C.D. Todd, A.M. Walker, M.G. Ritchie, J.A. Graves, and A.F. Walker

Transcription

1 1176 Population genetic differentiation of sea lice (Lepeophtheirus salmonis) parasitic on Atlantic and Pacific salmonids: analyses of microsatellite DNA variation among wild and farmed hosts C.D. Todd, A.M. Walker, M.G. Ritchie, J.A. Graves, and A.F. Walker Abstract: The copepod Lepeophtheirus salmonis is ectoparasitic on Atlantic and Pacific wild salmonids. It is a major pest to Atlantic salmon (Salmo salar) aquaculture and may be implicated in recent declines of certain European wild salmonid stocks. Variation at six microsatellite loci was assessed among L. salmonis from wild and farmed salmonids in Scotland, wild sea-run brown trout (Salmo trutta) in Norway, and farmed Atlantic salmon in eastern Canada. An outgroup North Pacific sample was obtained from farmed Atlantic salmon in British Columbia. No significant differentiation was found between L. salmonis from the host species or among samples from throughout the North Atlantic. This is consistent with long-distance oceanic migration of wild hosts and larval interchange between farmed and wild host stocks being sufficient to prevent genetic divergence of L. salmonis throughout the North Atlantic. These results have important management implications for both wild stock conservation and aquaculture in that genetically, L. salmonis in the North Atlantic comprises a single population: there is no evidence of isolation of populations on farmed hosts from those on wild fish. Comparison between North Pacific and North Atlantic L. salmonis populations showed significant but low differentiation (F ST = 0.06). Résumé : Le copépode Lepeophtheirus salmonis est un ectoparasite des salmonidés sauvages de l Atlantique et du Pacifique. C est un ravageur important dans l aquaculture du saumon de l Atlantique (Salmo salar) et il est peut-être impliqué dans le déclin récent de quelques stocks sauvages européens. Nous avons évalué la variation à six locus microsatellites des L. salmonis chez des saumons en nature et en pisciculture en Écosse, des truites de mer sauvages (Salmo trutta) en Norvège et des saumons de l Atlantique de pisciculture dans l est du Canada. Un échantillon du Pacifique nord obtenu chez les saumons de pisciculture de la Colombie-Britannique a servi de groupe externe. Il n existe aucune différence significative entre les L. salmonis, ni en fonction des hôtes, ni en fonction des échantillons, dans tout l Atlantique nord. Cela s explique par la migration océanique sur de grandes distances des hôtes sauvages; de plus, les échanges de larves entre les hôtes de stocks sauvages et cultivés suffisent à empêcher la divergence génétique chez L. salmonis dans tout l Atlantique nord. Ces résultats ont d importantes conséquences sur la gestion des stocks sauvages et des stocks de pisciculture, parce que les L. salmonis de l Atlantique nord forment une seule population génétique; il n y a aucune indication que les populations qui parasitent les hôtes de pisciculture soient isolées de celles qui vivent sur les poissons sauvages. Une comparaison des populations du Pacifique nord et de l Atlantique nord montre une différentiation faible, mais significative (F ST = 0,06). [Traduit par la Rédaction] Todd et al Introduction Caligid copepods are major ectoparasites of marine teleosts worldwide and can be important pathogens of cultured fish stocks (Boxshall and Defaye 1993; Scholz 1999). Lepeophtheirus salmonis is a specialist on salmonids in both the Atlantic and Pacific oceans (Nagasawa et al. 1993), and throughout its marine range, this Northern Hemisphere caligid is a major pest in aquaculture. In the North Atlantic, L. salmonis parasitizes cultured Atlantic salmon (Salmo salar), sea-run brown trout, or sea trout (Salmo trutta), rainbow trout (Oncorhynchus mykiss), and Arctic char (Salvelinus alpinus): the primary wild host species are Atlantic salmon and sea trout (Tingley et al. 1997; Todd et al. 2000), with Scandinavian Arctic char of comparatively minor numerical importance (Bjørn and Finstad 2002). In the Received 31 July Accepted 6 February Published on the NRC Research Press Web site at on 9 September J17677 C.D. Todd, 1 A.M. Walker, 2 M.G. Ritchie, and J.A. Graves. School of Biology, University of St Andrews, St Andrews, Fife, KY16 8LB, UK. A.F. Walker. Freshwater Fisheries Laboratory, Faskally, Pitlochry, Perthshire, PH16 5LB, UK. 1 Corresponding author ( cdt@st-and.ac.uk). 2 Present address: CEFAS, Pakefield Road, Lowestoft, Suffolk, NR33 0HT, UK. Can. J. Fish. Aquat. Sci. 61: (2004) doi: /F04-069

2 Todd et al North Pacific, L. salmonis infests cultured S. salar and all species of wild Pacific salmon (Oncorhynchus spp.), albeit with varying intensity (Nagasawa 2001). Sea lice (L. salmonis and Caligus elongatus) are by far the most economically limiting parasites in the European aquaculture industry and can debilitate or kill the host (farmed or wild) fish, either through disruption of host osmoregulation as a direct result of feeding damage to the skin and fins or from secondary pathological infection of wound sites (Pike and Wadsworth 1999). The annual economic cost (labour, chemical agents, reduced growth, and mortalities) of sea lice to the Scottish industry alone presently is ~ 25 million (Rae 2002). Salmon aquaculture has developed rapidly in Norway, Scotland, and Ireland since the 1970s, with present European annual production > t. Atlantic salmon are also cultured on both seaboards of Canada and in the Southern Hemisphere in Chile. During the late 1980s and early 1990s, the first reports of collapses in certain wild sea trout stocks, and possibly associated sea lice epizootics, arose in Ireland (Tully et al. 1993b), Scotland (Northcott and Walker 1996), and Norway (Birkeland 1996). Unlike Atlantic salmon, sea trout postsmolts generally spend all of their marine residence in shallow, coastal waters (Klemetsen et al. 2003) and will be especially vulnerable to any coastally occurring stressors (including pollutants, predators, and parasites), even if these are temporally or spatially heterogeneous. Throughout the United Kingdom, there is considerable geographic variation and plasticity in the duration, extent, and destination of the marine migration patterns of sea trout (Picken and Shearer 1990; Klemetsen et al. 2003), and declines or crashes in wild sea trout stocks almost certainly are attributable to the interaction of several independent ecological and climatic factors (McVicar 1997; Todd et al. 1997; Hawkins 2000). Within areas supporting an extensive salmon farming industry, some observers (e.g., Tully et al. 1999; Butler 2002) have explicitly correlated these declines with the industry s expansion, but direct and unequivocal evidence of a causal link is still lacking (e.g., Isdal et al. 1997; Bjørn et al. 2001; Tully and Nolan 2002). In marked contrast with sea trout, Atlantic salmon smolts are generally believed to quickly migrate through coastal waters and out into the open ocean (Holst et al. 2000) and possibly are only transiently exposed to nearshore sources of stress or mortality. However, the recent observations of Atlantic salmon smolts, captured by targeted trawling, having acquired potentially lethal infestations of L. salmonis within a few days of first migration to Norwegian fjords, and before gaining the open ocean (Finstad et al. 2000), will be of particular concern in the management of wild Atlantic salmon stocks if such rapid infections are unnatural. In terms of the effective management and conservation of wild salmonids, and for health maximization of cultured fish stocks, it is therefore essential to have a clear understanding of the levels of gene flow, connectedness of populations (or metapopulations) of the parasite, sources of larval infestations, and the relative fluxes of L. salmonis between farmed and wild salmonid hosts. Wild Atlantic salmon can migrate to geographically disjunct regions of the open ocean for the duration of their marine residence (Hansen and Jacobsen 1999). There appears to be a broad distinction between the oceanic destinations of European S. salar, which begin to mature and return to fresh water after only one winter at sea (one-sea-winter (1SW)), and those that spend two to five winters at sea (multi-sea-winter (MSW)). As far as can be judged, 1SW Scottish salmon tend to migrate primarily to waters near the Faroe Islands and to the southern Norwegian Sea, whereas MSW fish generally exploit coastal feeding areas off Greenland (Reddin and Friedland 1999), although some may remain to the north of Faroe. New infections of L. salmonis continue to be acquired, even in the open ocean (Jacobsen and Gaard 1997), with the result that 2SW fish carry significantly more L. salmonis on their migratory return than do 1SW fish (Todd et al. 2000). Given the large-scale geographic separation, the duration of the host s marine residence in relation to the longevity of the parasite (Nordhagen et al. 2000), and persistent open ocean reinfestation, L. salmonis populations infesting wild 1SW and 2SW Atlantic salmon may be genetically distinct. The quantification of any such differentiation would then permit an estimate of the levels of gene flow between those stock components. Adult female L. salmonis retain paired eggstrings and release free-swimming nauplius larvae. After two naupliar moults, host colonization is achieved by the infectively competent planktonic copepodid larva; the obligatory precompetent pelagic larval period ranges from 2 to 9 days (at 15 or 5 C, respectively; Johnson and Albright 1991). Salmon on their North Atlantic feeding grounds appear to primarily occupy waters ranging from 4 to 8 C (Reddin and Friedland 1993), where larval development of L. salmonis will be relatively protracted. In coastal waters, perhaps towards the higher development temperatures, export of precompetent nauplii from aquaculture cages located even in poorly flushed, semienclosed sea lochs or fjords is inevitable. Accordingly, populations of sea lice infesting captive salmonids will be demographically open, although there is a perception within the aquaculture industry that farm cage sites are selfreinfesting (e.g., Jaworski and Holm 1992; Costello 1993; also see Heuch et al. 2003). Self-reinfestation of demographically open farm populations is possible, but only as a result of reimportation of competent copepodids that had developed from the released nauplii. It is generally impractical to track individual invertebrate larvae in the marine environment (e.g., Todd 1998) and thereby to directly observe both dispersal and to measure gene flow among populations (Slatkin 1993; Bossart and Prowell 1998). Any indirect attempt to quantify the infestation interaction between farmed and wild salmonids in terms of larval source(s) and abundance(s) requires an assessment of the extent of demographic isolation for populations of sea lice on those host sources. Genetic markers would appear to offer the most promising opportunity to assess population substructuring and the levels of migration, if any, between sea lice infesting farmed and wild hosts. Estimates of gene flow can be derived from F ST assuming an n-island model (Wright 1969). Such indirect measures of gene flow do, however, theoretically require that exchange occurs between all populations and that they are in demographic and genetic equilibrium; this has to be presumed but is impossible to assess in the natural environment (Slatkin 1994) and probably seldom is actually met (Neigel 1997; Whitlock and McCauley 1999). Polymorphic microsatellite loci are usually selectively neutral markers that will trace genetic differentiation, pri-

3 1178 Can. J. Fish. Aquat. Sci. Vol. 61, 2004 Table 1. Lepeophtheirus salmonis. Details of the six polymorphic microsatellite loci. Locus Repeat structure Allele size range (bp) T a ( C) Primer sequences (5 3 ) LsalSTA1 (AY509254) TC F: CGT CGA AAT TCT CAT CCA A R: GGG AAA GAT TGG GAG TGA G LsalSTA2 (AY509255) TC F: TCG TGG TGG TTG ACT CTA CT R: AGG AAA TCA GGA GCA AGT G LsalSTA3 (AY509256) TC F: TTA TCC GAA TCC GTC TTA TG R: AGC CTG AAG TAG GTT AGT TGG LsalSTA4 (AY509257) (GA) 2 -A-(GA) 2 -GT-(GA) F: AAG GCG TGC GTT GTT AAG T R: CAA TGC GAT CCT GGA GTC T LsalSTA5 (AY509258) GA F: GGG ATA AGT GGC GAG CTA CC R: GTC TCA GCG GCA GAA GTC TC LSNUIG14 TA F: GTT CAC GGT CGG GCT ATC TA R: TTT GAG TTA ATT GGT AAG AAA AAT TGA Note: New primers were designed for LSNUIG14 (= Ls.NUIG.14b of Nolan et al. (2000); GenBank AJ249886). Repeat structure, allele size range across 1007 individuals, and optimal PCR annealing temperature (T a ) are provided along with forward (F) and reverse (R) primer sequences. The clone sequences have been registered with GenBank under the accession numbers in parentheses. marily by genetic drift (e.g., Goldstein and Schlötterer 1999). But the interpretation of genetic data for North Atlantic L. salmonis must take close account of the complex, and contrasting, life histories of the two major wild host species. Here, we present population genetic analyses comparing L. salmonis variation at six microsatellite DNA loci, five of which are novel and were developed for use in this study. Comparisons were drawn among samples from (i) wild 1SW and 2SW adult S. salar on their return to Scottish coastal waters, (ii) both wild and farmed salmon, farmed rainbow trout, and wild sea trout from Scottish coastal waters, and (iii) a wider geographic range of Atlantic samples, including Norway and Canada, with a single Canadian Pacific Ocean sample as an outgroup. The null hypotheses were that (i) there would be no significant genetic differentiation of populations of L. salmonis parasitizing wild 1SW and 2SW adult salmon, (ii) there would be no differentiation between host species, or between farmed and wild hosts, in the North Atlantic, and (iii) there would be no significant differentiation between Atlantic and Pacific L. salmonis. The respective alternatives were that (i) 1SW and 2SW salmon might bear populations of L. salmonis that differ genetically because of their distinct geographic migration destinations (Norwegian Sea versus Greenland), (ii) farmed stocks might be self-reinfesting and their parasite populations therefore relatively demographically closed and genetically distinct from those on wild fish, and (iii) the lack of migration between the Atlantic and Pacific oceans would lead to clear genetic differentiation of L. salmonis populations in the two oceans. Methods Microsatellite development Total DNA was extracted (Puregene tissue extraction kit) from bulk samples of newly hatched nauplius I larvae from eggstrings (20 females) cultured in 0.45-µm-filtered seawater. Tissue from (nonfeeding) larvae was considered optimal for microsatellite development because it precluded host DNA contamination potentially mediated by epidermal fish mucus or nucleated host blood in adult parasite gut contents. Enriched libraries for dinucleotide (GA/TC) and trinucleotide (CAT) repeats were produced by Genetic Information Services, Chatsworth, California, using magnetic bead capture. Naupliar genomic DNA was first digested with a cocktail of seven four-, and six-base blunt-end restriction enzymes and then enriched. The enrichment products were ligated into the HindIII site of puc19 and electroporated into Escherichia coli DH5alpha. Fragments of base pairs (bp) were selected for ligation to permit single-pass sequencing. Sequences were obtained (Perkin Elmer ABI 373 automatic sequencer) using dye-labelled terminators and either puc19 (GIS designed for LsalSTA1 4) or M13/pUC universal forward and reverse primers (GibcoBRL, LsalSTA5). Amplification and visualization of loci Polymerase chain reaction (PCR) primers were designed using Primer3 (Rozen and Skaletsky 2000). The libraries yielded five polymorphic dinucleotide loci but no polymorphic trinucleotide loci. Microsatellite specificity was confirmed by the lack of amplification products for any primer pair using DNA extracted from Lepeophtheirus pectoralis or C. elongatus. Five of the microsatellite loci (LsalSTA1 5) were newly developed; for the sixth (LSNUIG14 = Ls.NUIG.14B of Nolan et al. 2000; TA n ), primers were redesigned to shorten the amplification fragment and facilitate size fractionation by electrophoresis. For population screening, total DNA was extracted from the cephalothorax of individual adult or preadult L. salmonis. PCR amplifications (MJ thermal cycler, MJ Research Inc., Waltham, Mass.) were undertaken in a 25-µL final volume under the following conditions: (i) initial denaturation for 1 min at 95 C, (ii) cycles of denaturation at 95 C for 30 s and annealing at C (optimized for each primer pair; Table 1) for 30 s with extension at 72 C for 30 s, and (iii) a final extension at 72 C for 2 min. Each reaction contained approximately 37.5 pmol of forward and reverse primers, 2 µl of 10 buffer (160 mmol (NH 4 ) 2 SO 4, 670 mmol Tris HCl (ph 8.8), 0.1% Tween-20), 5 mmol each nucleotide, 0.75 µl of MgCl 2 (50 mmol), unit of BioTaq (Bioline), and 25 ng of target DNA. Final products were separated by electrophoresis on 6% denaturing polyacrylamide gels and visualized by silver staining

4 Todd et al Fig. 1. Map of the Northern Hemisphere showing the locations of the Scottish (inset), Norwegian (NT, sea trout), and Canadian (NB and BC, farmed Atlantic salmon) sample sites for L. salmonis. Scotland: AS and AT, Atlantic salmon and sea trout; UT, sea trout; DT, sea trout; SP and SPS, 1SW and 2SW Atlantic salmon; TS and TT, Atlantic salmon and sea trout. Host sample codes are as in Table 2. (Promega). For every gel, alleles were scored against 10-bp ladder fragments and known positives of heterozygous individuals of designated fragment sizes. Where alleles of single base pair differences in fragment size were noted, these were conservatively binned (up) to facilitate analyses. Population studies Sampling was focused primarily on Scottish coasts and included locations adjacent to and remote from salmon farms. For convenience, we refer here to all samples, irrespective of host species or location, as populations. For reasons of commercial confidentiality, the Scottish aquaculture sites are referred to as Farms 1 7 and attributed only to administrative districts in Fig. 1. Lepeophtheirus salmonis were collected from wild Atlantic salmon (three sites), wild sea trout (four sites), and farmed salmonids (seven sites: Atlantic salmon (six) and rainbow trout (one)) between September 1997 and December 1998 (Fig. 1; Table 2). As a protocol, wild Atlantic salmon samples were restricted to 1SW fish (AS and TS) but sufficient numbers of 1SW (SP) and 2SW fish (SPS) were sampled from Strathy Point, north Scotland, to allow a comparison between the two maturity groupings. Because it is a coastal interceptory trapping station, it is not possible to ascribe a natal river to fish captured at Strathy Point (SP and SPS) and these data were excluded from the initial geographic analysis around Scotland. SP and SPS were included in the wider North Atlantic and Atlantic Pacific comparisons. For certain sites (e.g., Rivers Tweed and Annan), samples were obtained both from wild 1SW Atlantic salmon (TS and AS) and sea trout (TT and AT). The sea trout ranged in sea age from 1 to 4 years (TT) to 0+ to 2 years (AT). The North Atlantic geographic range was extended by samples from wild sea trout trapped just upstream of the estuary of the River Halselva, Altafjord, northern Norway (NT), in August 1999 and from Atlantic salmon farmed in Canadian coastal waters at New Brunswick (NB) (Fig. 1). A single Pacific sample (BC) was obtained from cultured Atlantic salmon at a British Columbia farm in May Sea lice were subsampled from total clearances of adult parasites taken from known individual fish, except for Scottish Farm 6, NB, and BC, where pooled samples of sea lice were provided. A maximum of 10 lice was included here for any one host fish and, where possible, a balanced parasite sex ratio was maintained. Hosts were chosen at random from those sampled at each site (or maturity grouping) and para-

5 1180 Can. J. Fish. Aquat. Sci. Vol. 61, 2004 Table 2. Sample locations and host species utilized in the present study. Species Wild (W) or farmed (F) Site name/district Site code Capture method Date sampled N Size range (fork length (mm) or wet weight (kg)) nm nf n S W Strathy Point SP Trap 23 June mm S W Strathy Point SPS Trap 23 June mm S W River Tweed TS Seine net 12 Aug mm S W River Annan AS Rod-and-line 27, 28 July kg S F Argyll & Bute F1 Sampled individually 29 July kg S F Sutherland F2 Sampled individually 17 June kg S F Wester Ross F3 Sampled individually 9, 26 June 11 na S F Wester Ross F4 Sampled individually 24 Feb., 29 Mar., 23 Apr. 12 na S F Western Isles F5 Sampled individually 4 Sept kg S F Sutherland F6 Pooled sample 6 July 11 na na na 51 R F Argyll & Bute F7 Sampled individually 3 Dec. na na T W River Annan AT Rod-and-line 5 July mm T W River Tweed TT Seine net 8 June mm T W River Dionard DT Rod-and-line 16, 18 July mm T W River Dundonnell UT Fyke net trap June mm S W River Halselva NT Trap 10 Aug mm S F New Brunswick NB Pooled sample May 2000 na na S F British Columbia BC Pooled sample 3 May 2000 na na Note: S, Atlantic salmon; T, sea trout; R, rainbow trout. Samples from 1998 unless otherwise stated. Host sizes are given as fork length or wet weight. N, number of hosts; nm, nf, and n, number of male and female and total number of lice screened, respectively; na, not available.

6 Todd et al sites were subsampled at random when infection intensity exceeded five male and five female parasites. The majority of parasites screened were adults, but preadults were included to maintain sample sizes from UT and DT sea trout, Farms 3 and 7, and from BC. Data analyses Statistical analyses of multilocus structure within the geographic data set were conducted using Arlequin (version 2.0) (Schneider et al. 2000). Missing values (PCR failures or equivocal scoring) ranged from 0 to 28 per population and, to preclude bias for any given population, all analyses were conducted both including and excluding individuals with missing values. In no instance were the inclusive/exclusive analytical outcomes qualitatively different, and since they retain the maximum information, only results for the analyses including missing values are reported here. Deviations from Hardy Weinberg expectations for each locus and each population were tested using an analogy to Fisher s exact test, but extended to a triangular contingency, followed by a modified version of a Markov chain random walk algorithm (Guo and Thompson 1992). The inbreeding coefficient, F IS, was computed for all loci and populations using GENEPOP (Raymond and Rousset 1995). Linkage disequilibrium between pairs of loci was tested using a likelihood ratio test whereby the empirical distribution was obtained by permutation (Slatkin and Excoffier 1996). Pairwise F ST fixation indices were estimated summed over all loci (Weir and Cockerham 1984; Michalakis and Excoffier 1996) and genetic structure was inferred by hierarchical analysis of molecular variance (AMOVA) (Excoffier et al. 1992). Fixation index significance was tested using a nonparametric permutation approach (Excoffier et al. 1992). AMOVA was undertaken at the allloci level and Bonferroni correction applied to all pairwise comparisons. If there is limited or restricted dispersal among populations, then differences in allele frequencies will accumulate by genetic drift. This results in patterns of isolation by distance (IBD) expressible as correlations between geographic distance and genetic differentiation (Wright 1943, 1969; Slatkin 1993). The null hypothesis throughout was that there would be no significant differentiation among the populations under consideration; but even if levels of differentiation are formally not statistically significant, the process of gradual differentiation may be detectable as a positive correlation. Geographic distances among the Scottish and other North Atlantic sample sites were calculated as a network distance assuming a least distance (headland headland) coastal pathway. Pairwise F ST values were plotted against geographic distance and patterns of IBD were assessed by least squares regression of genetic distance on geographic differentiation, by rank correlation, and by sum of square Mantel tests (implemented in GenStat7; Payne et al. 2003) with significance percentages derived from randomization. Results Microsatellite characterization We present the repeat structure, range of allele sizes found across the 18 populations, optimal annealing PCR temperature, and forward and reverse primer sequences for each locus (Table 1). A total of 1007 L. salmonis from 18 Atlantic and Pacific populations of wild and farmed host fish were screened at the six loci. Among the 6042 amplifications, there were 117 (1.9%) PCR failures, or individual genotypes that could not be unequivocally scored. PCR failures occurred across 96 individuals (9.5% of the grand total) and among 16 of the 18 populations (excepting Farm 6 and BC). The mean and median individuals per population with missing values were, respectively, 9.3% and 8.5%. Of the 96 individuals showing missing values, nine were missing data for two loci and six for three loci. The amplification failures per locus ranged from 0.7% to 2.8% of the 1007 individuals. The numbers of alleles scored across all 18 populations for the five new loci ranged from 14 (LsalSTA4) to 43 (LsalSTA5) (Table 3). Eleven alleles were scored for LSNUIG14, the redeveloped locus of Nolan et al. (2000). Nolan et al. (2000) reported 10 alleles for 45 L. salmonis screened from three populations (15 individuals each from Ireland, Scotland, and Norway). Like-for-like comparisons between the alleles scored for this locus by Nolan et al. (2000) and the present study are not possible because of the reduction in fragment size with our redeveloped primers and the allele binning procedure applied in the present study. No single population included all possible alleles for a given locus: the proportion of alleles per population ranged from 0.24 (LsalSTA2) to 0.70 (LsalSTA1). Of the 166 alleles scored across the six loci, 35 were unique to single populations. Most of the latter were of very low frequency (<0.015), except for allele 192 at LsalSTA4 (0.324, BC). Observed heterozygosities across populations and loci ranged from 0.40 to 0.97 and F IS over all loci across the 18 populations ranged from to (Table 3). Two of the 108 population locus heterozygosity combinations deviated significantly from expectations after Bonferroni correction: Farm 2 displayed some heterozygosity deficit for LsalSTA1 and AT for LsalSTA4. Likelihood ratio tests for linkage disequilibrium between all pairs of loci undertaken for all populations revealed only the one significant association, between LsalSTA2 and LsalSTA5 for the SPS population (P = ). Therefore, for the present data, the loci were considered independent and null alleles were not a significant feature. Levels of population genetic differentiation Wild Atlantic salmon: 1SW versus 2SW maturity groupings The initial analysis of genetic differentiation was between L. salmonis sampled from 2SW (n = 70) (SPS) and 1SW fish (n = 59) (SP) captured at Strathy Point. This comparison assessed populations borne by wild hosts, which probably will have experienced the maximum geographic separation in the northwest (2SW) and northeast (1SW and 2SW) Atlantic during their oceanic migrations. The analysis upheld the first null hypothesis that there is no significant differentiation of L. salmonis populations borne by the two maturity groupings (3024 permutations; F ST = 0.003, P = 0.086). A nested AMOVA comparing 13 fish (six SP and seven SPS) sampled for multiple parasites revealed no significant among-fish variation for L. salmonis infesting either maturity grouping. The respective variance components (i) among fish within populations and (ii) between populations were 1.27% and +0.49%.

7 1182 Can. J. Fish. Aquat. Sci. Vol. 61, 2004 Table 3. Summary of genetic parameters characterizing six polymorphic microsatellite loci. Locus SP SPS TS AS F1 F2 F3 F4 F5 F6 LsalSTA1 (27) H o H e P F IS A n LsalSTA2 (33) H o H e P F IS A n LsalSTA3 (38) H o H e P F IS A n LsalSTA4 (14) H o H e P F IS A n LsalSTA5 (43) H o H e P F IS A n LSNUIG14 (11) H o H e P F IS A n Over all loci F IS Note: The number in parentheses following the locus code is the total number of alleles scored across all populations for that locus. H o and H e, Weinberg expectations (significant values, after Bonferroni correction, are indicated in bold type); F IS, Wright s inbreeding coefficient (f; Weir and screened per population per locus. Host sample codes are as in Table 2. Geographic analysis of populations around the coasts of Scotland A total of 727 L. salmonis were analysed from 11 sites (four river estuaries (Annan, Tweed, Dionard, and Dundonnell) plus seven farms). No account was taken in this particular analysis of host species, other than maintaining as separate the populations sampled from single sites for wild Atlantic salmon and sea trout. Differentiation was assessed first by an examination of estimates of weighted F ST over all loci generated for the pairwise comparisons. This yielded no evidence of genetic structure among L. salmonis from the various species/site sources: the maximum pairwise F ST was and none were significantly different from zero (Table 4). This apparent Scottish panmixia was further supported by global AMOVA with wild salmon (n = 249 parasites, including SP and SPS), wild sea trout (n = 240), and farmed salmonids (n = 367, Atlantic salmon and rainbow trout) treated as three separate host groups or populations. Global F ST was nonsignificant (Table 5). Geographic analysis of populations throughout the North Atlantic Further comparisons were drawn between L. salmonis samples pooled (i) for all Scottish salmonids (n = 856, including SP and SPS) versus (ii) sea trout from northern Norway (n = 58) and (iii) farmed Atlantic salmon from eastern Canada (n = 59). This analysis yielded neither pairwise F ST values (probability ranges ) nor a global multilocus F ST estimate significantly different from zero (Table 5). The lack of differentiation of L. salmonis populations throughout the

8 Todd et al F7 AT TT DT UT NT NB BC observed and expected heterozygosities, respectively; P, probability of deviations from Hardy Cockerham 1984); A, number of alleles scored per population; n, number of L. salmonis North Atlantic, and irrespective of farmed or wild host species, therefore conformed to the second null hypothesis. IBD for the North Atlantic Although the levels of population differentiation (F ST ) within Scotland and throughout the North Atlantic were not statistically significant, it was still expedient to ascertain whether or not there is any indication of a positive correlation between genetic differentiation and geographic distance. Analyses of IBD were confined to the North Atlantic because the distances measured through the Northwest Passage between the North Atlantic and North Pacific (BC) are of arguable ecological validity, given the present lack of migration of host fish around the coasts of North America. Despite the eastern Canadian (NB) and Norwegian (NT) sites being separated by almost 6000 km, F ST in the North Atlantic (excluding SP and SPS) was shown by regression to be totally unrelated to geographic distance (intercept and slope essentially zero; Fig. 2) (Mantel product-moment correlation 0.178; randomized P value 0.758; Mantel rank correlation 0.281, 0.928; sum of square Mantel tests 67.23, 0.774). North Atlantic versus North Pacific comparison The final assessment of Atlantic Pacific differentiation was possible from a comparison between the Atlantic L. salmonis (n = 973) and the one Pacific population (BC) (n = 34). The F ST here (Table 5) was highly significant, leading to rejection of the third null hypothesis, but only 6% of the overall variation was between oceans. The R st (computed with GENEPOP) across all loci for these same two populations

9 1184 Can. J. Fish. Aquat. Sci. Vol. 61, 2004 Table 4. Matrices of estimated F ST values (lower left) and associated P values (upper right, all nonsignificant) for pairwise comparisons between L. salmonis sampled from wild and farmed Atlantic salmon, wild sea trout, and farmed rainbow trout around the coasts of Scotland. TS AS F1 F2 F3 F4 F5 F6 F7 AT TT DT UT TS AS F F F F F F F AT TT DT UT Note: Strathy Point 1SW (SP) and 2SW (SPS) data were excluded from these analyses. Host sample codes are as in Table 2. Table 5. Hierarchical AMOVA of L. salmonis sampled in Scotland and pooled into three groups according to host species/source (wild Atlantic salmon, wild sea trout, farmed Atlantic salmon and rainbow trout), sampled from Scotland (group 1) versus those sampled from elsewhere in the North Atlantic (sea trout, Norway, NT (group 2) and farmed Atlantic salmon, eastern Canada, NB (group 3)), and sampled from the North Atlantic (Scotland, Norway, and eastern Canada; group 1) versus those from the North Pacific (British Columbia, BC (group 2)). Source of variation df SS Variance component % total variance F statistic P Scotland: wild Atlantic salmon vs. wild sea trout vs. farmed Atlantic salmon and rainbow trout Among groups F CT = Among populations within groups F SC = Within populations F ST = Total North Atlantic: Scotland vs. Norway vs. eastern Canada Among groups F CT = Among populations within groups F SC = Within populations F ST = Total North Atlantic vs. North Pacific Among groups F CT = Among populations within groups F SC = Within populations F ST = < Total was This clear pattern of interocean differentiation was confirmed by a hierarchical AMOVA for the two ocean groupings but with the Atlantic sites nested into Norwegian, Scottish, and eastern Canadian subgroups (Table 5). Given the small sample size, the BC data do not warrant further detailed analysis; but from inspection of the allele frequency distributions (Fig. 3), it is apparent that there were very clear contrasts between the Atlantic and Pacific. Thus, LsalSTA3 and LsalSTA5 in the Atlantic samples showed a shift towards alleles of increased repeat length, whereas LsalSTA1 showed the reverse. Across all six loci, only three alleles (LsalSTA1 202 and LsalSTA4 190 and 192) of the 166 recorded for the two oceans were unique to the Pacific (BC) sample. Although these data show a clear pattern of genetic distinction and isolation between the two oceans, whether alleles of the same repeat lengths have the same identity and evolutionary history is, of course, unknown (e.g., Estoup and Cornuet 1999). Discussion Biogeographic distribution and origin of L. salmonis There were no fundamental analytical problems, arising from null alleles or from linkage disequilibrium, for the six loci screened, and only the North Atlantic versus North Pacific comparison revealed significant population genetic differentiation in L. salmonis. Isolation of Pacific and Atlantic L. salmonis was expected, given that there is no exchange of host fish between the two oceans, but the foregoing does beg the question as to whether this is a North Pacific species

10 Todd et al Fig. 2. Lepeophtheirus salmonis. Scatterplot and least squares regression of pairwise comparisons of the fixation index, F ST,on network distance (km 10 3 ) for the North Atlantic. Data are for 13 site locations (AS + AT, TS + TT, Farms 1 7, DT, UT, NT, and NB). Data for Strathy Point (SP and SPS) were excluded because the sampled fish could not be attributed to a natal river. Regression: y = x; r = Host sample codes are as in Table 2. that has invaded the North Atlantic or vice versa. Although the North Pacific and North Atlantic basins have been largely isolated since the Cenozoic, the recent opening of the Bering Strait during the Pliocene ~5 million years ago (Marincovich and Gladenkov 2001) may well have facilitated the migration of L. salmonis (perhaps parasitizing Arctic char) around North America during warmer intervals. For example, recent studies of various molecular lineages of the shallow-water bivalve mollusc Macoma balthica complex (Väinölä 2003) do indicate repeated trans-arctic invasions from a Pacific origin. Such may well extend to L. salmonis if this crustacean conforms to the predominant pattern of Pacific species having invaded the Atlantic rather than vice versa (Vermeij 1991; Wares 2001). Within the North Atlantic, it is not clear which of the three candidates (S. salar, S. trutta, and S. alpinus) is the ancestral host species. Lepeophtheirus salmonis and its associated salmonids are widespread in both the North Atlantic and North Pacific, and there are marked contrasts in infection intensities across the spectrum of fish species (Nagasawa et al. 1993; Tingley et al. 1997; Todd et al. 2000). Nagasawa (2001) monitored North Pacific commercial longline catches of Oncorhynchus spp. for numbers of adult female L. salmonis: pink salmon (Oncorjhynchus gorbuscha) appears to be the primary Pacific host species, with 94% prevalence and an overall arithmetic mean abundance of 5.6 females per fish. This compares with the 100% prevalence and an antilogged annual mean abundance of adult female L. salmonis on Scottish 1SW Atlantic salmon (C.D. Todd, unpublished data). Population genetic differentiation, IBD, and North Atlantic panmixia Two of the three null hypotheses were upheld, indicating that there is effective panmixia of Scottish L. salmonis populations, irrespective of host species and of their being wild or cultured. The samples from Norway (NT) and eastern Canada (NB) confirmed a lack of significant differentiation or IBD of L. salmonis populations across the North Atlantic. The latter samples were especially important in this context because they derived from wild sea trout and farmed salmon, respectively. By definition, both of these host fish groups will remain close to (NT), or strictly within (NB), their sampled location and all L. salmonis will have been recruited locally as larvae. Lepeophtheirus salmonis throughout the North Atlantic therefore has to be considered a single, homogeneous population with an open demography. But it is important to note that this does not mean that local differentiation is impossible. Randomly chosen microsatellites will not reveal population genetic differences or structuring arising from possible selection for alleles associated with chemical resistance (see below). Selection-driven differentiation will affect only the region of the genome under selection (e.g., Chevillon et al. 1995) and recombination will strip away markers that are not tightly linked to loci under selection (e.g., Barton 1983; Barton and Gale 1993). Our results are therefore not incompatible with suppositions in the aquaculture industry of farm sites being essentially self-reinfesting (e.g., see Bron et al. 1993) and the possibility of localized resistance to chemical treatments developing among sea lice on farms (e.g., Treasurer et al. 2000; Tully and McFadden 2000; Sevatdal and Horsberg 2003). The obligatory planktonic larval phase in the life cycle of L. salmonis must result in its populations on salmon farms being demographically open and wild fish being vulnerable to infection from those sources. Nonetheless, Scottish farms do attempt to minimize self-reinfestation by a range of management strategies, including fallowing breaks in local production, maintenance of only single yearclasses of salmon at farm sites, and regionally coordinated lice treatment programmes (e.g., Butler 2002; Revie et al. 2002; Heuch et al. 2003). Although evidence for chemical resistance of L. salmonis is equivocal, increased resistance to pesticide treatments among invertebrates has a genetic basis (e.g., Lenormand et al. 1999), and if populations are large and free from genetic drift, as for L. salmonis throughout the North Atlantic, then even relatively weak selection could still lead to local increases in frequency of insecticide resistance genes. Outwith the farm environment, however, such resistance genes could be selected against. But irrespective of this, differentiation would relate only to those, and closely linked, loci and would not be detected in surveys of microsatellite variation. General genetic differentiation would be found only if resistance relied on multiple loci spread throughout the genome, and insecticide resistance generally seems to be effected by few loci (e.g., McKenzie and Batterham 1994). Genetic homogeneity of North Atlantic L. salmonis undoubtedly is largely explicable by the migratory behaviour of Atlantic salmon. Coupled with host mobility are the repeated production of eggstrings (Nordhagen et al. 2000),

11 1186 Can. J. Fish. Aquat. Sci. Vol. 61, 2004 Fig. 3. Lepeophtheirus salmonis. Frequency histograms of allelic diversity for the six microsatellite loci: (a) LsalSTA1, (b) LsalSTA2, (c) LsalSTA3, (d) LsalSTA4, (e) LsalSTA5, and (f) LSNUIG14. Data are for all 17 North Atlantic samples are pooled (open bars) and the single Pacific sample (BC) (solid bars). temporally extended gravid periods for individual female lice, and year-round reproductive capacity of the parasite at all sea temperatures experienced by the host. The copepodids that colonize smolts on their first migration to sea in April May will become reproductive within 1 2 months, according to temperature. By that time, the host fish will be well to the north in the open Atlantic and intermixing with fish of varied geographic origin. For returning adult MSW fish, and probably also for 1SW fish, their adult parasite burdens at return to natal coastal waters will all have been acquired as larvae in the open ocean. 1SW Atlantic salmon of Iberian, British Isles, Scandinavian, and Russian origin will readily cross-infect one another in the Norwegian Sea, and west European MSW fish can similarly cross-infect North American

12 Todd et al fish off Greenland. Russian fish may also be present at west Greenland, and North American fish have been captured in the Norwegian Sea (Hansen and Jacobsen 1999, 2003). Having cross-infected at more or less discrete areas (such as oceanographic fronts) on the open-ocean feeding grounds, the fish then separate for their return migration, bearing parasites that ultimately will release larvae some weeks or months later in coastal waters, both in North America and from Arctic to temperate latitudes in western Europe. These last released larvae will, therefore, be available for vertical transmission to descending salmonid smolts of all three host species in springtime and horizontal transmission to adult wild or farmed hosts in coastal waters. The F statistic analyses confirm that gene flow among the various host species throughout the North Atlantic is at sufficiently high levels to prevent divergence of L. salmonis populations as a result of random genetic drift. This overall result has considerable implications: first, for the management of sea lice infestations on wild and farmed salmonids (including the spread of genes associated with possible resistance) and, second, in relation to the controversy surrounding the infection interactions between these two stock components and especially the population declines of wild sea trout. The farmed versus wild debate with respect to declines of sea trout populations Coincident with the early 1990s population crashes of sea trout in the British Isles and Norway were the first observations of premature migratory return to fresh water, after only days or weeks, of large numbers of heavily lice-infested juvenile sea trout in poor physiological condition (Tully and Whelan 1993; Tully et al. 1993a). The bulk of the Scottish aquaculture industry is restricted to the fjordic sea lochs of the mainland west coast. It is there that the greatest declines in wild sea trout stocks and the occurrence of sea lice epizootics have been recorded. Gravid female L. salmonis occur on farmed salmonids in the sea throughout the year (e.g., Bron et al. 1993; Revie et al. 2002), and the recent controversy between the industry and wild salmonid interest groups has centred on the provenance of copepodid larvae that infest wild salmonids (and especially smolts as they enter seawater for the first time) and that might be responsible for epizootics and contribute to stock declines of wild sea trout (e.g., Butler 2002). Official Scottish catch statistics for wild sea trout (Anonymous 2003) show a long-standing decline in numbers since the 1960s, and following especially marked crashes of some west coast sea trout stocks in the early 1990s, numbers for both Atlantic salmon and sea trout in Scotland presently are at historical lows. Wild salmonids must have been the source of the initial infestations of farmed stocks in the 1970s, and wild fish are a persistent source of infection to farmed stocks. But it is unarguable that the numbers of wild salmonids migrating through Scottish coastal waters now are extremely low compared with the numbers of Atlantic salmon held year-round in aquaculture cages (see also Butler 2002). Current annual Scottish farmed salmon production is ~ t, which contrasts sharply with the most recently available (2002) data on the wild salmon catch in Scotland of ~280 t ( fish; Anonymous 2003). The annual numbers of returning wild salmon in Scotland presently total perhaps only ~ , or ~1% of the farmed total (see also Butler 2002). Are salmon farms the prime source of most sea lice larval output along the Scottish west coast? To what extent are farmed salmon responsible for the typical infestations of wild Atlantic salmon and sea trout? Sea trout smolts in west Scotland first migrate to sea in April May and most return in summer autumn to overwinter in fresh water, either as immatures or as spawning adults (Pemberton 1976). Juveniles appear to spend all or much of their marine residence period close inshore. Recent plankton surveys in one sea loch (M. McKibben and D. Hay, Freshwater Fisheries Laboratory, Pitlochry, Scotland, PH16 5LB, UK, personal communication) have shown that L. salmonis copepodids are temporally variable in their local distribution and abundance, but there are concentrations at salinity fronts and also close to the shoreline. Sea trout postsmolts might therefore be extremely vulnerable to rapid and repeated infestation. Wild Atlantic salmon return to UK coasts throughout the year, with the large majority returning between July and October. Because of the general scarcity of early-returning MSW ( spring ) salmon on the Scottish west coast, the presumption often has been drawn that the likeliest sources of copepodids infesting sea trout smolts are salmon farms (e.g., Butler 2002). There will, however, also be low numbers of salmon and sea trout kelts, which will have descended to coastal waters in the weeks prior to smolt migration, in addition to unknown numbers of marine-overwintering adult sea trout and escaped farm salmon. Todd et al. (1997) reported no significant differentiation (Weir and Cockerham s theta, two allozyme loci) of populations of L. salmonis sampled at two east Scotland river estuaries and two west Scotland salmon farms. Contrary to our previous studies of random amplified polymorphic DNA variation (Todd et al. 1997) and to inferences drawn by Tully and Nolan (2002) from a variety of perspectives (e.g., apparent self-reinfestation by farm sites, possible development of chemical resistance), the present data provide no evidence to suggest that there is substructuring or differentiation of L. salmonis populations infesting wild and farmed salmonids. Overall, gene flow and migration (crossinfection) between parasite populations on the two host components is at a high level. It is not possible to quantify the relative strengths of wild to farmed and farmed to wild migration, but given the inevitable export of nauplii from farm cages to adjacent waters and the striking contrasts in wild and farmed fish tonnages, it is apparent that L. salmonis gene flow (= larval colonization) from farmed to wild will be very much greater than from wild to farmed. The circumstantial evidence that sea lice are at least a major contributory factor to the recent declines and crashes of sea trout stocks in parts of Ireland, Scotland, and Norway is now considerable, and for wild Atlantic salmon, there is good evidence that L. salmonis may well be detrimental to smolts migrating through Norwegian fjords (Finstad et al. 2000). Our data show that there is no genetic separation between L. salmonis populations on wild and farmed salmonids in the North Atlantic, and given the indicated high net levels of production of infective lice larvae from fish farms, the management of L. salmonis by the international salmon