Bacterial associates of seed-parasitic wasps (Hymenoptera: Megastigmus)

Common heritable endosymbiont infections in Megastigmus

Three species tested positive in our inherited symbiont screens, with infection frequencies ranging from 33100% (Table1). Megastigmus milleri harbours a strain of Rickettsia from the bellii clade (Figure1) [GenBank:KJ353735]. Megastigmus amicorum and M. bipunctatus harbour a strain of Rickettsia that is allied with R. felis, i.e. in the `transitional group [65]. Rickettsia citrate synthase sequences from these two hosts were identical [GenBank:KJ353732 - KJ353734]. These two hosts also harboured supergroup A Wolbachia infections (Figure2) [GenBank:KJ353723 - KJ353731]. M. amicorum collected from different host plants and locations (Juniperus oxycedrus from Corsica and J. phoenicea from mainland France) were 2% divergent in mitochondrial COI [GenBank:KJ535736 - KJ535737] and infected with different Wolbachia strains. There was no significant difference in the frequency of infection in males and females, nor did we find an association between Wolbachia and Rickettsia in coinfected species (Fishers exact tests, data not shown). Arsenophonus, Cardinium, and Spiroplasma were not detected in Megastigmus samples screened using PCR with symbiont-specific primers.

Species	Host plant	Year	Location	N	Sample type	Wolbachiapositive	Rickettsiapositive
Family: Pinaceae
M. schimitscheki	Cedrus atlantica	2010	Petit Luberon, FR	15	Female
M. schimitscheki	Cedrus atlantica	2009	Mont Ventoux, FR	14	Female
M. schimitscheki	Cedrus atlantica	2010	Saou, FR	14	Female
M. schimitscheki	Cedrus atlantica	2010	Gap, FR	15	Female
M. schimitscheki	Cedrus atlantica	2008	Barjac, FR	15	Female
M. schimitscheki	Cedrus libani	2005	Turkey	9	Female
M. rafni	Abies alba	2009	Lespinassire, FR	15	Female
M. rafni	Abies alba	2009	Pardailhan, FR	15	Female
M. rafni	Abies alba	2010	Ventouret, FR	15	Female
M. rafni	Abies alba	2004	Doubs, FR	9	Female
M. rafni	Abies nordmanniana	2000	Rold Skov, DK	9	Female
M. rafni	Abies grandis	2012	Vancouver Island, CAN	16	Female
M. rafni	Abies grandis	2012	Vancouver Island, CAN	10	Male
M. milleri	Abies grandis	2012	Vancouver Island, CAN	16	Female		75% (12)
M. milleri	Abies grandis	2012	Vancouver Island, CAN	10	Male		90% (9)
M. spermotrophus	Pseudotsuga menziesii	2011	British Columbia, CAN	26	Female
M. spermotrophus	Pseudotsuga menziesii	2011	British Columbia, CAN	10	Larvae
Family: Cupressaceae
M. watchli	Cupressus sempervirens	2011	Sallles du Bosc, FR	15	Female
M. watchli	Cupressus sempervirens	2011	Monfavet, FR	15	Female
M. watchli	Cupressus sempervirens	2011	Ruscas, FR	16	Female
M. watchli	Cupressus sempervirens	1997	Aghois Ioannis, GR	10	Female
M. bipuncatatus	Juniperus sabina	2011	Brianon, FR	10	Female	90% (9)	100% (10)
M. bipuncatatus	Juniperus sabina	2011	Pallon, FR	13	Female	38% (5)	54% (7)
M. bipuncatatus	Juniperus sabina	2011	Pallon, FR	10	Male	50% (5)	60% (6)
M. amicorum	Juniperus phoenicea	2011	Petit Luberon, FR	8	Female	100% (8)	100% (8)
M. amicorum	Juniperus phoenicea	2011	Luberon, FR	15	Female	100% (15)	93% (14)
M. amicorum	Juniperus phoenicea	2011	Luberon, FR	10	Male	80% (8)	70% (7)
M. amicorum	Juniperus oxycedrus	2009	Corsica, FR	10	Female	70% (7)	80% (8)
M. amicorum	Juniperus oxycedrus	2011	Corsica, FR	10	Female	80% (8)	100% (10)
M. amicorum	Juniperus oxycedrus	2011	Corsica, FR	9	Male	33% (3)	56% (5)
Parasitoids of M. spermotrophus
Eurytoma sp.	-	2011	British Columbia, CAN	7	-
Mesopolobus sp.	-	2011	British Columbia, CAN	16	-

These species did not host Arsenophonus, Cardinium, or Spiroplasma. Spiroplasma was identified from Eurytoma sp. using 16S rRNA pyrosequencing.

Microbial associates of M. spermotrophus

16S rRNA bacterial amplicon pyrosequencing of M. spermotrophus (adult females, larvae and pupae), adult Eurytoma sp. and P. menziesii ovules generated 81,207 raw reads with an average length of 422bp (see Additional file 1) [BioProject: PRJNA239784]. Quality and chimera filtering removed approximately 27% of the reads. The assignment of operational taxonomic units (OTUs) resulted in 352 unique bacterial clusters after the removal of singletons. A total of 160 OTUs were assigned to the genus level. The average sequencing depth was 3,616 sequences per sample (minimum and maximum of 1,962 and 6,130 sequences per sample). Rarefaction analysis showed that for most of the M. spermotrophus samples the number of observed OTUs no longer exponentially increased after an approximate sampling depth of 3,000 sequences (see Additional file 2) and the average number of observed species was 60 ± 13 and the average Chao1 species diversity estimate was 71 ± 25.

The relative abundance of the major OTUs from the different developmental stages of M. spermotrophus was mostly conserved (Figure3), and there was no difference in the core microbiomes of the different developmental stages, based on principle coordinate analysis of weighted or unweighted UniFrac phylogenetic distances (see Additional file 3). The total relative abundance of OTUs from the class Betaproteobacteria (all in the genus Ralstonia) ranged from 46.4 - 72.3%. One female sample contained only a very small proportion of OTUs assigned to the class Gammaproteobacteria (0.36% relative abundance) while the total relative abundance of Gammaproteobacteria ranged from 12.7 - 33.1% in the remaining samples. The total relative abundance of all OTUs within the class Actinobacteria (all in the genus Corynebacterium) ranged from 1.9 - 7.1%.

A maximum likelihood phylogeny for Ralstonia was created using 16S rRNA sequence from the most abundant Ralstonia OTU in the pyrosequencing data set (Figure4). Strong bootstrap support (0.99) clusters the Ralstonia isolated from M. spermotrophus with the human pathogen R. pickettii (sequence divergence = 3.3%).

Ovule samples were dominated by chloroplast rRNA (99.0%); the remaining OTUs included Ralstonia (0.8%) and Acinetobacter (0.2%). The Eurytoma parasitoid samples were dominated by one OTU, which is allied with inherited Spiroplasma in the Ixodetis group (see Additional file 4) [GenBank:KJ535740], (99.6%). The remaining OTUs were Ralstonia.

Common heritable endosymbiont infections in Megastigmus

We found three sexual Megastigmus species infected with Rickettsia, and two of these same species infected with Wolbachia. None of the species was infected with Arsenophonus, Spiroplasma, or Cardinium. From this patchy distribution (i.e. high prevalence in some host populations and low prevalence or absence in others), we can likely conclude that none of these inherited symbionts are essential in host nutrition and/or manipulation.

It is not surprising that Wolbachia was detected, as it is the most common intracellular bacterial symbiont of insects [67]. Wolbachia are transmitted maternally, in the egg cytoplasm, and many strains have evolved strategies to increase the frequency of infected female hosts in the population. Reproductive manipulating strains of Wolbachia have been show to either cause cytoplasmic incompatibility or distort sex ratios by killing males or inducing parthenogenetic reproduction (i.e. clonal production of females) or feminization [68]. Parthenogenesis-inducing Wolbachia are common in Hymenoptera and have been characterized in several parasitoid [69] and cynipid gall wasps [70],[71]. A recent study implicated Wolbachia in parthenogenetic reproduction in Megastigmus, with 10/10 asexual species infected [64]. Treating M. pinsapinis with the antibiotic tetracycline restored the production of males, strongly suggesting that Wolbachia is the causative agent of thelytoky in asexual Megastigmus. No sexual Megastigmus species were infected with Wolbachia in the Boivin et al. study [64]; however, we found infections in M. amicorum and M. bipunctatus. The Wolbachia strains that we identified from sexual Megastigmus are closely allied with those in asexual Megastigmus. It would be interesting to determine if parthenogenesis-induction in Megastigmus is due to the host or the particular Wolbachia strain.

Rickettsia infections were discovered in three species. Bacteria in the genus Rickettsia are well known for being insect-vectored vertebrate pathogens, such as the causal agents of Rocky Mountain spotted fever (R. rickettsiae) and typhus (R. typhi). However, recent surveys have uncovered many Rickettsia that are vertically transmitted symbionts of diverse arthropods, most of which do not feed on vertebrates [72]. Some Rickettsia symbionts have been shown to distort host sex ratios via male-killing [73] or parthenogenesis-induction [74]. The presence of Rickettsia and Wolbachia in males likely rules out sex ratio distortion in our study. Alternatively, facultative symbionts may benefit their hosts under some circumstances. For example, some Wolbachia and Rickettsia increase host fitness by providing protection against natural enemies [75],[76].

Phylogenetic analysis shows that closely related Rickettsia and Wolbachia infect distantly related Megastigmus (Figures1 and 2). This provides strong evidence of horizontal transmission over evolutionary timescales, and is a common pattern in facultative inherited symbionts of insects [14]. In most cases, it is not known how inherited symbionts colonize novel hosts; shared hosts and shared natural enemies have both been implicated [77]-[80]. Interestingly, for some inherited symbionts, horizontal transmission over ecological timescales may be quite common [19],[81]. It would be useful to sequence more rapidly evolving Rickettsia genes, to determine if there was very recent transmission between M. amicorum and M. bipunctatus. Since both these species develop in junipers, we could speculate that horizontal transmission occurs via shared host plants; the Boivin et al. study of Wolbachia in asexual Megastigmus also found evidence for such host-plant-mediated transmission [64]. Plant-mediated transmission may be an important and underappreciated way for symbionts to colonize hosts. Indeed, a recent study showed that an inherited Rickettsia in the sweet potato whitefly can be transmitted via phloem [61]. Two strains of Arsenophonus that infect planthoppers are transmitted both transovarially and via plants, and both have been implicated in plant disease [82],[83]. However, as far as we are aware, interspecific transmission via plants has not yet been demonstrated in any inherited symbionts.

Microbial associates of M. spermotrophus

Our estimate of M. spermotrophus microbial species richness (60 ± 13 OTUs) fell within the range of other studies of insect microbiomes. Pollenivorous and predacious Hymenoptera (bees and wasps) harbour distinct bacterial communities with the lowest level of species richness (11.0 ± 5.4 OTUs/sample), while termites harbour the highest species diversity (89.5 ± 61.2 OTUs/sample), based on a recent meta-analysis [84]. A recent study estimated the diversity of bacteria associated with parasitoid wasps from the genus Nasonia ranged from 14 to 38 bacterial OTUs [85]. Pyrosequencing has been show to detect a greater number of OTUs compared to traditional methods, such as 16S rRNA clone sequencing [86]. This might explain why the estimated bacterial diversity associated with M. spermotrophus is comparably high because the Nasonia study and many previous insect microbiome surveys were done using 16S rRNA clone sequencing.

Despite a relatively high overall richness, only fifteen major OTUs are present with a total relative abundance of 0.5% or greater. The core bacterial community of M. spermotrophus can thus be considered to have a somewhat low diversity, characterized by bacterial OTUs that are commonly found associated with insect guts. The major OTUs associated with M. spermotrophus can be grouped into five distinct phylotypes: Betaproteobacteria (mostly Ralstonia), Gammaproteobacteria (mostly Acinetobacter), Actinobacteria (Corynebacterium), Firmicutes (mostly Anaerococcus) and Alphaproteobacteria (family Bradyrhizobiales). Most of these OTUs are related to bacteria that have been previously reported in insect guts, with Acinetobacter and Corynebacterium especially common (e.g. [85],[87]). All of the major OTUs identified below the order level are bacteria that commonly occur in the environment, such as in soil [88] and in the rhizospere [89]. Similar results are commonly found with microbial associates of insects. For example, the microbial symbionts of Tetraponera ants are closely related to nitrogen-fixing root nodule bacteria [40]. The giant mesquite bug, Thasus neocalifornicus acquires an important mutualistic gut symbiont de novo every generation from the soil [27]. The presence of the same major OTUs in M. spermotrophus in ovule and even Eurytoma samples provides clues to the distribution and transmission of the Megastigmus microbiome; it suggests that it is derived from the environment, which, for the developing wasp, is the ovule. Acinetobacter and Corynebacterium have been previously cultured from within surface-sterilized seeds and ovules [90]-[92].

The M. spermotrophus microbiome appears to be highly conserved across development, as demonstrated by the UniFrac analysis, with all of the samples tightly grouped. This contrasts with a recent survey of microbial associates of three Nasonia species that found that bacterial species richness increased with development [85]. Like most higher Hymenoptera, the larvae of M. spermotrophus have a blind digestive system with the midgut and hind gut only uniting during the last larval instar. Prior to pupation all of the built-up wastes are voided in a fecal pellet, termed the meconium [93]. During metamorphosis the larval midgut epithelium is discarded and replaced by a new pupal epithelium [94]. If these bacteria are associated with the gut, how M. spermotrophus maintains its major associates throughout development is not known. Some insects, like true bugs, termites and cockroaches, have crypts or paunches associated with the gut that are thought to enhance persistence of the microbiota [6]. This physiological feature is not well characterized in the Hymenoptera, with the exception of some ants [95].

A single OTU assigned to the genus Ralstonia comprised over 55% of all sequences from the M. spermotrophus samples. The high abundance and persistence of Ralstonia throughout host development is a strong indicator that this bacterium is an important associate of M. spermotrophus. Ralstonia was also found to be associated with Douglas-fir ovules and the parasitoid Eurytoma. The genus Ralstonia contains species from ecological diverse niches, such as the plant pathogen R. solanacearum, the opportunistic human pathogen R. pickettii and the environmental isolate R. eurytropha[96]. A maximum likelihood phylogeny placed M. spermotrophus associated Ralstonia in a cluster with the human pathogen R. pickettii (Figure4). To our knowledge, this is the first report of Ralstonia being a very abundant and potentially important component of an insect microbiome, although Ralstonia spp. have been previously reported from microbial surveys of insects, including the cotton bollworm (not published; accession # EU124821), Bartonella-positive fleas [97], an omnivorous carabid beetle [98] and the Potato Psyllid (as well as the faucet water used to water the potato plants) [99]. Recently, Husnik et al. also report the horizontal transfer of one Ralstonia gene into the genome of the mealybug Planococcus citri[100]. Also, R. oxalatica was isolated from the alimentary canal of an Indian earthworm [101].

A recent meta-analysis of 16S clone-library studies of insect associated microbes found that Betaproteobacteria contributed over 50% to all sequences from Hymenoptera [84]. The most common bacterial phylotype identified from solitary bee species, was a Betaproteobacteria from the genus Burkholderia[35], which is closely related to Ralstonia. Burkholderia spp. have also been identified as important mutualists of some phytophagous true bugs (suborder Heteroptera), where they reside in gut crypts [26]-[28],[102],[103].

The developing M. spermotrophus larva feeds on megagametophyte tissue, which contains all of the seed storage reserves, primarily in the form of starch, triacylglycerols, and nitrogen rich proteins [104],[105]. Therefore, Ralstonia and other microbial associates of M. spermotrophus would not likely play a role in supplementing this already rich diet with missing essential nutrients but instead may play a role in nutrient recycling. Parasitism by M. spermotrophus results in the formation of a nutrient sink, in which the larva and associated microbes are nourished by storage reserves of the megagametophyte. The reserves are intended to provide nourishment for the developing seedling or to be re-absorbed by the mother plant in the event of megagametophyte abortion. In loblolly pine, more than half of the nitrogen in megagametophytes comes from the amino acid arginine [106]. Insects use the enzyme arginase to hydrolyze arginine into ornithine and urea [107]. Excretion of urea would result in the substantial loss of nitrogen, especially since larvae must undergo extended periods of diapause. Very few insects are known to produce urease, the enzyme required to convert urea into ammonium for subsequent amino acid biosynthesis [108]. We speculate that Ralstonia or other microbial associates of M. spermotrophus might play an important role in nitrogen recycling by producing urease or other key enzymes missing from the host genome. Many insect symbionts have been suggested to promote increased availability of nitrogen in a variety of ways [5]. For example, Blochmannia and Blattabacterium, the obligate nutritional symbionts of carpenter ants and cockroaches, respectively, use ureases to recycle nitrogen from urea [109],[110]. Nitrogen recycling by symbionts has also been shown to be important during diapause in the shield bug, Parastrachia japonensi[111].

It is also tempting to speculate that Ralstonia could potentially play a role in plant manipulation. Another Ralstonia species, R. taiwanensi, has been shown to be capable of nodulating and fixing nitrogen in Mimosa spp. [112], which implies an ability to manipulate plant physiology. Alternatively, Ralstonia may not be a key associate of Megastigmus species in general, but rather a microbe that is found in the seed environment that encodes enzymes required for the catabolism of seed storage molecules or other essential pathways required for the seed feeding lifestyle of M. spermotrophus.

Now that Ralstonia has been identified as a likely symbiont of M. spermotrophus, further targeted surveys using Ralstonia-specific PCR primers would be helpful in determining its prevalence in other populations of M. spermotrophus, in other Megastigmus species, and in associated plants. The development of strain-specific markers for fluorescence in situ hybridization would also be useful for localizing Ralstonia on or within M. spermotrophus and the ovule, and following its transmission throughout its life cycle. It would also be interesting to examine Ralstonias role in nitrogen recycling, for example by identifying and following the expression of ureases and other key enzymes during M. spermotrophus development.

Insect samples

Several species of Megastigmus and their parasitoids were screened for common heritable symbionts using PCR. Adult insects were reared from seeds that were collected from forest stands in France, Greece, Denmark and Turkey from 1997 to 2011; detailed information on sample species is listed in Table1. Also, larvae of M. spermotrophus were dissected from infested seed collected in 2011 from seed orchards located throughout British Columbia. Adult M. spermotrophus were reared from this same seed. Any Eurytoma sp. parasitoids that emerged were also collected. Wild adult female M. spermotrophus were collected from trees located on the University of Victoria campus in Victoria, BC. Whole insect samples were stored in 95% ethanol at −20°C until DNA extraction.

For 16S rRNA bacterial amplicon pyrosequencing, M. spermotrophus and their parasitoids were obtained in 2011 from heavily infested seed from the Mt. Newton Seed Orchard, located in Saanichton, BC. The seeds were placed at room temperature to hasten the development of larvae and adult emergence. Larvae as well as approximately one-week-old pupae were extracted from surface-sterilized seeds. Adult female M. spermotrophus and adult Eurytoma sp. were collected upon emergence about two and three weeks later, respectively. Samples of uninfested ovules were also collected from surface-sterilized seeds.

DNA extraction

Whole insects were rinsed several times with sterile water and allowed to air dry. The samples were then placed individually into 2mL Micro tubes (Sarstedt) with 100 μL of PrepMan Ultra Reagent (Applied Biosystems, USA) and approximately twenty 1.0mm diameter zirconia or silica beads (BioSpec Products). Samples were homogenized using the Mini-Beadbeater-16 (BioSpec Products) on maximum (3450 oscillations/min) for two 2030 second cycles separated by 30seconds of centrifugation at 13,000 g. The samples were then incubated at 100C for ten minutes, then cooled to room temperature for one minute, then centrifuged for three minutes at 13,000 g and transferred into new Eppendorf tubes. DNA samples used for pyrosequencing were purified by precipitation in cold isopropanol and then washed with 70% ethanol and re-suspended in TE buffer (pH = 7.5). A NanoDrop 2000 Spectrophotometer (Thermo Scientific) was used to determine the DNA concentration and quality. The quality of the DNA extract was also checked by successful PCR amplification of the mitochondrial cytochrome oxidase subunit I (COI) gene using standard primers for invertebrates (see Additional file 5). All DNA extracts were stored at ?20C.

Directed PCR

Directed PCRs were conducted using either Invitrogen or ABM PCR Taq and reagents. Symbiont-specific primer-pairs were used to screen the samples for the presence of common heritable symbionts (see Additional file 5) with the following infected insects used as positive controls: Drosophila neotestacea (Wolbachia and Spiroplasma positive), Macrosteles quadrilineatus (Arsenophonus and Cardinium positive), and Ctenocephalides felis (Rickettsia positive). Sterile water was used as a negative control. Positive PCR products were separated on 1% agarose gel, stained with eithidium bromide and visualized under UV light.

Five microlitres of DNA from each individual extraction within a sample subset (individuals of the same species, sample type, location and year) were pooled (total of 32 pooled samples) and then screened using each primer set. If a positive PCR product was amplified from a pooled sample then each individual sample was screened for presence or absence of the corresponding symbiont using the same primer set. Positive PCR products were validated by sequencing representative amplicons in both directions. Purification and sequencing of PCR products were completed at Macrogen USA (Maryland). Forward and reverse sequences were aligned using MUSCLE and manually edited using the software Geneious (v6.1.3) (Biomatters) to create high-quality consensus sequences. A portion of mitochondrial COI was sequenced from one representative female of every symbiont-positive population, using Megastigmus-specific primers (see Additional file 5) and compared with other Megastigmus sequences deposited in GenBank. Percent divergence between COI sequences from M. amicorum populations was calculated using MEGA 5.1 [113].

Phylogenetic analysis of Rickettsia and Wolbachia infecting Megastigmus

A number of additional symbiont genes were amplified via PCR and sequenced: citrate synthase gene (gltA) for Rickettsia, and coxA, and gatB for Wolbachia (see Additional file 5). Phylogenies were re-constructed using sequences generated in this study and a sample of sequences obtained from GenBank. For Wolbachia, a sample of sequences obtained from an independent study of Wolbachia in parthenogenetic Megastigmus was also included [64]. Sequences were aligned using ClustalW, visually inspected and trimmed when necessary. A maximum-likelihood tree was generated using the Tamura 3-parameter model plus gamma distributed rates among sites (best substitution model identified by MEGA), with MEGA 5.1 [113], bootstrapped 500 times.

Bacterial tag-encoded FLX amplicon pyrosequencing

Three replicates of five sample types were submitted for bacterial tag-encoded FLX 454-pyrosequencing (bTEFAP): M. spermotrophus larvae, pupae and adult females, Eurytoma sp. adults and P. menziesii ovules. Although the 27F/519R primer set is not ideal for characterizing bacterial 16S rRNA sequence from plant tissue due to chloroplast DNA contamination [114],[115], we included ovule samples in order to see if any trace endophytic bacteria could be found after post-sequencing removal of plastid sequences. Inhibitor removal and bTEFAP were completed by MR. DNA Laboratories (Shallowater, TX). Inhibitor removal involved the use of the PowerClean DNA Clean-up kit (MO BIO Laboratories, Inc., Carlsbad, CA) according to the manufacturers protocol. The methods used for bTEFAP are previously described in Palavesam et al. (2012) and Shange et al. (2012) [116],[117] and were originally described by Dowd et al. (2008) [118]. Briefly, a single-step PCR was done using the following temperature profile: 94C for 3minutes, followed by 28cycles of 94C for 30seconds, 53C for 40seconds and 72C for 1minute, with a final elongation step at 72C for 5minutes using HotStarTaq Plus Master Mix Kit (Qiagen, Valencia, CA). The 16S universal bacterial primers 27Fmod (5-AGRGTTTGATCMTGGCTCAG-3) and 519Rmodbio (5-GTNTTACNGCGGCKGCTG-3) were used to amplify a 500bp region of the 16S rRNA gene spanning the V1-V3 regions. The PCR products from each of the different samples were mixed in equal concentrations and then purified using Agencourt Ampure beads (Agencourt Bioscience Corporation, MA, USA). Following the manufacturers guidelines, sequencing was conducted using the Roche 454 FLX titanium platform (Roche, Indianapolis, IN).

Qiime pipeline

The 454 generated Standard Format Flowgram (SFF) file was converted into a SFF text file using Mothur (v1.23.0) [119]. The open source software package Quantitative Insights Into Microbial Ecology (QIIME v1.6.0) was used to process the sequence data [120]. The raw sequencing data was filtered using the following parameters: minimum sequence length of 100bp, maximum sequence length of 2,000bp and maximum homopolymer region of eight. Also, any sequences with an average quality score below 25 or any ambiguous bases were discarded. This filtering step reduced the number of total sequences from 81,207 to 60,543. The 454 data were then denoised to reduce the number of erroneous OTUs [121]. Chimera detection was done independently of QIIME by implementing UCHIME through the USEARCH (v6.0.307) program [122]. The sequences were compared against the Gold database (http://www.drive5.com/usearch/manual/otupipe.html, downloaded February 13, 2013). Chimeric sequences (1,190 or 1.97%) were gleaned from the data set.

OTUs were picked with the UCLUST method with the optimal option indicated. Similar sequences were clustered at the default level of 0.97 [123]. Taxonomy was assigned to representative sequences using the RDP Classifier 2.2 method at the 0.9 confidence level [124]. Taxonomies were based on the Greengenes database (ftp://greengenes.microbio.me/greengenes_release/gg_12_10/, downloaded February 1, 2013) [125],[126].

Originally, the PyNast method was used to align the representative sequences to a pre-aligned database; however, this method resulted in poor overall alignment. Alternatively, representative sequences were aligned to a Stockholm format reference of pre-aligned sequences and secondary structures using Infernal [127]. The aligned sequences were filtered to remove common gap positions, with the gap filter threshold set to 0.8 and the entropy threshold set to 0.10. An approximately-maximum-likelihood phylogenetic tree was created using FastTree 2.1.3 [128]. An OTU table in Biom format was created and then split at the highest taxonomic ranking to remove unclassified OTUs (likely remnant chimeric sequences). Singletons were removed from the Biom table. Alpha diversity results were generated using a rarefaction depth of 5,000. In order to identify possible outliers (i.e., samples that contain unusual or unexpected OTUs), the microbiome data were visualized using a correspondence analysis biplot [129]. One pupal sample (P1) and one female sample (F4) were found to be associated with distinct OTUs that did not cluster with the remaining samples. Sample P1 had a relatively elevated species richness compared to the other samples, likely originating from environmental contamination (data not shown). Sample F4 contained bacteria typical of human contamination. Subsequently these two samples were removed from further analysis.

Data exploration, visualization and analyses were performed in R (v3.0.1) [130] on RStudio (v0.97.336) (http://www.rstudio.com, downloaded August 5, 2013), mainly using the Phyloseq R-package (v1.5.19) [131]. Data were rarefied to an equal sampling depth of 1,962 prior to community analysis. Initial correspondence analysis and biplots were generated using the Ade4 R-package (v1.5-2) [132]. Principle component analysis was completed using unweighted and weighted UniFrac distances [133],[134].

In order to obtain longer 16S rRNA fragments for phylogenetic analysis from the Spiroplasma strain infecting Eurytoma, general 16S rRNA amplicons were generated using the primers 63F (5-CAGGCCTAACACATGCAAGTC-3) [135] and 907R (5-CCGTCAATTCCTTTRAGTTT-3) [136]. Amplicons were then cloned using the Strataclone kit with Solopack Competent cells (Stratagene). Transformation was validated with PCR using M13F (5- CACGACGTTGTAAAACGAC-3) and M13R (5-GGATAACAATTTCACACAGG-3). Eight clones were sent for sequencing and one representative Spiroplasma 16S rRNA sequence was used for further analysis. Attempts to clone longer Ralstonia 16S rRNA fragments were not successful.

Ralstonia sequence from the most abundant OTU in the pyrosequencing data was used to generate a 16S rRNA phylogeny, along with representative Ralstonia species and outgroup sequences, obtained from GenBank. Maximum likelihood analysis was performed as above, except using the Tamura-Nei model with invariant sites and gamma rate distribution among sites.