Comparative evaluation of the genomes of common bacterial members of the Drosophila intestinal community

Drosophila melanogaster is an excellent model to explore the molecular exchanges that occur between an animal intestine and their microbial passengers. For example, groundbreaking studies in flies uncovered a sophisticated web of host responses to intestinal bacteria. The outcomes of these responses define critical events in the host, such as the establishment of immune responses, access to nutrients, and the rate of larval development. Despite our steady march towards illuminating the host machinery that responds to bacterial presence in the gut, we know remarkably little about the microbial products that influence bacterial association with a fly host. To address this deficiency, we sequenced and characterized the genomes of three common Drosophila-associated microbes: Lactobacillus plantarum, Lactobacillus brevis and Acetobacter pasteurianus. In each case, we compared the genomes of Drosophila-associated strains to the genomes of strains isolated from alternative sources. This approach allowed us to identify molecular functions common to Drosophila-associated microbes, and, in the case of A. pasteurianus, to identify genes that are essential for association with the host. Of note, many of the gene products unique to fly-associated strains have established roles in the stabilization of host-microbe interactions. We believe that these data provide a valuable starting point for a more thorough examination of the microbial perspective on host-microbe relationships.


INTRODUCTION
Environmental, microbial, and host factors act at mucosal barriers to establish a unique microclimate that shapes the lives of all participant species (1). For example, expression of a host genotype in gastrointestinal tissues works in concert with extrinsic factors to determine microbial associations (2). For their part, the metabolic outputs of the gastrointestinal microflora influence critical events in the host such as education of immune phenotypes (3,4), development of intestinal structures (5), and access to essential micronutrients (6). Given the intertwined relationship between host phenotype and microbial genotype, it is of some surprise that hosts often tolerate extensive alterations to their microflora in response to environmental shifts such as changes in diet (7). However, alterations to the gastrointestinal microflora are not invariably without consequence, and intestinal dysbiosis may lead to chronic, debilitating, and occasionally deadly diseases within the host (8)(9)(10)(11)(12). Our appreciation of the holobiont as an intricate network of biochemical and genetic transactions between multiple participants mandates a thorough evaluation of the microbial genomes that shape host physiology.
Unfortunately, such studies are tremendously complex in conventional mammalian models due to the size of the microbiome, and the lack of laboratory techniques for the isolation and manipulation of most intestinal microbes.
The simple invertebrate Drosophila melanogaster is an excellent model holobiont (13,14). From a developmental perspective, the Drosophila posterior midgut shares a number of important similarities with the small intestine of more complex mammalian counterparts (15). Both organs are endodermal in origin, and are surrounded by a sheath of mesodermal visceral muscle (16,17). The mammalian small intestine and Drosophila posterior midgut are maintained by regularly spaced, basal intestinal stem cells that generate transitory progenitor cells (18)(19)(20); the non-proliferative enteroblasts of Drosophila, and the transient-amplifying cells of mammals. In both systems, signals along the Notch-Delta axis promote differentiation of transitory progenitors into secretory enteroendocrine cells or absorptive enterocytes (15,21). In contrast to mammals, Drosophila lacks specialized basal paneth cells for the release of antimicrobial peptides. Nonetheless, the fly genome encodes antimicrobial peptides that actively contribute to the control of intestinal symbionts and pathogens (22), suggesting release of such factors into the Drosophila intestinal lumen. In both the mammalian small intestine and Drosophila midgut, host factors and biogeography favor association with members of the Lactobacillaceae family (2,23). In return, metabolites from Lactobacilli activate host response pathways that promote intestinal stem cell proliferation (24). Combined with the genetic accessibility of flies and their suitability for longitudinal studies of large populations in carefully defined environments, Drosophila is an excellent system to decipher the forces that determine genetic interactions within a holobiont (13,25,26).
In contrast to conventional vertebrate models, the Drosophila microbiome consists of a small number of bacterial species that are easily isolated and cultured (27). The adult Drosophila intestine hosts little to no bacteria immediately after emergence from the pupal case, and the microfloral population grows in number over time (28). Several studies established that environmental factors and host genotype influence the diversity of the microflora (22,29,30). It is unclear if bacteria establish stable associations with the host gut, or if they cycle from the intestine to the environment and back (31). Nonetheless, lab-raised or wild Drosophila frequently associates with representatives of the genii Lactobacillus and Acetobacter. These data suggest that the intestinal lumen of an adult fly favors the survival of specific bacteria, and that such bacteria encode the necessary factors to survive or proliferate within a Drosophila intestine.
Consistent with a long-term association between the fly intestine and specific microbes, many Drosophila phenotypes are influenced by individual Lactobacillus or Acetobacter species. For example, Lactobacillus plantarum, a common Drosophila-associated microbe, promotes larval development via regulation of the TOR signal transduction pathway (32), while Acetobacter pomorum regulates host Insulin Growth Factor signals to promote development and metabolic homeostasis (33). In addition, members of the Acetobacter and Lactobacillus populations regulate levels of essential nutrients in the host (34)(35)(36). Combined, these data present a compelling argument that Lactobacilli and Acetobacter are important members of the Drosophila-microbe holobiont.
Despite our advances in the elucidation of Lactobacillus and Acetobacter influences on their Drosophila host, we know remarkably little about the bacterial genomic elements that determine microbial responses to, and survival within, the Drosophila adult intestine. To address this issue, we prepared whole genome sequences of three bacterial species that regularly associate with Drosophila -Lactobacillus brevis, Lactobacillus plantarum, and Acetobacter pasteurianus. These sequences included those for a Lactobacillus plantarum strain isolated from our lab-raised flies, and a separate strain isolated from a wild Drosophila. For each species, we compared Drosophila-associated bacterial genomes, including ones reported previously, to the genomes of reference strains isolated from non-Drosophila sources. We focused on biochemical and signal transduction pathways that respond to environmental factors, as we reasoned that Drosophila-associated genomes will display signatures of adaptation to association with their fly host. Our studies identify several biochemical pathways that are enriched in the genomes of Drosophila-associated Lactobacilli, and uncover Acetobacter genomic elements that regulate viability within the host.

RESULTS
Intestinal bacteria face structural and chemical barriers that typically deter association with an animal intestine. In response, microbial genomes encode factors that overcome host defenses to permit bacterial growth or survival. Here, we examined the genomes of L. brevis, L. plantarum and A.
pasteurianus, common members of the Drosophila intestinal community. For each species, we studied whole-genome sequences of bacterial strains isolated from Drosophila intestines, and compared them to related strains isolated from the environment. As it is unclear if Lactobacilli or Acetobacter establish autochtonous or allochtonous relationships with the adult intestine we will refrain from use of the term commensal throughout this study. Instead, we will identify bacterial strains isolated from flies as associated with Drosophila, and we will identify all other strains as environmental isolates. Details on the respective genomes characterized in this study are presented in table 1.
We processed each genome in a similar manner. Where necessary, we used genomic databases to identify the bacterial species of newly sequenced genomes. We then annotated each genome with RAST, used PHAST to scan each genome for intact prophages, and searched for possible CRISPR arrays in the respective genomes. We scrutinized the annotated genomes for functions that facilitate microbial survival within the intestinal lumen, with a focus on genes involved in signal transduction, transcriptional responses, orchestration of stress responses, or induction of virulence factors. Finally, we compared environmental and Drosophila-associated genomes for each species to identify bacterial factors that are unique to Drosophila-associated genomes. We present the results for each species below.

Prophages and CRISPR Responses
In contrast to the similarities between environmental and Drosophila associated strains noted above, we detected large differences when we searched for prophages within the respective genomes.
The EW and EF genomes include four intact temperate prophages, compared to an absence of prophages from the environmental strain genome (Table 4). We found CRISPR sequences that target a common Lactobacillus phage within all three genomes, while the environmental strain encoded a separate CRISPR array that targets a Lactobacillus plasmid (Table 4). These results suggest an ongoing interaction between prophages and CRISPR defenses in the genomes of Drosophila-associated L. brevis strains.

Function-Based Comparisons of Commensal and Reference Strains of Lactobacillus brevis
The data above point to subtle differences between the genomes of Drosophila-associated and environmental strains of L. brevis. To characterize these differences in greater detail, we performed a function-based comparison of the 185 genes that are common to Drosophila-associated genomes, but absent from the environmental strain. This set of 185 genes describes thirteen distinct functional categories, with forty-seven unique roles (Table 5). Perhaps unsurprisingly, phage and CRISPR-associated gene products account for two of those categories, and cover eleven of the forty-seven unique roles.
Of the remaining gene products, the dominant functional categories are dedicated to roles suited for survival within an intestine. For example, both Drosophila-associated genomes encode a biochemical cascade that converts α-D-glucose-1-phosphate to dTDP-4-dehydro-6-deoxy-L-mannose, an exopolysaccharide that contributes to biofilm formation, and prokaryotic survival within a host intestine (38,39). Drosophila-associated strains of L. brevis also encode gene products that contribute to the formation of rhamnose-containing glycans, a cell membrane component of acid-fast bacteria that affects host-microbe interactions, such as adhesion, recognition, and biofilm formation (40).
We also identified gene products within Drosophila-associated genomes that facilitate nutrient acquisition from different sources (Table 5). Bacteria frequently respond to limitations in nutritional environments through activation of the cAMP Receptor Protein, a transcription factor that we did not identify in the environmental strain of L. brevis, but found in both Drosophila-associated strains. The cAMP Receptor Protein controls, among other things, the expression of gene products that coordinate metabolism of citrate (41), a function that is also enriched among associated Drosophila-associated L.
brevis genomes. In lactic acid bacteria, citrate lyase is activated in acidic environments such as those found in the gut, and increases carbon utilization and energy generation by blocking the inhibitory effects of the Lactobacillus fermentation product lactate (42). Finally, we detected an enrichment of genes involved in the transport and degradation of pectin in Drosophila-associated L. brevis genomes.
Pectin is an abundant source of energy and carbon for bacteria that grow on plant and vegetable surfaces, and microbial consumption of pectin accelerates the decay of organic matter. As Drosophila feed on bacteria and yeast that grow on decomposing fruit, we believe that pectin metabolism by the EW and EF strains of L. brevis increases the likelihood of their association with Drosophila in the wild.

General Genomic Features
Although Acetobacter frequently associate with the intestines of Drosophila in the wild and in the lab, we are unaware of any whole-genome sequences of A. pasteurianus strains derived from the intestines of adult Drosophila. We consider this a significant deficit given the impacts of Acetobacter on Drosophila development and metabolism. To address this shortcoming, we completed a whole-genome sequence of an A. pasteurianus strain (A. pasteurianus AD) that we isolated from the intestines of wildtype Drosophila. For comparative purposes, we examined the available genomic sequences of the NBRC 101655 strain, and the ATCC 33445 strain. The NBRC 101655 strain was originally isolated from pineapple in Thailand and successfully associates with adult Drosophila (Figure 2). In contrast, we found that the environmental ATCC 33445 isolate fails to survive within the intestine of adult Drosophila ( Figure 2).
At first glance, we did not observe substantial differences between the respective genomes. Each genome is approximately 3 MB in length, with similar GC content, and similar numbers of RNA, and predicted coding sequences ( Table 6). From an evolutionary perspective, A. pasteurianus AD appears more closely related to the NBRC 101655 strain than the ATCC 33445 strain ( Table 6). Consistent with a greater evolutionary distance to the ATCC strain, we found that the ATCC 33445 genome encodes 112 unique proteins, while the NBRC 101655 and AD strains share 112 genes that are absent from the ATCC 33445 genome ( Figure 3A).

Environmental Response Factors
As with L. brevis, we first compared the Drosophila-associated and environmental genomes for  (Table 7, and supplementary tables [10][11][12]. Across this series of comparisons, the most pronounced differences were commensurate with a closer relationship of the AD strain to the NBRC 101655 strain than to the ATCC 33445 strain. Thus, this admittedly limited comparison does not appear to identify genomic components that readily distinguish Drosophila-associated A. pasteurianus genomes from environmental counterparts. Nonetheless, these functional characterizations uncover differences between the AD and NBRC 101655 A. pasteurianus genomes that associate with Drosophila, and the ATCC 33445 genome that fails to do so.

Function-Based Comparisons of Commensal and Reference Strains of Acetobacter pasteurianus
Our fortuitous identification of an environmental A. pasteurianus strain that fails to survive passage through the fly gut allowed us to explore A. pasteurianus genomes for factors that permit association with Drosophila. As the AD and NBRC 101655 strains successfully pass through the adult intestine, we reasoned that the AD and NBRC 101655 genomes encode biochemical functions absent from ATCC 33445 that permit association with Drosophila, or that the ATCC 33445 genome encodes biochemical functions absent from the other strains that prevent association with Drosophila. This prompted us to identify biological subsystems shared exclusively by the AD and NBRC 101655 (Table 9), or unique to the ATCC 33445 genome (Table 10).
In this comparative analysis, we noted four subsystems exclusive to the AD and NBCR 101655 genomes that may explain their ability to survive passage through the fly intestine. Both strains encode polyamine metabolism factors that are frequently associated with cellular growth and survival, and have established roles in the formation of biofilms (43). The association-competent genomes also encode factors necessary for the conversion of urea to ammonium and carbon dioxide. A similar system operates in Helicobacter pylori where it raises the gastric pH to generate a more hospitable environment for microbial survival (44). We also detected the redox-sensitive transcriptional activator SoxR in both association-competent genomes. SoxR promotes microbial survival within the intestine by countering the antibacterial actions of host-derived superoxide anions (45). Finally, we detected several genes that contribute to organic sulfur assimilation in association-competent genomes. These gene products allow A. pasteurianus AD and NBRC 101655 to use alkanesulfonates as a source of sulfur during sulfate or cysteine starvation and may provide both strains a competitive advantage in the intestine if sulfur is limiting.
The association-incompetent ATCC 33445 strain also encodes products that contribute to generation of ammonia. However, the ATCC 33445 strain relies on respiratory nitrate reductase and nitrite reductase to generate ammonia, as well as assimilatory nitrate reductase to access nitrate for metabolic growth. This represents an entirely different strategy to use nitrogen as a fuel for metabolic energy and growth. Thus, it is feasible that the ATCC strain fails to grow within the adult intestine due to unsatisfied nitrogen requirements. We also observed two toxin-antitoxin systems unique to the associationincompetent ATCC 33445 genome -an addiction module toxin that ensures propagation of plasmids to progeny cells (46), and a MazE/MazF type toxin-antitoxin (47). The MazE/MazF system induces programmed cell death in prokaryotic cells in response to stressful environments. It is intriguing to speculate that the adult intestine activates this response, thereby inducing microbial death and preventing ATCC 33445 survival within the gut.

General Genomic Features
We used the environmental ATCC 14917 strain as a reference for our comparative analyses of L. plantarum genomes. L. plantarum ATCC 14917 was isolated from pickled cabbage and we confirmed viable association of this strain with the adult Drosophila intestine. We compared the ATCC 14917 strain to four Drosophila-associated genomes -WJL, DMCS_001, DF and KP. WJL and DMCS_001 were isolated from Drosophila raised in geographically separate labs (48,49). We isolated the KP strain from the intestines of our lab-raised wild-type strain, and the DF strain from an isofemale wild Drosophila melanogaster line that we captured in Edmonton, Canada in the summer of 2014. The DF and KP genomes encode one chromosome and three closely related plasmids each ( Figure 4). While all five genomes are closely related, genome-to-genome distance calculators suggest a greater degree of identity among the Drosophila-associated KP, DF, WJL and DMCS_001 strains (Table 11). In general, the environmental genome is smaller than the Drosophila-associated genomes ( Figure 5A), encodes fewer RNAs and coding sequences, and contains fewer phage-associated proteins (Table 11).

Environmental Response Factors
Contrary to our expectations, we did not observe pronounced differences between the genomes of environmental or Drosophila-associated strains in terms of modifiers of signal transduction, stress responses or virulence ( Figure (Table 12, supplementary tables [16][17][18]. In general, these findings suggest that Drosophila-associated strains of L. plantarum are not easily distinguished on the basis of functions that sense and respond to specific environmental cues.

Prophages and CRISPR Responses
Similar to our observations with L. brevis genomes, we observed a greater number of intact prophage genomes in Drosophila-associated L. plantarum strains than in the environmental strain (Table   13). Of all Lactobacillus genomes studied, we detected an average of four intact prophage genomes in Lactobacillus strains isolated from flies, and a maximum of two prophage genomes in environmental strains ( Figure 6). The main difference between the Drosophila-associated brevis and plantarum strains is that the plantarum strains do not appear to encode CRISPR-dependent anti-phage defenses within their genomes.

Function-Based Comparisons of Commensal and Reference Strains of Lactobacillus plantarum
We then used the whole-genome sequence data for environmental and Drosophila-associated strains to prepare an L. plantarum pangenome (Figure 7). Visual inspection indicated that three of the Drosophila-associated genomes (DF, KP and WJL) display similar organization, while the DMCS_001 genome appears more akin to the environmental genome. When we looked for factors exclusively common to the DF, KP and WJL genomes we detected several gene products that derive from prophages. However, we also detected functional elements that are consistent with support for microbial growth in the intestinal lumen. This includes access to methionine, an essential amino acid in L. plantarum; enzymes for the regulation of sialic acid, a comparatively rare bacterial molecule that is typically found in prokaryotes that live in association with deuterostomes (50); and a regulator of the biosynthesis of lipoteichoic acid, a regulator of cell wall autolysis in gram-positive bacteria (51,52).
Strikingly, we also detected uridine phosphorylase exclusively in the genomes of the Drosophilaassociated DF, KP and WJL strains. Uridine phosphorylase catalyzes conversion of uridine to uracil, a secreted metabolite that causes microbial-dependent inflammation in the intestine of adult Drosophila (53). Combined, these genes constitute a comparatively small number of molecular functions that may favor the propagation and establishment of L. plantarum populations in the fly intestine.
Finally, we examined the thirty-five genes exclusively observed in the genomes of DF, KP, WJL and DMCS_001. The majority (nineteen) were prophage genes, and an additional five were hypothetical proteins of unknown function. Rather strikingly, several of the remaining genes encode products that actively suppress the growth of competing microbes. These include the PlnMNO operon that encodes a bacteriocin and cognate immunity protein (54), and 1,3-propanediol dehydrogenase, an enzyme that converts propane-1,3,-diol to 3-hydroxypropanal. 3-hydroxypropanal is also known as reuterin, a Lactobacillus reuteri metabolite that exerts broad-spectrum microbicidal effects on intestinal microbes (55). Combined the actions of PlnMNO and reuterin give Drosophila-associated L. plantarum a competitive advantage against other microbes within the gut lumen.

DISCUSSION
The last decade witnessed a proliferation of elegant studies that uncovered critical host responses to microbial factors in the Drosophila intestine (Reviewed in (13)). Bacterial cues promote larval growth (32,33), direct innate immune responses (56,57), orchestrate the proliferation of intestinal stem cells (58,59), and regulate the uptake and storage of nutrients (36). Despite the importance of the intestinal microflora for Drosophila health and development, we know remarkably little about the biochemical events that permit bacterial survival within the hostile terrain of a fly intestine. A recent study identified microbial metabolism and stress response pathways that mediate interactions between intestinal bacterial and their Drosophila host (49). In this study, we examined the genomes of Drosophilaassociated strains of L. brevis, L. plantarum, and A pasteurianus. We were particularly interested in the identification of candidate signal transduction, virulence and defense factors that permit bacterial survival within the intestines of adult flies. To this end, we compared fly-associated genomes to environmental strains of the same species. For each species, we observed a small number of genetic pathways that were exclusive to the Drosophila-associated genomes. Many of the Drosophila-associated pathways encode products with established roles in host-microbe interactions, suggesting these products facilitate association of Drosophila with the individual strains.

Lactobacilli
For our studies of Lactobacillus genomes, we prepared whole-genome sequences of L. brevis or L.
plantarum strains that we isolated from lab-raised wild-type flies, and an L. plantarum strain that we isolated from a wild Drosophila. These genomes formed the cornerstones of a comparative study that included three previously reported Drosophila-associated genomes (37,48,49), and the genomes of environmental strains that successfully associate with the intestines of our wild-type Drosophila. In this manner, we identified bacterial functions that are unique to Drosophila-associated genomes of L. brevis and L. plantarum. The functions fall into four broad categories -antibacterial, structural, metabolic, and phage-related.
The most striking feature common to all four Drosophila-associated L. plantarum genomes was the presence of broad-spectrum bactericidal factors. For example, the DF, KP, WJL and DMCS_001 genomes all encode a complete PnlMNO operon, which encodes a bacteriocin and a corresponding immunity protein (54). Bacteriocins are produced by many lactic acid bacteria to kill neighboring bacteria, while the immunity protein protects L. plantarum from collateral damage (60). In addition, the Drosophila- The Drosophila-associated genomes of L. plantarum and L. brevis also encode structural components that stabilize associations with their fly host. In particular, we detected metabolic pathways for modifications to cell wells that permit host-microbe interactions and biofilm formation. These include the construction of exopolysaccharides by L. brevis and the regulation of sialic acid by L. plantarum.
Sialic acid is a comparatively rare microbial metabolite, but has been observed on microbes that associate with deuterostomes. Bacteria use sialic acid as a nutrient, but they also use it to mask detection by host immune responses. While the role of sialic acid in L. plantarum association with Drosophila requires further investigation, we feel that these elements merit consideration as hostmicrobe interaction factors.
The Drosophila-associated genomes of L. plantarum and brevis also includes gene products that address nutritional requirements. These functions include access to limited resources such as methionine by L. plantarum, and utilization of citrate as an energy source by L. brevis. We were particularly struck by the presence of pectin metabolism factors within the genomes of Drosophilaassociated strains of L. brevis. Pectin is an excellent source of carbon for bacteria that grow on plants.
However, bacterial utilization of pectin accelerates the ripening and decay of the same plants (61). Thus, Drosophila-associated L. brevis genomes express factors that contribute to the decay of organic substrates. This is particularly notable, as Drosophila preferably consumes decayed matter as a source of nutrients. The ability of L. brevis to generate meals for their Drosophila host provides a possible explanation for the fact that Drosophila frequently associate with L. brevis, even though intestinal bacteria do not appear to establish stable populations within the guts of adult Drosophila. As L. brevis generates palatable meals for fly hosts, we speculate that their chances of association with flies in the wild are greater than those for many other bacteria. This host-microbe relationship is similar to a proposed mechanism for association of Erwinia carotovora with Drosophila in the wild (62).
The final difference we noted between environmental and Drosophila-associated Lactobacillus genomes was an accumulation of temperate prophage genomes throughout Drosophila-associated Lactobacilli. Intestinal stresses such as high levels of reactive oxygen species are known to trigger lysogenic induction of temperate prophages (63). Thus, it is feasible that bacterial strains that pass though the fly intestine will release and integrate greater numbers of lytic prophages, explaining the increased numbers of prophage genomes in Lactobacillus strains that associate with adult Drosophila.

Acetobacter
In this study, we report the first genome of a Drosophila-associated strain of A. pasteurianus, and identified an A. pasteurianus strain that fails to survive passage through the fly gut. A comparison of association-competent and association-incompetent genomes is limited by the fact that only one Drosophila-associated genome is available for study. Nonetheless, our identification of an environmental A. pasteurianus strain that fails to associate with Drosophila yields a comparatively short list of candidate functions that regulate association of Acteobacter with adult Drosophila. This list includes gene products that process nitrogen, and gene products that directly control the induction of cell death in A. pasteurianus. To our knowledge, ours is the first study to identify a set of bacterial factors that regulate association of an Acetobacter with Drosophila. We believe that future characterization of mutations in the respective gene products will identify the bacterial factors that control viability of A. pasteurianus in the adult intestine. This approach has considerable value given the relationship between Acetobacter and Drosophila development.
In summary, our comparative study of bacterial genomes uncovers a number of genetic signatures of association with Drosophila. As many of the gene products have established roles in host-microbe interactions, we propose that these genes include factors that promote the frequent association of Drosophila with Lactobacillus and Acetobacter strains. Future characterization of mutations in the individual products will reveal the relationship between the individual factors and host physiology.

CONCLUSION
The intestinal microflora influences a number of physiological outcomes in their animal host. Microbial factors shape the establishment and execution of immune responses, influence the availability of critical nutrients and often determine the progression of debilitating inflammatory illnesses. Studies in Drosophila melanogaster have provided valuable insights into the host cellular machinery that coordinates responses to microbial presence. However, we know remarkably little about the microbial factors that permit long-term association with the host. In this study, we characterized the genomes of common Drosophila-associated microbes and identified putative bacterial products that favor association with Drosophila. We believe that products constitute valuable starting points to examine the microbial half of the host-microbe relationship.

Drosophila husbandry
All Drosophila assays were performed with virgin w 1118 male and female flies raised on standard corn-meal medium (Nutri-Fly Bloomington Formulation, Genesee Scientific) in a humidified incubator at 29°C. To generate germ-free flies, we transferred freshly eclosed (0 -16 hr old) adult flies to standard medium that we supplemented with an antibiotic (100 μg/ml ampicillin, 50 μg/ml vancomycin, 100 μg/ml neomycin and 100 μg/ml metronidazole dissolved in ethanol). This cocktail has been described previously (22). To generate gnotobiotic flies, we raised adult flies on the antibiotic cocktail for five days, starved flies for two hours, and transferred the flies to a vial containing an autoclaved fly vial cotton plug soaked with the respective bacteria resuspended in a sterile-filtered solution of 10% sucrose in PBS. We fed the flies the bacterial meal for 16 hours and then transferred the flies to vials of freshly autoclaved food. To confirm association, we periodically plated fly homogenates on bacterial medium selective for Acetobacter or Lactobacilli.

Bacterial Isolation and Sequencing
We plated homogenates of 15 day old adult Drosophila on Acetobacter and MRS culture plates. We

Genome Assembly and Annotation
For each sequencing project, we confirmed the individual species with the SpeciesFinder 1.2 algorithm (64) and calculated genome to genome distances with the genome to genome distance calculator of the Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures (65). We annotated each genome with RAST (66), identified intact prophage genomes with the PHAST server (67) and identified CRISPR arrays with CRISPRFinder (68). We used the CRISPRTarget algorithm to predict CRISP targets for the individual CRISPR arrays (69). To identify regulatory proteins within the respective genomes, we used the P2RP identifier (70). We used the GView tool to generate graphical representations of bacterial genomes (71).      19 15 20             with the corresponding genomes aligned in outer wheels.