Introduction
Recognising the conditions that favour speciation is critical if we are to understand the extent and structure of biodiversity. Moreover, species interactions, both between and within trophic levels, can be significant contributors to diversification processes and are sculpted by evolutionary forces, which in combination with abiotic drivers deliver an ecosystem or community’s (dynamic) state (Harmon et al. 2019). Thus, an in depth understanding of adaptive processes alongside their ecological contingencies (e.g., interaction strengths and polarities; Segaret al. 2020) is a fundamental component of the diagnostic tool kit essential for achieving standard objectives in ecology.
Apropos of this, increasing emphasis on the arthropod microbiome as a modifier of ecological interaction strength (e.g., Hansen & Moran 2014) underlines the need to consider endosymbionts as part of the extended phenotype. The rise and fall of microbial partners is an eco-evolutionary process, driving and being driven by ecological interactions between organisms and their environment. One such endosymbiotic bacterium, Wolbachia , infects up to 40% of arthropod species and often plays a key role in speciation (Werrenet al. 2008; Zug & Hammerstein 2015). Wolbachia commonly induces cytoplasmic incompatibility (CI) via sexual sterility between infected males and females that are either uninfected (unidirectional CI) or carry an alternative strain (bidirectional CI) (Beckmann et al. 2017, 2019; LePageet al. 2017). CI may therefore promote reproductive isolation (RI) between populations or incipient host species and increase the speed or likelihood of speciation by restricting geneflow (Bordensteinet al. 2001; Zimmer 2001; Telschow et al. 2007), a critical factor in sympatric and ecological speciation otherwise caused by correlations between divergent traits, mate choice and/or habitat choice (Gavrilets 2004).
In some non-arthropod taxa, Wolbachia is an essential mutualist and accordingly shows strict co-divergence with hosts (Casiraghiet al. 2001; Balvín et al. 2018). Among arthropods, Wolbachia lineages are mostly facultative and evolutionarily unstable symbionts generally exhibiting host co-phylogenetic incongruence (Shoemaker et al. 2002; Yang et al. 2012; Jäckel et al. 2013; Zug & Hammerstein 2015), although exceptions are known where essential mutualism appears likely (e.g., Dedeineet al. 2001; Raychoudhury et al. 2009; Hamm et al.2014). At broader taxonomic scales (e.g., families, orders), a non-random distribution of Wolbachia has been noted (Engelstädter & Hurst 2006; Weinert et al. 2015), viewed as the consequence of accelerated host switching among closely related species from highly speciose clades (Engelstädter & Hurst 2006). However, at reduced scales Wolbachia often appears idiosyncratically distributed (Shoemakeret al. 2002; Smith et al. 2012; Yang et al. 2012; Jäckel et al. 2013; Zug & Hammerstein 2015), as closely related hosts often harbour paraphyletic strains. These strains are paraphyletic within the context of the complete Wolbachia phylogeny, such that grouping strains from two closely related hosts renders them paraphyletic. Horizontal exchange also occurs between unrelated species (Shoemaker et al. 2002; Zug & Hammerstein 2015; Bailly-Bechetet al. 2017). Counterintuitively, this may not be readily predicted from close ecological contact (Haine & Cook 2005; Jäckelet al. 2013; Gerth et al. 2013) but incidences where it has been recorded (Sintupachee et al. 2006; McFrederick & Rehan 2016; Miraldo & Duplouy 2019) suggest that outcomes may be context dependant. Many studies conclude that infection status depends on the ability of Wolbachia to manipulate its arthropod hosts (Werrenet al. 2008; Zug & Hammerstein 2015), which may add to the sense that it is non-systemically distributed. Testable models linking eco-evolutionary processes to distribution patterns and ecological context remain critically absent.
As Wolbachia mediated CI results in post-zygotic mortality, initial fitness losses due to reduced fecundity are costly, meaning that selection may be expected to operate on hosts to purge Wolbachia . However, Wolbachia is posited to facilitate reproductive isolation between incipient species in combination with reduced hybrid fitness, even when only unidirectional pre-zygotic isolation operates (Shoemaker et al. 1999). The maladaptation of intermediate forms is central to models of sympatric/ecological speciation which may be likely under bi-directional CI as documented in closely-related, co-occurring Nasonia wasps (Bordenstein & Werren 2007). Thus, it is possible that Wolbachia represents a tolerable cost (contingent on ecological circumstances), rendering host fitness advantage (i.e., via hybrid avoidance) the prime determinant of infection status rather than the bacterium’s manipulative capability.
Predictive phylogenetic models of Wolbachia distribution have not previously incorporated the intensity of ecological contact between insect lineages that (a), provides a direct opportunity for horizontal exchange of microbes or genetic material, and (b), provides a contingency axis of whether RI is required. When speciation occurs in allopatry, specific mechanisms of RI may not necessarily evolve as the nascent species are not in contact (Coyne & Orr 2004). This may also be true if newly formed species specialise on different resources in sympatry (Nosil 2012). However, a mechanism of pre- or post-mating RI is required if ecological contact occurs, when the species use the same resources and overlap in space and time (e.g., Via & Hawthorne 2002).
Wolbachia typically drops out of host lineages after approximately 7 million years (±5.2-9.6) (Bailly-Bechet et al.2017), contributing to the lack of correlated host-symbiont divergence and adding weight to the idea that purging may occur. Compared with Wolbachia , alternative mechanisms of RI that require cytogenetic or morphological modification may take longer periods of time to evolve (Bordenstein et al. 2001), and thus may not be responsive enough to changing ecological circumstances that favour diversification, particularly in a sympatric setting. These lines of evidence suggest that observed lineage dropout (Bailly-Bechet et al. 2017) may result from temporal changes in the relative adaptive benefits of Wolbachia (as alternative mechanisms of RI evolve), that may subsequently become redundant and eradicated if hosts can mediate their own infection statuses (e.g., via physiological immune responses) – hereafter termed the adaptive decay hypothesis.
Fig wasps (Chalcidoidea), where Wolbachia prevalence is ca. 60%, appear to be a prime candidate for CI manipulation because many closely related and often cryptic species (both pollinating and non-pollinating) share an enclosed reproductive space (i.e., fig inflorescences), where they regularly come into contact giving potential for hybridisation (Molbo et al. 2003; Darwell et al. 2014; Yu et al.2019). Moreover, inbreeding is also common favouring female biased sex-ratios, strain fidelity through vertical transmission, and reduced allospecific (i.e., ex community) encounter rates – all increasing barriers to gene flow (Branca et al. 2009). Due to the confined nature of fig syconia (i.e., fig inflorescences), co-occurring incipient species must rapidly employ RI barriers (Nosil 2012) to avoid any hybridisation costs. Fig wasp studies often show paraphyletic Wolbachia infections across sister-species (Shoemaker et al. 2002; Haine & Cook 2005; Yang et al. 2012), while species occupying fig communities that do not contain congeners invariably display negative Wolbachia statuses (Haine & Cook 2005). Importantly, while these factors may obviously and measurably dominate fig wasp community structure, their influence may be apparent in all ecological systems to differing degrees.
Hybridisation between highly adapted lineages of wasps, with narrow abiotic niches and extreme matching for host fig interacting traits, presents a rather extreme cost. We develop a model which selects for ecologically contingent host tolerance of otherwise costly Wolbachia in this system, thus imposing pre-zygotic selection and reduced gene flow between lineages. We propose that sister populations/incipient species of wasps, associated with diverging fig hosts, should be infected with paraphyletic Wolbachia strains when in close ecological contact. Thus, Wolbachia should facilitate adaptive divergence. Subsequently, we model purging ofWolbachia after alternative mechanisms of RI are established across evolutionary time (see Fig. 1). This contact contingency hypothesis leads to a predictive system that would elicit an apparently stochastic distribution, with respect to the host phylogeny, similar to those frequently observed.
While the unusual ecological conditions of fig wasps, including the potential for complementary pre-zygotic (e.g., behavioural) barriers, may be sufficient to permit tolerance of post-zygotic fecundity reduction, we also develop a second model to singularly account for post-zygotic dynamics. We consider the heightened value of oviposition sites which are finite for pollinating fig wasps as they are unable to leave fig syconia after entry (Cook & Segar 2010). In monoecious fig species syconium oviposition sites are more valuable towards the centre where parasitoid wasp ovipositors typically do not penetrate (e.g., Dunnet al. 2008). As intermediate hybrid forms exhibit marked fitness reductions within co-evolved systems, the costs of reduced fecundity may prove tolerable if hybrid eggs are not wasted on premium oviposition sites. This could feasibly occur in two ways among fig wasps: (i) via preferential oviposition of favoured non-hybrid embryos (Hymenoptera are at least known to manipulate the oviposition order of haploid versus diploid eggs as well as adjust sex ratios; Raja et al. 2008); or (ii) via differential mortality affecting unviable hybrids before oviposition (an undocumented but plausible phenomenon). This is contingent on multiple mating events occurring within syconia (e.g., Murray 1990; Greeff et al. 2003), so that fig wasp foundresses carry egg loads of high versus low fitness embryos.
We model this oviposition trade-off hypothesis by simulating pre-oviposition egg mortality causing reduced egg load, meaning zero fitness is attributed to lost hybrid embryos. However, as fig wasps are known to prioritise oviposition into favourable sites, remaining non-hybrid eggs receive greater average fitness as they are probabilistically oviposited towards the syconium centre as opposed to their average position when mixed together with viable hybrids of reduced fitness (Fig. 2). We examine these trade-offs under different frequencies of conspecific-heterospecific mating opportunity scenarios.
As part of an ongoing study investigating patterns of co-speciation between several monophyletic fig (Ficus , Moraceae) species complexes and their pollinating wasps, we tested our primary ‘contact contingency’ hypothesis, which explains how Wolbachia infection may be an adaptive responses to diversification pressures in host wasps. This is accomplished by presentation of empirical data of pollinating wasps screened for Wolbachia , and then using Python programming to simulate our proposed mechanism incorporating ecological contact and phylogenetic relationships. We then test our secondary ‘oviposition trade-off’ hypotheses which accounts for Wolbachiapost-zygotic fecundity costs by modelling disparities in oviposition site quality again using Python programming. Our field site, located along a steep elevational gradient in Papua New Guinea, features a steep clinal turnover of Ficus species (Segar et al.2017) with species complexes comprising lowland/highland sister species or morphologically homogenous species with wide elevational ranges (Souto-Vilarós et al. 2018, 2019).