Rapid Evolution of Reproductive Isolation in the Wild: Evidence from Introduced Salmon

See allHide authors and affiliations

Science  20 Oct 2000:
Vol. 290, Issue 5491, pp. 516-518
DOI: 10.1126/science.290.5491.516


Colonization of new environments should promote rapid speciation as a by-product of adaptation to divergent selective regimes. Although this process of ecological speciation is known to have occurred over millennia or centuries, nothing is known about how quickly reproductive isolation actually evolves when new environments are first colonized. Using DNA microsatellites, population-specific natural tags, and phenotypic variation, we tested for reproductive isolation between two adjacent salmon populations of a common ancestry that colonized divergent reproductive environments (a river and a lake beach). We found evidence for the evolution of reproductive isolation after fewer than 13 generations.

Ecological speciation occurs when organisms exposed to divergent selective regimes evolve reproductive isolation as a by-product of adaptation (1–3). Mechanisms contributing to ecological speciation include mate choice based on traits under divergent selection (4, 5), hybrid or backcross inferiority (2), and reinforcement of assortative mating when hybrids are inferior (6,7). Ecological speciation appears to be relevant in allopatry and sympatry and has been supported by theoretical models, laboratory experiments, and studies of natural systems (1–9). Here we focus on an unknown aspect of ecological speciation: How quickly can reproductive isolation evolve?

Rapid evolution of adaptive traits often occurs in populations exposed to divergent ecological environments (10,11). Although this implies that reproductive isolation may also evolve rapidly, the best examples of ecological speciation are seen in groups that began diverging thousands of years ago (12, 13). Unfortunately, inferring evolutionary rates on the basis of long-standing groups is questionable, because averaging disparate rates across time will obscure biologically important short-term evolution (11). Thus, reproductive isolation might evolve in only a few generations, or it may require a long and gradual accumulation of isolating mechanisms. Some insects that colonized new host plants 100 to 200 years ago have evolved ecologically mediated reproductive isolation (14,15). We ask whether reproductive isolation can evolve even faster by testing for evidence of intrinsic barriers to gene flow between two populations of sockeye salmon (Oncorhynchus nerka) derived from a common source less than 13 generations previously.

Sockeye salmon form distinct reproductive “ecotypes,” with adults breeding in streams or along lake beaches (16,17). After continental glaciers receded about 10,000 years ago, sockeye salmon colonized hundreds of new lake systems, many of which now contain reproductively isolated populations that are adapted to beaches and streams (16). Postglacial reproductive isolation was presumably achieved through natal homing, divergent sexual ornamentation, and ecological selection against hybrids. Parallel evolution of these ecotypes within many different lake systems provides a robust interpretive framework.

Introductions of salmon to new locations have provided excellent opportunities to study rates and patterns of evolution (18–21). Sockeye salmon were introduced into Lake Washington, Washington, from Baker Lake, Washington, between 1937 and 1945. These introductions gave rise to a large (currently 100,000 to 350,000 breeders) self-sustaining population in the major tributary to Lake Washington (Cedar River). In 1957, a new population was first documented breeding along a Lake Washington beach (Pleasure Point) about 7 km north of Cedar River. The beach site was apparently colonized by fish from the river or, if not, both populations are at least of the same introduced lineage [inferred from historical records and allozyme variation (18–20)]. The two populations have been diverging for a maximum of 56 years (1937 to 1992), which is equivalent to approximately 13 generations (20), and the beach population is less abundant by about two orders of magnitude (22). As a starting point for divergence, we adopted 13 generations (from 1937) rather than 8 generations (from 1957) because the former is unambiguous and conservative.

We examined reproductive isolation between the beach (Pleasure Point) and river (Cedar River) populations on the basis of collections of breeding adults (23). We used population-specific natural marks to identify river fish that were immigrants to the beach, DNA microsatellite loci to estimate genetic differentiation between river and beach residents, population-genetic models to infer the relative reproductive success of immigrants, and adult life history and morphology to consider the ecological basis for isolation.

We quantified immigration of river-born adults to the beach using natural marks that are produced on otoliths (calcareous ear stones) while embryos incubate in the gravel. Because the incubation environment is isothermal at the beach but variable in the river (22), we could examine the otoliths of adults to determine whether each had been born (incubated) at the beach or river (24). This analysis, conducted blind with respect to collection location, microsatellite variation, and phenotypic traits, revealed that most breeding adults collected from the river had indeed incubated under a variable thermal regime (34 of 35 in 1992 and 30 of 38 in 1993) but that many adults collected from the beach had also incubated under a variable thermal regime (14 of 32 in 1992 and 12 of 34 in 1993). Thus, approximately 39% of adults breeding at the beach (48% of females and 34% of males) were immigrants from the river (22).

This estimated immigration rate of river fish into the beach population is so high that unless reproductive isolating mechanisms had evolved, the two populations could not have diverged at neutral genetic loci. We used allelic variation at six microsatellite loci (25) to quantify genetic differences between three groups categorized by otolith patterns and breeding location: beach residents (born and breeding at the beach), river residents (born and breeding in the river), and beach immigrants (born in the river but breeding at the beach). River residents and beach immigrants showed no evidence of genetic divergence (Table 1), which is consistent with the expectation that immigrants to the beach were from the river. In contrast, beach residents were genetically distinct from river residents and from beach immigrants (Table 1). This pattern of genetic differentiation could only have arisen if beach immigrants have reduced reproductive success relative to beach residents.

Table 1

Genetic differentiation at six microsatellite loci (25) between beach residents (N = 22), river residents (N = 35), and beach immigrants (N = 12). Nei's unbiased genetic distance (D) and F ST (bootstrapped 95% confidence intervals are shown in parentheses) were calculated with TFPGA (35). The significance of genotypic differentiation (GD) was determined with GENEPOP (36). Observed differentiation cannot be attributed to linkage of any one locus to a locus under divergent selection, because even after deletion of the locus that best differentiated river residents from beach residents (Ssa85, F ST = 0.054), divergence was still substantial.

View this table:

We considered the extent of reproductive isolation by comparing the proportion of immigrant breeding adults in the beach population (adult migration, determined from otoliths) to the proportion of immigrant genes (gene flow, determined from microsatellites). If gene flow were less than adult migration, we would have evidence for the evolution of reproductive isolation. The standard approach to estimating gene flow from genetic data requires assumptions that most natural populations violate (26). We therefore estimated gene flow using recursion equations that avoided these assumptions, allowing for two populations of different sizes, asymmetric gene flow, and nonequilibrium conditions (27). Gene flow from the river to the beach was less than adult migration (39%), as long as the beach effective population size was not exceptionally low (N e > 8; Fig. 1). Breeding population sizes range from 100 to 8180 at the beach (20), suggesting thatN e ≫ 8. Thus, the reproductive success of river fish breeding at the beach must be lower than that of beach residents, despite their recent common ancestry.

Figure 1

Estimated rates of gene flow from the river to the beach. Curves were determined with recursion equations (27) to estimate gene flow (m 1) that would lead to the observed F ST (0.025) after 13 generations. We assumed that beach colonizers were representative of the river population (that is, founder effects were minimal), that all gene flow was from the river to the beach (m 2 = 0), and that N e = 10,000 in the river. Recursions were started with an IBD of 1/10,000. Curves represent a range of possible N e/N ratios (0.1, 0.2, and 0.4).

We considered potential isolating mechanisms by examining two adult traits that are subject to divergent selection between rivers and beaches. Male body depth is sexually selected (28) and reaches extremes in beach populations where it is unopposed by predation, water flow, or water depth (29). In rivers, males have shallower bodies (29), presumably owing to selection for increased swimming efficiency. Female body size differs between beaches and rivers because large females dig deeper nests, thereby protecting their eggs from disturbance during flooding (30). Flooding is absent from beaches, and females are correspondingly smaller (17, 29). In our study, beach males had deeper bodies (standardized to average body length) and beach females were shorter, with beach immigrants somewhat intermediate for both traits (Fig. 2). These results suggest that beach immigrants are not as well suited for the beach environment as are beach residents, perhaps contributing to reduced mating success or offspring survival. Many other traits are subject to divergent selection between beaches and rivers (20), and beach immigrants are probably also compromised for those traits.

Figure 2

Differences in (A) standardized male body depth and (B) female body length (37) between beach residents (BR), beach immigrants (BI), and river residents (RR). Boxes contain 50% of the data and bars contain the remainder; horizontal lines indicate medians, arrows indicate means, and the circle indicates an outlier. On the basis of Tukey tests, river residents and beach immigrants had similar female lengths (P = 0.365) and male body depths (P = 0.076), river residents and beach residents had different lengths (P = 0.003) and body depths (P < 0.001), and beach residents and beach immigrants had similar lengths (P = 0.256) and body depths (P = 0.289). The morphological intermediacy of beach immigrants could arise because of phenotypic plasticity (if swimming in rivers reduces body depth), morphology-influenced site selection (if smaller river females and deeper bodied river males are more likely to breed at the beach), or site selection by hybrids (if hybrids were produced in the river and then bred at the beach).

Significant reproductive isolation (albeit partial) after fewer than 13 generations implies that much of the isolation observed in ecological speciation can arise soon after initial divergence. Our results may seem exceptional but are clearly biologically possible, as evidenced by laboratory studies in which reproductive isolation often evolves over similar time frames (8). Our study was based on indirect methods (patterns of genetic variation), which measure total isolation (postzygotic and prezygotic). Direct tests of reproductive isolation, such as mate preference, would be complementary because they quantify the prezygotic contribution to isolation (9). Experimental demonstrations of speciation in the wild have been considered intractable because isolation is assumed to require a long period. Our findings suggest that when organisms colonize different environments, experimental studies of speciation may prove feasible.

  • * To whom correspondence should be addressed. E-mail: ahendry{at}


View Abstract

Navigate This Article