Conservation Genetics

, Volume 15, Issue 5, pp 1095–1109

Origins and genetic diversity among Atlantic salmon recolonizing upstream areas of a large South European river following restoration of connectivity and stocking

Authors

    • Département de BiologieUniversité Laval
    • UMR 0985 ESEINRA
  • Jérôme Le Gentil
    • UMR 0985 ESEINRA
    • UMR 1224 EcobiopINRA
  • Virginie Ravigne
    • UMR BGPICIRAD
  • Philippe Gaudin
    • UMR 1224 EcobiopINRA
  • Jean-Claude Salvado
    • UMR 1224 EcobiopINRA
    • UMR 1224 EcobiopUniversité de Pau Et des Pays de l’Adour
Research Article

DOI: 10.1007/s10592-014-0602-3

Cite this article as:
Perrier, C., Le Gentil, J., Ravigne, V. et al. Conserv Genet (2014) 15: 1095. doi:10.1007/s10592-014-0602-3

Abstract

The restoration and maintenance of habitat connectivity are major challenges in conservation biology. These aims are especially critical for migratory species using corridors that can be obstructed by anthropogenic barriers. Here, we explored the origins and genetic diversity of Atlantic salmon (Salmo salar) recolonizing upstream areas of the largest South European Atlantic salmon population (Adour drainage, France) following restoration of connectivity and stocking. We genotyped 1,009 juvenile individuals, sampled either in continuously inhabited downstream sites or in recently reconnected and recolonized upstream locations, at 12 microsatellite loci. We found significant fine scale genetic structure, with three main genetic clusters corresponding to the Nive, Nivelle and Gaves rivers. Within each of these clusters, samples collected in continuously inhabited and recently recolonized sites had comparable allelic richness and effective population sizes and were only weakly differentiated. Genetic structure among basins was also similar among continuously inhabited and recently recolonized sites. The majority of the individuals sampled from recently recolonized sites were assigned to neighboring continuously inhabited downstream sites, but noticeable proportions of fish were assigned to samples collected in more distant sites or identified as putative hybrids. Overall, this study suggests that the restoration of accessibility to upstream areas can allow for the recolonization and effective reproduction of Atlantic salmon from proximate downstream refugia, which does not decrease local diversity or disrupt existing genetic structure.

Keywords

RecolonizationGenetic diversityDamConnectivityAssignmentSalmo salar

Introduction

The fragmentation of habitat is one of the major human threats to wild populations (Vitousek et al. 1997; Ewers and Didham 2006; Fahrig 2003) and is often due to the construction of artificial barriers, which can result in patch size reduction and patch isolation (Fahrig 2003). For a wide range of species, habitat fragmentation can also modify dispersal and gene flow (Coulon et al. 2010; Van Oort et al. 2011; Pépino et al. 2012). Moreover, population size and effective population sizes as well as genetic diversity can be affected (Couvet 2002; Blanchet et al. 2010; Dixo et al. 2009; Tsuboi et al. 2013; Whiteley et al. 2013). Eventually, fragmentation may affect the evolutionary trajectories of populations and lead up to local extinction. Therefore, to mitigate habitat fragmentation and its impacts on local populations, numerous conservation actions have been undertaken to conserve, restore (Clewell and Aronson 2006; Aronson 2011; De Groot et al. 2013), and increase habitat connectivity (Brown et al. 2013). Therefore, effective conservation and restoration of habitat connectivity requires knowledge of how fragmentation and reconnection impacts local species.

Freshwater ecosystems are particularly subject to fragmentation due to anthropogenic activities (Nilsson et al. 2005; Hall et al. 2011). One of the main causes of fragmentation of freshwater habitats has been the widespread construction of dams for irrigation purposes, drinking water retention, watermills, hydroelectric power plants, and recreational activities. Their impacts on freshwater ecosystems are well recognized, and range from modifications to population genetic diversity (Horreo et al. 2011; Neraas and Spruell 2001; Wofford et al. 2005) to changes in species assemblages (Poulet 2007; Boet et al. 1999; Brown et al. 2013; Grenouillet et al. 2008). Barriers sizes and permeability are important parameters affecting the movement of fish species (Raeymaekers et al. 2009) and thus their demographic and genetic characteristics. In addition, swimming ability of the fish and their capacity to disperse through barriers are often species specific (Haro et al. 2004). Accordingly, recent studies demonstrated that there are species-specific modifications of gene flow and genetic structure due to weirs (Blanchet et al. 2010) and waterfalls (Gomez-Uchida et al. 2009). In particular, anadromous species like Atlantic salmon, which migrate between spawning grounds located in rivers and feeding zones at sea (Jonsson and Jonsson 2011), can be highly impacted by reduced connectivity as a result of dams (Brown et al. 2013; Hall et al. 2012). Therefore, many conservation studies and programs aimed at documenting the impact of artificial barriers on the demography and evolutionary trajectories of anadromous fish populations have highlighted the importance of the restoration of habitat connectivity.

In many regions of the world, there are regulations to re-establish connectivity among freshwater habitats that have been disconnected by weirs and dams to restore fish movement within watersheds. Migratory fish species have been the main targets of these restoration regulations since they are highly impacted by habitat fragmentation and because they are of high biological, economic and societal interest. Partial restoration of river connectivity can be provided by the addition of fishways (Coutant and Whitney 2000; Brown et al. 2013; Larinier and Boyer-Bernard 1991). Even though fishways have proven their effectiveness, dam removal remains the best option to effectively improve fish movement (Brown et al. 2013; Oldani and Baigún 2002; Mallen-Cooper and Brand 2007). In addition to restoring fish movement, dam removal can also allow for the restoration of spawning grounds that were previously buried under sediment (Bednarek 2001). Genetic analyses of these recolonizing individuals together with baseline information from samples collected in geographically close and continuously inhabited sites may help to identify the origin and genetic diversity of these colonists and help us to better understand the recolonization process (Kiffney et al. 2009; Perrier et al. 2010; Griffiths et al. 2011; Winans et al. 2010; Ikediashi et al. 2012). Several recent studies have found that the Atlantic salmon recolonizing rivers mainly came from nearby rivers but that some fish also came from more distant source populations (Perrier et al. 2010; Griffiths et al. 2011; Ikediashi et al. 2012). However, there have not yet been any studies aiming at identifying the origin, among productive downstream areas, of salmon recolonizing recently reconnected upstream parts of rivers. In addition to determining the origin of the immigrants, such studies within a single river system will provide insight into the diversity and effective population size in recently recolonized river sections, two population genetic parameters predicted to influence the success of recolonization (Naish et al. 2013; Oakley 2013; Fraser et al. 2007).

As an alternative approach to sustain declining or re-establish extinct populations of migratory salmonids, the stocking of wild or captive hatchery-reared individuals is often used. However, although stocking may help reestablish populations, it has been highly criticized because of its potential negative effect on the fitness of wild populations in the long term (Aprahamian et al. 2003; Araki and Schmid 2010; Fraser 2008). Indeed, the fitness of stocked fish may be reduced due to the effects of unintentional selection and domestication during early life in hatcheries that affects survival, migration and reproductive success of stocked fish (Araki et al. 2008; Milot et al. 2013; Thériault et al. 2011). Moreover, stocking with individuals originating from highly differentiated stocks has resulted in the loss of neutral genetic integrity in salmonids (Marie et al. 2010; Bourret et al. 2011; Perrier et al. 2013a, b; Hansen et al. 2009). Since fine scale genetic structure may exist within large basins (Dionne et al. 2009; Ensing et al. 2011; Vaha et al. 2007; Primmer et al. 2006), even the use of local wild fish to produce stocked fry may ultimately result in the disruption of local genetic structure and local adaptation if subpopulations are not carefully considered (Eldridge et al. 2009; Pearse et al. 2011). Therefore, genetic analyses can be used not only to document the origin of fish recolonizing recently reconnected sites but also to estimate the relative contributions of “natural” recolonization versus stocking (Beaudou et al. 1994; Griffiths et al. 2011).

In this study, we explore the distribution of genetic diversity within the largest Atlantic salmon catchment in Southern Europe (Adour, France), to determine the origin of Atlantic salmon recolonizing upstream areas following recent restoration of connectivity and stocking. The Adour hydro-geographic basin has a surface of 16,880 km2 and harbours a large wild Atlantic salmon population, which is an important target of both recreational and commercial fishing (Vauclin 2007). On the basis of relatively few samples, Perrier et al. (2011b) revealed that Atlantic salmon samples collected in Nive, Nivelle and Adour rivers (see the map Fig. 1) clustered together and were genetically differentiated from other French stocks. Nevertheless, given the size of the Adour basin and the distance among spawning grounds where Atlantic salmon have been observed, a fine scale population genetic structure may exist. Of principal interests the connectivity of several upstream areas previously inaccessible to migratory fish due to the presence of impassable dams have been restored since 1986 (see Fig. 1; Table 1 and methods to locate impassable and passable dams). In parallel, salmon fry and parr, mainly the offspring of wild caught parents, have been released since the 1970s into various tributaries of the Adour catchment in order to sustain populations and to aid in the reestablishment of new populations in recently reconnected sites (Marty 1984; Beall et al. 1995; Davaine et al. 1996). How this history of human intervention has impacted the Salmon populations in this area is unknown. Therefore, the specific aims of the present study were to: (1) investigate the fine scale genetic population diversity and structure within the Adour catchment and identify conservation units, (2) determine whether Atlantic salmon recolonizing recently restored habitats originated from proximate downstream sites in the basin or from the other basins, (3) test whether the genetic diversity, effective population size, and genetic structure within recently recolonized areas are reduced compared to continuously inhabited sites, (4) test whether stocking impacted the distribution of genetic diversity within the basin, and (5) discuss the relevance of our results for the sustainable management and conservation of this important Southern-European Atlantic salmon stock and generally discuss the effects of the restoration of connectivity in freshwater fishes.
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-014-0602-3/MediaObjects/10592_2014_602_Fig1_HTML.gif
Fig. 1

Location of the sampling sites that were continuously inhabited (blue circles), or which have been recently recolonized (green circles). Dams which were passable (green stars) or impassable (red stars) at the time of sampling are also noted. For site number correspondence, see Table 1. (Color figure online)

Table 1

Characteristics of sampling locations and genetic diversity indices of their corresponding samples including FIS, expected heterozygosity (He), allelic richness (Ar), private allelic richness (PAr), effective size (Ne), and census size per basin (Nc corresponds to the average number of anadromous fish entering the basins, Lange et al. 2011)

Basin

River

Tributary

Coordinates

Location

Distance to river mouth (km)

Year of restoration of accessibility

N genotyped

N full sibs

% of full sibs

N after removing full sibs

FIS

He

Ar

PAr

Ne

Nc per basin

Nivelle

Nivelle

Nivelle

43°20′52″N/01°33′06″W

1

17

30

8

27

22

0.08

0.84

8.59

0.34

33.8

105

  

Nivelle

43°18′33″N/01°31′49″W

2*

22

1992

45

5

11

40

0.01

0.84

8.61

0.31

124.0

  

Lurgorieta

43°18′28″N/01°33′58″W

3*

22

1992

19

9

47

10

0.01

0.85

8.67

0.41

302.6

Adour

Nive

Nive

43°20′44″N/01°26′53″W

4

59

25

0

0

25

0.04

0.80

7.24

0.08

595.8

560

  

Laurhibar/Béhérobie

43°15′40″N/01°24′31″W

5*

67

2001

116

36

31

80

0.03

0.82

7.84

0.27

107.6

  

Arnéguy

43°12′89″N/01°26′74″W

6*

70

1992

174

50

29

124

0.05

0.80

7.28

0.18

595.5

 

Gaves

Saison

43°31′27″N/00°87′19″W

7*

88

1986

27

0

0

27

0.05

0.85

8.43

0.28

142.6

9,200

  

Saison

43°19′51″N/00°91′33″W

8*

105

1995

98

7

7

91

0.02

0.84

8.45

0.30.0

400.5

  

Gave d’Oloron

43°26′85″N/00°68′41″W

9

106

65

3

5

62

0.02

0.83

8.21

0.19

238.4

  

Gave d’Oloron

43°23′11″N/00°63′70″W

10

115

47

7

15

40

0.06

0.82

8.07

0.26

434.8

  

Vert

43°16′31″N/00°68′15″W

11*

123

2001

42

13

31

29

0.07

0.83

8.33

0.26

78.6

  

Gave d’Aspe

43°03′88″N/00°60′42″W

12*

141

1992

47

7

15

40

0.03

0.83

8.04

0.26

87.9

  

Lourdios

43°07′69″N/00°66′98′W

13*

134

1992

30

4

13

26

0.00

0.87

8.93

0.23

89.0

  

Gave d’Ossau

43°11′98″N/00°47′76′W

14

138

56

19

34

37

0.07

0.85

8.34

0.28

104.7

  

Gave d’Ossau

43°08′60″N/00°42′10′W

15*

149

1993

54

6

11

48

0.08

0.83

8.12

0.31

3583.8

  

Gave de Pau

43°09′60″N/00°17′10′W

16*

157

1992

49

13

27

36

0.03

0.82

7.78

0.15

84.0

Total

   

16

  

924

187

19

737

0.04

0.83

8.18

0.26

  

Materials and methods

Study site and Atlantic salmon population

The Adour catchment is a 16,880 km2 river-system. Atlantic salmon (Salmo salar) spawning sites can be found in several major sub-drainages including: the Nive (993 km2), the Saison (631 km2), the Gave d’Oloron (2,000 km2), the Gave d’Aspe (598 km2), the Gave d’Ossau (493 km2) and the Gave de Pau (2,600 km2). The Nivelle River (244 km2) estuary is twenty kilometres away from the Adour estuary and harbours a small Atlantic salmon population. On average over the last decade, 6,500 adults annually return to the Adour Drainage, and 300,000 parr 0+ are produced per year (Barracou 2008; Le Gentil et al. in prep). Thus, the Adour River is currently the most productive drainage in France as well as in Southern Europe. The Adour population is large when compared to most Northern-European populations, although much larger populations do exist (Tonteri et al. 2009).

Numerous dams have been built on the Adour River since the beginning of the eighteenth century, and habitat fragmentation reached its maximum during the 1940’s due to hydro-electric power plant construction. Dam construction decreased or prevented the migratory fish from reaching upstream spawning sites (Fig. 1; Table 1). As a result, the abundance and catch of Atlantic salmon dramatically declined in the 1960’s in the Adour basin (Barracou 2008). For example, more than 10,000 fish were caught annually at the beginning of the twentieth century but <500 fish were caught in 1976 (Marty and Bousquet 2001). The intensification of agriculture, urbanization, industrialization and overexploitation at sea may have also contributed to this decline in Atlantic salmon. Since 1986, in response to changes in regulations, much work has been done to improve the accessibility of upstream spawning sites to migratory fish, principally Atlantic salmon (Barracou 2008). Connectivity restoration mainly consisted of building fishways on dams isolating sectors 2, 3, 5, 7, 8, 11, 12, 13 and 16 (Fig. 1). The dam that isolated sector 6 was destructed. The Dam that isolated sector 15 was equipped with a fish lift. Thanks to these improvements in habitat connectivity sectors 7, 2–3–6–12–13–16, 15, 8 and 5–11 became accessible in 1986, 1992, 1993, 1995 and 2001, respectively.

To sustain the Atlantic salmon populations inhabiting the Adour drainage as well as facilitate the recolonization of upstream habitats, stocking programs were implemented in the 1970s (Marty 1984; Beall et al. 1995; Davaine et al. 1996; Marty and Bousquet 2001). Eggs, young of the year and smolts were stocked in several tributaries. Until 1990, eggs, fry and smolts produced from local and non-local parents (mainly originating from Scotland) were stocked in the Gaves, Nive and Nivelle. Since 1990, local parents were used in the Nive, Nivelle and Gaves Rivers (Marty and Bousquet 2001). At present, stocking is sustained only in the Gave de Pau. As presented in (Perrier et al. 2013b), little introgression by non-local strains occurred in the Adour populations and this introgression has remained stable over the last decades.

Sampling

During September and October 2005, 0+ and 1+ year old juvenile Atlantic salmon were sampled by cutting off a small part of their pectoral fins (non-lethal sampling). Fins were stored in 95 % ethanol. A total of 1,009 individuals were sampled in 16 locations distributed in the Adour drainage (Fig. 1; Table 1). To increase the number of individuals analysed per location, but to limit biases resulting from sampling individuals from the same family, each sampling site consisted in a section of 10–30 km long. Individuals sampled were wild fish for all but site #14, where a restocking operation occurred in 2004.

DNA extraction and genotyping

DNA extraction, amplification and genotyping were completed according to the procedures described in (Horreo et al. 2008). A total of twelve microsatellites were selected on the basis of consistency of amplification, ease of scoring and variability: Ssa197, Ssa202, Ssa171 (O’Reilly et al. 1996), SSsp1605, SSsp2210, SSspG7, SSsp2201 (Paterson et al. 2004), Ssosl417, Ssosl85 (Slettan et al. 1995), SsaD144b, SSa157a (King et al. 2005), Ssa289 (McConnell et al. 1995). A thirteenth microsatellite was used to distinguish S. salar from Salmo trutta and detect possible hybrids between these species (Perrier et al. 2011a).

Data quality

ML-relate (Kalinowski et al. 2006) was used to detect full-sibs. Full-sibs were subsequently removed from the dataset to avoid bias caused by the overrepresentation of individuals from the same family (Hansen et al. 1997). We used the software Micro-Checker to test for the presence of null alleles (Van Oosterhout et al. 2004). Linkage disequilibrium was estimated using Genepop 3.4 (Raymond and Rousset 1995b), with sequential Bonferroni correction (Raymond and Rousset 1995a). FIS and tests for Hardy–Weinberg disequilibrium were conducted with FSTAT2.9.3.2 based on 1,000 permutations.

Analysis of genetic diversity within samples

Observed (Ho) and expected heterozygoties (He) were estimated using Genepop 3.4. The number of alleles per locus and population, allelic richness (Ar) and private allelic richness (PAr) were estimated using Hp-rare 1.0 (Kalinowski 2005). Effective population size (Ne) was estimated for each location using the LDNe method (Waples and Do 2008) implemented in NeEstimator V2.0 (Do et al. 2013). We used an allele frequency threshold of 0.05. We tested for data normality using a Shapiro test, and then compared average Ar and Ne among recently recolonized and continuously inhabited sites using student’s t tests for normally distributed data and Mann–Withney tests for non-normal distributed data. Census size (Nc) were obtained from Lange et al.’s (2011) and corresponded to the average number of anadromous fish entering in the Nive, Nivelle and Gaves rivers.

Analysis of genetic structure among sites

We used Genepop 3.4 to estimate FST between sites. Significance of FST was estimated using 1,000 permutations. We compared average FST among: (A) recently recolonized sites and sites that were continuously inhabited within each basins, (B) recently recolonized sites from different basins and (C) sites from different basins that remained inhabited by Atlantic salmon. We tested for sstatistical significance using student’s t tests. We performed three AMOVAs using Arlequin v3.5 (Excoffier and Lischer 2010). The first AMOVA was conducted using Nivelle, Nive and Gaves as groups to test for hierarchical structure. The second AMOVA was conducted using the same grouping but on continuously inhabited sites only. The last AMOVA was conducted using the same grouping but on recently recolonized sites only. These last two AMOVAs were conducted to investigate whether genetic differentiation was lower among recently recolonized sites compared to continuously inhabited ones at both hierarchical levels. A neighbor-joining dendrogram based on pairwise Nei (Da) genetic distances (Nei et al. 1983) was constructed with Populations 1.2.30 (http://bioinformatics.org/~tryphon/populations/). Confidence estimates of tree topology were calculated by 1,000 bootstraps of loci. Dendrograms were visualized using TreeView (Page 1996).

Bayesian clustering and assignment of individuals

We examined the clustering of populations and individuals using Baps v2.0 (Corander et al. 2004) and Structure (Pritchard et al. 2000). While Baps rapidly and accurately finds main clusters, the estimation of individual admixture may be less accurate. Alternatively, Structure accurately estimates population and individual admixture but is relatively slow and may have difficulty finding main clusters when some populations are under or over-represented (Kalinowski 2010). First, we used the population clustering option implemented in BAPS to delineate main genetic clusters in the dataset, with a maximum number of potential clusters set to 10. Population and individual admixture was subsequently tested on the basis of these population-clustering results. Second, the Bayesian clustering method implemented in the software Structure was used to delineate K genetic clusters. The best K values were defined according to the ΔK procedure as described by (Evanno et al. 2005), using STRUCTURE HARVESTER (Earl and vonHoldt 2012). A total of 10 runs were computed for each value of K tested, from 1 to 10. Each run started with a burn-in period of 50,000 steps followed by 300,000 Markov Chain Monte Carlo (MCMC) replicates. We used an admixture model, without prior information regarding population clustering.

We used the software Geneclass2 (Piry et al. 2004) following the methods of (Baudouin and Lebrun 2000) to assign individuals sampled in either recently recolonized sites or continuously inhabited ones by Atlantic salmon to a baseline constituted of samples from continuously inhabited sites (sites 1, 4, 9, 10 and 14). This was done to identify the source populations for the newly founded populations. Individuals that had scores <70 % were considered as potential hybrids or migrants from un-sampled populations.

Results

Data quality

Of the 1,009 sampled individuals, 960 individuals were successfully genotyped with a minimum of 66 % of individual amplification success. The total amplification success was 99.68 %. Using the SSAD486 marker to identify species, we found that 5 individuals were S. trutta, 26 individuals were hybrids between S. trutta and S. salar, and 5 could not be identified (amplification failed at this locus). We discarded these 36 individuals and conducted the subsequent analyses on the 924 remaining Atlantic salmon. Following the results from ML-relate, we removed a total of 187 (20 %) individuals that were found to have full-sibs in our dataset. These 187 removed individuals represented 0–47 % of the total number of individuals at each site (median value of 15 %). We thus conducted all the subsequent analyses on a total of 737 individuals (Table 1). Micro-Checker detected no null alleles. No linkage disequilibrium was detected by Genepop among loci (p > 0.05 for all them after Bonferroni corrections), thus, all loci were considered to be genetically independent. Only eight out of 192 FIS computed in Fstat were significant. No locus presented significant deviation from Hardy to Weinberg Equilibrium over all populations. At the population level, locations 6, 11 and 15 yielded significantly smaller observed than expected heterozygosities (Table 1).

Analysis of genetic diversity within samples

Loci had 6 (Ssa289) to 49 (Ssa157a) alleles over the entire dataset, with a median value of 24 and a total of 322 alleles over all loci. Average He (Expected heterozygosity) over all loci per population varied from 0.80 to 0.87 (Table 1), 0.84 on average in the Nivelle basin, 0.81 in the Nive basin and 0.84 in the Gaves basin. He was on average of 0.83 in continuously inhabited locations and of 0.84 in recently recolonized sites. Average Ar (Allelic richness) over all loci varied from 7.28 to 8.93 depending on the location. Ar was on average 8.62 in the Nivelle basin, 7.45 in the Nive basin and 8.27 in the Gaves basin. Ar was on average 8.09 in continuously inhabited locations and 8.23 in recently recolonized sites (Fig. 2). Ar was not significantly different between continuously inhabited locations and recently recolonized ones (Table 2, t test, t = −0.50, df = 7.26, p value = 0.63). Average PAr (Private allelic richness) varied from 0.15 to 0.41 depending on the location. PAr was on average 0.35 in the Nivelle basin, 0.18 in the Nive basin and 0.25 in the Gaves basin. PAr was on average 0.23 in continuously inhabited locations and 0.27 in recently recolonized sites. Effective size (Ne) varied from 33.8 to 3,583.8 depending on the location (Table 1). Overall, the median value of Ne was 238.4 in continuously inhabited locations and 124.0 in recently recolonized sites (Fig. 2). These two medians of Ne estimates were not significantly different (Table 2, Mann–Whitney test, W = 30, p value = 0.83).
Table 2

Table of FST among sites. (Color table online)

https://static-content.springer.com/image/art%3A10.1007%2Fs10592-014-0602-3/MediaObjects/10592_2014_602_Tab2_HTML.gif

Significant values are indicated in bold and non-significant ones in italic. A color gradient help to visualize the hierarchical differentiation among populations

https://static-content.springer.com/image/art%3A10.1007%2Fs10592-014-0602-3/MediaObjects/10592_2014_602_Fig2_HTML.gif
Fig. 2

Boxplots of allelic richness and effective population size in continuously inhabited and recently recolonized sites

Analysis of genetic structure among sites

FST among sites ranged from 0.000 to 0.054 (Table 2). Table 2 and Fig. 3 show small FST among continuously inhabited sites and recently recolonized ones within each basin (FSTA = 0.012 on average), but relatively high FST among sites located in different basins, either recently recolonized ones (FSTB 0.035 on average) or continuously inhabited sites (FSTC 0.042 on average). These three FST were significantly different (FSTB vs FSTC t = −2.54, df = 10.99, p value = 0.03; FSTA vs FSTB t test, t = −10.30, df = 27.75, p value = 5.6e −11; FSTA vs FSTC t test, t = −10.36, df = 12.87, p value = 1.3e−07).
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-014-0602-3/MediaObjects/10592_2014_602_Fig3_HTML.gif
Fig. 3

Boxplots of FST: A between continuously inhabited and recently recolonized sites in each basin; B between recently recolonized sites among each basins; C between continuously inhabited sites among each basins

Considering all the sites, AMOVAs revealed that 2.53 % of the genetic variance was found among basins and 1.08 % among populations within basins (Table 3). When considering only continuously inhabited sites, AMOVAs revealed that 3.75 % of the variance was found among basins and 0.47 % among populations within basins. When considering only recently recolonized sites, AMOVAs revealed that 2.15 % of the variance was found among basins and 1.22 % among populations within basins. These two last AMOVAs showed that genetic structure was higher among continuously inhabited sites than among recently recolonized ones, according to the 5–95 % confidence intervals.
Table 3

Analysis of molecular variance partitioning genetic structure among and within the 3 main drainages (Gaves, Nivelle, Nive)

Population considered

Source of variation

% of variation

Φ-Statistic mean (5–95 %)

All, n = 16

Among groups

2.534

0.031

(0.021–0.041)

Among populations within groups

1.076

0.011

(0.009–0.013)

Continuously inhabited, n = 5

Among groups

3.749

0.037

(0.027–0.049)

Among populations within groups

0.469

0.005

(0.001–0.010)

Recently recolonized, n = 11

Among groups

2.145

0.021

(0.017–0.027)

Among populations within groups

1.222

0.012

(0.011–0.014)

A neighbor-joining dendrogram of Nei’s genetic distance revealed three genetically distinct populations corresponding to the Nivelle, Nive and Gaves basins (Fig. 4). This dendrogram also reveals relatively little differentiation among sites within each basin.
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-014-0602-3/MediaObjects/10592_2014_602_Fig4_HTML.gif
Fig. 4

Neighbor joining tree based on Nei 1983 genetic distances. Bootstrap values are indicated upside the node. Recently recolonized sites are indicated with an asterisk

Bayesian clustering and assignment of individuals

While the best k value found by BAPS was k = 3 (Fig. 5), the first delta k pick found in STRUCTURE result corresponded to k = 4, followed by a smaller pick at k = 8. However, the existence of several clusters in Gaves for k = 4 and in Gaves and Nive for k = 8 did not clearly corresponded to any geographic grouping and individuals were highly admixed. We therefore proposed that the most realistic number of genetic clusters was three, corresponding to the three main basins.
https://static-content.springer.com/image/art%3A10.1007%2Fs10592-014-0602-3/MediaObjects/10592_2014_602_Fig5_HTML.gif
Fig. 5

Bayesian individual clustering obtained using the software BAPS for k = 3 the software STRUCTURE for k = 8 and k = 4. Vertical bars represent proportions of membership of each individual to each cluster represented, which are represented by different colors. Recently recolonized sites are indicated with an asterisk. (Color figure online)

According to the assignment conducted using Geneclass, an average of 86 % of the individuals sampled in continuously inhabited sites were assigned to their site of origin, 95, 100 and 78 % on average in the Nivelle, Nive and Gaves rivers, respectively (Table 4), suggesting a relatively high power to assign fish to the three different basins. In contrast, an average of 54 % of putatively local individuals was found in recently recolonized sites (48, 71 and 50 % in the Nivelle, Nive and Gaves rivers, respectively). Within the continuously inhabited sites, 0 % of the individuals that could not be assigned to a basin were putative migrants from other basins and 12 % of individuals were hybrids or migrants from other unsampled populations. Within the recently recolonized sites, the individuals non-assigned to the basin were 4 % putative migrants from other basins and 38 % hybrids or migrants from other unsampled populations.
Table 4

Assignation of Atlantic salmon among rivers and tributaries (sampling sites grouped by basins)

River

Sites

1 (%)

4 (%)

9, 10, 14 (%)

Genetically admixed individuals (%)

Nivelle

1

95

0

5

0

 

2*

35

5

13

48

 

3*

60

0

10

30

Nive

4

0

100

0

0

 

5*

0

65

15

20

 

6*

1

77

9

13

Gaves

7*

7

7

48

37

 

8*

4

3

42

51

 

9

0

0

85

15

 

10

0

0

73

28

 

11*

0

14

41

45

 

12*

0

0

45

55

 

13*

0

0

65

35

 

14

0

0

78

22

 

15*

0

0

60

40

 

16*

0

0

50

50

* Recently recolonized sites

Discussion

Fine scale genetic diversity within continuously inhabited sites

The significant genetic differentiation (FST) observed among continuously inhabited sites within the Adour catchment was comparable to the FST observed among other Atlantic salmon populations at the within river scale using markers with similar level of polymorphism (Ellis et al. 2011; Dionne et al. 2009; Vaha et al. 2007; Primmer et al. 2006). This significant differentiation among populations from these three rivers contrasted with the results of Perrier et al. (2011b), who only found one major genetic cluster within the Adour River. However, Perrier et al. (2011b) analyzed a much smaller number of individuals. The clear clustering of individuals in three genetically and geographically distinct groups suggests limited dispersal and gene flow among the Nive, Nivelle and Gaves rivers. This is in line with the relatively strict homing behaviour of Atlantic salmon (Stabell 1984) that has been suggested to be accurate to the tributary level (Vaha et al. 2008; Dillane et al. 2008). This existing fine scale genetic structure may also be linked to fine scale local adaptation (Vaha et al. 2008, 2007) and should be taken into account for management (Fraser and Bernatchez 2001). Hence, these results support the local management strategy that has been applied since the 1990’s, which considers the Nive, Nivelle and Gaves rivers as three distinct conservation units.

Effective population sizes estimated for the continuously inhabited sites in these three rivers were relatively high compared to other rivers located in Southern Europe (Perrier et al. 2013b; Nikolic et al. 2009). In particular, the populations located at the edge of this species’ range (like the study population) have dramatically declined (Boylan and Adams 2006; Parrish et al. 1998; Dumas and Prouzet 2003; Prouzet 1990). While the size of Atlantic salmon populations inhabiting the Adour basin has declined during the past decades, the effective population sizes we detected appear may be high enough for maintaining genetic diversity on the short term. However, these effective population sizes are relatively low compared to what is needed for the long-term conservation of wild populations (Frankham 2002, 2005; Traill et al. 2010). Nevertheless, the role of migration among populations within the Adour basin or even from distant rivers should not be neglected in sustaining long-term effective population sizes and genetic diversity (Gomez-Uchida et al. 2013; Palstra and Ruzzante 2011; Kuparinen et al. 2010). Moreover, high proportions of mature male parr may also increase effective population size in these southern populations (Saura et al. 2008; Garcia-Vazquez et al. 2000; Martinez et al. 2000; Moran et al. 1996).

The low admixture and low proportions of putative migrants or hybrids within continuously inhabited sites suggests a relatively low impact of stocking. In particular, the use of geographically and genetically distant populations to stock the river from the Adour catchment may have resulted in noticeable admixture (Perrier et al. 2013b; Finnengan and Stevens 2008; Hansen et al. 2009) and in a lack of differentiation among locations due to a local homogenisation of the genetic diversity (Marie et al. 2010; Eldridge and Naish 2007). Similarly, since stocked fish can have a higher dispersal than their wild counterparts (Pedersen et al. 2007; Quinn 1993), even stocking local but hatchery-reared individuals may have led to admixture among local clusters. However, admixture appeared low within continuously inhabited sites and these locations were significantly differentiated. This may be due to a relatively low return rate (Perrier et al. 2013a) and fitness of non-native salmon stocked before the 90’s (Milot et al. 2013; Perrier et al. 2013b; Thériault et al. 2011). Relatively low survival and/or low straying of local fish stocked since the 90’s using Nive, Nivelle and the Gaves basins as conservation units may also explain low admixture among these populations. Overall, while we cannot rule out the impacts of historical stocking on present neutral genetic structure among populations or on their adaptive potential, our results suggests that stocking did not significantly affect genetic structure within continuously inhabited sites from the Adour basin.

The recolonization of reconnected sites by Atlantic salmon

A relatively high production of fry was found in recently reconnected spawning grounds (Barracou 2008), illustrating that the construction of fishways on dams may aid in the effective recolonization of Atlantic salmon populations (i.e. recolonization followed by successful reproduction). Even though noticeable proportions of fish sampled in recently recolonized sites were assigned to distant sites or were identified as putative hybrids, the majority of the individuals were assigned to neighboring downstream sites, suggesting a higher colonization success of local fish. Accordingly, the genetic structure among samples from a single river was small, suggesting a relatively high contribution of fish having local origin, either wild or stocked, to the establishment of new populations. This was also supported by our finding that there is a similar genetic structure among basins for both continuously inhabited and recently recolonized sites. These results agree with (Perrier et al. 2010; Ikediashi et al. 2012) data, which found that large proportions of fish recolonizing depopulated rivers originated from nearby rivers. However, in contrast with the previous studies documenting recolonization processes in Atlantic salmon by inferring the origin of adults’ individuals recolonizing rivers (Griffiths et al. 2011; Perrier et al. 2010; Ikediashi et al. 2012), here we genotyped juveniles caught in recolonized parts. Thus, we characterized both a recolonization and a successful reestablishment of Atlantic salmon populations. Nevertheless, it did not allow us to compare reproductive success among adult fish returning to upstream areas. Hence, the fact that large proportions of fry caught in recently recolonized habitats were assigned to downstream sites of the same rivers (Nive, Nivelle, Gaves, respectively) does not necessarily indicate that stocked or wild adult fish had high homing rates but instead may suggest that adults with local genetic characteristics had high reproductive success relative to potential migrants. This illustrates that even though dispersal of both wild and stocked fish might have contributed to the reproductive effort within recently reconnected sites, the parents with the highest reproductive success, overall, originated from close downstream sites. This result agrees with studies showing that immigrants often come from nearby sites (Perrier et al. 2010; Ikediashi et al. 2012) and with those finding locally adapted fish may have higher reproductive success (Dionne et al. 2008; McGinnity et al. 2007; Hendry 2004). However, it was not possible to disentangle the relative contributions of stocking and colonization of wild fish since recent stocking has used local parents for each of the three rivers. Pedigree reconstruction would have allowed us to address this question (Milot et al. 2013; Araki et al. 2007; Thériault et al. 2011), but this was not possible due to the prohibitively extensive and costly genotyping effort needed to address this issue in such a large population. Given that local fish might be locally adapted, as widely suggested for Salmonids (Bourret et al. 2013; Primmer 2011; Garcia de Leaniz et al. 2007; Taylor 1991), the fact that local fish may have had a higher reproductive contribution than non-local fish has important implications for the long term recolonization of the Adour basin. In particular, local fish might be more prone to establish a viable population than exogenous individuals.

We expected a relatively low effective population sizes in recently recolonized locations compared to continuously inhabited sites, as a result of founder effects. However, effective population sizes in these two groups were relatively similar. Such comparable effective sizes among recently recolonized and continuously inhabited sites could be explained by relatively high contemporary gene flow (Waples and England 2011). Of particular interest, effective population size tended to be larger for recently recolonized habitats within the Nivelle River. This result suggests that the new population that recolonized reopened areas is even larger than the populations located downstream in continuously inhabited sites. This may occur because upstream habitats are, in general, more suitable for Atlantic salmon than downstream sites (Barracou 2008). In turn, this observation may not be linked to stocking operations since hatcheries generally use a reduced numbers of parents harbouring reduced diversity compared to the population of origin (Araki and Schmid 2010). Indeed, these results likely suggest a relatively low impact of stocked Atlantic salmon, which is in line with recent studies on the impact of stocking on the breeding system in this species (Milot et al. 2013; Jonsson and Jonsson 2006). In the case of the Nivelle River the effective population size appeared larger than the census size. Such a result is difficult to explain without invoking a potentially large contribution of precocious parr (Johnstone et al. 2013; Saura et al. 2008; Jones and Hutchings 2001) but could also be linked to gene flow from the Nive and Gave rivers. Indeed, census size has been estimated through an exhaustive monitoring of adult anadromous salmon but did not included precocious parr (Lange et al. 2011). Within the Nive and the Gaves rivers, the effective population sizes tended to be smaller than the census sizes, which is more in lines with the expectations of a typical Atlantic salmon population (Palstra and Ruzzante 2011) in which a large variance among breeders exists (Richard et al. 2013; Fleming 1996). Overall, the relatively large effective population sizes estimated for recently recolonized sites might allow for the conservation of a relatively high level of genetic diversity in these new populations, which may limit short-term extinction risks (Traill et al. 2010; Frankham 2005). Accordingly, within each of the three rivers, samples collected in continuously inhabited and in recently recolonized sites had comparable allelic richness, suggesting no or only a weak loss of genetic diversity during the recolonization process. This is a critical observation because effective population size and genetic diversity is positively linked to the effectiveness of selection relative to drift (Charlesworth 2009; Olson-Manning et al. 2012).

Conclusion

Overall, our results suggest that restoring accessibility to upstream areas can allow for the recolonization of Atlantic salmon. This recolonization mainly comes from individuals from proximate downstream sites, with neither a decrease of local diversity nor disruption of existing genetic structure. Along with previous studies (Perrier et al. 2010; Ikediashi et al. 2012; Kiffney et al. 2009; Griffiths et al. 2011; Schreiber and Diefenbach 2005), this study shows connectivity restoration as an effective way to support recolonization of rivers from which salmonids have been previously extirpated. Nevertheless, the ecological restoration policy should also aim to reconnect several other upstream tributaries and improve water quality. Indeed, the actual census size of the Atlantic salmon population in the Adour basin is much smaller than the estimated potential capacity (Barracou 2008). While the implementation of fishways in the Adour drainage allowed an effective recolonization of Atlantic salmon within several upstream areas, little is known about the impacts of these fish passages on the recolonization dynamic of other migratory species. While Atlantic salmon can cross over relatively challenging fishways, several other migratory fish having lower swimming capacities may need more specific and less challenging fishways to effectively recolonize depopulated areas. More broadly, even though the recolonization of some fish migratory species can be enabled by fishways, such devices may not compensate the overall ecosystem-wide dramatic impacts of dams (Brown et al. 2013).

Acknowledgments

We acknowledge all participants to the collection of samples and of various historical and environmental data, with a special attention to A. Manicki, J. Chat, D. Barracou. We also thank P. Regnacq, JB Torterotot and Anne Dalziel for their help while analyzing data and writing the paper. We thank two anonymous reviewers and the associate editor C. Primmer for their very constructive comments. Authors also thank all French organizations that provided their technical assistance for electric fishing: the National Institute for Agricultural Research (INRA), the National Office of Water and Aquatic Media (ONEMA) and Migradour. This work was funded by the European Union INTERREG IIIB program [Atlantic Salmon Arc Project (ASAP)] and the European Union INTERREG IVB program [Atlantic Arc Resource Conservation (AARC)].

Copyright information

© Springer Science+Business Media Dordrecht 2014