Bacterial communities within Phengaris (Maculinea) alcon caterpillars are shifted following transition from solitary living to social parasitism of Myrmica ant colonies

Abstract Bacterial symbionts are known to facilitate a wide range of physiological processes and ecological interactions for their hosts. In spite of this, caterpillars with highly diverse life histories appear to lack resident microbiota. Gut physiology, endogenous digestive enzymes, and limited social interactions may contribute to this pattern, but the consequences of shifts in social activity and diet on caterpillar microbiota are largely unknown. Phengaris alcon caterpillars undergo particularly dramatic social and dietary shifts when they parasitize Myrmica ant colonies, rapidly transitioning from solitary herbivory to ant tending (i.e., receiving protein‐rich regurgitations through trophallaxis). This unique life history provides a model for studying interactions between social living, diet, and caterpillar microbiota. Here, we characterized and compared bacterial communities within P. alcon caterpillars before and after their association with ants, using 16S rRNA amplicon sequencing and quantitative PCR. After being adopted by ants, bacterial communities within P. alcon caterpillars shifted substantially, with a significant increase in alpha diversity and greater consistency in bacterial community composition in terms of beta dissimilarity. We also characterized the bacterial communities within their host ants (Myrmica schencki), food plant (Gentiana cruciata), and soil from ant nest chambers. These data indicated that the aforementioned patterns were influenced by bacteria derived from caterpillars’ surrounding environments, rather than through transfers from ants. Thus, while bacterial communities are substantially reorganized over the life cycle of P. alcon caterpillars, it appears that they do not rely on transfers of bacteria from host ants to complete their development.


| INTRODUC TI ON
Microbial symbionts can mediate diverse physiological processes in animals, particularly through adaptations that extend or enhance their trophic capacities. These symbioses can also lead to metabolic, developmental, and immunological adaptations in host animals, which facilitate their colonization of new environments and ultimately their evolution (McFall-Ngai et al., 2013;Moran, 2002Moran, , 2007. Many insects also benefit from microbial symbioses, and their vast diversity in form and function may have arisen in part through associations with beneficial microorganisms, particularly bacteria (Engel & Moran, 2013). Recently, gut bacteria have been shown to enhance digestive capabilities (Brune, 2014;Kwong & Moran, 2016;Russell et al., 2009), protect against pathogens and predators (Koch & Schmid-Hempel, 2011;Kwong, Mancenido, & Moran, 2017), and provide signals for inter-and intraspecific communication (Davis, Crippen, Hofstetter, & Tomberlin, 2013) and mating (Sharon et al., 2010) in insects.
Recently, Whitaker, Salzman, Sanders, Kaltenpoth, and Pierce (2016) found no clear link between trophic regime and gut bacterial composition, despite sampling a wide range of feeding strategies across 31 species of Lycaenid caterpillars. Hammer, Janzen, Hallwachs, Jaffe, and Fierer (2017) similarly found low densities of microbes in the guts of caterpillars spanning 124 species and 15 families.
Transient bacteria, which are excreted shortly after they are ingested with food, may dominate bacterial communities within caterpillars due to both physiological and ecological limitations. Highly alkaline conditions in the gut, coupled with relatively short and simple gut structures and a continuously replaced gut lining may limit or prevent the colonization of resident bacteria in caterpillars (Hammer et al., 2017). Development through several larval instars and metamorphosis may also dramatically reshape caterpillar digestive systems and any bacterial communities within them (Chen et al., 2016;Hammer et al., 2014). Moreover, many Lepidopterans engage in few social interactions outside of mating. This largely asocial development may also contribute to the apparent lack of beneficial resident bacteria within caterpillars, though until now, this has not been tested.
While social interactions may be uncommon for most caterpillars, many Lycaenid caterpillars engage in highly specialized interactions with eusocial ants. It is estimated that 75% of the approximately 6,000 Lycaenid species display some degree of myrmecophily (i.e., association with ants; reviewed in Pierce, 1995 andPierce et al., 2002). These are usually facultative mutualistic interactions, in which ants protect caterpillars from predators and parasitoids in exchange for nutritive secretions. However, obligate parasitic associations also occur in a small subset (<5%) of myrmecophilous Lycaenid species (Pierce et al., 2002), including in the genus Phengaris (formerly Maculinea). Parasitic Phengaris caterpillars enter host ant colonies and feed either through ant regurgitations (trophallaxis), or by directly preying upon ant larvae.
Our focal species is the Alcon blue (Phengaris alcon), a widely studied parasitic Lycaenid species with a "cuckoo" feeding strategy.
During the fourth instar, they fall off their host plant and are adopted by Myrmica worker ants, typically Myrmica schencki (Witek et al., 2008), though host ants can vary across the species distribution (Tartally, Nash, Lengyel, & Varga, 2008). Caterpillars utilize a combination of chemical (Akino, Knapp, Thomas, & Elmes, 1999;Nash, Als, Maile, Jones, & Boomsma, 2008) and acoustic (Barbero, Thomas, Bonelli, Balletto, & Schonrogge, 2009;Sala, Casacci, Balletto, Bonelli, & Barbero, 2014) signals to communicate with ants and avoid aggression, living in the colony for 1-2 years before pupating and emerging from the nest as an adult. bacterial community composition in terms of beta dissimilarity. We also characterized the bacterial communities within their host ants (Myrmica schencki), food plant (Gentiana cruciata), and soil from ant nest chambers. These data indicated that the aforementioned patterns were influenced by bacteria derived from caterpillars' surrounding environments, rather than through transfers from ants. Thus, while bacterial communities are substantially reorganized over the life cycle of P. alcon caterpillars, it appears that they do not rely on transfers of bacteria from host ants to complete their development.

K E Y W O R D S
16S amplicon sequencing, butterflies, Lepidoptera, Lycaenidae, myrmecophily, Spiroplasma While living inside ant colonies, P. alcon caterpillars are dependent on regurgitations from host ant workers for nutrition. These regurgitations are rich in protein; M. schencki regularly consume other ants, as well as honeydew, nectar, and pollen (Czechowski, 2008).
Regurgitations can be tailored to suit the nutritional needs of ant larvae (Dussutour & Simpson, 2009), and worker ants can play a role in the digestive processes of larvae directly, or by transferring beneficial gut symbionts with their regurgitations (Brown & Wernegreen, 2016). Consequently, when P. alcon rapidly shift from plant feeding to protein-rich ant regurgitations, they may be able to enhance their survival and integration within ant colonies by exploiting bacterial transfers from their ant hosts.
Here, we leverage the asocial-to-social transition of Phengaris alcon caterpillars and the associated shift in diet to test whether obligate myrmecophily reshapes their bacterial communities. To address this question, we surveyed populations of wild P. alcon caterpillars, both while they were feeding on G. cruciata buds and after they had entered M. schencki colonies, using high-throughput 16S rRNA amplicon sequencing. We also sequenced the bacterial communities within worker ants and ant larvae, and the surrounding environments of caterpillars (i.e., G. cruciata buds, and soil from inside ant nest chambers), to better understand the origins of any microbes present within caterpillars. Additionally, we used quantitative PCR to determine the total quantities of bacteria within P. alcon caterpillars and to test whether the number of bacteria within caterpillars shifted following their transition to living inside ant colonies. Together, these allowed us to fully assess the significance of bacterial symbioses as part of P. alcon caterpillars' complex life history.

| Sample collection
Samples were collected across the Alps (Switzerland and Northern Italy) and Pyrenees (Spain) mountain ranges between 2015 and 2016. We collected III instar Phengaris alcon caterpillars by dissecting G. cruciata buds and IV instar caterpillars by excavating M. schencki nests. All caterpillars were starved for 3-4 hr until they evacuated their gut contents, and were then individually preserved in RNAlater ® (Thermo Fischer Scientific) tubes. M. schencki workers and larvae were collected from all ant colonies hosting P. alcon caterpillars, and were starved, preserved, and stored under the same conditions as caterpillars. Environmental samples (whole G. cruciata buds that caterpillars were eating, and 250 mg of fresh soil from ant nest chambers containing caterpillars) were collected in tandem with the above samples and frozen at −80°C until extraction.

| 16S rRNA amplicon processing
DNA extraction, library preparation, and preprocessing steps are detailed in Supporting Information Appendix S1. To summarize, we (a) extracted bacterial DNA from surface-sterilized whole individuals, (b) amplified the V3/V4 region of the 16S rRNA gene in each sample, and (c) produced MiSeq-compatible libraries for 300 bp paired-end sequencing. Following these initial steps, we trimmed reads to 400 bp and performed open-reference OTU picking in QIIME v.1.9.1 (Caporaso et al., 2010), using UCLUST (Edgar, 2010) to cluster OTUs at 97% identity. We filtered out probable chimeric sequences using UCHIME (Edgar, Haas, Clemente, Quince, & Knight, 2011) and the GOLD reference database (Reddy et al., 2015).
We assigned taxonomies using UCLUST and two reference databases: Greengenes v13_8 (DeSantis et al., 2006;McDonald, Price et al., 2012a) and SILVA NR Small Subunit v128 (Quast et al., 2013). We then used QIIME to filter out low abundance OTUs (i.e., with fewer than two reads) and over-represented sequences (Gentiana chloroplast DNA and Wolbachia), produce biom (McDonald, Clemente et al., 2012b) tables for both the Greengenes-and SILVA-annotated datasets, and create a phylogenetic tree using FastTree (Price, Dehal, & Arkin, 2009). Next, we investigated whether the trophic shift and social association experienced by caterpillars in ant colonies led to more consistent bacterial communities, using assessments of beta dissimilarity.

| 16S rRNA amplicon diversity analyses
For these analyses, we rarefied the raw Greengenes-annotated biom tables to even sampling depth (1,000 reads per sample), calculated Bray-Curtis and unweighted UniFrac distance matrices and visualized the results with nonmetric multidimensional scaling (NMDS) and principal coordinate analysis (PCoA) ordinations, respectively.

| Determining the origins of bacterial communities within P. alcon caterpillars
Our final set of analyses using the 16S rRNA amplicon sequencing data investigated the relative contributions of social interactions and the environment on bacterial community composition and stability within P. alcon caterpillars. For these analyses, we CSS-normalized (Paulson, Stine, Bravo, & Pop, 2013) the raw Greengenes-annotated biom table using QIIME and used hclust2 (Segata, 2017) to visualize differences in abundances among the 40 most abundant OTUs (in terms of total read counts), clustering samples and features using Bray-Curtis dissimilarity. Then, we extracted the representative (i.e., most abundant) sequences for these 40 OTUs and performed BLAST searches of the NCBI nucleotide collection and 16S rRNA gene sequence databases to further improve the resolution of taxonomic identifications where possible. Then, we identified OTUs present in worker ants and caterpillars but not soil (i.e., OTUs that may have been exchanged between insects rather than environmentally derived) using the shared_phylotypes function in QIIME.
To determine which OTUs had the highest probability of being differentially abundant between all groups of caterpillars and ants, we performed a G-test on the CSS-normalized dataset using QIIME.
To test for an effect of geography on the observed abundances, we repeated the G-test using sample sites to group caterpillar and ant samples. We also used a Wilcoxon rank sum test to test for an effect of geography across caterpillars from Switzerland and Italy. All of the above-mentioned tests included Bonferroni correction for multiple testing.
Finally, we searched for differentially abundant bacteria with possible digestive roles within P. alcon caterpillars using PICRUSt (Langille et al., 2013). For these analyses, we generated a closed-reference OTU table from the Greengenes-annotated, CSS-normalized, dataset, and predicted metagenomic functions of OTUs in the form of KEGG Orthologs (Kanehisa & Goto, 2000). Then, we tested for the presence of differentially abundant features between caterpillars on plants and caterpillars in ant colonies using LEfSe (Segata et al., 2011).

| Quantitative PCR analyses
We assessed whether the total quantities of bacteria within P. alcon caterpillars shifted following their association with ants using quantitative PCR. Using universal 16S rRNA primers, we determined the absolute and relative quantities of total bacteria within individual caterpillar and ant samples. Additionally, we determined the quantities of Wolbachia and Spiroplasma species present within caterpillars and ants using custom primers, based on the sequences present in our 16S rRNA amplicon sequencing dataset. All primers, PCR conditions and additional details on absolute and relative quantification methods are detailed in Supporting Information Appendix S1.

| RE SULTS
Among our three sampling locations (Supporting Information Appendix S2: Figure S1), we successfully sampled P. alcon caterpillars before and after their trophic shift at two sites (Switzerland and Italy). We were unsuccessful in locating caterpillars within ant colonies in Spain, but still sampled and sequenced caterpillars feeding on plants (n = 4) there. We sampled similar numbers of caterpillars on plants in Switzerland and Italy (n = 4 and n = 5, respectively). We found caterpillars within one ant colony in Switzerland (n = 4), and within two ant colonies at the same site in Italy (n = 2 and n = 3).

Total numbers of samples for each group are detailed in Supporting
Information Appendix S3, Table S1.
We identified 27,630 operational taxonomic units (OTUs) in the Greengenes-annotated 16S rRNA amplicon sequencing dataset, and 28,504 OTUs in the SILVA-annotated dataset. Excluding the F I G U R E 1 Bacterial community composition within Phengaris alcon caterpillars and Myrmica schencki workers and larvae. There is a clear shift in community composition following P. alcon caterpillars' transition to parasitizing M. schencki colonies. We observed notable decreases in the abundances of Pseudomonadaceae and Enterobacteriaceae and an increase in Actinomycetales following caterpillars' transition to living inside ant colonies. Note: average relative abundances for each group, across the top 40 OTUs in terms of total read count (62.9% of the total dataset) are shown above Initial exploratory analyses revealed that the Greengenes taxonomic identifications were generally of higher resolution than those produced using SILVA, with more genus-level identifications and fewer unidentified OTUs. Thus, all results presented below will be based on Greengenes taxonomic identifications. However, we note that the SILVA-annotated dataset produced similar results overall (Supporting Information Appendix S2: Figure S2).

| Phengaris alcon caterpillars in ant colonies host more diverse and consistent bacterial communities
We observed a significant increase in the alpha diversity of bacterial communities within P. alcon caterpillars living in ant colonies (Nonparametric two-sample t test; p < 0.001). Caterpillars in ant colonies had the highest alpha (Shannon index) diversities, while caterpillars on plants appeared to be the most variable group (Supporting Information Appendix S2: Figure S3). In addition to producing more diverse bacterial communities within P. alcon caterpillars, the transition to living inside ant colonies also appeared to produce more consistent communities of bacteria in terms of beta diversity. In both Bray-Curtis/NMDS and unweighted UniFrac/ PCoA ordinations, caterpillars on plants covered a wider area on the plots (i.e., were more dissimilar to one another) than caterpillars in ant colonies (Figure 2). This pattern was most pronounced when phylogenetic distances between OTUs were considered using UniFrac distances, though only 26.7% of the variance was TA B L E 1 OTUs present in caterpillars throughout both life stages (i.e., both before and after their trophic shift and association with ants) (shared OTUs found IT only)   1025949_Mesorhizobium  1040713_Corynebacterium  1062748_Mycobacterium  928766_Chitinophagaceae  4394913_Sediminibacterium  168031_Erwinia  280799_Tepidimonas  590099_Sphingomonas  1091060_Sphingomonas_yabuuchiae  569952_Roseateles_depolymerans  544356_Polaromonas  136015_Delftia  136485_Methylobacterium_adhaesivum   963779_Agrobacterium  1093466_Agrobacterium  829523_Phyllobacteriaceae  816470_Bacillus  161287_Spiroplasma  698961_Spiroplasma  759061_Enterobacteriaceae  783638_Enterobacteriaceae  778478_Enterobacteriaceae  646549_Pseudomonas  967275_Stenotrophomonas  331752_Ralstonia  1108960_Sphingomonas  1104546_Rhizobiaceae  68621_Delftia  637901_Delftia   620684_Mesorhizobium  593555_Gluconobacter  1012112_Solirubrobacteraceae  622212_Spiroplasma  109263_Pseudomonas  836096_Pseudomonas  287032_Pseudomonas  279231_Pseudomonas  61192_Oxalobacteraceae  382348_Achromobacter  572643_Sinobacteraceae  1052559_Sphingomonadaceae  1091060_Sphingomonas  336364_Rhizobiaceae 210485_Comamonadaceae 323364_Delftia 525648_Rhizobiales

P. alcon on bud & P. alcon in ant colony
Note. The left-and rightmost columns contain the shared OTUs unique to Switzerland and Italy (respectively), while the center column contains the shared OTUs found in both countries. These OTUs represent the approximately 10% of bacterial taxa that persisted in P. alcon caterpillars following their trophic shift. Based on the Greengenes taxonomic identifications given above, most appear to be transient, environmentally derived bacteria.
explained by the first two axes of the PCoA. M. schencki workers and larvae also appeared to maintain distinct communities of bacteria, though ant samples from Switzerland did not cluster consistently.

| Phengaris alcon caterpillars share many OTUs with their surrounding environments
While P. alcon caterpillars inside ant colonies appear to host more diverse and similar communities than caterpillars on plants, environmentally derived and putatively transient bacteria likely contributed to the above patterns; Swiss and Italian P. alcon caterpillars in ant colonies shared 79% and 87% of their total microbial diversity with ant nest soil, respectively. This result is also apparent when clustering groups based on the 40 most abundant OTUs in terms of total read counts (Figure 3). After manually confirming taxonomies of the most abundant bacteria using BLAST, we found that most of the highly abundant bacteria in our dataset are common on plants, or in soil and water (though we also note that bacteria with similar taxonomic identities can be adapted to different environments  Phengaris alcon caterpillars living in ant colonies (n = 9) appear to host more similar bacterial communities than caterpillars on plants (n = 13), in terms of beta dissimilarity. Note: both distance matrices were calculated from the Greengenes-annotated dataset, with read counts rarefied to even sampling depth F I G U R E 3 Heatmap of the 40 most abundant OTUs, with Bray-Curtis clustering of sample types (X-axis; groups collapsed by averaging OTU abundances) and OTUs (Y-axis). Environmental and/or pathogenic bacteria appear to account for most of the differentiation between Phengaris alcon caterpillars on plants and caterpillars in ant colonies. However, Spiroplasma species also appear to be useful in distinguishing between groups. Note: OTUs with >97% identity were denoted with subscripts (i.e., 1a/1b), while those with <97% identity were separately numbered.

| Phengaris alcon caterpillars host relatively small total quantities of bacteria
the top of the range for quantities of bacteria known to be hosted by caterpillars. However, it should be noted that the total quantities of bacteria within P. alcon caterpillars, when scaled based on their size, are still lower than the quantities observed in other insects and animals.
Quantitative PCR analyses revealed that P. alcon caterpillars on plants hosted more variable, though overall not significantly different (Wilcoxon p > 0.05) absolute quantities of bacteria compared to caterpillars living in ant colonies (Figure 4). This variability within caterpillars on plants is also consistent with the patterns observed in our 16S amplicon sequencing data. When controlling for caterpillar size differences, we observed the same patterns in relative and absolute quantities of bacteria (Supporting Information Appendix S2: Figure   S4). Using species-specific qPCR primers, we also found that individual caterpillars and ants predominantly hosted either Wolbachia or Spiroplasma (Supporting Information Appendix S2: Figure S5).

| D ISCUSS I ON
Building on recent broad molecular surveys of microbial diversity within caterpillars (Hammer et al., 2017;Phalnikar et al., 2018;Whitaker et al., 2016), we characterized and compared bacterial communities within Phengaris alcon caterpillars before and after their trophic shift and social association with M. schencki ants.
We observed a compositional shift (Figure 1), increase in diversity (Supporting Information Appendix S2: Figure S3), and homogenization ( Figure 2) of bacterial communities within caterpillars following their transition to living inside M. schencki colonies. However, M. schencki workers and larvae shared relatively few bacteria with caterpillars living in their nests, and many of the most abundant bacteria within P. alcon were species common in soil and water ( Figure 3). Taken together, these results imply that most bacteria within caterpillars are derived from their food and surrounding environment. These findings are consistent with other recent characterizations of Lepidopteran microbiota (Hammer et al., 2017;Phalnikar et al., 2018;Staudacher et al., 2016;Whitaker et al., 2016).
Quantitative PCR analyses were also generally consistent with the patterns observed in the 16S amplicon sequencing dataset.
Notably, we did not detect significant differences in total quanti- While Phalnikar et al. (2018) recently reported that bacterial communities within caterpillars (including two Lycaenidae) generally did not change during development, and that dietary transitions had weak effects on bacterial communities, our focused sampling (fully replicated across Switzerland and Italy) found a more substantial shift. Few "core" bacteria appear to persist over P. alcon caterpillar development; 8%-10% of OTUs persisted across both stages of development and both sampling sites (see Figure 1 and Table 1). However, none of the caterpillars in Phalnikar et al.
(2018) underwent a trophic shift and change in environment as sudden and drastic as that experienced by P. alcon caterpillars.
Furthermore, two Lycaenid species (Leptotes plinius and Spalgis epius) in Phalnikar et al. (2018) were not obligate myrmecophiles (Common & Waterhouse, 1972;Venkatesha, 2005). Given this result, we set out to disentangle the influence of diet, surroundings, and ant association on the diversity, structure, and origins of bacteria within P. alcon caterpillars.
In our initial comparisons of alpha diversities, we observed greater variability in bacterial community richness within P. alcon caterpillars on plants (Supporting Information Appendix S2: Figure   S3). Some individuals were overwhelmingly dominated by one or a few bacteria not known to aid in digestion of plant material, suggesting that caterpillars do not crucially rely on metabolic associations with bacteria during most of their early development. This is not surprising, given that P. alcon caterpillars acquire 99% of their total biomass while living inside ant colonies (Thomas, Elmes, Wardlaw, & Woyciechowski, 1989). While some Lycaenidae are known to eat their eggshells, P. alcon caterpillars hatch basally, eating through the underside of the leaf their egg was laid on; they also do not eat F I G U R E 4 Boxplots representing total 16S rRNA gene copies per microlitre of DNA extraction in Phengaris alcon and Myrmica schencki samples. P. alcon caterpillars living inside ant colonies hosted more consistent, but not significantly different (Wilcoxon p > 0.05) total quantities of bacteria compared to P. alcon on plants. Note: two caterpillars living on plants from Switzerland, and all four caterpillars on plants from Spain are not shown above, as an insufficient quantity of DNA remained following 16S rRNA amplicon sequencing library preparation. Ant workers and larvae from the same nest were also (separately) pooled prior to extraction.  (Thomas, Munguira, Martin, & Elmes, 1991). This reduces the possibility for maternal transmission of bacteria to caterpillars, and thus it is likely that most bacteria within caterpillars on plants were derived from the G. cruciata buds they were eating.
While some P. alcon caterpillars on plants hosted diverse bacterial communities, many were dominated by Pseudomonadaceae, which include both plant-growth promoting and pathogenic species (Preston, 2004) and Enterobacteriaceae, which include many common, harmless symbionts, but also pathogenic species.
In general, it would appear that the dominant groups of bacteria within P. alcon caterpillars in ant colonies are also derived from their surrounding environment, rather than through transfers from ants. Following the transition to living inside ant colonies, Pseudomonadaceae and Enterobacteriaceae decreased in abundance, while several families within the order Actinomycetales, particularly Nocardiaceae, increased in abundance (Figure 1). These bacteria are commonly found in soil and water (Goodfellow, 2014).

The most abundant families in worker ants, Spiroplasmataceae and
Oxalobacteraceae, were not similarly abundant within caterpillars.
Caterpillars in ant colonies also hosted a greater diversity of bacteria than their host ants (see Figure 1 and Supporting Information Appendix S2: Figure S3). This implies a bacterial contribution from a source other than host ant regurgitations, such as soil. However, the lower diversity and quantities of bacteria within M. schencki may also be a consequence of more effective filtering of environmental bacteria, through immune defenses (Cremer, Armitage, & Schmid-Hempel, 2007) or colonization resistance (Spees, Lopez, Kingsbury, Winter, & Bäumler, 2013).
Our measures of beta dissimilarity revealed that P. alcon caterpillars on plants could be highly dissimilar to one another, even within the same site ( Figure 2). In contrast, caterpillars in ant colonies clustered more closely together and also clustered according to sampling location. Our qPCR data corroborated this finding, with caterpillars in ant colonies hosting more consistent (though not significantly greater) quantities of bacteria than caterpillars on plants (Figure 4).
Taken together, these results suggest a homogenization of bacterial communities occurs within caterpillars following their transition to living inside ant colonies. Homogenous bacterial communities are a hallmark of highly social species (Shropshire & Bordenstein, 2016), and P. alcon caterpillars' associations with ants seem to have led to consistent communities across a wide geographic range (i.e., across the Alps). However, environmentally derived bacteria likely remain the main driver of this pattern for P. alcon caterpillars (see Figure 3).
Given that P. alcon caterpillars in ant colonies shared 79%-87% of their OTUs with nest chamber soil, the observed shift in microbial communities following their transition from plants to ant colonies was certainly influenced by corresponding shifts in environmental bacteria. Some of these bacteria found in the environment could still have been acquired via trophallaxis, but we were unable to control for this when sampling wild populations of caterpillars. However, even with our more conservative analyses, further examination of the taxonomic identities of putatively transferred OTUs revealed that most were likely transient bacteria.
Some of the most consistently present bacteria in both caterpillars and ants are Spiroplasma and Wolbachia, two well-known insect endosymbionts. Pathogenic strains of both Spiroplasma and Wolbachia are known to cause cytoplasmic incompatibility, feminization, and male killing. Wolbachia are very common parasites of lepidopterans (Salunkhe, Narkhede, & Shouche, 2014), and some Spiroplasma may play similar parasitic roles in lepidopterans (Jiggins, Hurst, Jiggins, v. d. Schulenburg, & Majerus, 2000). However, potentially mutualistic symbiotic effects have also been uncovered for both Spiroplasma (Jaenike, Unckless, Cockburn, Boelio, & Perlman, 2010;Xie, Vilchez, & Mateos, 2010) and Wolbachia (Bian, Xu, Lu, Xie, & Xi, 2010;Hedges, Brownlie, Oneill, & Johnson, 2008;Hosokawa, Koga, Kikuchi, Meng, & Fukatsu, 2010) in other insect groups. However, no such mutualisms between caterpillars and Wolbachia are currently known, so we considered Wolbachia to be an intracellular parasite only. Both Wolbachia and Spiroplasma can co-occur within a host and have possible interactive effects on host immunity (Goto, Anbutsu, & Fukatsu, 2006;Shokal et al., 2016), though in our dataset, we observe a negative correlation between their abundances (Supporting Information Appendix S2: Figure S5). One explanation for this pattern is that Spiroplasma and Wolbachia may be respectively adapted to their ant and caterpillar hosts, and thus appear at lower abundances during cross-infections.
Spiroplasma are known to be enriched among predatory ant species, including many Myrmica species (Anderson et al., 2012;Funaro et al., 2010). Recent research has also detected possible mutualistic Spiroplasma associations with Myrmica, which may aid in nutrient uptake and immunity (Ballinger, Moore, & Perlman, 2018). Transfers of these Spiroplasma from ants to caterpillars may therefore also aid in their digestion of regurgitated materials. Here, we detected two Spiroplasma with <97% identity (i.e., different strains/species), with some geographic variation in their abundances across Switzerland and Italy (see Figure 3). This may suggest local, long-term mutualistic strains within host ants. However, our quantitative PCR results con- In addition to Spiroplasma, OTUs in the order Actinomycetales (e.g., Nocardioidaceae) were shared among ants and caterpillars in both Switzerland and Italy, but were not present in soil samples.
Here, we found that that Actinomycetales are more abundant within caterpillars than ants ( Figure 1); in fact, Actinomycetales account for <1% of all reads within ant workers and larvae. This may be due to our decision to surface sterilize both ants and caterpillars, which would eliminate bacteria colonizing the niche that Actinomycetales are most commonly associated with. However, surface-sterilization also revealed that Actinomycetales colonize caterpillar gut (and other noncuticular) tissues more effectively than in ants. While it is possible that Actinomycetales may protect caterpillars and ants against pathogens in the ant nest environment, this difference in localization and abundance reduces the likelihood that they play identical roles in both caterpillars and ants.

| CON CLUS ION
Microbes are increasingly being recognized as having a strong influence on the evolution of sociality (Archie & Theis, 2011; Archie & Tung, 2015;Lombardo, 2008). However, it remains difficult to disentangle the influences of shared diets, shared environments, and social interactions on microbial communities without controlled, long-term studies (e.g., Tung et al., 2015). We observed a homogenization of bacterial communities within P. alcon caterpillars following their social association with ants, and could identify possible transfers of a few species, notably Spiroplasma and Nocardiaceae, between ants and caterpillars. However, as observed in other caterpillars, the majority of bacteria characterized were not present in host ants, but were rather abundant in caterpillars' food and surroundings (i.e., G. cruciata buds and ant nest chamber soil).
Ultimately, it appears that bacterial symbionts are not essen-

ACK N OWLED G M ENTS
We thank three anonymous reviewers and the editors for their helpful comments on this manuscript. We also thank Nils Arrigo for helpful comments on an earlier version of this  (1309425). Funders had no role in study design or preparation of the manuscript.

CO N FLI C T O F I NTE R E S T
None declared.

DATA ACCE SS I B I LIT Y
Raw 16S