Cell-Specific Regulation of Intestinal Immunity in Drosophila

The intestinal epithelium contains secretory and absorptive cell lineages that are derived from undifferentiated progenitor cells. Despite the collective importance of these cells to host responses against microbial invaders, little is about the contributions of immune responses in individual cell types to the maintenance of intestinal homeostasis. In this study, we asked how inhibition of immune pathway activity exclusively in progenitor cells, or in differentiated enterocytes, affects midgut homeostasis in adult Drosophila. We found that blocking immune activity in enterocytes rendered flies more tolerant of Vibrio cholerae infection, had negligible effects on the gut bacterial microbiome, and significantly affected metabolism. In contrast, inhibition of immune activity in progenitors rendered flies less tolerant of Vibrio infections, modified host association with Lactobacillus symbionts, and blocked growth and renewal in the midgut epithelium. Together, these data uncover substantial cell type-specific contributions of epithelial immunity to the maintenance of adult intestinal homeostasis.


INTRODUCTION
The intestine processes a mixture of ingested materials that contain essential nutrients, but may also be tainted with harmful toxins or microbes. To extract nutrient and limit tissue damage, intestinal homeostasis requires the coordinated actions of immune and metabolic responses within the epithelium. The epithelium is maintained by progenitor cells that divide and differentiate to generate secretory, or absorptive cells that act in regionally specialized manners. For example, the anterior intestinal epithelium contains defined regions of low pH, and areas characterized by the expression of digestive enzymes that degrade complex, bulky material. In contrast, posterior epithelial cells are typically responsible for the uptake of small molecules, such as water and electrolytes.
In addition to digestion, the epithelium is also critical for the neutralization of threats posed by resident, or transitory microbes (1). Antimicrobial defenses include physical barriers constructed from polymeric materials, such as mucins or the chitinous peritrophic matrix, as well as tight cell-cell contacts that prevent paracellular leakage of potentially harmful molecules. In addition to barrier defenses, host-derived reactive oxygen species eliminate invasive microbes in a rather indiscriminate manner that can also lead to collateral damage of host tissue. Finally, immune systems rely on intra-and extracellular receptors that survey the intestinal environment for molecular patterns of the gut microbial content (2). Detection of microbial signatures activates defenses tailored to prevent microbial invasion of the host interior.
Although we understand many host signaling pathways that respond to intestinal microbes, we have less information about how the engagement of immune pathways in specific cell types contribute to the immune response. Drosophila melanogaster is a valuable system to address this question. The fly posterior midgut shares numerous similarities with the vertebrate small intestine, and is highly amenable to the modification of gene expression in defined intestinal cells (3). Similar to vertebrates, the fly midgut is maintained by a basal population of progenitor cells composed of intestinal stem cells that typically divide asymmetrically to generate a new stem cell and a transient cell type -the enteroblast (4,5). Enteroblasts frequently respond to signals from the Notch pathway to generate absorptive enterocytes (6,7), a differentiated cell type that undergoes several rounds of 4 endoreplication to produce large cells that occupy the majority of the midgut cellular volume. Less frequently, stem cells generate secretory enteroendocrine cells, either indirectly via enteroblasts that do not receive Notch signals, or more directly via an enteroendocrine intermediary (8)(9)(10). As with vertebrates, stem cell division is regulated by mitogenic cues from pathways that include EGF, Wnt, and JAK-STAT, and the rate of ISC division is tuned to the rate of epithelial damage (11,12). In the absence of an extrinsic stressor, stem cells divide approximately once every two weeks in adult females (4,13). However, damage to the midgut activates repair pathways that shift ISC proliferation rates. In this case, molecular cues from damaged cells accelerate ISC division.
This burst of progenitor growth generates a pool of cells that replenishes damaged structures and maintains the epithelial barrier.
In contrast to the advances made in understanding the contribution of individual cell types to epithelial renewal, we know less about cell-specific regulation of intestinal immunity. The Immune Deficiency (IMD) pathway, a germline-encoded antibacterial defense with numerous similarities to the mammalian TNF pathway (14), is activated by detection of bacterial DAP-type peptidoglycan (PGN) (15,16). IMD regulates approximately 50% of the intestinal transcriptional response to the microbiota (17), and alterations in immune activity are associated with effects on the composition and load of the gut microbiota (16)(17)(18)(19)(20). Similar to digestive processes, the IMD pathway displays hallmarks of cellular, and regional specialization (21)(22)(23). In the anterior of the intestine, IMD activation requires the extracellular receptor, PGRP-LC, that detects polymeric PGN, and antibacterial defenses appear to be predominantly activated in large, differentiated cells (24). In the posterior midgut, IMD activation requires PGRP-LE (24,25), an intracellular sensor that detects monomeric PGN (26), and IMD pathway components are expressed and activated in undifferentiated progenitor cells (27)(28)(29), as well as terminally differentiated enterocytes. Despite these regional and cellular distinctions in IMD activation, we do not know how IMD acts in individual cell types to regulate intestinal homeostasis. To address this question, we generated a Drosophila line that allows cell-specific inhibition of IMD. With this line, we determined the physiological consequences of IMD inactivation exclusively in progenitors or in enterocytes. Our results revealed significant differences between the contributions 5 of enterocyte IMD and progenitor IMD to intestinal homeostasis. Inhibition of IMD in enterocytes extended the lifespan of adults challenged with pathogenic Vibrio cholerae, did not affect composition of the intestinal microbiome, but had significant effects on host metabolism. In contrast, inhibition of IMD in progenitors shortened the lifespan of flies challenged with Vibrio cholerae, impacted the representation of Lactobacillus symbionts within the microbiota, and significantly impaired the growth of intestinal stem cells. Combined these data uncover fundamental differences between the impacts of immune activity in progenitor cells, and their differentiated progeny.

Cell-specific effects of IMD on host responses to bacterial infection
To determine cell-specific contributions of IMD to intestinal homeostasis, we developed a transgenic fly line that allows inducible inhibition of IMD in defined cell types. In Drosophila, IMD activation requires proteolytic removal of the N-terminal thirty amino acids from the Imd protein by the caspase Dredd, and expression a noncleavable Imd (ImdD30A) prevents signal transduction through the IMD pathway (30). Consistent with these data, we found that expression of ImdD30A in the adult fat body completely blocked the infection-dependent expression of the IMD-responsive antimicrobial peptide, diptericin (Fig. 1A). We used this transgenic fly line to ask how cell-sepcific inactivation of IMD impacts host survival after with oral infection with V.cholerae, a pathogenic bacteria that kills adult flies in a manner that includes contributions from the IMD pathway (31,32). To answer this question, we used the TARGET gene expression system (33) to express ImdD30A exclusively in intestinal progenitors (esg ts /D30A), or in differentiated enterocytes (Myo1A ts /D30A). When we blocked IMD exclusively in progenitor cells, we found that flies infected with V. cholerae died significantly faster than control esg ts /+ flies (Fig.   1B). Conversely, enterocyte-specific suppression of IMD extended the survival of infected flies relative to control Myo1A ts /+ flies (Fig. 1C). As esg ts /D30A flies and Myo1A ts /D30A flies respond differently to infection with V. cholerae, we asked if inhibition of IMD in progenitors or enterocytes had differential effects on host colonization by V. cholerae. Initially, we looked at the impacts of esg ts /D30A and Myo1A ts /D30A on feeding rates, a possible modifier of host colonization by V. cholerae. In Capillary Feeding (CAFE) assays, we found that esg ts /D30A flies and Myo1A ts /D30A flies consumed equal amounts of liquid food as their respective controls ( Fig. 1 D and E).
Furthermore, we found that expression of ImdD30A in enterocytes or in progenitors did not impact the consumption of solid food, as determined by a FlyPad assay ( Supplementary Fig. 1). Finally, we detected equal loads of V. cholerae in esg ts /D30A flies and Myo1A ts /D30A flies relative to their respective controls (Fig. 1 F and   G). In sum, our data suggests that IMD has cell-specific impacts on host tolerance of an enteric pathogen.

Effects of cell-specific IMD inactivation on the intestinal microbiome
The data above uncover cell-specific contributions of IMD to survival of infection with a pathogenic bacteria.
This led us to ask if IMD acts in cell-specific manners to regulate interactions between the host and symbiotic bacteria. To answer this question, we first generated axenic populations of esg ts /+, esg ts /D30A, Myo1A ts /+ and Myo1A ts /D30A embryos ( Fig. 2A). Next, we raised axenic embryos at 21°C in sterile environment until adulthood to prevent unwanted expression of imdD30A during development. We fed axenic adults a homogenate prepared from our lab wild-type flies to ensure that all genotypes had a similar intestinal microbiome. We transferred flies from each group to 29°C for ten days to drive ImdD30A expression in progenitors (esg ts /D30A) or enterocytes (Myo1A ts /D30A). Finally, we performed triplicate 16S deep-sequencing analyses of the intestinal microbiome in each genotype at days five and twenty-nine after association with a conventional microbiome. Of the 24 samples, 22 had sufficient read depth to allow us determine the overall composition of the gut bacterial microbiome.
Principle component analysis (Fig. 2B), phylogenetic diversity metrics (Fig. 2C), and operational taxonomic unit analysis ( Fig. 2D) all suggest that inactivation of IMD in differentiated enterocytes had minimal impact on the composition of the intestinal microbiome at both time points. In each case, we did not observe significant differences between bacterial populations in Myo1A ts /ImdD30A flies, and their Myo1A ts /+ controls at either day 5 or day 29. In contrast, inactivation of IMD in progenitors appears to affect both the diversity (Fig. 2 B andC) and composition of the intestinal microbiome ( Figure 2D), particularly at early stages. Here, we noted a relative increase in the amounts of Lactobacillales in five day-old esg ts /D30A intestines relative to age-matched esg ts /+ controls. In older intestines, we observed a reverse effect (Fig. 2D). In this case, the relative abundance of Lactobacillales was lower in esg ts >D30A populations (mean abundance: 8.1%), than in all other genotypes (mean abundance: 15.2%). We then generated balance trees to determine if cell-restricted inactivation of IMD affects the abundance of bacterial subcommunities within the intestinal microbiome. These studies identified a dominant balance that distinguished five day-old esg ts /D30A intestines from all other genotypes ( Fig. 2 E andF). Further examination of this group revealed a dominance of Lactobacillales and Psudomonadales within that balance.
Combined, these data suggest cell-dependent contributions of IMD to host associations with Lactobacillus species: 8 inhibition of IMD in progenitor cells has an impact on the growth of Lactobacillus symbionts, while inhibition of IMD in enterocytes appears to have negligible effects.

Loss of IMD in progenitors affects association with symbiotic Lactobacillus species
Our 16S deep-sequencing data suggested differential effects of cell-restricted IMD inactivation on the relative proportions of intestinal symbionts in the fly (Fig. 2). To determine if IMD affects the total amounts of intestinal bacteria, we measured bacterial loads in flies with either progenitor or enterocyte-specific inactivation of IMD.
For this experiment, we first standardized the bacterial composition associated with each fly lines. Specifically, we generated axenic embryos of all genotypes (esg ts /+, esg ts /D30A, Myo1A ts /+, and Myo1A ts /D30A) to eliminate the endogenous microbiome, and poly-associated the axenic adult flies with a 1:1:1 mix of three common fly symbionts -Acetobacter pasteurianus (Ap), Lactobacillus brevis (Lb), and Lactobacillus plantarum (Lp). We measured bacterial loads by serial dilution and counting of colony-forming units in adult flies that we raised at 29°C for 1, 10, 20, or 30 days. This approach allowed us to trace age-dependent changes in the microbiota of flies with compromised intestinal immune defenses. Consistent with our deep-sequencing data, enterocyte-specific suppression of IMD did not affect the load of any of the tested bacterial species at any of the tested time points, suggesting that loss of IMD in enterocytes alone does not have a detectable effect on either the load or relative proportions of common fly intestinal symbionts (Fig. 3A). Similarly, we did not observe measurable effects of IMD inhibition in progenitors on Ap loads at all times tested. In control esg ts /+ flies, the load of Ap gradually increased as flies aged to 20 days and dropped slightly in 30-day old flies (Fig. 3B). This pattern was also observed in flies with progenitor-specific suppression of IMD (esg ts /D30A), and there were no significant differences in intestinal Ap numbers between the two genotypes at any of the times measured. In contrast, we detected age-dependent effects of IMD inhibition in progenitors on intestinal Lactobacillus numbers. For example, Lb load remained approximately the same up to 10 days in esg ts /+ and esg ts /D30A flies and increased up to 1 X 10 6 CFU/fly in 30day-old control flies (esg ts /+). However, in esg ts /D30A flies there was a significant decrease in the colony-forming units per fly of Lb at 30 days compared to esg ts /+ control flies of the same age (Fig. 3B). We noted similar effects 9 of IMD inhibition in progenitors on Lp numbers. Here, there was a significant drop in Lp loads at 30 days in esg ts /D30A flies relative to their esg ts /+ controls (Fig. 3B). Combined with the data presented in Figure 2, our results suggest that IMD activity in progenitor cells contributes to the maintenance of intestinal Lactobacillus symbiont numbers as the fly ages.

Enterocyte IMD regulates metabolism and adult viability
Given the cell-specific impacts of IMD inactivation on host survival, microbiome composition, and symbiont load, we hypothesized that the IMD pathway exerts cell-specific controls on transcriptional events in the midgut.
To test this hypothesis, we performed side by side RNA-sequencing of the transcriptome of guts with IMD specifically inactivated in enterocytes or intestinal progenitors. To understand how enterocyte IMD modifies intestinal function, we compared the transcriptomes of purified midguts from Myo1A ts /D30A and control Myo1A ts /+ flies that we raised at 29°C for ten days (Fig. 4A). We found that inhibition of IMD in enterocytes had substantial effects on intestinal transcriptional activity (Fig. 4B, C), leading to modified expression of genes involved in a range of cellular processes (Fig. 4D). Unsurprisingly, the cohort of IMD-responsive genes includes immune effectors and modulators ( Supplementary Fig. 2). However, we also found that inhibition of IMD in enterocytes had significant effects on processes as diverse as the control of cell shape, transport of metabolites, and metabolism of amino acids (Fig. 4D). These wide-ranging impacts prompted us to ask if inhibition of IMD in enterocytes affects metabolism in the fly. To answer this question, we examined glucose, trehalose and lipid levels in the intestines and carcasses of Myo1A ts /D30A and Myo1A ts /+ flies that we incubated at 29°C for ten days.
Inhibition of enterocyte IMD did not affect glucose levels in the whole fly, or in the intestine (Fig. 4

E and F).
Likewise, inhibition of IMD in enterocytes did not affect trehalose, the main circulating disaccharide in the fly (Fig.   4G). In contrast, inhibition of IMD in enterocytes resulted in significant increases of total ( Figure 4H), and intestinal triglyceride levels (Fig. 4I), as well as an increase in adult weight (Fig. 4J). These observations implicate enterocyte IMD in the control of triglyceride storage in the adult. Interestingly, a loss of function mutation in imd leads to decreased triglyceride levels ( Supplementary Fig. 3), indicating enterocyte-specific roles for IMD in triglyceride storage. As enterocyte IMD modifies several key processes such as metabolism in the intestine, we reasoned that inactivation of IMD will have consequences for intestinal physiology and animal viability. Consistent with this prediction, we found that blocking IMD in enterocytes impaired intestinal stem cell proliferation (Fig. 4J), and shortened adult lifespan (Fig. 4K). Combined, these results show that enterocyte IMD activity contributes to the regulation of adult viability and metabolism.

Progenitor-specific roles for IMD
In parallel to a transcriptional characterization of Myo1A ts /D30A midguts, we determined the transcriptomes of purified midguts from esg ts /D30A and esg ts /+ flies that we raised at 29°C for ten days. Again, we noticed significant impacts of IMD inhibition on the transcriptional activity of the midgut (  (Fig. 5F). We also noticed particularly striking impacts on the expression of genes required for cholesterol absorption and signaling; piRNA biogenesis; components of Notch, Wnt and JAK/STAT pathways; modifiers of proteostasis; and regulators of adult lifespan (Fig. 5F). We consider these alterations to gene expression particularly interesting, as many of the pathways are involved in progenitor cell homeostasis and aging (34)(35)(36). In combination, our transcriptional studies reveal large differences between the contributions of enterocyte, or progenitor cell, IMD to intestinal homeostasis, and raise the possibility that progenitor cell IMD has roles in the growth and viability of intestinal stem cells.

IMD activity in progenitors promotes stem cell proliferation and Drosophila viability
Based on the effects of progenitor-specific suppression of IMD on pathways with roles in intestinal homeostasis and aging ( Figure 5F), we hypothesized that progenitor-specific inhibition of IMD will affect the proliferation of intestinal stem cells and adult Drosophila lifespan. To directly quantify the effects of ImdD30A on intestinal stem cell populations, we first determined the proportion of stem cells in young and aged flies using Delta as a stem cell marker. We did not observe differences in the number of ISCs between 5-day-old esg ts /D30A and control esg ts /+ flies (Fig. 6A). However, there was a significant decrease in the percent of Delta-positive cells in esg ts /D30A flies compared to esg ts /+ control flies when flies were aged for 30 days (Fig. 6A). These results suggest that IMD activity promotes ISC proliferation in adult flies. In agreement with this hypothesis, we detected a significant decrease in proliferating cells in esg ts /D30A flies when compared to esg ts /+ control flies (Fig. 6B).
Together, our data suggest that IMD in progenitors is essential for the maintenance and growth of ISCs.
Given the impacts of IMD inhibition on progenitor renewal, we next tested the effect of progenitor-specific IMD suppression on longevity. For this experiment, we used the RU486-inducible GeneSwitch GAL4 system (33) to block IMD in progenitors. Using two independent UASimdD30A lines, we monitored and compared the lifespans of control flies to flies with progenitor-specific suppression of IMD. In both cases, we found that flies died significantly faster when IMD was blocked in progenitors in comparison to control flies ( Fig. 6 C and D). Taken together, our data indicate that IMD activity in the progenitor compartment protects the fly from pathogenic Vibrio cholerae, supports the growth of symbiotic Lactobacillus species, and contributes to epithelial growth and adult longevity.

DISCUSSION
The intestine contains functionally distinct secretory and absorptive cell lineages that are replenished from a population of basal progenitor cells. These cells coordinate nutrient acquisition with immune defenses to support animal growth and survival. Disruptions to gut immunity permit the invasion of interstitial tissues by commensal microbes, leading to the onset of severe illnesses in the host (37). To fully understand how immune defenses restrict commensal microbes to the intestinal lumen, it is essential that we understand how individual cell types contribute to tissue-wide antibacterial defenses. We used Drosophila melanogaster to characterize the contributions of immune signaling in intestinal progenitors, or enterocytes, to intestinal homeostasis. Our study uncovered cell-specific contributions of IMD to host metabolism, intestinal growth, and the survival of enteric infection.
In Drosophila, DAP-type peptidoglycan activates the IMD response, a signal transduction pathway that displays hallmarks of regional specialization in the intestine (21)(22)(23). IMD has complex roles in the regulation of intestinal anti-bacterial responses (38). In the anterior, IMD activation occurs in differentiated epithelial cells upon detection of cell-extrinsic, polymeric peptidoglycan by PGRP-LC (24), leading to expression of antimicrobial peptides, presumably limiting the dissemination of ingested bacteria. In the posterior midgut, transcriptional and functional studies showed that IMD components are expressed and active in enterocytes and in progenitor cells (27)(28)(29), where IMD activation occurs upon detection of cytosolic, monomeric peptidoglycan by PGRP-LE (24,25).
Here, IMD contributes to the delamination of damaged epithelial cells (39), and the expression of immuneregulatory molecules such as the peptidoglycan amidase PGRP-LB (40). Alterations to IMD have been linked to effects on the size and composition of the gut microbiome (16)(17)(18)(19)(20), and the lifespan of the fly (41)(42)(43)(44). In the adult, IMD also has complex roles in responses to enteric infections. For example, IMD is necessary to survive infections with pathogenic Serratia marcescens (45), while IMD accelerates fly demise after infection with Vibrio cholerae (31,32).

13
Despite our progress in characterizing intestinal IMD, we know comparatively little about cell-specific aspects of IMD activity. For example, several lines of experimental evidence implicate IMD in metabolic homeostasis (17,(46)(47)(48), and alterations to lipid metabolism are hallmarks of the generation of anti-bacterial ROS in the intestine (49,50). However, we don't know if these activities are cell-intrinsic, or if they are secondary to IMD-mediated changes to the gut microbiota. Our results show that inactivation of IMD in distinct cell types has distinct effects on host metabolism. For example, inhibition of IMD in progenitors increases the expression of genes involved in lipid metabolism, and attenuates the expression of genes involved xenobiotic metabolism via glutathione or cytochrome p450. In contrast, inhibition of IMD in enterocytes increases expression of genes required for the transport of amino acids or carboxylic acid, and diminishes the expression of genes required for glycolysis or the metabolism of amino acids. Importantly, the effects of enterocyte IMD on metabolism appear to be independent of the microbiome, as inhibition of enterocyte IMD does not influence bacterial numbers, or microbiome composition. These data suggest that enterocyte IMD has cell-autonomous effects on metabolism. Modification of metabolism by enterocyte IMD has effects on distal sites, as inhibition of enterocyte IMD causes weight gain and hypolipidemia in the adult. Mutations in several IMD pathway genes are also linked to accelerated lipolysis in adult flies (51), further implicating the IMD pathway in the control and use of lipid stores.
The apparent absence of shifts in the microbiome after inhibition of enterocyte IMD was unexpected given the established effects of IMD pathway mutations on the microbiota (16)(17)(18)(19)(20). It is possible that these differences reflect a requirement for IMD to act in multiple intestinal cells to modify the gut microbiome. Alternatively, IMD activity in intestinal regions not examined in this study such as the acidic copper cells of the middle midgut may have more pronounced impacts on the microbiome. However, we note that a recent study reported similarities between the microbiome of conventionally reared flies, and immune-deficient dif ; key mutants (52).
Furthermore, immune defects have minimal impacts on the gut microbiome of fish or mice (53,54). In these models, familial transmission appear to be a major determinant of microbiome composition. As bacteria circulate from the food to the intestine of Drosophila (55), we consider it possible that mutations in IMD influence the recycling of symbiotic bacteria, with downstream consequences for the gut microbiome. For example, mutations 14 in the IMD-responsive NF-kB ortholog, rel, affect feeding rates in flies (51), possibly influencing the rate and extent of intestinal colonization by food-resident bacteria.
In contrast to enterocytes, inhibition of IMD in progenitors had detectable effects on intestinal levels of common fly symbitoic Lactobacillus species, and no apparent effects on Acetobacter pasteurianus. These observations are in keeping with a recent characterization of flies with mutations in PGRP-SD, an IMD pathway modifier that is expressed in posterior midgut progenitor cells (27). Loss of PGRP-SD specifically affected intestinal levels of Lactobacillus plantarum, but did not affect Acetobacter pomorum levels. Thus, it appears that IMD acts in progenitor cells of the midgut to modify host association with Lactobacillus symbionts. It will be interesting to see if the effects of IMD on intestinal Lactobacillus species are due to alterations in immune tolerance of cholerae infections are sensitive to metabolite availability (60), and IMD has cell-specific effects on intestinal metabolism, we consider immune-mediated regulation of gut metabolism a candidate mechanism by which IMD influences survival after infection with V. cholerae.
In summary, our data suggest that enterocyte IMD influences intestinal, and systemic, metabolism in the fly.      proteostasis; and longevity that are differentially regulated in esg ts /D30A midguts relative to esg ts /+ midguts.
Positive scores indicated genes that are upregulated in esg ts /D30A midguts and negative scores indicate genes that are downregulated. All numerical values indicate the fold-change on a log2 scale.

Fly Husbandry
We used w 1118 as a wildtype strain. We backcrossed UAS-imdD30A transgenic lines into the w 1118 background for eight generations prior to use, and used standard fly husbandry methods to ensure that our esg ts (esg-GAL4, tub- flies of the correct genotypes in a breeding cage with apple juice agar plates that contained yeast paste over night. The following morning, we collected eggs from the apple juice agar plates after visual inspection to confirm absence of larvae, and sterilized them by washing in a 10% solution of 7.4% sodium hypochlorite for 2.5min twice, followed by a 1min washed with 70% EtOH, and three washes with autoclaved water. Working in a tissue-culture hood, we transferred axenic eggs to the surface of autoclaved standard fly food, and transferred vials to a decontaminated incubator at 21˚C under a 12h:12h light:dark cycle. To confirm that flies were axenic, we regularly 21 sampled the food, by resuspending in MRS (Difco Lactobacilli MRS Agar, DB, 288210) and plating onto MRS-agar plates that we checked for bacterial growth after 2-3 days at 29°C.

Adult infections
For

Lifespan assay
For all longevity studies, we used the GeneSwitch (GS) gene expression system to inhibit IMD activity in defined cell types. This system has the advantage that control and experimental populations have identical background genotypes. To activate the GS system, we added 100µl of RU486 (Mifepristone, M8046, Sigma) dissolved in 80% EtOH (4mg/ml) to the surface of standard food and dried overnight prior to addition of flies. For control flies, we 22 added 100µl of 80% EtOH to the surface of standard food and dried overnight prior to addition of flies. For each treatment group, we placed thirty flies into four vials, for a total of 120 flies per group. All flies were maintained at 25˚C under a 12h:12h light:dark cycle throughout the experiment. We transferred flies to fresh food that contained either RU486, or vehicle, every Monday, Wednesday, and Friday, and counted dead flies at those times.

Establishment of poly-associated gnotobiotic adult Drosophila
We generated axenic flies as described above and starved the flies for 2h prior to bacterial poly-association. To generate poly-associated gnotobiotic adults, we fed axenic flies a mix of three Drosophila symbiotic bacteria strains (Lactobacillus plantarum KP , Lactobacillus brevis EF , and Acetobacter pasteurianus AD ) that were isolated from our wild-type laboratory flies and have described previously (31). Liquid cultures of each bacterial strain were prepared to an OD600 of 50 in 5% sucrose in PBS, then mixed at a 1

Feeding assay
We performed the Capillary Feeder (CAFE) assays as previously described (65). Briefly, we raised virgin females of the appropriate genotypes 29˚C for 7d. We placed ten flies per vial into vials with three capillaries (calibrated Pipets (5ul, VWR, Cat No. 53432-706)) that contained liquid food with 8.1% sucrose (D-sucrose, Fisher BioReagents, BP220-212) and 1.9% yeast extract (BD, 212750). We measured consumption for ten vials for each genotype. At the same time, we maintained a control vial that contained three food-bearing capillaries, but no flies. To quantify food consumption per fly, we calculated (food consumed in vials that contained flies -food consumed in control vial that lacked flies)/number of flies per fly. We measured total consumption every 24h for 3d. For the FlyPad assay, virgin female adults were aged for 23-24 days at 29˚C (25 flies per vial) with weekly renewal of food. Prior to the FlyPad assay, flies were starved on 1% (w/v) agar for 4h at 29˚C. Flies were placed 24 into arenas (one fly per arena) with 3μl of food containing 8.1% sucrose, 1.9% yeast extract, and 1% agar.
Drosophila feeding behavior was monitored for 1 hour on the FlyPad as described previously (66).

Metabolic assay
For metabolic measurements, adult virgin females were raised on standard medium for 10d at 29˚C. We measured the combined weight of five adult flies then mechanically homogenized in TE buffer (10mM Tris, pH 7.4, 1mM EDTA pH 8.0, 0.1% Triton X-100). The homogenates were used for measuring total triglyceride using the Serum TG Determination Kit (TR0100; Sigma) and total glucose by the GAGO Glucose Assay Kit (GAGO20; Sigma) following the manufacturer's instructions. We performed the same assays with 9-10 dissected intestines from adult virgin females. To measure circulating trehalose, 20 adult flies per sample were punctured on the thorax and spun at 9000 rcf for 5min at 4˚C through a filter tube (Zymo Research, C1006-50-F). Hemolymph was mixed in the trehalase buffer (5 mM Tris pH 6.6, 137 mM NaCl, 2.7 mM KCl) at a 1:100 dilution and heated at 70˚C for 5 min. The mixtures divided into two groups, and one was treated with Porcine Kidney Trehalase (T8778-1UN; Sigma), and another group was not. Both groups were incubated at 37˚C for 22h and treated with glucose assay reagent (GAGO20; Sigma) for 30min at 37˚C. To stop the reaction, we added 12N sulfuric acid and measured absorbance at 540nm.

Immunofluorescence
We used previously described immunofluorescence protocols to visualize posterior midguts (56). In brief, we used anti-phospho-histone H3 (PH3) immunofluorescence to quantify mitoses in the midguts, and anti-Delta immunofluorescence to quantify stem cells in the R4/R5 region of the posterior midguts of virgin female flies that we raised at 29˚C for 5d or 30d. The primary antibodies used in this study were as follows Technologies Inc.). Images were collected as z-slices and processed with Fiji software to generate a single z-stacked image.

Bioinformatics
Statistical analyses for metabolic assays, longevity studies, and survival analyses were performed with GraphPad Prism. Survival data and longevity data were analyzed with Log-rank (Mantal-Cox) tests. Feeding assays, CFU, quantification of PH-3 or Delta positive cells, and metabolic assays were analyzed with unpaired Student's ttests. Survival, and longevity curves were generated by GraphPad Prism and the R programming language was used for the remaining figures. All figures were assembled using Adobe Illustrator. We used QIIME2-2019.1 (67) for all analysis of 16S data, filtering all sequences that were present below a minimum frequency of 200, and using the DADA2 package for sequence quality control. We binned sequences at 99% sequence identity, and used the greengenes database to identify operational taxonomic units. Differential abundance analysis was performed using the gneiss (68) plugin for QIIME2. RNAseq was performed with RNA purified by TRIZOL (ambion, 15596-026) purification from the midguts of adult flies that we raised at 29°C for 10d. Purified RNA was sent on dry ice to Novogene (CA, USA) for library construction and sequencing with Illumina Platform PE150. For RNAseq studies, we obtained approximately 40 million reads per biological replicate. We used FASTQC (http://www.bioinformatics.bbsrc.ac.uk/projects/fastqc, version 0.11.3) to evaluate the quality of raw, pairedend reads, and trimmed adaptors and reads of less than 36 base pairs in length from the raw reads using Trimmomatic version 0.36 (69). We used HISAT2 version 2.1.0 (70) to align reads to the Drosophila transcriptome-bdgp6, and converted the resulting BAM files to SAM flies using Samtools version 1.8 (71). We counted converted files using Rsubread version 1.24.2 (72) and loaded them into EdgeR (73). In EdgeR, we filtered genes with counts less than 1 count per million and normalized libraries for size. Normalized libraries were used to call genes that were differentially expressed among treatments. Genes with P-value < 0.01 and FDR < 0.01 were defined as differentially expressed genes. Principle component analysis was performed on normalized libraries using