Mutational Patterns Cannot Explain Genome Composition: Are There Any Neutral Sites in the Genomes of Bacteria?

article has not abstract

Published in the journal: . PLoS Genet 6(9): e32767. doi:10.1371/journal.pgen.1001104
Category: Perspective
doi: 10.1371/journal.pgen.1001104


article has not abstract

The dissection of natural selection and neutral processes remains a core problem for molecular evolutionary biologists. One of the longest-standing controversies concerns the causes of genome base composition, notably the variation in the sum of G and C content (GC) between 17% and 75% in bacteria. Sueoka argued very early that GC content variation is driven by mutational biases and, as this bias affects non-synonymous sites, protein evolution might also be largely driven by neutral forces [1]. Later, Muto and Osawa showed that 4-fold degenerate positions in codons exhibit the largest range of GC content (GC4), whereas the non-degenerate second codon positions (GC2) exhibit the narrowest (Figure 1) [2]. As the footprint of genomic GC variation is most evident in those sites under the least selective constraint for amino acid composition, it has become accepted that GC content variation is primarily driven by neutral mutational effects and has little adaptive relevance [2].

The GC composition of genomes is strongly correlated with second codon (GC<sub>2</sub>) and 4-fold degenerate positions (GC<sub>4</sub>) <em class=&quot;ref&quot;>[<b>2</b>]</em>.
Fig. 1. The GC composition of genomes is strongly correlated with second codon (GC2) and 4-fold degenerate positions (GC4) [2].
Second codon positions show low variability due to purifying selection on non-synonymous changes. 4-fold degenerate positions vary between 5% and 97% GC among published genomes. In the classical neutral scenario (red), 4-fold degenerate positions are nearly neutral and their composition results essentially from mutational patterns. These patterns are modified in bacteria that lose repair genes, such as mutators, which show additional AT pressure (grey area) [19]. In the selectionist view (blue), the composition of 4-fold degenerate positions results from selection for GC content, the mutational patterns are AT-rich relative to genome composition, and there are no neutral positions. Naturally, this is an idealized view of genomes that code for many additional overlapping signals that are under selection, e.g., codon usage bias, regulatory signals, etc.

Two papers in the current issue of PLoS Genetics aim to test whether the variation in bacterial genomic GC content results directly from mutation biases. Far from observing variation in mutational patterns concordant with the range of GC content, Hildebrand et al. [3], and Hershberg and Petrov [4] independently point to a strong and consistent AT pressure on bacterial genomes, whereby de novo GC → AT mutations arise much more commonly than the reverse. Hershberg predicts that most bacterial genomes, if left entirely vulnerable to mutation, would approach an equilibrium GC content of 20%–30%, close to the highly reduced genomes of endosymbionts [5]. Discounting a rather implausible scenario whereby nearly all diverse GC-rich taxa are converging towards a low GC content, one is forced to conclude that the excess A and T generated by mutation bias (AT pressure) is lost over time. If so, mutational patterns are not strongly shaping genomes after all, and something else is keeping GC contents up.

Hildebrand and co-workers analyze polymorphism data from 149 phylogenetically diverse species corresponding to a wide range of GC content. A major strength of this analysis is that it tests for a number of possible confounders that might explain the excess of GC → AT changes, including variation in mutation rates, sequencing errors, and violations of the infinite sites assumption. The proportion of GC ↔ AT changes that are GC → AT (Z) is almost always >0.5, and is positively correlated with GC4. This means that AT pressure is strongest in GC-rich genomes. For the most GC-poor genomes, the ratio is reversed (Z<0.5), but this might result from violation of the infinite sites assumption at extreme GC content. In fact, the extreme AT-rich genomes of Buchnera do have Z = 0.5 [6].

Hershberg and Petrov exploit full genome data of five very recently evolved “clonal pathogens”, presumably under relaxed selection, allowing precise detection of mutational patterns. This more limited dataset includes no extreme GC-poor genomes. On the other hand, the availability of a large number of SNPs and of an outgroup allows the comparison of patterns within and between species. Consistent with the results of Hildebrand et al., Hershberg and Petrov find an excess of GC → AT mutations in synonymous, non-synonymous, and intergenic sites. Comparisons with the outgroup species suggest this is not caused by loss of repair genes, and that it abates over greater phylogenetic distances (i.e., between “species”). This pattern is similar to that previously found in E. coli [7], and reflects the action of purifying selection (or a process that mimics selection) preferentially removing AT-enriching mutations over time. Hershberg and Petrov's study also highlights the significance of weaker purifying selection in newly emerged pathogens, as shown in Shigella strains [7]. Strikingly, they find no evidence for a correlation between predicted GC contents at mutational equilibrium and extant base composition, suggesting that mutational bias might have no role in shaping genome composition. Hildebrand et al. show a similar qualitative bias, but predicted equilibrium values vary between 5% and 90% GC. As methods and datasets differ in the two studies, further analyses will be required to shed light on this issue.

Taken together, the evidence for a common mutational pressure towards low GC is clear. The process maintaining base composition in GC-rich genomes must be very strong, because a genomic GC content of 75% corresponds to a GC4 of nearly 100% (Figure 1). This represents a ∼70% gap with Hershberg and Petrov's predicted mutational equilibrium. Two distinct processes might be at work: biased gene conversion (BGC) and natural selection.

In certain eukaryotes, BGC results from recombination between heterologous sequences preferentially removing AT polymorphisms [8]. Contrary to sexual eukaryotes, allelic recombination in bacteria requires horizontal transfer. As a result, rates of recombination between, and even within, different bacterial species are notoriously variable. Consistent with the action of BGC, ecologically isolated endosymbionts do not recombine and have extremely rich AT genomes [5], and regions of high recombination in E. coli are also GC rich [9]. Yet, Hildebrand et al. found qualitatively similar results when excluding taxa with evidence for recombination. Hershberg and Petrov mostly use nearly clonal genomes and still find a large gap between mutation patterns and genome composition. While available evidence suggests a weak role for BGC in the variation of GC content in bacteria, it is very difficult to completely rule out a role for BGC because it purges AT polymorphisms just like natural selection. As a result, recently emerged pathogens with an excess of AT polymorphisms experience both weakened selection and decreased recombination, both of which could potentially explain a decrease in GC content. More research is needed on the impact of BGC in bacterial genomes.

The alternative to BGC is that high GC contents are selectively maintained. Many explanations for GC content variation have been proposed (summarized in Table 1). GC content variation is most marked at synonymous and intergenic sites. Hence, any selective explanation for this variation forces us to turn the traditional concept of the “neutral site” on its head (Figure 1). In this new view, no single position is evolving neutrally in genomes. As a result, 4-fold degenerate positions are not the closest proxy to mutational patterns, but the result of selection for genomic GC content. If so, we are facing a seismic shift of paradigm in molecular evolution. Detection of adaptive features such as codon bias or amino acid frequencies currently rely on a background null hypothesis assumed to reflect neutrality. Neutral models are also the basis of coalescent-based studies of bacterial demography. If there are no neutral positions, then there is no neutral null by which to detect adaptation and we are required to first superimpose selection leading to genome composition in evolutionary studies.

Tab. 1. Variables Historically Proposed to Explain GC Variation in Prokaryotes.
Variables Historically Proposed to Explain GC Variation in Prokaryotes.

Previous selective explanations for GC content variation are wide-ranging and include considerations of the cost and availability of nucleotides [10], aerobiosis [11], and genome length [12] (Table 1). Metagenomics analyses indicate a strong environmental component to GC content variation [13], [14], and it is intriguing that the most GC-rich taxa yet sequenced have very large genomes and live in the soil. Any selective explanation for GC content must tackle the problem of small selection coefficients at individual sites. This has been a long-standing argument against selection for temperature adaptation shaping mammalian isochores [8], [15]. However, bacteria have smaller genomes and supposedly much larger effective population sizes than mammals. This might facilitate the selection of mild-effect polymorphisms [16].

Even if one discovers a source of selection for GC content, basic questions will remain. For example, does GC variation reflect differences in the selective optima or just differences in the strength of selection? These and previous studies suggest that adoption of intimate associations with eukaryotes leads to a reduction in the effective population size and to AT enrichment, possibly due to less efficient purging of GC → AT mutations (but see [17]). But does it follow that GC-rich genomes are universally desirable, yet only achievable for taxa with a very large effective population size? Alternatively, intermediate GC contents might sometimes be optimal, e.g., because of trade-offs between traits associated with different explanatory variables. In this latter view, GC content variation would emerge through a combination of variation in selective optima and effective population sizes. One further intriguing question is, why haven't mutational patterns evolved towards generating the optimal composition in genomes? If it is confirmed that selection and mutation biases are always antagonistic in GC-rich genomes, what does this reveal about the mutation process?

Finally, are such biases peculiar to bacteria? In Arabidopsis thaliana, mutational patterns are also AT rich [18], and in mammals and birds there is evidence linking recombination rates with the rise in frequency of GC polymorphisms and isochore structure [8]. Could all such patterns be universally linked to the same biological processes? The ever-expanding sequencing output should soon allow extensive comparative studies to shed a great deal of light on these mysteries.


1. SueokaN

1961 Correlation between base composition of deoxyribonucleic acid and amino acid composiiton of protein. Proc Natl Acad Sci U S A 47 1141 1149

2. MutoA


1987 The guanine and cytosine content of genomic DNA and bacterial evolution. Proc Natl Acad Sci USA 84 166 169

3. HildebrandF



2010 Evidence of selection upon genomic GC-content in bacteria. PLoS Genet 6 e1001107 doi:10.1371/journal.pgen.1001107

4. HershbergR


2010 Evidence that mutation is universally biased towards AT in bacteria. PLoS Genet 6 e1001115 doi:10.1371/journal.pgen.1001115

5. MoranNA

1996 Accelerated evolution and Muller's rachet in endosymbiotic bacteria. Proc Natl Acad Sci U S A 93 2873 2878

6. MoranNA



2009 The dynamics and time scale of ongoing genomic erosion in symbiotic bacteria. Science 323 379 382

7. BalbiKJ



2009 The temporal dynamics of slightly deleterious mutations in Escherichia coli and Shigella spp. Mol Biol Evol 26 345 355

8. DuretL


2009 Biased gene conversion and the evolution of mammalian genomic landscapes. Annu Rev Genomic Human Gen 10 285 311

9. TouchonM





2009 Organised genome dynamics in the Escherichia coli species results in highly diverse adaptive paths. PLoS Genet 5 e1000344 doi:10.1371/journal.pgen.1000344

10. RochaEPC


2002 Competition for scarce resources might bias bacterial genome composition. Trends Genet 18 291 294

11. NayaH





2002 Aerobiosis increases the genomic guanine plus cytosine content (GC%) in prokaryotes. J Mol Evol 55 260 264

12. HeddiA





1998 Molecular characterization of the principal symbiotic bacteria of the weevil Sitophilus oryzae: a peculiar G+C content of an endocytobiotic DNA. J Mol Evol 47 52 61

13. FoerstnerKU

von MeringC



2005 Environments shape the nucleotide composition of genomes. EMBO R 6 1208 1213

14. RomeroH




2009 Oxygen and guanine-cytosine profiles in marine environments. J Mol Evol 69 203 206

15. BernardiG





1985 The mosaic genome of warm-blooded vertebrates. Science 228 953 958

16. DaubinV


2003 G+C structuring along the genome: a common feature in prokaryotes. Mol Biol Evol 20 471 483

17. McCutcheonJP



2009 Origin of an alternative genetic code in the extremely small and GC-rich genome of a bacterial symbiont. PLoS Genet 5 e1000565 doi:10.1371/journal.pgen.1000565

18. OssowskiS





2010 The rate and molecular spectrum of spontaneous mutations in Arabidopsis thaliana. Science 327 92 94

19. LindPA


2008 Whole-genome mutational biases in bacteria. Proc Natl Acad Sci U S A 105 17878 17883

20. BartonNH


1998 Why sex and recombination? Science 281 1986 1990

21. TraversAA

2004 The structural basis of DNA flexibility. Philos Transact A Math Phys Eng Sci 362 1423 1438

22. XiaX



2003 Effects of GC content and mutational pressure on the lengths of exons and coding sequences. J Mol Evol 56 362 370

23. OchmanH


2001 Genes lost and genes found: evolution of bacterial pathogenesis and symbiosis. Science 292 1096 1099

24. McEwanCE



1998 Nitrogen-fixing aerobic bacteria have higher genomic GC content than non-fixing species within the same genus. Hereditas 128 173 178

25. SeligmannH

2003 Cost-minimization of amino acid usage. J Mol Evol 56 151 161

26. Vieira-SilvaS


2009 An assessment of the impacts of molecular oxygen on the evolution of proteomes. Mol Biol Evol 25 1931 1942

27. MendezR




2010 Mutation bias favors protein folding stability in the evolution of small populations. PLoS Comput Biol 6 e1000767 doi:10.1371/journal.pcbi.1000767

28. GaltierN


1997 Relationships between genomic G+C content, RNA secondary structures, and optimal growth temperature in prokaryotes. J Mol Evol 44 632 636

29. ForsdykeDR

1996 Different biological species “broadcast” their DNAs at different (G+C)% “wavelengths”. J Theor Biol 178 405 417

30. KagawaY





1984 High guanine plus cytosine content in the third letter of codons of an extreme thermophile. DNA sequence of the isopropylmalate dehydrogenase of Thermus thermophilus. J Biol Chem 259 2956 2960

31. MustoH





2004 Correlations between genomic GC levels and optimal growth temperatures in prokaryotes. FEBS Lett 573 73 77

32. XiaX




2002 Genomic changes in nucleotide and dinucleotide frequencies in Pasteurella multocida cultured under high temperature. Genetics 161 1385 1394

33. SingerCE


1970 Sunlight ultraviolet and bacterial DNA base ratios. Science 170 822 826

34. PalmeiraL



2006 UV-targeted dinucleotides are not depleted in light-exposed prokaryotic genomes. Mol Biol Evol 23 2214 2219

Genetika Reprodukční medicína

Článek vyšel v časopise

PLOS Genetics

2010 Číslo 9

Nejčtenější v tomto čísle
Zapomenuté heslo

Nemáte účet?  Registrujte se

Zapomenuté heslo

Zadejte e-mailovou adresu, se kterou jste vytvářel(a) účet, budou Vám na ni zaslány informace k nastavení nového hesla.


Nemáte účet?  Registrujte se