Epistasis correlates to genomic complexity

Rafael Sanjuán* and Santiago F. Elena

Instituto de Biologı́a Molecular y Celular de Plantas, Consejo Superior de Investigaciones Cientı́ficas-UPV, 46022 València, Spain

Edited by Tomoko Ohta, National Institute of Genetics, Mishima, Japan, and approved July 28, 2006 (received for review June 5, 2006)

Whether systematic genetic interactions (epistasis) occur at the Results

genomic scale remains a challenging topic in evolutionary biology. We first compiled information from 21 previous works that
Epistasis should make a significant contribution to variation in sought for epistasis in fitness traits. These studies relied on a
complex traits and influence the evolution of genetic systems as variety of strategies. Many were based on mutation accumulation
sex, diploidy, dominance, or the contamination of genomes with experiments, sometimes combined with chemical mutagenesis,
deleterious mutations. We have collected data from widely differ- whereas in other cases, site-directed mutagenesis was done. In
ent organisms and quantified epistasis in a common, per-genera- some cases, mutations were combined by means of the appro-
tion scale. Simpler genomes, such as those of RNA viruses, display priate crosses, whereas in other cases, crosses were done to
antagonistic epistasis (mutations have smaller effects together quantify recombinational load or to characterize the offspring
than expected); bacterial microorganisms do not apparently devi- fitness distribution. Fig. 1 and Table 2, which is published as
ate from independent effects, whereas in multicellular eukaryotes, supporting information on the PNAS web site, provide a synopsis
a transition toward synergistic epistasis occurs (mutations have of these studies. Among viruses, four of six studies found
larger effects together than expected). We propose that antago- antagonistic epistasis, one was inconclusive, and one found weak
nistic epistasis might be a property of compact genomes with few synergistic epistasis. In bacteria, one study found antagonism,
nonpleiotropic biological functions, whereas in complex genomes, whereas another detected no average deviation in either sense.
synergism might emerge from mutational robustness. Among unicellular eukaryotes, one-half of the studies concluded
Downloaded from https://www.pnas.org by Universidad Autonoma de Nuevo Leon on August 19, 2024 from IP address 148.234.81.252.

that there was no average epistasis, whereas the other half found
mutation 兩 robustness 兩 antagonism 兩 evolution 兩 synergism some synergism. Finally, among pluricellular eukaryotes, seven
of nine studies detected variable degrees of synergism in at least

F ar from being independent units, genes usually interact in

complex networks whose expression is tightly regulated and
coordinated. As a consequence, the effect of mutations in
some fitness-related traits, one study reported antagonism, and
one was inconclusive. Hence, despite the existence of some
disparate cases, we found that most studies reporting antago-
different genes often deviates from what would be expected by nistic epistasis corresponded to organisms with simple genomes,
looking at their separate effects. This deviation, called epistasis, whereas as complexity increased, cases of synergistic epistasis
can be antagonistic or synergistic depending on whether muta- became more frequent (Kendall’s ␶b ⫽ ⫺0.5691; 17 df; P ⫽
tional effects overlap or reinforce each other, respectively (1). 0.001).
The average direction of epistasis determines the efficiency of However, it could be argued that the validity of the conclu-
natural selection in purging deleterious mutations from popu- sions drawn from the above studies might be limited by their
lations and hence, is important to many different evolutionary methodological diversity. The most efficient way to avoid this
theories, including those seeking to explain the origin of sexual potential problem is probably to focus on data sets for which
reproduction (2), the evolution of ploidy (3), dominance (4), the unambiguous determinations of the number of mutations per
accumulation of deleterious mutations in small populations (5), genotype were provided and the effect of single mutations is
or reproductive isolation (6). available. When the fitness effects of well defined mutations
Several studies have measured epistasis for fitness in model have been estimated alone and in combination, a quantitative,
organisms, but it is generally believed that no general conclu- standardized measure of epistasis can be obtained by calculating
sions about the average direction of epistasis can be drawn. the difference between the observed fitness and the product of
However, it should be noted that the genomic properties of these fitness values associated with the single mutations as:
model organisms widely differ and, therefore, a common pattern

写
of epistasis must not be necessarily expected. Higher complexity, n
in the form of increased number of genes, modularity, alternative ␧1,...i,...n ⫽ W1,...i,...n ⫺ W j, [1]
metabolic pathways, or multiple regulatory elements, could j⫽1
modify the effects of mutational perturbations and the proper-
ties of genetic interactions. Nonetheless, comparing reported where Wj is the relative fitness of the genotype carrying mutation
values for epistasis across widely different organisms may pro- j, W1,...i,...n is the relative fitness of the genotype carrying muta-
vide valuable insights into the evolution of gene interactions. tions 1,. . .i,. . .n and ␧ is the epistasis coefficient between loci
Here, we do this across-species comparison by focusing on 1,. . .i,. . .n (1). Positive and negative ␧ values indicate antago-
average epistasis; this does not preclude that variation in ep- nistic and synergistic epistasis, respectively.
istasis may also be important for many evolutionary processes, Five data sets provided fitness values for genotypes carrying
such as narrowing the range of parameter values at which sex and mutations alone and in combination. The model organisms used
recombination may evolve (7). We first searched the literature in these studies were the vesicular stomatitis virus (VSV) (8), the
for articles that studied epistasis in fitness traits. Then, we
analyzed a subset of data that allowed us to obtain a standardized
measure of epistasis and thus, to do a quantitative across-species Author contributions: R.S. designed research; R.S. and S.F.E. performed research; R.S. and
S.F.E. analyzed data; and R.S. wrote the paper.
comparison. The results showed that average epistasis correlates
The authors declare no conflict of interest.
to genome complexity. In simpler genomes, such as those of
viruses and some bacteria, interactions tend to be antagonistic. This paper was submitted directly (Track II) to the PNAS office.

In unicellular eukaryotes, there seems to be no average deviation Abbreviation: VSV, vesicular stomatitis virus.
from independent effects, whereas in higher eukaryotes, a *To whom the correspondence should be addressed. E-mail: [email protected].
transition toward synergism occurs. © 2006 by The National Academy of Sciences of the USA

14402–14405 兩 PNAS 兩 September 26, 2006 兩 vol. 103 兩 no. 39 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0604543103

 Table 1. Basic genomic properties for the five species involved in
this study, as well as the average selection coefficient against
single deleterious mutations (s៮ or h៮ s in the case of diploids)
Genome length, bp Genes Ploidy s៮ (or h៮ s)

VSV 1.1 ⫻ 104 5 Haploid 0.235

E. coli 4.6 ⫻ 106 4,398 Haploid 0.027
S. cerevisiae 1.3 ⫻ 107 5,863 Diploid 0.020
Aspergillus* 3.1 ⫻ 107 9,457 Diploid 0.019
D. melanogaster 1.8 ⫻ 108 14,651 Diploid 0.010

*The data correspond to A. nidulans.

coefficient and complexity was observed (Fig. 2; Spearman’s ␳S

⫽ ⫺0.315; 249 df; P ⬍ 0.001). The negative correlation and its
statistical significance held regardless (i) the five species were
removed from the data set one by one, indicating that our
conclusions did not depend on particularities of any single
species; (ii) all lethal genotypes were excluded from the analysis;
and (iii) for D. melanogaster, only productivity or only male
mating success were used instead of the total fitness measure.
Finally, although the above data were inevitably collected from
experiments varying in some methodological aspects, these
factors were very unlikely to have produced an artefactual
correlation between complexity and epistasis (Supporting Text,
Downloaded from https://www.pnas.org by Universidad Autonoma de Nuevo Leon on August 19, 2024 from IP address 148.234.81.252.

Fig. 1. Compilation of results from articles that studied epistasis in fitness

traits. Each long bar represents a study in which significantly antagonistic or
which is published as supporting information on the PNAS web
synergistic epistasis was found. Short bars represent cases in which some site).
nonsignificant epistasis was detected. Flat bars represent studies that found
no directional epistasis at all. Two studies are not included in the figure. One Discussion
analyzed data from vaccine design experiments in various RNA viruses and The sign of epistasis should depend on how loci interact to
found no general pattern of epistasis (45). The other analyzed data from express characters. If each of k loci is necessary for the expres-
sexual crosses in several fungi and plant species (46). In fungi, the data were sion of a fitness trait and, inasmuch as the fitness disadvantage
inconclusive, whereas in plants, they indicated synergistic epistasis. More
of the kth mutant would be similar to that of each single mutant,
detailed information about all 21 studies can be found in Table 2, which is
published as supporting information on the PNAS web site. FMDV, Foot-and-
epistasis between deleterious mutations should be antagonistic.
mouth disease virus; HIV-1, HIV type 1; C. elegans, the nematode Caenorhab- In general, antagonistic epistasis should preferentially take place
ditis elegans; M. guttatus, the monkeyflower Mimulus guttatus. in genomes with few genes and little redundancy, or few alter-
native metabolic pathways, because multiple mutations will
often hit essential bits of the same functional module. Genome
bacterium Escherichia coli (9), the yeast Saccharomyces cerevisiae compactness, with few nonfunctional regions, should enhance
(10), the mold Aspergillus niger (11), and the fruit fly Drosophila this effect. For example, in VSV, two random mutations have
melanogaster (12). In all cases, mutations were scattered more than a one-fifth chance of hitting the same gene. In good
throughout the genome and randomly or systematically com- agreement with the above hypothesis, it has been shown in silico
bined. After expressing fitness into a common per-generation that reducing genome compactness through the loss of some
scale, we calculated the mean epistatic coefficient between dispensable functions leads to a shift from antagonistic epistasis
deleterious mutations for each species by using Eq. 1. For VSV, toward multiplicative effects (13). Whereas the above multiple-
the average epistasis coefficient was significantly positive (␧៮ ⫽ hitting argument should be of clear relevance for short genomes,
0.109 ⫾ 0.041; P ⫽ 0.001), whereas for E. coli, ␧៮ was also positive
but did not significantly differ from zero (0.034 ⫾ 0.040; P ⫽
0.163). For S. cerevisiae, ␧៮ was nearly centered on zero (␧៮ ⫽
⫺0.001 ⫾ 0.001; P ⫽ 0.633). For A. niger, ␧៮ was significantly
negative for the whole data set (␧៮ ⫽ ⫺0.063 ⫾ 0.029; P ⫽ 0.010),
which included some presumably lethal genotypes. Even if the
putative lethal genotypes were removed from the analysis,
epistasis remained significantly synergistic on average (␧៮ ⫽
⫺0.009 ⫾ 0.005; P ⫽ 0.046). Finally, for D. melanogaster, two
fitness-related traits were measured: productivity and male
mating success. Using the product of both parameters as a total
fitness measure, we found that epistasis was significantly syner-
gistic on average (␧៮ ⫽ ⫺0.166 ⫾ 0.066; P ⫽ 0.004).
Some genomic properties of the above five species are shown
in Table 1. Whatever measure of complexity is adopted, it seems
unquestionable that the rank-order between species genomic
EVOLUTION

complexity is VSV ⬍ E. coli ⬍ S. cerevisiae ⬍ A. niger ⬍ D.

melanogaster. The magnitude of epistasis strongly differed
among the five species (Kruskal–Wallis’ H ⫽ 21.842; 4 df; P ⬍
0.001) and, using the rank-order classification for genome com- Fig. 2. Mean per generation epistasis coefficient (␧៮ ⫾ SEM) for five species
plexity, a significant negative correlation between the epistasis of increasing complexity.

Sanjuán and Elena PNAS 兩 September 26, 2006 兩 vol. 103 兩 no. 39 兩 14403
 as those of viruses, its power rapidly vanishes as the number of loci and, therefore, produces zero epistasis values. It is therefore
independent biological functions and genetic redundancy in- important to clarify that our multiple-hitting argument does not
crease. Although it remains unclear whether multiple-hitting apply to lethal mutations. (iii) In both types of organism,
could generate observable amounts of antagonistic epistasis in antagonism was more frequent than synergism. This observation
prokaryotes, a recent in silico study of the global transcriptional may reflect that even the most complex digital organisms are far
regulatory network of E. coli suggests that genes can indeed be less complex than most biological organisms. (iv) In agreement
grouped in very few functional modules (14). with our results, the frequency of synergistic pairs was larger in
In the contrary, if each of k loci is sufficient for the expression complex organisms.
of the character, epistasis should be synergistic, because the Several complex features of higher eukaryotes, such as large
fitness of the kth mutant would be disproportionately low genomes, increased numbers of replication rounds per genera-
compared with that of each single mutant. Therefore, synergistic tion, or small population sizes, may tend to inflate the mutational
epistasis could theoretically appear as a consequence of gene load, conferring a potential selective advantage to robustness
duplication and redundancy. However, this mechanism would (24, 25). Therefore, mutational robustness might be an adaptive
only be efficient in a system with very few genes. In large trait of complex organisms. In turn, buffering mechanisms could
genomes, although duplications could have a role in lessening provide the raw material to create new functions and increase
mutational damage (15), they should not generate significant genetic complexity (26). This kind of positive feedback, by which
amounts of synergism because the probability of simultaneously evolution could forge its own path, has been recently proposed
hitting two copies becomes vanishingly low as genome size to explain the evolution of sexual reproduction in regard to
increases. A better clue to how synergistic epistasis can be mutational robustness and synergistic epistasis (27).
generated in complex genomes comes from the observation that
many genes in higher eukaryotes code for sensors and regulators Data and Methods
that confer stability rather than functionality in optimal circum- VSV Data Set. Forty-seven combinations of single-nucleotide
stances. The ability of complex systems to adjust network random, nonlethal mutations were introduced by site-directed
organization and accommodate perturbation generates robust- mutagenesis in an infectious cDNA template (8). For informa-
ness to mutational or environmental perturbations (16, 17). As tion about the precise location of each mutation and raw fitness
Downloaded from https://www.pnas.org by Universidad Autonoma de Nuevo Leon on August 19, 2024 from IP address 148.234.81.252.

long as a robust genome accumulates a small number of muta- values, see supporting table 1 of ref. 8. It is easy to prove that the
tions, deleterious effects can be buffered by shuttling the met- fitness of each mutant genotype relative to a wild-type, defined
abolic flux through the unaffected parts of the network, whereas, as the size of the progeny per individual per wild-type genera-
no matter how robust the system is, for a sufficiently large tion, can be obtained as:
number of perturbations, buffering mechanisms would become
insufficient to prevent the system from collapsing. Therefore, Kri/r0 ⫺ 1
Wi ⫽ , [2]
synergistic epistasis can be viewed as an emerging property of K⫺1
complex and robust networks, with high levels of redundancy
and frequent alternative pathways (17). where K is the average per cell progeny for the wild-type and ri and
In general, the link between epistasis and complexity can be r0 stand for the mutant and wild-type growth rates, respectively.
better understood by acknowledging that epistasis and muta-
tional robustness are probably correlated. Strong mutational E. coli Data Set. Twenty-seven random insertions were assayed for
effects (low robustness) should be associated to more antago- fitness alone and in pairs (9). Southern blots confirmed that
nistic epistasis, whereas mild mutational (robustness) effects mutations were dispersed throughout the genome and that there
should be associated to synergism. This correlation has already were few, if any, genotypes with identical changes. Raw fitness
been observed in digital organisms (18), in in silico models of was calculated as the ratio between the growth rate of the mutant
bacteriophage T7 infectious cycle (19), and in simulated RNA and reference strains by S.F.E. and R. E. Lenski (Michigan State
folding (18, 20, 21). In good agreement, it seems that fitness University, East Lansing, MI). All genotypes were viable. The
effects associated with point mutations are larger in RNA fitness of each genotype was transformed into a per-generation
genomes than in DNA organisms (22). Table 1 shows estimations scale by setting K ⫽ 2 in Eq. 2.
of the selection coefficient against deleterious mutations for the
five species we analyzed in detail. Although these estimates are S. cerevisiae Data Set. Single-site random mutants were generated
difficult to obtain and sometimes controversial, the data seem by using ethyl methane-sulfonate (EMS) (10). EMS produces a
consistent with an increase in mutational robustness with in- majority of transitions, the remainder being transversions and,
creasing genomic complexity (␳S ⫽ ⫺1; 3 df; P ⬍ 0.001). In sporadically, small deletions (28). The EMS dose was low enough
conclusion, we propose that two hallmarks of simple genomes to ensure that no more than one mutation was incorporated per
would be that single mutations may have large fitness effects and individual. Haploids derived from two different heterozygous
epistasis among mutations shall be predominantly antagonistic. clones were mated to recover the wild-type, each single het-
Conversely, two characteristic properties of complex genomes erozygous mutant, and the double heterozygous mutant. Forty-
would be that single mutations may have mild effects on fitness nine combinations were tested. Raw fitness values were calcu-
and epistasis among mutations should be predominantly syner- lated as the ratio of colony growth rates and were provided by
gistic. R. Korona (Jagiellonian University, Krakow, Poland). Fitness
The relationship between epistasis, robustness, and complex- can be rescaled into generation units in the same way as for the
ity was previously studied by using digital organisms (23). The E. coli data set.
results of this study can be summarized as follows: (i) simple
digital organisms were much more sensitive to mutation than A. niger Data Set. Seven phenotypic markers, each located in a
their complex counterparts, in accordance with our above ar- different chromosome of the filamentous mold A. niger, were
gumentation. This was mainly due to the larger fraction of lethal combined by isolating segregants from heterokaryons (11). The
mutations in simpler digital organisms. (ii) The average intensity markers were five auxotrophic and two antibiotic resistance
of epistasis was weaker in simpler digital organisms. This finding mutations. Although ⬇2,500 haploids were screened, only 186 of
could be explained by noting that, except for compensatory 256 possible genotypes were recovered, despite the fact that,
cases, the presence of a lethal mutation in a locus determines the because each genotype had the same probability of appearance,
fitness of the organism independently of what alleles are at other ⬇10 individuals of each genotype were expected. This finding,

14404 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0604543103 Sanjuán and Elena

 together with the observation that the fraction of genotypes terious fitness were considered. Two alternative definitions of
successfully isolated decreased with mutation number, indicates the epistasis coefficient were also used: a relative epistasis
that absent genotypes were probably lethal or highly unviable. coefficient (see Supporting Text) and the scaled epistasis coef-
Raw fitness values were estimated as the increase in mycelial ficient (29). All three definitions led to identical conclusions.
surface and provided by J. A. G. M. de Visser (Wageningen
University and Research, Wageningen, The Netherlands). Fit- Selection Coefficients per Generation. Average selection coeffi-
ness per generation can be obtained as Wi ⫽ (⌬Ai(g)兾⌬Awt(g))1/g, cients (s៮) obtained from the above data sets plus mutation
where ⌬Ai and ⌬Awt are the increase in mycelial surfaces for the accumulation experiments were previously converted into a
mutant and the wild type, respectively, and g is the number of per-generation scale (22). In diploids, mutational effects against
generations elapsed during the fitness assay. The number of single mutations are measured as hs, where h is the dominance
spores at the initial and final time points were ⬇1.2 ⫻ 107 and coefficient. For VSV, E. coli, S. cerevisiae, and Aspergillus
⬇2 ⫻ 108, respectively, and therefore, g ⬇ log2(2 ⫻ 108兾1.2 ⫻ nidulans (a close relative to A. niger), values in Table 1 are the
107) ⫽ 4. The whole experiment was replicated in two different average of the values reported in the supporting tables of ref. 22.
genetic backgrounds. Fitness values were averaged across the For D. melanogaster, despite intensive work, estimations of the
two genetic backgrounds. selection coefficient remain controversial. Some authors argue
that the homozygous coefficient is in the order of s ⫽ 0.2 and the
D. melanogaster Data Set. Alleles at five different loci were put dominance coefficient in the order of h ⫽ 0.1 or lower (30),
together in all 26 different combinations by means of the whereas other authors argue that the homozygous coefficient is
appropriate crosses (12). All mutations were in the homozygous one order of magnitude lower (s ⫽ 0.02) and the dominance
state. These mutations were black (b) and plexus-speck (px兾sp) coefficient approximately h ⫽ 0.5 (31). Therefore, some con-
on the second chromosome and claret (ca), hairy (h), and sensus exists on the value of the selection coefficient against
ebony-stripe (e兾sr) on the third chromosome. Two measures of heterozygotes, hs ⬇ 0.01.
fitness relative to a reference strain were calculated: (i) produc-
tivity, which counts viable offspring, and (ii) male mating Statistical Analyses. Analyses were done with SPSS version 12.0
success, which counts the number of matings per male. Raw software. To avoid statistical problems stemming from the fact
Downloaded from https://www.pnas.org by Universidad Autonoma de Nuevo Leon on August 19, 2024 from IP address 148.234.81.252.

fitness measures are available in the supporting table of ref. 8. that some mutations were present in more than one genotype, P
Given that both fitness components refer to two nonoverlapping values were calculated by the bootstrap method. This latter
and necessary steps of the life cycle of a population, a total fitness analysis was done by using a Perl script.
can be obtained by multiplying them, which measures the fitness
that a population would have if all males and females carried the We thank J. A. G. M. de Visser and R. Korona and coworkers for sharing
their data. This work was supported by a grant from the Generalitat
mutation. Valenciana (to R.S.) and by grants from the Ministerio de Educación y
Ciencia Fondo Europeo de Desarrollo Regional, the Generalitat Va-
Calculation of Epistasis Coefficients. Per-generation fitness values lenciana, and the European Molecular Biology Organization Young
for single and multiple mutants were used to calculate the Investigator Program (to S.F.E.). R.S. is supported by a Consejo Superior
epistasis coefficient (Eq. 1). Only mutants with expected dele- de Investigaciones Cientificas I3P postdoctoral contract.

1. Wolf JB, Brodie ED, III, Wade MJ (2000) Epistasis and the Evolutionary Process 26. Wagner A (2005) Robustness and Evolvability in Living Systems (Princeton Univ
(Oxford Univ Press, Oxford). Press, Princeton).
2. Kondrashov AS (1988) Nature 336:435–440. 27. Azevedo RB, Lohaus R, Srinivasan S, Dang KK, Burch CL (2006) Nature
3. Kondrashov AS, Crow JF (1991) Nature 351:314–315. 440:87–90.
4. Omholt SW, Plahte E, Oyehaug L, Xiang K (2000) Genetics 155:969–980. 28. Davies EK, Peters AD, Keightley PD (1999) Science 285:1748–1751.
5. Butcher D (1995) Genetics 141:431–437. 29. Segrè D, DeLuna A, Church GM, Kishony R (2005) Nat Genet 37:77–83.
6. Coyne JA (1992) Nature 355:511–515. 30. Garcı́a-Dorado A, López-Fanjul C, Caballero A (1999) Genet Res 74:341–350.
7. Otto SP, Feldman MW (1997) Theor Pop Biol 51:134–147. 31. Lynch M, Blanchard JL, Houle D, Kibota TT, Schultz SC, Vassilieva LL, Willis
8. Sanjuán R, Moya A, Elena SF (2004) Proc Natl Acad Sci USA 101:15376–15379. J (1999) Evolution (Lawrence, Kans) 53:645–663.
9. Elena SF, Lenski RE (1997) Nature 390:395–398. 32. Burch CL, Chao L (2004) Genetics 167:559–567.
10. Szafraniec K, Wloch DM, Sliwa P, Borts RH, Korona R (2003) Genet Res 33. Elena SF (1999) J Mol Evol 49:703–707.
82:19–31. 34. Crotty S, Cameron CE, Andino R (2001) Proc Natl Acad Sci USA 98:6895–
11. de Visser JAGM, Hoekstra RF, van den Ende H (1997) Evolution (Lawrence,
6900.
Kans) 51:1499–1505.
35. Bonhoeffer S, Chappey C, Parkin NT, Whitcomb JM, Petropoulos CJ (2004)
12. Whitlock MC, Bourguet D (2000) Evolution (Lawrence, Kans) 54:1654–1660.
Science 306:1547–1550.
13. Edlund JA, Adami C (2004) Artif Life 10:167–179.
36. Maisnier-Patin S, Roth JR, Fredriksson A, Nystrom T, Berg OG, Andersson
14. Barrett CL, Herring CD, Reed JL, Palsson BO (2005) Proc Natl Acad Sci USA
DI (2005) Nat Genet 37:1376–1379.
102:19103–19108.
15. Gu Z, Steinmetz LM, Gu X, Scharfe C, Davis RW, Li WH (2003) Nature 37. de Visser JAGM, Hoekstra RF, van den Ende H (1996) Proc R Soc London B
421:63–66. 263:193–200.
16. Carlson JM, Doyle J (2002) Proc Natl Acad Sci USA 99(Suppl 1):2538–2545. 38. de Visser JA, Hoekstra RF, van den Ende H (1997) Genetics 145:815–819.
17. Wagner A (2005) BioEssays 27:176–188. 39. Kelly JK (2005) Genetics 171:1917–1931.
18. Wilke CO, Adami C (2001) Proc R Soc London B 298:1469–1474. 40. Peters AD, Keightley PD (2000) Genetics 156:1635–1647.
19. You L, Yin J (2002) Genetics 160:1273–1281. 41. Kitagawa O (1967) Genetics 57:809–820.
20. Wilke CO, Lenski RE, Adami C (2003) BMC Evol Biol 3:1–14. 42. Mukai T (1969) Genetics 61:749–761.
21. Sanjuán R, Forment J, Elena SF (2006) Mol Biol Evol, in press. 43. Seager RD, Ayala FJ, Marks RW (1982) Genetics 102:485–502.
22. Elena SF, Carrasco P, Daròs JA, Sanjuán R (2006) EMBO Rep 7:168–173. 44. Rivero A, Balloux F, West SA (2003) Evolution (Lawrence, Kans) 57:1698–
23. Lenski RE, Ofria C, Collier TC, Adami C (1999) Nature 400:661–664. 1703.
24. Krakauer DC, Plotkin JB (2002) Proc Natl Acad Sci USA 99:1405–1409. 45. Burch CL, Turner PE, Hanley KA (2003) J Evol Biol 16:1223–1235.
25. Proulx SR (2005) Am Nat 165:147–162. 46. de Visser JAGM, Hoekstra RF (1998) Genet Res 71:39–49.
EVOLUTION

Sanjuán and Elena PNAS 兩 September 26, 2006 兩 vol. 103 兩 no. 39 兩 14405