| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Genetics, Vol. 176, 501-511, May 2007, Copyright © 2007
doi:10.1534/genetics.106.063636
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Population and Quantitative Genomics Team, Genomic Sciences Center, RIKEN, Yokohama 230-0045, Japan
1 Address for correspondence: Population and Quantitative Genomics Team, Genomic Sciences Center, RIKEN, Yokohama 230-0045, Japan.
E-mail: kenzi{at}gsc.riken.jp
 |
ABSTRACT |
|---|
 TOP ABSTRACT ANALYSISDISCUSSIONAPPENDIX A: DIFFUSION MODELAPPENDIX B: BIRTH-AND-DEATH...ACKNOWLEDGEMENTSLITERATURE CITED |
|---|
Another aspect of evolution emphasized in the Wrightian theory is the genetic structure of geographically subdivided populations and its consequences on adaptive evolution (WRIGHT 1931, 1932). The Fisherian view holds that evolution should proceed in a large, panmictic population such that the effects of genetic subdivision may largely be ignored. In contrast, Wright envisaged that the individuals of a species are distributed into small, partially isolated breeding units, where the effects of random genetic drift and interdemic selection predominate. Evolutionary consequences of population subdivision have also been studied in relation to the fixation probability of a mutant allele under genic selection (MARUYAMA 1970; CHERRY and WAKELEY 2003). It has been shown that in the simplest case with conservative migration (i.e., migration that has no directional effects in changing the overall allele frequencies), fixation probabilities are not affected by subdivision (MARUYAMA 1970; CHERRY and WAKELEY 2003). Later, more complex (and presumably realistic) models of subdivision that incorporate local extinction and subsequent colonization were developed (BARTON 1993; CHERRY 2003b, 2004; ROZE and ROUSSET 2003; WHITLOCK 2003). The effect of dominance has also been included, leading to an alteration in the fixation probability under subdivision, even in the absence of local extinction and recolonization (SLATKIN 1981; CHERRY 2003a, 2004; ROZE and ROUSSET 2003; WHITLOCK 2003; NISHINO and TAJIMA 2004).
While these and other studies have seen only a marginal role for population subdivision in driving mutant alleles to fixation in single-locus systems, relatively little has been done formally to elucidate the joint effects of population subdivision and epistasis (WHITLOCK et al. 1993; WADE and GOODNIGHT 1998; GOODNIGHT 2000; WADE 2002). To investigate the possible role of population subdivision in the evolution of functionally integrated systems, we here extend our preceding formulation for a panmictic population (TAKAHASI and TAJIMA 2005) and develop two-locus models for the evolution of a coadapted system in a subdivided population. By focusing on the fixation process of two conditionally neutral mutations that interact epistatically to form a coadapted haplotype, we have previously demonstrated that when both dominance and linkage disequilibrium are absent in a panmictic population, the fixation probability of the coadapted haplotype is simply given by the product of two probabilities: the fixation probability of a selectively neutral mutant, which is 1/(2N) in a population of N diploids, multiplied by the fixation probability of a beneficial allele with a selective advantage s, which is approximately 2s (HALDANE 1927). The result holds irrespective of the time T between the two consecutive mutational events, implying that under random mating the epistatic effects of cosegregating variants during evolution do not induce any additional chance of double fixation.
Since we can expect that most natural populations should show some level of subdivision, the following analysis investigates the effect of limited migration by incorporating the finite-island model of population subdivision. The effects of local extinction and recolonization are not considered. Dominant allelic effects in fitness are not included either. Even with these simplifying assumptions, it is demonstrated by stochastic simulations that the fixation probability of a coadapted haplotype critically depends on migration rate, suggesting a role for population subdivision in promoting the evolution of epistatic systems. Analytical theories based on diffusion and birth-and-death models are also developed (in the APPENDIXES), which should approximate high and low migration limits, respectively.
|
ANALYSIS |
|---|
|
TOPABSTRACTANALYSISDISCUSSIONAPPENDIX A: DIFFUSION MODELAPPENDIX B: BIRTH-AND-DEATH...ACKNOWLEDGEMENTSLITERATURE CITED |
|---|
Diffusion approximation:
While the substantial part of this analysis is based on stochastic simulations, I also provide analytical theories based on diffusion and birth-and-death approximations for the limiting cases with high and low migration rates, respectively. The details of the analytical treatments are relegated to the APPENDIXES. I first study the probability of double fixation in a subdivided population using diffusion approximation. Recent applications of the diffusion theory to the related issues for the single-locus situations include CHERRY (2003a,b, 2004), CHERRY and WAKELEY (2003), ROZE and ROUSSET (2003), WHITLOCK (2003), and NISHINO and TAJIMA (2004, 2005). GAVRILETS and GIBSON (2002) and WHITLOCK and GOMULKIEWICZ (2005) have developed diffusion-based analytical methods to obtain fixation probabilities under spatially heterogeneous selection.
For a panmictic population of N diploids, we have previously shown that when the two loci are in linkage equilibrium and there is no dominance at each locus, the probability for the ultimate fixation of the coadapted haplotype A1B1 is given by
 | (1) |
 | (2) |
 |
 | (3a) |
 | (3b) |
 | (4) |
Mathematically, the condition (3) will always be satisfied if the allele frequencies at the two loci are statistically independent of each other. This situation is more likely to be met when the entire population is genetically homogenized by frequent migration. Later in this section we specify by simulations the range of migration rates where the diffusion approximation may be applicable.
For sufficiently strong selection (4NTs >> 1) on new mutations {yT = 1/(2NT) and E[xT] = 1/(2NT), respectively, assuming that the allele A1 has appeared in the population before the introduction of the mutant B1; see TAKAHASI and TAJIMA 2005}, the solution (2) reduces to E[u] = 2s/(2NT); here, the expectation is taken over the distribution of xT at the second mutational event.
Birth-and-death approximation:
For the opposite circumstances with extremely low migration, I develop a model based on the birth-and-death approximation (LANDE 1979; SLATKIN 1981; TAKAHATA 1991; NISHINO and TAJIMA 2004, 2005). In this limit, the migration process is so slow compared to the frequency changes within subpopulations that the fixation or loss of a new allele occurs independently within each subpopulation. As detailed in APPENDIX B, this implies that at any time during evolution each subpopulation may be considered monomorphic for one of the four possible allelic combinations, and the state of the entire population is uniquely determined by the numbers of subpopulations that are fixed for each of the four allelic combinations.
Consider two new mutations A1 and B1, each introduced randomly into a subdivided population. When the entire population is subdivided into L subpopulations of an equal size N, the two new alleles may initially arise in a single subpopulation with probability 1/L. Otherwise they arise in two distinct subpopulations (with probability 1 – 1/L). When the two mutations co-occur in a single subpopulation, the conditional probability of double fixation within the subpopulation is given approximately by 2s/(2N), assuming sufficiently strong selection (4Ns >> 1) and linkage equilibrium during the fixation process (TAKAHASI and TAJIMA 2005). The assumption of strong selection further entails that once an adaptive gene combination is established in a subpopulation, it is destined to proliferate throughout the entire population. This is because the stochastic loss of the favorable combination from the subpopulation should never happen under this assumption, in spite of the recurrent introduction of ancestral (and deleterious) alleles through migration. Therefore, when the two mutations are initially introduced into a single subpopulation, the contribution to the overall probability of double fixation becomes 2s/(2NL).
When the two mutations arise in distinct subpopulations, each of them must first get fixed independently in its resident subpopulation before they can be exported into other subpopulations through migration. This probability is simply given by {1/(2N)}2, which is the square of the fixation probability of a selectively neutral mutation in a population of N diploids. Summing up the two mutually exclusive possibilities, the expected fixation probability in the low migration limit is expressed as
 | (5) |
By numerically solving the system of linear equations that describes the transitions between possible states of the population (Equation B1), we obtain the solution for the conditional probability fL, and further by substituting this into (5), we arrive at the numerical solution for the fixation probability in the low migration limit.
Simulations:
For moderate migration, I conduct a set of simulations that follow the forward frequency dynamics of the two-locus system with recombination (at rate c). Initially, the entire population is monomorphic for the ancestral alleles. The two mutations are consecutively introduced, each as a single copy into a randomly chosen subpopulation. The second mutation appears T generations after the introduction of the first allele. No recurrent mutations are assumed afterward. While the waiting time T for the second mutation may in fact be determined by the mutability µ of the B0 allele and the population size NT {such that E[T] = 1/(2NTµ), assuming an exponential distribution of the second mutational event}, we here adopt a simplifying assumption that T is a fixed constant in each set of simulations. The outcome does not differ qualitatively if we assume that the time difference is a random variable. Random genetic drift is implemented in the simulation by the improved version of the pseudosampling (KIMURA and TAKAHATA 1983), whereas the expected frequency changes are obtained from the standard two-locus model with selection and recombination (EWENS 2004, Equation 2.94), with additional changes due to migration. Note that in the simulations two-locus genotype frequencies are followed explicitly in each subpopulation, without imposing linkage equilibria as assumed in the diffusion analysis. There is no dominance in fitness at each locus; more specifically, epistatic selection with cis-acting mutations is assumed (see TAKAHASI and TAJIMA 2005, Table 2).
Each run of simulations lasts until either of the mutant alleles is lost from the entire population or when both become fixed so that a coadaptation is established. The simulations are repeated until the fixation of the haplotype A1B1 is replicated for 1000 times. The total number of simulation runs, required to produce 1000 fixation events, is recorded to estimate the fixation probability.
Interaction between epistatic selection and migration:
To see how limited gene flow interacts with epistatic selection acting independently within each subpopulation, we first focus on two new mutations that arise simultaneously (T = 0 in the present notation) and investigate how subsequent migration should facilitate or inhibit the formation of coadaptation. In so doing, we consider two situations separately, where the two mutations together arise in a single subpopulation and where they arise separately in two distinct subpopulations, respectively. The probability for the ultimate fixation of the coadapted haplotype would be much different depending on the initial distribution of the new mutations.
When the two mutations initially co-occur in a single subpopulation, Figure 1 shows that with restricted migration the fixation probability may be increased by a factor of L (= 10 in Figure 1), roughly ranging from 2s/(2NL) (= 10–6) for m 
0.005 to 2s/(2N) (= 10–5) for m 
0.0002. Such dependence on migration rate is expected because limited gene flow will keep the two mutations clumped together in the subpopulation, and the evolutionary increase of the coadapted haplotype will be effectively promoted by epistatic selection, which favors the combination of two alleles that initially cosegregate at low frequencies in the subpopulation. Figure 2A shows that when m 0.0002 the correlation between allele frequencies at the two loci, computed across subpopulations (as in Equation 4), is kept high for a relatively long period after the introduction, suggesting that the two mutations share a greater chance to cosegregate within a same subpopulation during the initial stages of evolution. As the migration rate becomes larger, the correlation across subpopulations decays at a faster rate. This implies that too much migration would rapidly reallocate the two mutations into distinct localities, thereby limiting the role of epistatic selection in promoting the joint increase of mutant alleles within subpopulations. The conditional probability of double fixation should therefore increase monotonically with decreasing rates of migration.
|
|
0.0002, the correlation between allele frequencies at the two loci is established only after a substantial time has elapsed since the introduction of the two mutations. This result indicates that the two mutations should be less likely to cosegregate in the same subpopulation initially when they are still at low frequencies. The potential role of epistatic selection in driving the evolution of coadaptation would therefore be much constrained. In contrast, when the migration rate is slightly higher (m = 0.0005 or 0.001), a significant correlation is established rather rapidly (Figure 2B), and consequently the scaled conditional probability exceeds unity (Figure 1). This implies that in certain cases limited gene flow may promote the evolution of coadaptation even when the interacting alleles initially arise in distinct localities. (It should be emphasized here that in Figure 2 correlations were computed over a subset of simulation runs that eventually resulted in the joint fixation of the two new mutations.) Taken together, the unconditional probability for the ultimate fixation of the coadapted haplotype should be most facilitated by an intermediate rate of migration, roughly when m = 0.0005 with the parameter values studied here (Figure 1). Simulations also show that when m < 0.01 correlations between x and y(1 – y) [or equivalently between x(1 – x) and y, since we here assume T = 0 so that the two loci are indistinguishable] deviate significantly from zero (result not presented), suggesting that the condition (3) required for the diffusion approximation is satisfied only when the entire population is genetically homogenized by frequent migration, as expected reasonably.
Time difference between two mutational events:
As we have just seen for two mutations arising simultaneously (T = 0), simulations show that when the time interval between the two mutational events is relatively short (T = 1 or 100) there is an intermediate rate of migration (hereafter denoted mmax) that maximizes the fixation probability of the coadapted haplotype (Figure 3). This dependence of fixation probability on migration rate becomes less evident when the second mutation appears long after the first mutational event (T = 10,000). With a sufficiently long time interval T, the first mutation is more likely to be fixed or lost by the time of the second mutational event, and in the limiting case of T 
the probability for the ultimate fixation of the coadapted haplotype eventually converges to
 | (6) |
|
2s/(2NL), provides accurate estimates only for relatively high migration rates (roughly when m 0.01 or equivalently Nm 10, with the parameter values studied here). Hence, as expected, the diffusion model (A3) in APPENDIX A should be valid when the entire population is effectively panmictic. On the opposite extreme, in the low migration limit, the simulations show that the fixation probability is more or less increased by an amount corresponding to the second term in the right-hand side of (5).
Subpopulation number and size:
When the second mutation is introduced shortly after the first mutational event, the simulations show that the effect of population subdivision in promoting the fixation of the coadapted haplotype becomes more profound as the number of subpopulations L becomes larger and the entire population gets more fragmented. This may be observed either when the total population size NT is kept constant (so that the subpopulation size is altered along with the subpopulation number; Figure 4) or when the subpopulation size N is constant (so that the total population size is altered along with the subpopulation number; Table 1). In the former case, the subpopulation size is smaller in a more fragmented population (with larger L), implying that the relative strength of random genetic drift becomes more prevailing. In the latter case, as the subpopulation number L gets larger, the total population size also gets larger and accordingly the (absolute) fixation probability becomes smaller (Table 1). Still, the effect of subdivision is more pronounced for larger L; by contrasting the scaled fixation probabilities in units of 2s/(2NL), we can see that the fixation probability shows an almost threefold increase when L = 50, but the increment is only 1.7-fold when L = 10 (Table 1).
|
|
1 x 10–6), it appears that the fixation probabilities converge to (or become somewhat smaller than) the estimates obtained from the birth-and-death model. Additional simulations show that mmax is primarily affected by the selection intensity s and not by other factors such as the subpopulation size N (see Figures 4 and 5). Previously, we have demonstrated that there is an optimal rate of recombination, given approximately by c = s/2, that most facilitates the evolution of coadaptation in a panmictic population (TAKAHASI and TAJIMA 2005). While simulations suggest that mmax also changes monotonically with s, it appears that the relationship is rather complex.
|
|
4 (equality holds when L = 100 in Figure 4 or when s = 0.001 in Figure 5). When this condition is violated, simulations show that in contrast to the situations under strong selection the effect of population subdivision may be less pronounced in a more fragmented population. As shown in Figure 7, when the total population size is kept constant at NT = 2000 and s = 0.001 (hence 4NTs = 8), the probability of double fixation is increased only slightly in a population fragmented into L = 20 subpopulations (hence 4Ns = 0.4), whereas in a less fragmented population with L = 5 (4Ns = 1.6) or 10 (4Ns = 0.8), the effect of subdivision is more evident; still, the increment is only up to a factor of 1.2 at most (when L = 5). These simulation results also suggest that the primary role of restricted gene flow is to colocalize simultaneously segregating mutations that interact epistatically, thereby creating the condition favorable for the effective operation of epistatic selection within subpopulations. When 4Ns < 1 and the advantageous haplotype is only weakly selected in small subpopulations, random genetic drift may override the effect of epistatic selection in driving the evolution of coadaptation within each subpopulation, even if selection could be effective at the entire population level (4NTs > 1).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTANALYSISDISCUSSIONAPPENDIX A: DIFFUSION MODELAPPENDIX B: BIRTH-AND-DEATH...ACKNOWLEDGEMENTSLITERATURE CITED |
|---|
When the habitat range of an organism is fragmented into local subpopulations, the two mutations may initially arise in distinct localities. If migration among subpopulations is severely constrained (as in the low migration limit), it is unlikely that the two mutations should ever have a chance to see each other in a single locality before getting fixed (or lost) within a subpopulation. The fixation of a mutant allele within each subpopulation should then be driven completely by random genetic drift, and there is not much room left for epistatic selection to promote the joint increase of the two mutations. When the two mutations are initially introduced into a single subpopulation, limited migration will keep the two mutations together in the subpopulation, and the evolutionary increase of the coadapted haplotype will be effectively promoted by epistatic selection. However, too much migration would rapidly disperse the two mutations away from each other, thereby reducing the effect of epistatic selection in the formation of a favorable allelic combination within each subpopulation. Moderate migration and isolation allow the propagation of the mutant alleles across subpopulations, while at the same time preserving the favorable allelic combination established within each subpopulation. Together they create the condition most favorable for the ultimate fixation of the coadapted haplotype. Moreover, the facilitation of double fixation in a subdivided population should be expected only when the time difference between the two mutational events is sufficiently short (see Figure 3), suggesting the vital role played by epistatic selection during the initial stages of evolution, when both mutations are still segregating at low frequencies.
Our findings may seem to render support for the shifting-balance theory of evolution (WRIGHT 1931, 1932; reviewed in WRIGHT 1980), which advocates the primacy of epistasis and random genetic drift as the causes of evolution in small, partially isolated populations. Although the present study also focuses on the interplay between population subdivision and epistasis, there is an important difference between the Wrightian theory and the present theories. According to the conventional views, the Wrightian shifting-balance process proceeds through the rugged fitness landscape with peaks and valleys generated by epistatic interactions among alleles at multiple loci (WHITLOCK et al. 1995). The fitness valleys represent the class of unfavorable genotypes that may be formed during evolution. In contrast, the present formulation does not include unfit intermediate haplotypes; instead, it involves haplotypes connected by a neutral network, where evolution may proceed without crossing the intervening fitness valleys (GAVRILETS 2004; WAGNER 2005).
The role of conditionally neutral mutations in adaptation has recently renewed interest in the theories of evolution in multilocus systems, where, for example, biological systems achieve functional robustness by buffering genetic perturbations (e.g., VAN NIMWEGEN et al. 1999; RUTHERFORD 2000; WILKE 2001; GIBSON and DWORKIN 2004; HERMISSON and WAGNER 2004; WAGNER 2005). Classically, the adaptive significance of neutral variation has been noted in the neutral theory of evolution, when KIMURA (1983) introduced the Dykhuizen–Hartl effect (DYKHUIZEN and HARTL 1980) to describe the potential role of neutral mutations in creating the foundation of adaptation (see also HARTL et al. 1985; KIMURA 1986; ZHANG et al. 1998).
The stochastic dynamics of the shifting-balance process have also been studied by two-locus epistatic selection models, where mutations are individually deleterious but become neutral (or may even be better fit) when combined with the compensatory allele at the other locus (e.g., KIMURA 1990, 1991; PHILLIPS 1996; but see also STEPHAN 1996). In a subdivided population, mutant alleles would individually be kept at low frequencies in each subpopulation if sufficiently strong selection acts against the intermediate haplotypes. Consequently, it is highly unlikely that the two compensatory mutations, each formerly in a distinct locality, should be brought together into a single subpopulation through migration (PHILLIPS 1996). Instead, each subpopulation has to wait for the production of a new mutation that compensates the deleterious effect of the preexisting allele, without the aid of migration. Therefore, the evolutionary trajectories follow very different paths under selection models with or without the detrimental fitness effects on the intermediates, and the dependence of fixation probability on migration rate as demonstrated in the present study would not be expected when multiple fitness peaks are separated by valleys of low fitness. The conditional neutrality thus plays a decisive part in the present theory of evolution in a subdivided population.
It has been known that locally restricted interactions affect the global foundation of interactive systems in various ways (e.g., ESHEL and CAVALLI-SFORZA 1982; WADE and GOODNIGHT 1998; THOMPSON 1999; CHAVE et al. 2002; BRIGGS and HOOPES 2004). The present study also finds a primary role for subpopulation structure in creating the favorable condition for the evolution of coadaptation. As the genomewide polymorphism data accumulate, it is now possible to draw a detailed inference on the genetic structure and history of natural populations (HEY and MACHADO 2003; WAKELEY 2004). Together with our current knowledge on the genetic architecture of phenotypic diversity, insights into the historical demography will cast a new light on the role of population subdivision and migration in determining the course of evolution in multilocus systems.
|
APPENDIX A: DIFFUSION MODEL |
|---|
|
TOPABSTRACTANALYSISDISCUSSIONAPPENDIX A: DIFFUSION MODELAPPENDIX B: BIRTH-AND-DEATH...ACKNOWLEDGEMENTSLITERATURE CITED |
|---|
In the diffusion limit, the expected frequency changes within each subpopulation are given respectively by
 | (A1a) |
 | (A1b) |
 | (A2) |
Using these notations, fixation probability u(x1, ... , xL, y1, ... , yL) satisfies the equation
 | (A3) |
Here we want to know under what conditions the partial differential equation (A3) has a solution (2), which is mathematically equivalent to the panmictic solution (1). Given that the probability u is defined by (2), we have
 | (A4a) |
 | (A4b) |
 | (A4c) |
 | (A4d) |
Substituting (A1), (A2), and (A4) into the right-hand side of (A3), we arrive at the desired condition,
 | (A5) |
|
APPENDIX B: BIRTH-AND-DEATH MODEL |
|---|
|
TOPABSTRACTANALYSISDISCUSSIONAPPENDIX A: DIFFUSION MODELAPPENDIX B: BIRTH-AND-DEATH...ACKNOWLEDGEMENTSLITERATURE CITED |
|---|
 | (B1) |
0, L. Note also that fL = f(1, 1, 0). The conditional probability fL can be obtained by numerically solving a sparse system of (L – 1) (L – 2)/2 linear equations for (L – 1) (L – 2)/2 unknowns.
On deriving the relation (B1), it is assumed that if a subpopulation with (x, y) = (0, 1) receives immigrants with haplotype A1B0 from another subpopulation with (x, y) = (1, 0), the possible outcome is either (x', y') = (1, 1) (fixation of the incoming allele A1, with probability u+) or (x', y') = (0, 1) (no change, with probability 1 – u+), where the primes denote the frequencies after the migration event. An equivalent argument applies when a subpopulation with (x, y) = (1, 0) receives immigrants with haplotype A0B1. This simplifying assumption, which further entails the boundary condition f(n01, L – n01, 0) = 1 (for n01 0, L), neglects the possibilities of (x', y') = (0, 0) (loss of the preexisting allele B1) and (x', y') = (1, 0) (fixation of A1 accompanied by loss of B1) and should be valid only when the selection is sufficiently strong (4Ns >> 1). Therefore, when the entire population is fragmented into many small subpopulations such that the required condition 4Ns >> 1 does not hold, the present formulation based on (B1) should overestimate the probability of double fixation.
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTANALYSISDISCUSSIONAPPENDIX A: DIFFUSION MODELAPPENDIX B: BIRTH-AND-DEATH...ACKNOWLEDGEMENTSLITERATURE CITED |
|---|
|
LITERATURE CITED |
|---|
|
TOPABSTRACTANALYSISDISCUSSIONAPPENDIX A: DIFFUSION MODELAPPENDIX B: BIRTH-AND-DEATH...ACKNOWLEDGEMENTSLITERATURE CITED |
|---|
BARTON, N. H., 1993 The probability of fixation of a favoured allele in a subdivided population. Genet. Res. 62: 149–157.
BRIGGS, C. J., and M. F. HOOPES, 2004 Stabilizing effects in spatial parasitoid-host and predator-prey models: a review. Theor. Popul. Biol. 65: 299–315.[CrossRef][Medline]
BURCH, C. L., and L. CHAO, 1999 Evolution by small steps and rugged landscapes in the RNA virus 
6. Genetics 151: 921–927.
CHAVE, J., H. C. MULLER-LANDAU and S. A. LEVIN, 2002 Comparing classical community models: theoretical consequences for patterns of diversity. Am. Nat. 159: 1–23.[CrossRef]
CHERRY, J. L., 2003a Selection in a subdivided population with dominance or local frequency dependence. Genetics 163: 1511–1518.
CHERRY, J. L., 2003b Selection in a subdivided population with local extinction and recolonization. Genetics 164: 789–795.
CHERRY, J. L., 2004 Selection, subdivision and extinction and recolonization. Genetics 166: 1105–1114.
CHERRY, J. L., and J. WAKELEY, 2003 A diffusion approximation for selection and drift in a subdivided population. Genetics 163: 421–428.
COYNE, J. A., N. H. BARTON and M. TURELLI, 1997 Perspective: a critique of Sewall Wright's shifting balance theory of evolution. Evolution 51: 645–671.
COYNE, J. A., N. H. BARTON and M. TURELLI, 2000 Is Wright's shifting balance process important in evolution? Evolution 54: 306–317.[CrossRef][Medline]
DYKHUIZEN, D., and D. L. HARTL, 1980 Selective neutrality of 6PGD allozymes in E. coli and the effects of genetic background. Genetics 96: 801–817.
ESHEL, I., and L. L. CAVALLI-SFORZA, 1982 Assortment of encounters and evolution of cooperativeness. Proc. Natl. Acad. Sci. USA 79: 1331–1335 (corrigendum: Proc. Natl. Acad. Sci. USA 80: 1135).
EWENS, W. J., 1994 The changing role of population genetics theory, pp. 186–197 in Frontiers in Mathematical Biology, edited by S. A. LEVIN. Springer, Berlin.
EWENS, W. J., 2004 Mathematical Population Genetics I: Theoretical Introduction, Ed. 2. Springer, New York.
FISHER, R. A., 1930 The Genetical Theory of Natural Selection. Clarendon Press, Oxford.
GAVRILETS, S., 2004 Fitness Landscapes and the Origin of Species. Princeton University Press, Princeton, NJ.
GAVRILETS, S., and N. GIBSON, 2002 Fixation probabilities in a spatially heterogeneous environment. Popul. Ecol. 44: 51–58.[CrossRef]
GIBSON, G., and I. DWORKIN, 2004 Uncovering cryptic genetic variation. Nat. Rev. Genet. 5: 681–690.[Medline]
GILLESPIE, J. H., 1983 A simple stochastic gene substitution process. Theor. Popul. Biol. 23: 202–215.[CrossRef][Medline]
GILLESPIE, J. H., 1984 Molecular evolution over the mutational landscape. Evolution 38: 1116–1129.[CrossRef]
GOODNIGHT, C. J., 2000 Modeling gene interactions in structured populations, pp. 129–145 in Epistasis and the Evolutionary Process, edited by J. B. WOLF, E. D. BRODIE III and M. J. WADE. Oxford University Press, New York.
GOODNIGHT, C. J., and M. J. WADE, 2000 The ongoing synthesis: a reply to Coyne, Barton and Turelli. Evolution 54: 317–324.[CrossRef][Medline]
GRISWOLD, C. K., and M. C. WHITLOCK, 2003 The genetics of adaptation: the roles of pleiotropy, stabilizing selection and drift in shaping the distribution of bidirectional fixed mutational effects. Genetics 165: 2181–2192.
HALDANE, J. B. S., 1927 A mathematical theory of natural and artificial selection. V. Selection and mutation. Proc. Camb. Philos. Soc. 23: 838–844.
HAMMERSTEIN, P., 1996 Darwinian adaptation, population genetics and the streetcar theory of evolution. J. Math. Biol. 34: 511–532.[Medline]
HARTL, D. L., and C. H. TAUBES, 1996 Compensatory nearly neutral mutations: selection without adaptation. J. Theor. Biol. 182: 303–309.[CrossRef][Medline]
HARTL, D. L., and C. H. TAUBES, 1998 Towards a theory of evolutionary adaptation. Genetica 102/103: 525–533.
HARTL, D. L., D. E. DYKHUIZEN and A. M. DEAN, 1985 Limits of adaptation: the evolution of selective neutrality. Genetics 111: 655–674.
HERMISSON, J., and G. P. WAGNER, 2004 The population genetic theory of hidden variation and genetic robustness. Genetics 168: 2271–2284.
HEY, J., and C. A. MACHADO, 2003 The study of structured populations—new hope for a difficult and divided science. Nat. Rev. Genet. 4: 535–543.[CrossRef][Medline]
KAUFFMAN, S., and S. LEVIN, 1987 Towards a general theory of adaptive walks on rugged landscapes. J. Theor. Biol. 128: 11–45.[Medline]
KIMURA, M., 1983 The Neutral Theory of Molecular Evolution. Cambridge University Press, Cambridge, UK.
KIMURA, M., 1986 DNA and the neutral theory. Philos. Trans. R. Soc. Lond. Ser. B 312: 343–354.[Medline]
KIMURA, M., 1990 Some models of neutral evolution, compensatory evolution, and the shifting balance process. Theor. Popul. Biol. 37: 150–158.[CrossRef][Medline]
KIMURA, M., 1991 Recent development of the neutral theory viewed from the Wrightian tradition of theoretical population genetics. Proc. Natl. Acad. Sci. USA 88: 5969–5973.
KIMURA, M., and N. TAKAHATA, 1983 Selective constraint in protein polymorphism: study of the selectively neutral mutation model by using an improved pseudosampling method. Proc. Natl. Acad. Sci. USA 80: 1048–1052.
LANDE, R., 1979 Effective deme sizes during long-term evolution estimated from rates of chromosome rearrangement. Evolution 33: 234–251.[CrossRef]
MARUYAMA, T., 1970 On the fixation probability of mutant genes in a subdivided population. Genet. Res. 15: 221–225.[Medline]
MAYNARD SMITH, J., 1970 Natural selection and the concept of a protein space. Nature 225: 563–564.[CrossRef][Medline]
NISHINO, J., and F. TAJIMA, 2004 Effect of dominance on heterozygosity and the fixation probability in a subdivided population. Genes Genet. Syst. 79: 41–48.[CrossRef][Medline]
NISHINO, J., and F. TAJIMA, 2005 Effect of population structure on the amount of polymorphism and the fixation probability under overdominant selection. Genes Genet. Syst. 80: 287–295.[CrossRef][Medline]
ORR, H. A., 1998 The population genetics of adaptation: the distribution of factors fixed during adaptive evolution. Evolution 52: 935–949.[CrossRef]
ORR, H. A., 2002 The population genetics of adaptation: the adaptation of DNA sequences. Evolution 56: 1317–1330.[CrossRef][Medline]
PHILLIPS, P. C., 1996 Waiting time for a compensatory mutation: phase zero of the shifting-balance process. Genet. Res. 67: 271–283.[Medline]
POON, A., and S. P. OTTO, 2000 Compensating for our load of mutations: freezing the meltdown of small populations. Evolution 54: 1467–1479.[CrossRef][Medline]
RICE, S. H., 1990 A geometric model for the evolution of development. J. Theor. Biol. 143: 319–342.
ROZE, D., and F. ROUSSET, 2003 Selection and drift in subdivided populations: a straightforward method for deriving diffusion approximations and applications involving dominance, selfing and local extinctions. Genetics 165: 2153–2165.
RUTHERFORD, S. L., 2000 From genotype to phenotype: buffering mechanisms and the storage of genetic information. BioEssays 22: 1095–1105.[CrossRef][Medline]
SLATKIN, M., 1981 Fixation probabilities and fixation times in a subdivided population. Evolution 35: 477–478.[CrossRef]
STEPHAN, W., 1996 The rate of compensatory evolution. Genetics 144: 419–426.[Abstract]
TAKAHASI, K. R., and F. TAJIMA, 2005 Evolution of coadaptation in a two-locus epistatic system. Evolution 59: 2324–2332.[Medline]
TAKAHATA, N., 1991 Genealogy of neutral genes and spreading of selected mutations in a geographically structured population. Genetics 129: 585–595.[Abstract]
THOMPSON, J. N., 1999 Specific hypotheses on the geographic mosaic of coevolution. Am. Nat. 153: S1–S14.[CrossRef]
VAN NIMWEGEN, E., J. P. CRUTCHFIELD and M. HUYNEN, 1999 Neutral evolution of mutational robustness. Proc. Natl. Acad. Sci. USA 96: 9716–9720.
WADE, M. J., 2002 A gene's eye view of epistasis, selection and speciation. J. Evol. Biol. 15: 337–346.[CrossRef]
WADE, M. J., and C. J. GOODNIGHT, 1998 Perspective: the theories of Fisher and Wright in the context of metapopulations: when nature does many small experiments. Evolution 52: 1537–1553.[CrossRef]
WAGNER, A., 2005 Robustness and Evolvability in Living Systems. Princeton University Press, Princeton, NJ.
WAKELEY, J., 2004 Metapopulation models for historical inference. Mol. Ecol. 13: 865–875.[CrossRef][Medline]
WEINREICH, D. M., R. A. WATSON and L. CHAO, 2005 Perspective: sign epistasis and genetic constraint on evolutionary trajectories. Evolution 59: 1165–1174.[CrossRef][Medline]
WELCH, J. J., and D. WAXMAN, 2003 Modularity and the cost of complexity. Evolution 57: 1723–1734.[CrossRef][Medline]
WHITLOCK, M. C., 2003 Fixation probability and time in subdivided populations. Genetics 164: 767–779.
WHITLOCK, M. C., and R. GOMULKIEWICZ, 2005 Probability of fixation in a heterogeneous environment. Genetics 171: 1407–1417.
WHITLOCK, M. C., P. C. PHILLIPS and M. J. WADE, 1993 Gene interaction affects the additive genetic variance in subdivided populations with migration and extinction. Evolution 47: 1758–1769.[CrossRef]
WHITLOCK, M. C., P. C. PHILLIPS, F. B.-G. MOORE and S. J. TONSOR, 1995 Multiple fitness peaks and epistasis. Annu. Rev. Ecol. Syst. 26: 601–629.[CrossRef]
WHITLOCK, M. C., C. K. GRISWOLD and A. D. PETERS, 2003 Compensating for the meltdown: the critical effective size of a population with deleterious and compensatory mutations. Ann. Zool. Fenn. 40: 169–183.
WILKE, C. O., 2001 Adaptive evolution on neutral networks. Bull. Math. Biol. 63: 715–730.[CrossRef][Medline]
WRIGHT, S., 1931 Evolution in Mendelian populations. Genetics 16: 97–159.
WRIGHT, S., 1932 The roles of mutation, inbreeding, crossbreeding, and selection in evolution. Proc. 6th Int. Cong. Genet. 1: 356–366.
WRIGHT, S., 1980 Genic and organismic selection. Evolution 34: 825–843.[CrossRef]
ZHANG, J., H. F. ROSENBERG and M. NEI, 1998 Positive Darwinian selection after gene duplication in primate ribonuclease genes. Proc. Natl. Acad. Sci. USA 95: 3708–3713.
| ||||||||||
