Bacteriocyte-Associated Symbionts of Insects A variety of insect groups harbor ancient prokaryotic endosymbionts Nancy A. Moran and Aparna Telang

E

ndosymbiosis has been proposed as an evolutionary innovation that underlies the Endosymbiosis appears to success and diversification of some facilitate diversification groups of organisms by enabling them to exploit new ecological niches within ecological niches (e.g., Margulis and Fester 1991, Maynard Smith and Szathmary that would otherwise 1995). Three kinds of information are needed to evaluate this proposal. be inadequate, but First, it is necessary to know somedependence on an thing of the role played by symbionts in the lives of hosts: Are their effects endosymbiont may ones that allow use of novel ecological niches? Second, information is simultaneously enforce needed on the evolutionary history of endosymbioses: How old are they some restrictions on and, in particular, do endosymbiotic infections predate diversification of host evolution major host groups? Finally, the evolutionary consequences for hosts of acquiring new capabilities in the form 1984). Bacteriocyte associates are of symbiotic associations must be found in many insect orders and famievaluated: Are these novel traits lies and have been estimated to occur stable? Do they impose constraints in 10% of insect species (Douglas as well as new capabilities on hosts? Recent studies have begun to shed 1989). Before the past decade, much of light on these questions for bacthe knowledge of bacteriocyte assoteriocyte-associated endosymbioses ciations was based on extensive miin insects. Bacteriocytes (or mycecroscopical studies by Paul Buchner tocytes) are cells of animal hosts that and his associates. His fascinating are positioned at characteristic locacompilation of this work (Buchner tions in the body and appear to be 1965) details astonishing diversity specialized for housing bacteria (Figin animal-microbe associations. ure 1; Buchner 1965, Dasch et al. Most of this diversity is still unexplored, in large part because intraNancy A. Moran (e-mail: nmoran@ cellular symbionts typically cannot u.arizona.edu) is a professor in the De- be cultured outside of their hosts. partment of Ecology and Evolutionary Since 1990, however, molecular studBiology, and Aparna Telang is a gradu- ies on a handful of endosymbioses ate student in the Interdisciplinary Pro- have yielded considerable insight into gram in Insect Science, at the University their evolutionary histories and into of Arizona, Tucson, AZ 85721. © 1998 the specific adaptations to symbiosis American Institute of Biological Sciences. April 1998

that are found in the bacteria themselves. Some of the most revealing of these studies concern the bacteriocyte-associated, mutualistic endosymbionts of insects. The microscopical explorations performed by Buchner and his associates gave only hints regarding the relationships of insect endosymbionts to other bacterial groups and to other endosymbionts. The characterization of these bacteria and their phylogenetic placement relative to free-living prokaryotes remained almost a complete mystery until the development of molecular methods, in particular methods for determining nucleotide sequences. These methodological advances have revolutionized understanding of prokaryote evolution in general by providing a surplus of relatively reliable characters for phylogeny reconstruction. A major additional advantage of DNA sequences as phylogenetic characters is that they can be obtained for noncultivable organisms. In particular, sequences of the 16S ribosomal RNA (rRNA) genes have been used for the phylogenetic and taxonomic placement of many bacteria, including endosymbionts, that were previously unclassifiable (Maidak et al. 1996). In this article, we provide a brief overview of recent work on the nature ofthe relationships between insect hosts and bacteriocyte-associated organisms (see Dasch et al. 1984, Dadd 1985, Douglas 1989 for detailed reviews). We then consider in more detail the emerging picture of endosymbiont evolution presented

295

housing of endosymbionts by hosts. For many insect symbioses, beneficial effects of the bacteria on their hosts are supported by experimental evidence: Antibiotic or heat treatments causing loss of symbionts are accompanied by reductions in performance, including lowered growth and survivorship and loss of reproductive abilities (e.g., Nogge 1976, Ishikawa and Yamaji 1985, Ohtaka and Ishikawa 1991, Prosser and Douglas 1991, Sasaki et al. 1991, the aphid Uroleucon ja- Costa et al. 1993, Heddi et al. 1993; ceae; arrowhead points additional studies cited in Koch 1967 to host cell plasma and Dasch et al. 1984). Many hosts, membrane. Bar = 2 |im. including aphids and other insect b, Buchnera; e, midgut groups that feed exclusively on plant epithelium; 1, midgut sap, can grow or reproduce only with lumen; m, mitochon- intact symbionts. Others, including dria; n, host cell nucle- Sitophilus weevils, which feed on us; ov, ovariole; s, sec- grain, can survive and reproduce ondary endosymbiont. without the symbionts, but they grow more slowly and are smaller than fleeted in parallel symbiotic hosts (Heddi et al. 1993). phylogenies? What Although interpreting studies of symages can we infer for biont-free hosts is complicated by the endosymbioses the possibility that antibiotic or heat of insects? Third, treatment may have direct effects on what adaptations of host metabolism, the preponderance endosymbiotic bac- of evidence points to a beneficial teria function in the role of bacteriocyte associates for mutualistic interac- host insects. tion with hosts? Fourth, what are the A nutritional dependence of hosts long-term evolution- on their symbionts presumably exary consequences of plains why many insect taxa possessendosymbiosis, as ing bacteriocytes use narrow and reflected in genetic nutritionally unbalanced diets, such changes in the sym- as the phloem sap of plants or the biotic bacteria? blood of vertebrates. Because animals lack biosynthetic pathways that In the arena of molecular studies, are typically present in prokaryotes, the most intensive most animals must ingest a large research on insect number of essential nutrients. Alterendosymbioses has natively, a microbial associate can been on aphids, small insects that provide the missing nutrients, allowfeed by tapping phloem sap of vascu- ing animal hosts to exploit resources lar plants, and on their primary en- that would otherwise be nutritiondosymbionts, Buchnera aphidicola. ally deficient. Examples include Molecular studies of this association members of the Hemiptera that use have been reviewed recently in more their sucking mouthparts to feed solely on plant phloem or xylem sap detail (P. Baumann et al. 1997). (Dasch et al. 1984}. Plant phloem sap contains free amino acids but The mutualistic nature of lacks some amino acids that are essential nutrients for animals. Anbacteriocyte associations other dietary habit associated with Buchner considered bacteriocyte as- endosymbiosis is restriction to vertesociates to be mutualists that benefit brate blood, which is lacking in some insect hosts by providing limiting vitamins. Endosymbiosis connected nutrients. His views were based on to blood feeding throughout the life myriad observations of specialized cycle has evolved independently in a mechanisms for the transmission and Figure 1. Electron mi-

crographs oiBuchnera in aphid bacteriocytes. (a) Bacteriocytes adjacent to the midgut of the aphid Uroleucon sonchi. Bar = 10 |im. (b) Bacteriocytes adjacent to the ovarioles of U. sonchi. Bar = 5 |im. (c) Bacteriocytes adjacent to secondary endosymbionts housed in separate cells within

by molecular phylogenetic findings. Finally, we consider how such studies bear on the following sets of questions: First, what is the distribution of endosymbiotic bacteria on the prokaryotic tree of life? Which bacterial groups have produced lineages that are endosymhiotic in insects? Are these related to one another as members of one or a few monophyletic groups specialized for endosymbiotic living? Or do endosymbionts of different insect lineages represent independent origins of endosymbiotic bacteria? Second, do the bacteria show long-term patterns of cospeciation with their hosts, as re296

BioScience Vol. 48 No. 4

number of insect orders, including Diptera, Hemiptera, and Mallophaga (Buchner 1965, Dasch et al. 1984). Some insects with more generalized diets also have endosymbionts; these include cockroaches, some ants, and some weevils. In ants, however, endosymbiosis may be associated with dependence on nutritionally deficient plant fluids. The general conclusion of a nutritional role for symbionts has been accepted by almost all researchers who have studied bacteriocyte associates of insects (e.g., Buchner 1965, Koch 1967, Dasch et al. 1984, Dadd 1985). Yet definitive evidence that endosymbionts provide essential nutrients to insect hosts is rare. The best evidence pertains to the provision of tryptophan by Buchnera, the endosymbionts of aphids (Douglas and Prosser 1992, Lai et al. 1994). Also, some evidence suggests that tsetse symbionts provision hosts with B vitamins (Nogge 1981).

How are endosymbionts transmitted?

infection process vary tremendously among insect groups (e.g., Sacchi et al. 1988,Hypsal993).Forexample, in wbiteflies, an entire maternal bacteriocyte (or, in some species, several bacteriocytes) is transferred into each egg, witb the maternal nucleus later degenerating (Bucbner 1965, Costa etal. 1993).

Secondary endosymbionts Some insects harbor so-called "secondary" endosymbionts that coexist in the same individual hosts with the bacteriocyte-inbabiting primary endosymbionts. These secondary symbionts are maternally inherited but appear not to share a long evolutionary bistory witb their hosts. Tbey are usually not present in the bacteriocytes tbat bouse primary endosymbionts (Figure lc; Buchner 1965). Tsetse fly secondary symbionts occur in midgut cells, by contrast to tbe primary symbionts, wbich are found in specialized bacteriocytes in tbe anterior gut (Pinnock and Hess 1974). Aphid secondary endosymbionts occur mostly in a sbeatblike syncytium bordering the bacteriocytes tbat house Buchnera (Bucbner 1965, Fukatsu and Isbikawa 1993, Chen and Purcell 1997). Whitefly secondary endosymbionts do inhabit the bacteriocytes in wbich primary endosymbionts are found but are clumped irregularly within vacuoles (Costa etal. 1993,1996).

Bacteriocyte-associated endosymbionts are typically maternally inherited, often through complex developmental events that ensure transovariole transfer from the mother to the developing egg or embryo (Buchner 1965, Dasch et al. 1984). In the case of aphids and Bucbnera, the bacteria are housed within bacteriocytes that are grouped Effects of secondary endosyminto a loose organ witbin the body bionts on host fitness are not clear. cavity, in the vicinity of the develop- Bucbner (1965) described secondary ing ovarioles (Figure lb; Buchner endosymbionts of apbids as "recently 1965, Hinde 1971a, 1971b). The acquired guests wbich are still in maternal bacteriocyte adjacent to an need of adaptation." They are more embryo near tbe blastoderm stage sporadic in tbeir numbers and presforms a small opening through whicb ence among species and even in their a bacterial inoculum passes. Tbe in- distribution among individuals of the oculum then moves through tbe in- same species, as bas been sbown for tervening hemolymph and enters a aphids (Fukatsu and Ishikawa 1993, nearby opening on tbe oocyte sur- Chen and Purcell 1997), whiteflies face. During early embryonic devel- (Costa et al. 1993), and tsetse flies opment, the presumptive bacterio- (Pinnock and Hess 1974). Cben and cytes form, and the inoculating Purcell (1997) found tbat secondary bacteria migrate into tbese cells. endosymbionts are present in only some strains of pea apbid {AeyrthoThis apparent fine tuning of tbe siphon pisum) and are identical in infection process is typical of 16S rDNA sequence to secondary bacteriocyte-associated symbionts in endosymbionts of rose aphid (Mijcroinsects and suggests a long bistory of siphum rosfle). Vertical transmission selection favoring bost adaptations from the time of tbe common ancesthat help to maintain tbe associa- tor of tbese two bosts would almost tion. However, the particulars of tbe Aprill998

certainly have resulted in some sequence divergence; for example, the Buchnera 16S rDNA sequences for similarly closely related genera within Aphididae all differ by more tban 2% (Munsonet al. 1991). The observed sequence identity of the secondary symbionts implies the occurrence of either horizontal transfer between species or recent, independent infections by a bacterium that is widely distributed in tbe environment. Furthermore, the symbionts could be experimentally transferred by injecting hemolympb from infected clones into uninfected clones of the same or different species. Thus, although maternal transmission was observed (Chen and Purcell 1997) and is presumably the usual route of infection, tbese secondary endosymbionts must also undergo some transfer among bost lineages, including transfer between species. Their sporadic distribution within host species implies that unlike most primary endosymbionts, they are not required for bost development or reproduction. Currently, there is almost no evidence indicating whether tbeir effects on bost fitness are positive, negative, or neutral.

The phylogenetic distribution of insect endosymbionts Molecular pbylogenetic results based on 16S rDNA sequences show that within each of tbe insect groups examined so far, the primary endosymbionts descended from a single ancestor, as indicated by their forming a well-supported clade. This result has been obtained for the primary endosymbionts of aphids (Munson et al. 1991, Moran et al. 1993), mealybugs (Munson et al. 1993), wbiteflies (Clark et al. 1992), carpenter ants [Candidatus camponotii; Schroder et al. 1996),5;top/7/7M5 weevils (Campbell etal. 1992), tsetse flies (Aksoy 1995), and cockroaches plus termites (Bandi et al. 1994, 1995). Tbe endosymbionts of each of these insect groups form a distinct monophyletic group. The most obvious explanation for these findings is that the common ancestor to each of these ciades was also endosymbiotic in an ancestor to the same group of hosts. The positions of these endosymbiont clades on the bacterial phylo-

297

TJ U U

Proteobacteria

eubacteria Fieure 2. Phylogenetic placement of endosymhionts. (a) Phylogenetic relationships of major euhacterial groups (Woese 1987), showing the position of the Proteohacteria, which includes most hacteriocyte associates of insects, and the bacteroidesflavohacteria group, which includes the hacteriocyte associates of cockroaches and primitive termites, (b) Phylogenetic relationships of bacteriocyte-associated symhionts of insects and representative free-living bacteria withm the Proteobacteria based on published analyses of 16S ribosomal DNA (see text for references).

genetic tree show that insect endosymbionts have arisen a number of times independently from free-living bacterial groups (Figure 2). Most endosymbionts have originated from within the Proteobacteria, particularly the gamma and beta subdivisions. These two subdivisions are large sister clades within the Proteobacteria, each containing a wide variety of ecological types. 298

A number of insect endosymbionts fall within the gamma-3 subdivision ofthe Proteobacteria, a diverse group that includes Escberichia coU and other enteric bacteria. There are two main groups of insect symbionts within this subdivision. Clustered within the Enterobacteriaceae near £. coli are the secondary endosymbionts of pea aphids (Unterman et al. 1989), the secondary endosymbionts

of tsetse flies (Aksoy et al. 1995), and the primary endosymbionts of Sitopbilus weevils (Campbell et al. 1992). Together, these endosymbionts may form a clade, although variation in 16S rDNA sequences is not sufficient to resolve relationships within the gamma subdivision (Figure 2). Also within the gamma-3 subdivision, but outside of the EnterobacteBioScience VoL 48 No. 4

riaceae, is a clade consisting of pri- Table 1. G + C composition of 16S rDNA of free-living and symbiotic bacteria. niary endosymbionts of aphids, carG + Cin 16SrDNA(%) penter ants, and tsetse flies, with Taxon each of the three symbiont groups Aeromonas trota 55.2 forming a well-supported clade. Cur- Alcaligenes sp. 54.9 54.4 rent phylogenetic evidence raises the Escherichia coli 56.1 possibility that the shared ancestor Halomonas etongata 52.2 Haemophilus influenzae for tsetse, ant, and aphid primary 53.1 Legionella sp. symbionts was endosymbiotic or Neisseria gonnorheae 55.0 possessed habits causing it to readily Proteus uulgaris 52.8 54.1 enter into endosymbiotic associations Pseudomonas aeruginosa 54.6 with insects. However, this possibil- Pseudomonas testosteroni 53.8 ity is not yet firmly established; in- Vibrio parahaemolyticus 54.1 Yersinia pestis deed, it wonld be surprising in view Aphid primary symbiont {Buchnera aphidicola) 49.3 of major differences among the host Ant primary symbiont {Candidatus componotii) 47.7 47.9 taxa and among the developmental Tsetse primary symbiont {Wigglesworthia glossinidia) 47.8 patterns of the different symbioses. Whitefly primary symbiont 54.0 Weevil symbiont The 16S rDNA sequences of these 55.7 Mealybug symbiont endosymbionts have a low G+C con- Pea apbid secondary symbiont 54.1 tent compared with free-living mem- Wbitefly secondary symbiont 54.4 bers of the gamma-3 subdivision (Table 1), raising the possibility that phes that those ancestors lived at the their apparent close phylogenetic ber of taxa included and on the ro- same time. Consequently, minimum bustness of the phylogenies. Varying relationship results in part from conages of the infections can be inferred vergent evolution at some nucleotide levels of positive support have been from the fossil records of the insects found for aphids (Munson et al. positions. Also, undiscovered free(Table 2). The general conclusion living bacteria may belong to the 1991, Moran et al. 1993), tsetse flies from thus incorporating paleonto(Aksoy et al. 1995), carpenter ants same clade. Thus, phylogenetic evilogical evidence is that these infecdence from 16S rDNA is inconclu- (Schroder et al. 1996), and cock- tions are very old, dating back to roaches (Bandi et al. 1995). Even sive but consistent with the indepenperhaps 300 million years in the case dent derivation of aphid, ant, and more important, no study has yet of the clade containing cockroaches produced evidence contradicting the tsetse endosymbionts from free-livhypothesis of parallel phylogenesis plus termites and to 200 million years ing ancestors. for bacteriocyte-associated endosym- in the case of the clade represented by aphids (Table 2). Schroder et al. The best-known exception to the bionts in insects. Thus, the syndrome (1996) suggest that the ancestor of of long-term associations that are concentration of insect bacteriocyte all ants may have possessed an endoassociates within the Proteobacteria strictly maternally inherited may be symbiont that was subsequently lost typical of these associations. are the intracellular symbionts of in most lineages but retained in carThe support for long-term codiversi- penter ants and certain other species cockroaches and primitive termites (Bandietal. 1994,1995). These sym- fication of insects and their endosym- in the subfamily Formicinae. This bionts form a clade within the bionts is consistent with microscopical hypothesis remains to be tested bacteroides-flavobacteria group and and experimental observations of through examination of other are, thus, distantly related to other maternal transmission by Buchner formicines. If it is correct, the age of (1965) and others. However, the the infection would exceed 80 milknown bacteriocyte associates. phylogenetic studies extend these lion years. In addition to the eviresults by demonstrating that even dence based on congruent phylogAges and patterns rare instances of horizontal trans- enies and insect fossil dating, the of cospeciation mission are absent, a fact that could antiquity of the infections is supnever be established based on exAnother consistent pattern that has perimental observations of transmis- ported by levels of 16S rDNA seemerged for bacteriocyte-associated sion events alone. Even infrequent quence divergence within clades of bacteria in insects is congruence of horizontal transmission events would endosymbionts. the phylogenies of the bacterial and scramble patterns of congruence of The phylogenetic evidence for host clades. For every case that has host-symbiont phylogenies when bacteriocyte associates of insects been studied sufficiently, the phy- such phylogenies extend millions of contrasts with that for some other logeny for the bacteria matches the years into the past. endosymbioses. In Wolbachia pipiphylogeny for the corresponding inentis, a bacterium that infects insect For each insect bacteriocyte assosects. This congruence is strong evigerm line cells and causes various dence for parallel diversification aris- ciation, congruence of host and sym- reproductive abnormalities, molecuing from faithful and long-term biont phylogenies provides strong lar phylogenies give a clear indicavertical transmission of endosym- evidence that the original infection tion of noncongruence, indicating bionts along host lineages. Support predates the basal ancestor of the horizontal movement of bacteria for congruence varies among the symbiont and of the corresponding among lineages (Werren 1997, cases studied; it depends on the num- hosts. In addition, congruence imApril 1998

199

Table 2. Insect groups for which molecular phylogenetic studies support parallel cladogenesis of bacteriocyte-associated endosymhionts and hosts, implying ancient infection of a common ancestor of the host clade. Approximate minimum ages are based mainly on fossil evidence. Minimum age of symbiotic association (i.e., age of common ancestor of hosts)

Reference(s)

Aphids (Hemiptera: Aphidoidea)

150-250 million years

Munson etal. 1991, Moran etal. 1993

Cockroaches + termites (Isoptera + Blattaria)

135-300 million years

Bandi et al. 1995

Tsetse flies (Diptera: Glossinidae)

40 million years

Aksoy et al. 1995

Carpenter ants {Hymenopetera: Formicidae; genus Camponotus)

More than 50 million years?

Schroder et al. 1996

Host clade

Bourtzis and O'Neill 1998). Likewise, in some marine invertebrates, host and endosymbiont phylogenies also appear not to be congruent (Dubilier et al. 1995, Krueger and Cavanaugh 1997, Distel 1998). What is the basis for this difference? That is, why might bacteriocyte associates of insects lack the capacity for horizontal transfer? When symbionts are maternally transmitted and beneficial to hosts, hosts are selected to provide a continually favorable environment to ensure the persistence of the association. Consequently, selection on bacteria for the ability to withstand conditions outside bacteriocytes will be relaxed. The resulting intolerance of conditions outside hosts could effectively eliminate horizontal transfer, facilitating long-term cospeciation of host and symbiont. In the case of Wolbacbia^ by contrast, the typical infection route is maternal but the interaction appears to be more parasitic than mutualistic (Werren 1997, Bourtzis and O'Neill 1998). The bacteria do not reside in specialized bacteriocytes, and their maintenance and transmission do not appear to be facilitated by host adaptations. Rather, they inhabit a variety of cell types and invade host eggs using their own devices (Werren 1997). As a result, they may retain a more generalized ability to live under different conditions and thus have greater propensity for rare horizontal transmission. In some marine invertebrates, such as bivalves and marine annelids, the associations are 300

mutualistic, and hosts have evolved specialized cell types for housing symbiotic bacteria (Conway et al. 1992, Distel 1998). But long-term cospeciation may be prevented by the mode of infection: In at least some such associations, infecting stages are not maternally transferred but are waterborne, providing more opportunity for cross-infection of host lineages.

Adaptations of endosymbiotic bacteria Because bacteriocyte associates of insects cannot be cultured outside of hosts, it has been difficult to show how they differ from free-living bacteria. However, genetic characterization of Bucbnera, the primary endosymbiont of aphids, has demonstrated some striking adaptations to endosymbiotic hfe (P. Baumann et al. 1997). The best-documented contribution of aphid endosymbionts to their hosts is the provision of required amino acids that are rare or absent from the phloem sap diet. In particular, several lines of evidence support the role of Buchnera in provisioning hosts with tryptophan, an amino acid that is required by animals but is rare in phloem sap (Douglas and Prosser 1992, Lai et al. 1994). Tryptophan biosynthesis is regulated by feedback inhibition acting on anthranilate synthase, which is encoded by trpEG. In Buchnera that are endosymbiotic with members of the Aphididae, trpEG is excised from the chromosome and amplified on plasmids (Lai

et al. 1994). This amplification, which involves both the tandem duplication of the operon on each plasmid and the presence of multiple plasmids, appears to function in the overproduction of tryptophan through enhanced production of the limiting biosynthetic enzyme. Comparisons of phylogenies based on plasmidborne and chromosomal genes indicate that trpEG on plasmids originated from the chromosomal genes within the same bacterium and that the plasmids are transmitted strictly vertically (Rouhbaksh et al. 1996, 1997). The trpEG plasmid presumably was fixed through favorable selection at the host level; that is, host insects bearing endosymbionts with this feature experienced greater growth rates and became established at the expense of other aphid lineages of the same species in which endosymbionts possessed only a chromosomal trpEG copy. A similar instance of plasmid amplification enabling the overproduction by Buchnera of an essential aphid nutrient has been discovered for leucine production (Bracho etal. 1995). The four genes that underlie production of leucine are excised from the chromosome and transferred to plasmids in some Buchnera species. However, only a single set of the leucine genes is present on each plasmid. The presence of the replicating unit of the plasmid appears to be ancestral in Buchnera of all aphids, but its acquisition of the leucine biosynthetic genes appears to have occurred independently in different lineages (van Ham etal. 1997). Other unusual aspects of the Buchnera genome may represent adaptations to endosymbiotic life. Because symbionts are confined to host cells, their growth rates must be depressed and coordinated with the development of the host. For Bucbnera, there is a tight coupling of bacterial cell number and aphid growth, with Buchnera showing a doubling time of approximately two days, a growth rate that is much lower than the maximum exhibited by many free-living bacteria (Baumann and Baumann 1994). Probably in connection with this low maximal growth rate, both Bucbnera and Wiggleswortbia, the bacteriocyte associate of tsetse flies, possess only BioScience VoL 48 No. 4

a single copy of the rRNA genes, in contrast to free-living bacteria, which typically have several copies (Aksoy 1995, Baumann et al. 1995). Reduction in the number of copies of rRNA genes in prokaryotes is associated with lowered maximum growth rates. Endosymbiont adaptations for improved function as mutualists present the clear possibility of conflict between selection on individual cell lineages within a host aphid and selection on individual aphids. For example, the excess production of tryptophan must be beneficial to hosts but is probably detrimental to the bacterial cell lineage producing the nutrient. In related free-living bacteria, production of anthranilate synthase has been shown to be costly; indeed, it lowers fitness of bacteria when tryptophan is not limiting (Dykhuisen 1978}, suggesting that a mutant Buchnera lacking the excess copies would enjoy a short-term advantage. However, host lineages in which all symbionts were lacking the excess copies would suffer a nutritional handicap and would thus be at a selective disadvantage within aphid populations. The high mutation rate for loss of tandem repeats in bacteria (10^-10"^ per generation; Roth et al. 1996), in combination with the immediate selective advantage for mutant cells, would be expected to destabilize these host-level adaptations. In the long term, however, selection at the host level would favor adaptations that stabilize beneficial gene amplification. Some findings suggest that such host-level selection has resulted in reduced homologous recombination in Buchnera. Curiously, in certain Bucbnera of the family Aphididae, reduction in plasmid-borne functional trpEG has occurred through the transformation of some of the repeats into pseudogenes (Lai et al. 1996, L. Baumann et al. 1997). Typically, however, such unneeded repeats would instead be eliminated through recombination. A possible explanation for the transformation into pseudogenes is selection on hosts to stabilize beneficial gene amplification, resulting in loss of some recombination capabilities (P. Baumann et al. 1997). If certain aphid lineages subsequently evolved feeding habits or life cycles that reduced

April 1998

their need for tryptophan provision by Buchnera, selection then might have favored reduced trpEG expression due to the energy expense of making unneeded anthranilate synthase. If recombinational mechanisms for eliminating repeat copies had been lost earlier, gene silencing would have had to occur through the fixation of stop codons and frameshifts in the DNA sequence, resulting in transformation of the functional genes into pseudogenes. If this scenario is correct, then the loss of certain recombination functions would appear to be an adaptation to stabilize mutualistic traits of the endosymbiont that might otherwise be eliminated through a combination of high mutation rates and selection at the bacterial cell level.

Evolutionary consequences of endosymbiosis Endosymbiosis is sometimes considered to be an initial step in the evolution of organelles (e.g., Margulis 1993). However, organelles have arisen only a few times from freeliving prokaryotes, either once or twice each for mitochondria and chloroplasts (Cavalier-Smith 1992, Cray 1992}. By contrast, intracellular symbionts have arisen far more often, as a glimpse at Buchner's (1965} volume confirms. Thus, organelle characteristics are not an inevitable consequence of endosymbiosis combined with maternal transmission. One feature of mitochondrial and chloroplast genomes is the loss of most genes present in the ancestral prokaryotic genome and the dependence on host gene products for many of the basic functions required for replication. The only endosymbiont for which substantial genome characterization has been carried out is Buchnera, for which 85 kb has been sequenced in the Baumann laboratory (summary in P. Baumann et al. 1997). Despite substantial overall sequence divergence from related free-living bacteria, Buchnera coding gene sequences indicate selective constraint of the functioning polypeptides; the reading frame is maintained and nucleotide base substitutions are concentrated at silent, or synonymous, sites (i.e., sites that do

not affect amino acid identity}. Cenes with a wide range of "housekeeping" and biosynthetic functions have been identified in Buchnera (summaries in Baumann et al. 1995, P. Baumann et al. 1997, Clark and Baumann 1997, Clark et al. 1997}. This work provides strong evidence that Buchnera contains a complement of genes enabling a wide range of processes involved in cell growth and reproduction. This endosymbiotic association arose at least 200 million years ago, so the retention of most genes is not readily explained as the result of the association being too young for Buchnera to have developed characteristics of organelles. A more important limitation on the loss of genes might be that endosymbionts in animals reside primarily in somatic cells (the bacteriocytes}, eliminating most opportunities for gene transfer between symbiont and host nuclear genome. One apparent long-term consequence of endosymbiosis is increased rates of substitution in DNA sequences of endosymbionts relative to non-endosymbiotic relatives (Moran 1996}. This increase is seen in the 16S rDNA of endos.ymbionts of aphids, whiteflies, mealybugs, tsetse flies, and carpenter ants (Moran 1996, Lambert and Moran in press}. In Buchnera., in which many more genes have been sequenced than in any other endosymbiont, the increased rate extends to all genes examined (Moran 1996, unpublished analyses}. Positive natural selection for base substitutions seems unlikely to explain the faster evolution at all loci, because the effects of positive selection are expected to be locus and site specific within genes. In fact, it appears that the base substitutions are deleterious. Evidence for this hypothesis is provided by the changes in protein-coding genes in Buchnera. The net effect of the faster evolution in Buchnera is the accumulation, throughout the genome, of A and T at the expense of representation of G and C. This shift in nucleotide base composition is strong enough that most amino acid differences between polypeptides of Buchnera and those of E. coli are ones that allow increased A -H T in the corresponding codon and thus in the DNA sequence. 301

It is hard to imagine that the accumulation, in all polypeptides, of amino acids with A + T-rich codon families is adaptive at the level of protein function. A more plausible explanation is mutational bias favoring A + T combined with an increased rate of base substitution within lineages. If positive selection is excluded as the basis for the faster substitution rates of endosymbionts, we are left with two other categories of explanation. One is that the increased rate of substitution (fixed changes occurring in a lineage) is due to an increased rate of new mutations. This increased mutation rate could result from either an increased rate per generation or a faster average generation time. But if increased incidence of mutation were the sole explanation, the increase would have equivalent effects on substitution rates at sites that are subject to selection and at sites that are effectively neutral with respect to selection, and comparative sequence analysis indicates stronger effects on selected sites. A convenient way to categorize sites according to the intensity of selection is to separate silent (or synonymous) from replacement (or nonsynonymous) sites in protein coding sequences. At silent sites, where nucleotide changes do not affect the corresponding amino acid and thus do not affect the gene product, base substitutions have little consequence for fitness. At replacement sites, where a nucleotide change results in a change in the amino acid and thus in the selected phenotype, base substitutions are more likely to affect fitness. The increase in substitution rate in coding genes of Buchnera is heavily concentrated at replacement sites (Moran 1996). This observation rules out increased mutation as the primary basis for the acceleration in substitution rates. This result instead supports the second explanation, that the excess changes represent deleterious substitutions. According to this view, purifying selection is less effective at purging deleterious mutations. This reduction in the effects of selection could arise in either of two ways: from relaxed selection or from lower effectiveness of selection as a result of population structure. The first pos302

sibility, that selection is uniformly relaxed at all genes of Buchnera, regardless of function, is doubtful. Because endosymbionts show reduced effective population sizes than free-living bacteria, with a consequently greater susceptibility to genetic drift, the second possibility may apply. This proposal resembles the arguments made by Lynch (1996, 1997) for the accumulation of deleterious mutations in mitochondria and other organelles. Organelles have similar population structure to endosymbionts because they are asexual, intracellular, and maternally inherited.

Current mysteries and future problems The recent molecular studies of insect bacteriocyte-associated endosymbioses have given glimpses into the evolution of some of these associations. But most remain entirely unstudied, particularly the complex multiple infections found in some insects, such as the planthoppers and treehoppers (Hemiptera; Buchner 1965). Furthermore, this work has raised some new questions about the evolution of bacteriocyte associates in insects. For example, both Buchnera and tsetse endosymbionts exhibit unusually high levels of GroEL protein (Ishikawa 1989, Aksoy 1995, Baumann et al. 1996), a chaperonin that normally functions in refolding other polypeptides. The reason for this overexpression is not yet clear, although speculative explanations have been put forward (Ishikawa 1989, Moran 1996, P. Baumann et al. 1997). The finding of plasmidborne pseudogenes in some Buchnera presents another mystery: Why are nonfunctional copies not lost by recombination, as is typical for bacteria? Some basic difference involving recombination pathways may, as already discussed, distinguish Buchnera from free-living bacteria. Whether other endosymbionts also have atypical recombination pathways is not yet known. The basis for the nucleotide base compositional bias against G + C, a distinctive feature of numerous bacteriocyte associates (Table 1), is also unknown. Endosymbiotic interactions are shaped by forces acting at multiple levels, primarily that of the host in-

dividual and the bacterial cell lineage. Some mutations in endosymbionts will favor the individual symbiont lineage at the expense of the host. Such "selfish" mutations would include any that result in higher replication of the symbiont at the host's expense. If higher replication results in disproportionate transmission to progeny, these selfish symbionts could increase in frequency within a host population, at least in the short term. Michod (1997) presents a theoretical framework for evaluating this very situation that characterizes bacteriocyte associations in insects: a balance between frequent mutation in the direction of "selfish" traits favoring individual cell lineages and selection on the overall group (in this case, the host individual and its resident bacterial population). Buchner (1965) mentions cases in whiteflies in which symbionts spread beyond their usual limits within individual host insects, indicating the possibility of such selfish behavior that harms the host. Of course, selection favoring better-nourished (and faster-growing) host individuals will counter the spread of selfish bacterial lineages. One host adaptation that would limit the spread of selfish symbionts would be the sequestration of "germ line" symbionts from more rapidly dividing "somatic" symbionts that function in provisioning hosts with nutrients. Early microscopical studies suggest this kind of system in cicadas (Koch 1967). The infection of progeny with entire maternal bacteriocytes, as in whiteflies (Buchner 1965, Costa et al. 1996), would also have the effect of limiting the spread of "selfish" bacteria to single host progeny, provided that bacteria are confined to single bacteriocytes early in the development of the host. Frank (1996) proposes thatthe segregation of germ line symbionts may be a "policing" mechanism for the containment of selfish cell lineages within hosts of endosymbionts.

Role of endosymbiosis in host evolution Endosymbiosis has been argued to be a route for evolutionary innovation that underlies the origin and diversification of many groups of BioScience Vol. 48 No. 4

organisms that would otherwise not exist. Within the realm of the insects, this hypothesis appears to be true: Essentially all strict phloemfeeders, strict blood-feeders, and other groups with restricted diets possess endosymbionts. Recent molecular phylogenetic evidence indicates that endosymbiotic infections date to the origins of a variety of major insect clades, reinforcing the idea that the origin of endosymbiosis was an integral factor spurring diversification of some insect taxa. Because some of these groups are large adaptive radiations, endosymbiosis has probably had a massive effect on patterns of insect diversity. Although endosymbiosis appears to facilitate diversification within ecological niches that would otherwise be inadequate, dependence on an endosymbiont may simultaneously enforce some restrictions on host evolution. Cenes encoded within the endosymbiont are subject to a different set of population genetic parameters than genes encoded in the insect nuclear genome. As noted above, endosymbionts are subject to within-host selection, which often may counter selection between hosts. In addition, mutation rates and generation times differ between endosymhiont and host genomes. The distinct population genetics of endosymbiont genes may affect the speed of the response to selection. For example, adaptation may occur faster in symbiont genes than host genes due to the shorter generation time and consequent greater supply of mutations of the bacteria. Finally, symbiont genes are subject to a population structure in which genetic drift is more likely to limit the effectiveness of selection, resulting in increased fixation of mildly deleterious mutations. This increased rate of accumulation of deleterious mutations will be countered, to some extent, by selection at the level of host individuals. The adequacy of this counterbalancing will depend on the impact of selection at the host level, which is dependent in turn on the effective population size of the host. It is not yet clear what factors are most important in determining the net effect of these conflicting evolutionary forces. One intriguing possibility is that the accumulation of

April 1998

of the genus Buchnera: Intracellular symbionts of apbids. Annual Review of Microbiology 49; 55-94. Baumann P, Baumann L, Clark MA. 1996. Levels of Buchnera aphidicola cbaperonin GroEL during growth of tbe aphid Schizaphis graminum. Current Microbiology 32: 279-285. Baumann P, Moran NA, Baumann L. 1997. Tbe evolution and genetics of aphid endosymbionts. BioScience 47: 12-20. Bourtzis K, O'Neill S. 1998. Wolbachia infec^ tions and arthropod reproduction. BioScience 48: 287-293. Bracho AM, Martinez-Torres D, Moya A, Latorre A. 1995. Discovery and molecular Acknowledgments characterization of a plasmid localized in Buchnera sp., bacterial endosymbiont of Research on insect endosymbionts tbe aphid Rhopalosiphum padi. Journal of in the Moran lab is supported by Molecular Evolution 41: 63-73. grants from the National Science Buchner P. 1965. Endosymbiosis of Animals witb Plant Microorganisms. New York: Foundation (DEB-9527635) and the Interscience, US Department of Agriculture, NaBC, Bragg TS, Turner CE. 1992. tional Research Initiative (96-35302- Campbell Phylogeny of symbiotic bacteria of four 3817). Aparna Telang was supported weevil species (Coleoptera: Curculionidae) by a fellowship from the Plant-Insect based on analysis of 16S ribosomal RNA. Interactions Training Grant to the Insect Biochemistry and Molecular Biology 22: 415^21. University of Arizona from NSF, T. 1992. The origins, losses, USDA, and the US Department of Gavalier-Smith andgains of chloroplasts. Pages 291-349 in Energy. We thank Paul Baumann for Lewin RA, ed. Procholorophytes and comments on an early draft of this Symbiogenesis and tbe Origins of Chlororeview and Leah Kenaga and David plasts. New York: Chapman &c Hall. Bentley of the Imaging Facility (Ari- Chen D-Q, Purcell AH. 1997. Occurrence and transmission of facultative endosymbionts zona Research Laboratories) for asin aphids. Current Microbiology 34: 220sistance with electron microscopy. 225. Clark MA, Baumann P. 1997. The FIFO ATP syntVi^ie oi Buchnera aphidicola (tndosymbiont of aphids): Genetic analysis of the References cited putative ATP operon. Current Microbiology 35: 84-89. Aksoy S. 1995. Molecular analysis ofthe endosymbionts of tsetse flies: 16S rDNA locus Clark MA, BaumannL, Munson MA, Baumann and over-expression of a chaperonin. Insect P, Campbell BC, Duffus JE, Osborne JS, Molecular Biology 4; 23-29. Moran NA. 1992. Tbe eubacterial endosymbionts of wbiteflies (Homoptera: Aksoy S, Pourhosseini AA, Chow A. 1995. Aleyrodoidea) constitute a lineage separate Mycetome endosymbionts of tsetse flies from tbe endosymbionts of aphids and meaconstitute a distinct lineage related to Enlybugs. Current Microbiology 25:119-123. terobacteriaceae. Insect Molecular Biology 4: 15-22. Clark MA, Baumann L, Baumann P. 1997. Buchnera aphidicola (endosymbiont of Bandi C, Damiani G, Magrassi L, Grigolo A, aphids) contains nuoC(D) genes that enFani R, Sacchi L. 1994. Flavobacteria as code subunits of NADH dehydrogenase. ititracellular symbionts in cockroaches. ProCurrent Microbiology 35: 122-123. ceedings of the Royal Society of London Conway NM, Howes BL, McDowell-Capuzzo Series B Biological Sciences 257: 43-48. JE,TurnerRD,CavanaugbCM. 1992. CharBandi C, Sironi M, Damiani G, Magrassi L, acterization and site description oiSolemya Nalepa CA, Laudani U, Saccbi L. 1995. The borealis (Bivalvia: Solemyidae), another biestablishment of intracellular symbiosis in valve-bacteria symbiosis. Marine Biology an ancestor of cockroaches and termites. Proceedings ofthe Royal Society of London 112:601-613. B Biological Sciences 259: 293-299. Costa HS,JohnsonMW,UlImanDE,Tabasbnik BE. 1993. The antibiotic oxytetracycline Baumann L, Baumann P. 1994. Growth kinetinterferes with Bemesia tabaci oviposition, ics of tbe endosymbiont Buchnera aphidicola development, and ability to induce squash in tbe aphid Schizaphis graminum. Applied silverleaf. Annals of the Entomological Soand Environmental Microbiology 60:3440ciety of America 86: 740-748. 3443. Baumann L, Clark MA, Rouhbakhsh D, Costa HS, Toscano NC, Henneberry TJ. 1996 Mycetocyte inclusion in the oocytes of Baumann P, Moran NA, Voegtlin DJ. 1997. Bemesia argentifolU. Annals of tbe EntoEndosymbionts (Buchnera) of the aphid mological Society of America 89: 694-699. Uroleucon sonchi contain plasmids with trpEG and remnants of trpE pseudogenes. Dadd RH. 1985. Nutrition: Organisms. Pages 315-319 in Kerkut GA, Gilbert LI, eds. Current Microbiology 35: 18-21. Comprehensive Insect Physiology, BiochemBaumann P, Baumann L, Lai C-Y, Rouhbaksb istry, and Pharmacology. Vol. 4. Elmsford D, Moran NA, Clark MA. 1995. Genetics, (NY): Pergamon Press. physiology, and evolutionary relationships

deleterious mutations would overrun purifying selection at the host level, except in cases in which the host population size is large. This possibility may help to explain why maternally transmitted endosymbionts are rare in animals such as vertebrates, which have small populations, as compared with insects, many of which have unusually large populations.

303

Dasch GA, Weiss AE, Chang KP. 1984. Endosymbionts of insects. Pages 811-833 in Kreig NR, Holt JG, eds. Bergey's Manual of Systematic Bacteriology. Vol. 1. Baltimore (MD): Williams and Wilkins. Distel DL. 1998. Evolution of chemoautotrophic endosymbioses in bivalves. BioScience 48: 277-286. Douglas AE. 1989. Mycetocyte symbiosis in insects. Biological Review of the Cambridge Philosophical Society 64: 409-434. Douglas AE, Prosser WA. 1992. Synthesis of the essential amino acid tryptophan in the pea aphid {Acyrthosiphon pisum) symbiosis. Journal of Insect Pbysiology 38: 565-568. Dubilier N, Giere O, Distel DL, Cavanaugh CM. 1995. Characterization of chemoautotrophic bacterial symbionts in a gutless marine worm (Oligochaeta: Annelida) by phylogenetic 16S rRNA sequence analysis and in situ hybridization. Applied and Environmental Microbiology 61: 2346-2350. Dykhuisen D. 1978. Selection for tryptopban auxotropbs of Eschericbia coli in glucose limited chemostats as a test of the energy conservation hypothesis of evolution. Evolution 32: 125-150. Frank SA. 1996. Host control of symbiont transmission: The separation of symbionts into germ and soma. The American Naturalist 148:1113-1124. Fukatsu T, Ishikawa H. 1993. Occurrence of chaperonin 60 and chaperonin 10 in primary and secondary bacterial symbionts of aphids: Implications for the evolution of an endosymbiotic system in aphids. Journal of Molecular Evolution 36: 568-577. Gray MW. 1992. The endosymbiont hypothesis revisited. Pages 233-357 in Wolstenholme DR, Jeon KW, eds. International Review of Cytology. Vol. 141. Mitochondrial Genomes. New York: Academic Press. Heddi A, Lefebvre F, Nardon P. 1993. Effect of endocytobiotic bacteria on mitochondrial enzymatic activities in the weevil Sitopbilus oryzae (Coleoptera: Cuculionidae). Insect Biochemistry and Molecular Biology 23: 403-411. Hinde R. 1971a. The control of the mycetome symbiotes of the aphids Brevicoryne brassicae, Myzus persicae, and Macrosipbum rosae. Journal of Insect Physiology 17:1791-1800. . 1971b. The fine structure of the mycetome symbiotes of tbe aphids Brevicoryne brassicae, Myzus persicae, and Macrosiphum rosae. Journal of Insect Physiology 17: 2035-2050. Hypsa V. 1993. Endocytobionts of Triatoma infestans: Distribution and transmission. Journal of Invertebrate Pathology 61: 3238. Ishikawa H. 1989. Biochemical and molecular aspects of endosymbiosis in insects. International Review of Cytology 116: 1-45. Isbikawa H. Yamaji M. 1985. Symbionin, an aphid endosymbiont-specific protein; Production of insects deficient in Symbionin. Insect Biochemistry 15: 155-163. Koch A. 1967. Insects and their endosymbionts. Pages 1-106 in Henry SM, ed. Symbiosis. Vol. II. New York: Academic Press. Krueger DM, Cavanaugh CM. 1997. Phyloge-

304

netic diversity of bacterial symbionts of Sotemya hosts based on comparative sequence analysis of 16S rRNA genes. Applied and Environmental Microbiology 63: 91-98. Lai C-Y, Baumann L, Baumann P. 1994. Amplification of trpEG: Adaptation of Bwc/jKera aphidicola to an endosymbiotic association with aphids. Proceedings of the National Academy of Sciences of tbe United States of America 91: 3819-3823. Lai C-Y, Baumann P, Moran N. 1996. The endosymbiont {Bucbnera sp.) of the aphid Diuraphis noxia contains plasmids consisting of trpEG and tandem repeats of trpEG pseudogenes. Applied and Environmental Microbiology 62: 332-339. Lambert JL, Moran NA. In press. Deleterious mutations destabilize ribosomal RNA of endosymbionts. Proceedings of the National Academy of Sciences of the United States of America. Lynch M. 1996. Mutation accumulation in transfer RNAs: Molecular evidence for Muller's ratchet in mitochondrial genomes. Molecular Biology and Evolution 13: 209-220. , 1997. Mutation accumulation in nuclear, organelle, and prokaryotic transfer RNA genes. MolecularBiology and Evolution 14: 914-925. Maidak BL, Oisen GJ, Larsen N, Overbeek R, McCaughey MJ, Woese CR. 1996. The ribosomal database project (RDP). Nucleic Acids Research 24: 82-85. Margulis L. 1993. Symbiosis in Cell Evolution. 2nd ed. New York: W. H. Freeman. Margulis L, Fester R, eds. 1991. Symbiosis as a Source of Evolutionary Innovation. Cambridge (MA): MIT Press. Maynard Smith J, Szathmary E. 1995. Tbe Major Transitions in Evolution. New York: W. H. Freeman. Michod RE. 1997. Evolution of the individual. The American Naturalist 150 (supplement): S5-S21. Mittler TE. 1971. Some effects on the aphid Myzus persicae of ingesting antibiotics incorporated into artificial diets. Journal of Insect Physiology 17: 1333-1347. Moran NA. 1996. Accelerated evolution and Muller's ratchet in endosymbiotic bacteria. Proceedings of the National Academy of Sciences of the United States of America 93: 2873-2878. Moran NA, Munson MA, Baumann P, Ishikawa H. 1993. A molecular clock in endosymbiotic bacteria is calibrated using the insect hosts. Proceedings of the Royal Society of London B Biological Sciences 253: 167173. MunsonMA,BaumannP,ClarkMA, Baumann L, Moran NA, Voegtlin DJ. 1991. Evidence for the establishment of aphid-eubacterial endosymbiosis in an ancestor of four apbid families. Journal of Bacteriology 173:63216324. Munson MA, Baumann P, Moran NA. 1993. Phylogenetic relationships of the endosymbionts of mealybugs (Homoptera: Pseudococcidae) based on 16S rDNA sequences. Molecular Phylogenedcs and Evolution 1: 26-30. Nogge G. 1976. Sterility in tsetse flies caused by

loss of symbionts. Experientia 32: 995. . 1981. Significance of symbionts for the maintenance of an optimal nutritional state of successful reproduction in hematophagous arthropods. Parasitology 82:299-304. Ohtaka C, Ishikawa H. 1991. Effects of heat treatment on the symbiotic system of an aphid mycetocyte. Symbiosis 11: 19-30. Pinnock DE, Hess RT. 1974. Tbe occurrence of intracellular rickettsia-like organisms in the tsetse flies, Glossitia morsitans, G. fuscipes, G. brevipalpis, and G. pallidipes. Acta Tropica 31: 70-79. Prosser WA, Douglas AE. 1991. The aposymbiotic aphid: An analysis of chlortetracycline-treated pea apbid Acyrtbosipbon pisum. Journal of Insect Physiology 37: 713-719. Roth JR, Benson N, Galitski T, Haack K, Lawrence J, Miesel L. 1996. Rearrangements of tbe bacterial chromosome: Eormation and applications. Pages 2256-2276 in Neidhardt EC, ed. Escherichia coli and Salmonella: Cellular and Molecular Biology. 2iid ed. Washington (DC): American Society for Microbiology. Rouhbaksh D, Lai C-Y, von Dohlen CD, Clark MA, Baumann L, Baumann P, Moran NA, Voegtlin DJ. 1996. The tryptophan biosynthetic patbway of aphid endosymbionts {Bucbnera): Genetics and evolution of plasmid-associated antbranilate synthase {trpEG) within tbe Apbididae. Journal of Molecular Evolution 42: 414-421. Rouhbaksh D, Clark MA, Baumann L, Moran NA, Baumann P. 1997. Evolution of the tryptophan biosynthetic patbway in Buchnera (aphid endosymbionts): Studies of plasmid-associated trpEG witbin the genus Uro/eMcon. Molecular Phylogenetics and Evolution 8: 167-176. Sacchi L, Grigolo A, Mazzina M, Biliardi E, Baccetti B, Laudani U. 1988. Symbionts in the oocytes of Blattella germanica (L.) (Dictyoptera: Blattellidae): Their mode of transmission. International Journal of Insect Morphology and Embryology 17: 437-446. SasakiT,Hayashi H, Ishikawa H. 1991. Growth and reproduction of the symbiotic and aposymbiotic pea apbids, Acyrtbosiphon pisum, maintained on artificial diets. Journal of Insect Physiology 10: 749-756. Schroder D, Deppisch H, Obermeyer M, Krohne G, Stackebrandt E, Holldobler B, Goebel W, Gross R. 1996. Intracellular endosymbiotic bacteria of Camponotus species (carpenter ants): Systematics, evolution and ultrastructural characterization. Molecular Microbiology 21: 479-490. Untertnan BM, Baumann P, McLean DL. 1989. Pea aphid symbiont relationships established by analysis of 16S rRNAs. Journal of Bacteriology 171: 2970-2974. van Ham RCHJ, Moya A, Latorre A. 1997. Putative evolutionary origin of plasmids carrying the genes involved in leucine biosynz\\e.sismBucbneraapbidicola{en