Discussion
Symbiotic interactions between hosts and microbes are important drivers of host phenotype, with symbionts both contributing to, and degrading, host performance. Heritable microbes are particularly important contributors to arthropod biology, with marked attention focused onWolbachia , the most common associate (Hilgenboecker, Hammerstein, Schlattmann, Telschow, & Werren, 2008; Zug & Hammerstein, 2012). Members of the Rickettsiales, like Wolbachia , share an evolutionary history with mitochondria (Ferla, Thrash, Giovannoni, & Patrick, 2013; Wang & Wu, 2015), such that a previous screen of BOLD submissions of mtDNA submissions observed Wolbachia as the main bacterial contaminant associated with DNA barcoding (Smith et al., 2012). However, our BOLD screen found that Rickettsia were more likely to be amplified than Wolbachia (0.41% vs 0.17% of deposits). Furthermore, Torix group Rickettsia were overrepresented in barcode misamplifications (95%) when compared to other groups within the genus. A comparison of the most commonly used barcoding primers to Wolbachia and Rickettsia genomes suggest homology of the forward primer 3’ end was likely responsible for this bias towards Torix Rickettsia amplification. To gain a clearer understanding of the relative balance of Torix group to otherRickettsia within symbioses and habitats, a targeted screen and bioinformatic approach was also undertaken. Through these three screens, a broad range of host diversity associated with Torix Rickettsiawas uncovered.
As the in silico and empirical evidence suggests Rickettsia COI amplification is not uncommon (Řezáč et al., 2014; Ceccarelli et al., 2016; Park & Poulin, 2020), why has this phenomenon not been described more widely before? The conduction of a previous large-scale non-target COI study using BOLD submissions (Smith et al., 2012), revealed only Wolbachia hits. This screen involved comparison to a Wolbachia -specific reference library and was thus likely to miss Rickettsia . Additionally, there has been a lack of Torix Rickettsia COI homologues to compare barcodes to until recently, where a multilocus identification system, including COIwas devised (Pilgrim et al., 2017). Indeed, out of the contaminant dataset received in this study, some of the Rickettsia contaminants were tentatively described by BOLD as Wolbachia due to the previous absence of publicly available Rickettsia COI to compare.
Although Rickettsia will only interfere with barcoding in a minority of cases (~0.4%), it is likely that alternate screening primers for some studies will need to be considered. In a demonstration of how unintended Rickettsia amplifications can affect phylogeographic studies relying on DNA barcoding, aRickettsia COI was conflated with the mtDNA COI of a species of freshwater amphipod, Paracalliope fluvitalis(Lagrue et al., 2016). Subsequently, supposed unique mtDNA haplotypes were allocated to a particular collection site, whereas this merely demonstrated the presence of Torix Rickettsia in host individuals in this lake. Contrastingly, non-target Rickettsia amplification can also allow for the elucidation of a novel host range of the symbiont (Ceccarelli et al., 2016; Park & Poulin, 2020; Řezáč et al., 2014) and this has been exemplified with our probing of BOLD.
Previously, several host orders have been associated with TorixRickettsia , including Araneae, Coleoptera, Diptera, Hemiptera and Odonata (Goodacre et al., 2006; Küchler, Kehl and Dettner, 2009; Machtelinckx et al., 2012; Martin et al., 2013; Thongprem et al. 2020). However, newly uncovered host orders from this study include Gastropoda (snails), Trichoptera (caddis flies) and Trombidiformes (mites) (Table 2). Caution needs to be taken when interpreting what these newly found associations mean, as mere presence of Rickettsia DNA does not definitively indicate an endosymbiotic association. Indeed, parasitism or ingestion of symbiont-infected biota can also result in PCR detection (Le Clec’h et al., 2013; Plantard et al., 2012; Ramage et al., 2017). Additionally, by calculating barcode success rate at an order level, Hemiptera were deemed to fail barcoding (either lack of amplification and/or quality sequence) more commonly than Diptera despite having a similar estimated Rickettsia prevalence (Table S8). As an increased barcoding failure rate is correlated with non-targetCOI amplification, it is probable there is a higher overall proportion of Torix Rickettsia -associated Diptera than Hemiptera in BOLD and likely in nature.
Model-based estimation techniques suggest Rickettsia are present in between 20-42% of arthropod species (Weinert et al., 2015). However, targeted screens often underestimate the incidence ofRickettsia hosts due to various methodological biases including small within-species sample sizes (missing low-prevalence infections) and the use of non-conserved primers (Weinert, 2015). Importantly, the inclusion and exclusion of specific ecological niches can also lead to a skewed view of Rickettsia symbioses. A previous review of Rickettsia bacterial and host diversity by Weinert et al. (2009) suggested a possible (true) bias towards aquatic taxa in the Torix group. In accordance with this, our targeted screen demonstrated TorixRickettsia infections were more prevalent in aquatic arthropod species compared to terrestrial. However, our observed over-representation of Torix group Rickettsia (14/16 strains) contrasts with Weinert’s findings which show a predominance of Belli infections and is likely due to the latter study’s absence of screened aquatic taxa. Furthermore, through the additional use of a bioinformatics approach, our SRA search appears to confirm that Belli and Torix are two of the most common Rickettsia groups among arthropods. Overall, these multiple screening methods suggest TorixRickettsia are more widespread than previously thought and their biological significance underestimated.
Previous studies have used either one or two markers to identify the relatedness of strains found in distinct hosts. In this study, we use the multilocus approach developed in Pilgrim et al. (2017) to understand the affiliation of Torix Rickettsia from diverse invertebrate hosts. Our analysis of Torix strains indicates that closely related strains are found in distantly related taxa. Closely relatedRickettsia are also found in hosts from different niches and habitats – for instance, the Rickettsia strains found in terrestrial blood feeders do not lie in a single clade, but rather are allied to strains found in non-blood feeding host species. Likewise, strains in phloem feeding insects are diverse rather than commonly shared.
The distribution of Torix Rickettsia across a broad host range suggests host shifts are occurring between distantly related taxa. It is notable that parasitoid wasps are commonly infected withRickettsia and have been associated with enabling symbiont host shifts (Ahmed et al., 2015; Vavre, Fleury, Lepetit, Fouillet, & Bouletreau, 1999). Aside from endoparasitoids, it is also possible that plant-feeding can allow for endosymbiont horizontal transmission (Caspi-Fluger et al., 2012; Gonella et al., 2015; S.-J. Li et al., 2017). For example, Rickettsia horizontal transmission has been demonstrated in Bemisia whiteflies infected by phloem-feeding (Caspi-Fluger et al., 2012; Y.-H. Li et al., 2017). Finally, ectoparasites like the Torix-infected water mites of the Calyptostomatidae family, could also play a role in establishing novelRickettsia -host associations, as feeding by mites has been observed to lead to host shifts for other endosymbiont taxa (Jaenike, Polak, Fiskin, Helou, & Minhas, 2007). Indeed, if multiple horizontal transmission paths do exist, this could account for the diverse plethora of infected taxa, as well as arthropods identified in this study which harbour more than one strain of symbiont (Morrow, Frommer, Shearman, & Riegler, 2014; Vavre et al., 1999).
The finding that Torix Rickettsia are associated with a broad range of invertebrates leads to an obvious question: what is the impact and importance of these symbiotic associations? Previous work has established Torix Rickettsia represent heritable symbionts and it is likely that this is true generally. There have, however, been few studies on their impact on the host. In the earliest studies (Kikuchi & Fukatsu, 2005; Kikuchi et al., 2002), Torix spp. leeches infected with Rickettsia were observed to be substantially larger than their uninfected counterparts. Since then, the only observation of note, pertaining to the Torix group, is the reduced ballooning (dispersal) behaviour observed in infected Erigone atra money spiders (Goodacre et al., 2009). Overall, the incongruencies in host and Torix Rickettsiaphylogenies (suggesting a lack of co-speciation and obligate mutualism), along with the lack of observed sex bias in carrying the symbiont, indicate facultative benefits are the most likely symbiotic relationship (Jaenike, 2012; Weinert, 2015). However, Rickettsia induction of thelytokous parthenogenesis should not be discounted in Torix infected parasitoid wasps identified in this study (Giorgini, Bernardo, Monti, Nappo, & Gebiola, 2010; Hagimori, Abe, Date, & Miura, 2006). To add to the challenge of understanding Torix Rickettsiasymbioses, the challenges of laboratory rearing of many TorixRickettsia hosts has led to difficulties in identifying model systems to work with. However, the large expansion of our Torix group host knowledge can now allow for a focus on cultivatable hosts (e.g phloem-feeding bugs).
A particularly important group for study are haematophagous host species. Our discovery of Rickettsia -associated tabanid and simulid flies, alongside Anopheles plumbeus mosquitoes, add to existing blood-feeders previously identified as Torix group hosts which include sand flies (Kaili Li et al., 2016; Reeves, Kato, & Gilchriest, 2008), fleas (Song et al., 2018), ticks (Floris et al., 2008) bed bugs (Potts, Molina, Sheele, & Pietri, 2020) and biting midges (Pilgrim et al., 2017). Some Rickettsia strains are known to be transmitted to vertebrates via haematophagy (Parola, Paddock, & Raoult, 2005). However, there is no evidence to date for vertebrate pathogenic potential for the Torix group. Despite this, TorixRickettsia could still play a significant role in the ecology of vectors of disease. A key avenue of research is whether these endosymbionts alter vectorial capacity, as found for other associations (Bourtzis et al., 2014; Hoffmann, Ross, & Rašić, 2015). In contrast to the widely reported virus blocking phenotype observed in Wolbachia -infected vectors (Moreira et al., 2009; van den Hurk et al., 2012; Walker et al., 2011),Rickettsia have been associated with a virus potentiating effect in Bemisia white flies vectoring Tomato yellow leaf curl virus (Kliot, Cilia, Czosnek, & Ghanim, 2014). Additionally, we uncovered aRickettsia -infected psyllid (Cacopsylla melanoneura ) which is a vector of Phytoplasma mali (apple proliferation) (Tedeschi, Visentin, Alam, & Bosco, 2003). Thus, the question of TorixRickettsia vector-competence effects is clearly of widespread relevance and deserves further attention.
To conclude, we have shown that large-scale DNA barcoding initiatives of arthropods can include non-target amplification of TorixRickettsia . By examining these non-target sequences, alongside a targeted screen and SRA search, we have uncovered numerous previously undetected host associations. Our findings lay bare multiple new avenues of inquiry for Torix Rickettsia symbioses.