Aquaculture‐driven evolution of the salmon louse mtDNA genome

Abstract Resistance toward the antiparasitic pyrethroid, deltamethrin, is reported in the Atlantic salmon louse (Lepeophtheirus salmonis salmonis), a persistent ectoparasite of farmed and wild salmonids. The resistance mechanism is linked to mitochondrial DNA (mtDNA), where genetic markers for resistance have been identified. Here, we investigated how widespread pyrethroid use in aquaculture may have influenced mtDNA variation in lice, and the dispersion of resistant haplotypes across the North Atlantic, using historical (2000–2002 “pre‐resistance”) and contemporary (2014–2017 “post‐resistance”) samples. To study this, we sequenced ATPase 6 and cytochrome b, genotyped two genetic markers for deltamethrin resistance, and genotyped microsatellites as “neutral” controls of potential population bottlenecks. Overall, we observed a modest reduction in mtDNA diversity in the period 2000–2017, but no reduction in microsatellite variation was observed. The reduction in mtDNA variation was especially distinct in two of the contemporary samples, fixed for one and two haplotypes, respectively. By contrast, all historical samples consisted of close to one mtDNA haplotype per individual. No population genetic structure was detected among the historical samples for mtDNA nor microsatellites. By contrast, significant population genetic differentiation was observed for mtDNA among some of the contemporary samples. However, the observed population genetic structure was tightly linked with the pattern of deltamethrin resistance, and we therefore conclude that it primarily reflects the transient mosaic of pyrethroid usage in time and space. Two historically undetected mtDNA haplotypes dominated in the contemporary samples, both of which were linked to deltamethrin resistance, demonstrating primarily two origins of deltamethrin resistance in the North Atlantic. Collectively, these data demonstrate that the widespread use of pyrethroids in commercial aquaculture has substantially altered the patterns of mtDNA diversity in lice across the North Atlantic, and that long‐distance dispersion of resistance is rapid due to high level of genetic connectivity that is observed in this species.

tigated how widespread pyrethroid use in aquaculture may have influenced mtDNA variation in lice, and the dispersion of resistant haplotypes across the North Atlantic, using historical (2000-2002 "pre-resistance") and contemporary (2014-2017 "postresistance") samples. To study this, we sequenced ATPase 6 and cytochrome b, genotyped two genetic markers for deltamethrin resistance, and genotyped microsatellites as "neutral" controls of potential population bottlenecks. Overall, we observed a modest reduction in mtDNA diversity in the period 2000-2017, but no reduction in microsatellite variation was observed. The reduction in mtDNA variation was especially distinct in two of the contemporary samples, fixed for one and two haplotypes, respectively. By contrast, all historical samples consisted of close to one mtDNA haplotype per individual. No population genetic structure was detected among the historical samples for mtDNA nor microsatellites. By contrast, significant population genetic differentiation was observed for mtDNA among some of the contemporary samples. However, the observed population genetic structure was tightly linked with the pattern of deltamethrin resistance, and we therefore conclude that it primarily reflects the transient mosaic of pyrethroid usage in time and space. Two historically undetected mtDNA haplotypes dominated in the contemporary samples, both of which were linked to deltamethrin resistance, demonstrating primarily two origins of deltamethrin resistance in the North Atlantic. Collectively, these data demonstrate that the widespread use of pyrethroids in commercial aquaculture has substantially altered the patterns of mtDNA diversity in lice across the North Atlantic, and that

| INTRODUC TI ON
Pests and the evolution of pesticide resistance are regarded among the major challenges that global food production faces.
Understanding both resistance development and dispersal are important to provide better management strategies and prolong the efficacy of pesticides (Clark & Yamaguchi, 2002). The ectoparasitic salmon louse Lepeophtheirus salmonis has developed widespread resistance against the most used pesticides and has consequently become a persistent challenge to environmentally sustainable salmon aquaculture (Taranger et al., 2015).
The salmon louse is an obligate parasite of salmonid fishes and belongs to the Caligidae family of copepods (Boxaspen, 2006). The species exists as two genetically distinct subspecies residing in the Atlantic and Pacific . Their life cycle comprises eight developmental stages (Hamre et al., 2013). In the first two lecithotrophic planktonic larvae stages (Nauplius I and II) and the infective copepodid stage, salmon lice drift with the ocean currents. After attaching to a host, they molt into the chalimus I and II, pre-adult I, pre-adult II, and reproductive adult stages (Hamre et al., 2013). During the infective stages, lice feed on host epidermal cells, mucus, and blood (Boxaspen, 2006). For the host, this may give rise to osmoregulatory problems, stress, and wounds that increase the chance of infection by secondary pathogens (Birkeland & Jakobsen, 1997;Bjørn & Finstad, 1998;Fjelldal et al., 2020;Grimnes & Jakobsen, 1996;Nolan et al., 1999) that may ultimately result in death (Bjørn & Finstad, 1998;Fjelldal et al., 2020;Grimnes & Jakobsen, 1996).
During its larval planktonic stages, lice typically drift between 20 and 45 km before attaching to a host (Johnsen et al., 2016). Further dispersal depends on the type of host it attaches to (Johnsen et al., 2016), which in the Atlantic primarily means Atlantic salmon (Salmo salar) or anadromous sea trout (Salmo trutta) and in some areas Arctic charr (Salvelinus alpinus). While sea trout typically undergo shortdistance migrations in the coastal zone (Berg & Berg, 1987), Atlantic salmon from the western and eastern Atlantic migrate long distances from their coastal rivers to common oceanic feeding grounds (Gilbey et al., 2021). Thus, while attached to a wild salmon host, lice may disperse and exchange genetic material over long distances.
Population genetic studies, based on different types of markers, have revealed a lack of genetic structure in lice throughout the Atlantic and have largely attributed these observations to the long-distance exchange of genetic material while attached to a long-migrating salmon host (Dixon et al., 2004;Glover et al., 2011;Tjensvoll et al., 2006;Todd et al., 2004;Nolan and Powell, 2009). Some studies, however, have reported genetic differences in samples of lice from the North Atlantic when "outlier loci," that is, loci that are likely under selection, are included (Besnier et al., 2014;Glover et al., 2011;Jacobs et al., 2018). It is well documented that the analysis of loci under selection often identifies population genetic structure in highly marine species where presumed neutral markers often fail to reveal genetic structure (Breistein et al., 2022;Han et al., 2020;Quintela et al., 2020;Russello et al., 2012). In an extensive population genomic study on lice, multiple outlier loci involved in selective sweeps on chromosomes 1, 5, and 14 were identified (Besnier et al., 2014). However, these outliers did not reveal any geographically meaningful population structure. These authors identified two sweeps causatively linked to selection from the use of chemical delousing agents. Based on the distribution of resistance, it was therefore concluded by those authors that lice most likely consist of a single panmictic population in the North Atlantic, but that spatial and temporal patterns of pesticide usage may give rise to transient mosaic patterns of genetic differences among samples in markers linked to resistance.
A variety of delousing methods have been utilized to control lice on farmed salmonids (Overton et al., 2019). Chemical treatments have been the most important approach and include both orally administered drugs and bath treatments (Aaen et al., 2015).
However, access to only a few pesticides has created a strong selection pressure driving recurrent developments of resistance (Aaen et al., 2015). Among the pesticides rendered, largely inefficient is the pyrethroid deltamethrin, which is applied through bathing and was introduced in Norway in 1994 (Aaen et al., 2015). Several farms had reported loss in treatment efficiency (Aaen et al., 2015) and the peak of deltamethrin resistance measured by bioassays was reported in 2016 and 2017 (Jensen et al., 2020).
Several studies have demonstrated maternal inheritance of deltamethrin resistance in salmon lice, and since mitochondrial inheritance is commonly regarded as exclusively maternal, it supports the reported mitochondrial linkage of resistance (Bakke et al., 2018;Carmona-Antoñanzas et al., 2017;Giles et al., 1980;Tschesche et al., 2022). A study on resistant and sensitive strains of salmon lice collected in Norway revealed five mtDNA SNPs showing association with deltamethrin resistance (Nilsen & Espedal, 2015). Later on, when studying resistant and sensitive salmon lice collected in 2018-2019 in Scotland, three mtDNA haplotypes associated with deltamethrin resistance were identified (Tschesche et al., 2022). All resistant haplotypes shared one mutational SNP, T8600C located in the Cox1 gene, which was therefore suggested to be the causative marker for resistance (Tschesche et al., 2022). long-distance dispersion of resistance is rapid due to high level of genetic connectivity that is observed in this species.

K E Y W O R D S
aquaculture, deltamethrin, environmental impact, Lepeophtheirus salmonis, welfare Two recent studies have revealed almost uniform mtDNA haplotype composition among deltamethrin-resistant salmon lice collected in farms in Scotland (Carmona-Antoñanzas et al., 2017;Tschesche et al., 2022). This is presumably the result of selection driven by the pyrethroids deltamethrin and cypermethrin delousing agents. This pattern strongly contrasts the high number of mtDNA haplotypes observed in a study of salmon lice throughout the North Atlantic in the period 2000-2002, before extensive deltamethrin usage (Tjensvoll et al., 2006). These observations support earlier suggestions that aquaculture represents a major driver of the population dynamics of salmon lice (Fjørtoft et al., 2017(Fjørtoft et al., , 2020(Fjørtoft et al., , 2021 and can induce rapid evolutionary changes (Besnier et al., 2014;Dempster et al., 2021;Mennerat et al., 2017).
The present study was designed to address the emergence and dispersion of deltamethrin resistance across the North Atlantic, and to investigate how widespread selection for resistance has influenced mtDNA variation after two decades of extensive pyrethroid use. In order to address these aims, a collection of historical (year 2000-2002, "pre-resistance") and contemporary (year 2014-2017, "post-resistance") salmon lice sampled throughout the North Atlantic were assigned to resistant/non-resistant status based on markers for pyrethroid resistance, and, sequenced for two mitochondrial genes: ATPase 6 (A6) and Cytochrome B (Cyt b). Our most important findings were: an overall reduction in mtDNA variation of salmon lice with time, in addition to an emergence of two haplotypes that were not present in any of the historical samples but now dominate in the North Atlantic, both of which are linked to the causative mutation for deltamethrin resistance.

| Samples and data sources
The present study is based on genetic data from 493 salmon lice collected throughout the North Atlantic, including 180 salmon lice collected between the years 2000 and 2002 (hereafter referred to as the historical samples-that is, before widespread pyrethroid resistance was reported) and 313 salmon lice collected between the years 2014 and 2017 (hereafter referred to as the contemporary samples-that is, after widespread pyrethroid resistance was reported) (Table 1, Figure 1). Both the historical and contemporary samples originate from larger sample collections used in multiple research projects (Besnier et al., 2014;Fjørtoft et al., 2017Fjørtoft et al., , 2020Fjørtoft et al., , 2021Glover et al., 2011;Tjensvoll et al., 2006). Although some of the data presented here existed from previous studies, most of it was produced within the present study ( Table 2).
The data in this study include (i) the genotypes providing information on resistance to pyrethroids (deltamethrin), using two genetic markers for this. The first marker for deltamethrin resistance

TA B L E 1
Overview of the historical and contemporary samples of salmon lice used in the present study.
is the patented non-causative genetic marker C14065T that has been associated with resistance (Nilsen & Espedal, 2015). The data for this marker were obtained from Fjørtoft et al. (2021).
The second marker included the T8600C mutation that was recently discovered (Tschesche et al., 2022), and it is presumed to be causative for deltamethrin resistance. Data for this marker were acquired in the present study by genotyping all of the samples with this marker as described below; (ii) sequences for the mitochondrial genes A6 and Cyt b that were used to study spatial and temporal changes in mtDNA variation between the historical (Tjensvoll et al., 2006) and contemporary salmon lice samples; and (iii) genotypes at 15 microsatellite DNA loci. Microsatellites are highly effective in picking up strong demographic events such as inbreeding and strong population bottlenecks (Skaala et al., 2004;Skern-Mauritzen et al., 2013). Microsatellites were therefore included in the present study to provide a presumed selectively neutral nDNA control for potential temporal or spatial patterns observed in the mtDNA variation.

| Description of genetic analyses
The molecular analyses performed within the present study comprised mtDNA sequencing of the contemporary samples, genotyping microsatellites on the historical and contemporary samples, and genotyping the T8600C SNP on the contemporary samples.
These procedures are described below. Molecular analyses associated with previously published data are found in the original references.

F I G U R E 1
Map over sample locations for salmon lice collected in the North Atlantic. Sampling year is placed behind each locations shortened name. Map created using the R-package ggplot2 (Wickham, 2016).

TA B L E 2
Overview of the types of data used in the present study and their origins. and reverse sequences were then assembled into contigs, and consensus sequences were aligned against the historical sequences using MUSCLE (Edgar, 2004) default settings in Geneious Prime 2021.2.2. All polymorphisms (i.e., SNPs) were quality checked by visual inspection of chromatograms.

| Microsatellite genotyping
The historical and contemporary salmon lice samples were genotyped for 15 microsatellite DNA loci. Genotyping was conducted in three multiplex reactions consisting of the following markers: Multiplex one; Lsa1STA1, Lsa1STA2, Lsa1STA4, Lsa1STA5, LsNUIG14 (Todd et al., 2004). Multiplex two; Lsal103EUVC, Lsal109EUVC, Lsal110EUVC, Lsal111EUVC  and NUIG9 (Nolan et al., 2000). Multiplex three; Lsal104EUVC, Lsal105EUVC, Lsal106EUVC, Lsal108EUVC  and Lsa1STA3 (Todd et al., 2004). With minor concentration adjustments, PCR and amplification conditions followed those described previously (Glover et al., 2011). Microsatellites were independently scored by two different people, followed by a routine check of the genotyping consistency as recommended by Pompanon et al. (2005). This was performed by reanalyzing 144 individuals. After independent scoring of the reanalyzed individuals, 489 genotypes spread across the 15 microsatellite loci were compared between first and second genotyping. A total of two inconsistencies were observed giving an overall allelic genotyping consistency rate of 99.8%. Due to challenges relating to DNA quality, a 50% threshold for allowed missing data per individual as a combined measure for all 15 markers was adopted. Therefore, the final microsatellite dataset consisted of genotypes for 365 lice displaying >50% genotyping coverage (other thresholds were initially tested, all giving highly similar results, data not presented).

| Deltamethrin-resistant SNP amplification, genotyping, and processing
Data for resistance status according to a patented non-causative marker for pyrethroid resistance C14065T (Nilsen & Espedal, 2015) were obtained from Fjørtoft et al. (2021). The more recently discovered and presumed causative mutation for pyrethroid resistance, T8600C (Tschesche et al., 2022), was identified by genotyping in the present study on the contemporary samples. The historical lice samples used in the present study were checked for T8600C mutation in sequence data published previously by Tjensvoll et al. (2006) (Table 2) and were all found to be sensitive according to the T8600C marker.
Genotyping the T8600C marker for the contemporary samples was done by a 5´nuclease reverse transcriptase real-time quantitative PCR TaqMan assay. Primers used were targeting the mutant SNP T8600C (accession number; LT630766) (Tschesche et al., 2022).

Two-step amplification with 45 cycles was conducted on Applied
Biosystem 7500 real-time PCR following the manufacture instructions (ThermoFisher). The salmon lice were genotyped as either resistant or sensitive for deltamethrin. Positive controls belonging to the historical samples with known genotypes (Genebank accession number: AY602587-AY602589, AY602591, AY602596) were included, and resulted in expected sensitive genotypes for the T8600C genotyping.

| Genetic variation
For the mitochondrial sequences, differences in genetic variation among samples were measured for the combined haplotypes (i.e., where the combined gene haplotypes for A6 and Cyt B represented a single mitochondrial haplotype) using standard molecular indices in the software DnaSP v6 (Rozas et al., 2017). To account for uneven sample sizes, a rarefaction method was used to quantify haplotype richness (HR; number of haplotypes per sample adjusted for minimum sample size) and private haplotype richness (PR; level of unique haplotypes per sample adjusted for the minimum sample size) in the software HP-RARE v1.0 (Kalinowski, 2005). In the rarefaction analy-  (Dieringer & Schlötterer, 2003). For allelic richness, the minimum sample size was set to 20 individuals, and samples below the minimum sample size were excluded from the analysis of allelic richness.
FSTAT (Goudet, 1995) was used in the measure of allelic richness when comparing between the historical and contemporary dataset, with a minimum sample size set to 144 individuals.
The difference in genetic variation between the historical and contemporary samples was tested with a two-sided t-test using the t.test function from the stats package in R studio v1.4.1106 (R Core Team, 2021). Differences were considered statistically significant if p < 0.05.

| Population genetic structure
Although the analysis of population genetic structure was not a major part of the present study, temporal and spatial changes in population genetic structure for the combined mtDNA haplotypes across both of these genes, and for the 15 microsatellites, were measured using analysis of pairwise genetic differentiation (F ST for microsatellites and Φ ST for haplotypes (Weir & Cockerham, 1984)) implemented in Arlequin v3.5.2.2 (Excoffier & Lischer, 2010). Pairwise genetic differentiation for the microsatellites was measured using F-statistics to compare differentiation in allelic frequencies between samples (F ST ). For the mitochondrial genes, pairwise genetic differentiation was measured using F-statistics to calculate the genetic distance between haplotypes (Φ ST ).
Population genetic structure was also investigated for each mtDNA gene separately (see section 1.2 in Data S1). The significance of the calculated statistics measured by the analysis of pairwise differences was tested using the permutation approach in Arlequin v3.5.2.2 (Excoffier & Lischer, 2010). The significance level of 0.05 and 10,000 permutations was set to test the null hypothesis of a panmictic population of North Atlantic salmon lice for both mtDNA and microsatellites. The resulting p-values were corrected using the Holm-Bonferroni sequential correction method (Gaetano, 2013).

| Haplotype distribution
To visualize the distribution of haplotypes in all historical and contemporary samples, networks were generated using sequence data for A6 and Cyt b combined, with the software PopART (Leigh & Bryant, 2015). Haplotype networks were also generated for each mtDNA gene separately (section 1.3 in Data S1). All haplotype networks were constructed using the medium joint network approach, with the search for medium vectors set to a low minimum (ε being 0) to ease the presentation of the network.
Furthermore, haplotype networks based on resistance data were generated for the contemporary samples. Deltamethrin resistance based on the C14065T marker (Nilsen & Espedal, 2015) as well as the T8600C marker (Tschesche et al., 2022) were used separately to identify mtDNA haplotypes as being either resistant or sensitive. (historical samples) (Tjensvoll et al., 2006).

| Data description
Of the 249 contemporary salmon lice successfully sequenced for A6 and Cyt b, 237 were also successfully genotyped for the T8600C genetic marker to check resistant status for deltamethrin.
The 180 historical salmon lice were checked for resistance status according to the T8600C genetic marker by studying the COX-1 gene sequences presented in by Tjensvoll et al., 2006 (

| Pyrethroid resistance
None of the lice representing the historical samples were pyrethroid resistant according to either the C14065T or the T8600C markers.
In the contemporary samples, 39% of the lice were identified as resistant from at least one of the two markers, although this differed by marker (Table 3). According to C14065T, the highest percentages of deltamethrin resistance were observed in the SCO2016 sample (47%;

| Genetic variation in mtDNA sequences and microsatellites
Standard indices of genetic diversity within each of the historical and contemporary samples were calculated using data from the two mtDNA genes combined into haplotypes (Table 3), for the two mtDNA genes separately (section 1.1 in Data S1; Tables S1 and S2), and for the 15 microsatellite loci (Table 3).
For the two mtDNA genes combined, 249 salmon lice from the contemporary samples displayed 133 unique haplotypes (Table 3).  HR: t (9.07) = −4.70, p = 0.0001). Other estimates of haplotype diversity and richness were consistent with these results (Table 3).
Across the 15 pooled microsatellites, a total of 180 and 169 alleles were observed for the contemporary and historical samples, respectively (Table 3)

| Population genetic structure
To investigate spatial and temporal genetic structure among samples, a full matrix of pairwise comparisons were conducted using the two mtDNA genes combined (Table 4a), the mtDNA genes separately (section 1.2 in Data S1; Tables S3 and S4), and the 15 microsatellite loci (Table 4b).
For the analyses using the combined mtDNA sequences, no significant genetic differentiation was observed among the historical samples (Table 4a). However, among the contemporary samples, significant population genetic differentiation was observed. In particular, the FAR2016 and SCO2016 samples deviated significantly from each other and from all other samples both compared to the historical and contemporary samples. It is important to note that these samples comprised only one (FAR2016) and two haplotypes (SCO2016), respectively, which is the reason for the substantial genetic differentiation observed. By contrast, no genetic differentiation was observed among any historical or contemporary samples based on the microsatellite data (Table 4b).

| Haplotype distribution
Haplotype networks were generated for the two mtDNA genes combined (Figures 2-4) and separately (section 1.3 in Data S1; In the contemporary sample, deltamethrin-resistant salmon lice according to both the C14065T and the T8600C genetic marker were only found among five haplotypes (section 1.3 in Data S1; Figure S8c and Figure 3).
According to T8600C, which is regarded as the causative mutation for pyrethroid resistance, almost all individuals displaying haplotypes 165 and 174 were resistant (Figure 3). By contrast, according to C14065T, which is a non-causative marker for resistance, the majority of individuals that were of haplotype 165 were resistant while the majority of individuals that were of haplotype 174 were marked as sensitive (section 1.3 in Data S1; According to the T8600C marker, all lice found in the SCO2016 and FAR2016 samples were resistant (Table 3). By contrast, the highest percentage of deltamethrin-resistant salmon lice identified by the C14065T marker was 47% found in the SCO2016 sample (Table 3). The SCO2016 sample only included haplotypes 165 and 174 (Figure 3b). According to the C14065T marker, 12 of the 13 SCO2016 individuals that were of haplotype 165 were identified as resistant, whereas 16 of the 19 SCO2016 salmon lice that were of haplotype 174 were suggested to be sensitive (section 1.3 in Data S1; Figure S8a,b). All lice within the FAR2016 sample were of haplotype 174 (Figure 3b), and notably, all were resistant to deltamethrin based on the T8600C marker ( Figure 3b). By contrast, all were susceptible to deltamethrin based on the C14065T marker (section 1.3 in Data S1; Figure S8a).
Haplotype 174 dominated in lice from the Faroe Islands but was also well represented in Scotland and Ireland (Figure 4). Haplotype 165 was predominantly found along the Norwegian coastline, except for the southern-eastern part of Norway. At all sample locations, either both or one of the emerging haplotypes (haplotype 165 and haplotype 174) were found, except for the sample collected on the eastern coast of Canada (CAN2017).

| DISCUSS ION
Our analyses using an amphi-Atlantic set of historical (preresistance) and contemporary (post-resistance) samples of salmon lice show: (1) an overall reduction in mtDNA variation with time, (2) the emergence of two new haplotypes linked to the causative mutation for deltamethrin resistance that now dominate in the North Atlantic, and (3) no evidence of population genetic structure throughout the North Atlantic in genetic markers not associated with chemical resistance. Based on these results, we conclude that widespread use of deltamethrin in commercial fish farms has caused changes in the salmon louse mtDNA genome, leading to widespread dispersal of resistance from a minimum of two independent sources. We also suggest, based on these results and work from previous studies, that when the temporal and spatial mosaic of pesticide use is accounted for, this species displays little evidence of population genetic structure throughout the North Atlantic.

| Temporal changes in mtDNA variation
Historically, very high mtDNA variation has been observed in lice (Tjensvoll et al., 2006), as illustrated by the high proportion of singleton haplotypes in the historical samples analyzed here.
We observed a modest but nevertheless significant reduction in mtDNA variation in the contemporary samples, which was still ob-  Canada appeared to be 100% susceptible (Table 3) and displayed a high mtDNA diversity. This most likely reflects the limited deltamethrin usage in Canada (Fjørtoft et al., 2021), resulting in a reduced selection pressure on the haplotype resistant for deltamethrin. It is important to note that this does not suggest a lack of population genetic connectivity with lice in other regions of the North Atlantic (Besnier et al., 2014; and see discussion below), but is more likely to reflect the "time-lag" of resistance dispersion coupled with the lack of selection for resistance in that region. The same pattern has been observed for organophosphate resistance (Fjørtoft et al., 2017).

| Population genetic structure
Although not a major part of the current study, we found no evidence of population genetic structure among the historical samples for mtDNA nor among the historical or contemporary samples for microsatellites. By contrast, significant population genetic structure was observed for mtDNA among the contemporary samples collected from farmed hosts. However, this was largely driven by the samples collected in the Faroe Islands and Scotland, displaying just one and two haplotypes, respectively, and therefore strongly deviating haplotype frequencies. Neither of those two haplotypes were observed in the historical samples, and both were resistant to pyrethroids according to the marker causatively linked with resistance.
An earlier study based on analysis of 6000 SNPs throughout the genome concluded that lice were characterized by panmixia in the North Atlantic (Besnier et al., 2014 (Kaplan et al., 1989;Smith & Haigh, 2007). We therefore suggest that any observation of population genetic structure in this species, at least in regions of the genome associated with pesticide resistance, is likely to reflect the mosaic of chemical usage in aquaculture in time and space, as opposed to stable population genetic structure created through, for example, limited gene flow.

| Multiple origins of deltamethrin resistance in the North Atlantic
In the mitochondrial genes ATPase 6 and Cytochrome B, two haplotypes undetected in the historical samples (haplotypes 165 and 174) dominated in the contemporary samples, suggesting a strong selection for them. Only haplotype 165 was associated with the C14065T marker (section 1.3 in Data S1; Figure S8c). The synonymous C14065T SNP is regarded as a non-causative mutation (Nilsen & Espedal, 2015).  , where several haplotypes were detected having the causative SNP for organophosphate resistance. By contrast, one main haplotype was found to be under positive selection associated with emamectin benzoate resistance (Besnier et al., 2014) suggesting a primarily single origin of this resistance in the Atlantic.
Although findings of two distinct haplotypes support multiple origins of resistance, it is important to note that new haplotypes associated with resistance may theoretically result from mitochondrial recombination in heteroplasmid individuals without de novo emergence of resistance. Despite being very rare, recombination in mtDNA has been documented in invertebrates (Tsaousis et al., 2005) and may offer an alternative explanation for an origin of the resistant 165 and 174 haplotypes that does not require repetitive emergence of resistance per se. However, our haplotype networks based on the mtDNA genes separately (section 1.3 in Data S1; Figures S6 and S8) are similar to the results from the haplotype network based on the mtDNA genes combined, which suggests that mitochondrial recombination is not likely here.

| Dispersal of resistance
The combination of resistant mtDNA haplotypes being present in most of the contemporary samples (Figure 4), the lack of population genetic differentiation indicated by the microsatellite data (Table 4b), and the lack of population genetic differentiation indicated by mtDNA in the historical samples (Table 4a) all point toward to high connectivity among salmon lice in the North Atlantic. Previous studies have also reported a rapid spread of resistant haplotypes and genotypes across the North Atlantic (Besnier et al., 2014;Fjørtoft et al., 2017Fjørtoft et al., , 2021 which also bear testimony to a species displaying high genetic connectivity. It may be speculated that the absence of deltamethrin-resistant lice in the Canadian samples may result from either sampling limitations or a cost associated with resistance that eradicates resistant haplotypes when not under positive selection in that region, or a combination of these factors. In spite of anthropogenic selection evidently eliciting a pattern in the mitochondrial hereditome compartment, the observed arrangement will likely dissipate over time due to mixing and possible negative selection once deltamethrin use is discontinued. Efficient mixing is most likely facilitated by salmon lice attached to wild migrating salmon migrating large distances and meeting their conspecific relatives on common feeding grounds in the North Atlantic (Gilbey et al., 2021).
The resulting gene flow will prevent and/or erode the establishment of temporally consistent population structure (Besnier et al., 2014).

| Management implications and future investigation
The parasitic salmon louse, and its widespread resistance to delousing pesticides, provides a persistent challenge to commercial salmonid aquaculture, and represents the largest challenge to sustainable production in the Atlantic (Forseth et al., 2017). Therefore, a thorough understanding of the patterns of emergence and dispersal of resistance is imperative in order to develop informed pest management strategies for the future. Not least, this parasite has repeatedly demonstrated its ability to rapidly adapt to anthropogenic influence such as pesticide usage causing multiresistance (Fjørtoft et al., 2021), altered host parasite dynamics altering life history (Mennerat et al., 2012), non-chemical delousing treatments on farms potentially leading to changes in pigmentation (Hamre et al., 2021), and salinity and thermal tolerance (Ljungfeldt et al., 2017). In this regard, our data in association with earlier studies demonstrate two important principles: the remarkable adaptive capacity of this parasite and a high degree of connectivity and gene flow. Thus, it follows that there is a great need for amphi-Atlantic collaboration within management and aquaculture industry to control this parasite.

ACK N O WLE D G E M ENTS
This study was funded by the Norwegian Ministry of Industry and Fisheries.