Download Reductive evolution of resident genomes

REVIEWS
Reductive evolution of resident
genomes
Siv G.E. Andersson and Charles G. Kurland
T
he genomes of obligate Small, asexual populations are expected to
Despite the differences in
parasites, endosymbionts
accumulate deleterious substitutions and
evolutionary dynamics, the
and cellular organelles deletions in an irreversible manner, which transmission of both types of
evolve under conditions that
in the long-term will lead to mutational
resident genomes from one
are radically different from
meltdown and genome decay. Here, we
host to the next frequently
those of free-living organisms.
discuss the influence of such reductive
involves bottlenecks and repliOne distinguishing factor is processes on the evolution of genomes that cation in small populations,
that they are residents of an
replicate within the domain of a host
with little opportunity for reenvironment that is conditioned
genome.
combination between variants.
by a host genome. For examFor this reason, the resident
ple, it is possible to imagine an S.G.E. Andersson* and C.G. Kurland are in the Dept genomes accumulate deleteriof Molecular Biology, Biomedical Center, Uppsala
evolutionary sequence in which
ous mutations at a rate that is
University, S-751-24 Uppsala, Sweden.
a bacterium begins as a faculhigher than that for free-living
*tel: ⫹46 18 471 4379, fax: ⫹46 18 55 77 23,
tative resident within a cellular
organisms. Some of these dee-mail: [email protected]
domain. During the course of
leterious mutations lead to the
adaptation to the intracellular environment, the bac- loss of coding sequences, while others lead to a
terium can take one of two alternative evolutionary marked variability of resident genome architectures.
routes. On the one hand, the host cell can become de- This tendency of small asexual populations to accupendent on products provided by the activities of the mulate deleterious mutations is referred to as Muller’s
bacterial genome. In this case, the fates of the cell and ratchet3,4 (Box 1).
the resident are linked through a symbiotic dependence
The validity of Muller’s ratchet has been investithat is progressively strengthened by the tendency to gated in RNA viruses, which have extraordinarily
lose gene functions from one or other genome as a high mutation rates and are subjected to recurrent
consequence of the redundancy of their overlapping bottlenecks5–7. Co-infection provides opportunities
genomic functions. In the extreme, the two genomes for genetic exchange through recombination or segcan evolve a relationship as profound as that between mentation in some viral lineages, while others reprothe eukaryote nucleus and the mitochondria. Here, duce in a strictly asexual manner. When populations
the symbiosis initiated between the primitive eukary- of asexual and sexual viruses are forced through a seote nucleus and an aerobic bacterium has enabled the ries of bottlenecks, deleterious mutations accumulate
eukaryote to evolve an aerobic lifestyle. Gene content by genetic drift, resulting in a loss of fitness in both
is markedly reduced in mtDNAs, compared with populations. However, while this loss can be reversed
those of their eubacterial ancestors1, and some mito- in sexual viruses by recombination and/or segmentachondrial genomes have undergone so much reduc- tion during intervening periods of mass selection in
tive evolution that they seem to be on the verge of dis- large populations, loss of fitness is effectively irreversible in asexual viral populations if the rate of
appearing entirely from the eukaryotic cells2.
On the other hand, the bacterium may become an compensatory back-mutations is low7.
obligate parasite that offers no benefits to the host.
Because of their obligate intracellular lifestyles,
Thus, the fates of the host cell and the parasite are not there are few or no opportunities for resident genomes
reciprocally coupled. Instead, a more or less elaborate of bacteria and organelles to recover from the fitness
pattern of pathogenesis and defensive strategies evolves. decline caused by genetic drift during transmission
Here, too, losses of neutralized genes will strengthen from one host to the other. Indeed, mitochondrial
the dependence of the parasite on the host genome. genomes with low recombination frequencies underThe dependence can proceed so far that the parasite is go higher fixation rates for new mutations than their
unable to grow outside its host domain. However, sexually reproducing host genomes8,9. Thus, we may
whereas obligate endosymbionts are locked inside their wonder if some lineages of resident genomes, such as
hosts for ever, most pathogens must find new hosts those of obligate intracellular parasites, are driven to
regularly and, hence, are intrinsically more exposed extinction because of the effects of Muller’s ratchet.
to the external environment. Thus, the evolutionary
forces driving obligate endosymbionts are probably Mitochondria and their codes
dictated by selective factors acting on their hosts to a The most extreme examples of reductively evolved
much larger extent than those driving obligate intra- resident genomes are those of the cellular organelles.
cellular parasites.
Phylogenetic reconstructions suggest that mitochondrial
Copyright © 1998 Elsevier Science Ltd. All rights reserved. 0966 842X/98/$19.00
TRENDS
IN
MICROBIOLOGY
263
PII: S0966-842X(98)01312-2
VOL. 6 NO. 7 JULY 1998
REVIEWS
Box 1. Muller’s ratchet
Identifying the evolutionary forces that drive the fixation rates for
mutations is an important task of evolutionary biology. Nucleotide
substitutions consist of synonymous substitutions that do not
alter the protein sequence and nonsynonymous substitutions that
cause amino acid replacements. It is anticipated that some nonsynonymous substitutions will have a deleterious or slightly deleterious effect on fitness. A gradual accumulation of deleterious
mutations can induce an irreversible loss of the most-fit class of
organisms from the population5–7,56. This loss can be triggered by
high mutation rates, lack of recombination and/or small population sizes. Under these conditions, genetic drift can result in the
successive loss of the most-fit class, in a ratchet-like manner.
Such a gradual decline in the fitness of a population is referred to
as Muller’s ratchet3,4.
However, most nonsynonymous mutations are unlikely to be
fixed in the population. Indeed, Muller’s ratchet can be opposed
by high rates of compensatory mutations, sexual reproduction
and/or by large population sizes driven by purifying selection. Because resident genomes are effectively asexual and often subjected to bottlenecks during each transmission from one host to
the other, the loss of fitness induced by genetic drift will not be
slowed down to the same extent as in large, free-living bacterial
populations. Thus, abnormally higher fixation rates for nonsynonymous substitutions, as observed for intracellularly replicating
bacteria, are an indicator that Muller’s ratchet is operative in
these lineages.
U
U
C
A
G
A
C
G
A
Phe
Phe
Leu
Leu
Ser
Ser
Ser
Ser
Tyr
Tyr
Term
Term
Cys
Cys
Term
Trp
Leu
Leu
Leu
Leu
Pro
Pro
Pro
Pro
His
His
Gln
Gln
Arg
Arg
Arg
Arg
Thr
Thr
Thr
Thr
Asn
Asn
Lys
Lys
Ser
Ser
Arg
Arg
G
Ala
Ala
Ala
Ala
Asp
Asp
Glu
Glu
Gly
Gly
Gly
Gly
G
Ile
Ile
Ile
Met
Val
Val
Val
Val
C
U
G
U
G
A
A
A
C
C
C
U
U
Fig. 1. Codon assignments. Translation of the universal genetic code
requires a minimum of 24 tRNA species; one tRNA species per amino
acid for one-, two- and four-codon boxes (16 tRNAs) and two tRNA
species per amino acid for the three-codon box encoding isoleucine and
the six-codon boxes encoding leucine, serine and arginine (eight tRNAs).
In animal mitochondria, the codons AUA and AGA/G have been reassigned to methionine and serine, respectively. Consequently, the two
codons for methionine (AUA/G) and the four codons for serine (AGN)
can be translated by single tRNA species. Similarly, the two remaining
codons for isoleucine (AUU/C) and the four remaining codons for
arginine (CGN) can be translated by single tRNA isoacceptor species.
Thus, translation of the reassigned code in animal mitochondria requires
only 22 tRNA species.
TRENDS
IN
MICROBIOLOGY
genomes derive from much larger genomes. In particular, they appear to be most closely related to the
␣-proteobacteria and more specifically to an ancestor
of the Rickettsiaceae10–13. Some mitochondrial
genomes have retained ⬍1% of the gene repertoire of
modern bacteria. This marked reduction in genome
size can be partly explained by a massive transfer of
genes from the mitochondrial genome to the nuclear
genome14, a transfer that may have occurred as long
as 1–2 billion years ago when the ancestral bacterium
began to establish its symbiotic relationship with the
eukaryotes15. Nevertheless, examples of transfer events
from mitochondria to nuclear genomes as recently as
within the past 200 million years have been detected16,17,
suggesting that the process might be continuing.
A eukaryotic cell can contain hundreds of mitochondria and thousands of mitochondrial genomes.
These genomes are maternally inherited through the
cytoplasm of the egg with little or no paternal contribution. This mode of transmission effectively restricts
the exchange of genetic material between mitochondrial genomes; in effect, the egg is a bottleneck for mitochondrial genomes. Consequently, mitochondria from
multicellular organisms display characteristics that are
generally associated with small, asexual populations of
resident genomes8,9,18,19. They are the victims of drift
and Muller’s ratchet, as well as the neutralizing effect
of the host genome on redundant genes.
Animal mitochondria have diverged so far from
their free-living bacterial ancestors that they have
evolved new codes and a correspondingly specialized
transfer RNA (tRNA) ensemble. If the universal genetic code were to be translated unambiguously into
the canonical 20 amino acids, an absolute minimal set
of 24 tRNA species would be required to translate the
code. One tRNA species is required for each amino
acid that is coded by a single codon or by a two- or
four-codon box (16 in total). Two tRNA species are
required for each amino acid encoded by the threecodon box for isoleucine and by the six-codon boxes
for leucine, serine and arginine. In this minimal set,
the tRNA species would have a greater redundancy
than is normally seen; for example, each of the four
codon readers would be indifferent to the nucleotide
in codon position three. In some animal mitochondria, the number of tRNA genes has decreased below
the canonical minimum of 24. Thus, there may be as
few as 22 tRNA species20. The two missing tRNAs in
these systems correspond to an isoleucine–tRNA normally cognate to the codon AUA, and an arginine–
tRNA normally cognate to the codons AGA and
AGG. In addition, AUA is read by a methionine–
tRNA that also translates AUG. The codons AGA
and AGG now appear as termination codons or are
translated by a serine–tRNA that reads the AGN box.
In other words, amino acid reassignments in the
genetic code in animal mitochondria are demonstrably associated with changes in codon-recognition
patterns for tRNA species that permit all 64 codons
to be translated by a smaller number of tRNA species
than would be allowed by the universal code21
(Fig. 1). In these systems, the codon reassignments
264
VOL. 6 NO. 7 JULY 1998
REVIEWS
Table 1. Mutation and substitution rates in Escherichia coli, Salmonella typhimurium and Buchnera sp.a
Estimated mutation
rateb (ⴛ 10–9)
Synonymous ratec
(ⴛ 10–9)
Nonsynonymous ratec
(ⴛ 10–9)
Species
Lifestyle
Tuf
TrpB
Tuf
TrpB
Tuf
TrpB
Buchnera sp.
Intracellular
4–8
6–12
4–8
6–12
0.1–0.2
0.6–1.3
E. coli/
S. typhimurium
Free-living
4–5
9–12
0.2–0.3
3–4
<0.02
0.04–0.05
a
Data taken from Ref. 26.
Number of mutations per position per year.
c
Number of substitutions per position per year.
b
seem to be a reflection of the reductive impact of
Muller’s ratchet on the mitochondrial genomes18.
Dynamics of substitutions
A more subtle expression of the working of Muller’s
ratchet is the gradual accumulation of sublethal, deleterious mutations that lead to substitutions in the
gene sequences. Recently, the rate of sequence evolution has been determined for intracellular bacteria of
the genus Buchnera, which live as symbionts of
aphids22–25. These bacteria grow inside specialized
cells called bacteriocytes that are maternally transmitted via the egg in a manner resembling the transmission of mitochondria from one host generation to
the next. The endosymbionts supply the aphid hosts
with vital amino acids. Therefore, antibiotic treatment of the aphids results in sterility or death. As yet,
it has not been possible to cultivate the endosymbionts outside their aphid hosts, which suggests that
the symbiotic relationship is mutual. Phylogenetic reconstructions based on rRNA genes provide strong
support for the idea that the endosymbionts and their
hosts have diverged from their respective ancestors at
the same time and have co-speciated22. Indeed, the
aphids and their intracellular bacteria seem to have
coexisted in this symbiosis for ⬎100 million years22.
There is a fairly good fossil record available for the
aphids. By assuming that the symbionts are co-specific,
the dated nodes in the aphid subtree can be transferred to the bacterial subtree. These dated nodes can
then provide estimates of times of divergence, which
can, in turn, be used to calculate rates of evolution for
the endosymbionts. For example, analysis of the gene
coding for elongation factor Tu has been carried out
for five species of Buchnera26, suggesting that this
gene evolves at a rate of 1.3–2.5 ⫻ 10⫺10 substitutions
per position and per year at sites that result in amino
acid replacements (nonsynonymous substitutions). It
has also been found that there are 3.9–8.0 ⫻ 10⫺9
substitutions per position and per year at sites that have
no effects on amino acid composition (synonymous
substitutions)26.
Comparative analysis of codon-usage patterns in a
variety of genes from Buchnera provides no evidence
for selective constraints on synonymous third-codon
positions (S.G.E. Andersson, unpublished). Indeed,
synonymous substitution rates appear to be roughly
TRENDS
IN
MICROBIOLOGY
similar among genes with different expression levels,
in support of their neutrality26,27. Similar results have
been obtained from the obligate intracellular parasite
Rickettsia prowazekii, where selection does not appear
to influence base composition patterns at synonymous
third-codon positions28. Also, synonymous substitution rates appear to be similar among genes in this
species. This suggests that the genomes of Buchnera
and Rickettsia are under a randomizing mutational
pressure from Muller’s ratchet, which overwhelms
the weak selection pressure for preferred synonymous
codons. Thus, for both of these species, synonymous
rates should reflect the neutral, intrinsic mutation rate.
By contrast, measurable selective pressures for subsets of synonymous codons characterize free-living
organisms, such as Escherichia coli and Salmonella
typhimurium, and these are correlated to the abundance of the corresponding isoacceptor tRNAs29,30. In
these species, highly expressed genes have stronger
codon-usage biases and lower synonymous substitution rates than less-expressed genes. However, the mutational divergence between E. coli and S. typhimurium
has been calculated31, and the resulting estimates are
remarkably similar to the observed synonymous substitution rate in Buchnera26. This suggests that the intrinsic mutation rate may be very similar in Buchnera,
E. coli and S. typhimurium. By contrast, the substitution rates at nonsynonymous sites differ by more
than a factor of ten in these species26,27. Consequently,
amino acid replacements occur much faster in Buchnera
than in E. coli and S. typhimurium (Table 1).
These findings suggest that resident genomes, such
as those of organelles, endosymbionts and obligate
intracellular parasites, endure an enhanced rate of
deleterious substitutions8,9,18,19,26,27. This is a verification of the influence of drift and Muller’s ratchet on
their evolution.
Small, scrambled genomes
Obligate intracellular parasites, such as Rickettsia,
Chlamydia and Coxiella, have genome sizes in the
1-Mbp range19. The free-living bacteria from which
these parasites have evolved probably had genomes
that were 4–5 times larger13,19,32. Such genomic shrinkage is not a random process, in the sense that any and
all genes may be deleted from the genome: it is potentially less damaging to a parasite to lose a gene that is
265
VOL. 6 NO. 7 JULY 1998
REVIEWS
on the orientations of the repeated
sites36 (Fig. 2). Populations of freeA B
A C
C
B
living bacteria seem to accumulate
freakish genomic variants only at
D
A
very low frequencies37. In contrast,
because of their special circumstances,
genomes that replicate exclusively
D C
D B
as residents within the intracellular
or extracellular domain of another
(b)
genome evolve idiosyncratic gene
B C
C
arrangements that disrupt the conB
C D
served gene-order structures seen in
A
D
D
A
their free-living relatives.
A B
The most highly conserved geneorder structures in microbial genomes are those that encode components of the transcription and
translation systems. One such exFig. 2. Intragenomic recombination events at repeated sequences, demonstrating how the orientation of the repeats will affect the outcome of the recombination event. (a) Pairing of repeat
ample is the so-called super-ribomotifs in inverse orientation will generate an inversion of the segment (red) located between the
somal protein gene operon, which
repeat units. (b) Pairing of repeat motifs in direct orientation will generate an excision of the
comprises a total of ~40 genes: in
segment (red) located between the repeat units.
E. coli these are organized in seven
operons (secE–nusG, L11, L10,
also present in the host genome. Therefore, it is perhaps str, S10, spc and ␣) (Fig. 3). The genes within these
not surprising that many genes coding for proteins in- operons are located in close proximity to each other
volved in amino acid biosynthesis have been discarded in many bacterial and archaeal genomes38,39. Because
from the genomes of R. prowazekii and Mycoplasma there is a very small probability that 30–40 genes
genitalium33,34. Not unexpectedly, these bacterial have become linked independently of each other, it is
genomes also have a very limited set of regulatory generally believed that these shared structures repregenes33,34. The finding that a gene coding for S-adeno- sent ancient conserved genomic motifs38. Indeed,
sylmethionine synthetase is in different stages of mu- remnants of the super-ribosomal protein gene operon
tational meltdown in a number of Rickettsia species have been observed in plastid genomes, as well as in
suggests that elimination of biosynthetic genes is an the mitochondrial genome of the fresh-water protoongoing process in Rickettsia (J.O. Andersson and zoan Reclinomonas americana40, suggesting that these
S.G.E. Andersson, unpublished). A consequence of genes were clustered in an ancestral bacterial genome
these reductions in the metabolic repertoire of intra- prior to the divergence of mitochondria and chlorocellular and cell-surface parasites is that a relatively plasts from their bacterial ancestors.
larger fraction of the genome is devoted to basic
Despite the unusually high degree of gene-order
genetic functions, such as replication, transcription conservation in the ribosomal protein gene operons,
and translation.
we have identified both inversion and deletion events
Why should the depredations of Muller’s ratchet in this area of the R. prowazekii genome19,41. We suggenerate smaller genomes? The principle mechanism gest that an intrachromosomal recombination event
for excision and insertion of genomic sequences in between the two ancestral tuf genes resulted in a
bacteria seems to involve intrachromosomal recom- major inversion of the genome, after which one of the
bination at repeated sequences35. Duplications can two copies was deleted19,41. Deviations from the analways be reversed by an intrachromosomal recombi- cestral organization of these genes have also been
nation event, but once a string of nucleotides is deleted, detected in the tiny genome of M. genitalium. For this
the information cannot be recovered by an intrachro- species, numerous rearrangement events have to be
mosomal recombination event. Furthermore, the intra- inferred to explain the distribution of this set of genes
cellular habitat efficiently prevents the acquisition of into the six different gene blocks observed39.
lost genetic material via horizontal transfer from other
Additional examples of scrambled gene-order strucbacterial species. This means that deletions in the ab- tures in small parasitic genomes are provided by the
sence of effective counterselection will accumulate rRNA genes. Most of the deviations from the convenwith time. Many such deletions are probably weakly tional gene-order structure (16S–23S–5S) are associated
deleterious; these will spread to fixation relatively with bacterial genomes in the 1-Mbp range, having only
frequently in small asexual populations. Such reduc- one or two copies of each of the rRNA genes19,42,43. A
tive processes will occur at enhanced rates if there is a general survey of gene order conservation in the
selective advantage of a small genome size, owing R. prowazekii genome is consistent with the idea that
either to selection for genomes that replicate fast or to this genome is a highly derived, rearranged minimal
selection for small cell volumes19.
genome44. Thus, small sizes and scrambled gene organiIntrachromosomal recombination events result zation patterns seem to be a characteristic feature of
either in sequence deletions or inversions, depending the genomes of resident bacteria.
(a)
TRENDS
IN
MICROBIOLOGY
266
VOL. 6 NO. 7 JULY 1998
REVIEWS
(a)
nusG
L11
rif
L10
str
spc
S10
α
(c)
(b)
nusG L11 L10 rif str S10 spc α
Tu
nusG L11 L10 rif
90
ori 4
ori
str
S10 spc α
74
Escherichia coli
Bacillus subtilis
(e)
(d)
L10
29 str L11 rif
46
L11 str
L10
nusG
rif
S10 spc α
S10 spc α
Haemophilus influenzae
Mycoplasma genitalium
Tu
(f)
nusG
S10 spc α
str nusG L11 L10 rif
Rickettsia prowazekii
Fig. 3. Genomic rearrangements. (a) It is believed that an ancestor of Bacteria and Archaea had a set of transcriptional and translational genes located in close proximity to each other in the following order: nusG, L11 (rpsLA), L10 (rpsKJ), rif (rpoBC), str
(rpsLG–tuf–fus), S10 (rpsJ–rplCDWB–rpsS–rplV–rpsC–rplP–rpmC–rpsQ), spc (rplNXE–rpsNH–rplFR–rpsE–rpmD–rplO–prlA–rpmJ) and
␣ (rpsMKD–rpoA–rplQ). These segments are conservatively arranged in the genomes of (b) Bacillus subtilis and (c) Escherichia
coli, but are scattered around the genomes of (d) Haemophilus influenzae, (e) Mycoplasma genitalium and (f) Rickettsia
prowazekii. The organization of these genes in the R. prowazekii genome is consistent with an intrachromosomal recombination
event at the tuf genes41. The circumferences of the circles are directly proportional to the genome sizes. Data taken from Ref. 39.
Genomic extinction
Transitions to new environmental niches can induce
the loss of metabolic functions previously supplied by
a long-term endosymbiont. For example, photosynthesis has been secondarily lost in some plants, algae
and protists that live as parasites of other plants. Not
suprisingly, the chloroplast genomes of these species
are markedly reduced and the photosynthetic genes
have either been eliminated or are present as pseudogenes45–47. The presence of a residual plastid genome
is indicative of a role for chloroplast gene products in
metabolic processes other than photosynthesis. The
TRENDS
IN
MICROBIOLOGY
most extreme examples of such reductive processes
are the complete loss of mitochondrial genomes in
eukaryotes that have adapted to anaerobic or obligate
intracellular lifestyles, their past being evident only
from the presence of the mitochondrial heat-shock
proteins Hsp10, Hsp60 and Hsp70 (Refs 48–50).
Another interesting case of extreme reductive evolution is the incorporation of a green algae by a group of
marine protists called chlorarchinophytes, in which the
remnants of the invading nuclear genome, the nucleomorph, have been reduced to ⬍1 Mbp (Refs 51,52).
In some other groups of algae, unusual membrane
267
VOL. 6 NO. 7 JULY 1998
REVIEWS
Questions for future research
• How small must sexual populations be to behave as asexual
populations (i.e. populations where there is little or no recombination between individuals)?
• Are bottle-necked subpopulations, which start growth cycles from
a limited number of individuals, a valuable source of mutational
variation for large, sexual populations?
• Does the successful incorporation of resident genomes also
serve as a bottleneck for their hosts, by eliminating a large number of infected hosts, in the case of pathogens, or by boosting
the population with infected offspring, in the case of symbionts?
structures provide support for a similar kind of secondary endosymbiotic event, although in these species
the nucleomorph has become extinct53,54. Similarly,
the presence of four membranes enclosing the plastids
of apicomplexan parasites suggests that their origin can
be traced back to a secondary invasion event55. Transitions to intracellular environments are important
evolutionary processes that enable biological diversity
and novel life forms to be created over relatively short
time periods. However, the genetic isolation of the invading genome could ultimately lead to its extinction.
Conclusions
Duplication events can, in principle, be reversed by
compensatory deletion events. By contrast, deleted sequences cannot be restored by compensatory duplication events. Because of their asexual or nearly asexual
population structure, deletions in resident genomes will
have a higher probability of becoming fixed than in the
populations of free-living bacteria. The accumulation
of small deletion events and the gradual disruptions or
loss of genes in the resident genomes are, therefore, expected. The elevated fixation rates for deleterious substitution and deletion will adversely influence the fitness of the parasite5–7,56. Eventually, it might not be
able to survive the defensive stratagems of the host or
competition from younger, less-attenuated parasites.
Accordingly, the rich and protective cytoplasmic environment, which initially is so supportive for a newly established intracellular pathogen, might ultimately contribute to its decline and extinction. For endosymbionts
and cellular organelles, host dependence might limit
the accumulation of deleterious mutations. However, a
replacement of one resident with another could also
happen in mutualistic endosymbionts; indeed, the host
could be selected to facilitate such changes if a more
ancient symbiont is not as efficient as a more recent
one. In effect, the evolution of intracellular parasitism
and symbiosis might be characterized by waves of obligate intracellular bacteria that replace each other at
rates determined by Muller’s ratchet.
Acknowledgements
The support of the National Science Research Council of Sweden is
gratefully acknowledged.
References
1 Palmer, J.D. (1997) Nature 387, 454–455
2 Palmer, J.D. (1997) Science 275, 790–791
3 Muller, H.J. (1964) Mutat. Res. 1, 2–9
TRENDS
IN
MICROBIOLOGY
Felsenstein, J. (1974) Genetics 52, 737–756
Chao, L. (1990) Nature 348, 454–455
Chao, L. et al. (1992) Evolution 46, 289–299
Chao, L. (1997) Gene 205, 301–308
Lynch, M. (1996) Mol. Biol. Evol. 13, 209–220
Lynch, M. (1997) Mol. Biol. Evol. 14, 914–925
Yang, D. et al. (1985) Proc. Natl. Acad. Sci. U. S. A. 82, 4443–4447
Gray. M.W. et al. (1989) Proc. Natl. Acad. Sci. U. S. A. 86, 2267–2271
Viale, A. and Arakaki, A.K. (1994) FEBS Lett. 341, 146–151
Olsen, G.J., Woese, C.R. and Overbeek, R. (1994) J. Bacteriol. 176, 1–6
Brennicke, A. et al. (1993) FEBS Lett. 325, 140–145
Knoll, A. (1992) Science 256, 622–627
Covello, P.S. and Gray, M. (1992) EMBO J. 11, 3815–3820
Nugent, J.M. and Palmer, J.D. (1991) Cell 66, 473–481
Kurland, C.G. (1992) Bioessays 14, 709–714
Andersson, S.G.E. and Kurland, C.G. (1995) Biochem. Cell Biol.
73, 775–787
20 Andersson, S. et al. (1981) Nature 290, 457–465
21 Andersson, S.G.E. and Kurland, C.G. (1991) Mol. Biol. Evol. 8,
530–544
22 Moran, N.A. et al. (1993) Proc. R. Soc. London Ser. B 253,
167–171
23 Baumann, P. et al. (1995) Ann. Rev. Microbiol. 49, 55–94
24 Baumann, P. et al. (1997) BioScience 47, 12–20
25 Baumann, P. et al. (1998) ASM News 64, 203–209
26 Brynnel, E.U. et al. (1998) Mol. Biol. Evol. 15, 574–582
27 Moran, N.A. (1996) Proc. Natl. Acad. Sci. U. S. A. 93, 2873–2878
28Andersson, S.G.E. and Sharp, P.M. (1996) J. Mol. Evol. 42, 525–536
29 Ikemura, T. (1981) J. Mol. Biol. 146, 1–21
30 Dong, H., Nilsson, L. and Kurland, C.G. (1996) J. Mol. Biol.
260, 649–663
31 Berg, O.G. and Martelius, M. (1995) J. Mol. Evol. 41, 449–456
32 Weisburg, W.G. et al. (1989) J. Bacteriol. 171, 6455–6457
33 Fraser, C.M. et al. (1995) Science 270, 397–403
34 Andersson, S.G.E. et al. (1996) Microb. Comp. Genomics 1, 293–315
35 Krawiec, S. and Riley, M. (1990) Microbiol. Rev. 54, 502–598
36 Segall, A., Mahan, M. and Roth, J.R. (1988) Science 241, 1314–1318
37 Orston, L.N., Neidle, E.L. and Houghton, J.E. (1990) in
The Bacterial Chromosome (Drlica, K. and Riley, M., eds),
pp. 325–334, ASM Press
38 Keeling, P.J., Charlebois, R.L. and Doolittle, W.F. (1994) Curr.
Opin. Genet. Dev. 4, 816–822
39 Sicheritz-Pontén, T. and Andersson, S.G.E. (1997) Microb.
Comp. Genomics 2, 123–139
40 Lang, B.F. et al. (1997) Nature 387, 493–496
41 Syvänen, A.C. et al. (1996) J. Bacteriol. 178, 6192–6199
42 Andersson, S.G.E. et al. (1995) J. Bacteriol. 177, 4171–4175
43 Munson, M.A. et al. (1993) Gene 137, 171–178
44 Andersson, J.O. and Andersson, S.G.E. (1997) Microbiology 143,
2783–2795
45 Wolfe, K.H. et al. (1992) Proc. Natl. Acad. U. S. A. 89,
10648–10652
46 Gockel, G. et al. (1994) Curr. Genet. 26, 256–262
47 Wilson, R.J.M. et al. (1996) J. Mol. Biol. 261, 155–172
48 Bui, E.T.N. et al. (1996) Proc. Natl. Acad. Sci. U. S. A. 93,
9651–9656
49 Horner, D.S. et al. (1996) Proc. R. Soc. London Ser. B 263,
1053–1059
50 Roger, A.J. et al. (1996) Proc. Natl. Acad. Sci. U. S. A. 93,
14618–14622
51 Gilson, P.R. and McFadden, G.I. (1996) Proc. Natl. Acad. Sci.
U. S. A. 93, 7737–7742
52 Gilson, P.R. and McFadden, G.I. (1997) Bioessays 19, 167–173
53 McFadden, G. and Gilson, P. (1995) Trends Ecol. Evol. 10, 12–17
54 Palmer, J.D. and Delwiche, C.F. (1996) Proc. Natl. Acad. Sci.
U. S. A. 93, 7432–7435
55 Köhler, S. et al. (1997) Science 275, 1485–1489
56 Andersson, D.I. and Hughes, D. (1995) Proc. Natl. Acad. Sci.
U. S. A. 93, 906–907
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
268
VOL. 6 NO. 7 JULY 1998