Integrative Zoology 2016; 11: 162–172 doi: 10.1111/1749-4877.12186

1 SHORT COMMUNICATION 1 2 2 3 3 4 4 5 5 6 6 7 Is the ballan wrasse (Labrus bergylta) two species? Genetic 7 8 8 9 analysis reveals within-species divergence associated with plain 9 10 10 11 and spotted morphotype frequencies 11 12 12 13 13 14 María QUINTELA,1 Elin Annie DANIELSEN,1 Lua LOPEZ,2 Rodolfo BARREIRO,2 Terje 14 15 15 1 1,3,4 5 1 16 SVÅSAND, Halvor KNUTSEN, Anne Berit SKIFTESVIK and Kevin A. GLOVER 16 17 1Institute of Marine Research, Bergen, Norway, 2Grupo de investigación BIOCOST, University of A Coruña, Spain, 3Centre 17 18 18 for Ecological and Evolutionary Synthesis, Department of Biology, University of Oslo, Oslo, Norway, 4University of Agder, 19 19 Kristiansand, Norway and 5Austevoll Research Station, Storebø, Norway 20 20 21 21 22 22 23 Abstract 23 24 24 25 The ballan wrasse (Labrus bergylta) is a marine fish belonging to the family Labridae characterized by 2 main 25 26 morphotypes that occur in sympatry: spotty and plain. Previous studies have revealed differences in their 26 27 life-history traits, such as growth and maturation; however, the genetic relationship between forms is present- 27 28 ly unknown. Using 20 recently developed microsatellite markers, we conducted a genetic analysis of 41 and 48 28 29 spotty and plain ballan wrasse collected in Galicia (northwest Spain). The 2 morphotypes displayed highly sig- 29 30 nificant genetic differences to each other (FST = 0.018, P < 0.0001). A similar degree of genetic differentiation (FST 30 31 = 0.025, P < 0.0001) was shown using the STRUCTURE clustering approach with no priors at K = 2. In this 31 2 32 case, the frequency of spotty and plain morphotypes was significantly different ( = 9.46, P = 0.002). It is con- 32 33 cluded that there is significant genetic heterogeneity within this species, which appears to be highly associated 33 34 with the spotty and plain forms, but not completely explained by them. Given the previously demonstrated bio- 34 35 logical differences between morphotypes, and the present genetic analyses, we speculate about the convenience 35 36 of a taxonomic re-evaluation of this species. 36 37 37 Keywords: ballan wrasse, color morphotypes, Labrus bergylta, microsatellites, speciation 38 38 39 39 40 40 41 41 42 42 43 INTRODUCTION 43 44 44 45 Ballan wrasse (Labrus bergylta Ascanius, 1767, 45 46 Lab ri­ dae) is a protogynous hermaphrodite marine fish 46 47 showing highly variable color patterns that are not re- 47 48 lated with the sex of the individuals, as opposed to oth- 48 49 Correspondence: María Quintela, Institute of Marine Research, er wrasse species (Villegas-Ríos et al. 2013a). In Gali- 49 50 PO Box 1870, 5817 Bergen, Norway. cia (northwest Spain), 2 distinct morphotypes of this 50 51 Email: [email protected] species are known to coexist in sympatry. These include 51

162 © 2016 The Authors Integrative Zoology published by International Society of Zoological Sciences, Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Genetic analysis of ballan wrasse morphotypes

1 spotted colored individuals (Fig. 1a) that display a dark of the distribution range, from Azores Islands to Nor- 1 2 orange or reddish body patterned with white dots, and way (see Villegas-Ríos et al. [2013b] and references 2 3 plain colored individuals (i.e. non-spotted; Fig. 1b) that therein), and, thus, it has been recommended to consid- 3 4 are characterized by a uniform, although variable, body er plain and spotted morphotypes as 2 independent man- 4 5 color (mainly greenish, brownish or reddish), darker in agement units throughout its distribution range, at least 5 6 the back and whitish in the abdomen (Villegas-Ríos et while the taxonomic status of the species remains un- 6 7 al. 2013b). In Galicia, where this has been investigated, resolved (Villegas-Ríos et al. 2013b). The persistence 7 8 both types display overlapping but, nevertheless, dif- of color morphs within populations is usually attribut- 8 9 ferent life-history strategies, with plain morphs invest- ed to a number of factors, such as non-random mating, 9 10 ing more in reproduction at the expense of growing less fluctuations in selection regimes, and frequency-depen- 10 11 versus spotted morphs investing less in reproduction dent selection by predation or intrasexual competition 11 12 but growing bigger (Villegas-Ríos 2013). These differ- (e.g. Hoekstra et al. 2004). In addition, within-popula- 12 13 entiated types not only hold different common names tion sexual selection can generate negative frequency 13 14 in this region, but are commercialized separately (spot- dependence, initiating reproductive isolation and, thus, 14 15 ted individuals being more expensive) (Villegas-Ríos allowing color morphs to speciate even in sympatry (e.g. 15 16 et al. 2013b). Together, these observations indicate that Gray & McKinnon 2007). 16 17 the 2 morphotypes may belong to genetically different Recently, we developed a set of 20 microsatel- 17 18 groups. Indeed, the possibility of them representing 2 lite markers that enables genetic studies of this species 18 19 separate species has been put forward based upon their (Quintela et al. 2014). Here, the main aim was to evalu- 19 20 slightly different life-history strategies (Villegas-Ríos ate whether there were any genetic differences between 20 21 2013). However, this remains to be investigated. plain and spotted morphotypes of ballan wrasse in order 21 22 In spite of the growing body of literature about bal- to investigate the existence and, eventually, the degree 22 23 lan wrasse life history in recent times (e.g. D’Arcy et of disruptive selection between color patterns as well as 23 24 al. 2012; Villegas-Ríos et al. 2013a,b; Villegas-Ríos to measure the gene flow between them. A second aim 24 25 et al. 2014), the population structure and demograph- was to provide some insight about the taxonomic status 25 26 ic history of this species is still poorly known in the of both phenotypes. 26 27 wild, as it is the basis of differences between plain and 27 28 28 spotted individuals. These divergent patterns of color- MATERIAL AND METHODS 29 ation have also been observed in other geographic areas 29 30 Biological samples and DNA extraction 30 31 31 32 A total of 89 individuals belonging to the spotted (41) 32 33 and non-spotted (48) morphotypes were obtained in the 33 34 Galician coast (northwest Spain) from a local fish mar- 34 35 a ket in A Coruña (43°21″33.98″N, 8°24′7.20″W). No 35 36 specific permission for sampling was required as the in- 36 37 dividuals sampled for this study were obtained from 37 38 commercial artisanal inshore fishing. All samples were 38 39 photographed, weight and lengths were taken, and ge- 39 40 nomic DNA was extracted from muscle using the Qia- 40 41 gen DNeasy Blood & Tissue Kit following the manufac- 41 42 b turer’s instructions. 42 43 43 Genetic analyses with mtDNA 44 44 45 We used the barcoding approach to assess whether 45 46 there were differences in mtDNA sequence that could 46 47 support the hypothesis of ballan wrasse morphotypes 47 48 belonging to 2 different species. Thus, a 648-bp region 48 49 of mitochondrial COI gene was amplified in all the indi- 49 50 Figure 1 Ballan wrasse morphotypes: Pictures of (a) spotted viduals using the universal primers described by Folmer 50 51 and (b) non-spotted morphs. 51

© 2016 The Authors Integrative Zoology published by International Society of Zoological Sciences, 163 Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd M. Quintela et al.

1 et al. (1994). Amplifications were performed in 25 mL software (Foll & Gaggiotti 2008) and, second, we used 1 2 of a solution containing 0.5 mM of each primer, 0.2 mM the Fdist approach implemented in LOSITAN (Antao et 2

3 of each dNTP, 2 mM MgCl2, 1× AmpliTaq buffer, 1.25 al. 2008). However, as none of the outlier detection pro- 3 4 U AmpliTaq DNA polymerase (Applied Biosystems, cedures performed on the 89 fish divided according to 4 5 Foster City, CA, USA) and 1.5 mL DNA. Cycling con- morphotypes revealed any locus deviating from neutral 5 6 ditions were 2 min denaturing at 95°C followed by 30 expectations, the total set of 20 microsatellite loci was 6 7 cycles of 30 s at 95°C, 30 s at 45°C and 1 min at 72°C. regarded as neutral for all subsequent analyses. 7 8 After removing the excess of primers and nucleotides Samples sorted into their respective morphotypes (i.e. 8 9 (with shrimp alkaline phosphatase and exonuclease I en- spotty and plain) were used to estimate the total num- 9 10 zymes), samples were sequenced by Macrogen (Macro- ber of alleles, the number of private alleles, observed 10 11 11 gen, Seoul, Korea) on ABI 3730 instruments (Applied (HO) and unbiased expected heterozygosity (UHE), and 12 12 Biosystems). Sequences were checked and edited using the inbreeding coefficient FIS per morphotype using the 13 the program Genious v. 6.1.7 (Biomatters, Auckland, program GenAlEx (Peakall & Smouse 2006). The gen- 13 14 New Zealand). After alignment and trimming, the final otype distribution of each locus per morph and its di- 14 15 sequence length was 648 bp. rection (heterozygote deficit or excess) was compared 15 16 16 Genetic analyses with microsatellites with the expected Hardy–Weinberg distribution using 17 the program GENEPOP on the web (Rousset 2008) as 17 18 Twenty microsatellite loci (WrA103, WrA107, was linkage disequilibrium. Both were examined using 18 19 WrA111, WrA112, WrA113, WrA203, WrA223, the following Markov chain parameters: 10 000 steps 19 20 WrA224, WrA228, WrA236, WrA237, WrA254, of dememorization, 1000 batches and 10 000 iterations 20 21 WrA255, WrA256, WrA259, WrA261, WrB102, per batch. The same analyses were also conducted on 21 22 WrB212, WrB213, WrB215) were used as described STRUCTURE-produced clusters (see below). 22 23 23 in Quintela et al. (2014). Six multiplexed PCR reac- We used BAPS 6.0 (Corander et al. 2004) in a blind 24 24 tions were performed in final volume of 10 μL contain- manner to identify the most likely number of individual 25 25 ing 50 ng DNA template, 1× buffer, 2 mM MgCl2, 1.25 clusters (from K = 1 to K = 20) and then we performed 26 26 mM dNTPs, 0.06–0.12 μM of each primer and 1U Go- the most likely admixture of genotypes: an approach 27 27 Taq polymerase. PCR profiles included an initial 4-min that is powerful in identifying hidden structure within 28 28 denaturation at 94°C followed by 24 cycles of 50 s at populations (Corander & Marttinen 2006). 29 94°C, 90 s at an annealing temperature of 56°C, 1 min 29 30 of extension at 72°C and a final extension of 72°C for To investigate if the division according phenotypes 30 31 10 min. Forward primers were labeled with fluorescent would match any potential genetic differentiation, we 31 32 dyes and PCR products were electrophoresed on an ABI performed a blind clustering of the 89 samples (i.e. 32 33 Prism 377 Genetic Analyzer (Applied Biosystems). The without the use of any prior) using the Bayesian mod- 33 34 500LIZ size standard (Applied Biosystems) was used to el-based clustering algorithms implemented in STRUC- 34 35 accurately determine the size of the fragments and allel- TURE v. 2.3.4 (Pritchard et al. 2000). We chose a mod- 35 36 ic variation. Fragments were analyzed with the software el assuming admixture and correlated allele frequencies 36 37 GeneMapper v5 (Applied Biosystems). All samples without using population information and with all indi- 37 38 were genotyped twice to evaluate genotyping consisten- viduals regarded as 1 single population. Ten runs with 38 39 cy, and repeatability was found to be 99.86%. The labo- a burn-in period consisting of 100 000 replications and 39 40 ratory in which the microsatellite analysis was conduct- a run length of 1 000 000 Markov chain Monte Car- 40 41 ed performs analyses for the regulatory authorities, and lo (MCMC) iterations were performed for a number 41 42 routinely checks genotyping quality. of clusters ranging between K1 and K5. We then used 42 STRUCTURE Harvester (Earl & von Holdt 2012) to 43 Statistical analyses 43 44 calculate the Evanno et al. (2005) ad hoc summary sta- 44 45 In order to detect any sign of divergence between tistic ΔK, which is based on the rate of change of the 45 46 plain and spotted morphotypes, we conducted outlier “estimated likelihood” between successive K values. 46 47 detection tests aiming to identify loci under direction- Runs were averaged with CLUMPP version 1.1.1 (Ja- 47 48 al selection. To minimize the risk of detecting false pos- kobsson & Rosenberg 2007) using the LargeKGreedy 48 49 itives, we used 2 different procedures to detect loci with algorithm and the G′ pairwise matrix similarity statis- 49 50 a pattern deviating from neutrality. First, we used the hi- tics. Averaged runs were displayed using barplots. Fur- 50 51 erarchical Bayesian method implemented in BayeScan thermore, STRUCTURE was also conducted under al- 51

164 © 2016 The Authors Integrative Zoology published by International Society of Zoological Sciences, Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd Genetic analysis of ballan wrasse morphotypes

1 most identical conditions but with a slightly different 1 2 approach; that is, assisting the clustering with popinfo a 2 3 priori. 3 4 To compare the genetic differentiation between mor- 4 5 photypes on the one hand, and the clusters of individu- 5 6 6 als on the other, we computed traditional pairwise FST 7 7 (Weir & Cockerham 1984) analysis and FST per locus 8 using ARLEQUIN v.3.5.1.2 (Excoffier et al. 2005), and 8 9 9 G′ST using GenAlEx (Peakall & Smouse, 2006). G′ST is 10 useful in situations like the present one when the num- 10 11 ber of sampled populations is small, especially for pair- 11 12 wise comparisons. In all cases, signification was calcu- 12 13 lated using 10 000 permutations. In addition, a factorial 13 14 correspondence analysis was applied using the software 14 15 GENETIX (Belkhir et al. 2004) to further assess the dis- 15 16 tribution of samples by putative groups based on indi- 16 17 vidual genotypes. 17 18 18 The pattern and magnitude of gene flow between 19 19 morphotypes was assessed by using the Bayesian infer- 20 20 ence with the assistance of BayesAss v. 1.3 (Wilson & 21 21 Rannala 2003). This approach does not rely on a dataset 22 22 that is on Hardy–Weinberg or in migration-drift equi- 23 23 librium, and it uses a MCMC algorithm to estimate the 24 24 posterior probability distribution of the proportion of 25 25 migrants from one population to another (M). The anal- 26 6 26 ysis was conducted with 3 × 10 iterations, with a burn- 27 6 27 in of 10 iterations, and a sampling frequency of 2000 to 28 28 ensure that the model’s starting parameters were suffi- 29 29 ciently randomized. 30 30 31 Figure 2 Ballan wrasse mtDNA.- COI haplotype distribution 31 32 RESULTS according morphotypes: (a) spotted and (b) non-spotted. The 32 numbers of individuals per haplotype are depicted besides the 33 33 pie chart. 34 mtDNA 34 35 Eighty-seven individuals were successfully se- 35 36 quenced for COI (2 of the non-spotted ones failed). In 36 37 total, the 87 samples revealed 4 mutations in 4 polymor- plify for COI, they provided microsatellite profiles con- 37 38 phic sites (one of them parsimoniously informative), sistent with the species (Quintela et al. 2014). A total 38 39 which generated 5 different haplotypes. Haplotype 1, of 150 microsatellite alleles across the 20 markers were 39 40 present in 79 of the fish, clearly dominated in both mor- observed in the 89 samples. Individuals belonging to 40 41 photypes: 87% in the non-spotted versus 95.1% in the the non-spotted morphotype displayed smaller size than 41 42 spotted one (Fig. 2). Individuals belonging to the plain the spotted ones, higher number of total and private al- 42 43 morphotype showed 3 unique haplotypes not detected leles and slightly higher allelic richness (Table 1a). Ob- 43 44 in the spotted morph (Haplotypes 3, 4 and 5 that were served and unbiased expected heterozygosity togeth- 44 45 present in 1 individual each). er with the inbreeding coefficient were similar between 45 46 morphotypes. Analysis of Hardy–Weinberg equilibri- 46 47 Microsatellite genetic analysis sorting samples um revealed that, at the significance level of α 0.05, 5% 47 48 according to morphotypes of loci by sample combinations in the spotted-morph 48 49 displayed significant deviations, whereas no deviations 49 All 89 individuals were amplified for the suite of mi- 50 were registered at the significance level of α 0.001. In 50 crosatellites. Thus, while 2 of the samples failed to am- 51 the non-spotted morphotype, 10% of loci deviated from 51

© 2016 The Authors Integrative Zoology published by International Society of Zoological Sciences, 165 Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd M. Quintela et al.

1 - Hardy–Weinberg equilibrium at α 0.05 and half of those 1 2 at α 0.001. Linkage disequilibrium (LD) was detect- 2 3 ed in the spotted morphs 11 times (5.8%) at α 0.05 and 3 ) (average IS

4 F 6 (3.2%) at α 0.001, whereas in the non-spotted ones it 4 5 was detected 13 times (6.8%) at α 0.05 and 3 (1.6%) at 5 Number of deviations LD 11 7 6 Number of deviations LD 11 13 α 0.001. 6 7 P < 0.05. 7 Pairwise FST between morphotypes provided by the 8 20 microsatellite markers was found to be highly sig- 8 9 9 nificant (FST = 0.018, P < 0.0001; G′ST = 0.050, P < 10 0.0001) and the analysis of the allele frequencies be- 10 11 11 Number of deviations Hardy–Weinberg equilibrium 1 1 tween them revealed 10 alleles in 8 loci showing fre- 12 No deviations Hardy–Weinberg equilibrium 1 2 quency differences ranging between 0.2 and 0.3 (Table 12 13 13 2). The pairwise FST per locus (Table 3) revealed 9 loci 14 14 out of 20 showing an FST significantly different from 15 zero (WrA103, WrA203, WrA111, WrA254, WrA113, 15 16 WrA228, WrA261, WrA259, and WrB102), 8 of which 16 IS −0.015 ± 0.025 0.007 ± 0.027 17 F coincided with the ones that showed the highest al- 17 IS F 18 0.004 ± 0.027 −0.002 ± 0.027 lele-frequency differences between the plain and spotted 18 19 morphotyes. Factorial correspondence analyses further 19 20 confirmed the separation between morphotypes, despite 20 21 the low percentage of the total variation explained by 21 22 22 0.587 ± 0.045 0.686 ± 0.039 uHe the 2 first axes: 7.39% (Fig. 3). 23 23 The assessment of gene flow within and between uHe 24 0.619 ± 0.044 0.656 ± 0.041 24 morphotypes revealed an asymmetrical pattern. Hence, 25 25 the probability of being non-migrant was 68%, whereas 26 26 the gene flow between types was 32%. 0.589 ± 0.047 0.671 ± 0.041 27 Ho 27 28 Microsatellite analysis based upon clustering 28 29 29 Ho 0.616 ± 0.050 0.643 ± 0.039 methods 30 30 15 16 Number of private alleles BAPS reported one single genetic group (probabil-

31 P < 0.05 and number of deviations from genotypic linkage disequilibrium (LD) at 31 32 ity = 0.93) for a number of Ks ranging between 1 and 32 6.8 6.8 33 Ar 20. Likewise, STRUCTURE analyses blindly conduct- 33 Number of private alleles 12 21 34 ed on the 89 samples showed similar average likelihood 34 Ar 35 6.5 6.7 at K1, K2 and K3, although with K1 showing the high- 35 36 est value (Fig. 4a). However, the lack of striking differ- 36 134 135 Number of alleles 37 ences between likelihood at K1 and K2, combined with 37 38 an independent metric (i.e. the existence of 2 clearly dif- 38 Number of alleles 129 138 39 ferent morphotypes), encouraged us to explore the sce- 39 40 nario provided by STRUCTURE at K = 2 to assess the 40 29.3 ± 4.7 26.5 ± 5.6 Size (cm) 41 frequency of each morphotype per cluster in order to 41 42 investigate whether it coincided with the genetic data. 42 44 45 N Size (cm) 43 30.1 ± 5.1 25.8 ± 4.7 Thus, first we clumped 10 runs of STRUCTURE at K 43 44 = 2, and then, we set a threshold of inferred member- 44 45 ship to cluster of >0.50 to distribute the individuals into 45 N 41 48 46 the corresponding clusters. Despite the fact that at K = 46 47 2 STRUCTURE did not neatly divide the fish into spot- 47 48 ted and non-spotted morphs based upon the genetic data 48 49 alone, there were, nevertheless, significant differenc- 49 50 es in the frequencies of each phenotype in the 2 genetic 50 2 STRUCTURE cluster I STRUCTURE cluster II Group Group Spotted morph Plain morph ± SD), number of deviations from Hardy–Weinberg equilibrium at ± SD), number of deviations from Hardy–Weinberg Table 1 Ballan wrasse microsatellites Table a Division according phenotypic morphotypes b Division according STRUCTURE clustering with a threshold of >0.50 51 al number of alleles, ± SD), total size (mean showing number of individuals, per grouping method (color morphotypes and STRUCTURE clustering) Summary statistics ( coefficient inbreeding and SE), ± (average heterozygosity expected unbiased SE), ± (average heterozygosity observed alleles, of private number richness, lelic clusters ( = 9.46, P = 0.002). Hence, the distribution 51

166 © 2016 The Authors Integrative Zoology published by International Society of Zoological Sciences, Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd Genetic analysis of ballan wrasse morphotypes

1 Table 2 Ballan wrasse allele frequency 1 2 MORPHOTYPES STRUCTURE K2 2 3 3 Locus Allele Spotted Plain Cluster I Cluster II 4 4 5 WrA103 194 0.110 0.302 0.227 0.200 5 6 WrA103* 199* 0.317 0.083 0.273 0.111 6 7 WrA111 210 0.159 0.344 0.239 0.278 7 8 WrA111 222 0.220 0.052 0.170 0.089 8 9 9 WrA113* 197* 0.500 0.323 0.568 0.244 10 10 11 WrA203* 168* 0.549 0.281 0.523 0.289 11 12 WrA228* 167* 0.500 0.323 0.568 0.244 12 13 WrA254* 200* 0.793 0.576 0.791 0.568 13 14 WrA259* 195* 0.098 0.250 0.102 0.256 14 15 15 WrA261 234 0.256 0.083 0.159 0.167 16 16 17 Values for those alleles showing the highest difference in frequency (0.2-0.3) between morphotypes (spotted vs. plain) and genetic 17 18 clusters according STRUCTURE at K=2 without priors (I vs. II) are depicted in boldface type. *These alleles matched both criteria. 18 19 19 20 20 Table 3 Ballan wrasse pairwise F per locus between morphotypes (plain vs. spotted) and STRUCTURE K2 genetic clusters (I vs. 21 ST 21 22 II) 22 23 MORPHOTYPES STRUCTURE K2 23 24 24 Locus No alleles F P-value F P-value 25 ST ST 25 26 WrA103* 6 0.0652 0.0000 0.0212 0.0178 26 27 WrA107 3 0.0135 0.1391 0.0545 0.0055 27 28 WrA111* 9 0.0456 0.0006 0.0168 0.0441 28 29 WrA112 13 0.0002 0.4169 0.0202 0.0111 29 30 30 WrA113* 9 0.0236 0.0110 0.0716 0.0000 31 31 32 WrA203* 16 0.0531 0.0000 0.0369 0.0022 32 33 WrA223 5 0.0031 0.2336 0.0053 0.2017 33 34 WrA224 5 0.0000 0.4577 0.0000 0.4954 34 35 WrA228* 9 0.0236 0.0113 0.0716 0.0000 35 36 36 WrA236 9 0.0000 0.6750 0.0049 0.1887 37 37 38 WrA237 3 0.0000 0.5866 0.0000 0.4352 38 39 WrA254* 5 0.0553 0.0055 0.0573 0.0050 39 40 WrA255 7 0.0036 0.2914 0.0000 0.8386 40 41 41 WrA256 10 0.0033 0.2679 0.0018 0.3290 42 42 WrA259* 8 0.0173 0.0432 0.0186 0.0409 43 43 44 WrA261* 6 0.0348 0.0146 0.0000 0.9804 44 45 WrB102 2 0.0339 0.0457 0.0539 0.0170 45 46 WrB212 9 0.0000 0.4617 0.0694 0.0000 46 47 47 WrB213 10 0.0017 0.2469 0.0082 0.0926 48 48 49 WrB215 6 0.0000 0.8548 0.0013 0.4322 49

50 Values in boldface type were significantly different from zero. *These loci showed the highest difference in allele frequencies be- 50 51 tween morphotypes. 51

© 2016 The Authors Integrative Zoology published by International Society of Zoological Sciences, 167 Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd M. Quintela et al.

1 1 2 2 3 3 4 4 5 5 6 6 7 7 8 8 9 Figure 3 Two-dimensional representa- 9 10 tion of the Factorial Correspondence Anal- 10 11 ysis. Individuals are projected on the fac- 11 12 tor space defined by the similarity of their 12 13 allelic states. White dots depict plain mor- 13 14 photype whereas black dots depict spotted 14 15 individuals. 15 16 16 17 17 18 18 19 19 20 20 21 21 22 22 23 23 24 24 25 25 26 26 27 27 28 28 29 29 30 Figure 4 Plot of mean posterior probability values (L(K) ± SD) per cluster (K) based on 10 replicates per K generated by STRUC- 30 31 TURE (circles), and ΔK analysis (squares) of LnP(D) according to Evanno et al. (2005) for: (a) blind clustering approach and (b) 31 32 using morphotypes as popinfo. 32 33 33 34 34 35 35 of fish into clusters was as follows: 28 spotted individu- ter I 11 times (5.8%) at α 0.05 and 2 (1.1%) at α 0.001, 36 36 als (63.6%) were assigned to cluster I and 13 (28.9%) to whereas in cluster II it was detected 7 times (3.7%) at α 37 37 cluster II, whereas 16 of the plain fish (36.4%) were as- 0.05 and 1 (0.5%) at α 0.001. 38 38 signed to cluster I and 32 (71.1%) to cluster II (STRUC- 39 The genetic structure between clusters was high- 39 TURE barplots in Fig. 5). Furthermore, individuals dis- 40 ly significant F( ST = 0.025, P < 0.000; G′ST = 0.067, P < 40 tributed between clusters I and II were even in number, 41 0.0001), and this way of arranging the individuals re- 41 and showed similar sizes, and almost identical number 42 vealed a slightly higher genetic differentiation than 42 of total alleles, allelic richness and private alleles (Table 43 the division according to phenotypes (FST = 0.018, P < 43 1b). Only Ho, UHe and F took slightly lower values in 44 IS 0.0001; G′ST = 0.050, P < 0.0001). Likewise, the analy- 44 cluster I than in II. Analysis of Hardy–Weinberg equilib- 45 sis of allele frequencies revealed 6 alleles showing dif- 45 rium revealed that, at the significance level of α 0.01, 5% 46 ferences in frequencies between the 2 clusters ranging 46 of loci by sample combinations in cluster I displayed 47 between 0.2 and 0.3 (Table 2), in the same direction and 47 significant deviations; whereas no deviations were reg- 48 of similar magnitude as the ones between morphotypes. 48 istered at the significance level of α 0.001. In cluster II, 49 The latter point is important as it reveals that the 2 in- 49 only 1 locus (5%) showed deviations from Hardy–Wein- 50 dependent methods (sorting by morphotype and sorting 50 berg equilibrium at α 0.001. LD was detected in clus- 51 by clustering) pointed to the same markers as separat- 51

168 © 2016 The Authors Integrative Zoology published by International Society of Zoological Sciences, Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd Genetic analysis of ballan wrasse morphotypes

1 1 2 2 3 3 4 4 5 5 6 6 7 7 8 8 9 9 10 10 11 11 12 12 13 13 14 14 15 15 16 16 17 17 18 18 19 19 20 20 21 21 22 22 23 23 24 24 25 25 26 26 27 Figure 5 Inferred ancestry of plain and spotted morphotypes of ballan wrasse according STRUCTURE at K = 2 and K3 for: (a,c) 27 28 clustering using morphotypes as popinfo; and (b,d) blind clustering approach. The order of the individuals has been kept identical in 28 29 each of the 4 barplots. 29 30 30 31 31 32 32 33 ing the samples into 2 groups. Factorial correspondence < 0.0001). In addition, blind clustering methods based 33 34 analyses run on the total 20 markers confirmed the sep- upon genetic data alone, without the use of any kind 34 35 aration between clusters I and II and showed the same of prior, sorted the samples into 2 genetically distinct 35 36 percentage of total variation explained by the 2 first groups (FST = 0.025 averaged over 20 loci, P < 0.0001), 36 37 axes as in the morphotypes analysis (7.39%). where the frequencies of the spotty and plain morpho- 37 2 38 When assisting STRUCTURE clustering with mor- types were significantly different ( = 9.46, P = 0.002). 38 39 photypes as popinfo, L(K) took very similar values at Thus, the combination of these 2 analytical approaches 39 40 K2 and K3, and ΔK did not have enough grounds to dis- together with the asymmetric pattern of gene flow with- 40 41 criminate between K2 and K4 either (Fig. 4b). Howev- in and between morphs confirms that the genetic dif- 41 42 er, and as expected, individuals were neatly divided be- ferentiation within this species appears to be associated 42 43 tween cluster I and II according to their phenotypes. with, but not necessarily completely explained by, the 2 43 44 main morphotypes spotty and plain. 44 45 DISCUSSION The assessment of genetic differentiation between 45 46 morphotypes of ballan wrasse conducted in the pres- 46 47 This represents the first genetic study of the sympat- ent study aimed not only to investigate potential disrup- 47 48 ric color morphotypes of ballan wrasse. Highly signif- tive selection between color patterns, but also to provide 48 49 icant genetic differences were observed between plain some insight about any putative taxonomic difference 49 50 and spotted morphs sampled from the same region of between them. In contrast with the popular belief, sig- 50 51 north-west Spain (FST = 0.018 averaged over 20 loci, P 51

© 2016 The Authors Integrative Zoology published by International Society of Zoological Sciences, 169 Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd M. Quintela et al.

1 nificant homogeneity revealed by mitochondrial COI, ic data support this suggestion, there may be enough ge- 1 2 with haplotype 1 clearly dominating in both groups, did netic overlap between the 2 morphotypes to indicate that 2 3 not seem to provide strong support for the hypothesis it might not be as simple as spotty and plain forms being 3 4 of plain and spotted individuals belonging to 2 different potential subspecies. It is, for example, possible that the 4 5 species. However, and significantly, at the nuclear level, 2 genetic sub-groups differ in their frequency of these 5 6 40% of the studied loci showed large differences in al- morphotypes (i.e. one is primarily spotty and the other 6 7 lele frequencies between color morphotypes. Moreover, primarily plain). If this is the case, then some of the ob- 7 8 the loci flagged as displaying strong divergence between served differences in traits between the spotty and plain 8 9 the 2 morphotypes, and the 2 genetic clusters which morphotypes as reported by Villegas-Ríos et al. (2013b) 9 10 were associated with morphotype frequency differences, may actually represent underestimates of the differenc- 10 11 strongly coincided. es between the 2 genetic groups as there is no exact co- 11 12 The degree of structure between morphs in ballan incidence between morphotype and genetic grouping. 12 13 wrasse evokes a recurrent question on marine organ- Further studies using larger numbers of individuals and 13 14 isms; that is, the interpretation of low but still statisti- markers, and matching extensive life-history and trait 14 15 cally significant levels of genetic differentiation. In an data to the genetic data will be required to fully resolve 15 16 attempt to determine to what extent such observations this issue. 16 17 reflected a real phenomenon and not the result of con- The accurate evaluation of the taxonomic status of 17 18 founding factors (i.e. unrepresentative sampling, se- ballan wrasse is key for the sustainable management of 18 19 lection acting on the marker loci), a long-term study this species, which is commercially valuable through- 19 20 on coastal Atlantic cod combined genetics and exten- out its distribution range for recreational and artisanal 20 21 sive capture–mark–recapture to conclude that the low, fishing. The small home range of L. bergylta together 21 22 22 and yet highly significant, FST (average 0.0037) was bi- with its sedentary behavior highlights the potential use 23 ologically meaningful as it did correspond to separate of small marine protected areas as a management tool to 23 24 temporally persistent local populations (Knutsen et al. ensure a sustainable fishery (Villegas-Ríos et al. 2013c). 24 25 2011). The degree of differentiation found in the pres- In addition, the use of ballan wrasse as a biological 25 26 ent study was found to be 5-fold higher and matched, cleaner fish for the control of sea lice on farmed salmo- 26 27 for instance, the one reported for color morphotypes of nids can reduce the environmental impact of aquaculture 27 28 the rockfish Sebastes oculatus assessed with a suite of (e.g. Skiftesvik et al. 2013), and has been reported to be 28 29 24 microsatellites (Venerus et al. 2013). In this case, the an efficient alternative to the chemical treatment of sea 29 30 genetic differentiation added to the bathymetric segre- lice, thus resulting in the initiation of dedicated breeding 30 31 gation between color morphotypes (“dark” and “light” programs in Norway (Skiftesvik et al. 2013). Therefore, 31 32 fish) suggested the existence of speciation-by-depth a thorough understanding of the within-species genetic 32 33 in the absence of physical barriers, although the large structure associated with the morphotypes will be most 33 34 number of individuals displaying high levels of admix- valuable. Furthermore, the use of this species to con- 34 35 ture invoked incomplete reproductive barriers between trol sea lice infestations on marine farms still primar- 35 36 color morphotypes. Likewise, significant genetic differ- ily involves the capture and translocation of wild fish 36 37 37 ences of the same magnitude (FST = 0.032) were found over long distances. Thus, the lack of knowledge about 38 between littoral and profound morphotypes of the Arc- this species’ population genetic structure, the number of 38 39 tic charr (Salvelinus alpinus) using microsatellite DNA populations from which individuals are being harvest- 39 40 analysis, with morphs showing strong reproductive iso- ed and mixed in the farms, or even the existence of local 40 41 lation in time and space (Westgaard et al. 2004). adaptations can have dramatic consequences if hybrid- 41 42 42 A previous study on spotty and plain morphotypes ization between translocated and local populations takes 43 43 of ballan wrasse reported differences in growth trajec- place and hampers the local evolutionary adaptations 44 44 tories, size at age, otolith length to body length, and via maladaptive gene combinations or outbreeding de- 45 45 mortality (Villegas-Ríos et al. 2013b), and also in in- pression (e.g. McGinnity et al. 2003). Likewise, overex- 46 46 vestment in reproduction versus growth (Villegas-Ríos ploitation of wrasse in one area may leave a local popu- 47 47 2013). Based upon those results, which included exten- lation endangered or locally extinct if there is little gene 48 48 sive growth and maturation analyses, it was conclud- flow among regions (Hutchinson 2008). 49 49 ed that there was likely to exist large genetic differences Finally, a remarkable feature is that Galician bal- 50 50 between these 2 morphotypes. Clearly, while our genet- lan wrasse analyzed here showed a substantially high- 51 51

170 © 2016 The Authors Integrative Zoology published by International Society of Zoological Sciences, Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd Genetic analysis of ballan wrasse morphotypes

1 er number of alleles than an almost 3-fold larger set D’Arcy J, Dunaevskaya E, Treasurer JW et al. (2012). 1 2 of individuals sampled in the Norwegian coast (150 Embryonic development in ballan wrasse Labrus ber- 2 3 vs 115 alleles, respectively) (Quintela et al. 2014). Al- gylta. Journal of Fish Biology 81, 1101–10. 3 4 though this number of populations study does not allow Earl DA, von Holdt BM (2012). STRUCTURE HAR- 4 5 any further assumptions to be made, we would like to VESTER: A website and program for visualizing 5 6 point out the fact that also in the control region of mtD- STRUCTURE output and implementing the Evanno 6 7 NA, a lower genetic diversity has been formerly report- method. Conservation Genetics Resources 4, 359–61. 7 8 8 ed for this species in the northernmost areas when com- Evanno G, Regnaut S, Goudet J (2005). Detecting the 9 9 paring southern Norway and the British Isles, which number of clusters of individuals using the software 10 10 was interpreted as the result of historical demograph- STRUCTURE: A simulation study. Molecular Ecolo- 11 11 ic events (population bottleneck followed by expansion) gy 14, 2611–20. 12 (D’Arcy et al. 2013). The same pattern has also been 12 Excoffier L, Laval G, Schneider S (2005). Arlequin ver. 13 described for the corkwing wrasse (Symphodus mel- 13 3.0: An integrated software package for population 14 ops), with south European populations showing twice as 14 genetics data analysis. Evolutionary Bioinformatics 15 much allelic richness and average number of alleles per 15 Online 1, 47–50. 16 locus than the Scandinavian populations (Knutsen et al. 16 17 2013). Foll M, Gaggiotti O (2008). A genome-scan method to 17 18 identify selected loci appropriate for both dominant 18 and codominant markers: A Bayesian perspective. 19 ACKNOWLEDGMENTS 19 20 Genetics 180, 977–93. 20 21 Funding for this study was provided by the Norwe- Folmer O, Black M, Hoeh W, Lutz R, Vrijenhoek R 21 22 gian Ministry of Trade and Fisheries. We are grateful to (1994). DNA primers for amplification of mitochon- 22 23 Raquel Mosquera, Belén Carro, Cristina Pérez Corbei- drial cytochrome c oxidase subunit I from diverse 23 24 ra and Alfredo Veiga for their help with the collection of metazoan invertebrates. Molecular Marine Biology 24 25 samples. We thank Anne Grete Sørvik Eide, Bjørghild and Biotechnology 3, 294–9. 25 26 Breistein Seliussen and Geir Dahle for their valuable as- Gray SM, McKinnon JS (2007). Linking color polymor- 26 27 sistance in the laboratory. phism maintenance and speciation. Trends in Ecology 27 28 & Evolution 22, 71–9. 28 29 REFERENCES Hoekstra, Hopi E, Drumm, Kristen E, Nachman, Mi- 29 30 chael W (2004). Ecological genetics of adaptive color 30 31 Antao T, Lopes A, Lopes R, Beja-Pereira A, Luikart G 31 (2008). LOSITAN: A workbench to detect molecular polymorphism in pocket mice: Geographic variation 32 in selected and neutral genes. Evolution 58, 1329–41. 32 33 adaptation based on a FST-outlier method. BMC Bio- 33 Hutchinson WF (2008). The dangers of ignoring stock 34 informatics 9, 323. 34 complexity in fishery management: The case of the 35 Belkhir K, Borsa P, Chikhi L, Raufaste N, Bonhomme F 35 North Sea cod. Biology Letters 4, 693–5. 36 (2004). GENETIX 4.05, logiciel sous Windows TM 36 37 pour la génétique des populations. Jakobsson M, Rosenberg NA (2007). CLUMPP: A clus- 37 38 Corander J, Marttinen P (2006). Bayesian identifica- ter matching and permutation program for dealing 38 39 tion of admixture events using multilocus molecular with label switching and multimodality in analysis of 39 40 markers. Molecular Ecology 15, 2833–43. population structure. Bioinformatics 23, 1801–6. 40 41 Corander J, Waldmann P, Marttinen P, Sillanpaa MJ Knutsen H, Jorde PE, Blanco González E, Robalo J, Al- 41 42 (2004). BAPS 2: Enhanced possibilities for the anal- bretsen J, Almada V (2013). Climate change and ge- 42 43 ysis of genetic population structure. Bioinformatics netic structure of leading edge and rear end popula- 43 44 20, 2363–9. tions in a northwards shifting marine fish species, the 44 45 corkwing wrasse (Symphodus melops). PLoS ONE 8, 45 D’Arcy J, Mirimin L, Fitzgerald R (2013). Phylogeo- 46 e67492. 46 graphic structure of a protogynous hermaphrodite 47 47 species, the ballan wrasse Labrus bergylta, in Ireland, Knutsen H, Olsen EM, Jorde PE, Espeland SH, AndrÉ C, 48 48 Scotland, and Norway, using mitochondrial DNA se- Stenseth NC (2011). Are low but statistically signif- 49 49 quence data. ICES Journal of Marine Science: Jour- icant levels of genetic differentiation in marine fish- 50 50 nal du Conseil 70, 685–93. es ‘biologically meaningful’? A case study of coastal 51 Atlantic cod. Molecular Ecology 20, 768–83. 51

© 2016 The Authors Integrative Zoology published by International Society of Zoological Sciences, 171 Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd M. Quintela et al.

1 McGinnity P, Prodöhl P, Ferguson A et al. (2003). Fit- Villegas-Ríos D (2013). Life-history and behaviour of 1 2 ness reduction and potential extinction of wild popu- Labrus bergylta in Galicia. Department of Ecology 2 3 lations of Atlantic salmon, Salmo salar, as a result of and Marine Biodiversity, PhD thesis, 228. 3 4 interactions with escaped farm salmon. Proceedings Villegas-Ríos D, Alonso-Fernández A, Domínguez-Petit 4 5 of the Royal Society of London. Series B: Biological R, Saborido-Rey F (2013a). Intraspecific variability 5 6 Sciences 270, 2443–50. in reproductive patterns in the temperate hermaphro- 6 7 7 Peakall R, Smouse PE (2006). GenAlEx 6: Genet- dite fish, Labrus bergylta. Marine and Freshwater 8 Research 64, 1156–68. 8 9 ic analysis in Excel. Population genetic software for 9 Villegas-Ríos D, Alonso-Fernández A, Domínguez-Pe- 10 teaching and research. Molecular Ecology Notes 6, 10 tit R, Saborido-Rey F (2014). Energy allocation and 11 288–95. 11 reproductive investment in a temperate protogynous 12 Pritchard JK, Stephens M, Donnelly P (2000). Inference 12 hermaphrodite, the ballan wrasse Labrus bergylta. 13 13 of population structure using multilocus genotype Journal of Sea Research 86, 76–85. 14 data. Genetics 155, 945–59. 14 15 Villegas-Ríos D, Alonso-Fernández A, Fabeiro M, 15 Quintela M, Danielsen EA, Svåsand T, Knutsen H, Bañón R, Saborido-Rey F (2013b). Demographic 16 Skiftesvik AB, Glover KA (2014). Isolation and char- 16 17 variation between colour patterns in a temperate pro- 17 acterization of twenty microsatellite loci for the bal- 18 togynous hermaphrodite, the ballan wrasse Labrus 18 lan wrasse, Labrus bergylta. Conservation Genetics 19 bergylta. PLoS ONE 8, e71591. 19 Resources, 1–4. 20 Villegas-Ríos D, Alós J, March D, Palmer M, Mucien- 20 21 Rousset F (2008). GENEPOP′007: A complete re-imple- tes G, Saborido-Rey F (2013c). Home range and diel 21 22 mentation of the genepop software for Windows and behavior of the ballan wrasse, Labrus bergylta, deter- 22 23 Linux. Molecular Ecology Resources 8, 103–6. mined by acoustic telemetry. Journal of Sea Research 23 24 Skiftesvik AB, Bjelland RM, Durif CMF, Johansen IS, 80, 61–70. 24 25 Browman HI (2013). Delousing of Atlantic salm- Weir BS, Cockerham CC (1984). Estimating F-statistics 25 26 on (Salmo salar) by cultured vs. wild ballan wrasse for the analysis of population structure. Evolution 38, 26 27 (Labrus bergylta). Aquaculture 402–3, 113–8. 1358–70. 27 28 28 Venerus LA, Ciancio JE, Riva-Rossi C, Gilbert-Hor- Westgaard JI, Klemetsen A, Knudsen R (2004). Genet- 29 ic differences between two sympatric morphs of Arc- 29 vath EA, Gosztonyi AE, Garza JC (2013). Genetic 30 tic charr confirmed by microsatellite DNA. Journal 30 structure and different color morphotypes suggest the 31 of Fish Biology 65, 1185–91. 31 occurrence and bathymetric segregation of two in- 32 Wilson GA, Rannala B (2003). Bayesian inference of 32 cipient species of Sebastes off Argentina. Naturwis- 33 recent migration rates using multilocus genotypes. 33 senschaften 100, 645–58. 34 Genetics 163, 1177–91. 34 35 35 36 36 37 Cite this article as: 37 38 38 39 Quintela M, Danielsen E, Lopez L et al. (2016). Is the ballan wrasse (Labrus bergylta) two species? Genetic anal- 39 40 ysis reveals within-species divergence associated with plain and spotted morphotype frequencies. Integrative 40 41 Zoology 11, 162–72. 41 42 42 43 43 44 44 45 45 46 46 47 47 48 48 49 49 50 50 51 51

172 © 2016 The Authors Integrative Zoology published by International Society of Zoological Sciences, Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd