The experiment consisted of a fully reciprocal transplant experiment conducted in enclosures in the wild. In each of the three environments (lake, outlet, and inlet), experimental sites were chosen based on prior knowledge of appropriate stickleback habitat and their suitability for enclosures (water depths of 40–60 cm and low flow rates). In each of the environments, a single large enclosure (approximate sizes dictated by available space: inlet: 9.9 m2; lake: 13.5 m2; outlet: 7.8 m2) consisting of white nylon mesh (Delta, 10 mm diameter; Nylonnet Co., Memphis, TN) was erected. These enclosures and their construction were similar to – although larger and with more stickleback – than those in several previous studies (e.g., Hatfield and Schluter 1999; Rundle 2002; Hendry et al. 2002). Ideally, we would have used multiple (replicate) enclosures per environmental type. However, as enclosure number trades-off with enclosure size, we elected to use a single large enclosure per habitat (as opposed to many small as used by Hendry et al. 2002), to provide an environment that would more closely mimic a natural migration event (i.e., bracketing the local habitat heterogeneity, and allowing for habitat choice and competitive interactions within groups of stickleback).
The bottom edge of the mesh of the enclosures was buried in gravel, and the upper end was suspended well above the surface with posts hammered into the substrate. For several days, unbaited minnow traps were used to remove any stickleback from the constructed enclosures prior to experimental setup. Disturbance during enclosure setup, combined with the minnow traps, also ensured that fish predators were either absent or very rare in the enclosures (none were observed). However, aerial predation by piscivorous birds (terns or kingfishers) could contribute to performance effects.
Stickleback for the experiment were collected with unbaited minnow traps over several consecutive days from each of the three populations (Inlet, Lake, and Outlet). These fish were collected from a variety of locations near the experimental sites, but did not include the fish removed from the enclosures. We retained adult-size fish (min. ca. 50 mm total length) for the experiment – because they (as opposed to juveniles) are more likely to survive the tagging procedure and they could not escape from the enclosures. Adults are known to disperse in nature, including between lake and stream habitats (Bolnick et al. 2009; Moore and Hendry 2009), but the extent of juvenile dispersal is not known. Captured fish were transferred to 100L aquaria in the laboratory and held for a few days to conduct the pre-experimental procedures (see below) and to ensure their health prior to release into the enclosures.
Seventy-five fish from each population were haphazardly assigned to each of three experimental groups intended for the different enclosures. Prior to release into the enclosures, the fish were briefly anaesthetized with MS222, photographed with a digital camera (Nikon coolpix; Nikon Inc., Tokyo, Japan) on their left side, weighed on an electronic balance (to the nearest 0.01 g), and individually marked with coded wire tags (CWTs; North West Marinetechnology, Inc., Shaw Island, WA). The tags were inserted by injection into the muscle tissue on the left side of the body in front of the first dorsal spine. Although CWTs have not previously been used for stickleback, they are routinely used in a broad range of other fishes. Survival upon tagging was initially tested on stickleback in the laboratory, and no evidence was found for tagging-induced mortality. To ensure survival after handling, the fish were maintained overnight in holding tanks before being released to the enclosures.
Twenty-five fish of each of the three types were placed in each of the enclosures (i.e., 75 fish per enclosure, total N = 225) late May and left undisturbed for 21 days. This length of time has been used in several previous enclosure studies with stickleback (Hatfield and Schluter 1999; Hendry et al. 2002; Bolnick et al. 2010), where it has proven sufficient to reveal phenotype-specific performance differences. After this period, unbaited minnow traps were used intensively over several days to recapture the remaining fish (the tannic water in the Misty system prevented individual targeting for capture). The rarity of captures at the end of this period suggests that all or most of the surviving fish were recaptured. These fish were weighed (to nearest 0.01 g), euthanized with an overdose of MS222, photographed, and preserved in 95% ethanol. In the laboratory, they were dissected to determine sex and maturity status (male, female, and immature).
Of the released fish, 186 (82.7%) survived the experiment (i.e., were recaptured). Of these, 103 (55.4%) were males, 66 (35%) females, and 17 were immature (9.1%). The experimental period (late May) coincides with the breeding season of both Misty lake and stream stickleback and by the end of the experiment, 11 of 28 (39.2%) Inlet females (seven in the inlet, three in the lake, and two in the outlet enclosure), three of 12 (25%) Lake females (of which two in inlet and one in lake enclosure) were gravid. None of the 25 Outlet females was gravid. One Lake fish and one Outlet fish were clearly parasitized by Schistocephalus.
Our two response variables (fitness or “performance” measures) were survival and mass change (individual mass of a survivor at the end of the experiment minus its mass at the beginning of the experiment). Log(mass change) was analyzed with analyses of variance (ANOVA) and analyses of covariance (ANCOVA) in Proc Mixed in SAS 9.3 (SAS Inc., Carry, NC) (details below). Fixed factors were sex (male, female, or immature), ecotype (Lake, Inlet, and Outlet), environment (lake, inlet, and outlet), and the ecotype × environment interaction. To avoid bias owing to gravid state or parasitism, analyses of log(mass change) were conducted excluding the gravid (N = 14) and parasitized (N = 2) individuals.
Survival was analyzed with generalized linear models with a logit link function and binomial error structure (Proc Genmod in SAS 9.3). Fixed factors were ecotype (Lake, Inlet, and Outlet), environment (lake, inlet, and outlet), and ecotype × environment interaction. Survival analyses did not include sex, because fish could not always be reliably sexed before transfer to enclosures.
Log(initial mass) was used as a covariate in analyses of both survival and mass change to assess and control for effects of initial size on performance. We first conducted a full model including all three ecotypes and environments. Because our specific hypotheses relate to the Lake versus Inlet and the Lake versus Outlet contrasts, separate analyses were conducted within each ecotype contrast: one analysis considered Lake and Inlet fish in lake and inlet enclosures, and the other considered Lake and Outlet fish in lake and outlet enclosures. (Analyses of the Outlet-Inlet contrast are provided in the Supporting Information).
In these analyses, a significant main effect of ecotype would indicate that ecotypes differ in overall performance (independent of the environment), a significant main effect of environment would indicate that environments differ in their effects on stickleback performance (independent of ecotype), and a significant ecotype × environment interaction would indicate that performance differences between the ecotypes depended on the specific testing environment. This last effect is the one most relevant for testing environment-dependent selection against migrants. To quantify the magnitude of selection against migrants, differences in survival between local and immigrant ecotypes in a given environment were calculated as (WLOC−WMIG)/WSITE, where WLOC and WMIG are survival of local and immigrant individuals and WSITE is average survival across both ecotypes at a given site. These estimates are equivalent to those used by Hereford (2009) in his meta-analysis of local adaptation in reciprocal transplant experiments. We only calculated differences for survival as survival is more directly linked to fitness.