Key role of mitochondrial mutation Leu107Ser (COX1) in deltamethrin resistance in salmon lice (Lepeophtheirus salmonis)

The pyrethroid deltamethrin (DTM) is used to treat Atlantic salmon (Salmo salar) against salmon louse (Lepeophtheirus salmonis) infestations. However, DTM resistance has evolved in L. salmonis and is currently common in the North Atlantic. This study aimed to re-assess the association between DTM resistance and mitochondrial (mtDNA) mutations demonstrated in previous reports. Among 218 L. salmonis collected in Scotland in 2018–2019, 89.4% showed DTM resistance in bioassays, while 93.6% expressed at least one of four mtDNA single nucleotide polymorphisms (SNPs) previously shown to be resistance associated. Genotyping at further 14 SNP loci allowed to define three resistance-associated mtDNA haplotypes, named 2, 3 and 4, occurring in 72.0%, 14.2% and 7.3% of samples, respectively. L. salmonis strains IoA-02 (haplotype 2) and IoA-10 (haplotype 3) both showed high levels (~ 100-fold) of DTM resistance, which was inherited maternally in crossing experiments. MtDNA haplotypes 2 and 3 differed in genotype for 17 of 18 studied SNPs, but shared one mutation that causes an amino acid change (Leu107Ser) in the cytochrome c oxidase subunit 1 (COX1) and was present in all DTM resistant while lacking in all susceptible parasites. We conclude that Leu107Ser (COX1) is a main genetic determinant of DTM resistance in L. salmonis.

Lepeophtheirus salmonis strains and husbandry. In addition to salmon lice sampled at field sites (see below), different L. salmonis strains maintained in laboratory culture were studied, with an overview of strains provided in Supplementary Table S1. Strain IoA-00 was derived from an isolate collected in the Firth of Clyde system in 2003 and is susceptible to DTM, emamectin benzoate and azamethiphos 23,29 . Strain IoA-01, which originates from Sutherland and was taken into culture in 2008, is resistant to emamectin benzoate but susceptible to DTM 21,29 . Strain IoA-02 was established in 2011 from material collected in the Shetland islands and is resistant to emamectin benzoate, DTM, and azamethiphos 21,26,29 . IoA-03 (Sutherland, 2014), NA01-O and NA01-P (Argyll and Bute, 2012) are further DTM resistant strains 21  www.nature.com/scientificreports/ genotypes of the above strains have been characterised in an earlier report 21 . While different definitions of DTM resistance in L. salmonis based on bioassays results have been proposed in the literature 30,31 , strains referred to as DTM resistant in this report are characterised by DTM median effective concentration (EC 50 ) values obtained in traditional bioassays (see below for definition) greater than 2 µg/L DTM, which is the level typically used in bath treatments 30 . This rather stringent definition of DTM resistance was chosen to select isolates for study which showed susceptibility phenotypes differing markedly from those found in L. salmonis prior to selection 14 , and is not meant to imply lower levels of DTM hyposensitivity are not clinically relevant. During experiments of this study, the well-characterised strains IoA-00 and IoA-02, which are still maintained in laboratory culture, were included to provide L. salmonis of known DTM susceptibility status. In addition, archived specimens of the other strains, which had been conserved in absolute ethanol followed by storage at − 20 °C, were used in genetic analyses. Finally, experiments described in this report included DTM-resistant strain IoA-10, which was established during the current study. Strains were cultured at the Marine Environmental Research Laboratory of the University of Stirling (Machrihanish, UK), with husbandry conditions described in detail elsewhere 32 .
Collection of L. salmonis from field sites. Preadult-II and adult L. salmonis were collected at farm site 1 (Argyll and Bute) during routine veterinary procedures in November 2018. Lice were put into polyethylene bags containing cool (~ 10 °C) oxygen-saturated seawater, which were placed in insulated boxes equipped with cold packs and immediately shipped to the University of Stirling, where recipients containing lice were transferred to a temperature-controlled incubator set to 12 °C and linked to airlines. The next day, the DTM resistance phenotype of lice was characterised using bioassays (see below), after which parasites were conserved in absolute ethanol for later genotyping. In addition, genetic analyses of this study included ethanol-conserved lice collected at further sites (2: Argyll and Bute, November 2018; 3: Argyll and Bute, May 2019; 4: Sutherland, September 2019), for which DTM bioassay data have been reported previously 25 . Further parasites studied included adult L. salmonis collected in 2010 from wild Atlantic salmon caught in the river Esk, located on the East coast of Scotland 21 . No bioassay data were available for these lice. To obtain L. salmonis for copepodid bioassays, gravid lice with paired egg strings were collected from farm sites 5 (Argyll and Bute), 6 and 7 (both Inverness-shire) in December 2019, followed by shipment to the University of Stirling and transfer of lice to a temperature-controlled incubator, as described above. To obtain batches of copepodids derived from the same dam, paired egg strings of individual females were removed and incubated in beakers containing seawater and linked to airlines in a temperature-controlled chamber set to 12 °C, allowing hatching of eggs and progression of larval development to the copepodid stage.
Establishment of DTM resistant L. salmonis strain IoA-10. Gravid female L. salmonis (N = 30) were obtained from an aquaculture site located on the West coast of Scotland (Inverness-shire) in November 2019, and transported to the Marine Environmental Laboratory of the University of Stirling in Machrihanish, with details as given above. Paired egg strings were removed from dams and incubated, as described above, to produce batches of copepodids derived from the same mother. The corresponding dams were preserved in absolute ethanol, and subjected to DNA extraction and genotyping assays (see below) to establish mtDNA haplotypes. Copepodids derived from one dam showing the haplotype of interest (haplotype 3, see "Results" section) were selected and used to infect naïve host fish to establish strain IoA-10 after the mtDNA haplotype of the larval batch had been confirmed by repeating genotyping assays with a subset of copepodids.
Crossing experiment. To obtain insights into the mode of inheritance of DTM resistance expressed by strain IoA-10, a crossing experiment between this strain and the DTM susceptible strain IoA-00 was performed in July 2020, spanning two generations termed parental (P0) and first filial generation (F1). In addition to crosses, the experiment included treatments to produce IoA-00, IoA-02 and IoA-10 lice synchronised to hybrids generated in the cross. Experimental infections for each of the above strains were performed with Atlantic salmon smolts (IoA-00 and IoA-10: 14 host fish each; IoA-02: 7 host fish). Lice were maintained on fish until development had reached the adult male and preadult-II female stages, but lice had not yet mated. At this time point, half of the fish carrying IoA-00 lice and half of the fish carrying IoA-10 lice were removed from tanks to harvest lice required for the setup of batch crosses. The remaining fish in IoA-00 and IoA-10 tanks, as well as all fish carrying IoA-02 lice, were maintained further, allowing lice of the respective strains to mate and produce eggs. In the first batch cross the P0 generation consisted of IoA-00 males and IoA-10 females, whereas the second batch cross was set up with the inverse sex-strain orientation. Per cross, ten Atlantic salmon smolts were anaesthetised using 2-phenoxyethanol (100 mg/L; 99%; Sigma-Aldrich, Gillingham, UK) and each fish received two male and three female parasites. The two crosses, as well as the three tanks with strains IoA-00, IoA-02 and IoA-10, were maintained until dams produced egg strings, at which point they were removed, and egg strings incubated to produce copepodids. The resulting F1 larvae were used to inoculate tanks containing naïve Atlantic salmon. Infections were maintained until F1 parasites reached the adult male and preadult-II female stages. Subsequently, F1 parasites were subjected to traditional DTM bioassays to determine their susceptibility phenotype (see below).

Lepeophtheirus salmonis bioassays. Traditional bioassays involved several DTM concentrations and
had the purpose to characterise the DTM susceptibility of a parasite population or strain by estimating the compound's median effective concentration (EC 50 ). Single-dose bioassays used one diagnostic DTM concentration and served to characterise the susceptibility phenotype of individual lice. Bioassays with host-associated L. salmonis, i.e. adult stage males and preadult II stage or non-gravid adult stage females, had either a traditional or a single-dose design 30,33 , while all bioassays with copepodid larvae had a traditional design. In addition, exposure www.nature.com/scientificreports/ experiments to assess effects of DTM on whole body ATP levels were carried out. ATP experiments followed a design identical to that of single-dose bioassays, except that these trials were restricted to male parasites, used different exposure times to capture early physiological effects (see below), and included in addition to DTM one concentration of the acaricide fenpyroximate that is known to act by interference with the mitochondrial complex I 34 and has been to shown deplete ATP levels in L. salmonis 21 . DTM and fenpyroximate (Pestanal analytical standard grade, Sigma-Aldrich, Gillingham, UK) were initially dissolved in acetone, with further dilutions carried out in a way to reach a final solvent concentration of 0.05% (v/v). All bioassay incubations took place in a temperature-controlled chamber set to 12 °C. Bioassays were considered invalid if the number of affected animals in solvent controls exceeded 10%. After completion of bioassays, test animals were stored in absolute ethanol at − 20 °C pending DNA extraction and genetic analyses.
In bioassays with host-associated L. salmonis stages, animals were introduced to 300 mL crystallising dishes containing 100 mL of filtered (55 µm) seawater equilibrated to 12 °C. After allocation of test animals to vessels, dishes were randomly assigned to treatments. Traditional bioassays included a geometrical series of at least six DTM concentrations in the range of 0.125 and 32 µg/L and a solvent control (Supplementary Table S2) in duplicates, with each dish receiving about five females and five males. In single-dose bioassays, lice were exposed to 2 µg/L DTM, which is the concentration recommended for treatments 30 . Each dish contained about ten females and ten males. Both types of bioassays further included solvent controls and involved exposure of lice for 30 min followed by recovery in clean seawater for 24 h prior to behavioural responses being examined and lice being rated as "live", "weak", "moribund", or "dead", with these criteria having been defined previously 29 . Animals rated as live or weak were considered unaffected, while those rated moribund or dead were considered affected by treatments. Lice were considered DTM resistant if remaining unaffected after exposure to 2 µg/L DTM and recovery, and susceptible to DTM if found affected.
Lepeophtheirus salmonis copepodid bioassays were carried out using batches of larvae generated by incubation of paired egg strings derived from the same dam (see above). After the number of copepodids per unit volume of seawater had been ascertained by counting, approximately ten copepodids were added to embryo dishes with glass lids by dispensing an appropriate volume of the batch, followed by topping up to a total volume of 2 mL with filtered seawater. Chemical exposures were initiated by adding 1 µL of a 2000 × final concentration solution of DTM to dishes containing 2 mL seawater and parasites. Copepodids were exposed to a geometric series of at least four concentrations of DTM in the range of 0.10 to 100 µg/L and one solvent control, in duplicates. Exposures in larval bioassays were for 24 h, followed by examination and rating of the animals. Copepodids were rated "live" when attracted by light and swimming normally, "weak" when swimming irregularly (animals swim in a straight line within 2 min after stimulation using light and a fine brush), "moribund" when incapable of swimming away after stimulation by light and a fine brush (animals may twitch appendages), and "dead" when showing no movements in extremities, gut, or other organs as apparent from examination under a microscope. Copepodids rated live or weak were considered unaffected, while moribund and dead parasites were considered affected.
In order to assess drug effects on whole body ATP levels, an exposure experiment was designed, with exposure levels and periods being based on similar experiments reported earlier 21 . Male lice were exposed to DTM for 30 min followed by recovery in clean seawater for 300 min. Exposure to fenpyroximate was for 300 min, without seawater recovery. After drug exposure, and where applicable seawater recovery, behavioural responses of animals was rated according to categories given above, and animals deemed "live", "weak" or "moribund" were removed for ATP analyses.
Effects of DTM on L. salmonis ATP levels. Directly after experimental exposure and rating of behavioural responses, individual lice were added to plastic tubes containing 1 mL of Tris-EDTA-saturated phenol (10 mM Tris HCl and 1 mM EDTA; Sigma-Aldrich, Gillingham, UK; Thistle Scientific Ltd, Glasgow, UK). Samples were incubated at room temperature for 10 min and subsequently stored at − 70 °C pending ATP analyses. Whole-body ATP levels were measured using a commercially available luciferin-luciferase bioluminescence assay kit (A-22066, Molecular Probes, Thermo Fisher Scientific, Bishop's Stortford, UK), with details having been reported elsewhere 21 .
Extraction of genomic DNA. Genomic DNA (gDNA) was extracted from individual preadult-II or adult stage L. salmonis using a salt extraction method described in detail elsewhere 21 , producing high quality gDNA suitable for use in PCR reactions to amplify the mtDNA genome or genotyping assays. Alternatively, for samples only used in genotyping, DNA was extracted by a fast high throughput protocol 35 , with details having been reported elsewhere 25 . DNA from copepodids was extracted by the latter method and involved transfer of one specimen into a 0.2 mL tube containing 25 µL alkaline lysis buffer (25 mM NaOH, 0.2 mM EDTA, pH 12.0), followed by heating to 95 °C for 30 min and subsequently cooling to 4 °C for 5 min using a polymerase chain reaction (PCR) thermocycler. Then, 25 µL 40 mM Tris-HCL (pH 5.0) was added, and the sample vortexed briefly before being centrifuged at 4000×g for 1 min. DNA extracts were stored at − 20 °C pending use in genotyping analyses.

Mitochondrial haplotype network. A haplotype network was interfered from mitochondrial haplotypes
of L. salmonis. The network was constructed using the medium-joining method 36 implemented in the software PopArt v1.7 37 . Haplotypes were defined based on the combined occurrence of the 18 SNPs described above in preadult-II and adult L. salmonis collected at field sites for bioassay (see above).
Mitochondrial genome (mtDNA) amplification and sequencing. The mitochondrial genome of two individuals of L. salmonis strain IoA-10 was amplified by PCR using gDNA as the template and sequenced. Amplification involved generating six overlapping amplicons using specific oligonucleotide primers designed with Primer3 v4.1.0 (Supplementary Table S4). PCR reactions were performed using 50 ng template DNA, 2.5 μL (10 pmol) each of forward and reverse primers, 25 μL Q5 High-Fidelity 2× Master Mix (New England BioLabs Ltd, Hitchin, UK), and 19 μL nuclease-free water. PCR conditions for each product are listed in Supplementary Table S5. All PCR products were purified (NucleoSpin Gel and PCR Clean-up Kit, Macherey-Nagel, Düren, Germany) and sequenced (Supplementary Table S6) by Eurofins Genomics (Ebersberg, Germany). Prior to sequencing, PCR product 6 (Supplementary Table S4) was subcloned using the pGEM-T Easy Vector system (Promega, WI, USA), followed by plasmid isolation using the NucleoSpin Plasmid EasyPure kit (Macherey-Nagel, Düren, Germany). MtDNA sequences obtained for the same individual were manually trimmed, aligned and assembled.  To compare DTM-associated SNPs reported in previous studies using different reference mtDNA genomes 21,22 , nucleotide sequences of the respective assemblies were aligned using Needle 21,41 . An earlier study reported nucleotide sequences for four mtDNA genes in L salmonis salmonis collected in Norway, Scotland, Canada and Russia in 2000 and 2002 28 . Sequences from the study (NCBI accession numbers: Cyt b, AY602223-AY602402; A6, AY602407-AY602586; COX1, AY602587-AY602766 and 16S rRNA, AY602770-AY602949) were aligned to a mtDNA assembly 21 to derive SNP genotypes at mtDNA loci studied in this report (see above).

Identification of DTM
Data analyses and statistical tests. All statistical tests were carried out using the software R (v4.0.2), with the significance level set at P < 0.05. Dose-response relationships in L. salmonis bioassays were assessed by probit analysis using the packages drc (v3.0-1) and MASS (v7.3-55), assuming a log-normal distribution of drug susceptibility. Based on the fitted models, EC 50 and 95% confidence limits were derived and effects of sex and origin/strain on drug susceptibility assessed. Fisher's exact test of independence was used to examine whether DTM susceptibility phenotypes observed in single dose bioassays differed between mitochondrial haplotypes, employing the R package rcompanion (v2.4-1). Analyses of whole-body ATP data were conducted using the R packages car (v3.0-11) and PMCMR (v4.3). Normality and homogeneity of variances of data were assessed by Shapiro-Wilk's and Levene's tests, respectively. As the data violated these assumptions, the Kruskal-Wallis test was used to assess the effect of drug treatments on ATP levels, followed by post-hoc comparisons of treatments to the appropriate control group with Dunn's test. The experiment-wise type I error was controlled by sequential Bonferroni correction 42 .
Relationship of mtDNA haplotypes to DTM resistance of copepodids. The relationship between mtDNA haplotypes and DTM resistance was further investigated in 24 h behavioural bioassays with copepodid larvae derived from gravid females collected at field sites or obtained from laboratory strain cultures. The estimated DTM EC 50 of copepodids was 0.14 µg/L for haplotype 1 (strain IoA-00), 4.81 µg/L (strain IoA-02) or 3.98 µg/L (field derived larvae) for haplotype 2, and 6.90 µg/L for haplotype 3, field derived larvae (  1D-F, and 1H), but phylogenetically distant to resistance-associated haplotypes 2 and 4 ( Fig. 1). Compared  Table 2. Association of mitochondrial haplotypes with deltamethrin resistance. L. salmonis of laboratory strains IoA-00 and IoA-02 and parasites obtained from Scottish aquaculture production sites were subjected to deltamethrin bioassays to establish susceptibility status, followed by genotyping at four SNP loci to establish haplotypes (see Table 1 for details). a Susceptibility to deltamethrin was determined in single-dose bioassays, involving exposure (30 min) to 2 µg/L deltamethrin, followed by recovery in seawater (24 h) and subsequent rating as susceptible (affected) or resistant (unaffected). b 21 out of a total of 37 deltamethrin resistant individuals observed were successfully genotyped. www.nature.com/scientificreports/ to susceptibility-associated haplotype 1A expressed in strain IoA-00, the resistance-associated haplotype 3 differed in genotype at only one of 18 tested SNP loci, T8600C in COX1 (Supplementary Table S8).  Table S10). Comparison of the mtDNA genome of IoA-10 strain lice to that of DTM resistant and susceptible L. salmonis strains (IoA-00, IoA-01) revealed a limited number of mtDNA mutations that were present in all DTM resistant lice while lacking in all analysed susceptible isolates. These mutations comprised Table 3. Deltamethrin susceptibility and mitochondrial haplotype of L. salmonis copepodid larvae. Each bioassay was performed with the copepodid descendants of one female salmon louse. L. salmonis originated from the laboratory strains IoA-00 and IoA-02, or Scottish aquaculture production sites. Bioassays involved exposure (24 h) to deltamethrin and rating of copepodids as normal or affected. Haplotypes were established based on the genotyping of 10 individuals per bioassay. a For definition of haplotypes, see Table 1. b Raw data used to derive EC 50 values are provided in Supplementary Data S1.  Figure 1. Median-joining haplotype network interfered from mitochondrial haplotypes of L. salmonis.

Lepeophtheirus salmonis
Haplotypes were defined based on the combined occurrence of 18 mitochondrial SNPs, which have been associated with deltamethrin resistance in L. salmonis in a previous study 21 (Supplementary Tables S8 and  S9). Haplotypes that were identified in salmon lice from farm sites are represented by circles, while haplotypes that were identified in salmon lice from wild hosts are represented by grey rhombi. The size of each circle is proportional to the frequency of each haplotype in salmon lice from farm sites. Deltamethrin resistance associated haplotypes are represented by red circles and susceptibility-associated haplotypes are shown in blue. www.nature.com/scientificreports/ the non-synonymous SNP T8600C, corresponding to Leu107Ser in COX1, and eight polymorphisms in regions that are non-protein coding (Table 4).
Exposure of lice to fenpyroximate (100 µg/L), an acaricide known to block oxidative phosphorylation, caused toxicity, and decreased whole-body ATP levels in all strains assessed (Fig. 3, Supplementary Tables S5 and S12). Exposure to 2 µg/L DTM caused toxic effects accompanied by significantly decrease of ATP levels in IoA-00 lice (P = 0.008) but failed at causing toxicity or significant effects on the ATP levels in IoA-02 and IoA-10 lice.
Protein sequence comparisons of ND1, ND5, COX1, and COX3 among crustacean species. Amino acid sequences of ND1, ND5, COX1, and COX3 from L. salmonis and different crustacean species for which mtDNA sequences were available were aligned to reveal conserved residues within these mitochondrial polypeptides. Among four non-synonymous mtDNA SNP loci that have previously been reported from a DTM resistant strain 21 , G3338A (COX3 Gly33Glu), T5889C (ND5 Leu411Ser) and G8134A (ND1 Gly251Ser) cause residue changes at non-conserved positions, while T8600C (COX1 Leu107Ser) alters the amino acid sequence at a position that is conserved among all species assessed (Supplementary Fig. S1). Table 4. Mitochondrial sequence variations specific for deltamethrin resistant L. salmonis isolates. The mitochondrial genome (mtDNA) was compared between deltamethrin resistant (IoA-02, IoA-03, IoA-10, NA01-O, NA01-P) and susceptible (IoA-00, IoA-01) L. salmonis strains to establish mutations associated with resistance. The mtDNA sequence of strain IoA-10 was established by PCR followed by Sanger sequencing within the present study, while that of the other strains has been reported previously 21

Discussion
Previous studies have shown that DTM resistance in L. salmonis is transmitted predominantly by maternal inheritance and associated with mtDNA SNPs 20,21 . While an earlier study suggested that highly DTM resistant L. salmonis share virtually identical mtDNA sequences 21 , the present study identified three mtDNA haplotypes associated with resistance, one of which coincides with the previously reported resistance associated mtDNA sequence. L. salmonis strain IoA-10, which possesses one of the novel DTM resistance associated haplotypes, showed a high level of DTM resistance comparable to that of the previously characterised strain IoA-02 and, similar to IoA-02, passed on DTM resistance to the next generation through maternal inheritance. The phylogenetic analysis of mtDNA haplotypes provided evidence for multiple origins of mtDNA-associated DTM resistance in L. salmonis. Comparison of mtDNA sequences between DTM resistant and susceptible salmon louse strains suggested the association of DTM resistance with SNP T8600C, corresponding to Leu107Ser in COX1.
Lepeophtheirus salmonis obtained at aquaculture sites were subjected to bioassays to determine their DTM susceptibility status. Parasites were then genotyped at previously described mtDNA SNP loci 21 to identify mitochondrial haplotypes and assess haplotype association with DTM resistance. 93%, 87%, and 100% of lice possessing haplotypes 2, 3, and 4, respectively, were classified as DTM resistant, emphasising the association of mtDNA mutations with the resistance phenotype. The remaining 7-13% may have died due to interacting environmental factors or handling of parasites during sampling, transportation, and set-up of bioassays rather than drug toxicity 13,43 . As bioassays were performed on lice directly obtained from farmed salmon, environmental conditions prior sampling could not be controlled. Lice may have also been exposed to stressful conditions during sampling and due to abrupt changes of environmental conditions during transportation. The DTM resistance phenotype was less distinct for lice containing haplotype 1. While most lice with this haplotype were classified as DTM susceptible, 43% were rated resistant. The large number of resistant haplotype 1 individuals suggests the contribution of genetic factors other than mitochondrial mutations to the DTM resistance phenotype, which is in line with findings by Carmona-Antoñanzas et al. 21 who attributed the presence of about 20% resistant F2 parasites in a family descending from a DTM susceptible dam and a resistant sire to nuclear genetic determinants of resistance.
To further assess the relationship between mtDNA haplotypes and DTM resistance, bioassay experiments were carried out with batches of copepodid larvae derived from individual females. Compared to bioassays with field collected preadult-II and adult lice, advantages of F1 bioassays with copepodid larvae include the fact that sibling clutches share the same mitochondrial haplotype 44 and were hatched and reared under standardized laboratory conditions less likely to bias the outcome of the experiment. While insufficient gravid females of appropriate genotype precluded analysis of haplotype 4 in copepodid bioassays, testing of larvae of the remaining haplotypes resolved significant differences in DTM susceptibility between L. salmonis of haplotypes 2 and 3 as compared to parasites of haplotype 1, confirming results from conventional bioassays and underpinning the association of mtDNA mutations with DTM resistance. Moreover, the study showed that for DTM resistant Figure 3. Effect of deltamethrin and fenpyroximate on ATP levels in L. salmonis. Male adult salmon lice of the drug susceptible strain IoA-00, the multi-resistant strain IoA-02, and strain IoA-10 were exposed to deltamethrin (2 µg/L), fenpyroximate (100 µg/L), or a solvent acetone control (0.05% v/v, control) for 300 min before behavioural effects were recorded (secondary y-axis; orange data points) and alive animals were sampled for whole-body ATP analysis (N = 15 per group). ATP concentrations in drug treated lice are expressed relative to those of the control group (relative units, RU ± SE). Stars indicate significant differences to the control group (Kruskal-Wallis test followed by post hoc comparisons using Dunn's test; * P < 0.05, ** P < 0.01, *** P < 0.001). www.nature.com/scientificreports/ lice, hyposensitivity to the drug is already apparent at the larval stage, which would be expected for resistance conferred by mtDNA mutations 21 .
Genotyping of field isolates collected in 2018 and 2019 revealed three DTM resistance-associated mtDNA haplotypes, with haplotypes 2, 3, and 4 being found in 75%, 14% and 8% of all resistant lice, respectively. While an earlier study from our laboratory 21 found that all Scottish DTM resistant isolates shared mtDNA sequences consistent with haplotype 2, re-analysis of a larger number of archived samples from these isolates during this study provides evidence for haplotypes 2, 3 and 4 being present in DTM resistant strain NA01-O, which was established in 2013, while haplotype 4 was expressed by some specimens among a sample of lice removed from wild salmon in 2010. In contrast, re-analysis of sequence variation data for four mitochondrial genes in L. salmonis populations from Norway, Scotland, Canada and Russia sampled in 2000-2002 28 failed to detect in any of the samples the DTM resistance-associated mutation Leu107Ser in COX1, which is shared among the three DTM resistance associated mtDNA haplotypes identified in this report. In accordance with these findings, another study, which scanned L. salmonis populations from sites across the North Atlantic for one DTM resistance associated mtDNA SNP 45 , found that the resistance-associated genotype first appeared in 2009 in populations from Shetland, Norway and Ireland, but was absent in samples from 2002 and before. Phylogenetic analyses of mtDNA haplotypes identified in this study showed that the three DTM resistance-associated haplotypes fall into two clusters, with haplotype 3 being phylogenetically distant to haplotypes 2 and 4, which shared 17 out of 18 tested SNP loci. This suggests that mtDNA-associated DTM resistance in L. salmonis evolved at least twice independently followed by spread of resistant lice as a consequence of the extensive use of pyrethroids. However, more research is required to resolve at greater detail the roles of different DTM resistance associated mtDNA haplotypes in the emergence and spread of L. salmonis pyrethroid resistance across the North Atlantic.
Resistance-associated haplotypes 2 and 3 differed maximally in sequence, with only one out of 18 tested SNP loci being shared. However, despite haplotypes 2 and 3 being very different in sequence, the resistance phenotype conferred is very similar. Both IoA-10 (haplotype 3) and IoA-02 (haplotype 2) lice were highly DTM resistant (EC 50 values > 24.0 µg/L; P = 0.845) and reciprocal crosses of strains IoA-10 (present study) and IoA-02 21 with the drug-susceptible IoA-00 lice revealed that both strains transmit their resistance to the next generation through maternal inheritance. Moreover, DTM exposure caused behavioural toxicity and whole-body ATP depletion in DTM susceptible IoA-00 parasites, but not resistant IoA-10 and IoA-02 lice. These findings are in line with an earlier experiment, which compared the effect of DTM exposure on behavioural toxicity and ATP levels between IoA-00 and IoA-02 lice 21 . Depletion of ATP levels in DTM susceptible lice may be related to the toxic effect of DTM on the mitochondria, and mtDNA mutations in haplotypes 2 and 3 may have a protective effect.
As the inheritance of mtDNA is linear and lacks recombination through meiosis, relevant SNPs for DTM resistance are transmitted together with irrelevant hitchhiking SNPs 44 . Thus, SNPs that are truly linked to DTM resistance are expected to be present in all resistance-associated haplotypes but lacking in all susceptibilityassociated haplotypes. The non-synonymous mtDNA SNP T8600C, corresponding to Leu107Ser in COX1, was the only mutation shared by the resistance-associated haplotypes 2, 3, and 4 and lacking in all susceptibilityassociated haplotypes. When comparing mtDNA sequences of DTM resistant and susceptible lice, T8600C was also the only non-synonymous mutation differentiating between resistant and susceptible individuals. Sequencing analyses further revealed eight additional SNPs in non-coding regions (NCR) of the mtDNA that were common to all resistant lice and lacking in all susceptible lice. NCR sequences are the most variable mtDNA sequences, which may explain the high number of SNPs found within this region in the present study 46 . Seven of these SNPs were found in the mitochondrial control region, also known as displacement loop (D-loop). Its function is not yet fully understood but seems to be critical in regulating replication and transcription of mtDNA 47 . However, D-loop mutations are not known to confer drug resistance. Another SNP, A10178G, was found within a mitochondrial ribosomal RNA (rRNA) gene and has also been described by Bakke et al. 22 . Mitochondrial rRNAs are assembled with ribosomal proteins encoded by nuclear genes to form mitochondrial ribosomes, which are responsible for translating mitochondrial proteins 48 . Thus, mutation within the mitochondrial rRNA may lead to ribosome dysfunction and may result in respiratory chain defects 49 . However, to our knowledge, there are no reports of mitochondrial rRNA mutations associated with drug resistance.
Findings of the present study raise questions about the mechanism of DTM resistance and by inference the mechanism of DTM toxicity in L. salmonis. While it is generally accepted that pyrethroids target Na v in terrestrial arthopods 16 , several studies with terrestrial arthropods and mammals provide evidence for pyrethroid effects on mitochondrial functions. Due to their lipophilic nature, pyrethroids can pass and interact with biological membranes, making mitochondrial membranes and membrane proteins candidate targets for toxic action 50 . For example, pyrethroids have been shown to affect mitochondrial membrane structures and dynamics, which can impair oxidative phosphorylation 51,52 . Mitochondrial oxidative phosphorylation can also be impaired by intracellular Ca 2+ accumulation 53,54 , which can result from interactions of pyrethroids with Na v and consequent Ca 2+ influx 55 and direct effects of pyrethroids on voltage-gated Ca 2+ channels 56 . Pyrethroid induced disruption of mitochondrial membrane integrity and inhibition of respiratory complexes, as well as intracellular Ca 2+ accumulation can cause the generation of reactive oxygen species (ROS) in mitochondria [57][58][59] . ROS can trigger a cascade of reactions that induces lipid peroxidation and damage of macromolecules [60][61][62] .
Pyrethroids have been shown to induce intrinsic mitochondrial apoptosis 63,64 , which involves mitochondrial outer membrane permeabilization, release of cytochrome C into the cytosol, activation of caspases, and ultimately DNA fragmentation 65 . In particular, this pathway can be triggered by pyrethroid induced oxidative stress and Ca 2+ accumulation, as well as low levels of ATP that lead to disruption of the mitochondrial transmembrane potential [66][67][68][69] . Interestingly, DTM exposure increased apoptosis in mitochondria-rich skeletal muscle, subcuticular tissue, and central ganglion cells in salmon lice of a drug susceptible strain, but not or to a lesser degree in a DTM resistant strain 22  www.nature.com/scientificreports/ Taken together, in salmon lice, DTM may induce toxicity through disruption of mitochondrial membranes, direct inhibition of mitochondrial respiratory complex(es), intracellular Ca 2+ accumulation, or by causing oxidative stress. An obvious explanation for DTM toxicity in salmon lice would be that DTM or its metabolites are binding to a mitochondrial respiratory complex and lead to disruption of the mitochondrial ATP production, which has been observed in the present study. In addition, inhibition of respiratory complexes can lead to the formation of ROS 57 . Both, low levels of ATP and oxidative stress can in turn induce intrinsic mitochondrial apoptosis 66,67,69 , which has been described by Bakke et al. 22 . Resistance may be conferred by mitochondrial SNP(s) that changes the amino-acid sequences of the complex and impair binding of DTM. For reasons explained above, T8600C leading to Leu107Ser in COX1 is the most probable mutation for conferring DTM resistance in salmon lice. Alternatively, DTM might impair the mitochondrial ATP production in susceptible lice by causing disruptions of the mitochondrial membrane or by secondary effects arising from DTM toxicity. In these scenarios, mtDNA mutation(s) may have functional effects on the efficiency of electron transfer or proton translocation, counteracting ATP deficits.
While the association of DTM resistance with mtDNA mutations is well documented in L. salmonis [20][21][22] , implying mitochondrial targets for DTM toxicity, evidence for roles of mitochondrial genes in pyrethroid resistance is lacking in other arthropods. If point mutations of mtDNA are sufficient to confer DTM resistance in L. salmonis, a possibility suggested by this and previous studies, this would further imply that Na v is of secondary importance in the mode of action of DTM in L. salmonis. Interestingly, DTM toxicity in L. salmonis showed a delayed onset in earlier studies. In bioassays, no effects of DTM on lice of a susceptible strain were apparent directly after 30 min of exposure to 0.5 µg/L DTM and three hours of recovery in clean seawater were required for effects to fully develop 21 , while in experimental treatments of fish carrying susceptible strain lice, the median survival time of parasites after a 30 min bath treatment with 2.0 µg/L DTM was 16.8 hours 70 . While the delayed onset of DTM toxic effects observed in these studies is unexpected for a drug assumed to cause disruption of neurotransmission, more research is required to characterise the molecular mode of action of DTM in L. salmonis.

Conclusion
DTM resistance in L. salmonis is associated with multiple mtDNA haplotypes which have multiple origins. Non-synonymous mtDNA mutation T8600C, corresponding to Leu107Ser in COX1, was common to all DTM resistant mtDNA haplotypes but lacking in haplotypes not associated with resistance. Parasites possessing a mtDNA haplotype in which T8600C was the only non-synonymous mutation are highly DTM resistant and pass on their resistance to the next generation through maternal inheritance. The results suggest the association of DTM resistance with SNP T8600C (Leu107Ser in COX1).

Data availability
The sequencing data were submitted to the EBI ENA database under the Project number PRJEB47839.