Parasite infection directly impacts escape response and stress levels in fish

Parasites can reach high biomass in marine communities (Kuris et al., 2008) and make up around 40% of the total biodiversity on Earth, making them one of the most successful modes of life (Poulin and Morand, 2000; Hatcher and Dunn, 2011). As such, parasites play a significant role in ecosystem functioning as they exert sub-lethal effects on their host where they can modify and manipulate behavioural and physiological phenotypes (for review, see McElroy and de Buron, 2014). Unlike macropredators such as piscivores, micropredators (which we define broadly to include both parasites and micropredators as defined more narrowly by Kuris and Lafferty, 2000, and Lafferty and Kuris, 2002) typically do not cause direct mortality, but rather cause a constant drain on energetics, thus significantly affecting host physiology and behaviour (for review, see Barber, 2007). However, the magnitude of this change depends on the parasite type, parasitic loading and the size and ontogenetic stage of the host (Sun et al., 2012). For example, larval and juvenile fish are reported to be more vulnerable to the effects of infection than their adult counterparts, owing to low body reserves and high metabolism (Strathmann et al., 2002; Grutter et al., 2011). Moreover, parasitic infection can also affect behaviours and physical attributes important for fleeing predators, such as reducing visual acuity (Seppälä et al., 2005), limb malformation causing reductions in maximum jumping distance, burst swimming speed and endurance (Goodman and Johnson, 2011), and reducing critical swimming speeds in adult and newly settled coral reef fishes (Binning et al., 2013; Grutter et al., 2011).

One of the most prevalent ectoparasites on coral reefs is gnathiid isopods (Crustacea) (Grutter, 1994; Sikkel and Welicky, 2019). Gnathiids, mobile temporary parasites of fish, feed using a trophic strategy that might best be referred to as micropredation (Kuris and Lafferty, 2000; Lafferty and Kuris, 2002). Micropredators attack multiple prey (hosts), much like predators do, but an individual micropredator’s effect on its prey tends to be small. Micropredators of vertebrate hosts, such as ticks, mosquitos and gnathiids, briefly feed on blood and are not transmitted trophically. Because micropredators feed on several prey individuals, they also do not benefit from minimising damage to prey (Barber et al., 2000) and can rapidly abandon their prey if it is incapacitated (Murray, 1990; Lehmann, 1993). These reef-based parasites feed on a variety of coral reef fish hosts from teleosts to elasmobranchs and on all host ontogenetic stages (Grutter and Poulin, 1998; Grutter et al., 2017). As such, they can cause significant physiological stress such as increased oxygen consumption (Grutter et al., 2011), reduced haematocrit (Jones and Grutter, 2005), increased cortisol load (Triki et al., 2016) and even mortality (Hayes et al., 2011). Previous work by Grutter et al. (2011) estimated that a single gnathiid can consume up to 85% of the blood volume of a late-stage larval damselfish, which has the potential to significantly affect behaviours that rely on aerobic activities, such as swimming (Gallaugher et al., 1995; Grutter et al., 2011). Reduced swimming performance can affect the way in which a fish interacts with conspecifics and predators and whether it can settle successfully to the benthic environment (Allan et al., 2013; Grutter et al., 2011).

When coral reef fishes recruit to the benthic environment, it is reported that predator-induced mortality can be absolute, but averages 60% within the first few days of settlement (Almany and Webster, 2006). Predator avoidance and evasion are key ecological traits that are directly related to growth and survival. When a predator attacks, prey are faced with a series of decisions, such as how fast to respond, which direction to turn, and how fast and how far to escape in an overall whole-organism behaviour called a fast start (for review, see Domenici and Blake, 1997). Fast-start escape behaviour can significantly increase the probability of prey escape (Walker et al., 2005; Allan et al., 2013, 2015, 2017). The effectiveness of fast-start escape behaviour is a consequence of body morphology, muscle mass and muscle cell physiology and energy reserves (Langerhans, 2009). Fast starts are characterised by rapid acceleration, which is driven by the rapid anaerobically powered contraction of large myotomal blocks of fast glycolytic muscle (Rome et al., 1988; Josephson, 1993). Although anaerobically powered, fast starts are a strenuous form of activity in which the active muscles require more oxygen than can be supplied during the period of activity. Therefore, an oxygen debt is accrued that needs to be repaid via aerobic metabolism (Scarabello et al., 1991).

To date, few studies have addressed the effects of parasitic load on fast-start escape behaviours. Blake et al. (2006) examined the effects of parasite load on the C-start performance of the three-spined stickleback, Gasterosteus aculeatus, and found negative effects on escape kinematics (Blake et al., 2006). By contrast, Binning et al. (2014) tested the escape performance of the monocle bream, Scolopsis bilineata, following infection by the large ectoparasitic cymothoid isopod Anilocra nemipteri, and observed no change in the escape performance of parasitised fish, suggesting that infection may not compromise escape performance. However, these studies used adult fish (overall range in body length of 4 to 13 cm) to measure the effects of parasite infection on escape performance, and it seems likely, given the physiological cost of parasitic infection (Grutter et al., 2011; Sun et al., 2012), that the escape performance of coral reef fish recruits would be negatively affected. Therefore, the main goal of the current study was to understand whether gnathiid infection would compromise the fast-start escape kinematics of newly settled, coral reef fish recruits. Furthermore, we evaluated whether experimental exposure to gnathiids induced changes in cortisol levels. The physiological processes by which fish respond to a stressor can be grouped into primary, secondary and tertiary responses (Barton and Iwama, 1991). Initially, catecholamines from chromaffin tissue are released, thus stimulating the hypothalamic–pituitary–interrenal (HPI) axis, which causes the release of corticosteroid hormones. This is followed by a secondary response, which involves haematological preparations to increase the efficiency of metabolic and immune responses (for review, see Barton, 2002). Finally, tertiary responses manifest as changes in whole-animal performance, such as changes in condition and behaviour. Increased cortisol following infection by haematophagous parasites has been observed across multiple taxa, including birds (Quillfeldt et al., 2010), rodents (St Juliana et al., 2014) and fishes (Triki et al., 2016). These variables were selected as they are key metrics of individual performance and are predictors of fish survival in the wild (McCormick et al., 2018). Newly settled fish were chosen as prey because the life-history shift between pelagic larvae and settled juveniles represents an important bottleneck where mortality is intense and selective.

All work carried herein was in accordance with the James Cook University Animal Ethics guidelines (JCU Animal Ethics approvals A2080, Great Barrier Reef Marine Park Authority collection permit G12/35117.1.). During December 2016, newly metamorphosed ambon damselfish, Pomacentrus amboinensis Bleeker 1868 (Pomacentridae) [standard length range 9–12 mm, mean±s.d. 10.3±0.05 mm] were collected using light traps (Meekan et al., 2001) in the waters off Lizard Island (14°40′S, 145°28′E) in the northern Great Barrier Reef, Australia. This species is a common component of the benthic fish fauna of Indo-Pacific reefs, and adults inhabit sandy areas of lagoons and inshore reefs. Pomacentrus amboinensis naturally settle on patch reef environments near the continuous reef. In this habitat, juveniles are exposed to reef-associated gnathiids and macropredators that use a variety of feeding modes from ambush (lizardfish Synodus dermatogenys and the small grouper Cephalopholis microprion) to pursuit (dottybacks Pseudochromis fuscus and wrasse Thalassoma lunare). These fishes can be observed to prey on juveniles that venture too far from shelter (McCormick, 2012), including the species used in this study, P. amboinensis. After capture, P. amboinensis were transferred from light traps to aquaria (65×35×30 cm) with aeration and water flow for a minimum of 48 h before use in trials. Coral reef fish recruits, when captured using light traps, habituate to life in aquaria extremely quickly and will feed within several hours following removal from light traps.

In the evening, prior to behavioural trials (17:00 h), well-fed P. amboinensis were individually transferred to randomly assigned 700 ml black aquaria filled with filtered seawater. Fish were left to habituate for 1 h, after which a single, unfed, stage three gnathiid (∼1.5 mm long, harvested from a well-established gnathiid culture tank at the Lizard Island Research Station; Grutter et al., 2010) was carefully transferred to each aquarium via a pipette. Control fish were treated in the same way and transferred into 700 ml black aquaria filled with filtered seawater. However, instead of a gnathiid, filtered seawater was added via a pipette. After transfer, the gnathiids were observed to be swimming freely in the aquaria. Fish were exposed to the gnathiids during the night, as gnathiids tend to be nocturnally active when their fish hosts are less active (Grutter and Hendrikz, 1999; Sikkel et al., 2009). Fish were left undisturbed for 2 h and were subsequently checked at 2 h intervals (using a red light to minimise disturbance) throughout the night, and the status of the gnathiid (fed, unfed or gnathiid missing – presumably eaten by the fish) recorded. The next day, the fish were tested for swimming behaviour and fast-start responses in the order in which they had been parasitised, meaning that they were tested no more than 10 h after the gnathiid was observed to be attached. To control for a temporal effect, control fish and non-parasitised fish (i.e. the gnathiid remained unfed at the end of infection exposure) were also tested throughout the day. For sample sizes per treatment, see Fig. 1 legend.

Routine swimming and fast starts were examined using individual fish in a transparent circular acrylic arena (diameter 200 mm, height 70 mm) within a large opaque-sided plastic tank (585×420×330 mm; 60 l) with a transparent Perspex bottom to allow responses to be filmed from below using the fish's silhouette. The water level was maintained at a height of 60 mm to reduce movements in the vertical plane, and the water in the arena was emptied and refilled with fresh seawater after approximately every 20 min to maintain water quality and temperature. The arena was illuminated by an LED light strip wrapped around the outside of the holding tank with light penetrating with even illumination through the white plastic sides. At the end of the 5 min habituation period, routine activity (used to determine routine swimming) was recorded as a silhouette from below, at 30 frames s⁻¹ for 2 min (Casio EX-ZR1000). Routine swimming was analysed on these 2 min video sequences and measured by tracking the distance (in metres) covered by the fish every second, resulting in 120 data points per fish. From this distance measure, average speed was also calculated (m s⁻¹).

A fast start was then stimulated by the release of a conical weight with a tapered end into the testing arena and was recorded at 480 frames s⁻¹ (Casio EX-ZR1000). This was only carried out when fish had moved to the middle portion of the tank, allowing an individual to move an equal distance in any direction and standardising for fish position relative to the stimulus. The weight was released from an electromagnet and was governed by a piece of fishing line that was long enough such that the tapered tip of the weight only just touched the surface of the water. To avoid a premature fast-start response associated with visual stimulation occurring, the weight was released through a 550 mm piece of 48.5 mm diameter PVC pipe with the bottom edge at a distance of 10 mm above the water level. To ensure a standardised protocol, fast-start variables were only measured when fish performed a C-start (commencement of fast-start that results in the individual forming a C-shape, sensu Domenici and Blake, 1997). A minimum of 27 replicates (individual fish) per treatment group were startled to ensure statistical robustness (n=34 controls, n=34 non-parasitised and n=27 parasitised). Trials were conducted between 08:00 h and 16:00 h. Kinematic variables associated with the fast-start response were analysed using ImageJ with a manual tracking plug-in. The centre of mass of each fish was tracked for the duration of the response. The following kinematic variables were measured. (1) Response latency (s), measured as the time interval between the stimulus onset and the first detectable movement leading to escape of the animal. (2) Response distance (m), a measure of the total distance covered by the fish during the first two flips of the tail [the first two axial bends, i.e. stages 1 and 2 defined based on Domenici and Blake (1997), which is the period considered crucial for avoiding ambush predator attacks (Webb, 1976)]. (3) Response speed (m s⁻¹), measured as the distance covered within a fixed time (25 ms). This fixed duration was based on the average duration (22.8 ms) of stage 1 and 2 (as defined above). (4) Maximum response speed (m s⁻¹), measured as the maximum speed achieved at any time during stage 1 and stage 2.

After fish had been assessed for their routine swimming and fast-start responses, they were killed by cold shock, blotted dry, immediately frozen in liquid nitrogen, and then transferred back to James Cook University, Townsville, QLD, Australia, where samples were analysed for cortisol (n=14 controls, n=13 non-parasitised, n=14 parasitised).

Individual fish were freeze-dried (0.2 mbar, >16 h, Alpha 1-2 LDplus, Martin Christ, Osterode am Harz Germany) and weighed (Mettler Toledo UMX2 Ultra-Microbalance, 0.1 µg readability) then homogenised in 2 ml Eppendorf vials, using a glass bead, 0.5 ml 1× phosphate-buffered saline (PBS, pH 7.4) and a shaking mill (3 min, FastPrep24, MP Biomedicals, Santa Ana, CA, USA). Homogenised tissue was transferred to a 10 ml glass vial and rinsed with 0.4 ml PBS. Ethyl acetate (Ajax Finechem, Thermo Fisher Scientific) was added (1:9 ratio), and samples were vortexed (1 min, Vortex Mixer, Ratek, Boronia, VIC, Australia) and centrifuged (3500 rpm, 5 min, 4°C, Eppendorf 5810 R). Ethyl acetate has been shown to be an effective organic solvent for extracting whole-body cortisol from early life stages of fish (Yeh et al., 2013). The supernatant was collected and transferred to a 28.5 ml glass vial, and this extraction step was performed 4 times with all collected supernatants being pooled. The ethyl acetate was dried off in glass reaction tubes using a centrifugal vacuum concentrator (43°C, Savant SpeedVac SC110A, Thermo Fisher Scientific). The samples were reconstituted on the same day with 1 ml assay buffer and processed following the enzyme-linked immunosorbent assay (ELISA) protocol provided by Cayman Chemical (Cortisol ELISA Kit 500360, Cayman Chemical). The samples were analysed in triplicate with a spectrophotometer (SpectraMax Plus 384 Microplate Reader, Molecular Devices; average absorbance calculated from readings at 405 to 420 nm).

Several assay validation steps were performed to test for parallelism, accuracy and precision of the cortisol ELISA kit, following recommendations by Metcalfe et al. (2018). Parallelism was confirmed by comparing dose–response curves of diluted samples against a standard curve (ANCOVA, P>0.05, n=3). Briefly, reconstituted samples (n=3) were diluted (1:4, 1:8, 1:12, 1:16, 1:20 and 1:24) and compared against the cortisol standard curve (Cayman Chemical ELISA kit, 6.6–4000 pg ml⁻¹ range). An optimal dilution for the samples (20×) was observed at 50% relative maximum binding, and sample dilutions falling within 20–80% B/B₀ relative maximum binding were considered as acceptable (Metcalfe et al., 2018). The accuracy of the method (i.e. the recovery of a known amount of added cortisol) was assessed by spiking three samples with 800 pg cortisol ml⁻¹, more than half of the samples' cortisol concentration and within the detection limit of the ELISA kit (see Guest et al., 2016). For each of the three samples, two fish were homogenised, pooled and the homogenate split into halves, with one half receiving the spike and the other the assay buffer. Both parts were then processed in the same way as all other samples. The spike's recovery (percentage) was expressed as spiked sample result−unspiked sample result×100/known spike (800 pg ml⁻¹), and the mean recovery (94.3%, n=3) was used as a correction factor for calculating the samples' cortisol concentration. Intra-assay precision of triplicate samples was determined using the coefficient of variation (CV), and found to be 5.5±4.9 (mean±s.d., n=41).

A preliminary analysis of covariance (ANCOVA) found that latency to respond to the startle was positively related to distance to the stimulus, and the slope of the relationship did not differ between the two treatments (i.e. homogeneous slopes; F_2,84=1.77, P=0.177). To remove the influence of distance to the stimulus on latency (F_1,84=11.29, P=0.001), the residuals of the relationship were used for subsequent analyses. No other variable was affected by distance of the fish to the startle stimulus. A multivariate analysis of variance (MANOVA) was undertaken to determine whether there was a difference in the routine swimming or fast-start kinematics of P. amboinensis after exposure to a single gnathiid. Dependent variables included were: the fast-start variables distance, speed, maximum speed and latency (residuals), and the routine swimming variables distance and speed. The nature of significant differences found by MANOVA in relation to the original variables values was then compared between treatments using canonical discriminant analyses (CDAs) to determine how escape and swimming kinematics differed between treatments. Trends in the behavioural variables were represented as vectors, which were plotted on the first two canonical axes, together with treatment centroids and their 95% confidence clouds (Seber, 1984). The strength or importance of each of the original variables in discriminating among groups was displayed graphically as the length and direction of these vectors. To further explore the differences between treatments, one-way ANOVA were used to identify significant differences within individual behaviours of interest. When significant, differences were further examined using Tukey's HSD means comparison tests. Pairs of fish were successively tested in the same water; however, in doing this, it is possible that the behaviour of the second fish may have been influenced by chemical signals excreted from the first fish. To remove this potential risk, we suggest using clean water for each trial. To account for this possible bias, we undertook a repeated-measures approach to test the potential effect of trial order influencing the behaviour of the fish, while still allowing us to determine whether there was an effect of gnathiid exposure. Here, a two-way repeated-measures MANOVA was undertaken on a subset of pairs of fish to test the effect of trial order (1st or 2nd trial) and treatment (n=8 control pairs, n=7 non-parasitized pairs, n=5 gnathiid pairs) on the routine swimming and fast-start kinematics of P. amboinensis. All assumptions of normality and homogeneity of variance were visually inspected and found to have been met. Analyses were carried out in Statistica version 13.

The cortisol results were tested for homogeneity of variance, which was found to be violated (Bartlett's test, P<0.001). Data were subsequently analysed using a Kruskal–Wallis test with Dunn's test and Holm–Šidák adjustment as post hoc tests. All statistical analyses were performed in R, version 3.5.1.

Exposure to a single gnathiid affected nearly all measured kinematic traits (Figs 1 and 2, Table 1). The MANOVA revealed a significant difference in the overall change in behaviour in response to gnathiid exposure (Pillai's trace 0.414, F_12,164=3.568, P<0.0001). A CDA displayed the nature of the differences found among treatment centroids and shows a clear separation of the three treatments into two distinct groups with respect to the six behavioural measurements, with the parasitised treatment being separate from the non-parasitised and control treatments (Fig. 1). Control fish and non-parasitised fish were differentiated from parasitised fish along the first canonical axis, which accounted for 90.4% of the difference among treatments. This axis was principally driven by trends in fast-start kinematics, which indicated that control fish and non-parasitised fish travelled further, had higher average speeds and exhibited a more rapid response to the drop stimulus (i.e. lower response latency) than parasitised fish. This suggestion was statistically confirmed by the results of the one-way ANOVA, with the parasitised group exhibiting reductions in performance in nearly all measured traits (Fig. 2). For example, parasitised fish were slower to respond to the stimulus, with increased latency in this group (F_2,86=11.425, P=0.001). The distance achieved during stage 1 and 2 and the speed achieved during this same period were significantly reduced (F_2,88=3.871, P=0.0025; F_2,88=3.987, P=0.0022) in fish that had been parasitised. In addition, the distance and speed over a 2 min period were significantly reduced, with parasitised fish covering half the distance covered by the control and the non-parasitised groups (F_2,91=9.929, P=0.001; speed F_2,91=9.997, P=0.001). There was, however, no significant difference among treatments in the maximum speed achieved during an escape (F_2,87=1.818, P=0.160). The repeated measures MANOVA revealed a significant effect of treatment (Wilks 0.337, F_8,28=2.523, P=0.033). However, the order in which the trial occurred was insignificant (Wilks 0.814, F_4,14=0.7861, P=0.547). There was also an insignificant interaction between order of trial and treatment (Wilks 0.646, F_8,28=0.851, P=0.566). These results suggest that despite being tested in the same water as a previous trial, there was no effect of this on routine swimming or fast-start escape behaviour.

Cortisol concentrations were significantly different among treatments (Kruskal–Wallis test, P<0.001) but highest in ambon damselfish that were parasitised by gnathiids (Dunn's post hoc test, P<0.001; see Fig. 3, Table 2). Non-parasitised ambon damselfish showed comparable cortisol levels to fish maintained under control conditions (Dunn's post hoc test, P=0.244).

Predation is a central tenet in ecology – predators capture, kill and consume their prey (Lima and Dill, 1990). By contrast, micropredators attack multiple hosts, may briefly feed on blood, and can influence the mortality schedules of fish through changes in physiology, morphology and behaviour (Grutter et al., 2011, 2017; Binning et al., 2013, 2014; Artim et al., 2015; Triki et al., 2016; Sellers et al., 2019). Here, we demonstrate that experimental infection by a single gnathiid has a marked influence on the fast-start escape kinematics and the routine swimming behaviour of settlement-stage ambon damselfish. For example, latency to respond when startled increased following gnathiid infection, and high latencies have been associated with lower survival (McCormick et al., 2018). In addition to latency, all locomotory behaviours, with the exception of maximum speed, were found to be reduced when compared against the control and non-parasitised groups, indicating that there was a kinematic cost associated with infection.

Fast-start escape behaviour is a measure of whole-organism performance and is influenced by intrinsic (i.e. physiological and biochemical) and extrinsic processes (i.e. habitat degradation, predation stress, temperature and oxygen) (McCormick et al., 2017; Allan et al., 2015; Domenici et al., 2019). It is the interaction between these processes that can trigger and modify how an escape is undertaken (Breed and Sanchez, 2010). Any factor that disrupts these processes can lead to increased mortality rates (Allan et al., 2013). Grutter et al. (2011) quantified the cost of infection by a single gnathiid on newly recruited ambon damselfish, using metabolic performance measured as oxygen uptake, and found infected fish had reduced performance, probably driven by blood loss. Consequently, fish infected with strongly debilitating parasites may exhibit markedly reduced activity levels to conserve energy; this may explain the observed decrease in fast-start behaviour in the current study. Infected fish may have substantially decreased energy reserves (via blood loss), thus reducing the ability to recover after eliciting an energetically costly escape. In addition, we also observed a 50% decrease in routine swimming and average speed following infection by a single gnathiid.

Our results contrast those of Binning et al. (2014), who found that the escape performance of S. bilineata was unaffected following infection by the cymothoid isopod A. nemipteri, with little difference in escape kinematics between non-infected and infected fish. However, these contrasting results may be driven by ontogeny. For example, Binning et al. (2014) used infected adult fish (∼130 mm body length) that may have a higher physiological tolerance to infection than the newly recruited fish (∼15 mm body length) used in the current study. By examining adult fish, the results may have been biased toward those individuals that could cope with infection. Those that could not cope with infection may have been removed from the population, thus underestimating the cost of infection. Moreover, the life history strategies of the parasites used in both studies are markedly different. Anilocra nemipteri remain on their host for between 12 and 16 months and may not exert a major cost to the host, owing to their dependence on host survival. By contrast, gnathiids have a larval phase consisting of three stages and associated moults during which they feed on the blood of their host before releasing from their host (Tanaka, 2007). Therefore, the fitness cost exerted on their host is much greater (i.e. 85% blood loss, sensu Grutter et al., 2011) and depends on the size of the juvenile host (Grutter et al., 2017). Given an individual parasite is large, relative to its small hosts (a 1:10 ratio of gnathiid to a newly recruited ambon damselfish), it is not surprising that we observed a significant reduction in the effectiveness of fast-start escape behaviour in the ambon damselfish as a result of infection. Aside from gnathiid and cymothoid isopods, other isopods are known to feed on blood or fluids of marine fishes, including cirolanid, coralanid and aegeid isopods (Poore and Bruce, 2012; Smit et al., 2019).

To date, few studies have explored how short-term infections with gnathiids affect coral reef fish host stress physiology (Grutter and Pankhurst, 2000; Grutter et al., 2011; Binning et al., 2014; Triki et al., 2016). We quantified total body cortisol levels following exposure to a parasite and found that infection led to nearly a 2-fold increase in cortisol levels. The effects of elevated cortisol on behaviour in fish have been well documented (Barton and Iwama, 1991). However, to the best of our knowledge, this is the first study to investigate the relationship between elevated cortisol and fast-start escape performance in fish. Increased glucocorticoids prime animals for a number of activities, including reproduction, competition and avoiding predation. Therefore, it seems likely that glucocorticoids would play an important role in fast-start escape behaviour. However, if the stressor is severe, the ability of the fish to cope may be reduced, and the overall effect of stress may become maladaptive (Barton and Iwama, 1991).

Increased cortisol may be due to either the physiological cost of infection or the discomfort caused by attachment of the parasite. For example, gnathiids were observed to be attached around the anterior region of the fish, which is often dense with nociceptors that, when stimulated, lead to quantifiable changes in neurological activity (Sneddon et al., 2014) indicative of pain. To date, the effect of parasite attachment on nociception has not been examined. However, it is possible that attachment could cause the release of cortisol via nociceptive system hormones (for review, see Galhardo and Oliveira, 2009). By contrast, it is possible that attachment could trigger an immune response with a resulting increase in cortisol. The immune system and the release of glucocorticoids are tightly coupled. Glucocorticoids have a strong anti-inflammatory effect and can induce relevant changes in immune cells as well as cytokines having the power to stimulate cortisol production (Wikel and Alarcon-Chaidez, 2001; Fulford and Harbuz, 2005). Regardless of the mechanism(s), our results suggest that short-term exposure to a gnathiid ectoparasite causes the release of cortisol. Whether the release of cortisol following attachment has long-term effects is unknown. However, this seems unlikely, given that cortisol rises quickly, within the first 4–10 min of an experienced stress, and lasts for only a few hours (Foo and Lam, 1993; Sumpter, 1997).

We found that experimentally exposing coral reef fish recruits to gnathiids negatively affected their fast-start escape performance. We also observed increased cortisol levels following infection. A loss of fitness can decrease survival during metamorphosis as fish transition from the pelagic to the benthic environment where they face myriad predators (Hoey and McCormick, 2004). Therefore, any external stressor (i.e. parasitism) that reduces condition, affects behaviour and/or alters physiology may indirectly increase mortality rates. For example, Grutter et al. (2017) examined the effect of gnathiid infection on 14 species of pre-settlement coral reef fish and found that, for small fish (<12 mm), there was significant mortality following infection by a single gnathiid. This suggests that micropredators may contribute to size-selective mortality during settlement. Moreover, parasites can interact with other ecological drivers such as habitat degradation (Sikkel et al., 2019), resulting in an increase in infection rate with potentially detrimental effects on biodiversity and ecosystem health. The early life-history stages of marine fishes are critical for the replenishment and abundance of keystone species to marine ecosystems (Almany et al., 2007). As such, any changes at this stage can compromise the integrity of adult populations.

We thank all the staff at the Lizard Island Research Station, and all the students and volunteers who helped with the light traps and sorting of fish and with the maintenance of the gnathiid culture.