Host social behaviours alter parasite ecology (Arrow A)

Parasites spread via close contact between conspecifics over time or space (which we term ‘socially transmitted parasites’ hereafter for simplicity; Box 1) are hypothesized to pose a greater risk for host species that exhibit social behaviours such as group living (Krause and Ruxton, Reference Krause and Ruxton2002). Classic mathematical models for socially transmitted parasites [e.g. susceptible-infectious-recovered (SIR) compartmental models] often assume that the rate of contact between susceptible and infectious individuals increases with host density (Begon et al., Reference Begon, Bennett, Bowers, French, Hazel and Turner2002). On a local scale, this results in higher contact rates, and thus parasite transmission, for animals in larger social groups. Indeed, two meta-analyses support the hypothesis that larger social groups generally harbour higher prevalence and/or infection intensity (Box 1) of parasites spanning diverse transmission modes (Rifkin et al., Reference Rifkin, Nunn and Garamszegi2012; Patterson and Ruckstuhl, Reference Patterson and Ruckstuhl2013). In contrast, however, there is some evidence that group living can dilute host risk of infection with highly mobile parasites by reducing per capita attack rates (the encounter-dilution effect; Côté and Poulin, Reference Côté and Poulin1995). The encounter-dilution effect primarily applies to parasites that actively seek hosts by flying or swimming; the likelihood of being singled out by these parasites can decrease with increasing group size (Côté and Poulin, Reference Côté and Poulin1995; Patterson and Ruckstuhl, Reference Patterson and Ruckstuhl2013).

Recent work suggests that social group substructure may in some cases be equally or more important than group size in predicting parasite risk (Griffin and Nunn, Reference Griffin and Nunn2012; Nunn et al., Reference Nunn, Jordan, McCabe, Verdolin and Fewell2015; Sah et al., Reference Sah, Mann and Bansal2018). If the majority of close social interactions in large groups occur between subsets of individuals (e.g. ‘cliques’), this modularity (Box 1) can act as a ‘social bottleneck’ that contains parasite spread within subgroups and reduces spread to the group at large (e.g. Nunn et al., Reference Nunn, Jordan, McCabe, Verdolin and Fewell2015). In support of this idea, the social networks of eusocial insect colonies can be highly structurally subdivided, and epidemiological models show that this constitutive modularity dampens the transmission of an entomopathogenic fungus within colonies (Stroeymeyt et al., Reference Stroeymeyt, Grasse, Crespi, Mersch, Cremer and Keller2018). Similarly, a comparative study of 19 non-human primate species found that higher levels of modularity may help ameliorate the heightened risk of parasite spread in large social groups, as higher modularity was associated with lower parasite richness (Griffin and Nunn, Reference Griffin and Nunn2012). However, perhaps because of its protective function, social group modularity tends to increase with group size across taxa (Nunn et al., Reference Nunn, Jordan, McCabe, Verdolin and Fewell2015), making it challenging to tease apart whether resulting patterns of parasitism are a function of group size, modularity or both.

Individual variation in social behaviours can also have important effects on transmission risk. As shown through descriptive network approaches that quantify social connections among conspecifics using direct behavioural interactions or physical proximity, individuals that have ties to multiple social ‘cliques’ (VanderWaal et al., Reference VanderWaal, Obanda, Omondi, McCowan, Wang, Fushing and Isbell2016) or those highly connected to neighbouring conspecifics (e.g. Bull et al., Reference Bull, Godfrey and Gordon2012) can have an increased likelihood of parasite infection (but see Drewe, Reference Drewe2010 for the importance of type and directionality of interactions). Similarly, bold or ‘pro-active’ personality traits, which correlate with social network centrality in some taxa (e.g. Aplin et al., Reference Aplin, Farine, Morand-Ferron, Cole, Cockburn and Sheldon2013), may influence social parasite transmission: two studies of mammalian species found that bolder individuals had a higher seroprevalence of viruses largely spread via aggressive interactions (Natoli et al., Reference Natoli, Say, Cafazzo, Bonanni, Schmid and Pontier2005; Dizney and Dearing, Reference Dizney and Dearing2013). While these correlational studies suggest the effects of variation in social behaviour on parasite risk, field studies generally cannot directly elucidate cause and effect (Arrow A vs B: does behaviour affect parasites or vice versa?). Further, it is challenging to disentangle the relative contributions of individual variation in exposure vs susceptibility to field patterns of transmission [VanderWaal and Ezenwa, Reference VanderWaal and Ezenwa2016; see ‘Synthesis: ecological feedbacks between social behaviours and parasite infection’ section], particularly when traits relevant for both exposure and susceptibility can simultaneously be influenced by social context (e.g. Müller-Klein et al., Reference Müller-Klein, Heistermann, Strube, Franz, Schülke and Ostner2019). Experimental studies, while not possible for all host–parasite systems, can isolate the effects of host social behaviour per se on parasite transmission risk. For example, Keiser et al. (Reference Keiser, Pinter-Wollman, Augustine, Ziemba, Hao, Lawrence and Pruitt2016) used experimental epidemics to show that bolder female social spiders (Stegodyphus dumicola) had a higher risk of acquiring a cuticular microbe. Future studies could examine how individual differences in ‘social personalities’, which are seldom quantified in themselves (e.g. Kulahci et al., Reference Kulahci, Ghazanfar and Rubenstein2018), influence the transmission dynamics of socially transmitted parasites.

Overall, the social behaviours of groups and individuals appear to strongly influence parasite transmission risk (Arrow A). However, in order to fully elucidate the effects of social behaviours on parasite transmission, it is critical to also consider how parasite infection affects host social behaviours (Arrow B), as both processes together will ultimately underlie the dynamics of socially transmitted parasites.

Parasite infection influences host social behaviours (Arrow B)

The way in which parasite infection alters the social behaviours of both infected hosts and their uninfected conspecifics (Arrow B) has received relatively less attention than the effects of social behaviours on parasite risk (Arrow A) ]. This is somewhat surprising given that it has long been recognized that hosts often behave differently during infection (reviewed in Moore, Reference Moore2002). Changes in social behaviours during infection can broadly result from parasite-mediated manipulation of host behaviours to promote transmission to new hosts (reviewed in Klein, Reference Klein2003), or from host-mediated behavioural changes, which typically occur from one of three mechanisms: (1) as side-effects of tissue damage or energy needs associated with infection, (2) via expression of ‘sickness behaviours’ that are part of a host's broader, adaptive immunological responses to infection (Hart, Reference Hart1988), or (3) as active self-isolation to prevent ongoing spread, a behaviour largely seen in eusocial insects (Shorter and Rueppell, Reference Shorter and Rueppell2012). All four possibilities, whether parasite- or host-mediated, can lead to notable changes in social behaviours of hosts, with important consequences for parasite transmission. For example, three-spined sticklebacks (Gasterosteus aculeatus) infected with the socially transmitted parasite Glugea anomala are more likely than their uninfected counterparts to be attracted to conspecifics, a behaviour predicted to augment transmission (Petkova et al., Reference Petkova, Abbey-Lee and Løvlie2018). Whether behavioural changes in that system are parasite- or host-mediated remains unclear, but in this section, we focus on changes in behaviour during infection that are likely host-mediated, and consider parasite-mediated behavioural changes in ‘Host social behaviours influence parasite evolution (Arrow D)’ section.

Host-mediated changes in behaviour during infection, such as self-isolation and sickness behaviours, often reduce the degree of interaction with conspecifics and thus the spread of socially transmitted parasites. While active self-isolation is rare outside of eusocial insects, sickness behaviours are a conserved component of vertebrate immune responses that include general reductions in activity levels and specific reductions in non-essential activities (Hart, Reference Hart1988), such as many forms of social interaction (e.g. allogrooming). For example, Lopes et al. (Reference Lopes, Block and König2016) stimulated sickness behaviours in wild house mice (Mus musculus domesticus) by injecting individuals with bacterial endotoxin, and found that immune activation resulted in lower activity levels and fewer direct social interactions with conspecifics relative to controls. Similarly, work in two other mammalian systems found that infected individuals (or those expressing sickness behaviours) are less likely than control individuals to engage in affiliative allogrooming with conspecifics [banded mongooses (Mungos mungo), Fairbanks et al., Reference Fairbanks, Hawley and Alexander2014; vampire bats (Desmodus rotundus), Stockmaier et al., Reference Stockmaier, Bolnick, Page and Carter2018]. In vampire bats, these changes in allogrooming during sickness behaviour expression, potentially in combination with reduced contact calling (Stockmaier et al., Reference Stockmaier, Bolnick, Page, Josic and Carter2020a), result in significant reductions in several measures of social connectedness relative to controls (Ripperger et al., Reference Ripperger, Stockmaier and Carter2020). Overall, host-mediated reductions in social interactions during infection, particularly when they occur during the host's infectious period, likely reduce transmission of socially transmitted parasites.

The extent to which infected hosts alter their social behaviour is likely to depend on the energetic costs of a given parasite infection and the importance of that social behaviour for maintaining host fitness (Ezenwa et al., Reference Ezenwa, Ghai, McKay and Williams2016b). In some systems, social behaviours of hosts appear to be maintained during infection (Powell et al., Reference Powell, Wallen, Miketa, Krzyszczyk, Foroughirad, Bansal and Mann2020), which may be common for infections by low-virulence parasites. In other cases, infected animals may maintain a subset of social interactions potentially most important to host recovery, including those with high inclusive fitness benefits. For example, vampire bats injected with endotoxin to induce sickness behaviours continued to groom close kin (mother or offspring) at levels similar to controls, but reduced the extent to which they groomed non-kin (Stockmaier et al., Reference Stockmaier, Bolnick, Page and Carter2020b). In some systems, social behaviours of hosts can even be augmented during infection. For example, male guppies (Poecilia reticulata) with high loads of a socially transmitted ectoparasite showed higher sociality relative to males with lower parasite loads (Stephenson, Reference Stephenson2019), and rhesus monkeys (Macaca mulatta) given low-dose endotoxin injection show marked increases in social behaviours with conspecifics (Willette et al., Reference Willette, Lubach and Coe2007). The ultimate mechanisms underlying these patterns remain unknown, but in some systems, the maintenance or even augmentation of sociality during infection may be a form of tolerance (Box 1), allowing hosts to minimize the fitness impacts of infection via group living (Ezenwa et al., Reference Ezenwa, Ghai, McKay and Williams2016b). For example, recent work in Grant's gazelle (Nanger granti) suggests that association with larger groups benefits gazelle infected with gastrointestinal parasites by allowing them to better ameliorate the costs associated with infection-induced anorexia (Ezenwa and Worsley-Tonks, Reference Ezenwa and Worsley-Tonks2018). Given that infected hosts experience anorexia (e.g. Adelman et al., Reference Adelman, Kirkpatrick, Grodio and Hawley2013) and higher predation risk (e.g. Alzaga et al., Reference Alzaga, Vicente, Villanua, Acevedo, Casas and Gortazar2008; Stephenson et al., Reference Stephenson, Kinsella, Cable and van Oosterhout2016) in many social taxa, future work should examine whether enhanced gregariousness during infection is a common mechanism of tolerance across taxa, with important consequences for ecological feedbacks between social behaviour and parasite transmission.

Parasite infection can also alter social interactions by changing the behaviour of uninfected hosts towards their infected conspecifics. Among taxa spanning fish, birds, crustaceans, social insects and mammals, infected or immune-activated individuals display visual cues of infection (e.g. lethargy: Zylberberg et al., Reference Zylberberg, Klasing and Hahn2012) or release distinct chemical cues that conspecifics can use to avoid them (e.g. Arakawa et al., Reference Arakawa, Arakawa and Deak2009; Anderson and Behringer, Reference Anderson and Behringer2013; Stephenson and Reynolds, Reference Stephenson and Reynolds2016) or, in the case of honey bees (Apis mellifera), remove them from the colony (Baracchi et al., Reference Baracchi, Fadda and Turillazzi2012). Intriguingly, recent work in mice suggests that the scent of uninfected hosts themselves can change when they are housed with an infected conspecific (Gervasi et al., Reference Gervasi, Opiekun, Martin, Beauchamp and Kimball2018), suggesting the potential for complex downstream effects of infection status on social group dynamics and resulting transmission.

In some highly social animals, uninfected groupmates continue to engage in intimate interactions such as allogrooming with conspecifics that are infected or expressing sickness behaviours. At the extreme are some eusocial insects, where individuals care for infected conspecifics, likely because their high degree of relatedness favours the evolution of seemingly ‘altruistic’ behaviours via kin selection [see ‘Parasites and the evolution of host social behaviour (Arrow C)’ section]. But even in systems where groupmates are not as closely related, uninfected individuals often maintain intimate social interactions with infected conspecifics. For example, uninfected conspecifics in two social mammals groom visibly diseased groupmates or those expressing sickness behaviours at a similar intensity to controls, even when allogrooming reciprocity from these individuals is greatly reduced (e.g. mongooses: Fairbanks et al., Reference Fairbanks, Hawley and Alexander2014; vampire bats: Stockmaier et al., Reference Stockmaier, Bolnick, Page and Carter2018); furthermore, uninfected vampire bats continue to share food with conspecifics expressing sickness behaviours (Stockmaier et al., Reference Stockmaier, Bolnick, Page and Carter2020b). In mandrills (Mandrillus sphinx), the degree to which uninfected individuals maintain social interactions with infected conspecifics appears to depend on kinship: mandrills reduce grooming towards parasitized partners that are non-kin, but maintain grooming if these potentially contagious partners are offspring or close maternal kin (Poirotte and Charpentier, Reference Poirotte and Charpentier2020). Finally, in other systems, uninfected conspecifics are attracted to feed near (male house finches, Haemorhous mexicanus: Bouwman and Hawley, Reference Bouwman and Hawley2010) or socially explore (mice: Edwards, Reference Edwards1988) infected conspecifics. Understanding heterogeneity in the behaviour of uninfected hosts towards infected conspecifics (Fig. 2B), which can vary from avoidance to attraction, will help predict the conditions in which parasite-induced changes in sociality lead to positive or negative ecological feedbacks that ultimately maintain or dampen parasite epidemics (Fig. 1).

The effects of infection on social interactions between groups are also key to understanding pathogen transmission dynamics (Cross et al., Reference Cross, Lloyd-Smith, Johnson and Getz2005), but have generally received less attention than within-group social interactions. Because infected individuals or those expressing sickness behaviours are less likely to explore their surroundings than uninfected individuals (e.g. Lopes et al., Reference Lopes, Block and König2016), they may be less likely to interact with other social groups, either temporarily or permanently (as occurs in banded mongooses; Fairbanks et al., Reference Fairbanks, Hawley and Alexander2014). In other cases, infected individuals may be more likely to leave an existing group, as has been observed among European badgers (Meles meles meles) with bovine tuberculosis (Cheesman and Mallinson, Reference Cheesman and Mallinson1981; Weber et al., Reference Weber, Bearhop, Dall, Delahay, McDonald and Carter2013). Whether infected individuals join new social groups, either temporarily or permanently, will also depend on whether infected individuals are ‘accepted’ by conspecifics in the new social group (Butler and Roper, Reference Butler and Roper1996). Uninfected guppies appear to largely prevent the integration of experimental intruders with ectoparasite infections into existing shoals (Croft et al., Reference Croft, Edenbrow, Darden, Ramnarine, van Oosterhout and Cable2011). In contrast, honey bee colonies were more likely to accept entry by foreign bees infected with Israeli acute paralysis virus than foreign controls, which may represent a unique case of pathogen manipulation of chemical signals that mediate aggressive interactions in this species [Geffre et al., Reference Geffre, Gernat, Harwood, Jones, Morselli Gysi, Hamilton, Bonning, Toth, Robinson and Dolezal2020; see ‘Host social behaviours influence parasite evolution (Arrow D)’ section]. The movement or dispersal of uninfected individuals between groups can also be driven by conspecific infection or disease status, as occurs in western lowland gorillas (Gorilla gorilla gorilla), where adult females are more likely to emigrate from social groups with a higher prevalence of facial lesions associated with a contact-transmitted skin disease (Baudouin et al., Reference Baudouin, Gatti, Levréro, Genton, Cristescu, Billy, Motsch, Pierre, Le Gouar and Ménard2019). Overall, more studies are needed on how parasite infection influences among-group movements for both infected hosts and uninfected conspecifics, particularly for taxa where social group composition is relatively fluid, such as fission–fusion societies.

Studies have only recently begun to address how changes in social behaviours of both infected and uninfected conspecifics scale up to influence host social networks and disease dynamics. Chapman et al. (Reference Chapman, Friant, Godfrey, Liu, Sakar, Schoof, Sengupta, Twinomugisha, Valenta and Goldberg2016), for example, used a deworming approach to examine how parasite infection in vervet monkeys (Chlorocebus pygerythrus) influenced social interactions in ways relevant to population-level spread. Dewormed individuals (particularly juveniles) had more frequent social interactions with more total conspecifics, suggesting that uninfected individuals may generally be more central in vervet monkey social networks, thereby attenuating parasite spread. Likewise, two recent studies combined experimental manipulations of infection status or sickness behaviour with network modelling to examine how parasite infection might influence the dynamics of socially transmitted pathogens (Lopes et al., Reference Lopes, Block and König2016; Stroeymeyt et al., Reference Stroeymeyt, Grasse, Crespi, Mersch, Cremer and Keller2018). Lopes et al. (Reference Lopes, Block and König2016) used empirical contact data from mice induced to express sickness behaviours to simulate disease outbreaks across social networks, showing that changes in social interactions associated with sickness behaviours resulted in highly attenuated disease outbreaks. Although Lopes et al. (Reference Lopes, Block and König2016) did not find evidence of conspecific avoidance in their system, recent work in Lasius niger ants showed that responses of both parasite-contaminated ants and their uncontaminated nestmates contributed together to changes in group social networks that inhibited the spread of pathogens through colonies (Stroeymeyt et al., Reference Stroeymeyt, Grasse, Crespi, Mersch, Cremer and Keller2018). Thus, understanding the behaviour of both infected hosts and the uninfected conspecifics they interact with is key for elucidating ecological feedbacks that dampen or augment disease spread within and among social groups.

Synthesis: ecological feedbacks between social behaviours and parasite infection

The bidirectional feedbacks between host social behaviours and parasite infection make it challenging to determine whether ecological patterns such as group size–parasitism relationships [‘Host social behaviours alter parasite ecology (Arrow A)’ section] result from the effect of social interactions on parasite risk (Arrow A), the effect of infection on social behaviours (Arrow B) or both. Experimental manipulation of parasite infection allows direct elucidation of causality. For example, Ezenwa and Worsley-Tonks (Reference Ezenwa and Worsley-Tonks2018) treated a subset of Grant's gazelles with anti-helminthic drugs and found that individuals in larger social groups re-acquired gastrointestinal parasites more rapidly, supporting the idea that larger group sizes augment the risk of acquiring parasites (Arrow A). Because they also found that parasitized gazelle benefit from larger group sizes where they can spend more time foraging [see ‘Parasite infection influences host social behaviours (Arrow B)’ section], parasitized Grant's gazelles may actively seek out larger social groups (Arrow B), further contributing to patterns of higher parasite prevalence in larger groups. Although such changes in sociality with parasitism have not yet been explicitly examined in this system, the ability of gregariousness to augment host tolerance of infection may produce positive feedbacks between infection and social behaviour, facilitating longer persistence of parasite loads in larger groups.

The strength of ecological feedbacks between social behaviour and infection will be influenced by the degree of heterogeneity in the behaviour of both infected and uninfected hosts (Fig. 2), as well as the way in which behavioural heterogeneity covaries with physiological resistance to parasites. Recent studies reveal that individual variation in social behaviour among uninfected individuals often covaries with their susceptibility to infection (Fig. 2B), a pattern with unknown causality but hypothesized to result from hosts balancing their investment in behavioural vs physiological immunity. Individual hosts with less effective physiological defences against parasites appear to avoid behaviours entailing high infection risk (Barber and Dingemanse, Reference Barber and Dingemanse2010): mice (Filiano et al., Reference Filiano, Xu, Tustison, Marsh, Baker, Smirnov, Overall, Gadani, Turner, Weng, Peerzade, Chen, Lee, Scott, Beenhakker, Litvak and Kipnis2016) and zebrafish (Danio rerio; Kirsten et al., Reference Kirsten, Fior, Kreutz and Barcellos2018) that express lower levels of interferon γ (and are therefore potentially more susceptible to intracellular parasites) are less social, and house finches with lower levels of circulating immune proteins more strongly avoid conspecifics expressing sickness behaviours (Zylberberg et al., Reference Zylberberg, Klasing and Hahn2012). Stephenson (Reference Stephenson2019) built on these findings by demonstrating that the pattern is similar, with the most susceptible individuals showing strongest conspecific avoidance, when considering susceptibility to the most prevalent parasites in an animal's environment, rather than a general immune component. Intraspecific variation in parasite susceptibility can therefore covary with intraspecific variation in behaviour, leading to potential dampening of ecological feedbacks, and reduced epidemic potential, if individuals that are the most social are also least likely to acquire infection (Hawley et al., Reference Hawley, Etienne, Ezenwa and Jolles2011).

Once transmission occurs, behavioural changes of parasite-contaminated or actively infected hosts are also heterogeneous (Fig. 2A). Factors extrinsic to the host, such as social context (Lopes, Reference Lopes2014) and seasonality (Owen-Ashley and Wingfield, Reference Owen-Ashley and Wingfield2006), as well as factors intrinsic to the host, such as sex (Silk et al., Reference Silk, Weber, Steward, Hodgson, Boots, Croft, Delahay and McDonald2018; Stephenson, Reference Stephenson2019), social caste (Stroeymeyt et al., Reference Stroeymeyt, Grasse, Crespi, Mersch, Cremer and Keller2018) and previous exposure to the parasite (Walker and Hughes, Reference Walker and Hughes2009), can dramatically affect behavioural changes in response to infection. Additionally, behavioural changes of infected animals often positively covary with infection intensity (Edwards, Reference Edwards1988; Houde and Torio, Reference Houde and Torio1992; Barber and Dingemanse, Reference Barber and Dingemanse2010), which is naturally highly variable in host populations (Shaw et al., Reference Shaw, Grenfell and Dobson1998). Thus, hosts that harbour the highest infection intensity (a potential proxy for infectiousness) are also typically the ones most likely to alter their social behaviours (and thus contact rates) in ways that result in ecological feedbacks relevant for parasite transmission. Hawley et al. (Reference Hawley, Etienne, Ezenwa and Jolles2011) used an SIR model to show that positive covariation among individuals between their infectiousness and contact rate, whereby the most heavily infected individuals are the most social, can lead to rapid epidemic spread. Recent work demonstrating that infected animals can benefit from living in groups (Almberg et al., Reference Almberg, Cross, Dobson, Smith, Metz, Stahler and Hudson2015; Ezenwa and Worsley-Tonks, Reference Ezenwa and Worsley-Tonks2018) suggests that this positive covariation may occur broadly in systems where animals use social behaviour to increase tolerance. Conversely, when the most infectious individuals elicit the strongest avoidance in uninfected conspecifics (e.g. in guppies: Stephenson et al., Reference Stephenson, Perkins and Cable2018), this negative covariation can lead to rapid fade-out of a parasite from a host population. Experimental probing of individual-level relationships (e.g. Stephenson, Reference Stephenson2019) will ultimately allow a better understanding of the potential ecological feedbacks that arise from bidirectional relationships between social behaviour and parasite infection, and the way in which these feedbacks are influenced by sources of heterogeneity both intrinsic and extrinsic to hosts (Fig. 2; Hawley et al., Reference Hawley, Etienne, Ezenwa and Jolles2011; VanderWaal and Ezenwa, Reference VanderWaal and Ezenwa2016; White et al., Reference White, Forester and Craft2018).

Parasites and the evolution of host social behaviour (Arrow C)

Akin to parasite-induced changes in social behaviour via ecological processes (Arrow B), the social behaviours of both infected and uninfected individuals can evolve in response to parasites (Townsend et al., Reference Townsend, Hawley, Stephenson and Williams2020). Here, we focus on evolutionary changes in the social behaviours of uninfected hosts that are likely to reduce the fitness costs imposed by their socially transmitted parasites. These include reductions in overall individual gregariousness (mechanism 1) that manifest as lower average group sizes for group-living taxa; reductions in social interactions with some but not all conspecifics (mechanism 2), which often manifest as increases in modularity; and reductions or augmentation in specific social behaviours that either increase or decrease parasite risk, respectively (mechanism 3). While these three mechanisms involve fixed phenotypic changes in social behaviours in response to parasite-mediated selection, the costs associated with reduced sociality for many taxa may favour the evolution of conspecific avoidance only in the presence of specific cues of infection (mechanism 4; Amoroso and Antonovics, Reference Amoroso and Antonovics2020; Townsend et al., Reference Townsend, Hawley, Stephenson and Williams2020). We briefly explore each of these four mechanisms and discuss constraints associated with evolving phenotypic changes in social behaviours in the face of parasites.

Host social behaviours influence parasite evolution (Arrow D)

For socially transmitted parasites, host social behaviours shape transmission opportunities (Arrow A), which in turn determine a parasite's population structure and evolutionary dynamics. The relatively short generation time of parasites means that host social behaviours may lead to genetic changes in parasite populations within just one or a few host generations. Here, we consider the influence of host social behaviours on (1) fundamental population genetic processes and (2) adaptive evolution of parasites. Our scope of social behaviours includes a diversity of host interactions (Box 1) that may have distinct effects on parasite evolution (Schmid-Hempel, Reference Schmid-Hempel2017). We focus on social behaviours that change the size and connectivity of host groups, with a brief consideration of behaviours that might change host relatedness.

We first consider the role of host behaviour in shaping the population genetics of parasites and thereby their potential to respond to selection. Increases in the size and connectivity of host social groups can decrease parasite population structure, increase gene flow and promote genetic diversity, leading to overall increases in the effective size of parasite populations. This prediction applies particularly when parasite prevalence increases with host group size, and when transmission opportunities increase with host connectivity. Because larger host groups often maintain larger parasite populations [see Host social behaviours alter parasite ecology (Arrow A) section; Rifkin et al., Reference Rifkin, Nunn and Garamszegi2012; Patterson and Ruckstuhl, Reference Patterson and Ruckstuhl2013], host social grouping can contribute to the maintenance of parasite genetic diversity at neutral loci and loci under selection by limiting the probability of stochastic extinction of parasite populations (Barrett et al., Reference Barrett, Thrall, Burdon and Linde2008). In addition, connectivity of social groups can increase connectivity of groups of parasites (i.e. demes) if parasite transmission increases alongside direct contacts of hosts. Increased connectivity means increased gene flow and reduced genetic differentiation between parasite groups, both at the level of host individual and population (e.g. Nadler et al., Reference Nadler, Hafner, Hafner and Hafner1990). In a test of these predictions, Van Schaik et al. (Reference van Schaik, Kerth, Bruyndonckx and Christe2014) compared the parasites of greater mouse-eared bats (Myotis myotis) and Bechstein's bats (M. bechsteinii), congeners which differ in their social system: maternal colonies of M. myotis mix readily, and individuals hibernate in large clusters, mate in harems and migrate relatively long distances, while maternal colonies of M. bechsteinii never mix, and individuals hibernate alone, meet briefly during mating and migrate relatively short distances. Their respective Spinturnix wing mite species differ accordingly in their population genetic structure: nuclear genetic diversity of S. myoti is very high, with little genetic differentiation between mites in different bat colonies, while nuclear genetic diversity of S. bechsteini is lower, with marked differentiation between colonies, suggesting strong genetic drift in small, isolated mite populations. This work demonstrates that larger, more connected social groups host parasite populations that are more genetically diverse.

Increasing host connectivity can also reduce parasite aggregation, with parasites more uniformly distributed rather than clumped on a subset of hosts. Reducing parasite aggregation lowers within-host competition and variance in reproductive success, increasing effective population size for parasites (Whitlock and Barton, Reference Whitlock and Barton1997; Poulin, Reference Poulin2007). Empirical data support reduced aggregation for ectoparasites with increased host sociality: comparative studies show reduced aggregation of lice in colonial bird species relative to territorial species (Rózsa et al., Reference Rózsa, Rékási and Reiczigel1996; Rékási et al., Reference Rékási, Rózsa and Kiss1997) and in large vs small social groups of Galapagos hawks for amblyceran lice (Buteo galapagoensis; Whiteman and Parker, Reference Whiteman and Parker2004). Taking these processes of parasite connectivity and aggregation together, we generally expect increases in the size and connectivity of host social groups to decrease the effects of genetic drift and promote responses to selection in parasite populations (reviewed in Nadler, Reference Nadler1995; Barrett et al., Reference Barrett, Thrall, Burdon and Linde2008). However, in both bat and avian systems, the sensitivity to host social system varied among parasite taxa, with the structure of some parasites (bat flies and avian ischnoceran lice) unresponsive to differences in group size and connectivity of the same bat (M. bechsteinii) and bird (B. galapagoensis) hosts (Whiteman and Parker, Reference Whiteman and Parker2004; Reckardt and Kerth, Reference Reckardt and Kerth2009; van Schaik et al., Reference van Schaik, Dekeukeleire and Kerth2015) that produced notable changes in the population structure of wing mites and amblyceran lice, respectively. This contrast between parasite taxa highlights the fact that host social behaviour is but one of many factors that can shape parasite population genetics, and it would be valuable to weigh its relative importance across a broader diversity of host–parasite systems.

In addition to shaping the population genetic structure of their parasites, host group size and connectivity may impose direct selection on virulence, a key parasite trait (Box 1). The common assumption of a trade-off between transmission and virulence predicts that reduced connectivity, or increased modularity, of host groups selects against virulence. The ecological structure of host groups means that parasites with high transmission and virulence should end up with low effective transmission rates because they rapidly deplete the local density of susceptible hosts. This process of ‘self-shading’ favours mutants with low transmission and low virulence, which maintain a higher average density of susceptible hosts and lower probability of extinction (Boots and Sasaki, Reference Boots and Sasaki1999). Genetic structure could also lead to ‘kin shading’: within host groups, nearby parasites are likely kin, such that reduced transmission also confers an inclusive fitness benefit (Wild et al., Reference Wild, Gardner and West2009; Lion and Boots, Reference Lion and Boots2010). Moreover, Lipsitch et al. (Reference Lipsitch, Herre and Nowak1995) proposed a ‘law of diminishing returns’: repeated contact between hosts selects for lower virulence because the increased opportunities for transmission between individuals make the benefits of increasing transmission rate too small to offset the cost of increased virulence. By these arguments, the clustering associated with the modularity of social groups should select for parasites with low virulence.

Though they do not directly consider social behaviour, theoretical models support the evolution of reduced virulence with increased modularity of host populations (e.g. Claessen and de Roos, Reference Claessen and de Roos1995; Rand et al., Reference Rand, Keeling and Wilson1995; Boots and Sasaki, Reference Boots and Sasaki1999). In models that explicitly incorporate spatial structure, transmission ranges from global to local, either by modifying transmission of the parasite (e.g. Boots and Sasaki, Reference Boots and Sasaki1999) or by varying host contact structure from random interactions between hosts to clustered, regular interactions, modelling modularity within social groups (e.g. Van Baalen, Reference Van Baalen, Diekmann, Metz, Sabelis and Sigmund2002). Generally, as transmission becomes increasingly local, or host contacts become more clustered, the evolutionary optima for transmission rate and correlated virulence shift lower (though see Read and Keeling, Reference Read and Keeling2003). Consistent with theory, Boots and Mealor (Reference Boots and Mealor2007) found that, in experimental populations of the host Plodia interpunctella, a granulosis virus (PiGV) evolved reduced infectivity when host mobility was reduced (for further experimental support from other systems, see Kerr et al., Reference Kerr, Neuhauser, Bohannan and Dean2006; Dennehy et al., Reference Dennehy, Abedon and Turner2007; Berngruber et al., Reference Berngruber, Lion and Gandon2015). In contrast to modularity, other characteristics of social groups – such as size – may select for increased virulence. Indeed, increasing the size of host modules in spatial models brings the evolutionary dynamics closer to that of well-mixed host populations (Van Baalen, Reference Van Baalen, Diekmann, Metz, Sabelis and Sigmund2002). With transmission and/or host interactions less clustered and regular, the cost of self-shading falls, boosting the evolutionary optima for transmission and virulence. While these models generally assume that host mobility or contact networks (and by extension, modularity) do not vary with parasite status, it is important to also consider infection-induced changes in behaviour and their inherent heterogeneity [‘Synthesis: ecological feedbacks between social behaviours and parasite infection ’ section; Fig. 2]. These dynamic behavioural feedbacks in response to infection (Arrow B) may alter predictions for virulence evolution (e.g. see Pharaon and Bauch, Reference Pharaon and Bauch2018 on human social behaviour).

Virulence may also evolve indirectly in response to selection that host social behaviour imposes on parasite transmission mode. For parasites with genetic variation in transmission mode, frequent transmission opportunities in host social groups are expected to select for an increased rate of horizontal transmission, whereas among solitary or territorial hosts, reduced transmission opportunities should favour vertical transmission, which ensures transmission from parent to offspring (Antonovics et al., Reference Antonovics, Wilson, Forbes, Hauffe, Kallio, Leggett, Longdon, Okamura, Sait and Webster2017). Selection on transmission mode may in turn impose selection on virulence: experimental studies show that parasite lineages evolve higher virulence with increased opportunities for horizontal transmission (Bull et al., Reference Bull, Molineux and Rice1991; Turner et al., Reference Turner, Cooper and Lenski1998; Messenger et al., Reference Messenger, Molineux and Bull1999; Stewart et al., Reference Stewart, Logsdon and Kelley2005), whereas a recent comparative study suggests that vertical transmission favours the evolution of obligate mutualisms (Fisher et al., Reference Fisher, Henry, Cornwallis, Kiers and West2017). Thus, assuming a trade-off between transmission modes, social grouping may indirectly select for increased virulence via evolutionary shifts in transmission mode. It is not clear, however, how many host–parasite systems have a significant genetic variation in transmission mode (Antonovics et al., Reference Antonovics, Wilson, Forbes, Hauffe, Kallio, Leggett, Longdon, Okamura, Sait and Webster2017). Moreover, in a key proof of principle study, Turner et al. (Reference Turner, Cooper and Lenski1998) did not find that transmission mode evolved in response to host density, a potential proxy for host social behaviour.

A further indirect mechanism through which host social behaviour may affect parasite virulence evolution is through its effects on the likelihood of coinfection, which is hypothesized to alter the costs and benefits of virulence for parasites (Bremermann and Pickering, Reference Bremermann and Pickering1983; Alizon et al., Reference Alizon, de Roode and Michalakis2013). Several studies have found that larger, more connected host groups support richer, more genetically diverse parasite communities (Ranta, Reference Ranta1992; Griffin and Nunn, Reference Griffin and Nunn2012) and populations (e.g. van Schaik et al., Reference van Schaik, Kerth, Bruyndonckx and Christe2014). These studies suggest that hosts in such groups are more likely to be co-infected with multiple species or strains of parasites (though see Bordes et al., Reference Bordes, Blumstein and Morand2007). Coinfection could select for increased virulence if virulence stems from the depletion of host resources: in this case, within-host competition favours more virulent parasites that draw more aggressively on host resources (Bremermann and Pickering, Reference Bremermann and Pickering1983; Frank, Reference Frank1992; de Roode et al., Reference de Roode, Pansini, Cheesman, Helinski, Huijben, Wargo, Bell, Chan, Walliker and Read2005). Alternatively, coinfection could lead to reduced virulence if virulence stems from collective action, like the production of public goods: in this case, competition between unrelated strains favours cheaters, limiting the growth of the parasite population and suppressing virulence (Turner and Chao, Reference Turner and Chao1999; Chao et al., Reference Chao, Hanley, Burch, Dahlberg and Turner2000; Brown et al., Reference Brown, Hochberg and Grenfell2002). As of yet, these predictions are untested in the context of host sociality.

Overall, there is a substantial body of theory and data indicating that host social behaviours likely drive virulence evolution through several interacting pathways: host group size and modularity affect parasite population genetics, and impose both direct and indirect selection on virulence. In contrast, there is surprisingly little research investigating the effect of host social behaviours on the evolution of other parasite traits (Schmid-Hempel, Reference Schmid-Hempel2017). Here we highlight two topics – host specialization and manipulation – that have received some attention, in hopes of stimulating more research in these areas. First, behaviours that dictate how social groups or modules assemble may determine parasite prevalence and selection for specialization. In many systems, individual hosts preferentially interact with kin due to active choice or physical proximity (e.g. Grosberg and Quinn, Reference Grosberg and Quinn1986; Archie et al., Reference Archie, Moss and Alberts2006; Davis, Reference Davis2012). Parasitism may even enhance kin grouping if, for example, individuals actively avoid parasitized non-kin but continue to associate with parasitized kin [see ‘Parasite infection influences host social behaviours (Arrow B)’ section]. Kin association boosts the mean relatedness of hosts encountered by a parasite lineage, above that predicted if hosts met at random. Taken to its extreme, socializing with kin could create conditions for a parasite akin to host monoculture (King and Lively, Reference King and Lively2012; Lively, Reference Lively2016): on average, increased relatedness, or decreased genetic diversity, of host groups promotes parasite transmission (i.e. the monoculture effect as in Baer and Schmid-Hempel, Reference Baer and Schmid-Hempel1999; Altermatt and Ebert, Reference Altermatt and Ebert2008; Ekroth et al., Reference Ekroth, Rafaluk-Mohr and King2019). Moreover, host relatedness can mimic the selection parasites face under serial passage (Ebert, Reference Ebert1998): generations of transmission within relatively homogeneous host groups may lead to the evolution of host specialization (Bono et al., Reference Bono, Smith, Pfennig and Burch2017), either due to trade-offs or relaxed selection for performance on alternate hosts (Kassen, Reference Kassen2002). In systems where hosts do not associate with kin (e.g. Russell et al., Reference Russell, Kelley, Graves and Magurran2004; Riehl, Reference Riehl2011; Godfrey et al., Reference Godfrey, Ansari, Gardner, Farine and Bull2014), we expect the opposite: increased genetic diversity of interacting hosts should limit parasite spread and maintain parasite populations with relatively broad host ranges. This argument makes the interesting prediction that parasites that jump to novel host populations or species may preferentially derive from diverse host groups. We emphasize that there are few tests of these ideas – our predictions for the impact of group assembly on parasite evolution are based on studies of non-social systems and a few social insect systems (Sherman et al., Reference Sherman, Seeley and Reeve1988; Schmid-Hempel, Reference Schmid-Hempel2017).

Finally, behavioural manipulation of hosts, which includes any parasite-induced change in host behaviour that promotes parasite transmission (Poulin, Reference Poulin2010), is a trait that may experience selection in the context of social behaviour. Parasites transmitted socially could increase their probability of transmission by increasing the rate at which infected hosts interact with susceptible hosts. By this argument, selection of parasite manipulation would intensify host social behaviour. Nonetheless, there is little evidence in support of this hypothesis. Although there is strong evidence of host manipulation in parasites with other transmission modes such as trophic (e.g. trematodes – Carney, Reference Carney1969) or vector-borne transmission (e.g. Leishmania – Rogers and Bates, Reference Rogers and Bates2007), there are few accounts of socially transmitted parasites manipulating host contact rates (Poulin, Reference Poulin2010). Some socially transmitted viruses, including rabies, can increase aggression and thereby physical contact, but whether this constitutes adaptive manipulation remains under review due to the variable manifestation of symptoms (Lefevre et al., Reference Lefevre, Adamo, Biron, Missé, Hughes and Thomas2009; Poulin, Reference Poulin2010). In fact, across parasites, it is far more common that parasitism leads to reduced activity and social isolation (Poulin, Reference Poulin2019). An exception are the microsporidia and cestode parasites of brine shrimp (Artemia franciscana and A. parthenogenetica): these parasites increase swarming of brine shrimp near the water surface, which may increase trophic transmission of the cestode to its avian host and direct transmission of microsporidia to nearby Artemia (Rode et al., Reference Rode, Lievens, Flaven, Segard, Jabbour-Zahab, Sanchez and Lenormand2013). Poulin (Reference Poulin2010) hypothesizes that evidence for host manipulation in socially-transmitted parasites is limited because the benefits of manipulation are smaller than the costs: for host taxa with high degrees of sociality, many factors already promote interactions with conspecifics, so parasites may gain relatively little in the way of additional transmission opportunities by augmenting contact within groups. Recent work, however, suggests that parasites may induce behavioural changes that increase an infected host's probability of acceptance into new social groups. Geffre et al. (Reference Geffre, Gernat, Harwood, Jones, Morselli Gysi, Hamilton, Bonning, Toth, Robinson and Dolezal2020) found that honey bees infected with Israeli acute paralysis virus (IAPV) are accepted into foreign colonies at higher rates than control bees, even though bees can detect and avoid IAPV-infected nestmates. In comparison, colonies did not show higher acceptance of foreign bees that were immune-stimulated but not infected, suggesting a specific manipulation by IAPV to increase between-colony transmission. The authors speculate that these results point to a coevolutionary battle between parasite manipulation of host social behaviour and hosts' own social defences.

Synthesis: evolutionary feedbacks between host social behaviour and parasite traits

The evolution of host social behaviours in response to parasites (Arrow C) and parasites in response to hosts (Arrow D) support the potential for coevolutionary feedback between social behaviour and parasite traits. Although direct examination is challenging, theoretical models have begun to explore reciprocal adaptation between host social behaviour and parasite traits, and the impact of the behavioural environment on coevolutionary trajectories. For example, Bonds et al. (Reference Bonds, Keenan, Leidner and Rohani2005) examined feedback between virulence and social behaviour, measured as variation in host contact rate. They made the key assumption that more gregarious hosts live longer, so increased contact carries both a fitness benefit and cost (parasite transmission). As a result, increasing contact rates select against virulence: the lower death rate of more gregarious hosts prolongs the window for parasite transmission, reducing the advantage of parasites with high transmission rates and, by correlation, high virulence. Decreasing virulence reduces the cost of social behaviour, thereby selecting for host contact. These changes in virulence and contact rate increase parasite prevalence, which, at its highest level, further selects for host contact: hosts may as well reap the benefits of socializing when there is no hope of avoiding infection. Prado et al. (Reference Prado, Sheih, West and Kerr2009) extended this work to incorporate spatial structure, showing that sociality selects for high parasite virulence and that high virulence, in turn, selects against sociality. Though their results differ somewhat, both models suggest that coevolutionary feedbacks between social behaviour, parasite prevalence and virulence could generate either positive or negative correlations between parasitism and social traits, such as group size, depending upon the life history and coevolutionary history of the study populations.

Other studies suggest that social behaviour is a contextual variable that alters the trajectory of coevolution between host resistance and parasite traits. Best et al. (Reference Best, Webb, White and Boots2011) explored the evolution of host resistance and parasite virulence in a coevolutionary model with spatial structure. As in the above models of virulence evolution, Best et al. (Reference Best, Webb, White and Boots2011) did not explicitly consider social behaviour, but drew parallels between social grouping of hosts and the treatment of host reproduction and parasite transmission as local (i.e. host offspring or new infections are placed in neighbouring sites, forming clusters) or global (i.e. placed randomly across the network). They found that local host reproduction and transmission select for increased host resistance and reduced parasite virulence. Similar to prior evolutionary models, the explanation for these coevolutionary patterns lies in the spatial distribution of susceptible and infected hosts (ecological structure) and the clustering of kin (genetic structure). A key result from Best et al. (Reference Best, Webb, White and Boots2011) is that reproduction and transmission within local (e.g. social) groups could lead to heavily defended hosts with parasites that have low transmission rates and low virulence. Interestingly, this theoretical result matches Hughes et al. (Reference Hughes, Pierce and Boomsma2008)'s verbal prediction for social insects and their parasites. Given the importance of the scale of host interactions and transmission for these predictions, further understanding of the among-group movements of infected hosts [see ‘Parasite infection influences host social behaviours (Arrow B)’ section, Geffre et al., Reference Geffre, Gernat, Harwood, Jones, Morselli Gysi, Hamilton, Bonning, Toth, Robinson and Dolezal2020] would facilitate prediction of coevolutionary outcomes.

Host social behaviour may further alter coevolutionary trajectories if behavioural defences negatively covary with physiological defences against parasites . Physiological defences may decline in the presence of behavioural defences if there are trade-offs between defence components (Sheldon and Verhulst, Reference Sheldon and Verhulst1996; Parker et al., Reference Parker, Barribeau, Laughton, de Roode and Gerardo2011) or if physiological defences prove redundant and thus experience relaxed selection (Evans et al., Reference Evans, Aronstein, Chen, Hetru, Imler, Jiang, Kanost, Thompson, Zou and Hultmark2006; Amoroso and Antonovics, Reference Amoroso and Antonovics2020). There is some support for negative covariance of behavioural and physiological defences in social insect systems (Evans et al., Reference Evans, Aronstein, Chen, Hetru, Imler, Jiang, Kanost, Thompson, Zou and Hultmark2006; Viljakainen et al., Reference Viljakainen, Evans, Hasselmann, Rueppell, Tingek and Pamilo2009; Harpur and Zayed, Reference Harpur and Zayed2013; López-Uribe et al., Reference López-Uribe, Sconiers, Frank, Dunn and Tarpy2016) and more broadly [Klemme et al., Reference Klemme, Hyvärinen and Karvonen2020; see ‘Synthesis: ecological feedbacks between social behaviours and parasite infection ’ section]. A key implication of covariance between defence traits is that host social behaviours could fundamentally alter the host defences against which parasites battle and thereby change the traits predicted to be under coevolutionary selection. Given the potential for behavioural defences to alter not only host evolution but also the strength and nature of reciprocal adaptation, it would be valuable to use an experimental evolution approach to directly test the trade-offs between behavioural and physiological defences.

Finally, host social behaviour may structure coevolutionary dynamics via its effect on parasite population genetics. Specifically, data from natural host–parasite interactions suggest that the size and connectivity of host social groups contribute to determining genetic diversity and gene flow in their associated parasite populations [see ‘Host social behaviours influence parasite evolution (Arrow D)’ section]. Coevolutionary models show that gene flow and genetic variation define the capacity for parasite populations to adapt to their evolving host populations and thereby drive coevolution (Lively, Reference Lively1999; Gandon, Reference Gandon2002; Gandon and Michalakis, Reference Gandon and Michalakis2002). In particular, experimental evolution studies (Forde et al., Reference Forde, Thompson and Bohannan2004; Morgan et al., Reference Morgan, Gandon and Buckling2005) and meta-analyses of tests with natural host–parasite populations (Greischar and Koskella, Reference Greischar and Koskella2007; Hoeksema and Forde, Reference Hoeksema and Forde2008) show that relatively low rates of gene flow can prevent parasites from adapting to their local host populations. While social behaviour entails its own complexities, the parallels we highlight suggest that the extensive body of work on the geography and spatial structure of host–parasite coevolution may prove valuable in formulating hypotheses and experiments on the evolution and coevolution of host sociality and parasites (Thompson, Reference Thompson2005).