Evidence for Genetic Drift in Endosymbionts (Buchnera): Analyses of
Protein-Coding Genes
J. J. Wernegreen and N. A. Moran
Department of Ecology and Evolutionary Biology, University of Arizona
Buchnera, the bacterial endosymbionts of aphids, undergo severe population bottlenecks during maternal transmission through their hosts. Previous studies suggest an increased effect of drift within these strictly asexual, small
populations, resulting in an increased fixation of slightly deleterious mutations. This study further explores sequence
evolution in Buchnera using three approaches. First, patterns of codon usage were compared across several homologous Escherichia coli and Buchnera loci, in order to test the prediction that selection for the use of optimal
codons is less effective in small populations. A x2-based measure of codon bias was developed to adjust for the
overall A1T richness of silent positions in the endosymbionts. In contrast to E. coli homologues, adaptive codon
bias across Buchnera loci is markedly low, and patterns of codon usage lack a strong relationship with gene
expression level. These data suggest that codon usage in Buchnera has been shaped largely by mutational pressure
and drift rather than by selection for translational efficiency. One exception to the overall lack of bias is groEL,
which is known to be constitutively overexpressed in Buchnera and other endosymbionts. Second, relative-rate tests
show elevated rates of sequence evolution of numerous protein-coding loci across Buchnera, compared to E. coli.
Finally, consistently higher ratios of nonsynonymous to synonymous substitutions in Buchnera loci relative to the
enteric bacteria strongly suggest the accumulation of nonsynonymous substitutions in endosymbiont lineages. Combined, these results suggest a decreased effectiveness of purifying selection in purging endosymbiont populations
of slightly deleterious mutations, particularly those affecting codon usage and amino acid identity.
Introduction
The rate of fixation of mutations with fitness consequences depends not only on the strength of selection
for or against them but also on the effectiveness of such
selection as influenced by effective population size. In
populations with low rates of recombination and small
effective sizes, slightly deleterious mutations may experience increased rates of fixation through drift (Ohta
1973). This predicted relationship between population
structure and rate of fixation of slightly deleterious mutations can be tested among prokaryotes. Free-living
bacteria are thought to have large effective population
sizes (Selander, Caugant, and Whittam 1987), and even
clonal groups experience recombination that is important in their evolutionary dynamics (Maynard Smith,
Dowson, and Spratt 1991; Dykhuizen and Green 1993;
Maynard Smith et al. 1993).
In contrast, endosymbiotic bacteria associated with
several insect groups have relatively small effective population sizes and have restricted opportunities for interstrain recombination because of their mode of transmission. Bacteria associated with specialized insect cells
(i.e., mycetocytes) are maternally transmitted by the infection of ovaries or of internally developing embryos
(reviewed in Buchner 1965; Moran and Baumann 1994;
Baumann et al. 1995). The effective population size of
the bacteria is reduced by the bottleneck at each inoculation of progeny, where relatively few bacteria are
Abbreviations: CAI5 Codon Adaptation Index; Nc5 effective number of codons; GC35 percent G1C content at third-codon positions.
Key words: Buchnera, endosymbionts, codon bias, drift, population
size.
Address for correspondence and reprints: Jennifer Wernegreen, Department of Ecology and Evolutionary Biology, University of Arizona,
Biological Sciences West, Room 310, Tucson, Arizona 85721. E-mail:
[email protected].
Mol. Biol. Evol. 16(1):83–97. 1999
q 1999 by the Society for Molecular Biology and Evolution. ISSN: 0737-4038
transmitted (Hinde 1971; A. Mira, personal communication). In addition, modeling indicates that insect host
population sizes may be the primary determinant of the
effective population size of intracellular genomes (C.
Rispe, personal communication). Insect population sizes, while relatively large among animals, are much
smaller than those of free-living bacteria (reviewed in
Lambert and Moran 1998). Finally, any lateral gene
transfer among endosymbionts would be confined to the
bacterial genotypes present in the same host individual,
and the tight bottleneck at transmission implies that
these would be similar or identical. Buchnera, the endosymbionts of aphids, are particularly well characterized, and the perfect congruence between symbiont and
host phylogenies supports anatomical evidence for their
stable, vertical inheritance (Munson et al. 1991; Moran
and Baumann 1994). The goals of this study are to explore the effects of this strict asexuality and small population size on sequence evolution in Buchnera and to
test the hypothesis that Buchnera lineages experience
increased rates of substitution of slightly deleterious mutations.
Codon Bias
The use of alternative codons may be shaped by
biases in mutation rates among the four bases (Suoeka
1961; Muto and Osawa 1987), by selection for the use
of optimal codons to maximize rates and efficiency of
translation (Ikemura 1981, 1985), or by a combination
of these processes. Studies of codon usage often attempt
to distinguish the relative importance of genome nucleotide composition and selection for translational efficiency by testing alternative predictions of these models.
In cases where patterns of codon usage largely reflect
mutational pressure and drift rather than translational selection, codon bias is expected to correspond with local
base compositional biases. This pattern characterizes genomes of vertebrates and some bacterial genomes with
83
84
Wernegreen and Moran
strong A1T or G1C mutational biases (Sharp et al.
1988; Andersson and Sharp 1996). In contrast, an effect
of translational selection is evidenced by a positive relationship between the extent of codon bias and level of
gene expression. In Escherichia coli and in yeast, for
example, a correlation between the degree of codon bias
and the gene expression level (Gouy and Gautier 1982;
Sharp, Tuohy, and Mosurski 1986; Sharp et al. 1988) is
thought to reflect selection for the rapid translation of
highly expressed genes through the use of optimal codons and a lack of such selection on lowly expressed
genes and their retention of nonoptimal codons (Shields
1990). Among bacterial genomes with base compositional biases, patterns of codon usage often reflect a
combination of mutational pressure and translational selection (Shields and Sharp 1987; Ohtaka, Nakamura,
and Ishikawa 1992; Wright and Bibb 1992; Ohtaka and
Ishikawa 1993).
Population genetic models indicate that effective
population size influences the balance between the effects of any mutational bias and selection for optimal
codons (Li 1987; Bulmer 1991). In particular, translational selection will be effective in a haploid population
only when the selective coefficient (s) for a particular
codon is greater than 2/Ne (Li 1987). In large populations, relatively weak selection may produce strong codon biases. However, in small populations, weak translational selection is countered by a strong effect of drift,
which may allow the maintenance of nonoptimal codons. In addition, drift may also allow rare transitions
in coadapted states (Wright 1931), such as the links between codon frequencies and tRNA abundances (Shields
1990). Therefore, small, asexual populations are expected to exhibit lower overall levels of codon bias because of weakness of translational selection. Where they
do exhibit bias, the bias may favor synonyms other than
the optimal codons of related lineages because of
switches in codon preferences.
Relatively few empirical studies explicitly test the
effect of population size on the balance between these
two processes. One such study compares levels of codon
bias of chloroplast genes across several algal and angiosperm lineages (Morton 1998). The high levels of codon
bias in chloroplasts of most algal lineages, relative to
chloroplasts of flowering plants, supports the hypothesis
that selection for translational efficiency is more effective in large (algal) populations than in relatively small
(angiosperm) populations. Likewise, previous analyses
of protein-coding genes in Buchnera show a general
A1T richness at synonymous sites and a lack of strong
preferences for the optimal codons of the closely related
species E. coli (Clark, Baumann, and Baumann 1992;
Ohtaka and Ishikawa 1993; Brynnel et al. 1998; Clark,
Baumann, and Baumann 1998). These studies suggest
that codon usage in Buchnera largely reflects strong
A1T mutational bias and fixation of nonoptimal codons
through drift. However, previous estimates of codon bias
in Buchnera do not account for local base composition,
so it is difficult to identify any preferences for particular
codons, given the strong A1T bias at synonymous sites.
In the analysis presented in this study, a more extensive
sample of Buchnera and E. coli homologues are included to represent a wide range of gene expression levels,
and the A1T richness of the Buchnera genome (Ishikawa 1987) is considered in testing for codon bias.
Previous studies of codon usage in A1T– and
G1C–rich genomes highlight methods for assessing codon usage in genomes with strong mutational biases
(Shields and Sharp 1987; Ohtaka, Nakamura, and Ishikawa 1992; Wright and Bibb 1992; Ohtaka and Ishikawa
1993; Andersson and Sharp 1996). For example, the effective number of codons, Nc, is reduced by preferences
for particular codons or biased base composition. In order to test the null hypothesis that codons are used randomly except for the influence of local mutational bias,
expected values of Nc may be adjusted to account for
local base composition. In the A1T–rich Rickettsia genome, ‘‘Nc-plots’’ show an agreement between observed
Nc values and those expected, given the GC3, indicating
that codon usage reflects local base composition and
may therefore be attributed largely to mutational bias
(Andersson and Sharp 1996). Likewise, similar levels of
codon bias across Rickettsia genes with very different
expression levels indicate that mutational bias has a
stronger effect than translational selection. In other taxa,
the combined effects of mutational bias and translational
selection are apparent. Across several Streptomyces loci,
a strong effect of mutational bias is suggested by the
correspondence of GC3 and Nc and by a correlation between the GC3 of a locus and the locus’ position along
the major axis in correspondence analysis of codon usage (Wright and Bibb 1992). A slight effect of translational selection on the highly expressed Streptomyces tuf
gene is supported by the relatively low Nc for this locus,
its clear distinction from other loci in correspondence
analysis, and the fact that, apparently, preferred codons
in tuf are also preferred by another G1C–rich bacterium, Micrococcus luteus (Wright and Bibb 1992). This
combination of mutational bias and translational selection is also apparent for other genomes with mutational
biases, such as Micrococcus luteus (Ohtaka, Nakamura,
and Ishikawa 1992; Ohtaka and Ishikawa 1993), Dictyostelium discoideum (Sharp and Devine 1989), and
Bacillus subtilis (Shields and Sharp 1987). Organelle genomes may also show strong nucleotide biases. The relative importance of selection and genome composition
in shaping codon usage of several A1T–biased chloroplast genomes was recently tested by comparing an
observed CAI (Sharp and Li 1987), or bias toward a
pool of preferred codons (here, on the basis of a highly
expressed chloroplast gene), to an expected distribution
of CAIs based on genome-wide nucleotide composition
(Morton 1998).
The analyses above test the null hypothesis that
codon usage may be explained solely by local base composition. However, Nc plots, correspondence analysis,
and CAI estimates may fail to detect slight preferences
among synonyms, since these methods derive a single
estimate across all amino acids in a locus. In addition,
CAI estimates are possible only when the optimal codons for a particular genome are known. In this study,
estimates of codon bias across Buchnera loci are also
Evidence for Drift in Endosymbionts
adjusted for local base composition. However, in contrast to previous estimates, the x2-based method developed here tests for nonrandom-use codons for single
amino acids and does not require prior knowledge of
preferred codons. This approach may be generally applicable to other organisms in which codon preferences
may be absent or subtle, such as in taxa with small effective population sizes and/or strong mutational biases.
Similar to the scaled x2 (Shields et al. 1988), as modified
by Akashi and Shaeffer (1997) to adjust for A1T content at silent positions, we compared observed codon
frequencies to those expected if codon usage reflects local base composition at synonymous sites. By applying
this method to several homologous loci in Buchnera and
in their free-living relative, E. coli, we test the hypothesis that translational selection is relatively ineffective
in the endosymbionts, so that codon usage in Buchnera
is shaped by A1T mutational bias and by the fixation
of nonoptimal codons through drift.
Rates of Sequence Divergence
The decreased effectiveness of selection in small,
asexual populations is also expected to accelerate the
fixation of replacement substitutions. Previous studies
provide strong evidence for differences in rates and patterns of sequence evolution of endosymbiotic and freeliving bacterial lineages. For example, compared to freeliving relatives in the enterics, the 16S rRNA gene of
several endosymbiotic lineages has been shown to
evolve 1.5–2 times faster (Moran, von Dohlen, and Baumann 1995), and observed changes in endosymbiont
16S rRNA genes destabilize the secondary structure of
the molecule (Lambert and Moran 1998). In addition,
several protein-coding genes in Buchnera have been
shown to evolve more rapidly than their E. coli homologues (Moran 1996; Brynnel et al. 1998). Relatively
high ratios of nonsynonymous divergence (Ka) to synonymous divergence (Ks) imply that these substitutions
are concentrated at sites that affect amino acid sequences (Moran 1996; Brynnel et al. 1998).
This study combines several types of analyses to
further explore whether Buchnera loci experience increased rates of fixation of deleterious mutations, as
would be expected in small, asexual populations. Here
we assess patterns of codon usage across several Buchnera loci, compare levels of bias with homologues in E.
coli, and explore the possibility of subtle, possibly different, codon preferences in Buchnera. Since purifying
selection against nonoptimal codons is likely to be
weaker than selection against replacement substitutions,
previous evidence for the accumulation of nonsynonymous substitutions through drift strongly suggests that
nonoptimal codons will accumulate in Buchnera lineages. Therefore, adaptive codon bias is expected to be
much lower than that observed in E. coli. In addition,
rates of sequence evolution of Buchnera and E. coli are
compared across an extensive sample of available protein-coding loci, in order to test whether previously observed rate elevation is a general phenomenon across
the Buchnera genome. Previous estimates of the ratio of
nonsynonymous to synonymous divergence in Buchnera
85
were limited by high levels of synonymous divergence
(Moran 1996). Values near saturation are known to have
high standard errors, which may be exacerbated by the
strong A1T bias of Buchnera genomes (Berg 1995). In
this study, the consideration of shallower taxonomic levels allows for more reliable estimates of ratios of nonsynonymous to synonymous substitutions.
Methods
Loci Sampled
Several loci from Buchnera taxa and the enteric
bacteria were included (table 1) in estimates of codon
bias, rates of sequence evolution, and patterns of nonsynonymous and synonymous substitutions.
Alignments
Inferred protein sequences of homologous loci
were aligned using Pileup of GCG (Wisconsin Sequence
Analysis Program, Genetics Computer Group, Madison,
Wis.), and nucleotide alignments were adjusted to conform to amino acid alignments. Regions of loci with
ambiguous amino acid alignments were excluded from
the analysis.
Codon Bias
Estimation of Codon Bias at Fourfold
Degenerate Sites
For each Buchnera locus, nonrandom use of U- and
A-ending codons within fourfold degenerate codon families was assessed using a x2 analysis. (Arginine, leucine, and serine were treated as fourfold degenerate by
considering only the fourfold degenerate synonyms.) Cand G-ending codons were excluded from the analysis
because of their small sample sizes and low expected
values in this A1T-rich genome (see below).
The x2 analysis involved several steps: first, nucleotide composition at fourfold degenerate sites was determined for each Buchnera locus (by the computer
package Molecular Evolutionary Analysis [MEA], E.
Moriyama, personal communication). Expected relative
frequencies of U- and A-ending codons for each fourfold codon family were based on the relative frequencies
of A’s and T’s at fourfold degenerate sites (calculated by
MEA). Second, observed and expected relative values
of U- and A-ending codons were compared by a twoclass x2 test adjusted for a small sample size, as suggested by Sokal and Rohlf (1981) for sample sizes
,200. Amino acids with less than five residues in a
given locus were omitted.
Similar to the scaled x2 of Shields et al. (1988), the
magnitude of the x2 value reflects the deviation of random use of synonymous codons. However, the x2 values
in this study reflect the deviation of relative frequencies
of U- and A-ending codons from frequencies expected
if codon usage reflects local base composition. In order
to estimate overall bias at a given locus, x2 values were
averaged across the fourfold codon families after excluding those amino acids with fewer than five representatives.
86
Wernegreen and Moran
Table 1
Genetic Loci of Buchnera Strains Included in Study
Taxona
Gene Name
GenBank
Accession
Number
Acyrthosiphon pisum. . . . . . . . . . . . .
Schlechtendalia chinensis. . . . . . . . .
Schizaphis graminum . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
A. pisum. . . . . . . . . . . . . . . . . . . . . . .
A. pisum. . . . . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
A. pisum. . . . . . . . . . . . . . . . . . . . . . .
Myzus persicae . . . . . . . . . . . . . . . . .
Rhopalosiphum padi . . . . . . . . . . . . .
Salmonella typhimurium. . . . . . . . . .
Sitobion avenae. . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
M. persicae . . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
Diuraphis noxia . . . . . . . . . . . . . . . .
R. padi . . . . . . . . . . . . . . . . . . . . . . . .
S. typhimurium . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
Thelaxes suberi . . . . . . . . . . . . . . . . .
D. noxia . . . . . . . . . . . . . . . . . . . . . . .
R. padi . . . . . . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. typhimurium . . . . . . . . . . . . . . . . .
T. suberi. . . . . . . . . . . . . . . . . . . . . . .
D. noxia . . . . . . . . . . . . . . . . . . . . . . .
R. padi . . . . . . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. typhimurium . . . . . . . . . . . . . . . . .
T. suberi. . . . . . . . . . . . . . . . . . . . . . .
D. noxia . . . . . . . . . . . . . . . . . . . . . . .
R. padi . . . . . . . . . . . . . . . . . . . . . . . .
S. typhimurium . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
T. suberi. . . . . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
R. padi . . . . . . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
argS
argS
aroA
aroE
aroH
atpA
atpB
atpC
atpD
atpE
atpF
atpG
atpH
cysE
cysS
ddlB
dnaA
dnaG(pt)c
dnaJ
dnaK
dnaN
dnaQ
fdx
ftsA
ftsZ
gapA
gidA
groEL
groEL
groEL
groEL
groEL
groEL
groES
groES
gyrB
himD
hscA
hscB
ilvC
ilvD
infC
leuA
leuA
leuA
leuA
leuA
leuB
leuB
leuB
leuB
leuB
leuC
leuC
leuC
leuC
leuC
leuD
leuD
leuD
leuD
leuD
murC
nifS
pfs
rep
rep
L18933b
L18932b
L43549b
U09230b
U11066b
2827020b
2827024b
2827018b
2827017b
2827023b
2827022b
2827019b
2827021b
M90644b
U09230b
2738587b
M80817b
M90644b
D88673b
D88673b
M80817b
L18927b
2827028b
2738589b
2738588b
U11045b
2827025b
X61150b,d
2754808b,d
U77380b,d
U01039c
U77379b,d
D85628b,d
2754807b,d
D85628b,d
M80817b
L43549b
2827029b
2827030b
2827034b
2827033b
U11066b
AF041837b,d
X71612b,d
47968d
AF041836b,d
Y11966b,d
AF041837b,d
X71612b,d
AF041836b,d
X53376d
Y11966b,d
AF041837b,d
X71612b,d
AF041836b,d
M31047d
Y11966b,d
AF041837b,d
X71612b,d
47764d
AF041836b,d
Y11966b,d
AF012886b
2827032b
AF01288b
X71612b
2827035b
Table 1
Continued
Taxona
D. noxia . . . . . . . . . . . . . . . . . . . . . . .
R. padi . . . . . . . . . . . . . . . . . . . . . . . .
Tetraneura caerulescens. . . . . . . . . .
T. suberi. . . . . . . . . . . . . . . . . . . . . . .
D. noxia . . . . . . . . . . . . . . . . . . . . . . .
R. padi . . . . . . . . . . . . . . . . . . . . . . . .
T. suberi. . . . . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. chinensis . . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. chinensis . . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
A. pisum. . . . . . . . . . . . . . . . . . . . . . .
D. noxia . . . . . . . . . . . . . . . . . . . . . . .
Macrosiphoniella ludovicianae . . . .
Melaphis rhois . . . . . . . . . . . . . . . . .
Rhopalosiphum maidis . . . . . . . . . . .
R. padi . . . . . . . . . . . . . . . . . . . . . . . .
S. typhimurium . . . . . . . . . . . . . . . . .
S. chinensis . . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
ECOR 17 . . . . . . . . . . . . . . . . . . . . . .
ECOR 29 . . . . . . . . . . . . . . . . . . . . . .
ECOR 31 . . . . . . . . . . . . . . . . . . . . . .
ECOR 37 . . . . . . . . . . . . . . . . . . . . . .
ECOR 46 . . . . . . . . . . . . . . . . . . . . . .
ECOR 50 . . . . . . . . . . . . . . . . . . . . . .
ECOR 51 . . . . . . . . . . . . . . . . . . . . . .
ECOR 60 . . . . . . . . . . . . . . . . . . . . . .
ECOR 71 . . . . . . . . . . . . . . . . . . . . . .
ECOR 72 . . . . . . . . . . . . . . . . . . . . . .
Uroleucon aeneum . . . . . . . . . . . . . .
Uroleucon ambrosiae . . . . . . . . . . . .
Uroleucon astronomus . . . . . . . . . . .
Uroleucon caligatum . . . . . . . . . . . .
Uroleucon erigeronense . . . . . . . . . .
Uroleucon helianthicola . . . . . . . . . .
Uroleucon jaceae . . . . . . . . . . . . . . .
Uroleucon jaceicola . . . . . . . . . . . . .
Uroleucon obscurum . . . . . . . . . . . .
Uroleucon rudbeckiae . . . . . . . . . . .
Uroleucon rapunculoidis . . . . . . . . .
Uroleucon rurale . . . . . . . . . . . . . . .
Uroleucon solidaginis. . . . . . . . . . . .
Uroleucon sonchi . . . . . . . . . . . . . . .
S. chinensis . . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. chinensis . . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
A. pisum. . . . . . . . . . . . . . . . . . . . . . .
D. noxia . . . . . . . . . . . . . . . . . . . . . . .
R. maidis . . . . . . . . . . . . . . . . . . . . . .
R. padi . . . . . . . . . . . . . . . . . . . . . . . .
S. typhimurium . . . . . . . . . . . . . . . . .
S. chinensis . . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
U. caligatum . . . . . . . . . . . . . . . . . . .
Gene Name
repA1
repA1
repA1
repA1
repA2
repA2
repA2
rho
rmph
rnh
rnpA
rpoB
rpoC
rpoD
rpsA
secB
sohB
thrS
tpiA
trmE
trpA
trpA
trpB
trpB
trpB
trpB
trpB
trpB
trpB
trpB
trpB
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpB(pt)
trpC
trpC
trpD
trpD
trpE
trpE
trpE
trpE
trpE
trpE
trpE
trpE
GenBank
Accession
Number
AF041837b,g
X71612b,g
Y11972b,g
Y11966b,g
AF041837b,g
X71612b,g
Y11966b,g
2827037b
M80817b
L18927b
M80817b
Z11913b
Z11913b
M90644b
L43549b
M90644b
U09185b
U11066b
L43549b
2827009b
U09185b
Z19055b
L46355b,d
AF038565b,d
AF058428e
L46357b,d
L46356b,d
L46358b,d
J01810d
U09185b,d
Z19055b,d
U23489e
U25425e
U23494e
U23496e
U23495e
U23497e
U25884e
U23499e
U23500e
U25429e
AF058431d,e
AF058432d,e
AF058433d,e
L81150d,e
L81151d,e
AF058434d,e
AF058435d,e
AF058436d,e
AF058437d,e
AF058439d,e
AF058438d,e
L81149d,e
AF058440d,e
1137716b,d,e
U09185b
Z19055b
U09185b
Z19055b
L43555b,d
L46769b,d
L43550b,d
L43551b,d
V01378d
U09184b,d
Z21938b,d
L8124d
Evidence for Drift in Endosymbionts
Table 1
Continued
Taxona
U. erigeronense. . . . . . . . . . . . . . . . .
U. rurale . . . . . . . . . . . . . . . . . . . . . .
U. sonchi . . . . . . . . . . . . . . . . . . . . . .
A. pisum. . . . . . . . . . . . . . . . . . . . . . .
D. noxia . . . . . . . . . . . . . . . . . . . . . . .
R. maidis . . . . . . . . . . . . . . . . . . . . . .
R. padi . . . . . . . . . . . . . . . . . . . . . . . .
S. chinensis . . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
A. pisum. . . . . . . . . . . . . . . . . . . . . . .
M. rhois . . . . . . . . . . . . . . . . . . . . . . .
Pemphigus betae . . . . . . . . . . . . . . . .
S. typhimurium . . . . . . . . . . . . . . . . .
S. chinensis . . . . . . . . . . . . . . . . . . . .
S. graminum . . . . . . . . . . . . . . . . . . .
Aeromonas salmonecida. . . . . . . . . .
Pseudomonas putida. . . . . . . . . . . . .
Haemophilus ducreyi . . . . . . . . . . . .
Vibrio cholerae . . . . . . . . . . . . . . . . .
P. putida . . . . . . . . . . . . . . . . . . . . . .
P. aeruginosa . . . . . . . . . . . . . . . . . .
Azotobacter vinelandii . . . . . . . . . . .
Azotobacter vinelandii . . . . . . . . . . .
P. putida . . . . . . . . . . . . . . . . . . . . . .
Gene Name
trpE
trpE
trpE
trpG
trpG
trpG
trpG
trpG
trpG
trxA
tufA
tufA
tufA
tufA
tufA
tufA
aroA
dnaG
dnaJ
dnaK
dnaN
leuB
leuC
leuD
rpoB
GenBank
Accession
Number
L81123d
L8112d
1137712d
L43555b,g
L46769b,g
L43550b,g
L43551b,g
U09184b,g
Z21938b,g
2827036b
2369691b,d
2369697b,d
2369695b,d
X55116d
L43549b,d
2369693b,d
L05002f
U85774f
U25996f
Y14237f
X14791f
U29655f
Y11280f
Y11280f
X15840f
a Buchnera strains are labeled by the aphid species from which they were
isolated. All sequences from Escherichia coli and Haemophilus influenzae are
accessible from the full genome sequences of these two species (GenBank accession numbers U00096 and L42023, respectively). Individual loci of these two
species are not listed.
b Sequences of Buchnera taxa (listed) and E. coli compared in codon usage
analysis.
c ‘‘(pt)’’ indicates that only a partial sequence is available in GenBank.
d Sequences of Buchnera and enteric bacteria (including E. coli and the enteric species listed) used in comparison of Ka/Ks.
e Sequences used in mapping of nucleotide changes across phylogenies, in
addition to the trpB sequence of E. coli K12.
f Taxon 3 in relative-rate tests.
g No homologue in E. coli with sufficient similarity; locus excluded from
comparison of CAI values.
Estimates of Codon Bias in Homologous E. coli Loci
Codon bias in E. coli loci was estimated by the
deviation from random use of all four synonyms (not
just U- and A-ending codons) at fourfold degenerate
sites. Expected values were based on the nucleotide
composition at fourfold degenerate sites for a particular
locus (calculated using MEA), and deviation from expected was tested using a four-class x2 test. Only amino
acids with five or more representatives in a given locus
were considered. The magnitudes of x2 values are not
directly comparable between E. coli and Buchnera, since
the two tests have different numbers of classes and
therefore different degrees of freedom.
Analyses of Codon Usage Based on x2 Test Results
and Values
E. coli and Buchnera loci were compared in terms
of the proportion of total x2 tests with significant results,
the average x2 values of homologous loci, and the x2
values for individual amino acids across homologous
loci. Although the actual x2 values are not directly com-
87
parable between E. coli and Buchnera (see above), the
latter two analyses allow a visual comparison of levels
and patterns of codon bias. Codon usage in Buchnera
was also explored by testing for the overrepresentation
of particular loci among significant x2 test results and
by identifying consistent preferences for U- or A-ending
codons for particular amino acids.
Rates of Nucleotide Divergence
Relative-Rate Tests
Relative-rate tests were used to compare rates of
nonsynonymous divergence at homologous loci of
Buchnera and E. coli. Tests were performed as described
previously (Moran 1996).
Estimation of Ka/Ks
Estimates of nonsynonymous and synonymous
pairwise divergence and standard deviations were calculated using Li’s (1993) method (DIVERGE, GCG).
Compared to other commonly used estimates of nucleotide divergence, Li’s (1993) method has been shown to
be reliable for loci with biased base composition (Ina
1995). In order to avoid inaccurate estimates due to saturation, only pairwise comparisons with Ks , 1.0 were
included in the calculation of Ka/Ks ratios.
Mapping of Nucleotide Changes Across Genealogies
A portion of trpB is available for Buchnera of several Uroleucon species and for several E. coli isolates
in the ECOR collection. Within each of these two
groups, Ka values were too low to estimate reliable Ka/
Ks ratios for pairwise divergences. Instead, the percent
change at first- and second-codon positions was used to
approximate nonsynonymous divergence, and percent
change at third positions was used to approximate synonymous divergence. Changes at each codon position
were mapped with parsimony across genealogies of the
partial trpB sequence and summed across each phylogeny. The tree length at first- and second-codon positions,
divided by the length at third positions, adjusted for the
number of sites, roughly approximates the ratio of divergence at replacement versus silent positions.
Phylogenies of trpB partial sequences were estimated by parsimony analysis (by PAUP, Swofford
1993). For the ECOR and Buchnera trpB data sets, two
most-parsimonious trees were found, and a single mostparsimonious tree was selected. For each data set, nucleotide changes at each codon position were mapped
across the selected tree (by MacClade, Maddison and
Maddison 1992). Changes at first- and second-codon positions were summed across each tree and divided by
the total number of first- and second-codon positions
(452 nt) to approximate percent change at nonsynonymous sites. Likewise, changes at third positions were
summed and divided by the total number of third-codon
positions (227 nt) to approximate percent change at synonymous sites. The ratios of these estimates were compared across the ECOR phylogeny and subsets of the
Buchnera phylogeny.
88
Wernegreen and Moran
Results and Discussion
Evidence for Strong A1T Mutational Pressure Across
Buchnera Loci
All Buchnera loci sampled were extremely A1T
rich at fourfold degenerate sites (the average across loci
was 88.4%). This is consistent with previous observations of A1T richness across the Buchnera genome
(Ishikawa 1987; Clark, Baumann, and Baumann 1992;
Ohtaka and Ishikawa 1993; Clark, Baumann, and Baumann 1998).
Comparison of E. coli and Buchnera
Number of Significant x2 Tests
The null hypothesis that codon usage in Buchnera
reflects local base composition was tested by performing
a x2 analysis of observed and expected numbers of Uor A-ending codons for eight fourfold degenerate families across several loci. Of a total of 772 x2 tests performed across Buchnera loci, fewer tests were significant than expected by chance alone. Only 23 tests were
significant at the 5% level, and 42 more were significant
at the 10% level (table 2). This paucity of significant x2
tests indicates that codon usage generally reflected local
base composition. In contrast, the majority of E. coli
loci showed significant nonrandom use of alternative codons for most amino acids. Of a total of 528 x2 tests
performed across E. coli genes, 274 were significant at
the 5% level, and 35 more were significant at the 10%
level (data not shown).
Comparison of x2 Averaged Across Amino Acids
The depression of codon bias reflected in Buchnera
is also apparent in the low x2 values, averaged across
amino acids for each locus (fig. 1). Although the magnitude of x2 values is not directly comparable for E. coli
and Buchnera (see Methods), a large proportion of E.
coli loci have average x2 values greater than the critical
value for significance at the 5% level (7.815 for df 5
3), whereas no Buchnera loci have an average x2 value
that exceeds the critical value significance at 5% (3.841
for df 5 1).
In addition, a strong relationship exists between the
average x2 values for loci of E. coli and the CAI on the
basis of preferred codons for this species, which is
known to be highly correlated with gene expression level (Sharp and Li 1987). In contrast, the lack of correspondence between average x2 estimates of Buchnera
genes and the CAI of homologous E. coli loci indicates
that codon bias in Buchnera does not correspond with
levels of gene expression of E. coli homologues and
provides further evidence against effective translational
selection in Buchnera.
Comparison of x2 Values for Individual Amino Acids
The contrast between levels of bias in Buchnera
and in E. coli is also evident in the narrow range of x2
values for each fourfold degenerate family across Buchnera loci (fig. 2). Buchnera loci rarely show significantly nonrandom use of U- or A-ending codons for
individual amino acids, even for groEL, which is known
to be highly expressed in Buchnera (Baumann, Bau-
mann, and Clark 1996). In contrast, x2 values are relatively high for individual amino acids across most E.
coli loci, including those considered low expression in
E. coli, such as trp genes (Sharp, Tuohy, and Mosurski
1986).
Evidence for Preferential Use of U-ending and Aending Codons in Buchnera
Despite the severe depression of codon bias in
Buchnera, codon usage cannot be attributed solely to
mutational bias. In particular, serine and arginine are
generally encoded by U-ending codons, and alanine
tends to be encoded by the A-ending codon. For these
amino acids, significant x2 tests across loci almost always reflect a higher frequency of one particular (U- or
A-ending) codon (table 2). Since expected values of
these tests are based on gene-specific base composition,
these significant results cannot be attributed to local variation in A1T content. Likewise, for loci showing slight
(nonsignificant) codon preferences, the preferred codon
within the fourfold degenerate family for serine tends to
be UCU, and alanine tends to be encoded by GCA (P
, 0.01 in each case; table 3). These preferences agree
with the general trends found in a previous analysis of
codon usage across several loci of the endosymbiont of
Schizaphis graminum, many of which are included in
the genes sampled here (Clark, Baumann, and Baumann
1998).
These apparent codon preferences across several
Buchnera loci and taxa suggest that purifying selection
may effectively reduce the frequency of A-ending codons for serine and arginine and reduce the U-ending
codon for alanine. Buchnera populations may occasionally go through periods where selection is more effective, possibly because of an expansion of aphid population sizes. Strong mutational bias and drift may largely
shape patterns of codon use, but these occasional periods of effective purifying selection may reduce the frequency of nonoptimal codons.
Preference for the U-ending codon of the fourfold
degenerate families of serine and arginine agrees with
the codon preferences of highly expressed genes of E.
coli (Sharp et al. 1988). However, codon preferences for
alanine in E. coli depend on gene expression levels, as
highly expressed loci are biased toward GCU, which is
thought to represent the optimal codon for this amino
acid, but lowly expressed loci are biased toward GCA.
The preference for GCA across Buchnera loci may reflect a change in the optimal codon, possibly precipitated by drift in small populations. It has been suggested
that such switches in codon preference among the enteric bacteria may relate to differences among lineages
in effective population sizes (Shields 1990).
Evidence for Bias in Buchnera groEL Genes
A second line of evidence that codon use may be
under effective (although weak) selection in Buchnera
is the slightly higher levels of bias observed at the overexpressed gene, groEL. This gene had the greatest overrepresentation in the pool of significant x2 tests, relative
to its frequency in the original sample (4.1% in original
Evidence for Drift in Endosymbionts
Table 2
Significant Nonrandom Use of U- or A- Ending Codons Across Several Buchnera Loci
for Each of Eight Fourfold Degenerate Families
Taxona
Locus
Amino
Acid
Schizaphis graminum . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
Schlechtendalia chinensis . . . . . .
Acyrthosiphon pisum . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
Sitobion avenae . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. chinensis . . . . . . . . . . . . . . . . .
S. chinensis . . . . . . . . . . . . . . . . .
S. chinensis . . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
Diuraphis noxia . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
Thelaxes suberi . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. chinensis . . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
Rhopalosiphum maidis . . . . . . . .
S. chinensis . . . . . . . . . . . . . . . . .
A. pisum . . . . . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. avenae . . . . . . . . . . . . . . . . . . .
T. suberi . . . . . . . . . . . . . . . . . . . .
T. suberi . . . . . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
A. pisum . . . . . . . . . . . . . . . . . . . .
A. pisum . . . . . . . . . . . . . . . . . . . .
A. pisum . . . . . . . . . . . . . . . . . . . .
S. chinensis . . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
D. noxia . . . . . . . . . . . . . . . . . . . .
S. avenae . . . . . . . . . . . . . . . . . . .
R. padi . . . . . . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. chinensis . . . . . . . . . . . . . . . . .
A. pisum . . . . . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
T. suberi . . . . . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
A. pisum . . . . . . . . . . . . . . . . . . . .
Melaphis rhois . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
Pemphigus betae . . . . . . . . . . . . .
S. graminum. . . . . . . . . . . . . . . . .
M. rhois . . . . . . . . . . . . . . . . . . . .
D. noxia . . . . . . . . . . . . . . . . . . . .
S. chinensis . . . . . . . . . . . . . . . . .
tpiA
groES
trpG
leuA
tufA
groEL
trpD
trpA
groEL
rpoD
atpA
trpE
trpA
tuf
rpoB
rep
leuB
trpE
leuD
dnaN
trpA
ftsA
hscA
trpE
trpB
groEL
dnaA
aroH
groEL
leuA
leuC
dnaG
murC
ilvD
groEL
trpB
rpoB
dnaA
infC
dnaK
groEL
dnaJ
tuf
ddlB
trpE
groEL
repAl
ftsZ
rpoB
trpG
trpE
groEL
rhn
leuA
ilvD
argS
tufA
rpoB
ilvC
trpA
tuAf
gidA
tufA
leuD
trpA
A
A
A
A
A
A
A
A
A
A
A
G
G
G
G
G
G
G
G
L
L
L
L
L
P
P
P
P
P
P
P
P
P
R
R
R
R
R
R
R
R
R
S
S
S
S
S
S
S
S
S
S
T
T
T
T
T
T
T
V
V
V
V
V
V
a
b
No. of
Amino Acid
Residues
in Protein
x2 Value for
Nonrandom
Useb
Codon
Ending
14
8
7
35
22
54
12
11
51
27
35
18
17
30
90
17
19
16
8
9
5
9
15
18
11
15
10
10
15
7
12
8
11
14
17
6
35
17
11
9
18
10
10
16
20
29
11
17
64
7
33
30
10
18
26
5
22
58
18
7
30
26
31
9
11
6.401
4.814
4.45
3.817
3.532
3.487
3.484
3.424
2.803
2.734
5.174
9.044
4.2
3.452
3.227
3.039
2.754
2.858
3.125
3.875
2.995
4
6.317
6.368
8.27
4.158
2.9
2.756
2.708
2.734
2.76
3.265
3.777
3.435
2.7299
2.7963
3.0635
3.184
3.5566
5.0547
6.5715
9.2069
2.915
3.0218
3.1565
3.2248
3.5667
4.5219
5.8242
5.9269
6.7696
7.9462
3.784
3.613
3.2
3
2.8356
3.8263
7.2636
5.293
4.417
3.81
3.624
3.531
2.967
A
A
A
A
A
A
A
A
A
A
U
A
A
A
A
A
A
U
U
A
U
U
U
U
A
A
A
A
A
U
U
U
U
A
U
U
U
U
U
U
U
U
U
U
U
U
U
U
U
U
U
U
A
A
A
A
U
U
U
A
A
A
A
A
U
Buchnera taxa are labeled by the aphid species from which they were isolated.
Critical values for these two-class x2 tests are 2.706 (P , 0.1) and 3.841 (P , 0.05).
89
90
Wernegreen and Moran
FIG. 1—Relationship between the x2-based estimate of codon bias and the Codon Adaptation Index (CAI) for several E. coli or Buchnera
loci. Points represent a single locus and are positioned on the y-axis by the x2 value averaged across each fourfold degenerate amino acid.
Points are positioned on the x-axis by the CAI of either the same gene (E. coli) or the homologous E. coli gene (Buchnera). Average x2 includes
only x2 values for amino acids with five or more residues in a given locus. In contrast to Buchnera genes, E. coli loci have consistently higher
average x2 values and a strong relationship with CAI.
sample versus 12.3% in the pool of significant x2 tests;
table 4). In particular, for symbionts of Acyrthosiphon
pisum, four of the eight amino acids considered showed
significant x2 values (table 2). Slight codon bias at
groEL is also suggested by the relatively high x2 values
at this locus (figs. 1, 2). Compared to other Buchnera
loci, groEL of Buchnera from Acyrthosiphon pisum has
the highest average x2 (see fig. 2).
Relative-Rate Tests
In agreement with previous studies, relative-rate
tests demonstrated higher rates of nonsynonymous substitution in Buchnera loci relative to E. coli homologues.
Rates of sequence evolution in Buchnera genes were 1.3
to 6.9 times faster than in E. coli genes (table 5).
Comparisons of Ka/Ks
For several protein-coding loci, pairwise divergences at nonsynonymous and synonymous sites were
calculated across several Buchnera lineages and across
E. coli versus Salmonella typhimurium. These pairwise
divergences were used to estimate ratios of nonsynonymous to synonymous divergence (Ka/Ks) for Buchnera
and for the enterics. Across each locus sampled, values
of Ka/Ks were generally higher in Buchnera than in E.
coli (fig. 3). The maximum value for Ka/Ks for Buch-
Evidence for Drift in Endosymbionts
91
FIG. 2—Comparison of x2 values for individual amino acids, across several homologous E. coli [x] and Buchnera [.] loci. Buchnera loci
are labeled by the taxon from which a gene was sampled (Sg 5 Schizaphis graminum, Ap 5 Acyrthosiphon pisum, Mr 5 Melaphis rhois, Pb
5 Pemphigus beta, Sc 5 Schlechtendalia chinensis, Rp 5 Rhopalosiphum padi, Sa 5 Sitobion avenae, Mp 5 Myzus persicae, Dn 5 Diuraphis
noxia, Rm 5 R. maidis, Usn 5 Uroleucon sonchi, Ts 5 Thelaxes suberi). For illustrative purposes only, x2 values of E. coli are graphed as
negative values if the preferred codon is a nonoptimal codon as defined by the E. coli Relative Synonymous Codon Usage table (Sharp et al.
1988). x2 values of Buchnera are graphed as negative values if the A-ending codon is preferred. For each locus, levels of bias in E. coli are much
higher than those in Buchnera. The locus showing the strongest evidence for bias in Buchnera is groEL, particularly that of Acyrthosiphon pisum.
nera always exceeded that of the enterics. Except for
groEL, the minimum Ka/Ks value also exceeded the estimate for the enterics (fig. 3b). When several loci were
available for a given pair of Buchnera taxa, groEL generally proved to have the lowest Ka and the lowest Ks
(fig. 3a). The relatively low Ka/Ks ratios for this gene
may be attributed to decreased nonsynonymous divergence that is more extreme than the observed depression
at synonymous sites. Compared to other genes in Buchnera, purifying selection is apparently more effective
against replacement substitutions at this highly expressed locus.
92
Wernegreen and Moran
Table 3
Number of Buchnera Loci Showing a Slight Preference for U- or A-Ending Codons for
Each of Eight Fourfold Degenerate Families
AMINO ACID
xxU
xxA
CONSIDERED
EXPECTED NUMBER OF LOCI
SHOWING A PREFERENCE FOR AENDING CODON
A.........
G.........
L .........
P .........
R.........
S .........
T .........
V.........
37
52
33
37
32
72
41
55
53
47
37
48
23
31
54
51
90
99
70
85
55
103
95
106
39.8
43.8
31.0
37.6
24.3
45.6
42.0
46.9
NO.
OF
BUCHNERA LOCIa
TOTAL
LOCI
x2 VALUEb
7.77
0.42
2.09
5.12
0.13
8.32
6.07
0.64
P VALUE (,)
0.01
0.025
0.01
0.01
a Number of Buchnera loci showing slight preference for the A-ending codon of that amino acid and number showing
a preference for the U-ending codon.
b The x2 value expresses the deviation of the observed from the expected number of loci with preference for A-ending
codons.
Mapping of Base Changes Across Genealogies
Sequence evolution at a portion of trpB was compared across two very shallow taxonomic groups: the
ECOR collection of E. coli strains and Buchnera associated with the aphid genus Uroleucon. Because of low
levels of divergence within each group, nucleotide
changes were summed across genealogies of trpB instead of calculated pairwise (fig. 4). Given the higher
divergence among Uroleucon isolates, distinct trpB
clades were considered separately. Each clade, however,
gave comparable values for the ratio of the tree length
at first and second positions, divided by the length at
third positions adjusted for the number of sites. This
approximation of nonsynonymous divergence divided
by synonymous divergence is considerably higher for
the Uroleucon isolates, relative to the E. coli strains
(0.162 to 0.285 across subsets of the Buchnera trpB
phylogeny and 0.0328 across the E. coli trpB phylogeny;
fig. 4). The ECOR and Buchnera trees each include
nodes with relatively weak support (see bootstrap values, fig. 4). For the purposes of mapping nucleotide
changes at each codon position, these relatively weak
nodes do not affect the results obtained. It should be
noted, however, that the resolved trees presented are not
intended to represent exact relationships among E. coli
or Buchnera isolates.
Estimates of Synonymous Substitution in E. coli versus
Buchnera
Under the assumption that rates of synonymous
substitutions reflect mutation rates, high ratios of nonsynonymous to synonymous divergence indicate that replacement substitutions accumulate at faster rates in a
particular lineage (Brynnel et al. 1998). This assumption
may be violated by differing levels of codon bias in
Buchnera and E. coli. Selection for translational efficiency in E. coli may depress rates of synonymous substitution and thus elevate Ka/Ks estimates. However, this
discrepancy in codon bias would only dispose the Ka/
Ks comparison away from the previously observed
trend: higher Ka/Ks ratios in Buchnera. Therefore, low
levels of adaptive codon bias in Buchnera only strengthen the interpretation that relatively high Ka/Ks ratios
across Buchnera loci reflect an elevation of nonsynonymous substitutions.
Table 4
Frequencies of Buchnera Loci in the Original Sample Compared to Their Representation
in the Pool of Loci for Which There Is Evidence of Significant Codon Bias
Locus
No. of
x2 Tests
Fraction of
Total x2 Tests
Performeda
No. Significant
x2 Tests
Fraction of
Significant
x2 Testsb
groELc . . . . . . . . . .
groES . . . . . . . . . . .
leu genes . . . . . . . .
trpB . . . . . . . . . . . .
trpA . . . . . . . . . . . .
trpD . . . . . . . . . . . .
trpE . . . . . . . . . . . .
trpG . . . . . . . . . . . .
tuf . . . . . . . . . . . . . .
32
8
119
64
15
14
46
36
39
0.041
0.010
0.154
0.083
0.019
0.018
0.060
0.047
0.051
8
1
7
2
5
1
5
2
6
0.123
0.015
0.108
0.031
0.077
0.015
0.077
0.031
0.092
A total of 772 x2 tests were performed.
The total number of significant x2 tests was 65.
c groEL had the greatest overrepresentation in the pool of significant tests compared to its frequency in the original
sample.
a
b
Evidence for Drift in Endosymbionts
93
Table 5
Relative-Rates Test for Substitutions at Nondegenerate Sites in Loci of Buchnera Versus Escherichia coli
Gene
ilvC . . . . . . . .
ilvD . . . . . . . .
ilvI . . . . . . . . .
leuA . . . . . . . .
leuB . . . . . . . .
leuC. . . . . . . .
leuD. . . . . . . .
aroA . . . . . . .
cysE. . . . . . . .
dnaJ. . . . . . . .
dnaK . . . . . . .
dnaN . . . . . . .
dnaG(pt)c . . .
secB. . . . . . . .
rpoB . . . . . . .
rpsA. . . . . . . .
atpD . . . . . . .
tpiA . . . . . . . .
gapA . . . . . . .
gidA. . . . . . . .
Function
No.
Codons
Isoleucine valine biosynthesis
492
Isoleucine valine biosynthesis
617
Isoleucine valine biosynthesis
568
Leucine biosynthesis
507
Leucine biosynthesis
365
Leucine biosynthesis
465
Leucine biosynthesis
208
Aromatic amino acid biosynthesis
226
Serine/glycine family amino acid biosynthesis 251
Heat-shock protein
380
Heat-shock protein
638
Chromosome replication
368
Replication
321
Protein export, molecular chaperonin
155
Transcription (RNA polymerase)
1360
Translation (ribosomal protein)
557
Energy metabolism (ATP synthase)
465
Electron transport
253
Electron transport
336
Chromosome replication
629
Taxon 3a
K12
K13
K23 K13–K23
Haemophilus influenzae
H. influenzae
H. influenzae
H. influenzae
Pseudomonas aeruginosa
Azotobacter vinelandii
A. vinelandii
Aeromonas salmonicida
H. influenzae
Haemophilus ducreyi
Vibrio cholerae
Pseudomonas putida
P. putida
H. influenzae
P. putida
H. influenzae
H. influenzae
H. influenzae
H. influenzae
H. influenzae
0.27
0.25
0.22
0.24
0.30
0.31
0.42
0.25
0.43
0.21
0.12
0.59
0.45
0.44
0.12
0.16
0.11
0.53
0.18
0.31
0.29
0.28
0.25
0.29
0.45
0.58
0.55
0.42
0.41
0.37
0.16
0.71
1.06
0.58
0.25
0.22
0.14
0.56
0.23
0.32
0.12
0.17
0.22
0.24
0.40
0.37
0.34
0.36
0.21
0.30
0.10
0.38
0.77
0.36
0.21
0.14
0.07
0.20
0.09
0.18
0.17
0.10
0.02
0.06
0.05
0.20
0.21
0.05
0.20
0.07
0.06
0.33
0.29
0.21
0.04
0.09
0.07
0.36
0.14
0.14
zb
K01/K02
7.65***
5.24***
1.20
2.36**
1.33
5.18***
3.66***
1.21
4.83***
2.23*
3.84***
6.26***
3.34**
3.13**
3.37**
5.01***
4.48***
7.21***
6.00***
6.38***
4.3
2.4
1.3
1.6
1.4
4.7
2.9
1.5
2.8
2.0
2.6
3.6
4.7
2.9
2.0
3.4
3.7
5.3
6.9
2.6
a In each test, taxon 1 is Buchnera of Schizaphis graminum, except for dnaJ and dnaK, which are Buchnera of Acyrthosiphon pisum, and taxon 2 is always
Escherichia coli. Taxon 3 is a more distantly related reference taxon.
b z scores were calculated as described by Muse and Weir (1992). Probabilities for one-tailed t-test (H :K
0
01 # K02) are * P , 0.05, ** P , 0.01, *** P ,
0.0001.
c ‘‘(pt)’’ indicates that only a partial sequence is available in GenBank.
More problematic is the possibility that Ks in Buchnera is underestimated because of the strong A1T bias
across loci and more rapid saturation at silent sites.
However, the calculation of Ka/Ks ratios across shallow
taxonomic levels avoided the problem of saturation at
synonymous sites and the large standard errors that accompany high divergence estimates. In addition, comparisons of changes at first- and second- vs. third-codon
positions across very shallow taxonomic levels (Buchnera isolates of Uroleucon and members of the ECOR
collection of E. coli) also suggest higher rates of fixation
at replacement sites, relative to synonymous sites, in
Buchnera.
Conclusions
Because of their small population sizes and limited
opportunities for recombination, vertically inherited endosymbionts provide a good model system to test the
effects of increased drift on sequence evolution in bacteria. In this study, the lack of adaptive codon bias
across several Buchnera loci suggests that codon usage
is shaped primarily by A1T mutational bias rather than
by translational selection. In addition, relative-rate tests
and comparisons of Ka/Ks ratios support previous conclusions that Buchnera lineages experience rapid sequence evolution at nonsynonymous sites, compared to
their free-living relative, E. coli. These results suggest
that selection is ineffective in eliminating two types of
weakly deleterious mutations from Buchnera populations: those resulting in nonoptimal codons and those
resulting in amino acid replacements.
A set of loci that might be suspected of experiencing unusual selection in Buchnera are those encoding
enzymes for biosynthesis of essential amino acids.
Buchnera provisions host insects with these nutrients,
which are limiting in the plant phloem on which aphids
feed. In several Buchnera lineages, genes for anthranilate synthase (trpEG), the first and limiting enzyme in
the tryptophan biosynthetic pathway, are amplified in
tandem repeats on multicopy plasmids (Lai, Baumann,
and Baumann 1994; Rouhbakhsh et al. 1996, 1997; Baumann et al. 1997, 1998a). This increased copy number
is considered an adaptation that benefits the aphid hosts,
which do not synthesize tryptophan and which depend
on Buchnera to provide this essential nutrient (Baumann
et al. 1995; Douglas 1998). Likewise, observed duplications of leucine genes and their occurrence on plasmids may represent an adaptation for overexpression
(Bracho et al. 1995; van Ham et al. 1997; Baumann et
al. 1998b).
In addition, groEL is known to be constitutively
overexpressed in several intracellular bacteria, both mutualistic (Aksoy 1995; Baumann, Baumann, and Clark
1996) and pathogenic (Garduno et al. 1998). In Buchnera, groEL comprises about 10% of the total protein
produced (Ishikawa 1984; Hara et al. 1990). The mechanism for this overexpression is uncertain but likely involves changes in gene regulation rather than a gene
duplication, since groEL is apparently single copy in at
least one Buchnera species in which it is overexpressed,
the symbiont of Acyrthosiphon pisum (Ohtaka and Ishikawa 1993). The protein is known in E. coli to be a
heat shock chaperonin involved in the folding, assembly,
and translocation of other polypeptides (Bochdareva,
Lissen, and Girshovich 1988; Goloubinoff et al. 1989;
Goloubinoff, Gatenby, and Lorimer 1989). While its role
94
Wernegreen and Moran
FIG. 3—Comparison of levels of nonsynonymous substitutions (Ka) and synonymous substitutions (Ks) across several loci of the enteric
bacteria and Buchnera. (A) Pairwise divergence at nonsynonymous and synonymous sites, calculated across Buchnera [.] taxa and across E.
coli versus Salmonella typhimurium [x]. Ks values higher than 1.0 were excluded. (B) Ratios of nonsynonymous to synonymous substitutions,
on the basis of pairwise divergence across Buchnera taxa [.] and across E. coli versus S. typhimurium [x]. With the exception of groEL, all Ka/
Ks estimates for Buchnera exceed the ratio calculated for the homologous gene of the enteric bacteria. All Ka/Ks ratios are based on Ks values
less than one.
in endosymbionts is less certain, it may function to stabilize proteins that have accumulated amino acid substitutions, as suggested by Moran (1996). In this study,
sequence evolution of these functionally important
Buchnera genes shows the same patterns as housekeeping genes that are not overexpressed. In particular, genes
in the tryptophan (trpABC(F)DE) and leucine (leuABCD) biosynthetic pathways and groEL each show
low levels of codon bias and rapid rates of nonsynonymous substitution, relative to E. coli homologues.
The fact that the observed trends of depressed codon bias and accelerated evolutionary rates occur across
all Buchnera loci included, even functionally important
genes, argues for an increased effect of drift within endosymbiont populations. One alternative explanation for
elevated evolutionary rates at nonsynonymous sites is
positive selection for amino acid changes; however, such
selection typically acts at specific loci rather than across
the genome. In addition, there is no evidence that the
acceleration of nonsynonymous substitution rates is any
higher across Buchnera biosynthetic genes than across
other Buchnera loci. Combining loci from table 5 with
those examined previously (Moran 1996; table 2), 14
loci encode amino acid biosynthetic genes and 14 encode genes for other functions. Of the first set, seven
have K01/K02 . 2 and seven have K01/K02 , 2. Of the
second set, 12 have K01/K02 . 2, and 2 have K01/K02 ,
1. Thus, amino acid biosynthetic loci tend to be less
accelerated than other Buchnera genes.
A second alternative hypothesis for lack of codon
bias and accelerated evolutionary rates in Buchnera is a
relaxation of purifying selection against nonoptimal codons and replacement substitutions. Perhaps the intracellular environment is more constant than that experienced by free-living bacteria. While the results here cannot exclude the possibility of relaxed selection, two features of the data argue against it. First, selection is
unlikely to be relaxed across all loci, and the trends we
observed occur consistently across each locus examined.
In addition, selection is not likely to be relaxed at loci
that are functionally important in the symbiosis and that
are known to be overexpressed in Buchnera. Given their
apparent significance in the symbiosis, it is difficult to
imagine relaxed selection on trp genes, leu genes, and
Evidence for Drift in Endosymbionts
95
FIG. 4—Mapping of nucleotide changes across genealogies of trpB for (A) strains in the ECOR collection of E. coli isolates and (B) several
Buchnera isolates from the aphid genus Uroleucon (Uae 5 Uroleucon aeneum, Uja 5 U. jaceae, Usl 5 U. solidaginis, Uam 5 U. ambrosiae,
Uas 5 U. astronomus, Urd 5 U. rudbeckiae, Usn 5 U. sonchi, Uo 5 U. obscurum, Urp 5 U. rapunculoidis, Ujl 5 U. jaceicola, Uc 5 U.
caligatum, Uh 5 U. helianthicola, Urr 5 U. rurale, Ue 5 U. erigeronense). TrpB phylogenies were estimated using parsimony analysis of all
sites (679 nucleotides). For both data sets, trees presented are one of two most-parsimonius trees. Confidence in nodes was assessed using
bootstrapping (1,000 replicates). The number of unambiguous nucleotide changes along branches is given in parentheses. The number of changes
at first- and second-codon positions, is followed by the number of changes at third-codon positions (in b). Nucleotide changes were summed
across the entire ECOR tree, and across subsets of taxa in the Buchnera tree (see figure insert). The percent change at first and second positions,
divided by the percent change at third positions, roughly approximates Ka/Ks.
groEL. While the addition of more loci would be desirable, the consistency of the results across a variety of
loci best supports the view that the effects of mutational
bias and drift in Buchnera are sufficiently strong to
override the effect of purifying selection against the
nonoptimal codons and amino acid replacements. In
contrast to trp and leu genes, groEL does not appear to
be duplicated in Buchnera (Ohtaka and Ishikawa 1993),
and its constitutive overexpression may depend on rapid,
efficient translation of a limited number of mRNA molecules. It is not surprising, therefore, that slight codon
bias was detected for Buchnera groEL, albeit at much
lower levels than for the E. coli homologue.
Acknowledgments
We thank Joana Silva for her comments on an earlier version of this paper, and we thank two anonymous
reviewers for their helpful suggestions. This work was
supported by a Center for Insect Science postdoctoral
fellowship to J.J.W. and by an NSF grant to N.M. (DEB9527635).
LITERATURE CITED
AKASHI, H., and S. W. SCHAEFFER. 1997. Natural selection and
the frequency distributions of ‘‘silent’’ DNA polymorphism
in Drosophilia. Genetics 146:295–307.
AKSOY, S. 1995. Molecular analysis of the endosymbionts of
tsetse flies: 16S rDNA locus and over-expression of a chaperonin. Insect Mol. Biol. 4:23–29.
ANDERSSON, S. G. E., and P. M. SHARP. 1996. Codon usage
and base composition in Rickettsia prowazekii. J. Mol.
Evol. 42:525–536.
BAUMANN, P., L. BAUMANN, and M. CLARK. 1996. Levels of
Buchnera aphidicola chaperonin groEL during growth of
the aphid Schizaphis graminum. Curr. Microbiol. 32:279–
285.
96
Wernegreen and Moran
BAUMANN, P., L. BAUMANN, M. A. CLARK, and M. L. THAO.
1998a. Buchnera aphidicola: the endosymbiont of aphids.
ASM News 64:203–209.
BAUMANN, P., L. BAUMANN, C. LAI, D. ROUHBAKHSH, N. MORAN, and M. CLARK. 1995. Genetics, physiology, and evolutionary relationships of the genus Buchnera: intracellular
symbionts of aphids. Ann. Rev. Microbiol. 49:55–94.
BAUMANN, L., P. BAUMANN, N. A. MORAN, J. SANDSTROM, and
M. L. THAO. 1998b. Genetic characterization of plasmids
containing genes encoding enzymes of leucine biosynthesis
in endosymbionts (Buchnera) of aphids. J. Mol. Evol.,in
press.
BAUMANN, L., M. A. CLARK, D. ROUHBAKHSH, P. BAUMANN,
N. A. MORAN, and D. J. VOEGTLIN. 1997. Endosymbionts
(Buchnera) of the aphid Uroleucon sonchi contain plasmids
with trpEG and remnants of trpE pseudogenes. Curr. Microbiol. 35:18–21.
BERG, O. G. 1995. Kinetics of synonymous codon change for
an amino acid of arbitrary degeneracy. J. Mol. Evol. 41:
345–352.
BOCHDAREVA, E. S., N. M. LISSEN, and A. S. GIRSHOVICH.
1988. Transient association of newly synthesized unfolded
proteins with the heat-shock GroEL protein. Nature 336:
254–257.
BRACHO, A. M., D. MARTINEZ-TORRES, A. MOYA, and A. LATORRE. 1995. Discovery and molecular characterization of
a plasmid localized in Buchnera sp. bacterial endosymbiont
of the aphid Rhopalosiphum padi. J. Mol. Evol. 41:67–73.
BRYNNEL, E. U., C. G. KURLAND, N. A. MORAN, and S. G. E.
ANDERSSON. 1998. Evolutionary rates for tuf genes in endosymbionts of aphids. Mol. Biol. Evol. 15:574–582.
BUCHNER, P. 1965. Endosymbiosis of animals with plant microorganisms. Wiley and Sons, New York.
BULMER, M. 1991. The selection-mutation-drift theory of synonymous codon usage. Genetics 129:897–907.
CLARK, M. A., L. BAUMANN, and P. BAUMANN. 1992. Sequence analysis of an aphid endosymbiont DNA fragment
containing rpoB (B-subunit of RNA polymerase) and portions of rplL and rpoC. Curr. Microbiol. 25:283–290.
. 1998. Sequence analysis of a 34.7-kb DNA segment
from the genome of Buchnera aphidicola (endosymbiont of
aphids) containing groEL, dnaA, the atp operon, gidA, and
rho. Curr. Microbiol. 36:158–163.
DOUGLAS, A. 1998. Nutritional interactions in insect-microbial
symbioses: aphids and their symbiotic bacteria Buchnera.
Annu. Rev. Entomol. 43:17–37.
DYKHUIZEN, D. E., and L. GREEN. 1993. Recombination in E.
coli and the definition of biological species. J. Bacteriol.
173:7257–7268.
GARDUNO, R. A., G. FAULKNER, M. A. TREVORS, N. VATS,
and P. S. HOFFMAN. 1998. Immunolocalization of Hsp60 in
Legionella pneumonphila. J. Bacteriol. 180:505–513.
GOLOUBINOFF, P., J. T. CHRISTELLER, A. A. GATENBY, and G.
H. LORIMER. 1989. Reconstitution of active dimeric ribulose
bisphosphate carboxylase from an unfolded state depends
on two chaperonin proteins and Mg-ATP. Nature 342:884–
889.
GOLOUBINOFF, P., A. A. GATENBY, and G. H. LORIMER. 1989.
GroE heat-shock proteins promote assembly of foreign prokaryotic ribulose bisphosphate carboxylase oligomers in
Escherichia coli. Nature 337:44–47.
GOUY, M., and C. GAUTIER. 1982. Codon usage in bacteria:
correlation with gene expressivity. Nucleic Acids Res. 10:
7055–7074.
HARA, E., T. FUKATSU, K. KAKEDA, M. KENGAKU, C. OHTAKA,
and H. ISHIKAWA. 1990. The predominant protein in an
aphid endosymbiont is homologous to an E. coli heat shock
protein. Symbiosis 8:271–283.
HINDE, R. 1971. The control of the mycetocyte symbiotes of
the aphids Brevicoryne brassicae, Myzus persicae and Macrosiphum rosae. J. Insect Physiol. 17:1791–1800.
IKEMURA, T. 1981. Correlation between the abundance of E.
coli transfer RNA’s and the occurrence of respective codons
in its protein genes. J. Mol. Biol. 146:1–21.
. 1985. Codon usage and tRNA content in unicellular
and multicellular organisms. Mol. Biol. Evol. 2:13–34.
INA, Y. 1995. New methods for estimating the numbers of synonymous and nonsynonymous substitutions. J. Mol. Evol.
40:190–226.
ISHIKAWA, H. 1984. Characterization of the protein species
synthesized in vivo and in vitro by an aphid endosymbiont.
Insect Biochem. 14:417–425.
. 1987. Nucleotide composition and kinetic complexity
of the genomic DNA of an intracellular symbiont in the pea
aphid Acyrthosiphon pisum. J. Mol. Evol. 24:205–211.
LAI, C. Y., L. BAUMANN, and P. BAUMANN. 1994. Amplification of trpEG: adaptation of Buchnera aphidicola to an endosymbiotic association with aphids. Proc. Natl. Acad. Sci.
USA 91(38):19–23.
LAMBERT, J. D., and N. A. MORAN. 1998. Deleterious mutations destabilize ribosomal RNA in endosymbiotic bacteria.
Proc. Natl. Acad. Sci. USA 95:4458–4462.
LI, W. H. 1987. Models of nearly neutral mutations with particular implications for nonrandom usage of synonymous
codons. J. Mol. Evol. 24:337–345.
. 1993. Unbiased estimation of the rates of synonymous
and nonsynonymous substitution. J. Mol. Evol. 36:96–99.
MADDISON, W. P., and D. R. MADDISON. 1992. MacClade: analysis of phylogeny and character evolution. Version 3.0. Sinauer Associates, Sunderland, Mass.
MAYNARD SMITH, J., C. B. DOWSON, and B. G. SPRATT. 1991.
Localized sex in bacteria. Nature 349:29–31.
MAYNARD SMITH, J., N. SMITH, M. O’ROURKE, and B. SPRATT.
1993. How clonal are bacteria? Proc. Natl. Acad. Sci. USA
90:4384–4388.
MORAN, N. 1996. Accelerated evolution and Muller’s rachet in
endosymbiotic bacteria. Proc. Natl. Acad. Sci. USA 93:
2873–2878.
MORAN, N., and P. BAUMANN. 1994. Phylogenetics of cytoplasmically inherited microorganisms or arthropods. Trends
Ecol. Evol. 9:15–20.
MORAN, N. A., C. D. VON DOHLEN, and P. BAUMANN. 1995.
Faster evolutionary rates in endosymbiotic bacteria than in
cospeciating insect hosts. J. Mol. Evol. 41:727–731.
MORTON, B. R. 1998. Selection on the codon bias of chloroplast and cyanelle genes in different plant and algal lineages. J. Mol. Evol. 46:449–459.
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.
MUSE, S. V., and B. S. WEIR. 1992. Testing for equality of
evolutionary rates. Genetics 132:269–275.
MUTO, A., and S. OSAWA. 1987. The guanine and cytosine
content of genomic DNA and bacterial evolution. Proc.
Natl. Acad. Sci. USA 84:166–169.
OHTA, T. 1973. Slightly deleterious mutant substitutions in
evolution. Nature 246:96–98.
OHTAKA, C., and H. ISHIKAWA. 1993. Accumulation of adenine
and thymine in a groE-homologous operon of an intracellular symbiont. J. Mol. Evol. 36:121–126.
Evidence for Drift in Endosymbionts
OHTAKA, C., H. NAKAMURA, and J. ISHIKAWA. 1992. Structures
of chaperonins from an intracellular symbiont and their
functional expression in E. coli groE mutants. J. Bacteriol.
174:1869–74.
ROUHBAKHSH, D., M. A. CLARK, L. BAUMANN, N. A. MORAN,
and P. BAUMANN. 1997. Evolution of the tryptophan biosynthetic pathway in Buchnera (aphid endosymbionts):
studies of plasmid-associated trpEG within the genus Uroleucon. Mol. Phylogenet. Evol. 8:167–176.
ROUHBAKHSH, D., C. Y. LAI, C. D. VON DOHLEN, M. A.
CLARK, L. BAUMANN, P. BAUMANN, N. A. MORAN, and D.
J. VOEGTLIN. 1996. The tryptophan biosynthetic pathway of
aphid endosymbionts (Buchnera): genetics and evolution of
plasmid-associated anthranilate synthase (trpEG) within the
aphididae. J. Mol. Evol. 42:414–421.
SELANDER, R. K., D. A. CAUGANT, and T. S. WHITTAM. 1987.
Genetic structure and variation in natural populations of
Escherichia coli. Pp. 1625–1648 in F. NEIDHARDT, ed. Escherichia coli and Salmonella typhimurium: cellular and molecular biology. American Society of Microbiology, Washington, D.C.
SHARP, P., E. COWE, D. G. HIGGINS, D. C. SHIELDS, K. H.
WOLFE, and F. WRIGHT. 1988. Codon usage patterns in E.
coli, B. subtilis, S. cerevisiae, S. pombe, D. melanogaster
and H. sapiens: a review of the considerable within-species
diversity. Nucleic Acids Res. 16:8207–8211.
SHARP, P. M., and K. M. DEVINE. 1989. Codon usage and gene
expression level in Dictyostelium discoideum: highly expressed genes do ‘prefer’ optimal codons. Nucleic Acids
Res. 17:5029–39.
SHARP, P. M., and W. H. LI. 1987. The codon adaptation index—a measure of directional synonymous codon usage
bias, and its potential applications. Nucleic Acids Res. 15:
1281–1295.
97
SHARP, P., T. M. F. TUOHY, and K. R. MOSURSKI. 1986. Codon
usage in yeast: cluster analysis clearly differentiates highly
and lowly expressed genes. Nucleic Acids Res. 14:5125–
5143.
SHIELDS, D. C. 1990. Switches in species-specific codon preferences: the influence of mutation biases. J. Mol. Evol. 31:
71–80.
SHIELDS, D. C., and P. M. SHARP. 1987. Synonymous codon
usage in Bacillus subtilis reflects both translational selection
and mutational biases. Nucleic Acids Res. 16:8023–8040.
SHIELDS, D. C., P. M. SHARP, D. G. HIGGINS, and F. WRIGHT.
1988. ‘‘Silent’’ sites in Drosophila genes are not neutral:
evidence of selection among synonymous codons. Mol.
Biol. Evol. 5:704–716.
SOKAL, R. R., and F. J. ROHLF. 1981. Biometry. W.H. Freeman
and Co., New York.
SUOEKA, N. 1961. Correlation between base composition of
DNA and amino acid composition of protein. Proc. Natl.
Acad. Sci. USA 47:1141–1149.
SWOFFORD, D. L. 1993. PAUP: phylogenetic analysis using
parsimony. Version 3.1.1. Illinois Natural History Survey,
Champaign.
VAN HAM, R., A. MOYA, and A. LATORRE. 1997. Putative evolutionary origin of plasmids carrying the genes involved in
leucine biosynthesis in Buchnera aphidicola. J. Bacteriol.
179:4768–4777.
WRIGHT, F., and M. J. BIBB. 1992. Codon usage in the G1Crich Streptomyces genome. Gene 113:55–65.
WRIGHT, S. 1931. Evolution in Mendelian populations. Genetics 16:97–159.
GEOFFREY I. MCFADDEN, reviewing editor
Accepted September 24, 1998