Draft Genetic risks of supplementing trout populations with native stocks : a simulation case study from current Pyrenean populations

The risks of supplementation must be examined to assess the genetic effects to native wild populations before full implementation or exclusion of programs that involve captive breeding and release. Real genetic data can be applied to simulations of genetic changes in populations of interest and subsequently used in risk assessment. Ancestral Mediterranean brown trout (Salmo trutta) lineages exhibit complex population structure among native populations. Genetically divergent Atlantic stocks were maintained and released in the Mediterranean rivers as recreational fish, which resulted in hybridization and introgression with local populations. Therefore, we designed a new supplementation program based on native stocks and evaluated the genetic risks associated with releasing native fish in recreational fisheries. Our simulation was delimited by the observed population genetic structure and available hatchery facilities in the study region. Supplementation with native stocks maintained estimates of gene divers...


Introduction
Freshwater fish are among the most endangered species groups due to various anthropogenic impacts, including habitat fragmentation and contamination, and water exploitation and diversion (Freyhof and Brooks 2011).In addition, several freshwater fish species have exhibited notable declines in recent decades from exotic species introductions, overfishing, and/or release of non-native stocks (Cowx and Gerdeaux 2004).In temperate and cold river basins, salmonids have experienced declines due to each of these threats on native freshwater biodiversity (Lewin et al. 2006;Naish et al. 2007), and on-going climate change has increased the vulnerability and endangered status of salmonid species worldwide (e.g.Hari et al. 2006;Almodovar et al. 2012;Vera et al. 2013).In addition to habitat recovery, several other strategies have been suggested to improve local degraded populations.Extensive release of hatchery cultured fish to enhance recreational opportunities has been widely applied, and supported by anglers (Brown and Day 2002;Arlinghaus and Mehner 2005;Cowx et al. 2010).Despite the widespread biological concerns of applying supplementation practices from hatcheryreared fish to restore wild populations (Laikre and Ryman 1996;Lewin et al. 2006;Naish et al. 2007;Araki and Schmid 2010), these practices continue; and anglers show a positive response by increasing their fishing efforts soon after fish are released (e.g.Baer et al. 2007), which subsequently promotes revenues to local economies (Arlinghaus et al. 2002).
In European basins, brown trout are an integral target for a socio-economically important recreational fishery (Elliot 1989).The economic value of the fishery mandates the development of management strategies focused on achieving and maintaining the delicate balance between exploitation and conservation of the resource D r a f t (Araguas et al. 2009;Arlinghaus et al. 2010;Cowx and Van Anrooy 2010).Due to a reduction in the abundance of catchable trout, anglers perceived efforts to reinforce populations as a positive management objective (Arlinghaus and Mehner 2005).Such circumstances also promoted supplementation initiatives among fisheries managers, because recreational fisheries were traditionally managed based on the quality of the fishing experience (Cowx and Gerdeaux 2004), and trout anglers associated quality with catch quantity, and satisfaction followed (Arlinghaus and Mehner 2005).However, management decisions based on anglers' perception could perpetuate stocking as a panacea to maintain both, sustained fishery and the anglers' satisfaction, but at the risk of replacing local fish diversity by hatchery stocks (van Poorten et al. 2011).
For example, Araguas et al. (2004) reported extensive foreign stock releases compromised the genetic integrity and differentiation among native trout populations in eastern Pyrenean rivers.In this region, up to 8 million fish were released in year 1995, but later hatchery releases lowered and have been stabilized in recent times to 2 million fish per year (Araguas et al. 2008).Because local salmonid populations often exhibit a fitness advantage relative to foreign populations (Fraser et al. 2011;Perrier et al. 2013), supplementation with foreign stocks is likely to reduce the mean population fitness in recipient locations due to outbreeding depression and maladaptation of released fish (Rhymer andSimberloff 1996, Baskett et al. 2013).Studies on the adaptive value of local brown trout differences remain scarce, and not all phenotypic traits respond similarly.For example, Jensen et al. (2008a) reported local adaptive variation for length D r a f t 5 at hatching, and length at first feeding among four Danish trout populations within close geographic proximity.However, adaptive evidence for distinct hatching and survival times was not detected among five brown trout populations in the Swiss Rhine Basin (Stelkens et al. 2012).Nevertheless, a precautionary approach in conservation and fisheries management mandates the preservation of local types, at least until clear scientific evidence is collected to assess the biological significance of local differences (e.g.FAO 1995).
European brown trout have several evolutionary lineages (revised in Kottelat and Freyhoff 2007).The Atlantic one (now considered as Salmo trutta sensu stricto in IUCN red list) was native in the Atlantic river basins northward from the Pyrenees.Several morphological and genetically distinct Mediterranean lineages have been described (e.g.S.rhodanensis, S.cettii, S.macrostigma), and some of them are now identified as threatened species in the IUCN red list (Freyhoff and Brooks 2011).However, a taxonomic revision of the trout types inhabiting Mediterranean Iberian rivers is still lacking; despite they could represent several unnamed species (Kottelat and Freyhoff 2007).As a result, the Spanish inland fisheries agencies consider all trout populations in Spain as S.trutta, according with former taxonomic studies (Lelek 1980).However, since the 1980s, the Autonomous Government of Catalonia, as well as other regional administrations, recognized declines in endemic trout biodiversity for the region; and since 1987 introduced management measures to recover native populations.These measures included an increase in length requirement for catchable fish from 18 to 22 cm, a reduction in hatchery releases, which are now banned in some upstream stretches declared as genetic refuge locations to protect native gene pools, and a change to catch and release management in some previously fished areas (Araguas et al. 2009).

D r a f t
6 However, such restrictive measures typically received low angler support (Arlinghaus and Mehner 2005).In addition, on-going assessments indicated that establishment of genetic refuge did not result in a significant decrease in foreign stock alleles in the Mediterranean Spanish and French wild populations (Araguas et al. 2008(Araguas et al. , 2009;;Caudron et al. 2011Caudron et al. , 2012)).Therefore, complementary active measures, including nonnative fish removal, supplementation with local native breeding stocks, and/or translocation of wild native individuals were required to mitigate the genetic effects of past releases with foreign fish (Caudron et al. 2012;Vera et al. 2013).
Hatchery strains derived from regional or local sources (hereafter, native stocks) have been used to reinforce wild fish populations when conservation and fisheries management goals were in common, because these stocks preclude damage to local populations.For example, increased production with limited ecological and genetic effects was observed in Hamma Hamma River steelhead (Oncorhynchus mykiss) populations following supplemental stocking (Berejikian et al. 2008;Van Doornik et al. 2010).Hess et al. (2012) detected minimal negative impacts on fitness in a Chinook salmon (O.tshawytscha) population in the Columbia River enhanced with supportive captive breeding.Native stock reinforcement following habitat recovery is an accepted measure within the European Inland Fisheries Advisory Commission (EIFAC) code of practice for recreational fisheries (FAO 2008).Nevertheless, released fish from native stocks and their wild progeny often exhibited decreased performance compared to wild populations (reviewed in Araki et al 2008).For instance, Christie et al (2012) reported genetic changes in the steelhead (O.mykiss) induced during a single generation of hatchery culture resulting in maladaptation to the wild.Consequently more empirical data is needed to evaluate the genetic and fitness effects of supplementation from native D r a f t 7 stocks (Naish et al 2007;Fraser 2008;Araki and Schmid 2010).
The risks inherent with a supplementation program involving captive stocks must be carefully assessed prior to implementation of the program with native populations (Naish et al. 2007;FAO 2008).Theoretical studies demonstrated reduced Ne and fitness of wild populations following supportive breeding (Wang and Ryman 2001;Ford 2002), and a decline in local populations following long-term supplementation by native domestic stocks (Satake and Araki 2012).The effect occurred through densitydependent overcompensation during recruitment, resulting in fewer wild fish recruits.
Simulations from real genetic data, augment these theoretical population genetics approaches (Hoban et al. 2012), which facilitated assessments of genetic changes in populations of interest (e.g.Perrier et al 2013).
In the present study, we assess patterns of genetic diversity within and among populations at 13 sampling locations along four river basins in the Pyrenean Mountains as a basic pre-requisite to design a regional native hatchery stocks program.We subsequently simulated a supplementation program that involved replacement of foreign by native trout stocks using the observed population structure to select source locations, while integrating the existing regional hatchery facilities and hatchery personal expertise.We assessed the long-term genetic effects of the releases, including the target population gene pools but also the population structure at the intra-and inter-basin levels.Finally, we discussed the social and economic benefits of the results relative to regional trout fisheries.

River network, sampled locations and estimates of genetic diversity
In the Spanish eastern Pyrenees, the river network is organized into two major units for water supply and hydroelectric production, which includes the Ebro River basin and the coastal rivers.The Ebro basin includes the easternmost Segre River drainage, composed of the mainstem, and its tributaries the Noguera Pallaresa River (2820.1 km 2 drainage surface), and the Noguera Ribagorzana River (2045.6 km 2 drainage surface).The Segre basin supports the most preserved native populations of Mediterranean brown trout in the Iberian Peninsula (Sanz et al. 2002).The largest coastal rivers are the Llobregat (4948.3km 2 ) and Ter (3010.5 km 2 ) Rivers.Trout populations only inhabit the headstreams of these rivers, with trout habitats more abundant in the Ter River, with headstreams up to 2000 m a.s.l., where the Llobregat are below 1300 m a.s.l.
A total of 854 wild brown trout were collected during summer 2006 by electro-fishing at 13 localities along the Noguera Pallaresa, Noguera Ribagorzana, Llobregat, and Ter River basins (Table 1, Fig. 1) to assess patterns of genetic diversity within and among Pyrenean populations.Each fish was anesthetized with tricaine methane-sulphonate (MS-222) to biopsy a piece of the adipose fin.The sample was stored in an eppendorf tube containing 96% ethanol and transported to the laboratory until DNA extraction.In the field, the fork length of each fish was measured (to the nearest 0.5 cm) to estimate fish age (0+, 1+, 2+, 3+, and older fish) based on FISAT II software (Gayanilo et al. 2005) modal progression analysis and length-age relationships provided by Rocaspana et al. (2006) for Pyrenean trout populations.Once recovered from anaesthesia, fish were returned to the streams alive.We also analysed 96 fish from the 2003-year cohort of the foreign stock at Bagà hatchery (HAT1, see Fig1) used to reinforce Pyrenean brown trout populations in the region.We only analysed fish from this hatchery because trout D r a f t 9 culture at HAT2 is discontinued when trout production at HAT1 is enough to supplement regional fisheries.Only when an increased hatchery production is required, HAT2 receives fish from HAT1 to regenerate a new stock.
Conformance of genotype distributions with Hardy-Weinberg expectations were tested by exact probability tests (Guo and Thompson 1992) using the computer package GENEPOP 3.3 (Raymond and Rousset 1995).For each location, the minimum number of homogenous units (K = 1, 2, and 3) was determined using the Bayesian Markov Chain Monte Carlo (MCMC) approach in STRUCTURE 2.3.3 (Pritchard et al. 2000) as indicated in Sanz et al. (2009).FSTAT2.9.3 software (Goudet 1995) was used to summarize genetic diversity within samples as follows: mean unbiased expected heterozygosity (H E ), mean direct count heterozygosity (H O ), and average allele richness per locus (A R ).To measure the level of current foreign stock introgression in each wild location, we estimated the average proportion of foreign stock ancestry (q) following Sanz et al. (2009), however Araguas et al. (2008) already reported estimates for 10 of these locations based on LDH-C* locus polymorphisms.Effective population size (Ne) at each study location was estimated using linkage disequilibrium between loci in the LDNe 1.31 program (Waples and Do 2008).Ne was estimated by removing rare alleles with frequency less than 0.02, due to a sample size of less than 50 fish in some study locations.The LDNe method assumes discrete generations, which are not the case in D r a f t 10 brown trout; but the method roughly estimates Ne when the number of sampled cohorts approximates the suspected generation time (Waples and Do 2010).At our study locations, three cohorts (0+, 1+, and 2+, see results) were the most abundant, as similarly reported in the Iberian Peninsula (Lobón-Cervia et al. 1986;Nicola et al. 2008;Parra et al. 2009), suggesting generation lengths between 2-3 y for these trout populations.Additional confidence on estimated Ne from our locations resulted from genetic stability reported among five consecutive cohorts (year 1998 to 2002) studied in an earlier work in the Vallfarrera stream (Vera et al. 2010).

Population structure
Genetic divergence between locations was examined using a matrix of pairwise genetic differentiation, F ST (Weir and Cockerham 1984), and its significance (based on 1000 permutations) in FSTAT.Patterns of genetic diversity within and among the study river basins were quantified by gene diversity analyses (Nei 1987) using FSTAT and Analysis of Molecular Variance (AMOVA) using Arlequin 3.5.1.3(Excoffier and Lischer 2010).Hierarchical levels followed hydrogeographical criteria within (F SC ) and among (F CT ) basins or tributaries, and involved an analysis spanning the entire study region, and separate analyses for each river basin.Additionally, the pairwise genetic distance matrix among samples (D a , Nei 1987) computed using MSA 4.05 (Dieringer and Schlötterer 2003) was used to generate a multidimensional scaling (MDS) using SPSS program to depict genetic similarity between study locations.Finally, we examined the most likely number of genetically homogeneous groups (K = 1 to 13) in the study region from the Bayesian method implemented in STRUCTURE 2.3.3 following Evanno et al. (2005) after 20 replicate STRUCTURE runs for each K value.

D r a f t
11 Each run used a burn-in of 10000 iterations, a run length of 10000 iterations, and all other parameters set to default model.

Design of the native stocks
The current foreign hatchery stock originated from central European trout, and regional fishery management could require its replacement with native stocks in the future.
However, we should keep in mind that despite the stock origin, the current normative framework involving hatchery trout stocks in the study region serves to support recreational fisheries, and not conservation genetics goals.In addition, legislation on other water uses (e.g.industrial, agricultural, and domestic consumption, or hydroelectric production) often takes prevalence over recreational fisheries.The Environmental Services of the Autonomous Government of Catalonian manage trout populations in all the regional basins.However, water supplies and hydroelectric production in the Ebro River basin are under the jurisdiction of the C.H.E.
(Confederación Hidrográfica del Ebro), a Spanish State agency; while in the coastal rivers are jurisdiction of the A.C.A (Agència Catalana de l'Aigua), an agency of the Autonomous Government of Catalonian.Because our genetic results suggest "basin" as a primary source for population structure, we focused on "basin" (Ribagorzana, Pallaresa, Llobregat and Ter River basins) as a work unit, which also facilitates the administrative collaboration between the environmental and water agencies.The location (see Fig. 1) and facilities of the two regional hatcheries should easily accommodate and separately manage several stocks, despite their previous expertise, which focused on one stock.
For the simulation, we first selected a source location for each river basin, which was a D r a f t 12 challenging task (Allendorf and Ryman 1987;Griffith et al. 2009;Laikre et al. 2010).
The selection criteria involved Hardy-Weinberg and gametic equilibria; and low impacts involved with on-going hatchery release of the foreign stock (see results in Table 1).Furthermore, genetic refuge locations were preferred (Araguas et al. 2008).
We finally prioritized sources located on river drainage mainstems, because brown trout in mainstem populations were typically larger than in small tributaries (Parra et al. 2009), which can increase offspring abundance in the first generation at hatchery facilities, i.e. large trout should have increased fecundity (Lobón-Cerviá et al. 1997).
Abundant progeny from as many parents as possible might facilitate further F1 manipulation to prevent undesirable genetic diversity losses in hatchery stocks (Araki et al. 2008, Fraser 2008).A large amount of stock diversity was also desirable to prevent further changes by selection and domestication (Fraser 2008;Araki and Schmid 2010).
Because of the high amount of foreign stock introgression in the single location studied at the Llobregat River basin we decided to join this basin to the Ter River for the simulated supplementation program as a single management unit for all coastal basins.Subsequently, three captive native stocks were simulated, one to stock the Noguera Ribagorzana River, one the Noguera Pallaresa, and one to stock the coastal basins.
According with results on genetic structure, the simulated native stocks supplementation program allowed assessment of genetic effects at intra and inter basins scales covering aspects as: (i) maintenance of the population structure at regional scale, (ii) preservation of local singularities, and (iii) recovery of native gene pools in heavily introgressed locations with foreign alleles.In this later case, at Filià location we simulated the recovery using a native stock derived from a location in the same river, but at the D r a f t 13

Simulated supplementation program
Our simulation evaluated the long-term effects (100 generations) on genetic population structure of a potential supplementation program using the three native stocks characterized above.Initially, a Ne of 100 was assumed for each location, well in the range of estimated brown trout Ne at each study location (see results in Table 1).The supplementation effort was adjusted to accept the one-migrant-per-generation (OMPG) rule, considered enough to maintain polymorphism and heterozygosity levels within wild locations, while facilitating divergence (Mills and Allendorf 1996).This level of hatchery introgression per generation was consistent with previous estimates in the study area resulting from release of foreign hatchery fish (1 to 5%, Araguas et al. 2004).
For each study location using HYBRIDLAB software (Nielsen et al. 2006), we first simulated samples of 100 individuals at generation 0 from current allele frequencies.In addition, for each native stock we simulated 1000 individuals from the current genotypes representing the source locations.Although a Ne size of 1000 fish for stock might be perceived as unrealistic due to the abundant evidence for genetic changes in salmonid hatchery stocks (Naish et al. 2007;Araki and Schmid 2010), it allowed us to address genetic change in wild populations resulting from fish released from native stocks kept genetically similar to the native source.In addition, this approach might also have utility in evaluating genetic change induced by repetitive supportive breeding from a few local sources.
All wild and stocked fish simulated at generation 0 were aged as adults, and randomly sexed as male or female.Input data were subsequently converted in an FSTAT format file used for simulations in NEMO v2.2 software (Guillaume and Rougemont 2006).In each management unit (Noguera Pallaresa, Noguera Ribagorzana and Coastal basins), D r a f t 14 and in each of their wild population, we simulated 100 generations of native stock reinforcement by setting the local immigration rate (m) from the assigned stock to 0.01 using the NEMO software "breed-disperse" life cycle event.Ne for wild and stock populations was maintained each generation at 100 and 1000 individuals, respectively.
All loci assorted independently (recombination rate adjusted to 0.5), and simulated genotypes at each generation were stored for further analyses.
The above reinforcement scenario was compared with simulated management in the absence of supplementation, and drift as a single evolutionary force during 100 generations based on negligible dispersal and gene flow reported between resident trout populations even at hydrogeographic distances of a few kilometres (Knouft and Spotila 2002;Vera et al. 2010;Vollestad et al. 2012).Gene pool evolution within each location was subsequently simulated as indicated above for the supplementation program, but local immigration from the assigned stock was not permitted (m = 0).Additional simulations were performed for Ne of 50, and 1000 fish in wild populations for both scenarios, with and without supplementation with native stocks.Finally, three additional data sets were simulated based on maintaining current foreign stock releases into all wild locations (m = 0.01 per generation), and respectively considering Ne of 50, 100, and 1000 fish.In summary, 3 x 3 (three scenarios x three Ne) simulated genotype data sets were collected.
Genetic diversity levels at Generation 0 and 100 for each location and scenario were estimated using FSTAT.In all simulations, gene pool stability between generations 0 and 100 in each wild location, and in each simulated stock, was estimated by F ST values (Weir and Cockerham 1984).Significance levels were obtained from 1000 permutation D r a f t 15 tests in FSTAT.Patterns of population structure among locations followed after 100 simulated generations, and for each scenario were estimated by gene diversity analyses (FSTAT).The proportion of divergence between locations within (SC) and among (CT) river basins was examined using hierarchical AMOVAs (Arlequin Software).In addition, population relationships among locations in simulated scenarios (with and without reinforcement) were depicted by nonparametric multidimensional scaling (MDS) from the Da genetic distance matrix between location pairs, as indicated above in the study of captured wild samples.Furthermore, to assess local genetic changes in scenarios with and without supplementation with native stocks, and Ne of 100, we estimated the remaining proportion of the simulated generation 0 and the contribution of the native stock , both as q-values following Sanz et al. (2009) as indicated above for estimating the current contribution of the foreign hatchery stock on wild locations.We should note that the estimated native stock q-value associated with each location under drift was not related to any effects of fish release, but to historical evolutionary processes within and among basins, driving the current genetic divergence between locations.The estimated native stock q-value under drift provides only a reference value to compare with q-value obtained under reinforcement.

Genetic diversity and population structure of wild populations
Brown trout populations at the study locations were short-lived, with a clear dominance of age groups 0+ to 2+, with the exception of the Vallter location, where older fish (> 2+) were also abundant (Table 1 Genetic variability in the current hatchery stock was also high (A R =7.78, H E =0.682), and some abundant allele variants detected in this foreign stock (Str73*146,Str591INRA*158,SsHaeIII14.20*312,SsHaeIII14.20*324,SsHaeIII14.20*322,and SsoSL438*105) were rare in wild trout.
Following Bonferroni correction, genotype distributions at three study locations (Tor, Erta, and Nuria) deviated from Hardy-Weinberg expectations due to homozygote excess (supplementary Table S1).MICRO-CHECKER suggested the presence of null alleles in the Nuria Str15INRA locus, and the Tor SsoSL417 locus.Nevertheless, evidence of null alleles at these two loci was not detected at other study locations.In Nuria, Riutort, Tor, Filià, Cavallers, and Erta, significant gametic disequilibria was observed after applying a Bonferroni correction for multiple tests.High hatchery ancestry (q-values) were detected in fish sampled at Riutort and Filià (0.30 at both study locations) (Table 1), but in the other study locations the estimated hatchery ancestry for captured fish was below 5% (0.05).STRUCTURE results suggested two genetically distinct units presented in Riutort, Filià, Nuria and Tor locations.The abundance of rare homozygotes for alleles common among hatchery fish indicated recent releases in these streams responsible for observed Hardy-Weinberg and gametic disequilibria.Estimated Ne ranged from 18.7 in Filià to 293.0 in Ermita, and a larger indeterminate estimate in Ainet, i.e. 243.9 to α.
Reduced Ne were related to locations with evidence for recent release of hatchery fish (K = 2; Table 1).
Significant pairwise differentiation was detected in all but one comparison, i.e.Ermita-Palomera (Table 2).Despite high hatchery q-values observed in Riutort and Filià, all STRUCTURE (Fig. 2) and MDS analyses based on D a distances (Fig. 3, 2006 collections) were largely congruent.Gene pools from study locations were grouped following a hydrogeographic pattern (Ter, Pallaresa, and Ribagorzana), however two outlier samples from different basins were grouped together, i.e.Manyanet and Erta, and a fifth group comprised the high hatchery ancestry locations Riutort and Filià (qvalue = 0.300, see Table 1).

Simulated long-term effects of supplementation with native stocks
None of the study locations had the complete requirements to be an ideal source location for native stocks (Table 1).The selected source locations for simulated native D r a f t 18 stocks were then determined as follows: Ermita for the Noguera Pallaresa basin due to the location inclusion as a genetic refuge, reduced average hatchery ancestry, and lack of evidence for Hardy-Weinberg or gametic disequilibria; Conangles for the Noguera Ribagorzana River due to its location in a genetic refuge area in the river mainstem, and Vallter for Coastal basins due to low incidence of hatchery releases, the largest Ne estimate among the three study locations in the coastal management unit, and inclusion as a genetic refuge.Permutation tests in FSTAT indicated that average genetic diversity from source locations (see Table 3) was not significantly different from the average diversity at other study locations (P > 0.05), and divergence among source locations (G ST ) resulted in the observed population differentiation among the eastern Pyrenean trout populations (P > 0.05).Despite Ne for stocks fixed at 1000 fish, small but significant (P < 0.05) F ST value between generation 0 and 100 was observed in all simulated stocks (including the foreign one).However, these allelic changes did not significantly modify stock diversity levels (allele richness and gene diversity) as indicated by FSTAT permutation tests.
Following 100 generations, simulated genetic drift in wild populations of 100 individuals caused loss in genetic diversity, and significant allele frequency changes (P < 0.001) at all locations, indicated by F ST coefficients (Table 4).Simulations showed supplementation with native stocks maintained levels of diversity (heterozygosity and allele richness) at most locations, but again, significant F ST values were detected (P < 0.001), even at the native stock source location.However, genetic changes resulting from supplementation with native stocks were generally less severe (decreased F ST value) than alterations observed from maintaining supplementation with the current exogenous stock (Table 4).Short time supplementation with native stocks (10 D r a f t 19 generations) did not significantly alter diversity patterns within and among locations (Table 3, Fig. 3 G10 N100).Estimates of ancestries in the populations following 100 generations in scenarios with and without (drift) supplementation with native stocks added information on genetic changes occurred in each location (Table 5).Despite losses in genetic diversity for simulations involving genetic drift, the estimate remaining ancestry for the local original (generation 0) gene pool was close to 1.0 at each location, and the estimated native stock ancestry was only high for the respective source location except in Palomera.Here, the estimated native stock ancestry was high because of its current genetic similarity with the Ermita trout, which was used as source for the simulated native stocks in the Noguera Pallaresa River basin (F ST = 0.003; Table 2).
However in all basins, supplementation with native stock produced declines of the original gene pool ancestry for all locations, particularly in locations currently showing large genetic divergence from the source population (Tor and Manyanet in the Noguera Pallaresa basin; Cavallers, Nicolau, and Erta in the Ribagorzana basin, and Riutort in the coastal basins).As expected, these declines were related to increased estimates in native stock ancestry.Altogether indicated that supplementation with native stocks was efficiently eroding local gene pools.
In simulations of Ne of 100 individuals, FSTAT comparisons confirmed reduced average allele richness within locations in the simulated scenario with and without (drift) native stocks supplementation (Table 3).The simulation without supplementation maintained total diversity (H T ) in the region by significantly reducing local diversity  3).Hierarchical AMOVAs also showed that without supplementation, drift alone was not sufficient to completely erode genetic differences within (SC component) and among (CT component) into the management units.Nevertheless, supplementation with native stocks resulted in increased genetic homogenization of populations within management units (strong reduction in the SC component).MDS results from Da genetic distance matrices between locations (Fig. 3) clearly depicted all changes in population structure in the study area resulting from supplementation with exogenous and native stocks.
Consequently, following long-term supplementation by native stocks, populations were clustered based on management units, while maintaining releases from a common foreign stock, resulted in genetic relationships among populations from distinct management units.
In simulations involving population sizes of 50 fish, the effects of drift were evidenced by reduced average allele richness and increased divergence among locations under all simulation scenarios (Table 3), however a large "among management units" component (CT) was obtained from supplementing with native stocks (Fig. 3).This diversity component (CT) was not preserved in the other two scenarios (with or without supplementation with the foreign stock).Nevertheless, simulations with population sizes of 1000 fish exhibited decreased genetic changes in gene diversity and hierarchical population structure under the drift scenario compared to any supplementation program (Table 3).

Current wild population status
Brown trout in the eastern Pyrenean rivers showed levels of diversity at microsatellite loci well within the range detected in Mediterranean populations from other countries (Jug et al. 2005;Apostolidis et al. 2008).Populations showing the greatest genetic impacts following releases of the Atlantic foreign stock (Riutort and Filià locations, Table 1) also exhibited the highest estimates of heterozygosity.Recent foreign stock releases remained common evidenced by the two breeding units detected at four study locations (Nuria, Riutort, Tor, and Filià), which were supported by high estimates of hatchery ancestry or Hardy-Weinberg and gametic disequilibria.Despite estimated introgression rates below 5% from all study locations, with the exception of Riutort and Filià, wider geographical surveys suggested average introgression rates from Atlantic foreign stocks exceeded 10% in eastern Iberian trout populations (Sanz et al. 2002;Aparicio et al. 2005;Almodovar et al. 2006).Recent releases likely contributed to the small Ne estimated in Tor and Filià, as expected from theoretical predictions on hatchery releases and long-term supportive breeding programs (Ryman and Laikre 1991;Wang andRyman 2001, Waples andEngland 2011).However, a large amount of native genetic diversity was still preserved within the eastern Pyrenean rivers, where native brown trout remained morphologically distinct from hatchery fish (Aparicio et al.

2005).
Ne estimates suggested adequately self-sustained trout populations in some locations, including Ter, Palomera, Ermita, Ainet, Manyanet, Nicolau, and Erta.In the Iberian Peninsula, age at maturity for brown trout is between 1+ and 2+ for females, and 2+ to D r a f t 22 3+ for males (Lobón-Cerviá et al. 1986;Parra et al. 2009).The adult fish older than 2+ at time of capture represented less than 40% in all study locations with the exception of Ter.In addition, fish available for angling should be even less abundant, because at the study locations, the oldest fish (> 3+) reach catchable length (22 cm) i.e. in Iberian rivers (Lobón-Cerviá et al. 1986;Rocaspana et al. 2006;Parra et al. 2009).A decrease in catchable fish (< 10%) abundance is common among fished trout populations compared with increased abundance in unfished stretches (Almodovar and Nicola 2004).Nevertheless, environmental factors and anthropogenic-mediated perturbations, in addition to angling contribute to reductions in catchable trout throughout Spanish rivers (Almodovar and Nicola 1999;Nicola et al. 2009;Ayllón et al. 2012), including Pyrenean waters (Garcia de Jalon et al. 1988Jalon et al. , 1996;;Alonso et al. 2011).Cumulatively, these factors promote social support for supplementation practices.

Genetic effects of the supplementation with native stock
Our results indicated that following short-term supplementation, sporadic supportive breeding programs from local sources would not markedly change the gene pools of the supplemented populations.In addition, as Caudron et al. (2011Caudron et al. ( , 2012) ) reported in French brown trout populations following supportive breeding programs, simulations predicted recovery of native diversity in the Filià and Riutort populations highly impacted by current releases of a foreign hatchery stock.In the French streams, trout densities increased 20-to 55-fold, and the majority of juvenile trout (78-89%) were first-generation descendants of released trout.However, recovering native alleles in these previously highly introgressed populations might not restore native local ancestry, but result in genetic swamping of the local populations along the management unit by the native stock gene pool.Swamping was particularly relevant in locations reinforced D r a f t 23 with a native stock originated from a source in other river basin as Riutort in our simulations; in this location a notorious reduction of the original local ancestry was observed.Within management units, such an effect was even more evident in genetically unique native populations, as Manyanet and Erta in our study.In these outlier populations, the native local gene pool were reduced following long-term supplementation, while a notable increase in the estimate native stock ancestry was observed.The expected result following long-term supplementation with native stocks was then a significant reduction in diversity among locations within management units.
However, in the study region, current genetic differentiation was related more with local divergence within rather than isolation between basins.Similar results have already been observed in the wild; for example, populations of Coho salmon (Oncorhynchus kisutch) in the Puget Sound of Washington State that underwent extensive hatchery propagation share more of their ancestry recently than they did historically (Eldridge et al. 2009).Hansen et al. (2009) reported substantial local changes in Danish brown trout populations following long-term supplementation with Danish stocks.Supportive breeding efforts in French Mediterranean trout populations showed over short-term time scales, the genetic and demographic effects were restricted to river stretches 2 km downstream from the release locations (Caudron et al. 2012), however other evidence indicated expansion of hatchery genes by hybridized fish occurring over longer periods of time following releases (García-Marín et al. 1998;Allendorf et al. 2004;Araguas et al. 2008;Perrier et al. 2013).
Fishery managers should be aware that simply maintaining global estimates of gene diversity indices (H T , H S and G ST ) does not insure conservation of local genetic variation.For example, significant changes were not detected in diversity indices for D r a f t 24 supplementation simulations, either using current foreign or native stocks, despite losses of native alleles in all locations.Therefore, native stocks do not mean local stocks, even at short hydrogeographical distances.Increased genetic differentiation between Pyrenean brown trout populations suggested isolated populations at hydrogeographical separation of just a few kilometres.In fact, the only non-significant divergence detected in the entire study was between Palomera and Ermita locations, 1.2 km apart along the same mainstem of the Vallfarrera River.Hierarchical partitioning of genetic diversity revealed another relevant source of divergence between tributaries in the same river basin.For example, the Tor River sample was collected 6 km from the Ermita location in a tributary of the Vallfarrera River, and the estimated F ST value between Tor and Ermita samples was 0.175.

Selective forces
Our study was based on microsatellite loci, which are typically considered selective neutral markers.However, there might be adaptive and selective processes modulating and modifying our observations.First, we should consider the adaptive value of trout genetic singularities in the Pyrenean locations.Certainly it remains to be demonstrated, yet the observed divergences at microsatellite loci among these trout populations were congruent with distinctions based on protein coding loci variation reported earlier by Araguas et al. (2004).Local adaptation in salmonids occurs at several spatial scales (from a few to thousands of kilometers), with local populations often manifesting a fitness advantage over foreign populations (Fraser et al. 2011;Perrier et al. 2013).In brown trout, selective processes have been detected at small spatial scales for loci related to immune systems, i.e.MHC or TAP (Hansen et al. 2007;Jensen et al. 2008b; D r a f t 25 (Meier et al. 2011).Therefore, the high divergence (F ST ) observed between wild Pyrenean populations could be related with local adaptation.In this sense, the divergence observed among locations within each management unit is often as high as values reported among wild locations and the current foreign stock, questioning whether the choice of a single native source for each management unit would be a successful management decision.Unfortunately, a supplementation program involving all requested and necessary native stocks to preserve all local genetically differentiated populations at the intra-and inter-basin levels currently appears unfeasible due to the reduced infrastructure (two hatcheries), and high economic costs of maintaining several native stocks.
Selective domestication induced by culture conditions is often detected in hatchery stocks (Araki and Schmid 2010).Putative diversifying selection between wild populations and hatchery trout stocks (e.g.Hansen et al. 2010) can result in additional risk due to reduced average fitness of reinforced wild populations, even at low fitness differences between wild and hatchery fish (weak selection), facilitated for instance by reiterated immigration of wild fish into captive stock (Ford 2002).Surprisingly, releasing hatchery stocks phenotypically differentiated from wild populations might result in less harmful effects because the phenotype divergence could be associated with traits maladaptive to wild conditions, and hence strong selective pressures purging released fish before the reproductive season (Baskett and Waples 2012;Baskett et al. 2013).In the study territory, the current hatchery stock is phenotypically divergent from wild populations (Aparicio et al. 2005), and its foreign origin and long-time maintenance in captive conditions might result in maladaptation to wild conditions.
Based on the simulation results of Baskett et al. (2013), such maladaptation could D r a f t justify the reduced estimates of hatchery introgression detected in Pyrenean trout populations despite the long period of intensive releases (Sanz et al 2002, Aparicio et al 2005), and questions the suitability of replacing foreign stock by native.Nevertheless, reduced impact of hatchery releases has been reported among North-European populations phylogenetically close to source populations of our foreign stock (e.g.Ruzzante et al. 2001).

Conclusions: management prospectives
Results of our simulations indicated that in wild populations comprised of an effective population size of 100 individuals, genetic drift would result in significant changes in the study basin gene pools during the next one hundred years.However, drift alone might better conserve distinct populations among locations within management units.
Based on simulation predictions, an increase in effective population size to 1000 individuals decreased the likelihood of significant changes in the study basin gene pools.In this case, within and among population changes were lower than those caused by regional reinforcement policies.Such results suggested that habitat restoration to insure large effective population sizes might be enough to protect native genetic diversity.Sociological studies on German anglers indicated that limiting brown trout management to improve or maintain good habitat quality would receive anglers' support, primarily when catches were increased, and fishing experiences were positive (Baer and Brinker 2010), but even then, a large proportion of anglers were not opposed to supplementation practices (Arlinghaus and Mehner 2005).Therefore, avoiding hatchery-releases might compromise angler support for further management measures.
In the eastern Pyrenean rivers, a combination of genetic refuges to protect native trout diversity in some stretches, and stocking practices in other river sections undergoing D r a f t 27 intensive fishing efforts, has favoured restrictive measures in a per day bag and increased minimum catchable size limits, and additional fishing river stretches designated as catch-and-release areas (Araguas et al. 2009).In addition, the current normative framework involving trout populations in the Spanish eastern Pyrenean serves to support recreational fisheries, and anglers' societies are the key stakeholders and one of the main lobbyists in decision-making on regulatory measures in the Pyrenean river basin.Because of pending taxonomic revision, any distinct trout taxa from the study region were not included in the most recent red list of European freshwater fish (Freyhoff and Brooks 2011), limiting conservationists' arguments against extensive hatchery trout releases and translocations.
Often criticized due to the conservation risks on remnant biodiversity (Laikre et al. 2010), hatchery stock supplementation is maintained as a traditional fisheries practice, because it is perceived as a prophylactic measure for human induced damage, including fisheries itself, on wild populations (Arlinghaus et al. 2002).From a put-and-take fish stocking point of view, which is maintained in some eastern Pyrenean river stretches, it may be irrelevant which stocks, foreign or native, are used for release.In fact, current foreign hatchery stocks are well-adapted to hatcheries, and hatchery personnel have sufficient expertise to generate large output to enhance regional trout fisheries.
However, compared with maintaining supplementation from foreign stock, supplementation from native stocks might serve as a much better balance between the social benefits of angling and biological damage to native diversity in the regional river basins.Together with measures that limit native releases to locations of intensive fishing that preclude the survival of released fish before the spawning season, the replacement of current foreign with native trout stocks provides an important Alonso, C., García de Jalón, D., Álvarez, J., and Gortázar, J. 2011.A large-scale approach can help detect general processes driving the dynamics of brown trout populations in extensive areas.Ecol. Freshw. Fish,20,[449][450][451][452][453][454][455][456][457][458][459][460] Aparicio, E., García-Berthou, E. Araguas, R.M., Martínez, P., and García-Marín, J.L.
significant genetic differentiation from current hatchery foreign stock (F ST range 0.116-0.426).Hierarchical AMOVA results for the entire study area reflected higher significant divergences within (SC component = 86.14%,Table3) than among basins (CT component = 13.86%).Due to the increased number of study locations in the Pallaresa and Ribagorzana basins, a hierarchical AMOVA was possible for each of the two basins.In both basins, results showed increased differentiation within compared to among tributaries.In the Pallaresa River (G ST = 0.160), where the two more hydrogeographically separated study locations were Palomera and Filià (91.6 km), only 23.35 per cent of the total differentiation was assigned to differences among tributaries.In the Ribagorzana River, increased differentiation was observed among locations (G ST = 0.357), even though within this basin, 64.95% of the population differentiation was assigned within tributaries.In this basin, the largest hydrogeographic separation among study locations occurred between Conangles and Cavallers, 50.1 km apart.

(
H S ), and increasing divergence (G ST ) among locations.The native stocks supplementation program showed reduced effects on gene diversity indices (H T , H S and foreign stock did not result in significant changes in allele richness and gene diversity indices.However, hierarchical AMOVAs indicated releases from foreign stock into all native population locations nearly eliminated the genetic distinction among management units (CT component close to 0; Table the value of preserving local diversity among anglers, and with the presently implemented genetic refuge policy, should be an additional step in the albeit slow transition to a fisheries model focused on local self-sustaining trout populations, and regional habitat management.Native stock development can stimulate regional hatchery expertise and hatchery personnel to manage native fish, and facilitate short-term supportive breeding programs to recover endangered native trout populations.., and Nicola, G.G. 1999.Effects of a small hydropower station upon brown trout Salmo trutta L. in the River Hoz Seca (Tagus basin, Spain) one year after regulation.Regul.Rivers, 15, 477-484.Almodovar, A., and Nicola, G.G. 2004.Angling impact on conservation of Spanish stream-dwelling brown trout Salmo trutta.Fish.Manag.Ecol., 11, 173-182.Almodóvar, A., Nicola, G.G., Elvira, B., and García-Marín, J.L. 2006.Introgression variability among Iberian Brown trout Evolutionary Significant Units: the influence of local management and environmental features.Freshw.Biol., 51, 1175-1187.Almodovar, A., Nicola, G.G., Ayllón, D., and Elvira, B. 2012.Global warming threatens the persistence of Mediterranean brown trout.Glob.Change Biol., 18, 1549-1560.

Fig. 2 .
Fig. 2. Current individual sample relationships indicated by STRUCTURE analysis considering 2, 3, 4 and 5 genetic groups.Each individual is represented by a vertical bar partitioned into segments according to the proportion of the genome assigned to each of the identified clusters.Location codes are defined in Table1.

Fig. 1 .Fig. 2 .
Fig. 1.Geographic locations of the brown trout collections and the two hatcheries (HAT) in the study region.Location codes are defined in Table1.254x190mm (96 x 96 DPI) ).On average, 77.0 % of fish sampled were in age classes 0+ to 2+, and ranged from 55.4% in Vallter to 99.5% in Nuria.Genetic diversity differed among locations (Table 1), with the highest variability exhibited in Filià =8.40, H E =0.769), and the lowest variability in Vallter (A R =3.23, H E =0.393).

Table 2 .
Weir and Cockerham, 1984)tiation (F ST values,Weir and Cockerham, 1984)between brown trout captured in study locations and the current foreign stock.Location code as in Table1.* P < 0.05

Table 3 .
Gene diversity analyses and population structure in the study region using current genotypes, or simulated after 100 generations in scenarios without (Drift), and with supplementation by native stocks, or foreign stock.A R : Average allelic richness. http://mc06.manuscriptcentral.com/cjfas-

Table 4 .
Diversity levels in each location and stock at simulated Generations 0 and 100, in scenarios without (drift), and with supplementation by native stocks or by a foreign stock.H E : expected heterozygosis (A R , Average allele richness).F ST : estimated genetic divergence between simulated Generation 0 and 100 at each location and stock.In bold, source location for each simulated native stocks.Effective population sizes of 100 fish for wild locations and 1000 fish for stocks.*P < 0.05. http://mc06.manuscriptcentral.com/cjfas-

Table 5 .
Estimate ancestries in each populations after 100 generations in scenarios without (Drift) and with supplementation with native stocks, and Ne of 100 fish.Np: average remaining ancestry of the local gene pool; Ss: average ancestry of the simulated native stock.In each basin, location in bold was the source of the native stock.