The immune system is crucial in responding to disease-causing pathogens. However, immune responses also generally cause stereotypical changes in behaviour known as sickness behaviours, which often include reduced activity. Sickness behaviours are thought to have an important role in conserving energy required to support the immune response, however little is known about how they manifest over time or in relation to risk, particularly in fishes. Here, we induced an immune response in mosquitofish (Gambusia holbrooki) by inoculating them with E.coli lipopolysaccharide (LPS). We subsequently tested batches of fish at 24-hour intervals and examined: locomotory behaviour, tendency to use a refuge and response to a threat stimulus. There was no difference in behaviour between control and LPS-treated fish 1, 3, or 4 days post-inoculation. However, 2 days post-inoculation, LPS fish swam more slowly and spent more time in the refuge than control fish, although no difference in peak acceleration was found. Our findings suggest that the expression of sickness behaviours peaks roughly 2 days following exposure to LPS and are relatively short-lived. Specifically, immune-challenged individuals exhibit reduced locomotion and exploratory behaviour. They also become more risk averse, but retain the ability to respond acutely to a threat stimulus.

Study Species

Mosquitofish (Gambusia holbrooki) measuring 20.4 ± 4.9 mm (mean ± standard deviation) were collected for use in the experiments in May 2023 from Manly Dam, Sydney (-33.806360 S, 151.235144 E). The mosquitofish is a small species of fish that inhabits shallow freshwater habitats and is preyed upon by a wide variety of predators, including fish, birds and insects (Pyke 2005). Prior to experiments, the fish were held for 1 week in 180 L vats held at a temperature of 23.1 ± 1°C and a 12:12 light:dark cycle in the animal holding rooms at the University of Sydney. Fish were fed daily ad libitum with commercially available fish flakes (Nutrafin).

Immune Challenge

Fish were randomly assigned to either LPS or control treatments (LPS-treated: N = 64, Control: N = 64). Groups of 8 individuals were placed in a 500mL bath of either LPS solution or plain aged tap water and left for 60 minutes before being transferred in their groups to separate 5L holding aquaria. Fish assigned to the LPS treatment were bathed in a solution of aged tap water and LPS at a concentration of 100mg/L (Sigma-Aldrich; Lipopolysaccharides from Escherichia coli serotype O111:B4). This concentration was selected based on previous studies which induced immune reactions in fish using similar dosages. LPS offers a valuable means of inducing an inflammatory response in vertebrates and has been widely used on subjects from various taxa for this purpose. The use of LPS allows the effects of innate immune responses to be decoupled from the variable and disease-specific effects of live pathogens, allowing broader inferences to be made. Following treatment, fish were monitored for signs of stress prior to experiments. Previous studies show that the immune response of mosquitofish peaks at approximately 48hrs post-exposure. The testing schedule (24, 48, 72, 96hrs) was chosen on this basis. The bath method was chosen over an intraperitoneal or intramuscular injection due to the small size of the fish and the success of this method in other trials using small teleosts.

Experimental Protocol

A single fish was netted from its holding tank and introduced to an adjacent circular test arena, which had a diameter of 64cm, and was filled to a depth of 8 cm with aged tap water at the same temperature as the holding tank (24°C). The arena was constructed from white Perspex. A single artificial aquarium plant was added to the arena. This had a circular base of diameter 3cm, spreading to a diameter of 11cm at the water’s surface. The plant was positioned 23 cm from the arena centre. The purpose of this was to provide a refuge for the fish. The arena was surrounded by white screens to minimise external disturbance and was lit by 2 x 20W cool white LED strips. Following its introduction, the fish was allowed to acclimate to the arena for 10 minutes, after which we filmed for 5 minutes, using a Canon GX camera set into a recess in the ceiling of the screens, 1.2m above the arena, and filming at 24 fps at a resolution of 1920p. Halfway through the filming process (i.e. 12.5 minutes after introduction and 2.5 minutes after we began filming) we dropped a 3g cubic lead weight into the centre of the arena. The purpose of this was to simulate an attack by an aerial predator. At the completion of 5 minutes’ filming, the fish was removed and transferred to a new holding tank. We conducted tests on each of four days, respectively 24 hours, 48 hours, 72 hours, and 96 hours following control inoculation, or inoculation with LPS. On each day, we tested 32 fish, comprising 16 LPS treated fish and 16 control fish. Each test was done with a naïve fish. Each fish was used only once. After testing, the sex and body length of the fish were recorded.

Analysis and Data Extraction

Each video was tracked using TRex software, yielding a total of 7200 x,y coordinates for each fish (300 seconds at 24 fps). From this, we were able to calculate our two key response variables, which were speed, and time in the refuge. We defined time in the refuge as being when a fish was within a radius of 5.5 cm of the centre of the artificial plant. We calculated speed based only on the time the fish spent outside the refuge. For the purpose of our analysis, we divided the trial into two halves: before the introduction of the lead weight, and after the introduction of the lead weight. The delineation between the two was determined as the point at which the lead weight first made contact with the water surface. Acceleration was measured by examining the trajectories of fish in the 10 seconds immediately following the threat stimulus. We took the 10 highest values for acceleration (i.e. peak acceleration) in this time and calculated the median of these.

Statistical analysis was performed using R, and using packages lme4, car, MuMIn, and ggplot. We used the model.sel function in MuMIn to identify the most appropriate model for the error distributions in both cases, We subsequently used a generalised linear mixed model, specifying a gamma distribution, to analyse time in the refuge, accounting for the fact that this response variable was non-normal and right skewed. We used a general linear mixed model to analyse speed. Neither sex, nor size were statistically significant (p > 0.6 in all cases), and so both were excluded from the final models. In each of our two models (‘Time in Refuge’ and ‘Mean Speed’), we specified three fixed effects: Treatment (Control or LPS), Day (1, 2, 3, or 4), and Threat (Pre-threat and Post-threat). We specified Fish ID as our random effect, to account for the repeated measures nature of the data in respect of quantifying behaviour before and after the threat stimulus. We used a general linear model to analyse acceleration, with Day and Treatment as our fixed factors.