Loss of genetic integrity and biological invasions result from stocking and introductions of Barbus barbus: insights from rivers in England

Barbus barbus in England and Wales

Barbus barbus is the only Barbus species present in England and Wales and thus translocated fish pose no risk of hybridisation with endemic Barbus species as has occurred elsewhere in Europe (Meraner et al. 2013; Zaccara et al. 2014). The indigenous range of B. barbus in England covers the Yorkshire Ouse, Trent, and Thames river basins (Table 1; Wheeler and Jordan 1990). Although there a number of other eastern-flowing rivers that could theoretically have also held natural stocks, there is some doubt over whether this would be the case, as these rivers are generally small and habitat limiting, such as the Wensum, Yare, and Suffolk Stour (Table 1; Wheeler and Jordan 1990). The redistribution of B. barbus in England commenced in the 1890s, when the Hampshire Avon in Southern England had fish introduced from the Thames catchment, with subsequent releases into this river using fish from the Rivers Kennet and Lea. The River Severn had B. barbus introduced in 1956 with a release of 509 adult fish from the River Kennet and remains the only known release of fish in either the Severn or its tributary, the River Teme (Table 1; Wheeler and Jordan 1990). This introduction was very successful, and the Severn and Teme have been important recreational fisheries for B. barbus since the 1970s. The Warwickshire Avon, also in the Severn catchment, received a stocking of fish in the 1960s from the River Swale (Table 1). This river is highly regulated and it appears unlikely that the river could have been colonized from the Severn due to impassable blockages. Although the focus here was on B. barbus in England, the species is also present in Scotland (River Clyde; W. Yeomans pers. comm.) and Wales (River Taff), but samples were not available to the study. The species remains absent from Ireland (Wheeler and Jordan 1990).

The use of hatchery-reared B. barbus to supplement populations or extend their range in England became more prevalent from the 1980s and continues to present. Fish are usually reared up to the age of 1+ or 2+ years before their release into the wild at lengths of 120–250 mm. Although primarily involving broodstock from the River Trent and completed by Government agencies, some stocking has also been completed using fish cultured in other sites and completed legally by individuals and angling associations, usually using fish of Kennet broodstock (C. Seagrave personal communication). Although these releases were regulated, there is less detail on the numbers of fish released (Table 1). Finally, there is anecdotal evidence suggesting some unregulated movements of B. barbus might have occurred between river basins via anglers. There is, however, no documented evidence of this.

Sampling

In this study, scale samples were available for genetic analyses from 18 rivers in nine river catchments (basins), of which seven catchments were in the indigenous range (but included rivers where it was uncertain if B. barbus were found naturally) and two were in the nonindigenous range (Table 1). The scales had been collected from fish sampled either during fish population surveys completed by the Environment Agency (a Government agency in England) between 2001 and 2014, or from fish captured by anglers, with scales removed by a competent person. An exception was scales for the River Great Ouse, where scales were also available from 1994. In all cases, the scales were removed for the purposes of age and growth analysis to support fishery management programmes, rather than specifically for this study. In addition, samples were also available from the River Trent hatchery (10 fish per year from 1997, 2002, 2003 & 2004 and 20 fish from 2000).

Molecular data

Total genomic DNA was extracted from scales using DNA extraction kit (DNeasy Blood & Tissue Kit and QIAamp DNA Mini Kit, Qiagen), under manufacturer instructions. Phylogenetic and population genetic analyses were performed on (1) mitochondrial DNA Control Region (CR) gene and (2) two nuclear genes (S7 ribosomal protein [S7] and growth hormone [Gh]) (Table 2). These nuclear were selected as suitable markers as they have been extensively used in population studies of barbel species (e.g., Gante et al. 2011, 2015; Meraner et al. 2013; Zaccara et al. 2014; Buonerba et al. 2015). The mtDNA CR was amplified in 268 individuals using primer pair dloop-sxF and dloop-dxR (Rossi et al. 2013). As B. barbus is tetraploid, the S7 ribosomal protein (S7-1 and S7-2) and growth hormone (Gh-2) genes were amplified using paralog-specific primers (Gante et al. 2011), methods recently applied in their population genetic analysis elsewhere (Zaccara et al. 2014; Buonerba et al. 2015) (Table 2). Nuclear loci S7-1 and S7-2 were successfully amplified with 3–8 and 10–13 primer pairs, while Gh-2 with 24–30 primer pairs (see Gante et al. 2011). Gh-1 was not used in this analysis, as we were not able to obtain sequences for a high proportion of the samples. Polymerase chain reaction (PCR) amplifications were performed with Multiplex PCR kit (Qiagen) in 10 μl reaction volume containing approximately 10 ng of template DNA and 0.2 μM of each primer pair. Thermal cycling was performed as follows: denaturation of 15 min at 95°C, followed by 30 cycles of 94°C for 30 sec, 90 sec at 56°C of annealing temperature and the extension step at 72°C for 90 sec, the final elongation was at 72°C for 10 min. The appropriate annealing temperatures were as follows: 58°C for S7-1, 62°C for S7-2, and 56°C for Gh-2 and mtDNA CR. PCR products were purified using kit Illustra^™ Exostar (GE Healthcare) and sequenced in both directions with amplification primers on an ABI 3130xl Genetic Analyzer using Big Dye 3.1 terminator (Applied Biosystem). The nucleotide sequences of nuclear alleles and CR haplotypes were deposited in GenBank database under accession numbers (KT766197-KT766290; KT766373-KT766378) (Table S1).

DNA polymorphism

Mitochondrial and nuclear sequences were both manually aligned using BioEdit ver. 5.0.9 (Hall 1999) to eliminate ambiguities and to check polymorphic sites. For the nuclear S7 paralog 2, specimens heterozygous for insertions or deletions (indels) were manually phased using the complementary information carried by the forward and the reverse sequences (Flot et al. 2006). Then, nuclear heterozygous alleles for single nucleotide polymorphism (SNPs) were phased using DnaSP v. 5 (Librado and Rozas 2009); following, the number of haplotypes was calculated using Non Redundant Data Base (Gish 2004). DNA polymorphism indices for each locus, like the number of polymorphic sites (S), the haplotype diversity (Hd), and the percentage nucleotide diversity (π %), were calculated using DnaSP v. 5 (Librado and Rozas 2009).

Phylogenetic analyses

In phylogenetic analyses of mtDNA CR, four sequences from previously sampled individuals of B. barbus and B. plebejus from Northern Italy (Zaccara et al. 2014) and one available B. barbus sequence (GenBank acc. No. AB238965) were added to the dataset, that was rooted with Barbus meridionalis (GenBank acc. No. AJ388417).

Phylogenetic analyses were performed on mtDNA CR using three different optimality criteria: maximum likelihood (ML), neighboring-joining (NJ) and Bayesian analysis. The ML analysis was performed through GARLI v1.0 software (Zwickl 2006) using Trn+I model of sequence evolution (Tamura and Nei 1993), as estimated with ModelTest 3.7 software (Posada and Crandall 1998). Neighboring-joining was performed trough PAUP 4.0b10 program (Swofford 2002). Bayesian analysis was performed using MrBayes v.3.1.2 software (Ronquist and Huelsenbeck 2003), with a Markov chain Monte Carlo algorithm (MCMC): four simultaneous and independent Markov chains from random trees were started and run for 1,000,000 generations, with the first 25,000 generations (2,500 trees) discarded as the burn-in (P < 0.01).

Population genetic analyses

A preliminary description of genetic variability at the catchment scale was provided by the mtDNA CR haplotype distribution. Following, each unique nuclear allele was numerically coded and used to genotype each specimen. The genetic variability within populations was quantified (i.e., expected (H_E), observed (H_O) heterozygosity, and all loci were tested for deviation from Hardy–Weinberg equilibrium) using GenePop (Raymond and Rousset 1995; Rousset 2008). In addition, populations were tested for inbreeding by calculating F_IS using GenePop. Computation of pairwise multilocus F_ST values (Weir and Cockerham 1984) among populations was performed using Arlequin ver. 3.11 (Excoffier et al. 2005) with 1000 permutation procedure.

Covariation among nuclear loci was assessed using the Bayesian clustering method implemented in the software STRUCTURE v.2.3.4 (Pritchard and Wen 2002) in order to detect the presence of distinct genetic clusters, assign individuals to populations, and to identify migrants and admixed individuals. Each STRUCTURE run consisted of 100,000 MCMC generations as burn-in, followed by 500,000 MCMC replicates to estimate the posterior sample distribution, using the admixture, correlated allele frequency models. To assess reliability, 20 iterations were run for each K cluster. The number of groups (K) identified by STRUCTURE was estimated by a combination of changes in log-likelihood of consecutive K-values and ΔK (Evanno et al. 2005) using the program STRUCTURE HARVESTER (Earl and von Holdt 2012). Results were summarized using CLUMPP (Jakobsson and Rosenberg 2007) and displayed using Distruct (Rosenberg 2004). Finally, in order to better highlight clusters between the populations, a phenetic tree, based on the F_ST matrix, was built using PAUP 4.0b10 program (Swofford 2002).

mtDNA sequence variation at basin scale

A total of six different haplotypes were identified in the 941 bp of CR region in the 268 individuals analyzed (Table 2). In the sequence alignment, 6 variable sites were recorded that were all parsimony informative. ML, NJ, and BI phylogenetic analysis of the mitochondrial sequences confirmed that all six UK haplotypes clustered with B. barbus (Fig. 1), characterized by weak genetic distance (P-distance = 0.32%), while the interspecies, B. plebejus, distance was over 4%. The geographic distribution of the haplotypes revealed a homogenous pattern (Fig. 2). Haplotype Hap_2 was the most abundant, found in 199 individuals and being widespread in all sampled basins and the hatchery. A similar pattern was recorded in haplotype Hap_3 that differed from Hap_2 at 2 nucleotides and was found in all basins (Fig. 2). Haplotype Hap_4 was present only in the Yorkshire Ouse catchment and the hatchery, while Hap_5 was specific to the Yorkshire Ouse catchment only (Fig. 2).

Nuclear DNA genetic variability

Sequence analysis of three nuclear loci yielded 1998 sequences, with a whole 2058-bp-long alignment (S7-1: 467 bp; S7-2: 562 bp and Gh-2: 1029 bp). A total of 1813 SNPs were found, mainly concentrated on the Gh-2 locus, while only one indel was assumed in the S7-2 locus. Levels of sequence polymorphism for each marker are summarized in Table S1. Among the 1998 nuclear sequences, 94 haplotypes were scored prevalently expressed by Gh-2 locus, characterized both by the higher haplotype (H) and by nucleotide (π %) diversity (see Table S1).

Genotyping and population genetic structure analyses

Following genotyping, populations were characterized by H_E and H_O values averaged over three loci, ranging from 0.64 to 1.0, respectively; F_IS values ranged from −0.56 (River Kennet) to −0.14 (River Medway) and were all significant except for two populations (Table 3). These negative values represent an excess of heterozygotes that could be interpreted as lack of inbreeding or genetic drift. Only four populations (Rivers Nidd, Trent, Severn, and Hampshire Avon) had loci that were at Hardy–Weinberg equilibrium.

Among the 171 F_ST values of pairwise population comparisons, 70 values were significant (P < 0.05). Pairwise F_ST values that were not significantly different (P > 0.05) were among nonindigenous (NI) populations (Rivers Teme, Severn, Warwickshire Avon and Hampshire Avon), while all values were significant among populations in the indigenous range but where there is conjecture over whether their rivers have natural populations (i.e., Rivers Witham, Wensum and Medway (I*); Table 1). Between the 55 comparisons in the indigenous (I) populations, 21 were significant, mainly in the Yorkshire Ouse catchment (Table 1). Among the 18 comparisons between populations and the hatchery data, 9 were significant, shared out equally among the three population groupings (Table 4).

To infer population structure on a finer scale, a STRUCTURE analysis, completed on the entire nDNA dataset, was performed from K = 1 to K = 20. Using the Evanno et al. (2005) method, two clusters (K = 2) were identified and the two statistics used to infer the number of clusters, LnP(D) (−2492.79) and ΔK, were consistent for K = 2. The populations were then grouped using their K assignment, with a threshold of K > 0.7 (Fig. 3). Two river populations (Great Ouse and Swale) and the hatchery population did not reach the K > 0.7 threshold. A concordant output was found in the neighbor-joining phenetic tree, where populations with the same K assignment resulted in the same cluster, with the only exception of four populations (River Witham, Warwickshire Avon, Dove, and Trent) (Fig. 4).

Indigenous range

The indigenous range of B. barbus in the UK only covers some eastern-flowing rivers in England, a relic of the last glacial period when this landmass was still joined to the European mainland, enabling their colonization by a number of fish species in the late-glacial and postglacial period from western-flowing rivers such as the Rhine, Meuse, and Elbe (Wheeler and Jordan 1990). In the B. barbus indigenous range in England, two principal groups were identified indicating that the primary separation between these were at river catchment levels: Yorkshire Ouse and Thames, thus representing a north: south pattern of grouping across this range. Indeed, populations in the rivers of the Yorkshire Ouse catchment generally showed relatively high levels of genetic differentiation with populations elsewhere and had at least one specific (unique) haplotype.

The analyses of genetic distance between the populations suggested that policies to enhance populations that involved releasing hatchery-reared fish have resulted in some genetic-level homogenization of B. barbus populations from different river catchments. For example, across the River Thames catchment, there is significant genetic structuring whereby the Kennet and Lea rivers (both tributary rivers of the River Thames) cluster together but are significantly different from the main River Thames population. Unexpectedly, the main River Thames population is genetically similar to the Yorkshire Ouse and River Trent, possibly due to the introduction of hatchery-reared fish, with stocking records for the river showing releases in recent years of B. barbus from a hatchery that uses River Trent broodstock (Environment Agency, unpublished data). Similarly, the fish from the River Ure (in the Yorkshire Ouse catchment) grouped with the River Kennet and Lee, suggesting that fish were introduced there that originated from these rivers. Losses of population genetic integrity following stocking events have also been detected in populations of salmonid fishes in England when hatchery-reared fish from other catchments have been released (e.g., Sušnik et al. 2004; Eldridge and Naish 2007). For example, in a study involving 27 UK populations (including England) of grayling Thymallus thymallus, while there was considerable population specific genetic diversity evident, it was also revealed that this had been eroded by long-term stocking of hatchery-reared fish using broodstock from other catchments (Dawnay et al. 2011).

There are a number of smaller river catchments in this indigenous range in which there is some uncertainty as to whether B. barbus populations would have been able to survive there naturally due to, for example, catchment sizes being small and rivers of slow flow that would limit habitat availability and result in small populations vulnerable to environmental changes (Wheeler and Jordan 1990). Such rivers include the Wensum and Medway. In the last 25 years, these small rivers have received stockings of considerable numbers of hatchery-reared B. barbus, particularly using River Trent broodstock, in order to enhance their populations (Environment Agency, unpublished data). The outputs here suggested their populations comprised fish of similar genetic origin. While this could be through the rivers having ancestral populations of similar genetic composition, Wheeler and Jordan (1990) suggested that it was doubtful that these rivers could have historically supported sustainable natural B. barbus populations due to their small sizes and variable flow regimes. These rivers have also been subjected to considerable habitat disruptions in the last 50 years through, for example, flood defense works that can substantially affect habitat connectivity and recruitment patterns of cyprinid fishes (Peirson et al. 2008). Consequently, it was considered probable that the similar genetic patterns detected here across these rivers were due more to the hatchery-rearing stocking activities of recent years rather than their original genetic origins. This then suggests that these stocking activities have been relatively successful with, as a minimum, these resulting in the persistent presence of stocked B. barbus in these rivers that enabled their capture and analysis here. This is contrary to many studies on stocking hatchery-reared fishes of the Cyprinidae family that tend to suggest either their poor survival (e.g., Aprahamian et al. 2004) or relatively low proportions in subsequent samples (Britton 2010). Nevertheless, it does suggest that if there were original, genetically differentiated B. barbus populations in these catchments, these have now been lost due to the introgression of the hatchery-reared fish.

Nonindigenous range

The translocation of B. barbus from their indigenous range to their nonindigenous range has been successful, with invasive populations evident in a number of western-flowing rivers (Wheeler and Jordan 1990; Britton and Gozlan 2013). Commencing over 100 years ago, it originally involved the direct movement of adult fish between catchments, whereas today it is reliant on releasing hatchery-reared individuals. The genetic data presented here corroborated historical stocking records (Wheeler and Jordan 1990). They revealed high genetic similarity between the indigenous fish of the River Kennet and the nonindigenous fish of the River Severn and Hampshire Avon, where written records suggest the Kennet was the original source of the introduced B. barbus (Wheeler and Jordan 1990). For example, the only recorded B. barbus translocation into the River Severn was from the Kennet when 509 adult fish were released in 1956. Moreover, these fish have since proved highly invasive by colonizing much of the river and its major tributary, the River Teme, as well as providing a source of fish for translocations into other nonindigenous catchments (Wheeler and Jordan 1990; Britton and Gozlan 2013).

This successful invasion of the River Severn is similar to the invasion success of B. barbus observed in other European rivers following introductions. An example is in the River Po basin, Northern Italy, where apparent barriers to migration were unable to prevent the dispersal of invasive B. barbus throughout the basin. This was related to intentional releases of fish for angling purposes, resulting in relatively high propagule pressure (Meraner et al. 2013; Zaccara et al. 2014). By contrast, the invasion of the River Severn basin arose from a single founding event, and the invasion of the Hampshire Avon was initiated by an unknown number of fish released in 1896 from the Thames catchment and then 124 fish in the 1960s from the same catchment (Wheeler and Jordan 1990). Thus, these data suggest that in the nonindigenous range in the UK, the release of relatively small numbers of adult fish from either an individual river or a single catchment was sufficient to initiate very successful invasions, with colonization at the catchment level then achieved through natural dispersal and recruitment (Wheeler and Jordan 1990; Britton and Pegg 2011).

In the River Po basin, the B. barbus invasion has resulted in population declines of the endemic Barbus plebejus, particularly via introgressive hybridization (Meraner et al. 2013; Zaccara et al. 2014). In the invasive UK range of B. barbus, there are no other species of the Barbus genus present and so their genetic impacts have been negligible. Potential impacts thus relate only to ecological concerns, although these have received little attention to date (Britton and Pegg 2011).

Evolutionary applications of B. barbus genetic data to their population management

Phylogenetic clarification of species assemblages has important implications for conserving and managing populations (Moritz 1994), as unless there is knowledge on what constitutes a species, population or subpopulation, management resources are difficult to assign and strategies difficult to design and implement (Dawnay et al. 2011). Thus, the genetic outputs for B. barbus in the UK can be applied to the management of their populations across both their indigenous and their nonindigenous ranges by ensuring management practices that involve fish stocking events always consider their potential for causing detrimental evolutionary consequences in receiving populations.

The mtDNA data revealed B. barbus in the UK were within the same clade as populations in mainland Europe that originate in the River Danube catchment, the Western European B. barbus glacial refuge (Kotlik and Berrebi 2001). Consequently, they do not constitute a historically isolated unit that could be considered as a single evolutionary significant unit (ESU). However, as inland fish populations tend to be managed within countries, irrespective of their wider ESU status (Dawnay et al. 2011), their populations can then be managed at national levels, with this management informed by their genetic status. For example, as the B. barbus of the Yorkshire Ouse catchment (except the River Ure) had at least one specific haplotype and were significantly genetically differentiated from other catchments, it can be argued that they require protection from this being further disrupted by more stockings that originate from outside the catchment. For all other river catchments now containing B. barbus in the UK, the genetic data can be applied to either ensuring populations are managed to facilitate the maintenance of their existing levels of genetic differentiation (e.g., the Rivers Kennet, Lea, Severn and Hampshire Avon) or engage in active management to re-establish populations to their prestocking genetic variability (e.g., River Thames).

Given the lack of overall conservation concern for B. barbus across their range, with their IUCN Red Listing being of least concern (IUCN 2015, www.iucnredlist.org), then it can be argued that it remains appropriate that UK populations are managed primarily for recreational, catch-and-release angling, especially given the considerable socioeconomic benefits their fisheries can generate (Britton and Pegg 2011). Within this management, measures are thus recommended to either restore or maintain their original genetic diversity, as this should help ensure populations can maintain their ability to adapt to changing environmental conditions, while emphasizing that management approaches should focus more on habitat improvement, such as through improving the longitudinal connectivity of rivers to enable greater access to spawning grounds (e.g., Lucas and Batley 1996; Lucas and Frear 1997) and improving nursery areas (Gordon and Bennetts 1996). Stocking then becomes a last resort to enhance or maintain a population, and given the population structuring detected here, it should focus on only using fish from that river or catchment, especially in the Yorkshire Ouse catchment. In England, this is entirely consistent with current policy and practice for the salmonid fishes brown trout Salmo trutta and T. thymallus (Environment Agency 2003) and thus also appears highly appropriate for B. barbus.