The Immune Deficiency Pathway Regulates Metabolic Homeostasis in Drosophila

The gastrointestinal tract processes ingested material in a manner that prevents microbial penetration of the host interior and allows an orderly flow of essential nutrients to metabolic organs. Once inside the host, nutrients initiate signal transduction pathways that control growth and development. Prominent metabolic regulators appeared very early during animal evolution—an estimated billion years ago in the case of the insulin peptides (1)—and execute conserved functions across the animal kingdom. For example, insulin peptides control the uptake and storage of nutrients, activate cellular growth, and influence longevity in animals as diverse as worms, flies, and rodents (2–4).

Metabolism undergoes a fundamental shift upon infection (5). At this point, germline-encoded pattern recognition receptors detect the molecular signatures of alien microbes and initiate physiological responses designed to neutralize the invader and optimize host survival. We tend to focus on immunity as the generation of molecules that eliminate the invader and eradicate infected cells. However, microbial detection initiates a complex spectrum of responses that may include elements as diverse as increased body temperature, lethargy, loss of appetite, social isolation, and tolerance mechanisms that neutralize pathogens without affecting their numbers (6). Metabolic adaptations are a common theme in the host response to infection (7). In this case, hosts balance traditional metabolic needs against the immediate threat presented by the microbe and alter metabolic pathway activity accordingly.

The integration of immune and metabolic pathways is particularly apparent in insects such as Drosophila melanogaster, in which the fat body simultaneously regulates energy storage and humoral immunity. Under optimal conditions, the larval fat body detects circulating sugars and amino acids in the hemolymph to control the release of Drosophila insulin-like peptides (dILPs) from the brain (8). dILPs enter circulation and orchestrate the actions of metabolic organs such as muscle and fat. At the same time, pattern recognition receptors survey the hemolymph for microbe-associated molecular patterns that indicate infection. Several types of microbe-associated molecular patterns activate Toll-mediated responses in the fat body (9), whereas the immune deficiency (IMD) pathway, an evolutionary relative of the TNF pathway (10), responds to bacterial diaminopimelic acid–containing peptidoglycan (11). The host integrates signals from immune and metabolic pathways to determine the net output of the fat body. For example, in times of high nutrient availability and limited microbial detection, the Drosophila MEF2 transcription factor is phosphorylated and promotes lipogenesis and glycogenesis, molecular pathways that support growth in the animal (12). However, bacterial infection causes a loss of MEF2 phosphorylation, an event that shifts fat body activity from the accumulation of energy stores to the release of antimicrobial peptides. Such metabolic shifts are common in Drosophila responses to infection and frequently include alterations to the activity of insulin/target of rapamycin (TOR) pathway elements (13, 14).

Molecular links between immune and metabolic pathways are conserved across vast evolutionary distances, and abnormal immune–metabolic signals are linked to several pathological states. For example, inflammation is involved in the development of chronic metabolic disorders, such as insulin resistance and type two diabetes (15, 16). In experimental models of obesity, adipose tissue–resident macrophages produce TNF (17, 18), and TNF contributes to the development of obesity-induced insulin resistance (19–21). Indeed, treatment with anti-inflammatory salicylates improves obesity-induced insulin resistance and type two diabetes (22, 23). However, despite the impact of inflammatory cues on metabolic homeostasis, we do not fully understand how the respective pathways communicate.

We used the Drosophila model to characterize the contributions of IMD to immune–metabolic homeostasis. We found that activation of IMD in the fat body has the molecular, genetic, and phenotypic signatures of alterations to host metabolism. Transcriptionally, activation of IMD resulted in a gene expression signature consistent with diminished insulin/TOR activity. Physiologically, IMD activation caused a depletion of lipid stores, hyperglycemia, delayed development, and a reduction in adult size. In follow-up studies, we found that loss-of-function imd mutants weigh more and have deregulated insulin signaling, hyperlipidemia, and impaired glucose tolerance. The apparent links between IMD and metabolism led us to speculate that loss of key of metabolic regulators, such as insulin pathway components, will have a measurable impact on the ability of Drosophila to survive microbial infection. To test this hypothesis, we determined the impact of common fly pathogens on the survival of wild-type or insulin pathway mutant flies. We found that mutations in the insulin pathway significantly impacted host survival and bacterial loads in a manner that depended on the route of infection and the identity of the infectious microbe. Our results support a model in which integrated immune–metabolic activity is critical for host responses to microbial infection.

Adult flies and larvae were raised on standard corn meal medium (Nutri-Fly Bloomington formulation, https://bdsc.indiana.edu/information/recipes/bloomfood.html; Genesse Scientific). All adult experiments were performed using virgin male and female flies. For experiments using flies maintained on the holidic diet, the holidic medium was prepared following the published protocol and recipe using the original amino acid solution (Oaa) at 100 mM biologically available nitrogen (24). The sugar/yeast (SY) diet consists of 0.15 M sucrose, 100 g/l (w/v) yeast extract, and 1% (w/v) agar. The high-SY version with elevated sucrose levels (SYS) diet consists of 1 M sucrose, 100 g/l (w/v) yeast extract, and 1% (w/v) agar. Flies that were used in this study are as follows: w¹¹¹⁸, R4-GAL4, UAS-ImdCA, impl2^def2, ilp2,3,5, Ilp2HF, and imd ^EY08573 (null) mutants. The imd mutants used in this study were back-crossed to the w¹¹¹⁸ flies for eight generations prior to use. To measure developmental rates, 25 age-matched feeding third instar larvae were cultured at 25°C and monitored for the formation of wandering third instar larvae, pupae, and eclosed adults. For pupariation timing, 25 age-matched third instar larvae were cultured at 25°C and monitored for the length of time required for development to the P13 pupal stage. Developmental and pupariation assays were performed in quadruplicate. For total triglyceride (TG) measurement, 10 third instar larvae or 5 adult flies were weighed and homogenized in TE buffer with 0.1% Triton X-100. TG content was measured in larval homogenate using the serum TG determination kit (TR0100; Sigma) according to the manufacturer’s instructions. Total glucose was measured by homogenizing 10 third instar larvae or 5 adult flies in TE buffer and measuring glucose using the GAGO glucose assay kit (GAGO20; Sigma) according to the manufacturer’s instructions. For trehalose hemolymph measurements, groups of 15 third instar larvae were dipped in halocarbon oil 700 (Sigma), and the epidermis was punctured to start hemolymph bleeding. Accumulated hemolymph on the oil drop was aspirated using a glass pipette and immediately frozen on dry ice. One microliter of hemolymph was mixed with 99 μl of trehalase buffer (5 mM Tris pH 6.6, 137 mM NaCl, 2.7 mM KCl) and heated at 70°C for 5 min to inactivate endogenous trehalase. The samples were treated with or without Porcine Kidney Trehalase (T8778-1UN; Sigma) and incubated at 37°C for 16 h, and then the reaction was started by adding glucose assay reagent (GAGO20; Sigma), incubated at 37°C for 30 min, and stopped by adding 12 N sulfuric acid. Absorbance was measured at 540 nm. To calculate trehalose levels, we subtracted glucose levels in untreated samples from glucose levels of samples that were treated with trehalase. Capillary feeder (CAFE) assays were performed as described previously (25). We delivered previously mentioned holidic liquid food by leaving out the agar to the capillaries. Each vial contained three capillaries with 10 adult flies. Total consumption was calculated every 24 h for 5 d. For Nile red staining, 10 third instar larvae were dissected in PBS and fixed in 4% formaldehyde for 30 min. After twice washing with 1× PBS, fat tissues were stained with 1:1000 of a Nile red stock (0.5 mg/ml in acetone) and 1:500 of Hoechst 33258 for 30 min. Stained tissue was mounted on slides and visualized using a spinning disc confocal microscope (Quorum WaveFX). Lipid area was quantified with Columbus software (PerkinElmer). Pupal volume was calculated as previously described (26). In brief, 24 h after egg laying, larvae were collected and put into food vials in groups of 50 larvae. Using a paintbrush, 1-d-old pupae were picked off the side of the vial. Pupae were imaged using a Zeiss Stereo Discovery V8 microscope using a ×14 magnification. AxioVision software was used to measure the length and width of each pupae. Pupal volume was calculated with the assumption that the pupae are cylindrical using the formula (4/3π) × (length/2) × (diameter/2)².

We acquired the FlyPAD instrument from Dr. P.M. Itskov (27). We raised male w¹¹¹⁸ and imd flies on a holidic diet for 20 d. For the FlyPAD experiment, we used a holidic medium with agarose substituted for the agar. Prepared food was melted at 95°C and then maintained at 60°C to facilitate pouring. Individual flies were placed in each FlyPAD arena with a mouth aspirator at n = 32 for each genotype. Eating behavior was recorded for 1 h.

w¹¹¹⁸ and imd males were starved overnight for 16 h on 1% agar, switched to vials containing 10% glucose and 1% agar for 2 h, and then restarved on vials of 1% agar. Samples of five flies were obtained after initial starvation, after 2 h on 10% glucose, and then at both 2 and 4 h following restarvation. Samples of five flies were weighed and then mashed in 125 μl TE of buffer (10 mM Tris, 1 mM EDTA, 0.1% Triton X-100, pH 7.4). Glucose was measured using the Glucose Oxidase (GO) Assay Kit (GAGO20; Sigma).

To measure circulating and total ILP2 levels, we used the ilp2¹ gd2HF fly stock and protocols acquired from Dr. S.K. Kim (28). For sample preparation, we dissected the black posterior end of the abdomen away and transferred 10 dissected male bodies to 60 μl of PBS, followed by a 10-min vortex at maximum speed. We centrifuged these tubes at 1000 × g for 1 min and transferred 50 μl of the supernatant to a PCR tube as our circulating ILP2HF sample. We added 500 μl of PBS with 1% Triton X-100 to the tubes with the remaining flies and mashed the samples using a pestle and cordless motor (VWR 47747-370), followed by a 5-min vortex at maximum speed. We centrifuged these tubes at maximum speed for 5 min and then transferred 50 μl of the supernatant to a PCR tube as our total ILP2HF sample. For standards, we used FLAG(GS)HA peptide standards (DYKDDDDKGGGGSYPYDVPDYA amide, 2412 d; LifeTein). We added 1 μl of the stock peptide standards (0–10 ng/ml) to 50 μl of PBS or PBS with 1% Triton X-100. We coated wells of a Nunc MaxiSorp plate (44-2404-21; Thermo Fisher Scientific) with 100 μl of anti-FLAG Ab diluted in 0.2 M sodium carbonate/bicarbonate buffer (pH 9.4) to 2.5 μg/ml and then incubated the plate at 4°C overnight. The plate was washed twice with PBS with 0.2% Tween 20 and then blocked with 350 μl of 2% BSA in PBS at 4°C overnight. We diluted Anti-HA–Peroxidase, High Affinity (clone 3F10, 25 μg/ml, no. 12013819001; Roche) in PBS with 2% Tween at a 1:500 dilution. We then added 5 μl of the diluted anti-HA–peroxidase to the PCR tubes containing 50 μl of samples or standards, vortexed, and centrifuged briefly. Following blocking, we washed the plate three times with PBS with 0.2% Tween 20. Samples and standards were transferred to wells of the plate, and the plate was sealed with adhesive sealer (MSB-1001; Bio-Rad) and then placed in a humid chamber at 4°C overnight. Samples were removed with an aspirator, and the plate was washed with PBS with 0.2% Tween 20 six times. We added 100 μl of 1-Step Ultra TMB-ELISA Substrate (no. 34028; Thermo Fisher Scientific) to each well and incubated at room temperature for 30 min. The reaction was stopped by adding 100 μl of 2 M sulfuric acid, and absorbance was measured at 450 nm on a SpectraMax M5 (Molecular Devices).

For microarray studies, we used the GeneChip Drosophila Genome 2.0 Array (Affymetrix) to measure gene expression in triplicate assays. Total RNA was extracted from third instar larvae using TRIzol. We used 100 ng of purified RNA to make labeled cRNA using the GeneChip 3′ IVT Plus Reagent Kit (Affymetrix). We used Transcriptome Analysis Console software (Affymetrix) for preliminary analysis of gene expression data. Array data have been submitted to the National Center for Biotechnology Information Gene Expression Omnibus database (accession number GSE109470, http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE109470). Transcriptome data from R4/ImdCA relative to R4GAL4/+ (R4/+) larvae were analyzed using Gene Set Enrichment Analysis (29) to identify KEGG pathways that were differentially regulated upon activation of IMD. The data from the GSEA analysis were then visualized using the EnrichmentMap plugin in Cytoscape (version 3.6.1) to generate the gene interaction network (30). The resulting network map was curated to remove uninformative nodes, resulting in the simplified network shown in Fig. 1A. We used Panther (31) to identify biological process that were affected by IMD activation and FlyMine (32) to determine tissue enrichment of the respective genes in third instar larvae. GOrilla (Gene Ontology Enrichment Analysis and Visualization Tool) was used to identify biological processes influenced by R4/ImdCA (33). From the transcriptome data, two gene lists were created that contained significantly upregulated or downregulated genes in response to ImdCA. Each of these lists was run in GOrilla against the background gene set (all microarray genes) with a p-value cutoff of 10⁻⁴. The top 15 Gene Ontology (GO) terms sorted by p value were selected for both upregulated and downregulated analyses, ranked by enrichment score, and visualized using the easyggplot2 package in R (version 1.1.442).

For infection experiments, the following bacteria were used: Providencia sneebia, P. rettgeri, Enterococcus faecalis, Serratia marcescens DB 11, and Vibrio cholerae (C6706 strain). For oral infections, all bacteria except E. faecalis were streaked from glycerol stocks onto lysogeny broth (LB)-agar plates and grown overnight at 37°C. E. faecalis was streaked from glycerol stocks onto brain heart infusion (BHI) plates and grown overnight at 37°C. The following day, we grew single colonies in medium to an OD₆₀₀ of 0.245 and soaked a sterile cotton plug with 3 ml of the bacterial culture in LB or BHI medium (for E. faecalis). Six- to seven-day-old virgin female flies were fed on the cotton plug, and death was recorded at the indicated time points. We used a cotton plug soaked with LB medium or BHI medium (for E. faecalis) for our control in oral infection experiments. For bacterial load quantification, at indicated time points 25 flies from five biological replicates (five flies from each biological replicate) were collected and surface-sterilized by rinsing in 20% bleach, 70% EtOH, and distilled water. Then, we randomly distributed these 25 flies into five groups (five flies in each 1.5-ml tube) and then homogenized in respective media. Serial dilutions of fly homogenates were made in 96-well plates, and 10 μl of spots were plated on LB agar supplemented with 100 μg/ml streptomycin (to select for V. cholerae), BHI agar (to select for E. faecalis), and LB agar for the rest of the bacteria. For calculating CFUs per fly, CFU/ml calculated for each bacterial culture was divided by five. To normalize the CFUs for weight, the CFUs for ilp2,3,5 flies were divided by the ratio of the average weights of w¹¹¹⁸ and ilp2,3,5 mutants. For septic infection, 0.15-mm minutin pins (Fine Science Tools) were dipped into the OD₆₀₀ = 1 dilution of bacterial, which were grown overnight in media at 37°C and then pricked into the thorax of 6–7-d-old virgin female flies. A sterile 0.15-mm minutin pin was used to prick flies in the thorax and served as control. Flies were then transferred to normal food and kept at 29°C for the rest of the experiments.

Quantitative PCR (qPCR) measurements were performed with RNA purified from whole larvae using TRIzol, and the ΔΔ cycle threshold method was used to calculate relative expression values. For adult fly qPCR measurements, 10 heads were homogenized in TRIzol. Gene expressions were normalized to actin. The following primers were used in this study: wisp (forward [F]: 5′-CAACAACAGTCACTCGTGGG-3′, reverse [R]: 5′-TGGAAGAACGAAGATGGTTGC-3′), pathetic (F: 5′-TACTACAGAACTCGCCGCAC-3′, R: 5′-CAGACCAAACAGGATGGAGAAC-3′), odc1 (F: 5′-ATCTGCGACCTGTCTAGCGT-3′, R: 5′-CATTGGATCGTCATTGCACTTG-3′), tep1 (F: 5′-AGTCCCATAAAGGCCGACTGA-3′, R: 5′-CACCTGCATCAAAGCCATATTG-3′), tsf1 (F: 5′-CGATTGTGTGGTGGCTCTGACCAAG-3′, R: 5′-AAGGACATCATCCTGAGCCCTCTGC-3′), diptericin (F: 5′-ACCGCAGTACCCACTCAATC-3′, R: 5′-ACTTTCCAGCTCGGTTCTGA-3′), dilp2 (F: 5′-TCCACAGTGAAGTTGGCCC-3′, R: 5′-AGA TAATCGCGTCGACCAGG-3′), dilp3 (F: 5′-AGAGAACTTTGGACCCCGTGAA-3′, R: 5′-TGAACC GAACTATCACTCAACAGTCT-3′), dilp5 (F: 5′-GAGGCACCTTGGGCCTATTC-3′, R: 5′-CATGTG GTGAGATTCGGAGCTA-3′), and actin (F: 5′-TGCCTCATCGCCGACATAA-3′, R: 5′-CACGTCACCAGGGCGTAAT-3′). For Western blots, larvae were lysed in lysis buffer (20 mM Tris-HCl [pH 8], 137 mM NaCl, 1 mM EDTA, 25% glycerol, 1% NP-40, 50 mM NaF, 1 mM PMSF, 1 mM DTT, 5 mM Na₃VO₄, protease inhibitor mixture [cat. no. 04693124001; Roche], and phosphatase inhibitor [cat. no. 04906845001; Roche]), and protein concentrations were measured using the Bio-Rad DC Protein Assay Kit II. For each experiment, equal amounts of protein lysates (usually 15–40 μg) were subjected to Western blot analysis. Primary Abs used were anti–α-tubulin (α-tubulin E7; Drosophila Studies Hybridoma Bank), anti–phospho-Drosophila Akt Ser⁵⁰⁵ (no. 4054; Cell Signaling Technology), and anti–phospho-S6K Thr³⁹⁸ (no. 9209; Cell Signaling Technology). For immunoblot quantifications, the area under each peak, subtracting the background, was quantified. The p-Akt was normalized to total Akt, and the p-S6K was normalized to tubulin.

All statistical analyses were performed with GraphPad Prism. qPCR data were analyzed with unpaired Student t tests (p < 0.05). Survival data were analyzed with log-rank (Mantel–Cox) tests. For pupariation timing and pupae counting, Kolmogorov–Smirnov tests and unpaired Student t tests were used (p < 0.05), respectively. Pupal volumes were compared with unpaired Student t tests (p < 0.05). For analyzing the bacterial load difference, we used one-way ANOVA with Sidak correction.

Chronic inflammation is a hallmark of metabolic disorders such as type two diabetes. However, we do not fully understand the extent to which host immunity controls metabolic homeostasis. To directly examine the effects of persistent immune signaling on a metabolic organ, we used the R4GAL4 driver line to express a constitutively active IMD (ImdCA) construct exclusively in the fat body (R4/imdCA). This approach allowed us to ask how persistent immune activity influences host physiology without collateral damage through the introduction of pathogenic microbes.

Initially, we used whole-genome microarrays to compare the transcription profiles of R4/imdCA larvae to control, age-matched R4/+ larvae. Activation of IMD deregulated the expression of 1188 genes in third instar larvae by a factor of 1.5 or more (Supplemental Table I). We confirmed deregulated expression for six representative genes in subsequent qPCR assays (Fig. 1D). As expected, many response genes, such as antimicrobial peptides, have established roles in the elimination of microbial invaders (Fig. 1A, 1B). However, we also noted substantial effects of IMD activation on the expression of genes that control metabolism (Fig. 1A–C, Supplemental Table I). Of the 807 IMD response genes with annotated biological functions, 247 are classified as regulators of metabolic processes (Supplemental Table I). For example, activation of IMD diminished expression of TOR pathway genes (Fig. 1A), decreased expression of dilp3 (Fig. 1C), increased expression of the insulin pathway antagonists dilp6 and impl2 (Fig. 1C), and elevated expression of the FOXO-responsive transcripts thor and tobi (Fig. 1C). Consistent with effects of IMD activation on host metabolism, we observed significant reduction in the expression of enzymes involved in glycolysis, the TCA cycle, mitochondrial ATP production, and fatty acid β oxidation (Fig. 1A, 1B, Supplemental Table I). We also noted that host responses to IMD activation extend beyond the fat body, as IMD activation suppressed the expression of intestinal peptidases and chitin-binding proteins and lowered the expression of hormone signaling molecules in the salivary glands (Supplemental Table I). We previously determined the consequences of imdCA expression in adult intestinal progenitor cells (34). This allowed us to compare host responses to IMD activation in the fat body, the principal regulator of humoral immunity, to immune activation in the intestine, a first line of defense against oral infection. Interestingly, we observed minimal overlap between the two responses (Supplemental Fig. 1A–C).

Only 9.8% of genes affected by IMD activation in the fat body were affected by IMD activation in intestinal progenitors (Supplemental Fig. 1A). These observations suggest broad tissue autonomy in IMD responses. In contrast, 29.8% of genes affected by IMD activation are similarly affected by loss of the insulin receptor in fat tissue (35) (Supplemental Fig. 1D, 1E), suggesting an overlap between IMD and insulin-dependent transcriptional outputs in the fat body. This possibility is further supported by a recent examination of the fly transcriptional response to systemic infection with 10 distinct bacteria (36). Similar to ImdCA expression, bacterial infection modified the expression of a large number of host genes involved in the regulation of metabolism (Supplemental Fig. 2A, Supplemental Table II). Combined, these data sets support a model for the coordinated expression of immune and metabolism regulatory genes in the fat body.

To test the possibility that IMD activation impacts metabolic homeostasis, we measured carbohydrate levels in R4/imdCA and R4/+ larvae. We did not detect obvious effects of IMD on total glucose levels in larvae (Fig. 2B). However, we found that activation of IMD resulted in significantly higher levels of trehalose, the primary form of circulating carbohydrate in Drosophila (Fig. 2A). On average, trehalose levels were approximately twice as high in R4/imdCA larvae as in R4/+ larvae (Fig. 2A). We then examined the effects of IMD activation on TG stores. Drosophila larvae store TG in large lipid droplets in the fat body. The lipid storage droplet 1 (Lsd1) and 2 (Lsd2) proteins are involved in lipid storage and lipolysis control, respectively (37), and we found that activation of IMD suppressed expression of both (Fig. 1C). To determine if IMD activation affects TG stores, we measured total TG levels and lipid droplet size in third instar R4/imdCA and R4/+ larvae. We found that activation of IMD decreased total TG levels by ∼50% (Fig. 2C) and caused a significant drop in the total volume of lipid stores in the fat body (Fig. 2D, 2E). Together, these results demonstrate that persistent activation of IMD in the fat body causes hyperglycemia and a depletion of lipid stores.

As nutrient availability influences larval development (38), we tested the effects of IMD activation on larval growth. We were specifically interested in the length of time to pupariation, the size of pupae, and the rate of pupal eclosion, as each of these factors is sensitive to metabolite availability. In each assay, we noted significant effects of IMD activation on development. Specifically, IMD activation delayed the average duration of development from feeding third instar larvae to the P13 stage of pupal development by approximately 18 h (Fig. 3A), led to a roughly 10% drop in pupal volume (Fig. 3B), and caused a significant reduction in adult eclosion rates compared with R4/+ controls (Fig. 3C). In Drosophila, dietary nutrients promote cellular and organismal growth by activation of insulin-PI3Kinase-AKT and TOR-S6 kinase pathways (39). To determine if IMD affects insulin/TOR signaling, we measured the extent of S6 kinase and AKT phosphorylation in R4/ImdCA and R4/+ larvae in four biological replicates. Here, we noticed a significant reduction in the phosphorylation of S6 kinase and AKT in R4/ImdCA larvae relative to R4/+ controls (Fig. 3D, 3E). Combined, these data indicate that activation of IMD in the fat body disrupts metabolism in Drosophila.

To determine if IMD contributes to metabolic homeostasis in the absence of an activating signal, we measured insulin expression, body weight, glucose, and TG levels in imd mutant adult flies raised for 10 d on an SY mix, and on an SYS version. In these assays, we noticed sex-dependent effects of IMD on host responses to the respective foods. In males, we observed significantly less expression of dilp2, dilp3, and dilp5 in imd mutants relative to wild-type controls raised on the SY diet and significantly less dilp3 expression in imd males raised on the SYS diet (Fig. 4A–C). Furthermore, whereas wild-type males had lower total body weight (Fig. 4E) and higher TG on SYS food relative to wild-type males raised on SY food (Fig. 4G), we did not see any effects of food change on the weight or TG levels of imd mutants. Instead, we observed a significant increase in glucose levels in imd mutants relative to wild-type controls raised on SYS (Fig. 4F). In contrast to wild-type controls, imd mutant females displayed a significant drop in dilp5 expression levels in flies raised on SYS food relative to flies raised on SY food (Fig. 4J). Likewise, we noticed elevated body weight at day 0 (Fig. 4K) and a significant drop in glucose levels of mutants raised on the SY diet in comparison with the wild-type controls (Fig. 4M).

To follow the effects of IMD on metabolism more closely, we monitored metabolic activity in adult males raised on a previously described holidic food (24) for 20 d. This food provides all nutrients needed to sustain adult life and allows investigators to monitor host physiology on a chemically defined food. We found that mutation of imd significantly impairs the expression of dilp3 (Fig. 5B), diminishes total dILP2 levels (Fig. 5D), and increases the amount of circulating dILP2 (Fig. 5E). As insulin is an essential regulator of glucose metabolism, we determined the effect of imd mutation on the ability of 1- or 20-d-old flies to tolerate a glucose meal after a period of fasting. In both cases, we detected significantly higher levels of glucose in mutants immediately upon conclusion of the fast and during the first 2 h after feeding (Fig. 5F, 5G). Furthermore, we found that imd mutants weighed more (Fig. 5H) and had elevated glucose (Fig. 5I) and TG levels (Fig. 5J) relative to wild-type controls. As imd mutants weigh more, we asked if IMD influenced feeding in adult males. To address this question, we raised adult males on holidic food for 20 d and then quantified their consumption of solid holidic food in a FlyPAD assay or liquid holidic food in a CAFE assay. Although we observed an increase in total feeding bouts on solid food (Fig. 5K), mutation of imd did not have a significant effect on the total number of sips (Fig. 5l) or on the consumption of liquid food (Fig. 5N), suggesting that IMD does not influence food consumption. In short, our data uncover a wide range of effects of IMD on insulin, metabolism, and energy storage in adult flies raised on three distinct foods, supporting a role for IMD in the maintenance of metabolic homeostasis.

We showed that IMD impacts metabolic homeostasis in both larvae and adult flies. However, it is not clear if metabolic deregulation influences the ability of Drosophila to combat bacterial pathogens. As insulin is one of the principal regulators of metabolic homeostasis, we examined the immune responses of ilp2,3,5 mutant flies to oral and septic challenges with a panel of bacteria that range from low to high pathogenicity in Drosophila infection models. For septic infections, we pricked flies in the thorax with a fine needle that deposits the bacteria into the body cavity. As a control, we monitored the survival of flies that we pricked with an uncontaminated, sterile needle. We measured survival and bacterial load as indicators of pathogenicity in our study. In each assay, we found that a sterile wound had minimal effects on the viability of ilp2,3,5 mutants or of wild-type flies (Fig. 6A, 6D, 6G, 6K). In contrast, we found that ilp2,3,5 mutants showed different response toward each pathogen. Systemic infection with P. sneebia, a microbe that fails to activate IMD (40), did not result in a significant difference between survival of ilp2,3,5 mutants and control w¹¹¹⁸ flies (Fig. 6A). In contrast, the survival of ilp2,3,5 mutants infected with P. rettgeri, S. marcescens Db11, or E. faecalis was significantly reduced compared to the survival of control w¹¹¹⁸ flies (Fig. 6D, 6G, 6K). As ilp2,3,5 mutants weigh significantly less than wild-type flies (Supplemental Fig. 2B), we determined the bacterial load per fly and per fly normalized to weight in the respective infection assays. There was no significant difference between bacterial load in ilp2,3,5 mutants and w¹¹¹⁸ flies 6 or 12 h postinfection with P. sneebia before or after we corrected bacterial load for host weight (Fig. 6B, 6C). In contrast, the bacterial load at 12 h of septic infection with P. rettgeri and E. faecalis was significantly higher for ilp2,3,5 mutants (Fig. 6E, 6H). When we corrected bacterial load for host weight, the difference between ilp2,3,5 mutants and w¹¹¹⁸ flies remained unchanged (Fig. 6F, 6I). Thus, loss of insulin has microbe-dependent effects on bacterial pathogenicity in Drosophila.

For oral infection experiments, we raised 6–7-d-old virgin female flies on cotton plugs soaked with bacterial pathogens or with LB medium. Bacterial medium alone had minimal effects on fly viability for the first 4 d (Fig. 7A, 7D, 7G, 7J). In this study, we found that, with the exception of E. faecalis (Fig. 7J), loss of insulin had bacteria-dependent effects on CFUs and host survival. For example, ilp2,3,5 mutants showed a reduced survival after oral infection with P. sneebia, whereas there was no significant difference between bacterial load in insulin mutant and control flies (Fig. 7A–C). In contrast, we did not observe a significant difference between the survival of ilp2,3,5 mutants and w¹¹¹⁸ flies after oral infection with P. rettgeri (Fig. 7D). The bacterial load of ilp2,3,5 mutants infected with P. rettgeri was significantly lower compared with control flies after 48 h of infection; however, this difference was not significant once we normalized for the weight of ilp2,3,5 mutants (Fig. 7E, 7F). ilp2,3,5 mutants lived significantly longer than w¹¹¹⁸ flies after oral infection with S. marcescens Db11 (Fig. 7G), and the bacterial load in ilp2,3,5 mutants was lower than in control flies after 48 h of infection (Fig. 7H). However, upon correction for weight, we found approximately equal numbers of S. marcescens Db11 in wild-type and insulin mutant flies (Fig. 7I). Thus, it appears that, controlling for weight, bacterial load is not affected by insulin mutation for each microbe tested. Nonetheless, insulin mutation has distinct effects on the ability of the host to survive bacterial challenges. In combination, these observations implicate insulin signaling in the regulation of Drosophila immunity to bacterial infection.

It has been reported previously that enteric infection of w¹¹¹⁸ flies with V. cholerae suppresses insulin signaling (41). To determine if insulin is important for host immunity to V. cholerae, we measured host survival and bacterial loads in Drosophila that we infected with V. cholerae. In this study, we found that ilp2,3,5 mutants have a significantly different survival response toward V. cholerae through oral or septic infection. Oral infection with V. cholerae resulted in a significantly improved survival of ilp2,3,5 mutants compared with w¹¹¹⁸ flies (Fig. 8A). Improved viability was accompanied by reduced bacterial load in ilp2,3,5 mutants 48 h postinfection (Fig. 8B). After we normalized the CFUs for host weight, there was still a reduced bacterial load in insulin mutants compared with control flies (Fig. 8C). In contrast, systemic infection with V. cholerae led to a significantly reduced survival in ilp2,3,5 mutants (Fig. 8G).

As insulin deficiency improves host survival after oral challenge with V. cholerae, we then asked what effect increased insulin activity will have on the immune response to V. cholerae. To answer this question, we examined the immune response of impl2^def20 mutants to V. cholerae. Impl2 is an antagonist of insulin signaling, and impl2^def20 mutants contain an amorphic mutation that results in an increased insulin signaling activity. As impl2^def20 flies weigh slightly more than w¹¹¹⁸ flies (Supplemental Fig. 2C), we also corrected total CFUs for weight in these assays. We found that oral infection of impl2^def20 flies with V. cholerae results in a significantly reduced survival compared with wild-type counterparts (Fig. 8D). The bacterial load in flies with increased insulin signaling was significantly higher compared with control flies (Fig. 8E), and normalization of CFUs for the weight of impl2^def20 flies showed an increased bacterial load in impl2^def20 flies compared with w¹¹¹⁸ (Fig. 8F). In combination, our data indicate that insulin signaling regulates host immunity to oral infection with V. cholerae.

We have summarized the results of all infection assays as a heat map in Fig. 8H. Dark blue indicates challenges in which ilp2,3,5 mutants have improved survival rates or diminished bacterial loads, and red indicates challenges in which mutants have an impaired survival or elevated bacterial loads. These data reveal that loss of insulin affects host survival and bacterial load in flies challenged with a range of bacterial pathogens. However, the magnitude and nature of the effect depend on the identity of the infectious microbe and the route of bacterial introduction to the fly.

Eukaryotic life emerged in an environment dominated by microbes and unpredictable availability of nutrients. In response to these challenges, eukaryotes evolved growth and defense responses that support the complexities of multicellular existence. Both responses act in concert to sustain homeostasis, and disruptions to one pathway often impact the other (42). This relationship is particularly evident in insects such as Drosophila that rely on their fat body to simultaneously coordinate humoral immunity and nutrient utilization. The fat body detects the nutritional and microbial content of the hemolymph and dictates systemic responses designed to maximize host viability. These dual functions require an integrated response that facilitates the allocation of resources to support growth, reproduction, or antimicrobial responses as needed. The situation is more complex in higher vertebrates, in which distinct tissues execute metabolic and immune duties. Nonetheless, the organs in question remain in close communication, and metabolic reprogramming is a critical aspect of immune activation in vertebrate lymphocytes (43). Obesity results in recruitment of macrophages to adipose tissue (17, 18) and induces expression of the TNF-α adipokine in adipocytes, with consequences for metabolism and inflammation (21). We used Drosophila to examine the relationship between IMD activity and metabolic homeostasis. The IMD and TNF signal transduction pathways are closely related, and recent genomic studies implicated IMD in the regulation of metabolism (34, 44, 45). We showed that experimental manipulation of IMD has a substantial effect on host metabolism. Persistent activation of IMD depletes fat reserves, causes hyperglycemia, delays development, and impairs larval growth. In contrast, loss of IMD activity leads to weight gain, elevated storage of TG and glucose, and impaired glucose tolerance. In total, these observations suggest that IMD influences metabolic regulatory pathways. In support of this model, a recent study showed that systemic infection with 10 distinct bacterial pathogens has immediate effects on the expression of genes involved in the control of host metabolism (36).

The fat body communicates with the brain through mediators such as CCHamide2 (46), the TGF/activin-like ligand dawdle (47), the leptin-like cytokine Upd2 (48), dILP6 (49), and stunted (26) to maintain metabolic homeostasis in Drosophila. For example, inhibition of upd2 expression in the Drosophila fat body prevents release of ILPs into the hemolymph, negatively affecting systemic growth and energy metabolism (48). Our data suggest that IMD activation in the fat body affects the cross-talk between fat body and brain through deregulation of systemic insulin signaling. This hypothesis is consistent with the phenotypic overlaps between larvae with elevated IMD activity in the fat body and insulin-deficient flies. However, the exact molecular basis by which the IMD pathway contributes to the regulation of the fat–brain axis needs to be further investigated.

Our finding that IMD affects metabolism corroborates a report that the Drosophila TNF-α homolog Eiger regulates production of insulin peptides in the brain (50) and that the FOXO homolog Forkhead regulates intestinal metabolism and survival postinfection in adult Drosophila (51). In addition, several studies identified interaction points between immune and insulin responses in the fly. For example, depletion of the insulin receptor from the fat body alters the expression of immune response genes and alters sensitivity to infection (35). Furthermore, mutations of the IRS homolog chico increase survival after infection with Pseudomonas aeruginosa and E. faecalis (52), challenges with Mycobacterium marinum lower AKT phosphorylation and diminish systemic insulin activity (14), activation of TOR blocks AMP expression (53), infection increases expression of the FOXO-responsive transcript 4E-BP ortholog thor (54), and FOXO regulates the expression of intestinal antimicrobial peptides (55). Combined, these findings suggest a direct relationship between bacterial challenges and insulin-sensitive pathways in the fly. This hypothesis is supported by observations that infection with M. marinum lowers TG stores and increases the amount of circulating sugar (14), whereas challenges with Listeria monocytogenes lower TG and glycogen stores and inhibit glycolysis (13), and that starvation or protein restriction increases the expression of antimicrobial peptides (56, 57). An earlier study established that activation of the IMD-responsive NF-κB family member Relish fails to affect insulin signaling in the fly (58), suggesting that IMD acts downstream of Relish as a metabolic regulator. We consider the IMD-sensitive JNK a likely nexus of immune–metabolic control, as several reports link JNK and insulin activity in Drosophila (59, 60).

Connections between insulin and immune activity appear conserved through evolution. In Caenorhabditis elegans, mutations in the insulin receptor homolog age-1, the PI3-kinase homolog daf-2, or the FOXO homolog daf-16 impact survival after bacterial infection (61, 62). Furthermore, protein restriction improves survival against malaria infections in mice (63), whereas TNF-α regulates glucose and lipid levels in vertebrates (64). Experimentally induced obesity increases the levels of circulating TNF-α (21), and TNF-α makes mice less sensitive to insulin signaling, possibly through the regulation of GLUT4 and IRS-1 (65). In humans, nutrient excess leads to an inflammatory state characterized by excess TNF production and increased insulin resistance (66). Our present work, together with these findings, indicates that activation of immune responses in a metabolic organ leads to systemic immune–metabolism alterations in the host.

Mechanistically, it is unclear how an immune–metabolic axis influences host responses to bacterial infection. Immunity encompasses resistance mechanisms that kill infectious microbes and tolerance mechanisms that mitigate disease severity without effects on microbial load. In our study, mutations of the insulin pathway affected bacterial load and host survival in a manner that depends on bacterial identity and the route of infection. However, further studies are needed to determine the extent to which insulin regulates tolerance or resistance to the individual pathogens. We consider it likely that the impact of metabolism on host responses to infection is a function of the infectious microbe and the route of infection. For example, glucose supplementation improves survival outcomes in mice challenged with influenza virus but has the opposite effect on mice infected with L. monocytogenes (67). This study outlines an accessible model to characterize the relationship between insulin and IMD/TNF-dependent containment of infectious microbes.

impl2^def20 was provided by Young Kwon. ilp2,3,5 and Ilp2-HF flies were provided by Seung Kim. The imd ^EY08573 (null) mutant flies were provided by Dr. Bruno Lemaitre. The C6706 strain of V. cholerae was provided by Dr. Stefan Pukatzki. The following bacterial stocks were provided by Dr. Nicolas Buchon: E. faecalis, P. sneebia, P. rettgeri, and S. marcesecens DB11. FlyPAD was built and assembled by Pavel Itskov. We acknowledge microscopy support from Dr. Stephen Ogg at the University of Alberta. Microarrays were processed at the Alberta Transplant Applied Genomics Center.

The authors have no financial conflicts of interest.