Evolutionary Rates for tuf Genes in Endosymbionts of Aphids E. Ulrika Brynnel,*1 Charles G. Kurland,* Nancy A. Moran,† and Siv G. E. Andersson* *Department of Molecular Biology, Biomedical Center, Uppsala University, Uppsala, Sweden; and † Department of Ecology and Evolutionary Biology, University of Arizona The gene encoding elongation factor Tu (tuf) in aphid endosymbionts (genus Buchnera) evolves at rates of 1.3 3 10210 to 2.5 3 10210 nonsynonymous substitutions and 3.9 3 1029 to 8.0 3 1029 synonymous substitutions per position per year. These rates, which are at present among the most reliable substitution rates for protein-coding genes of bacteria, have been obtained by calibrating the nodes in the phylogenetic tree produced from the Buchnera EF-Tu sequences using divergence times for the corresponding ancestral aphid hosts. We also present data suggesting that the rates of nonsynonymous substitutions are significantly higher in the endosymbiont lineages than in the closely related free-living bacteria Escherichia coli and Salmonella typhimurium. Synonymous substitution rates for Buchnera approximate estimated mutation rates for E. coli and S. typhimurium, as expected if synonymous changes act as neutral mutations in Buchnera. We relate the observed differences in substitution frequencies to the absence of selective codon preferences in Buchnera and to the influence of Muller’s ratchet on small asexual populations.

Introduction Although molecular sequence data have greatly improved our knowledge of prokaryotic phylogeny, progress in dating ancestral divergence events has been hampered by the lack of identifiable prokaryotic fossils. Since some DNA sequences appear to evolve in a clocklike manner, they might be used to date divergence events. In order to realize this ambition, the clock must be calibrated. One method for calibrating molecular clocks in bacteria exploits endosymbionts that have codiversified with animal hosts that have a fossil record (Moran et al. 1993; Bandi et al. 1995). Associations between insects and intracellular prokaryotes are widespread among members of the insect orders Homoptera (aphids, whiteflies, mealybugs, psyllids, cicadas), Blattaria (cockroaches), Diptera (flies), and Coleoptera (beetles) (Buchner 1965; Baumann et al. 1995). The phylogenetic placement of these bacteria on the basis of their ribosomal RNA genes indicates that they are independent descendants of a variety of freeliving bacteria (Clark et al. 1992; Munson, Baumann, and Moran 1992; Aksoy, Pourhosseini, and Chow 1995; Bandi et al. 1995; Moran and Telang 1998). For example, Buchnera aphidicola, which are maternally inherited endosymbionts of aphids, are members of the gamma subdivision of Proteobacteria and closely related to free-living bacteria such as Escherichia coli (Baumann et al. 1995). The congruence between the host and endosymbiont phylogenies suggests that this particular association originated 100–250 MYA and that subsequent divergence events of both host and endosymbiont have been synchronous (Moran et al. 1993). This implies that dates estimated for ancestral aphids can be assigned to the corresponding ancestral Buchnera. 1 Present address: Malmo ¨ Museer, Department of City Archaeology, Malmo¨, Sweden.

Key words: Buchnera aphidicola, elongation factor Tu, mutation bias, molecular evolution, substitution rates. Address for correspondence and reprints: Siv G. E. Andersson, Department of Molecular Biology, Box 590, BMC, S-751 24 Uppsala, Sweden. E-mail: [email protected]. Mol. Biol. Evol. 15(5):574–582. 1998 q 1998 by the Society for Molecular Biology and Evolution. ISSN: 0737-4038

574

Using this approach, the rRNA genes in Buchnera have been estimated to evolve at a rate of 0.01–0.02 substitutions/site per 50 Myr (Moran et al. 1993). This rate is about 1.5–2-fold greater than the corresponding rate estimates for free-living bacteria (Moran 1996). Protein-coding genes also appear to exhibit an accelerated rate of sequence evolution in Buchnera, as well as a comparatively low ratio of synonymous-to-nonsynonymous substitutions (Moran 1996). These results have been explained as an effect of an increased rate of fixation of slightly deleterious mutations in combination with mutational bias toward A1T (Moran 1996). The tuf gene, coding for elongation factor Tu, has also been used as a reference gene to address questions concerning deep evolutionary relationships (Iwabe et al. 1989; Cammarano et al. 1992; Creti et al. 1994; Baldauf, Palmer, and Doolittle 1996). Like the rRNA gene, it has a universal occurrence and a high degree of sequence conservation, allowing alignments across a wide range of taxa. In this study, we determined the rate at which the tuf gene is evolving in Buchnera from five different aphids using estimated divergence times for the aphid hosts as calibration points. Our data suggest that the rate of nonsynonymous substitutions in the tuf gene is roughly the same as the overall rate of sequence evolution of the gene coding for 16S rRNA. Furthermore, we observe a much greater rate of synonymous substitutions for the tuf gene in Buchnera endosymbionts than in freeliving bacteria. The high rate of synonymous substitutions is correlated with the absence of the marked codon preference normally seen in tuf genes of free-living bacteria. We interpret this correlation as support for the hypothesis that, compared with free-living relatives, symbionts or parasites in small intracellular populations with low recombination frequencies undergo higher fixation rates for new mutations. Materials and Methods Laboratory Methods Genomic DNA was isolated as described previously (Munson et al. 1991) from Buchnera inhabiting the following aphid hosts: Acyrthosiphon pisum, Schizaphis graminum, Pemphigus betae, Melaphis rhois, and

Substitution Rates in Buchnera

Schlechtendalia chinensis (referred to as Buchnera (Ap), Buchnera (Sg), etc.). Primers used for the PCR amplification were degenerate oligomers (RP-Tu-1, RP-Tu-2, RP-Tu-3, and RP-Tu-4) that have previously been used to successfully amplify the tuf gene from the A1T-rich genome of Rickettsia prowazekii (Syva¨nen et al. 1996). The primary PCR amplifications were carried out under standard conditions using 5 mg/ml of genomic DNA and 10–30 mM of the primers RP-Tu-1 and RP-Tu-2. Reamplifications with the nested primer pairs RP-Tu-1/RP-Tu4 and RP-Tu-2/RP-Tu-3 were carried out using 1 ml of a 1:300 or 1:600 dilution of the primary PCR product. The annealing temperatures were 548C (RP-Tu-1/RP-Tu4) and 568C (RP-Tu-1/RP-Tu-2 and RP-Tu-2/RP-Tu-3). PCR reactions were carried out separately for each species to avoid cross contamination. The PCR products were purified using the ‘Wizardy PCR Preps DNA purification System (Promega) directly from the PCR reaction mixture or after separation on 1% agarose gels according to the instructions of the manufacturers. The double-stranded Buchnera PCR products were sequenced by an autocycle sequencing protocol using the Thermo Sequenase fluorescent labelled primer cycle sequencing kit (Amersham). In each reaction, 0.5 mg of PCR product and 10–20 pmol of fluorescent primers were used. The products of the sequencing reactions were separated and analyzed on an A.L.F sequencer (Pharmacia Biotech, Norden). Sequence Analysis The alignment of the tuf genes was performed using CLUSTAL W (Thompson, Higgins, and Gibson 1994). Pairwise distances for replacement substitutions (nucleotide substitutions that result in an amino acid change) and for silent substitutions (nucleotide substitutions that do not result in an amino acid change) were calculated as KA and Ks, respectively, using Li’s (1993) method. Phylogenetic relationships among the aligned sequences were estimated by maximum parsimony, and by the neighbor-joining method (Saitou and Nei 1987) using the interface program Phylopwin (Galtier, Gouy, and Gautier 1996). To assess the level of statistical confidence in the tree, 500 bootstraps were performed. The nucleotide sequences reported in this paper are available in the EMBL, GenBank, and DDBJ nucleotide sequence databases under the following accession numbers: Buchnera (Ap) Y12307; Buchnera (Sg), Y12308; Buchnera (Pb), Y12309; Buchnera (Mr), Y12310; and Buchnera (Sc), Y12311. Accession numbers for tuf gene sequences of free-living bacteria used in the analyses are Haemophilus influenzae, U32692; Escherichia coli, M10459; and Salmonella typhimurium, X55116. Accession numbers for the trpB gene sequences are Buchnera (Ap), L46355; Buchnera (Mr), L46357; Buchnera (Sg), Z19055, Buchnera (Sc), U09185; E. coli, V00365; and S. typhimurium, J01810. Results Alignments and Trees The nucleotide sequences corresponding to the E. coli amino acid residues 19–383 of elongation factor Tu

575

(EF-Tu) were determined for five different Buchnera species. These five EF-Tu sequences are aligned in figure 1 with the corresponding sequences from two other gamma Proteobacteria, E. coli and H. influenzae. The observed substitutions within the tuf gene of these gamma Proteobacteria appear to be more or less evenly distributed along the gene. It has previously been suggested that there may be an increased rate of fixation of deleterious mutations within the Buchnera lineages (Moran 1996). Out of the 364 aligned positions, there are 38 sites in total that are conserved between E. coli and H. influenzae but that differ in one or more of the Buchnera lineages. Of these, 25 correspond to conservative replacements, i.e., to substitutions expected to have little or no effect on protein function. We utilized two different methods, distance neighbor-joining and parsimony, to establish the phylogenetic relationships among the various Buchnera EF-Tu sequences (fig. 2). Both methods generated unrooted trees with similar topologies. The inferred tree separates endosymbionts of the Aphididae (Buchnera (Ap) and Buchnera (Sg)) from endosymbionts of the Pemphigidae (Buchnera (Pb), Buchnera (Mr) and Buchnera (Sc)). Furthermore, endosymbionts of the tribe Melaphidini (Buchnera (Mr) and Buchnera (Sc)) are sister taxa. This phylogeny matches the tree topology previously obtained with small-subunit ribosomal RNA sequences from the Buchnera endosymbionts as well as the proposed phylogeny of their aphid hosts (Heie 1987; Munson, Baumann, and Kinsey 1991; Moran et al. 1993; Moran and Telang 1998). Base Composition Buchnera has a genomic G1C content of 28%– 31% (Ishikawa 1987; Baumann et al. 1995). In order to determine whether the bias toward A1T has affected the amino acid content of EF-Tu, we examined the overall base composition pattern of the tuf gene (table 1). We observed that the nucleotide sequences of the tuf gene in each of the five species reflect the A1T mutation bias, particularly at synonymous third codon positions (GC3S). In Buchnera, the tuf genes have GC3S values ranging from 10% (Buchnera (Sg)) to 14% (Buchnera (Ap)). Similarly, GC3s values for the trpB gene range from 7% (Buchnera (Sg)) to 15% (Buchnera (Sc)). These values appear to be no different from those observed for other genes in Buchnera (unpublished data). In contrast, there is a clear difference in amino acid composition patterns of the tuf and trpB gene products (table 1). For EF-Tu, the average ratio of amino acids coded by A1T-rich codons over those coded by G1Crich codons (AT/GC) is 1.03 in Buchnera, which can be compared to an AT/GC value of 0.95 in E. coli (table 1). The corresponding ratio for the trpB gene product is 1.27, compared with 0.68 for E. coli (table 1). The values for E. coli are presumed to reflect the optimal amino acid compositions for these gene products, since the large population sizes of this taxon result in highly efficient selection. It has previously been shown that genes are resistant to biased mutation pressures to an extent

576

Brynnel et al.

FIG. 1.—Alignment of elongation factor Tu sequences from Buchnera (Sg) (BSg), Buchnera (Ap) (BAp), Buchnera (Pb) (BPb), Buchnera (Mr) (BMr), Buchnera (Sc) (BSc), Escherichia coli (Eco), and Haemophilus influenzae (Hin). Symbols beneath the aligned sequences indicate (*) identical residues and (.) sites with conservative replacements only.

Substitution Rates in Buchnera

577

Table 1 Base Frequencies for the tuf and trpB Genes in Buchnera and Escherichia coli Speciesa

G1Cb (tuf)

G1Cb (trpB)

GC3Sc GC3Sc AT/GCd AT/GCd (tuf) (trpB) (tuf) (trpB)

Buchnera (Sg) . . . Buchnera (Ap) . . Buchnera (Pb) . . . Buchnera (Mr) . . Buchnera (Sc) . . . Average . . . . . . . .

0.36 0.37 0.37 0.37 0.37 0.37

0.33 0.36 NDe 0.36 0.36 0.35

0.10 0.14 0.13 0.14 0.13 0.13

0.07 0.13 ND 0.12 0.15 0.12

1.01 1.07 1.04 1.04 1.01 1.03

1.21 1.43 ND 1.20 1.25 1.27

E. coli . . . . . . . . . 0.54

0.57

0.56

0.57

0.95

0.68

a Sg 5 Schizaphis graminum, Ap 5 Acyrthosiphon pisum, Pb 5 Pemphigus betae, Mr 5 Melaphis rhois, Sc 5 Schlechtendalia chinensis. b Gene G1C content. c G1C content at silent third codon positions. d Ratio of amino acids coded for by A1T-rich codons (Asn, Ile, Lys, Phe, Tyr) over amino acids coded for by G1C-rich codons (Ala, Gly, Pro). e Data not available.

FIG. 2.—Phylogenetic relationships of representative aphid endosymbionts, derived from tuf protein sequences. Neighbor-joining and maximum-parsimony methods gave identical topologies. Branch lengths are proportional to those reconstructed under neighbor-joining. Values at nodes are bootstrap values (neighbor-joining, parsimony) indicating the degree of support for individual clusters for each method. Letters at nodes (A–C) refer to estimated divergence times (table 4).

that is correlated with their overall degree of sequence conservation (Andersson and Sharp 1996). Here, we observe that the tuf gene product, which is one of the most highly conserved bacterial proteins known (Iwabe et al. 1989; Cammarano et al. 1992; Creti et al. 1994; Andersson and Sharp 1996) is much less affected by the mutation bias than is the trpB gene product (table 1). Therefore, we propose that the minor enrichment for amino acids coded by A1T-rich codons in elongation factor Tu in Buchnera is the result of an interplay between directional mutation pressures driving base composition patterns toward A1T-rich codons and counterselection acting to preserve a functionally optimal amino acid sequence for EF-Tu. Relative-Rate Tests Rates of replacement substitutions are 1.5- to 3-fold greater for Buchnera trp sequences than for homologous sequences from free-living relatives such as E. coli and S. typhimurium (Moran 1996). In order to determine whether the tuf gene is similarly evolving at a faster rate in the endosymbiont species, relative-rate tests were performed. In these tests, E. coli was chosen as the most closely related free-living species, and H. influenzae was chosen as a more distantly related species. Since synonymous sites are saturated even in comparisons between Buchnera and its closest relatives (e.g., E. coli), our analysis is based exclusively on nondegenerate sites

(at which any base substitution results in an amino acid change). The results suggest that the rates of substitution at nondegenerate sites are, on average, 2.25-fold greater in Buchnera lineages than in E. coli or S. typhimurium lineages (table 2). The increased rate of sequence evolution was apparent in all five Buchnera lineages. The smallest increase was observed in Buchnera (Sg), with rates increased by factors of 1.84 and 1.68 relative to the E. coli and S. typhimurium lineages. The largest increase was observed in Buchnera (Mr), with ratios of 2.82 and 2.41. The rate of evolution of the tuf gene in the lineage leading to Buchnera (Sg) is 1.84-fold faster than that in the lineage leading to E. coli (table 2). This value is in good agreement with a corresponding value of 1.82 averaged over five genes coding for tryptophan biosynthetic enzymes (trpA, trpB, trpC, trpD, and trpE) (Moran 1996). Data discussed below suggest that the tuf gene is evolving at a significantly slower rate than are the trp genes. Therefore, the relative increase in the substitution rates within genes in Buchnera, compared with those in E. coli is independent of the absolute rate of sequence evolution in the tuf and trp gene families. Synonymous and Nonsynonymous Substitutions In all pairwise comparisons, KA values for the tuf genes were found to be significantly lower, by a factor of 10–40, than Ks values (table 3). That the tuf gene is one of the more slowly evolving protein-coding genes (Iwabe et al. 1989; Cammarano et al. 1992; Creti et al. 1994; Andersson and Sharp 1996) is consistent with the observation that the nonsynonymous substitution rate for the tuf gene is about fourfold lower than that for the trpB gene (table 3). In contrast, the KS values for the tuf and the trp genes differ by less than a factor of 1.5. Absolute Rates of Nucleotide Substitution The dates for the nodes (A, B, C) describing the divergence of the different endosymbionts of Buchnera have previously been estimated using fossil information and biogeographic inference (Heie 1985, 1987; Moran 1989). These dates were used to calculate minimum and

578

Brynnel et al.

Table 2 Relative-Rate Tests for Substitutions at Nondegenerate Sites in the tuf Genes of Buchnera Versus Free-Living Relatives Taxon 1a

Taxon 3

K12b

K13

K23

K13 2 K23 (SD)

zc

K01/ K02d

Haemophilus influenzae H. influenzae H. influenzae H. influenzae H. influenzae H. influenzae H. influenzae H. influenzae H. influenzae H. influenzae

0.046 0.048 0.057 0.056 0.057 0.056 0.060 0.060 0.051 0.052

0.060

0.046 0.048 0.046 0.048 0.046 0.048 0.046 0.048 0.046 0.048

0.014 (0.0084) 0.012 (0.0085) 0.024 (0.0093) 0.022 (0.0093) 0.024 (0.0094) 0.023 (0.0093) 0.029 (0.0097) 0.027 (0.0096) 0.019 (0.0089) 0.018 (0.0089)

1.62 1.43 2.55** 2.41** 2.59** 2.45** 2.97** 2.83** 2.15* 1.97*

1.84 1.68 2.44 2.32 2.48 2.35 2.82 2.41 2.20 2.02

Taxon 2

Buchnera (Sg). . . . Escherichia coli Salmonella typhimurium Buchnera (Ap) . . . E. coli S. typhimurium Buchnera (Pb). . . . E. coli S. typhimurium Buchnera (Mr) . . . E. coli S. typhimurium Buchnera (Sc). . . . E. coli S. typhimurium

0.070 0.071 0.075 0.065

a In each test, taxon 1 is the endosymbiont, taxon 2 is a related free-living bacterium, taxon 3 is a more distantly related reference taxon, and taxon 0 is the most recent common ancestor of taxa 1 and 2. See footnote to table 1 for abbreviations. b K is the pairwise distance at nondegenerate sites between taxon i and taxon j. ij c z scores were calculated as described (Muse and Weir 1992). Probabilities for one-tailed test (H :K 0 01 # K02) are * P , 0.05 and ** P , 0.01. d K 01 and K02 were calculated as K01 5 (K13 2 K23 1 K12)/2; K02 5 K12 2 K01 (Wu and Li 1985).

maximum rates of evolution for the tuf gene along the branches descending from ancestors A, B, and C (table 4). We obtain estimates ranging from 1.3 3 10210 to 2.5 3 10210 nonsynonymous substitutions per position per year, and from 3.9 3 1029 to 8.0 3 1029 synonymous substitutions per position per year. The rates of sequence evolution for the tuf genes of free-living bacteria such as E. coli and S. typhimurium are much lower. For these organisms, which are thought to have diverged between 120 and 160 MYA (Ochman and Wilson 1987), the approximate rate of synonymous substitutions per position per year is 2.5 3 10210 to 3.3 3 10210, and the substitution rate at nonsynonymous positions is not measurable (i.e., less than 2.3 3 10211) (table 5). Thus, at both synonymous and nonsynonymous sites, the nucleotide sequence of the tuf gene is evolving at a more than 10-fold higher rate in Buchnera than in E. coli or S. typhimurium.

irreversible decline in fitness. This process is referred to as Muller’s ratchet (Muller 1964; Felsenstein 1974). Large bacterial populations driven by purifying selection are able to escape the irreversible effects of the ratchet. Recombination or sex is an efficient way of producing progeny with a reduced number of deleterious mutations in populations of small sizes (Muller 1964; Maynard Smith 1978; Pamilo, Nei, and Li 1987; Charlesworth, Charlesworth, and Morgan 1993; Lynch, Conery, and Bu¨rger 1995). However, organelles and obligate intracellular bacteria are effectively asexual, and they experience recurrent bottlenecks during transmission from one host to the other. Intracellularly replicating genomes are therefore expected to be particularly vulnerable to the effects of Muller’s ratchet, which will be expressed as increased fixation rates for mutations (Kurland 1992; Andersson and Kurland 1995; Moran 1996; Lynch 1996, 1997). The present data show that the rates of evolution for the tuf genes of the Buchnera are much faster than those of their homologs in free-living bacteria such as E. coli and S. typhimurium (Sharp and Li 1986, 1987).

Discussion The gradual accumulation of deleterious mutations in small asexual populations is expected to result in an

Table 3 Substitutions at Synonymous (KS) and Nonsynonymous (KA) Sites for the tuf and trpB Genes TRPB

TUF

COMPARISONa Buchnera (Sg)–Buchnera (Ap). . . . . . . . . . . . . . . Buchnera (Mr)–Buchnera (Sc). . . . . . . . . . . . . . . Buchnera (Pb)–Buchnera (Mr). . . . . . . . . . . . . . . Buchnera (Pb)–Buchnera (Sc) . . . . . . . . . . . . . . . Buchnera (Sg)–Buchnera (Pb) . . . . . . . . . . . . . . . Buchnera (Sg)–Buchnera (Mr). . . . . . . . . . . . . . . Buchnera (Sg)–Buchnera (Sc) . . . . . . . . . . . . . . . Buchnera (Ap)–Buchnera (Pb). . . . . . . . . . . . . . . Buchnera (Ap)–Buchnera (Mr) . . . . . . . . . . . . . . Buchnera (Ap)–Buchnera (Sc). . . . . . . . . . . . . . . Escherichia coli–Salmonella typhimurium . . . . . E. coli–Haemophilus influenzae . . . . . . . . . . . . . . S. typhimurium–H. influenzae. . . . . . . . . . . . . . . . a b

See footnote to table 1 for abbreviations. Data not available.

KS 0.74 0.61 0.56 1.15 1.41 1.83 1.38 1.93 1.30 1.47 0.08 1.11 1.27

6 6 6 6 6 6 6 6 6 6 6 6 6

0.11 0.07 0.08 0.23 0.20 1.17 0.19 1.26 0.29 0.22 0.02 0.13 0.18

KA

KS

KA

0.015 6 0.004 0.026 6 0.006 0.033 6 0.007 0.044 6 0.008 0.056 6 0.008 0.047 6 0.008 0.051 6 0.008 0.052 6 0.008 0.049 6 0.008 0.058 6 0.008 ,0.003 0.047 6 0.008 0.049 6 0.008

0.93 6 0.21 0.82 6 0.13 NDb ND ND 1.24 6 0.22 1.72 6 0.40 ND 1.48 6 0.29 3.48 6 3.87 0.99 6 0.14 ND ND

0.093 6 0.014 0.096 6 0.015 ND ND ND 0.178 6 0.020 0.184 6 0.020 ND 0.169 6 0.019 0.168 6 0.019 0.012 6 0.005 ND ND

Substitution Rates in Buchnera

Table 4 Rate Estimates Based on Substitution Frequencies in the tuf Gene of Buchnera SUBSTITUTIONS/POSITION/YEAR ESTIMATED AGE (Myr) ANCESTORa A............ B............ C............ Average (tuf) . . a

Mini- Maximum mum 30 80 48

80 160 70

NONSYNONYMOUS 210

(310

)

Maxi- Minimum mum

Table 5 Absolute Substitution Rates at Synonymous and Nonsynonymous Sites in the tuf and trpB Genes of Buchnera and of Escherichia coli and Salmonella typhimurium RATEa

SYNONYMOUS (310210) Maximum

Minimum

2.5 2.7 2.4

0.9 1.9 1.2

123 53.4 63.5

46.2 26.7 43.6

2.5

1.3

80.0

38.8

A, B, and C refer to the nodes in figure 2.

579

COMPARISON

TUF

(31029) RATEa

TRPB

(31029)

Synony- Nonsyn- Synon- Nonsynmous onymous ymous onymous

Substitution rate Buchnera . . . . . . . . . . . . . . 4–8 E. coli–S. typhimurium. . . 0.2–0.3 Mutation rateb E. coli–S. typhimurium. . .

4–5

Ratio

1.0c

0.1–0.2 ,0.02

6–12 3–4

0.6–1.3 0.04–0.05

9–12 .10.0d

1.0c

.10.0d

a

In this respect, the data are consistent with our expectations that intracellularly replicating genomes such as those of organelles, endosymbionts, and obligate intracellular parasites experience elevated fixation rates for mutations due to the combined effects of low recombination frequencies and small effective population sizes (Kurland 1992; Andersson and Kurland 1995; Lynch 1996, 1997; Moran 1996). Although we expect the faster evolutionary rates to be driven by the fixation of deleterious mutations, some of the substitutions may involve compensatory mutations that are fixed during intervals of greater effective selection due to temporary increases either in population size or in selective pressures. In fact, the occasional incidence of compensatory mutations is indicated by substitutions that restore base-pairing within stems of 16S rRNA structures in Buchnera and other endosymbionts (unpublished data). However, in the absence of recombination, deletions of genetic material will be difficult or impossible to restore. It may therefore not be surprising that organelles and intracellular parasites have severely reduced genome sizes (Andersson and Kurland 1995). The finding that a gene, metK, is subjected to mutational meltdown in several different Rickettsia lineages independent of each other suggests that the accumulation of mutations is an ongoing process, at least in Rickettsia (unpublished data). In this study, we have shown that there is a much higher (10- to 40-fold) rate of synonymous codon substitution rate in tuf genes of Buchnera lineages than is observed in tuf genes of commonly studied free-living bacteria (table 5). This result requires more attention, because it suggests that for some genes in some organisms, synonymous substitutions are not neutral, while for homologous genes in other organisms, they are effectively neutral. This means that synonymous substitution rates in organisms such as Buchnera and E. coli may not be directly comparable, as was assumed in a previous study (Moran 1996). Indeed, if there is a strong codon preference for a gene, even synonymous substitutions will be under selective pressure (Sharp and Li 1986, 1987). Thus, Ks for the tuf gene is only 0.08 for the taxon pair E. coli and S. typhimurium, while the corresponding average value for the trpB gene is 1.0 (table 3). This marked difference in divergence at synonymous sites can be attributed to

Number of substitutions per position per year. The mutation rate constants for E. coli and S. typhimurium have been taken from Berg and Martelius (1995), and the divergence time for E. coli and S. typhimurium has been taken from Ochman and Wilson (1987). c The observed synonymous substitution rate for Buchnera relative to the estimated mutation rates for E. coli and S. typhimurium. d The observed nonsynonymous substitution rate for Buchnera relative to the observed nonsynonymous substitution rates for E. coli and S. typhimurium. b

the fact that the highly expressed tuf genes in free-living bacteria such as E. coli and S. typhimurium are subject to pronounced codon selection, while the trp genes are not (Sharp 1991). The selective pressure for codon preferences seems in these cases to be associated with the efficiency of the translational machinery, an interpretation that is supported by recently observed correlations between codon usage patterns in mRNA pools and tRNA isoacceptor frequencies under a variety of growth conditions (Dong, Nilsson, and Kurland 1996; Berg and Kurland 1997). Comparisons of synonymous substitution rates between endosymbionts and free-living species are also complicated by a technical problem which concerns the degree of saturation for synonymous divergences of the taxon pairs that were studied. Thus, many of the pairwise comparisons show synonymous substitution frequencies per position close to or above one per site. When there is strong selection for codon bias, and both species are biased toward the same subset of codons, the estimated Ks values may be reliable only when the true substitutions are lower than one per site (Li, Wu, and Luo 1985; Sharp and Li 1987; Sharp 1991). On the other hand, if there is no selection for codon bias, Ks should become saturated at a value closer to 1.5–2 substitutions per site (K. Wolfe, personal communication). When pairwise comparisons are made across the two groups of endosymbionts (i.e., between the Aphididae and the Melaphidini), they produce Ks values ranging from 1.1 to 1.9, and several of these estimates have very large standard errors (table 3). For these comparisons, synonymous substitutions are saturated. However, when the comparisons are made within the Aphididae and within the Melaphidini, the Ks values are in the range of 0.5–0.7 for the tuf gene (table 3), suggesting that synonymous divergences are not saturated for these pairs. However, because of the strong nucleotide bias in Buchnera, synonymous substitution rates may be un-

580

Brynnel et al.

derestimated by Kimura’s two-parameter model. Indeed, the Ks values for transitional substitutions in twofold degenerate codon families for the species pair Buchnera (Mr-Sc) is estimated to be 0.25 using Kimura’s twoparameter model and 0.41 using Berg’s (1995) model, which adjusts for biases in nucleotide frequencies. A closer examination of substitution rates at two- and fourfold degenerate sites suggests that transitional substitutions may be approaching saturation in the species pairs Buchnera (Mr-Sc) and Buchnera (Ap-Sg). The pair Buchnera (Mr-Pb) is farthest away from saturation; for this pair, estimates of synonymous divergence should accurately reflect the time elapsed since they shared a common ancestor. In free-living organisms, even the genes with strong codon preferences are characterized by Ks values larger than the corresponding KA values. This means, in effect, that selective forces responsible for even the most extreme codon preferences at synonymous sites of highly expressed genes are relatively weak compared with those that maintain the functional structures of proteins at nonsynonymous sites. Therefore, selection for codon bias at synonymous sites is maximally expressed in large populations of free-living organisms with significant rates of sexual recombination. In contrast, in small asexual populations, there will be a strong tendency to fix random mutations. The result is, we believe, that the selective forces operating at synonymous sites cannot maintain codon preferences in asexual populations. Indeed, we have been unable to detect a comparable difference in codon preference for putative high-expression-level and low-expression-level genes from Buchnera (unpublished data). Furthermore, synonymous substitution rates are roughly similar for the tuf and the trpB genes in Buchnera (table 5). Accordingly, we suggest that the augmented synonymous substitution rates for the tuf gene in the Buchnera lineages are associated with the loss of selective codon preferences characteristic of free-living bacteria. We believe that the absence of strong codon preferences and the greatly enhanced synonymous substitution frequencies for the tuf genes of Buchnera are both consequences of the population structure of the endosymbionts. Indeed, the genomes of endocellular parasites such as the Rickettsia resemble those of endosymbionts such as the Buchnera in that they do not have a discernible codon preference (Andersson and Sharp 1996), and synonymous substitution rates appear to be roughly similar among genes expected to show low as well as high expression levels (unpublished data). Thus, synonymous substitutions in the tuf genes of Rickettsia and Buchnera apparently behave as neutral or nearly neutral substitutions. This implies that, with all other parameters equal, these synonymous substitution rates not only should be equivalent to those seen in the trp operons of the same organisms, but also should be equivalent to the intrinsic mutation rates of free-living organisms such as E. coli and S. typhimurium. This prediction is based on Kimura’s (1968) notion that the substitution rate for neutral mutations equals the mutation rate. An important assumption here is that mutation rates

on an absolute timescale are similar in symbiotic and free-living bacteria. If the mutation rate is dependent on numbers of replication events, this assumption would be affected by the relative generation times of the lineages giving rise to E. coli and Buchnera. Unfortunately, this ratio is not known; however, there is no reason to expect an overwhelming difference in average generation time. Buchnera divides about 50–200 times per year (based on about 12 divisions per aphid generation and 5–15 generations per year [Moran 1992; Baumann and Baumann 1994]), a number that is not unreasonable for enteric bacteria, which may multiply as rarely as once every 2 days in nature (Ochman and Wilson 1987). For the pairwise comparison between E. coli and S. typhimurium, the mutational divergence for synonymous transitions from T to C and from A to G in twofold degenerate codons has been estimated to be 0.2– 0.5 mutations per site, with highly expressed genes having a lower mutation divergence than genes that are not highly expressed (Berg and Martelius 1995). These estimates are based on a systematic analysis of the relationship between substitution rate constants and selection pressure using a kinetic model that takes selection bias into account (Berg and Martelius 1995). The mutational divergence for transitions from T to C and from A to G in the tuf genes of Buchnera (Pb) and Buchnera (Mr) has been estimated at 0.046 mutations per site, and the reverse rate is about sixfold higher, 0.29 mutations per site, because of the direction of the mutation bias. The overall rate for transitions in Buchnera is 0.17 mutations per site based on Berg’s (1995) four-parameter model, compared with 0.13 mutations per site based on Kimura’s (1980) two-parameter model. By using a divergence time of 120–160 Myr for E. coli and S. typhimurium (Ochman and Wilson 1987), we estimate the mutation rate for transitions at synonymous sites for the tuf gene to be 1.2 3 1029 to 1.7 3 1029 mutations per site per year. These estimates are in perfect agreement with experimental results which suggest a mutation rate for transitions of 10211 or less per generation (D. Hughes, personal communication). The corresponding rate for the tuf gene in Buchnera is 1.0 3 1029 to 2.0 3 1029 mutations per site per year. Thus, the estimated mutation rates for transitions at synonymous sites are remarkably similar in these species. If we assume that the rate of transversions is approximately similar to the rate of transitions, the neutral mutation rate for the tuf gene in E. coli and S. tyhimurium can be estimated to be 3.6 3 1029 to 5.1 3 1029 substitutions per position per year, and that for the trpB gene can be estimated to be 9.4 3 1029 to 12.5 3 1029 substitutions per position per year (table 5). The absolute rate of substitutions at synonymous sites in Buchnera was found to be 3.9 3 1029 to 8.0 3 1029 substitutions per position per year for the tuf gene and 5.8 3 1029 to 12.0 3 1029 substitutions per position per year for the trpB gene (table 5). This suggests that the intrinsic mutation rates for the endosymbionts (Buchnera) are very similar to the intrinsic mutation rates for their closely related free-living species (E. coli and S. typhimurium).

Substitution Rates in Buchnera

In contrast, there is a 10-fold or greater difference in the absolute substitution rates at nonsynonymous sites for the tuf and the trpB genes for the same set of species (table 5). Taken together, our data suggest that the rates of mutational fixation at both synonymous and nonsynonymous sites are highly augmented in the endosymbiotic lineages. We suggest that these observations are expressions of one and the same phenomenon, namely Muller’s ratchet operating on small endocellular populations with low recombination frequencies (Muller 1964; Felsenstein 1974; Ohta 1987; Lynch 1996). By using the estimates for rates of nonsynonymous substitutions in the tuf gene for Buchnera and assuming approximate rate constancy along the branches, we can estimate the age of the most recent ancestor of all aphid endosymbiont species; this corresponds to the node separating ancestors A and B in figure 2. Accordingly, we calculate a divergence time of 100–200 MYA. This dating is in excellent agreement with the rRNA-based dating, which suggests that the minimum age of the endosymbiotic association is 100–250 MYA (Moran et al. 1993). Because of its slow evolutionary rate, the tuf gene, along with rRNA, may be a particularly useful gene for phylogenetic studies of old groups. In fact, the tuf gene has a potential advantage over the rRNA gene. By studying codon frequencies and synonymous substitution frequencies in tuf lineages of a taxon group, it may be possible to infer something about the population structures of these taxa. Acknowledgements We thank Otto Berg and Kenneth Wolfe for helpful discussions; Paul Baumann for preparation of genomic DNA; and Kristina Na¨slund, Ann-Sofie Eriksson, and Morten Andersen for technical assistance. This work was supported by grants from the Natural Sciences Research Council, the Swedish Cancer Society, and the Knut and Alice Wallenberg Foundation. LITERATURE CITED

AKSOY, S., A. A. POURHOSSEINI, and A. CHOW. 1995. Mycetosome endosymbionts of tsetse flies constitute a distinct lineage related to Enterobacteriaceae. Insect Mol. Biol. 4: 15–22. ANDERSSON, S. G. E., and C. G. KURLAND. 1995. Genomic evolution drives the evolution of the translation system. Biochem. Cell Biol. 73:775–787. ANDERSSON, S. G. E., and P. M. SHARP. 1996. Codon usage and base composition in Rickettsia prowazekii. J. Mol. Evol. 42:525–536. BALDAUF, S. L., J. P. PALMER, and W. F. DOOLITTLE. 1996. The root of the universal tree and the origin of eukaryotes based on elongation factor phylogeny. Proc. Natl. Acad. Sci. USA 93:7749–7754. BANDI, C., M. SIRONI, G. DAMIANI, L. MAGRASSI, C. A. NALEPA, U. LAUDANI, and L. SACCHI. 1995. The establishment of intracellular symbiosis in an ancestor of cockroaches and termites. Proc. R. Soc. Lond. Biol. Sci. 259:293–299. BAUMANN, L., and P. BAUMANN. 1994. Growth kinetics of the endosymbiont Buchnera aphidicola in the aphid Schizaphis gramium. Appl. Environ. Microbiol. 60:3440–3443.

581

BAUMANN, P., L. BAUMANN, C. Y. LAI, and D. ROUHBAKSH. 1995. Genetics, physiology, and evolutionary relationships of the genus Buchnera: intracellular symbionts of aphids. Annu. Rev. Microbiol. 49:55–94. BERG, O. G. 1995. Kinetics of synonymous codon change for an amino acid of arbitrary degeneracy. J. Mol. Evol. 41: 345–352. BERG, O. G., and C. G. KURLAND. 1997. Growth rate-optimised tRNA abundance and codon usage. J. Mol. Biol. 270: 544–550. BERG, O. G., and M. MARTELIUS. 1995. Synonymous substitution-rate constants in Escherichia coli and Salmonella typhimurium and their relationship to gene expression and selection pressure. J. Mol. Evol. 41:449–456. BUCHNER, P. 1965. Endosymbiosis of animals with plant microorganisms. Interscience Publishers, New York. CAMMARANO, P., P. PALM, R. CRETI, E. CECCARELLI, A. M. SANANGELANTONI, and O. TIBONI. 1992. Early evolutionary relationships among known life forms inferred from elongation factor EF-2/EF-G sequences: phylogenetic coherence and structure of the archaeal domain. J. Mol. Evol. 34:396– 405. CHARLESWORTH, B., D. CHARLESWORTH, and M. T. MORGAN. 1993. Genetic loads and estimate of mutation rates in highly inbred plant populations. Nature 347:380–382. CLARK, M. A., L. BAUMANN, M. A. MUNSON, P. BAUMANN, B. C. CAMPBELL, J. E. DUFFUS, L. S. OSBORNE, and N. A. MORAN. 1992. The eubacterial endosymbionts of whiteflies (Homoptera: Aleyrodoidea) constitute a lineage distinct from the endosymbionts of aphids and mealybugs. Curr. Microbiol. 25:119–123. CRETI, R., E. CECCARELLI, M. BOCCHETTA, A. M. SANANGELANTONI, O. TIBONI, P. PALM, and P. CAMMARANO. 1994. Evolution of translational elongation factor (EF) sequences: reliability of global phylogenies inferred from EF-1alph(Tu) and EF-2(G) proteins. Proc. Natl. Acad. Sci. USA 91:3255– 3259. DONG, H., L. NILSSON, and C. G. KURLAND. 1996. Co-variation of tRNA abundance and codon usage in Escherichia coli at different growth rates. J. Mol. Biol. 260:649–663. FELSENSTEIN, J. 1974. The evolutionary advantage of recombination. Genetics 78:157–159. GALTIER, N., M. GOUY, and C. GAUTIER. 1996. SEAVIEW and PHYLOpWIN: two graphic tools for sequence alignment and molecular phylogeny. CABIOS 12:543–548. HEIE, O. E. 1985. Fossil aphids, a catalogue of fossil aphids with notes on systematics and evolution. Pp. 101–133 in H. SZELEGIEWICZ, ed. Evolution and biosystematics of aphids. Polska Acad. Nauk, Ossolineum, Warsaw. . 1987. Paleontology and phylogeny. Pp. 367–391 in A. K. MINSK and P. HARREWIJN, eds. Aphids: their biology, natural enemies and control. World crop pests. Elsevier, Amsterdam. ISHIKAWA, H. 1987. Nucleotide composition and kinetic complexity of the genomic DNA of an intracellular symbiont of the pea aphid Acyrthosiphon pisum. J. Mol. Evol. 24: 205–211. IWABE, N., K.-I. KUMA, M. HASEGAWA, S. OSAWA, and M. TAKASHI. 1989. Evolutionary relationship of archaebacteria, eubacteria, and eukaryotes inferred from phylogenetic trees of duplicated genes. Proc. Natl. Acad. Sci. USA 86:9355– 9359. KIMURA, M. 1968. Evolutionary rate at the molecular level. Nature 217:624–626. . 1980. A simple method for estimating evolutionary rates of base substitutions through comparative studies of nucleotide sequences. J. Mol. Evol. 16:111–120.

582

Brynnel et al.

KURLAND, C. G. 1992. Evolution of mitochondrial genomes and the genetic code. BioEssays 14:709–714. LI, W.-H. 1993. Unbiased estimation of the rates of synonymous and nonsynonymous substitution. J. Mol. Evol. 36: 96–99. LI, W.-H., C. I. WU, and C.-C. LUO. 1985. A new method for estimating synonymous and nonsynonymous rates of nucleotide substitution considering the relative likelihood of nucleotide and codon changes. Mol. Biol. Evol. 2:150–174. LYNCH, M. 1996. Mutation accumulation in transfer RNAs: molecular evidence for Muller’s ratchet in mitochondrial genomes. Mol. Biol. Evol. 13:209–220. . 1997. Mutation accumulation in nuclear, organelle, and prokaryotic transfer RNA genes. Mol. Biol. Evol. 14: 914–925. LYNCH, M., J. CONERY, and R. BU¨RGER. 1995. Mutational meltdowns in sexual populations. Evolution 49:1067–1080. MAYNARD SMITH, J. 1978. The evolution of sex. Cambridge University Press, Cambridge, U.K. MORAN, N. 1989. A 48-million-year-old aphid-host plant association and complex life cycle: biogeographic evidence. Science 245:173–175. . 1992. The evolution of life cycles in aphids. Annu. Rev. Entomol. 37:321–348. . 1996. Accelerated evolution and Muller’s ratchet in endosymbiotic bacteria. Proc. Natl. Acad. Sci. USA 93: 2873–2878. MORAN, N. A., M. A. MUNSON, P. BAUMANN, and H. ISHIKAWA. 1993. A molecular clock in endosymbiotic bacteria is calibrated using the insect hosts. Proc. R. Soc. Lond. Biol. Sci. 253:167–171. MORAN, N. A., and A. TELANG. 1998. Bacteriocyte associated endosymbionts of insects. Bioscience (in press). MULLER, H. J. 1964. The relation of recombination to mutational advance. Mutat. Res. 1:2–9. MUNSON, M. A., P. BAUMANN, M. A. CLARK, L. BAUMANN, N. A. MORAN, D. J. VOEGTLIN, and B. C. CAMPBELL. 1991. Evidence for the establishment of aphid-eubacterium endosymbiosis in an ancestor of four aphid families. J. Bacteriol. 173:6321–6324. MUNSON, M. A., P. BAUMANN, and M. G. KINSEY. 1991. Buchnera gen. nov. and Buchnera aphidicola sp nov., a taxon consisting of the mycetocyte-associated primary endosymbionts of aphids. Int. J. Syst. Bacteriol. 41:566–568.

MUNSON, M. A., P. BAUMANN, and N. A. MORAN. 1992. Phylogenetic relationships of the endosymbionts of mealybugs (Homoptera: Pseudococcidae) based on 16S rRNA sequences. Mol. Phylogenet. Evol. 1:26–30. MUSE, S. V., and B. S. WEIR. 1992. Testing for equality of evolutionary rates. Genetics 132:269–275. OCHMAN, H., and A. C. WILSON. 1987. Evolutionary history of enteric bacteria. Pp. 1649–1654 in F. NEIDHARDT, ed. Escherichia coli and Salmonella typhimurium: cellular and molecular biology. American Society for Microbiology, Washington, D.C. OHTA, T. 1987. Very slightly deleterious mutations and the molecular clock. J. Mol. Evol. 26:1–6. PAMILO, P., N. NEI, and W.-H. LI. 1987. Accumulation of mutations in sexual and asexual populations. Genet. Res. 49: 135–146. SAITOU, N., and M. NEI. 1987. The neighbor-joining method: a new method for reconstructing phylogenetic trees. Mol. Biol. Evol. 4:406–425. SHARP, P. M. 1991. Determinants of DNA sequence divergence between Escherichia coli and Salmonella typhimurium: codon usage, map position and concerted evolution. J. Mol. Evol. 33:23–33. SHARP, P. M., and W.-H. LI. 1986. An evolutionary perspective on synonymous codon usage in unicellular organisms. J. Mol. Evol. 24:28–38. . 1987. The rate of synonymous substitution in enterobacterial genes is inversely related to codon usage bias. Mol. Biol. Evol. 4:222–230. SYVA¨NEN, A.-C., H. AMIRI, A. JAMAL, S. G. E. ANDERSSON, and C. G. KURLAND. 1996. A chimeric disposition of the elongation factor genes in Rickettsia prowazekii. J. Bacteriol. 178:6192–6199. THOMPSON, J. D., D. G. HIGGINS, and T. J. GIBSON. 1994. CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Res. 22:4673–4680. WU, C.-I., and W.-H. LI. 1995. Evidence for higher rates of nucleotide substitutions in rodents than in man. Proc. Natl. Acad. Sci. USA 82:1741–1745.

MANOLO GOUY, reviewing editor Accepted February 10, 1998