Ecological speciation occurs when reproductive isolation arises due to adaptive divergence between populations inhabiting ecologically different environments (Schluter 2000; Nosil 2012). This process now has considerable empirical support from a wide range of taxa (Schluter 2000; Rundle and Nosil 2005; Funk et al. 2006; Nosil 2012; Shafer and Wolf 2013) – yet it is clearly not ubiquitous. For instance, a growing number of studies seeking evidence of reproductive barriers between populations in different environments have failed to find them or have found that they are very weak (reviews: Hendry 2009; Nosil et al. 2009). At the same time, a growing number of studies report speciation in the apparent absence of ecological differences (Rundell and Price 2009; Svensson 2012). These variable results highlight the value of considering the relative contributions of multiple reproductive barriers in taxa that vary in their progress toward speciation – whether ecological or otherwise. Such analyses should prove useful in attempting to delineate the conditions that do and do not promote ecological speciation – and the combinations of reproductive barriers that are most important.
A reproductive barrier that should be particularly powerful and ubiquitous is selection against migrants, which occurs when individuals adapted to one environment (or habitat or resource) immigrate to (or start to use) another environment (Hendry 2004; Nosil et al. 2005). In such cases, local adaptation is expected to reduce the fitness of immigrants through either lower survival (“immigrant inviability”: Nosil et al. 2005) or lower fecundity or mating success (“immigrant infecundity”: Smith and Benkman 2007). Selection against migrants is expected to be important and common for several reasons. First, it often acts early in the life cycle (e.g., cross-type mating would usually occur later) and so is expected to capture more of the total isolation (Nosil et al. 2005). Second, it acts as an “automatic magic trait” in that divergent selection acts directly on the traits that also influence reproductive isolation (Servedio et al. 2011). Third, the many reciprocal transplant experiments that have been conducted across diverse organisms frequently find lower survival or fecundity in individuals moved between environments (Schluter 2000; Nosil et al. 2005; Hereford 2009).
Two contrasting predictions can be made in the context of selection against migrants. First, populations living in very different environments and showing strong divergence in adaptive traits should show local superiority (e.g., survival and growth should be higher for local than for immigrant individuals in a given environment; Kawecki and Ebert 2004; Nosil et al. 2005; Hereford 2009; Blanquart et al. 2013). Second, populations showing low divergence in adaptive traits should show small (if any) differences in fitness between local and immigrant individuals in a given environment. The second situation might occur if the environments are not very divergent or if adaptive divergence is constrained for some reason, such as high gene flow (Räsänen and Hendry 2008). The latter context is particularly interesting, because it presents a case where ecological speciation might be predicted (divergent selection is strong) but cannot be achieved. We test these two predictions in lake/stream threespine stickleback (Gasterosteus aculeatus; Fig. 1) from the Misty Lake watershed in British Columbia, Canada.
Threespine stickleback are a good model for studying progress toward ecological speciation because they show dramatic adaptive divergence between populations in different environments (review: Bell and Foster 1994; McKinnon and Rundle 2002), but highly variable progress toward ecological speciation (Berner et al. 2009; Hendry et al. 2009). That is, in some population contrasts, reproductive barriers can be very strong, whereas in others, similar barriers can be weak or absent (e.g., Jones et al. 2008; Hendry et al. 2009; Raeymaekers et al. 2010; Räsänen et al. 2012). This variation provides excellent opportunities to uncover the factors that promote and constrain progress toward ecological speciation (Hendry et al. 2009).
Our investigation focuses on parapatric stickleback populations in lake versus stream environments. Suitable properties of this system include (1) independent evolutionary origins of lake/stream pairs in many different watersheds (Hendry and Taylor 2004; Berner et al. 2009), (2) strongly divergent foraging environments that generate strong divergent selection (Lavin and McPhail 1993; Berner et al. 2008; Kaeuffer et al. 2012), (3) high gene flow that can sometimes constrain lake/stream divergence (Hendry et al. 2002; Hendry and Taylor 2004; Moore et al. 2007), and (4) highly variable progress toward ecological speciation (Berner et al. 2009, 2010; Roesti et al. 2012). Ecological reproductive barriers are likely: habitat preferences can be important (Bolnick et al. 2009), reproductive timing differences are likely (A. Hendry & D. Bolnick, pers. obs.), and Lake fish have difficulty swimming upstream (Hendry et al. 2002). In contrast, strong genetic incompatibilities are unlikely, given that among ecotype crosses can be successfully conducted and hybrids from these crosses are viable in the laboratory (Lavin and McPhail 1993; Raeymaekers et al. 2010; Berner et al. 2011). Overall, many of these reproductive barriers appear rather weak in at least some places, and assortative mating has not been found (Raeymaekers et al. 2010; Räsänen et al. 2012).
The present study was conducted in the Misty watershed, where two stream populations (Inlet and Outlet) are found in parapatry with the Lake population. The Inlet and Lake populations show very low gene flow (as inferred from neutral markers) and strong genetically based adaptive divergence in a broad suite of phenotypic traits (Hendry et al. 2002; Delcourt et al. 2008; Sharpe et al. 2008; Raeymaekers et al. 2009; Berner et al. 2011; Hendry et al. 2011; Kaeuffer et al. 2012; Baker et al. 2013). Despite this genetic and phenotypic divergence, strong symmetric reproductive barriers have not yet been found in the Misty system, leading Räsänen et al. (2012) to pose the “conundrum of missing reproductive isolation.” The Outlet and Lake populations also experience divergent selection but, in contrast to the Inlet and Lake populations, do not show much adaptive divergence owing to very high gene flow (Hendry et al. 2002; Moore et al. 2007; Delcourt et al. 2008; Sharpe et al. 2008; Berner et al. 2008, 2009; Roesti et al. 2012).
We here use a reciprocal transplant experiment in the wild with individually marked fish placed in enclosures to test for selection against migrants between these two Misty lake/stream population pairs. First, we ask whether trait differences predict performance differences: that is, Lake and Inlet fish should perform differently (measured as survival and mass change), whereas Lake and Outlet fish should perform similarly. Second, we ask whether selection against migrants is evident: that is, Lake fish should perform better than Inlet fish in the lake whereas Inlet fish should perform better than Lake fish in the inlet (no such differences should be evident in Lake/Outlet contrasts). Importantly, our study used wild-caught fish. Although this means that any performance differences cannot be conclusively ascribed to genetic differences, genetic differences do seem likely given the documented genetic basis for adaptive trait divergence in this system (Lavin and McPhail 1993; Hendry et al. 2002; Delcourt et al. 2008; Raeymaekers et al. 2009, 2010; Berner et al. 2011; Hendry et al. 2011; Baker et al. 2013). Moreover, the use of wild-caught fish in reciprocal transplant experiments is usually the starting point for such studies (e.g., 66.7% of the studies in the meta-analysis of Hereford 2009). Finally, and most importantly, the use of wild-caught fish (as opposed to common garden fish) encompasses the effects of the whole phenotype and, hence, is most directly relevant to selection against migrants in nature.