By definition, parasites incur a cost to their hosts and therefore, hosts are expected to behave in such a way to minimize the chance of acquiring parasites and maximize defensive behaviours (Veitch et al. 2021). Acquisition of ectoparasitic arthropods (hereafter “parasites” unless otherwise stated) is influenced by the hunting strategy used by the parasite. While some parasites, such as mosquitoes, actively seek out their hosts, other parasites “sit and wait” in ambush. Some parasites take on yet another strategy; nidicolous ectoparasitic arthropods seek out host nests. Hence, the acquisition of “sit-and-wait” and nidicolous parasites is expected to be dependent on host habitat choice, exploratory activity, and nesting behaviour. While hosts need to be active and explore their environment to increase their chances of finding resources and potential mates, and build one or more nests for thermoregulation and reproduction, these activities likely expose them to clusters of awaiting parasites. In chipmunks (Tamias) for example, increased activity and exploration have been associated with larger parasitic burdens (Boyer et al. 2010; Bohn et al. 2017; Paquette et al. 2020). Host behaviour might influence not only the rate of encounter with parasites but also the probability of acquiring them after an encounter. Indeed, hosts actively prevent parasites from feeding by removing them with defensive behaviour such as grooming (Hart and Hart 2018). Presumably, parasites that are found on hosts have already successfully evaded the host defensive behaviour. Moreover, host behaviour might also be altered by the acquisition of parasites. Once parasites are feeding, their hosts may change their behaviour to compensate for the energetic cost of bearing parasites, such as decreasing the use of costly locomotion when highly parasitized (Finnerty et al. 2018; Hicks et al. 2018). These changes in behaviour may be one of the many factors that could make parasitized hosts more or less likely to encounter other parasites or vulnerable to other parasites. Hence, host behaviour plays a potentially important yet often overlooked role in parasite co-occurrence. White-footed mice (Peromyscus leucopus (Rafinesque, 1818); hereafter referred to as mice) and black-legged ticks (Ixodes scapularis Say, 1821) form a host–parasite system that has received considerable attention given its importance for human health (Barbour 2017). Indeed, the black-legged tick is the primary vector for Borrelia burgdorferi, the pathogen responsible for Lyme disease, and mice are one of the most important reservoirs for the maintenance of B. burgdorferi in black-legged tick populations (Brunner et al. 2008; Barbour 2017). Black-legged ticks have three active life stages (adult, nymph, and larval stages), each with different behaviour and host preferences. Larvae and nymphs feed once before molting to the next life stage. Once they molt into adults, ticks can feed multiple times to reproduce. While nymphs have no host preference, adult ticks prefer large mammals, whereas larvae mostly feed on smaller animals such as birds or rodents (Keirans et al. 1996). Tick larvae acquire B. burgdorferi from infected mice and may then transmit the bacterium to other mice (and humans) once they molt into nymphs, continuing the infection cycle. Adult black-legged ticks only actively seek out hosts within a few meters, while their larvae are even less active, making black-legged tick larvae a prime example of a “sit-and-wait” parasite (Falco and Fish 1991; Stafford 1992). Another tick species, Ixodes angustus Neumann, 1899, is potentially important to consider since it has been shown to be a competent vector of B. burgdorferi (Peavey et al. 2000). Ixodes angustus is a coastal hard tick with a nidicolous lifestyle, spending most of its lifespan within its host’s nest (Hadwen 1911). Although its lifecycle remains partially unknown, I. angustus is rarely collected from flagging, which indicates that unlike black-legged ticks, they rarely engage in questing or “sit-and-wait” behaviour on the ground (Foley et al. 2011; Stephenson et al. 2016). While it may be found on a large variety of animals, the majority of I. angustus are usually found in all life stages on a few species such as squirrels, woodrats, chipmunks, and Peromyscus spp. (Stephenson et al. 2016). Therefore, the lifestyle of I. angustus seems to be well suited for feeding on infected female hosts and then feeding on the host’s pups (Hadwen 1911), and thus it could be potentially responsible for the vertical transfer of B. burgdorferi. The role of habitat choice and activity on parasite acquisition has been relatively well studied in Peromyscus mice. It has been shown that mice with larger mixed species tick burdens have smaller home ranges, a relationship that could be mediated through activity and exploratory behaviour (Gaitan and Millien 2016). Individual P. leucopus also vary in their nesting behaviour in ways that may affect their tick acquisition, including site selection (e.g., tree hollows vs. underground burrows) and the use of multiple nesting sites (Wolff and Hurlbutt 1982). More active mice that explore their environment over larger areas (i.e., with a larger home range) may have more site options for nesting, and therefore have a lower nest fidelity. This could affect acquisition of nidicolous ticks (I. angustus), because they depend on hosts returning to the same nest after ecdysis. Ixodes angustus could also be affected by host nest habitat choice because it is often found on arboreal species in northeastern climates (Stephenson et al. 2016). While the relationship between grooming behaviour and parasitic burden has been studied in other wild animals (Eads et al. 2017; Heine et al. 2017), it has been rarely studied in mice. This is due to the difficulty in observing grooming behaviour in wild mouse populations, since mice are secretive and nocturnal. Most studies on wild mice are limited to directly observing the animals during standard manipulations upon capture (e.g., identification, sexing, weighing, parasite count, etc.) or when applying an experimental treatment (e.g., applying acaricide; Ostfeld et al. 1996; Hersh et al. 2014; Gaitan and Millien 2016; Larson et al. 2018). Although one study quantified the number of ticks removed by grooming in wild-caught individuals within a laboratory setting, it did not measure individual grooming behaviour (Keesing et al. 2003). Given that grooming behaviour is often cited as an important factor that might mitigate parasite feeding success (Keesing et al. 2003; Calabrese et al. 2011; Ostfeld et al. 2018; Mowry et al. 2019), it seems crucial to quantify individual differences in mice grooming behaviour and test whether there is any relationship with tick parasitism. Here, our goal was to test for a relationship between exploratory and grooming behaviours of mice and tick parasitism. We first use data from a capture-mark-recapture study conducted over five summers (2016–2020) in eastern Ontario (Canada) to determine the influence of various intrinsic and extrinsic factors on tick burdens in mice. Studies conducted so far on the mouse/tick system have led to mixed results about the role of host intrinsic factors (i.e., age, sex, reproductive status, and body mass) on tick burden (Moore and Wilson 2002; Halliday et al. 2014; Mowry et al. 2019; Butler et al. 2020). The number of ticks found on a mouse may also vary as a function of many extrinsic factors, including the abundance of active ticks and mice in the environment, which both fluctuates over time according to annual and seasonal variations in acorn mast production, precipitations, and temperature (Jones et al. 1998; Ostfeld et al. 2001). Competition with other parasite species may also determine tick burdens because hosts have a limited amount of space and resources that can be exploited by parasites without heavily impacting the host, and therefore parasites feeding on similar host body parts and host tissues may compete against one another (Gobbin et al. 2021). When two or more species of parasites find themselves competing or interfering with one another, they will be less likely to co-occur, whereas if one of the parasites is beneficial to another parasite, then the abundance of the second parasite will increase with the presence of the first (Lello et al. 2004, 2008). Hence, in addition to simultaneously testing for the effect of several intrinsic (sex, age, reproductive status, body mass) and extrinsic (mouse population density, precipitations, and temperature) factors on the likelihood of tick parasitism in mice, we also considered the presence of other ectoparasites (mites, fleas, and botfly larva) to verify the possibility of parasite co-occurrence.
Host behaviour should be a key determinant of the acquisition of “sit-and-wait” and nidicolous ectoparasitic arthropods that ambush the host in their environment and/or nest. Here, we tested the association between parasitism and individual differences in exploratory and grooming behaviours in white-footed mice (Peromyscus leucopus (Rafinesque, 1818)), the primary host for the black-legged tick (Ixodes scapularis Say, 1821) in its larval stage. During 4333 captures of 1035 individual mice over five summers (2016–2020) in Ontario, Canada, the probability and intensity of tick parasitism were both significantly higher when the mouse was also parasitized by fleas, suggesting co-occurrence of these two parasites on host mice. Distance moved by mice in a novel environment was negatively and positively affected by tick and flea parasitism, respectively. Interestingly, there was a significant “tick × flea” statistical interaction on distance moved, such that fleas were positively associated with activity/exploration only when ticks were absent. There was no relationship between grooming behaviour and parasite presence. This study suggests that co-occurring parasite species (ticks and fleas) may differentially affect their host's behaviour depending on the presence/absence of the other parasite on the host. Alternatively, host behaviour may differentially affect individual susceptibility to being infested with ticks, fleas, or both.
Study site The study was conducted at the Queens University Biological Research Station (QUBS) in Chaffey’s Lock, Ontario, Canada (44◦33 5N; 76◦19 27W). There were three trapping grids, all covered by deciduous forest, populated primarily with sugar maple (Acer saccharum Marsh.) and oak trees (Quercus rubra L. and Quercus alba L.). Each trapping grid consisted of a network of trapping sites and nest boxes laid out in a checkboard pattern (alternating trap and nest boxes) in a 15 m × 15 m mesh. The first trapping grid was established in 2016 and contains 110 trapping sites and 117 nest box sites. Another trapping grid was established in 2016 and contains 63 trapping sites and 61 nest box sites. A third trapping grid was established in 2019 and contains 50 trapping sites and 20 nest box sites. Finally, mice were also sometimes captured outside the systematic grids using traps that were set at specific locations around the laboratory (see below), where the habitat is overall similar to the rest of the study area. Captures Trapping occurred from early May to late August, using Longworth traps set at sundown and checked at sunrise. Each night, traps were baited with sunflower seeds and diced apples, with cotton batting added for nesting material. The number of traps set overnight in 2016, 2017, 2018, 2019, and 2020 were 5337, 6258, 4723, 4537, and 3920, respectively. Grid sampling alternated, such that grids were usually not sampled more than two nights in a row. Nest boxes were checked every two weeks from April to October, weather permitting. Newly captured mice had permanent ear tags installed for individual identification. For each capture, the following information was noted: sex, age, body mass, and reproductive status. Sex was determined based on the distance between the genitals and anus, with males having a larger distance than females. Age was determined based on pelage colour: while adults were completely brown above the middorsal molt line, juveniles exhibit brown and grey pelage above the molt line (Collins 1923). Body mass was measured with a 50 g Pesola (±0.3%). Reproductive status was categorized as active or not depending on external sexual characteristics (i.e., males with an enlarged scrotum were considered reproductively active, and females were considered lactating when their mammae were enlarged and clearly visible due to lack of surrounding hair). Each mouse captured was checked for parasites (mites, botflies, fleas, and ticks) following a standardized protocol that consisted of the close inspection of the ears, face, tail, shoulders, toes, genitals, and by softly blowing through the fur on their backs and bellies. Mites, botflies, and fleas were simply counted and not removed nor identified to the species. Ticks were not removed from heavily pregnant mice to prevent additional handling stress. All other ticks found on mice that could be removed in a timely manner were collected with fine tweezers and preserved in 90% alcohol for future studies. Tick removal in our study implies that every tick counted on a mouse is a different tick (otherwise, there would be no way of knowing how many ticks are newly acquired by a given mouse in between two captures). A total of 112 ticks collected over 3 years (2016–2018) were identified; 37.5% were I. scapularis, 1% were Dermacentor variabilis (Say, 1821), and 61.5% were I. angustus. Although we used a relatively standard parasite checking/counting protocol, we cannot exclude the possibility that annual differences in parasite counts partly reflect changes in the composition of the fieldwork team. However, of a total of 17 different handlers that counted parasites, the 5 handlers that were present in multiple years collectively gathered 49.4% of the observations. After the standard manipulation procedure was completed at the capture site, all mice caught in a nest box were placed back in and left in their nest box. Individuals caught in a trap were either released or placed back in their trap and transported to a laboratory for behavioural measurement. Up to 16 individuals per day were brought to the laboratory, prioritizing mice that were caught less often, and avoiding testing individuals that had been tested within 7 days prior. Moreover, mice that were below 14 g, lactating, or pregnant were released. All mice were released on site. Behavioural tests In the laboratory, mice were kept in their trap until it was their turn for behavioural testing, which consisted of a 10 min open-field test. The test involved placing a mouse in a circular arena, which consisted of a white plastic tank with 1.4 m in diameter and 86 cm walls to prevent animals from escaping. The arena was surrounded by black curtains to prevent any disturbance from the surroundings and standardize the environment. The mouse to be tested was held by the scruff and tail and was then placed into the arena along the wall, facing away from the handler. Noise was kept to a minimum while the test was running. The arena was cleaned before and after each test with Accel and then rinsed with water. Each test was recorded using a camera (Basler ace IR) positioned above the arena and connected to the EthovisionXT software. Tracking immediately began when the mouse was detected inside the arena, and continued for 10 min at which point the test ended. During the whole test, an observer watched the video recording on the screen, manually documenting the beginning and end of all events of grooming directly in EthovisionXT. Tracking files were analyzed in EthovisionXT to extract the total distance moved and total time spent grooming while in the arena. The animal was then weighed using a precision scale (Mettler Toledo, Model ML1602T/00). Ticks found during these manipulations were also removed. Once all open-field tests were completed, mice were replaced in their trap (with a small amount of seeds and peanut butter), transported back to their capture site, and then released. Given that we worked on wild animals captured overnight, it was impossible to standardize the conditions prior to behavioural testing. For example, the time elapsed between field manipulations and behavioural testing varied according to various factors such as the number of mice captured that day and the order in which mice were manipulated in the field and tested in the laboratory. On average, mice spent 150 min in the trap between capture and behavioural testing, but this varied between 14 and 413 min. Statistical analysis The statistical analysis was conducted in R version 4.03, and involved two models. The first model was based on all capture data and included the number of ticks counted on each capture as the dependent variable. The glmmTMB package (Brooks et al. 2017) was used to fit a hurdle mixed model to simultaneously estimate the effects of multiple extrinsic and intrinsic factors on tick burdens. Hurdle models are usually used to account for zero inflation, and include twocomponent models: a binary model that treats the data as zeroes or nonzeroes, which accounts for zero inflation, and a zero-truncated negative binomial distributed model fitting only the nonzero values (Zuur et al. 2009). A hurdle model is therefore useful to get an approximation of the factors affecting parasite prevalence (with the binary model) and intensity (with the binomial model) of tick parasitism in the same model, while still being able to interpret either separately from the other (Markle et al. 2020; Obiegala et al. 2021). Note that the binary model is predicting the odds of obtaining a zero; thus, negative estimates represent an increase in the odds of ticks being present on mice. Two hurdle models were fitted: one with a Poisson distribution and another with negative binomial distribution and the latter was retained because it had a higher likelihood than the former (Zuur et al. 2009). The two-component models had the same set of independent variables, which included extrinsic variables such as yearly mouse density (high or low), month, daily temperature, daily mouse capture numbers, and daily precipitation, and intrinsic variables such as body mass, sex, age, and reproductive status. The model also included the abundance of other parasites found on the mouse (i.e., number of fleas, botfly larvae, or mites). March and April only had two and five captures with ticks, respectively; both months were therefore combined under April. We also included a dummy variable coding for whether or not the individual had been previously captured within the last 3 days to take into account the removal of ticks from the mice. This period was selected because the average feeding time for I. angustus is 2.5 days and 3 days is the minimum amount of time necessary for larval black-legged ticks to feed and drop off from their host (Hadwen 1911; Nuss et al. 2017) (note: we also conducted a sensitivity analysis, see below). All continuous variables (temperature, precipitation, body mass, daily mouse capture numbers, and flea/botfly/mite abundance) were scaled to a mean of 0 and a variance of 1. Categorical factors with two categories were treated as centered continuous variables (i.e., captured 3 days prior: no = −0.5, yes = 0.5; mouse density for the year: low = −0.5, high = 0.5; sex: female = −0.5, male = 0.5; age: juvenile = −0.5, adult = 0.5; reproductive status: inactive = −0.5, active = 0.5). Daily temperature and precipitation were taken from the nearby Lyndhurst Shawmere weather station, with missing data filled with data from the Grenadier weather station (data from Government of Canada’s past weather and climate historical data). Three random effects were included in the model: year, handler identity, and mouse identity. Year was included to take into account any differences between years (e.g., the start of the trapping season was delayed by a month in 2020). The identity of the handler was also included as a random effect to account for possible differences in observation between handlers for parasite counts. Since many individual mice were repeatedly captured, mouse identity was included as a random effect to account for this source of nonindependence and quantify among-individual variance in tick parasitism. We used the DHARMa package to access goodness of fit using a quantile–quantile plot of simulated residuals (Hartig 2020) and confirm that overdispersion and zero inflation were not significant. Individual repeatability was calculated using the rptBinary package only for the binary model (there is no specific function for zero-truncated negative binomial distribution; Nakagawa and Schielzeth 2010). The trigamma function from the r.squaredGLMM package was used to calculate marginal and conditional pseudo R2 for the binary model (Nakagawa and Schielzeth 2013). The marginal pseudo R2 only includes the fixed effects, while the conditional pseudo R2 also includes the random effects in its calculation. Both are summaries of the variance explained in a model. For the binomial model component, no pseudo R2 could be found since it is currently unknown how to calculate the index within zero-truncated negative binomial distributions. Given that ticks were removed at each capture, and that a large portion of our dataset included recaptures, we conducted a sensitivity analysis to get an approximation of how long it takes for a mouse to reacquire ticks. We fitted the same model as above, but each time with a different dummy variable coding for whether or not the individual had been previously captured within the last 1–20 days (instead of 3 days as in the initial model). The second part of the analysis was restricted to a subset of captures for which the mouse was brought back to the laboratory for behavioural testing. Two separate linear mixed models were run using the lme4 package in R 4.03, with distance moved or time spent grooming during the open-field test as the dependent variable. Distance moved and time spent grooming were square root-transformed to reach normality of residuals. After transformation, both variables were scaled to a mean of 0 and a variance of 1. The fixed effects were mouse density of the year (high/low), month, test sequence, time of day, daily mouse capture numbers, time spent in the trap between field manipulations and testing, mouse mass, sex, age, reproductive status, and tick/flea/botfly/mite presence. Because results from the first part of the analysis suggested that ticks and fleas co-occur (see below), we also included a “tick × flea” statistical interaction. Mass, test sequence, and time of day were scaled to a mean of zero. The presence of each parasite was centered by setting absence/presence to −0.5/0.5. Mouse identity and year were included as random effects.