Bacteriophage-Host Association in the Phytoplasma Insect Vector Euscelidius variegatus

Insect vectors transmit viruses and bacteria that can cause severe diseases in plants and economic losses due to a decrease in crop production. Insect vectors, like all other organisms, are colonized by a community of various microorganisms, which can influence their physiology, ecology, evolution, and also their competence as vectors. The important ecological meaning of bacteriophages in various ecosystems and their role in microbial communities has emerged in the past decade. However, only a few phages have been described so far in insect microbiomes. The leafhopper Euscelidius variegatus is a laboratory vector of the phytoplasma causing Flavescence dorée, a severe grapevine disease that threatens viticulture in Europe. Here, the presence of a temperate bacteriophage in E. variegatus (named Euscelidius variegatus phage 1, EVP-1) was revealed through both insect transcriptome analyses and electron microscopic observations. The bacterial host was isolated in axenic culture and identified as the bacterial endosymbiont of E. variegatus (BEV), recently assigned to the genus Candidatus Symbiopectobacterium. BEV harbors multiple prophages that become active in culture, suggesting that different environments can trigger different mechanisms, finely regulating the interactions among phages. Understanding the complex relationships within insect vector microbiomes may help in revealing possible microbe influences on pathogen transmission, and it is a crucial step toward innovative sustainable strategies for disease management in agriculture.


Introduction
Like all other organisms, insects harbor a rich, dynamic, and interactive community of microorganisms, collectively known as the microbiome, which comprises not only living members (microbiota), but also elements considered as not living organisms (viruses, plasmids, prions, viroids, and free DNA) and a whole spectrum of molecules produced by the microorganisms [1]. Both microbiome composition and its modification influence insect ecology, physiology, evolution, and behavior through genetic and metabolic interactions. In many cases, an insect's survival depends on its microbiome composition [2]. For these reasons, both pest control and insect protection may take advantage by deciphering the relationships between insects and their microbiome as well as among microbiome components [3].
In the past decade, many studies have been devoted to characterizing the microbiome of insects, essentially through next-generation sequencing approaches. Even though bacteria, fungi, protozoa, and viruses may be associated with their insect host permanently or transiently, the vast majority of these works have focused on bacterial communities [2,4].

Bacteriophage-Like Particles in Euscelidius variegatus
During the electron microscope observation of a partial viral purification obtained from a Euscelidius variegatus Torino (EvaTO) population [36], aimed at revealing insect virus particles, bacteriophage-like particles were serendipitously observed. Those phages Pathogens 2021, 10, 612 3 of 16 had head-and-tail morphology typical of the viral order Caudovirales and in particular of the Siphoviridae family; they had a prolate (elongated) oval head 132 nm long (SE = 0.7 nm; n = 48) and 59 nm wide (SE = 0.3 nm; n = 48) and a flexuous thin non contractile tail 179 nm long (SE = 4.4 nm; n = 16) (Figure 1a). Frequently, heads and tails appeared detached (not shown). from a Euscelidius variegatus Torino (EvaTO) population [36 virus particles, bacteriophage-like particles were serendipitou had head-and-tail morphology typical of the viral order Cau the Siphoviridae family; they had a prolate (elongated) oval he n = 48) and 59 nm wide (SE = 0.3 nm; n = 48) and a flexuous nm long (SE = 4.4 nm; n = 16) (Figure 1a). Frequently, heads (not shown).

Selection of Expressed Bacteriophage Sequences
In an effort to identify the siphovirus-like phage observ from EvaTO adults was used to construct two cDNA libraries merged and depleted of reads matching to the insect transcri de novo metatranscriptome assembly of E. variegatus microbio assembled sequences revealed the presence of 12 expressed se as hallmark bacteriophage structural genes, i.e., major caps

Selection of Expressed Bacteriophage Sequences
In an effort to identify the siphovirus-like phage observed by TEM, RNA extracted from EvaTO adults was used to construct two cDNA libraries. The two datasets were then merged and depleted of reads matching to the insect transcriptome in order to generate a de novo metatranscriptome assembly of E. variegatus microbiome. A blastx analysis of the assembled sequences revealed the presence of 12 expressed sequences that were identified as hallmark bacteriophage structural genes, i.e., major capsid, minor capsid, baseplate, major tail, minor tail, portal, tail fiber, tail sheath, collar, and head-tail joining proteins. After a further annotation with the ViPTree server [37], 5 out of 12 sequences appeared to be polycistronic mRNAs (Table 1), i.e., they encoded at least two putative proteins. The selected transcripts included three different complete major capsid/head proteins, suggesting that the selected expressed sequences belonged to at least three different phages, thereafter referred to as EVP-1 (Euscelidius variegatus phage 1, associated with MW965291 ORF3), EVP-2 (associated with MW965292) and EVP-3 (associated with MW965289 ORF2).

Phylogenetic Analysis of the Identified Major Capsid Proteins
The deduced amino acid sequences of the three major capsid proteins were aligned to the first hit retrieved by blastx against the NCBI nr database and the first ten hits identified by blastp against the NCBI RefSeq protein limited to the taxon "Viruses (taxid:10239)". We considered for the analysis only complete major capsid proteins from phages assigned to one of the nine Caudovirales families recognized by the International Committee on Taxonomy of Viruses (ICTV). The phylogenetic analysis ( Figure 2) showed that EVP-1 was part of a completely separate cluster from the one that included the other two identified major capsid proteins. Such a cluster was formed only by major capsid proteins from members of the family Siphoviridae. The other cluster was far more heterogeneous, including members of both the families Siphoviridae and Myoviridae. EVP-2 and a protein identified in Serratia marcescens (best blastx hit) formed a separate branch that diverged from a clade including both siphoviruses and myoviruses. Finally, EVP-3 formed a strongly supported clade (100% bootstrap value) with major capsid proteins from Candidatus Symbiopectobacterium sp. Dall1.0 and from a member of the family Myoviridae.

Detection and Prevalence of the Three Phages
Specific primers designed on the three genes (Table 2), coding for the identified major capsid proteins, were used on DNA extracted from insect whole bodies and a partial viral purification. All the three primer pairs produced amplicons of the expected size and sequence from 20 individuals randomly selected from the EvaTO population (data not shown). By contrast, only primers designed on EVP-1 major capsid protein gave an amplification product from the DNA extracted from viral particles ( Figure 1b). Therefore, we could reasonably associate the phage observed by TEM to EVP-1.

Isolation of EVP-1 Bacterial Host
In an attempt to identify the EVP-1 bacterial host, EvaTO hemolymph was cultivated on chocolate agar and purple agar plates, yielding several colonies ( Figure S1). Few colonies appeared after one/three days of incubation and showed a fast growth. After 7-10 days, many other whitish small colonies (almost forming a layer) appeared; they looked all similar and grew slowly ( Figure S1a,b). The fast growing (1C-3C, 5C-7C, 1P, 2P) and some of the slow growing (4C, 8-15C, 3-5P) colonies were isolated and sub-cultivated on both chocolate and purple agar, irrespective of their original selection medium. Ten out of the 20 isolated colonies were positive to the PCR amplification with EVP-1 primers ( Figure S1b). All the fast-growing colonies were negative; almost all the slow growing colonies (except 10C and 11C) were positive and they were considered as EVP-1 bacterial hosts. The growth of EVP-1 positive colonies on chocolate and purple agar was scanty, with low viability. Attempts at subculturing these colonies on Tryptic soy agar (TSA) resulted in faster growth, with similar viability (about two weeks). Three colonies were selected and maintained on TSA for further investigations: 4C, 12C, and 14C. At each subculture step, the presence of EVP-1 was checked by PCR with EvaTOphage1 primers (not shown). The 16S ribosomal sequences of the three colonies were 100% identical to each other and 99% identical (95% query coverage) to the one isolated from the "Bacterial parasite of Euscelidius variegatus" (BEV, GenBank accession number: Z14096 [38]) ( Figure S2), which has been recently renamed Candidatus Symbiopectobacterium [35]. Based on this result, primers BEV3/BEV4 [39], specifically designed on the BEV 16S ribosomal sequence, were used hereinafter. In particular, they were used to exclude the presence of bacterial host DNA after DNA extractions from viral particles (Figure 1b).

Detection and Prevalence of the Three Phages
Specific primers designed on the three genes (Table 2), coding for the identified major capsid proteins, were used on DNA extracted from insect whole bodies and a partial viral purification. All the three primer pairs produced amplicons of the expected size and sequence from 20 individuals randomly selected from the EvaTO population (data not shown). By contrast, only primers designed on EVP-1 major capsid protein gave an amplification product from the DNA extracted from viral particles (Figure 1b). Therefore, we could reasonably associate the phage observed by TEM to EVP-1. Previous literature has reported that BEV can colonize multiple insect host organs, including the midgut [40,41]. PCR and RT-PCR experiments on E. variegatus dissected guts not only confirmed the presence of EVP-1 but also revealed that the phage was transcriptionally active in this tissue (Figure 3a). In addition, TEM observations of negatively stained crude extract from guts demonstrated the presence of siphovirus-like phage particles similar to those observed in the viral purification from the whole insect (Figure 3b).

Multiple Phages in EVP-1 Host
According to a preliminary survey of BEV genome size and content [42], there was evidence for an extrachromosomal element that could represent a prophage and 65 partial coding sequences that could be ascribed to phages. A blastn analysis against all the BEV sequences submitted to the GenBank Trace Archive revealed that 11 out of the 12 putative phage transcripts identified in this work found a hit with percentages of identity >88%, Pathogens 2021, 10, 612 7 of 16 but with low percentages of query coverage (average coverage < 41%) (Table S1). Such result could be explained by the fact that some deposited sequences are shorter than the identified transcripts and all of them contain ambiguous nucleotides (Ns), which in the case of gnl|ti|2292005461 represented more than 50% of its length. Nevertheless, those results suggested that most of the identified sequences could be assigned to phages that have BEV as their bacterial host. A colony PCR confirmed that EVP-2 and EVP-3 were also likely to be associated with BEV (Figure 4e). each other and 99% identical (95% query coverage) to the one isolated from the "Bacterial parasite of Euscelidius variegatus" (BEV, GenBank accession number: Z14096 [38]) ( Figure  S2), which has been recently renamed Candidatus Symbiopectobacterium [35]. Based on this result, primers BEV3/BEV4 [39], specifically designed on the BEV 16S ribosomal sequence, were used hereinafter. In particular, they were used to exclude the presence of bacterial host DNA after DNA extractions from viral particles (Figure 1b).
Previous literature has reported that BEV can colonize multiple insect host organs, including the midgut [40,41]. PCR and RT-PCR experiments on E. variegatus dissected guts not only confirmed the presence of EVP-1 but also revealed that the phage was transcriptionally active in this tissue (Figure 3a). In addition, TEM observations of negatively stained crude extract from guts demonstrated the presence of siphovirus-like phage particles similar to those observed in the viral purification from the whole insect (Figure 3b).

Multiple Phages in EVP-1 Host
According to a preliminary survey of BEV genome size and content [42], there was evidence for an extrachromosomal element that could represent a prophage and 65 partial coding sequences that could be ascribed to phages. A blastn analysis against all the BEV sequences submitted to the GenBank Trace Archive revealed that 11 out of the 12 putative phage transcripts identified in this work found a hit with percentages of identity >88%, but with low percentages of query coverage (average coverage <41%) (Table S1). Such result could be explained by the fact that some deposited sequences are shorter than the identified transcripts and all of them contain ambiguous nucleotides (Ns), which in the case of gnl|ti|2292005461 represented more than 50% of its length. Nevertheless, those results suggested that most of the identified sequences could be assigned to phages that have BEV as their bacterial host. A colony PCR confirmed that EVP-2 and EVP-3 were also likely to be associated with BEV (Figure 4e).
Electron microscopy observation of negative-stained bacteria from TSA cultures revealed rods from 2.1 to 5.3 μm long (with a mean value of 3.1 μm, n = 60), from 0.60 to 0.79 μm wide (with a mean value of 0.68 μm, n = 60), without flagella and sometimes curved (Figure 4a). In the proximity of bacterial cells, some bacteriophage particles were noted; some were siphovirus-like and similar to those observed in the insect extract ( Figure 4c). Moreover and unexpectedly, myovirus-like ( Figure 4b) and podovirus-like ( Figure 4d) particles were also present. In particular, the number of podovirus-like particles were approximately 100-times higher than the number of siphovirus-and myovirus-like ones ( Figure 5).  Electron microscopy observation of negative-stained bacteria from TSA cultures revealed rods from 2.1 to 5.3 µm long (with a mean value of 3.1 µm, n = 60), from 0.60 to 0.79 µm wide (with a mean value of 0.68 µm, n = 60), without flagella and sometimes curved (Figure 4a). In the proximity of bacterial cells, some bacteriophage particles were noted; some were siphovirus-like and similar to those observed in the insect extract (Figure 4c). Moreover and unexpectedly, myovirus-like (Figure 4b) and podovirus-like (Figure 4d) particles were also present. In particular, the number of podovirus-like particles were approximately 100-times higher than the number of siphovirus-and myovirus-like ones ( Figure 5).  a and b, asterisks), a siphovirus-like particle without tail (a, #), a myovirus-like particle with contracted tail (b, @). Bars represent 200 nm.  a,b, asterisks), a siphovirus-like particle without tail (a, #), a myovirus-like particle with contracted tail (b, @). Bars represent 200 nm.

Discussion
The bacterial colony was subjected to an enrichment of viral particles and a DNase treatment prior to DNA extraction to ensure the purification of encapsidated DNA only. Given the different phage morphologies observed by TEM, the three phage primer pairs were used in PCR. Amplicons of the expected size were obtained with specific primers for EVP-1 and EVP-3, whereas a faint band was observed with EVP-2 primers (Figure 4e). BEV primers failed to amplify bacterial DNA, so we could exclude the presence of contaminating phage DNA integrated into the host chromosome. PCR results confirmed that the siphovirus-like phage observed by TEM was EVP-1 and suggested that the myovirus-like particles could be associated with EVP-3.

Discussion
Phage "omics" studies are represented mostly by metagenomic shotgun analyses applied to a wide range of environments and locations, from uncultured marine samples [43][44][45] to human gut microbiome [44,[46][47][48][49]. Differential filtrations and densitydependent gradient centrifugations are the usual enrichment steps taken to concentrate viral DNA and limit bacterial constituents and other contaminants before sequencing.
While metagenomics is the most rapid and efficient approach used to describe the overwhelming diversity of phages, transcriptomics has been generally used to investigate the transcriptional response of bacteria isolated in pure cultures upon phage infection [50][51][52]. Only a few untargeted metatranscriptomic studies, which explored mainly soil microbial communities, reported the discovery of novel RNA phages [53,54] and phage-related mRNA sequences [55,56].
Metatranscriptomics has here been integrated with classical microbiological and microscopy techniques to identify a DNA tailed-phage and its bacterial host within the microbiome of a phytoplasma insect vector. To this end, we re-analyzed two RNA-seq libraries that were originally constructed without any prior phage enrichment step to explore E. variegatus transcriptome [57]. De novo identification of phage sequences can be an extremely challenging task, especially from a background of genes expressed by the insect host and all the active microorganisms that constitute its microbiome. Nevertheless, the stringent selection of phage-hallmark genes resulted in a reliable identification of the observed phage. Such a bioinformatic approach was clearly a non-exhaustive way to retrieve all the expressed phage sequences in the libraries. Some were probably overlooked due to the lack of similarity with known phage sequences (the so-called "dark matter") and/or the wrongful identification as bacterial genes in public databases. The absence of a biomarker gene among DNA phages and the polyphyletic origin of most viral lineages pose a hindrance for identifying all the putative phages. In any case, the characterization of the whole E. variegatus phageome was beyond the scope of this work, and the chosen approach proved effective in the detection of the transcribed phage genes associated with the observed phage particles.
Although RNA-seq data provided information about the active fraction of phages within the insect microbiome, they were not sufficient to apply the new computational approaches developed to predict phage-host relationships [58,59]. Therefore, classical microbiological techniques were applied for the unambiguous phage-host identification. All the bacterial hosts of the phages identified by the BLAST analyses belonged to the order Endobacterales, so it was reasonable to hypothesize that bacteria of the same order could be part of the E. variegatus microbiome and hosted the identified phages. Only two cultivable bacterial endosymbionts, BEV (Endobacterales [40]) and Asaia sp. (Rhodospirillales [33,60]), are known to be facultatively associated with E. variegatus. Because comprehensive studies of the bacterial fraction of its microbiome had never been undertaken, we did not have any a priori knowledge about either the possibility of isolation in axenic culture of the other endosymbionts or the most suitable media for their cultivation. The chosen growth media, usually adopted to identify enteric bacteria or to isolate fastidious bacteria, were effective in isolating the EVP-1 host. Interestingly, the EVP-1 host was shown to be the already known E. variegatus endosymbiont BEV, i.e., Candidatus Symbiopectobacterium.
Moreover, colony-PCR proved that BEV was also the host of the other two identified major capsid protein-coding genes. Interestingly, the best blastx hits of the EVP-3 major capsid protein was the one identified in Candidatus Symbiopectobacterium sp. Dall1.0 during a metagenomic study of Diachasma alloeum, a parasitoid of the apple maggot Rhagoletis pomonella [61]. Such result further supported the proposal made by Martinson et al. [35] that the genus "Candidatus Symbiopectobacterium" represents a monophyletic group of invertebrate host-associated microbes. With the huge production of whole genome sequencing data, it is known that most bacterial genomes carry multiple prophages, a phenomenon called polylysogeny [62]; however, inferred prophage sequences do not always correspond to active temperate phages [63]. The association of multiple active phages to BEV was confirmed by electron microscope observations and PCR reactions on the bacterial colonies. The intense phage lytic activity observed in vitro may explain the low viability of the bacterium in culture. It is likely that, after a few days in pure culture, most bacterial cells undergo phage lysis as a result of the massive production of podovirus-like particles. By contrast, neither viral particles nor transcripts that could be associated with podoviruses were identified in the insect microbiome. This may indicate that these are in the prophage status in the insect environment, while they switch to the lytic phase in plate culture condition. A transcript coding for a myovirus-like major capsid protein was retrieved during the RNA-seq analysis, even if myovirus-like particles were not observed in the viral partial purification obtained from the whole insect. Hence, we can suppose either that the phage is present at very low concentration in the insect body (below the detection limit) or that the particles were, for some reason, destroyed during the purification process. Nevertheless, on the one hand, some regulation mechanisms should operate in the insect to prevent the podovirus-like phage from producing the same massive release of viral particles observed in pure culture, which could potentially cause the lysis of the whole bacterial population. On the other hand, the growth of the BEV population within the insect could be kept under control by the action of EVP-1 and, maybe, EVP-3.
It is known that temperate phages have an important role in shaping microbial diversity and community structure; in fact, not only do they alter the biology of their hosts (i.e., regulating gene expression, introducing novel functions), but they also influence the surrounding hosts and non-host bacterial cells (i.e., entering the lytic cycle and killing susceptible bacteria, liberating intracellular contents used as nutrients by neighboring cells) [64,65]. Moreover, bacteriophage-bacteria interactions are considered by Refardt [66] as an ideal system for studying the competitive interactions within hosts. In particular, he considered the competition among phages in the same host, which is still an unexplored area in phage ecological research, and showed that multiple infection in Escherichia coli often resulted in a decreased lytic productivity.
At present, we do not know whether BEV active phages can influence the lysogenic status of the other prophages or whether they can infect other insect bacterial endosymbionts. These two phenomena deserve to be elucidated in view of using this system as a model both for among-hosts competition studies and for developing microbiome-based new plant pest control strategies. In fact, unravelling the microbiome of insect vectors and understanding the complex relationships within its components may help to reveal possible microbe influences on pathogen transmission, and it is a crucial step toward an innovative sustainable strategies for disease management in agriculture.

Insect Population
Euscelidius variegatus of the Torino (Italy) phytoplasma-free laboratory colonies (EvaTO) were originally collected in Piedmont (Italy) and reared on oat, Avena sativa (L.), inside plastic and nylon cages in growth chambers at 20-25 • C with a L16:D8 photoperiod.

DNA and RNA Extraction
DNA and total RNA were extracted from either whole bodies or dissected organs of emerged EvaTO adults, as described by Marzachì et al. [67] and Ottati et al. [68], respectively.
In order to distinguish the encapsidated phage DNA (lytic infection) from the phage DNA integrated into the host chromosome (lysogenic infection), we proceeded as follows. Bacterial colonies were resuspended in sterile water and filter-sterilized with a 0.22 µm filter. Such step should eliminate bacterial hosts and enrich the aqueous phase with phage particles. The suspension was then treated with TURBO™ DNase (Thermo Fisher Scientific, Waltham, MA, USA) for 1 h at 37 • C to digest any residual free DNA and subjected to DNA extraction with one volume of phenol:chloroform:isoamyl alcohol (25:24:1 v/v) followed by a wash step with one volume of chloroform:isoamyl alcohol (24:1 v/v) to remove any trace of phenol. Finally, DNA was precipitated with sodium acetate/ethanol, washed with 70% ethanol, and resuspended in 10 mM Tris-HCl pH 8.2. The same DNase treatment and DNA extraction protocol were also used on semipurified viral particles.

RNA-Seq and Bioinformatic Analysis
Six micrograms of total RNA were sent to Macrogen (Seoul, Korea) for cDNA library construction and sequencing, as detailed in [36]. At least 100 million 100-nt paired-end reads were obtained for each dataset. The two datasets were merged and the pre-assembly steps were performed using BBTools suite v38.70, as previously described [11]. BBMap, in particular, was used to remove reads mapping to the Euscelidius variegatus transcriptome shotgun assembly (TSA Accession number: GFTU00000000.1) before the assembling step with Trinity v2.9.1 [69]. The resulting sequences were further assembled by CAP3 v3 (overlap length cutoff = 60; overlap percent identity cutoff n = 90) [70]. As a result, around 120,000 assembled sequences were obtained and queried against the NCBI non redundant "nr" protein database (Last access February 2021) with DIAMOND v0.9.24.125 [71] using an E-value cut-off of 0.0001. Bowtie2 [72] was used to map reads against the putative phage transcripts with default parameters. Reads mapping onto the selected transcripts were expressed as RPK (Reads Per Kilobase of transcript).
ViPTree [37] was used to automatically generate the annotation of the selected transcripts. Given that the tool used the RefSeq release 93, the predicted coding sequences and the corresponding deduced proteins were analyzed using NCBI RefSeq Release 205 (February 2021).

Accession Numbers
Reads were deposited into the NCBI's Sequence Read Archive (SRA) database with BioSample accessions SAMN18744878 and SAMN18744879 as part of BioProject PRJNA393620. The partial 16S ribosomal RNA sequence of Candidatus Symbiopectobacterium strain EvaTO and the 12 phage transcripts were submitted to NCBI GenBank with accessions MW936016 (16S rRNA) and MW965281 MW965292 (phage sequences).

Phylogenetic Analysis
Phylogenetic relationships were inferred on the basis of the amino acid sequences of phage major capsid proteins. The three newly identified major capsid proteins were aligned with MUSCLE [73] to their best blastx hits and the first ten hits identified by blastp analysis against the NCBI RefSeq protein limited to the taxon "Viruses (taxid:10239)". Phylogenetic trees were then generated using the maximum likelihood (ML) approach, implemented in IQ-TREE [74] with default parameters through the CIPRES Science Gateway V. 3.3 [75]. Bootstrap analyses involving 1000 replicates were used with the Dayhoff substitution matrix to estimate the pairwise distances.

PCR and RT-PCR Amplifications
PCR was used to verify the presence of the three genes, encoding the major capsid proteins in EvaTO insects, phage DNA, and bacterial colonies. Universal 16S rRNA bacterial primers were first used, and then BEV3/BEV4 primers were used to identify the bacterial host harboring EVP-1 phage by colony PCR. BEV primers were also used to exclude the presence of residual bacterial host DNA whenever it was necessary to distinguish the encapsidated phage DNA from the phage DNA integrated into the bacterial chromosome.
For each sample, cDNA was synthesized from total RNA (500 ng) with EvaTO_phage1r primer using the RevertAid Reverse Transcriptase (Thermo Fisher Scientific, Waltham, MA, USA). The absence of contaminating genomic DNA was verified including, in the PCR step, samples without the reverse transcription step.
All primer sequences and amplification conditions used in this work are listed in Table 2. The resulting amplicons were validated by Sanger sequencing at BMR Genomics (Padua, Italy).

Bacterial Isolation
CO 2 -anesthetized leafhoppers were surface-sterilized by submerging them first in 95% ethanol for 1 to 2 min, then in 1.2 to 1.5% sodium hypochlorite solutions for 2 min and, eventually, rinsing them 2 or 3 times in sterile water [40]. Under a dissecting microscope, the hemolymph was aspirated with a fine, flame-drawn needle inserted into the insect body, between the thorax and the abdomen. The fluid was transferred to a tube containing 1X PBS. The hemolymph of five individuals was combined, then split in two and plated onto chocolate (Blood Agar Base-Sigma Aldrich-added with 7% horse defibrinated blood) and purple agar (Bromocresol Purple Broth-Sigma Aldrich-added with 1.5% agar) solid medium. The hemolymph of a total of 40 individuals was obtained. Plates were kept at 26 • C in the dark for up to 14 days. Colonies were numbered (using C when isolated from chocolate agar and P when isolated from purple agar plates) and transferred onto new chocolate and purple plates. For subcultures and maintenance, a Tryptic Soy Agar (TSA) medium (Sigma Aldrich) was eventually used.

Transmission Electron Microscopy
Viral particles were partially purified following the protocol previously described by [36]. Insect guts were collected under a dissecting microscope and crushed in 0.1 M phosphate buffer pH 7, added with 2% PVP, to obtain a crude extract. A portion of the bacterial colonies was collected with a toothpick from cultures on solid medium plates and suspended in 20 µL of liquid growth medium. A drop of the viral partial purification, the gut crude extract, or the bacterial suspension was deposited on carbon and formvar coated copper-palladium grids and left to stand for about 3 min. Grids were washed with water and negatively stained with 0.5% aqueous uranyl acetate. Observations and image acquisition were done using a CM 10 electron microscope (Philips, Eindhoven, The Netherlands) operating at 80kV. Micrograph films were developed and then digitally acquired at high resolution with a D800 Nikon camera. Images were trimmed and adjusted for brightness and contrast using GIMP 2 software. Particle measurements were done using Fiji software.