Where art thou louse? A snapshot of attachment location preferences in salmon lice on Atlantic salmon hosts in sea cages.

Problematic sea lice infestations on farmed Atlantic salmon (Salmo salar) have motivated extensive research and development into new methods to prevent, monitor and control sea lice. Most of these technologies require detailed information on the behaviour, spatial distribution and demography of lice on host fish. This study investigated how salmon lice (Lepeophtheirus salmonis) infestation density varies across the host's surface under sea cage farming conditions. Lice abundance, demography and attachment location were tracked over time, with repeated sampling of 300 individually tagged salmon across three replicate experimental sea cages. The data reveal clear differences in attachment locations according to sex and stage, but with an overall preference for the dorsal surface among mobile stages-dorsal head for adult females and dorsal-posterior section for males and pre-adults. Total lice abundance was highly variable between repeated measures of individual fish, consistent with frequent host-switching or mortality. Total lice numbers also declined between sampling dates, likely due to handling, with lost mobile lice being almost exclusively adult males. As the distribution of sea lice on hosts is likely determined by numerous factors, future image-based automated detection systems should be validated in settings that reflect the complex host-parasite interactions that occur in open farming systems.

behavioural studies, and some emerging technologies will also benefit from this knowledge (e.g. image-based automated lice detection).
There is an increasing focus on more efficient and accurate methods to enumerate lice on salmon in sea cages without the need to physically capture and count individuals. The drive for a solution is so great that multiple competing companies, in parallel, are developing camera systems for in situ automated counts (Grøntvedt, 2019). Although these innovations require specific technological advancements, there is also a need for biological data to validate any system's accuracy. Specifically, developers and users of the system must understand size range of stages and how these change with temperature (i.e. Hamre, Bui, Oppedal, Skern-Mauritzen, & Dalvin, 2019), patterns of their distribution on the host (Todd et al., 2000) and expected number of lice on the unimaged areas of the host.
This study aimed to describe the distribution of lice stages and attachment locations on a cohort of salmon hosts held in sea cages, with individuals sampled twice, 15 days apart. These fish were not deloused for ~4 months, which allowed a "natural" development and between-host distribution of lice, undisturbed by treatment efforts.

| ME THODS
A cohort of ~7,500 salmon was held in a single sea cage (12 m × 12 m × 15 m) at the Austevoll Research Station (Institute of Marine Research, Norway). After sea transfer four months prior to trial start, salmon were held under standard husbandry protocols where they acquired lice infestations naturally (Figure 1). Five weeks prior to trial start, nine groups of 400 salmon were transferred to smaller cages (5 m × 5 m × 6 m) for a separate experiment where fish were not manipulated, but held with low quantities (7.5% stocking density) of cleaner fish (ballan wrasse, Labrus bergylta) for 3 weeks.
After this period, subsets of these fish were transferred to new cages (5 m × 5 m × 6 m) and held for 2 weeks without cleaner fish present; this fish group formed the stock population for the current study. One day prior to trial start, three cages were stocked with 500 salmon weighing 271 ± 16.3 g (±SE).
Of the 500 salmon per cage, 100 were tagged for individual identification. Before transfer into the cage, fish were anaesthetized with tricaine methanesulfonate (Finquel, 0.1 g/L). Within 4 min of complete sedation, fish were tagged with a unique coloured anchor T-bar tag (5 cm), photographed, measured for length, and assessed for lice attachment locations. Each fish had a diagrammatic representation of the exact location of lice on the body (Figure 2), marked with the stage of each louse (for detailed biological description of L. salmonis, see Costello, 2006, Hull, Pike, Mordue Luntz, & Rae, 1998, Johnson & Albright, 1991b. Precopulating pairs of adult males and pre-adult II females were common (Ritchie, Mordue, Pike, & Rae, 1996) and labelled when observed. Only chalimus II and mobile lice stages were recorded (pre-adults not sex-distinguished), as chalimus II could moult into the mobile stage within this time frame (Hamre et al., 2019). After processing, fish were placed in a recovery tank and monitored for full recovery, after which they were transferred to the cage. Any lice that dislodged in the recovery tank were discarded as they could not be allocated to their original host. This time point, when louse distributions were first recorded, is termed Sample 1.
The groups were left in the cages undisturbed for 15 (cage 1) or 16 days (cage 2 and 3). After this period, tagged fish were hand-netted and anaesthetized in individual buckets so that any lice that were displaced could be assigned to their original host. Individuals were identified and photographed, and lice attachment locations were recorded in the same format. This time point, the second assessment of louse distribution, is termed Sample 2.
For the duration of their time at sea, and throughout this experimental period, the salmon were not deloused, as the study had approval for lice abundances exceeding the standard regulative threshold (Mattilsynet reference 2018/185649). An environmental profile was taken daily at a reference location on the sea facilities using a profiling CTD probe (Model SD204, SAIV AS, Norway).

F I G U R E 1
Timeline outlining the handling of subjects prior to and during the present study, across months and weeks Sea temperature between 0 and 5m was on average 14.3°C (range: 11-16.8°C) and salinity on average 31.2 ppt (range: 29.7-32.5 ppt) during the study period. Between 5 m and 10 m, temperature was on average 3.3°C cooler than the top layer (range: 8.4-14.4°C) and salinity 32.2 ppt (31.2-32.9 ppt).

| Data handling and statistical analyses
The abundance of lice stages and their attachment location preference on the host were examined through a series of directed comparisons, either across both time points (mean abundance, sex ratio and attachment location) or between times (repeatability of lice abundance within a host).
To assess repeatability of lice infestation density of a given fish between Samples 1 and 2, lice counts at Sample 2 were regressed against lice counts at Sample 1 using the lm() function in R. Fish were treated as replicates. A cage ID factor (3 levels) was also tested, but did not explain a significant amount of variation (p = .3) and was therefore omitted from the final model. Sex ratios were tested for deviation from 50:50 using a Wilcoxon signed-rank test (wilcox.test function in R). The null hypothesis was 50%, with a 2-sided test.
Dorsal-ventral and left-right preferences in attachment locations were tested using a chi-square test of proportions (H 0 = equal numbers). To test whether dorsal-ventral and leftright preference varied according to lice stage, we fitted binomial generalized linear mixed models (GLMMs) to binary response data describing attachment location (2 models: dorsal versus. ventral; left versus. right side). Lice were treated as replicates. Both models were fitted using the glmmTMB package for R (Brooks et al., 2017), and included fixed terms for lice stage (5 levels: L. salmonis stages chalimus II, pre-adult, adult male, adult female, or Caligus elongatus), sample (two levels), cage ID (three levels), and a random intercept term for fish ID. Type II analysis of deviance tables was computed using the ANOVA() function in the car package for R (Fox & Weisberg, 2011). Lice found on the head were omitted from the dorsal-ventral model (remaining: 2,431 lice on 330 fish), while lice found on the dorsal or ventral midline were omitted from the left-right model (remaining: 2,627 lice on 331 fish).

| Abundance
Overall, fish had a mean infestation density of 5.9 lice per fish at Sample 1 and 5.0 lice per fish at Sample 2. Abundance of male lice halved between Samples 1 and 2, from 2.0 to 0.9 lice per fish.
Abundance of adult female lice (including those with or without egg strings) was similar between Samples 1 and 2 (1.1 lice per fish at both samples), suggesting that any adult females lost between samples were replaced by maturation of pre-adult females. Of all of the lice that were recorded at both samples, 17% of pre-adults, 9% of adult males and 5% of adult females were dislocated from the host and found in the sedation vessel.

| Sex ratios
Of the 319 fish that had adult lice attached at Sample 1, 70% had adult females present. 34% of adult lice were female (significant deviation from 50:50 sex ratio: Wilcoxon V = 6,187, p < .0001).
At Sample 2, only 236 fish had adult lice attached. 75% of fish had adult female lice present, with females making up 52% of the adult lice population (no significant deviation from 50:50 sex ratio: V = 9,375, p = .6). Across Samples 1 and 2, 42% of adult lice were female.

| Repeatability of lice abundance within fish
A total of 265 fish were assessed at both the sample points; of these, the difference in total lice abundance between the two sample points varied between a loss of 11 lice and a gain of 26. Within fish, infestation density at Sample 1 predicted infestation density at Sample 2 significantly (t 331 = 2.3, p = .02) but weakly (adjusted R 2 = .01; Figure 3). In other words, the number of lice on a fish at

| Attachment location
Lice abundance varied considerably across quadrants, with lice being most abundant on the left side of the head and the posterior dorsal midline ( Figure 5). Adult females, with or without egg strings, showed an especially strong preference for the head (53% were on the head compared to 26% of pre-adults and adult males: Table 1).
Among lice that were not attached to the host's head, the dorsal surface was preferred despite a similar available surface area on the ventral half (Table 1). This pattern held for Samples 1 and 2, but was significantly influenced by louse stage ( Table 2): 57% of pre-adults and adult males were found in the dorsal half of the body compared with only 27% of adult female lice, corresponding to a stronger preference for the head among adult females (Table 1). There was a slight overall preference for the left side of the fish (41%) compared to the right (36%) or midline (24%) ( Table 3). Left-right preference did not depend on louse stage and did not vary significantly between Samples 1 and 2 (Table 2).

| Repeatability of lice abundance
Total louse abundance on a given host was not similar between the two sample points, indicating a large effect of either host transfer, intrinsic mortality, handling, new infections or recruitment from nearby cages or farms. The repeatability of lice abundance, as recorded in this study, reflects the synergistic effect of multiple influences on infection levels and suggests that at least in a cage environment with multiple samplings, there is little reliability in using observed abundances in predicting levels for the near future. Although sources for change in lice abundance within hosts could not have been accurately determined here, we explore potential mechanisms that have been The reduction of lice numbers between sample points could be partially due to natural mortality; natural loss of mobile lice stages is not uncommon, even in tank environments which are relatively contained. In communal and single fish tanks, between 36% and 51% of mobile lice can be lost over 24 days (Hamre & Nilsen, 2011).
Interestingly, farmed strains of salmon lost 27%-32% of attached lice over 34 days in tanks; however, this was not the case with wild and landlocked Atlantic salmon in the same tanks (Bui, Dalvin, et al., 2018), suggesting a genetic effect on lice developmental success or movement between hosts. However, experience in unspecific loss of lice over development and between replicate tanks has been reported as quite high and very variable (Bjørn & Finstad, 1998;Hamre, Glover, & Nilsen, 2009) The reduced lice abundance in Sample 2 could be driven by fish handling during capture and counting, rather than natural attrition.

F I G U R E 5
Distribution of lice stages on the body of host salmon. Bars represent the percentage of lice of a given stage that occurred in the corresponding quadrant (percentages for each stage across all five quadrants sum to 100%). Stages shown are chalimus II (CH2), combined pre-adults 1 and 2 of both sexes (PA), adult males (AM) and adult females (AF). Data are pooled across Samples 1 and 2 and are irrespective of side (left, right, midline). Quadrants were defined as head (delineated at the opercula), dorsal-anterior section (including the dorsal fin), dorsal-posterior section (including the adipose fin and dorsal part of the caudal fin), ventral-anterior section (including the pectoral and pelvic fins) and ventral-posterior section (including the anal fin and ventral part of the caudal fin) [Colour figure can be viewed at wileyonlinelibrary.com] Notably, some individual salmon lost >10 mobile lice over the 15day trial-this is unlikely to be due to natural attrition (Figure 4). The choice of capture method when collecting subsamples of fish to assess infestation status has a large impact on observed lice abundance (Holst, Nilsen, Hodneland, & Nylund, 1993;Nagasawa, 1985;Tully, Gargan, Poole, & Whelan, 1999). In this study, fish were gently crowded in the cage and hand-netted, which is likely to result in increased rubbing and abrasion against the net before transfer into the sedation vessel. Previous work indicates that immobilizing fish using sedatives reduces lice loss during handling compared to mechanical euthanasia (Copley, O'Donohoe, McGrath, & Jackson, 2005;Jackson & Minchin, 1993). Depending on the sampling conditions and procedure, up to 68% of mobile lice can be displaced from the host during handling and therefore were found in the sedation or anaesthesia bucket (Table 4; Berntsen et al., 2018). This occurs with a variety of sedative compounds (personal observation S. Bui). In fact, Copley et al. (2005) reported that 71% of pre-adult male lice were found in the sedation vessel on one sampling occasion, with lice of all mobile stages also found displaced from the host. In comparison, the present study found only 5%-17% of mobile lice where dislodged, likely due to the use of individual sedation buckets where lice cannot experience mechanical disturbance from other salmon hosts.

| Sex ratios
Females were underrepresented among adult lice at Sample 1 (37% of adult lice). However, the decline in total abundance at Sample 2 was almost entirely due to loss of male lice, bringing the female prevalence up to 56% of adult lice. This follows previous reports of higher ratios of male to female lice on farmed hosts (Bron, Sommerville, Wootten, & Rae, 1993;Todd et al., 2000), whereas females have been found predominant (up to 72%) on wild hosts (Copley et al., 2005;Jacobsen & Gaard, 1997;Todd et al., 2000). The question of why adult male lice are more prevalent on farmed salmon has received little attention since those early reports; previous hypotheses attributed this observation to the faster developmental rate of male lice (Hamre et al., 2019), sex-biased impact of delousing therapeutants (specifically, to dichlorvos; Jaworski & Holm, 1992) or greater mobility of male versus female lice between hosts (e.g. Stephenson, 2012).
An alternative mechanism for biased sex ratios could be differential effects of host handling on males versus females (see below for further discussion of handling effects). Females are larger than males at all mobile stages (Johnson & Albright, 1991b) and therefore may be more robust to mechanical disturbances. If so, male lice would be at higher risk of detachment during fish handling. Two studies conducted at commercial sites (14-39 sample dates with similar sampling procedures for each) reported that 41%-76% of adult lice were found in the sedation vessel (Table 4). Adult males were 17% (Site 1) and 22% (Site 2) more likely to be found in the sedation vessel than females (Bui, Stien, Nilsson, Trengereid, & Oppedal, 2020;Geitung et al., 2019); in the present study, adult males were 91% more likely.
The likelihood of lice being detached during handling is likely to be influenced by the sedation or euthanasia protocol-the fish sampled at the above commercial sites were quickly killed, probably reducing mechanical stress on attached lice relative to the present study, where fish were sedated individually in smaller buckets.

| Preference of lice attachment locations on the host
The mobile nature of pre-adult and adult lice stages results in a distribution around the host body that varies according to environmental TA B L E 1 Dorsal-ventral distribution of lice stages for pre-and post-trial samples. p < .05 indicates significant deviation from 50:50 dorsal:ventral distribution TA B L E 2 Summary of fitted GLMMs. Tests are Wald's chi-square (type II) performed using the car package for R factors and the presence of conspecifics. This study is among the first to describe this behaviour in a sea cage rather than tank setting and therefore reflects responses to complex environmental conditions experienced in the field.
Mobile stages were most densely aggregated on the head and dorso-posterior sections of the host in this study, indicating that these areas are optimal for fitness under the prevailing conditions.
Interestingly, the distributions observed in this study are almost exactly opposite to those reported by Todd et al. (2000). That study also described a significant difference between distribution of adult males and females on wild Atlantic salmon hosts, but with favourable attachment locations for females being the areas adjacent and posterior to the anal fin, and the males predominating the head and anterior dorsal midline. Jackson and Minchin (1993) similarly noted a more posterior-preferred distribution of adult females on wild salmonids, whereas adult males and pre-adults of both sexes were found more uniformly across the body. This contrasts to observations on farmed Atlantic salmon held in sea cages (both in this study and anecdotally), where there was an equal preference for the head compared to the posterior region of the host (Wootten, Smith, & Needham, 1982), albeit dependent on louse stage and sex. Jaworski and Holm (1992) reported that 42%-81% of adult females were found solely in the post-anal body region on salmon held in sea cages, whereas the largest proportion (54%) were found on the head in this study. Females are thought to situate themselves to exploit areas of optimal resources and maximize hydrodynamic streamlining due to their larger body size (although this was suggested for the posterior preference of females on wild salmon; Todd et al., 2000).
Differences between farmed and wild salmon may be vast from a parasitic perspective, and variable distribution preferences may also be influenced by host-specific traits such as mucosal thickness Any of the above may affect the fitness outcomes resulting from specific attachment locations and host-switching behaviour for lice.
Previous studies have not highlighted lateral asymmetry in attachment preferences in lice; however, in this cohort there was a greater abundance of lice found on the left side of the host. This preference was also not specific for any developmental stages, indicating that the benefits of being situated on the left side are likely equal across stage and sex. As Atlantic salmon are essential bilaterally symmetrical, competition or attraction with conspecifics could drive this preference, but we consider it more likely due to site-specific environmental factors, whereby the exposed outer side of a salmon is more affected by hydrodynamic forces or potential scraping against the cage net. It is interesting to note that in this study, fish were not continuously schooling tightly as is often observed in TA B L E 4 Total lice found across all sample points in the study, and the proportion of those lice recorded that were found dislodged from the host into the sedation vessel. Mobile stages are represented (PA = pre-adult, A = adult) with sexes (M = male, F = female) differentiated. Note that the present study did not distinguish between pre-adults 1 and 2, or between sexes; all pre-adults are pooled The total number recorded of pre-adults 1 and 2, and adult stages of both sexes. d Ratio of the proportion of adult male and adult female lice found in the sedation vessel Note that the present study did not distinguish between pre-adults 1 and 2, or between sexes; all pre-adults are pooled (grey shaded values).
salmon aquaculture (Oppedal, Dempster, Stien, 2011), but rather that individuals swam in varied directions in a loose shoal, which is often observed when group size is small or just after sea transfer (e.g. Glaropoulos, Stien, Folkedal, Dempster, & Oppedal, 2019). The lateral preference observed here should be explored across a range of hosts and environmental conditions to discover whether this pattern is generalizable and, if so, elucidate the drivers of such a preference. This will be particularly important for in situ lice assessments using camera technologies that will typically image hosts from one side only. Users of these systems will need to predict or track any asymmetrical distribution in order to estimate cage lice abundances with confidence.

| CON CLUS IONS
Innovative techniques to prevent, monitor and remove sea lice infes-

E TH I C S S TATEM ENT
This study adhered to the Norwegian regulations for animal experimentation, as assessed by the Norwegian Food Safety Authority (application ID: 16289).

ACK N OWLED G EM ENTS
The authors thank the staff at the Sauganeset research sea facilities for their technical assistance, in addition to Nina Frogg and Kim Levik for their contributions to the sampling work. This study was indirectly funded by MSD Animal Health Norge.

CO N FLI C T O F I NTE R E S T
The authors have no conflict of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.