodites performs as well as pollen from the males, this needs to be tested in pollination experiments both in vitro and in vivo. Fraxinus longicuspis is the first documented case of an androdioecious and wind-pollinated species within Fraxinus. There is another case of an androdioecious and wind-pollinated species in the family Oleaceae: the Mediterranean shrub Phillyrea angustifolia L. (Lepart & Dommée 1992, Aronne & Wilcock 1992, Vassiliadis 1999, Pannell & Ojeda 2000). Outside Oleaceae, three wind-pollinated herbs have been shown to be functionally androdioecious, Datisca glomerata (Presl) Baill. (Datiscaceae) (Liston et al. 1990), Mercurialis annua L. (Euphorbiaceae) (Pannell 1997) and Schizopepon bryoniaefolius (Cucurbitaceae) (Akimoto et al. 1999). The two species of Fraxinus that have proved to be functionally androdioecious are the ambophilous F. ornus (see above) and Fraxinus lanuginosa (Ishida & Hiura 1998). According to evolutionary stable strategy (ESS) models developed by Lloyd (1975) and Charlesworth (1984), the evolution and maintenance of androdioecy requires that males have at least twice the fitness of the male function in hermaphrodite plants. Are there any indications that such differences exists in the three androdioecious Fraxinus species? In Table 10 we compare some characteristics of male function in the three species. There are no differences in anther size and pollen content between the sexes in either of the species, but in both F. lanuginosa and F. longicuspis there are more than twice as many flowers per inflorescence in males as compared to hermaphrodites 22

(Table 9). Unfortunately, unprepared as we were for this aspect at the field site of F. ornus, our collected material was not complete enough to allow counts of number of flowers per inflorescence to be made, and Dommée et al. (1999) did not count the number of flowers per inflorescence either. However, observations on photos and voucher material indicate no differences. In F. ornus there is no detected difference in pollen viability and siring success between males and hermaphrodites. In F. lanuginosa there is no difference in pollen grain volume (Ishida & Hiura 1998) (although we found a significant difference in size in our small sample, see Table 5), but male pollen shows a significantly higher germination rate than hermaphrodite pollen, and pollination experiments have also shown that male siring success is higher. In contrast, preliminary data on F. longicuspis show that males have larger pollen grains, but no difference in germination rate in vitro was detected. Thus it seems that males in these species may have the necessary pollen fertility required by the ESS models. On the other hand, unisexual males may have other functional advantages over hermaphrodites that lowers the stringent demands on male fertility predicted by the ESS models. This hypothesis is developed further in Wallander (ms. 2). The three androdioecious species of Fraxinus have a sex ratio of about 50% males (Table 9). This value cannot be explained by the phenotypic selection models of Lloyd (1975) and Charlesworth and Charlesworth (1978), as suggested by Vassiliadis et al. (2000). On the other hand, if these species have gametophytic selfincompatibility linked with a sex determination locus, then the sex ratio may be explained by the frequency dependent selection model of Vassiliadis et al. (2000). Thus we need further studies to clarify whether the species has gametophytic self-incompatibility.

In summary, in most morphological variables there are no differences between males and hermaphrodites in phenotypically androdioecious species. However, in at least two of the functionally androdioecious species (F. lanuginosa and F. longicuspis) males produce more pollen (through more flowers per inflorescence) and in at least F. lanuginosa male pollen have higher siring success, and the case might be the same in F. longicuspis. Thus, males in these two species have a higher paternal success than hermaphrodites. In F. ornus, however, there is yet nothing that shows how males would attain the required doubled fitness. Even with equal pollen quality in the two genders, male pollen might be at an advantage because of the indication that male trees flower earlier and the hermaphrodites are protogynous for about a week. This hypothesis needs quantification and testing before determining its importance for the maintenance of males in this breeding system. In connection with this, we wish to point out that although these results point towards superior male function in males, the hermaphrodites still retain a potential male function. Thus, they are not functionally female and they have a potential to sire seeds.

Comparisons between polygamy and androdioecy in Fraxinus In Fraxinus, polygamy and androdioecy appear to be similar strategies in the anemophilous and ambophilous taxa. It is true that there is a difference in that gender is stable in the androdioecious taxa but rather continuous in the polygamous, and that the different forms of flowers never coexist in an individual of an androdioecious species. But apart from the constancy of sex expression, the function of hermaphrodite flowers in androdioecious species appears to be similar to the continuum of hermaphrodites and females with rudimentary stamens in the polygamous taxa. Both groups of species have adaptations to pollination by wind, a rather unpredictable pollination mode that may not always result in cross-pollination. In such cases, it would be adaptive for ‘females’ to retain functional stamens as a means of reproductive assurance, especially in colonising populations (Dommée et al. 1999). Another explanation may be that selection pressures for reducing male function in ‘female’ flowers are

not as strong as the selection pressures that act on reducing fruiting ability for optimal male function (Wallander ms. 1 and 2). However, the selection pressures for a pure female function may be stronger in anemophilous than ambophilous species. Some reasons for this, e.g. structural conflicts in anemophiles versus pollinator attraction in ambophiles, are developed further in Wallander (ms. 2). As stated before, despite the proven superior male function in males contra cosexuals in several species of both breeding systems, it would be wrong to call them functionally dioecious since the hermaphrodites do have a male function. Even though hermaphrodite pollen may have low fertilisation success in competition with male pollen on a stigma or in the style, without this competition it would occasionally sire seeds on other individuals and hermaphrodite flowers would sometimes even self.

Conclusions and future studies Although the long-term study on Fraxinus excelsior has attempted to answer some of the questions about its seemingly complicated polygamous breeding system, a number of unanswered questions still remain. The floral morphology has been quantified, but crossing experiments are needed in order to determine the function of the different sex morphs in a population. We have shown that hermaphrodite flowers are self-compatible and protogynous, but the extent of actual self-pollination and selffertilisation is unknown. The flowering and fruiting pattern during this seven-year study indicated a 2-3 year cycle, which seems to have some sort of environmental component determining the apparent synchronicity among trees. An ongoing study will try to identify the proximate factors determining the flowering and fruiting pattern. Our preliminary results suggest that F. longicuspis is functionally androdioecious but further field studies, e.g., crossing experiments, are needed in order to determine the male function in hermaphrodites. In F. ornus, all present evidence points towards functional androdioecy, but one important piece of information is still missing. It would be very valuable to compare the number of flowers in male and hermaphrodite inflorescences. In contrast to most of the previous in23

formation in the literature, but in accordance with observations on floral morphology, F. ornus was shown to be ambophilous. However, the relative importance of insects and wind in the pollination of this species is unknown. For example, pollen dispersal distances are unknown, as well as the extent to which the pollen-feeding insects actually contribute to any cross-pollination. The incidence of phenotypic androdioecy in Fraxinus is high and so far two species (or three if including F. longicuspis) have been shown to be functionally androdioecious. The common feature for all androdioecious species in Fraxinus (and other cases documented hitherto as well) is that they are ambophilous or anemophilous. This is no coincidence and the significance of this pollination system for the evolution and maintenance of androdioecy will be discussed in relation to phylogenetic data in a forthcoming paper.

ACKNOWLEDGEMENTS We thank Margit Fredrikson and Jan Helgesson for assistance in the lab, and Cilla Odenman and Karin H. Persson for assistance in the field in Sweden. For assistance in the field in Sicily we thank Kerstin Helgesson and for local support Giuseppe Venturella and Pietro Mazzola at the University of Palermo, and Giuseppe Piro, Maria Cinquegrani, and Gioacchino Genchi for help to find populations of F. ornus. For help with determining insects collected in flowers of F. ornus we thank Thomas Pape and colleagues at the Swedish Museum of Natural History. For help in locating and translating Novak’s information about insects on F. ornus we are indebted to Boze Kokan (Natural History Museum Split, Croatia). For help with carrying out the field studies in Japan and for permission to include the preliminary results on F. longicuspis in this paper, we wish to thank Tsutom Hiura (Tomakomai Research Station, Hokkaido University Forests) and Kiyoshi Ishida (Forestry and Forest Products Research Institute, Kyoto). We also thank the staff at the Tokyo University Forest at Chichibu for all valuable help in the field, especially locating and climbing the trees! We thank Lennart Andersson, Kiyoshi Ishida, Ulf Molau, and Johan Wallander for helpful comments on the manuscript, the lat24

ter two also for help with the statistics. Funding for this doctoral research has come from The Royal Swedish Academy of Sciences, Kungliga & Hvitfeldtska Stiftelsen, Wilhelm & Martina Lundgrens Vetenskapsfond, Helge Ax:son Johnsons stiftelse, Stiftelsen Botaniskas Vänner, Kungliga Vetenskaps- & VitterhetsSamhället in Göteborg, Uddenberg-Nordingska Stiftelsen, and Collianders stiftelse.

REFERENCES Akimoto, J. & Fukuhara, T. & Kikuzawa, K. 1999. Sex ratios and genetic variation in a functionally androdioecious species, Schizopepon bryoniaefolius (Cucurbitaceae). American Journal of Botany 86: 880-886. Aronne, G. & Wilcock, C.C. 1992. Breeding system of Phillyrea latifolia L. and Phillyrea angustifolia L: evidence for androdioecy. Giornale Botanico Italiano 126: 263. Binggeli, P. & Power, A. J. 1991. Gender variation in ash (Fraxinus excelsior L.) In: Proceedings of the Irish Botanist Meeting, p 42. University College Dublin, Dublin. (Abstract). Bosch, J. 1992. Floral biology and pollinators of three co-occurring Cistus species (Cistaceae). Botanical Journal of the Linnean Society 109: 39-55. Brewbaker, J. L. & Kwack, B. H. (1963) The essential role of calcium ion in pollen germination and pollen tube growth. American Journal of Botany 50: 747-858. Charlesworth B., Charlesworth D. 1978. A model for the evolution of dioecy and gynodioecy. American Naturalist 112: 975-997. Charlesworth, D. 1984. Androdioecy and the evolution of dioecy. Biological Journal of the Linnean Society 22: 333-348. Dahl, Å. E. & Fredrikson, M. 1996. The timetable for development of maternal tissues sets the stage for male genomic selection in Betula pendula (Betulaceae). American Journal of Botany 83: 895-902. Dommée, B. & Geslot, A. & Thompson, J. D. & Reille, M. & Denelle, N. 1999. Androdioecy in the entomophilous tree Fraxinus ornus (Oleaceae). New Phytologist 143: 419-426. Englund, R. 1993. Movement patterns of Cetonia beetles (Scarabaeidae) among flowering Viburnum opulus (Caprifoliaceae). Oecologia 94: 295-302. Galen, C. & Plowright, R. C. 1987. Testing the accuracy of using peroxidase activity to indicate stigma receptivity. Canadian Journal of Botany 65: 107111. Gardner, G. 1977. The reproductive capacity of Fraxinus excelsior on the Derbyshire limestone. Journal of Ecology 65: 107-118. Hegi, G. 1924. Oleaceae - Fraxinus L. Illustrierte Flora von Mittel-Europa 5: 1919-1934.

Hiura, T. & Ishida, K. 1994. Reproductive ecology in Fraxinus lanuginosa. I. Breeding structure. Transactions of the Hokkaido Branch of the Japanese Forest Society 42: 58-60. Hyde, H. A. 1951. Pollen output and seed production in forest trees. Quarterly Journal of Forestry 45: 172-175. Ilardi, V. & Raimondo, F. M. 1999. The genus Fraxinus L. (Oleaceae) in Sicily. Flora Mediterranea 9: 305318. Ishida, K. & Hiura, T. 1994. Reproductive ecology in Fraxinus lanuginosa. II. Pollen vectors and pollen viability. Transactions of the Hokkaido Branch of the Japanese Forest Society 42: 61-63. Ishida, K. & Hiura, T. 1998. Pollen fertility and flowering phenology in an androdioecious tree, Fraxinus lanuginosa (Oleaceae), in Hokkaido, Japan. International Journal of Plant Sciences 159: 941-947. Lamb, A. & Boshier, D. 1994. The reproductive biology of common ash (Fraxinus excelsior): a literature review. Oxford Forestry Institute, University of Oxford. Larsen, C. S. 1945. Ash flowering and improvement. Dansk Skovforenings tidskrift 3: 49-89. Lee, T. D. 1988. Patterns of fruit and seed production. In: Lovett Doust, J. & Lovett Doust, L. (eds.) Plant reproductive biology: patterns and strategies, pp. 179-202. Oxford University Press. Lepart, J. & Dommee, B. 1992. Is Phillyrea angustifolia L. (Oleaceae) an androdioecious species? Botanical Journal of the Linnean Society 108: 375-387. Liston, A. & Rieseberg, L. H. & Elias, T. H. 1990. Functional androdioecy in the flowering plant Datisca glomerata. Nature 343: 641-642. Lloyd, D. G. 1975. The maintenance of gynodioecy and androdioecy in Angiosperms. Genetica 45: 325-339. Lloyd, D. G. 1980. Sexual strategies in plants. III. A quantitative method for describing the gender of plants. New Zealand Journal of Botany 18: 103108. Lloyd, D. G. & Bawa, K. J. 1984. Modification of the gender of seed plants in varying conditions. Evolutionary Biology 17: 255-338. Mincigrucci, G. & Bricchi, E. & Romano, B. & Frenguelli, G. 1987. Studio preliminare sulla pollinazione di Oleaceae. Annali della Facoltà di Agraria XLI: 929-940. Nakaike, T. 1972. A synoptical study on the genus Fraxinus from Japan, Korea and Formosa. Bulletin of the National Science Museum of Tokyo 15: 475-512. Novak, P. 1952. Kornjasi Jadranskog Primorja (Coleoptera). Jugoslavenska Akademija Znanosti i Umjetnosti, Zagreb, Croatia (now Croatian Academy of Science and Art). Pannell, J. 1997. Widespread functional androdioecy in Mercurialis annua L. (Euphorbiaceae). Biological Journal of the Linnean Society 61: 95-116. Pannell, J. R. & Ojeda, F. 2000. Patterns of flowering

and sex-ratio variation in the Mediterranean shrub Phillyrea angustifolia (Oleaceae): implications for the maintenance of males with hermaphrodites. Ecology Letters 3: 495-502. Picard, J.-F. 1982. Contribution a l’étude de la biologie florale et de la frutification du Frêne Commun (Fraxinus excelsior L.). Revue Forestiere Francais 34: 97-107. Porsch, O. 1958. Alte Insektentypen als Blumenausbeuter. Österreichische Botanische Zeitschrift 104: 115-164. Richards, A. J. 1997. Plant breeding systems. 2nd edition. Chapman & Hall, London. Rohmeder, E. 1952. Untersuchungen über die Verteilung der Geschlechter bei den Blüten von Fraxinus excelsior. Forstwissenschaft Centralblatt 71: 17-29. Sawada, T. 1937. Plantae Hakonenses Sawadanae. Journal of Japanese Botany 13: 29-33. Schultz, A. 1892. Beiträge zur Morphologie und Biologie der Blüten: Fraxinus excelsior L. Berichte der Deutschen Botanischen Gesellschaft 10: 395-409. Scotti, L. 1911. Contribuzioni alla Biologia fiorale delle “Contortae”. Annali di Botanica 9: 199-314. Tapper, P-G. 1992. Irregular fruiting in Fraxinus excelsior. Journal of Vegetation Science 3: 41-46. Tapper, P-G. 1996. Long-term patterns of mast fruiting in Fraxinus excelsior. Ecology 77: 2567-2572. Tollsten, L. & Knudsen, J. 1992. Floral scent in dioecious Salix (Salicaceae) – a cue determining the pollination system? Plant Systematics and Evolution 182: 229-237. Vassiliadis, C. 1999. Evolution et maintien de l’androdioécie: Etude théorique et approaches expérimentales chez Phillyrea angustifolia L. PhD Dissertation, Université des Sciences at Technologies de Lille, France. Vassiliadis, C. & Valero, M. & Saumitou-Laprade, P. & Godelle, B. 2000. A model for the evolution of high frequencies of males in an androdioecious plant based on a cross-compatibility advantage of males. Heredity 85: 413-422. Vroege, P. W. & Stelleman, P. 1990. Insect and wind pollination in Salix repens L. and Salix caprea L. Israel Journal of Botany 39: 125-132. Wallander, E. Evolution of pollination and breeding systems in Fraxinus (Oleaceae), and a new classification. (manuscript 1) Wallander, E. Evolution of wind-pollination and gender specialisation in Oleaceae – exaptations and adaptations. (manuscript 2) Wallander, E. & Albert, V. A. 2000. Phylogeny and classification of Oleaceae based on rps16 and trnLF sequence data. American Journal of Botany 87: 1827-1841. Wardle, P. 1961. Biological Flora of the British Isles: Fraxinus excelsior L. Journal of Ecology 49: 739-751. Wei, Z. & Green, P. S. (1996) Fraxinus. In: Wu, Z. & Raven, P. H. (eds). Flora of China 15: 273-279. Science Press & Missouri Botanical Garden. Yamazaki, T. 1993. Fraxinus. In: Iwatsuki et al. (eds.) Flora of Japan IIIa: 124-128. Kodansha Ltd., Japan.

25

Paper I

Paper II

Paper III

Paper IV

Paper V

Paper VI

Evolution of pollination and breeding systems in Fraxinus (Oleaceae), and a new classification Eva Wallander

ABSTRACT Phylogenetic relationships among 40 of the 43 recognised species of Fraxinus L. (Oleaceae) were estimated in order to trace the evolution of pollination and breeding systems in the genus. Cladistic analyses of nuclear ribosomal ITS sequences, complemented in part with sequences from two noncoding chloroplast regions (the rps16 intron and trnL-F region), resulted in a well-resolved tree. Although the molecular support for the branching order among the main clades is low, they have themselves high molecular and morphological support. A revised classification of the genus is given, consisting of three sections and four subsections. Anemophily has three independent origins, two of them correlated with a shift in breeding system. In one clade, wind-pollinated species have evolved dioecy via androdioecy in ambophilous species. In another wind-pollinated clade, dioecy has evolved twice via polygamy. Evolution of dioecy in anemophilous species is interpreted mainly as a result of selection for sexual specialisation. Key words: androdioecy, anemophily, breeding system, dioecy, Fraxinus, ITS, Oleaceae, phylogeny.

INTRODUCTION The genus Fraxinus, the ashes, comprises 43 species occurring in temperate and subtropical regions of the northern hemisphere. It is one of 24 extant genera of Oleaceae (the olive family) and sole member of the subtribe Fraxininae, sister group to Oleinae in the tribe Oleeae (Wallander and Albert 2000). The genus was described by Linnaeus in 1753 and since then over 400 taxa have been described, most of which, obviously, are synonyms. The latest monograph of the genus (Lingelsheim 1920) lists 63 species, and more taxa have been described since then. Some species are economically important as timber, e.g., F. excelsior, F. americana, and F. mandshurica, and some are cultivated for manna production, e.g. F. ornus (authors of names are given only if not listed in Table 2 or 3).

The genus is monophyletic and unique in the Oleaceae by mostly having relatively large imparipinnate leaves and one-seeded samaras. Most of the species are trees, but some are shrubs in dry areas. There is much variation in leaf morphology (shape, number of leaflets, petiolules, indumentum, papillae, rachis wings, etc.) and variation in these features has been the cause of most synonyms. As is characteristic of nearly all taxa of Oleaceae, the small flowers have only one pistil and two stamens. The corolla may be wanting (apetalous) or consists of four (rarely two) white, free (rarely fused) petals. The calyx is small, cup-shaped, and usually dentate, or wanting. Petaliferous flowers are (with two exceptions) borne in large showy panicles, emerging with the leaves from terminal buds. Apetalous flowers, which are wind1

pollinated, occur in lateral or terminal inflorescences and emerge before the leaves unfold. The syncarpous ovary contains four ovules, but develops into a one-seeded samara. Existing infrageneric classifications are shown in Table 1. In all previous classifications, the circumscription of each group has remained more or less the same, although different authors have chosen different ranks and names for the groups. Lingelsheim (1920) provided the first monographic revision of the entire genus and divided it into two sections, Fraxinaster (with lateral inflorescences) and Ornus (with terminal). Ornus was further divided into the subsections Euornus (with petals) and Ornaster (without petals). Fraxinaster, consisting mainly of apetalous taxa, was divided into five subsections based on calyx (present or absent), inflorescence type (panicle or raceme), leaf rachis morphology (winged or not), and bud scale morphology. He also described many new species and infraspecific taxa (Lingelsheim 1907, 1920). Most subsequent authors have only regional coverage and have left out one or more sections from their treatments. Anemophily (wind-pollination) is a derived condition within the angiosperms and has evolved independently in several families (Whitehead 1968). It is correlated with a number of environmental and morphological characters, such as temperate climate, deciduousness, unisexual flowers (monoecy or dioecy), smooth pollen, and high pollen-to-ovule (P/O) ratio. Nowadays there is no doubt that anemophily is derived from zoophily (animal pollination) in angiosperms (Cox 1991). However, it is unclear which characters that generally preceded the transition to anemophily, which were coincidental with or evolved later as a consequence of that system (Linder 1998). The first type of trait has been termed exaptation (preadaptation) by Gould and Vrba (1982), and is interpreted as those characters that facilitate or are a prerequisite for a transition. Traits that follow in evolutionary sequence have evolved in response to anemophily and are adaptations for increasing the efficiency and function of that system. Particularly, in this regard, it is unclear whether dioecy, which is strongly correlated with anemophily, evolves prior to or after the shift to anemophily (Charlesworth 1993).

2

The genus Fraxinus has a variety of pollination and breeding systems, and combinations thereof, and offers an interesting case for studying evolution of traits related to reproductive biology. About one third of the taxa are entomophilous (insect-pollinated), most of which occur in Eurasia (two in North America), and two thirds are anemophilous and occur in both North America and Eurasia (Table 4). Only five of the insect-pollinated taxa are hermaphrodites and the rest are described as androdioecious, i.e., having male and hermaphrodite flowers on separate trees. This is considered to be a very rare breeding system (Charlesworth 1984), but in Oleaceae, and particularly in Fraxinus, there are several species that are morphologically androdioecious. Lately, three species have been shown to be functionally androdioecious: F. lanuginosa (Ishida and Hiura 1998), Fraxinus ornus (Dommée et al. 1999), and Phillyrea angustifolia L. (Vassiliadis 1999). About half of the number of wind-pollinated taxa are polygamous (male, hermaphrodite and female flowers on the same or different trees) and the other half, particularly the North American taxa, are dioecious (separate male and female trees). Some of the dioecious taxa have vestigial organs of the opposite sex and it is clear that they, as all other dioecious angiosperms, have arisen from hermaphrodite ancestors — but by which route? In 1997, Jeandroz et al. published the first molecular phylogeny of Fraxinus, which was based on ITS-1 and ITS-2 sequences of 20 species. Since they only included half the number of species, and left out representatives of the sections Pauciflorae and Sciadanthus, their phylogeny could not be used for the purpose of this study. Although they used their phylogenetic tree to map some floral characters (presence or absence of calyx and/or corolla), a number of errors in character coding made their interpretation of character evolution incorrect. The initial purpose of the present study was twofold: first, to estimate the phylogeny of the entire genus Fraxinus based on molecular data and, second, to use this phylogenetic estimate to study the evolution of wind-pollination and related traits in the genus. In the process, interesting relationships among taxa were discovered which led to a revised infrageneric classification.

Table 1. Different classification schemes of Fraxinus, including the revised one proposed in this study. Mistakes of author citation are shown within quotation marks and corrected within square brackets. Vassiljev (1952) and Wei (1992) did not include Dipetalae or Pauciflorae because representatives from these sections are not native to the regions they covered.

Lingelsheim (1920), Rehder (1951), Miller (1955) sect. Fraxinaster DC. subsect. Bumelioides “Endl.” [(Endl.) Lingelsh.] subsect. Melioides “Endl.” [(Endl.) Lingelsh.] subsect. Sciadanthus “Coss. et Dur.” [(Coss. et Dur.) Lingelsh.] subsect. Dipetalae Lingelsh. subsect. Pauciflorae Lingelsh. sect. Ornus (Neck.) DC. subsect. Euornus “Koehne et Lingelsh.” [Lingelsh.] subsect. Ornaster “Koehne et Lingelsh.” [(Koehne et Lingelsh.) Lingelsh.] Vassiljev (1952) subgenus Fraxinaster (DC.) V. Vassil. sect. Melioides “(Endl.) V. Vassil.” [(Endl.) Pfeiff.] sect. Bumelioides “(Endl.) V. Vassil.” [(Endl.) Pfeiff.] subgenus Ornus “(DC.) V. Vassil.” [(Boehm.) Pers.] sect. Euornus “(Koehne et. Lingelsh.) V. Vassil.” [Koehne et. Lingelsh.] sect. Ornaster “(Koehne et. Lingelsh.) V. Vassil.” [Koehne et. Lingelsh.] Nikolaev (1981) subgenus Fraxinus sect. Fraxinus subsect. Paniculatae E. Nikolaev subsect. Racemosae E. Nikolaev sect. Melioides “(Endl.) V. Vassil.” [(Endl.) Pfeiff.] subsect. Melioides “(Endl.) E. Nikolaev” [(Endl.) Lingelsh.] subsect. Sciadanthus (Coss. et Dur.) Lingelsh. (incl. subsect. Pauciflorae Lingelsh.) sect. Dipetalae (Lingelsh.) E. Nikolaev subgenus Ornus “(Boehm.) Peterm.” [(Boehm.) Pers.] sect. Ornus “DC.” [(Boehm.) DC.] sect. Ornaster “(Koehne et Lingelsh.) V. Vassil” [Koehne et Lingelsh.] Wei (1992) subgenus Fraxinus sect. Fraxinus sect. Melioides “(Endl.) V. Vassil.” [(Endl.) Pfeiff.] sect. Sciadanthus “(Coss. et Dur.) Z. Wei” [Coss. et Dur.] subgenus Ornus “(Boehm.) Peterm.” [(Boehm.) Pers.] sect. Ornus [(Boehm.) DC.] sect. Ornaster “(Koehne et Lingelsh.) V. Vassil” [Koehne et Lingelsh.] Wallander (this study) sect. Dipetalae (Lingelsh.) E. Nikolaev sect. Ornus (Boehm.) DC. sect. Fraxinus subsect. Fraxinus subsect. Melioides (Endl.) Lingelsh. subsect. Pauciflorae Lingelsh. subsect. Sciadanthus (Coss. et Dur.) Lingelsh.

3

MATERIALS AND METHODS Materials First, a thorough inventory of most of the taxa accepted in various recent treatments, and their synonyms, were undertaken in order to come up with a reasonable representation. This inventory was based on original descriptions, floristic treatments (Little 1952, Vassiljev 1952, Miller 1955, Murray 1968, do Amaral Franco and da Rocha Afonso 1972, Nakaike 1972, Grohmann 1974, Scheller 1977, Yaltirik 1978, Nikolaev 1981, Yamazaki 1993, Wei and Green 1996), regional monographs (mainly Standley 1924, Hara 1956 and 1982, Kitagawa 1979, Sun 1985, and Green 1991), and monographs of the genus (Lingelsheim 1920). Many taxa were studied in the field, and in botanical gardens and arboreta (Göteborg Botanical Garden and Arboretum, New York Botanical Garden, Missouri Botanical Garden and Shaw Arboretum, Palermo Botanical Garden, Kyoto Botanical Garden, and the Royal Botanic Gardens at Kew). Herbarium material from BM, C, E, GB, K, MO, NY, S, and UPS (acronyms according to Index Herbariorum), were also studied. This work resulted in a provisional list of about 50 taxa, most of which I eventually accepted (Table 3). Based on this list, at least two representatives of each recognised species, and in some cases subspecies, were chosen for DNA sequencing. In addition, some taxa of uncertain standing (synonyms) were chosen to evaluate their relationships. For the molecular work, I used fresh or silicagel dried material from many taxa cultivated in botanical gardens, and from field collections in Spain, Italy, USA, and Japan. Vouchers for these collections are deposited in the Göteborg Herbarium (GB). Herbarium material from C, E, GB, MO, NY, S, and UPS, was also frequently used. DNA extracted from cultivated plants at the Royal Botanic Gardens at Kew, and herbarium specimens at K, were kindly provided by Dr. Mark W. Chase of the Jodrell Laboratory at Kew. The final number of ingroup specimens selected for this study was around 90, of which I was able to obtain sequences from 83. In addition to this, 18 ITS-1 and ITS-2 sequences from the study by Jeandroz et al. (1997) were taken from GenBank and included in the full analysis. Outgroup taxa were chosen from the sister groups, Ligustrinae and Oleinae, based 4

on the Oleaceae phylogeny of Wallander and Albert (2000). All vouchers and GenBank accession numbers are listed in Table 2.

Molecular methods DNA extraction, PCR amplification, and automated sequencing were done using methods and equipment described by Wallander and Albert (2000). Primers designed by Nickrent et al. (1994) and Wojciechowski et al. (1993) (ITS4 and ITS5) were used to amplify the entire ITS region of the nuclear ribosomal DNA (i.e. the internal transcribed spacer 1, the 5.8S gene, and the internal transcribed spacer 2). In some difficult cases, the ITS-1 and ITS-2 were amplified and sequenced separately using the internal primers (ITS2 and ITS3) of Wojciechowski et al. (1993). The forward and reverse ITS sequences were assembled and edited using the Sequencher™ software version 3.1.1 (Gene Codes Corporation, Ann Arbor, MI, USA). Consensus sequences were aligned using the alignment feature in Sequencher and then manually adjusted with gap insertions according to the criteria specified by Andersson and Rova (1999). For all GenBank taxa and a few others, sequence data from the more conservative 5.8S gene were missing and replaced by ‘N’ in the data matrix. Furthermore, the rps16 and trnL-F intron sequences of 10 Fraxinus species obtained by Wallander and Albert (2000) were combined with five new ones (obtained in the same way) and 15 ITS sequences of the same taxa into one alignment, representing all sections and subsections. The alignments are available from the author upon request.

Cladistic analyses The final ITS data matrix contained 101 ingroup taxa, representing all but three of the 43 recognised species, and five outgroup taxa. The cladistic analyses were performed using heuristic searches in PAUP* 4.0b8 (Swofford 2000) with the optimality criterion set to maximum parsimony. These searches consisted of TBR branch-swapping in ten random addition sequence replicates limited to a maximum of 1000 trees saved per replicate. All characters were given equal weight (=1), gaps were treated as missing data, and multistates interpreted as polymorphisms (according to the IUPAC-IUB ambiguity set). Furthermore, parsimony jack-

Table 2. Voucher and GenBank accession numbers for 85 specimens of Fraxinus and 5 outgroup taxa used for ITS sequencing. In addition, 18 ITS sequences from the study by Jeandroz et al. (1997) are included (no vouchers). Auctors of all taxa are given in Table 3. [Note: GenBank accession numbers for sequences from this study will be added when accepted for publication.] Taxon Ingroup F. americana F. americana F. americana F. angustifolia ssp. angustifolia F. angustifolia ssp. oxycarpa F. angustifolia ssp. syriaca F. anomala F. anomala F. anomala F. apertisquamifera F. apertisquamifera F. berlandieriana F. biltmoreana F. bungeana F. bungeana F. caroliniana F. caroliniana F. chiisanensis F. chiisanensis F. chinensis F. chinensis F. chinensis var. rhynchophylla F. cubensis F. cuspidata F. cuspidata var. macropetala F. cuspidata var. macropetala F. dipetala F. dipetala F. dubia F. excelsior F. excelsior var. diversifolia F. floribunda F. gooddingii F. greggii F. greggii F. hubeiensis F. japonica F. japonica F. jonesii F. lanuginosa F. lanuginosa F. lanuginosa F. latifolia F. latifolia F. longicuspis F. longicuspis F. longicuspis F. mandshurica F. mandshurica F. mandshurica F. micrantha F. micrantha F. nigra

Voucher

GenBank accession number

Wallander 101 (GB) Wallander 99 (GB) U82906+U82907 Wallander 135 (GB) Wallander 2 (GB) Samuelsson 1 (GB) Rollins 1899 (GB) Pinzl 10931 (NY) U82914+U82915 Wallander 274 (GB) Kinoshita sn 1999-07-14 (GB) Jones 3595 (NY) U82910+ U82911 King 168 (S) Tianwei & Zhaofen 228 (MO) Massey & Boufford 4500 (MO) Hill 11048 (NY) Min 304 (SNUA) Min 264 (SNUA) U82884+ U82885 Wallander 87 (GB) Wallander 116 (GB) Rova 2261 (GB) U82916+U82917 Reichenbacher 1716 (MO) Barneby 18368 (NY) Wallander 180 (GB) Walker 1287 (NY) García 1456 (MO) Wallander 159 (GB) Wallander 1 (GB) Wallander 240 (GB) McGill & Lehto 20365 (NY) Diaz 406 (MO) Annable 2379 (NY); Arizona Xu Youming (Forestry Exp. Station, Wuhan Univ, China) Wallander 235 (GB) Wallander 115 (GB) Thorne 58757 (NY) Wallander 110 (GB) Wallander 266 (GB) Seino 2 (GB) U82912+ U82913 Wallander 182 (GB) U82888+ U82889 Im 10518 (NY) Wallander 256 (GB) U82874+ U82875 Wallander 113 (GB) Seino 1 (GB) Polunin et al. 4299 (UPS) Bist 96 (S) U82878+ U82879

[cont’d p. 6]

5

Table 2. Continued from previous page. Taxon

Voucher

F. nigra F. nigra F. ornus F. ornus F. ornus F. ornus F. oxyphylla F. pallisiae F. pallisiae F. papillosa F. paxiana F. pennsylvanica F. pennsylvanica F. pennsylvanica F. platypoda F. platypoda F. potamophila F. pringlei? F. profunda F. purpusii F. purpusii F. quadrangulata F. quadrangulata F. quadrangulata F. raibocarpa F. raibocarpa F. retusus F. rhynchophylla F. rufescens F. schiedeana F. sieboldiana F. sieboldiana F. sieboldiana F. sikkimensis F. sogdiana F. spaethiana F. spaethiana F. syriaca F. texensis F. texensis F. tomentosa F. trifoliolata F. uhdei F. uhdei F. velutina F. velutina F. velutina F. velutina F. xanthoxyloides F. xanthoxyloides var. dimorpha

Wallander 105 (GB) Rickson 239 (GB) Wallander 31 (GB) Wallander 38 (GB) Wallander 146 (GB) Wallander 216 (GB)

Chase 3886 (K) Broome et al. 1829 (NY) Smith 3299 (NY) Dubuis 12424 (C) Wallander 141 (GB)

Outgroup Ligustrum vulgare L. Syringa vulgaris L. Forestiera acuminata Poir. Osmanthus fragrans Lour. Phillyrea latifolia L.

Wallander 168 (GB) Wallander 111 (GB) Wallander 100 (GB) Wallander 28 (GB) Chase 3880 (K)

6

GenBank accession number

U82868+ U82869 U82870+ U82871 Rechinger 10066 (S) Tucker 2597 (S) Wallander 187 (GB) U82902+ U82903 Wallander 103 (GB) Wallander 83 (GB) U82876+ U82877 Wallander 114 (GB) Wallander 88 (GB) Martínez & Soto 3718 (MO) Wallander 104 (GB) Breedlove 42154 (MO) Breedlove & Thorne 30445 (NY) U82880+ U82881 Wallander 98 (GB) Wallander 94 (GB) Cadupob s.n. 1955-08-17 (MO) Regel s.n. 1982-07-?? (S) Wallander 249 (GB) U82886+ U82887 Zamudio 3673 (MO); Mexico Villanueva 274 (NY) Takahashi et al. 1708 (MO) Wallander 112 (GB) Wallander 265 (GB) Wallander 188 (GB) Elias 10008 (C) Wallander 142 (GB) Wallander 259 (GB) U82872+ U82873 Chase 3887 (K) Walker 1692 (NY) U82896+ U82897 Forrest 15313 (E) McPherson 805 (NY) Wallander 192 (GB) U82904+ U82905

knifing was performed using XAC (James S. Farris, Swedish Museum of Natural History, Stockholm), doing 1000 replicates, each with ten random addition sequence replicates and nonrotational branch-swapping. First, an analysis of the complete data set was undertaken in order to investigate possible mistakes in the taxon identity or in the alignment. The identities of taxa with material from more than one specimen were verified in this way. Unfortunately, an analysis including such a large number of taxa, where many of the species duplicates or triplicates are very similar or even identical, takes long time to run and yields many shortest trees of equal length, all of which are not possible to save in the limited searches. To shorten run time and reduce the number of trees obtained, the same analyses were conducted on a reduced ITS matrix only including one sequence from each of the recognised species. The chloroplast data set (sequences of the rps16 intron and trnL-F region combined) was analysed separately and in combination with ITS data for 15 ingroup species, representing all subsections, and two outgroup species. The 17 ITS sequences were also analysed separately. In all PAUP analyses, I also explored consequences of removing oddly positioned taxa or those with uncertain alignment from the analyses. Mapping of reproductive characters, viz. pollination system (wind or insect), breeding system (hermaphroditism, androdioecy, polygamy, or dioecy), flowering time (before or after leafing), inflorescence position (lateral or terminal), calyx (present or absent), and habit (shrub, small tree, or large tree), onto the phylogenetic tree was done with MacClade version 3.08a (Maddison and Maddison 1999). All characters were considered to be unordered. Transformations between character states were compared between the trees using both ACCTRAN and DELTRAN optimisations. In taxa were experimental demonstration of anemophily have not been conducted, anemophily has been inferred on the basis of a number of traits considered to be typical of anemophilous taxa (the so-called anemophilous syndrome, explained in the Discussion). Here anemophily is defined by apetalous flowers and flowering before leafing. Therefore, these traits are not traced in the ingroup.

RESULTS Molecular phylogeny The full ITS data set with 106 taxa contained 666 nucleotide sites of which 217 were phylogenetically informative. The cladistic analysis of this data set resulted in 4000 most parsimonious (MP) trees and the strict consensus of these is shown in Fig. 1, with jackknife values at the supported branches. The retention index (RI) of these trees is 0.85 and the length is 2703 steps. All major clades receive strong jackknife support, except the weakly supported relationship of F. raibocarpa as sister to the rest of section Ornus. There is no resolution between any of the major clades, except that Dipetalae is weakly supported as sister to the rest of the genus. The reduced ITS data set, including 43 taxa and 176 informative characters, resulted in a single and fully resolved MP tree (RI=0.75, 772 steps) (Fig. 2). However, there is still no jackknife support for the branching pattern among the major clades, only section Dipetalae is weakly supported as sister group to the rest of the genus. There is no jackknife support in any of the trees for the positions of F. chiisanensis, F. spaethiana, and F. cuspidata. The results from these first analyses seemed to make good sense, in terms of taxa within the main clades, except for the very odd position of F. cuspidata together with F. spaethiana and F. chiisanensis in the reduced ITS tree (Fig. 2). In the full analysis (Fig. 1) their relationships are unresolved. After excluding the sequences of F. cuspidata, the effect on the full analysis was that the branching order of the main clades became resolved, but still without jackknife support. The combined rps16 and trnL-F chloroplast (cp) matrix of 17 taxa contained 1737 characters of which 14 were phylogenetically informative among the ingroup (15 taxa). The heuristic searches found 18 MP trees of 70 steps and the strict consensus of them is shown in Fig. 3a. When combined with ITS data for the same taxa, two most parsimonious trees were obtained (not shown). The 17-taxon ITS matrix analysed separately (110 informative characters) resulted in three trees of 403 steps (Fig. 3b). The ITS and cp trees are similar in their supported groupings, but there is no supported position for F. chiisanensis, F. spaethiana or F. cus7

71

100 96 97

52

80 85 100

100

100 65 98

96 100 95

100

100

100

100

100

84

68

97 93 70 61

63

74 98

93

100 90 88 53

95 96 89

93 63

84

65 81

82

Ligustrum vulgare Syringa vulgaris OUTGROUP Forestiera acuminata Osmanthus fragrans Phillyrea latifolia F. jonesii RFT58757 F. dipetala EW180 F. dipetala JBW1287 F. anomala API10931 DIPETALAE F. anomala RCR1899 F. quadrangulata EW94 F. quadrangulata EW98 F. quadrangulata genbank F. chiisanensis WKM304 264 F. spaethiana EW142 F. spaethiana EW259 F. cuspidata genbank F. cuspidata FWR1716 F. cuspidata RBA18368 F. dubia AGR1456 F. pringlei? EM3718 F. greggii RD406 F. schiedeana RVG274 F. rufescens SZ36673 PAUCIFLORAE F. gooddingii EL20365 F. greggii CRA2379 F. purpusii DEB30445 F. purpusii DEB42154 F. hubeiensis XUx1 F. xanthoxyloides ADB12424 SCIADANTHUS F. xanthoxyloides EW141 F. nigra EW105 F. nigra FRP239 F. nigra genbank F. platypoda genbank F. mandshurica EW113 F. mandshurica genbank F. mandshurica TAS1 F. platypoda EW114 F. excelsior var diversifolia EW1 FRAXINUS F. excelsior EW159 F. oxyphylla genbank F. syriaca genbank F. angustifolia ssp angustifolia EW135 F. angustifolia ssp oxycarpa EW2 F. pallisiae genbank F. pallisiae KHR10066 F. sogdiana TSE10008 F. angustifolia ssp syriaca JAS1 F. potamophila EW88 F. biltmoreana GenBank F. latifolia EW182 F. latifolia genbank F. papillosa JMT2597 F. uhdei EW192 F. americana EW101 F. americana genbank F. uhdei GMP805 F. anomala genbank F. americana EW99 F. berlandieriana SGJ3595 F. caroliniana JRM4500 MELIOIDES F. caroliniana SH11048 F. cubensis JHR2261 F. pennsylvanica EW103 F. pennsylvanica EW83 F. texensis JBW1692 F. texensis MWC3887 F. profunda EW104 F. tomentosa genbank F. velutina EL1829 F. velutina JFS3299 F. velutina MWC3886 F. pennsylvanica genbank F. velutina genbank F. raibocarpa AREx1 F. raibocarpa CADx1 F. trifoliolata GF15313 F. bungeana LT228 F. bungeana TFK168 F. floribunda EW240 F. retusus EW249 F. paxiana EW187 F. sikkimensis EW188 F. ornus EW216 F. ornus EW146 F. ornus EW31 F. ornus EW38 F. apertisquamifera EKx2 F. apertisquamifera EW274 F. lanuginosa EW110 ORNUS F. lanuginosa EW266 F. lanuginosa TAS2 F. sieboldiana EW112 F. sieboldiana EW265 F. sieboldiana TT1708 F. micrantha HSB96 F. micrantha OP4299 F. longicuspis EW256 F. longicuspis HTI10518 F. rhynchophylla genbank F. chinensis EW87 F. chinensis genbank F. chinensis ssp rhynchophylla EW116 F. japonica EW115 F. japonica EW235 F. longicuspis genbank

Figure 1. Strict consensus of 4000 equally most parsimonious trees from the cladistic analysis of ITS sequences from 106 Fraxinus taxa. Jackknife values exceeding 50% are shown above the supported branches. The length of the trees is 2703 steps and the RI=0.85.

8

72

100

100

Ligustrum vulgare Syringa vulgaris OUTGROUP Forestiera acuminata Osmanthus fragrans Phillyrea latifolia F. dipetala 98 SECTION DIPETALAE F. anomala F. quadrangulata F. raibocarpa 82 58 F. lanuginosa F. sieboldiana 87 F. ornus F. trifoliolata SECTION ORNUS F. paxiana F. bungeana F. floribunda F. micrantha 91 64 F. longicuspis 96 F. chinensis F. japonica F. cuspidata F. spaethiana F. chiisanensis F. americana F. uhdei 98 F. latifolia 66 F. papillosa MELIOIDES F. texensis 98 F. caroliniana F. velutina F. pennsylvanica F. profunda F. dubia 97 F. rufescens 100 PAUCIFLORAE F. greggii F. purpusii 100 F. hubeiensis SCIADANTHUS F. xanthoxyloides 100 F. nigra 62 F. angustifolia 95 F. excelsior FRAXINUS 95 F. mandshurica 5 changes F. platypoda

?

SECTION

FRAXINUS

SUBSECTIONS

Figure 2. Single most parsimonious tree resulting from the cladistic analysis of ITS sequences from 43 taxa (38 ingroup taxa in Fraxinus), shown as a phylogram. The length is 772 steps and RI=0.75. Jackknife values exceeding 50% are shown at the supported branches. Sectional and subsectional assignments are according to the revised classification in this study.

pidata, which is the major explanation for the multiple MP trees. The results from the chloroplast and combined analyses differ from the ITS tree (Fig. 2)

in one major respect — some species of section Fraxinus are nested within section Ornus. This is a dubious result that, although with jackknife support in the cp tree, is not supported by non9

Syringa vulgaris

Syringa vulgaris

Phillyrea latifolia

Phillyrea latifolia

Fraxinus chiisanensis

F. dipetala

99

Fraxinus cuspidata

78

F. quadrangulata

Fraxinus greggii

F. chiisanensis

Fraxinus spaethiana Fraxinus americana

70 97

100

F. cuspidata

Fraxinus pennsylvanica

F. spaethiana

Fraxinus dipetala

94 68

98

88 89

F. pennsylvanica

Fraxinus quadrangulata

F. raibocarpa 94

F. chinensis

Fraxinus xanthoxyloides

F. ornus

Fraxinus chinensis

F. greggii

Fraxinus ornus 65

F. americana

Fraxinus anomala

Fraxinus raibocarpa 86

F. anomala

98

F. xanthoxyloides

Fraxinus excelsior

F. excelsior

Fraxinus nigra

F. nigra

(a) chloroplast tree

(b) ITS tree

Figure 3. (a) Consensus of the 18 MP trees from the analysis of chloroplast DNA sequences (rps16 intron and trnL-F region) from 17 taxa, and (b) consensus of the three MP trees from the analysis of ITS sequences from the same taxa. Parsimony jackknife values exceeding 50% are shown above the supported branches.

molecular data. Also the 17 ITS sequences that were analysed separately (Fig. 3b) gave the same strange position of section Fraxinus as in the cp tree (Fig. 3a). This prompted me to try including some more taxa, and when I had restored a few more taxa of sections Ornus and Fraxinus, the topology changed to the same as in Fig. 2. The ITS tree (Fig. 2) identifies five main wellsupported clades (Dipetalae, Ornus, Melioides, Pauciflorae, and Fraxinus/Sciadanthus) that correspond well to morphological data and the existing classification, with a few exceptions. A revised classification based on the molecular result is given in Table 1 and 3. I recognise three sections (Dipetalae, Ornus, and Fraxinus) and four subsections (Pauciflorae, Fraxinus, Sciadanthus, and Melioides), which are shown in Fig. 2 and discussed further below. Three taxa have no supported position and their possible relationships are discussed in the light of non-molecular data. 10

Evolution of reproductive systems A tree based on the 43-taxon ITS tree was used to trace reproductive traits (Fig. 4). The topology of this tree is the same as the ITS tree, except for the position of F. cuspidata. The rationale for moving it down one node from its grouping with F. spaethiana and F. chiisanensis is explained in the systematic discussion below. Excluding F. cuspidata or not makes no difference for the interpretation of the character state changes relating to reproductive biology. In the ingroup, character state optimisations are equivocal only in two characters (calyx and habit) between the ACCTRAN and DELTRAN optimisations. However, these no relevance for the implied number of origins of anemophily and correlated shifts in breeding system. The ancestral character states in Fraxinus are shrubs or small trees with sympetalous bisexual flowers, emerging with or after leafing. The two genera of subtribe Ligustrinae are both insectpollinated, but among the 12 genera in subtribe

FRAXINUS DIPETALAE

FRAXINUS

ORNUS

OUTRGOUPS

Lig LIGUSTRINAE Sy ustru Foringa m r e No stie Phtelae ra OLEINAE Osillyre a qu man a a t an dranhus dipoma gula l ta rai etalaa sieboca ap bold rpa e i lan rtisqana ornugin uam trif us osa ifera paoliola buxianata n flo gea micribunna lon ranthda japgicu a ch onicspis cusinensa chi pida is spaisan ta tex ethiensis a carensis na vel olinia proutina na pe fund uh nnsy a padei lvani ca p lat illos amifolia a xanerica hu thoxna nigbeienyloid ex ra sis es ancelsio plagusti r matypofolia du ndshda rufbia urica puesce n r gre pusii s go ggii od din gii

SECTION

SCIAD -

"ORNASTER " ? ?

MELIOIDES

ANTHUS

FRAXINUS

PAUCIFLORAE

SUBSECTION

dioecy

dioecy calyx loss

large tree

shrub

androdioecy

Geographical distribution

dioecy

terminal inflorescences free petals

New World

polygamy large tree

Old World

Pollination system

lateral inflorescences

entomophily anemophily

hermaphrodites sympetalous terminal inflorescences shrubs or small trees flowering with or after leafing

Figure 4. Origins of characters related to pollination system optimised on a tree with 40 taxa of Fraxinus showing the evolution of anemophily. Squares under taxon names refer to geographical distribution. Only pollination system is shown in the outgroup and not all reversals or losses of other characters are shown in the ingroup. The revised classification of this study is also shown.

Oleinae, of which only four are shown in Fig. 4, there have been multiple transitions in pollination and breeding systems (Wallander ms.). In Fraxinus, section Dipetalae is resolved as sister to the other two sections. The three taxa of section Dipetalae are hermaphrodites, two of which have evolved adaptations to wind-pollination (e.g. loss of petals and scent). In the sister group there are two major divisions: the sections Ornus and Fraxinus. The ancestral states in Ornus are entomophily or ambophily and hermaphroditism, but androdioecy originated early in this group. An increase in tree height preceded the transition to anemophily (loss of petals and delayed leafing). Two of the anemophilous taxa, F. micrantha and F. longicuspis, are still androdioecious, but F. baroniana (not shown), F. chinensis, and F. japonica have

evolved dioecy through loss of the male function in morphologically hermaphrodite flowers. The taxa in section Fraxinus are all windpollinated and like those in section Dipetalae they have flowers in lateral inflorescences, but in contrast they flower before the terminal leaf buds open. The trees are usually quite large and the taxa are either polygamous or dioecious. The transition to dioecy in F. platypoda and F. mandshurica is probably rather recent, since the polygamous past is still evident in the rudimentary stamens of functionally female flowers. In contrast, not a single dioecious species in subsection Melioides has vestigial organs of the opposite sex in the flowers. To summarise: Within Fraxinus there are three independent origins of anemophily that in three separate cases have been followed by 11

Table 3. Revised infrageneric classification of the genus Fraxinus. Only synonyms that are commonly used as good taxa, or sequenced in this study, are given. Taxa

Distribution

Fraxinus L. sect. Dipetalae (Lingelsh.) E. Nikolaev F. anomala Torr. ex S. Wats. W USA F. dipetala Hook. et Arn. W USA F. quadrangulata Michx. C & E USA, C Canada Fraxinus L. sect. Ornus (Boehm.) DC. F. apertisquamifera Hara F. bungeana DC. F. floribunda Wall. F. griffithii C. B. Clarke

Japan China Himalayas, E Asia SE Asia

F. lanuginosa Koidz. F. malacophylla Hemsl. F. ornus L. F. paxiana Lingelsh.

Japan China, Thailand C & E Mediterranean Himalayas, China

F. raibocarpa Regel F. sieboldiana Blume F. trifoliolata W. W. Smith F. baroniana Diels F. chinensis Roxb. F. japonica Blume ex K. Koch F. longicuspis Sieb. et Zucc. F. micrantha Lingelsh.

C Asia China, Japan, Korea China China E Asia Japan Japan Himalayas

Fraxinus L. sect. Fraxinus subsect. Fraxinus F. angustifolia Vahl ssp. angustifolia F. angustifolia Vahl ssp. oxycarpa (Willd.) Franco & Rocha Afonso F. angustifolia Vahl ssp. syriaca (Boiss.) Yalt. F. excelsior L. F. mandshurica Rupr. F. nigra Marsh. F. platypoda Oliv. subsect. Melioides (Endl.) Lingelsh. F. americana L. F. caroliniana Mill. F. latifolia Benth. F. papillosa Lingelsh. F. pennsylvanica Marsh. F. profunda (Bush) Bush F. texensis (Gray) Sarg. F. uhdei (Wenzig) Lingelsh. F. velutina Torr.

SW Europe SE Europe

F. retusa Champ. ex Benth., F. insularis Hemsl. F. ferruginea Lingelsh., F. formosana Hayata, F. philippinensis Merr. F. retusifoliolata Feng ex P. Y. Bai F. sikkimensis (Lingelsh.) Hand.-Mazz., F. suaveolens W. W. Smith, F. depauperata (Lingelsh.) Z. Wei

F. rhynchophylla Hance

F. pallisiae A. J. Willmott, F. oxycarpa Willd., F. oxyphylla M. Bieb. (nom. illeg.) F. potamophila Herder, F. sogdiana Bunge

E USA & E Canada SE USA N & C USA, Mexico N & C USA, Mexico C & E USA, Canada E USA USA (Texas) C America SW USA, N Mexico

F. biltmoreana Beadle F. cubensis Griseb. F. oregona Nutt.

Mexico, Guatemala

F. schiedeana Schlecht. & Cham.

SW USA, N Mexico SW USA, Mexico Mexico, Guatemala Mexico

subsect. Sciadanthus (Coss. et Dur.) Lingelsh. F. xanthoxyloides (G. Don) DC. F. hubeiensis S. Z. Qu, C. B. Shang & P. L. Su

N Africa to China China

12

F. jonesii Lingelsh.

Middle East N. & C. Europe China, Japan, Korea, E Russia E USA, E Canada China

subsect. Pauciflorae Lingelsh. F. dubia (Willd. ex Schult. & Schult. f.) P. S. Green & M. Nee F. gooddingii Little F. greggii A. Gray F. purpusii Brandegee F. rufescens Lingelsh.

Incertae sedis F. cuspidata Torr. F. chiisanensis Nakai F. spaethiana Lingelsh.

Synonym(s)

SW USA, Mexico Korea Japan

F. berlandieriana DC. F. tomentosa Michx. f.

F. dimorpha Coss. & Dur.

evolution of dioecy. Important traits that precede the evolution of anemophily have been the androdioecious or polygamous breeding systems (with unisexual males) and an increase in tree size.

DISCUSSION Molecular phylogeny The ITS phylogeny (Fig. 1 and 2) has five main clades (Dipetalae, Ornus, Melioides, Pauciflorae, and Fraxinus/Sciadanthus), each with strong jackknife support, and which roughly correspond to groups in the existing classifications. Although the analysis of the reduced ITS matrix resulted in a single most parsimonious and completely resolved tree (Fig. 2), there is no jackknife support for any particular resolution among these clades. Only section Dipetalae has low support as sister to the rest of Fraxinus. Previously, Gielly and Taberlet (1994) have shown that the noncoding chloroplast trnL-F region has very low interspecific variation in Fraxinus, which was concluded to be too low to resolve phylogenetic relationships within the genus. In my work, trnL-F sequences from 15 species of Fraxinus were combined with sequences of the rps16 intron, and together these data sets contained 14 characters that were informative within the ingroup. The analyses of chloroplast data alone gave a poorly resolved tree (Fig. 3a), but in combination with ITS data the resolution increased (not shown). The chloroplast tree and the ITS tree with the same taxon sampling do not differ in the main groups, only in the resolution between them, but they do differ significantly from the 43taxon ITS tree in nesting species from section Fraxinus within section Ornus. However, when adding only a few more taxa to the smaller data set (Fig. 3b), the resulting topology became identical to the one produced by the larger data set (Fig. 2). Thus it appears that it was the low taxon sampling that caused the spurious result. Although without supported resolution among the clades in section Fraxinus, I conclude that the tree in Fig. 2 has the correct topology with regard to the relative placements of section Ornus and taxa in section Fraxinus. Except for the doubtful position of F. cuspidata, the ITS phylogeny shows good congruence with morphological data (discussed below), and seems

to be a reliable estimate of the phylogeny of the genus Fraxinus. This result differ somewhat from the one of Jeandroz et al. (1997), mainly because of a nearly complete taxon sampling. They did not include any representatives from the subsections Pauciflorae or Sciadanthus, and only one representative of the petaliferous taxa in section Ornus. Conclusions about the monophyly of subsections Ornaster and Euornus (sensu Lingelsheim) within section Ornus cannot be made with only F. ornus representing the latter subsection. It can also be seen in the complete tree (Fig. 1) that the GenBank sequence of Fraxinus anomala, from their study, is misplaced. This is probably due to a misidentification of the sampled specimen. Unfortunately, the identification of this specimen could not be checked because their publication lacks references to vouchers.

Systematics In previous classifications (Table 1), Fraxinus has been divided into two subgenera or sections. However, accepting the ITS tree (Fig. 2) as a reliable estimate of the phylogeny of Fraxinus, with section Dipetalae as sister to the rest of Fraxinus, makes subgenus Fraxinus (section Fraxinaster sensu Lingelsheim 1920) paraphyletic. If one wants to have a natural classification of the genus, there are two options. One is to abandon the subgenera and accept only six sections (Dipetalae, Ornus, Melioides, Pauciflorae, Sciadanthus, and Fraxinus) without indicating any relationships among them. Considering the weak or unsupported resolution among the groups, this is not a bad idea. However, taking into account also non-molecular data, there are several characters that support a group consisting of Melioides, Pauciflorae, Sciadanthus, and Fraxinus (i.e. subgenus Fraxinus excluding Dipetalae). Thus, a more informative classification would be achieved by having three sections, Dipetalae, Ornus, and Fraxinus, the latter one including four subsections (Melioides, Pauciflorae, Sciadanthus, and Fraxinus). One could of course also choose the subgenus level, but that would require two new combinations (subgenus Dipetalae and section Pauciflorae). The alternative with sections and subsections requires no new combinations, as all groups have previously had sectional or subsectional status. 13

Based on the molecular result that correlates well with morphological data, I propose a revised classification with three sections and four subsections (Table 1, Table 3, Fig. 2). The apetalous taxa of section Ornus, previously constituting the subsection Ornaster sensu Lingelsheim, are not recognised separately as it would

make the previous subsection Euornus paraphyletic. There are a few necessary transfers of taxa between the sections and subsections, and a discussion of the new sectional delimitation, recognised taxa, and their reproductive ecology follows. Below is a diagnostic key to the sections and subsections:

1. Inflorescences emerging with or after leaves from terminal buds; hermaphrodites or androdioecious (at least morphologically); Old World species .............................. section Ornus 1. Inflorescences emerging before or with leaves from lateral buds on previous year’s shoot; hermaphrodites, polygamous, or dioecious; New World or Old World species 2. Stems quadrangular; flowers mostly bisexual, with two united petals or without petals; New World species .............................................................................................. section Dipetalae 2. Stems not quadrangular; flowers uni- or bisexual, apetalous; New World or Old World species, section Fraxinus 3. Shrubs or small trees; leaf rachis winged; polygamous 4. Few-flowered panicles; small samaras; New World species ....... subsection Pauciflorae 4. Many-flowered panicles; large samaras; Old World species .... subsection Sciadanthus 3. Large trees; leaf rachis not winged; dioecious or polygamous 5. Flowers without calyx or with small and/or deciduous calyx; polygamous (at least morphologically); mainly Old World species (1 New World) ...... subsection Fraxinus 5. Flowers always with calyx; dioecious (flowers always unisexual); New World species ................................................................................................................ subsection Melioides

Fraxinus sect. Dipeta lae Section Dipetalae comprises three American species, previously widely scattered in the old classifications of the genus. F. quadrangulata belonged to section Fraxinus (Bumelioides), F. anomala to section Melioides, and F. dipetala used to be the sole member of section Dipetalae. Having found strong molecular support for the unexpected relationship between the three taxa, I investigated their morphological characteristics more closely. In Fraxinus, they are unique in having quadrangular branches and hermaphrodite flowers in lateral inflorescences. F. dipetala (two-petal ash, foothill ash or California flowering ash) is a shrub or small tree restricted to western USA. It is the only ash having two petals, which are united and tubular by fusion with the filaments. The flowers have a mild sweet fragrance and occur in manyflowered and showy inflorescences, which are probably attractive to insects. The anthers are relatively large and protrude from the corolla, an indication that the flowers might be ambophilous (both wind- and insect-pollinated). 14

F. anomala (singleleaf ash) is a shrub or small tree, with simple leaves or occasionally 3-5 leaflets, and predominantly occurs in south-western USA. It differs from the species in section Melioides (where it was formerly classified) in having quadrangular branches and bisexual or sometimes unisexual flowers (Sargent 1949, Vines 1984). The flowers have a persistent calyx, but lack corolla, and appear in lateral panicles before or with the young leaves. They are apparently wind-pollinated. F. quadrangulata (blue ash) is a large tree with conspicuously quadrangular branches, occurring in eastern and central North America. It was previously classified in section Fraxinus (Bumelioides). The flowers are mostly hermaphrodite (Sargent 1949, Miller 1955, Barnes and Wagner 1981), and I have only seen hermaphrodite flowers, but according to Vines (1984) the species is hermaphrodite, dioecious or polygamous. They are apetalous and presumably wind-pollinated. The calyx is small and deciduous.

Fraxinus sect. Ornus Section Ornus, as circumscribed by Lingelsheim (1920) and others, encompassed two subsections: Ornus (Euornus) and Ornaster (Table 1). All taxa in this section are characterised by flowers that emerge in terminal panicles together with the leaves. Subsection Euornus comprised insect-pollinated species that have flowers with four free petals, and the species being either hermaphrodites or androdioecious. Subsection Ornaster was characterised by apetalous and wind-pollinated flowers, the species being either androdioecious or dioecious. The molecular support for section Ornus is high, although with low support for inclusion of F. raibocarpa, and there is no doubt that it is a monophyletic group. However, although subsection Ornaster (F. micrantha, F. longicuspis, F. chinensis, F. japonica, and the non-sequenced F. baroniana) is monophyletic and has high jackknife support, there is no support for recognising the two subsections as it would make Euornus paraphyletic. In spite of this situation, I will for practical reasons use “Ornaster” as an informal name for the wind-pollinated clade within section Ornus in the following discussion about the evolution of anemophily. Section Ornus comprises 16 recognised species, all distributed in Eurasia and with a concentration in eastern Asia. Three species are hermaphrodites: F. griffithii, F. malacophylla (Nakaike 1972, Wei and Green 1996), and F. raibocarpa (Grohmann 1974). F. griffithii is the only evergreen species of Fraxinus and occurs in China, Japan, Taiwan and the Philippines, F. malacophylla is endemic to the mountains of northern Thailand and southern China (Yunnan), and F. raibocarpa occurs in mountainous areas in central Asia. All three species are small trees with large and showy inflorescences that are insect-pollinated. Although I only succeeded in getting DNA sequences from one of them, F. raibocarpa, I am sure that the other two belong here too. The jackknife support for F. raibocarpa in section Ornus is not high. However, there are several non-molecular synapomorphies (e.g. terminal inflorescences and free petals) that unite these species. A survey of phenotypic sex expression and measurements of floral parts in all but one taxon of this section (Wallander and Dahl ms.) showed that ten taxa are morphologically androdioecious. Two of these have so far been shown to be functionally androdioecious, F.

ornus (Dommée et al. 1999) and F. lanuginosa (Ishida and Hiura 1998). F. ornus (manna ash) occurs in the Mediterranean area and eastwards to Iraq. It has showy, white-flowered and fragrant panicles that attract numerous insects. The anthers of both males and hermaphrodites contain much pollen, a lot of which is also spread by wind, i.e. it is ambophilous (Wallander and Dahl ms.). F. lanuginosa is endemic to Japan and has a similar reproductive system. Both male and hermaphrodite flowers produce functional pollen but there are 2.6 times as many flowers in the male inflorescences. This species is also ambophilous (Ishida and Hiura 1994). The other six petaliferous and androdioecious taxa of section Ornus are distributed mainly in China and Japan, several of them being endemic to either country. F. floribunda and F. paxiana have a wider distribution, extending into the Himalayas as well. They all have inflorescences similar to F. ornus and F. lanuginosa and they may conceivably turn out to be functionally androdioecious and ambophilous as well. There are five apetalous taxa in section Ornus (the “Ornaster” group): F. micrantha in the Himalayas, F. japonica and F. longicuspis (Japanese ash) in Japan, F. baroniana that is endemic to China, and the more widespread F. chinensis (Chinese ash). These taxa form a well-supported monophyletic group (Fig. 2), distinguished from the rest of section Ornus in having no corolla. Their flowers are presumably wind-pollinated. One of the taxa has been studied in the field: F. longicuspis in Japan (Wallander and Dahl ms.). Compared to most of the petaliferous taxa, the trees are much bigger and the leaves are still very small when flowering takes place. The preliminary studies show that F. longicuspis is probably functionally androdioecious, but F. japonica and F. chinensis might be functionally dioecious as the anthers in the hermaphrodite flowers are much smaller than in the males (Yamazaki 1993, Wallander and Dahl ms.). I have not seen any flowering material of F. baroniana, but Wei and Green (1996) describe both F. baroniana and F. chinensis as dioecious. However, for F. baroniana male and female flowers are depicted, but for F. chinensis (both subspecies chinensis and rhynchophylla) male, female and hermaphrodite flowers are illustrated. In Nakai (1921), hermaphrodite flowers of F. rhynchophylla are illustrated with smaller an15

thers than in the male flowers, and the same is the case for F. japonica in Kitamura and Murata (1971). Thus it appears that, despite the somewhat incongruent descriptions of floral morphology, both F. chinensis and F. japonica are functionally dioecious. Notable is this context are the identical ITS sequences of F. japonica and F. chinensis, and Wei and Green (1996) treat F. japonica as a synonym of F. chinensis ssp. rhynchophylla.

Fraxinus sect. Fraxinus Section Fraxinus comprises of four subsections: Fraxinus, Sciadanthus, Pauciflorae, and Melioides. Together with F. chiisanensis and F. spaethiana they comprise 23 recognised species that are wind-pollinated and characterised by apetalous flowers in inflorescences that emerge from lateral buds before the terminal leaf buds open. Like section Fraxinus, section Dipetalae is also characterised by flowering on the previous year’s shoots, but in contrast, the species of section Fraxinus differ from the two apetalous and hermaphrodite species of section Dipetalae in that they are either polygamous, subdioecious or dioecious. They also differ in having terete stems. Subsection Fraxinus Subsection Fraxinus contains five recognised species and two subspecies and all except F. nigra are distributed in Eurasia. In previous classifications, this group was characterised by polygamous flowers without calyx, but now I have included F. platypoda, which has a small calyx. F. nigra has a deciduous calyx, but the other three species have asepalous flowers. The male flowers consist of two stamens only, the hermaphrodite flowers of one pistil and two (functional) stamens, and the female flowers of one pistil and sometimes rudimentary stamens. The species are all large, wind-pollinated trees. The reproductive ecology of the European ash (common ash), F. excelsior, has been studied in detail (Wallander and Dahl ms.). It has a polygamous breeding system with a continuous phenotypic gender expression, with male, hermaphrodite, and female individuals occurring in approximately equal proportions in the populations. Phenotypic gender can also change to some extent (within the continuum) between years. F. angustifolia and F. excelsior are very closely 16

related and they have also been shown to be able to hybridise (Jeandroz et al. 1996). F. angustifolia is also polygamous, but its reproductive ecology has so far not been studied in any detail. It is subdivided into three geographical subspecies: ssp. angustifolia (narrow-leaved ash) distributed in southwestern Europe, ssp. oxycarpa (Caucasian ash) in southeastern Europe, and ssp. syriaca (Syrian ash) in the Middle East. In contrast to all other taxa of the genus, which have paniculate inflorescences, F. angustifolia has racemes. After considering the opinions of several authors (Anderson and Turrill 1938, Metcalfe 1938, Vassiljev 1952, Scheller 1977, de Jong 1990), studying herbarium material, and having seen the similarity between the ITS sequences, I have come to the conclusion that F. potamophila (Turkestan ash) and F. sogdiana (distributed from Turkestan to China) should be synonymised under F. angustifolia ssp. syriaca, and F. pallisae should be synonymised under F. angustifolia ssp. oxycarpa. F. pallisae (Pallis’ ash) is morphologically very similar to F. angustifolia ssp. oxycarpa and also occurs in south-eastern Europe. It is distinguished only by the distribution of its much denser indumentum on the leaves and shoots (Anderson and Turrill 1938). There is not much variation in the ITS sequences between any of the subspecies, and the differences are much less among them than between them and F. excelsior. F. mandshurica (Manchurian ash) is functionally dioecious, but the polygamous origin is evident in the presence of rudimentary stamens in the female flowers (personal observation, Min et al. 2001). It is sister to F. platypoda and both species occur in China, although F. mandshurica has a wider distribution into eastern Russia and Korea. F. platypoda is described as polygamodioecious (Wei and Green 1996), but is probably functionally dioecious as the stamens are reduced in the functionally female flowers. Despite its geographical distribution, it was previously placed in Melioides because of the presence of a small calyx, at least in the hermaphrodite flowers (Wei and Green 1996). Nevertheless, it exhibits all other characters shared by subsection Fraxinus and molecular data give strong support for its placement here. In fact, the ITS sequences of F. platypoda and F. mandshurica are nearly identical. Many authors (e.g., Nakaike 1972, Wei and Green 1996) place F.

spaethiana (endemic to Japan) in synonymy with F. platypoda (endemic to China), but their ITS sequences are quite dissimilar and I recognise them as separate species (see further below). F. nigra (black ash) is closely related to F. mandshurica and F. platypoda. In fact, F. mandshurica has been referred to as F. nigra ssp. mandshurica (Rupr.) S. S. Sun. Although an eastern North American species, it definitely belongs to this group. The calyx is small and deciduous, or wanting. It is variously described as polygamous (Sargent 1949, Miller 1955), polygamo-dioecious (Barnes and Wagner 1981), or dioecious (Rehder 1951). Barnes and Wagner (1981) illustrate the hermaphrodites with and without rudimentary stamens, which might indicate a functionally dioecious breeding system. Regardless of functionality of the sexes, it is phenotypically polygamous, and male and hermaphrodite flowers can occur on the same tree (andromonoecy). Subsection Sciadanthus Subsection Sciadanthus consists of only two species: F. xanthoxyloides (Afghan or Algerian ash) distributed from Morocco and Algeria in north Africa over the Middle East to the Himalayas and China, and F. hupehensis which is a threatened species endemic to the Hubei province in China (Ming and Liao 1998). They are small trees or shrubs, and characterised by apetalous flowers with calyx, except that the male flowers of F. xanthoxyloides lack calyx. The flowers are polygamous and wind-pollinated. They resemble New World Pauciflorae, but have many more flowers in their congested, cymose panicles. They form a well-supported sister group to subsection Fraxinus and share the same geographical distribution area (disregarding that F. nigra is American). Subsection Pauciflorae Subsection Pauciflorae is a monophyletic group consisting of five New World species that all occur in arid regions of south-western USA, Mexico, and Guatemala. They are shrubs or small trees with small coriaceous leaves. In common with the two species of section Sciadanthus, they have winged rachides, but in contrast to these, they are described as having fewflowered panicles. The flowers are polygamous, apetalous, and wind-pollinated. To my knowledge, there are no studies on the reproductive biology of this group and especially the Mexi-

can species are poorly described in the literature. This group requires a thorough revision, but here I recognise five taxa. Two taxa occur in SW USA and Mexico: F. greggii (Gregg’s ash) in Texas, Arizona, New Mexico and Mexico and F. gooddingii (Goodding’s ash) restricted to Arizona and Sonora in Mexico. F. gooddingii has an ITS sequence identical to that of F. greggii, and has recently been referred to as Fraxinus greggii A. Gray ssp. gooddingii (Little) E. Murray (Murray 1982). Three taxa occur only south of the USA/Mexican border: F. dubia, F. purpusii, and F. rufescens, the first two extending into Guatemala. Subsection Melioides There is little variation in the ITS sequence among most of the taxa in subsection Melioides, and in the strict consensus tree (Fig. 1) most of them occur in a polytomy. There is no molecular support (Fig. 2) for the so-called ‘white ash’ and ‘red ash’ complexes (Miller 1955). Many taxa have been described in Melioides and different authors accept different numbers of species (Sargent 1949, Little 1952, Dayton 1954, Miller 1955, Kartesz 1994, USDA 2001). The delimitation of taxa is also complicated by the presence of polyploids (e.g. in F. americana; Wright 1944, Black-Schaefer and Beckmann 1989) and possible hybrid species (Miller 1955). The group is in need of a revision, perhaps in the light of more informative molecular data, but this lies outside the aims of the present study. Provisionally, I follow Kartesz (1994) and USDA (2001) and accept nine species (Table 3), with a wide distribution in Central and North America. They are all medium-sized to large trees, deciduous, and dioecious. The unisexual flowers have a persistent calyx, but are apetalous and wind-pollinated. Female flowers consist of a calyx and one pistil, and male flowers of two stamens with elongated anthers and a small calyx. There are no rudimental organs of the opposite sex in the flowers. Taxa incertae sedis Three taxa, F. chiisanensis, F. spaethiana, and F. cuspidata have not been reliably placed by the molecular data. The insect-pollinated American species F. cuspidata has previously been placed in section Ornus (Lingelsheim 1920) and in section Dipetalae (Nikolaev 1981). The results from molecular data are inconclusive about either 17

position, but it is connected to F. chiisanensis and F. spaethiana, which is a very doubtful position. The other two wind-pollinated species are distributed in East Asia (Japan and Korea) and belong to section Fraxinus, without doubt, but they share morphological and molecular features with both subsection Fraxinus (mainly Asia) and Melioides (all American). F. chiisanensis is a tree endemic to Korea. The fruits have a persistent calyx and the leaf rachis is not winged. Surprisingly, this fact led Nakai (1929) to conclude that it belonged to section Dipetalae, which previously comprised only the American F. dipetala. However, the combination of calyx and non-winged rachis is also a feature of subsections Melioides and Fraxinus (as circumscribed here), where it fits much better. The relationships of F. chiisanensis were discussed by Nakaike (1972), who believed that the subsection to which this species belongs could not be determined without flowers (which he had not seen). Hypotheses on the putative hybrid origin of F. chiisanensis from F. mandshurica and F. chinensis ssp. rhynchophylla have been refuted by Noh et al. (1999) on the basis of RFLP patterns and by Min et al. (2001) on the basis of foliar flavonoids. Although the apetalous flowers of F. chiisanensis appear to be intermediate in morphology between F. mandshurica and F. chinensis ssp. rhynchophylla (Min et al. 2001), they have no other close similarities. Based on examinations of male and hermaphrodite flowers of F. chiisanensis (sent to me by W-K Min, Seoul National University, Korea), and photos in Min et al. (2001), I noted that the lateral inflorescences have either apetalous male flowers or hermaphrodite flowers with elongated anthers, quite like those of subsection Melioides. Although androdioecious in appearance, the anthers in the hermaphrodite flowers seem to be smaller than in the male ones, suggesting functional dioecy. The geographical distribution and presence of bisexual flowers, on the other hand, suggest a relationship with F. platypoda and F. mandshurica in subsection Fraxinus. Inspection of the sequence alignment shows that different parts of the ITS sequence of F. chiisanensis resemble sequences from either subsection Melioides or Fraxinus. Both the ITS and the chloroplast data are unclear regarding the position of this species, in either subsection Fraxinus or Melioides, and the consensus is a position as sister to F. spaethiana (and F. cuspidata) outside both these subsec18

tions. However, chemical support for a shared ancestry of F. chiisanensis and subsection Melioides has come from studies of foliar flavonoids. A recent study by Min et al. (2001) found the presence of advanced flavones in F. chiisanensis, in common with F. americana and F. pennsylvanica of subsection Melioides (Harborne and Green 1980, Black-Schaefer and Beckmann 1989). Other taxa of Fraxinus have only the plesiomorphic flavonols. The same unresolved situation exists for F. spaethiana, endemic to Japan. It has polygamous, apetalous, and wind-pollinated flowers with a calyx present only in pistillate flowers. It has been suggested to be a synonym of F. platypoda (Nakaike 1972, Wei and Green 1996), but the molecular data place these taxa in widely separated positions. F. spaethiana has not yet been investigated for the possible presence of flavones. Although the molecular data are not conclusive, morphological and chemical data suggest a close relationship for F. chiisanensis and F. spaethiana with subsection Melioides. Pending further data, I will leave them as incertae sedis within section Fraxinus. However, in the sectional key above they will both key out under subsection Fraxinus. F. cuspidata (fragrant ash) is a small entomophilous tree occurring in Mexico and southwestern USA. Based on its possession of a corolla it has been included in the otherwise Eurasian section Ornus. Although having four petals, the flowers are not similar to those of section Ornus. The corolla differs in being united at the base and with the two stamens, which are shorter than the petals. The flowers are borne terminally in lateral, leafy panicles developed from the axils of the leaves of the previous year, not in terminal panicles on current year shoots as in section Ornus. The flowers have a strong fragrance and are insect-pollinated. Nikolaev (1981) may have been close to its ancestry when he included F. cuspidata with F. dipetala in section Dipetalae (then monotypic). Although differing in petal number (F. dipetala have two and F. cuspidata four petals), they are both hermaphrodites with the plesiomorphic sympetalous corolla fused with the filaments, and a similar distribution area (the only two petaliferous species of Fraxinus in America). The odd position in the ITS tree is most certainly erroneous, probably because of a divergent ITS sequence, and based on its plesiomorphic flo-

ral morphology I have pushed it down one node as sister to section Fraxinus in Fig. 4. This procedure does not affect any of the evolutionary conclusions drawn from this study.

Evolution of wind-pollination in Fraxinus Anemophily has rarely been experimentally verified in species where it is assumed to occur (e.g. as by Bernardello et al. 1999, Goodwillie 1999, Otegui and Cocucci 1999, de Figueiredo and Sazima 2000). Instead, anemophily has been inferred on the basis of a number of traits considered to be typical of anemophilous taxa, the so-called anemophilous syndrome (Whitehead 1968 and 1983, Faegri and van der Pijl 1979). The most obvious traits in plants adapted to pollination by wind are flowers without (or with a very reduced) corolla, nectar, or scent. The flowers are often dichogamous and/or unisexual and borne either on the same (monoecy) or different individuals (dioecy). The anthers are usually large and well exposed on long slender filaments or in hanging catkins that can easily be moved by wind. The stigmata and/ or surrounding structures often have a morphology favouring efficient pollen trapping (Niklas 1985). Many are trees, but there are also shrubs and non-woody plants, e.g. the whole of Poaceae, that are wind-pollinated. They usually flower earlier than other plants, when they and/or the surrounding plants have no leaves, i.e., many are deciduous and occur in deciduous forests or open environments. Anemophily is especially common among temperate trees and there is a striking increase with latitude, and decrease with species diversity, in percentages of anemophilous trees (Regal 1982). Anemophily in habitats like tropical rainforests are rare (Bawa et al. 1985, Linskens 1996, Renner and Feil 1993, but see Williams and Adam 1999). High pollen-to-ovule (P/O) ratios are also a feature of wind-pollinated plants (Cruden 1977, Tormo et al. 1996). A number of plants assumed to be entomophilous may in fact be ambophilous, i.e., pollinated by both insects and wind. Some examples are Piper (de Figueiredo and Sazima 2000), Calluna (cited by Faegri and van der Pijl 1979), Salix (Vroege and Stelleman 1990, Tollsten and Knudsen 1992), Erica arborea (Aronne and Wilcock 1994), Luzula, Acer, Tilia, and other taxa reviewed by Cox (1991). All species of Fraxinus assumed to be anemo-

philous exhibit the above features. However, even some species assumed to be entomophilous display one or more of the anemophilous features. The species of section Ornus produce large quantities of pollen which attract a number of different unspecialised pollinators, mainly small beetles, and their many-flowered inflorescences are situated so that the wind can catch and transport pollen between the flowers (Wallander and Dahl ms.). So far, only F. ornus (Wallander and Dahl ms.) and F. lanuginosa (Ishida and Hiura 1994) have been shown to release pollen into the air, but whether any significant proportion of this pollen successfully reach other conspecific stigmata is unknown. The P:O ratios of F. ornus and F. lanuginosa are high (Wallander and Dahl ms.) and lie well within the range of typical anemophiles (Tormo et al. 1996). Regardless of the relative efficacy of the ambophilous system, it appears to be a clear exaptation for the transition to anemophily. In subtribe Oleinae, the sister group to Fraxinus, entomophilous, ambophilous, and anemophilous taxa are present, and it appears that anemophily has arisen independently at several occasions there as well (Wallander ms.). The ancestral condition in Fraxinus may be entomophily or ambophily, and there are three independent origins of anemophily, in the sections Dipetalae, Ornus, and Fraxinus (Fig. 4). Anemophilous taxa within these three groups display somewhat different adaptations to wind-pollination, but apetalous flowers are common to all of them. In section Fraxinus, all 23 recognised species are wind-pollinated and they constitute a monophyletic group (including F. chiisanensis and F. spaethiana). They clearly represent one origin of anemophily, but it is not clear what the immediate ancestral states might have been. However, all extant taxa have apetalous flowers in panicles that develop from lateral buds on the shoot of the previous year. The current year’s leafy shoots develop from terminal buds. The floral buds open before the leaf buds and the flowering is completed by the time the leaves reach full size. Thus, in this group there is no overlap between flowering and leafing. Most taxa are morphologically polygamous, but some are dioecious (discussed in more detail below). In a few taxa of the subsection Fraxinus, the floral reduction has gone even farther in that some species lack calyx as well. 19

However, the calyx is quite small in all taxa of the genus and even where it is present it is unlikely to interfere with pollen dispersal or capture. Most of the species are large trees but the five taxa of subsection Pauciflorae and F. xanthoxyliodes of subsection Sciadanthus are shrubs or small trees. These six taxa are xerophytes and their small size (and other traits such as coriaceous leaves) is probably an adaptation to the dry or almost desert-like areas they inhabit. Although in subtropical areas, they experience a pronounced dry season during which they are leafless and flowering takes place. The large trees occur in temperate, deciduous forests and flower during the leafless spring. Within section Ornus there are five anemophilous species (“Ornaster”) and they represent an independent origin of adaptations to windpollination. Like their petaliferous and ambophilous ancestor, they have combined floral and vegetative terminal buds. Consequently, the inflorescences and leaves emerge at the same time. The entomophilous or ambophilous taxa do not open their flowers until the leaves have reached their full size. In contrast, the anemophilous species begin flowering while the leaves are still not fully expanded. Another difference is tree size. All anemophilous taxa and two of the closely related ambophilous taxa (F. floribunda and F. paxiana) are much larger trees than the rest in section Ornus. Thus, the evolution of anemophily in this group was preceded by an increase in tree height, a trait that, in addition to ambophily, can be interpreted as an exaptation. The third case involves two species in section Dipetalae, F. anomala and F. quadrangulata. They have lost their petals and scent, but have to some extent retained the plesiomorphic breeding system. Although mostly hermaphrodites, it appears that their sexuality is labile

and under selection for unisexuality. In comparison to F. anomala and the entomophilous F. dipetala, F. quadrangulata has attained a much larger size and occur in the deciduous forests in central and eastern USA. In Fraxinus, four intergrading breeding systems are present (Table 4), which are clearly correlated with the evolution of anemophily. Of the 16 species in section Ornus, three are hermaphrodites, ten morphologically androdioecious (Wallander and Dahl ms.), and three dioecious or functionally dioecious. The three species of section Dipetalae and F. cuspidata are hermaphrodites. In section Fraxinus, the three subsections Fraxinus, Sciadanthus, and Pauciflorae, and F. chiisanensis and F. spaethiana, consist of polygamous species, but in Melioides all species are dioecious. In subsection Fraxinus, two of the morphologically polygamous species are functionally dioecious. How are these breeding systems correlated with the pollination systems? It seems that after two of the three independent origins of anemophily, in section Ornus and Fraxinus, there have been no less than three independent shifts to dioecy. Within Ornus, three of the five wind-pollinated species (“Ornaster”) have shifted from androdioecy to a functionally dioecious breeding system. In section Fraxinus, dioecy have originated in subsection Melioides and in the species pair F. mandshurica and F. platypoda in subsection Fraxinus. In addition, a fourth case might be on its way in section Dipetalae. The common feature for all these cases is the presence of a dimorphic breeding system with unisexual males (either androdioecy or polygamy) before the transition to anemophily. In all cases, dioecy have evolved after the transition to anemophily and can be interpreted as an adaptation that evolved in response to selection for increased efficiency and function of that system.

Table 4. Number of Fraxinus species displaying various breeding systems, in connection with either anemophily or entomophily and ambophily. Pollination system

anemophily entomo/ambophily Total

20

Breeding system

Total no. of species

hermaphroditism

androdioecy

polygamy

dioecy

2 5 7

2 8 10

12 0 12

14 0 14

30 13 43

Dioecy – an adaptation to anemophily Dioecy occurs in about 6 % of the flowering plants and has been correlated with a number of characters such as fleshy fruits, woodiness, climbing habit, small and white to yellowish or greenish flowers, unspecialised insect pollinators, wind-pollination, etc. (reviewed by Renner and Ricklefs 1995). The evolution of dioecy from hermaphroditism has generated much debate and a number of hypotheses have been put forward to explain these associations (e.g. Charlesworth and Charlesworth 1978, Bawa 1980, Lloyd 1980 and 1982, Bawa and Beach 1981, Thomson and Barrett 1981, Givnish 1980 and 1982, Ross 1982, Charlesworth 1984, Thomson and Brunet 1990, Freeman et al. 1997, Charlesworth 1999). There are two main explanations: selection for outcrossing or influences of ecological factors (e.g. sexual selection, division of labour leading to optimal resource allocation, decreased intraspecific competition, pollinator attraction to massive pollen crops and frugivore attraction to massive fruit crops) (Givnish 1982). Selection for outcrossing is the traditional explanation (see discussion in Bawa 1980, Givnish 1982), but this may not be the main factor driving the evolution of unisexual flowers in wind-pollinated plants. I have previously showed that in Fraxinus, all species with hermaphrodite flowers are protogynous (Wallander and Dahl ms.). The ones that have been investigated are self-compatible, but it is yet unknown whether selfing results in any inbreeding depression. In at least three species, pollen from male flowers has been shown to have a competitive advantage over pollen from hermaphrodite flowers (Ishida and Hiura 1998, Wallander and Dahl ms.). Furthermore, in one of the wind-pollinated species, F. excelsior, it has even been shown that at the time of pollination there are no mature megagametophytes in the ovules (Wallander and Dahl ms.), further accentuating the crosspollen advantage through pollen tube competition in the style. Because hermaphrodites are protogynous, neither stigma clogging nor selfing is any problem. Only if cross-pollination fails may selfing occur, which may be viewed as a means of reproductive assurance. Therefore, because they already possess mechanisms ensuring outcrossing, I argue that selection for outcrossing in Fraxinus has not been

the main selective force in the evolution of dioecy. Another indication is given by the presence of the monoecious breeding system among many other wind-pollinated plants, e.g. in Alnus, Betula, Corylus, and Quercus. Since monoecy may still lead to selfing (Bertin 1993), this fact shows that selection for outcrossing is not the main factor. Therefore, it appears that something else is driving the evolution of unisexual flowers. In a study similar to the present, on the evolution of anemophily in Oleaceae (Wallander ms.), I review the hypotheses that are based on various ecological factors favouring unisexual flowers in wind-pollinated plants. The main conclusion from that study is that selection for sexual specialisation has played a major role in the evolution of dioecy. The arguments leading to that conclusion is summarised below. In wind-pollinated plants, the requirements on floral structure and position for optimal pollen dispersal and pollen capture differ between the two sexes (Frankel and Galun 1977, Niklas 1985, Freeman et al. 1997). In addition, there may also be a spatial interference between stamens and pistil in a bisexual and wind-pollinated flower (Faegri and van der Pijl 1979). This may be directly disadvantageous and lead to conflicts between optimal male and female function (Lloyd 1982). Furthermore, because wind-pollination is a rather inefficient (in the sense that it is wasteful or imprecise) means for plants to spread their male gametes, there is a strong selection pressure for maximising pollen production, which increases the likelihood for pollen to reach conspecific stigmas (Frankel and Galun 1977). Consequently, those plants that divert more resources to pollen production, either through more pollen per anther, more anthers per flower and/or more flowers per individual, may have higher paternal success. In fact, in several wind-pollinated species of Fraxinus it has been shown that male inflorescences have significantly more flowers than hermaphrodite and female ones (Wallander and Dahl ms.). Increased pollen production, in terms of more flowers, may be in conflict with investments to female function. Thus, there may also be a conflict in optimal allocations to the paternal and maternal functions. It is thus obviously impossible for one flower to optimise at the same time pollen dispersal and pollen capture. Consequently, anemophilous plants may respond to the selection for 21

separate sexes due to the different demands on optimal male and female function in a flower, since they are no longer affected by selection pressures for optimising floral displays and reward and maximising the returns from insect visits to bisexual flowers. Thus, selection for sexual specialisation leads to a dioecy through total or partial reduction of the opposite sex in the resulting unisexual flowers. This study shows that dioecy is clearly an adaptation in wind-pollinated species of Fraxinus. However, the numerous correlations that have been found between the dioecy and anemophily (Linder 1998), have not been able to tell which evolved first, i.e. if dioecy is an adaptation to anemophily, or vice versa. Except for the large-scale study by Linder (1989), I know of no studies that explicitly have shown the order of events relating to the evolution of anemophily. However, de Jong (1976) believes that dioecy preceded anemophily in Acer L. (Aceraceae), based on the fact that there are dioecious species in Acer that are not anemophilous. Kaplan and Mulcahy (1971) also believe that dioecy precedes anemophily in Thalictrum L. (Ranunculaceae), based on the illogical argument that dioecy evolved to promote outcrossing and when females were not visited by insects (because they lacked the pollen attractant) anemophily was selected for. The few other studies that have been based on phylogenies are more or less inconclusive (e.g. Weller et al. 1998, Linder 1998). However, Linder (1988) showed that dioecy have evolved several times in the wind-pollinated group Poales.

Androdioecy – an adaptation to ambophily? It is striking that most of the species in the section Ornus are (phenotypically) androdioecious and only three are hermaphroditic. Unfortunately, I only succeeded in getting DNA sequences from one of the hermaphrodite species, F. raibocarpa, which is sister to the rest of this group. The three hermaphrodites may or may not be closely related, but regardless of which, the most probable evolutionary scenario with regard to the origin of androdioecy in section Ornus is an evolution from hermaphroditic ancestors. In most of the androdioecious species, males retain a vestigial pistil (Wallander and Dahl ms.), an evidence of a hermaphrodite ancestry. This fact, together with the sister 22

group relationship of F. raibocarpa to the rest of this group, makes this the most probable evolutionary route. According to ESS models developed by Lloyd (1975) and Charlesworth (1984), the evolution and maintenance of androdioecy requires that males have at least twice the fitness of the male function in hermaphrodite plants, and the fitness demands increase as the selfing rate increases. Because of these high demands on male fitness, Charlesworth (1984) believe functional androdioecy is very unlikely to evolve, and most cases of morphologically androdioecious taxa have been argued to be cryptically dioecious, with dysfunctional male organs in hermaphrodite flowers rendering them functionally female (Charlesworth 1984, Mayer and Charlesworth 1991). In most of the taxa they list, many are dependent on pollinators attracted by pollen as a reward, and it is suggested that one function for the retained stamens in ‘female’ flowers is to attract pollinators. However, two species of Fraxinus have recently been confirmed to be functionally androdioecious: F. ornus (Dommée et al. 1999, Wallander and Dahl ms.) and F. lanuginosa (Ishida and Hiura 1998). Two other taxa have been investigated in the field and are suspected to be functionally androdioecious as well, viz. F. sieboldiana (Dr. Junko Okazaki, pers. comm.) and F. longicuspis (Wallander and Dahl ms.). These partly wind-pollinated and partly insectpollinated plants seem to have benefited from having a dimorphic breeding system with unisexual male and bisexual flowers. The flowers do not produce nectar and are visited mainly by pollen-feeding or pollen-collecting insects (Wallander and Dahl ms.). Therefore, an increase in pollen production would increase the attractiveness to insects (Bawa 1980), and at the same time make the ambophilous strategy more efficient (Frankel and Galun 1977). Following the same explanation as for evolution of dioecy in anemophilous plants, purely male plants may be selected for if they can reallocate more resources to increased pollen production, either through more flowers per inflorescence and/ or more pollen per flower. In all species that have been investigated, there is no difference in anther size between males and hermaphrodites, and the hermaphrodites have been shown to produce viable and fully functional pollen (Wallander and Dahl

ms.). However, although pollen germination experiments have revealed no significant differences in germination or growth rates in vitro, there seem to be a differential pollination and fertilisation success in vivo. Pollen from hermaphrodite plants is less successful in competition with pollen from male plants in F. lanuginosa (Ishida and Hiura 1998). Thus, pollen from males seems to have a higher viability and this, in connection with a higher pollen production, may contribute to the necessary doubled fitness. However, the superior male function in male plants compared to hermaphrodite ones should not be used as an argument that the hermaphrodites are in fact functionally female. Hermaphrodites in the investigated species of Fraxinus do have a potential male function, even if it is inferior in competition with males, and pollen from hermaphrodites may sire seeds occasionally. Why then do the hermaphrodite flowers in androdioecious ‘female’ plants retain stamens? There are two basic explanations for this: either the stamens are selected for, or they have not been selected against. A simple argument for the latter explanation is that since all of the investigated species are protogynous, stigma clogging and/or self-pollination is not a problem. On the contrary, since at least some of them are self-compatible (F. ornus and F. lanuginosa), they may self-pollinate if cross-pollination fails. This may be one factor favouring retention of stamens a s a means of reproductive assurance. The self-compatible hermaphrodites may also be favoured during episodes of colonisation (Pannell 1997b). Another selection pressure for keeping the stamens, in these nectarless speices, may be pollinator attraction (Charlesworth 1984). But more importantly, if hermaphrodites actually do contribute to part of the progeny in the population, this may be a strong enough selection pressure for keeping the male function, even if it is inferior compared to the specialised males. Other cases of functional androdioecy in wind-pollinated plants have lately been confirmed, Datisca glomerata (Presl) Baill. (Datiscaceae) (Liston et al. 1990), Mercurialis annua L. (Euphorbiaceae) (Pannell 1997a) and Schizopepon bryoniaefolius (Cucurbitaceae) (Akimoto et al. 1999). The two former are wind-pollinated herbs and the latter an insect-pollinated annual, all three with colonising habits. Male frequencies in these populations are much less than

50%, sometimes zero, and they conform nicely to patterns predicted by the metapopulation model by Pannell (2000). However, this model cannot explain the coexistence of equal proportions of males and hermaphrodites in androdioecious species of Oleaceae, and Pannell (2000) suggest that different processes are driving these systems. In Oleaceae, several other genera than Fraxinus contain species that are androdioecious and possible factors for this high incidence are discussed elsewhere (Wallander ms.).

Conclusions The ITS sequence data have (with the exception of F. cuspidata) yielded a reliable estimate of the phylogeny of Fraxinus. Based on the phylogenetic estimate, in combination with morphology, I propose a revised infrageneric classification of Fraxinus (Table 3), consisting of three sections and four subsections. Mapping reproductive data on the phylogenetic tree revealed interesting trends and patterns of evolutionary history in the genus Fraxinus. This contributes new empirical data relevant to general theories on evolution of wind-pollination and breeding systems. I have shown that wind-pollination has three separate origins in Fraxinus and that it is preceded by ambophily. Dioecy is correlated with anemophily and has evolved after the origin of anemophily in three instances. In the section Ornus, dioecy has originated via androdioecy, which in turn has a hermaphroditic origin. This is the first study that shows the evolution of dioecy from hermaphrodites via the rare androdioecious pathway. In the wind-pollinated section Fraxinus, dioecy has evolved twice from polygamous ancestors. The interpretations of the results are that ambophily and a dimorphic breeding system with unisexual males, either androdioecy or polygamy, were some of the exaptations for the evolution of anemophily in Fraxinus. Dioecy, on the other hand, has followed in evolutionary sequence and is interpreted as an adaptation that evolved in response to anemophily for increasing the efficiency and function of that system. Having lost the selection pressure for cosexual flowers imposed by the previous dependence on pollinators, this has lead to a dioecious system through total or partial reduction of the opposite sex in the resulting uni23

sexual flowers. Thus, dioecy in anemophilous plants is here interpreted as a selection for sexual specialisation, not selection for inbreeding avoidance, due to the different requirements for optimal male and female function in anemophilous flowers. The presence of the rare androdioecious breeding system in many of the ambophilous Fraxinus species is interpreted as a compromise between selection for maximum pollen production and dispersal (unisexual males) and the selection pressure imposed by the pollen-feeding insects on visits to ‘female’ flowers (retention of partial or fully functional stamens in the hermaphrodite flowers as an attractive pollen source). Furthermore, the retention of stamens in the self-compatible and protogynous hermaphrodites may provide reproductive assurance in case of failure of cross-pollination and during colonisation episodes. They may also contribute to cross-pollination when there is low competition with males.

ACKNOWLEDGEMENTS I am grateful to the following individuals and institutions for help and providing me with floral and/ or leaf material: Dr. Tsutomu Enoki (University of the Ryukyus, Okinawa, Japan), Woong-Ki Min (Seoul National University, Korea), Satoshi Nanami and Takashi Osono (Kyoto University, Japan), Dr Junko Okazaki (Osaka Kyoiku University, Japan), Xu Youming (Huazhong Agricultural University, Hubei, P. R. China), and Johanne Maad and Lars-Gunnar Reinhammar (Uppsala University, Sweden). Herbaria (C, E, GB, MO, NY, S, and UPS) that lent material and gave permission to extract DNA are gratefully acknowledged, and especially Mark Chase at the Jodrell Laboratory at Kew Gardens, who provided DNA extracts from living and herbarium material at K. I thank Mari Källersjö and Steve Farris for performing the jackknife analyses, and Olga Khitun for translating information on Russian species of Fraxinus. For critically commenting on previous drafts of this paper, I thank Lennart Andersson, Åslög Dahl, and Roger Eriksson. I also thank Victor. A. Albert for valuable discussions and for inviting me to use the DNA sequencing facilities of the Cullman Program for Molecular Systematics, the New York Botanical Garden and the American Museum Natural of History. This doctoral research was supported by the Lewis B. and Dorothy Cullman Foundation, The Royal Swedish Academy of Sciences, Kungliga och Hvitfeldtska Stiftelsen, Wilhelm och Martina Lundgrens Vetenskapsfond, and Collianders stiftelse.

24

REFERENCES Akimoto J., Fukuhara T., Kikuzawa K. (1999) Sex ratios and genetic variation in a functionally androdioecious species, Schizopepon bryoniaefolius (Cucurbitaceae). American Journal of Botany 86: 880-886. do Amaral Franco J., da Rocha Afonso M. L. (1972) In: Tutin T. G. et al. (eds.) Flora Europaea 3: 53-54. Cambridge University Press. Anderson E., Turrill W. B. (1938) Statistical studies on two populations of Fraxinus. New Phytologist 37: 160-172. Andersson L., Rova J. H. E. (1999) The rps16 intron and the phylogeny of the Rubioideae (Rubiaceae). Plant Systematics and Evolution 214: 161-186. Aronne G., Wilcock C. C. (1994) Reproductive characteristics and breeding system of shrubs of the Mediterranean region. Functional Ecology 8: 6976. Barnes B. V., Wagner W. H. (1981) Michigan Trees. The University of Michigan Press, USA. Bawa K. S. (1980) Evolution of dioecy in flowering plants. Annual Review of Ecology and Systematics 11: 15-39. Bawa K. S., Beach J. H. (1981) Evolution of sexual systems in flowering plants. Annals of the Missouri Botanical Garden 68: 254-274. Bawa K. S., Bullock S. H., Perry D. R., Coville R. E., Grayum M. H. (1985) Reproductive biology of tropical lowland rain forest trees. II. Pollination systems. American Journal of Botany 72: 346-356. Bernardello G., Anderson G. J., Lopez S. P., Cleland M. A., Stuessy T. F., Crawford D. J. (1999) Reproductive biology of Lactoris fernandeziana (Lactoridaceae). American Journal of Botany 86: 829-840. Bertin R. I. (1993) Incidence of monoecy and dichogamy in relation to self-fertilization in angiosperms. American Journal of Botany 80: 557-560. Black-Schaffer C. L., Beckmann R. L. (1989) Foliar flavonoids and the determination of ploidy and gender in Fraxinus americana and Fraxinus pennsylvanica (Oleaceae). Castanea 54: 115-118. Charlesworth B., Charlesworth D. (1978) A model for the evolution of dioecy and gynodioecy. American Naturalist 112: 975-997. Charlesworth D. (1984) Androdioecy and the evolution of dioecy. Biological Journal of the Linnean Society 22: 333-348. Charlesworth D. (1993) Why are unisexual flowers associated with wind pollination and unspecialized pollinators? American Naturalist 141: 481-490. Charlesworth D. (1999) Theories on the evolution of dioecy. In: Geber M. A., Dawson T. E., Delph L. F. (eds.) Gender and sexual dimorphism in flowering plants, pp. 33-60. Springer-Verlag, Berlin Heidelberg, Germany. Cox P. A. (1991) Abiotic pollination: an evolutionary escape for animal-pollinated angiosperms. Philosophical transactions of the Royal Society of London (B) 333: 217-224.

Cruden R. W. (1977) Pollen-ovule ratios: a conservative indicator of breeding systems in flowering plants. Evolution 31: 32-46. Dayton W. A. (1954) Some more notes on United States ashes (Fraxinus). Journal of the Washington Academy of Science 44: 885-390. Dommée B., Geslot A., Thompson J. D., Reille M., Denelle N. (1999) Androdioecy in the entomophilous tree Fraxinus ornus (Oleaceae). New Phytologist 143: 419-426. Faegri K., van der Pijl L. (1979) Abiotic pollination. In: The principles of pollination ecology, pp. 3440. Pergamon Press. de Figueiredo R. A., Sazima M. (2000) Pollination biology of Piperaceae species in southeastern Brazil. Annals of Botany 85: 455-460. Frankel R., Galun E. (1977) Pollination mechanisms, reproduction and plant breeding. Monographs on theoretical and applied genetics, 2. SpringerVerlag, Berlin. Freeman D. C., Lovett Doust J., El-Keblawy A., Migla K. J., McArthur E. D. (1997) Sexual specialization and inbreeding avoidance in the evolution of dioecy. The Botanical Review 63: 65-92. Gielly L., Taberlet P. (1994) Chloroplast DNA polymorphism at the intrageneric level and plant phylogenies. Comptes Rendus de l’Academie des Sciences Serie III Sciences de la Vie 317: 685-692. Givnish T. J. (1980) Ecological constraints on the evolution of breeding systems in seed plants: dioecy and dispersal in gymnosperms. Evolution 34: 959972. Givnish T. J. (1982) Outcrossing versus ecological constraints in the evolution of dioecy. American Naturalist 119: 849-865. Goodwillie C. (1999) Wind pollination and reproductive assurance in Linanthus parviflorus (Polemoniaceae), a self-incompatible annual. American Journal of Botany 86: 948-954. Gould J. S., Vrba E. S. (1982) Exaptation – a missing term in the science of form. Paleobiology 8: 4-15. Green P. S. (1991) Notes on Oleaceae from “Flora Mesoamericana”. Kew Bulletin 46: 273-276. Grohmann F. (1974) Oleaceae. Flora of West Pakistan 59: 1-27. Hara H. (1956) Critical notes on some type specimens of East-Asiatic plants in foreign herbaria (7). Journal of Japanese Botany 31: 57-60. Hara H. (1982) Oleaceae. In: Hara H., Chater A. O., Williams L. H. J. (eds.) An enumeration of the flowering plants of Nepal 3: 79-82. Trustees of British Museum, London. Harborne J. B., Green P. S. (1980) A chemotaxonomic survey of flavonoids in leaves of the Oleaceae. Botanical Journal of the Linnean Society 81: 155167. Ishida K., Hiura T. (1994) Reproductive ecology in Fraxinus lanuginosa. II. Pollen vectors and pollen viability. Transactions of the Hokkaido Branch of the Japanese Forest Society 42: 61-63. Ishida K., Hiura T. (1998) Pollen fertility and flower-

ing phenology in an androdioecious tree, Fraxinus lanuginosa (Oleaceae), in Hokkaido, Japan. International Journal of Plant Sciences 159: 941-947. Jeandroz S., Frascaria-Lacoste N., Bousquet J. (1996) Molecular recognition of the closely related Fraxinus excelsior and F. oxyphylla (Oleaceae) by RAPD markers. Forest Genetics 3: 237-242. Jeandroz S., Roy A., Bousquet J. (1997) Phylogeny and phylogeography of the circumpolar genus Fraxinus (Oleaceae) based on internal transcribed spacer sequences of nuclear ribosomal DNA. Molecular Phylogenetics and Evolution 7: 241-251. de Jong P. C. (1976) Flowering and sex expression in Acer L. – a biosystematic study. Medelingen Landbouwhogeschool Wageningen Nederland 76: 1201. de Jong P. C. (1990) The genus Fraxinus. Dendroflora 26: 30-39. Kaplan S. M., Mulcahy D. L. (1971) Mode of pollination and floral sexuality in Thalictrum. Evolution 25: 659-668. Kartesz J. T. (1994) A synonymized checklist of the vascular flora of the United States, Canada and Greenland, vol. 1 (checklist), pp. 393-394. 2nd edition. Timber Press, Portland, Oregon, USA. Kitagawa M. (1979) Neo-Lineamenta Florae Manshuricae, pp. 510-511. A. R. Ganther Verlag K.-G., Vaduz. Kitamura S., Murata G. (1971) Coloured illustrations of woody plants of Japan. Vol. 1, pp. 74-93. Hoikusha Publ., Osaka, Japan. Linder H. P. (1998) Morphology and the evolution of wind pollination. In: Owens S. J., Rudall P. J. (eds.) Reproductive Biology, pp. 123-135. Royal Botanic Gardens, Kew. Lingelsheim A. (1907) Vorarbeiten zu einer Monographie der Gattung Fraxinus. Engler’s Botanische Jahrbücher 40: 185-223. Lingelsheim A. (1920) Oleaceae-Oleoideae-Fraxineae. In: Engler, A. Das Pflanzenreich IV 243: 1-61. Linskens H. F. (1996) No airborne pollen within tropical rain forests. Proceedings of the Koninklijke Nederlandse Akademie van Wetenschappen 99: 175-180. Liston A., Rieseberg L. H., Elias T. H. (1990) Functional androdioecy in the flowering plant Datisca glomerata. Nature 343: 641-642. Little E. L. (1952) Notes on Fraxinus (ash) in the United States. Journal of the Washington Academy of Science 42: 369-380. Lloyd D. G. (1975) The maintenance of gynodioecy and androdioecy in Angiosperms. Genetica 45: 325-339. Lloyd D. G. (1980) The distributions of gender in four angiosperm species illustrating two evolutionary pathways to dioecy. Evolution 34: 123-134. Lloyd D. G. (1982) Selection of combined versus separate sexes in seed plants. American Naturalist 120: 571-585. Maddison W. P., Maddison D. R. (1999) MacClade version 3.08a. Sinauer Associates, Inc. Publishers,

25

Sunderland, MA, USA. Mayer S. S., Charlesworth D. (1991) Cryptic dioecy in flowering plants. Trends in Ecology and Evolution 6: 320-325. Metcalfe C. R. (1938) Anatomy of Fraxinus oxycarpa and F. pallisae. Kew Bulletin 0: 258-262. Miller G. N. (1955) The genus Fraxinus, the ashes, in North America, North of Mexico. Cornell University Agricultural Experiment Station Memoir 335: 1-64. Min W. K., Jeon J. I., Chang, C. S. (2001) A taxonomic reconsideration of Fraxinus chiisanensis (Oleaceae) in Korea. Journal of the Korean Forest Society 90: 266-267. Ming J., Liao H. R. (1998) On the present situation of Fraxinus hupehensis and its sustainable utilization. Journal of Plant Resources and Environment 7: 1922. Murray E. (1968) Oleaceae. In: Rechinger, K. H. (ed.) Flora Iranica 52: 1-11. Murray E. (1982) Notae Spermatophytae. Kalmia 12: 20. Nakai T. (1921) Oleaceae. Flora Sylvatica Koreana 10: 13-61. Nakai T. (1929) Plantae Japanice et Koreanae. Botanical Magazine (Tokyo) 43: 446. Nakaike T. (1972) A synoptical study on the genus Fraxinus from Japan, Korea and Formosa. Bulletin of the National Science Museum Tokyo 15: 475512. Nickrent D. L., Schuette K. P., Starr E. M. (1994) A molecular phylogeny of Arceuthobium (Viscaceae) based on nuclear ribosomal DNA internal transcribed spacer sequences. American Journal of Botany 81: 1149-1160. Niklas K. J. (1985) The aerodynamics of wind pollination. The Botanical Review 51: 328-386. Nikolaev E. V. (1981) The genus Fraxinus (Oleaceae) in the flora of the USSR. Botanicheskii Zhurnal 66: 1419-1432. Noh E. W., Choi Y. I., Lee E. J., Song W. S. (1999) RFLP analysis of nuclear ribosomal DNA repeats in several selected species of Fraxinus in Korea. FRI Journal of Forest Science 61: 1-8. Otegui M., Cocucci A. (1999) Flower morphology and biology of Myrsine laetevirens, structural and evolutionary implications of anemophily in Myrsinaceae. Nordic Journal of Botany 19: 71-85. Pannell J. (1997a) Widespread functional androdioecy in Mercurialis annua L. (Euphorbiaceae). Biological Journal of the Linnean Society 61: 95-116. Pannell J. (1997b) The maintenance of gynodioecy and androdioecy in a metapopulation. Evolution 51: 10-20. Pannell J. (2000) A hypothesis for the evolution of androdioecy: the joint influence of reproductive assurance and local mate competition in a metapopulation. Evolutionary Ecology 14: 195-211. Regal P. J. (1982) Pollination by wind and animals: Ecology of geographic patterns. Annual Review of Ecology and Systematics 13: 497-524.

26

Rehder A. (1951) Oleaceae. Manual of cultivated trees and shrubs hardy in North America, exclusive of the subtropical and warmer temperate regions. 2nd ed., pp. 765-793. The Macmillan company, New York, USA. Renner S. S., Feil J. P. (1993) Pollinators of tropical dioecious angiosperms. American Journal of Botany 80: 1100-1107. Renner S. S., Ricklefs R. E. (1995) Dioecy and its correlates in the flowering plants. American Journal of Botany 82: 596-606. Ross M. D. (1982) Five evolutionary pathways to subdioecy. American Naturalist 119: 297-318. Sargent C. S. (1949) Oleaceae. In: Manual of the trees of North America (exclusive of Mexico), vol. 2, pp. 832-858. Scheller H. (1977) Kritische Studien über die kultivierten Fraxinus-Arten. Mitteilungen der Deutschen Dendrologischen Gesellschaft 69: 49162. Standley P. C. (1924) Oleaceae. In: Trees and shrubs of Mexico. Contributions from the US National Herbarium 23: 1132-1141. Sun S. S. (1985) Studies on the genus Fraxinus L. (Oleaceae) in China (I). Bulletin of Botanical Research 5: 37-67. Swofford D. L. (2000) PAUP*: phylogenetic analysis using parsimony (* and other methods), version 4. Sinauer, Sunderland, Massachusetts, USA. Thomson J. D., Barrett S. C. H. (1981) Selection for outcrossing, sexual selection and the evolution of dioecy in plants. American Naturalist 118: 443-449. Thomson J. D., Brunet, J. (1990) Hypotheses for the evolution of dioecy in seed plants. Trends in Ecology and Evolution 5: 11-16. Tollsten L., Knudsen J. (1992) Floral scent in dioecious Salix (Salicaceae) – a cue determining the pollination system? Plant Systematics and Evolution 182: 229-237. Tormo M. R., Munoz R. A., Silva P. I., Gallardo L. F. (1996) Pollen production in anemophilous trees. Grana 35: 38-46. USDA, NRCS. (2001) The PLANTS Database, Version 3.1 (http://plants.usda.gov/). National Plant Data Center, Baton Rouge, LA, USA. Vassiliadis C. (1999) Evolution et maintien de l’androdioécie: Etude théorique et approaches expérimentales chez Phillyrea angustifolia L. PhD Dissertation, Université des Sciences at Technologies de Lille, France. Vassiljev V. N. (1952) Fraxinus. In: Komarov V. L. (ed.) Flora URSS 18: 485-502. (English translation by N. Landau (1967) In: Flora of the USSR 18: 357-370.) Vines R. A. (1984) Fraxinus. In: Trees, shrubs, and woody vines of the Southwest, pp. 859-869. 5th ed. University of Texas Press, Austin. Vroege P. W., Stelleman P. (1990) Insect and wind pollination in Salix repens L. and Salix caprea L. Israel Journal of Botany 39: 125-132. Wallander E. (manuscript) Evolution of wind-pollination and gender specialisation in Oleaceae –

exaptations and adaptations. Wallander E., Albert V. A. (2000) Phylogeny and classification of Oleaceae based on rps16 and trnL-F sequence data. American Journal of Botany 87: 1827-1841. Wallander E., Dahl Å. (manuscript) Studies on the reproductive ecology of F. excelsior, F. ornus and F. longicuspis, and a comparison with some other species of Fraxinus (Oleaceae). Wei Z. (1992) Fraxinus. In: Chang M., Qui L. (eds.). Flora Reipublicae Popularis Sinicae 61: 5-39. Wei Z., Green P. S. (1996) Fraxinus. In: Wu Z., Raven P. H. (eds). Flora of China 15: 273-279. Science Press & Missouri Botanical Garden. Weller S. G., Sakai A. K., Rankin A. E., Golonka A., Kutcher B., Ashby K. E. (1998) Dioecy and the evolution of pollination systems in Schiedea and Alsinidendron (Caryophyllaceae: Alsinoideae) in the Hawaiian Islands. American Journal of Botany 85: 1377-1388. Whitehead D. R. (1968) Wind pollination in the angiosperms: evolutionary and environmental considerations. Evolution 23: 28-35.

Whitehead D. R. (1983) Wind pollination: Some ecological and evolutionary perspectives. In: Real L. (ed.) Pollination Biology, pp. 97-108. Academic Press, London, U.K. Williams G., Adam P. (1999) Pollen sculpture in subtropical rain forest plants: Is wind pollination more common than previously suspected? Biotropica 31: 520-524. Wojciechowski M. F., Sanderson M. J., Baldwin B. G., Donoghue M. J. (1993) Monophyly of aneuploid Astragalus (Fabaceae): evidence from nuclear ribosomal DNA internal transcribed spacer sequences. American Journal of Botany 80: 711-722. Wright J. W. (1944) Genotypic variation in white ash. Journal of Forestry 42: 489-495. Yaltirik F. (1978) Oleaceae. In: Davis P. H. (ed.) Flora of Turkey 6: 145-158. Edinburgh University Press, Edinburgh. Yamazaki T. (1993) Oleaceae. In: Iwatsuki et al. (eds.) Flora of Japan IIIa: 122-135. Kodansha Ltd., Japan.

27

Paper I

Paper II

Paper III

Paper IV

Paper V

Paper VI

American Journal of Botany 87(12): 1827–1841. 2000.

PHYLOGENY

AND CLASSIFICATION OF

OLEACEAE

BASED ON RPS16 AND TRNL-F SEQUENCE DATA1

EVA WALLANDER2,4

VICTOR A. ALBERT3

AND

2 Botanical Institute, Go¨teborg University, P.O. Box 461, SE-405 30 Go¨teborg, Sweden; and Department of Biological Sciences, The University of Alabama, Tuscaloosa, Alabama 35487 USA.

3

Phylogenetic relationships among 76 species of Oleaceae, representing all 25 recognized genera of the family, were assessed by a cladistic analysis of DNA sequences from two noncoding chloroplast loci, the rps16 intron and the trnL-F region. Consensus trees from separate and combined analyses are congruent and agree well with nonmolecular data (chromosome numbers, fruit and wood anatomy, leaf glycosides, and iridoids). The two debated genera Dimetra and Nyctanthes, previously suggested to belong to Verbenaceae (sensu lato) or Nyctanthaceae, are shown to belong to Oleaceae, sister to the hitherto genus incertae sedis Myxopyrum. This clade is also supported by anatomical and chemical data. The subfamily Jasminoideae is paraphyletic, and a new classification is presented. The subfamily level is abandoned, and the former Jasminoideae is split into four tribes: Myxopyreae (Myxopyrum, Nyctanthes, and Dimetra), Fontanesieae (Fontanesia), Forsythieae (Abeliophyllum and Forsythia), and Jasmineae (Jasminum and Menodora). The tribe Oleeae (previous subfamily Oleoideae) is clearly monophyletic, comprising the subtribes Ligustrinae (Syringa and Ligustrum), Schreberinae status novus (Schrebera and Comoranthus), Fraxininae status novus (Fraxinus), and Oleinae (12 drupaceous genera). An rps16 sequence obtained from Hesperelaea, known only from the type specimen collected in 1875, confirmed the placement of this extinct taxon in the subtribe Oleinae. Key words:

cpDNA; Dimetra; Myxopyrum; Nyctanthes; phylogeny; Oleaceae; rps16; trnL-F.

The Oleaceae is a medium-sized family of ;600 species in 25 genera (Table 1). The family is distributed on all continents except the Antarctic, from northern temperate to southern subtropical regions and from low to high elevations. Some genera are widespread and occur on more than one continent, e.g., Chionanthus, Menodora, and Fraxinus (authors of names are given only if not listed in the Appendix or Table 3, and only the first time they are mentioned). The genus Jasminum is the largest with over 200 species. Many of the genera are economically important, e.g., the olive (Olea europaea) is cultivated for its fruit and oil, species of Fraxinus are grown for timber, and Jasminum, Forsythia, Syringa, and Ligustrum are planted as ornamentals. The family is considered monophyletic on the basis of several morphological synapomorphies and is easily circumscribed. The members of the family are trees, shrubs, or woody climbers with opposite, simple, or compound leaves without stipules. The flowers are hypogynous and four-merous, generally with two stamens, but with four stamens in some species. The corolla is actinomorphic and usually sympetalous. Free petals occur in Chionanthus and Fraxinus, and apetalous 1 Manuscript received 9 September 1999; revision accepted 22 February 2000. ˚ slo¨g Dahl, Roger ErThe authors thank Lennart Andersson, Pieter Baas, A iksson, Peter S. Green, Bengt Oxelman, and Jens G. Rohwer for comments on previous drafts of this paper; Peter S. Green for valuable advice regarding taxon sampling; Mari Ka¨llersjo¨ and James S. Farris for performing the jackknife analysis; herbaria that lent material and gave permission to extract DNA (BKF, BM, C, GB, MO, and NY); Mark W. Chase at the Jodrell Laboratory at Kew, who provided many DNA extractions; and botanical gardens that gave permission to collect specimens (Go¨teborg Botanical Garden, Palermo Botanical Garden, the New York Botanical Garden, the Missouri Botanical Garden, Tokyo Botanical Garden, and the Botanical Garden of Kyoto University). This research was supported by the Lewis B. and Dorothy Cullman Foundation, The Royal Swedish Academy of Sciences, Adlerbertska Forskningsfonden, Helge Ax:son Johnsons Stiftelse, Stiftelsen Wilhelm och Martina Lundgrens Vetenskapsfond, Ra˚dman och fru Ernst Collianders Stiftelse, and Kungliga och Hvitfeldtska Stiftelsen. 4 Author for reprint requests (e-mail: [email protected]).

flowers are known in Nestegis, Forestiera, and wind-pollinated species of Fraxinus. The ovary is syncarpous, consisting of two carpels. Fruit types range from loculicidal capsules, woody schizocarps, and samaras to berries and drupes. The Oleaceae have by recent molecular studies been placed in Lamiales, sister group to the rest of the order (Wagstaff and Olmstead, 1997), and APG (1998) classified it in this order. The family has also been treated in an order of its own, Oleales, by, e.g., Takhtajan (1997). Most classifications of Oleaceae divide the family into two subfamilies, Jasminoideae and Oleoideae (Table 2). Knoblauch (1895) based his division on the point of attachment of the ovules and the presence of a constriction through the apex of the fruit. Taylor (1945) rearranged some genera on the basis of chromosomal data and fruit morphology. The most recent review of the entire family is that of Johnson (1957). His division of Oleaceae into subfamilies and tribes follows Taylor (1945), with a few exceptions. The members of the subfamily Oleoideae apparently form a monophyletic group. They all have x 5 23 and are thought to be of an allopolyploid origin (Taylor, 1945). In addition, they share a number of anatomical, morphological, and chemical apomorphies. In contrast, the Jasminoideae are a heterogeneous assemblage of those genera that do not fit in the Oleoideae. Except for the family-wide characters, the tribes of Jasminoideae share no apomorphies, but they are well distinguished from the Oleoideae. Therefore, most authors have placed them in a separate subfamily. The phylogenetic position of the genus Nyctanthes and its close relative Dimetra has been much debated. Nyctanthes was placed in Oleaceae by Bentham (1876) and Dimetra next to Nyctanthes by Kerr (1938). Later, both genera were suggested to belong to the Verbenaceae (Airy Shaw, 1952; Stant, 1952), or in a family of their own, Nyctanthaceae (Kundu and De, 1968). The exclusion from the Oleaceae was based mainly on the plants’ ‘‘Verbenaceous appearance’’ (Airy Shaw, 1952). Since then, the morphology of these genera has been investigated and compared to Oleaceae and Verbenaceae in a number

1827

1828

AMERICAN JOURNAL

OF

BOTANY

[Vol. 87

TABLE 1. The 25 genera recognized for this study, the number of representatives sequenced, the approximate number of species in the genus, and their world distribution. * denotes two individuals of the only species in that genus.

Genus

Abeliophyllum Chionanthus Comoranthus Dimetra Fontanesia Forestiera Forsythia Fraxinus Haenianthus Hesperelaea Jasminum Ligustrum Menodora Myxopyrum Nestegis Noronhia Notelaea Nyctanthes Olea Osmanthus Phillyrea Picconia Priogymnanthus Schrebera Syringa Sum

No. representatives sequenced

2* 4 2 2* 2 4 2 10 2 1 8 4 2 3 4 1 3 2 4 5 3 2* 2 2 4 80

No. of species

1 ca 100 3 1 1–2 ca 15 11 40–50 3 1 (extinct) 2001 45 24 4 5 41 12 2 401 30 2 1(22) 2 4 20 6001

Distribution

Korea Tropical and subtropical Africa, America, Asia, and Australia Madagascar and the Comores Thailand SW Asia (and Sicily) and China Subtropical North America, West Indies, and N South America E Asia and SE Europe (one sp.) Mainly temperate and subtropical regions of the Northern Hemisphere West Indies Mexico (was endemic to Guadalupe Island) Tropical and subtropical parts of the Old World Temperate to tropical parts of the Old World, except Africa Subtropical North and South America and S Africa Tropical SE Asia New Zealand and Hawaii (one sp.) Madagascar Australia and Tasmania Tropical and subtropical SE Asia Tropical and subtropical parts of the Old World Subtropical parts of E Asia and North America (1-2 spp.) Mediterranean region to W Asia Macaronesia South America (Bolivia, Brazil, Paraguay, Ecuador) Tropical parts of Africa and India Mainly subtropical parts of Eurasia

of studies (cf. Kiew and Baas, 1984), all reaching the conclusion that they should belong to Oleaceae. In an attempt to test this hypothesis with molecular data, the two genera were included in this study. The position of the distinct genus Myxopyrum within Oleaceae has also been uncertain. It was first assigned to the Oleineae by Bentham (1876), kept in the Oleoideae-Oleineae by Knoblauch (1895), then Taylor (1945) thought that it was ‘‘not Oleineae,’’ and Johnson (1957) put it in a tribe of its own in the Jasminoideae. Later, arguments for placement in the subfamily Oleoideae have come from Kiew (1983, 1984), but Baas et al. (1988) and Rohwer (1996) have doubted this. One of the aims of the present study has been to let molecular data shed new light on possible relationships for this genus incertae sedis. Despite containing such well-known and economically important genera, no recent classification of the entire family based on an explicit phylogeny has been published. The first author to present a ‘‘phylogeny’’ for the Oleaceae was Taylor (1945), who drew a phylogenetic chart based on cytological data. Later, Johnson (1957) made an important contribution to the systematics of the family by reviewing its taxonomy and classification. So far, only two studies have employed cladistic methods in evaluating phylogenetic hypotheses for Oleaceae. Baas et al. (1988) studied wood anatomy of the whole family and based cladistic and phenetic analyses on wood anatomical characters. Rohwer (1996) based his cladistic analyses mainly on fruit and seed characters. Our molecular phylogeny is the first to be documented [see also Kim and Jansen (1993) and Kim (1999)] and contributes new insights towards a revised classification of the family.

MATERIALS AND METHODS Material—Table 1 lists the 25 genera we recognize and used in this study (based on previous classifications; mainly Johnson, 1957), the approximate number of species in each genus, the number of species sequenced, and the world distribution of the genera. At least two representatives from each genus in the family were sequenced, including Nyctanthes and Dimetra. Where possible we tried to use material from the type-bearing species of the genus. In the monotypic genera Abeliophyllum, Dimetra, and Picconia two different individuals of each species were sequenced. The genus Hesperelaea is also monotypic, but because it is extinct and known solely from the type collection, only a sample from this could be used. About a third of the material studied was silica-gel dried plant material, and a few fresh samples from Go¨teborg Botanical Garden and the New York Botanical Garden, collected by Eva Wallander. Silica-gel dried material of three Australian and New Zealand taxa were received from Wayne K. Harris (BRI), a sample of Nestegis sandwicensis from Timothy J. Motley (NY), and a recent collection of Dimetra craibiana from S. Suddee (by courtesy of the Bangkok Forestry Department, Thailand). Another third of the DNA was isolated from herbarium specimens held at BM, C, GB, MO, and NY. DNA extracts from plants cultivated at the Royal Botanic Gardens at Kew and herbarium specimens held at K were received from Mark W. Chase. Vouchers for all sequenced taxa are listed in Table 3 along with their GenBank accession numbers. As outgroup taxa we chose species of Verbenaceae and Myoporaceae (Lamiales) to test the position of Dimetra and Nyctanthes, and members of Rubiaceae, Loganiaceae, Strychnaceae, and Gelsemiaceae (Gentianales) were included to provide a root hypothesis for Oleaceae. Except for two Verbenaceae sequences, outgroup sequences were received from various authors, which are also listed in Table 3. DNA extraction—Fresh leaf tissue was manually ground with a pestle in an Eppendorf tube immersed in liquid nitrogen, and dried tissue was homogenized using the FastPrept instrument (BIO 101, Vista, California, USA). Total DNA was extracted using a lysis buffer consisting of 2% CTAB (cetyltrimethylammonium bromide), 1% PEG 6000 (polyethylene glycol, molec-

Oleoideae-Oleineae Oleoideae-Syringeae Oleoideae-Syringeaeb Oleoideae-Fraxineae Oleoideae-Oleineaec Oleoideae-Oleineae Oleoideae-Oleineae Oleoideae-Oleineae Oleoideae-Oleineae Oleoideae-Oleineae Oleoideae-Oleineae Oleoideae-Oleineae Oleoideae-Oleineae

Oleineae Syringeae Syringeae Fraxineae Oleineaea Oleineae

Oleineae Oleineae Oleineae Oleineae Oleineae

Taylor (1945)

This study

Fontanesieae Forsythieae Forsythieae Jasmineae Jasmineae Myxopyreae Myxopyreae Myxopyreae Oleeae-Ligustrinae Oleeae-Ligustrinae Oleeae-Schreberinae Oleeae-Schreberinae Oleeae-Fraxininae Oleeae-Oleinae Oleeae-Oleinae Oleeae-Oleinae Oleeae-Oleinae Oleeae-Oleinae Oleeae-Oleinae Oleeae-Oleinae Oleeae-Oleinae Oleeae-Oleinae Oleeae-Oleinae Oleeae-Oleinae Oleeae-Oleinae

AND CLASSIFICATION OF

c

b

Johnson (1957)

Jasminoideae-Fontanesieae Jasminoideae-Forsythieae Jasminoideae-Forsythieae Jasminoideae-Jasmineae Jasminoideae-Jasmineae Jasminoideae-Myxopyreae Verbenaceae Verbenaceae Oleoideae-Oleeae Oleoideae-Oleeae Jasminoideae-Schrebereae Jasminoideae-Schrebereae Oleoideae-Fraxineae Oleoideae-Oleeaee Oleoideae-Oleeae Oleoideae-Oleeae Oleoideae-Oleeae Oleoideae-Oleeaef Oleoideae-Oleeae Oleoideae-Oleeae Oleoideae-Oleeaeg Oleoideae-Oleeaeh Oleoideae-Oleeae Oleoideae-Oleeae

ALBERT—PHYLOGENY

Oleoideae-Oleineaed Oleoideae-Oleineaed Oleoideae-Oleineae Oleoideae-Oleineae Oleoideae-Oleineae

Oleoideae-Oleineae Oleoideae-Syringeae Oleoideae-Oleineaed Oleoideae-Syringeaed Oleoideae-Fraxineae Oleoideae-Oleineae Oleoideae-Oleineae Oleoideae-Oleineaed Oleoideae-Oleineaed

Jasminoideae-Fontaneseae Jasminoideae-Fontaneseae Jasminoideae-Forsytheae Jasminoideae-Jasmineae Jasminoideae-Jasmineae incertae sedis (‘‘not Oleineae’’)d Jasminoideae-incertae sedisd

AND

Including Linociera. as Nathusia. incl. Mayepea and Tessarandra. d genera not in Taylor’s classification based on cytological studies, but relationships were discussed in the text. e incl. Linociera and Tessarandra. f as Gymnelaea. g incl. Tetrapilus. h incl. Amarolea and Siphonosmanthus.

Oleoideae-Syringeae Jasminoideae-Jasmineae Jasminoideae-Jasmineae Oleoideae-Oleineae Jasminoideae-Jasmineae

Syringeae Jasmineae Jasmineae Oleineae Jasmineae

Knoblauch (1895)

Oleoideae-Fraxineae

Fraxineae

Bentham (1876)

WALLANDER

a

Fontanesia Abeliophyllum Forsythia Jasminum Menodora Myxopyrum Nyctanthes Dimetra Ligustrum Syringa Comoranthus Schrebera Fraxinus Chionanthus Forestiera Haenianthus Hesperelaea Nestegis Noronhia Notelaea Olea Osmanthus Phillyrea Picconia Priogymnanthus

Genus

TABLE 2. Previous classifications of recognized genera of Oleaceae into subfamilies and tribes by different authors. The genera are arranged in order based on our new classification into tribes and subtribes.

December 2000] OLEACEAE 1829

Ingroup Abeliophyllum distichum Nakai* Chionanthus filiformis (Vell.) P. S. Green Chionanthus ramiflorus Roxb. Chionanthus retusus Lindley & Paxton Chionanthus virginicus L.* Comoranthus madagascariensis H. Perrier Comoranthus minor H. Perrier Dimetra craibiana Kerr* Dimetra craibiana Kerr* Fontanesia phyllyreoides Labill. ssp. fortunei (Carr.) Yait Fontanesia phillyreoides Labill.* Forestiera acuminata (Michx.) Poir. Forestiera eggersiana Krug & Urban Forestiera neo-mexicana A. Gray Forestiera segregata (Jacq.) Krug & Urban var. pinetorum (Small) E. Murray Forsythia suspensa (Thunb.) Vahl* Forsythia x intermedia Zabel Fraxinus americana L. Fraxinus anomala Torr. Fraxinus chinensis Roxb. var. rhynchophylla (Hance) E. Murray Fraxinus cuspidata Torr. var. macropetala (Eastw.) Rehd. Fraxinus dipetala Hook. & Arn. Fraxinus excelsior L.* Fraxinus excelsior L. var. diversifolia Fraxinus greggii A. Gray Fraxinus ornus L. Fraxinus quadrangulata Michx. Fraxinus xanthoxyloides (G. Don) DC. var. dimorpha (Coss. & Dur.) Lingelsh. Haenianthus incrassatus Griseb.* Haenianthus salicifolius Griseb. var. obovatus (Krug & Urban) Knobl. Hesperelaea palmeri A. Gray Jasminum fluminense Vell. Jasminum humile L. Jasminum mesnyi Hance Jasminum nitidum Skan Jasminum nudiflorum Lindl. Jasminum odoratissimum L. Jasminum officinale L.* Jasminum sinense Hemsl. Ligustrum ovalifolium Hassk. Ligustrum sempervirens (Franch.) Lingelsh. Ligustrum sinense Lour. Ligustrum vulgare L.* Menodora africana Hook. Menodora integrifolia (Cham. & Schltdl.) Steud. Myxopyrum nervosum Blume* Myxopyrum smilacifolium Blume var. confertum (Kerr) Kiew Myxopyrum smilacifolium Blume var. confertum (Kerr) Kiew Nestegis apetala (Vahl) L. Johnson* Nestegis cunninghamii (Hook. f.) L. Johnson

Taxon

BOTANY

GBAN-AF231837 GBAN-AF231838 GBAN-AF231839 GBAN-AF231840 GBAN-AF231841 GBAN-AF231842 GBAN-AF231843 GBAN-AF231844 GBAN-AF231845 GBAN-AF231846 GBAN-AF231847 GBAN-AF231848 GBAN-AF231849 GBAN-AF231850 GBAN-AF231851 GBAN-AF231852 GBAN-AF231853 GBAN-AF231854 GBAN-AF231855

GBAN-AF225241 GBAN-AF225242 GBAN-AF225243 GBAN-AF225244 GBAN-AF225245 GBAN-AF225246 GBAN-AF225247 GBAN-AF225248 GBAN-AF225249 GBAN-AF225250 GBAN-AF225251 GBAN-AF225252 GBAN-AF225253 GBAN-AF225254 GBAN-AF225255 GBAN-AF225256 GBAN-AF225257 GBAN-AF225258 GBAN-AF225259 GBAN-AF225260 GBAN-AF225261 GBAN-AF225262 GBAN-AF225263 GBAN-AF225264

GBAN-AF225238 GBAN-AF225239 GBAN-AF225240

OF

GBAN-AF231833 GBAN-AF231834 GBAN-AF231835 GBAN-AF231836

GBAN-AF231831 GBAN-AF231832

GBAN-AF231808 GBAN-AF231809 GBAN-AF231810 GBAN-AF231811 GBAN-AF231812 GBAN-AF231813 GBAN-AF231814 GBAN-AF231815 GBAN-AF231816 GBAN-AF231817 GBAN-AF231818 GBAN-AF231819 GBAN-AF231820 GBAN-AF231821 GBAN-AF231822 GBAN-AF231823 GBAN-AF231824 GBAN-AF231825 GBAN-AF231826 GBAN-AF231827 GBAN-AF231828 GBAN-AF231829 GBAN-AF231830

rps16

GBAN-AF225216 GBAN-AF225217 GBAN-AF225218 GBAN-AF225219 GBAN-AF225220 GBAN-AF225221 GBAN-AF225222 GBAN-AF225223 GBAN-AF225224 GBAN-AF225225 GBAN-AF225226 GBAN-AF225227 GBAN-AF225228 GBAN-AF225229 GBAN-AF225230 GBAN-AF225231 GBAN-AF225232 GBAN-AF225233 GBAN-AF225234 GBAN-AF225235 GBAN-AF225236 GBAN-AF225237

GenBank accession numbera trnL-F

AMERICAN JOURNAL

R. F. Thorne & G. R. Proctor 48278 (NY) B. Sta˚hl & J. Knudsen 2302 (GB) Palmer 81 (K) L. Struwe 1098 (NY) ˚ . Dahlstrand 2073 (GB) K. A ˚ . Dahlstrand 37 (GB) K. A E. Wallander 195 (GB) E. Wallander 193 (GB) E. Wallander 130 (GB) E. Wallander 194 (GB) Sino-American Guizhou Bot. Exp. no. 228 (NY) E. Wallander 197 (GB) cult. Kew, voucher # 000.73.10104 (K) E. Wallander 198 (GB) E. Wallander 168 (GB) ˚ . Dahlstrand 1081 (GB) K. A S. G. Tressens et al. 546 (GB) M. J. E. Coode 6845 (K) C. F. van Beusekom & T. Santisuk 2859 (C) C. Wang 34636 (NY) M. W. Chase 3940 (K) M. W. Chase 3884 (K)

M. W. Chase 3881 (K) P. I. Oliveira 659 (GB) Collected in Australia, without voucher E. Wallander 82 (GB) E. Wallander 81 (GB) Capuron 20913 (K) L. J. Dorr 4135 (K) A. F. G. Kerr 20476 (BM) S. Suddee et al. 1000 (BKF, K, TCD) M. W. Chase 3878 (K) E. Wallander 20 (GB) E. Wallander 100 (GB) W. G. D’Arcy 5135A (C) A. Carter 1045 (GB) E. Wallander 191 (GB) J. Jutila 556 (NY) ˚ . Dahl 702 (GB) A E. Wallander 101 (GB) R. C. Rollins 1899 (GB) E. Wallander 116 (GB) F. W. Reichenbacher 1716 (MO) E. Wallander 180 (GB) E. Wallander 159 (GB) E. Wallander 1 (GB) Rafael Diaz 406 (MO) E. Wallander 31 (GB) E. Wallander 98 (GB) E. Wallander 141 (GB)

Source of DNA/voucher

TABLE 3. Vouchers, or references, and GenBank accession numbers for taxa sequenced. * indicate type species of a genus. Herbaria acronyms follow Index Herbariorum (Holmgren, Holmgren, and Barrett, 1990).

1830 [Vol. 87

Taxon

a

GBAN-AF102405 GBAN-AF102453

GBAN-AF152684

The prefix GBAN- has been added to link the online version of American Journal of Botany to GenBank but is not part of the actual accession number.

Rova et al. (unpubl.); Andersson and Rova (1999) Rova et al. (unpubl.) Andersson and Rova (1999) Rova et al. (unpubl.) Andersson and Rova (1999) Rova et al. (unpubl.); Andersson and Rova (1999) Rova et al. (unpubl.)

GBAN-AF004035 GBAN-AF004038 GBAN-AF242974

GBAN-AF004094

GBAN-AF004091

GBAN-AF004092

GBAN-AF225294 GBAN-AJ299259 GBAN-AF225295 GBAN-AJ299258

AND CLASSIFICATION OF

GBAN-AJ299257 GBAN-AF102428 1 GBAN-AF159696 GBAN-AF102379 GBAN-AF102484

GBAN-AF231884 GBAN-AJ299260 GBAN-AF231885

rps16

ALBERT—PHYLOGENY

GBAN-AF231877 GBAN-AF231878 GBAN-AF231879 GBAN-AF231880 GBAN-AF231881 GBAN-AF231882 GBAN-AF231883

GBAN-AF231856 GBAN-AF231857 GBAN-AF231858 GBAN-AF231859 GBAN-AF231860 GBAN-AF231861 GBAN-AF231862 GBAN-AF231863 GBAN-AF231864 GBAN-AF231865 GBAN-AF231866 GBAN-AF231867 GBAN-AF231868 GBAN-AF231869 GBAN-AF231870 GBAN-AF231871 GBAN-AF231872 GBAN-AF231873 GBAN-AF231874 GBAN-AF231875 GBAN-AF231876

GBAN-AF225265 GBAN-AF225266 GBAN-AF225267 GBAN-AF225268 GBAN-AF225269 GBAN-AF225270 GBAN-AF225271 GBAN-AF225272 GBAN-AF225273 GBAN-AF225274 GBAN-AF225275 GBAN-AF225276 GBAN-AF225277 GBAN-AF225278 GBAN-AF225279 GBAN-AF225280 GBAN-AF225281 GBAN-AF225282 GBAN-AF225283 GBAN-AF225284 GBAN-AF225285 GBAN-AF225286 GBAN-AF225287 GBAN-AF225288 GBAN-AF225289 GBAN-AF225290 GBAN-AF225291 GBAN-AF225292 GBAN-AF225293

GenBank accession numbera trnL-F

AND

C. Vazquez Yanes 526 (GB) Olmstead and Reeves (1995) H. Kalheber 78–506 (GB) B. Oxelman (unpublished) Olmstead and Reeves (1995) Rova et al. (unpubl.); Andersson and Rova (1999)

M. Pole s.n. (BRI) T. Motley 2014 (NY) J. Pruski, B. Stein & S. Zona 2885 (NY) K. Egero¨d 9264 (GB) W. K. Harris 91 (BRI) W. K. Harris 89 (BRI) A. F. G. Kerr 13501 (K) ˚ . Dahlstrand s.n., 2 Oct. 1952 (GB) K. A H. Y. Liang 66470 (GB) E. Wallander 26 (GB) ˚ . Dahl 703 (GB) A M. W. Chase 3882 (K) S. A. Spongberg et al. 17159 (NY) E. Wallander 28 (GB) E. Wallander 223 (GB) E. Wallander 231 (GB) E. Wallander 230 (GB) J. Lewalle 10113 (C) M. W. Chase 3880 (K) H. C. Stutz 3140 (NY) E. Wallander 132 (GB) X. Cornejo & C. Bonifaz 5303-B (GB) E. Hassler 11889 (C) M. W. Chase 3883 (K) D. C. Plowes 1303 (NY) Lancaster 1623 (K) E. Wallander 228 (GB) E. Wallander 111 (GB) E. Wallander 170 (GB)

Source of DNA/voucher

WALLANDER

Antonia ovata Pohl. (Loganiaceae) Strychnos tomentosa Benth. (Strychnaceae) Strychnos nux-vomica L. (Strychnaceae) Cinchona pitayensis Wedd. (Rubiaceae) Cinchona pubescens Vahl (Rubiaceae) Coffea arabica L. (Rubiaceae) Luculia gratissima Sweet x tsetensis (Rubiaceae)

Nestegis lanceolata (Hook. f.) L. Johnson Nestegis sandwicensis (Gray) O. & I. Deg. & L. Johnson Noronhia emarginata (Lam.) Thouars* Notelaea longifolia Vent.* Notelaea microcarpa R. Brown Notelaea punctata R. Brown Nyctanthes aculeata Craib Nyctanthes arbor-tristis L.* Olea brachiata (Lour.) Merrill Olea capensis L. Olea europaea L.* Olea paniculata R. Brown Osmanthus americanus (L.) A. Gray Osmanthus fragrans Lour.* Osmanthus heterophyllus (G. Don) P. S. Green Osmanthus insularis Koidz. Osmanthus rigidus Nakai Phillyrea angustifolia L. Phillyrea latifolia L.* Phillyrea media L. Picconia excelsa (Aiton) DC.* Priogymnanthus apertus (B. Sta˚hl) P. S. Green Priogymnanthus hasslerianus (Chodat) P. S. Green* Schrebera alata (Hochst.) Welw. Schrebera mazoensis S. Moore Syringa pekinensis Rupr. Syringa reticulata (Blume) H. Hara Syringa vulgaris L.* Syringa yunnanensis Franchet Outgroup Lantana camara L. (Verbenaceae) Stachytarpheta dichotoma (Ruiz & Pav.) Vahl (Verbenaceae) Verbena officinalis L. (Verbenaceae) Myoporum insulare R. Br. (Myoporaceae) Myoporum mauritianum DC. Gelsemium sempervirens (L.) Aiton (Gelsemiaceae)

TABLE 3. Continued.

December 2000] OLEACEAE 1831

1832 TABLE 4. Name

tRNc tRNd tRNe tRNf rpsF rpsMRP rpsMF2 rpsR2

AMERICAN JOURNAL

OF

BOTANY

[Vol. 87

Primer sequences used for PCR and sequencing. Primer sequence (59–39)

CGAAATCGGTAGACGCTACG GGGATAGAGGGACTTGAACC GGTTCAAGTCCCTCTATCCC ATTTGAACTGGTGACACGAG GTGGTAGAAAGCAACGTGCGACTT GGATCCCAAAACAAGGAAACACC GGGTATGTTGCTGCCATTTTGAAA TGCGGATCGAACATCAATTGCAAC

Position

trnL fwd trnL rev trnL-F igs fwd trnL-F igs rev rps 16 end fwd rps 16 internal rev rps 16 internal fwd rps 16 end rev

ular weight 6000), 1.4 mol/L NaCl, 10 mmol/L Tris-HCl, and 20 mmol/L EDTA. Lysis was performed at 748C with 2% mercaptoethanol added, followed by cleaning with the Genecleant II kit (BIO 101). Cleaned DNA was transferred to 10 mmol/L Tris and kept in freezer. cpDNA regions and primers—For our study, we chose two noncoding chloroplast regions, the trnL-F region and the intron of rps16. The trnL-F region consists of the trnL intron and the trnL-trnF intergenic spacer (Taberlet et al., 1991). The primer pair tRNc/tRNf (Table 4) was used to amplify the entire region of ;900 bp in one PCR (polymerase chain reaction). In some cases, the tRNc/tRNd and the tRNe/tRNf primer pairs were used to amplify the intron and the spacer, respectively. The intron of rps16 is a group II intron that was first used for phylogenetic studies by Oxelman, Lide´n, and Berglund (1997). The primer pair rpsF/rpsR2 was used to amplify the entire 800–900 bp region. For DNA of low quality, internal primers were used with each of the end primers to split that region into two approximately equal halves. The position of the internal forward primer (rpsMF2) is located ;50 base pairs downstream of the internal reverse primer (rpsMRP), giving sufficient overlap for determining a full sequence. Amplification—Most PCRs were performed in a 25-mL reaction volume using the Taq kit from Boehringer Mannheim (now Roche Molecular Biochemicals, Indianapolis, Indiana, USA). These amplifying reactions were run on a Perkin Elmer GeneAmpt PCR System 9600 version 2.01, using the same program for both chloroplast regions (30 cycles of [958C 50 sec, 608C 50 sec, 728C 1 min 50 sec]). A second round of PCR was sometimes performed using the first PCR product as template. In this case, the first PCR products were run out on a low-melting-point agarose gel, stained with ethidium bromide to visualize the bands, then cut out and dissolved in water. No further cleaning was done before the second PCR. The PCR products (from first or second PCR) were purified before sequencing using the Genecleant II kit (BIO 101). Some PCR reactions were performed in 50-mL volumes using polymerase and buffer of the Thermoprime 1 kit (Advanced Biotechnologies Ltd., Surrey, UK), or in 25-mL reactions using Ready-To-Goy PCR beads (Amersham Pharmacia Biotech AB, Uppsala, Sweden), following the manufacturer’s instructions. In these cases, both chloroplast regions were amplified on a Perkin Elmer Cetus 480 version 1.1, using the same cycling program (27 cycles of [948C 1 min, 608C 1 min, 728C 1 min]). The PCR products were purified before sequencing using the QIAquicky PCR Purification Kit (QIAGENt GmbH, Hilden, Germany). Sequencing—Sequencing reactions, using the same primer sequences as in the PCR, were performed on a Perkin Elmer GeneAmpt PCR System 9600 version 2.01 (1 min at 958C, followed by 32 cycles of [958C 10 sec, 508C 5 sec, 608C 3 min]), using the dRhodamine Terminator Cycle Sequencing Ready Reaction DNA sequencing kit with AmpliTaqt DNA polymerase (Perkin Elmer Applied Biosystems, Foster City, California, USA) and HT1000 halfTERM Dye Terminator Reagent (GENPAK Inc., Stony Brook, New York, USA). Before gel separation, the sequence reaction products were cleaned using Sephadext G-50 Fine DNA Grade (Amersham Pharmacia Biotech AB) in Centrisep Spin Columns (Princeton Separations, Philadelphia, Pennsylvania, USA). Separation of the fragments was done on a 5% Long Rangery gel (FMC BioProducts, Rockland, Maine, USA) on an ABI Prismy 377 DNA

Reference

Taberlet et al. (1991) Taberlet et al. (1991) Taberlet et al. (1991) Taberlet et al. (1991) Oxelman, Lide´n, and Berglund (1997) Persson (2000) this study Oxelman, Lide´n, and Berglund (1997)

Sequencer (Perkin Elmer Applied Biosystems). The ABI Prismy 377 Collection software version 2.1 was used to evaluate the sequences. Some sequencing reactions were also performed on an ALFexpressy DNA Sequencer (Amersham Pharmacia Biotech AB). Reactions were then performed using the ThermoSequenase fluorescent labeled primer cycle sequencing kit with 7deaza-dGTP (Amersham Pharmacia Biotech AB) and Cy5-labeled primers. Cycle sequencing reactions were performed on a Perkin Elmer Cetus 480 version 1.1 (2 min at 968C, followed by 18 cycles of [958C 30 sec, 608C 40 sec]). The reaction products were loaded without further cleaning on a 0.5 mm 5.28% Page-Plus gel (Amrescot, Solon, Ohio, USA). Sequences were evaluated with the ALFwiny software version 1.10. Alignment and indel coding—The forward and reverse sequences were checked and edited using the Sequenchery software version 3.1 (Gene Codes Corporation, Ann Arbor, Michigan, USA). Consensus sequences from each of the two chloroplast loci were aligned separately. The two sequences of each of Abeliophyllum and Picconia were found to be identical so only one of them was included. Alignment was done using the assembly feature in Sequencher, and then manually adjusted using criteria described in Andersson and Rova (1999). Adding new sequences to the alignment was relatively easy because of conserved regions and shared indels. Twenty-two indels in the rps16 matrix and 20 in the trnL-F were considered informative, and indel characters were added to the combined matrix (using A/T for present/absent). A few insertions, which did not contain informative characters, were then deleted. Autapomorphic insertions were also removed. The alignment is available from the corresponding author upon request. Cladistic analyses—The combined matrix consisted of 78 ingroup taxa and ten outgroup taxa. Two taxa in this matrix, Hesperelaea palmeri and Priogymnanthus apertus, were represented by the rps16 sequence only, and in the case of Hesperelaea, only the first 423 bp of the sequence were included (because of sequencing problems with the second part). The two data sets were subjected to parsimony analyses separately, and in combination, and in the latter case with and without indel characters, using PAUP* version 4.0b4a (Swofford, 2000) on a Power Macintosh. All characters were analyzed using equal weights (51), and gaps were treated as missing data. Initial rounds of PAUP analyses yielded tree overflow with maximum memory settings so the following search strategy was adopted: first a search for multiple tree islands was conducted by doing 100 random addition sequence replicates, limited to only ten saved trees from each. The resulting most parsimonious trees were then used as starting trees for TBR (tree bisectionreconnection) branch swapping in an additional heuristic search for shorter trees. Up to 5000 additional trees of equal or shorter length were allowed to be saved and were then compared to the starting trees as consensus trees. Another strategy was also adopted, namely excluding some taxa thought to cause most of the problems with thousands of equally parsimonious trees. By starting out with only two of the closest outgroup taxa (Verbenaceae) and 16 ingroup ‘‘backbone taxa,’’ and then restoring a few taxa in successive runs, we were able to determine which were causing the problem. It was also evident from the first runs, and from inspecting the alignment, that the taxa of Johnson’s tribe Oleeae are very closely related and not many characters are available to support any particular interrelationship. By only representing each genus in this tribe with one sequence, but excluding taxa with an in-

December 2000]

WALLANDER

AND

ALBERT—PHYLOGENY

complete sequence (e.g., Hesperelaea palmeri), the number of trees obtained drastically decreased. In an analysis that made the best compromise between computational time and fewest excluded taxa, 64 of the 78 ingroup taxa were included, and 4312 trees were found with complete TBR branch-swapping. Parsimony jackknifing (Farris et al., 1996) was performed on the combined matrix, with and without indel characters added, using XAC (J. S. Farris, Swedish Museum of Natural History, Stockholm, Sweden). One thousand jackknife replicates, each with ten random addition sequences and nonrotational branch-swapping (J. S. Farris, Swedish Museum of Natural History, Stockholm, Sweden, personal communication), were conducted.

RESULTS The rps16 data set contained 1212 characters, of which 265 were informative, and the trnL-F 1211 characters, of which 240 were informative. The combined matrix with the indel characters included (and autapomorphic and other uninformative indels removed) contained 1890 characters, of which 524 were informative. The resulting consensus trees of the most parsimonious trees from the separate analyses were compatible, although not equally well resolved (not shown). The limited analysis of the combined matrix with indel characters resulted in 810 most parsimonious trees of length 1509. In the additional analysis, no shorter trees were found. Strict consensus trees computed for the first 810 trees and for the 5000 extra trees were identical, shown in Fig. 1 with jackknife support values exceeding 50%. In Fig. 2, one of the most parsimonious trees (randomly chosen) from this analysis is shown as a phylogram. The strict consensus trees from the analyses of the combined data sets, compared to those from the separate analyses, are resolved to a higher degree, but there were no differences in topology between the strict consensus trees from the analyses with and without indel characters. The only difference was in the amount of jackknife support, i.e., clades that shared informative indels received slightly higher support values. The trees from the alternative search strategy were, although containing fewer taxa, congruent with the trees from the full analyses. The RI of the trees from all analyses varied between 82 and 84%. DISCUSSION The molecular result—Although the Oleaceae traditionally are divided into two subfamilies, this may not be a phylogenetically natural representation. In the consensus tree (Fig. 1), the Jasminoideae are paraphyletic because the tribe Jasmineae is sister to the Oleoideae. The jackknife analysis gives 76% support for this resolution, and all tribal clades are given strong support (95–100%). The phylogram (Fig. 2) shows that within Fraxinus and its sister group, branch lengths are very short, explaining the relatively low support in this group. Like many other noncoding chloroplast regions, rps16 and trnL-F have too little variation to resolve phylogenies at an infrageneric level, at least for relatively recently diverged groups (cf. Small et al., 1998). Previously, Gielly and Taberlet (1994) have shown that the variation in the trnL-F region is too low to resolve relationships within Fraxinus. For the Oleaceae family, the rps16 intron is more informative than the trnL-F region, but still, the combined data set does not contain enough informative characters to resolve the inter-tribal relationships outside Jasmineae and Oleeae (i.e., branch lengths between the basal tribes are almost zero).

AND CLASSIFICATION OF

OLEACEAE

1833

Congruence between molecular data and other characters—The result of the molecular analyses agrees well with nonmolecular data, e.g., chromosomal data (Taylor, 1945), wood anatomy (Baas et al., 1988), ovule number and position (Taylor, 1945; Rohwer, 1996), fruit anatomy (Rohwer, 1996), flavonoid glycosides (Harborne and Green, 1980), and iridoids (Jensen, 1992). The monophyly of the Oleoideae is not only supported by the present molecular analysis, but also by numerous morphological, anatomical, chemical, and chromosomal synapomorphies. The subfamily Jasminoideae, on the other hand, is not supported by this study, nor by any nonmolecular synapomorphies. Some of the many characters supporting the results from this study are shown on a summary tree (Fig. 3) and are discussed under the tribal and subtribal sections further below. Fruit type, which varies considerably in the Oleaceae (Rohwer, 1996), is also discussed. A new classification—Based on the molecular phylogeny, supported by nonmolecular data, some changes in the classification are necessary to accord with these results. The subfamily Jasminoideae is clearly paraphyletic, and it is now time to abandon the subfamilial classification. It has been convenient to put all genera excluded from the Oleoideae in another subfamily, even though they do not share any apomorphies. Already Johnson (1957) saw the need to ‘‘ultimately abandon the subfamilies and to treat the allotetraploid Oleoideae as equivalent to the other tribes.’’ Rohwer (1996) stated that ‘‘the Jasminoideae is so heterogeneous in its present circumscription that it seems advisable to dismember it as a taxonomic unit,’’ and Qin (1996), who based his conclusions on leaf peroxidases and morphology of a few genera of the family, revoked ‘‘the subfamily rank because the tribes in subfamily Jasminoideae have no points in common.’’ Kiew and Baas (1984) proposed to abandon the use of subfamilies and revert to the old tribes sensu Bentham (1876). In this case, the Oleoideae would fall apart to Fraxineae, Syringeae, and Oleeae, and the jasminoids would stay in their assigned tribes (sensu Johnson). However, our present findings suggest that only dropping the subfamily rank and keeping the tribes unaltered is unsatisfactory. Because chromosomal data and a number of morphological characters support the monophyly of Oleoideae, we think it is important to recognize this. We suggest that changing rank of the subfamily Oleoideae to tribe Oleeae, and changing all previous tribes of Oleoideae to subtribes, is a better solution. In this way the monophyly of this group is shown, equal in status to the jasminoid tribes. Therefore, we present a revised classification of Oleaceae (Table 2; Appendix), shown on a summary tree in Fig. 3. We recognize five tribes: Myxopyreae (Myxopyrum, Nyctanthes, and Dimetra), Fontanesieae (Fontanesia), Forsythieae (Forsythia and Abeliophyllum), Jasmineae (Jasminum and Menodora) and Oleeae. The Oleeae now contain four subtribes: Ligustrinae (Syringa and Ligustrum), Schreberinae (Schrebera and Comoranthus), Fraxininae (Fraxinus), and Oleinae (the remaining 12 genera). The subtribes Schreberinae and Fraxininae are new (Appendix). From this point forward, when we discuss and compare our results with those of other studies, there are cases where it is simpler to refer to the old taxonomy, i.e., subfamilial groupings. In order not to cause confusion when using the tribal name Oleeae, we will state whether it is our new tribe Oleeae (former subfamily Oleoideae) or subtribe Oleinae (former tribe Oleeae). The term ‘‘jasminoids’’ is used to refer to the tribes

1834

AMERICAN JOURNAL

OF

BOTANY

[Vol. 87

Fig. 1. Strict consensus tree of the most parsimonious trees from the analyses of the combined data set with indels coded as separate characters. Jackknife support values over 50% are shown above the branches. Tribal delimitations follow this study.

Myxopyreae, Fontanesieae, Forsythieae, and Jasmineae, i.e., the former subfamily Jasminoideae. As can be seen by comparing our classification with that of Johnson (1957), apart from alterations in ranks, the changes are: (1) transfer of the tribe Schrebereae (as subtribe Schreberinae) back to our tribe Oleeae, (2) reinstatement of the subtribe Ligustrinae (with Syringa and Ligustrum) in tribe Oleeae, and (3) inclusion of the formerly incertae sedis Nyctanthes and Dimetra with Myxopyrum in Myxopyreae. The new tribe Oleeae—The Oleeae are clearly a monophyletic group, supported by numerous data (Fig. 3). The haploid

chromosome number n 5 23 is basic in all genera of the new tribe Oleeae. In contrast, the basic number of the jasminoids is either x 5 11, (12), 13, or 14. Chromosome numbers served as one of the fundamental characters upon which Taylor (1945) based his division of Oleaceae into subfamilies and tribes. It was suggested by Taylor that the x 5 23 group has an allopolyploid origin (from two unknown and now probably extinct jasminoids with x 5 11 and 12). The ovaries of all Oleaceae are bilocular and the number of ovules in each locule vary from one to many. All genera of Oleaceae have pendulous ovules, except Myxopyreae, which has ascending ovules (see below). The synapomorphy for the

December 2000]

WALLANDER

AND

ALBERT—PHYLOGENY

AND CLASSIFICATION OF

OLEACEAE

1835

Fig. 2. A randomly selected phylogram from the analyses of the combined data set with indel characters. Because branches are quite long in the outgroup, only the closest outgroup (Verbenaceae) is shown. Numbers above branches indicate number of changes. The scale bar represents five changes.

1836

AMERICAN JOURNAL

OF

BOTANY

[Vol. 87

Fig. 3. Summary of the molecular phylogeny of Oleaceae, with the new classification shown on top. Some nonmolecular characters that support this phylogeny are plotted onto the tree and discussed in the text. Synapomorphies for clades are shown with a bar. Some plesiomorphic characters and others with uncertain polarity are shown within parentheses for comparison. A number after each character corresponds to data from the following authors: (1) karyology (Taylor, 1945), (2) wood anatomy (Baas et al., 1998), (3) fruit anatomy (Rohwer, 1996), (4) flavonoid glycosides (Harborne and Green, 1980), (5) verbascoside and iridoid glucosides (Jensen, 1992), and (6) iridoid glucosides (H. Franzyk, S. R. Jensen, and C. E. Olsen, Technical University of Denmark, unpublished data).

new tribe Oleeae is two pendulous ovules per locule (except Schrebera, which has four; Taylor, 1945). In contrast, the tribes Fontanesieae and Forsythieae have varying numbers of ovules per locule, but never two. Jasminum and Menodora in the Jasmineae have 1–2 and 2–4 ovules per locule, respectively, but their position is more horizontal. Harborne and Green (1980) carried out an investigation of flavonoid glycosides in leaves of all genera of Oleaceae. The pattern they found was clear: all the jasminoid genera have only the plesiomorphic flavonols present, but more complex flavonoids, including flavones and flavanones, were found to be a synapomorphy for taxa with x 5 23, i.e., the new tribe Oleeae. Unfortunately, they did not recognize Nyctanthes and Dimetra in Oleaceae, which were excluded from the study despite containing two common flavonols. Harborne and Green, also investigating flavonoid patterns in other closely

related families came to the conclusion that keeping Oleaceae in an order of its own (Oleales) was justified based on the fact that the flavonoid pattern in this family differed from other sympetalous families. Baas et al. (1988) studied wood characters for the whole family (including Nyctanthes), and made both phenetic and cladistic analyses of the data. Trees from both analyses agree in principal with our results. The distribution of fiber and vessel characters especially agrees with the molecular phylogeny presented here, and libriform fibers and multiple vessels form synapomorphies for the new tribe Oleeae (exceptions in Ligustrinae, see below). The jasminoids—The tribes of the former subfamily Jasminoideae, viz. Fontanesieae, Forsythieae, Myxopyreae, and Jasmineae, share no apparent morphological apomorphies with

December 2000]

WALLANDER

AND

ALBERT—PHYLOGENY

each other, nor with the new tribe Oleeae (except Jasmineae, see under this subheading below). Taylor (1945), in his cytological study of the family, found varying low basic chromosome numbers (x 5 11, 12, 13, or 14) in the jasminoids. George, Geethamma, and Ninan (1989) have proposed that x 5 11 (found in Menodora and Myxopyrum) is the basic chromosome number of the family, and that all other low numbers have originated by aneuploidy. In the study of flavonoid glycosides by Harborne and Green (1980), the jasminoids were shown to contain only the plesiomorphic flavonols (except Myxopyrum that also contained advanced flavones, but not of the same type as in Oleoideae). In the wood anatomical study by Baas et al. (1988), fiber-tracheids and solitary vessels were shown to be plesiomorphic characters in common for the jasminoids. Tribe Jasmineae—In the strict consensus tree (Fig. 1), Jasmineae are resolved as sister group to Oleeae, supported by eight steps and a jackknife value of 76%. Nonmolecular support has also come from Jensen (1992), who investigated iridoids in a number of species of Oleaceae, and the results fit very well with the molecular phylogeny. The tribes Jasmineae and Oleeae both contain oleoside, whereas in Fontanesieae (Damtoft, Franzyk, and Jensen, 1995), Forsythieae (Damtoft, Franzyk, and Jensen, 1994), and Myxopyreae (S. R. Jensen, Technical University of Denmark, unpublished data) this compound is absent. Our results also indicate that Jasminum is paraphyletic, as Kim and Jansen (1993) and Rohwer (1996) have suggested, because Menodora is nested within it. There is, however, no doubt that the tribe is monophyletic. The phylogram (Fig. 2) shows that the clade is supported by 61 steps, and based on fruit anatomy the Jasmineae are unique in the family in having bilobed fruits. Jasminum has a bilobed berry (each lobe one-to-two-seeded, one lobe frequently aborted) and Menodora, the New World counterpart of Jasminum, has a bilobed circumscissile capsule. The development of these two seemingly different fruit types is in fact very similar except for the final stages (Rohwer, 1995, 1997). The position of Nyctanthes and Dimetra—The molecular results presented here clearly show that both Nyctanthes and Dimetra belong to Oleaceae. Their inclusion in the family is supported by a jackknife value of 100%. Nyctanthes arbortristis L. was placed in Oleaceae by Bentham (1876), Knoblauch (1895), and Taylor (1945) (Table 2). Takhtajan (1997) placed Nyctanthes in its own subfamily in Oleaceae (Nyctanthoideae). The second species of Nyctanthes, N. aculeata Craib, was described in 1916 and placed by the author in Oleaceae-Jasmineae. When Kerr (1938) described the new monotypic genus Dimetra, he assigned it to Oleaceae without hesitation. He stated that its closest alliance clearly was with Nyctanthes. Later, Airy Shaw (1952) transferred both of them to Verbenaceae (in subfam. Nyctanthoideae) because ‘‘the Verbenaceous facies of Nyctanthes almost hits one in the eye.’’ Stant (1952) supported this view with a study of some anatomical characters, and Johnson (1957) agreed. This transfer generated a number of papers investigating various morphological aspects of Nyctanthes and Dimetra. Kundu and De (1968) investigated cytology, palynology, and leaf, wood, and floral anatomy of Nyctanthes and compared it with members of Oleaceae, Verbenaceae, and Loganiaceae. They came to the conclusion that it should be placed in a family of its own, Nyctanthaceae, because of differences with both Oleaceae and

AND CLASSIFICATION OF

OLEACEAE

1837

Verbenaceae. They described Nyctanthaceae as a new family, not knowing that it had already been described by Agardh in 1858 (as Nyctantheae). Support for placing Nyctanthes in Oleaceae has come from studies of embryology (Kapil and Vani, 1966), structure and vascular anatomy of the gynoecium (Kshetrapal and Tiagi, 1970), vessel anatomy (Murthy et al., 1978), leaf morphology (Mohan and Inamdar, 1983), wood anatomy (Baas et al., 1988), ultrastructure and morphology of intranuclear proteinic inclusions in the mesophyll parenchymatic cells (Bigazzi, 1989), and fruit anatomy (Kuriachen and Dave, 1989; Rohwer, 1994, 1996). These and other studies are reviewed in detail by Kiew and Baas (1984), who summarized the overwhelming evidence that Nyctanthes belongs to Oleaceae. Because Nyctanthes shares a number of characters with Jasminum and Menodora (Kiew and Baas, 1984), and because they did not want to erect a monogeneric tribe, they proposed that Nyctanthes should be kept in Jasmineae sensu Bentham. Although no one has disputed a close relationship between Dimetra and Nyctanthes, Dimetra was not included in most of the studies and was not mentioned in the review by Kiew and Baas (1984). Since our results clearly point to the close relationship between Nyctanthes and Dimetra, grouped with Myxopyrum rather than with Jasmineae, we argue for placing them in the tribe Myxopyreae. The pertinent node is supported by a jackknife value of 100%. The position of Myxopyrum—The genus Myxopyrum consists of four species distributed in subtropical and tropical east Asia (Kiew, 1984). They are scandent shrubs with quadrangular stems and conspicuously triplinerved leaves. They share the common basic characters with other Oleaceae, but some divergent features have made the genus difficult to place, and there has therefore been different opinions on where it belongs (Table 2). Bentham (1876) and Knoblauch (1895) put it in the Oleineae (sensu Bentham), but according to Taylor (1945) it differed in so many characters that it should probably be separated from the Oleineae. Johnson (1957) erected a new tribe for it, Myxopyreae, and placed it in the heterogeneous Jasminoideae. The results from this study strongly support the placement of Myxopyrum as sister to Nyctanthes and Dimetra, as discussed above, even though there are no apparent outer morphological similarities between them. However, the three genera share the apomorphic character of ascending ovules (Fig. 3), and Nyctanthes and Myxopyrum both have quadrangular stems with cortical bundles in the corners (Kiew, 1984; Kiew and Baas, 1984). Rohwer’s (1996) investigation of fruit and seed characters of the Oleaceae showed that Myxopyrum and Nyctanthes, apart from ascending ovules, also share a deep stylar canal and the presence of a distinctive tissue in the center of the ovary septum. In contrast to the ovary, the fruit of Myxopyrum (a one-to-four-seeded berry) is not similar to that of Nyctanthes (a dry schizocarp that splits into two one-seeded mericarps), and Myxopyrum has varying one to three ovules per locule, whereas Dimetra and Nyctanthes have only one. Apart from the above synapomorphies, it is difficult to find morphological characters that unite these quite distinct genera. Most characters are either plesiomorphic and found in other jasminoid genera as well, or autapomorphic. For example, the wood anatomical study by Baas et al. (1988) showed that Myxopyrum only shared plesiomorphies with the other jasminoids, and the phytochemical study by Harborne and Green (1980) showed that Myxopyrum contains three apigenin glycosides that are not found in any of the other genera of Ole-

1838

AMERICAN JOURNAL

aceae. This finding of advanced glycosides in Myxopyrum led Kiew (1984), together with her own investigation of the morphology, to conclude that Myxopyrum should be retained within the Oleoideae. The chromosome number of Myxopyrum was unknown at that time, but now there are two reports: 2n 5 22 for M. hainanense Chia (synonym to M. pierrei Gagnep.) (Weng and Zhang, 1992) and 2n 5 24 for M. smilacifolium Blume (George and Geethamma, 1983). At least the former fit well with 2n 5 44 reported for Nyctanthes arbor-tristis (George and Geethamma, 1984), which would suggest that the ancestor of Nyctanthes (and Dimetra) arose by polyploidy from the ancestor in common with Myxopyrum. The chromosome number of Dimetra is not known. Chromosome counts in Nyctanthes are notoriously variable (Rohwer, 1996), however, so one should not draw any conclusions based on chromosome number alone. New chemical evidence (S. R. Jensen, Technical University of Denmark, unpublished data) on two new carbocyclic iridoid glucosides in Myxopyrum smilacifolium shows that these are very similar to the compounds found in Nyctanthes and structurally represent the same biosynthetic pathway (myxopyroside). Also, the three genera do not contain oleoside, a compound that only occurs in the two tribes Jasmineae and Oleeae (Jensen, 1992). These findings, together with chromosome numbers, further strengthens the conclusion that Myxopyrum does not belong in the former Oleoideae. To conclude, a number of nonmolecular synapomorphies do support the Myxopyreae clade, despite no obvious outer morphological similarities. Tribes Fontanesieae and Forsythieae—The molecular result shows a closer relationship between Forsythia and Abeliophyllum than between Fontanesia and Abeliophyllum, as might have been expected on the basis of fruit morphology (Taylor, 1945; Rohwer, 1996). Fontanesia and Abeliophyllum both have the same type of samara (differing from the one in Fraxinus, see below), but Forsythia has loculicidal capsules. Other characters, e.g., karyology (Fontanesia has x 5 13 and Forsythieae x 5 14; Taylor, 1945) and chemical data (only Forsythieae contains cornoside; Damtoft, Franzyk, and Jensen, 1994), also support the close relationship between Forsythia and Abeliophyllum. But as can be seen in Fig. 2, Fontanesia is resolved as sister group to the Forsythieae clade. This is the fact in most of the equally parsimonious trees and, although the branch length is extremely short (one step!), this relationship can be expected to be phylogenetically most probable, because Fontanesia and Abeliophyllum share fruit characters that are much easier interpreted as synapomorphies than parallelisms (J. G. Rohwer, University of Hamburg, Germany, personal communication). Because the strict consensus tree does not resolve the position of Fontanesia and because of the conflict between characters, we continue to leave Fontanesia alone in its own tribe. Subtribe Ligustrinae—Syringa and Ligustrum form a wellsupported basal clade within the new tribe Oleeae. They have dry bilocular capsules and one-to-four-seeded berries (except Ligustrum sempervirens that has dehiscent drupes), respectively. Their fruits are quite similar in development, the only differences being in the development of the mesocarp and fruit dehiscence (Taylor, 1945). Johnson (1957) also stated that Ligustrum is undoubtedly more closely related to Syringa than to the rest of Oleeae, but instead of including Ligustrum in the

OF

BOTANY

[Vol. 87

Syringeae (sensu Taylor), he placed both of them in the Oleeae. Because they form a distinct and well-supported clade in our tribe Oleeae, we have reinstated subtribe Ligustrinae Koehne to accommodate them. In Fig. 3, libriform fibers and multiple vessels are plotted as synapomorphies for the tribe Oleeae (with the plesiomorphic states fiber-tracheids and solitary vessels). This is true in the sense that no taxa outside this clade have this type of wood anatomy, but these features are poorly developed in some taxa of both Ligustrum and Syringa. They are unique in having both fiber types and have most of their vessels solitary rather than in multiples (Baas et al., 1988), i.e., the plesiomorphic states are retained alongside with the apomorphic. All taxa in the Oleeae clade, excluding Ligustrum and Syringa, always have vessel multiples and exclusively libriform fibers. This condition supports the position of Ligustrinae as basal in the Oleeae. The molecular results also indicate that Syringa might be paraphyletic (Fig. 2). Subtribe Schreberinae—The genus Schrebera has a disjunct distribution in Africa and India, but Comoranthus occurs only on Madagascar and the Comores. There is also a report of Schrebera americana Gilg. from Peru. Both genera have bivalved woody capsules (Rohwer, 1996), and, based on overall morphology, it is obvious that they are closely related, if not congeneric (P. S. Green, Royal Botanic Gardens, Kew, personal communication). Johnson (1957) grouped Schrebera and Comoranthus in the new tribe Schrebereae and, pending chromosomal data, provisionally referred it to his subfamily Jasminoideae. Briggs (1970) determined the haploid chromosome number of Schrebera to be 23, but the chromosome number of Comoranthus is still unknown. The present study clearly shows that these genera form a distinct clade that belongs to the same group as the other genera with x 5 23, and we have therefore placed them in the subtribe Schreberinae status novus in the new tribe Oleeae. The chemotaxonomic survey by Harborne and Green (1980) and the wood anatomical study by Baas et al. (1988) also give support to this placement. Subtribe Fraxininae—This new subtribe contains only the genus Fraxinus. It is a circumpolar genus of the northern hemisphere, comprising ;50 species of mainly trees. The genus is characterized by large pinnate leaves and samaras, and there is no doubt that it represents a monophyletic group. Because of the fruit type, Fontanesia was included in the Fraxineae by Bentham (1876) and Knoblauch (1895). However, the samaras in Fontanesia and Abeliophyllum, compared with those of Fraxinus, are neither morphologically nor developmentally similar. Instead, the samara of Fraxinus shows an internal structure very similar to that of the loculicidal capsule of Syringa (Rohwer, 1996). The fruit of Fraxinus has two ovules per locule but usually only one ovule develops, making the samara one-seeded. In contrast, Fontanesia and Abeliophyllum have only one ovule per locule, and although both ovules start to develop, the mature fruit is usually one-seeded (Rohwer, 1996). There are also differences in the morphology of the wing. In the long terminal wing of Fraxinus the fibers run longitudinally, and in Fontanesia’s short lateral wings, they run obliquely perpendicular (Rohwer, 1993). Subtribe Oleinae—The subtribe Oleinae, former tribe Oleeae, is characterized by drupes. Although this group does

December 2000]

WALLANDER

AND

ALBERT—PHYLOGENY

not receive strong jackknife support, there is nevertheless no doubt that this is a monophyletic group. Relationships between the genera in this subtribe are difficult to elucidate, with neither the cpDNA data nor morphology giving a clear answer. Several studies have found that some genera in this group may be polyphyletic as presently circumscribed, e.g., Olea (Altamura, Altamura, and Mazzolani, 1985, 1987; Kiew, 1979) and Osmanthus (Johnson, 1957), and our study can only confirm this suspicion. For example, Olea brachiata (formerly placed in the separate genus Tetrapilus Lour.; Johnson, 1957) seems to be more related to Chionanthus, and so does Osmanthus americanus, the only New World species of Osmanthus (once treated separately in Amarolea Small; Johnson, 1957). These results are also supported by wood anatomy (Baas et al., 1988). Within this subtribe lies a complex of five supposedly more closely related Old World genera, distributed mainly in the subtropics: Osmanthus (except O. americanus), Phillyrea, Picconia, Nestegis, and Notelaea. There is no jackknife support for this grouping, but it is shown in the strict consensus tree (Fig. 1). Green (1958) mentioned this generic complex and Baas et al. (1988) found some support for its monophyly in wood anatomical characters. The synapomorphies are dendritic vessel distribution and vascular tracheids. The similarity in fruit morphology between Phillyrea and Picconia has been pointed out by Taylor (1945), and other characters by Johnson (1957). W. K. Harris (University of Queensland, personal communication) has found that, based on nuclear ITS sequences, the Australian, New Zealand, and New Caledonian taxa of Osmanthus and Nestegis should be included in Notelaea. Generic delimitations in this complex are admittedly difficult (P. S. Green, Royal Botanic Gardens, Kew, personal communication) and further studies, using loci with more variation than in the present study (e.g., ITS), are needed to clarify relationships within the entire subtribe Oleinae. Hesperelaea—The genus Hesperelaea is now extinct (Moran, 1996), but we were successful in obtaining an rps16 intron sequence from the type specimen, the one and only collection from 1875. It is only known from its type locality on Guadalupe, a Mexican island off Baja California. Hesperelaea was collected by Edward Palmer and described by Asa Gray (Watson, 1876) as H. palmeri, a new monotypic genus of Oleaceae. When collected, Palmer found only three old trees alive, no young trees, but several dead ones. The area was heavily grazed by goats, which presumably led to Hesperelaea’s extinction (Moran, 1996). Not much is known about the genus; Gray’s description was rather short, but noteworthy was that its flowers had four stamens. In other genera of Oleaceae, the most common condition is two stamens, but four stamens occasionally occur, e.g., in Chionanthus, Osmanthus, Noronhia, Schrebera, and Forestiera. The fruit was a drupe and so it was placed by Johnson (1957) among the other genera with drupes in his tribe Oleeae (the new Oleinae). Despite having only part of the rps16 intron sequence to confirm this placement, we feel sure that this is correct. Conclusions—This study has presented molecular evidence, congruent with other data, that requires a revised classification of the Oleaceae. (1) The subfamily level is abandoned because Jasminoideae is paraphyletic. (2) The monophyly of the former Oleoideae—here recognized as tribe Oleeae—is strongly supported and treated equal in status to the former jasminoid

AND CLASSIFICATION OF

OLEACEAE

1839

tribes Fontanesieae, Forsythieae, Myxopyreae, and Jasmineae. (3) The tribe Jasmineae is sister to Oleeae. This relationship is supported by chemical data. (4) The long-debated genera Nyctanthes and Dimetra clearly belong to the Oleaceae. (5) The position of the hitherto genus incertae sedis Myxopyrum is supported as sister to Nyctanthes and Dimetra. All three genera are placed in Myxopyreae. (6) The monophyly of the subtribe Oleinae, characterized by drupes, is supported. (7) The rps16 sequence of Hesperelaea palmeri, known only from the type specimen collected in 1875, confirms the placement of this extinct taxon in the subtribe Oleinae. (8) A closer relationship between a group of five genera in the Oleinae, viz. Osmanthus, Picconia, Phillyrea, Nestegis, and Notelaea, is suggested by molecular data and has morphological and wood anatomical support. (9) The two noncoding chloroplast loci, the rps16 intron and the trnL-F region, have proven useful for this infrafamilial study, in combination giving over 500 informative sites. In contrast, the variation at infra- and intergeneric level in the Oleeae, especially in the genus Fraxinus and in the subtribe Oleinae, is too low to be useful. LITERATURE CITED AGARDH, J. 1858. Nyctantheae. Theoria Systematis Plantarum: 284. AIRY SHAW, H. K. 1952. Note on the taxonomic position of Nyctanthes L. and Dimetra Kerr. Kew Bulletin (1952): 271–272. ALTAMURA, L., M. M. ALTAMURA, AND G. MAZZOLANI. 1987. Elements for the revision of the genus Olea (Tourn.) L. VII. The taxa present in Asia which can be ascribed to Olea and allied genera. Annali di Botanica 45: 119–134. ALTAMURA, M. M., L. ALTAMURA, AND G. MAZZOLANI. 1985. Elements for the revision of the genus Olea (Tourn.) L. VI. The taxa present in Oceania which can be ascribed to Olea and allied genera. Annali di Botanica 43: 45–52. ANDERSSON, L., AND J. H. E. ROVA. 1999. The rps16 intron and the phylogeny of the Rubioideae (Rubiaceae). Plant Systematics and Evolution 214: 161–186. APG (ANGIOSPERM PHYLOGENY GROUP). 1998. An ordinal classification for the families of flowering plants. Annals of the Missouri Botanical Garden 85: 531–553. BAAS, P., P. M. ESSER, M. E. T. VAN DER WESTEN, AND M. ZANDEE. 1988. Wood anatomy of the Oleaceae. IAWA Bulletin 9: 103–182. BENTHAM, G. 1876. Oleaceae. In G. Bentham and J. D. Hooker [eds.], Genera Plantarum 2: 672–680. BIGAZZI, M. 1989. Ultrastructure of nuclear inclusions and the separation of Verbenaceae and Oleaceae (including Nyctanthes). Plant Systematics and Evolution 163: 1–12. BRIGGS, B. G. 1970. Some chromosome numbers in the Oleaceae. Contributions from the New South Wales National Herbarium 4: 126–129. CRAIB, W. G. 1916. Contributions to the flora of Siam. Additamentum IX. Kew Bulletin (1916): 265. DAMTOFT, S., H. FRANZYK, AND S. R. JENSEN. 1994. Biosynthesis of iridoids in Forsythia spp. Phytochemistry 37: 173–178. ———, ———, AND ———. 1995. Biosynthesis of secoiridoids in Fontanesia. Phytochemistry 38: 615–621. FARRIS, J. S., V. A. ALBERT, M. KA¨LLERSJO¨, D. LIPSCOMB, AND A. G. KLUGE. 1996. Parsimony jackknifing outperforms neighbor-joining. Cladistics 12: 99–124. GEORGE, K., AND S. GEETHAMMA. 1983. Lactopropionic orcein as a suiatble stain for chromosomes of Oleaceae. Current Science 52: 733–734. ———, AND ———. 1984. Cytological and other evidences for the taxonomic position of Nyctanthes arbor-tristis L. Current Science 53: 439– 441. ———, ———, AND C. A. NINAN. 1989. Chromosome evolution in Oleaceae. Journal of Cytology and Genetics 24: 71–77. GIELLY, L., AND P. TABERLET. 1994. Chloroplast DNA polymorphism at the intrageneric level and plant phylogenies. Comptes Rendus de l’Academie des Sciences, Serie III, Sciences de la Vie 317: 685–692. GREEN, P. S. 1958. A monographic revision of Osmanthus in Asia and America. Notes from the Royal Botanical Garden Edinburgh 22: 439–542.

1840

AMERICAN JOURNAL

HARBORNE, J. B., AND P. S. GREEN. 1980. A chemotaxonomic survey of flavonoids in leaves of the Oleaceae. Botanical Journal of the Linnean Society 81: 155–167. HOLMGREN, P. K., N. H. HOLMGREN, AND L. C. BARRETT. 1990. Index Herbariorum, part I, The herbaria of the world. New York Botanical Garden Press, Bronx, New York, USA. JENSEN, S. R. 1992. Systematic implications of the distribution of iridoids and other chemical compounds in the Loganiaceae and other families of the Asteridae. Annals of the Missouri Botanical Garden 79: 284–302. JOHNSON, L. A. S. 1957. A review of the family Oleaceae. Contributions from the New South Wales National Herbarium 2: 395–418. KAPIL, R. N., AND R. S. VANI. 1966. Nyctanthes arbor-tristis Linn.: embryology and relationships. Phytomorphology 16: 553–563. KERR, A. F. G. 1938. XIX Contributions to the flora of Siam. Kew Bulletin (1938): 127–133. KIEW, R. 1979. Florae Malesianae praecursores LX. The Oleaceae of Malesia. II. The genus Olea. Blumea 25: 305–313. ———. 1983. Two unusual Chionanthus species from Borneo and the position of Myxopyrum in the Oleaceae. Journal of the Arnold Arboretum 64: 619–626. ———. 1984. The genus Myxopyrum L. (Oleaceae). Blumea 29: 499–512. ———, AND P. BAAS. 1984. Nyctanthes is a member of Oleaceae. Proceedings of the Indian Academy of Science 93: 349–358. KIM, K.-J. 1999. Phylogeny of the olive family (Oleaceae). In Abstracts. XVI International Botanical Congress, St. Louis. (abstract number 19.2.6, p. 227). ———, AND R. K. JANSEN. 1993. Phylogeny of Oleaceae based on ndhF sequence variation and chloroplast genome rearrangements. In Abstracts. XV International Botanical Congress, Tokyo. (abstract number 1065, p. 209). KNOBLAUCH, E. 1895. Oleaceae. In A. Engler [ed.], Die Natu¨rlichen Pflanzenfamilien IV, 2: 1–16. KSHETRAPAL, S., AND Y. D. TIAGI. 1970. Structure, vascular anatomy and evolution of the gynoecium in the family Oleaceae and their bearing on the systematic position of genus Nyctanthes L. Acta Botanica Academiae Scientiarum Hungaricae 16: 143–151. KUNDU, B. C., AND A. DE. 1968. Taxonomic position of the genus Nyctanthes. Bulletin of the Botanical survey of India 10: 397–408. KURIACHEN, P. M., AND Y. S. DAVE. 1989. Structural studies in the fruits of Oleaceae with discussion on the systematic position of Nyctanthes L. Phytomorphology 39: 51–60. MOHAN, J. S. S., AND J. A. INAMDAR. 1983. Studies of the leaf architecture of the Oleaceae with a note on the systematic position of the genus Nyctanthes. Feddes Repertorium 94: 201–211. MORAN, R. 1996. The Flora of the Guadalupe Island, Mexico. Memoirs of the California Academy of Sciences 19: 28, 40–43, 128–129 (pages about Hesperelaea). MURTHY, G. S. R., K. M ALEYKUTTY, V. S. RAO, AND J. A. INAMDAR. 1978. Vessels of Oleaceae and Verbenaceae. Feddes Repertorium 89: 359–368. OLMSTEAD, R. G., AND P. A. REEVES. 1995. Evidence for the polyphyly of the Scrophulariaceae based on chloroplast rbcL and ndhF sequences. Annals of the Missouri Botanical Garden 82: 176–193. OXELMAN, B., M. LIDE´N, AND D. BERGLUND. 1997. Chloroplast rps16 intron phylogeny of the tribe Sileneae (Caryophyllaceae). Plant Systematics and Evolution 206: 393–410. PERSSON, C. 2000. Phylogeny of Gardenieae (Rubiaceae) based on chloroplast DNA sequences from the rps16 intron and trnL (UAA)-F(GAA) intergenic spacer. Nordic Journal of Botany 20: 257–269. QIN, X.-K. 1996. The use of peroxidases in the systematics of Oleaceae. Acta Botanica Yunnanica 18: 159–166. ROHWER, J. G. 1993. A preliminary survey of the fruits and seeds of the Oleaceae. Botanische Jahrbu¨cher fu¨r Systematik, Pflanzengeschichte und Pflanzengeographie 115: 271–291. ———. 1994. Fruits and seeds of Nyctanthes arbor-tristis L. (Oleaceae): a comparison with some Verbenaceae. Botanische Jahrbu¨cher fu¨r Systematik, Pflanzengeschichte und Pflanzengeographie 115: 461–473. ———. 1995. Fruit and seed structures in Menodora (Oleaceae): a comparison with Jasminum. Botanica Acta 108: 163–168. ———. 1996. Die Frucht- und Samenstrukturen der Oleaceae. Bibliotheca Botanica 148: 1–177. ———. 1997. The fruits of Jasminum mesnyi (Oleaceae), and the distinction between Jasminum and Menodora. Annals of the Missouri Botanical Garden 84: 848–856.

OF

BOTANY

[Vol. 87

SMALL, R. L., J. A. RYBURN, R. C. CRONN, T. SEELANAN, AND J. F. WENDEL. 1998. The tortoise and the hare: choosing between noncoding plastome and nuclear ADH sequences for phylogeny reconstruction in a recently diverged plant group. American Journal of Botany 85: 1301–1315. STANT, M. Y. 1952. Anatomical evidence for including Nyctanthes and Dimetra in the Verbenaceae. Kew Bulletin 7: 273–276. SWOFFORD, D. L. 2000. PAUP*: phylogenetic analysis using parsimony (* and other methods), version 4. Sinauer, Sunderland, Massachusetts, USA. TABERLET, P., L. GIELLY, G. PANTOU, AND J. BOUVET. 1991. Universal primers for amplification of three non-coding regions of chloroplast DNA. Plant Molecular Biology 17: 1105–1109. TAKHTAJAN, A. 1997. Oleanae—Oleales—Oleaceae. In Diversity and classification of flowering plants, 449–451. Columbia University Press, New York, New York, USA. TAYLOR, H. 1945. Cyto-taxonomy and phylogeny of the Oleaceae. Brittonia 5: 337–367. WAGSTAFF, S. J., AND R. G. OLMSTEAD. 1997. Phylogeny of Labiatae and Verbenaceae inferred from rbcL sequences. Systematic Botany 22: 165– 179. WATSON, S. 1876. Botanical contributions. I. On the flora of Guadalupe Island, Lower California. Proceedings of the American Academy of Arts and Sciences XI: 105–112. WENG, R.-F., AND M.-Z. ZHANG. 1992. Chromosome numbers in Chinese Oleaceae I. Investigatio et Studium Naturae 12: 66–77.

APPENDIX A new generic and suprageneric classification of Oleaceae. Source for most of the nomenclatural information is the Indices Nominum Supragenericorum Plantarum Vascularium Project Database (http://matrix.nal.usda.gov:8080/star/ supragenericname.html). Fam. Oleaceae Hoffmanns. & Link, Fl. Portug. 1: 385. 1813–1820 (Oleinae), nom. cons. Synonyms: Bolivariaceae Griseb., Gen. Sp. Gent.: 20. Oct 1838. Forestieraceae Endl., Ench. Bot.: 174. 15–21 Aug 1841 (Forestiereae). Fraxinaceae Vest, Anleit. Stud. Bot.: 269, 288. 1818 (Fraxinoideae). Jasminaceae Adans., Fam. Pl. 2: 220. Jul–Aug 1763 (Jasmina). Ligustraceae G. Mey., Chloris Han.: 245, 254. Jul–Aug 1836 (Ligunstrinae). Lilacaceae Vent., Tabl. Re`gne Ve´g. 2: 307. 5 Mai 1799 (Lilaceae), nom. illeg. Nyctanthaceae J. Agardh, Theoria Syst. Pl.: 284. Apr–Sep 1858 (Nyctantheae). Schreberaceae (Wight) Schnizl., Iconogr. Fam. Regni Veg. 2: ad t. 151*. 1857–1870. Syringaceae Horan., Char. Ess. Fam.: 115. 1847. Tribe Oleeae (Hoffmanns. & Link ex R. Br.) Dumort., Fl. Belg.: 52. 1827 (Oleineae). Subtribe Oleinae Synonyms: Tribe Chionantheae DC., Prodr. 8: 294. mid Mar 1844. Tribe Forestiereae Horan., Char. Ess. Fam.: 80. 1847 (Forrestiereae). Subtribe Forestierinae Koehne, Deut. Dendrol.: 500. Mai 1893 (Forestiereae). Tribe Hesperelaeeae Baill., Hist. Pl. 11: 242, 249. Sep–Oct 1891. Tribe Notelaeeae G. Don, Gen. Hist. 4: 44, 51. 1837–8 Apr 1838 (Notelaeiae). Accepted genera: Chionanthus L. Forestiera Poir. Haenianthus Griseb. Hesperelaea A. Gray Nestegis Rafin. Noronhia Stadtmann ex Thouars Notelaea Vent. Olea L. Osmanthus Lour. Phillyrea L. Picconia DC. Priogymnanthus P.S. Green

December 2000]

WALLANDER

AND

ALBERT—PHYLOGENY

Subtribe Fraxininae (Vest) E. Wallander & V. A. Albert, stat. nov. Basionym: Fam. Fraxinaceae Vest, Anleit. Stud. Bot.: 269, 288. 1818 (Fraxinoideae). Synonyms: Subfam. Fraxinoideae Kostel., Allg. Med.-Pharm. Fl. 3: 998. Apr–Dec 1834 (Fraxineae). Tribe Fraxineae Bartl., Ord. Nat. Pl.: 218. Sep 1830 (Fraxinea). Accepted genus: Fraxinus L. Subtribe Schreberinae (Wight) E. Wallander & V. A. Albert, stat. nov. Basionym: Subfam. Schreberoideae Wight, Ill. Ind. Bot. 2: 185. 1850 (Schreberaceae). Accepted genera: Schrebera Roxb. Comoranthus Knobl. Subtribe Ligustrinae Koehne, Deut. Dendrol.: 500. Mai 1893 (Ligustreae). Synonyms: Tribe Ligustreae Burnett, Outl. Bot.: 1022, 1102. Jun 1835. Subfam. Syringoideae Leurss., Handb. Syst. Bot. 2: 1041. Nov 1882 (Syringeae). Tribe Syringeae Burnett, Outl. Bot.: 1022, 1103. Jun 1835. Accepted genera: Syringa L.

AND CLASSIFICATION OF

OLEACEAE

1841

Ligustrum L. Tribe Jasmineae Lam. & DC., Syn. Pl. Fl. Gall.: 216. 30 Jun 1806. Synonym: Tribe Bolivarieae Horan., Char. Ess. Fam.: 116. 1847. Accepted genera: Jasminum L. Menodora Humb. & Bonpl. Tribe Myxopyreae Boerl., Handl. Fl. Nederl. Ind. 2: 324. 1 Jan 1899. Synonym: Subfam. Myxopyroideae Boerl., Handl. Fl. Nederl. Ind. 2: 324. 1 Jan 1899 (Myxopyreae). Accepted genera: Myxopyrum Blume Nyctanthes L. Dimetra Kerr Tribe Forsythieae H. Taylor ex L. Johnson, Contrib. N.S.W. Natl. Herb. 2: 397. 1957. Accepted genera: Abeliophyllum Nakai Forsythia Vahl Tribe Fontanesieae H. Taylor ex L. Johnson, Contrib. N.S.W. Natl. Herb. 2: 397. 1957. Accepted genus: Fontanesia Labill.

Paper I

Paper II

Paper III

Paper IV

Paper V

Paper VI

Evolution of wind-pollination and gender specialisation in Oleaceae – exaptations and adaptations Eva Wallander

ABSTRACT In Oleaceae, six of 24 extant genera have wind-pollinated species. These are the genus Fraxinus (subtribe Fraxininae) and five genera belonging to the subtribe Oleinae (Forestiera, Nestegis, Olea, Phillyrea, and Priogymnanthus). The ancestral condition in the family is entomophily (insect-pollination). Adaptations to anemophily (wind-pollination) appear to have seven independent origins, including three within the genus Fraxinus. All of the wind-pollinated species occur either in temperate forests or other seasonally deciduous areas. The ancestral breeding system in the family is hermaphroditism and most of the insect-pollinated taxa are hermaphrodites. The rare androdioecious breeding system (separate male and hermaphrodite individuals) occurs in four genera (Chionanthus, Fraxinus, Phillyrea, and the polyphyletic Osmanthus) and appears to have evolved from hermaphroditism on five occasions. Six genera have dioecious species: Chionanthus, Forestiera, Fraxinus, Nestegis, Olea, and Osmanthus, and most of these species are anemophilous. In at least one case (Fraxinus), dioecy has evolved via androdioecy. One clear trend in the floral evolution of Oleaceae is the shift from conspicuous, white or yellow, fragrant and usually nectariferous flowers, to smaller and inconspicuous, white to yellowish-green flowers offering only pollen as a reward. The latter type of flowers is generally pollinated by unspecialised insects, and its presence is a common feature in the subtribe Oleinae. This is interpreted as an exaptation that may select for ambophily (both wind- and insect-pollination) in suitable habitats and ultimately may lead to anemophily. Androdioecy and dioecy, where unisexual males specialise in pollen production and dispersal, are adaptations to ambophily and anemophily. Hermaphrodite flowers, in androdioecious and ambophilous species, have partially or fully functional stamens. These may have a potential male function, as reproductive assurance, or may be retained for pollinator attraction. The presence of unisexual flowers, especially male flowers, in anemophilous and ambophilous species is interpreted mainly as selection for sexual specialisation, not selection for outcrossing. The paper includes a short summary of ecological factors favouring unisexual flowers in anemophilous or ambophilous species, which support these arguments

1

INTRODUCTION The olive family, Oleaceae, is a medium-sized family of about 600 species in 24 extant and world-wide genera (Wallander and Albert 2000). Many of the insect-pollinated genera are well-known and are extensively cultivated, e.g. the jasmines (Jasminum), lilacs (Syringa), privets (Ligustrum), and Forsythia (authors of names are only given if not listed in Table 1 or 2). The family also contains a number of genera with species that are wind-pollinated, of which the most well-known is perhaps the genus Fraxinus. Examples of other wind-pollinated species are the American genus Forestiera and some olives, particularly Olea europaea L. Although flowers of the latter species possess small corollas and appear to be visited by unspecialised insects, they produce large amounts of pollen that are dispersed by wind. Other genera with windpollinated flowers are Phillyrea in the Mediterranean and Nestegis in New Zealand. The genus Fraxinus comprises 43 species (Wallander ms.), two thirds of which are anemophilous (wind-pollinated) and one third of which is entomophilous (insect-pollinated). Studies on the evolution of wind-pollination in the genus, based on a molecular phylogeny of nearly all species, revealed no less than three independent origins of anemophily (Wallander ms.). The origin of anemophily was preceded by ambophily (both insect- and wind-pollination), which was interpreted as an exaptation (as defined by Gould and Vrba 1982). In two of these instances, anemophily was followed by three independent shifts in breeding system to dioecy, once from hermaphrodites via androdioecy and twice via polygamy, and dioecy was interpreted as an adaptation that evolved through selection for sexual specialisation of the wind-pollinated flowers. Apart from the suspicion that there might have been multiple origins of anemophily in the family as well, there is also a fascinating diversity of breeding systems that seems to be connected to the pollination mode. Particularly, it is remarkable that so many species exhibit phenotypic and in some cases even proven functional androdioecy (male and hermaphrodite flowers on separate individuals), which is considered to be a very rare breeding system (Charlesworth 1984). The aim of this study was to estimate the number of times adaptations to anemophily 2

have evolved in Oleaceae and which other traits that are correlated to the shift in pollination mode, e.g. breeding system, floral morphology, etc. The study is based on an estimate of phylogenetic relationships among all genera of the family Oleaceae and I develop a hypothesis to explain why anemophily has evolved repeatedly in this family and what might be exaptations and adaptations for this pollination system. In addition, I summarise hypotheses on the evolution of dioecy and androdioecy in order to explain their prevalence in Oleaceae.

MATERIALS AND METHODS The phylogenetic relationships among all genera of Oleaceae (Wallander and Albert 2000), based on two noncoding chloroplast loci, is not resolved in the subtribe Oleinae. Therefore, a well-resolved ITS phylogeny of this group (Wallander, Green, and Harris, unpublished data) has been combined with the chloroplast phylogeny to yield a resolved tree of the Oleaceae. The positions of two genera missing in the ITS phylogeny (Hesperelaea and Haenianthus) were inferred from other molecular data (Lee and Kim 2000) and flavonoid data (Harborne and Green 1980). Some genera in the Oleinae, e.g. Osmanthus and Chionanthus, are polyphyletic as presently circumscribed and where it has bearing on the hypotheses tested here, the genera will be split on the tree accordingly. Information gathered on floral morphology, breeding and pollination system for the 25 genera of Oleaceae, through field, herbarium and literature studies, is summarised in Table 1. More detailed information on the floral morphology for taxa exhibiting a dimorphic breeding system, dioecy or androdioecy, is shown in Table 2. These data were optimised onto the phylogenetic tree of Oleaceae using MacClade version 3.08 (Maddison and Maddison 1999). Both ACCTRAN and DELTRAN optimisations were tested. The genus Fraxinus has been studied in detail (Wallander ms.) and the phylogeny from this study has here been included in order to map the occurrences of traits in different sections of the genus. For a more detailed resolution and sequence of evolutionary events, reference should be made to Wallander (ms.)

Nyctanthes L.

Dimetra Kerr

Fontanesia Labill.

Abeliophyllum Nakai

Forsythia Vahl

Menodora Humb. & Bonpl. 24

Jasminum L.

Ligustrum L.

Syringa L.

Myxopyreae

Myxopyreae

Fontanesieae

Forsythieae

Forsythieae

Jasmineae

Jasmineae

Oleeae: Ligustrinae

Oleeae: Ligustrinae

Insect

Long-tubed & sweet-scented or shorttubed & inodorous, yellow or white

43

Fraxinus L.

Chionanthus L.

Forestiera Poir.

Haenianthus Griseb.

Hesperelaea A. Gray

Nestegis Rafin.

Noronhia Stadm. ex Thou.

Notelaea Vent.

Olea L.

Osmanthus Lour.

Phillyrea L.

Picconia DC.

Priogymnanthus P.S. Green 2

Oleeae: Fraxininae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Oleeae: Oleinae

Insect

Small, white or yellow, corolla lobes divided to the base or with short tube

Total number of species: 600+

1(-2)

2

30

40+

12

41

5

S America (Bolivia, Brazil, Paraguay, Ecuador)

Macaronesia

Mediterranean region to W Asia

Subtropical parts of E Asia and N America (1 sp.)

Tropical and subtropical parts of the Old World

Australia and Tasmania

Madagascar

New Zealand and Hawaii (1 sp.)

Insect

Insect

Wind

Insect

Insect

Petals very early caducous

Small whitish corolla

Small whitish corolla

Small, white or yellowish corolla, extrafloral nectaries

Wind?

Insect

Wind

Insect

Small, whitish , lobes shorter than corolla Insect & tube, stamens exserted in some spp. wind

Small corolla

Small, fleshy & globular, nectar

Apetalous

Yellow corolla, four stamens

Small, white, deeply 4-lobed corolla

Subtropical N America, C America, and West Indies Apetalous West Indies

Wind

Insect & Wind

Small, with 4 (or 2) free or fused white petals, some scented, or without petals

Mainly temperate and subtropical regions of the Northern Hemisphere Tropical and subtropical parts of Africa, America, Asia, and Australia

Insect

White or pinkish corolla

Insect

Insect

Tropical parts of Africa and India

1 (extinct) Mexico (was endemic to Guadalupe Island)

3

ca 15

ca 100

4

Oleeae: Schreberinae Schrebera Roxb.

White or pinkish corolla

White to purple, nectariferous

Mainly subtropical parts of Eurasia Madagascar and the Comores

20 3

Oleeae: Schreberinae Comoranthus Knobl.

Insect

Insect

Insect

Yellow, strongly scented, stamens not exserted

White or yellow, scented, nectariferous

Insect

Insect

Insect

Insect

Insect

Hermaphroditic

Hermaphroditic

Androdioecious

Hermaphroditic, androdioecious

Hermaphroditic, a few dioecious or andromonoecious

Hermaphroditic

Hermaphroditic

Dioecious

Hermaphroditic

Hermaphroditic

Dioecious, a few hermaphroditic

Hermaphroditic, androdioecious or dioecious

Hermaphroditic, androdioecious, polygamous, or dioecious

Hermaphroditic, heterostylous

Hermaphroditic

Hermaphroditic

Hermaphroditic

Hermaphroditic, some heterostylous

Hermaphroditic

Hermaphroditic

Hermaphroditic

Hermaphroditic

Hermaphroditic

Hermaphroditic

Hermaphroditic

Pollination Breeding system(s) system(s)

White or pinkish corolla

Small, deeply lobed corolla

Tubular, white with yellow mouth

Tubular with sessile anthers, white or orange, night-scented

Small yellow or pink corolla

Floral morphology

Temperate to tropical parts of the Old World, except Usually white, nectariferous Africa

Tropical and subtropical parts of the Old World

Subtropical N and S America and S Africa

E Asia and SE Europe (one sp.)

Korea

SW Asia (and Sicily) and China

Thailand

Tropical and subtropical SE Asia

Tropical SE Asia

Distribution

45

200+

11

1

1-2

1

2

4

Myxopyrum Blume

Myxopyreae

No. of species

Genus

Tribe: subtribe

Table 1. The 25 genera of Oleaceae, with tribal and subtribal assignments according to Wallander and Albert (2000), the approximate number of species and their world distribution, short floral descriptions, and occurrences of different pollination and breeding systems.

Table 2. Taxa in Oleaceae exhibiting a dimorphic breeding system and their geographical distribution. Taxon (no. species)

Distribution

Breeding system and floral morphology

References

Chionanthus retusus Lindl. & Paxt.

Japan, Korea, China

Androdioecious; white corolla, fragrant?, apparently both sexual functions in hermaphrodites

Ueda 1996

Chionanthus virginicus L.

N America

Dioecious; white corolla, small and indehiscent stamens in hermaphrodites

Rehder (1904)

Forestiera Poir. (8 spp.)

N & C America

Dioecy; apetalous flowers with rudimentary stamens in female flowers, anemophilous

Brooks (1977)

Nestegis Rafin. (4 spp.)

New Zealand

Dioecy; unisexual flowers rudimentary organs of the opposite sex, anemophilous

Green (1963b)

Olea L. (~10 spp.)

Tropical Asia and China

Dioecy or polygamodioecy; small whitish flowers, 2 apetalous

Kiew (1979), Chang et al. (1996)

Olea L. (2-? spp.)

Africa

Andromonoecious; small whitish flowers, exserted stamens, ambophilous

Dyer (1991)

Osmanthus Lour.

Androdioecious; small white and usually fragrant flowers, functionally male ones with abortive pistil

Green (1958)

Sect. Linocieroides (1 sp.)

Eastern Asia (from northern India and Nepal to China and Japan)

Osmanthus Lour. Sect. Notosmanthus

New Caledonia

Androdioecious; small white flowers, functionally male ones with abortive pistil and larger stamens than hermaphrodites

Green (1963a)

USA and Mexico (1 sp.), East Asia (5-6 spp.)

Dioecious; female flowers with rudimentary stamens and male ones with abortive pistil

Green (1958)

Mediterranean area to West Asia

Functionally androdioecious, wind-pollinated male and hermaphrodite flowers with a small whitish corolla, anemophilous

Lepart & Dommée 1992, Aronne & Wilcock 1992, Herrera et al. 1994, Vassiliadis 1999, Vassiliadis et al. 2000

Fraxinus L. (10 spp. in sect. Ornus)

Eurasia

Androdioecious; male flowers with a vestigial pistil, petalous or apetalous

Ishida & Hiura (1998), Dommée et al. (1999), Wallander (ms.)

Fraxinus L. (14 spp. in sect. Fraxinus and sect. Ornus)

Eurasia and North America

Dioecious; apetalous, with or without vestigial organs of the opposite sex

Wallander (ms.)

Fraxinus L. (12 spp. in sect. Fraxinus)

Eurasia and North America

Polygamous; apetalous, some asepalous

Wallander (ms.)

Sect. Osmanthus (17 species) Sect. Siphosmanthus (2 spp.)

(4 spp.)

Osmanthus Lour. Sect. Leiolea (6-7 spp.) Phillyrea angustifolia L. Phillyrea latifolia L.

Total no. of dimorphic species: 97

Fraxinus sect. Dipetalae

Fraxinus sect. Fraxinus

TRIBE

OLEEAE

wind-pollination

OLEINAE

insect-pollination

SUBTRIBE

Pollination system

Hesperelaea Haenianthus Forestiera Priogymnanthus Noronhia Picconia Osmanthus sect. Osmanthus Osmanthus sect. Siphosmanthus Osmanthus sect. Linocieroides Phillyrea Notelaea Osmanthus sect. Nothosmanthus Nestegis Osmanthus sect. Leiolea Olea Chionanthus Chionanthus retusus & Ch. virginicus

FRAXININAE

Fraxinus sect. Ornus

SUBTRIBE

Myxopyrum Nyctanthes Dimetra Fontanesia Abeliophyllum Forsythia Menodora Jasminum Ligustrum Syringa Schrebera Comoranthus

Figure 1. Occurrences of wind-pollination optimised onto a phylogenetic tree of all genera of Oleaceae, based on molecular data (Wallander and Albert 2000, Wallander ms., and Wallander, Green, and Harris, unpublished data). Not all species of these genera are wind-pollinated. For more detailed optimisation of traits in the genus Fraxinus, reference should be made to Wallander (ms.), where all species are included. DELTRAN optimisation shown.

RESULTS AND DISCUSSION Figure 1 and 2 show the occurrences of anemophily and different breeding systems, respectively, in the family. Insect-pollination is ancestral in the family and all taxa of genera outside subtribes Fraxininae and Oleinae have hermaphrodite flowers. These are usually quite showy and fragrant and many provide nectar as reward for its pollinators, mainly bees and butterflies. Examples are the flowers of Jasminum, Syringa, Ligustrum, and Forsythia. The trend in reduction of floral showiness is shown in Figure 3.

Anemophily has evolved seven times in the family (Figure 1), including three times in the genus Fraxinus (see Wallander ms. for a more detailed tree and character mapping). Within subtribe Oleinae there are five genera that contain species with adaptations to wind-pollination (Nestegis, Olea, Phillyrea, Forestiera, and Priogymnanthus). All these represent independent origins of anemophily, except Forestiera and Priogymnanthus that may be sister taxa and represent one origin (see under Priogymnanthus below). 5

Fraxinus sect. Dipetalae

Fraxinus sect. Fraxinus

dioecy

OLEEAE

androdioecy

TRIBE

polygamy

OLEINAE

hermaphroditism

SUBTRIBE

Breeding system

Hesperelaea Haenianthus Forestiera Priogymnanthus Noronhia Picconia Osmanthus sect. Osmanthus Osmanthus sect. Siphosmanthus Osmanthus sect. Linocieroides Phillyrea Notelaea Osmanthus sect. Nothosmanthus Nestegis Osmanthus sect. Leiolea Olea Chionanthus Chionanthus retusus & Ch. virginicus

FRAXININAE

Fraxinus sect. Ornus

SUBTRIBE

Myxopyrum Nyctanthes Dimetra Fontanesia Abeliophyllum Forsythia Menodora Jasminum Ligustrum Syringa Schrebera Comoranthus

Figure 2. Occurrences of different breeding systems optimised (DELTRAN) onto a phylogenetic tree of all genera of Oleaceae, based on molecular data (Wallander and Albert 2000, Wallander ms., and Wallander, Green, and Harris, unpubl. data). For more detailed optimisation of traits in the genus Fraxinus, reference should be made to Wallander (ms.), where all species are included.

Before discussing evolutionary trends in Oleaceae, I will give a more detailed account on the floral morphology of Fraxinus and the genera of subtribe Oleinae with a dimorphic breeding system, than what is summarised in Table 2.

Genera with dimorphic species in Oleaceae Fraxinus Fraxinus is the sole member of subtribe Fraxininae, sister to subtribe Oleinae (Wallander and Albert 2000), and the 43 recognised species (Wallander ms.) are distributed in temperate 6

and subtropical areas of the northern hemisphere. About one third of the species are entomophilous (or ambophilous), most of which occur in Eurasia (two in North America), and two thirds are anemophilous and occur in both the Americas and Eurasia. Most of the windpollinated species are large trees, but some are also shrubs in arid areas. Seven species are hermaphroditic and most of them are entomophilous. Ten species in section Ornus appear to be functionally androdioecious. Two of them are wind-pollinated and the others appear to be ambophilous (Wallander and Dahl ms.). Two of the ambophilous species have been shown to be functionally androdioecious, F. ornus L.

Fraxinus sect. Dipetalae

Fraxinus sect. Fraxinus

OLEEAE

apetalous

TRIBE

inconspicuous

OLEINAE

showy

SUBTRIBE

Flowers

Hesperelaea Haenianthus Forestiera Priogymnanthus Noronhia Picconia Osmanthus sect. Osmanthus Osmanthus sect. Siphosmanthus Osmanthus sect. Linocieroides Phillyrea Notelaea Osmanthus sect. Nothosmanthus Nestegis Osmanthus sect. Leiolea Olea Chionanthus Chionanthus retusus & Ch. virginicus

FRAXININAE

Fraxinus sect. Ornus

SUBTRIBE

Myxopyrum Nyctanthes Dimetra Fontanesia Abeliophyllum Forsythia Menodora Jasminum Ligustrum Syringa Schrebera Comoranthus

Figure 3. Generalised floral morphology optimised (DELTRAN) onto a phylogenetic tree of Oleaceae, based on molecular data (Wallander and Albert 2000, Wallander ms., and Wallander, Green, and Harris, unpublished data). There may be exceptions in some genera. For more detailed optimisation of traits in the genus Fraxinus, reference should be made to Wallander (ms.), where all species are included.

(Dommée et al 1999, Wallander and Dahl ms.) and F. lanuginosa Koidz. (Ishida and Hiura 1998). There are five wind-pollinated species in section Ornus. One of them, F. longicuspis Sieb. & Zucc., has been studied as well and the preliminary results suggests functional androdioecy (Wallander and Dahl ms.). Three of them are dioecious, with or without rudimentary stamens in the functionally female flowers, and have evolved via androdioecy. Twelve of 23 wind-pollinated species in section Fraxinus are polygamous, with male, female, and hermaphrodite flowers occurring on the same or different trees. The other 11 species are

dioecious and dioecy has two separate origins from polygamous ancestors. Fraxinus is the genus best studied in the family. Androdioecy is derived from hermaphroditism and males have a vestigial pistil, further showing its hermaphrodite ancestry. The transition from polygamy to functional dioecy in two Asian species of subsection Fraxinus is probably rather recent, since their polygamous past is still evident in the rudimentary stamens of functionally female flowers. In contrast, not a single dioecious species in the American species (subsection Melioides) has vestigial organs of the opposite sex in the flowers. 7

Forestiera Forestiera is an American genus and its approximately ten species extend from southern USA south to Panama and West Indies. Typical habitats include stream-sides, usually on limestone, and vary from extremely dry, near desert-like conditions, to moist cloud forest localities (Brooks 1977). The species are deciduous or semi-evergreen shrubs or small trees, two of which are hermaphroditic and the rest dioecious (Brooks 1977). The flowers are apetalous. A minute calyx is present in some species and absent or rare in others, particularly in male flowers. One hermaphroditic and three dioecious species are described as fragrant. Female flowers of dioecious species have 0-4 abortive stamens and male flowers have a rudimentary or wanting pistil. F. rhamnifolia Griseb. has morphologically androdioecious flowers (Bullock 1994), although they are probably functionally dioecious as the stamens in hermaphrodite flowers are much smaller than in male ones. Male flowers have no pistil vestiges. Although I am not aware of any studies on the pollination systems of this genus, they are believed to be mostly wind-pollinated (Bullock 1994). The ITS phylogeny of this genus (Wallander, Green, and Harris, unpublished data) suggests that hermaphroditism here constitutes two reversals from dioecy. Priogymnanthus Priogymnanthus is a newly described genus (Green 1994) and comprises two species in South America: P. apertus in Ecuador and P. hasslerianus in Bolivia, Brazil, and Paraguay. Although differing in number of perianth parts, Green (1994) thought it was related to Forestiera because of its similarity in inflorescence structure (a racemoid reduced dichasium). The chloroplast phylogeny of Oleaceae (Wallander and Albert 2000) also indicated that this might be the case. However, the ITS phylogeny (Wallander, Green, and Harris, unpublished data) showed no close relationship. There are four stamens, a condition that also occurs in some species of Forestiera. Although this tropical genus has hermaphrodite flowers, there is no calyx and the corolla is shed at the beginning of anthesis (Priogymnanthus means “early naked flower”), thus making the flowers essentially apetalous. Both species are deciduous or semideciduous trees that occur in semi-deciduous dry forests and they flower when leafless dur8

ing the dry season. Although none of these species has been studied in the field, floral morphology and habitat suggest that they are windpollinated.

Chionanthus The genus Chionanthus comprises about 80 taxa. As far as I know, Ch. retusus and Ch. virginicus are the only two with a dimorphic breeding system. Ch. virginicus was investigated by Rehder (1904) who found that the male flowers had a normally developed ovary (with ovules) and style but no stigma. The hermaphrodite flowers had functional pistils, but narrower anthers that fell off indehisced with the fading corolla. The hermaphrodite anthers contained numerous pollen grains, but these were smaller compared to those of the male flowers and without the granular structure of the exine. Occasionally, he found anthers in hermaphrodite flowers that opened and shed pollen. Thus the hermaphrodite flowers seem to be functionally more female than male, and Rehder (1904) termed the species imperfectly dioecious. The inflorescences of male plants are showier and have larger panicles with more flowers than females. In contrast, Green (1979) and Ueda (1996) showed that Ch. retusus is at least morphologically androdioecious and that there are no differences in stamen size between males and hermaphrodites. Here the male flowers have even more reduced pistils (Rehder 1904) than in Ch. virginicus. Phillyrea The genus Phillyrea consists of two shrub species distributed in the Mediterranean area to West Asia, Ph. angustifolia and Ph. latifolia. The former has been the subject of extensive field studies by Aronne and Wilcock (1992), Vassiliadis (1999), and Pannell and Ojeda (2000), and the latter by Aronne and Wilcock (1992 and 1994) and Gal Pollak (pers. comm.). Both species have been shown to be functionally androdioecious. The inconspicuous, creamy-whitish flowers have exserted anthers and are windpollinated. In male flowers of Ph. angustifolia there is a rudimentary pistil, whereas it is completely absent in Ph. latifolia (Aronne and Wilcock 1992). Reported sex ratios are 30-50% males (Aronne and Wilcock 1992, Pannell and Ojeda 2000). Male plants of Ph. latifolia produce greater numbers of flowers per node than her-

maphrodites (Aronne and Wilcock 1994). Phillyrea is closely related to Osmanthus (Green 1958, Wallander and Albert 2000, Wallander, Green, and Harris, unpublished data), a genus that includes many morphologically androdioecious species.

Osmanthus The about 30 species in Osmanthus are evergreen shrubs or trees and most of them occur in temperate to subtropical parts of eastern Asia, mainly China, and New Caledonia. One species, O. americanus, is distributed in subtropical parts of USA and Mexico (including O. mexicanus Lundell and O. megacarpus (Small) Small). The flowers are small, whitish to yellowish-greenish, usually fragrant, and most of the species exhibit sexual dimorphism, either as androdioecy or as dioecy. Both chloroplast data (Wallander and Albert 2000) and ITS data (Wallander, Green, and Harris, unpublished data) indicate that the genus, as presently circumscribed, is polyphyletic. The genus has been divided into five sections (Green 1958, 1963 a): Osmanthus (17 species), Linocieroides (2 species), Siphosmanthus (2 species), Nothosmanthus (4 species), and Leiolea (8 species). All taxa in the first three sections occur in Asia and form a monophyletic group. The species of Nothosmanthus occur in New Caledonia and are most closely related to the Australian and New Zealand genera Notelaea and Nestegis. Section Leiolea occurs in the Americas (one species) and eastern Asia (mainly China), and is more closely related to Chionanthus and Olea. The species in the first three sections are androdioecious, male flowers having an abortive pistil and the hermaphrodites having both functions. Joshi (1942) investigated the male flowers of O. suavis King (sect. Siphosmanthus) and describes the ovary as incompletely developed: the gynoecium is not completely bilocular and not closed at the top, the lateral carpel traces are atrophic, there are no ovules (not even rudiments), style, or stigma. In section Nothosmanthus (the New Caledonian species), the species are also androdioecious but the hermaphrodite flowers usually have smaller stamens than the male flowers, suggesting that they might be functionally dioecious. Six of eight species of section Leiolea are dioecious and two hermaphroditic. Male flowers have abortive pistils and the stamens in female flow-

ers are also abortive. Sections Osmanthus, Siphosmanthus and Nothosmanthus consist of shrubs or small trees whereas the species of sections Linocieroides and Leiolea mainly are larger trees. The pollination systems of these taxa are unknown, but they appear to be pollinated by generalist insects, or ambophilous. The larger trees of section Leiolea might perhaps be wind-pollinated. According to Mabberley (1997), extrafloral nectaries are present in Osmanthus, but these have probably no function in pollination (Metcalfe 1938).

Nestegis Four species of Nestegis occur in New Zealand and are apparently wind-pollinated and functionally dioecious (Green 1963 b). They are either large canopy trees or smaller coastal trees (Salmon 1980). The flowers are apetalous but have a small calyx. They are mostly unisexual, male flowers with a vestigial pistil and female flowers with reduced and dysfunctional stamens with empty anthers. There is a variation in the size of the pistil in functionally male flowers, making those with pistils almost as large as in females look like hermaphrodite flowers. These flowers also have slightly smaller stamens than ‘pure’ males, suggesting that there is a trade-off in resource allocation to the sex functions in a flower, and they appear to be functionally dioecious. Green (1963 b) also noted that the functional gender of individual trees seems to vary between years. Some specimens bearing flowers with abortive pistils had fruits from the previous year on the same shoot. This variability in sex expression between years is also known in wind-pollinated species of Fraxinus, e.g. F. excelsior (Wallander and Dahl ms.). One hermaphroditic species, N. sandwicensis (A. Gray) O. & I. Degener & L. Johnson, occurs on Hawaii and its presumably insect-pollinated flowers have a small, white corolla. The genus Nestegis is not monophyletic but included in a monophyletic complex together with Australian and New Caledonian taxa of Notelaea (hermaphroditic; Green 1968) and Osmanthus section Nothosmanthus (androdioecious), where Nestegis sandwicensis appears as sister to the rest (Wallander, Green, and Harris, unpublished data).

9

Olea A number of tropical species of Olea are reported to be dioecious (Kiew 1979), e.g. O. borneensis Boerlage, O. brachiata (Lour.) Merrill, O. decussata (Heine) Kiew (=O. rubrovenia (Elmer) Kiew), O. dentata (Wall.) DC., and O. javanica (Bl.) Knobl. All these taxa are evergreen shrubs or trees, with dull white corolla, some exceptionally small, and with stamens usually shorter than the corolla. Some of these species, and O. polygama Wight, show a close relationship with O. dioica Roxb. (Altamura et al. 1987). I have no information on pollination of these tropical forest trees, but they do not appear to be wind-pollinated. However, O. dioica, which is described as having small white unisexual flowers and with the corolla wanting or caducous in female flowers, is related to O. gamblei Clarke, which is dioecious and without corolla in both sexes (Brandis 1906). Both occur on hills in India and Nepal, together with O. polygama, which is also dioecious but with corolla. Perhaps some of the apetalous and/or dioecious species are wind-pollinated, especially those occurring in open habitats. In the flora of China, two species of Olea, with corolla, are reported to be polygamodioecious (Chang et al. 1996). There are nine recognised species of Olea in Africa and all are hermaphroditic, outcrossing and self-incompatible (Dyer 1991). Floral morphology suggests pollination by generalist insects, but many were found to be ambophilous or anemophilous (Dyer 1991). The pollen morphology in all taxa suggests an anemophilous pollination mode, but anemophily is probably more important in savanna species and entomophily in forest species (Dyer 1991). The fleshy fruits are mainly bird dispersed. Olea europaea, with an abundance of airborne pollen in the Mediterranean region during the flowering season (e.g. Liccardi et al. 1996), is mainly wind-pollinated. In fact, pollen from O. europaea has been identified as one of the most important cause of seasonal respiratory allergy in the Mediterranean area (Liccardi et al. 1996). O. europaea is mostly described as hermaphroditic but also as andromonoecious (Brooks 1948, Uriu 1959). Dyer (1991) suggested that the low fruit to flower ratio observed in at least two of the morphologically hermaphroditic species of Olea may be attributed to functional andromonoecy, where some flowers only function as males. He suggested that the aborted flowers 10

acted as pollinator attractants to enhance effective outcrossing by insects and as pollen donors for effective wind-pollination. Other genera The species of other genera of subtribe Oleinae generally possess hermaphrodite and small, whitish to greenish coloured flowers. None of them offer nectar reward but many are fragrant. I am not aware of any studies showing what pollinates them, but based on pollination syndromes they appear to be pollinated by unspecialised insects. Some taxa may be ambophilous in suitable habitats. All genera of Oleinae are characterised by drupes, a synapomorphic fruit type. The genus Fraxinus is unique in having a one-seeded samara.

Evolution of anemophily in Oleaceae – exaptations and adaptations Anemophily is a derived condition within the angiosperms and has evolved independently in several families (Whitehead 1968, Cox 1991). It is correlated with a number of environmental and morphological characters, such as temperate climate, deciduousness, unisexual flowers (monoecy or dioecy), smooth pollen and high pollen-to-ovule (P/O) ratio. Despite that there is nowadays no doubt that anemophily is derived from zoophily (animal pollination) in angiosperms (Cox 1991), it is unclear which characters generally preceded the transition to anemophily, which were coincidental or which evolved later as a consequence of that system (Linder 1998). The first type of trait has been termed exaptation (preadaptation) by Gould and Vrba (1982) and exapted features are interpreted as those characters that facilitate or are a prerequisite for a transition. Traits that follow in evolutionary sequence have evolved in response to anemophily and are adaptive for increasing the efficiency and function of that system. Particularly, in this regard, it is unclear whether dioecy, which is strongly correlated with anemophily, evolves prior to or after the shift to anemophily (Charlesworth 1993). Anemophily has rarely been experimentally verified in species where it is assumed to occur (e.g. as by Bernardello et al. 1999, Goodwillie 1999, Otegui and Cocucci 1999, de Figueiredo and Sazima 2000). Instead anemophily has been inferred on the basis of a number of traits con-

sidered to be typical of anemophilous taxa, the so-called anemophilous syndrome (Whitehead 1968 and 1983, Faegri and van der Pijl 1979). The most obvious traits in plants adapted to pollination by wind are flowers without (or with a very reduced) corolla, nectar or scent. The flowers are often dichogamous and/or unisexual and borne either on the same (monoecy) or different individuals (dioecy). The anthers are usually large and well exposed on long slender filaments or in hanging catkins that can easily be moved by wind. The stigmata and/ or surrounding structures often have a morphology favouring efficient pollen trapping (Niklas 1985). Many are trees, but there are also shrubs and non-woody plants, e.g. the whole of Poaceae, that are wind-pollinated. The woody species usually flower earlier than other plants, when they and/or the surrounding plants have no leaves, i.e., many are deciduous and occur in deciduous forests or open environments (Dowding 1987). Anemophily is especially common among temperate trees and there is a striking increase with latitude and decrease with species diversity in percentages of anemophilous trees (Regal 1982). Anemophily in habitats like tropical rainforests are rare (Bawa et al. 1985, Linskens 1996, Renner and Feil 1993, but see Williams and Adam 1999). High pollen/ovule (P/O) ratios are also a feature of wind-pollinated plants (Cruden 1977, Tormo et al. 1996). A number of plants assumed to be entomophilous may in fact be ambophilous, i.e., pollinated by both insects and wind, and ambophily is common among neotropical dioecious trees (Bullock 1994). Ambophilous species are characterised by numerous simple flowers with small and open perianth (Linder 1998). In this study, anemophily has been inferred on the basis of possession of apetalous flowers in combination with a suitable habitat, such as open coastal habitats (Nestegis), seasonally deciduous and/or open habitats (the two species of Priogymnanthus, some species of Forestiera), temperate forests (many species of Fraxinus), and/or evidence for airborne pollen (e.g. Olea europaea, Phillyrea, and species of Fraxinus). The flowers of subtribe Oleinae are generally small (smaller than the ones in other taxa of the family), whitish to yellowish-green and apparently visited by unspecialised insects. Many show evidence of being ambophilous, e.g. Fraxinus section Ornus and African species

of Olea (Dyer 1991) and, indeed, Olea europaea is the main source of airborne Oleaceae pollen in the Mediterranean area (Liccardi et al. 1996). There is a trend from sympetalous and longtubed flowers to short-tubed and more open flowers, or flowers with free or almost free petals. Many of the anemophilous taxa have no corollas at all, e.g. in Forestiera, Fraxinus and Nestegis, and in some species of Fraxinus and Forestiera not even calyces. Thus, an exaptation for anemophily were the small whitish flowers and the trend in reduction of corolla size is viewed as an adaptation for enhanced efficiency of wind-pollination. In connection with the changes in floral features, there is a general trend in shift of pollinator reward, from nectar to pollen. The pollen-rewarding flowers were probably also an exaptation for ambophily, which in turn was an exaptation for anemophily. The shift from pollen to nectar reward was probably selected for in habitats where generalist pollinators were more common or reliable than specialised and nectar-feeding pollinators. Since there are wind-pollinated taxa that possess small corollas (e.g. Olea europaea and Phillyrea), loss of petals may have appeared after anemophily in some cases and can be viewed as an adaptation for more effective wind dispersal and reception of pollen. Unisexual flowers are also interpreted as an adaptation to anemophily and the significance of this will be discussed further below. These trends in floral evolution in Oleaceae fit nicely with general trends and patterns found among angiosperm families (Linder 1998). In Linder’s (1998) model of the evolution of anemophily, dry pollen and simple, open perianths are structures or innovations that have to be in place before wind-pollination can evolve. The other characteristics of the anemophilous syndrome evolve later as adaptations to the system. Those adaptations are, e.g., smooth pollen, reduction in perianth, reduction in the number of ovules, and dioecy. Wind-pollination is seen as a specialisation from a basic ambophilous system and has most often evolved in lineages where numerous small and open flowers are common features. In no case has anemophily evolved in taxa with zygomorphic flowers, but always in taxa with actinomorphic flowers (Linder 1998).

11

Incidence of dimorphic breeding systems in Oleaceae In Oleaceae, about 100 of 600 species have a breeding system that is not hermaphroditism (Table 2). There are two dimorphic breeding systems present, dioecy and androdioecy, as well as several polygamous (Fraxinus) or andromonoecious taxa (Olea). The high incidence of the rare androdioecious breeding system in Oleaceae is remarkable. As can be seen in table 2, I have found 37 taxa (in four genera) that exhibit morphological androdioecy, including 10 species in the genus Fraxinus. Six genera have dioecious species, Fraxinus, Forestiera, Nestegis, Olea and Osmanthus (section Leiolea) and Chionanthus. Four species of Nestegis and the 14 dioecious species of Fraxinus are wind-pollinated trees, and Forestiera consists mainly of wind-pollinated shrubs. The species of Osmanthus section Leiolea and Olea are mostly large trees and the flowers are small and inconspicuous. It is unknown how they are pollinated, but the syndrome certainly points towards ambophily or anemophily. Why does dioecy and androdioecy appear to be adaptations to anemophily and ambophily? In order to find some answers to this question, I will review some theories and models for the evolution of dioecy.

Dioecy – why is it so common among wind-pollinated plants? Dioecy occurs in about 6 % of flowering plants (Renner and Ricklefs 1995) and has been correlated with a number of characters such as fleshy fruits, woodiness, climbing habit, small and white to yellowish or greenish flowers, unspecialised insect pollinators, wind-pollination, etc. (reviewed by Renner and Ricklefs 1995). Although the great majority of dioecious taxa are entomophilous, in temperate regions many dioecious species are anemophilous (Bawa 1980). The evolution of dioecy from hermaphroditism in plants has generated much debate and a number of hypotheses have been put forward to explain the above associations (Darwin 1877, Bawa and Opler 1974, Charlesworth and Charlesworth 1978, Bawa 1980, Lloyd 1980 and 1982, Bawa and Beach 1981, Thomson and Barrett 1981, Givnish 1980 and 1982, Ross 1982, Charlesworth 1984, Fox 1985, Thomson and Brunet 1990, Charlesworth 1993, Renner and 12

Ricklefs 1995, Sakai et al. 1995, Freeman et al. 1997, Richards 1997, Charlesworth 1999). Organisms with separate sexes have no problem with selfing and inbreeding depression (at least not in large populations). In cosexual plants, on the other hand, which may suffer from inbreeding depression after selfing, there is a substantial selective pressure as to promote outcrossing. Ever since Darwin’s (1877) comprehensive work, the existence of a variety of breeding systems in plants have traditionally been viewed mainly as consequences of selection for outbreeding (see discussion in Bawa 1980, Givnish 1982). Givnish (1982) identifies two main groups of models for the evolution of dioecy: selection for outcrossing and influences of ecological factors. The ecological mechanisms invoked are quite diverse and include, for example, sexual selection, division of labour leading to optimal resource allocation, decreased intraspecific competition, pollinator attraction to massive pollen crops and frugivore attraction to massive fruit crops. Bawa (1980) and Bawa and Beach (1981) emphasise that viewing plant breeding systems as the result of regulation of genetic recombination is unlikely to account fully for the evolution of breeding systems, and that the key to understanding them lies in considering patterns of sexuality as means of optimising male and female reproductive success in different ways within the constraints imposed by the pollination system. Dioecy should be considered as more than a simple mechanism to promote outcrossing (Bawa 1980). Before attempting to explain the evolution of dioecy in Oleaceae (and other wind-pollinated plants), I would like to turn the question around and ask: Why do most plants have only cosexual flowers? Among sexual organisms that can move around and find mates without the aid of other organisms, the dominating sexual condition is unisexuality, i.e. separate male and female individuals (Richards 1997). However, among angiosperms the most common breeding system is hermaphroditism (e.g. Lloyd 1982, Richards 1997). Why is this? The obvious explanation is their dependence on external pollinators, visiting their flowers for food and in the process transferring the male gametes of the plants to other conspecific individuals. This fact imposes a very strong selective pressure for combined sexes in zoophilous plants, partly because costly floral displays and

pollinator reward can be shared between the two sexes (Lloyd 1982), and partly because one pollinator can then perform two services in one visit, both delivering and picking up pollen (Richards 1997). Another advantage is that a cosexual and self-compatible plant may selfpollinate when growing at low population density or where pollinators are absent (Lloyd 1982). The same is true if a single individual colonises a new area after successful long-distance dispersal (Baker’s law; Baker 1967). Bawa and Beach (1981) have argued that the evolution of sexual systems is constrained by the way pollinators interact with flowers. Although there are advantages of having cosexual flowers, such as shared costs of floral display, increased chances of both pollen dispersal and reception, and reproductive assurance, there are also disadvantages. If the pollen is presented at the same time as the receptive stigma in a flower (no dichogamy) and it is capable of selffertilisation (no incompatibilty mechanism), this may lead to inferior progeny if there is inbreeding depression. Therefore, while maintaining the advantageous cosexual flowers in zoophilous species, selection has favoured other mechanisms that promote outcrossing, e.g. dichogamy, herkogamy, and pre-zygotic or post-zygotic self-incompatibility mechanisms. So, what happens when a plant is no longer dependent on animal pollen vectors? The abiotic vectors, wind and water, is ubiquitous (in most habitats) and there is no longer any need for the plant to maximise both sexual functions for one visit. Neither do they have costly displays that it would be advantageous to share. On the contrary, the spatial interference between the male and female structures in a windpollinated flower may be directly disadvantageous and lead to conflicts between optimal male and female function (Faegri and van der Pijl 1979, Lloyd 1982). There are different structural requirements for optimal pollen dispersal and pollen capture in wind-pollinated flowers (Frankel and Galun 1977, Niklas 1985, Freeman et al. 1997). Males do better with anthers on long filaments, exposed to the wind in lax inflorescences, where the flowers may be tightly clustered. Females, on the other hand, catch pollen more efficiently on their stigmata if the individual flowers are a bit separated in more rigid inflorescences, situated on less exposed parts of the branches. Furthermore, because wind-pollination is a rather inefficient (in the

sense that it is wasteful or imprecise) means for plants to spread their male gametes, there is a strong selection pressure for maximising pollen production, which increases the likelihood for pollen to reach conspecific stigmas (Frankel and Galun 1977). Consequently, those plants that divert more resources to pollen production, either through more pollen per anther, more anthers per flower and/or more flowers per individual, may have higher paternal success. In fact, in several wind-pollinated species of Fraxinus it has been shown that male inflorescences have significantly more flowers than hermaphrodite and female ones (Wallander and Dahl ms.). This is in accordance with the emphasised importance of male function in selection for inflorescence size (Queller 1983) and male plants in dioecious species are known to bear more flowers than female plants (Bawa 1980). Increased pollen production, in terms of more flowers, may be in conflict with female function, where resource limitation may impose a limit on the optimum number of fruit-producing flowers to carry (Lloyd and Bawa 1984). It is thus obviously impossible for one flower to optimise at the same time pollen dispersal and pollen capture. Consequently, anemophilous plants may respond to the selection for separate sexes due to the different demands on optimal male and female function in a flower, since they are no longer affected by selection pressures for optimising floral displays and reward and maximising the returns from insect visits to bisexual flowers. Thus, selection for sexual specialisation leads to a dioecy through total or partial reduction of the opposite sex in the resulting unisexual flowers. Darwin (1877) recognised that there is division of labour in unisexual plants and suggested that this is possibly a factor in the evolution of dioecy. Another factor is sexual selection. The basis of intrasexual competition is the numerical inequality of the two types of gametes, which itself is a result of greater investment per sex cell by females than by males (Bateman cited in Bawa 1980). The reproductive success of males is therefore limited by their access to the female gametes, while that of the females is limited by the resources available for egg production and parental care of the offspring. Thus males tend to optimise the quantity of matings while females tend to optimise the quality (Janzen 1977). Willson (1979) implicated the role of sexual selection in the evolu13

tion of dioecy and the model below is also based on male competition for mates. Male and female allocations are expected to be equal or nearly so (Charnov 1979). However, measurements of allocations in wind-pollinated plants have typically differed from these expectations (cases cited by Burd and Allen 1988). It has been suggested that paternal fitness in wind-pollinated plants might be linearly related to male reproductive investment, because wind will not be saturated as a pollen vector even at high levels of pollen production (Charnov 1979, Charlesworth and Charlesworth 1981). In contrast, Burd and Allen (1988) have proposed that the allocation patterns found in anemophilous species can be understood in the context of current theory if the relation between male fitness and male investment is not linear but, rather, follows a pattern of saturation or diminishing marginal returns (Charnov 1979). They suggest that the rate of saturation is determined largely by the height of a plant, because pollen dispersal distances (and thus access to mates) are dependent on the height of the pollen emitting source. A taller individual should experience reduced local mating competition and have a less saturating male-fitness curve due to wider dispersal of its pollen. Thus, contrary to the traditional expectation of greater female effort in larger plants (Lloyd and Bawa 1984), Burd and Allen (1988) expect larger wind-pollinated plants to have a relatively greater male investment. In addition, models of plant mating systems (e.g. Lloyd and Bawa 1984) show that male allocation should increase as the potential mating-group size increases (based on the arguments of sexual selection). Selfing, on the other hand, is expected to shift allocation in favour of female function, because the reduced pool of outcrossing ovules restricts potential paternal reproductive success (Lemen 1980, Charlesworth and Charlesworth 1981, Schoen 1982). An associated explanation for gender specialisation involves fruit type. The correlation between dioecy and fleshy fruits is strong (Bawa 1980, Renner and Ricklefs 1995). In general, many dioecious species are tropical trees with fleshy, one- or few-seeded fruits. These fruits or seeds are generally distributed by birds, which are common in the tropics and effective as long-distance dispersers. The high incidence of dioecy in the tropics and on islands may thus be explained in part by the correla14

tion between dioecy and fruit dispersal by birds (Bawa 1980, Givnish 1980). In Oleaceae, many genera have fleshy and bird-dispersed fruits, e.g. Ligustrum and all genera of subtribe Oleinae. Fraxinus have wind-dispersed samaras. How does fruit type and dispersal influence gender specialisation? If the range of seed dispersal increases, the degree of local resource competition among sibling seedlings will decrease, favouring investment in female function. Wind dispersal of fruits or seeds, such as the samaras of Fraxinus, may be enhanced by tree height for reasons similar to those affecting pollen dispersal and lead to selection for increased female allocation (Burd and Allen 1988). Frugivore dispersal of large crops of fleshy fruits, such as the drupes of Oleinae, could create disproportionate gains for female reproduction among those plants capable of more massive fruit set (Burd and Allen 1988). The conflicts between selection for increased male allocation in outcrossing wind-pollinated plants and selection for increased female allocation in plants with effective long-distance fruit dispersal (by wind or birds), select for separation of sexes. Thus, these opposing selective pressures on sexual allocation will lead to males specialising in production and dispersal of pollen, and females specialising in fruit production. If these selection pressures are strong enough, they will override the selection for combined sexes in animal-pollinated plants. According to Givnish (1980), dioecy should predominate only in those wind-pollinated taxa that have fleshy fruits dispersed by animals. In a study on the breeding systems of shrubs in the Mediterranean region, Aronne and Wilcock (1994) noted that the dioecious taxa were smallflowered, fleshy-fruited, few- to one-seeded and vegetatively spreading, whereas hermaphrodite taxa were large-flowered, dry-fruited, many-seeded and generally non-vegetatively spreading. Many of the dioecious species were also anemophilous (e.g. Phillyrea). As stated above, the general explanation for the presence of dioecy in plants has been selection for outcrossing. It is evident that dioecy confer obligate outcrossing, but the question is if it has been the main selective force? The presence of the monoecious breeding system among many wind-pollinated plants, e.g. in Alnus, Betula, Corylus, and Quercus, show that selection for outcrossing is not the main factor, since monoecy may still permit selfing (Bertin 1993).

Many of these plants are dichogamous, which reduce the risk of selfing. Protandry is more common than protogyny, even though protogyny is a more effective way to prevent selfpollination in that it gives foreign pollen a head start over self-pollen (Faegri and van der Pijl 1979). In Fraxinus, all cosexual individuals are protogynous and not only does protogyny promote outcrossing, it also prevents self-pollen from clogging the stigma. It is therefore my suggestion that selection for sexual specialisation is the main factor that has repeatedly led to the evolution of unisexual flowers in Oleaceae, as well as in other windpollinated plants. Even when outcrossing is advantageous, is not the main selective force and has merely conferred a beneficial side effect. Sexual dimorphism is probably the result of independent selection on the two sexes separately, rather than the result of selection specifically favouring sexual dimorphism (Meagher 1984).

Evolution of androdioecy Androdioecy is considered to be a very rare breeding system in flowering plants (Charlesworth 1984). Theoretical ESS (evolutionary stable strategy) models developed by Lloyd (1975) and Charlesworth (1984) show that androdioecy can not evolve in a selfing species as a means to avoid inbreeding depression. However, in an outcrossing population males can be established if they have more than twice the pollen fertility (Charnov et al. 1976, Charlesworth and Charlesworth 1978). The fitness necessary for males to establish in a hermaphrodite population increases as the selfing rate increases. Because of these high demands on male fitness, Charlesworth (1984) believes androdioecy is very unlikely to evolve. She reviews a number of claimed cases of androdioecy and dismisses the nectarless species as being cryptically dioecious. The hermaphrodites are in these cases viewed as functionally female with indehiscent anthers or dysfunctional pollen that is thought to function only in pollinator attraction (Mayer and Charlesworth 1991). So, these are the theoretical conditions for the evolution of androdioecy. Males need to be at least twice as fit compared to the hermaphrodites in order for them to be maintained in a completely outcrossing population, and even more fit if there is partial selfing, and it is there-

fore not strange that there are so few documented cases. The listing by Yampolsky and Yampolsky (1922) is now outdated (Bawa 1980, Renner and Ricklefs 1995). The older work includes extremely few cases of androdioecy, however, none of them in Oleaceae and none of the other cases that have been studied since then (Table 3). There are probably more androdioecious taxa to discover, if one knows where to look, and I predict that new cases will be found in ambophilous taxa (as will be explained below). Phenotypic androdioecy occurs in ten species of Fraxinus section Ornus, 24 species of Osmanthus, both species of Phillyrea, and one species of Chionanthus (Table 2). Few of these taxa have been studied in the field, but up to date at least three of them have been shown to be functionally androdioecious; Fraxinus ornus (Dommée et al. 1999, Wallander and Dahl ms.), Fraxinus lanuginosa (Ishida and Hiura 1998), and Phillyrea angustifolia (Lepart and Dommée 1992, Aronne and Wilcock 1992, Vassiliadis 1999, Pannell and Ojeda 2000). Preliminary studies of the wind-pollinated F. longicuspis suggest that it is functionally androdioecious as well (Wallander ms.). In addition, three other species have lately been confirmed to be functionally androdioecious (Table 3). The feature in common for all but one of these species is the anemophilous or ambophilous pollination system. The differences, on the other hand, are that the nonoleaceous species are herbs, and male frequencies in these populations are around 30% or less, as predicted by the models of Lloyd (1975) and Charlesworth and Charlesworth (1978), or sometimes zero. Moreover, they are colonising ruderals and conform nicely to patterns predicted by the metapopulation model by Pannell (2000). However, these models cannot explain the coexistence of equal proportions of males and hermaphrodites in androdioecious species and Pannell (2000) suggest that different processes are driving the systems in Oleaceae. A more than doubled pollen fertility in males may seem implausible (Charlesworth 1984). However, such a dramatic increase in pollen production may be possible if pollination is effected by pollen collecting insects (Bawa and Beach 1981) and, as I suggest based on Burd and Allen’s (1988) model outlined above, when the unsaturated wind is the pollination agent. Darwin (1877) also suggested that an increased pollen supply in some plants might be benefi15

References

habit high outcrossing rate strong inbreeding depression higher male pollen production higher male pollen performance protogynous earlier flowering males wind-pollinated colonising habit male frequency sex determination evolution from…

Feature

√

√ √ √ √ 17-24% ≥2 loci dioecy?

Liston et al. 1990, Fritsch & Rieseberg 1992, Rieseberg et al. 1993, Philbrick & Rieseberg 1994, Spencer & Rieseberg 1995, Wolf et al. 1997, Swensen et al. 1998

ca 4-10x

ca 3x

Pannell 1997 a, b, c

√ √ ca. 30% 1 locus dioecy

no

no

Akimoto et al. 1999

dioecy

no √ 5-28%

>2x flowers

annual herb

Schizopepon bryoniaefolius

annual herb

Mercurialis annua

perennial herb 65-92% √

Datisca glomerata

no

√ √ ? √ both √ √ ca. 50% ca. 50% genetic hermaphrohermaphroditism ditism Lepart & Dom- Dommée et al. 1999, mée 1992, Aronne & Wil- Wallander & Dahl cock 1992, (ms.) Vassiliadis 1999, Pannell & Ojeda 2000, Vassiliadis et al. 2000

no

√

√

no

tree

Fraxinus ornus

no

shrub

Phillyrea angustifolia

Table 3. Comparison of features of functionally androdioecious species. √ means trait present.

hermaphroditism Hiura & Ishida 1994, Ishida & Hiura 1994, Ishida & Hiura 1998

√ no both √ 14-66%

√

2.6x

tree √ √

Fraxinus lanuginosa

hermaphroditism Wallander & Dahl (ms.)

ca. 50%

√ √ √

?

2.2x

tree

Fraxinus longicuspis

cial when there was a change in pollinators or when the species was becoming wind-pollinated. Are there any indications that males in these androdioecious species have the necessary increased male fitness? In F. lanuginosa there are 2.6 more flowers per inflorescence in males (Hiura and Ishida 1994) and in F. longicuspis 2.2 more flowers (Wallander and Dahl ms.) (Table 3). The former species also shows a significantly higher pollen germination rate of male versus hermaphrodite pollen (Ishida and Hiura 1998) and the latter 40% larger male pollen grains (Wallander and Dahl ms.). In all other variables there are no significant differences between males and hermaphrodites. In Phillyrea, males also produce more flowers (Aronne and Wilcock 1994). Thus, males in all these cases have a more than doubled pollen production compared to hermaphrodites. The critical question is then: do they also sire more than twice the amount of available ovules? Yes, I suggest they do, because the hermaphrodites are protogynous (shown in several species of Fraxinus by Wallander and Dahl [ms.]), which in combination with the higher germination rate and fertilisation success of male pollen in competition with hermaphrodite pollen, may contribute to the high demands on male fertility. None of the models (Lloyd 1975, Charlesworth and Charlesworth 1978, Charlesworth 1984) has incorporated any of the ecological selection pressures that have been mentioned above for the evolution of unisexual flowers. Genetic models that predict the rate and conditions for the spread of unisexual mutants in response to selective pressures for outcrossing must take into account how the dynamics of sexual specialisation, pollination system, mode of dispersal, and differential predation influence the fitness of the mutants (Bawa 1980). When this is taken into consideration, males may be selected for on less stringent conditions because of the advantages of being a specialist of pollen production and dispersal in ambophilous or anemophilous species. Because hermaphrodites, even if they are self-compatible, are protogynous, outcrossing is ensured and males (and other hermaphrodites) will have access to most of the ovules (Wallander ms.). The possible explanations for maintenance of androdioecy in partially or fully anemophilous species are much the same as for dioecy, with increased male investment selected for in

large outcrossing populations. Why then do the hermaphrodite flowers in androdioecious ‘female’ plants retain their stamens? There are two basic explanations for this: either the stamens are selected for, or they have not been selected against. A simple argument for the latter explanation is that since all of the investigated species are protogynous, stigma clogging and/ or self-pollination is not a problem. On the contrary, since at least some of them are self-compatible (F. ornus and F. lanuginosa), they may self-pollinate if cross-pollination fails. This may be one factor favouring retention of stamens as a means of reproductive assurance. The selfcompatible hermaphrodites may also be favoured during episodes of colonisation (Pannell 1997d). Another selection pressure for keeping the stamens, in these nectarless speices, may be pollinator attraction (Charlesworth 1984). But more importantly, if hermaphrodites actually do contribute to part of the progeny in the population, this may be a strong enough selection pressure for keeping the male function, even if it is inferior compared to the specialised males. Most reports claiming functional androdioecy have demonstrated the superior function of pollen from males over pollen from hermaphrodites in siring success. This has been interpreted as the hermaphrodites being in fact functional females, i.e. cases of cryptic dioecy (Mayer and Charlesworth 1991), even though the pollen of hermaphrodites may be fertile. However, since the requirements for male fitness are so high, compared to hermaphrodites, by definition it automatically renders the hermaphrodites a poorer male function, and consequently they are said to be functionally female. Still, if hermaphrodite pollen is fertile, even though it may have low success in competition with male pollen (Wallander ms.), it will potentially have a function if there is no competition. Thus, if the hermaphrodites have a potential to function as males, I argue that the coexistence of such hermaphrodites and males (with higher male fertility) should be regarded as functional androdioecy, regardless of the sex ratio.

Andromonoecy It is particularly interesting that some species of Olea are andromonoecious (Brooks 1948, Uriu 1959, Dyer 1991). The advantage of this 17

system over hermaphroditism in anemophilous or ambophilous species is suggested to lie in a more efficient resource allocation (Bertin 1982). The production of male flowers in place of some hermaphrodite flowers does not lower the total female fitness, because fruit production in these species is not limited by the number of functional pistils. Instead, it may increase male fitness through increased donation of pollen to other plants. In other words, if females cannot mature the fruit of all flowers, resources that would have been wasted on those flowers will contribute more to reproductive success if they are placed elsewhere. Andromonoecy thus appears to be one evolutionary response to fruits that are large relative to the size of the flowers, and whose production is limited by factors other than the number of bisexual flowers per plant. The high male flower production appears to be the response to selection for pollinator attraction (in ambophilous plants) and/or selection for increased likelihood of fertilising ovules on other plants (Bertin 1982). If the two sexual functions in a bisexual flower interfere with other, as in wind-pollinated plants, monoecy may replace andromonoecy (Bertin 1982). These hypotheses are based on the same factors as suggested above for dioecy and androdioecy. Andromonoecy and androdioecy are obviously not the result of selection for outcrossing. Therefore, even for dioecy, I suggest that the primary cause for the presence of these breeding systems in ambophilous or anemophilous taxa is not selection for outcrossing, but instead the above described ecological factors.

Anemophily and dimorphism in other taxa There are very few studies on the evolution of dimorphic breeding systems in anemophilous plants. In Acer (Sapindaceae), dioecy is believed to have evolved before anemophily (de Jong 1976), although Hesse (1979) does not draw this conclusion. In Thalictrum Ranunculaceae), dioecy is thought to precede anemophily (Kaplan and Mulcahy 1971). Neither of these studies are based on phylogenetic data and the reasons for postulating that dioecy preceded anemophily do not hold. Kaplan and Mulcahy (1971) believe that dioecy evolved first as a result of selection for outcrossing and when females did not receive enough pollinators (be18

cause they did not have the attractive pollen source), anemophily was selected for. Their conclusion is that in taxa that are originally offering pollen as the insect attractant, wind-pollination appears to be a response to dioecy, and dioecy itself being the response to selective forces favouring greater degree of outbreeding. This is contrary to the general pattern which argues for environmental factors selecting for anemophily, rather than the morphology of the plants (Linder 1998). I also argue that natural selection can not have favoured a state where females do not get cross-pollinated, unless they are regularly selfing, in which case anemophily would not be selected for. Linder (1998) mapped pollination and breeding systems and other traits on a molecular phylogeny of angiosperm families. The results showed that dioecy have evolved independently in several lineages, following anemophily, e.g. within the poalean group. The study by Weller et al. (1998) was based on a phylogeny and showed a strong correlation between anemophily and dimorphism in Schiedea (Caryophyllaceae) in Hawaii, but the conclusions are unclear as to which evolved first.

General trends and conclusions The main conclusion of this study is that androdioecy and dioecy are adaptations to ambophily and anemophily, respectively, and these dimorphic breeding systems with unisexual males are mainly the result of selection for male specialisation in pollen production and dispersal. The presence of ambophily and/or androdioecy may be an exaptation for the evolution of anemophily and ultimately dioecy. The presence of more or less functional stamens in hermaphrodite flowers is regarded as the result of selection for either pollinator attraction in entomophilous or ambophilous species, or reproductive assurance, or both. It may also be that they lack a function but have not been selected against (yet). Some trends in the evolution of traits related to anemophily can be discerned in Oleaceae. • Nectar reward --> pollen reward • Reduction in flower ‘showiness’ and size, from conspicuous, white or yellow, fragrant and nectariferous flowers --> small and inconspicuous, whitish-green flowers without fragrance and nectar • Small whitish flowers attracting unspecial-

ised pollen-feeding insects, plus suitable habitat --> ambophily and/or androdioecy • Ambophily and/or androdioecy --> anemophily --> dioecy Anemophily has been suggested to be an abiotic escape in habitats where insects are unpredictable but where wind is more predictable and accessible (Cox 1991). Anemophily may also be favoured by short growing seasons, which favours early flowering before animal vectors become common (Regal 1982). In habitats with low species diversity, such as temperate forests, selection for biotic pollination syndromes may be relaxed because of the higher probability for wind-dispersed pollen to hit a conspecific stigma than for pollen dispersed by insects. These may be some reasons why anemophily have evolved on so many occasions in Oleaceae. Since many taxa of the Oleinae and Fraxinus are distributed in subtropical and temperate parts of the world, where these preconditions may be common, especially at higher altitudes, the shift toward abiotic pollination may have arisen more easily.

ACKNOWLEDGEMENTS I thank Prof. emeritus Peter S. Green at the Royal Botanic Gardens at Kew for valuable help, advice and discussions. I thank Lennart Andersson, Åslög Dahl, and Claes Gustafsson for critical comments on this manuscript. Funding for this doctoral research have come from The Royal Swedish Academy of Sciences, Stiftelsen Botaniskas Vänner, Helge Ax:son Johnsons stiftelse, and Collianders stiftelse.

REFERENCES Akimoto, J. & Fukuhara, T. & Kikuzawa, K. 1999. Sex ratios and genetic variation in a functionally androdioecious species, Schizopepon bryoniaefolius (Cucurbitaceae). American Journal of Botany 86: 880-886. Altamura, L. & Altamura, M. M. & Mazzolani, G. 1987. Elements for the revision of the genus Olea (Tourn.) L. VII. The taxa present in Asia which can be ascribed to Olea and allied genera. Annali di Botanica 45: 119-134. Aronne, G. & Wilcock, C. C. 1992. Breeding system of Phillyrea latifolia L. and Phillyrea angustifolia L: evidence for androdioecy. Giornale Botanico Italiano 126: 263.

Aronne, G. & Wilcock, C. C. 1994. Reproductive characteristics and breeding system of shrubs of the Mediterranean region. Functional Ecology 8: 6976. Baker, H. G. 1967. Support for Baker’s law – as a rule. Evolution 21: 853-856. Bawa, K. S. 1980. Evolution of dioecy in flowering plants. Annual Review of Ecology and Systematics 11: 15-39. Bawa, K. S. & Beach, J. H. 1981. Evolution of sexual systems in flowering plants. Annals of Missouri Botanical Garden 68: 254-274. Bawa, K. S. & Bullock, S. H. & Perry, D. R. & Coville, R. E. & Grayum, M. H. 1985. Reproductive biology of tropical lowland rain forest trees. II. Pollination systems. American Journal of Botany 72: 346-356. Bawa, K. S. & Opler, P. A. 1974. Dioecism in tropical forest trees. Evolution 29: 167-179. Bernardello, G. & Anderson, G. J. & Lopez, S. P. & Cleland, M. A. & Stuessy, T. F. & Crawford, D. J. 1999. Reproductive biology of Lactoris fernandeziana (Lactoridaceae). American Journal of Botany 86: 829-840. Bertin, R. I. 1982. The evolution and maintenance of andromonoecy. Evolutionary Theory 6: 25-32. Bertin, R. I. 1993. Incidence of monoecy and dichogamy in relation to self-fertilization in angiosperms. American Journal of Botany 80: 557-560. Brandis, D. 1906. Oleaceae. In: Indian Trees - an account of the trees, shrubs, bamboos and palms indigenous or commonly cultivated in the British Indian Empire, pp. 442-452. Archibald Constable & Co. Ltd., London. Brooks, C. J. 1977. A revision of the genus Forestiera (Oleaceae). Dissertation, University of Alabama, USA. Brooks, R. M. 1948. Seasonal incidence of perfect and staminate olive flowers. Proceedings of the American Society for Horticultural Science 52: 213-218. Bullock, S. H. 1994. Wind pollination of neotropical dioecious trees. Biotropica 26: 172-179. Burd, M. & Allen, T. F. H. 1988. Sexual allocation strategy in wind-pollinated plants. Evolution 42: 403407. Chang, M. & Qiu, L. & Green, P. S. 1996. Olea L. In: Wu, Z. & Raven, P. H. (eds). Flora of China 15: 295-299. Science Press & Missouri Botanical Garden. Charlesworth, B. & Charlesworth, D. 1978. A model for the evolution of dioecy and gynodioecy. American Naturalist 112: 975-997. Charlesworth, D. & Charlesworth, B. 1981. Allocations of resources to male and female function in hermaphrodites. Biological Journal of the Linnean Society 15: 57-74. Charlesworth, D. 1984. Androdioecy and the evolution of dioecy. Biological Journal of the Linnean Society 22: 333-348. Charlesworth, D. 1993. Why are unisexual flowers associated with wind pollination and unspeciali-

19

zed pollinators? American Naturalist 141: 481-490. Charlesworth, D. 1999. Theories on the evolution of dioecy. In: Geber, M. A., Dawson, T. E. and Delph, L. F. (eds.) Gender and sexual dimorphism in flowering plants, pp. 33-60. Springer-Verlag, Berlin Heidelberg, Germany. Charnov, E. L. & Maynard Smith, J. & Bull, J. J. 1976. Why be an hermaphrodite? Nature 263: 125 -126. Charnov, E. L. 1979. Simultaneous hermaphroditism and sexual selection. Proceedings of the National Academy of Science, USA 76: 2480-2484. Cox, P. A. 1991. Abiotic pollination: an evolutionary escape for animal-pollinated angiosperms. Philosophical Transactions of the Royal Society London B 333: 217-224. Cruden, R. W. 1977. Pollen-ovule ratios: a conservative indicator of breeding systems in flowering plants. Evolution 31: 32-46. Darwin, C. 1877. The different forms of flowers on plants of the same species. Appleton, New York. Dommée, B. & Geslot, A. & Thompson, J. D. & Reille, M. & Denelle, N. 1999. Androdioecy in the entomophilous tree Fraxinus ornus (Oleaceae). New Phytologist 143: 419-426. Dowding, P. 1987 Wind pollination mechanisms and aerobiology. International Review of Cytology 107: 421-437. Dyer, C. 1991. A biosystematic study of the African species of Olea L. (Oleaceae). Dissertation, University of the Witwatersrand, Johannesburg, South Africa. Faegri, K. & van der Pijl, L. 1979. Abiotic pollination. In: The principles of pollination ecology, pp. 3440. Pergamon Press. de Figueiredo, R. A. & Sazima, M. 2000. Pollination biology of Piperaceae species in southeastern Brazil. Annals of Botany 85: 455-460. Fox, J. F. 1985. Incidence of dioecy in relation to growth form, pollination and dispersal. Oecologia 67: 244-249. Frankel, R. & Galun, E. 1977. Pollination mechanisms, reproduction and plant breeding. In: Monographs on theoretical and applied genetics 2. SpringerVerlag. Freeman, D. C. & Lovett Doust, J. & El-Keblawy, A. & Migla, K. J. & McArthur, E. D. 1997. Sexual specialization and inbreeding avoidance in the evolution of dioecy. The Botanical Review 63: 65-92. Fritsch, P. & Rieseberg, L. H. 1992. High outcrossing rates maintain male and hermaphrodite individuals in populations of flowering plant Datisca glomerata. Nature 359: 633-636. Givnish, T. J. 1980. Ecological constraints on the evolution of breeding systems in seed plants: dioecy and dispersal in gymnosperms. Evolution 34: 959972. Givnish, T. J. 1982. Outcrossing versus ecological constraints in the evolution of dioecy. American Naturalist 119: 849-865. Goodwillie, C. 1999. Wind pollination and reproductive assurance in Linanthus parviflorus (Polemonia-

20

ceae), a self-incompatible annual. American Journal of Botany 86: 948-954. Gould, S. J. and Vrba, E. S. 1982. Exaptation - a missing term in the science of form. Paleobiology 8: 415. Green, P. S. 1958. A monographic revision of Osmanthus in Asia and America. Notes from the Royal Botanic Garden Edinburgh 22: 439-542. Green, P. S. 1963a. A revision of the New Caledonian species of Osmanthus. Journal of Arnold Arboretum 44: 268-283. Green, P. S. 1963b. The genus Nestegis from New Zealand. Journal of Arnold Arboretum 44: 377-389. Green, P. S. 1968. Revision of the genus Notelaea (Oleaceae). Journal of Arnold Arboretum 49: 333369. Green, P. S. 1979. Chionanthus retusus. Botanical Magazine 182: 145-148. Green, P. S. 1994. A revision of Chionanthus (Oleaceae) in S. America and the description of Priogymnanthus, gen. nov. Kew Bulletin 49: 261-286. Harborne, J. B. & Green, P. S. 1980. A chemotaxonomic survey of flavonoids in leaves of the Oleaceae. Botanical Journal of the Linnean Society 81: 155167. Hesse, M. 1979. Ultrastructure and distribution of pollenkitt in the insect- and wind-pollinated genus Acer (Aceraceae). Plant Systematics and Evolution 131: 277-289. Hiura, T. & Ishida, K. 1994. Reproductive ecology in Fraxinus lanuginosa. I. Breeding structure. Transactions of the Hokkaido Branch of the Japanese Forest Society 42: 58-60. Ishida, K. & Hiura, T. 1994. Reproductive ecology in Fraxinus lanuginosa. II. Pollen vectors and pollen viability. Transactions of the Hokkaido Branch of the Japanese Forest Society 42: 61-63. Ishida, K. & Hiura, T. 1998. Pollen fertility and flowering phenology in an androdioecious tree, Fraxinus lanuginosa (Oleaceae), in Hokkaido, Japan. International Journal of Plant Science 159: 941-947. Janzen, D. H. 1977. A note on optimal mate selection by plants. American Naturalist 111: 365-371. de Jong, P. C. 1976. Flowering and sex expression in Acer L. - a biosystematic study Mededelingen Landbouwhogeschool Wageningen Nederland 76: 1-201. Joshi, A. C. 1942. Structure of the gynoecium in male flowers of Osmanthus suavis King. Journal of the Indian Botanical Society 21: 331-337. Kaplan, S. M. & Mulcahy, D. L. 1971. Mode of pollination and floral sexuality in Thalictrum. Evolution 25: 659-668. Kiew, R. 1979. Florae Malesianae praecursores LX. The Oleaceae of Malesia. II. The genus Olea. Blumea 25: 305-313. Lee, H-L. & Kim, K-J. 2000. Intergeneric relationships of Oleaceae based on cpDNA noncoding sequences and nuclear ITS sequences. Abstract, Recent Topics Poster, Botany 2000 meeting, Portland,

Oregon, USA. Lemen, C. 1980. Allocation of reproductive effort to the male and female strategies in wind-pollinated plants. Oecologia 45: 156-159. Lepart, J. & Dommee, B. 1992. Is Phillyrea angustifolia L. (Oleaceae) an androdioecious species? Botanical Journal of the Linnean Society 108: 375-387. Liccardi, G. & D’Mato, M. & D’Amato, G. 1996. Oleaceae pollinosis: A review. International Archives of Allergy and Immunology 111: 210-217. Linder, H. P. 1998. Morphology and the evolution of wind pollination. In: Owens, S. J & Rudall, P. J. (eds.) Reproductive Biology, pp. 123-135. Royal Botanic Gardens, Kiew. Linskens, H. F. 1996. No airborne pollen within tropical rain forests. Proceedings of the Koninklijke Nederlandse Akademie van Wetenschappen 99: 175-180. Liston, A. & Rieseberg, L. H. & Elias, T. H. 1990. Functional androdioecy in the flowering plant Datisca glomerata. Nature 343: 641-642. Lloyd, D. G. 1975. The maintenance of gynodioecy and androdioecy in Angiosperms. Genetica 45: 325-339. Lloyd, D. G. 1980. The distributions of gender in four angiosperm species illustrating two evolutionary pathways to dioecy. Evolution 34: 123-134. Lloyd, D. G. 1982. Selection of combined versus separate sexes in seed plants. American Naturalist 120: 571-585. Lloyd, D. G. & Bawa, K. J. 1984. Modification of the gender of seed plants in varying conditions. Evolutionary Biology 17 255-338. Mabberley, D. J. 1997. The plant-book. Cambridge University Press. Maddison, W. P. and Maddison, D. R. 1999. MacClade version 3.08a. Sinauer Associates, Inc. Publishers, Sunderland, MA, USA. Mayer, S. S. & Charlesworth, D. 1991. Cryptic dioecy in flowering plants. Trends in Ecology and Evolution 6: 320-325. Meagher, T. R. 1984. Sexual dimorphism and ecological differentiation of male and female plants. Annals of the Missouri Botanical Garden 71: 254-264. Metcalfe, C. R. 1938. Extra-floral nectaries on Osmanthus leaves. Kew Bulletin 0: 254-256. Niklas, K. J. 1985. The aerodynamics of wind pollination. The Botanical Review 51: 328-386. Otegui, M. & Cocucci, A. 1999. Flower morphology and biology of Myrsine laetevirens, structural and evolutionary implications of anemophily in Myrsinaceae. Nordic Journal of Botany 19: 71-85. Pannell, J. 1997a. Mixed genetic and environmental sex determination in an androdioecious population of Mercurialis annua. Heredity 78: 50-56. Pannell, J. 1997b. Variation in sex ratios and sex allocation in androdioecious Mercurialis annua. Journal of Ecology 85: 57-69. Pannell, J. 1997c. Widespread functional androdioecy in Mercurialis annua L. (Euphorbiaceae). Biological Journal of the Linnean Society 61: 95-116.

Pannell, J. 1997d. The maintenance of gynodioecy and androdioecy in a metapopulation. Evolution 51: 10-20. Pannell, J. 2000. A hypothesis for the evolution of androdioecy: the joint influence of reproductive assurance and local mate competition in a metapopulation. Evolutionary Ecology 14: 195-211. Pannell, J. R. & Ojeda, F. 2000. Patterns of flowering and sex-ratio variation in the Mediterranean shrub Phillyrea angustifolia (Oleaceae): implications for the maintenance of males with hermaphrodites. Ecology Letters 3: 495-502. Philbrick, C. T. & Rieseberg, L. H. 1994. Pollen production in the androdioecious Datisca glomerata (Datiscaceae): Implications for breeding system equilibrium. Plant Species Biology 9: 43-46. Queller, D. C. 1983. Sexual selection in a hermaphroditic plant. Nature 305: 706-707. Regal, P. J. 1982. Pollination by wind and animals: ecology of geographic patterns. Annual Review of Ecology and Systematics 13: 497-524. Rehder, A. 1904. The pseudo-monoclinism of Chionanthus virginica. Rhodora 6: 18-20. Renner, S. S. & Feil, J. P. 1993. Pollinators of tropical dioecious angiosperms. American Journal of Botany 80: 1100-1107. Renner, S. & Ricklefs, R. E. 1995. Dioecy and its correlates in the flowering plants. American Journal of Botany 82: 596-606. Richards, A. J. 1997. Plant breeding systems. Chapman & Hall. Rieseberg, L. H. & Philbrick, C. T. & Pack, P. E. & Hanson, M. A & Fritsch, P. 1993. Inbreeding depression in androdioecious populations of Datisca glomerata (Datiscaeae). American Journal of Botany 80: 757-762. Ross, M. D. 1982. Five evolutionary pathways to subdioecy. American Naturalist 119: 297-318. Sakai, A. K. & Wagner, W. L. & Ferguson, D. M. & Herbst, D. R. 1995. Biogeographical and ecological correlates of dioecy in the Hawaiian flora. Ecology 76: 2530-2543. Salmon, J. T. 1980. The native trees of New Zealand, pp. 286-291. A. H. & A. W. Reed Ltd. Wellington. Schoen, D. J. 1982. Male reproductive effort and breeding system in an hermaphroditic plant. Oecologia 53: 255-257. Spencer, S. C. & Rieseberg, L. H. 1995. The importance of flowering time and flower number in the relative fitness of males and hermaphrodites in Datisca glomerata (Datiscaceae). Plant Species Biology 10: 65-69. Swensen, S. M. & Luthi, J. N. & Rieseberg, L. H. 1998. Datiscaceae revisited: monophyly and the sequence of breeding system evolution. Systematic Botany 23: 157-169. Thomson, J. D. & Barrett, S. C. H. 1981. Selection for outcrossing, sexual selection and the evolution of dioecy in plants. American Naturalist 118: 443-449. Thomson, J. D. & Brunet, J. 1990. Hypotheses for the evolution of dioecy in seed plants. Trends in Ecol-

21

ogy and Evolution 5: 11-16. Tormo, M. R. & Munoz, R. A. & Silva, P. I. & Gallardo, L. F. 1996. Pollen production in anemophilous trees. Grana 35: 38-46. Ueda, K. 1996. Androdioecism in Chionanthus retusus (Oleaceae). Journal of Phytogeography and Taxonomy 44: 93-94. Uriu, K. 1959. Periods of pistil abortion in the development of the olive flower. Proceedings of the American Society for Horticultural Science 73: 194-202. Vassiliadis, C. 1999. Evolution et maintien de l’androdioécie: Etude théorique et approaches expérimentales chez Phillyrea angustifolia L. PhD Dissertation, Université des Sciences at Technologies de Lille. Vassiliadis, C. & Lepart, J. & Saumitou-Laprade, P. & Vernet, P. 2000. Self-incompatibility and male fertilization success in Phillyrea angustifolia (Oleaceae). International Journal of Plant Sciences 161: 393-402. Wallander, E. Evolution of pollination and breeding systems in Fraxinus (Oleaceae), and a new classification. (Manuscript) Wallander, E. & Albert, V. A. 2000. Phylogeny and classification of Oleaceae based on rps16 and trnLF sequence data. American Journal of Botany 87: 1827-1841. Wallander, E. & Dahl, Å. Studies on the reproductive

22

ecology of F. excelsior, F. ornus and F. longicuspis, and a comparison with some other species of Fraxinus (Oleaceae). (Manuscript) Weller, S. G. & Sakai, A. K. & Rankin, A. E. & Golonka, A. & Kutcher, B. & Ashby, K. E. 1998. Dioecy and the evolution of pollination systems in Schiedea and Alsinidendron (Caryophyllaceae: Alsinoideae) in the Hawaiian Islands. American Journal of Botany 85: 1377-1388. Whitehead, D. R. 1968. Wind pollination in the angiosperms: evolutionary and environmental considerations. Evolution 23: 28-35. Whitehead, D. R. 1983. Wind pollination: Some ecological and evolutionary perspectives. In: Pollination Biology, chapter 5, pp. 97-108, (ed: Real, L.), Academic Press, London, U.K. Williams, G. & Adam, P. 1999. Pollen sculpture in subtropical rain forest plants: Is wind pollination more common than previously suspected? Biotropica 31: 520-524. Willson, M. F. 1979. Sexual selection in plants. American Naturalist 113: 777-790. Wolf, D. E. & Rieseberg, L. H. & Spencer, S. C. 1997. The genetic mechanism of sex determination in the androdioecious flowering plant, Datisca glomerata (Datiscaceae). Heredity 78: 190-204. Yampolsky, C. & Yampolsky, H. 1922. Distribution of sex forms in the phanerogamic flora. Bibliotheca Genetica 3: 1-62.