- Email: [email protected]
Reinforcement Can Overcome Gene Flow during Speciation in Drosophila
Current Biology 20, 2229–2233, December 21, 2010 ª2010 Elsevier Ltd All rights reserved
DOI 10.1016/j.cub.2010.11.036
Report Reinforcement Can Overcome Gene Flow during Speciation in Drosophila Daniel R. Matute1,* 1Department of Ecology and Evolution, The University of Chicago, 1101 East 57 Street, Chicago, IL 60637, USA
Summary Reinforcement, the strengthening of prezygotic reproductive isolation by natural selection in response to maladaptive hybridization [1–3], is one of the few processes in which natural selection directly favors the evolution of species as discrete groups (e.g., [4–7]). The evolution of reproductive barriers via reinforcement is expected to evolve in regions where the ranges of two species overlap and hybridize as an evolutionary solution to avoiding the costs of maladaptive hybridization [2,3,8]. The role of reinforcement in speciation has, however, been highly controversial because population-genetic theory suggests that the process is severely impeded by both hybridization [8–11] and migration of individuals from outside the contact zone [12,13]. To determine whether reinforcement could strengthen the reproductive barriers between two sister species of Drosophila in the face of these impediments, I initiated experimental populations of these two species that allowed different degrees of hybridization, as well as migration from outside populations. Surprisingly, even in the face of gene flow, reinforcement could promote the evolution of reproductive isolation within only five generations. As theory predicts, high levels of hybridization (and/or strong selection against hybrids) and migration impeded this evolution. These results suggest that reinforcement can help complete the process of speciation. Results and Discussion The process of reinforcement, or the strengthening by natural selection of prezygotic isolation between closely related taxa as an evolutionary response to maladaptive hybridization, was once seen as the inevitable last stage of speciation [14] but was later deemed to be extremely unlikely [15,16]. Although its importance remains contentious [2,3], reinforcement is supported by more recent data. Reinforcement is usually documented by observing a biogeographic pattern in which a reproductive isolating barrier is stronger in areas where two species overlap (‘‘sympatric’’) than in areas outside each other’s range (‘‘allopatric’’ [4–7]). But there are other explanations for such a pattern [17–19], and there has been little evidence that reinforcement can strengthen reproductive isolation in laboratory experiments [3,20]. Drosophila yakuba and its sister species D. santomea have several characteristics that make them ideal candidates for experimental studies of reinforcement. First, they hybridize within a well-demarcated hybrid zone [21,22] on the slopes of the African volcanic island of Sa˜o Tome´. Second, I have previously found evidence suggesting that natural selection
*Correspondence: [email protected]
has acted in the wild to reduce maladaptive hybridization between these species: D. yakuba females from the hybrid zone show higher gametic isolation from males of D. santomea than do D. yakuba females from outside the hybrid zone. This elevated isolation in sympatric D. yakuba females results from their faster depletion of heterospecific sperm [7]. In contrast, there is no evidence for reinforced behavioral (sexual) isolation in the wild [7]. Third, when D. yakuba lines derived from allopatric populations are exposed to experimental sympatry with D. santomea and no hybridization is allowed (that is, when all hybrids are removed), there is an evolutionary increase in both behavioral and gametic isolation after only four generations [7]. Finally, both pure species and their reciprocal F1 hybrids are distinguishable by the degree of abdominal pigmentation [23,24]. This morphological species difference allows experimentation on naturally collected isofemale lines (i.e., derived from a single wild-caught female) rather than on inbred mutant stocks whose hybrids are identifiable by their wild-type phenotype. Taking advantage of these features, I created experimental ‘‘sympatry populations’’ (bottles that contained both D. yakuba and D. santomea) and allowed these populations to experience different degrees of selection against F1 hybrids, as well as migration from ‘‘allopatric’’ bottles of flies. This models a natural situation in which two species with overlapping ranges produce hybrids. The disadvantage of these hybrids is taken to be a byproduct of evolution that occurred when their parental species were previously completely isolated geographically. Different numbers of F1 females in the population model different degrees of selection against the hybrids and hence in favor of reinforcement (in this case, selection imposed on the F1 hybrids was determined by the number of F1 females that were allowed to survive every generation). Varying levels of migration from stock bottles of the two species represent differential influx of migrants from allopatric populations. To study the effect of migration on reinforcement, I allowed different levels of movement into the sympatry bottles of individuals from ‘‘allopatric’’ populations that were never exposed to the other species. These individuals were collected as virgins from stock bottles. The five different migration treatments involved transferring every generation 0, 4, 8, 12, or 16 allopatric individuals (half virgin females, half virgin males) into the sympatry bottles. To vary the strength of selection against hybrids, I allowed different numbers of F1 hybrid females (easily identifiable by pigmentation) to survive each generation (F1 male hybrids were destroyed because they are sterile). There were five treatments, involving survival of 0, 4, 8, 14, or 20 female hybrids. Levels of migration were measured as the proportion of migrants per generation per species each generation (mpgps), and levels of hybridization were measured as the proportion of surviving F1 female hybrids in the population each generation (F1fh). In total, there were 25 different combinations of migration and hybridization. I estimated the effect of hybridization and migration on the population composition of species by determining how many F1 hybrids of both sexes appeared each generation in every experimental bottle. Treatments with low levels of migration
Current Biology Vol 20 No 24 2230
A
B
C
D
E
F
G
H
I
Figure 1. Hybrid Production in 9 of the 25 Experimental Combinations of Migration and Hybridization (A–I) Red lines represent the proportion of total hybrids for both reciprocal crosses observed in the experimental population. Blue lines represent the proportion of \ yak 3 san _ hybrid males among all males in each population. yak 3 san hybrid males were used as a proxy for the total number of yak 3 san hybrids because it is not possible to distinguish san 3 yak and yak 3 san female hybrids. Yellow lines represent the proportion of \san 3 yak _ hybrid males among all males in each population. Dotted lines represent 61 standard error. Each regression line is shown with its r2 (i.e., coefficient of determination) in the legend. Treatments with low levels of hybridization and migration (A, B, D) showed a reduced frequency of hybrids over time (i.e., the evolution of reinforcement), whereas those with high levels of gene flow showed either no significant change (C, E, F, G) or an increase (H, I) in the proportion of hybrids over time (i.e., the breakdown of preexisting reproductive isolation). Figure S1 shows the hybrid production for the other 16 treatments.
and hybridization showed a reduction in the observed frequency of hybrids over time, whereas those experiencing high levels of gene flow showed an increase in the proportion of hybrids (Figure 1). Both the amount of hybridization and the amount of migration showed significant effects on the number of hybrids appearing in each treatment (linear mixed model [LMM]: hybridization, F1,71 = 282.0, p < 0.0001; migration: F1,71 = 264.3, p < 0.0001; interaction between factors: F1,1 = 52.2, p < 0.0001; Figure 1; see also Figure S1 available online). To establish what form of reproductive isolation caused the reduction in the number of F1 hybrids in those treatments in which migration and hybridization were low, I measured, after five and ten generations of experimental sympatry, the levels of sexual and gametic isolation in the 25 treatments with different levels of hybridization and migration. Females exposed to experimental sympatry from both species evolved increased sexual and gametic isolation within five generations, but only in those treatments in which levels of migration were fairly low and in which selection against F1 hybrids was strong (reinforcement for both gametic and sexual isolation required that migration be lower than 0.13% mpgps and hybridization be lower than 0.07 F1fh; Figure 2; Figure 3;
Tables S1 and S2). After ten generations, reinforced reproductive isolation in D. yakuba persisted in those treatments in which selection against hybrids was fairly strong and persisted in D. santomea when migration was low (reinforced sexual and gametic isolation in D. yakuba females evolved when hybridization was lower than 0.12 F1fh and migration was lower than 0.20 mpgps. Reinforcement of sexual and gametic isolation in D. santomea occurred when hybridization was lower than 0.12 F1fh, as well as at several levels of migration; Figure 2; Figure 3; Tables S1 and S2). To analyze the influence of selection against hybrids and of migration on reproductive isolation, I fitted LMMs for sexual and gametic isolation for D. yakuba and D. santomea separately [25], establishing which model best explained the heterogeneity on both types of reproductive isolation among treatments (combinations of different levels of hybridization and migration). The strength of sexual isolation in both species and the gametic isolation in D. santomea after five generations of experimental sympatry were best explained by a model incorporating both main factors (migration and selection against hybrids, as well as the interaction term in gametic isolation in D. santomea), whereas heterogeneity in levels of
Can Gene Flow Prevent Speciation by Reinforcement? 2231
A
C
Figure 2. Reproductive Isolation after Five and Ten Generations of Experimental Sympatry in D. yakuba Females under Different Levels of Migration and Hybridization (A and C) Sexual isolation. (B and D) Gametic isolation. Key to colors: black: reproductive isolation (sexual or gametic) is significantly higher than control bottles (i.e., those not exposed to D. santomea); gray: reproductive isolation is not significantly different from controls; white: reproductive isolation is significantly lower than controls. Different combinations of hybridization and migration levels produce substantial differences in the levels of reproductive isolation at generations 5 and 10.
B
D
sexual isolation in D. yakuba is best explained by a model that includes only migration (Table 1). On the other hand, after ten generations, the full-factorial model better explained both types of reproductive isolation in both species (Table 1). These results show that both the strength of selection against hybrids and the level of migration played a significant and substantial role in the evolution of both types of reproductive isolation. Not surprisingly, the strongest increase in isolation over time, compared to controls, was seen when selection was complete (that is, when all F1 hybrids were removed) and when there was no migration from allopatric areas. This kind of selection is sometimes called ‘‘reproductive character displacement’’ (RCD) rather than ‘‘reinforcement,’’ because it mimics what happens when two species meet but when postzygotic isolation is already complete—that is, when they are already good biological species. Nevertheless, in both species, sexual and gametic isolation also increased significantly under low levels of migration and strong—but not complete—selection against hybrids (Figure 2; Figure 3); this situation corresponds to true speciation through reinforcement. The magnitude of reproductive isolation within treatments showed changes in 85 out of 100 treatments between generations 5 and 10 (Table S3). In general, reproductive isolation increased in those treatments in which levels of hybridization and migration were low, and reproductive isolation decreased in those treatments in which these two factors were high. In D. yakuba, regardless of the level of migration, high levels of hybridization (i.e., more than 0.12 F1fh) overwhelmed the effect of reinforcing selection, actually reducing the level of reproductive isolation below that seen in control bottles (Figure 2; Figure 3; Table 1; Tables S1–S3). This suggests that if the species boundary becomes more permeable (i.e., if there is
not selection against female hybrids), the two species will collapse into a hybrid swarm. This experimental design leads to two caveats. First, to allow gene flow between species, we kept the number of surviving F1 female hybrids each generation constant. Maintaining constant numbers of surviving F1 females required varying the magnitude of selection against the hybrids, for as selection increased the reproductive isolation between species, fewer hybrids were produced. Accordingly, a greater fraction of produced hybrids were allowed to survive over time, which in turn meant that as selection succeeded, its intensity decreased. Second, I was unable to control the number of surviving backcross flies in the population (and thus was unable to judge their impact on the evolution of reproductive isolation) because their pigmentation is often indistinguishable from that of pure-species individuals. Because I could not follow backcross flies, the number of surviving F1 hybrids is a proxy for—and not the exact value of—the amount of selection against hybrid genotypes. Given that females from backcrosses have similar levels of sexual and gametic isolation as do the pure-species females from the their father’s species (e.g., (san 3 yak) 3 yak and yak females have similar behavioral and gametic isolation levels as pure-species yak and san females; [26,27]), this situation is a fairly good approximation to what happens when hybrid incompatibility is caused by a Dobzhansky-Muller incompatibility involving two (or more) epistatically dominant factors (this is because the deleterious genic interaction occurs mainly in F1 hybrids and is rarer among backcross individuals because of recombination). The results shown here raise additional questions about the evolution of reinforcement in the D. yakuba-D. santomea species pair. Natural populations of D. yakuba that are sympatric with those of D. santomea show stronger gametic isolation than do allopatric populations, implying reinforcement. Oddly, however, geography seems to have no effect on levels of sexual isolation. Yet this experimental study clearly shows that allopatric D. yakuba populations have genetic variation for sexual isolation that could form the basis for reinforced mate discrimination in the wild [7]. It is a mystery why we have not observed this in nature. Additionally, I found that D. santomea also shows genetic variation for increased gametic and sexual isolation from D. yakuba males, but I have observed no reinforcement for either trait in the wild [7,27]. This disparity between the laboratory and field results might reflect the greater proportion of D. santomea than D. yakuba in the hybrid zone on Sa˜o Tome´,
Current Biology Vol 20 No 24 2232
A
Figure 3. Reproductive Isolation after Five and Ten Generations of Experimental Sympatry in D. santomea Females under Different Treatments of Migration and Hybridization
C
(A and C) Sexual isolation. (B and D) Gametic isolation. Key to colors: black: reproductive isolation (sexual or gametic) is significantly higher than control bottles (i.e., those not exposed to D. yakuba); gray: reproductive isolation is not significantly different from controls; white: reproductive isolation is significantly lower than controls. Different combinations of hybridization and migration levels produce substantial differences in the levels of reproductive isolation at generations 5 and 10.
B
D
which would make selection for strengthened reproductive isolation on D. santomea weaker (the consequences of maladaptive hybridization are less severe in the more common species [2,3]). Alternatively, because reproductive isolation with D. yakuba is already high in natural populations of D. santomea (reproductive isolating mechanisms measured under laboratory conditions reduce interspecific gene flow [compared to intraspecific controls] up to 93% for \ yak 3 san _ crosses and up to 98% for \ san 3 yak _ crosses [22]), the selective pressure to increase reproductive isolation in D. santomea is weaker than in D. yakuba. And, of course, both the population-frequency and preexisting-isolation explanations are possible.
How do these results compare with previous work? Several earlier studies [7,28,29] examined mixed populations of two sister species of Drosophila to determine whether complete selection against hybrids could yield reproductive isolation. In these cases, reproductive isolation evolved quickly. In all of these studies, however, no gene flow was possible, because hybrids were completely eliminated (or were totally unfit) from populations each generation before reproduction. So although these studies demonstrated that artificial sympatry can indeed promote the rapid evolution of prezygotic isolation (both premating [7,28,29] and/or postmating-prezygotic [7]), the lack of gene flow meant that these studies were models not of reinforcement, but of reproductive character displacement, a postspeciation phenomenon. Related work includes experimental studies of sympatric speciation (divergence with gene flow). Sympatric speciation requires the evolution of mate discrimination in the face of gene flow between the divergent populations [30–33]. These studies have demonstrated that assortative mating can evolve as a byproduct of a strong selection regime involving traits of potential ecological relevance. But these studies do not constitute a test of reinforcement, because the increased reproductive isolation does not evolve as a way to avoid the fitness costs associated with postzygotic isolation (in sympatric speciation, the reduction of gene flow and isolation
Table 1. Different Levels of Selection against F1 Hybrids and Migration Yield Differences in Gametic and Sexual Isolation in D. yakuba and D. santomea after Five and Ten Generations of Experimental Sympatry D. santomea
D. yakuba
Sexual Isolation
Gametic Isolation
Sexual Isolation
Gametic Isolation
2498.52 2516.23 2491.56 2512.1
23420.5 23405.69 23412.57 23229.79
2456.35 2473.52 2457.39 2474.9
22233.85 22243.5 21791.86 21751.46
277.3 242.65 24.87 211.73
22094.98 22066.25 21842.75 21891.54
2225.67 2140.97 106.05 222.08
2561.44 2482.24 10.2 2396.04
Generation 5 Selection + migration + interaction Selection + migration Selection Migration Generation 10 Selection + migration + interaction Selection + migration Selection Migration
To estimate the importance of each effect (migration, hybridization, interaction between migration and hybridization), I formulated four mixed-effect linear models that differed in their fixed effects and in which differences among individuals within each replicate were considered to be random effects. To determine which model was more likely to explain the data, I used the Akaike information criterion (AIC [20]) scores of each model. The model with the lowest AIC for each type of reproductive isolation is highlighted in bold. See also Tables S1–S3.
Can Gene Flow Prevent Speciation by Reinforcement? 2233
between subpopulations evolves while they adapt to different habitats, not as a response to reduce the production of inferior hybrids [34,35]). Mathematical theory has shown that the evolution of reinforcement depends critically on the amount of hybridization: with even small amounts of gene flow, populations fuse rather than becoming more isolated [2,8–10,36,37]. This is because gene flow breaks up associations between alleles that can cause postzygotic isolation, between characters that can cause prezygotic isolation, and between characters that cause pre- and postzygotic isolation [3,9,11]. My experimental results confirm this conclusion: even moderate levels of gene flow, in the form of either hybridization or migration, overwhelmed the effects of reinforcing selection. This pattern was also seen in observations of Timema walking sticks in nature [38,39]. Taken together, the results described here confirm theoretical formulations asserting that reinforcement can be important in completing the speciation process, especially when gene flow is not too high. Other factors that might affect the probability of reinforcement remain to be studied; these include migration from the hybrid zone back to allopatric populations, which will eliminate the pattern of reinforcement by spreading it throughout allopatric populations, heterogeneity in habitat structure across the geographic range of a species, and population structure. But although these aspects of reinforcement remain unexplored, this study confirms that the process is at least a plausible factor in the final stages of speciation. Supplemental Information Supplemental Information includes Supplemental Experimental Procedures, one figure, and four tables and can be found with this article online at doi:10.1016/j.cub.2010.11.036. Acknowledgments I thank J.A. Coyne, T.D. Price, D. Kennedy, M.A.F. Noor, M.A. Sprigge, and N. Bloch for critical discussions and I.A. Butler and J. Gladstone for technical help. This work was funded by National Institutes of Health grant R01GM058260 to Jerry A. Coyne. Received: August 25, 2010 Revised: October 28, 2010 Accepted: November 4, 2010 Published online: December 2, 2010 References 1. Dobzhansky, Th. (1940). Speciation as a stage in evolutionary divergence. Am. Nat. 74, 312–321. 2. Servedio, M.R., and Noor, M.A.F. (2003). The role of reinforcement in speciation: Theory and data. Annu. Rev. Ecol. Evol. Syst. 34, 339–364. 3. Coyne, J.A., and Orr, H.A. (2004). Speciation (Sunderland, MA: Sinauer Associates). 4. Rundle, H.D., and Schluter, D. (1998). Reinforcement of stickleback mate preferences: Sympatry breeds contempt. Evolution 52, 200–208. 5. Saetre, G.P., Moum, T., Bures, S., Kral, M., Adamjan, M., and Moreno, J. (1997). A sexually selected character displacement in flycatchers reinforces premating isolation. Nature 387, 589–592. 6. Jaenike, J., Dyer, K.A., Cornish, C., and Minhas, M.S. (2006). Asymmetrical reinforcement and Wolbachia infection in Drosophila. PLoS Biol. 4, e325. 7. Matute, D.R. (2010). Reinforcement of gametic isolation in Drosophila. PLoS Biol. 8, e1000341. 8. Servedio, M.R. (2000). Reinforcement and the genetics of nonrandom mating. Evolution 54, 21–29. 9. Servedio, M.R., and Kirkpatrick, M. (1997). The effects of gene flow on reinforcement. Evolution 51, 1764–1772.
10. Kirkpatrick, M., and Servedio, M.R. (1999). The reinforcement of mating preferences on an island. Genetics 151, 865–884. 11. Sanderson, N. (1989). Can gene flow prevent reinforcement? Evolution 43, 1223–1235. 12. Liou, L.W., and Price, T.D. (1994). Speciation by reinforcement of premating isolation. Evolution 48, 1451–1459. 13. Caisse, M., and Antonovics, J. (1978). Evolution in closely adjacent plant populations. Heredity 40, 371–384. 14. Dobzhansky, T. (1951). Genetics and the Origin of the Species (New York: Columbia University Press). 15. Spencer, H.G., Mcardle, B.H., and Lambert, D.M. (1986). A theoretical investigation of speciation by reinforcement. Am. Nat. 128, 241–262. 16. Marshall, J.L., Arnold, M.L., and Howard, D.J. (2002). Reinforcement: The road not taken. Trends Ecol. Evol. 17, 558–563. 17. Howard, D.J. (1993). Reinforcement: Origin, dynamics, and fate of an evolutionary hypothesis. In Hybrid Zones and the Evolutionary Process, R.G. Harrison, ed. (New York: Oxford University Press), pp. 46–69. 18. Noor, M.A.F. (1999). Reinforcement and other consequences of sympatry. Heredity 83, 503–508. 19. Price, T.D. (2008). Speciation in Birds (Greenwood Village, CO: Roberts & Co. Publishers). 20. Harper, A.A., and Lambert, D.M. (1983). The population genetics of reinforcing selection. Genetica 62, 15–23. 21. Llopart, A., Lachaise, D., and Coyne, J.A. (2005). Multilocus analysis of introgression between two sympatric sister species of Drosophila: Drosophila yakuba and D. santomea. Genetics 171, 197–210. 22. Llopart, A., Lachaise, D., and Coyne, J.A. (2005). An anomalous hybrid zone in Drosophila. Evolution 59, 2602–2607. 23. Llopart, A., Elwyn, S., Lachaise, D., and Coyne, J.A. (2002). Genetics of a difference in pigmentation between Drosophila yakuba and Drosophila santomea. Evolution 56, 2262–2277. 24. Carbone, M.A., Llopart, A., deAngelis, M., Coyne, J.A., and Mackay, T.F.C. (2005). Quantitative trait loci affecting the difference in pigmentation between Drosophila yakuba and D. santomea. Genetics 171, 211–225. 25. Pinheiro, J., and Bates, D. (2000). Mixed-Effects Models in S and S-PLUS (New York: Springer). 26. Coyne, J.A., Kim, S.Y., Chang, A.S., Lachaise, D., and Elwyn, S. (2002). Sexual isolation between two sibling species with overlapping ranges: Drosophila santomea and Drosophila yakuba. Evolution 56, 2424–2434. 27. Matute, D.R., and Coyne, J.A. (2010). Intrinsic reproductive isolation between two sister species of Drosophila. Evolution 64, 903–920. 28. Koopman, K.F. (1950). Natural selection for reproductive isolation between Drosophila pseudoobscura and Drosophila persimilis. Evolution 4, 135–148. 29. Higgie, M., Chenoweth, S., and Blows, M.W. (2000). Natural selection and the reinforcement of mate recognition. Science 290, 519–521. 30. Thoday, J.M., and Gibson, J.B. (1970). The probability of isolation by disruptive selection. Am. Nat. 104, 219–230. 31. Thoday, J.M., and Gibson, J.B. (1962). Isolation by disruptive selection. Nature 193, 1164–1166. 32. Rice, W.R., and Salt, G.W. (1988). Speciation via disruptive selection on habitat preference: Experimental evidence. Am. Nat. 131, 911–917. 33. Rice, W.R., and Salt, G.W. (1990). The evolution of reproductive isolation as a correlated character under sympatric conditions: Experimental evidence. Evolution 44, 1140–1152. 34. Bush, G.L. (1994). Sympatric speciation in animals: New wine in old bottles. Trends Ecol. Evol. 9, 285–288. 35. Diehl, S.R., and Bush, G.L. (1989). The role of habitat preference in adaptation and speciation. In Speciation and Its Consequences, D. Otte and J. Endler, eds. (Sunderland, MA: Sinauer), pp. 345–365. 36. Kelly, J.K., and Noor, M.A. (1996). Speciation by reinforcement: A model derived from studies of Drosophila. Genetics 143, 1485–1497. 37. Servedio, M.R. (2004). The evolution of premating isolation: Local adaptation and natural and sexual selection against hybrids. Evolution 58, 913–924. 38. Nosil, P., Crespi, B.J., and Sandoval, C.P. (2003). Reproductive isolation driven by the combined effects of ecological adaptation and reinforcement. Proc. Biol. Sci. 270, 1911–1918. 39. Nosil, P. (2005). The role of selection and gene flow in the evolution of sexual isolation in Timema walking-sticks and other Orthopteroids. J. Orthoptera Res. 14, 247–253.
DOI 10.1016/j.cub.2010.11.036
Report Reinforcement Can Overcome Gene Flow during Speciation in Drosophila Daniel R. Matute1,* 1Department of Ecology and Evolution, The University of Chicago, 1101 East 57 Street, Chicago, IL 60637, USA
Summary Reinforcement, the strengthening of prezygotic reproductive isolation by natural selection in response to maladaptive hybridization [1–3], is one of the few processes in which natural selection directly favors the evolution of species as discrete groups (e.g., [4–7]). The evolution of reproductive barriers via reinforcement is expected to evolve in regions where the ranges of two species overlap and hybridize as an evolutionary solution to avoiding the costs of maladaptive hybridization [2,3,8]. The role of reinforcement in speciation has, however, been highly controversial because population-genetic theory suggests that the process is severely impeded by both hybridization [8–11] and migration of individuals from outside the contact zone [12,13]. To determine whether reinforcement could strengthen the reproductive barriers between two sister species of Drosophila in the face of these impediments, I initiated experimental populations of these two species that allowed different degrees of hybridization, as well as migration from outside populations. Surprisingly, even in the face of gene flow, reinforcement could promote the evolution of reproductive isolation within only five generations. As theory predicts, high levels of hybridization (and/or strong selection against hybrids) and migration impeded this evolution. These results suggest that reinforcement can help complete the process of speciation. Results and Discussion The process of reinforcement, or the strengthening by natural selection of prezygotic isolation between closely related taxa as an evolutionary response to maladaptive hybridization, was once seen as the inevitable last stage of speciation [14] but was later deemed to be extremely unlikely [15,16]. Although its importance remains contentious [2,3], reinforcement is supported by more recent data. Reinforcement is usually documented by observing a biogeographic pattern in which a reproductive isolating barrier is stronger in areas where two species overlap (‘‘sympatric’’) than in areas outside each other’s range (‘‘allopatric’’ [4–7]). But there are other explanations for such a pattern [17–19], and there has been little evidence that reinforcement can strengthen reproductive isolation in laboratory experiments [3,20]. Drosophila yakuba and its sister species D. santomea have several characteristics that make them ideal candidates for experimental studies of reinforcement. First, they hybridize within a well-demarcated hybrid zone [21,22] on the slopes of the African volcanic island of Sa˜o Tome´. Second, I have previously found evidence suggesting that natural selection
*Correspondence: [email protected]
has acted in the wild to reduce maladaptive hybridization between these species: D. yakuba females from the hybrid zone show higher gametic isolation from males of D. santomea than do D. yakuba females from outside the hybrid zone. This elevated isolation in sympatric D. yakuba females results from their faster depletion of heterospecific sperm [7]. In contrast, there is no evidence for reinforced behavioral (sexual) isolation in the wild [7]. Third, when D. yakuba lines derived from allopatric populations are exposed to experimental sympatry with D. santomea and no hybridization is allowed (that is, when all hybrids are removed), there is an evolutionary increase in both behavioral and gametic isolation after only four generations [7]. Finally, both pure species and their reciprocal F1 hybrids are distinguishable by the degree of abdominal pigmentation [23,24]. This morphological species difference allows experimentation on naturally collected isofemale lines (i.e., derived from a single wild-caught female) rather than on inbred mutant stocks whose hybrids are identifiable by their wild-type phenotype. Taking advantage of these features, I created experimental ‘‘sympatry populations’’ (bottles that contained both D. yakuba and D. santomea) and allowed these populations to experience different degrees of selection against F1 hybrids, as well as migration from ‘‘allopatric’’ bottles of flies. This models a natural situation in which two species with overlapping ranges produce hybrids. The disadvantage of these hybrids is taken to be a byproduct of evolution that occurred when their parental species were previously completely isolated geographically. Different numbers of F1 females in the population model different degrees of selection against the hybrids and hence in favor of reinforcement (in this case, selection imposed on the F1 hybrids was determined by the number of F1 females that were allowed to survive every generation). Varying levels of migration from stock bottles of the two species represent differential influx of migrants from allopatric populations. To study the effect of migration on reinforcement, I allowed different levels of movement into the sympatry bottles of individuals from ‘‘allopatric’’ populations that were never exposed to the other species. These individuals were collected as virgins from stock bottles. The five different migration treatments involved transferring every generation 0, 4, 8, 12, or 16 allopatric individuals (half virgin females, half virgin males) into the sympatry bottles. To vary the strength of selection against hybrids, I allowed different numbers of F1 hybrid females (easily identifiable by pigmentation) to survive each generation (F1 male hybrids were destroyed because they are sterile). There were five treatments, involving survival of 0, 4, 8, 14, or 20 female hybrids. Levels of migration were measured as the proportion of migrants per generation per species each generation (mpgps), and levels of hybridization were measured as the proportion of surviving F1 female hybrids in the population each generation (F1fh). In total, there were 25 different combinations of migration and hybridization. I estimated the effect of hybridization and migration on the population composition of species by determining how many F1 hybrids of both sexes appeared each generation in every experimental bottle. Treatments with low levels of migration
Current Biology Vol 20 No 24 2230
A
B
C
D
E
F
G
H
I
Figure 1. Hybrid Production in 9 of the 25 Experimental Combinations of Migration and Hybridization (A–I) Red lines represent the proportion of total hybrids for both reciprocal crosses observed in the experimental population. Blue lines represent the proportion of \ yak 3 san _ hybrid males among all males in each population. yak 3 san hybrid males were used as a proxy for the total number of yak 3 san hybrids because it is not possible to distinguish san 3 yak and yak 3 san female hybrids. Yellow lines represent the proportion of \san 3 yak _ hybrid males among all males in each population. Dotted lines represent 61 standard error. Each regression line is shown with its r2 (i.e., coefficient of determination) in the legend. Treatments with low levels of hybridization and migration (A, B, D) showed a reduced frequency of hybrids over time (i.e., the evolution of reinforcement), whereas those with high levels of gene flow showed either no significant change (C, E, F, G) or an increase (H, I) in the proportion of hybrids over time (i.e., the breakdown of preexisting reproductive isolation). Figure S1 shows the hybrid production for the other 16 treatments.
and hybridization showed a reduction in the observed frequency of hybrids over time, whereas those experiencing high levels of gene flow showed an increase in the proportion of hybrids (Figure 1). Both the amount of hybridization and the amount of migration showed significant effects on the number of hybrids appearing in each treatment (linear mixed model [LMM]: hybridization, F1,71 = 282.0, p < 0.0001; migration: F1,71 = 264.3, p < 0.0001; interaction between factors: F1,1 = 52.2, p < 0.0001; Figure 1; see also Figure S1 available online). To establish what form of reproductive isolation caused the reduction in the number of F1 hybrids in those treatments in which migration and hybridization were low, I measured, after five and ten generations of experimental sympatry, the levels of sexual and gametic isolation in the 25 treatments with different levels of hybridization and migration. Females exposed to experimental sympatry from both species evolved increased sexual and gametic isolation within five generations, but only in those treatments in which levels of migration were fairly low and in which selection against F1 hybrids was strong (reinforcement for both gametic and sexual isolation required that migration be lower than 0.13% mpgps and hybridization be lower than 0.07 F1fh; Figure 2; Figure 3;
Tables S1 and S2). After ten generations, reinforced reproductive isolation in D. yakuba persisted in those treatments in which selection against hybrids was fairly strong and persisted in D. santomea when migration was low (reinforced sexual and gametic isolation in D. yakuba females evolved when hybridization was lower than 0.12 F1fh and migration was lower than 0.20 mpgps. Reinforcement of sexual and gametic isolation in D. santomea occurred when hybridization was lower than 0.12 F1fh, as well as at several levels of migration; Figure 2; Figure 3; Tables S1 and S2). To analyze the influence of selection against hybrids and of migration on reproductive isolation, I fitted LMMs for sexual and gametic isolation for D. yakuba and D. santomea separately [25], establishing which model best explained the heterogeneity on both types of reproductive isolation among treatments (combinations of different levels of hybridization and migration). The strength of sexual isolation in both species and the gametic isolation in D. santomea after five generations of experimental sympatry were best explained by a model incorporating both main factors (migration and selection against hybrids, as well as the interaction term in gametic isolation in D. santomea), whereas heterogeneity in levels of
Can Gene Flow Prevent Speciation by Reinforcement? 2231
A
C
Figure 2. Reproductive Isolation after Five and Ten Generations of Experimental Sympatry in D. yakuba Females under Different Levels of Migration and Hybridization (A and C) Sexual isolation. (B and D) Gametic isolation. Key to colors: black: reproductive isolation (sexual or gametic) is significantly higher than control bottles (i.e., those not exposed to D. santomea); gray: reproductive isolation is not significantly different from controls; white: reproductive isolation is significantly lower than controls. Different combinations of hybridization and migration levels produce substantial differences in the levels of reproductive isolation at generations 5 and 10.
B
D
sexual isolation in D. yakuba is best explained by a model that includes only migration (Table 1). On the other hand, after ten generations, the full-factorial model better explained both types of reproductive isolation in both species (Table 1). These results show that both the strength of selection against hybrids and the level of migration played a significant and substantial role in the evolution of both types of reproductive isolation. Not surprisingly, the strongest increase in isolation over time, compared to controls, was seen when selection was complete (that is, when all F1 hybrids were removed) and when there was no migration from allopatric areas. This kind of selection is sometimes called ‘‘reproductive character displacement’’ (RCD) rather than ‘‘reinforcement,’’ because it mimics what happens when two species meet but when postzygotic isolation is already complete—that is, when they are already good biological species. Nevertheless, in both species, sexual and gametic isolation also increased significantly under low levels of migration and strong—but not complete—selection against hybrids (Figure 2; Figure 3); this situation corresponds to true speciation through reinforcement. The magnitude of reproductive isolation within treatments showed changes in 85 out of 100 treatments between generations 5 and 10 (Table S3). In general, reproductive isolation increased in those treatments in which levels of hybridization and migration were low, and reproductive isolation decreased in those treatments in which these two factors were high. In D. yakuba, regardless of the level of migration, high levels of hybridization (i.e., more than 0.12 F1fh) overwhelmed the effect of reinforcing selection, actually reducing the level of reproductive isolation below that seen in control bottles (Figure 2; Figure 3; Table 1; Tables S1–S3). This suggests that if the species boundary becomes more permeable (i.e., if there is
not selection against female hybrids), the two species will collapse into a hybrid swarm. This experimental design leads to two caveats. First, to allow gene flow between species, we kept the number of surviving F1 female hybrids each generation constant. Maintaining constant numbers of surviving F1 females required varying the magnitude of selection against the hybrids, for as selection increased the reproductive isolation between species, fewer hybrids were produced. Accordingly, a greater fraction of produced hybrids were allowed to survive over time, which in turn meant that as selection succeeded, its intensity decreased. Second, I was unable to control the number of surviving backcross flies in the population (and thus was unable to judge their impact on the evolution of reproductive isolation) because their pigmentation is often indistinguishable from that of pure-species individuals. Because I could not follow backcross flies, the number of surviving F1 hybrids is a proxy for—and not the exact value of—the amount of selection against hybrid genotypes. Given that females from backcrosses have similar levels of sexual and gametic isolation as do the pure-species females from the their father’s species (e.g., (san 3 yak) 3 yak and yak females have similar behavioral and gametic isolation levels as pure-species yak and san females; [26,27]), this situation is a fairly good approximation to what happens when hybrid incompatibility is caused by a Dobzhansky-Muller incompatibility involving two (or more) epistatically dominant factors (this is because the deleterious genic interaction occurs mainly in F1 hybrids and is rarer among backcross individuals because of recombination). The results shown here raise additional questions about the evolution of reinforcement in the D. yakuba-D. santomea species pair. Natural populations of D. yakuba that are sympatric with those of D. santomea show stronger gametic isolation than do allopatric populations, implying reinforcement. Oddly, however, geography seems to have no effect on levels of sexual isolation. Yet this experimental study clearly shows that allopatric D. yakuba populations have genetic variation for sexual isolation that could form the basis for reinforced mate discrimination in the wild [7]. It is a mystery why we have not observed this in nature. Additionally, I found that D. santomea also shows genetic variation for increased gametic and sexual isolation from D. yakuba males, but I have observed no reinforcement for either trait in the wild [7,27]. This disparity between the laboratory and field results might reflect the greater proportion of D. santomea than D. yakuba in the hybrid zone on Sa˜o Tome´,
Current Biology Vol 20 No 24 2232
A
Figure 3. Reproductive Isolation after Five and Ten Generations of Experimental Sympatry in D. santomea Females under Different Treatments of Migration and Hybridization
C
(A and C) Sexual isolation. (B and D) Gametic isolation. Key to colors: black: reproductive isolation (sexual or gametic) is significantly higher than control bottles (i.e., those not exposed to D. yakuba); gray: reproductive isolation is not significantly different from controls; white: reproductive isolation is significantly lower than controls. Different combinations of hybridization and migration levels produce substantial differences in the levels of reproductive isolation at generations 5 and 10.
B
D
which would make selection for strengthened reproductive isolation on D. santomea weaker (the consequences of maladaptive hybridization are less severe in the more common species [2,3]). Alternatively, because reproductive isolation with D. yakuba is already high in natural populations of D. santomea (reproductive isolating mechanisms measured under laboratory conditions reduce interspecific gene flow [compared to intraspecific controls] up to 93% for \ yak 3 san _ crosses and up to 98% for \ san 3 yak _ crosses [22]), the selective pressure to increase reproductive isolation in D. santomea is weaker than in D. yakuba. And, of course, both the population-frequency and preexisting-isolation explanations are possible.
How do these results compare with previous work? Several earlier studies [7,28,29] examined mixed populations of two sister species of Drosophila to determine whether complete selection against hybrids could yield reproductive isolation. In these cases, reproductive isolation evolved quickly. In all of these studies, however, no gene flow was possible, because hybrids were completely eliminated (or were totally unfit) from populations each generation before reproduction. So although these studies demonstrated that artificial sympatry can indeed promote the rapid evolution of prezygotic isolation (both premating [7,28,29] and/or postmating-prezygotic [7]), the lack of gene flow meant that these studies were models not of reinforcement, but of reproductive character displacement, a postspeciation phenomenon. Related work includes experimental studies of sympatric speciation (divergence with gene flow). Sympatric speciation requires the evolution of mate discrimination in the face of gene flow between the divergent populations [30–33]. These studies have demonstrated that assortative mating can evolve as a byproduct of a strong selection regime involving traits of potential ecological relevance. But these studies do not constitute a test of reinforcement, because the increased reproductive isolation does not evolve as a way to avoid the fitness costs associated with postzygotic isolation (in sympatric speciation, the reduction of gene flow and isolation
Table 1. Different Levels of Selection against F1 Hybrids and Migration Yield Differences in Gametic and Sexual Isolation in D. yakuba and D. santomea after Five and Ten Generations of Experimental Sympatry D. santomea
D. yakuba
Sexual Isolation
Gametic Isolation
Sexual Isolation
Gametic Isolation
2498.52 2516.23 2491.56 2512.1
23420.5 23405.69 23412.57 23229.79
2456.35 2473.52 2457.39 2474.9
22233.85 22243.5 21791.86 21751.46
277.3 242.65 24.87 211.73
22094.98 22066.25 21842.75 21891.54
2225.67 2140.97 106.05 222.08
2561.44 2482.24 10.2 2396.04
Generation 5 Selection + migration + interaction Selection + migration Selection Migration Generation 10 Selection + migration + interaction Selection + migration Selection Migration
To estimate the importance of each effect (migration, hybridization, interaction between migration and hybridization), I formulated four mixed-effect linear models that differed in their fixed effects and in which differences among individuals within each replicate were considered to be random effects. To determine which model was more likely to explain the data, I used the Akaike information criterion (AIC [20]) scores of each model. The model with the lowest AIC for each type of reproductive isolation is highlighted in bold. See also Tables S1–S3.
Can Gene Flow Prevent Speciation by Reinforcement? 2233
between subpopulations evolves while they adapt to different habitats, not as a response to reduce the production of inferior hybrids [34,35]). Mathematical theory has shown that the evolution of reinforcement depends critically on the amount of hybridization: with even small amounts of gene flow, populations fuse rather than becoming more isolated [2,8–10,36,37]. This is because gene flow breaks up associations between alleles that can cause postzygotic isolation, between characters that can cause prezygotic isolation, and between characters that cause pre- and postzygotic isolation [3,9,11]. My experimental results confirm this conclusion: even moderate levels of gene flow, in the form of either hybridization or migration, overwhelmed the effects of reinforcing selection. This pattern was also seen in observations of Timema walking sticks in nature [38,39]. Taken together, the results described here confirm theoretical formulations asserting that reinforcement can be important in completing the speciation process, especially when gene flow is not too high. Other factors that might affect the probability of reinforcement remain to be studied; these include migration from the hybrid zone back to allopatric populations, which will eliminate the pattern of reinforcement by spreading it throughout allopatric populations, heterogeneity in habitat structure across the geographic range of a species, and population structure. But although these aspects of reinforcement remain unexplored, this study confirms that the process is at least a plausible factor in the final stages of speciation. Supplemental Information Supplemental Information includes Supplemental Experimental Procedures, one figure, and four tables and can be found with this article online at doi:10.1016/j.cub.2010.11.036. Acknowledgments I thank J.A. Coyne, T.D. Price, D. Kennedy, M.A.F. Noor, M.A. Sprigge, and N. Bloch for critical discussions and I.A. Butler and J. Gladstone for technical help. This work was funded by National Institutes of Health grant R01GM058260 to Jerry A. Coyne. Received: August 25, 2010 Revised: October 28, 2010 Accepted: November 4, 2010 Published online: December 2, 2010 References 1. Dobzhansky, Th. (1940). Speciation as a stage in evolutionary divergence. Am. Nat. 74, 312–321. 2. Servedio, M.R., and Noor, M.A.F. (2003). The role of reinforcement in speciation: Theory and data. Annu. Rev. Ecol. Evol. Syst. 34, 339–364. 3. Coyne, J.A., and Orr, H.A. (2004). Speciation (Sunderland, MA: Sinauer Associates). 4. Rundle, H.D., and Schluter, D. (1998). Reinforcement of stickleback mate preferences: Sympatry breeds contempt. Evolution 52, 200–208. 5. Saetre, G.P., Moum, T., Bures, S., Kral, M., Adamjan, M., and Moreno, J. (1997). A sexually selected character displacement in flycatchers reinforces premating isolation. Nature 387, 589–592. 6. Jaenike, J., Dyer, K.A., Cornish, C., and Minhas, M.S. (2006). Asymmetrical reinforcement and Wolbachia infection in Drosophila. PLoS Biol. 4, e325. 7. Matute, D.R. (2010). Reinforcement of gametic isolation in Drosophila. PLoS Biol. 8, e1000341. 8. Servedio, M.R. (2000). Reinforcement and the genetics of nonrandom mating. Evolution 54, 21–29. 9. Servedio, M.R., and Kirkpatrick, M. (1997). The effects of gene flow on reinforcement. Evolution 51, 1764–1772.
10. Kirkpatrick, M., and Servedio, M.R. (1999). The reinforcement of mating preferences on an island. Genetics 151, 865–884. 11. Sanderson, N. (1989). Can gene flow prevent reinforcement? Evolution 43, 1223–1235. 12. Liou, L.W., and Price, T.D. (1994). Speciation by reinforcement of premating isolation. Evolution 48, 1451–1459. 13. Caisse, M., and Antonovics, J. (1978). Evolution in closely adjacent plant populations. Heredity 40, 371–384. 14. Dobzhansky, T. (1951). Genetics and the Origin of the Species (New York: Columbia University Press). 15. Spencer, H.G., Mcardle, B.H., and Lambert, D.M. (1986). A theoretical investigation of speciation by reinforcement. Am. Nat. 128, 241–262. 16. Marshall, J.L., Arnold, M.L., and Howard, D.J. (2002). Reinforcement: The road not taken. Trends Ecol. Evol. 17, 558–563. 17. Howard, D.J. (1993). Reinforcement: Origin, dynamics, and fate of an evolutionary hypothesis. In Hybrid Zones and the Evolutionary Process, R.G. Harrison, ed. (New York: Oxford University Press), pp. 46–69. 18. Noor, M.A.F. (1999). Reinforcement and other consequences of sympatry. Heredity 83, 503–508. 19. Price, T.D. (2008). Speciation in Birds (Greenwood Village, CO: Roberts & Co. Publishers). 20. Harper, A.A., and Lambert, D.M. (1983). The population genetics of reinforcing selection. Genetica 62, 15–23. 21. Llopart, A., Lachaise, D., and Coyne, J.A. (2005). Multilocus analysis of introgression between two sympatric sister species of Drosophila: Drosophila yakuba and D. santomea. Genetics 171, 197–210. 22. Llopart, A., Lachaise, D., and Coyne, J.A. (2005). An anomalous hybrid zone in Drosophila. Evolution 59, 2602–2607. 23. Llopart, A., Elwyn, S., Lachaise, D., and Coyne, J.A. (2002). Genetics of a difference in pigmentation between Drosophila yakuba and Drosophila santomea. Evolution 56, 2262–2277. 24. Carbone, M.A., Llopart, A., deAngelis, M., Coyne, J.A., and Mackay, T.F.C. (2005). Quantitative trait loci affecting the difference in pigmentation between Drosophila yakuba and D. santomea. Genetics 171, 211–225. 25. Pinheiro, J., and Bates, D. (2000). Mixed-Effects Models in S and S-PLUS (New York: Springer). 26. Coyne, J.A., Kim, S.Y., Chang, A.S., Lachaise, D., and Elwyn, S. (2002). Sexual isolation between two sibling species with overlapping ranges: Drosophila santomea and Drosophila yakuba. Evolution 56, 2424–2434. 27. Matute, D.R., and Coyne, J.A. (2010). Intrinsic reproductive isolation between two sister species of Drosophila. Evolution 64, 903–920. 28. Koopman, K.F. (1950). Natural selection for reproductive isolation between Drosophila pseudoobscura and Drosophila persimilis. Evolution 4, 135–148. 29. Higgie, M., Chenoweth, S., and Blows, M.W. (2000). Natural selection and the reinforcement of mate recognition. Science 290, 519–521. 30. Thoday, J.M., and Gibson, J.B. (1970). The probability of isolation by disruptive selection. Am. Nat. 104, 219–230. 31. Thoday, J.M., and Gibson, J.B. (1962). Isolation by disruptive selection. Nature 193, 1164–1166. 32. Rice, W.R., and Salt, G.W. (1988). Speciation via disruptive selection on habitat preference: Experimental evidence. Am. Nat. 131, 911–917. 33. Rice, W.R., and Salt, G.W. (1990). The evolution of reproductive isolation as a correlated character under sympatric conditions: Experimental evidence. Evolution 44, 1140–1152. 34. Bush, G.L. (1994). Sympatric speciation in animals: New wine in old bottles. Trends Ecol. Evol. 9, 285–288. 35. Diehl, S.R., and Bush, G.L. (1989). The role of habitat preference in adaptation and speciation. In Speciation and Its Consequences, D. Otte and J. Endler, eds. (Sunderland, MA: Sinauer), pp. 345–365. 36. Kelly, J.K., and Noor, M.A. (1996). Speciation by reinforcement: A model derived from studies of Drosophila. Genetics 143, 1485–1497. 37. Servedio, M.R. (2004). The evolution of premating isolation: Local adaptation and natural and sexual selection against hybrids. Evolution 58, 913–924. 38. Nosil, P., Crespi, B.J., and Sandoval, C.P. (2003). Reproductive isolation driven by the combined effects of ecological adaptation and reinforcement. Proc. Biol. Sci. 270, 1911–1918. 39. Nosil, P. (2005). The role of selection and gene flow in the evolution of sexual isolation in Timema walking-sticks and other Orthopteroids. J. Orthoptera Res. 14, 247–253.