Journal of Biogeography (J. Biogeogr.) (2015) 42, 1626–1638

ORIGINAL Phylogeography of the prickly sculpin ARTICLE (Cottus asper) in north-western North America reveals parallel phenotypic evolution across multiple coastal–inland colonizations Stefan Dennenmoser1,2*, Arne W. Nolte2, Steven M. Vamosi1 and Sean M. Rogers1

1Department of Biological Sciences, University ABSTRACT of Calgary, Calgary AB T2N 1N4, Canada, Aim Glacial cycles during the Pleistocene may have frequently contributed to 2Max-Planck Institute for Evolutionary parallel evolution of phenotypes across independently evolving genetic lineages Biology, 24306 Pl€on, Germany associated with separate glacial refugia. Previous studies based on morphology suggested that the prickly sculpin (Cottus asper) survived the Last Glacial Maxi- mum (LGM) in southern coastal and inland refugia, favouring allopatric diver- gence between coastal and inland prickling phenotypes, which vary in the degree to which spine-like scales cover the body of the fish. Herein, we aimed to test whether parallel evolution across multiple genetic lineages rather than a single-lineage origin of highly prickled inland sculpins could serve as an expla- nation for the biogeographical distribution of prickling phenotypes. Location North-western North America, Southeast Alaska and Canada (). Methods We used data from mitochondrial haplotypes and 19 microsatellite loci to identify distinct genetic lineages as a basis to interpret patterns of phe- notypic evolution. Results The occurrence of multiple mtDNA groups suggests that highly prick- led inland phenotypes comprise more than one genetic lineage. Both mtDNA and microsatellite data are consistent with post-glacial dispersal along the coast and repeated coastal to inland colonization events, as opposed to inland dis- persal of a single lineage from a southern refugium to northern regions. Main conclusions Our results suggest that highly prickled inland phenotypes evolved repeatedly following multiple inland colonization events, probably via coastal rivers. The prickly sculpin therefore provides an example of recent (post-glacial) parallel evolution, potentially facilitated by standing genetic varia-

*Correspondence: Stefan Dennenmoser, tion already present in the ancestral coastal populations. Max-Planck Institute for Evolutionary Biology, € Keywords August Thienemann Str. 2, 24306 Plon, Cottus asper Germany. , fish, morphometrics, multiple colonization events, North Amer- E-mail: [email protected] ica, Pacific Northwest, parallel evolution, phylogeography, Pleistocene.

creates unique opportunities to investigate evolutionary con- INTRODUCTION sequences of allopatric separation in glacial refugia, rapid The north-west of North America is characterized by a com- range expansions into recently deglaciated areas, and mixing plex geomorphological history associated with multifaceted of genetic lineages in zones of secondary contact (Shafer phylogeographical patterns largely shaped by repeated glacia- et al., 2010). Such investigations elucidate the relative impor- tions during the Pleistocene (c. 2.5 Ma–12,000 years ago; tance of historical and ecological effects on intraspecific geo- reviewed in Shafer et al., 2010). The increasing knowledge of graphical variation (Avise et al., 1987; Shikano et al., 2010), the post-glacial history of numerous taxa in this region and are therefore integral towards testing predictions about

1626 http://wileyonlinelibrary.com/journal/jbi ª 2015 John Wiley & Sons Ltd doi:10.1111/jbi.12527 Phylogeography reveals parallel evolution the adaptive divergence of local populations (Knowles, Phenotypic divergence between highly prickled inland scul- 2009). Ecological adaptation through natural selection can be pins and weakly prickled coastal sculpins has therefore been inferred when the repeated colonization of a new habitat interpreted as a consequence of allopatric divergence between across independently evolving genetic lineages results in par- coastal and inland glacial refugia, followed by northward dis- allel phenotypic evolution (Lu & Bernatchez, 1999; Schluter, persal during deglaciation (Krejsa, 1965; McPhail, 2007). 2000). However, recent studies have demonstrated that the However, assumptions about post-glacial dispersal routes genetic basis of parallel phenotypes may vary among popula- and phenotypic divergence of C. asper are based on the tra- tions, ranging from identical or different mutations in the ditional view that species persisted solely in glacial refugia same gene to mutations in different genes (Elmer & Meyer, south of the ice sheets (Hewitt, 2004). Meanwhile, phylogeo- 2011; Elmer et al., 2014). For example, historical differentia- graphical studies have increasingly revealed the existence of tion in standing genetic variation may lead to different more northern, cryptic glacial refugia (reviewed in Shafer genetic pathways causing parallel evolution across the distri- et al., 2010) that may have provided additional sources for bution range of a species (Prunier et al., 2012), and genetic allopatric divergence and post-glacial range expansions. Thus, introgression among divergent lineages could increase geo- highly prickled inland populations may have evolved in par- graphical variation of genomic architectures (Bernatchez allel across multiple glacial lineages originating from both et al., 2010), potentially favouring different genetic pathways southern and northern refugia. during the parallel evolution of a similar phenotype. There- Here, we use multi-locus data (mitochondrial DNA and fore, it is increasingly important to consider the historical microsatellites), as well as phenotypic data (prickling catego- context of parallel evolution and its underlying genetic basis ries, morphometrics), to elucidate the phylogeographical to better understand targets and mechanisms of natural context for the disjunct inland distribution of highly prickled selection. As mitochondrial DNA presents a biased view of phenotypes in the prickly sculpin. We aimed to distinguish demographic history and may not be selectively neutral between two main hypotheses. (Dowling et al., 2008), phylogeographical studies that use 1. Our first hypothesis is that highly prickled inland popula- both mitochondrial and nuclear DNA markers for investigat- tions originated from a single southern inland refugium, fol- ing the historical background of parallel phenotypes across lowed by post-glacial colonization of inland areas through the range of a species are needed. proglacial lakes temporarily connecting neighbouring water- The prickly sculpin, Cottus asper Richardson, 1836, is a sheds. Accordingly, a long glacial phase of allopatric diver- euryhaline fish species characterized by both amphidromous gence between coastal and inland populations should result and purely freshwater populations. Consistent phenotypic in pronounced coastal–inland differentiation in both genetic variants have been observed across a wide distributional and morphological data. A decrease of genetic diversity range between southern Alaska and California, representing towards northern latitudes would be expected if founder potentially diverse historical backgrounds (Krejsa, 1965; events characterized post-glacial colonization. McPhail, 2007). Post-glacial colonization of inland regions 2. Our second hypothesis is that highly prickled inland scul- has resulted in the predominance of a highly prickled pheno- pins evolved following colonization from multiple inland type characterized by a dense lateral coverage with spiny, and coastal refugia. In this case, highly prickled inland phe- scale-like epidermal structures (Krejsa, 1965). A heritable notypes should be associated with multiple genetic lineages genetic basis of prickling probably exists, because a major rather than a southern inland lineage. If inland groups do quantitative trait locus (QTL) that affects prickling has been not share a common glacial history of long-time coastal– found in the European sculpin, Cottus perifretum (Cheng, inland separation, morphological divergence between coastal 2013). Geographically disjunct distributions in prickling and inland groups is not expected to be strong. Under this intensity have been found in additional Cottus species in scenario, decreases in genetic diversity towards northern Europe and North America (Koli, 1969; Freyhof et al., 2005; regions are not necessarily expected, because of the presence McPhail, 2007), but we are not aware of any apparent of northern glacial refugia and admixture zones. phenotype–environment associations that would indicate a common adaptive function of prickling. MATERIALS AND METHODS Cottus asper is thought to have survived the Last Glacial Maximum (LGM) in two glacial refugia south of the ice Sample collection and DNA sequence amplification sheets: a coastal refugium located in the Chehalis River region (south of Puget Sound, Washington State, USA) and Prickly sculpins were sampled with baited minnow traps an inland refugium in the system (British during spring and summer 2009 and 2010 from 29 sites in Columbia, Canada, and Washington State, USA) (McAllister British Columbia and one site in Alberta (Fig. 1). A total of & Lindsey, 1961; Krejsa, 1965). Accordingly, two northward 1053 sculpins were collected, euthanized with an overdose of post-glacial colonization routes have been proposed: (1) clove oil, and preserved in 95% ethanol. Additionally, 14 fin coastal along-shore dispersal through salt-tolerant planktonic clips from Auke Creek (Juneau, SE Alaska) were generously larvae; and (2) inland through proglacial lakes creating provided by David Tallmon (University of Alaska South- ephemeral connections between river systems (Krejsa, 1965). east). Extractions of genomic DNA were obtained from

Journal of Biogeography 42, 1626–1638 1627 ª 2015 John Wiley & Sons Ltd S. Dennenmoser et al.

Figure 1 Map of north-western North America showing sampling sites of the prickly sculpin (Cottus asper). Filled squares represent sites with both sequence and microsatellite data available, open squares sites with microsatellite data only, and black circles additional sampling sites where morphometric data were collected. Abbreviations: CL, Cowichan Lake; GL, Graham Lake; KR, Keogh River; MCL, McCreight Lake; NKL, Nimpkish Lake; NL, North Lake; NR, ; P, Prospect Lake; PAQ, Paq Lake; PI, Poirier Lake; PR, Parsnip River; RL, Ruby Lake; SL, Skidegate Lake; WC, Williams Creek. either abdominal muscle or fin-clip tissue using a standard conducted to test whether coastal and inland sculpin popula- phenol–chloroform technique. Polymerase chain reactions tions differ in frequencies of prickling categories. (PCRs) were used to amplify a total of 1585 base pairs (bp) We analysed body shape using a geometric morphometrics of mitochondrial DNA (mtDNA) sequence for 141 sculpins approach to assess morphological divergence among coastal from 12 sites: Alaska (Auke Creek, n = 14), Meziadin Lake and inland sculpins. We applied 23 homologous landmarks (n = 11), Mosquito Lake (n = 10), Bella Coola River digitized on the left lateral body side of photographed scul- (n = 11), Nimpo Lake (n = 12), McLeod Lake (n = 13), pins using tpsDIG 2.14 and included a ruler in the image in British Columbia (n = 11), Peace River in for scaling (Rohlf, 2009a) (see Appendix S2). Because preser- Alberta (n = 10), Falls Creek (n = 10), Little Campbell River vation caused pronounced arching of the fish bodies, we (n = 16), (n = 10), and included only samples showing minimal bending in analyses (n = 13) (Fig. 1, Table 1; for primer sequences and PCR (coastal: n = 548; inland: n = 183). To remove remaining protocols see Appendix S1 in Supporting Information). effects of bending from the landmark configuration, we used Sequence proofreading was conducted with 4peaks 1.7.2 the ‘unbend’ function implemented in tpsUTIL 1.44 (Rohlf, (Griekspoor & Groothuis, 2006), and alignments were car- 2009b). For this purpose, landmarks 8 and 21–23 were cho- ried out with ClustalX 2.1 (Larkin et al., 2007). Sequences sen to represent a straight line (Appendix S2). A quadratic of mtDNA haplotypes are deposited in GenBank (accession curve was then fitted along these landmarks and used to cor- numbers KR024578–KR024637). rect for the arch in the complete landmark set. After remov- ing landmarks 21–23, we used MorphoJ 1.05d (Klingenberg, 2011) to conduct a generalized least squares Procrustes Morphology superimposition to remove the effects of rotation, scale and Sculpins were bleached for 2–10 days (3 parts of 0.5% potas- non-allometric effects of size. The resulting partial warp sium hydroxide to 1 part glycerol, containing 50 lL of 50% scores and uniform components were used as shape variables hydrogen peroxide per 10 mL KOH:glycerol solution), and in subsequent analyses. To test for allometric effects within stained for 48 h in a 1% potassium hydroxide and Alizarin coastal and inland groups, we used a multivariate analysis of red solution. We assessed prickling pattern and body shape covariance (MANCOVA) with (log) centroid size as a covari- variation among coastal and inland populations. Prickling ate, and tested for equal slopes among coastal and inland intensity was characterized as ‘high’, ‘medium’ or ‘low’ for groups using a MANCOVA as implemented in tpsRegr 1.38 each individual, which corresponds to prickling categories (Rohlf, 2011). We obtained size-corrected shape variables by 2–4 as described by Nolte et al. (2005b) and reflects the cov- performing a MANCOVA of Procrustes coordinates on (log) erage of the lateral body side with prickle-shaped, scale-like centroid size using a permutation test for each subgroup bony structures (Fig. 2). A Pearson’s chi-square test was (10,000 permutations), and used the regression residuals for

1628 Journal of Biogeography 42, 1626–1638 ª 2015 John Wiley & Sons Ltd Phylogeography reveals parallel evolution

Table 1 Sampling sites and genetic diversity estimates for microsatellite loci and mitochondrial DNA of the prickly sculpin (Cottus asper) from north-western North America (n, number of samples; HE, expected heterozygosity; AR, allelic richness; s, number of polymorphic sites; h, haplotype diversity; p, nucleotide diversity). Summary statistics for the three major mitochondrial DNA lineages ‘A’, ‘B’ and ‘C’ (excluding admixture populations in Auke Creek and Bella Coola River) are given at the end of the table.

Microsatellites mtDNA

Location Lat./Long. nHE AR n sh p

Coastal Alaska (Auke Creek, A) 58° 22.960N 14 0.60 3.54 14 5 0.85 ( 0.07) 0.0012 ( 0.0008) 134° 38.130W Bella Coola River (BC) 52° 22.020N 29 0.68 5.34 11 5 0.76 ( 0.1) 0.0012 ( 0.0008) 126° 44.150W Little Campbell River (LC) 49° 0.90 N, 122° 46.70 W 30 0.73 5.66 16 3 0.44 ( 0.14) 0.0004 ( 0.0004) Falls Creek (FC) 48° 34.020N 29 0.77 6.30 10 4 0.78 ( 0.13) 0.0008 ( 0.0006) 124° 21.250W Mosquito Lake (MOL) 53° 4.20 N, 132° 4.20 W 30 0.55 4.00 10 2 0.53 ( 0.18) 0.0003 ( 0.0003) Martins Lake (MAL) 52° 10.330N 51 0.70 5.24 128° 8.830W Martins Lake outlet (MALO) 52° 10.310N 31 0.69 5.18 128° 8.740W Tlell River (TR) 53° 34.880N 20 0.66 4.9 131° 56.610W Inland Meziadin Lake (MEL) 56° 5.140N 30 0.45 3.45 11 1 0.67 ( 0.12) 0.0003 ( 0.0003) 129° 18.310W Nimpo Lake (NIL) 52° 20.020N 12 0.15 1.80 12 1 0.45 ( 0.17) 0.0001 ( 0.0002) 125° 9.220W McLeod Lake (ML) 54° 59.40 N, 123° 2.10 W 13 0.34 3.10 13 3 0.54 ( 0.16) 0.0003 ( 0.0003) Peace River, AB (PA) 55° 55.10 N, 118° 36.40 W 12 0.35 2.99 10 2 0.2 ( 0.15) 0.0003 ( 0.0003) Peace River, BC (PB) 56° 3.90 N, 121°50.50 W 30 0.33 2.89 11 0 0.18 ( 0.14) 0 Harrison Lake (HL) 49° 32.190N 30 0.63 5.02 10 3 0.84 ( 0.1) 0.0005 ( 0.0005) 121° 52.450W Okanagan Lake (OL) 50° 12.50 N, 119° 27.90 W 14 0.46 3.66 13 3 0.53 ( 0.15) 0.0004 ( 0.0004) Lakelse Lake (LAL) 54° 22.60 N, 128° 33 ‘ W 30 0.41 3.30 mtDNA groups Lineage ‘A’ (MOL, MEL, NIL, ML, PA, PB) 67 8 0.44 ( 0.08) 0.0002 ( 0.0002) Lineage ‘B’ (FC, LC, HL) 36 6 0.66 ( 0.09) 0.0006 ( 0.0005) Lineage ‘C’ (OL) 13 3 0.53 ( 0.15) 0.0004 ( 0.0004) a size-corrected analysis (Berner, 2011). We used a jackknifed without making a priori assumptions about group assign- discriminant function analysis (DFA) on both size-corrected ments. Genetic diversity estimates were calculated for each and uncorrected data to determine whether a priori defined putative population and geographical mtDNA group, which coastal and inland populations could be distinguished by were derived from the SAMOVA analyses. body shape. A principal components analysis (PCA) on the In order to characterize genetic lineages and to estimate variance–covariance matrix of shape coordinates was con- phylogenetic relationships among them, phylogenetic trees ducted to visualize patterns of shape variability. Landmark were built from concatenated mtDNA sequences using a data are deposited under DRYAD (doi: 10.5061/ maximum-likelihood (ML) approach implemented in dryad.8ht04). RAxML 7.3.2 (Silvestro & Michalak, 2012) with 1000 boot- strap runs, and a Bayesian Inference (BI) approach in MrBa- yes 3.2 (Ronquist et al., 2012). The best-fit nucleotide Genetic diversity and phylogenetic analysis of substitution model was determined in jModetlTest 0.1.1 sequence data (Posada, 2008), based on the corrected Akaike information

We used Arlequin 3.5.1.2 (Excoffier et al., 2005) for calcu- criterion (AICc). A Bayes Factor test implemented in Tracer lating population pairwise FST values for mtDNA sequences 1.5 (Rambaut & Drummond, 2009) indicated a better fit of a (h; Weir & Cockerham, 1984) and genetic diversity estimates. codon-partitioned model (by gene and 3rd codon) as We performed a spatial analysis of molecular variance (SAM- opposed to an unpartitioned model. Analyses in MrBayes OVA) using the software samova 1.0 (Dupanloup et al., consisted of four Markov chain Monte Carlo (MCMC) 2002), to test for geographical groupings that are genetically chains running for 30 million generations (average standard homogeneous and maximally differentiated from each other, deviation of split frequencies < 0.01). Trees were sampled

Journal of Biogeography 42, 1626–1638 1629 ª 2015 John Wiley & Sons Ltd S. Dennenmoser et al.

(a) (b)

Figure 2 Overview of the geographical distribution of major mitochondrial haplotypes and prickling morphotypes of the prickly sculpin (Cottus asper) from north-western North America. (a) Haplotype network (mtDNA), with circles (c) (d) representing haplotypes and circle sizes corresponding to haplotype frequencies. Black shading refers to lineage ‘A’, grey shading to lineage ‘B’, and white to lineage ‘C’. (b) Map showing the geographical distribution of the three groups of mtDNA haplotypes. The geographical distribution of prickling categories is shown in (c): high (black), medium (grey) and low (white) (numbers indicate sample sizes). Prickling categories are depictured in (d), showing bleached and stained representatives of highly prickled, medium prickled and weakly prickled sculpins (from top to bottom). every 1000th generation with a default burn-in of 25% of phylogenetic relationships between haplotypes without samples. A 50% majority-rule consensus tree was built from enforcing bifurcating tree branching that may be inappropri- the remaining trees and used to calculate posterior probabili- ate for describing intraspecific patterns (Posada & Crandall, ties of tree branches. Convergence and stationarity of likeli- 2001). This method collapses sequences into haplotypes and hood scores was confirmed in Tracer. Average uncorrected calculates pairwise distances to construct connections among pairwise genetic distances (p-distances) within and between haplotypes where the probability of parsimony exceeds 95% mtDNA tree groups were calculated in mega 5.05 (Tamura (Templeton et al., 1992). et al., 2011). To better distinguish alternative colonization scenarios of Microsatellite amplification and analyses either single origin or multiple coastal–inland colonizations of highly prickled inland populations, we used a constrained We amplified a total of 19 microsatellite loci, including 18 tree analysis as implemented in paup* 4.0b1 (Swofford, microsatellite loci from Nolte et al. (2005a) and one newly 1998). After determining the most appropriate substitution designed marker ‘EDA1’. This locus is linked to the Ectodys- model in jModelTest, a ML search under the HKY+I model plasin (EDA) region in Cottus perifretum and was potentially in paup* was performed for each of three tree topologies: associated with prickling in C. asper, but showed no devia-

(1) unconstrained tree; (2) constrained tree with monophyly tions from neutrality in a FST outlier test or gametic phase of highly prickled inland populations (Okanagan Lake, Har- disequilibrium with the prickling phenotype (S.D., unpub- rison Lake, Nimpo Lake, McLeod Lake); and (3) constrained lished data). We genotyped 273 individuals from the 12 pop- tree with independent colonization histories reflected by ulations used for the sequence analyses (including the same monophyly of northern populations (Mosquito Lake, Mezia- individuals), and 132 individuals from four additional sites din Lake, McLeod Lake), southern populations (Falls Creek, (Fig. 1, Table 1): Lakelse Lake (n = 30), Martins Lake on Little Campbell River, Harrison Lake), as well as monophyly Campbell Island (n = 51), Martins Lake outlet (n = 31), and of Okanagan Lake assuming an independent colonization Tlell River (n = 20). Alleles were separated and genotyped history via the Columbia River watershed. Resulting tree via electrophoresis on a ABI 3500xl sequencer (Applied Bio- topologies were compared using a one-tailed Shimodaira– systems, Carlsbad, CA, USA), with allelic sizes (in base pairs) Hasegawa likelihood ratio test with full likelihood parameter determined by reference to an internal sizing standard in the estimation using a RELL approximation with 1000 bootstrap software GeneMapper 4.1 (Applied Biosystems). For primer replicates (Shimodaira & Hasegawa, 1999) sequences and PCR conditions, see Appendix S1. A haplotype network under statistical parsimony was con- We used an extended version of Fisher’s exact test for test- structed with tcs 1.13 (Clement et al., 2000) to visualize ing microsatellite loci for deviations from Hardy–Weinberg

1630 Journal of Biogeography 42, 1626–1638 ª 2015 John Wiley & Sons Ltd Phylogeography reveals parallel evolution equilibrium (HWE, 105 MCMC chain steps, 104 dememori- (Procrustes distance: 0.0088, P < 0.001). Classification accu- zation steps) and linkage disequilibrium (LDE, 105 MCMC racy in the cross-validated DFA was 81.9% for coastal and chain steps, 104 dememorization steps) with Arlequin 77% for inland sculpins for size-corrected data. Inland scul- 3.5.1.2 (Excoffier et al., 2005). Sequential Bonferroni pins showed a subtle tendency towards having a shorter and correction was applied to the P-values to correct for multiple higher caudal peduncle, a smaller eye and a slightly stouter comparisons in the same dataset. Pairwise FST values, as well snout than coastal sculpins (Appendix S2). These differences as observed (HO) and expected (HE) heterozygosities were among coastal and inland groups were also visible when per- calculated in Arlequin. Genetic diversity and allelic richness forming a DFA on highly prickled (category 4) sculpins only (mean number of alleles, corrected for smallest sample num- (not shown). ber) was calculated in fstat 2.9.3.2 (Goudet, 2001). We used a permutation test (15,000 permutations) in fstat to Mitochondrial DNA compare allelic richness and genetic diversity between coastal and inland populations. The possibility of null alleles, large The geographical distribution of mtDNA haplotypes did not allele drop-out and scoring errors was evaluated using support monophyly of highly prickled inland populations. micro-checker 2.2.3 (van Oosterhout et al., 2004). We detected 20 mitochondrial haplotypes (20 variable posi- An individual-based assignment approach implemented in tions, with five being parsimony informative) in 141 individ- structure 2.3.3 (Pritchard et al., 2000) was used to visual- uals from 12 populations. The substitution model HKY + I ize geographical clustering of microsatellite data and infer was chosen as the most appropriate model, as determined by genetic population structure. Following initial test runs, we the lowest AIC in jModelTest. used an admixture model to test for number of populations The SAMOVA supported five distinct geographical groups: (K) from 4 to 16. We applied a 100,000-iteration burn-in (1) Alaska (Auke Creek), (2) Bella Coola River, (3) Okana- followed by 900,000 iterations, and repeated each K four gan Lake, (4) a central British Columbian group (Meziadin times to ensure stability. The most likely number of genetic Lake, McLeod Lake, Mosquito Lake, Nimpo Lake, Peace clusters was determined by using the ΔK method (Evanno River), and (5) a southern coastal group (Falls Creek, Harri- et al., 2005) implemented in structure harvester (Earl & son Lake, Little Campbell River) (FCT = 0.638, P < 0.001). In vonHoldt, 2012). Additionally, individual-based chord dis- total, 63.88% of the variance was explained by differences tance measures (DC; Cavalli-Sforza & Edwards, 1967) were among geographical regions (Table 2). Two of the five calculated and used for the construction of a 1000 times SAMOVA groups [Alaska (Auke Creek) and Bella Coola bootstrapped neighbour-joining (NJ) tree in populations River] exhibited pronounced admixture of haplotypes from 1.2.32 (Langella, 1999). Concordance of genetic variability different mtDNA haplotype clusters (Fig. 2a,b). Consistently, between microsatellite data and mtDNA groups was assessed mtDNA haplotype and nucleotide diversity was higher in with an analysis of molecular variance (AMOVA) using the Auke Creek and Bella Coola River than in the three main three mtDNA groups as a grouping variable, and excluding groups (Table 1), supporting their potential role as admix- genetically admixed populations [Auke Creek (Alaska) and ture zones. Bella Cool River, see Results]. For comparison, a second Highly prickled inland populations were present in all AMOVA was conducted using the clusters inferred from the three major mtDNA groups, which was also evident in the structure analysis as a grouping variable. Genotype data mtDNA haplotype network (Fig. 2a,b). Here, major groups are deposited under DRYAD (doi: 10.5061/dryad.8ht04). are visible as star-like clusters around a frequent haplotype and denoted as lineages A, B and C. Importantly, lineages ‘A’ (central BC) and ‘B’ (southern BC) include highly prick- RESULTS led inland populations as well as coastal populations that are weakly to moderately prickled (Fig. 2c). Morphology The same genetic groups could be resolved in a phyloge- Consistent with Krejsa’s (1965) findings, prickling intensity netic tree, and exhibited identical topologies and similar pos- differed significantly between inland and coastal sculpins terior probability support for both maximum likelihood (v2 = 208.2, d.f. = 2, P < 0.001). Inland populations were (ML) and Bayesian inference (BI) constructions (see Appen- dominated by highly prickled sculpins, whereas coastal scul- dix S3). Uncorrected average pairwise distances within and pins exhibited high frequencies of medium and low prickling between major tree clusters were low, ranging from 0.09% to categories (Fig. 2c, Appendix S2). Allometric effects of cen- 0.21% between clusters, and from 0.02% to 0.03% within troid size on body shape variation were significant, but each cluster. The constrained tree analysis supported an explained only 3.91% (coastal: MANCOVA, 10,000 permuta- unconstrained tree (negative log-likelihood, ln tions, P < 0.001) and 3.88% (inland: MANCOVA, 10,000 L = 2363.49) over monophyly of highly prickled inland pop- permutations, P < 0.001) of shape variation, respectively. A ulations (ln L = 2405.18, Shimodaira–Hasegawa (SH) test, discriminant function analysis (DFA) on size-corrected and P = 0.023). In contrast, a constrained tree topology mirror- uncorrected data gave similar results and indicated a small, ing multiple independent coastal–inland colonizations in but significant difference between coastal and inland groups northern (central BC) and southern populations was not

Journal of Biogeography 42, 1626–1638 1631 ª 2015 John Wiley & Sons Ltd S. Dennenmoser et al. significantly different from an unconstrained ML tree (ln microsatellites (Fig. 3c). The AMOVA of microsatellite L = 2363.6, SH-test, P = 0.66). genetic variation revealed that most of the variation existed within populations (61.3%), while grouping according to the three mtDNA lineages explained 20.25% of the variation, Microsatellite loci overall indicating a high degree of concordance between data In 405 individuals, the degree of polymorphism in the 19 sets (Table 3). Similarly, the AMOVA using structure K microsatellite loci was highly variable, with the total number clusters as grouping variable also revealed elevated within- of alleles ranging from 2 to 31. Allelic richness (standardized population variability (71%), with genetic clusters explaining to a minimum sample size of seven) of polymorphic loci 26.25% of the variation (Table 3). ranged from 1.8 to 6.3, and gene diversity ranged from 0.15 to 0.77 (Table 1). Significant departures from HWE were DISCUSSION found for locus Cott153 in Okanagan Lake and Martins Lake, and for locus CottE30 in Martins Lake outlet. Evidence for In this study, we presented phenotypic and multi-locus linkage disequilibrium was found in 38 (out of 2736) com- genetic data to evaluate the phylogeographical history of the parisons, and potential null alleles occurred in nine locus- prickly sculpin (Cottus asper) and the evolutionary context of population pairs. We did not exclude any loci from the a geographical discontinuity in prickling intensity described analyses because no consistent deviations across the majority by Krejsa (1965). The pronounced coastal–inland split of of sites were found. Coastal populations showed overall prickling phenotypes was accompanied by weak, yet signifi- higher genetic diversity estimates than inland populations, cant differentiation in body shape, which supports divergent with an average allelic richness of 5.04 and 3.23 (fstat per- evolutionary trajectories in alternative environments. Our mutation test, two-sided, P = 0.004), respectively, and a gene finding that highly prickled inland phenotypes are genetically diversity of 0.665 and 0.422 (P = 0.001), respectively. Signifi- distinct from coastal populations but associated with multi- cant pairwise FST values ranged from 0.038 (Falls Creek– ple genetic lineages implies parallel evolution subsequent to Martins Lake outlet) to 0.570 (McLeod Lake–Okanagan wide coastal dispersal and repeated coastal to inland coloni- Lake). zations, rather than a single origin from a southern inland Genetic population structure was more pronounced under refugium as has been assumed since Krejsa (1965). microsatellites compared to mitochondrial DNA, yet largely concordant with the inferred mtDNA groups. Clustering Post-glacial colonization results in structure and a highest ΔK value (Evanno et al., 2005) supported 11 clusters by grouping Peace River, McLe- The widespread occurrence of northern and southern mito- od Lake and Nimpo Lake in one cluster, Martins Lake and chondrial haplotypes along the coast underlines the impor- outlet in another cluster, and forming a third cluster includ- tance of coastal colonization routes. Inland routes through ing Little Campbell River and Falls Creek (Fig. 3a,b). These ephemeral connections between river systems may explain clusters were consistent with a NJ- tree of DC distances from the occurrence of shared coastal and inland haplotypes in

Table 2 Results of spatial analyses of molecular variance (SAMOVA) of mitochondrial DNA assessed for grouping schemes of 2–9 groups of prickly sculpin (Cottus asper) populations from north-western North America (*P < 0.001).

Sum of squares Variance components Percentage of variation a Grouping schemes d.f. (among groups) (among groups) (among groups) FCT

2 groups 1 46.23 0.81 59.04 0.59* 3 groups 2 61.18 0.73 62 0.62* 4 groups 3 66.17 0.68 62.92 0.629* 5 groups 4 67.83 0.69 63.88 0.638* 6 groups 5 68.47 0.67 63.46 0.634* 7 groups 6 69.6 0.63 63.01 0.63* 8 groups 7 69.76 0.62 62.62 0.626* 9 groups 8 69.5 0.59 61.36 0.614* aSAMOVA grouping schemes were as follows: 2 groups: (Falls Creek, Harrison Lake, Little Campbell River), (Alaska, Bella Coola, Meziadin, McLeod, Mosquito, Nimpo, Okanagan, Peace River); 3 groups: (Falls Creek, Harrison Lake, Little Campbell River), (Bella Coola, Okanagan), (Alaska, Meziadin, McLeod, Mosquito, Nimpo, Peace); 4 groups: (Alaska, Bella Coola), (Okanagan), (Falls Creek, Harrison Lake, Little Campbell River), (Meziadin, McLeod, Mosquito, Nimpo, Peace); 5 groups: (Alaska), (Bella Coola), (Okanagan), (Falls Creek, Harrison Lake, Little Camp- bell River), (Meziadin, McLeod, Mosquito, Nimpo, Peace River); 6 groups: (Alaska), (Bella Coola), (Okanagan), (Harrison), (Falls Creek, Little Campbell River), (Meziadin, McLeod, Mosquito, Nimpo, Peace); 7 groups: (Alaska), (Bella Coola), (Okanagan), (Harrison), (Falls Creek, Little Campbell River), (Meziadin), (McLeod, Mosquito, Nimpo, Peace); 8 groups: (Alaska), (Bella Coola), (Okanagan), (Harrison), (Falls Creek), (Lit- tle Campbell River), (Meziadin), (McLeod, Mosquito, Nimpo, Peace); 9 groups: (Alaska), (Bella Coola), (Okanagan), (Harrison), (Falls Creek), (Little Campbell River), (Meziadin), (Mosquito), (McLeod, Nimpo, Peace).

1632 Journal of Biogeography 42, 1626–1638 ª 2015 John Wiley & Sons Ltd Phylogeography reveals parallel evolution

(a)

(b)

(c)

Figure 3 Genetic structuring of the prickly sculpin (Cottus asper) in north-western North America. (a) Log-likelihood plot supporting K = 11. (b) structure plot showing assignment of 405 individuals of Cottus asper from 16 sampling sites into 11 clusters. (c) Neighbour-joining tree based on DC distances among populations based on 19 microsatellite loci (numbers indicating 10009 bootstrap support). now disconnected inland watersheds such as Peace River. were lower in inland populations, which is presumably a However, the lack of shared haplotypes among southern and consequence of founder effects associated with coastal to northern inland regions does not support far northward dis- inland colonizations, as well as larger effective population persal of southern inland haplotypes. Gene flow among sizes and more gene flow among coastal populations (Tea- coastal populations of C. asper is facilitated by a high dis- cher et al., 2011). persal capability of salt-tolerant, planktonic larval stages Inland dispersal was facilitated during deglaciation, when (Krejsa, 1965; McPhail, 2007), which is well known from drainage of large proglacial lakes allowed crossing watersheds many amphidromous species including sculpins (Whiteley through temporal connections between river systems such as et al., 2009; Dennenmoser et al., 2014). Accordingly, second- Columbia/Fraser, Fraser/Peace, Nass/Skeena or Skeena/Peace ary contact between coastal northern and southern genetic rivers (McPhail & Carveth, 1993). Similar to other species lineages may have contributed to the occurrence of elevated (e.g. pygmy whitefish, Prosopium coulterii; Witt et al., 2011), genetic diversities and the admixture of haplotypes in Auke the prickly sculpin may have accessed the Peace River Creek (Southeast Alaska) and Bella Coola River (central BC). watershed via the Fraser and Skeena rivers (McPhail, 2007). Genetic diversity did not decrease towards northern latitudes, Far inland dispersal and an exclusively southern inland as would be expected if founder events accompanied a long origin of all highly prickled inland sculpins via multiple period of northward colonization. Instead, genetic diversities inland watershed crossings is unlikely, given the high genetic

Journal of Biogeography 42, 1626–1638 1633 ª 2015 John Wiley & Sons Ltd S. Dennenmoser et al.

Table 3 Analyses of molecular variance (AMOVAs) for microsatellite data of the prickly sculpin (Cottus asper) sampled in north- western North America (* P < 0.001, 10,100 permutations).

Sum of Variance Percentage of Source of variation d.f. squares component total variation Fixation indices

3 groups Among groups 2 326.2 0.76 20.25 FCT = 0.2* (mtDNA lineages)

Among populations, within 10 387.4 0.74 19.71 FSC = 0.25* groups

Among individuals, within 302 667.97 0 0 FIS = 0 populations

Within individuals 315 726.5 2.3 61.3 FIT = 0.39* 11 groups Among groups 10 887.94 1.0 26.25 FCT = 0.26* (structure clusters)

Among populations, within 7 65.5 0.13 3.45 FSC = 0.05* groups

Among individuals, within 442 1177.5 0 0 FIS = 0 populations

Within individuals 460 1250 2.72 71 FIT = 0.29* differentiation among northern and southern inland areas. ‘Okanagan group’ in the Okanagan Valley that connects to a Overall, our data highlight the importance of large coastal glacial refugium in the Columbia River (Larson et al., 2012). rivers such as the Skeena and Fraser rivers for the coloniza- tion of inland freshwater habitats, such as the highly prickled Comparison of mitochondrial and nuclear genetic populations present in Lakelse Lake or Harrison Lake. This data suggests that post-glacial colonization of inland populations has most likely advanced by separate coastal–inland coloniza- Analyses of microsatellite loci revealed pronounced popula- tions across different genetic lineages, resulting in parallel tion structuring, which was overall consistent with the char- evolution of highly prickled inland populations. acterization of the three mtDNA lineages and two mtDNA This interpretation contradicts the previous assumption of admixture zones. Genetic divergence among mtDNA lineages monophyly of highly prickled inland sculpins, associated has not been eroded by high migration rates during more with only southern glacial refugia and post-glacial dispersal recent times, as has been suggested for other species (e.g. from southern towards northern areas. Instead, our results Atlantic herring, Clupea harengus; Gaggiotti et al., 2009). show a pattern of mitochondrial haplotypes forming three Moreover, the structure in the microsatellite data indicates genetic clusters, albeit weakly differentiated but consistent that male-biased dispersal or female-biased philopatry are with divergence among three glacial lineages. These patterns not likely major drivers of mtDNA structuring. The finding could suggest the presence of a northern coastal refugium, as of additional, more fine-scaled genetic structure revealed by has been suggested for other species in the Queen Charlotte microsatellites is expected in the absence of high migration Islands and the Alexander Archipelago area, such as three- rates and promoted by the relatively higher microsatellite spine stickleback (Gasterosteus aculeatus), rainbow trout (On- mutation rates (Koskinen et al., 2002). While elevated muta- corhynchus mykiss) and salmon (Oncorhynchus nerka, tion rates and homoplasy render microsatellite markers less Oncorhynchus kisutch) (O’Reilly et al., 1993; Wood et al., informative for more ancient divergences, and have fre- 1994; McCusker et al., 2000; Smith et al., 2001). While our quently failed to find structure in the presence of mitochon- limited sampling scheme and possibly confounding effects of drial genetic structuring (Zink, 2010), they still may be post-glacial admixture does not allow pinpointing the exact useful to support phylogeographical patterns that reflect location of a northern coastal refugium, the microsatellite more recent (late Pleistocene) divergences in some taxa genetic structure among northern coastal populations could (Koskinen et al., 2002; H€anfling et al., 2002). indicate the presence of a mosaic of separate, partly discon- An unexpected discordance between nuclear and mito- nected coastal refugia (‘refugia within refugia’; Shafer et al., chondrial DNA was found in a pronounced genetic differen- 2010). Regarding southern refugia, a strong genetic break tiation between northern coastal and inland populations in between southern coastal (Fraser Valley) and inland (Okana- the microsatellite loci but not the mtDNA, which is domi- gan Lake) populations supports the previous assumption of nated by the most common northern haplotype (lineage ‘A’). separate coastal and inland refugia in this region (Krejsa, This could suggest mitochondrial introgression following 1965; McPhail, 2007). A similar coastal–inland break has secondary contact between glacial lineages, as has been frequently been attributed to refugium areas located in the reported in other studies (Nolte et al., 2005b; Redenbach & Chehalis River and lower Columbia River valleys (e.g. McP- Taylor, 2003; Toews & Brelsford, 2012). A possible double- hail & Carveth, 1993; Taylor et al., 1999), including an colonization scenario allowing mitochondrial introgression

1634 Journal of Biogeography 42, 1626–1638 ª 2015 John Wiley & Sons Ltd Phylogeography reveals parallel evolution upon secondary contact could be post-glacial dispersal of the CONCLUSIONS southern-coastal lineage ‘B’ into the Peace River watershed via the , followed by the introgressive invasion of In summary, we provide an example of a parallel phenotype the northern mtDNA lineage ‘A’ via the . We that has evolved across multiple genetic lineages, and repeat- recognize the speculative nature of such a double-coloniza- edly became the predominant phenotype after multiple tion scenario, and propose that future studies should test for inland colonizations via coastal rivers. Future studies should introgression of a northern mitochondrial haplotype that test whether mitochondrial introgression between two lin- might have replaced a southern haplotype (Glemet et al., eages may have facilitated colonization of northern inland 1998). In general, selection on the mitochondrial genome regions, and whether standing genetic variation of the highly may be common (Dowling et al., 2008), and selective sweeps admixed ancestral coastal populations could have resulted in could play an important yet frequently overlooked role in lineage-specific genetic polymorphisms underlying highly shaping phylogeographical patterns (Rato et al., 2011). prickled phenotypes. While the adaptive function of parallel phenotypes often remains to be experimentally tested, the possibility of alternative parallel evolutionary pathways Parallel evolution (genetic constraints, ancient selective regimes) deserves more Our results suggest that multiple genetic lineages have con- attention. tributed to the occurrence of highly prickled inland popula- tions following colonizations via coastal rivers, which indicates parallel phenotypic evolution (Elmer & Meyer, ACKNOWLEDGEMENTS 2011). The high variability in prickling intensity in the pre- We are grateful to David Tallmon (University of Alaska sumed ancestral coastal populations may indicate a role for Southeast) for providing fin-clip samples and to Susan standing genetic variation rather than new mutations as a Cousineau, Andrew Rezansoff, Stevi Vanderzwan and Lisa source of parallel evolution. This distinction is important, Duke for field assistance. Comments from Daniel Ruzzante because consequences for genomic architecture and speed of and two anonymous referees improved the manuscript. We adaptive divergence can differ dramatically (Elmer & Meyer, thank the Alberta Innovates Technology Futures for a gradu- 2011; Rogers et al., 2013). For example, the availability of ate scholarship to S.D. This research was supported by standing genetic variation allows faster fixation of alleles NSERC Discovery Grants (S.M.R. and S.M.V.), a European compared to low-frequency new mutations, which may Research Council starting grant (A.W.N.) and an Alberta increase the probability of finding parallel phenotypes in Innovates Technology Futures New Faculty Award (S.M.R.). recently deglaciated, post-glacial environments such as those found in north-western North America (e.g. threespine stick- leback; Colosimo et al., 2004). If high gene flow promotes REFERENCES the wide dispersal of pre-existing adaptive alleles, parallel Avise, J.C., Arnold, J., Ball, R.M., Bermingham, E., Lamb, T., evolution can rapidly occur in derived populations across a Neigel, J.E., Reeb, C.A. & Saunders, N.C. (1987) Intraspe- large geographical area as has been shown for the parallel cific phylogeography: the mitochondrial DNA bridge evolution of freshwater ecotypes in threespine stickleback between population genetics and systematics. Annual (Schluter & Conte, 2009). Similarly, our study suggests paral- Review of Ecology and Systematics, 18, 489–522. lel evolution of highly prickled inland groups following colo- Bernatchez, L., Renaut, S., Whiteley, A.R., Campbell, D., nization through coastal rivers, which is based on standing Derome,^ N., Jeukens, J., Landry, L., Lu, G., Nolte, A.W., genetic variation and wide dispersal among the ancestral Østbye, K., Rogers, S.M. & St-Cyr, J. (2010) On the origin coastal populations. The parallel fixation of a highly prickled of species: insights from the ecological genomics of white- phenotype across multiple genetic lineages suggests a poten- fish. Philosophical Transactions of the Royal Society B: Bio- tial role of natural selection (Schluter, 2000), which assumes logical Sciences, 365, 1783–1800. a heritable genetic basis of prickling. This seems plausible Berner, D. (2011) Size correction in biology: how reliable are given the finding of a major prickling QTL in Cottus perifre- approaches based on (common) principal component tum (Cheng, 2013). While we currently do not know analysis? Oecologia, 166, 961–971. whether prickling is adaptive, the finding of weak differentia- Cavalli-Sforza, L.L. & Edwards, A.W.F. (1967) Phylogenetic tion in body shape among coastal and inland groups besides analysis. Models and estimation procedures. The American prickling intensity also suggests that inland freshwater habi- Journal of Human Genetics, 19, 233–257. tats represent a different selective environment. Further stud- Cheng, J. (2013) Genomic mapping of speciation-related traits ies are needed to test for possible functions of prickling, and in hybridizing sculpins (Cottus). PhD Thesis, Christian- to reveal whether parallel evolution of prickling phenotypes Albrechts-Universit€at, Kiel. in C. asper is based on a single mutation in the EDA path- Clement, M., Posada, D. & Crandall, K.A. (2000) TCS: a way as suggested by Cheng (2013) for European Cottus,or computer program to estimate gene genealogies. Molecular whether historical differentiation has diversified genetic vari- Ecology, 9, 1657–1660. ation underlying highly prickled phenotypes.

Journal of Biogeography 42, 1626–1638 1635 ª 2015 John Wiley & Sons Ltd S. Dennenmoser et al.

Colosimo, P.F., Peichel, C.L., Nereng, K., Blackman, B.K., Hewitt, G.M. (2004) Genetic consequences of climatic oscil- Shapiro, M.D., Schluter, D. & Kingsley, D.M. (2004) The lations in the Quaternary. Philosophical Transactions of the genetic architecture of parallel armor plate reduction in Royal Society B: Biological Sciences, 359, 183–195. threespine sticklebacks. PLoS Biology, 2, e109. Klingenberg, C.P. (2011) MorphoJ: an integrated software Dennenmoser, S., Rogers, S.M. & Vamosi, S.M. (2014) package for geometric morphometrics. Molecular Ecology Genetic population structure in prickly sculpin (Cottus Resources, 11, 353–357. asper) reflects isolation by environment between two life Knowles, L.L. (2009) Statistical phylogeography. Annual history ecotypes. Biological Journal of the Linnean Society, Review of Ecology, Evolution, and Systematics, 40, 593–612. 113, 943–957. Koli, L. (1969) Geographical variation of Cottus gobio L. Dowling, D.K., Friberg, U. & Lindell, J. (2008) Evolutionary (Pisces, Cottidae) in northern Europe. Annales Zoologici implications of non-neutral mitochondrial genetic varia- Fennici, 6, 353–390. tion. Trends in Ecology and Evolution, 23, 546–554. Koskinen, M.T., Nilsson, J., Veselov, A.J., Potutkin, A.G., Dupanloup, I., Schneider, S. & Excoffier, L. (2002) A simu- Ranta, E. & Primmer, C.R. (2002) Microsatellite data lated annealing approach to define the genetic structure of resolve phylogeographic patterns in European grayling, populations. Molecular Ecology, 11, 2571–2581. Thymallus thymallus, Salmonidae. Heredity, 88, 391–401. Earl, D.A. & vonHoldt, B.M. (2012) STRUCTURE HAR- Krejsa, R.J. (1965) The systematics of the prickly sculpin, Cot- VESTER: a website and program for visualizing STRUC- tus asper: an investigation of genetic and non-genetic varia- TURE output and implementing the Evanno method. tion within a polytypic species. PhD Thesis, University of Conservation Genetic Resources, 4, 359–361. British Columbia, Vancouver. Elmer, K.R. & Meyer, A. (2011) Adaptation in the age of Langella, O. (1999) Populations 1.2.32. Available at: http:// ecological genomics: insights from parallelism and conver- bioinformatics.org/~tryphon/populations/. gence. Trends in Ecology and Evolution, 26, 298–306. Larkin, M.A., Blackshields, G., Brown, N.P., Chenna, R., Elmer, K.R., Fan, S., Kusche, H., Spreitzer, M.L., Kautt, A.F., McGettigan, P.A., McWilliam, H., Valentin, F., Wallace, Franchini, P. & Meyer, A. (2014) Parallel evolution of Nic- I.M., Wilm, A., Lopez, R., Thompson, J.D., Gibson, T.J. & araguan crater lake cichlid fishes via non-parallel routes. Higgins, D.G. (2007) ClustalW and Clustal X version 2.0. Nature Communications, 5, 5168. Bioinformatics, 23, 2947–2948. Evanno, G., Regnaut, S. & Goudet, J. (2005) Detecting the Larson, E.R., Abbott, C.L., Usio, N., Azuma, N., Wood, K.A., number of clusters of individuals using the software Herborg, L.M. & Olden, J.D. (2012) The signal crayfish is structure: a simulation study. Molecular Ecology, 14, not a single species: cryptic diversity and invasions in the 2611–2620. Pacific Northwest range of Pacifastacus leniusculus. Fresh- Excoffier, L., Laval, G. & Schneider, S. (2005) Arlequin ver. water Biology, 57, 1823–1838. 3.0: an integrated software package for population genetics Lu, G. & Bernatchez, L. (1999) Correlated trophic specializa- data analysis. Evolutionary Bioinformatics Online, 1,47–50. tion and genetic divergence in sympatric lake whitefish ec- Freyhof, J., Kottelat, M. & Nolte, A. (2005) Taxonomic otypes (Coregonus clupeaformis): support for the ecological diversity of European Cottus with description of eight new speciation hypothesis. Evolution, 53, 1491–1505. species (Teleostei: Cottidae). Ichthyological Explorations of McAllister, D.E. & Lindsey, C.C. (1961) Systematics of the Freshwaters, 16, 107–172. freshwater sculpins (Cottus) of British Columbia. Bulletin Gaggiotti, O.E., Bekkevold, D., Jørgensen, H.B.H., Foll, M., of the National Museum of Canada, Contributions to Zool- Carvalho, G.R., Andre, C. & Ruzzante, D.E. (2009) Disen- ogy (1959), 172,66–89. tangling the effects of evolutionary, demographic, and McCusker, M.R., Parkinson, E. & Taylor, E.B. (2000) Mito- environmental factors influencing genetic structure of nat- chondrial DNA variation in rainbow trout (Oncorhynchus ural populations: Atlantic herring as a case study. Evolu- mykiss) across its native range: testing biogeographical tion, 63, 2939–2951. hypotheses and their relevance to conservation. Molecular Glemet, H., Blier, P. & Bernatchez, L. (1998) Geographical Ecology, 9, 2089–2108. extent of Arctic char (Salvelinus alpinus) mtDNA intro- McPhail, J.D. (2007) The freshwater fishes of British Colum- gression in brook char populations (S. fontinalis) from bia. University of Alberta Press, Alberta. eastern Quebec. Canada. Molecular Ecology, 7, 1655–1662. McPhail, J.D. & Carveth, R. (1993) A foundation for conser- Goudet, J. (2001) FSTAT, version 2.9.3: a program to estimate vation: the nature and origin of the freshwater fish fauna of and test gene diversities and fixation indices. Available at: British Columbia. Queen’s Printer for British Columbia, http://www2.unil.ch/popgen/softwares/fstat.htm. Victoria. Griekspoor, A. & Groothuis, T. (2006) 4Peaks, version 1.7. Nolte, A.W., Stemshorn, K.C. & Tautz, D. (2005a) Direct Available at: http://mekentosj.com/4peaks. cloning of microsatellite loci from Cottus gobio through a H€anfling, B., Hellemans, B., Volckaert, F.A. & Carvalho, G.R. simplified enrichment procedure. Molecular Ecology Notes, (2002) Late glacial history of the cold-adapted freshwater 5, 628–636. fish Cottus gobio, revealed by microsatellites. Molecular Nolte, A.W., Freyhof, J., Stemshorn, K.C. & Tautz, D. Ecology, 11, 1717–1729. (2005b) An invasive lineage of sculpins, Cottus sp. (Pisces,

1636 Journal of Biogeography 42, 1626–1638 ª 2015 John Wiley & Sons Ltd Phylogeography reveals parallel evolution

Teleostei) in the Rhine with new habitat adaptations has sheds new light on the phylogeography of northwestern originated from hybridization between old phylogeograph- North America. Molecular Ecology, 19, 4589–4621. ic groups. Proceedings of the Royal Society B: Biological Sci- Shikano, T., Shimada, Y., Herczeg, G. & Meril€a, J. (2010) ences, 272, 2379–2387. History vs. habitat type: explaining the genetic structure of van Oosterhout, C., Hutchinson, W.F., Wills, D.P.M. & European nine-spined stickleback (Pungitius pungitius) Shipley, P. (2004) micro-checker: software for identify- populations. Molecular Ecology, 19, 1147–1161. ing and correcting genotyping errors in microsatellite data. Shimodaira, H. & Hasegawa, M. (1999) Multiple compari- Molecular Ecology Notes, 4, 535–538. sons of log-likelihoods with applications to phylogenetic O’Reilly, P., Reimchen, T.E., Beech, R. & Strobeck, C. (1993) inference. Molecular Biology and Evolution, 16, Mitochondrial DNA in Gasterosteus and Pleistocene glacial 1114–1116. refugium on the Queen Charlotte Islands, British Colum- Silvestro, D. & Michalak, I. (2012) raxmlGUI: a graphical bia. Evolution, 47, 678–684. front-end for RaxML. Organisms Diversity and Evolution, Posada, D. (2008) Phylogenetic model averaging. Molecular 12, 335–337. Biology and Evolution, 25, 1253–1256. Smith, C.T., Nelson, R.J., Wood, C.C. & Koop, B.F. (2001) Posada, D. & Crandall, K.A. (2001) Intraspecific phylogenet- Glacial biogeography of North American coho salmon ics: trees grafting into networks. Trends in Ecology and (Oncorhynchus kisutch). Molecular Ecology, 10, 2775–85. Evolution, 16,37–45. Swofford, D.L. (1998) PAUP*: phylogenetic analysis using par- Pritchard, J.K., Stephens, M. & Donnelly, P. (2000) Inference simony (*and other methods). Version 4. Sinauer Associ- of population structure using multilocus genotype data. ates, Sunderland, MA. Genetics, 155, 945–959. Tamura, K., Peterson, D., Peterson, N., Stecher, G., Nei, M. Prunier, J., Gerardi, S., Laroche, J., Beaulieu, J. & Bousquet, & Kumar, S. (2011) MEGA5: molecular evolutionary J. (2012) Parallel and lineage-specific molecular adaptation genetics analysis using maximum likelihood, evolutionary to climate in boreal black spruce. Molecular Ecology, 21, distance and maximum parsimony methods. Molecular 4270–4286. Biology and Evolution, 28, 2731–2739. Rambaut, A. & Drummond, A.J. (2009) Tracer v1.5. Avail- Taylor, E.B., Pollard, S. & Louie, D. (1999) Mitochondrial able at: http://beast.bio.ed.ac.uk/Tracer. DNA variation in bull trout (Salvelinus confluentus) from Rato, C., Carranza, S. & Harris, D.J. (2011) When selection northwestern North America: implications for zoogeogra- deceives phylogeographic interpretation: the case of the phy and conservation. Molecular Ecology, 8, 155–170. Mediterranean house gecko, Hemidactylus turcicus (Linna- Teacher, A.G.F., Shikano, T., Karjalainen, M.E. & Meril€a, J. eus, 1758). Molecular Phylogenetics and Evolution, 58, 365– (2011) Phylogeography and genetic structuring of Euro- 373. pean ninespined sticklebacks (Pungitius pungitius) – mito- Redenbach, Z. & Taylor, E.B. (2003) Evidence for bimodal chondrial DNA evidence. PLoS ONE, 6, e19476. hybrid zones between two species of char (Pisces: Salveli- Templeton, A.R., Crandall, K.A. & Sing, C.F. (1992) A cladis- nus) in northwestern North America. Journal of Evolution- tics analysis of phenotypic associations with haplotypes ary Biology, 16, 1135–1148. inferred from restriction endonuclease mapping and DNA Rogers, S.M., Mee, J.A. & Bowles, E. (2013) The conse- sequence data. III. Cladogram estimation. Genetics, 132, quences of genomic architecture on ecological speciation 619–633. in postglacial fishes. Current Zoology, 59,53–71. Toews, D.P.L. & Brelsford, A. (2012) The biogeography of Rohlf, F.J. (2009a) tpsDig 2.14. Available at: http:// mitochondrial and nuclear discordance in animals. Molec- life.bio.sunysb.edu/morph/. ular Ecology, 21, 3907–3930. Rohlf, F.J. (2009b) tpsUtil version 1.44. Available at: http:// Weir, B.S. & Cockerham, C.C. (1984) Estimating F-statistics life.bio.sunysb.edu/morph/. for the analysis of population structure. Evolution, 38, Rohlf, F.J. (2011) tpsRegr 1.38. Available at: http:// 1358–1370. life.bio.sunysb.edu/morph/. Whiteley, A.R., Gende, S.M., Gharret, A.J. & Tallmon, D.A. Ronquist, F., Teslenko, M., van der Mark, P., Ayres, D., Dar- (2009) Background matching and color-change plasticity ling, A., Hohna,€ S., Larget, B., Liu, L., Suchard, M.A. & in colonizing freshwater sculpin populations following Huelsenbeck, J.P. (2012) MrBayes 3.2: efficient Bayesian rapid deglaciation. Evolution, 63, 1519–1529. phylogenetic inference and model choice across a large Witt, J.S., Zemlak, R. & Taylor, E.B. (2011) Phylogeography model space. Systematic Biology, 61, 539–542. and the origins of range disjunctions in a north temperate Schluter, D. (2000) The ecology of adaptive radiation. Oxford fish, the pygmy whitefish (Prosopium coulterii), inferred University Press, Oxford, UK. from mitochondrial and nuclear DNA sequence analysis. Schluter, D. & Conte, G.L. (2009) Genetics and ecological Journal of Biogeography, 38, 1557–1569. speciation. Proceedings of the National Academy of Sciences Wood, C.C., Riddell, B.E., Rutherford, D.T. & Withler, R.E. USA, 106, 9955–9962. (1994) Biochemical genetic survey of sockeye salmon (On- Shafer, A.B.A., Cullingham, C.I., Cote, S.D. & Coltman, corhynchus nerka) in Canada. Canadian Journal of Fisheries D.W. (2010) Of glaciers and refugia: a decade of study and Aquatic Sciences, 51, 114–131.

Journal of Biogeography 42, 1626–1638 1637 ª 2015 John Wiley & Sons Ltd S. Dennenmoser et al.

Zink, R.M. (2010) Drawbacks with the use of microsatellites are deposited under GenBank (accession numbers in phylogeography: the song sparrow Melospiza melodia as KR024578-KR024637). a case study. Journal of Avian Biology, 41,1–7.

BIOSKETCH SUPPORTING INFORMATION

Additional Supporting Information may be found in the Stefan Dennenmoser is broadly interested in the online version of this article: intersection of biogeography, natural history and local adap- tation and its implications for the genetic basis of adaptive Appendix S1 Primer sequences and PCR protocols. change. This study was part of his PhD research on the Appendix S2 Prickling and body shape differences between phylogeography of Cottus asper in north-western North inland and coastal sculpins. America. Appendix S3 Fifty per cent majority-rule consensus tree of mtDNA sequences. Author contributions: S.D., S.M.V., A.W.N. and S.M.R. designed the study. S.D. performed the research and analysed the data, and all authors contributed to writing the paper. DATA ACCESSIBILITY

Landmark and genotype data are deposited under DRYAD Editors: John Stewart and Robert Whittaker (doi: 10.5061/dryad.8ht04), and mitochondrial haplotypes

1638 Journal of Biogeography 42, 1626–1638 ª 2015 John Wiley & Sons Ltd