Community-wide analysis of horizontal symbiont transmission: oak gallwasps as a model system

The maternally inherited bacteria Wolbachia and Cardinium infect 20-70% and 6-7% respectively of insect species, have a wide range of phenotypic impacts on their hosts, and have been seen as possible agents of insect control (Hughes et al. 2004; Hurst & Jiggins 2000; Zchori-Fein & Perlman 2004; Kittayapong et al. 2003). Wolbachia is the best studied of these symbionts, and an important feature of its biology is the non-concordance observed between host and Wolbachia phylogenies (O’Neil et al. 1992; Rousset et al. 1992; Werren et al. 1995; Schilhuizen & Stouthamer 1997). This pattern is most easily explained by horizontal transmission between phylogenetically distant host species, a possibility that has received considerable attention in recent decades (Hughes et al. 2004; Hurst & Jiggins 2000; Zchori-Fein & Perlman 2004; Kittayapong et al. 2003). Several mechanisms for transmission among closely related species have been proposed, including contact during attempted predation or after injury (Rigaud & Juchault 1993). An important potential transmission route in insect communities is thought to be the intimate interactions between parasitoids and their hosts (West et al. 1998; Rokas et al. 2002), particularly during stinging or oviposition (Werren et al. 1995; Vavre et al. 1999). While horizontal transfer has been demonstrated in nature and in the laboratory (Heath et al. 1999), its frequency in natural communities and the role of transmission across trophic levels remain unclear. This frequency is central to understanding the evolution of Wolbachia and similar symbionts, and to predicting impacts of their potential use in insect control (e.g. Sinkins et al. 2005). In this project, we used natural insect communities consisting of gallwasp hosts and parasitoid natural enemies to assess the significance of three host attributes in predicting horizontal symbiont transmission: host relatedness within a trophic level, interaction frequency between trophic levels, and membership of a given geographical community.

Oak gallwasps provide a unique opportunity to test these hypotheses, as our laboratory has collected large samples of virtually the entire ecological community and we have detailed knowledge of the links between community members. Oak gallwasps (Hymenoptera, Cynipidae) exist as species-rich assemblages on oaks (Schönrogge & Crawley 2000; Stone et al. 2002). Their larvae develop within galls that are also home to another group of herbivorous wasps, termed inquilines. Both gallwasps and inquilines are attacked by a rich community of chalcid parasitoids (Stone et al. 2002). These three groups exist in close proximity within galls, and the same parasitoid may attack both gallwasp and inquiline in a single gall (Stone et al. 2002). Previous work has shown Wolbachia to be present in two of the trophic groups addressed in this study (gall inducers and inquilines) (Rokas et al. 2002). Although prior to this project we had not yet looked for Cardinium in oak gall communities, it is most frequent in Hymenoptera, and the Hymenoptera-dominated communities in oak galls were thus appropriate for examination of the frequency and form of symbiont transmission between species. We used PCR to screen for Wolbachia and Cardinium infection in the gall inducers, inquilines and parasitoids in two replicate oak gallwasp communities, one in Hungary and one in Iran. We extended our work to include three further symbiont taxa - Flavobacteria, Rickettsia and Spiroplasma, providing the first comprehensive survey of all 5 symbionts in any natural community.

In contrast to a previous U.K. community survey of symbiont infection (West et al. 1998), each of our replicates represents a long-isolated glacial refugium (Stone & Sunnucks 1993; Rokas et al. 2003; Stone et al. 2001, 2007; Challis et al. 2007), within which communities members have been interacting for at least tens of thousands of years. For each host shown to be infected in initial screens, we established strain identity by sequencing appropriate symbiont genes. Our intent was to represent similarity in strains present in each infected host in a matrix generated using genetic distance data. We aimed to test the following hypotheses using matrix correlation techniques, which assess the concordance between symbiont strain identity (as the dependant variable) and explanatory variable matrices summarising similarity in the host attributes under test.
We aimed to test the following hypotheses.

Hypothesis 1. That strain similarity is positively correlated with host relatedness. This analysis tests the prediction arising from bacterial inoculation studies that symbionts are more likely to become established in closely related hosts. This analysis will be carried out for gallwasps and inquilines, and explanatory variable matrices will contain genetic similarities derived from our own phylogenetic analyses of these groups (Cook et al. 2002; Rokas et al. 2003; Ács et al. 2006).

Hypothesis 2. That strain similarity is positively correlated with interaction frequency within (gallwasp-inquiline) and across (gallwasp-parasitoid) trophic levels. This analysis tests the prediction that higher interaction rates are associated with higher rates of symbiont inoculation. Our previous NERC-funded research has quantified trophic interactions between community members in Hungary (NERC GR/12847, 2001-2003), allowing estimation of the frequency of interactions between given community members. The explanatory variable matrix will be the relative frequency of association between gallwasp and inquiline, or gallwasp and parasitoid. Meaningful application of this approach requires many species in at least two trophic groups to be infected with a given symbiont at a single site.

Hypothesis 3. Strain similarity is higher among members of given glacial refugium than it is between populations of the same species in different refugia. This analysis tests the prediction that geographically defined sets of interacting hosts will accumulate shared symbiont strains. The explanatory variable matrix will code refugial membership for populations of all species common to the communities being compared.

We have extended our previous work on Wolbachia in oak gallwasps and their inquilines (Rokas et al. 2001, 2002) to show that Wolbachia is widespread across trophic levels in oak gallwasp communities. We also provide the first evidence from these communities for infection with Cardinium, Spiroplasma, Flavobacteria and Rickettsia.
For Wolbachia, we find that closely related hosts often do not have similar symbiont sequences, and symbiont lineages contain taxonomically diverse hosts, arguing against our Hypothesis 1. Support for parallel symbiont-host diversification is only provided by symbiont sequence similarity in two congeneric pairs of parasitoids. We find that patterns of symbiont incidence and prevalence differ dramatically between sites, suggesting (a) that dynamics of infection and prevalence occur on regional spatial scales, and (b) that infections persist for less than the timescales required for range expansion by hosts common to both sampled communities. This is an important result because it underlines the need to take geographic variation into analyses of symbiont infection in any given group of host taxa. However, there is also no evidence that infected hosts from the two sampled communities represent mutually monophyletic clades of Wolbachia sequences, arguing against our Hypothesis 3.
The incidence of symbionts within the two sampled communities was too low to allow a meaningful food web based analysis of potential horizontal transfer, and hence testing of Hypothesis 2. Our demonstrated similarity in Wolbachia wsp sequence between hosts in different trophic levels is potentially compatible with horizontal symbiont transmission, but support for this conclusion is strongly dependent on the placement in the phylogeny of sequences for species not associated with gall-inhabiting hosts. If sequences from gall inhabitants form a monophyletic group, within which sequences from different trophic groups do not form monophyletic lineages, we have strong support for horizontal symbiont transfer. We are in the process of assembling an appropriate sequence dataset to test this possibility formally.
One striking pattern for Wolbachia is that the greatest diversity of wsp lineages (Iran) is associated with areas of ancient intraspecific genetic diversity for geographically widespread gallwasps (Rokas et al. 2003; Challis et al. 2007) and associated parasitoids (Eurytoma brunniventris, Megastigmus dorsalis, M. stigmatizans, Ormyrus nitidulus: unpublished data). Iran and Turkey probably represent ancient origins for these widespread insect species. The geographic distribution of Wolbachia wsp sequence diversity parallels this pattern, suggesting that longitudinal westwards range expansion by the insect hosts may have been associated with loss of symbiont sequence diversity as well as host intraspecific genetic diversity. This pattern is potentially compatible with the maternal inheritance pattern of both the host genetic diversity marker (mitochondrial sequence data) and the symbionts studied here (e.g. Jiggins 2003; Zchori-Fein & Perlman 2004). If the symbionts do not bias their inheritance in these hosts, then we might expect loss of lineage diversity during the sequential population bottlenecks commonly associated with range expansion events (Stone & Sunnucks 1993; Stone et al. 2001; Hayward & Stone 2006). However, we do not know the phenotypic impact of infection for any of the sampled hosts or symbionts. Ongoing analyses for the other symbionts in this project will show the extent to which they show similar patterns.