The nature of the relationship between hosts and their parasites has been the focus of a wide range of studies (reviewed in [1]). One particular area of current interest is in the types of behavioural change that can be induced in a host following an infection. Studies of horizontally acquired nematode, trematode or protozoan parasites in a range of arthropods indicate that behavioural changes may be induced through altered levels of gene expression in the host CNS [2,3]. The precise mechanisms are not fully understood but in many instances the behavioural change is predicted to be the consequence of selection acting on the parasite to promote its own transmission. Behavioural changes are also induced in hosts following infection with vertically acquired, intracellular bacteria such as Wolbachia, Cardinium or Spiroplasma. These changes in host biology may also be driven by selection acting to promote bacterial transmission, but in this case the primary route is from mother to offspring. Horizontal transmission, whilst it undoubtedly occurs, is not the primary transmission route for these bacteria [4] and is therefore likely only to be important on an evolutionary timescale.

There are several main intracellular, primarily vertically acquired, bacterial lineages that are known to infect arthropods. These include bacteria from the Rickettsiales such as Wolbachia, and Rickettsia, members of the Bacteroidetes group such as Cardinium or Flavobacteria, Mollicutes such as Spiroplasma and Arsenophonus, a newly described bacterium lineage belonging to the Gamma-Proteobacteria. Each of these is inherited primarily through the female germ line and as such is always found within female reproductive tissue, but many are found within a much wider range of host tissue types and/or are present throughout the haemocoel. These bacteria act as selfish genetic elements, promoting their own transmission by increasing the reproductive success of infected females. They achieve this through a variety of mechanisms: (1) by causing cytoplasmic incompatibility (where uninfected females produce no offspring when mating with an infected male) (2) by biasing the sex ratio in favour of females via either male-killing (either early or late, i.e., as embryos or as larvae) or through feminisation of genetically male individuals or (3) through inducing parthenogenesis [4,5]. A summary of bacterial phenotypes and examples of taxa in which these have been found is given in Table 1. Thus far only Wolbachia has been shown to cause all these phenotypes, whereas for Cardinium there is evidence for all of these phenotypes except male-killing. In contrast, the other microbes are all known to act as male-killers and Rickettsia can also be associated with parthenogenesis.

Vertically acquired bacterial symbionts are extremely common in arthropods, with the most intensively studied bacterium in this context, Wolbachia, potentially infecting more than 60% of insect species [6]. Although Wolbachia could well represent the most widespread symbiont in insects, various surveys of insects and arachnids show that other symbiont types may also be highly prevalent [7,8,9]. Some strains, such as Cardinium, appear to be more common in arachnids than in insects, an observation that indicates that individual bacterial types do not infect all arthropod groups equally [10,11].

The wide range of hosts that can be infected with symbionts such as Wolbachia is matched by a correspondingly wide range of known additional effects on host physiological and fitness traits such as female fecundity and male fertility (Table 2). Although the physiological consequences of an endosymbiont infection in any given species are not always fully understood it is clear that these bacteria can interfere in a diverse range of processes and in some cases may even be essential to host survival or reproduction. For example the presence of Wolbachia has been shown to be necessary for host oogenesis in the hymenopteran parasitoid Asobara tabida [12]. Similar effects are known from associations between symbionts and their nematode hosts. Physiological effects are however not restricted to reproductive processes; symbionts may also alter host susceptibility to a range of biotic and abiotic factors (summarised in Table 2). For example, some bacteria have been shown to provide benefits to their host through conferring increased resistance to pathogens, elevating protection from parasitoids, altering susceptibility to insecticides and improving thermotolerance. These kinds of beneficial effects are perhaps not unexpected given that bacterially mediated resistance against pathogens and parasitoids is already known from studies of other types of bacteria to those that are the focus of this study. For example, the Hamiltonella defensa bacterial symbiont that infects the pea aphid Acyrthosiphon pisum can block larval development of the hymenopteran parasitoids Aphidius ervi, A. eadyi [13] and Lysiphlebus fabarum [14].

It is known that the speed at which phenotypic changes in the host can be effected by endosymbionts may be rapid [15] and it is also known that bacterial phenotypes have the potential to change [16]. Features of the bacterial genomes themselves may play a part in the latter, for instance mobile genetic elements and the presence of numerous phage are proposed to contribute to the rate at which the phenotype that is induced in the host changes from one state to another, for example from pathenogenicity to protective mutualism [13]. There is clearly, however, also a role for the interaction between host and bacterial genomes in determining phenotype because the same type of symbiont can cause different effects in different hosts. As an example, Spiroplasma causes male-killing in several Drosophila species, yet other spiroplasmas, such as that which infect the pea aphid, Acyrthosiphon pisum [17], are not male-killers. In fact, even the same bacterial strain need not always have the same effects. This is illustrated in Table 3, which lists a range of different phenotypes exerted by identical Wolbachia strains infecting different hosts. A single infecting strain may even cause different phenotypes depending upon its precise location within the host. For example the effect of the same infecting Rickettsia strain in whiteflies is shown to be dependent upon whether it is found widely throughout the host’s hemocoel or if it is restricted to within bacteriocytes [18].

In addition to the physiological effects already described (Table 1,Table 2,Table 3) it is known that vertically acquired endosymbiont infections can result in altered behaviour of the infected host. A summary of the types of behavioural shifts that have been observed thus far is given in Table 4. The types of change that are documented can be grouped into ‘reproductive behaviour’ (i.e., those behaviours directly involved in reproduction) and non-reproductive behaviour (i.e., those not directly involved in reproduction). Reproductive behaviours that are known to be directly under the influence of endosymbiont infections include mating preference, courtship, mating duration, mating frequency and post-mating behaviours of females such as oviposition. Non-reproductive behaviours, which are less frequently reported in the literature as being influenced by endosymbiont infections than reproductive behaviours, include dispersal in a linyphiid spider [19] and competitive behaviour of larvae in Drosophila [20].

In many of the cases described above a direct link is established between the infecting bacterium and the behaviour itself. Wild-type behaviour is restored by curing the host of its infection using antibiotics or by treatment at elevated temperatures. In this sense, the behavioural response is a direct result of bacterial interaction with its host and the behavioural change itself may be one of the mechanisms through which increased bacterial transmission is achieved. One illustrative example is the aggregation of eggs by spider mites [21], which is proposed to increase the chances of mating between individuals carrying compatible bacterial strains. Behavioural changes that increase the competitive ability of offspring are also likely to increase bacterial transmission relative to uninfected individuals [20]. The precise mechanisms underlying these behavioural effects on the host remain unknown but it is interesting to note that in the case of the linyphiid spider bacteria were found specifically in regions of nervous tissue associated with motor function [19].

In addition to direct influences of bacteria on the behaviour of their host there are also instances when host behaviour is altered indirectly, for example in response to manipulation of the population sex ratio. One extreme example of this is observed in a species of Acraea butterfly where a male killing endosymbiont results in a highly female-biased sex ratio [22]. In response to this there is sex-role reversal of lekking behaviour where females form leks around the remaining males. The case of female lekking in Acraea represents an extreme and rapid behavioural shift in response to a bacterial phenotype. Similarly, strongly female-biased sex ratios caused by Wolbachia infection in another butterfly, Hypolimnas bolina, lead to increased female mating frequency, and simultaneously to increased male fatigue associated with more prudent sperm investment per mating [23]. However there may be many more instances of subtle (and potentially transient) shifts in behaviour that might also have long-term consequences. For example, dispersal of female Erigone atra spiders is shown to be under the influence of a Rickettsia infection [19]. This is expected to result in localised perturbations of the sex ratio within the spider meta-population and as a consequence, localised variation in the strength of selection acting on traits that are sex-ratio dependent such as those involved in male-male competition or in the levels of male investment in attracting a mate.

Perhaps the ultimate evolutionary novelty is the creation of a new species. Endosymbiont infections have been shown to be mechanisms by which gene flow can be restricted, for example through causing cytoplasmic incompatibility. Such restrictions on gene flow, even if transient or spatially restricted, may be an important component of the speciation process in particular cases [4,24]. It is unclear whether endosymbiont driven shifts in behaviour alone can also ultimately create new species although traits such as reduced dispersal e.g., as shown in the spider Erigone atra [19], likely reduce genetic exchange within the wider meta-population. Such effects, even if transient and/or spatially restricted, potentially contribute to the development of reproductive isolation and thus influence the broader process of adaptation and speciation.

Whilst the hypothesis that endosymbiont driven changes in behaviour can lead directly to reproductive isolation is difficult to test, it is certainly true that changes in both reproductive and non-reproductive behaviours such as mating preference or dispersal likely influence genetic diversity at a local scale. For instance restriction in the number of successfully reproducing individuals via male-killing or CI will lower the effective population size [25]. Such impacts on population size and the various impacts on reproductive traits will have further consequences for the local intensity and dynamics of sexual selection and conflict [26,27,28]. Similarly, a skewed population sex ratio will also lead to shifts in the strength and shape of sexual selection and as a consequence altered female behaviour [22,23]. Changes in mating preference may also alter the predicted linkage disequilibria amongst physically unlinked loci. Migrants canalso perturb linkage disequilibria if incoming genotypes differ from those in the resident population. These and other factors influence the ability of populations locally to adapt through natural selection in two ways. First, incoming genes or genotypes may be maladapted to the local environment. Second, changes in effective population size affect the sensitivity of local populations to the effects of random genetic drift.

Endosymbiont shifts in arthropod behaviour potentially reduce the ability of a population to adapt and persist if the outcome is reduced intraspecific variability, but in certain instances they may also favour population expansion, for instance through inducing parthenogenesis or by altering female mating rate [29]. Such rapid shifts can allow a population to expand into an empty niche. Shifts in behaviour caused by endosymbionts may also promote and maintain rather than reduce variation within a population, for instance if they increase levels of assortative mating that reduce the loss of rarer variants.

In many cases it is difficult to assess whether or not there has been sufficient time since an infection was acquired for the host and bacterium to reach equilibrium. Examples that illustrate that equilibrium has not always been reached include that of Hypolimnas butterflies in Polynesia, where the frequency of a male-killer decreased rapidly over a short time-frame due to the proliferation of a suppressor gene [30,31,32]. Assessing whether equilibrium has or has not been reached is sometimes further complicated by the many instances where there are co-infecting endosymbionts that may not have been acquired at the same time and that have different, potentially conflicting phenotypes where some increase host fitness and others appear detrimental [99]. Changes in host behaviour (and other host traits) that result directly from an infection or as a response to bacterial phenotypes such as sex ratio bias, are thus not only potentially transient but also rapidly changing. The magnitude of the effects that they have had at a population level may thus be difficult to establish.

Studies of individual species indicate that endosymbionts are a potential source of behavioural modification of their arthropod hosts either directly through their interaction with the host or indirectly as a consequence of perturbations in sex ratio. Endoymbionts may influence precopulatory reproductive isolation by altering preferences and mating rates of male and female hosts. Behavioural changes induced by endosymbionts can also potentially alter levels of genetic variation within species and the degree of population subdivision. They thereby influence the potential of a species to adapt in response to changes in the environment.