Parasitic infections are often followed by changes in host behaviour. Numerous and exquisite examples of such behavioural alterations are known, covering a broad spectrum of parasites and hosts. Most descriptions of such parasite-induced changes in host behaviour are observational reports, while experimentally confirmed examples of parasite genes inducing these changes are limited. In this study, we review changes in invertebrate host behaviour observed upon infection by parasites and discuss such changes in an evolutionary context. We then explore possible mechanisms involved in parasite-induced changes in host behaviour. Genes and pathways known to play a role in invertebrate behaviour are reviewed, and we hypothesize how parasites (may) affect these pathways. This review provides the state of the art in this exciting, interdisciplinary field by exploring possible pathways triggered in hosts, suggesting methodologies to unravel the molecular mechanisms that lead to changes in host behaviour.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Parasites often have profound effects on the animal hosts they invade by affecting development, physiology, morphology, evolution and ecology (Price 1980; Beckage 1997; Ros et al. 2008; Lefèvre et al. 2009a,b; Thomas et al. 2010). In many cases, host behaviour is also altered upon parasitic infection (Beckage 1997; Moore 2002; Lefèvre & Thomas 2008; Libersat et al. 2009). These changes range from slight alterations of already existing behavioural traits to the exhibition of completely new activities. Fascinating examples include Toxoplasma-infected rodents that lose their innate aversion to cats (Berdoy et al. 2000), Gordian worm–infected crickets committing suicide (Thomas et al. 2002) and lancet liver fluke-infected ants climbing into grass blades (Hohorst & Graefe 1961; Moore 1995; Libersat et al. 2009). These observed changes in host behaviour are often thought to be beneficial to the parasite as they may increase the rate of transmission and survival. Parasites that change host behaviour are diverse and comprise viruses, bacteria and a wide range of eukaryotes, including fungi, protozoa, parasitoids (including parasitic wasps) and parasitic worms. The latter encompass nematodes (roundworms), trematodes (flukes), cestodes (tapeworms), nematomorphs (Gordian worms or horsehair worms) and acanthocephalans (thorny-headed or spiny-headed worms). Besides the ability of parasites to manipulate host behaviour, an increasing body of evidence suggests that nonparasitic microbes (e.g. so-called beneficial microbes like gut microbes) can also alter host behaviour (see Ezenwa et al. 2012, for a review).
Most descriptions of parasite-induced changes in host behaviour are observational reports, while experimentally confirmed examples of parasite genes inducing these changes are limited (Kamita et al. 2005; Lefèvre et al. 2009a,b; Libersat et al. 2009; Hoover et al. 2011; van Houte et al. 2012). As a consequence, the underlying mechanisms behind parasite-induced behavioural alterations remain enigmatic. In addition, little is known on host pathways translating the parasite-induced signal to a particular behaviour. Insights into such mechanisms should provide important knowledge on the evolution of parasitic manipulative strategies and on animal behaviour in general. However, unravelling these mechanisms is a challenging task as both the processes by which the behavioural changes are induced and the behaviour itself are often complex and multidimensional (Thomas et al. 2010). For example, a single host may be infected by multiple parasites, and a single parasite can induce multiple behavioural alterations in a host.
This review describes different aspects of changes in invertebrate host behaviour observed upon infection by parasites, illustrated with several appealing examples. The adaptiveness of changes in host behaviour is addressed, and the concepts of multidimensionality and convergence of behavioural alterations are discussed. Possible mechanisms involved in parasite-induced changes in host behaviour are explored, and genes and pathways known to play a role in invertebrate behaviour (in nonparasitized conditions) are discussed. As the components of such pathways provide excellent entrees for parasites to change host behaviour, we hypothesize how parasites (may) affect these pathways. We did not aim to provide an exhaustive research of all available examples of behavioural alterations upon parasitic infection (see Moore 2002, for an excellent review). As the majority of the published data on this topic concerns invertebrates, vertebrate behaviour will not be addressed. For the latter, we refer to reviews by Klein (2003) and Tomonaga (2004).
On the extended phenotype and (non)adaptiveness
In an evolutionary context, changes in host behaviour upon parasite infection are examples of an extended phenotype, a concept introduced by Dawkins (1982). He stated that the observed host phenotype is a consequence of a parasite's gene being expressed. As such, parasitic manipulation can be defined as the alteration by the parasite of a host phenotypic trait in a way that enhances the parasite's probability of transmission and survival (Lefèvre & Thomas 2008; Thomas et al. 2012; Hughes 2013). This can be achieved for example by directly enhancing host-to-host transmission, by increasing the chance of finding a mate, or by dissemination of the parasite in a suitable location. In this view, changes in host behaviour are adaptive to the parasite.
However, observed changes in host behaviour are not necessarily beneficial to the parasite. Such changes may as well be adaptive to the host, aimed at reducing the fitness costs of infection, or pathological side effects (Poulin 2010; Thomas et al. 2012; Moore 2013). Host adaptive changes are for example seen in carpenter ants (Camponotus aethiops) that upon infection by the pathogenic fungus Metarhizium brunneum become unsociable to reduce the risk of dissemination of infection in the entire colony (Bos et al. 2012). Other host adaptive strategies include self-medication (e.g. feeding on compounds toxic to parasites) or behavioural fever (seeking temperature conditions unfavourable for parasite growth) to reduce parasitic load (see de Roode & Lefèvre 2012, for a review).
Other theories elaborate on the existence of alternative adaptive strategies. For example, infected hosts might apply a ‘mafia-like’ strategy where they cooperate with manipulative parasites instead of resisting them, thereby reducing fitness costs associated with manipulation (when cooperation is less costly than resistance; Ponton et al. 2006; Lefèvre et al. 2009a). Likewise, parasites might exploit host compensatory responses to their own benefit (Lefèvre & Thomas 2008; Lefèvre et al. 2009a). Although well-described theoretical frameworks exist for such strategies, experimental evidence for their existence is lacking and hard to obtain.
In many cases, behavioural alterations upon infection are intuitively interpreted as being adaptive to the parasite. Ideally, such assumptions require experimental evidence to be critically evaluated. However, this may be a challenging task, because measuring fitness of host (e.g. survival rates or fecundity) and parasite (e.g. transmission rates) is difficult in many host–parasite systems (Poulin 2010; Moore 2013). Throughout this article, we will use the term ‘parasitic manipulation’ in cases where this seems applicable, even in cases where evidence that changes in host behaviour are actually adaptive to the parasite might be lacking.
Even if host behavioural changes are a result of manipulation by parasites, the observed changes should be regarded as a shared phenotype resulting from the expression of host and parasite genes. This shared phenotype is the result of an evolutionary arms race and is determined by changes induced by the parasite and counteracted by the host (Poulin et al. 1994; Lefèvre & Thomas 2008; Lefèvre et al. 2008). The actual outcome and magnitude of changes might vary over time and space and are dependent on many factors, including host and parasite genetics and nongenetic factors like age, developmental stage, parasitic load, time after infection and environmental conditions (Poulin et al. 1994; Thomas et al. 2011; Moore 2013).
Multidimensionality: multiple alterations and multiple parasites
Instead of regarding parasitic manipulation as a ‘simple’ alteration of a single host behavioural trait, it should be seen as a multidimensional phenomenon (Thomas et al. 2010, 2012; Cézilly et al. 2013). Single parasites may cause multiple behavioural alterations within a host (simultaneously or sequentially), and, likewise, a single host can be infected by more than one parasite at the same time (Cézilly et al. 2000, 2013; Perrot-Minnot 2004; Lefèvre & Thomas 2008). For example, the cockroach Periplaneta americana can be parasitized by the parasitoid Moniliformis moniliformis, leading to increased locomotion activity (Moore 1983; Wilson & Edwards 1986), while the same cockroach species can be parasitized by the tropical wasp Ampulex compressa, inducing a zombie-like state and preventing it from any spontaneous locomotion (Gal & Libersat 2008). These parasites may have conflicting interests, and in case of coparasitation, this might be expressed in different manipulation patterns than for singly infected hosts. Parasites could for instance sabotage the manipulation induced by other parasites (Haine et al. 2005; Thomas et al. 2011; Cézilly et al. 2013). On the other hand, effects may become additive, leading to enhanced (behavioural) changes, or some parasites may profit from changes induced by other parasites. Also, the presence of one parasite might prevent infection or manipulation by another parasite. Wolbachia, for example, has been reported to reduce (Hedges et al. 2008; Martinez et al. 2012) or increase (Graham et al. 2012) the susceptibility of its host to virus infections. Therefore, when studying host manipulation, it should be taken into account that hosts might harbour more parasites that can influence the outcome of the studied manipulation.
Convergence: similar manipulations in different systems
Different parasites may encounter similar selective pressures when infecting a host and consequently may develop similar strategies to manipulate host behaviour (Poulin 1998; Ponton et al. 2006). Such convergent manipulative behaviour can be a consequence of similar proximate mechanisms, but can also be achieved by different mechanisms. Gordian worms (phylum Nematomorpha) induce suicidal behaviour in some arthropods, forcing these hosts to drown themselves by jumping into water, where the worms are released to mate (Thomas et al. 2002). A similar behaviour is observed for ants infected by mermithids (phylum Nematoda, unrelated to nematomorphs; Maeyama et al. 1994). Another example concerns two other unrelated parasites, the trematode Microphallus papillorobustus and the acanthocephalan Polymorphus minutus that both manipulate the behaviour of gammarid species (small freshwater crustaceans) by inducing them to move towards the water surface, where they are easily visible to aquatic birds that serve as the final hosts. In both parasite–gammarid associations, similar proteins are involved, suggesting molecular convergence in the proximate mechanisms of these manipulations (Ponton et al. 2006). Climbing behaviour upon infection is also observed in different systems, including caterpillars infected by baculoviruses (Hoover et al. 2011) and ants infected by Ophiocordyceps fungi (Hughes et al. 2011) or by lancet liver flukes (Dicrocoelium dendriticum; Hohorst & Graefe 1961; Libersat et al. 2009). Whether these alterations are caused by similar mechanisms is not yet clear.
In contrast to the above, related parasites may rely on different mechanisms to alter host behaviour. Baculoviruses induce enhanced locomotion activity and climbing behaviour in caterpillars (Fig. 1A). For two of these viruses [Bombyx mori nucleopolyhedrovirus (BmNPV) and Autographa californica nucleopolyhedrovirus (AcMNPV)], the viral protein tyrosine phosphatase (ptp) gene was found to be responsible for inducing enhanced locomotion activity in insect hosts (B. mori and Spodoptera exigua larvae, respectively; Kamita et al. 2005; van Houte et al. 2012). However, the underlying mechanism appears to be different; while in AcMNPV the phosphatase function of the encoded PTP protein is crucial for behavioural manipulation (van Houte et al. 2012), this is not the case for BmNPV. For the latter, absence of the PTP protein affected viral gene expression levels, possibly leading to the observed behavioural change through an unknown mechanism (Katsuma et al. 2012). Although the involvement of PTP in behavioural changes might have a common origin, the exact mechanism might have diverged over time.
Phylogenetic analyses can give valuable insights into the evolutionary history of behavioural manipulations and may help to understand whether certain manipulations have a common origin or evolved independently. In addition, it can be useful to assess the adaptive significance of such manipulations (Moore & Gotelli 1996; Poulin 1998). If similar changes in behaviour are observed in two distantly related host species, induced by two distantly related parasite species, this strongly directs at an adaptive significance (to either host or parasite) of the behavioural change.
Changes in host behaviour
The bodyguard: using the host to avoid enemies
Some parasites turn their host into a bodyguard to ensure the protection of the parasite from enemies like predators or hyperparasitoids (Grosman et al. 2008; Harvey et al. 2008; Maure et al. 2011). For example, Thyrinteina leucocerae caterpillars protect pupae of the braconid parasitoid Glyptapanteles sp. (Grosman et al. 2008). Once the parasitoid larvae leave the host to pupate, the host defends the pupae by knocking off predators with violent head-swings, resulting in reduced mortality rates of the parasitoid pupae. A similar phenomenon is observed with Pieris brassicae caterpillars parasitized by the braconid parasitoid Cotesia glomerata. Upon parasitoid egression from the host, the caterpillar spins a silk web over the parasitoid pupae and responds aggressively when disturbed, thus protecting the pupae from hyperparasitoids and predators (Brodeur & Vet 1994; Harvey et al. 2008). Bracoviruses and ichnoviruses (Polydnaviridae) appear to play beneficial roles in the development of some braconid and ichneumonid parasitoids (Burke & Strand 2012). The genomes of these viruses contain many ptp genes (although they seem to be evolutionarily unrelated to the baculovirus ptp gene inducing hyperactive behaviour), of which at least some play a role in immunomodulation of the caterpillar host (Falabella et al. 2006; Ibrahim & Kim 2008; Suderman et al. 2008). It is unknown whether these polydnaviral ptp genes play a role in the observed behavioural alterations in caterpillars infected by parasitoids.
A change in behaviour presumably aimed to reduce hyperparasitism and predation is also seen in parasitized Macrosiphum euphorbiae aphids, with a different behavioural change depending on the physiological state of the parasitoid Aphidius nigripes. Aphids containing nondiapausing parasitoids move to the upper surface of leaves to mummify, while those containing diapausing parasitoids leave the host plant and mummify in concealed sites. In both cases, the parasitoids appear to seek the optimal microhabitat for survival (Brodeur & McNeil 1989, 1992).
It should be noted, however, that host predation is not always detrimental to the parasite and in some cases could even be advantageous. Host predation can increase parasite transmission, as is seen in baculovirus-infected caterpillars that climb to the top of plants, where they are more visible to birds that predate on caterpillars. Baculoviruses survive a passage through the bird gut (where the pH is lower than in the larval midgut), and consequently the viruses disseminate more widely in the environment (Vasconcelos et al. 1996).
Where to go? Changes in locomotion behaviour and phototactic or geotactic behaviour
Many parasites alter their host's locomotion behaviour, by changing the speed or the direction of locomotion (including climbing behaviour, described below). This might increase the area over which the parasite (e.g. fungal spores or virus particles) is spread and increase the probability of finding a new host or a suitable place for survival. A change in host behaviour can also be adaptive to the host, for example, to prevent contamination of conspecifics. Changes in the direction of locomotion might be a response to, for example, gravity (geotaxis), light stimuli (phototaxis) or odours. A clear example of manipulation of host locomotion is observed in caterpillars infected with baculoviruses (see above).
Another fascinating and well-described case is the previously mentioned water-seeking behaviour of arthropods infected with a Gordian worm (Thomas et al. 2002; Fig. 1B). A recent study demonstrated that the water-seeking behaviour is the consequence of an altered response to light (positive phototaxis), combined with an increase in locomotion activity (Ponton et al. 2011). The behaviour is time-regulated and is only observed at night (not during the day or after 2–3 am). Using a parasitoproteomics approach, Biron et al. (2005) found that during a Gordian worm (Spinochordodes tellinii) infection of a grasshopper (Meconema thalassinum), a protein [CG31732-PD (isoform D)] was differentially expressed in the host's central nervous system (CNS). This protein is known to be involved in control of geotactic behaviour, suggesting that it may play a role in the observed behaviour (Biron & Loxdale 2013).
Several examples of other parasitic worms manipulating host locomotion behaviour have been reported (reviewed in Adamo 2002). The family Gammaridae comprise small freshwater crustaceans, which are intermediate hosts for parasitic worms, including the acanthocephalans P. minutus and Pomphorynchus laevis, and the trematode M. papillorobustus (Bethel & Holmes 1977; Helluy 1984; Cézilly et al. 2000). Uninfected gammarids dive downwards when disturbed and cover themselves in the mud. On the other hand, gammarids infected by parasitic worms (Fig. 1C) glide along the water surface to seek a solid support. This typical clinging behaviour exposes the gammarid to predators such as frogs, fish and birds, which are the parasites' next host. In gammarids infected with M. Papillorobustus, such a behaviour is characterized by positive phototaxis, negative geotaxis and aberrant evasive behaviour (Ponton et al. 2006). Polymorphus minutus does not induce positive phototaxis, but negative geotaxis and aberrant evasive behaviour were clearly observed in parasitized Gammarus pulex (Cézilly et al. 2000). A comparative proteomics study revealed an increased expression of a protein (aromatic l-amino acid decarboxylase) involved in serotonin synthesis in Gammarus insensibilis infected with M. papillorobustus, but not in G. pulex infected with P. minutus. This suggests that serotonin may function in positive phototaxis (Ponton et al. 2006). Another gammarid species, Echinogammarus stammeri, carrying P. laevis worms showed increased locomotion activity compared with conspecifics without worms (Maynard et al. 1998). This behavioural change is likely to be adaptive to the parasite, because there is increased consumption of infected compared with uninfected E. stammeri by the bullhead host, Cottus gobio (Lagrue et al. 2007).
An increase in locomotion activity has also been observed in hosts parasitized by parasitoids or infected with bacteria. The cockroach P. americana parasitized by the parasitoid M. moniliformis shows an increased locomotion activity and becomes positively phototactic (Moore 1983; Wilson & Edwards 1986). In the parasitoid Leptopilina heterotoma (Fig. 1D), the endosymbiotic bacterium Wolbachia decreases locomotion activity (Fleury et al. 2000), while in flies of the genus Drosophila, it either decreases or increases activity, depending on the Drosophila species and the bacterial strain (Peng et al. 2008). Wolbachia infection of the mosquito Aedes aegypti leads to increased mosquito locomotion, accompanied by an increase in metabolic rate. As Wolbachia infects brain tissue (Dobson et al. 1999), it was speculated that the increase in activity may be due to physiological changes in the CNS (Evans et al. 2009).
Towards the top: enhancing parasite dispersal
Remarkable are some examples of hosts showing climbing behaviour upon infection. This behaviour has been observed for different host–parasite associations, where the induced climbing is thought to contribute to enhanced transmission of the parasite. Caterpillars infected with a baculovirus climb to the top of plants, where they eventually die and liquefy, and in the process release progeny virus (Fig. 1A; Smirnoff 1965; Vasconcelos et al. 1996; Goulson 1997; Kamita et al. 2005). The virus has a higher chance to disseminate to lower parts of the plant when it is released from the top rather than from the lower branches (Vasconcelos et al. 1996; Goulson 1997). This behaviour has been described as ‘Wipfelkrankheit’ or ‘tree top disease’ (Smirnoff 1965; Evans 1986). Hoover et al. (2011) recently showed that a viral gene, ecdysteroid UDP-glucosyl transferase (egt), is involved in inducing this climbing behaviour of Lymantria dispar caterpillars infected with L. dispar multiple nucleopolyhedrovirus.
Another remarkable example is the fungus Ophiocordyceps that turns ants into zombies (Hughes et al. 2011). Infected ants drop down from canopy nests and climb up again in understory vegetation to bite into leaf veins before dying (Fig. 1E). This death grip is timed around solar noon and occurs at approximately 25 cm above the soil surface. This position is optimal for fungal sporulation, where the fungal fruiting body grows from the ant's brain and releases its spores (Andersen et al. 2009). Alterations in climbing behaviour occur in several other fungus-infected arthropods as well (see Roy et al. 2006, for a review). An example is the arctiid caterpillar Chionarctia nivea infected by the entomopathogenic fungus Entomophaga aulicae (Yamazaki et al. 2004). Sick caterpillars crawled up dead grass and herb stems to die at the highest parts, while healthy larvae wandered on the ground.
Lancet liver flukes use ants, including Formica fusca, as intermediate hosts between snails and cattle. The flukes encyst in the ant's hemocoel, except for one individual fluke, which enters the subesophageal ganglion (SEG) of the ant. Similar to the fungus Ophiocordyceps, lancet liver flukes induce the ants to climb and to anchor onto grass blades with their mandibles (Schneider & Hohorst 1971; Romig 1980). This behavioural change appears to be regulated in a clockwise manner: ants move to the top of grass blades in the evening and, if not consumed by cattle, descend again the next morning (Hohorst & Graefe 1961; Libersat et al. 2009).
Climbing behaviour might be related to geotactic or phototactic behaviour. In a dense canopy, more light is often present at the top than at the lower parts of the canopy. Moving towards the light might be involved in climbing behaviour, although this would only be relevant during daytime. Alternatively, negative geotaxis might be involved.
Under the spell: paralysing the host
Paralysing the host is a widespread strategy among parasitoid wasps (Libersat et al. 2009). Some of these cases involve the injection of (neuro-)toxins or venoms directly into the host's nervous system (Adamo 2002; Gal et al. 2005; Libersat et al. 2009). The tropical wasp A. compressa induces a zombie-like state in the cockroach P. americana, preventing it from any spontaneous locomotion. This enables the wasp to lead the cockroach to the wasp's nest, where it lays an egg on one of the cockroach's legs and subsequently buries the cockroach in the nest. The cockroach finally serves as a food supply for the emerging wasp larva (Gal & Libersat 2008). The authors hypothesized that this hypokinetic state is induced by neurotoxins present in the wasp's venom affecting the SEG (Gal & Libersat 2010).
The parasitoid Cotesia congregata feeds on larvae of the moth Manduca sexta. On the day prior to the emergence of the wasps' larvae, feeding and locomotion activity of parasitized M. sexta larvae decreases and ceases completely until its death (Adamo et al. 1997; Adamo 1998). Removal of the host SEG restores locomotion activity, indicating that the parasite inhibits locomotion activity via a neural pathway (Adamo 2002).
Sex and the parasite: changes in reproductive behaviour
Manipulation of reproductive behaviour has been reported for several parasites. Leptopilina boulardi filamentous virus (LbFV) infects the parasitoid L. boulardi and increases the tendency of L. boulardi females to superparasitize Drosophila larvae (i.e. larvae that have already been parasitized by another female wasp; Varaldi et al. 2003, 2009; Patot et al. 2009). Superparasitism is supposedly beneficial for the virus (Varaldi et al. 2006, 2009), because it permits horizontal transmission between L. boulardi embryos from different mothers. Leptopilina boulardi wasps use their ovipositors to detect chemical cues reminiscent of previous infestations, and it was hypothesized that LbFV affects chemoreceptor neurons in the ovipositors (Varaldi et al. 2009). In addition, virus infection reduces locomotion activity of L. boulardi females (Varaldi et al. 2005, 2006).
The endosymbiotic bacterium Cardinium manipulates the oviposition choice of the parasitoid wasp Encarsia pergandiella (Kenyon & Hunter 2007). These parasitoid wasps are autoparasitoids; in sexual (haplodiploid) forms of this species, female (diploid) eggs are laid in nymphal whiteflies (the primary hosts), while male (haploid) eggs are laid in conspecific or heterospecific pupal parasitoid wasps developing within the whitefly cuticle (secondary hosts). Cardinium induces thelytokous parthenogenesis in E. pergandiella, by which females develop from unfertilized, haploid, eggs. Uninfected wasps lay unfertilized eggs in secondary hosts (developing into males), but for Cardinium-infected hosts, a shift is seen towards primary hosts, where successful development into females is possible. This change in oviposition behaviour seems to be induced by Cardinium (Kenyon & Hunter 2007). Even so, the endosymbiotic bacterium Wolbachia has been shown to influence mate choice in spider mites (Vala et al. 2004) and fruit flies (Miller et al. 2010).
Helicoverpa zea nudivirus 2 (HzNV-2; Nudiviridae) infects the corn earworm moth H. zea and has been found to alter the moth's mating behaviour (Burand et al. 2005, 2012; Burand & Tan 2006). HzNV-2 (a.k.a. gonad-specific virus, Hz-2V) is sexually transmitted when males mate with an infected female and transfer the virus to uninfected females in subsequent mating attempts. While uninfected females stop calling after mating, infected females continue calling (Burand et al. 2005). Similarly, healthy females mating with healthy males cease mate calling, while females mating with infected males continue calling (Burand et al. 2005; Burand & Tan 2006). Most likely, the transfer of anti-calling factors, part of the seminal fluid, is blocked by a ‘virus-plug’ (Burand & Tan 2006). In addition, healthy males are more attracted to infected females, which exhibit a five- to sevenfold increased pheromone production than uninfected females (Burand et al. 2005). Infected males show no preference for infected over uninfected females (Burand & Tan 2006). The observed alterations in physiology and behaviour likely enhance virus transmission (Burand et al. 2005; Burand & Tan 2006).
Some parasites are known to completely feminize their male hosts. Wolbachia bacteria, for example, are transmitted by females only and manipulate host reproductive behaviour to increase the number of females in a population. One way is by inducing feminization: hosts that would develop as males develop into functional females (Engelstädter & Hurst 2009), changing them morphologically and behaviourally. The exact mechanism is unknown, but somehow Wolbachia interferes with the sex-determination pathway, probably by suppressing an androgenic gland (Werren et al. 2008). Feminization is also observed for the mayfly Baetis bicaudatus infected with the nematode Gasteromermis sp., where male hosts are feminized resulting in the formation of intersexes and complete sex reversals (Vance & Peckarsky 1996).
Modifying the vehicle: vector manipulation
Arthropods not only suffer from parasitic infections themselves, but often serve as vectors for transmitting human and veterinary parasites from one host to another. Evidence is accumulating that such vector-borne parasites not simply use their vector as a vehicle, but also alter vector traits to increase transmission rates, including alterations in feeding and probing, locomotion, host seeking and reproductive behaviour (Hurd 2003; Lefèvre et al. 2006; Bennett et al. 2008; Lefèvre & Thomas 2008). A better understanding of such alterations of vector behaviour and associated changes in parasite transmission may benefit the development of disease control strategies.
Arthropod-borne viruses (arboviruses) increase probing time and/or frequency of blood feeding by mosquitoes (Grimstad et al. 1980; Platt et al. 1997; Bennett et al. 2008; Lima-Camara et al. 2011). Aedes triseriatus infected with La Crosse virus (Bunyaviridae) probes more, while engorging less blood per feeding (Grimstad et al. 1980). Dengue virus (Flaviviridae) causes a longer probing and feeding time in A. aegypti females (Platt et al. 1997), and recently, Lima-Camara et al. (2011) reported a 50% increase in locomotion activity of A. aegypti mosquitoes carrying this virus.
Several studies show that the protozoan Plasmodium (Fig. 1F), a causative agent of malaria, increases feeding frequency and feeding persistence of its mosquito vector (Moore 1983; Koella & Packer 1996; Koella et al. 1998; Anderson et al. 1999, 2000). Infection with Leishmania parasites increases feeding persistence and enhances feeding on multiple human individuals (Rogers & Bates 2007). Trypanosoma parasites infecting tsetse flies (Diptera: Glossinidae) promote their transmission by manipulating tsetse feeding behaviour (van den Abbeele et al. 2010). The parasites achieve this by modifying saliva composition, resulting in a reduced antihaemostatic potential of the saliva. This hampers feeding performance and, consequently, prolongs feeding. Such behavioural alterations are thought to enhance parasite transmission by an increased contact rate between vector and host.
Examples showing a reduction in feeding or locomotion behaviour of insect vectors are also known. Vesicular stomatitis virus (Rhabdoviridae) reduces blood feeding of Culicoides sonorensis midgets significantly at the peak of the virus titre (Bennett et al. 2008). Decreased flight activity was observed for Culex tarsalis mosquitoes infected with Western equine encephalomyelitis virus (Togaviridae; Lee et al. 2000). Flight ability is an important epidemiological factor in arbovirus transmission. Whether the last two examples are the consequence of parasitic manipulation or rather reflect pathological side effects of infection or vector adaptations is not known.
Behavioural changes have not only been found for vectors transmitting human or veterinary parasites, but also for insect vectors transmitting plant viruses. Recently it was reported that male Frankliniella occidentalis thrips infected with Tomato spotted wilt virus (Bunyaviridae) show an increase in feeding frequency, thereby enhancing the chance of virus transmission from vector to plant (Stafford et al. 2011). Although many examples of parasites affecting vector behaviour have been described (see Lefèvre & Thomas 2008, for an overview), direct evidence that the observed changes indeed increase parasite transmission is still scarce.
Despite the wide range of examples of manipulation of host behaviour, little is known concerning mechanisms behind these manipulations. The question arises how parasites manipulate the behaviour of their host, or more specifically, which parasite and host genes and/or proteins are involved in this change in behaviour? To date, only two parasite genes (the baculovirus genes ptp and egt) inducing host behavioural changes have been identified, alluded to above (Kamita et al. 2005; Hoover et al. 2011; van Houte et al. 2012). Also, wasp venoms are known to induce host behavioural changes (Gal & Libersat 2010), although responsible components within the venom still have to be identified. To date no host genes involved in behavioural manipulation have been identified. To comprehend mechanisms of behavioural manipulation by parasites, a thorough understanding of the genetic pathways underlying invertebrate behaviour as such – that is, not in the context of parasitic manipulation – is required. Studying the interplay between genes and behaviour is a relatively new discipline that is rapidly expanding. A wealth of genetic information has become available with the whole-genome sequencing of invertebrate model organisms, including Drosophila melanogaster, Caenorhabditis elegans and A. aegypti. This knowledge, combined with insights from neurobiology, provides a firm base for research on behavioural genetics. In recent years, several genes have been found to play a role in the modulation of specific behaviours (reviewed in Sokolowski 2001), and many pathways of behavioural adaptation appear to be conserved across species and, in some cases, even across phyla (Beck 2006). This high degree of conservation has led to the implementation of the candidate gene approach in the field of behavioural genetics (Fitzpatrick et al. 2005; Hoedjes et al. 2011). This is an experimental approach based on the assumption that the involvement of a specific gene in a behavioural phenotype is conserved among organisms. The foraging (for) gene, encoding protein kinase G (PKG), was the first example of a gene implicated in feeding-dependent locomotion across species (Fitzpatrick & Sokolowski 2004; Fig. 2 and see below).
The fact that many pathways underlying particular behavioural traits are highly conserved renders these pathways suitable targets for manipulation by parasites. We therefore propose to extend the candidate gene approach to the level of parasitic manipulation. Not only the role of a specific gene in determining a behavioural phenotype is assumed to be conserved among different species, but the pathways that are targets for behavioural manipulation by parasites may also be conserved. Although exact mechanisms of interference with these pathways may vary among parasites, this approach aids in identifying potential manipulation targets exploited by parasites.
The main drawback of this approach is that it provides a relatively narrow view and only those pathways already known to be implicated in behaviour can be considered. To reduce the risk of overlooking (unknown) pathways, additional genome- and proteome-wide approaches therefore need to be exploited. Highly sensitive and quantitative transcriptomic and proteomic analyses are relatively new disciplines that have become very important in studying parasite–host interactions at a molecular level. Techniques used for gene expression profiling include serial analysis of gene expression (SAGE) or deep sequencing of transcriptomes that may be combined with microarrays for more precise quantification (Ramsay 1998; Knox & Skuce 2005). Such approaches provide differential expression profiles between parasitized/infected and uninfected individuals, revealing genes affected by parasitism, although not necessarily with a relation to host behaviour. If the parasite gene(s) that induces a behavioural trait is known, a more directed search is possible by comparing individuals infected with a ‘wild-type’ parasite with individuals infected with a ‘mutant’ parasite in which the gene of interest is knocked out.
Proteomic quantification methods have the advantage not only to show whether a gene is expressed, but also provide information on protein levels and/or modifications. Such data are highly relevant because many signalling pathways use alterations in the phosphorylation status of proteins as a signal. ‘Parasitoproteomics’ aiming to identify peptides expressed by the host and the parasite, and their possible crosstalk, has already been applied in several parasite–host systems (Biron et al. 2005; Ponton et al. 2006; Lefèvre et al. 2009a; Kariithi et al. 2011). An example relevant to behavioural manipulation is the proteomic analysis of insects infected and manipulated by Gordian worms (see above), revealing the presence of parasite proteins that mimicked host proteins belonging to the Wnt family, parasite proteins involved in neurotransmitter release and a host protein involved in geotactic behaviour (Biron et al. 2005, 2006).
Both genome- and proteome-wide approaches often render complex data sets, from which the genes/proteins/pathways that are related to changes in behaviour, need to be extracted. Whatever strategy is chosen, the candidate gene approach may assist in selecting particular pathways exploited by parasites for the purpose of manipulating host behaviour. Below an overview is given of possible points of interference and a schematic on how the various pathways could be interconnected (Fig. 3).
Genes involved in (feeding-related) locomotion behaviour
Research on invertebrate behavioural genetics led to the discovery of a number of genes involved in the induction of locomotion behaviour. For some of these genes, their role as behavioural determinant is conserved across several species, and in one case, even across phyla [neuropeptide Y (NPY), see below]. One of the most extensively studied examples is the for gene, encoding a cGMP-dependent protein kinase (PKG). This gene was first discovered in D. melanogaster, where allelic variation in for was associated with a natural polymorphism in foraging behaviour. ‘Rover’ larvae show a higher locomotion activity than ‘sitter’ larvae and move over a larger distance when placed on a nutritive source. This increase in activity is accompanied by higher for mRNA and PKG protein levels in rovers compared with sitters (Osborne et al. 1997). In Apis mellifera, a similar link exists between the for homolog Amfor and a behavioural transition from nursing to foraging bees (Ben-Shahar et al. 2003), and subsequently a link between PKG and variation in locomotion was also identified in ants (Ingram et al. 2005, 2011; Lucas & Sokolowski 2009), locusts (Newland & Yates 2006), beetles (Garabagi et al. 2008), and even beyond the order Insecta, in several nematode species (Fujiwara et al. 2002; Hong et al. 2008). These data plus phylogenetic analyses suggest that modulation of PKG levels is a highly conserved mechanism for the regulation of (feeding-related) locomotion behaviour, because a link between PKG and behaviour has so far been found in four different insect genera and in nematodes (Fig. 2 and (Fitzpatrick & Sokolowski 2004).
Another gene involved in the regulation of locomotion behaviour is neuropeptide Y (NPY, a.k.a. NPF; Brown et al. 1999). Its role seems to be conserved across animal phyla: NPY regulates foraging and social behaviour in worms and flies (de Bono & Bargmann 1998; Wu et al. 2003), while its mammalian homolog is responsible for the regulation of feeding behaviour in vertebrates (Stanley & Leibowitz 1985; Beck 2006). A mutation in a C. elegans NPY receptor-like gene is associated with a behavioural variation between social and solitary foragers (de Bono & Bargmann 1998), while in D. melanogaster NPY is implicated in the transition from feeding to wandering behaviour during larval development (Wu et al. 2003).
The metabolic neuropeptide adipokinetic hormone (AKH) plays a role in starvation-induced locomotion as was first described in D. melanogaster (Lee & Park 2004; Isabel et al. 2005). Upon starvation, wild-type flies display increased locomotion behaviour prior to death, while AKH-deficient flies lack this increase in locomotion and are more resistant to starvation-induced death (Lee & Park 2004; Isabel et al. 2005). AKH also mediates transport of lipids from the insect fat body to the hemolymph (Beenakkers et al. 1985) and increases hemolymph sugar levels (Gade et al. 1997). The double function of AKH in regulating locomotion activity and raising hemolymph lipid and sugar levels may be considered complementary, because AKH-induced locomotion probably results from AKH-regulated lipid and sugar metabolism, thereby maximizing survival chances for the fly (Lee & Park 2004; Isabel et al. 2005). In moths, cockroaches and firebugs, injection of AKH into the CNS induced locomotion (Milde et al. 1995; Kodrik et al. 2000; Wicher et al. 2006). Variation in akh mRNA levels also coincides with the natural occurrence of rover and sitter D. melanogaster larvae that display an allelic variation in the for gene (Kaun et al. 2008). It is hypothesized that for directly regulates akh transcription or alternatively that the above-mentioned npy acts as a link between for and akh expression.
Tachykinins form a large family of vertebrate neuropeptides, implicated in numerous processes in the nervous, gastrointestinal and vascular systems (Otsuka & Yoshioka 1993). In gerbils, activity of a brain-specific tachykinin receptor is involved in locomotion behaviour (Nordquist et al. 2008). In invertebrates, an ancestrally related family of tachykinin-related peptides (TKRPs) has been identified (Nässel 1999), which function in modulation of muscular activity, regulation of diuresis in the Malphigian tubules, and as release factors for AKH (Nässel 2002). A role for TKRPs in locomotion was shown in D. melanogaster, where TKRP-deficient flies displayed aberrant olfactory perception and enhanced locomotion activity (Winther et al. 2006) and aberrant spatial orientation behaviour (Kahsai et al. 2010). These studies suggest a function for TKRP in the modulation of locomotion activity in flies (Nässel & Winther 2010).
Given the apparent similarity between PKG, NPY, AKH and tachykinin in the modulation of feeding-related locomotion behaviour, a single pathway may exist in which (some of) these molecules act together (Sokolowski 2003; Fig. 3). The highly conserved role that these genes play in modulating behaviour across species or even taxa renders them optimal targets for parasites to intervene with host behaviour. Although the evidence is scarce, several studies indicate that (some of) these pathways could indeed be altered by parasites, as discussed below.
Parasitic manipulation of AKH has not been reported so far; however, it is well established that many parasites affect host metabolic processes, some of which could be AKH-mediated. Several studies have shown increased sugar levels in the hemolymph of parasitized insects (Dahlman 1975; Dahlman & Vinson 1975, 1976; Thompson 1982). For example, Heliothis virescens larvae parasitized by the braconid wasp Microplitis croceipes showed elevated hemolymph sugar levels and higher fat body glycogen levels until 5 and 6 days after parasitic oviposition, respectively. In addition, hemolymph lipid levels were elevated (Rivers & Denlinger 1995; Nakamatsu & Tanaka 2004). AKH may also be involved in manipulation of the beetle Tenebrio molitor by the tapeworm Hymenolepis diminuta, a parasite that affects oviposition behaviour (Hurd 2009). Parasitized hosts showed lower levels of vitellogenin, a protein thought to be regulated by AKH (Hurd 2009).
In the freshwater snail Lymniae stagnalis, increased expression of the npy gene is observed in the CNS upon parasitization by the trematode Trichobilharzia ocellata, a parasite that alters egg-laying behaviour of its host (Hoek et al. 1997). Direct injection of Lymniae NPY peptide in nonparasitized snails inhibits egg laying, indicating that the parasite may alter host reproductive behaviour by modulation of host npy expression (de Jong-Brink et al. 1999).
A parasite protein (e.g. the above-mentioned baculovirus PTP protein) could target one of these host proteins by interfering directly with its expression or synthesis or by targeting proteins more upstream in a pathway. For example, PTP could target PKG, which could subsequently lead to changes in neurohormones such as NPY, AKH and tachykinin (Fig. 3).
Neurotransmitters and locomotion behaviour
Locomotion in invertebrates also involves the action of several monoamine molecules, including dopamine, serotonin and octopamine (Bicker 1999; Saraswati et al. 2004; van Swinderen & Andretic 2011; Fig. 3) that may act as neurotransmitters or neurohormones. For example, dopamine and octopamine are known to enhance locomotion activity in D. melanogaster, while tyramine decreases locomotion (Saraswati et al. 2004; Draper et al. 2007; Akasaka et al. 2010), and in A. mellifera dopamine enhances locomotion activity (Fussnecker et al. 2006), but injection of octopamine and its precursor tyramine led to decreased locomotion activity (Fussnecker et al. 2006).
In several cases of behavioural manipulation, neurotransmitter levels were found to be altered (Adamo 2002, 2013). Lefèvre et al. (2009a) hypothesized that host phototactic behaviour can be altered by interference with the serotonergic system. For example, higher levels of brain serotonin seem to be functionally linked to the positive phototactic behaviour observed in G. pulex infected with P. laevis (see above) and Pomphorhynchus tereticollis (Tain et al. 2006) and in G. insensibilis infected with M. papillorobustus (Ponton et al. 2006; see above). Manduca sexta larvae parasitized by the wasp C. congregata have increased octopamine levels (Adamo et al. 1997), which may coincide with the observed decrease in movement (see above). In the snail Biomphalaria glabrata, levels of dopamine, its precursor l-dopa, and serotonin decreased upon parasitization by the human blood fluke Schistosoma mansoni (Manger et al. 1996). Serotonin levels also decreased due to the uptake of host serotonin by these larval parasites (Yoshino et al. 2001). In P. americana cockroaches stung by the parasitoid A. compressa, dopamine was detected in the wasp venom injected into the cockroach's CNS (Gal et al. 2005). A proteomic analysis of tsetse flies (Glossina) infected with Trypanosoma shows the presence of two pyridoxal-dependent decarboxylases, enzymes that catalyse the final step in the synthesis of dopamine and serotonin, suggesting a modification of dopamine and/or serotonin in the brain of the tsetse flies (Lefèvre et al. 2007). Parasites that manipulate host behaviour via neurotransmitters (e.g. parasitic worms infecting gammarids, see above) most likely do this by interfering with the synthesis or release of these agents in the CNS (Lefèvre et al. 2009a; Fig. 3).
Circadian-clock regulated genes involved in feeding-related locomotion and geotaxis
Insect behaviour is controlled in a 24-h cycle by the circadian system, allowing the organism to respond to rhythmic environmental changes, related to for example light, temperature and food availability (reviewed in Allada & Chung 2010). The circadian system consists of three components: (i) the pacemaker, representing the core circadian clock; (ii) the input pathway, synchronizing the clock to the environment; and (iii) the output pathway, comprising genes that are involved in circadian-clock-regulated behaviour, thereby representing a link between circadian rhythm and behaviour (Jackson et al. 2001; Allada & Chung 2010). Parasitic alterations of host behaviour may also be regulated in a clockwise manner. For example, water-seeking behaviour in arthropods infected with Gordian worms is observed at night only (Ponton et al. 2011) and ants parasitized by lancet liver flukes climb to the top of grass blades in the evening (Libersat et al. 2009). Clock genes are possibly involved in such behavioural manipulations.
Takeout (to) (Fig. 3) is a circadian-clock-regulated gene that regulates feeding behaviour in D. melanogaster (Sarov-Blat et al. 2000). Flies kept under starvation conditions displayed increased locomotion activity, which was accompanied by an increase in to mRNA and protein levels. This increase in activity under starvation was not observed in mutant flies carrying a deletion in the 3′UTR of the to gene. In addition, to expression was blocked in all arrhythmic central clock mutants, indicating that to acts as a molecular link between the circadian clock and the regulation of feeding behaviour under starvation conditions (Sarov-Blat et al. 2000; Meunier et al. 2007). Takeout also acts as a modulator of juvenile hormone (JH) levels (Meunier et al. 2007), which is a hormone also involved in locomotion (Belgacem & Martin 2002; Fig. 3). Although no reports on parasitic manipulation of the takeout pathway are currently known, parasites may influence circadian rhythms in their hosts. Recently, Biernat et al. (2012) showed that a DNA repair protein (PHR2) encoded by the baculovirus Chrysodeixis chalcites nucleopolyhedrovirus (ChchNPV; van Oers et al. 2005) can mimic the function of mammalian cryptochromes, essential regulators of the circadian clock. Whether this affects insect circadian rhythms remains to be elucidated.
The neuropeptide pigment-dispersing factor (pdf; Fig. 3) is highly expressed in a subset of pacemaker neurons and is needed for proper circadian locomotion activity (Renn et al. 1999). PDF is involved in geotactic behaviour in D. melanogaster as was demonstrated with two fly lines displaying opposing geotactic behaviour, in which pdf was identified as one of the differentially expressed genes in a cDNA microarray analysis (Toma et al. 2002). In addition, pdf-null mutants showed strong negative geotaxis, while transgenic insertion of an extra copy of the pdf gene resulted in a modest increase in positive geotaxis (Toma et al. 2002). This is in line with the finding that a neuronal PDF receptor is involved in both circadian rhythmicity and geotaxis in Drosophila (Mertens et al. 2005).
Besides inducing hypermobile behaviour (see above), baculoviruses also induce larval hosts to move to elevated positions of plants or trees (Smirnoff 1965; Evans 1986; Goulson 1997). Some baculovirus infections, however, give opposite effects: infected Operophtera brumata larvae tend to migrate downwards (Raymond et al. 2005). If the climbing behaviour of baculovirus-infected larvae is an example of modulation of geotaxis (alternatively it could be a modulation of phototaxis), then the pdf gene provides an excellent candidate to play a role in the induced climbing behaviour. This can also be the case for ants infected with the lancet liver fluke D. dendriticum (Schneider & Hohorst 1971; Romig 1980) and Ophiocordyceps-infected ants (Hughes et al. 2011).
Conclusions and outlook
Research on behavioural manipulation by parasites is at a turning point; a shift from mere observations to unravelling molecular mechanisms by which parasites alter host behaviour is dawning. In this review, we discussed different examples of parasitic manipulation of invertebrates, and we explored possible mechanisms behind such manipulations, confirming that experimental data on underlying mechanisms are still scarce. Relatively novel techniques such as transcriptomics and proteomics made important initial contributions to this field and will be invaluable in the future. Combining the information obtained in such genome-/proteome-wide analyses with what is known from the genetic basis of behaviour in unparasitized animals is a relevant next step in unravelling mechanisms of parasitic manipulation, and several studies indeed support the hypothesis that genes known to have a conserved role in insect behaviour could be targets for manipulation by parasites. Therefore, the ‘parasite-extended’ candidate gene approach, based on the assumption that parasites hijack existing signalling pathways involved in behavioural traits, will be an important new tool to deepen our understanding of the intricate strategies by which parasites alter host behaviour. Vast amounts of genomic information have become available for many organisms and will aid in this research, providing novel insights into the molecular basis of behavioural manipulation by parasites.
The authors thank Katja Hoedjes, Hans Smid, Just Vlak and Louise Vet for useful discussions and reading previous versions of this manuscript and four anonymous reviewers for constructive and helpful comments. Basil Arif is sincerely thanked for critically reading the manuscript. SVH and VIDR are both supported by the Program Strategic Alliances of the Royal Dutch Academy of Sciences (project 08-PSA-BD-01), and VIDR is supported by a VENI grant of the Netherlands Organisation for Scientific Research (project 863.11.017).
S.v.H. and V.I.D.R. are researchers interested in host-parasite interactions, working with baculoviruses and lepidopteran insects. M.M.v.O. leads the Virology research group, which focuses on virus-host interactions in insects and plants.