Divergent selection and adaptive divergence can increase phenotypic diversification amongst populations and lineages. Yet adaptive divergence between different environments, habitats or niches does not occur in all lineages. For example, the colonization of freshwater environments by ancestral marine species has triggered adaptive radiation and phenotypic diversification in some taxa but not in others. Studying closely related lineages differing in their ability to diversify is an excellent means of understanding the factors promoting and constraining adaptive evolution. A well-known example of the evolution of increased phenotypic diversification following freshwater colonization is the three-spined stickleback. Two closely related stickleback lineages, the Pacific Ocean and the Japan Sea occur in Japan. However, Japanese freshwater stickleback populations are derived from the Pacific Ocean lineage only, suggesting the Japan Sea lineage is unable to colonize freshwater. Using stable isotope data and trophic morphology, we first show higher rates of phenotypic and ecological diversification between marine and freshwater populations within the Pacific Ocean lineage, confirming adaptive divergence has occurred between the two lineages and within the Pacific Ocean lineage but not in the Japan Sea lineage. We further identified consistent divergence in diet and foraging behaviour between marine forms from each lineage, confirming Pacific Ocean marine sticklebacks, from which all Japanese freshwater populations are derived, are better adapted to freshwater environments than Japan Sea sticklebacks. We suggest adaptive divergence between ancestral marine populations may have played a role in constraining phenotypic diversification and adaptive evolution in Japanese sticklebacks.
Citation: Ravinet M, Takeuchi N, Kume M, Mori S, Kitano J (2014) Comparative Analysis of Japanese Three-Spined Stickleback Clades Reveals the Pacific Ocean Lineage Has Adapted to Freshwater Environments while the Japan Sea Has Not. PLoS ONE9(12): e112404. https://doi.org/10.1371/journal.pone.0112404
Editor: John A. Craft, Glasgow Caledonian University, United Kingdom
Received: April 7, 2014; Accepted: October 6, 2014; Published: December 2, 2014
Copyright: © 2014 Ravinet et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The authors confirm that all data underlying the findings are fully available without restriction. Data are available from DRYAD using the following identifier: doi:10.5061/dryad.s8f74.
Funding: This research was supported by the Japan Science and Technology Agency PRESTO program, Center for the Promotion of Integrated Sciences (CPIS) of Sokendai, and Grant-in-Aid for Scientific Research on Innovative Areas (23113007 and 23113001) from the Ministry of Education, Science, Sports, and Culture to JK. MR was kindly funded by the Japanese Society for the Promotion of Science on a Short-term Postdoctoral Fellowship and the NIG Collaborative Research Program (2013-A33 and A34). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Colonisation of new environments can lead to adaptive divergence, the evolution of reproductive isolation and progression towards speciation within evolutionary lineages –. Divergent selection between populations in different habitats that leads to adaptive divergence occurs when ecological opportunity arises – i.e. previously unused niche space becomes available. Yet adaptive divergence and diversification is not ubiquitous; for example Darwin's finches have evolved remarkable phenotypic diversity in foraging behaviour and morphology, while Galápagos' mockingbirds have not , . Similarly, the extent of adaptive divergence is variable in three-spined stickleback lineages; with strong reproductive isolation and phenotypic divergence in some regions but not others , .
Many potential factors may constrain phenotypic diversification and adaptive divergence . Intrinsic factors such as differences in rate of dispersion and adaption to novel niche space can shape both colonisation history and rates of evolutionary diversification. The genetic basis of adaptation is also important; adaptation from standing genetic variation is more rapid than that from de novo mutations . Extrinsic factors such as competition with earlier colonizers may also prevent lineages from establishing , , . Alternatively, constraint may simply arise due to historical contingency; e.g. the chance formation of novel island and lake environments and in turn, chance colonisation , , . Most likely, differences in the extent of adaptive divergence arise because of a combination of these factors and considering interactions between them is a more informative approach for understanding constraint , , . For example, when resources in the new environment are similar to those in the ancestral habitat, probability of survival and establishment of a species upon colonization is greater , . Comparative analyses of genetic and phenotypic variation between lineages undergoing different rates of adaptive divergence and phenotypic diversification may help us better understand these constraints.
The formation of freshwater lakes and rivers by glacial and interglacial cycles has repeatedly created novel unoccupied niche space during the late Quaternary . Several marine taxa, such as crustaceans, molluscs, annelids, and teleosts have colonized these newly formed freshwater environments and undergone adaptive divergence, phenotypic diversification and in some cases parallel speciation ,  The three-spined stickleback (Gasterosteus aculeatus) offer an excellent example with repeated colonisation of freshwater environments throughout the Pleistocene resulting in parallel phenotypic adaptation and genomic divergence , , . Adaptive divergence between stickleback species pairs is widespread throughout the distribution of the species ; however, the nature and extent this divergence can vary across spatial scales and between evolutionary lineages , , , . For example, stronger phenotypic differences between freshwater ecotypes in Canada relative to Europe has led some authors to suggest genomic constraints arising from different evolutionary histories may be influencing divergence , .
Despite freshwater colonization being characteristic of three-spined sticklebacks , , , the Japan Sea three-spined stickleback lineage contains no freshwater populations at all , . Freshwater populations consist only of individuals from the Pacific Ocean lineage and demonstrate considerable phenotypic diversity in gill raker number, body size and armour traits, much like freshwater populations occurring elsewhere in the stickleback distribution , . Japan Sea stickleback populations in contrast show little phenotypic diversity throughout their distribution and are characterised by a smaller body size and smaller lateral armour plates in comparison to similarly anadromous Pacific Ocean fish , . In Japan and Eastern Asia, the Japan Sea and Pacific Ocean stickleback clades co-exist , , . These two divergent lineages likely experienced a period of allopatric divergence during the geographical isolation of the Sea of Japan due to sea-level change 1.5–2 million years ago . Phylogenetic analyses show that all Japanese freshwater populations analysed thus far are repeatedly derived from the Pacific Ocean lineage only , . The apparent lack of freshwater colonisation within the Japan Sea lineage is intriguing as all other known evolutionary lineages of three-spined stickleback are able to adapt to freshwater –. This is also surprising because there are many freshwater lakes and rivers surrounding the Sea of Japan. Comparisons between the Japan Sea and Pacific Ocean lineages may therefore shed light on the factors that have caused differences in the ability to colonise empty niche space and evolve adaptive divergence in foraging traits in this stickleback system.
Both lineages have extant marine forms that breed in brackish waters and rivers, the Pacific Ocean and Japan Sea anadromous forms (PA and JA hereafter). These lineages are reproductively isolated from one another due to hybrid male sterility , . Pacific Ocean freshwater populations (PF herein) also occur. Some spatial isolation during spawning also occurs as PA fish migrate to freshwater while JA remain at higher salinities , . The two anadromous forms also differ in size, shape, diet and trophic traits such as gill raker number , . However, detailed analyses of habitat, resource use and niche width have not been carried out. Furthermore, it is not known whether the magnitude of habitat divergence differs between sympatric PA and JA populations and allopatric populations of both forms (i.e. where only a single form occurs).
Focusing on the Japan Sea and Pacific Ocean stickleback lineages we first asked whether phenotypic and ecological diversification rates are higher in the Pacific Ocean lineage compared to the Japan Sea lineage. Using phylogenetic comparative methods we quantified diversification in trophic traits, including gill raker number and resource use in both anadromous and freshwater populations. We expected that divergence in trophic ecology would be consistent with foraging trait divergence between lineages. This first part of our study aimed to explicitly test the hypothesis that significant adaptive divergence in freshwater foraging traits – i.e. gill raker morphology and resource use - has occurred between marine and freshwater populations in the Pacific Ocean lineage. In contrast, we expected the Japan Sea lineage to lack phenotypic diversification because of a lack of adaptive divergence and our results strongly indicate this is the case.
Second, we asked how anadromous forms of the two lineages differ in trophic morphology, ecology and feeding behaviour, specifically focusing on sympatric (i.e. both forms co-occurring) and allopatric (i.e. only one form present) populations of JA and PA fish. Forms from both lineages migrate to coastal regions to spawn and previous research has suggested PA migrate further upstream than JA . We hypothesised that ancestral and on-going adaptive divergence within the Pacific Ocean lineage has predisposed PA populations to exploit freshwater resources more frequently and efficiently than their JA counterparts. We also hypothesised that interactions between JA and PA forms in sympatry may further increase divergence in resource use. Our results confirm that PA populations do indeed exploit greater freshwater resources than JA sticklebacks in both allopatry and sympatry. However, we also found that JA populations exploited a greater proportion of freshwater resources when the PA was absent. We suggest that if adaptive divergence between marine and freshwater habitats occurred ancestrally in the Pacific Ocean lineage but not the Japan Sea lineage, this may have played a role in constraining phenotypic diversification.
Animal use protocols were approved by the Institutional Animal Care and Use Committee of the National Institute of Genetics (23-15). Fish sampling in Hokkaido was conducted under a permit issued by Hokkaido Prefecture.
Anadromous (PA, JA) and freshwater stickleback populations (PF) were sampled across Northern Japan using minnow traps and seine nets between June 2006 and May 2012 (Fig. 1a; Tables S1 and S22 in File S1). JA and PA fish were sampled from sites where both forms were present (sympatric) and also where only one form occurred (allopatric). In the Bekanbeushi (Akkeshi) system, Eastern Hokkaido (Fig. 1b), fish were collected clinally at three sites with increasing distance from the lake. Fish were immediately euthanized upon capture using MS-222, preserved in ethanol and classified as JA, PA or PF using morphologically identifying features .
The upper panels are based on the Nei's D tree, while the lower panels are based on the δμ2 tree.
Benthic macroinvertebrates were collected clinally from the lake in the Bekanbeushi and Shiomi river systems and marine invertebrates from Akkeshi Bay (Fig. 1b). Samples included putative prey items for stickleback as well as molluscs and bivalves to provide a representation of the baseline isotopic values of benthic and pelagic food webs .
Trophic trait variation and stable isotope analysis
Dietary inference is an important means of determining habitat and resource use between divergent populations , , , . While stomach content analysis provides a high-resolution indication of prey items  this method can only provide a temporal snapshot of resource use , . In contrast, stable isotope analysis provides a long-term signal of diet and habitat use –. Using both stable isotope and stomach content analysis ensures characterisation of long-term diet and high resolution identification of prey items , .
The long-term resource use signal from stable isotope analysis is informative for anadromous species making large distance migrations across salinity gradients. As ectotherms, δ13C and δ15N values from fish muscle tissue typically reflect dietary assimilation during spring/summer growth as during winter nutrients are used to sustain basal metabolic processes . For sticklebacks sampled in May-June shortly following the peak migration times for both forms , δ13C and δ15N values may indicate habitat divergence throughout life history. Three-spined sticklebacks from each of the sample sites were processed for stable isotope analysis (n = 314). Dorsal muscle was dissected from each fish, dried for 48 hours at 60°C, ground and weighed. Benthic macroinvertebrates (n = 66) were similarly processed. Samples were analysed for δ13C, δ15N, % C and % N on a Carlo Erba Elemental Analyser and a Thermo Finnigan Delta Plus XL mass spectrometer at the Duke Environmental Isotope Analysis laboratory (DEVIL) at Duke University, North Carolina, USA and at UC Santa Cruz Stable Isotope Laboratory, California, USA. Prior to analysis, fish muscle δ13C values were lipid-normalised .
Carbon and nitrogen isotopes can also be used to quantify the isotopic niche, a proxy for ecological niche , . Niche quantification using Bayesian estimation of standard ellipses provides an accurate representation of niche width while quantifying sampling error . We used SIBER (Stable Isotope Bayesian Ellipses in R) to estimate the corrected standard ellipse area (SEAC) and Bayesian estimated standard ellipse area (SEAB).
We additionally counted the number of gill rakers, a functionally important trait closely correlated with trophic ecology , . Gill rakers were counted on the left first gill arch under a dissecting microscope (Tables S1 and S22 in File S1).
Comparison of evolutionary rates
Two phylogenetic trees were estimated from microsatellite data in anadromous and freshwater populations from the Japan Sea and Pacific Ocean lineages using both Nei's D and δμ2 , . Briefly, fish (n = 249) were genotyped using 10 microsatellite markers (Stn170, Stn233, Stn64, Stn159, Stn46, Stn90, Stn120, Stn278, Stn332 and Stn384) located on different three-spined stickleback linkage groups not linked to sex , . While coalescent methods for estimation of population history from microsatellite markers are available , they do not integrate phylogenies across multiple markers and are not suitable for large numbers of populations. We used two metrics of genetic distance in order to account for the shortcomings of each metric; Nei's D performs best when divergence time is relatively recent whereas δμ2 performs better when divergence is older , . Pairwise matrices of genetic distances and UPGMA trees were estimated and bootstrapped 200 times using Populations . Phylogenetic trees were then pruned using the R package ape so that only populations with ecological data remained (n = 19 for stable isotope data, n = 24 for gill rakers) .
To test for lineage specific rates of diversification for gill raker number and niche use (i.e. δ13C and δ15N) we first used the method developed by O'Meara et al. (2006). This allows phenotypic traits to evolve along a phylogeny under Brownian motion (BM) and estimates the likelihood of two models; a single rate only and separate rates (σ2) for the Japan Sea (JS) and Pacific Ocean (PO) lineages. Lineage was mapped onto each tree and nested Brownian motion models were fitted using the brownie.lite function in phytools , .
While widely applied, BM is a neutral model and may not be applicable when examining adaptive traits as it does not account for selection . The Ornstein-Uhlenbeck (OU) model is an extension of BM including the parameters è and α; the optimum trait mean and the strength of selection against deviations from the optimum , . Using Butler & King's (2004) method we tested OU models with single optimal trait value for both lineages (OU1), lineage specific optimal trait values (OU2) and lineage specific values with third optimal value for Pacific Ocean freshwater populations (OU3). If adaptive divergence has occurred between lineages, multiple optimum value models would be supported. As before, lineage was mapped onto the tree and nested OU models were fitted using the hansen function in the ouch R package .
Phylogenetic model choice is not straightforward as uninformative data may result in false positives using information criteria . Through simulations under contrasting models, parametric bootstrapping produces likelihood ratios distributions which the observed data can be compared to, providing an estimate of power and a means to distinguish models . Using the pmc R package  and custom functions, we performed parametric bootstrapping for the BM, OU and BM vs. OU tests based on 1000 simulated datasets. R scripts and datasets used to perform these analyses are available at the Dryad repository (doi:10.5061/dryad.s8f74).
Our final strategy was to perform both BM and OU tests on all trait and tree combinations choosing either a single or multiple parameter model based on the bootstrapped distributions. We then used bootstrapping to test whether it was possible to distinguish between the best-supported BM and OU models. Support for either a multiple rate BM model or a three optimum OU model would indicate a difference in diversification and adaptive divergence between the two stickleback lineages.
Divergence in diet and feeding behaviour in a sympatric pair
Baseline-corrected δ15N values can provide an estimate of the trophic position of a consumer . We collected molluscs (n = 15) and bivalves (n = 5) from the Bekanbeushi system to characterise the benthic and pelagic food webs and calculated trophic position (TPOS) using Post's  method.
Isotope mixing models can estimate source proportions, providing a time-averaged indication of dietary preference . To assess the contribution of marine vs. freshwater foraging environments to Japanese stickleback, we used mixing-models implemented in SIAR . SIAR uses Bayesian inference to account for variation in sources and fractionation values allowing estimation of error and uncertainty. We estimated mean percentage contributions to each form at each site; contribution posterior probabilities were compared to test for differences.
Correlation between dietary preferences and ecomorphological traits is an proxy for detecting divergent natural selection between environments . Stomach content analysis was performed on individuals sampled from Bekanbeushi (n = 284, see Table S20 in File S1 for site specific sample sizes). Fish captured in Akkeshi Bay had empty stomachs and were not included. Prey items were identified counted, weighed to the nearest 0.001 mg and then classified into seven categories; terrestrial insects, zooplankton, benthic macroinvertebrates, fish, fish eggs, plant material and other. Using frequency, weight and number we calculated the index of relative importance (%IRIi) .
We also conducted benthic foraging experiments to test whether freshwater foraging efficiency was greater in PA. Fish used were captured from Bekanbeushi system, returned to the laboratory and kept for one month. Experiments were conducted in a 63-litre clear glass tank (H: 35 cm, L: 60 cm, W: 30 cm) filled with 10% seawater. The tank was placed on the floor of a well-lit room with an ambient of temperature of 16°C. All sides apart from the front were covered to prevent startling the fish. Substrate consisted of fine sand and gravel, spread thinly to prevent benthic prey items from burying themselves beyond reach. Before the first trial on a given day, 60 live chironomid larvae were added to the test arena and spread at random across the substrate. Trials were then filmed using a SONY HVR HD 1000 placed at 1.5 m from the test tank. Trials were observed remotely and could be initiated, monitored and recorded without disturbing or startling the fish. A total of 42 trials were conducted over four days in June 2012 and the test arena was cleared, cleaned and refilled on each day.
Fish were fed a diet consisting of live Artemia and frozen chironomids for one week and were then starved for 24 hours before a trial to ensure feeding. Trials were conducted using the following protocol; a fish was chosen at random from a holding tank and then placed in the test arena containing food items. Once a fish made a vertical strike at a prey item, recording was started and a ten-minute foraging trial initiated. If no strike was made within 10 min after introduction, the trial was ended and the fish removed. Following trial completion, successful or not, fish were removed and their standard length recorded. Following each trial approximately five chironomid larvae were added to the test tank and the substrate raked to ensure prey visibility.
Measures of foraging efficiency were recorded from video footage. All vertical and horizontal strikes were counted; vertical strikes were defined as strikes made at a prey item resting on the substrate; horizontal strikes were defined as strikes made at prey suspended in the water column. The former captures the number of attempts to feed on new prey while the latter captures the number of strikes required to handle prey. Videos were reviewed again to calculate the number of chironomid larvae consumed per trial. The number of chironomids handled or abandoned was recorded and the difference equalled the number of chironomids consumed.
We calculated ratios of the number of vertical and horizontal strikes as well as the number of abandoned prey items to the number of chironomids handled in order to give an indication of handling efficiency. We additionally calculated a measure of foraging efficacy.Where, Tprey is total number of prey items consumed and SV and SH are vertical and horizontal strikes. This measure ranges from 0 to 1, with larger values indicating lower numbers of strikes per prey item consumed. We further calculated the number of vertical strikes per second to obtain a measure of foraging rate.
For SIA data, generalised linear mixed models (GLMMs) were used with species and site set as fixed or random factors depending on the test (specified in the Results section). Standard length and efficacy values were log10 transformed prior to analysis; all other foraging efficiency measures were square root transformed. Foraging efficiency measures were tested using GLMs with species as a factor and standard length as a covariate. All statistical analysis was conducted using R 2.15.1 .
Higher phenotypic and ecological diversification rates in the Pacific Ocean clade than in the Japan Sea clade
Both Nei's D and δμ2 trees clearly indicated two monophyletic groups with high bootstrap support, consistent with the Pacific Ocean and Japan Sea clades (Fig. 1c; Figure S2 in File S1). All freshwater populations occurred within the Pacific Ocean clade (Fig. 1c; Figure S2 in File S1), which is consistent with our previous findings that freshwater colonization occurs in the Pacific Ocean clade only .
Parametric bootstrapping indicated both phylogenies had high power to distinguish between single and separate rate Brownian motion models (i.e., different diversification rate between lineages; mean power 83%, Figure S3 & Tables S7 & S9 in File S1). BM2 models were highly supported for gill raker number and mean δ13C and δ15N values for both Nei's D and δμ2 trees (mean power 99%, Table S7 in File S1). Therefore, diversification rates (σ2) for all three of these trophic traits were significantly higher for the Pacific Ocean lineage than the Japan Sea lineage (Fig. 1d, Table S9 in File S1), irrespective of phylogeny and sub-clade topology. Furthermore, this difference in diversification rate was not an artefact of sample size differences between lineages. Pacific Ocean σ2 values were higher than Japan Sea even when sample sizes for the two lineages were equal (Figure S11 in File S1). Nonetheless, power to detect these rate differences did increase with the number of Pacific Ocean populations included in the analysis (S13).
Power was also high in OU1 vs. OU2 comparisons (mean power 77%, Figure S4 & Table S7 in File S1), but less so for OU2 vs. OU3 (mean power 48%, Figure S5 & Table S7 in File S1). Multiple optimum models (OU2 & OU3) were highly supported for all trophic traits (Figures S5–S6 and Tables S7–S8 in File S1), suggesting that optimum trait values differed between lineages (Table S9 in File S1). For gill raker number, there was strong support for an OU3 model (Tables S7 & S8 in File S1) - i.e. lineage specific optimal trait values with an additional value for Pacific Ocean freshwater populations. Comparative results and power estimates were again consistent across topologies (Fig. 1C,D, Figures S11 & S12 in File S1) suggesting that low-topographic support within sub-clades (Figure S2 in File S1) did not influence our ability to detect higher diversification within the Pacific Ocean lineage. As with the BM models, sample size increased the variance between repeated analyses but did not alter the main findings that mean trait values differed between the lineages; Japan Sea gill raker length was consistently higher than Pacific freshwater populations when the number of Pacific Ocean populations included in the analysis was varied (Figure S12 in File S1).
Distinguishing between the best supported BM and OU models suggested that an OU2 model was preferable (mean power 97%, Figure S6 & Table S8 in File S1). Despite extremely high power (100%), it was not possible to distinguish between BM2 and OU3 models for gill raker number on either tree as the observed log likelihood ratio fell between test distributions (Figure S6 in File S1). This supports either increased trophic trait diversification or multiple adaptive optima within the Pacific Ocean lineage (Fig. 1d).
Mean gill raker number (è, lower – upper 95% CI) was lower in Pacific Ocean freshwater populations (18.04, 17.07–19.05) than in Pacific Ocean (21.75, 20-.50–22.99) and Japan Sea (24.51, 23.44–25.64) anadromous forms (Table S9 in File S1; also supported by GLMM with population as a random factor, P<0.0001). A strong positive correlation between mean gill raker number and δ15N (r = 0.70, t = 3.98, df = 17, P<0.001) was found, indicating a functional link to trophic ecology, although there was no δ13C correlation (P = 0.17).
Ecological divergence in sympatry
To investigate ecological divergence between marine forms we first focused on a sympatric anadromous pair in the Akkeshi catchment. Stable isotope analysis on macroinvertebrate prey items indicated a clear transition from marine to freshwater environments in the Akkeshi catchment (Text S14 in File S1). Mean δ13C and δ15N values differed considerably between JA and PA forms (GLMMs, site as a random factor, δ13C: F1, 100 = 188.47, R2 = 0.87, P<0.0001; δ15N: F1, 100 = 183.74, R2 = 0.81, P<0.0001, Fig. 2a). Mean δ13C for PA was more depleted than that of JA, suggesting greater use of freshwater resources; in contrast JA showed general enrichment for δ15N over PA individuals (Table 1, Fig. 2a). Accordingly, mean trophic position was higher in JA stickleback (GLMMs with site as a random factor, R2 = 0.81, F1, 100 = 119.93, P<0.0001, Fig. 2b).
NB: we are not able to explain why a single JA individual is present amongst the PA. Assignment was performed using species-specific microsatellites but we cannot rule out the possibility this sample was mislabelled.
Spatial variation in stable isotope values was also apparent within forms (Fig. 2a, c). JA stickleback captured at the midstream site were more δ13C depleted than conspecifics captured in the bay and lake (see Table 1, GLMM, collection year as random factor; R2 = 0.31, F2, 42 = 9.01, P<0.0001). PA fish varied spatially for both δ13C (R2 = 0.66, F2, 55 = 11.87, P<0.0001) and δ15N (R2 = 0.52, F2, 55 = 4.87, P = 0.01). This was largely driven by δ13C variation between fish captured at the bay site and those in the river (P<0.05 in both cases). Focusing on the bay and midstream sites where both species co-occur, δ13C differences occurred between sites (GLM; R2 = 0.76, F3, 61 = 67.34, P = 0.02) and forms (P<0.0001) but no significant interaction could be detected (P = 0.48), suggesting the difference between species did not vary between sites.
Bayesian source estimation revealed a greater contribution of marine benthic sources to JA fish (Table S15 in File S1, P = 0.002) while freshwater benthic prey was a more important prey resource for PA fish (P = 0.00, Tables S15 & S17, Figure S16 in File S1). Freshwater benthic contributions increased in both forms upstream (Table S1 & Text S14 in File S1). SEAC values suggested no isotopic niche overlap for JA and PA fish (Tables S18 & S19 in File S1). SCA revealed some overlap (PSI = 0.20) and that both forms fed on planktonic and benthic prey items (Table S20 in File S1). However JA fish fed on larger proportions of zooplankton (%IRI = 73.35) than PA fish (%IRI = 0.14, X2 = 72.94, df = 1, P<0.0001), consistent with a greater marine contribution to diet.
To further confirm the divergence in trophic ecology between the sympatric PA and JA fish, we conducted a total of 42 benthic foraging trials (17 JA, 25 PA) on wild fish captured from the Bekanbeushi system (Fig. 1b) of which 35 were successful (6 PA failures, 1 JA). PA fish were larger than JA (standard length mm ± SD; 68.2±3.5 and 53.6±2.4, respectively; GLM R2 = 0.86, F1, 32 = 205.9, P<0.0001) so size was included as a factor in the analysis to test for size specific effects. PA fish consumed a greater number of chironomids per trial than JA fish (GLM R2 = 0.39, F1, 32 = 22.02, P<0.0001, Fig. 3a, Table S21 in File S1) and also had greater efficacy values (R2 = 0.48, F1, 32 = 30.95, P<0.0001, Fig. 3c). Although the number of strikes per second did not differ (P = 0.62), a significant interaction indicated a size effect (R2 = 0.16, F3, 30 = 3.03, P = 0.009; Fig. 3b).
Patterns of ecological divergence between two marine forms in Japan
Comparing anadromous forms from both lineages across Japan, mean δ13C values for PA were lower than JA populations at both sympatric and allopatric sites, indicating greater freshwater foraging in the former (GLMMS with site as a random factor: R2 = 0.86, F1, 261 = 77.46, P<0.0001, Table 2, Fig. 4a, b). JA fish also had higher mean δ15N values, suggesting feeding at a higher trophic level (R2 = 0.62, F1, 261 = 79.28, P<0.0001, Table 2, Table S1 in File S1, Fig. 4a, c). These results suggest that trophic divergence and greater freshwater resource use in PA is consistent across the distribution range.
We additionally tested whether allopatric (i.e. only a single lineage is present at a site) and sympatric (i.e. both lineages are present) JA and PA populations differed in trophic ecology. Distribution did account for δ13C variation (P = 0.02) and a significant species * distribution interaction (P = 0.05; GLM, R2 = 0.24, F3, 270 = 29.42, P<0.0001) indicated that JA fish occurring in sympatry with PA fish had a more marine δ13C signal than allopatric JA populations; no difference occurred between allopatric and sympatric PA populations (Fig. 4b and c). Isotopic niche size was also larger in allopatric JA, allopatric PA and sympatric PA populations compared to sympatric JA (P = 0.99, 0.96 & 1 respectively, Table 2), suggesting a possible role of competition when JA and PA forms co-occur.
The results of our comparative phylogenetic analyses show that rates of phenotypic and ecological diversification are higher in the Pacific Ocean stickleback lineage compared to the closely related Japan Sea lineage. Since marine-freshwater adaptive divergence has only occurred in the former, our findings support the hypothesis that colonization of freshwater environments facilitates the evolution of increased diversity in stickleback foraging traits between marine-freshwater populations within lineages. Different mean optimal gill raker numbers between JA and PA populations indicate divergent ecological selection also occurs between these anadromous forms. To further investigate differences in optimal trophic trait values, we quantified divergence in resource use, foraging morphology and behaviour between the anadromous forms (PA and JA) in these two lineages. PA exploited more benthic freshwater resources, consistent with fewer gill rakers and improved prey handling on benthic macroinvertebrates compared to JA in both sympatry and allopatry. Stable isotope analysis also suggested allopatric JA populations exploit more freshwater resources when PA are not present, suggesting the possibility that competition may occur between the two anadromous forms. Together, our results confirm that only the Pacific Ocean lineage has undergone extensive diversification in foraging behaviour, ecology and morphology as a result of marine-freshwater adaptation and that substantial ecological divergence has occurred between Pacific Ocean and Japan Sea anadromous forms.
Adaptive divergence has occurred in the Pacific Ocean lineage but not the Japan Sea lineage
Sticklebacks have been extensively studied as model organisms for studies of adaptive divergence and evolution , . However most studies have focused on the ecological and genetic mechanisms underlying phenotypic diversification occurring after freshwater colonization . In comparison we know little about the ecological and genetic mechanisms that might have constrained adaptive evolution in these species. Having failed to colonize freshwater environments and evolve increased phenotypic diversification, the Japan Sea lineage provides a unique opportunity to study why this is the case. Examining factors limiting adaptive evolution in this lineage may provide additional insight to factors promoting it in others.
Selection between different habitats, resources and niches is the major determinant of adaptive divergence between populations , . Such adaptive evolution can result in different optimal traits amongst adaptive peaks, leading to an increase in the mean phenotypic diversity of a set of populations , . Strong support for a BM2 model for trophic ecology and foraging morphology indicates that phenotypic diversity has evolved in the Pacific Ocean lineage but not the Japan Sea (Fig. 1D). Similarly, our comparative phylogenetic analysis also supported an OU3 model. This confirms adaptive divergence has occurred between both lineages and within the Pacific Ocean only with different adaptive optima for JA, PA and PF populations respectively (Fig. 1D). It should be noted that it was not possible for our analysis to effectively distinguish which of these two models provided a better fit for trophic morphology. However, either scenario - higher rates of evolutionary diversification or multiple adaptive optima - suggests foraging traits such as gill raker number have diversified within the Pacific Ocean lineage.
In addition to our comparative phylogenetic results, our other analyses and previous research provides a strong case for adaptive divergence between anadromous and freshwater resident populations of the Pacific Ocean lineage. Our stable isotope data showed clear structuring along a marine-freshwater axis, with obligate freshwater PF populations at one end and JA populations at the other (Fig. 4). While PA populations have a greater freshwater signal than JA populations, there is also dietary structuring between PA and PF populations. Furthermore, this marine-freshwater dietary structuring within the Pacific Ocean lineage is accompanied by phenotype-environment associations; ecologically functional traits such as reduction of gill raker number and body size in freshwater populations have been shown in both present and previous studies , . Further work is necessary to directly confirm that these phenotype-environment associations are indicative of a direct functional link between foraging traits and improved fitness in marine and freshwater environments , , . However, our stable isotope and behavioural data indicate improved freshwater foraging efficiency in the Pacific Ocean lineage. Given that there is considerable evidence that gill raker number and body size influences foraging efficiency in other stickleback populations , , , it seems likely this is also the case between PF and PA populations in Japan.
Adaptive divergence between ancestral forms may have constrained colonisation of novel environments
Evolutionary diversification is closely linked to the colonization of new environments and establishment success often depends on the similarity of these environments to the source habitat as colonisers may already possess suitable adaptations , . Divergence in habitat-specific adaptations between the ancestral anadromous populations of these lineages may therefore have played a role in constraining adaptive divergence in this case; i.e. adaptation to marine foraging in the Japan Sea lineage might result in an intrinsic constraint limiting freshwater resource use.
Our stable isotope and stomach content analyses demonstrated divergent resource use between anadromous forms from the two lineages. PA stickleback exploited a greater proportion of freshwater resources than JA across their distribution with mean isotopic values more similar to freshwater resident populations (PF). Spatial isolation of spawning sites following migration upstream between the PA and JA forms at sites where both forms co-occur is supported by our present stable isotope data and also by previous longitudinal demographic studies . Habitat isolation arising as a by-product of divergent natural selection occurs quite readily between stickleback species pairs , ,  and the consistent pattern of divergence across the distribution of the Japan Sea and Pacific Ocean lineages suggests divergent selection might act on their anadromous forms.
Our comparative OU analysis supported a three optimum model for gill raker morphology – i.e. Japan Sea fish have a higher mean number of gill rakers than Pacific Ocean anadromous and freshwater populations. Support for an additional optimum in Pacific freshwater populations suggests freshwater colonization is characterised by a reduction in gill raker number and a shift towards a new adaptive peak . Gill raker number is closely associated with determining foraging efficiency throughout the stickleback species complex , , . Fish with numerous long gill rakers show greater efficiency when foraging for pelagic prey; whereas fewer, shorter gill rakers occur in benthic ecotypes . This additional optimum value also indicates that adaptive diversification has occurred with the Pacific Ocean lineage but not the Japan Sea lineage.
Our behavioural experiment supports a functional link between gill raker morphology and foraging efficiency in the sympatric Bekanbeushi population. PA fish consumed more benthic prey and demonstrated improved prey handling. As variation in gill raker number was not included in our experimental design, we cannot conclude that lower efficiency with benthic prey in JA fish is as a result of morphological adaptation to feeding on pelagic prey. However gill raker number has been shown to correlate closely with foraging efficiency in stickleback and other fish species , . Furthermore, our stable isotope data supports a strong correlation between foraging morphology and behaviour in Japanese sticklebacks; enriched mean δ15N values and higher trophic level in Japan Sea anadromous fish indicate increased pelagic diet. Correlations between morphology and behaviour are compelling but further work is necessary to test whether gill raker morphology directly increases foraging efficiency in the Japanese stickleback system.
Body size can also influence prey capture success in fishes and may be more important in determining benthic foraging success than trophic morphology , , . A significant relationship between strikes per second and size suggests larger JA individuals exhibit improved prey handling. This was not the case for PA fish however as adult PA show very little body-size variation (J Kitano, personal observation) and our sample may have lacked the variation necessary to show such a relationship. Preliminary results from foraging experiments using smaller fish support this conclusion as both forms show poor benthic prey handling (Ravinet & Kitano, unpublished data). Smaller stickleback have a greater handling cost when feeding on larger food items such as benthic macroinvertebrates . Since JA stickleback are smaller than PA , this may contribute to lower benthic prey foraging efficiency, suggesting size may act as a competitive advantage for PA when colonizing freshwater environments.
Although our present study has focused on ecological factors arising from differences in foraging and morphology, physiological constraint may have also played a role in preventing marine-freshwater adaptive divergence in the Japan Sea lineage. ‘Key innovations’ are adaptive traits which allow new niches to be exploited and can increase diversification, potentially leading to adaptive radiations when they arise , . In contrast, the loss or failure to evolve key innovations would prevent a lineage from being exposed to divergent selection necessary for adaptive divergence and diversification. The Pacific Ocean lineage and all other known stickleback lineages have been able to colonise freshwater environments , , , . Furthermore, adaptive divergence between stickleback species pairs occurs either between contrasting freshwater environments or along a marine-freshwater axis. Freshwater tolerance may therefore be a key innovation that has allowed increased diversification within the Pacific Ocean and other stickleback lineages. JA however exhibit lower freshwater tolerance with low survival rates in freshwater than PA , . The loss or reduction of freshwater tolerance probably plays a major role in preventing JA fish from undergoing adaptive divergence.
Possible extrinsic constraints of adaptive divergence: competition
Ecological interactions, such as competition for resource use, may also play an important role in shaping colonization of and adaptation to novel environments , . More depleted δ13C values in allopatric JA populations suggest that the JA form is able to make greater use of freshwater resources when PA fish are absent. Competition for resources can also have a negative effect on population densities during colonization . Where JA and PA fish occur in sympatry, numbers of JA fish are small and fluctuate yearly . PA fish also display much better benthic prey handling than JA fish, suggesting that competition in sympatry may restrict JA habitat use, limiting freshwater adaptation and diversification.
Priority effects may also explain the difference in the ability of the two lineages to adapt to freshwater. Priority effects are species-specific interactions that influence establishment and fitness –. In short, if the Pacific Ocean lineage was able to reach freshwater environments before the Japan Sea lineage, this may have given it a fitness advantage. This seems unlikely as current evidence suggests the two lineages diverged 1.5–3 million years in the Japanese archipelago when the Sea of Japan was isolated from the Pacific Ocean during the Quaternary , , . Furthermore, fossil evidence suggests sticklebacks have been present in East Asia and around Japan for ∼10 million years . Additionally, freshwater environments occur on both the Pacific and Sea of Japan coasts of the Japanese Islands, indicating both lineages would have been able to access freshwater environments during and following divergence.
Competition may play some role in determining differences in resource use when the two lineages occur in sympatry, however it seems unlikely this is the sole explanation for the failure of the Japan Sea lineage to diversify by colonizing freshwater environments, Resource competition would only prevent colonisation where the two forms overlap (Fig. 1a) and there is no evidence of freshwater colonisation in regions where Japan Sea fish are present and Pacific Ocean absent, despite suitable freshwater habitats being available . Nonetheless the nine-spined stickleback (Pungitius spp.) could also act as a competitor, having adapted to freshwater environments connected to the Sea of Japan , . Nine-spined stickleback are also absent from the Pacific coast, except in Hokkaido potentially allowing greater ecological opportunity for the Pacific Ocean lineage although this has not been explicitly studied. Further work is necessary to test whether inter-specific competition occurs between Japanese stickleback lineages.
Internal and external constraints on adaptation and phenotypic diversification should not be considered in isolation as the interaction between both classes of constraint is often more informative . Intrinsic constraints for example may play a role in mediating ecological interactions increasing the effect size of potential external constraints. Low freshwater tolerance in Japan Sea sticklebacks ,  for example would explain the lack of diversification in freshwater environments even in areas where the two forms do not overlap. An interesting avenue for further work is to test whether intrinsic constraints such as freshwater tolerance can influence ecological interactions such as competition and foraging efficiency. A combination of multiple factors, including intrinsic constraints and interspecific competition probably constrain freshwater colonisation and morphological diversification in the Japan Sea lineage .
Importance of comparative studies of adaptive divergence
Genomic constraints on phenotypic diversification remain uncertain, although the loss of allelic variants underlying traits may constrain adaptive divergence . Mutation effect sizes influence the probability of adaptation as genes of small effect are less likely to shift populations closer to distant adaptive optima , . Divergent selection between lineages leading to stronger adaptation to the marine environment by the Japan Sea lineage may have driven it further from a freshwater optimum. Since multiple genes of small effect play a role in stickleback freshwater adaptation  this may have led to genomic constraint on diversification. Selection from standing genetic variation is also important for stickleback adaptation , . Loss of genetic variation in genomic regions underlying adaptive traits during a period of isolation may have lowered adaptive potential in the Japan Sea lineage. Further work combining biogeographical information and next-generation sequencing is now necessary to identify the roles of selection and drift in the loss of adaptive genomic variation in the Japan Sea stickleback lineage.
Understanding factors that facilitate or constrain adaptive divergence, phenotypic diversification and adaptive radiation is a fundamental question for evolutionary biologists. Both phylogenetic and experimental data have demonstrated that ecology of founders can influence the patterns of adaptive radiation , . Although an increasing number of studies have examined the genomic basis for adaptive evolution , , few have focused on lineages that are unable to diversify or undergo divergence. Comparing closely related lineages differing in the magnitude of phenotypic divergence is fertile ground for developing an understanding of the genetic constraints on adaptive evolution. Different mechanisms might constrain divergence and diversification in different taxa, particularly when chance and historical contingency play a greater role than the deterministic influences of selection . Further studies on the differences in phenotypic diversification across diverse taxa are required for a better understanding the constraints on adaptive divergence following the colonisation of novel environments.
Ecological opportunity is thought to be key for phenotypic diversification , . Adaptive traits allowing the exploitation of new environments, are fundamental for driving adaptive divergence, radiation and phenotypic diversification . Indeed, when ancestral species make transitions into new environments similar to their current habitat, survival and diversification is more likely . Failure to maintain or evolve adaptive traits may constrain the colonization and invasion of new environments, ultimately limiting the range and evolutionary diversification of a species. Further comparative and genomic studies of closely related lineages with differing evolutionary rates will provide a promising means of understanding the constraints on diversification and adaptive evolution.
We are grateful to the Akkeshi Bay fishermen, Minoru Kanaiwa, and Satoru Takeyama for their assistance in providing stickleback and macroinvertebrate samples. We thank Shuji Shigenobu at NIBB for use of his laboratory. We thank Roger Butlin, Kohta Yoshida and Asano Ishikawa for their helpful comments during this study and to two anonymous reviewers for their constructive criticisms on the manuscript. We are also grateful to Carl Boettiger for his assistance with reproducing the pmc method.
Conceived and designed the experiments: MR JK. Performed the experiments: MR NT. Analyzed the data: MR. Contributed reagents/materials/analysis tools: MK SM JK. Wrote the paper: MR JK.
- 1. Schluter D (2000) The ecology of adaptive radiation. Oxford: Oxford University Press.
- 2. Berner D, Grandchamp AC, Hendry AP (2009) Variable progress toward ecological speciation in parapatry: stickleback across eight lake-stream transitions. Evolution (N Y) 63-7:1740–1753.
- 3. Jones FC, Grabherr MG, Chan YF, Russell P, Mauceli E, et al. (2012) The genomic basis of adaptive evolution in threespine sticklebacks. Nature 484:55–61.
- 4. Losos JB, Mahler DL (2010) Adaptive Radiation: The Interaction of Ecological Opportunity, Adaptation, and Speciation. In: Bell MA, Futuyma DJ, Eanes W, Levinton J, editors. Evolution Since Darwin: the First 150 Years. Sunderland, Massachusetts: Sinauer Associates. pp. 381–420.
- 5. Losos JB (2010) Adaptive radiation, ecological opportunity, and evolutionary determinism. Am Nat 175:623–639.
- 6. Berner D, Roesti M, Hendry AP, Salzburger W (2010) Constraints on speciation suggested by comparing lake-stream stickleback divergence across two continents. Mol Ecol 19:4963–4978.
- 7. Ravinet M, Prodöhl PA, Harrod C (2013) Parallel and nonparallel ecological, morphological and genetic divergence in lake - stream stickleback from a single catchment. J Evol Biol 26:186–204.
- 8. Barrett RDH, Schluter D (2008) Adaptation from standing genetic variation. Trends Ecol Evol 23:38–44.
- 9. Waters JM (2011) Competitive exclusion: phylogeography's elephant in the room? Mol Ecol 20:4388–4394.
- 10. Wiens JJ, Ackerly DD, Allen AP, Anacker BL, Buckley LB, et al. (2010) Niche conservatism as an emerging principle in ecology and conservation biology. Ecol Lett 13:1310–1324.
- 11. Wagner CE, Harmon LJ, Seehausen O (2012) Ecological opportunity and sexual selection together predict adaptive radiation. Nature 487:366–369.
- 12. Kaeuffer R, Peichel CL, Bolnick DI, Hendry AP (2012) Parallel and nonparallel aspects of ecological, phenotypic and genetic divergence across replicate population pairs of lake and stream stickleback. Evolution (N Y) 66:402–418.
- 13. Crisp MD, Arroyo MTK, Cook LG, Gandolfo MA, Jordan GJ, et al. (2009) Phylogenetic biome conservatism on a global scale. Nature 458:754–756.
- 14. Lee CE, Bell MA (1999) Causes and consequences of recent freshwater invasions by saltwater animals. Trends Ecol Evol 14:284–288.
- 15. Schluter D, Nagel L (1995) Parallel speciation by natural selection. Am Nat 146:292–301.
- 16. Bell MA, Foster SA (1994) Introduction to the evolutionary biology of the threespine stickleback. In: Bell MA, Foster SAeditors. The Evolutionary Biology of the Threespine Stickleback. Oxford: Oxford University Press. pp. 1–27.
- 17. McKinnon JS, Rundle HD (2002) Speciation in nature: the threespine stickleback model systems. Trends Ecol Evol 17:480–481.
- 18. Hendry AP, Bolnick DI, Berner D, Peichel CL (2009) Along the speciation continuum in sticklebacks. J Fish Biol 75:2000–2036.
- 19. Lucek K, Sivasundar a, Roy D, Seehausen O (2013) Repeated and predictable patterns of ecotypic differentiation during a biological invasion: lake-stream divergence in parapatric Swiss stickleback. J Evol Biol 1–19.
- 20. Higuchi M, Goto A (1996) Genetic evidence supporting the existence of two distinct species in the genus Gasterosteus around Japan. Environ Biol Fishes 47:1–16.
- 21. Cassidy L, Ravinet M, Mori S, Kitano J (2013) Are Japanese freshwater populations of threespine stickleback derived from the Pacific Ocean lineage? Evol Ecol Res 15:295–311.
- 22. Kitano J, Mori S, Peichel CL (2007) Phenotypic divergence and reproductive isolation between sympatric forms of Japanese threespine sticklebacks. Biol J Linn Soc 91:671–685.
- 23. Kitano J, Ross JA, Mori S, Kume M, Jones FC, et al. (2009) A role for neo-sex chromosomes in stickleback speciation. Nature 461:1079–1083.
- 24. Spoljaric MA, Reimchen TE (2007) 10,000 years later: evolution of body shape in Haida Gwaii three-spined stickleback. J Fish Biol 70:1484–1503.
- 25. Mäkinen HS, Merilä J (2008) Mitochondrial DNA phylogeography of the three-spined stickleback (Gasterosteus aculeatus) in Europe - Evidence for multiple glacial refugia. Mol Phylogenet Evol 46:167–182.
- 26. DeFaveri J, Zanella LN, Zanella D, Mrakovcic M, Merilä J (2012) Phylogeography of isolated freshwater three-spined stickleback Gasterosteus aculeatus populations in the Adriatic Sea basin. J Fish Biol 80:61–85.
- 27. Deagle BE, Jones FC, Absher DM, Kingsley DM, Reimchen TE (2013) Phylogeography and adaptation genetics of stickleback from the Haida Gwaii archipelago revealed using genome-wide single nucleotide polymorphism genotyping. Mol Ecol 22:1917–1932.
- 28. Ortí G, Bell MA, Reimchen TE, Meyer A (1994) Global survey of mitochondrial DNA sequences in the threespine stickleback: evidence for recent migrations. Evolution (N Y) 48:608–622.
- 29. Kume M, Kitamura T, Takahashi H, Goto A (2005) Distinct spawning migration patterns in sympatric Japan Sea and Pacific Ocean forms of threespine stickleback Gasterosteus aculeatus. Icthyological Res 52:189–193.
- 30. Kume M, Kitano J, Mori S, Shibuya T (2010) Ecological divergence and habitat isolation between two migratory forms of Japanese threespine stickleback (Gasterosteus aculeatus). J Evol Biol 23:1436–1446.
- 31. Post DM (2002) Using stable isotopes to estimate trophic position: Models, methods, and assumptions. Ecology 83:703–718.
- 32. Schluter D (1993) Adaptive radiation in sticklebacks: size, shape and habitat use efficiency. Ecology 74:699–704.
- 33. Schluter D (1995) Adaptive radiation in sticklebacks - trade-offs in feeding performance and growth. Ecology 76:82–90.
- 34. Berner D, Adams DC, Grandchamp AC, Hendry AP (2008) Natural selection drives patterns of lake-stream divergence in stickleback foraging morphology. J Evol Biol 21:1653–1665.
- 35. Hyslop EJ (1980) Stomach contents analysis - a review of methods and their application. J Fish Biol 17:411–429.
- 36. Schindler DE, Lubetkin SC, Polis GA, Power MA, Huxel GR (2004) Using stable isotopes to quantify material transport in food webs. In: Polis GA, Power ME, Huxel GReditors. Food webs at the landscape level. Chicago: University of Chicago Press, Vol. 3pp. 25–43.
- 37. Grey J (2006) The use of stable isotope analyses in freshwater ecology: current awareness. Polish J Ecol 54:563–584.
- 38. Hesslein RH, Hallard KA, Ramlal P (1993) Replacement of sulfur, carbon and nitrogen in tissue of growing broad whitefish (Coregonus nasus) in response to change in diet traced by δ34S, δ13C and δ15N. Can J Fish Aquat Sci 50:2071–2075.
- 39. Grey J (2000) Trophic fractionation and the effects of diet switch on the carbon stable isotope “signatures” of pelagic consumers. Verhandlungen der Int Vereinigung für Theortische und Angew Limnol 27:3187–3191.
- 40. Perga ME, Gerdeaux D (2005) “Are fish what they eat” all year round? Oecologia 144:598–606.
- 41. Kiljunen M, Grey J, Sinisalo T, Harrod C, Immonen H, et al. (2006) A revised model for lipid-normalizing δ13C values from aquatic organisms, with implications for isotope mixing models. J Appl Ecol 43:1213–1222.
- 42. Newsome S, Martínez Del Rio C, Bearhop S, Phillips DL (2007) A niche for isotopic ecology. Front Ecol Environ 5:429–436.
- 43. Jackson AL, Inger R, Parnell A, Bearhop S (2011) Comparing isotopic niche widths among and within communities: SIBER - Stable Isotope Bayesian Ellipses in R. J Anim Ecol 80:595–602.
- 44. Robinson BW (2000) Trade offs in habitat-specific foraging efficiency and the nascent adaptive divergence of sticklebacks in lakes. Behaviour 137:865–888.
- 45. Nei M, Tajima F, Tateno Y (1983) Accuracy of estimated phylogenetic trees from molecular data II. Gene frequency data. J Mol Evol 19:153–170.
- 46. Goldstein DB, Ruiz Linares A, Cavalli-Sforza LL, Feldman MW (1995) Genetic absolute dating based on microsatellites and the origin of modern humans. Proc Natl Acad Sci 92:6723–6727.
- 47. Peichel CL, Nereng KS, Ohgi KA, Cole BLE, Colosimo PF, et al. (2001) The genetic architecture of divergence between threespine stickleback species. Nature 414:901–904.
- 48. Wu C-H, Drummond AJ (2011) Joint inference of microsatellite mutation models, population history and genealogies using transdimensional Markov Chain Monte Carlo. Genetics 188:151–164.
- 49. Nei M, Kumar S (2000) Molecular Evolution and Phylogenetics. Oxford, UK: Oxford University Press.
- 50. Langella O (2013) Populations 1.2.31.
- 51. Paradis E, Claude J, Strimmer K (2004) APE: Analyses of Phylogenetics and Evolution in R language. Bioinformatics 20:289–290.
- 52. O'Meara BC, Ané C, Sanderson MJ, Wainwright PC (2006) Testing for different rates of continuous trait evolution using likelihood. Evolution (N Y) 60:922–933.
- 53. Revell LJ (2012) phytools: an R package for phylogenetic comparative biology (and other things). Methods Ecol Evol 3:217–223.
- 54. Butler MA, King AA (2004) Phylogenetic comparative analysis: a modeling approach for adaptive evolution. Am Nat 164:683–695.
- 55. Hansen T (1997) Stabilizing selection and the comparative analysis of adaptation. Evolution (N Y) 51:1341–1351.
- 56. Boettiger C, Coop G, Ralph P (2012) Is your phylogeny informative? Measuring the power of comparative methods. Evolution (N Y) 2240–2251.
- 57. Phillips DL, Gregg JW (2003) Source partitioning using stable isotopes: coping with too many sources. Oecologia 136:261–269.
- 58. Parnell A, Inger R, Bearhop S, Jackson AL (2010) Source partitioning using stable isotopes: coping with too much variation. PLoS One 3:1–5.
- 59. R Development Core Team (2012) R: A language and environment for statistical computing.
- 60. Endler JA (1986) Natural selection in the wild. New Jersey, USA: Princeton University Press.
- 61. Rogers SM, Tamkee P, Summers B, Balabahadra S, Marks M, et al. (2012) Genetic Signature of Adaptive Peak Shift in Threespine Stickleback. Evolution (N Y) 66:2439–2450.
- 62. Mori S (1987) Geographical variations in freshwater populations of the three-spined stickleback, Gasterosteus aculeatus, in Japan. Japanese J Ichthyol 34:33–46.
- 63. Räsänen K, Hendry AP (2008) Disentangling interactions between adaptive divergence and gene flow when ecology drives diversification. Ecol Lett 11:624–636.
- 64. Ibrahim AA, Huntingford FA (1988) Foraging efficiency in relation to within-species variation in morphology in threeispined sticklebacks, Gasterosteus aculeatus. J Fish Biol 33:823–824.
- 65. Hendry AP, Taylor EB (2004) How much of the variation in adaptive divergence can be explained by gene flow? An evaluation using lake-stream stickleback pairs. Evolution (N Y) 58:2319–2331.
- 66. Bolnick DI, Snowberg LK, Patenia C, Stutz WE, Ingram T, et al. (2009) Phenotype-dependent native habitat preference facilitates divergence between parapatric lake and stream stickleback. Evolution (N Y) 63-8:2004–2016.
- 67. Robinson BW, Schluter D (2000) Natural selection and the evolution of adaptive genetic variation in northern freshwater fishes. In: Mousseau T, Sinervo B, Endler JAeditors. Adaptive genetic variation in the wild. New York: New York. pp. 65–94.
- 68. Roesch C, Lundsgaard-Hansen B, Vonlanthen P, Taverna a, Seehausen O (2013) Experimental evidence for trait utility of gill raker number in adaptive radiation of a north temperate fish. J Evol Biol 26:1578–1587.
- 69. Gill A, Hart PJB (1996) How feeding performance and energy intake change with a small increase in the body size of the three-spined stickleback. J Fish Biol 48:878–890.
- 70. Rutschmann S, Matschiner M, Damerau M, Muschick M, Lehmann MF, et al. (2011) Parallel ecological diversification in Antarctic notothenioid fishes as evidence for adaptive radiation. Mol Ecol 20:4707–4721.
- 71. Near TJ, Dornburg A, Kuhn KL, Eastman JT, Pennington JN, et al. (2012) Ancient climate change, antifreeze, and the evolutionary diversification of Antarctic fishes. Proc Natl Acad Sci U S A 109:3434–3439.
- 72. Hamada K (1975) Excessively enlarged thyroid follicles of the threespine stickleback, Gasterosteus aculeatus aculeatus, reared in freshwater. Japanese J Ichthyol 21:183–190.
- 73. Honma Y, Shioda S, Yoshie S (1977) Changes in the thyroid gland associated with the diadromous migration of the threespine stickleback, Gasterosteus aculeatus. Japanese J Ichthyol 24:17–25.
- 74. Losos JB, Spiller DA (1999) Differential Colonization Success and Asymmetrical Interactions between Two Lizard Species. Ecology 80:252.
- 75. Shorrocks B, Bingley M (1994) Priority effects and species coexistence : experiments with fungal-breeding Drosophila. J Anim Ecol 63:799–806.
- 76. Almany G (2003) Priority Effects in Coral Reef Fish Communities. Ecology 84:1920–1935.
- 77. Alford R, Wilbur H (1985) Priority Effects in Experimental Pond Communities : Competition between Bufo and Rana. Ecology 66:1097–1105.
- 78. Nazarkin MV, Yabumoto Y, Urabe A (2013) A New Miocene Three-Spined Stickleback (Pisces: Gasterosteidae) from Central Japan. Paleontol Res 16:318–328.
- 79. Goto A, Mori S (2003) Natural history of stickleback. Sapporo: Hokkaido University Press.
- 80. Ikeda K (1933) Distribution and variation of sticklebacks. Zool Mag 45:141–173.
- 81. Tsuruta T, Goto A (2006) Fine scale genetic population structure of the freshwater and Omono types of nine-spined stickleback Pungitius pungitius (L.) within the Omono River system, Japan. J Fish Biol 69:155–176.
- 82. Orr HA (1998) The population genetics of adaptation: the distribution of factors fixed during adaptive evolution. Evolution (N Y) 52:935–949.
- 83. DeFaveri J, Shikano T, Shimada Y, Goto A, Merilä J (2011) Global analysis of genes involved in freshwater adaptation in threespine sticklebacks (Gasterosteus aculeatus). Evolution (N Y) 65:1800–1807.
- 84. Schluter D, Conte GL (2009) Genetics and ecological speciation. Proc Natl Acad Sci U S A 106:9955–9962.
- 85. Ord TJ (2012) Historical contingency and behavioural divergence in territorial Anolis lizards. J Evol Biol 25:2047–2055.
- 86. Flohr RCE, Blom CJ, Rainey PB, Beaumont HJE (2013) Founder niche constrains evolutionary adaptive radiation. Proc Natl Acad Sci U S A 110:20663–20668.
- 87. Hoekstra HE, Hirschmann RJ, Bundey RA, Insel P a, Crossland JP (2006) A single amino acid mutation contributes to adaptive beach mouse color pattern. Science (80-) 313:101–104.
- 88. Colosimo PF, Peichel CL, Nereng K, Blackman BK, Shapiro MD, et al. (2004) The genetic architecture of parallel armor plate reduction in threespine sticklebacks. PLoS Biol 2:E109.
- 89. Gould SJ (1990) Wonderful Life: The Burgess Shale and the nature of history. London: Vintage.