Discussion

Amplicon sequencing as a powerful method of Wolbachia strain determination

Strain determination is a key step in studying Wolbachiadistribution and host-shifting among a given host group that needs to be performed using an efficient method. Given that infection with more than one Wolbachia strain is common in various arthropod groups (Werren et al. 1995; Perrot-Minnot et al. 1996; Hiroki et al. 2004; Narita et al. 2007; Hou et al. 2020), strain determination methods should be able to distinguish and identify strains in both singly and multiply-infected samples. The traditional method of using Sanger sequencing is not effective in dealing with co-infected arthropod samples, and improvements such as using different primers and cloning (Schuster 2008; Vo & Jedlicka 2014)) are costlier and more labour-intensive, and also have limitations (Van Borm et al. 2003; Schuler et al. 2011). High-throughput whole genome sequencing (WGS) would seem to be the most accurate available methodology for strain identification, but this approach has its own difficulties (Bleidorn & Gerth 2018). First, given that Wolbachia is not culturable, it is challenging to obtain genetic material enriched for Wolbachiarelative to host DNA, possibly resulting in low sequencing depth. Second, even with high sequencing depth, assembling Wolbachiagenomes can be difficult due to a high density of mobile elements (Wang et al. 2019) and thus only draft genomes can be recovered. Finally, the still relatively high costs of WGS make this approach less applicable in large Wolbachia surveys. Due to these limitations, only 33Wolbachia annotated whole genomes have been publicly available on GenBank so far (as of October 2021). As suggested by Bleidorn and Gerth (2018), instead of whole genomes, sequencing and assembling aWolbachia draft genome is sufficient for strain determination. However, many more draft genomes should be publicly available first to provide a reliable reference bank for strain determination. Although a draft genome can indeed be mapped to the selective marker amplicons (e.g. MLST), generating such data for large surveys is still time- and cost-intensive. To overcome these technical obstacles, we suggest Illumina multi-target amplicon sequencing as a middle-ground, efficient and affordable method that can be applied to large surveys and is also capable of dealing with multiple infections. In particular, the fiveWolbachia MLST genes along with wsp and 16S used in our study appear to be well suited to distinguish between strains, as has also been shown in a recent comparative study of available whole genomes of Wolbachia (Wang et al. 2020).

Wolbachia diversity in scale insects

This study revealed that a substantial portion of tested scale insects are infected with more than one strain of Wolbachia (27% double and 5% triple infected). We also found Wolbachia multiple infections in associate species (including wasps and ants), indicating co-infection might be a common phenomenon in most of these insect groups. However, it is important to caution that detecting a given Wolbachia strain in a given host is not conclusive evidence of a stable infection, and laboratory assays should be conducted to ascertain Wolbachia maternal transmission and establishment within the host population (Chrostek et al. 2017). Moreover, in the case of parasitoids and predators, a detected strain may derive from their undigested prey rather than the screened insect itself (Ross et al. 2020). Unfortunately, laboratory rearing of collected samples is not feasible for large Wolbachiasurveys such as the current study. Therefore, any interpretation from this type of data should be treated with caution.
Based on the MLST database (as of 31st August 2021), 24 strains with complete MLST gene sequences had previously been reported from the Australian fauna (https://pubmlst.org/organisms/wolbachia-spp). Here, we report 63 new strains (belonging to 31 strain groups) for Australia, including the first three Supergroup F strains in Australasia. Apart from two strains (w Sph4.1 = ST 289, andw Cal = ST 357), none of the strains in the current study were 100% identical to any registered in the MLST database. As our sequenced regions were slightly (~5%) smaller than the MLST amplicons available on MLST online database, there is a possibility that the two strains that were identical to the MLST profiles were different in the remaining part of the gene fragments. We found w Sph1 to be the most common and widely distributed strain group in Australia (detected in seven scale insects, four wasps and one ant species). Based on the phylogenetic tree of all reported strains in the MLST database and the current study strains, there are six registered strains (STs) within the w Sph1 strain group (Figure S2). These strains seem to be globally distributed across various insect orders. For example, one of the strains in this group, registered as ST=19, has been reported in 16 different host species belonging to four insect orders. This broad host range may be an indicator of an extraordinary host-shifting ability of w Sph1. Mostly based on the number of infected host species, several Wolbachia strains have been reported with a similar ability (e.g., HVR-2 in ants (Tolley et al. 2019), ST41 in Lepidoptera (Ilinsky & Kosterin 2017), and w Hypera in weevils (Sanaei et al. 2019)). Among all the superspreaders, w Ri is one of the best-studied Wolbachia strain groups that has rapidly (within 14,000 years) naturally infected five Drosophila species (Turelli et al. 2018). w Ri can also be introduced to mosquitoes by transinfection, corroborating this strain’s potential to infect new host species (Fraser et al. 2017). Compared to w Ri, it seems thatw Sph1 has been reported in a higher number of host species that are taxonomically more diversified (belonging to various insect orders). Although the w Ri group has an extensive genomic diversity (Ishmael et al. 2009; Turelli et al. 2018), low variation has been observed within its MLST profiles (https://pubmlst.org/organisms/wolbachia-spp). Four strains have been reported in the w Ri group and only one strain (ST=17) has been reported in more than one species of Drosophila (based on MLST website as of 31st August 2021). Therefore, w Sph1 might have a higher diversity than w Ri and may therefore have a potential to be artificially introduced to other insects for human applications (e.g., controlling vector born disease). However, transinfection studies are necessary to ascertain the host-shifting ability of w Sph1 in laboratory conditions.

Phylogenetic distance effect can explain host-shifting

As is typical of Wolbachia infection in an arthropod family, non-independence between the scale insect and Wolbachiaphylogenetic trees were not statistically supported, i.e. no signal of congruence was detected between them. Instead, the current distribution of Wolbachia in scale insects was most likely shaped by host-shifting. Among many potential factors determining host shifts, it seems that host phylogeny and geographic distributions are two major players (Sanaei et al. 2021a). Combining data from 25 transinfection studies, Russell et al. (2009) showed that there is a positive correlation between host phylogenetic relatedness and success of theWolbachia transinfection. In addition, by focusing only on a part of the host phylogenetic tree, several studies uncovered a pattern of host-shifting among closely related species (Haine et al. 2005; Guz et al. 2012; Turelli et al. 2018). On the other hand, the observation of identical Wolbachia strains in species that live in the same area points to a role of geography in host-shifting (Kittayapong et al. 2003; Stahlhut et al. 2010; Morrow et al. 2014; Gupta et al. 2021). The relative contributions of the host phylogenetic and geographic distance effect on Wolbachia host shifts are poorly understood. Here, we tried to evaluate these two factors in Wolbachia host shifting by using a powerful statistical method. The results of our GAMM indicate that host shifts in scale insects can be mainly explained by the phylogenetic distance effect (host shifting is more feasible between closely related species compared to distantly related) (Figure 3). This result is in line with numerous examples of finding the sameWolbachia strain group in congeneric species (e.g.,wHypera1 in the genus Hypera (Coleoptera) (Sanaei et al. 2019), w Lev in the genus Lutzomyia (Diptera) (Vivero et al. 2017), ST19 in the genus Bicyclus (Lepidoptera) (Duplouy & Brattström 2018)).
Horizontal transfer of parasites/symbionts among closely related species can generate a phylogenetic signal similar to host-parasite co-speciation (De Vienne et al. 2007). However, there is indirect evidence advocatingWolbachia sharing patterns in scale insects that can be explained best by recent host-shifting. In contrast to horizontal transmission which occurs rapidly, co-speciation happens in an evolutionary timeframe which allows Wolbachia genes to be mutated. By investigatingWolbachia infection in Nasonia species complex, it is estimated that Wolbachia MLST genes mutation rate is one third of their host nuclear genes (from nine single copy nuclear regions) (Raychoudhury et al. 2009). Although this ratio can be slightly different among various host species and Wolbachia strains (see also Conner et al. (2017)), it can be adopted as a tool to distinguish co-diversifications from recent host shifting. Given that the lowest pairwise distance between host species nuclear genes that we have in our dataset is 2%, in the case of Wolbachia co-speciation, at least 17 bp differences (out of 2608 bp) should be observed between two closely related strains. We infer host-shift events based on sharing either identical strains or identical strain groups (which includes strains with up to only 5bp differences across all Wolbachiaamplicons) (File S1). In addition, in 73% of determined host-shift events in scale insects, shared Wolbachia strains have identical wsp which is less conserved compared to the MLST genes.
Interestingly, the significant impact of host phylogenetic distance on Wolbachia strain group sharing was not mirrored by statistically significant tests for host andWolbachia phylogenetic tree non-independence (Parafit and Paco), as might have been expected under host shifting with a PDE. We speculate that these tests for tree independence are not sufficiently powerful to detect minor departures from tree independence as caused by host shifting under a PDE. It would be useful to verify this using computer simulations of host shifting with PDE (e.g., de Vienne et al. (2007); Engelstädter & Fortuna (2019)).
Hybridisation between closely related species may lead to introgression ofWolbachia into a new species by vertical transmission. This has been demonstrated in the Nasonia  species complex (Raychoudhury et al. 2009) and some species of Drosophila (Turelli et al. 2018; Cooper et al. 2019). Introgression can be considered a special case of host shifting (referred to as ‘hybridisation-mediated host shifts’ by Sanaei et al., (2021a), but also has similarities to cospeciation. Since this type of host shifting can only occur between closely related species, it is expected to produce a PDE signal. We believe, however, that hybridization is unlikely to explain the observed patterns of strain group sharing in our dataset. There are only two con-generic species in our dataset that share the same Wolbachia strain (Cystococcus pomiformis /C. echiniformis andEriococcus sp1 /E. sp2 ), and hybridisation between species from different genera seems unlikely. Given that Wolbachia and mitochondria are co-transmitted, introgression of Wolbachia would lead to mtDNA hitchhiking and hence be expected to leave a signature of similar mtDNA sequences (Jiggins 2003; Cooper et al. 2019; Miyata et al. 2020). By contrast, none of our congeneric species have very similar COI sequences. (The difference between C. pomiformis and C. echiniformis is 6.5%, and the difference between E. sp1 and E. sp2 is 3%).