Salmon lice (Lepeophtheirus salmonis) pose a major challenge to the sustainability of salmon aquaculture due to their capacity to rapidly evolve resistance to parasite control methods. As the effectiveness of chemical treatments has declined, the industry has increasingly relied on preventive strategies to limit initial infections. One such approach is depth-based farming, where fish are held deeper in the water column using submerged cages. These systems reduce exposure to lice, which typically concentrate near the surface. However, there is growing concern that such practices may inadvertently select for lice that are better adapted to deeper swimming, potentially enabling resistance to depth-based interventions.

In this study, we investigated whether vertical swimming behaviour in salmon lice larvae is influenced by the depth at which their parents were collected. We sampled the first generation of 122 adult female lice carrying egg strings from commercial salmon farms using either standard cages (0–20 m) or submerged cages (20–40 m). Larvae were reared under controlled conditions, and the vertical positioning of 11,291 copepodid larvae was tested in pressure columns simulating a depth of 10 m. Our results revealed a significant interaction between larval depth distribution and the cage type from which the parental lice were sourced (χ² = 278.85, df = 1, p < 0.001). Larvae from standard cages showed a greater tendency to ascend (35% vs. 23%) and were less likely to sink (19% vs. 27%) compared to larvae from submerged cages. These findings suggest that vertical swimming behaviour may be heritable, with submerged cages potentially selecting for deeper-dwelling lice over time. This study provides the first evidence that the depth preference of salmon lice larvae may be influenced by their parents’ environment. Understanding this behavioural inheritance is crucial for evaluating the long-term sustainability of submerged cage systems and for developing lice management strategies that anticipate evolutionary responses.

Louse Collection and Larval Production

From early April to late May 2024, we collected female Lepeophtheirus salmonis from five salmon farms during routine lice counts within all cages at the sites (Fig. 1). We gathered 25 pairs of egg strings from adult females at two standard Atlantic salmon cages in Smørdalen and Brattavika, and 22 pairs from a standard cage in Uforø. For the submerged farms at Gjengane and Hestabyneset, we collected lice from recently deceased fish, as live fish sampling before harvest is difficult in submerged technology. The dead fish sank to the bottom of the submerged cages, where a lift-up suction pump brought them to the surface. The fish were then placed into a bin with a grid to allow water drainage, making it easy to collect still-attached lice. We obtained 25 pairs of egg strings from Gjengane and 11 pairs of egg strings from Hestabyneset, as well as 14 females and 4 males for laboratory reproduction.

After collection, lice were transported in chilled seawater to the Institute of Marine Research’s station in Matre, Masfjorden. Upon arrival, egg strings were carefully removed from the lice using tweezers. Each pair of egg strings from a female adult was incubated separately under controlled conditions (13 °C, 34 PSU), with daily monitoring of development.

12 salmon of about 400 g, approximately two years old and from the Aquagen strain, were sourced from holding tanks at the station. These salmon were placed in 400 L tanks at the research station and maintained at 12 °C and 34 PSU for two months. Adult lice (both male and female) from submerged cages at Hestabyneset were transferred onto these salmon under controlled conditions. The fish were sedated with Metomidate Hydrochloride (0.01 g/L, C₁₃H₁₅ClN₂O), and the lice were applied to the fish’s skin using tweezers and waterproof paper. The lice were monitored until new egg strings were produced. Once this occurred, the salmon were sedated again, and the egg strings were carefully removed and incubated. All egg strings, whether collected directly from farms or produced in the lab, were incubated under the same controlled conditions (13 °C, 34 PSU) until they developed into copepodids. In total, egg strings from 122 individual females were used in the pressurised column experiments: 108 collected directly from standard and submerged farms, and 14 produced in the lab from adult lice collected at Hestabyneset (submerged cage). These additional egg strings were produced to supplement the larvae from Hestabyneset, where egg string availability was insufficient.

For the experiments, we used three-day-old copepodids, the stage at which they are most infectious (Tucker et al., 2000; Skern-Mauritzen et al., 2020). Due to natural variation in level of egg string maturity, the copepodids hatched in a staggered manner. In total, 11,291 copepodids were assayed to investigate their vertical distribution in response to hydrostatic pressure, with a focus on how their swimming behaviour might be influenced by the parental generation, either from standard or submerged cages.

Experimental Columns

Experiments were conducted using small-scale water columns, as illustrated in the schematic diagram in Fig 1, following procedures established by Coates et al. (2020) and Crosbie et al. (2019, 2020). Each column consisted of a clear polyvinyl chloride (PVC) tube, 85 cm in height and 4.5 cm in internal diameter, positioned vertically and sealed at the bottom. Prior to each experiment, the tube was filled with seawater to a height of 80 cm, with a salinity of 34 PSU. The tube was then placed inside a transparent PVC box (85 cm x 20 cm x 20 cm) acting as a water bath maintained at a constant temperature of 13 °C. A white backdrop behind the column was used in order for there to be enough contrast to be able to see the copepodids clearly with the naked eye. To minimise light exposure, the area was surrounded by black plastic panels, keeping the experimental space dark. A PVC lid (28 cm²) was securely fastened over the apparatus with bolts, creating a 5 cm air pocket above the water. Rubber O-rings ensured an airtight seal. A 6 mm hose, inserted through a hole in the lid, connected the column to a scuba tank of compressed air via a pressure reducer and adjustable regulator (Festo LFR-D-MINI-A, Festo, Australia). The scuba tank introduced air to pressurise the air pocket, increasing the hydrostatic pressure of the seawater (Stake & Sammarco, 2003). Desired pressure levels were achieved in less than 4 seconds, after which no additional air was introduced or lost. To validate the apparatus, an inverted graduated test tube containing an air bubble was submerged in the column. When 1 bar of additional pressure was applied, the air bubble compressed to half its original volume, consistent with Boyle’s law.

Experimental assays

The swimming behaviour of three-day old infectious copepodids (all lice within a single replicate came from the same egg string), were tested under 2 atm of pressure (ambient pressure at 10 m depth) for a 5-minute interval. This duration was chosen based on research by Coates et al. (2020), as it captures the maximum behavioural response after which little to no further change was observed in the vertical distribution of copepodids.

Before each experiment, the tube was filled halfway with seawater at 34 PSU. Live lice were carefully transferred into the tube using a pipette to ensure deceased specimens or any viscous residues at the bottom of the incubator were not added. The remaining volume of the tube was then filled with seawater, allowing the lice to mix evenly throughout the height of the column. After being transferred, the lice were left to acclimatise in darkness for 15 minutes. All columns were run in darkness (underwater light levels = 0.006–0.03 µmol m² s¹, measured with a LI-COR LI-1500 light sensor, LI-COR Biosciences, Germany). After 5 minutes of pressurisation, the lice distribution was recorded on a sheet of paper stuck to the front of the column with water. With the aid of a head torch, larvae were visible to the naked eye. The water level in the tube was marked, and the distribution of lice was recorded on the sheet at 10 cm intervals from 10 cm to 80 cm. The counting was done within 45 seconds to minimise the effect of the head torch light on the lice distribution (Szetey et al., 2021). The same scientist recorded the vertical distribution each time to avoid sampling bias. In total, 122 columns were run (25 columns each for 5 locations, excluding 3 from one location) over a period of 2 months.

Statistical Analyses

First, we assessed whether the source of the eggs from lice sampled from Hestabyneset (whether collected directly from the submerged cage or produced after lice were transferred to the laboratory) influenced the results. We tested this using a Mann-Whitney U Test. The interaction between depth and egg string source (submerged farm vs. laboratory-reared) was not significant (W = 11808, p = 1), and patterns of distribution with depth were similar (see Supplementary Fig 1). Larvae from both sources for Hestabyneset were included in subsequent analyses.

To analyse the vertical distribution of copepodid-stage salmon lice larvae, we calculated the proportion of larvae observed at each 10 cm depth interval within the experimental columns. The distribution was bimodal, with higher proportions at the surface (0 - 10 cm; hereafter 10 cm)) and near the bottom (70 - 80 cm, hereafter 80 cm), which accounted for 53% of the observed lice across trials. Intermediate depths (10 – 70 cm) showed minimal variation and were excluded from further analysis (as per Coates et al., 2020).

A generalised linear model (GLM) with a binomial distribution was initially used to test the effect of cage type (submerged vs. standard) on vertical swimming behaviour at 10 cm and 80 cm. However, model diagnostics revealed significant overdispersion (dispersion ratio = 8.17), indicating that the variance exceeded what a standard binomial model could account for. To correct for this, we refitted the model using a quasibinomial distribution, which accounts for overdispersion and provides more reliable standard errors and significance testing. All analyses were performed in R version 4.3.1 (2023-06-16).