Evolution and Ecology of Parasite Avoidance

Abstract

Parasite avoidance is a host defense that reduces the contact rate with parasites. We investigate avoidance as a primary driver of variation among individuals in the risk of parasitism and the evolution of host-parasite interactions. To bridge mechanistic and taxonomic divides, we define and categorize avoidance by its function and position in the sequence of host defenses. We also examine the role of avoidance in limiting epidemics and evaluate evidence for the processes that drive its evolution. Throughout, we highlight important directions to advance our conceptual and theoretical understanding of the role of avoidance in host-parasite interactions. We emphasize the need to test assumptions and quantify the effect of avoidance independent of other defenses. Importantly, many open questions may be most tractable in host systems that have not been the focus of traditional behavioral avoidance research, such as plants and invertebrates.

I. Introduction

Avoidance reduces the rate at which victims make contact with their enemies. In human populations, avoidance of parasites has proven highly effective as an individual defense and a population-level intervention for reducing disease transmission. Avoidance measures like social distancing, hygiene, and quarantine have played critical roles in controlling infectious diseases. Although avoidance is considered to be a central component of predator-prey and plant-herbivore interactions (Box 1), it has yet to be recognized as a fundamental force driving the ecology and evolution of host-parasite interactions. However, as the first defense hosts can mount, avoidance likely makes an outsized contribution to protecting hosts from infection. Moreover, avoidance may strongly influence the evolution of defenses that act after parasites contact hosts.

Box 1: Avoidance in Other Antagonisms.

Avoidance is not unique to host-parasite interactions: prey avoid predators, and plants avoid herbivores. There are clear parallels and important differences in the ecology and evolution of avoidance across these distinct antagonisms.

Predator-prey:

While the most salient consequence of predation is mortality, predators also have non-consumptive effects on prey, of which avoidance is a prominent example (Lima & Dill 1990). Avoidance reduces predator-prey contact, thereby reducing the rate at which predators initiate an attack (Daversa et al. 2021). Induced behavioral defense is the most studied form of predator avoidance. For example, increased predation risk can drive prey to decrease their movement (Kohler & McPeek 1989) and seek refuge (Kotler et al. 1991). This form of avoidance can carry a cost - lost foraging opportunities - that varies with context: e.g., hungrier animals spend more time foraging outside refuges (Kohler & McPeek 1989). Parasite avoidance can directly conflict with predator avoidance, pointing to a context-dependent cost of both defenses. For example, when green frogs avoid parasites by increasing activity, visual predators can more easily locate them (Marino & Werner 2013).

How does the strength of parasite avoidance compare to that of predator avoidance? Daversa et al. (Daversa et al. 2021) conducted a meta-analysis to compare the degree to which tadpoles avoided predators vs. parasites. They found evidence suggesting that predators may select more strongly for avoidance than do parasites. The relatively rapid lethality of predators shortens the defense cascade against predation as compared to a more protracted and possibly less lethal parasitic interaction, suggesting stronger selection on avoidance of predators than of parasites.

Plant-herbivore:

In the field of plant-herbivore interactions, avoidance refers to strategies that reduce the probability that an herbivore feeds on a plant (Tiffin 2000). Structural or morphological traits may increase avoidance (Lev-Yadun 2021): plants with more prostrate growth are less likely to be grazed by livestock (Fujita & Koda 2015, Rotundo & Aguiar 2007). Avoidance may be phenological: seed predation of Geum urbanum induced delayed flowering, such that more flowers set seed when seed predators were no longer active (Sercu et al. 2020). Herbivore avoidance is also used to refer to post-contact defenses that reduce initiation of herbivore feeding. For example, prior herbivory can induce shifts in plant chemistry that drive later herbivores to leave the plant prior to feeding (Sauge et al. 2006).

As for host-parasite interactions, avoidance constitutes the first step in the defense cascade against herbivores. With the exception of grazing, herbivory does not immediately result in plant death. In this sense, we expect the defense cascade against parasites to be more similar in length and structure to that against herbivores than that against predators.

Herbivore avoidance may, however, differ fundamentally from parasite avoidance in its effect on the fitness of the enemy. Herbivores are generally more mobile than their hosts and can readily shift between plants. Thus avoidance could shift the distribution of herbivores to non-avoidant plants without necessarily reducing herbivore fitness (Tiffin 2000).

In both the above interaction types, avoidance can also work in reverse, with predators or herbivores avoiding toxic victims. It is possible that parasites with mobile infectious stages may similarly avoid defended hosts.

In this review, we focus on avoidance of parasites (see glossary). We define the term “parasite” broadly to capture the intimate relationship in which one organism (the parasite) benefits by living in or on another (the host), at a cost to the host. Thus, the term “parasite” encompasses organisms that some fields separate as macroparasites and microparasites/pathogens, including viruses, bacteria, protozoa, helminths, and arthropods (Anderson & May 1979, Lafferty & Kuris 2002). We adopt this general terminology to synthesize diverse research areas. We specifically focus on variation in avoidance to address a long-standing paradox: if defenses are beneficial to hosts, we expect natural selection to favor defended individuals, leading to a loss of variation. Yet some individuals are much more parasitized than others. Why do hosts vary so much, and how does avoidance contribute to this variation?

In this review, we contextualize avoidance of parasites as the first step in a sequence of defenses that hosts mount in succession. Our approach builds on the foundation laid by prior reviews of avoidance, which have highlighted animal behaviors that limit contact with parasites (Behringer et al. 2018, Buck et al. 2018, Curtis 2014, Hart 2011, Loehle 1995, Stockmaier et al. 2021). We synthesize these classic examples with case studies that introduce tractable systems and non-behavioral mechanisms of avoidance. Doing so emphasizes the diverse forms avoidance can take and its wide taxonomic distribution. Moreover, in treating avoidance as a key component of the defense cascade, rather than as a taxonomically-isolated phenomenon, we integrate avoidance with theory on the epidemiology and evolution of host defenses.

We begin by defining avoidance with respect to other defenses, highlighting examples to showcase its general features. We then consider the epidemiological consequences of avoidance by reviewing theory and data on its contribution to variation in transmission. In the second half, we review theory relevant to the evolution of avoidance, pointing out the scarcity of data on the genetic basis and fitness consequences of avoidance independent of other defense modes. We also consider whether avoidance shapes the evolution of other defenses, notably resistance, and imposes reciprocal selection on parasites. In the final section, we outline directions for future research that will fill significant theoretical and empirical gaps in our understanding of the evolution and ecology of parasite avoidance. We conclude that the privileged position of avoidance as a host’s first defense justifies a concerted effort to quantify its contributions to epidemiological dynamics and host fitness.

II. What is avoidance?

We define avoidance as any host defense that reduces the contact rate between host and parasite. Avoidance limits both the risk of contact with parasites and the number of parasites transmitted upon contact, or exposure. By this definition, avoidance is not a binary trait (i.e., avoidant or not), but rather one that varies quantitatively: hosts with higher avoidance contact fewer parasites on average than hosts with lower avoidance (Figure 1a).

Figure 1: Defining avoidance.

A) Avoidance reduces the rate at which hosts (grey circles) contact parasites (red triangles). Therefore, in the same environment, hosts with high avoidance (dark grey) will contact fewer parasites on average than hosts with low avoidance (light grey). B) Avoidance is the first in a series of host defenses that parasites must overcome - it acts to halt progress of the parasite by limiting contact in the first place. Once contact has been made (i.e., avoidance has been overcome), defenses act to limit establishment of the parasite and then limit growth of the established infection. We refer to post-contact defenses that reduce parasite fitness collectively as resistance. For simplicity, we do not consider other strategies, such as tolerance, in this illustration; see Hall et al. (2017) for further details.

This definition places avoidance at the beginning of a cascade of host defenses that occur in sequence as an individual’s infection advances (Hall et al. 2017). By definition, these diverse defense strategies reduce fitness losses to parasitism, and we differentiate them based on when they act (Schmid-Hempel 2011). Avoidance sits at the start of the defense cascade, acting as the gatekeeper by limiting initial contact with parasites (Figure 1b). In doing so, avoidance reduces the number of parasites a host encounters and by extension a host’s probability of infection.

When avoidance fails and contact occurs, parasites encounter a host’s post-contact defenses. Some post-contact defenses limit parasite fitness by preventing establishment and impeding parasite proliferation post-establishment. We refer to these strategies as mechanisms of resistance (Hall et al. 2017, Rigby et al. 2002). Other post-contact defenses preserve host fitness by limiting the damage inflicted by parasites without reducing parasite fitness. We refer to these strategies as mechanisms of tolerance (Råberg et al. 2009). Because they occur after contact, resistance and tolerance can only influence host and parasite fitness after parasites overcome a host’s avoidance defenses (Hall et al. 2017) (Figure 1b). Thus, variation in each of these defenses can drive individual differences in the probability and fitness consequences of infection, but variation in avoidance dictates the degree to which subsequent defenses contribute to infection outcomes.

Major syntheses of this topic have primarily focused on behavioral mechanisms of avoidance used by animals (Behringer et al. 2018, Buck et al. 2018, Curtis 2014, Sarabian et al. 2018). In defining avoidance instead by its position in the defense cascade, we emphasize that avoidance need not be behavioral, and behavioral defenses are not necessarily avoidance. For example, behavioral thermoregulation reduces the fitness of established parasites and is thus a mechanism of resistance (de Roode & Lefèvre 2012). In defining avoidance by its role in the sequence of defense, rather than by its mechanisms, we demonstrate that avoidance need not be restricted to animals. Moreover, some aspects of the evolution and ecology of avoidance may be more easily studied in non-animal systems.

Studying avoidance

Interest in parasite avoidance grew with observations of the behavior of large, charismatic mammals. Notably, Freeland (1976) hypothesized that primates’ social structures, degree of promiscuity, and habitat use reflect adaptations to avoid parasites. Taylor (Taylor 1954) proposed that ungulates avoid grazing in areas contaminated with feces to avoid helminth infections. The taxonomic scope of avoidance research has since grown to include a wider diversity of hosts, including humans, fish, invertebrates, and plants. Possible mechanisms now include life history and structural traits as well as behaviors (Table 1).

Table 1:

Avoidance reduces the probability of contact with:

1) Infectious conspecifics, via changes to: • Social contact: Healthy guppies (Poecilia reticulata) use visual and chemical cues to avoid ectoparasite-infected guppies when they are most transmissible (Stephenson et al. 2018). • Sexual contact: Female olive baboons (Papio anubis) reject mating attempts by males with conspicuous genital ulcerations caused by the sexually-transmitted bacterium Treponema pallidum (Paciência et al. 2019).
2) Parasites that actively seek hosts, via changes to: • Induction: Host roots induce germination of seeds of the parasitic plant broomrape (Orobanche spp.). Several cultivated plants show genetic variation in induction of broomrape germination, likely due to variation in root exudates (Fernández-Aparicio et al. 2014, Sillero et al. 2005). • Detectability: The parasitoid fly Ormia ochracea locates its host, the Pacific field cricket (Teleogryllus oceanicus), via male mating calls. Wing mutations that render male hosts silent have arisen independently on multiple Hawaiian islands, rising to high frequency (>90%) in some places (Pascoal et al. 2014, Zuk et al. 2006).
3) Food and water contaminated with parasites, via changes to: • Foraging: After selection in the presence of a water-borne protozoan parasite (Perkinsus marinus), two lines of eastern oyster (Crassostrea virginica) display increased shell closure and decreased filter feeding in the presence of parasites. This foraging modification results in a large reduction in parasite loads (Ben-Horin et al. 2018). • Cannibalism: Healthy adult amphipods (Gammarus duebeni) choose to cannibalize uninfected juveniles twice as frequently as juveniles infected with a microsporidian parasite (Pleistophora mulleri). Infected adults show higher rates of cannibalism and little discrimination between infected and uninfected juveniles (Bunke et al. 2015).
4) Parasite stages in the environment, via changes to: • Physical activity: Green pea aphids (Acyrthosiphon pisum) drop from their host plant when they detect natural enemies, including parasitoid wasps. Fill et al. (Fill et al. 2012) demonstrated that these defensive drops reduce aphid population growth, independent of the fitness costs of infection. • Dispersal: The presence of ectoparasitic mites induces backswimmers (Notonecta undulata) to disperse further. In experimental conditions, cues signaling the presence of mites can also induce higher dispersal rates (Baines et al. 2020). • Phenology: Male Silene latifolia of late flowering families have lower rates of infection with anther smut (Microbotryum violaceum) in the field, likely due to reduced contact with pollinators that vector the fungus (Biere & Antonovics 1996). • Habitat construction: Urban house sparrows (Passer domesticus) and house finches (Carpodacus mexicanus) incorporate cigarette butts in their nests (Suárez-Rodríguez et al. 2013). Nests with more cigarette material attracted fewer parasitic mites and had higher hatching and fledging success (Suárez-Rodríguez & Garcia 2017).
• Host structure: Splashing rain drops carry pycnidiospores of the fungus Zymoseptoria tritici from the lower leaves of wheat plants to healthy upper leaves (Suffert et al. 2011). Increased vertical spacing causes leaves of taller plants to avoid contact with inoculum from below (Tavella 1978), which may explain the explosion of Z. tritici following the introduction of dwarf wheat varieties in the 1980s (Robert et al. 2018). • Personality: More active and exploratory Siberian chipmunks (Tamias sibiricus) have higher loads of ticks, suggesting higher exposure rates due to their traveling over greater geographic ranges (Boyer et al. 2010).

Photo credits: 1) Male and female Wild Form guppies (Poecilia reticulata), PH Olsen, Wikimedia Commons, CC BY 3.0; 2) Used with permission of Justa Heinen-Kay; 3) Gammarus duebeni (Hindrem - The Trondheim Fjord), Used with permission of Kåre Telnes, www.seawater.no, © Kåre Telnes; 4) Adult pea aphid on alfalfa, Jpeccoud, Wikimedia Commons, CC BY 3.0; 5) Silene latifolia with Microbotryum violaceum, A Gibson, author’s own picture;

By definition, these diverse mechanisms of avoidance act to reduce a host’s rate of contact with parasites. How effective are avoidance defenses at limiting the probability of infection? Relatively few studies have quantified the effect of avoidance, but these report a massive effect. For example, tadpoles accelerate, twist and turn in response to free-swimming trematodes (Taylor et al. 2004). Daly and Johnson (2011) compared Pacific chorus frog (Pseudacris regilla) tadpoles that were active vs. anesthetized and found that such evasive movements reduced the probability of infection by 20-39% and the number of established parasites per host by 65%. Kiesecker et al. (1999) demonstrated that bullfrog (Rana catesbeiana) tadpoles physically move away from conspecifics infected with a parasitic yeast. Caging individuals showed that this distancing reduced exposure by ~80%. In these examples, avoidance rivals, or even surpasses, more commonly studied resistance mechanisms in their effect on infection outcomes.

These examples also highlight a key challenge to studying avoidance: its effect cannot be inferred from infection outcomes alone, because these result jointly from avoidance and all subsequent defenses (Figure 1b). Isolating the effect of avoidance requires experimental comparison of an intact infection cascade and a disrupted cascade that bypasses avoidance by, for example, caging, immobilizing, or artificially exposing hosts. Eakin et al. (2015) offers a rare exception by quantifying the feeding behavior of individual gypsy moth larvae (Lymantria dispar) on oak leaves and fitting simulation models to variation in feeding and infection outcomes. Their results indicated that larvae avoided baculovirus-laden cadavers, and avoidance reduced infection risk by up to 7% per transmission round (see also Biere & Antonovics 1996, Strauss et al. 2019).

Experiments like these are feasible for many invertebrate, plant, and small vertebrate hosts, but they are rarely possible in the wild ungulate and primate systems that have historically dominated the study of avoidance. As a result, the benefit of avoidance is often assumed rather than quantified. This assumption is reasonable, but intuition cannot measure the magnitude of an effect, nor establish its evolutionary basis. Avoidance of infected conspecifics may, for example, reflect selection to avoid general, conserved “alarm” cues that signal multiple threats, including predators and physical hazards (Candia-Zulbarán et al. 2015). Direct estimates of the fitness and transmission consequences of avoidance in diverse systems are required to integrate avoidance into existing epidemiological and evolutionary frameworks. Moreover, phylogenetic and microevolutionary studies are required to determine if avoidance evolves in response to selection by parasites or as a consequence of selection for other traits.

Categorizing avoidance

Parasites vary in their transmission mode (Antonovics et al. 2017), which dictates the avoidance mechanisms that can reduce a host’s rate of contact with parasites. Avoiding close contact with infected conspecifics will limit contact with parasites transmitted directly between individuals (e.g., by touching) but not parasites transmitted by vectors (e.g., mosquitoes). In Table 1, we highlight the form and function of prominent mechanisms to avoid horizontally transmitted parasites by categorizing them according to the type of contact they reduce (building on (Curtis 2014, de Roode & Lefèvre 2012). Broadly, avoidance mechanisms decrease the probability that hosts contact 1) infectious conspecifics; 2) parasites that actively seek hosts; 3) food or water contaminated with parasites; or 4) parasite stages in the environment. Some mechanisms may have generalized effects and thus could be assigned to more than one of these categories.

Avoidance mechanisms reflect dichotomies that are typically used to describe other parasite defenses. First, an avoidance defense may be constitutive or induced. For example, some male field crickets (Teleogryllus oceanicus) are physically unable to call; these silent males constitutively avoid parasitoid flies (Ormia ochracea) that locate cricket hosts by mating calls (Pascoal et al. 2014, Zuk et al. 2006). Avoidance of these eavesdropping flies can also be induced (Lewkiewicz & Zuk 2004): non-silenced male field crickets from heavily parasitized populations take longer to resume calling after a disturbance. Induced defenses may carry fewer costs than constitutive ones, but their contribution to host fitness hinges upon whether hosts can accurately estimate their risk of parasitism using cues from the environment (Yousefi & Fouks 2019) or conspecifics (Behringer et al. 2006). Host avoidance can also be cued by damage signals from the early stages of its own infection (Zhang et al. 2005), raising the possibility that a host’s downstream defenses induce upstream avoidance defenses.

Second, an avoidance defense may be either innate or acquired based on prior experience (e.g., learned), akin to the distinction between innate and acquired immunity (Velagapudi et al. 2020). For example, several studies show that foraging individuals vary innately in their tendency to avoid areas likely to be contaminated with parasites, irrespective of prior experience (Hutchings et al. 2007, Parker et al. 2010). Individuals can also acquire avoidance defenses based on direct experience (Meisel & Kim 2014) or observation of conspecifics (Kavaliers et al. 2001). Babin et al. (2014) found that flies (Drosophila melanogaster) learned to avoid artificial odors associated with infection with a virulent strain of a bacterial parasite (Pseudomonas entomophila) but not a harmless strain. The significance in nature of learning as a mechanism for acquired avoidance remains unclear; it typically requires that reliable parasite cues coincide with the host’s experience of exposure or infection (Amoroso 2021).

Third, avoidance defenses can be specific or general. The nematode Caenorhabditis elegans rapidly vacates patches of the virulent bacteria Serratia marcescens by detecting the biosurfactant serrawettin W2 produced by some strains of S. marcescens (Pradel et al. 2007). Pea aphids (Acyrthosiphon pisum) avoid leaves where virulent Pseudomonas syringae strains are present, responding to the emission spectrum of pyoverdine, a fluorescent compound produced by pseudomonads (Hendry et al. 2018). Many induced avoidance defenses are, however, linked to general cues, such as the presence of feces (Hutchings et al. 2003), host cadavers (Capinera et al. 1976), or sick conspecifics (Boillat et al. 2015), for which the underlying mechanisms are poorly characterized. These general cues are frequently studied because they are relatively easy for researchers to manipulate. For example, avoidance of infected conspecifics can be elicited by inducing general sickness behaviors with immune system stimulants (Stockmaier et al. 2020, Zylberberg et al. 2013).

The examples above emphasize that induced and acquired avoidance defenses can contribute to host defense only to the extent that cues accurately signal parasite presence. Thus, much research has focused on linking induced behaviors to the cues that trigger them. For example, Poirotte et al. (Poirotte et al. 2017) found that parasites modified the fecal odors of their mandrill hosts, potentially explaining why mandrills spent more time in proximity to feces from healthy vs. parasitized conspecifics. Research with the host C. elegans has even dissected the genetic and neurological basis of avoidance (Meisel & Kim 2014): Chang et al. (2012) identified two polymorphisms in a single gene whose localized function in a sensory neuron pair explains variation in avoidance of lawns of P. aeruginosa. However, many parasites may not have – or perhaps even strategically lack – reliable cues signaling their presence (see Section VI). For example, Svalbard reindeer (Rangifer tarandus platyrhynchus) prefer to feed in wetter sites where fecal density is lower, but feces turn out to be an unreliable cue: nematode survival, and thus infection risk, is actually higher in wetter sites (van der Wal et al. 2000). These studies of induced responses reveal that a key challenge in understanding avoidance is the difficulty for both the host and the researcher in accurately identifying parasite risk.

Our approach to avoidance unites taxa and traits that are rarely treated as analogous (Table 1). These distinct mechanisms of avoidance all reduce rates of contact between hosts and parasites and are thus likely to reduce the rate at which hosts become infected in a population. This shared epidemiological effect in turn suggests that we can make general predictions about the evolution of parasite avoidance, irrespective of host taxon or mechanism. We elaborate upon the epidemiological effect of avoidance in the next section, followed by a treatment of the evolution of avoidance.

III. Epidemiology of avoidance

Here, we highlight the role avoidance can play in epidemiological dynamics. We review studies showcasing the insights gained by quantifying the unique effect of avoidance on the spread of parasites through host populations.

In standard epidemiological models, the rate at which uninfected individuals become infected depends upon the number of infected individuals in the host population, or parasite propagules in the environment, and β, the transmission parameter. β can be broken down into two components of transmission: 1) that attributable to contact: β_c, the rate of physical contact with parasites, via uninfected hosts contacting infected hosts or free-living parasites, and 2) that attributable to physiology: β_p, the probability of infection given that contact has occurred (Dobson 1995, Hawley et al. 2011). These two terms are complex coefficients that capture multiple features of the transmission process. β_c is expected to vary with features of the habitat, such as the spatial distribution of free-living parasites, and traits of the hosts, such as social behavior. β_p is expected to vary with attributes of the infectious agents, such as the parasite load of the infected host, as well as post-contact defenses that limit parasite establishment (Lloyd-Smith et al. 2006). We can thus categorize defenses according to their effect on these two components of transmission: avoidance reduces β_c, while resistance reduces β_p.

Though β is commonly treated as a “black box” (Lello & Fenton 2017), breaking β down into its constituent parts allows measurement of the impact of avoidance and resistance on rates of transmission. This typically requires experiments. For example, Strauss et al. (2019) quantified variation in β_c and β_p using experimental data on variation in the rate of infection of Daphnia dentifera with the fungus Metschnikowia bicuspidata. Daphnia dentifera contact fungal spores via feeding, and the presence of spores induced many host genotypes to reduce their feeding rate (i.e., the contact rate). For some host genotypes, increasing the density of spores in the environment enhanced both this avoidance response and resistance, resulting in their having lower infection rates at the highest spore density than at intermediate density. This mechanistic approach models how to experimentally disentangle the relative importance of defenses that jointly determine variation in infection outcomes. However, in natural settings, it is challenging to parse mechanisms from the endpoints of the transmission process, because it is impossible to determine if an observed infection resulted from a high contact rate coupled with a low per-contact infection probability, or a low contact rate coupled with a high per-contact infection probability. In one approach to this problem, Aiello et al. (2016) measured transmission of Mycoplasma agassizii in experimental groups of desert tortoises (Gopherus agassizii) to predict transmission rates in a natural population. Proximity loggers showed that most contacts in the natural population were short-lived (low β_c), suggesting that transmission risk was low except when contacting the most heavily infected hosts (high β_p).

Accounting for variation in avoidance improves the accuracy of epidemiological predictions for at least two reasons. First, theoretical work suggests that epidemiological dynamics are sensitive to covariation between contact rate and per-contact infection probability within or among hosts. Hawley et al. (2011) demonstrated that the sign and magnitude of covariation between contact rate and infection probability can determine whether a disease will spread or fade out. For example, Stephenson et al. (2018) found that uninfected hosts specifically avoid the infected hosts that pose the greatest infection risk: uninfected guppies (Poecilia reticulata) avoided contact with guppies infected with Gyrodactylus turnbulli only in the later days of their infection, when the infected hosts were most infectious. This correlated behavior should reduce disease spread far more than if the propensity to avoid conspecifics varied independently of conspecific infectiousness. Thus, explicitly modeling contact rate and per-contact infection probability may be necessary to capture a realistic range of variation in infection outcomes.

Second, inducible avoidance defenses can be deployed relatively quickly, and the speed of this response can dramatically alter epidemiological predictions. Stroeymeyt et al. (2018) tracked social networks in a colony of Lasius niger ants following exposure of foragers to the fungus Metarhizium brunneum. Ants quickly modified their social behavior, resulting in changes to the network that isolated potentially infected hosts. Failure to account for induced avoidance will lead epidemiological predictions to overestimate the probability and size of an epidemic. For example, a mass die-off of the sponge community in the Florida Keys could have resulted in an epidemic of PaV1 virus in Caribbean spiny lobsters (Panulirus argus). Groups of juvenile lobsters shelter in sponges, so fewer sponges created larger social aggregates, where PaV1 could transmit readily. Yet, rates of PaV1 infection did not increase after the sponge die-off. Combining field assays of behavior and an epidemiological model, Butler et al. (2015) suggested that avoidance explained the lack of PaV1 spread. Even when sponge shelters were relatively scarce, lobsters vacated dens upon introduction of an infected conspecific (Behringer et al. 2006).

Recent human epidemics also speak to the power of induced avoidance. For example, the U.S. Centers for Disease Control projected the number of cases of Ebola virus in Liberia and Sierra Leone by January 2015, but their projection over-estimated actual case numbers by 70-fold. Arthur et al. (2017) argued that the lower case numbers reflected the adoption of avoidance practices in communities experiencing Ebola (Hewlett & Amola 2003). The COVID-19 pandemic provides more evidence of the significance of avoidance: epidemiological models of SARS-CoV-2 included parameters representing avoidance (e.g., social distancing) because of its exceptional power to rapidly change disease dynamics, particularly prior to mass vaccination (Ferguson et al. 2020). These case studies suggest that induced avoidance can generate rapid, extreme reductions in β_c, making avoidance a uniquely powerful strategy when physiological defenses are absent.

Variation in the traits of individuals drives the trajectory of epidemics (e.g., (Lloyd-Smith et al. 2005). This individual-level variation is as likely, perhaps even more likely, to stem from variation in traits that affect contact rates (e.g., avoidance) than from variation in traits that affect per-contact infection probability (e.g., resistance) (Hethcote & Yorke 1984). If epidemiological models do not account for the contribution of avoidance to individual variation in infection risk, then predictions will likely overestimate the rate of spread and size of epidemics. On the other hand, in the absence of avoidance, epidemics have the potential to run rampant, especially in the case of emerging infectious diseases where post-contact defenses may be limited. Broadly resolving the relative contributions of avoidance and resistance to individual-level heterogeneity in transmission patterns would advance our understanding of their ecological significance and our prediction of their evolutionary trajectories, which we turn to in the next section.

IV. How does avoidance evolve?

Despite the potential contribution of avoidance to variation in infection outcomes, we have very little theory that can explain how avoidance evolves and why variation in avoidance is maintained. If avoidance has a genetic basis and a fitness benefit, parasites are expected to select against hosts that fail to avoid them, reducing variation for avoidance in the host population. Evolutionary theory has addressed a similar paradox for resistance, which varies extensively in host populations (Boots et al. 2009, Buckling & Rainey 2002, Frank 1993). Avoidance and resistance both decrease parasite transmission, thereby reducing the prevalence of infection. This basic similarity provides a starting point for theoretical expectations and empirical predictions about avoidance evolution (Amoroso 2021, Amoroso & Antonovics 2020).

Theory finds that the evolution of resistance depends fundamentally on the nature and magnitude of its costs, which can maintain variation in resistance among individuals in a population (Antonovics & Thrall 1994, Boots et al. 2009, Hoyle et al. 2008). Amoroso and Antonovics (2020) applied a similar theoretical approach to make predictions for the evolution of avoidance. Modeling the transmission of parasites from infected conspecifics, they found that the nature of costs determines whether avoidance of conspecifics can evolve. If individuals gained fitness from interacting with conspecifics, as we might expect in social taxa, then hosts that avoided interactions gave up these benefits. Under these conditions, avoidance was unlikely to evolve, in spite of the benefits of reduced transmission. Though an important first step, this first theoretical treatment of avoidance evolution makes assumptions that have yet to be substantiated with data: that there is heritable variation in avoidance, and that avoidance has costs of some form. We address the evidence for these assumptions.

Is there heritable variation in avoidance?

Most of the existing research on avoidance has been descriptive, identifying mechanisms of avoidance and interpreting these in an adaptive context. Variation in avoidance in host populations is rarely quantified, and accordingly the genetic basis of avoidance remains poorly characterized. Such basic research may be especially lacking because many of the species and contexts in which avoidance is most apparent to human observers are challenging to manipulate. However, phenotypic variation in traits related to avoidance has been documented in several species. For example, bolder or more exploratory individuals have higher rates of parasite infection in pumpkinseed sunfish (Lepomis gibbosus: (Wilson et al. 1993) and deer mice (Peromyscus maniculatus: (Dizney & Dearing 2013), suggesting that variation in animal personality could generate variation in avoidance (Lopes 2017). Similarly, behavioral experiments in Japanese macaques (Sarabian & MacIntosh 2015), chimpanzees (Sarabian et al. 2017), and even humans (Shook et al. 2019) have uncovered a high degree of variation in avoidance across individuals (see Sidebar: Landscape of disgust).

Sidebar: The “landscape of disgust” and parasite avoidance.

In human psychology, psychiatry, and evolutionary psychology, the emotion of disgust has been proposed to motivate avoidance of parasites or their cues (Curtis et al. 2011, Rozin & Fallon 1987, Tybur et al. 2013). The pervasiveness of parasite avoidance across animals has led some researchers to suggest that the emotion of disgust may not be limited to humans (Kavaliers et al. 2019, Sarabian et al. 2017) disgust in non-human animals.

Recently, the idea has been put forth that non-consumptive effects of parasites could include changes in spatial use and habitat selection (Buck et al. 2018). Similar to the concept that predation risk can prevent herbivores from grazing in open areas (“the landscape of fear;” Gaynor et al. 2019), this concept has been termed the “landscape of disgust” (Buck et al. 2018). Research is just beginning to address how risk of parasite infection could drive host avoidance of particular habitats and cascading effects on ecosystems (Koprivnikar & Penalva 2015).

There is some evidence that phenotypic variation in avoidance has a genetic basis. Researchers have identified differences in avoidance among family lineages (gypsy moths: Parker et al. 2010; Silene latifolia: Biere & Antonovics 1996) and genotypes (Daphnia: Strauss et al. 2019). In two experiments, hosts evolved enhanced avoidance when exposed to parasites for multiple generations (sheep: Hutchings et al. 2007; C. elegans: Penley & Morran 2018). Knockout experiments in laboratory models have identified genes involved in avoidance of specific parasites (C. elegans: Nakad et al. 2016, Pradel et al. 2007), and of olfactory cues of infected conspecifics (Mus musculus: Kavaliers et al. 2005). However, determining the heritability of avoidance can be challenging, because avoidance traits are often plastic, polygenic, and difficult to quantify, and the hosts in question can be long-lived and challenging to rear in controlled environments. To lay the foundation for research on avoidance evolution, it will be necessary to expand the systems and types of traits that we study to characterize genetic variation in avoidance.

What are the costs of avoidance?

Variation in avoidance may be maintained by costs that offset its benefits, similar to resistance (Antonovics & Thrall 1994, Bowers et al. 1994). Costs are expected to dictate the evolutionary trajectory of avoidance, yet explicit quantification of avoidance costs is rare. Biere and Antonovics (1996) demonstrated that late-flowering genotypes of S. latifolia avoid the pollinator-transmitted disease anther smut by flowering later in the season. For male plants, later flowering comes with a constitutive fitness cost - fewer reproductive opportunities. Other costs may only be incurred upon induction of avoidance. For example, Drosophila hydei produced 70% more CO₂ (a measure of host metabolism) than controls when ectoparasitic mites were nearby but not in direct contact with hosts. Direct contact between host and parasite further increased CO₂ production by 35% (Luong et al. 2017). The costs of avoidance may also vary with context, as observed for costs of resistance (Bartlett et al. 2018, Kraaijeveld & Godfray 1997, Meaden et al. 2015). For example, choice experiments indicate that sheep fed on a reduced diet were less likely than well-fed sheep to avoid grass swards with fecal contamination (Hutchings et al. 2001). These examples highlight the potentially diverse and complex nature of avoidance costs. It remains unclear how to scale up from costs measured on an experimental timescale to infer relative fitness outcomes that would inform theory.

The evolution of avoidance may also hinge on its cost relative to other defense modes. Avoidance is commonly described as a “cheap alternative” to resistance and thus more “cost effective” (Curtis 2014, Klemme & Karvonen 2017). The only study to our knowledge to track the evolutionary trajectory of avoidance indeed found no measurable fitness cost in C. elegans, and high levels of avoidance were maintained after multiple generations in the absence of the parasite (Penley et al. 2018). However, in other cases, the absence of avoidance has been taken as evidence that the cost of avoidance is exceedingly high. For example, highly social vampire bats (Desmodus rotundus) continue to share food with sick partners, perhaps because limiting contact with infected individuals bears too high a cost in the context of this species’ cooperative social system (Stockmaier et al. 2020).

Integrating theoretical and empirical work on resistance evolution with evidence from the avoidance literature demonstrates that we currently lack the data to develop an evolutionary framework for avoidance. Critically, we do not know to what degree natural variation in infection can be attributed to individual variation in avoidance, independent of resistance. If variation in avoidance contributes to differences in infections among individuals, it remains to be tested how commonly this variation has a genetic basis, and why it exists in the first place. One possible explanation is a fitness cost of avoidance, but research is needed to test this prediction and evaluate alternatives. In the next section, we consider whether the evolutionary interaction of avoidance and resistance could maintain variation in these defense traits.

V. Role of avoidance in the defense cascade

In prior sections, we differentiated pre-contact (avoidance) from post-contact (e.g., resistance) defenses, emphasizing their independent contributions to reduced infection outcomes. However, the evolutionary trajectories of different defenses are connected at the moment of contact between host and parasite. Theory suggests that alleles controlling the steps in a defense cascade can become associated, even if they involve mechanistically independent pathways (Fenton et al. 2012). We focus on the joint evolution of avoidance and resistance, highlighting several reasons they may covary evolutionarily. These arguments could be extended to other defenses, such as tolerance.

First, avoidance and resistance may covary positively, such that lineages with increased resistance also show increased avoidance. Because avoidance and resistance both reduce fitness losses to parasites, high parasite risk could select for increased investment in both defenses. Some evidence supports this positive covariance. Late-flowering Silene latifolia families have both higher avoidance of anther smut infection and higher biochemical resistance (Biere & Antonovics 1996). Similarly, experimental selection of C. elegans for defense against a bacterial parasite results in populations with higher avoidance and resistance (Penley & Morran 2018, Penley et al. 2017, 2018). Dispersal stages of some C. elegans strains also have greater avoidance and resistance relative to other life stages, another indication of positive covariance (White et al. 2019).

Alternatively, avoidance and resistance may covary negatively. If hosts have limited energy to invest in defense, investment in resistance might select against avoidance, and vice-versa. This trade-off would generate a negative covariance. In addition, because avoidance acts earlier than resistance, avoidance could relax selection on later steps in the defense cascade (Fig. 2). By this argument, strong avoidance can render resistance selectively neutral, with selection for resistance only operating in the absence of avoidance. This “strategy blocking” effect could generate a negative covariance between avoidance and resistance (Hall et al. 2017).

Figure 2: Evolution of the defense cascade.

Avoidance and resistance can be conceptualized as filters that prevent the passage of parasites (Combes 2001). These filters occur in sequence, so that only parasites that bypass the avoidance filter reach the resistance filter. In bypassing both avoidance and resistance, a parasite can establish infection. Because avoidance occurs first in the cascade, its presence may reduce the selective pressure on resistance, such that resistance can only be the target of selection when avoidance fails.

Some evidence supports a negative covariance of defense strategies. Though this empirical pattern is typically attributed to trade-offs, current data cannot rule out strategy blocking as an explanation. Zylberberg et al. (2013) found that house finches (Carpodacus mexicanus) that avoided sick conspecifics had lower immune responses than individuals that did not avoid. Klemme et al. (2020) found evidence of negative correlations between avoidance and resistance against a trematode eye fluke across populations of Atlantic salmon (Salmo salar) and sea trout (Salmo trutta). However, in a similar study, avoidance was not found to covary with resistance (Klemme & Karvonen 2017). Another contradiction emerges from two studies of defenses against parasitoid wasps in Drosophila. When comparing two species, there was a negative relationship between avoidance and resistance: D. simulans larvae kill parasitoid wasp eggs more efficiently than D. melanogaster larvae, while adults of D. melanogaster, but not of D. simulans, reduce production of susceptible offspring in the presence of parasites (Lefèvre et al. 2012). A subsequent study of eight Drosophila species found no evidence of this negative correlation (Lynch et al. 2016), calling into question the scale at which we should test for trait associations.

Theory suggests that defense strategies may show distinct evolutionary dynamics when considered in isolation (Boots & Bowers 1999), yet considering them separately is only a first, albeit necessary, step. We expect that the evolutionary trajectories of avoidance and resistance are mutually dependent, such that the mean and variance of one defense trait drives the mean and variance of the other. Indeed, some of the outstanding cases in which variation in resistance evolution does not fit theoretical predictions may be explained in part by unaccounted-for variation in avoidance. More work is needed to understand the scale and circumstances in which positive or negative covariances occur. One important step will be to test basic assumptions about the fitness costs of avoidance, especially in relation to costs of resistance.

VI. The parasite’s perspective on avoidance

We have largely taken the host’s perspective on avoidance, but the evolution of any host defense depends critically on the evolution of the parasite’s strategy. We assume that, in reducing the rate at which parasites contact hosts, avoidance reduces parasite fitness. The degree to which this assumption holds true may vary with the mobility of infective stages and their ability to replicate outside the host. If avoidance reduces parasite fitness, do parasites evolve to overcome host avoidance defenses (Fig. 2)?

Parasites have strategies to increase contact with hosts. Parasites with free-living stages display active host-seeking behaviors, including changes in activity and trajectory in response to host cues (Chaisson & Hallem 2012, Mikheev et al. 2015). Complex-life cycle parasites show adaptations to increase transmission by manipulating vector feeding behavior (Bacot & Martin 1914, Jefferies et al. 1986) and making infected prey more vulnerable to predation by the parasite’s next host (Carney 1969, Thomas & Poulin 1998). Parasites may also evolve to avoid detection. Sexual transmission can theoretically select for parasites that disguise infection from potential mates (Knell 1999). Indeed, several studies of insect hosts find no evidence of avoidance of mates with sexually-transmitted parasites (Abbot & Dill 2001, Luong & Kaya 2005). Infective stages in the environment can also disassociate themselves spatially or temporally from cues that elicit host avoidance (Buck et al. 2018). Infective stages of the nematode Dictyocaulus viviparus disperse away from feces to grass patches more palatable to cattle by hitching rides on the projectile sporangia of the fungus Pilobolus (Jørgensen et al. 1982, Robinson 1962). It remains unclear, however, to what extent these parasite strategies have evolved in response to host avoidance.

If parasites can evolve directly in response to host avoidance, then avoidance defenses have the potential to drive coevolutionary dynamics via negative frequency-dependent selection or arms races. The nature of this potential coevolution hinges on the specificity of avoidance defenses. Do host genotypes vary in which parasite genotypes they can avoid (i.e., genotype-by-genotype interactions)? In an experimental coevolution study, Penley et al. (Penley et al. 2017) found that C. elegans hosts that coevolved with the parasite S. marcescens for 30 generations evolved increased avoidance of ancestral parasites. These same hosts did not, however, show greater avoidance of their contemporary, sympatric parasites. These results indicated that avoidance of S. marcescens varied with parasite genotype, with coevolving parasites able to adapt to overcome host avoidance. Many other avoidance defenses seem to be more general, such as reductions in feeding rate. These general defenses may be more likely to drive coevolutionary escalations (i.e., an arms race) or be so general that they impede pairwise reciprocal adaptation altogether. We currently lack the data required to fully evaluate avoidance as a driver of coevolution.

More fundamentally, avoidance may shape the potential for parasites to respond to selection. By reducing the number of parasites that contact hosts, strong avoidance defenses could sharply limit the effective population size of parasites (Hawley et al. 2021). For example, a solitary bat species hosts populations of parasitic mites that are less diverse and more subject to genetic drift than those found on a related gregarious bat species (van Schaik et al. 2014). If this effect is generalizable, then avoidance could impede parasites’ response to selection and the potential for reciprocal adaptation. Experimental systems will be critical for tracing the evolutionary trajectory of parasite populations in the presence and absence of avoidance.

VI. Open questions

Now that studies have documented parasite avoidance in a diverse array of host species, research on avoidance is well-positioned to address deeper questions that build on the theoretical literature developed to understand epidemiology, disease ecology, and the evolution of resistance. We highlight key questions, and emphasize how addressing them could lay the foundation for fruitful lines of inquiry on avoidance and the interaction of defense modes.

1. How does avoidance contribute to epidemic suppression in natural systems?

Avoidance strategies have proven extremely effective at suppressing epidemics in human populations. How does the presence or absence of avoidance change the trajectory of epidemics in other taxa? How do different types of avoidance – induced or constitutive, general or specific – change these expectations?

2. What is the genetic basis of parasite avoidance?

The degree to which variation in avoidance is genetically based dictates its potential to evolve in response to selection by parasites. Similarly, the evolutionary interplay between avoidance and downstream defenses depends in part on the evolutionary lability of avoidance, which will be determined to some degree by its genetic architecture. For avoidance mechanisms that are not genetically based, how is variation generated and maintained?

3. What are the fitness consequences of parasite avoidance?

How effective is avoidance at reducing infection, and how does this effectiveness translate into a measurable fitness benefit? Is avoidance costly, and does this cost vary predictably across the different forms avoidance can take (Table 1)? How do the efficacy and cost of avoidance compare to those of resistance?

4. How do avoidance and resistance influence one another on an evolutionary scale?

Given that these two defenses occur sequentially, does avoidance dictate selection on resistance? Does avoidance evolve faster than resistance as a result of its primary position? Is it more or less evolutionarily labile than resistance? How would evolutionary dynamics between avoidance and resistance change if the cascade is not strictly linear and induction of post-contact defenses feeds back to trigger avoidance?

5. To what extent does avoidance impose selection on parasites?

Because it acts before contact between host and parasite, avoidance may select for parasite traits that influence the steps leading up to transmission, including a parasite’s transmission mode, survival in the environment, or level of transmissibility. Do parasites evolve in response to avoidance? What are the parasite traits that are most likely to evolve in response to avoidance? How does widespread avoidance shape parasite populations?

VII. Conclusion

Avoidance has now been described in many forms across a wide diversity of host-parasite interactions. Avoidance has the potential to suppress epidemics, influence downstream selection on resistance, and impose selection on parasites. However, the relative importance of avoidance in the ecology and evolution of host-parasite relationships remains to be tested explicitly. The first step must be to quantify the effects of avoidance independently of other defenses. As theory develops around the evolution of avoidance and the defense cascade, it will be necessary to test assumptions of models empirically, including measuring the fitness consequences of avoidance for hosts and parasites. These are basic pursuits that will likely require more tractable study systems than have been the focus of avoidance research traditionally. In defining avoidance by its primary position in the defense cascade, we highlight the diversity of forms that host avoidance strategies can take, providing an avenue for future research on parasite avoidance to bridge taxonomic and mechanistic divides.

Terms and definitions

Avoidance

any host defense that reduces the contact rate between host and parasite

Contact rate

the rate at which uninfected hosts make physical contact with parasites, typically via infectious hosts or free-living parasites

Constitutive

always present or active; contrast with induced

Defense

any mechanism that reduces the loss of host fitness to parasites

Defense cascade

a series of host defenses that occur in sequence as an individual’s infection advances

Exposure

physical contact with parasites, typically with infectious stages in the environment

Horizontal transmission

transmission between individuals who may be unrelated, in contrast to vertical transmission, from parent to offspring

Induced defense

turned on or activated in response to parasite cues, not always present

Non-consumptive effects

the impact of enemies (predators, herbivores, parasites) on their victims aside from the interaction’s direct effect (e.g. death, infection)

Parasite

organism that reduces the fitness of a victim (the host), typically attacking a single host individual per life stage

Post-contact defenses

host defenses encountered by parasites after contact, once parasites have overcome host avoidance

Resistance

any post-contact defense that limits parasite fitness by reducing establishment and growth of an infection

Tolerance

any post-contact defense that preserves host fitness by limiting damage inflicted by a parasite, but does not reduce parasite fitness

Transmission mode

the method by which parasites make contact with new hosts