Horizontal transfers of polydnavirus segments extend the known range of parasitoid attacks to stick insects and orthopterans

Abstract

Horizontal gene transfer occurs beyond anecdotal frequencies in metazoans. Among insects, some parasitoid wasps even carry gene delivery agents called polydnaviruses (PDVs). These domesticated viral elements mediate the integration of wasp genes into the genome of parasitized hosts, thereby protecting developing larvae from immune defenses. The frequency of PDV-mediated transfers is sufficiently high that it could be exploited to better characterize the range of organisms attacked by parasitoid wasps. Here, we apply this rationale by screening for the specific molecular footprints of these transfers in 6,814 protostome genomes. We found a total of 6,556 PDV-mediated integrations, all of which were in insects. The distribution of these integrations is highly consistent with the known host range of PDV-encoding parasitoid wasps. Most were found in lepidopterans (6,260 integrations in 303 species)—the main hosts of PDV-encoding wasps—and a few were retrieved in sawflies (139 integrations in 14 species) and leaf beetles (four integrations in two species), also known to be parasitized by some of these wasps. Remarkably, we found a total of 232 integrations in three species of stick insects and one integration in an orthopteran, two insect lineages that have never been reported to be attacked by PDV-encoding wasps. We show that these integrations are mostly recent and that stick insects and sawflies were attacked recurrently by multiple wasp lineages. Overall, our study warrants accounting for stick insects and orthopterans as possible new targets of parasitoid attacks, both in community ecology and in assessments of biological control strategies.

Introduction

An in-depth knowledge of parasite host range is essential for predicting community structure and ecosystem stability (Hudson et al. 2006; Lafferty et al. 2008; Pilosof et al. 2014; Hasik et al. 2023; de Angeli Dutra and Poulin 2024). Characterizing host range is also crucial for evaluating the extent to which parasite-based biocontrol may affect nontarget species (Heimpel et al. 2021). This is most critical for parasitoid wasps, which exert strong top-down control in many ecosystems (Lafferty et al. 2008; Burke and Sharanowski 2024) and are used worldwide to control agricultural pests (van Lenteren 2012). The life cycle of parasitoid wasps begins when a female lays eggs on (or in) a host larva (Godfray 1994; Pennacchio and Strand 2006). Wasp embryos develop by feeding on host tissues and eventually pupate within the host, or at the surface of its body, usually killing it in the process. A common strategy to characterize the host range of parasitoid wasps relies on collecting potentially parasitized insects in the wild and recording which wasp species emerge (Sertkaya and Bayram 2005; Smith et al. 2008). Such field surveys can be complemented by experimental parasitism assays in the laboratory (Kaiser et al. 2017). To some extent, host range can also be predicted from host phylogeny, which is known to strongly structure current associations (Hambäck et al. 2013; Heimpel et al. 2021). Although effective, these methods may overlook certain elusive host–wasp associations, due to incomplete sampling of potential host taxa. In this study, we leverage comparative genomics to further characterize the range of hosts attacked by two wasp lineages from the Ichneumonoidea superfamily. The approach relies on identifying molecular footprints typically left when horizontal gene transfer (HGT) occurs from wasps to their hosts via domesticated viruses, which can be used to uncover previously unrecognized host lineages.

Within Ichneumonoidea, several wasp lineages produce viral particles that females inject into host larvae together with their eggs, leading to suppression of host immunity and facilitating the development of wasp embryos (Vinson and Scott 1975; Webb 1998; Bézier et al. 2009; Bitra et al. 2011; Tan et al. 2018; Burke and Sharanowski 2024). The structural components of these particles, including capsid and envelope, are encoded by viral genes that underwent chromosomal integration in ancestors of today's parasitoid wasps, followed by molecular domestication. The distribution and diversity of these elements suggest that domestication events occurred independently in several wasp lineages, yielding viral particles that differ in terms of structure and gene content. Because of the viral origin of their structural proteins, the viral particles encoded by parasitoid wasp genomes were originally referred to as viruses and classified into their own viral family, the Polydnaviridae (PDVs). Two genera of PDVs have thus been defined: Bracovirus and Ichnovirus (also named Bracoviriform and Ichnoviriform [Kuhn and Koonin 2023]), respectively, produced by wasps from the Braconidae and Ichneumonidae families. Bracoviruses result from the domestication of an ancestral nudivirus that took place ∼100 million years ago in an ancestor of the microgastroid wasp complex (Murphy et al. 2008; Bézier et al. 2009; Herniou et al. 2013). Ichnoviruses were likely acquired independently in at least two ichneumonid lineages (Béliveau et al. 2015; Legeai et al. 2020; Santos et al. 2022). Their structural organization strongly suggests a viral origin, although no closely related free-circulating viruses have yet been observed.

Although originating from different viruses, PDVs from these two groups share many features. In particular, they all package multiple circular DNA molecules that are produced by distinct loci scattered throughout the wasp genome—the so-called proviral segments—whose genetic content is likely not of viral origin (Bézier et al. 2009; Strand and Burke 2014). These segments are amplified in specialized ovarian cells and circularized through homologous recombination involving DNA motifs known as direct repeat junctions (DRJs) (Gruber et al. 1996; Savary et al. 1997; Beck et al. 2011). Once in host cells, wasp DNA circles are expressed, producing virulence factors that are crucial to parasitism success. Remarkably, some circles include a specific ≈150-bp DNA sequence called the host integration motif (HIM), which drives their systematic chromosomal integration in host somatic cells infected by PDVs (McKelvey et al. 1996; Gundersen-Rindal and Dougherty 2000; Beck et al. 2011; Chevignon et al. 2018). The detailed mechanism underlying integration is not known, but molecular signatures observed at wasp/host breakpoints reveal that it always involves two double-strand breaks occurring at conserved motifs within the HIM, called junction 1 and 2 (J1 and J2) (Beck et al. 2011; Chevignon et al. 2018; Muller et al. 2021; Wang et al. 2021). The small DNA fragment separating J1 and J2 is lost upon integration, so that the integrated linearized circles are delimited by J1 and J2 in the host somatic genome. An earlier study quantified these integration events by bulk short-read sequencing of DNA extracted from parasitized lepidopteran hosts, revealing between 12 and 85 bracovirus integrations and between 23 and 40 ichnovirus integrations per haploid host cell depending on the tissue (Muller et al. 2021; Heisserer et al. 2023).

In addition to the systematic and massive integration of PDV observed in somatic cells during parasitism, PDV sequences resulting from HIM-mediated integration have also been identified in assembled genomes of lepidopterans, the main hosts of parasitoid wasps (Gasmi et al. 2015; Di Lelio et al. 2019; Heisserer et al. 2023). No less than 2,387 PDV sequences bordered by either the J1 or J2 motif were found by similarity search in a total of 124 moth and butterfly species (15% of all species surveyed) belonging to 15 lepidopteran families. The number of integrations was low in most species (less than 10 integrations, median = 2), but several species underwent many integrations, including one, the wainscot hooktip (Ypsolopha scabrella, Ypsolophidae) in which no less than 1,756 integrations were identified. These sequences clearly result from ancient, HIM-mediated PDV integrations, having occurred in the germline genome of hosts that survived the parasitism and were able to transmit them vertically to their offspring (Heisserer et al. 2023). These findings not only establish HIM-mediated PDV integration as a major route of HGT from parasitoid wasps to lepidopterans, but also indicate that past events of parasitism between wasps and their hosts can leave specific signatures in host genomes.

Here, we explore such diagnostic footprints in the genomes of other insects to further characterize the potential host range of these parasitoid wasps. Our results not only confirm the presence of HIM-mediated horizontal transfers across all known host clades, but also reveal an extension of the range of insects attacked. As expected, many integration events were detected in the genomes of coleopterans and sawflies (Symphyta, Hymenoptera) (Hinz 1969; Sharanowski and Zhang 2014; Shaw et al. 2016; Robin et al. 2019). More surprisingly, we report dozens of HIM-mediated integrations of PDV in multiple species of stick insects, as well as one in the oak bush cricket (Meconema thalassinum, Orthoptera), two insect lineages not previously recognized as targets of PDV-encoding parasitoid wasps. Overall, our study reveals that PDV-mediated HGT affects a wider range of insects than previously thought, suggesting that stick insects and orthopterans could potentially serve as hosts for parasitoid wasps. These putative ecological interactions should be further evaluated in future studies of insect communities, and considered when assessing the possible impact of wasp-based biocontrol strategies.

Results

PDV integrations in lepidopterans

In order to assess the extent of PDV-mediated HGT from parasitoid wasps to other species, we performed a large-scale systematic search for PDV integrations in all protostome genomes available as of May 2025 (n = 6,814). Specifically, we used a total of 87 PDV segments containing a HIM from 10 wasps as queries to perform blastn searches on protostome genomes. PDV segments contain both genes of wasp origin and transposable elements (TEs), which can also be found elsewhere in the wasp genome. While some hits involving such genes and TEs may result from PDV-mediated HGT, many others may be due to vertical transmission from a common ancestor, or to HGT not mediated by PDVs. Thus, in a conservative approach, we only retained hits containing at least a fraction of the HIM sequence, which is specific to PDV segments, and absent from the rest of the wasp genomes. Our search revealed a total of 6,636 blastn hits corresponding to HIM-containing sequences. Their lengths ranged from 165 bp to 4,345 bp (mean = 620 bp), with similarity to a PDV segment varying from 63.1% to 96.9% at the nucleotide level (mean = 77.9%). The vast majority of hits (n = 6,556) were bordered on one end by the J1 or J2 motif and thus most likely resulted from an HIM-mediated integration event, akin to those described in parasitized hosts (Table S1) (Beck et al. 2011; Chevignon et al. 2018; Muller et al. 2021; Wang et al. 2021). The remaining hits (n = 80) contained an intact, full-length HIM, suggesting that they resulted from rare, noncanonical integrations, involving DNA breaks outside the HIM. Our blastn search did not directly retrieve full-length integrated circles, but manual assembly of blastn hits made it possible to identify several such sequences, bordered by the J1 motif on one side and by the J2 motif on the other (see below).

HIM-containing PDV integrations were distributed among 323 insect species, with no integration detected in protostomes other than insects (Fig. 1a). This highly structured distribution confirms the specificity of our approach, yielding no false positives. Among insects, the majority of integrations (94%) were found in lepidopterans, which are the primary hosts of PDV-encoding parasitoid wasps (Shaw and Huddleston 1991; Sharanowski and Zhang 2014; Fernandez-Triana et al. 2020). In total, PDV integrations were found in 303 of the 1,398 lepidopteran species included in the search (ie 22%, compared to 15% in a previous study [Heisserer et al. 2023]), from 32 families (out of 52 surveyed) (Fig. 1b). The median number of integrations per lepidopteran species is 2 (as in a previous study [Heisserer et al. 2023]), with a range from 1 (in 110 species) to 2,129 (in Y. scabrella), surpassing the previously observed maximum of 1,756 integrations (Heisserer et al. 2023). Consistent with Heisserer et al. (2023), most PDV integrations in lepidopterans derive from Ichnoviruses (4,011 out of 5,074). Overall, the present results reinforce earlier results showing the widespread nature of HIM-mediated PDV integration in lepidopterans (Fig. 1c) (Heisserer et al. 2023). These analyses further indicate that increasing the number and diversity of PDV circles (used as queries in similarity searches) leads to higher integration counts across a greater proportion of lepidopteran species. This trend suggests that the true fraction of species involved will continue to expand as more PDV-encoding parasitoid wasp genomes will be sequenced.

Figure 1.

Distribution of HIM-mediated polydnavirus integrations among protostomes. In blue, the number of species investigated, in red, the number of species with integrations. Trees and divergence times were recovered from timetree.org (Kumar et al. 2017). a) Number of species per phylum among protostome species for which whole genome sequence was available as of May 2025. b) Number of species per order among insect species for which whole genome sequence was available as of May 2025. c) Number of species per family of insect where integrations were found.

Discovery of PDV integrations in coleopterans, sawflies, stick insects, and a grasshopper

Beyond lepidopterans, we detected PDV integrations in two other insect lineages also known to be parasitized by a few PDV-carrying wasps: leaf beetles (Chrysomelidae, Coleoptera) and sawflies (Symphyta, Hymenoptera). In leaf beetles, one integration was found in the common asparagus beetle (Crioceris asparagi) and three in Lochmaea crataegi. In sawflies, integrations were observed in 14 species from eight genera belonging to four families (Tenthredinidae, Diprionidae, Athaliidae, and Cimbicidae) (Fig. 1c). The number of integrations per sawfly species ranges from 1 to 20 (median = 3.5) (Fig. 2). All PDV fragments found in sawflies and leaf beetles derive from ichnoviruses, consistent with these insects being parasitized by Campopleginae wasps, which carry ichnoviruses (Hinz 1969; Quednau and Lim 1983; Sharanowski and Zhang 2014; Shaw et al. 2016; Robin et al. 2019), and not by bracovirus-carrying wasps.

Figure 2.

Number of HIM-mediated polydnavirus integrations across species. The number of integrations is shown for each species. When more than one genome is available for a species, we provide the mean number of integrations. The name of the species having integrations are in red, while species devoid of integrations are in black. GenBank accession number of the genome of these species as well as their taxonomy are provided in Table S6. The red triangle indicates the integration timing of the three orthologous integrations shared by the Neodiprion species. The diversity of PDV circles is color-coded with a blue palette for circles homologous to bracoviruses (BV) and a pink palette for circles homologous to ichnoviruses (IV).

Our similarity search also retrieved PDV integrations in two insect orders not known to be parasitized by PDV-carrying wasps: orthopterans and stick insects (Phasmatodea). Specifically, we found one integration in the oak bush cricket (M. thalassinum, Tettigoniidae), 10 in the Lord Howe Island stick insect (Dryococelus australis, Phasmatidae), 38 in the smooth stick insect (Clitarchus hookeri, Phasmatidae), and 148 in the European stick insect (Bacillus rosius, Bacillidae) (Figs. 1c and 2b). As observed in lepidopterans, most integrations found in the oak bush cricket and stick insects derive from ichnoviruses (164 out of 196). In fact, all integrations found in the Lord Howe Island stick insect and in the oak bush cricket derive from ichnoviruses. Only the smooth stick insect and the European stick insect harbor integrations from both ichnoviruses and bracoviruses (Fig. 2).

To verify that the similarity between non-lepidopteran PDV-like sequences and segments from wasps is not restricted to the HIM region, we manually assembled full-length or nearly full-length PDV integrations in two species of stick insects, one leaf beetle and the oak bush cricket (Fig. 3). As observed earlier in lepidopterans (Heisserer et al. 2023), this shows that integrations are not restricted to a small region surrounding the J1 or J2 motifs. Instead, similarity to PDV segments can extend over their complete sequence (Fig. 3).

Figure 3.

Schematic representation of four complete or nearly complete HIM-mediated integrations. a) Integration homologous to segment F3 of Diadegma fenestrale ichnovirus (IV12-like) in Bacillus rossius (Phasmatodea) genome. b) Integration homologous to segment 30 of Glyptapanteles indiensis bracovirus (BV26-like) in Clitarchus hookeri (Phasmatodea) genome. c) Integration homologous to segment 25 of Glyptapanteles indiensis bracovirus (BV1-like) in Meconema thalassinum (Orthoptera) genome. d) Integration homologous to segment Hd27 of Hyposoter didymator ichnovirus (IV27-like) in Lochmaea crataegi (Coleoptera) genome. The host genome is represented by a black line, the integration is in yellow, best blastn hit polydnavirus genes are in orange, the junctions J1 and J2 are in red, and the host genes are in green. The position in the host genome, the GenBank accession number of the host contig, and the length of the integration are indicated below the integration. The coordinates of the J1 and J2 junctions as well as the taxonomic assignment of the host genes occurring in the flanking regions are provided in Table S4.

In order to estimate the diversity of wasp circles integrated in non-lepidopteran genomes, we clustered all proviral segments that we used as queries in our similarity searches into groups of homologous sequences, yielding 25 and 44 groups for bracoviruses and ichnoviruses, respectively. PDV sequences found in non-lepidopteran species originate from 16 of these groups (9 from bracoviruses and 7 from ichnoviruses) (Fig. 2). Most of this diversity is concentrated in the three stick insect species, where PDV sequences are derived from 5 segments in D. australis and up to 13 in B. rossius (Fig. 2). By contrast, PDV sequences from sawflies, leaf beetles, and the oak bush cricket all derive from only one to two segments depending on the species (Fig. 2). In terms of quantity, one ichnovirus segment (IV27, homologous to segment 27 of Hyposoter dydimator) appears to be the source of the vast majority of PDV insertions in non-lepidopteran genomes (217 out of 303). This pattern mirrors what is observed for somatic integrations during parasitism by ichnovirus-encoding wasps, with IV27 being by far the most frequently integrated circle (Wang et al. 2021; Heisserer et al. 2023).

PDV integrations are vertically inherited in non-lepidopteran insects

When finding PDV sequences in the genome of insects other than parasitoid wasps, two scenarios must be excluded before concluding that these sequences indeed stem from ancient integrations in the germline of a surviving host, followed by vertical transmission to its offspring. First, one must rule out the possibility that these sequences result from contamination, which could arise, for instance, if the DNA of parasitoid wasps was co-sequenced with that of their hosts. Second, one must assess whether PDV integrations are present in all cells of the sequenced individual (as expected if they were inherited from its parents), or, in contrast, only in one or a few somatic cells (as expected if the sequenced specimens had been parasitized) (Heisserer et al. 2023).

Regarding contamination, all PDV sequences uncovered here in non-lepidopteran genomes are flanked by J1 and J2 motifs. This differs from their main form in wasp genomes, which contains a full HIM and is flanked by DRJs. Furthermore, for five non-lepidopteran species, two genomes were produced at different dates and from different insect samples. For all five species, we confirmed the presence of most PDV integrations in the two genomes (Table S2). We further verified the taxonomic origin of all non-lepidopteran contigs bearing PDV integrations by extracting 10-kb regions upstream and downstream of each integration and checking the taxonomic affiliation of the best blastx hit found in these regions. This analysis yielded at least one blastx hit on the NCBI clusteredNR protein database (length > 100 amino acids, percent of identity > 39%) for one or the two flanking regions of 273 out of the 303 integrations. For the vast majority of the 273 integrations (n = 239), the blastx hit of at least one flanking region was from an arthropod species other than parasitoid wasps, many of them belonging to the same family (n = 146) or order (n = 181) as the query species (Fig. 3 and Tables S3 and S4). Importantly, evidence for the presence of a protein from a non-parasitoid wasp arthropod species was found in the flanking regions of at least one PDV integration for 19 out of the 20 non-lepidopteran species in which we found PDV integrations. For the remaining species (the sawfly Tenthredo livida), blastx search performed using a longer fragment (30 kb) of the regions flanking one of the two PDV integrations yielded a 728-amino acid long hit showing 82% identity to a protein (ribonuclease 3-like) from another sawfly (Athalia rosae). Together, these results show that PDV integrations lie on contigs that are not composed only of non-HGT wasp (parasite) genomic DNA. It is thus highly unlikely that the PDV sequences reported here in non-lepidopteran species come from wasps, or any other contaminating DNA.

To assess whether PDV integrations are somatic (or instead occurred in the germline of an ancestor of the individuals used for genome sequencing), we downloaded raw sequencing reads for five species having more than three integrations and computed sequencing depth across the last 100 bp of the PDV sequence and 300 bp of the flanking region. As in Heisserer et al. (2023), we reasoned that sequencing depth should be homogeneous across PDV-host junctions if they were vertically inherited because for these junctions, both the PDV sequence and its flanking region should be present in all cells. Moreover, the breakpoint of such junctions should be covered by multiple sequencing reads aligning on one side on the PDV sequence and on the other side on the flanking host genome. On the contrary, sequencing depth is not expected to be homogenous over PDV-host junctions that would result from somatic integrations because these are typically present in only one or a few cells (Muller et al. 2021). Furthermore, the PDV–host breakpoints in such somatic integrations are usually covered by only one or very few sequencing reads (Muller et al. 2021). Here, in the five surveyed species, we see a homogeneous average sequencing depth across PDV–host junctions and find that the breakpoint between PDV and host sequences is covered by many reads (mean = 5.4 to 78.6 per integration, depending on the species) (Fig. 4). Altogether, these results indicate that the PDV integrations uncovered in non-lepidopteran species indeed took place in the germline, at some point in the ancestry of the individuals that were sampled for genome sequencing.

Figure 4.

Sequencing depth and number of reads covering host–PDV junctions in five non-lepidopteran species. Raw reads were downloaded for five species harboring integrations: Bacillus rossius (Phasmatodea), Clitarchus hookeri (Phasmatodea), Lochmaea crataegi (Coleoptera), Neodiprion pinetum (Hymenoptera), and Rhogogaster chlorosoma (Hymenoptera) to calculate the sequencing depth across all junctions. The black lines show the mean sequencing depth over all junctions for each position: 100 bp in the integration (coordinate 1 to 100; red rectangle) and 300 bp in the flanking region (coordinate 101 to 400). The standard deviation is shown in gray. Number of integrations are given after species names in parentheses. Mean numbers of chimeric reads covering host–PDV junctions are given below species names.

Using PCR screening of new field samples from the oak bush cricket and the sawfly Diprion similis, we further ruled out contamination in these two species, and assessed at the same time whether the observed PDV integrations were rare in natural populations. In the oak bush cricket, we confirmed that the unique detected integration is shared between the individual initially used for genome sequencing (sampled near Oxford, UK), and two individuals newly sampled in Gif-sur-Yvette (France) (Figure S1). Regarding D. similis, the integration selected from the individual used for genome sequencing (collected from an unknown location) is also shared with four individuals collected in Normandie (France) and five individuals from Ontario and Quebec (Canada). This indicates that the two PDV sequences are probably fixed in the two species.

Leaf beetles, sawflies, and stick insects were attacked by multiple parasitoid wasp lineages in the recent past

We reasoned that the molecular diversity of PDV sequences found in non-lepidopteran genomes could inform us on the diversity of donor parasitoid wasp species that were involved in PDV-mediated HGT. To this end, we obtained multiple alignments of PDV sequences homologous to the J1 and J2 motifs of IV27, the most widespread circle among non-lepidopteran species. To these multiple alignments, we added IV27-like sequences from four Campopleginae wasp species (Hyposoter didymator, Hyposoter fugitivus, Diadegma semiclausum, Diadegma fenestrale). The unrooted trees revealed that PDV sequences from each insect orders are not monophyletic (Fig. 5 and Figure S2). Rather, several well-supported groups include sequences from widely divergent insects. For example, groups 1, 3, and 4 (in Fig. 5) include closely related sequences from sawflies and stick insects. This pattern is consistent with a scenario whereby a given wasp lineage (ie a wasp genus or a group of closely related wasp species) was able to attack multiple insect species at a given time or underwent recurrent switches between widely divergent target lineages. The tree further shows that a group of PDV sequences found in stick insects and sawflies derives from wasps related to H. didymator (group 1 on Fig. 5), while another group of sequences found in the smooth stick insect (C. hookeri) derives from wasps closely related to H. fugitivus (group 2 on Fig. 5). However, the long branches separating all other PDV sequences from those of the four wasps indicate that the HGT of PDV sequences uncovered in this study originate from a large diversity of wasp lineages, for which no genome sequence is available.

Figure 5.

Unrooted phylogeny of HIM-mediated integrations homologous to junction 2 of Hyposoter didymator circle 27 in non-lepidopteran species. The alignment used to reconstruct this phylogeny included all integrations found in non-lepidopteran genomes homologous to junction 2 of circle 27 from Hyposoter dydimator. We also included homologous circles from other parasitoid wasps marked with an asterisk, as potential donors of PDV integrations (DsIV15 of Diadegma semiclausum, C2 of Hyposoter fugitivus, and D4 of Diadegma fenestrale). Integrations are colored per species and each insect order is depicted using a silhouette for illustration purposes. Full gray circles at internal nodes indicate SH-aLRT support values higher than 70%. Five monophyletic groups mentioned in the Result section are numbered from 1 to 4.

Finally, we evaluated the age of PDV sequences found in non-lepidopteran genomes by searching for integrations shared at orthologous loci across different species. Our search revealed only three such sequences, shared between three or four closely related species of the Neodiprion genus (Fig. 2 and Figure S3 and Table S5). This indicates that most PDV sequences uncovered here are species-specific and result from relatively recent integration events. Of note, the dN/dS ratio (calculated on the two genes contained in the PDV sequences shared at orthologous loci among Neodiprion species) did not reveal any sign of purifying selection (dN/dS = 1).

Discussion

In this study, we have shown that horizontal transfers of PDV sequences have occurred in at least 22% of lepidopteran genomes, ie a larger fraction of lepidopteran species than previously estimated by Heisserer et al. (2023). Furthermore, we report PDV-mediated horizontal transfers in multiple non-lepidopteran species. Some of these are known to be parasitized by PDV-encoding parasitoid wasps (leaf beetles and sawflies) while others are not (stick insects and orthopterans). These results thus extend the range of putative targets for PDV-carrying parasitoids.

Our results reinforce the previous finding that HIM-mediated integrations of PDV sequences constitute a major route of HGT from parasitoid wasps to other insects. The high prevalence of these transfers raises the question of their functional impact on recipient species. Previous studies have shown that two C-type lectins—acquired by lepidopterans from parasitoid wasps through PDV-mediated HGT—are now involved in immunity in these host species (Gasmi et al. 2015; Di Lelio et al. 2019). Being present across several species, following an ancient HGT event, these two genes were good candidates for functional analysis. This made it possible to show that they evolved under purifying selection, having retained an intact open reading frame after integration. The characterization of these genes resulted from manual curation performed in a limited number of species and did not rely on the presence of the HIM motif (which was assumed to be lost through mutational decay after integration). In contrast, the large-scale screening presented here only considered PDV integrations in which the HIM motif is still recognizable, implying that only relatively recent integrations could be identified. In agreement with this, few integrations were shared by two or more species and thus amenable to selection analysis. Future assessment of the global impact of PDV sequences acquired through recent HGT could be done through expression analysis of these sequences in a large panel of species, coupled with knockout experiments of some candidates. More generally, a full understanding of the consequences of these transfers would entail developing alternative approaches to systematically screen for older PDV-mediated HGT events, including those for which HIM motifs are no longer recognizable.

While the finding of PDV sequences integrated in the genome of stick insects and an orthopteran suggests that some species within the Phasmatodea and Orthoptera clades may be suitable hosts of PDV-encoding wasps, this interpretation is at odds with current knowledge and it is important to assess whether alternative hypotheses can be ruled out. Indeed, parasitism of stick insects and orthopterans by Campopleginae or Microgastrinae wasps has never been conclusively reported. Wasps from these two clades are only known to parasitize holometabolous hosts at the larval stage. Given that stick insects and orthopterans are hemimetabolous (ie they do not undergo full metamorphosis), we can only speculate that they can be parasitized at juvenile stages, where their cuticle is less robust. It is interesting to note that other endoparasitoid wasps can attack stick insects and orthopterans at the egg stage (Irshad et al. 1978; Baker et al. 1996; Baker 2016) while orthopteran juveniles can also be targeted by ectoparasitoid wasps (Portman et al. 2009). However, none of these cases involve wasps encoding PDVs. For stick insects, we did find two brief notes reporting observations of ichneumonid-like wasps on Anisomorpha buprestoides (King 1867; Bragg 1993) and Paraphasma rufipes (Brasse 1995), but these were not confirmed by subsequent studies.

To challenge the main interpretation of our results—that of a truly extended host range—a first set of alternative hypotheses can be proposed, not even implying that these insects were ever touched by PDV-encoding wasps. First, HGT of PDV sequences could have involved intermediate vectors, such as large double-stranded DNA viruses. A wide diversity of these viruses infect parasitoid wasps and other insects (Strand et al. 2026), and several of them have been proposed to act as vectors of horizontal transfer of transposable elements (Gilbert et al. 2016). Given that integrated circles are no longer mobile, such a scenario would imply that PDV circles were encapsidated in a viral particle (without being integrated into the viral genome), and that the viruses subsequently reached the germline of stick insects and orthopterans. Since HIM-mediated integration involves wasp factors that are not encoded by PDV circles (Wang et al. 2021), this scenario also implies that such factors were shuttled by the virus. Altogether, we see this scenario as very unlikely. In the same vein, a second hypothesis could be that while feeding on a host plant, stick insects and orthopterans may occasionally and accidentally ingest remains of dead caterpillars containing PDV particles. This would imply that PDV particles are able to cross the insect midgut, which is not the natural route of PDV infection and thus also seems unlikely.

Yet two other alternative scenarios may be considered. These still do not require that orthopterans and stick insects fall within the range of suitable hosts for full wasp development, but they do involve puncture by female wasps. First, PDV circles may have been accidentally injected into stick insects and orthopterans during host-feeding by adult parasitoid wasps, whereby female wasps puncture the host with their ovipositor and feed upon host hemolymph and other tissues. While this behavior is widespread among some wasp subfamilies such as Braconinae and Pimplinae (Jervis and Kidd 1986) and could in principle lead to accidental release of PDV particles into targeted insects, it is notable that there is no clear report of such behavior in PDV-encoding wasps (ie Microgastrinae and Campopleginae). Another hypothesis would be that PDV-encoding wasps regularly oviposit and inject PDV particles into stick insects and orthopterans, but that this never results in successful development of the wasp eggs. In instances where the density of the preferred host is low, parasitoid wasps may attempt to attack alternative hosts, as is observed for several species both in lab and wild conditions (Houseweart et al. 1984; Henneman and Memmott 2001; Beckage and Tan 2002; Hiroyoshi et al. 2017). Host shortage may be common in the field, and aberrant oviposition behaviors may occur frequently on a large diversity of unsuitable hosts. In this context, the presence of PDV sequences in stick insects could be interpreted as the result of a greater permissiveness of these insects, compared to other unsuitable hosts, to germline integration of these sequences. However, this scenario suffers from at least two caveats. First, while alternative host seeking may be common, all known alternative hosts of ichneumonid and braconid parasitoid wasps are relatively closely related to the preferred host, ie they are lepidopterans. Second, it is unclear why stick insects and orthopterans would be more permissive to HIM-mediated germline integration than the many other holometabolous insect lineages that are more closely related to lepidopterans and may also be occasionaly targeted by PDV-encoding wasps.

While none of these alternative hypotheses can be fully discarded, we see three arguments supporting the view that stick insects and orthopterans may occasionally serve as hosts to PDV-carrying parasitoid wasps. First, the distribution of PDV integrations we uncovered in protostome genomes is highly structured and coherent with the known host range of PDV-encoding parasitoid wasps. The only non-lepidopteran insects known to be occasionally parasitized by PDV-encoding wasps (sawflies, leaf beetles) and in which we were expecting to find HIM-mediated HGT of PDV sequences, were indeed positive for these sequences. Furthermore, only ichnovirus sequences were found in these two lineages, in agreement with the fact that bracovirus-encoding wasps (Microgastrinae) are not known to parasitize sawflies and coleopterans. Given that HGT events of PDV sequences are found in all known hosts and in a limited number of other lineages, it seems reasonable to hypothesize that, in return, these other lineages may sometimes be targeted by PDV-carrying wasps. Second argument: the junctions between PDV sequences and their flanking regions in the orthopteran and stick insect genomes display the exact same molecular signature of HIM-mediated integration as those found in the somatic tissues of lepidopteran larvae during parasitism, ie they are delimited by the J1 and J2 motifs of the HIM. As proposed for lepidopterans here and in a previous study (Heisserer et al. 2023), it appears parsimonious to infer that stick insects and the orthopteran PDV sequences were introduced under similar parasitism circumstances. Finally, the high number and diversity of PDV sequences found in several stick insects indicate recurrent attacks by multiple ichnovirus- and bracovirus-encoding wasps. Such repeated attacks raise the possibility that oviposition in stick insects may be adaptive; in other words, that these insects may indeed represent suitable hosts of PDV-encoding wasps.

In conclusion, our study supports previous claims that host–parasite relationships favor HGT (Gilbert et al. 2010; Kuraku et al. 2012; Wijayawardena et al. 2013; Davis and Xi 2015), and that HIM-mediated integration of PDV circles is a major route of HGT from parasitoid wasps to other insects (Gasmi et al. 2015; Heisserer et al. 2023). It also further shows that characterizing HGT can be useful to reveal hitherto unknown ecological interactions (Suh et al. 2016; Cai et al. 2021). The finding of HIM-mediated integration of PDV sequences in stick insects and an orthopteran suggests these lineages have been attacked by—and may be hosts of—PDV-encoding parasitoid wasps belonging to the Microgastrinae and/or Campopleginae subfamilies. However, alternative scenarios cannot be excluded, and clarifying the nature of this ecological relationship will ultimately require direct field observations or controlled laboratory experiments. In any case, these interactions warrant consideration in future studies of the community ecology of these insects. They should also be taken into account when evaluating the risks associated with parasitoid-based biocontrol strategies.

Materials and methods

Polydnavirus segments and protostome genomes

The PDV segments we used as queries to search for HIM-mediated PDV integrations in protostome genomes were extracted from the genome of 10 parasitoid wasp species in which segment annotation could be retrieved, including six Microgastrinae (Chelonus inanitus, Cotesia congregata, Cotesia flavipes, Cotesia icipe, Cotesia typhae, Microplitis demolitor), three Campopleginae (D. semiclausum, H. didymator, Tranosema rostrale), and one Banchinae (Glypta fumiferanae). The sequences are provided in Dataset S1, and the coordinates of the HIMs are available in Dataset S2. HIM-mediated PDV integrations were searched in the 6,814 protostome genome assemblies from 4,746 species available in GenBank as of September 2024. Accession numbers, taxonomy identifiers, sizes, assembly levels, dates, and number of scaffolds are provided in Dataset S3. All genomes were downloaded from GenBank using esearch v.19.2 (https://www.ncbi.nlm.nih.gov/books/NBK179288/) based on accession numbers.

Similarity search for HIM-mediated integrations of PDV segments

We used blastn to perform a first similarity search between PDV segments and arthropod genomes only (-task blastn). The resulting blastn hits were treated as follows. First, in cases where multiple hits corresponded to overlapping arthropod genome regions retrieved by different (but homologous) segments, we only retained the hit involving the longest genome region. Then we removed hits shorter than 160 bp and/or with an e-value < 0.0001, and only retained hits containing a full-length or partial HIM motif. Moreover, since low complexity regions can induce false positives, we annotated such regions in all PDV segments used as blastn queries using RepeatMasker v.4.1.9 (-low) (Smit et al. 2013) and removed them from the blastn hits based on their coordinates. We then extracted sequences showing homology to PDV sequences from arthropod genomes using Seqtk subseq (https://github.com/lh3/seqtk) and clustered these sequences using mmseqs2 v. 17.b804f with the easy-cluster command and the following options: --min-seq-id 0.9, -c 0.2, --cov-mode 1 (Mirdita et al. 2019). The representative sequence from each cluster was added to the initial PDV segments, and we used this enlarged list of sequences as queries to perform a second blastn similarity search on all protostome genomes. The blastn hits resulting from this second search were submitted to the same filtering steps as in the first blastn search.

Computing sequencing depth over PDV–insect breakpoints

To validate that the PDV sequences found in insect genomes are present in all cells of the sequenced individuals, and thus result from HIM-mediated integration in the germline genome of an ancestor of these insects, we downloaded raw sequencing long reads for five species having at least three PDV integrations (two stick insects, two sawflies, one leaf beetle). We then mapped these reads on the species genome using minimap2 (default options) (Li 2018) and computed the per-base sequencing depth over the PDV–insect breakpoints with bedtools genomecov (Quinlan and Hall 2010), ie over the last 100 bp of each PDV sequence and 300 bp of flanking region. In addition, we identified chimeric reads by using blastn (-task megablast), ie reads in which a portion aligns on the PDV sequence (at least 28 bp) and the other portion aligns on the insect genome (at least 28 bp).

Reconstructing a phylogeny of insects to map PDV integrations

To illustrate the distribution of PDV integrations in non-lepidopteran insects, we reconstructed a phylogeny using a set of selected species harboring such integrations and others in which no PDV integration was found. Our selection of species was designed to best represent the diversity of available genomes from each taxon in which PDV integrations were found (Table S6). For Phasmatodea, we used all species for which a whole genome sequence is available, given the limited of available genomes, except for the genus Timema for which we only included 4 out of the 11 species. For Orthoptera, we included one species per family in addition to M. thalassinum. For Coleoptera, given the many genomes available, we only included species from the family in which we found PDV integrations (Chrysomelidae) and selected one species per genus. For sawflies (Symphyta), we selected all species from families in which we found PDV integrations (Athaliidae, Cimbicidae, Diprionidae, Tenthredinidae). Phylogenetic relationships among species were inferred using orthologous genes identified through BUSCO (Waterhouse et al. 2018) analysis, employing the insecta_odb10 dataset. A random subset of 300 BUSCO genes that were present as complete orthologs across all species was selected for downstream analyses (Figure S4). Multiple sequence alignments were performed individually for each gene using Clustal Omega (Sievers et al. 2011). Alignments were subsequently trimmed to remove poorly aligned regions using TrimAl (Capella-Gutiérrez et al. 2009) with the automated1 setting. The resulting high-quality alignments were concatenated into a single supermatrix to construct a comprehensive dataset for phylogenetic inference. Phylogenetic tree reconstruction was conducted using IQ-TREE (Minh et al. 2020) under a maximum likelihood (ML) approach. The -m MFP option was employed to select the best-fitting substitution model for each partition using ModelFinder. To assess node support, we performed an approximate likelihood ratio test (aLRT; -alrt 2,500) and ultrafast bootstrap resampling (-bb 2,500). Partitions exhibiting poor model fit or compositional heterogeneity were filtered out using the --symtest-remove-bad option. To root the resulting phylogeny, Allacma fusca was included as an outgroup.

Phylogenetic analysis of PDV sequences

Sequences homologous to HdIV27 (segment 27 of H. didymator) and DsIV15 (segment 15 of D. semiclausum), containing a J1 or J2 boundary, were extracted from insect genomes using Seqtk (https://github.com/lh3/seqtk). Sequences corresponding to the J1 and J2 boundaries were aligned separately. Homologous segments from parasitoid wasps (Hd27 and Ds15) were added to each alignment as well as homologous segments from D. fenestrale and H. fugitivus. Multiple sequence alignments were performed using the MUSCLE algorithm implemented in Geneious v11.0.5 (https://www.geneious.com/) (available in Datasets S4 and S5) and trimmed using TrimAL (Capella-Gutiérrez et al. 2009) with the automated1 setting. ML phylogenetic trees were reconstructed using IQ-TREE v2.0.3 (Minh et al. 2020). The best-fit substitution model for each alignment was determined with ModelFinder, and branch support was assessed using the SH-aLRT test with 1,000 replicates. Resulting phylogenies were visualized and formatted for presentation using R v4.4.1 with the ggtree package (Yu et al. 2017).

Searching for PDV integrations shared at orthologous loci between species

To identify orthologous PDV integrations, we extracted all PDV sequences along with 5,000 bp of flanking genomic regions from insect genomes using Seqtk (https://github.com/lh3/seqtk). An all-versus-all BLAST (-task megablast) alignment was then performed. Self-megablast hits were removed, and we filtered hits based on alignment coverage criteria: ≥50% of the flanking region and ≥90% of the integration sequence had to be aligned. Candidate orthologous integration events were then manually inspected to retain only those PDV loci that were inserted at the same genomic position in two or more species, suggesting shared ancestry of the integration event. Position information and alignments of the three orthologous PDV integration sequences found using this approach are provided in Table S5 and Datasets S6 to S8.

PCR screening of selected PDV integrations

We experimentally verified the presence of two PDV circle integrations in wild-caught individuals of the sawfly D. similis and of the grasshopper M. thalassinum. For D. similis, a total of nine specimens were collected from three geographic locations: four individuals from Tours, France, four from Ontario, Canada, and one from Quebec, Canada. Genomic DNA was extracted using the NucleoMag Tissue Genomic DNA Extraction Kit (Macherey-Nagel), following the manufacturer’s protocol. Primers flanking both junctions (J1 and J2) of the candidate PDV integration were designed using Primer3web (version 4.1.0) (Table S7 and Figure S1). PCR amplification was performed using a T100 Thermal Cycler (Bio-Rad), and the resulting PCR products were Sanger-sequenced.

For M. thalassinum, two specimens (one male and one female) were collected in Gif-sur-Yvette, France. Primers were designed using Geneious v11.0.5 (https://www.geneious.com/) to amplify the single PDV integration found in the M. thalassinum genome produced by the Darwin Tree of Life Project (UK, 2023; GenBank under accession number GCA_943193665.1) (Figure S1). Given that only one junction (corresponding to the J1 motif) could be identified in this species, only one primer pair was designed, with one primer located within the PDV sequence and the other in the flanking host region (Table S7). Genomic DNA was extracted independently from each individual using the NucleoBond AX 20 columns and the NucleoBond Buffer Set IV (Macherey-Nagel), following the manufacturer's protocol. PCR screening for the presence or absence of the integration was performed using DreamTaq PCR Master Mix (Thermo Fisher Scientific), with an annealing temperature of 58 °C. PCR products corresponding to the junction region were subsequently purified using ExoStar, and Sanger-sequenced. The resulting sequenced individuals were aligned to the reference sequence, and the alignments are provided in Datasets S9 and S10.

Supplementary material

Supplementary material is available at Molecular Biology and Evolution online.

Funding

This work was supported by Agence Nationale de la Recherche, project VIRHOZFER ANR-24-CE02-1004 and project HORIZON ANR-17-CE02-0021.

Data and resource availability

All scripts used in this study have been deposited on GitHub: https://github.com/InesMatrougui/Find_Polydnavirus_Integration

Data Availability

All data are incorporated into the article and its online supplementary material.