Development of a Piscirickettsia salmonis immersion challenge model to investigate the comparative susceptibility of three salmon species

Abstract Piscirickettsia salmonis, the aetiological agent of salmonid rickettsial septicaemia (SRS), is a global pathogen of wild and cultured marine salmonids. Here, we describe the development and application of a reproducible, standardized immersion challenge model to induce clinical SRS in juvenile pink (Oncorhynchus gorbuscha), Atlantic (Salmo salar) and sockeye salmon (O. nerka). Following a 1‐hr immersion in 105 colony‐forming units/ml, cumulative mortality in Atlantic salmon was 63.2% while mortality in sockeye salmon was 10%. Prevalence and levels of the bacterium in kidney prior to onset of mortality were lower in sockeye compared with Atlantic or pink salmon. The timing and magnitude of bacterial shedding were estimated from water samples collected during the exposure trials. Shedding was estimated to be 82‐fold higher in Atlantic salmon as compared to sockeye salmon and peaked in the Atlantic salmon trial at 36 d post‐immersion. These data suggest sockeye salmon are less susceptible to P. salmonis than Atlantic or pink salmon. Finally, skin lesions were observed on infected fish during all trials, often in the absence of detectable infection in kidney. As a result, we hypothesize that skin is the primary point of entry for P. salmonis during the immersion challenge.

. Between 2002 and 2016, there were a total of 36 farm-level diagnoses of SRS in British Columbia (BC) (Jones, 2019).
There have been no reports of SRS outbreaks in sockeye salmon (O. nerka), and the susceptibility of sockeye salmon to P. salmonis is unknown.
Controlled exposure studies have been used to understand transmission characteristics of P. salmonis, host-pathogen interactions and the evaluation of potential treatments (Rozas & Enríquez, 2014).
The controlled laboratory infection is initiated in naïve fish by intraperitoneal (ip) injection or by cohabitation, in which recipient fish are held together with injected donors. However, ip injection bypasses primary host defences and typically elicits a rapid and severe infection, whereas mortality rates of cohabitation-infected fish are variable, time to death is unpredictable, and the dose is difficult to quantify or standardize. In earlier attempts to develop a waterborne P. salmonis challenge model, the challenge organism was obtained from infected cell cultures and mortality rates were highly variable (Birkbeck, Rennie, Hunter, Laidler, & Wadsworth, 2004;Smith et al., 1997Smith et al., , 2015. The advent of enriched blood agar media for plate-based culture of the bacterium (Mauel, Ware, & Smith, 2008) has provided an opportunity to obtain bacterial cells at concentrations sufficiently high for immersion challenges in combination with plate-based methods for bacterial quantification.
The current study evaluates a reproducible waterborne P. salmonis challenge model in pink, sockeye and Atlantic salmon smolts. We document clinical symptoms and pathogenesis, report species-specific differences in mortality and estimate the bacterial burden in tank water during the infections.  Table 1.

| Fish care
Pink salmon fry were transferred to PBS from the Quinsam River Hatchery (BC) and held in seawater for approximately 12 months prior to Trial 1. Juvenile Atlantic salmon were obtained from a commercial hatchery on Vancouver Island and reared in brackish water at PBS. Four weeks prior to Trial 2, fish were transitioned to seawater. Sockeye salmon (Pitt River stock) were obtained from the Inch Creek Hatchery (BC), reared in 5°C dechlorinated freshwater for 12 months and transitioned to seawater 12 weeks prior to Trial 3. Atlantic and sockeye smolts previously transitioned to seawater were used in Trial 4.
During an individual trial, fish were held in 400-L tanks provided with UV-treated flow-through seawater (7-8 L/min) under a 12-hr:12-hr photoperiod and fed a commercial diet (EWOS Canada) at daily rates of 1% (Trials 1 and 2) or 0.5% (Trials 3 and 4) total biomass. Dissolved oxygen levels were monitored daily and were ≥8 mg/L throughout the trial.

| Bacterial culture
The SR-1 isolate of P. salmonis used in all challenges was isolated in 2000 from Atlantic salmon during an SRS outbreak at a net-pen aquaculture site near Vancouver Island. For all challenges, a low-pass stock previously stored at −80°C was thawed and inoculated onto a monolayer of CHSE-214 cells (ATCC CRL-1681) (Lannan, Winton, & Fryer, 1984). Cells were grown in a 25-cm 2 flask with minimal essential medium (MEM) with Earle's salts (Gibco®) supplemented with 10% foetal bovine serum (MEM-10; Gibco®) and 2.5 µg/ml Fungizone® (Amphotericin B; Gibco®). Flasks were incubated at 15°C for 15-17 d to allow for development of confluent cytopathic effect. The infected cells were then gently scraped off the bottom of the flask into the TA B L E 1 Salmon species, water quality and experimental parameters for individual trials media using a cell scraper, and 100 µl of the harvested material was spread-plated on BCG agar (Mauel et al., 2008). Plates were incubated at 15°C for 14-21 d to allow for development of a bacterial lawn.

| Waterborne challenge
For all trials, P. salmonis colonies were resuspended in 105 ml of BM4 liquid medium (Henríquez et al., 2013)  Dead, moribund and surviving fish were examined for evidence of infection and/or pathology. In Trial 2, fish were frozen until necropsy, and in all other trials, fresh fish were examined. At the conclusion of Trial 1, kidney samples were collected from 25% of fish (n = 8) in the 10 1 , 10 2 and 10 3 cfu/ml treatments and from all fish in the 10 4 cfu/ml treatment (n = 32). In Trial 2, kidney and liver samples were collected and preserved in 95% ethanol from all survivors (n = 27) and from 34% of mortalities (n = 16), and kidney, liver and spleen from 5 mortalities were streaked on BCG agar. In Trial 3, kidney samples from all mortalities (n = 9) were preserved in 95% ethanol, and kidney, liver and spleen from 6 mortalities were streaked on BCG agar. In addition, subsamples of kidney from 20% of survivors in each treatment (n = 10) were streaked on BCG agar and preserved in 95% ethanol, respectively. At the conclusion of Trial 4, samples of kidney, spleen, liver, heart, brain, gill, intestine and skin lesions from each fish were preserved in 10% neutral-buffered formalin (NBF) for histological examination. In addition, subsamples of kidney were streaked on BCG agar and preserved in 95% ethanol. BCG plates were incubated at 15°C for 14 -21 d and monitored for bacterial growth.

| Water sampling for bacteria
In Trials 2 and 3, water samples were collected every 2 d. To do so, the water flow to each tank was stopped for 30 min after which a 40 ml water sample was collected. Fish were monitored throughout this period, and supplemental aeration was provided. Water samples were kept at −80°C pending DNA extraction.

| Histology
NBF-fixed tissue samples were dehydrated in an alcohol gradient, clarified in xylene and infiltrated with paraffin wax for subsequent sectioning at 3 µm, staining and examination by light microscopy.

| DNA extraction and quantitative PCR
Thawed water samples were centrifuged at 12 000 x g for 10 min at 4°C, and the supernatant was removed. DNA was extracted from the resulting pellet and from cultured bacteria by resuspending in 180 µl Buffer ATL and 20 µl Proteinase K (Qiagen DNeasy® Blood and Tissue kit) following the manufacturer's instructions for Gramnegative bacteria. DNA was purified from ethanol preserved kidney using the same kit and following the manufacturer's instructions for purification of total DNA from animal tissues. Eluted DNA was stored at −20°C.
Bacterial loads in tissue and tank water samples were determined using a qPCR assay designed for the P. salmonis 23S gene (Corbeil, McColl, St, & Crane, 2003). The identity of cultured bacteria isolated from infected fish (1 sockeye and 1 Atlantic salmon) was confirmed by amplifying a fragment of the 16S ribosomal gene by conventional nested PCR (Mauel, Giovannoni, & Fryer, 1996) followed by sequencing (Eurofins Genomics). The resulting sequences were compared to NCBI nucleotide archives by BLAST.

| Statistical analysis
For Trial 1 data, a chi-square test of independence was done in R version 3.6.1 to determine the statistical significance of differences in the proportion of fish with lesions between treatments. For post hoc analysis, pairwise chi-square tests using Fisher's exact test were employed and Bonferroni-adjusted P values were generated using the rcompanion package (Mangiafico, 2020). Results were considered significant if p ≤ .05.
Graphs were prepared in R using the ggplot2 package (Wickham, 2016).

| Challenge trials
The nucleotide sequence of the 16S ribosomal gene from both bacterial isolates was identical to the corresponding region in the recently sequenced P. salmonis SR-1 genome (GenBank accession number CP039227.1).

| Trial 1 (pink salmon)
The trial was terminated at 30 d post-immersion (dpi), and there were no mortalities. At this time, external lesions such as scale loss, areas of redness and ulcers were observed on fish in all groups with the exception of the negative control. Ulcers were typically shallow and bloody. Erosion of the dermal layer with exposure of the underlying muscle was rarely observed. Lesion frequency increased significantly (χ 2 = 66.2, df = 3, p < .001) with dose (

| Trial 2 (Atlantic salmon)
The trial was terminated at 50 dpi. Mean cumulative mortality and morbidity (CMM) were 63.2% ± 10.8 (Figure 1a), and mean days to death (MDD) was 39.6 dpi. The external lesions observed on all dead fish included scale loss, ulcers, darkening of skin around the ulcer, petechial haemorrhaging and areas of redness. Ulcer severity ranged from shallow, in which the dermal layer was still intact, to deep in which the underlying muscle was exposed. The bacterium was detected by qPCR in 33% of survivors sampled at 50 dpi and in all mortalities (Table 3). In addition, pure cultures of P. salmonis were re-isolated on BCG agar from 5 mortalities sampled at 44 dpi.

| Trial 3 (sockeye salmon)
The trial was terminated at 58 dpi. Mean CMM values of 8% and 10% were observed in the 10 4 and 10 5 cfu/ml groups, respectively (Table 4), and MDD values were 37.8 (10 4 cfu/ml) and 42.4 (10 5 cfu/ ml). P. salmonis was re-isolated on BCG agar from all 6 dead salmon examined. Among the dead fish, white subcapsular hepatic lesions were observed in one without external symptoms of disease.
Mottled livers (n = 2), splenomegaly (n = 1) and pale intestine (n = 3) were observed among the remaining 5 dead fish. Ulcer severity (n = 5) ranged from shallow and pale to deep with muscle tissue exposed, and 8 of 9 mortalities were positive by qPCR (Table 4). The median genome equivalents (GEq)/µg DNA in mortalities from the 10 4 cfu/ml treatment was 8.9 × 10 4 and 5.2 × 10 3 for mortalities in the 10 5 cfu/ml treatment. The bacterium was not isolated from survivors, and prevalence, as determined by qPCR, was low (Table 4).

| Trial 4 (Atlantic and sockeye salmon)
The trial was terminated at 29 dpi. The CMM for Atlantic salmon and sockeye salmon was 13% and 0%, respectively. The trial was terminated

| Quantification of P. salmonis in seawater
Piscirickettsia salmonis burden in tank water was estimated from samples collected during Trials 2 and 3. In Trial 2 (Atlantic salmon), the bacterial burden peaked at 36 dpi with a median value of 37.1 GEq/ml SW/fish (interquartile range, IQR = 16.9) ( Figure 1b). In Trial 3 (sockeye salmon), the bacterial burden was consistently low in all treatments (Figure 3a,b). The bacterium was detected in the 10 5 cfu/ml treatment between 30 and 42 dpi, with a peak of 0.5 GEq/ml SW/fish at 38 dpi (Figure 3a). In the 10 4 cfu/ml treatment, P. salmonis was detected between 32 and 38 dpi and the bacterial burden peaked at 38 dpi at a median value of 0.19 GEq/ml SW/fish (Figure 3b). Fewer than 0.22 GEq/ ml SW/fish were detected at 24 and 54 dpi in the 10 3 treatment, and the bacterium was not detected in the 10 2 treatment (data not shown).

| D ISCUSS I ON
In the current study, we demonstrate that 1-hr immersion in a sus-  (Smith et al., 1997). Despite the lack of mortality, infections were acquired and the authors suggested longer exposure times or higher bacterial concentrations may have been necessary to elicit mortality. A later study with 2.5 g rainbow trout reported 92% mortality following immersion in 10 5 TCID 50 /ml of another Chilean strain (SLGO-95) (Smith et al., 2015).
In a study with 100 g Atlantic salmon post-smolts, 1 of 10 fish died following 1-hr immersion in 10 5 TCID 50 /ml of a Scottish P. salmonis isolate (SCO-95A) (Birkbeck et al., 2004). These observations indicate the importance of fish size and bacterial strain as considerations in immersion challenge development. This earlier work also highlights the limitations in calculating absolute bacterial infectious doses from cytopathic effect in cultured cells.
Our research has begun to shed light on the relative susceptibility of sockeye salmon to P. salmonis infection and SRS. Following exposure to 10 5 cfu/ml, mortality in sockeye salmon (Trial 3) was 10% compared with 63% in Atlantic salmon (Trial 2). Similarly, at this dose the bacterium was detected in fewer sockeye survivors (10%, Trial 3) compared with Atlantic salmon survivors (33%, Trial 2). In Trial 4 in which sockeye and Atlantic salmon were concurrently challenged with 10 5 cfu/ml, despite similarities in skin lesion frequency, bacterial isolation rates (7% versus 53%) and qPCR prevalence (27% versus 67%) were lower in sockeye salmon. Data from pink salmon exposed to 10 4 cfu/ml in Trial 1 (qPCR prevalence = 59%, median GEq/µg F I G U R E 3 Bacterial burden in water following immersion of sockeye salmon in Piscirickettsia salmonis (Trial 3). P. salmonis genome equivalents in the (a) 1 × 10 5 cfu/ ml and (b) 1 × 10 4 cfu/ml treatments. Data are presented in box plots in which the inner horizontal line is the median, and the upper and lower boundaries of the box correspond to the first and third quartiles. The upper and lower whiskers denote the largest and smallest values no further than 1.5 times the interquartile range. Dashed horizontal line denotes assay limit of detection DNA = 290) suggest higher susceptibility in this species. Together, these data are consistent with the hypothesis that sockeye salmon are less susceptible to P. salmonis than pink or Atlantic salmon and represent the first evidence of species-specific differences in salmonid susceptibility to this pathogen when exposed via immersion. In an earlier study (Smith et al., 1996), size-matched coho salmon and rainbow trout were ip-injected with P. salmonis LF-89 and susceptibility compared. Lower levels of cumulative mortality and reduced bacterial burden in kidney led to the conclusion that rainbow trout were less susceptible than coho salmon. In another study (Garcés et al., 1991), mortalities of up to 100% in coho and Atlantic salmon following ip injection with high doses (> 10 3.3 TCID 50 /fish) of an unidentified Chilean P. salmonis isolate precluded any distinction of susceptibility between these species. In western Canada, outbreaks of SRS or a similar disease have been reported in cultured Atlantic salmon, Chinook salmon and coho salmon (Brocklebank et al., 1992), and the infection has been recognized in pink salmon (Brocklebank et al., 1992;Jones et al., 2020). In contrast, there have been no reports of SRS or P. salmonis infection in sockeye salmon and our data suggest a reduced likelihood of infection or SRS in migratory sockeye salmon.
Pathogen shedding data provide useful indications of temporal shifts in infectiousness within a population. In the current study, we estimated the timing and magnitude of P. salmonis shedding from water samples. In tanks containing infected Atlantic salmon, the bacterium was first detected at 12 dpi and levels peaked at 36 dpi, 3 d prior to the MDD. From 44 dpi onwards, levels of P. salmonis in the water column remained close to the LOD and coincided with the cessation of acute mortality. Similarly, peak shedding in sockeye salmon occurred at 38 dpi, 4 d prior to the MDD. The peak shedding rate in Atlantic salmon was 82-fold higher than the peak shedding rate in sockeye salmon exposed to 10 5 cfu/ml. We can therefore conclude that, in infected fish, the highest levels of P. salmonis are shed shortly before death, and Atlantic salmon experiencing an SRS outbreak are most infectious between 18 and 42 dpi. Our estimates of bacterial shedding in this study reflect the accumulation of bacteria in tank water from all infected fish. Shedding is unlikely to be synchronized either in timing or magnitude within a population and further studies are required to estimate the contributions of individual fish to better define infectious characteristics within the population. There are few data describing shedding of bacterial pathogens from fish. In Renibacterium salmoninarum, another facultative intracellular bacterium, McKibben and Pascho (1999) reported shedding from chinook salmon first occurred at 12 dpi and consistently occurred by 20 dpi.
Skin has previously been identified as one of the primary routes of entry of P. salmonis (Smith et al., 1999). The skin lesions described here were similar to those described in previous studies in which fish were exposed by skin patch, injection, oral inoculation or gill inoculation (Almendras, Fuentealba, Markham, & Speare, 2000;Smith et al., 1999). Skin lesions due to P. salmonis progressed from a slight raised area to decolouration and scale loss and finally to ulceration (Smith et al., 1999). When examined at or before the onset of mortality, more pink, sockeye and Atlantic salmon (Trials 1 and 4) had skin lesions than had detectable infections in the kidney. Almendras et al. (2000) had previously hypothesized that P. salmonis infects leucocytes and is disseminated throughout the circulatory system whereupon it infects vascular endothelial cells, eventually resulting in systemic infection. Leucocyte proliferation has been observed at the site of P. salmonis-induced skin lesions (Smith et al., 1999), and P. salmonis survives and replicates in leucocytes (McCarthy et al., 2008;Rojas, Galanti, Bols, & Marshall, 2009). This information, combined with premortality shedding measured in Trials 2 and 3, is consistent with the delayed onset of a systemic infection after an infection is acquired by external exposure. Although we did not monitor the route of entry of the bacterium, our data are consistent with the hypothesis that skin serves as a portal of entry for the bacterium during the immersion exposure. The epithelia of gill, the oral cavity and the oesophagus have also been suggested as potential routes of entry (Almendras, Fuentealba, Jones, Markham, & Spangler, 1997;Rozas-Serri et al., 2017;Smith et al., 1999), and further work is needed to elucidate their relative importance in uptake of the bacterium.
In conclusion, we describe the development and application of a standardized, reliable and controlled quantitative immersion challenge method, using a western Canadian isolate of P. salmonis, which results in infection and SRS in Atlantic salmon and two species of Pacific salmon. We link the timing and magnitude of bacterial shedding with acute mortality and provide evidence that sockeye salmon have reduced susceptibility to the infection and to SRS compared with Atlantic and pink salmon.

ACK N OWLED G EM ENTS
We would like to acknowledge the assistance of Jessica Low, Ashley Burton and the Aquarium Services staff at the Pacific Biological Station. This research was funded by Fisheries and Oceans Canada's Program for Aquaculture Regulatory Research.

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

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. Pathogenesis of liver lesions caused by experimental infection with