Downregulation of Perilipin1 by the Immune Deficiency Pathway Leads to Lipid Droplet Reconfiguration and Adaptation to Bacterial Infection in Drosophila

Immune activation is essentially accompanied by metabolic reprogramming, which redistributes accessible energy to prioritize immune protection against pathogenic infections (1, 2). Thus, stringent regulation of metabolic machinery in response to immunoreaction is critical for host fitness. Besides carbohydrates, lipids provide another important bioenergetic and synthetic resource to the host. In turn, a number of lipid metabolites have also been reported to play key roles in pro- or anti-inflammatory pathways (3–5). In all eukaryotic and some prokaryotic cells, there are important intracellular organelles, lipid droplets (LDs), which provide a major place for the synthesis, lysis, transfer, and storage of lipids or their derived metabolites (6). LDs contain a hydrophobic core of neutral lipids, such as di/triacylglycerols or sterol esters, which are surrounded by a phospholipid monolayer decorated with different proteins (7). Previously, LDs were considered to get involved in many physiological and pathological processes just because of their main functions in storing/providing energy and/or buffering toxic lipid species through modulating enzymic or autophagic lipolysis (8, 9). So far, emerging pieces of evidence have shown that LDs also take part in immune regulation. For instance, LDs modulate the functions of myeloid cells through immune–metabolic reprogramming (10). LDs facilitate hosts to combat pathogens’ infections through selectively recruiting immune proteins (11, 12). Recently, a study indicated that mammalian LDs respond to bacterial LPS and function as innate immune hubs to coordinate host defense and cell metabolism (13). However, the role of LDs as pro- or anti-inflammatory modulators is still controversial (14) because of the fact that LDs are highly dynamic organelles. The number, size, and anchored proteins of LDs change quickly in response to infection or stress (15–17). These pieces of evidence indicate that the status of LDs should be tightly controlled. Defects in the biogenesis and mobilization of LDs not only result in lipotoxicity (18, 19) but also exacerbate inflammatory responses and organelles dysfunction (20, 21). However, the role and dynamic pattern of LDs during the immune process is still barely described. Especially, the factors mediating the transformation of LDs underlying immunometabolic switches have not been well identified.

LDs are nonhomogeneous organelles, which accommodate hundreds of variable proteins (22, 23). Perilipins (Plins) are the most prominent proteins that span the surface of LDs (24, 25). Each LD is usually decorated with two or more members of Perilipin family proteins, and no LDs without perilipins have been identified in mammalian cells so far (26). There are five major perilipins in mammals, named PLIN1-5. These perilipins differ in the expression and cellular localization in different tissues, and they have essential roles in the regulation of LDs’ structure and morphology (27, 28). However, whether perilipins get involved in immune functions, probably through mediating LDs’ reconfiguration, is still obscure. In human or mouse adipocyte tissue, PLIN1 deficiency leads to uncontrollable LDs lipolysis and infiltration of inflammatory cells (29, 30). Inhibition of lipases, such as adipose triglyceride lipase or hormone-sensitive lipase, can alleviate this metaflammation (31, 32). Plin1 knockout also promotes secretion of PGs, the proinflammatory lipid metabolites, and elevates proinflammatory M1-type adipose tissue macrophages in mice (33). In bacterial infection, the expression and localization of perilipins on LDs changed in response to LPS stimulation (13). All these pieces of evidence suggest a possible link between plins and immune–metabolic regulations.

Drosophila melanogaster has emerged as a productive organism to investigate immune–metabolism because of the advantages of powerful genetic manipulation and highly conserved mechanisms in both innate immunity and metabolism (34–36). Especially, the fat body (analogous to human liver and adipose tissue), as a major organ mediating systemic innate immunity, is an ideal place for studying the interaction between metabolism and inflammation of LDs because of its richness in LDs (37, 38). Furthermore, plins are evolutionarily conserved from fungi to humans. Not like more redundant Plins in mammalian, there are only two Plins in Drosophila, lipid storage droplet-1 (plin1) and lsd2 (plin2) (39). Plin2 acts to promote lipid storage and LDs’ growth as a barrier for lipase (40–42), whereas Plin1 modulates protein flux on LDs (27, 28). They have opposite functions on the control of LDs’ morphology (27). In Drosophila, the immune deficiency (IMD) pathway is a dominant innate immune signaling mechanism against Gram-negative bacterial infections, which is homologous to the mammalian NF-κB/TNF pathway (35). In brief, the diaminopimelic acid–type peptidoglycan (PGN) of Gram-negative bacteria binds to PGN recognition protein LC (PGRP-LC) on the cell membrane and initiates the IMD signaling cascade where transcriptional factor, Relish (p105-like NF-κB in mammals), is phosphorylated and then cleaved to release the Rel domain into nuclei to drive the expression of antimicrobial peptide (AMP) genes (35). A couple of studies have revealed that IMD signaling modulates lipolysis in either the fat body or intestine of flies (43, 44). LDs’ accumulation was once reported in fly guts after IMD activation (17). However, as if LDs adapt to immune response, whether and how morphological changes occur on LDs remain unclear. More importantly, the contribution of these adaptive altered LDs to infectious pathogenesis and the underlying mechanisms mediating by LD-anchored factors such as Plins are still poorly understood.

In this study, the alteration in morphology and number of LDs were traced dynamically during bacterial infection. Plin1 was found to respond to IMD activation and then modify LDs’ morphology to alleviate inflammatory stress. Our data reveal that adaptive modification of LDs acts as an active modulator of infection-induced pathogenesis.

All flies were propagated at 25°C on standard cornmeal food (1 L food contains 77.7 g cornmeal, 32.19 g yeast, 10.6 g agar, 0.726 g CaCl₂, 31.62 g sucrose, 63.2 g glucose, 2 g potassium sorbate, and 15 ml 5% Tegosept), 30–55% humidity with a 12 h/12 h light/dark cycle. Fly resources that were used in this study are as follows: w¹¹¹⁸ were used as wild type (WT) controls if there was no additional indication. plin1³⁸, UAS-plin1 mcherry, UAS-plin1-RNA interference (RNAi), and ppl-GAL4 were kindly gifted from Dr. Xun Huang (Institute of Genetics and Developmental Biology, Chinese Academy of Sciences [CAS]). gstD-GFP was kindly gifted from Dr. Zhiwei Liu (Shanghai Ocean University). w¹¹¹⁸, w[1118];P{w[+mW.hs]=Switch1}106, PGRP-LCΔ5, and Relish^E20 were obtained from Bloomington Drosophila Stock Center. UAS-plin1 RNAi-TH (TH04820.N) was from Tsinghua Fly Center. All flies used in this study were male. Two bacterial strains, Escherichia coli (DH5a) and Salmonella typhimurium (SR-11) (a gift from Dr. Zhihua Liu, Institute of Biophysics, CAS), were used in this study.

To construct the mrt and pzg reporter vector (mrt-luc and pzg-luc), the mrt and pzg promoter sequence (about −1500 or −1000 to 0 bp) was PCR amplified from Drosophila genomic DNA and introduced into pGL3 vector (Promega) at HindIII restriction site by using recombination technology (Hieff Clone Plus One Step Cloning Kit, YEASEN Biotechnology). All the plasmid constructs were verified by nucleotide sequencing. pAC5.1-Renilla plasmid was a normalized reporter. dsRNAs against Relish or GFP used in the luciferase reporter assay were synthesized using MEGAscript T7 kit (Invitrogen). Primers used for PCR amplification are listed in Supplemental Table I.

Bacterial strains used in this study are E. coli (DH5a) and S. typhimurium. Two days before infection, both bacteria from glycerol stocks were streaked onto Luria Broth (LB) agar plates and grown overnight at 37°C. The plate could be stored at 4°C for up to 1 wk. A single colony was inoculated to 6 ml fresh LB medium and grown at 37°C with shaking (200 rpm). The bacteria was grown to an OD600 of 0.7 to 0.8 (∼3.5 h). The bacterial culture was pelleted with sterile PBS to the desired concentration. We injected 50.6 nl of bacterial suspension into dorsal prothorax of each fly with Nanoject II injector (Drummond Scientific). All flies used were 1 wk after eclosion. The final OD (OD/ml) at 600 nm for injection were E. coli (OD 10) and S. typhimurium (OD 6 or OD 3). For E. coli infection, each fly obtained ∼1 × 10⁶ CFUs. For S. typhimurium infection, each fly obtained the lower dose (∼2 × 10⁵ CFUs) or the higher dose (∼1 × 10⁶ CFUs) according to the experiment design. Infected flies ∼23 per vial were maintained at 25°C. Death was recorded at the indicated time point, and alive flies were transferred to fresh food every day for the survival analysis and CFUs assay.

To monitor bacterial loads of the flies during infection, the number of CFUs grown on an LB agar plate was determined as follows: five living flies were randomly collected in a 1.5 ml Eppendorf (EP) tube, rinsed with 70% ethanol two times by vortex for 10 s to sterile the surface adherent bacteria, then rinsed with sterile deionized water two times by vortex for 10 s, and then homogenized in 200 μl of sterile PBS with three fly body volumes of ceramic beads (diameter: 0.5 mm) in the Minilys apparatus (Bertin Technologies) at the highest speed for 30 s. The suspensions obtained were then serially diluted in PBS and plated on LB agar. For S. typhimurium plating, PBS was substituted with PBS + 1% Triton X-100. For the bacterial load at zero time point, flies were allowed to rest for 10 min after bacterial injection before plating as described above. The agar plate was maintained at 37°C for 18 h before CFUs counting. CFUs were log10 transformed.

S2* cells (a gift from Dahua Chen, Institute of Zoology, CAS) were maintained in Drosophila Schneider Medium (Invitrogen) supplemented with 10% heat-inactivated FBS (Life Technologies), 100 U/ml of penicillin, and 100 mg/ml of streptomycin at 28°C. Transient transfection of various plasmids and dsRNA was performed with lipofectamine 3000 (Invitrogen), according to the manufacturer’s manual. Luciferase reporter assays were carried out using a dual-luciferase reporter assay system (Promega). Where indicated, cells were treated with PGN (35 µg/µL, 6 h) purified from Erwinia carotovora carotovora 15, referring to a previous study (45).

To generate V5-labeled Plin1 protein, the coding sequence of plin1 was PCR amplified from Drosophila cDNA and introduced into pAC5.1 vector. S2 cells (2 × 10⁶ per well) were seeded in a six-well plate and transfected with 2.5 μg plasmid DNA for 36 h. For cycloheximide (CHX) treatment, transfected cells were treated with 0.1 mM CHX (YEASEN Biotechnology, 40325ES03) or 0.1% DMSO (solvent of CHX) for the indicated time. After PBS washing, cells were lysed by cell lysis buffer (Beyotime Biotechnology, P0013) with PMSF and cOmplete Protease Inhibitor Cocktail (50×, Roche, 4693116001) for 30 min on ice. The supernatant was collected and mixed with protein loading buffer (Tanon, 180-8210D) heating at 95°C for 7 min. Western blot was used to detect V5-labeled Plin1, and the following Abs were used: mouse anti-V5 (1:3000, Invitrogen, R960-25), mouse anti-tubulin (1:10000, Beyotime Biotechnology, AT819), and anti-mouse IgG linked with HRP (1:3000, Cell Signaling Technology, 7076).

For quantification of mRNA level, ∼20 flies carcass/fat body tissue were dissected in sterile PBS buffer on ice at the indicated time points postinfection, immediately homogenized in 200 μl cold TRIzol with three fly body volumes of ceramic beads (diameter: 0.5 mm) and then supplied an additional 300 μl TRIzol to reach total 500 μl volume, and samples were stored at −80°C. RNA extraction was referred to the manual of commercial kit (Magen, Hipure Total RAN Plus Micro Kit); this kit can effectively remove genomic DNA contamination. cDNA was synthesized by using the kit (abm, 5× All-In-One MasterMix) with total 1 μg isolated RNA as a template in a 20 μl reaction system. Quantitative real-time PCR (qRT-PCR) was performed using a SYBR Green Kit (abm, EvaGreen Super Master Mix) on an Applied Biosystems 7500 or ViiATM 7 thermocycler (Life Technologies). Samples from at least four independent biological replicates per genotype were collected and analyzed. Housekeeping gene rp49 was the reference gene for data normalization. Primer data for qRT-PCR are provided in Supplemental Table I.

For LD staining, we used a thin tungsten needle to peel off the sheets of fat body tissues from the cuticle inside the male adult flies’ abdomen. At least eight flies of each genotype were dissected randomly at each time point. The tissues from different flies were mixed and fixed in 4% freshly prepared paraformaldehyde (pH 7.5) in PBS for 10 min on ice. Tissues were then rinsed twice with PBS (3 min each time) and then incubated in PBS containing 1 μg/ml of BODIPY 493/503 (Invitrogen) dye or 0.5 μg/ml Nile Red (Sigma) for 30 min on ice; DAPI (1 μg/μl, final concentration) was added to stain nuclei at the last 5 min of the staining process. After staining, tissues were rinsed three times with PBS (3 min each time) and then mounted in mounting medium (Vector Laboratories, H-1000) for microscopy analysis. Random 8 to 10 confocal images (1024 × 1024 pixel) of each genotype were selected for image analysis. Over 30 adult fat body cells/nuclei were randomly chosen from these images for LDs’ size statistics by using ImageJ. The area of the three largest LDs per cell (or around per nuclei) was measured to quantify the relative size of LDs, as described previously (27). Except for plin1 mutant flies, only the biggest LD was measured (27, 28) because the largest LD almost occupies the entire cell space in these mutants. To count the size distribution of LDs, the average percentage of the indicated size range of LDs per cell from 30 fat body cells was determined by using the “Analyze Particles” tool embedded in ImageJ software (https://imagej.nih.gov/ij/). To quantify the fluorescence intensity of GFP on the surface of LDs, confocal images acquired from eight fat bodies were measured by ImageJ software.

Glyceride amounts were measured using a Triglyceride Quantification Kit (BioSino). Briefly, for whole body glyceride quantification, groups of 12 1-wk-old male flies were collected and weighted (∼10 mg) in a 1.5 ml EP tube and then immediately stored at −80°C for a subsequent assay. Stored flies were homogenized in 200 μl lysis buffer (10 mM KH₂PO₄ and 1 mM EDTA [pH = 7.4]) with three fly body volumes of ceramic beads and inactivated in a water bath at 75°C for 15 min. The inactivated homogenate was homogenized again for 30 s and kept on ice ready for an assay. For each glyceride measurement, 3 μl of homogenate was incubated with 250 μl reaction buffer at 37°C for 10 min. After removal of debris by centrifugation (2000 rpm for 2 min), 150 μl of clear supernatant was used to perform a colorimetric assay in a 96-well plate (Corning Costar) for absorbance reading at 505 nm. Glyceride level was normalized with fly weight in each homogenate (unit: nmol/mg fly). For fat body glyceride quantification, 25 fly carcass/fat body tissues were dissected and completed following the assay described above. Glyceride level was normalized per 25 flies (unit: nmol/25 fly).

RU486 induction was described as before (46). Briefly, a 10 mg/ml stock solution of RU486 (mifepristone; Sigma) was dissolved in DMSO. Appropriate volumes of RU486 stock solution were diluted with water containing 2% ethanol to a final concentration of 50 μg/ml. Then, 100 μl of the diluted RU486 solution was dipped onto the surface of fresh food in vials (diameter: 2 cm). The vials were then allowed to dry at room temperature for half a day or 4°C overnight. Flies were transferred to RU486-contained food and raised at 25°C, and fresh food was changed every 2 d.

N-acetyl-l-cysteine (NAC) (Beyotime Biotechnology) fresh solution was prepared by dissolving 0.5 g of NAC powder in 10 ml distilled water; the solution could be aliquoted into 1 ml per EP tube and frozen or stored at −80°C. Then, 100 μl of NAC solution was dipped onto the surface of fresh food in vials (diameter: 2 cm). The vials were then allowed to dry at room temperature for half a day or 4°C overnight. Flies were transferred to NAC-contained food and raised at 25°C, and fresh food was changed every day.

We used two methods to detect reactive oxygen species (ROS) in the fat body, which are gstD-GFP reporter flies and dichlorofluorescein diacetate (DCFH-DA) labeling. The oxidative stress reporter construct gstD-GFP for evaluating cellular ROS levels has been described before (47). Briefly, the carcass/fat bodies of transgenic flies containing a gstD-GFP reporter construct were dissected in sterile PBS, fixed in 4% formaldehyde for 10 min on ice, and rinsed twice with ice-chilled PBS (3 min each time), and then the flaky fat body cells attached to the inner carcass shell were dissected out to mount and confocal image (Vector Laboratories, H-1000). DCFH-DA (Beyotime Biotechnology, ROS Assay Kit) labeling of fresh dissected carcass/fat body tissues, performed according to the manufacturer’s manual, which is based on the ROS-dependent oxidation of DCFH-DA to fluorescent molecule 2′-7′dichlorofluorescein. In brief, the tissues were incubated with PBS containing 20 μM DCFH-DA for 30 min at 30°C and washed with sterile PBS three times (3 min each) to remove free DCFH-DA that do not uptake by the cell, and then the flaky fat body cells attached to the inner carcass shell immediately were dissected out to mount and confocal image (Vector Laboratories, H-1000). It should be noted that the slices were confocal imaged using the exact same settings for control and experimental groups. The fluorescence intensity is proportional to the ROS levels, and the fluorescence intensity of GFP or 2′-7′dichlorofluorescein was quantified by using ImageJ software.

LSM700 (Leica Biosystems) and Olympus FV-1200 confocal laser scanning microscopy were used for imaging. Captured images were analyzed by implemented soft. ImageJ (https://imagej.nih.gov/ij/) was used for the analysis of fluorescence intensity and LDs’ size.

All replicates are shown as the mean ± SD or mean with range. Statistical significance was determined using an unpaired Student t test for two measurements, one-way ANOVA (Tukey honestly significant difference) with multiple t tests and multiple t tests for pairwise comparisons. Kaplan–Meier test for survival curves comparison was also performed. All data processing was performed with GraphPad Prism 7.0.

The sample size was determined according to the number of data points. Batches of the experiment were carried out to ensure repeatability and the use of enough animals for each data point.

Measures were taken to ensure randomization. Each experimental batch contained more animals than the number of data points to ensure randomization and to prevent the accidental exclusion of animals. In vitro analyses were usually performed on a specimen from animals at each data point to ensure a minimum of three biological replicates.

Data collection and data analysis were routinely performed by different people to blind potential bias. All measurement data are expressed as mean ± SD to maximally show the derivations, unless otherwise specified.

In Drosophila, the fat body is not only a central organ mediating systemic immune responses but also the epicenter for lipid metabolism. Thus, to decipher the mechanistic connections between innate immunity and lipid metabolism, the kinetics of fat content was tested in the fat body of Drosophila after systemic infection. E. coli, a nonpathogenic Gram-negative bacterium for flies, was used to perform nanoinjection to infect adult male fruit flies. The IMD pathway is a dominant innate immune signaling against Gram-negative bacterial infections that regulates Relish-dependent transcription of AMPs, such as diptericin (Dpt) (35). Thus, by measuring the expression level of Dpt, IMD signaling activity could be monitored (48, 49). Consistent with the previous report that E.coli injection resulted in a transient innate immune response within 48 h postinfection (hpi) (50), a gradient increase in IMD activity in the fat body was observed from 0 to 12 hpi, and then this activity subsided to the basal level after 48 hpi (Fig. 1A). Interestingly, compared with mock injection control (Supplemental Fig. 1A), the fat levels in the fat body of flies with E. coli infection steadily increased from 4 to 16 hpi and then almost recovered after 48 hpi (Fig. 1A). Therefore, these results hint at a possible correlation between lipid metabolism and IMD signaling activation in the fat body of flies.

LDs are the main sites for lipid metabolism, mobilization, and storage (51), which prompted us to investigate whether LDs change in the fat body in response to bacterial infection. BODIPY staining of fat body cells revealed that compared with the PBS injection group (Fig. 1B, 1C), E. coli infection increased the percentage of intracellular small LDs (diameter < 2 μm) at 6 hpi (Fig. 1B, 1C). And then, LDs grew bigger at 16 hpi as indicated by a decrease in the percentage of small LDs and a concurrent increase in the percentage of large LDs (diameter > 4 μm). Finally, this size distribution of LDs was restored to basal levels at 24 hpi (Fig. 1B, 1C). Accordingly, the size of LDs in fat body cells had a similar changing trend (Fig. 1D). These results indicate that more large LDs tend to form during the initial 16 h after E. coli infection.

To determine whether IMD signaling activation rather than live bacterial growth is responsible for the modification of LDs during infection, heat-killed E. coli was applied to repeat infection in WT flies. The elevated fat levels in fat bodies were still observed at 12 hpi in WT flies (Fig. 2A). In Drosophila, PGN from Gram-negative bacteria sensed by PGRP-LC, the receptor of IMD signaling, results in IMD pathway activation, which is dependent on the transcriptional regulator Relish (35). Thus, the flies with homozygous mutation of PGRP-LC (PGRP-LC^Δ5) or Relish (Relish^E20) were used for infection. In contrast to WT flies, the phenotype of elevated fat contents in fat bodies disappeared and even was reversed in these mutant flies at 12 h after heat-killed E. coli injection (Fig. 2B). Moreover, IMD signaling pathway deficiency also restricted the increase in LDs size at 16 hpi compared with WT controls (Fig. 2C, 2D). It is interesting to note that Relish^E20 fat bodies contained less fat content and smaller LDs than WT at the basal level. However, these results suggest that IMD signaling activation is required to modify LDs’ morphology in response to bacterial infection.

Plins, a group of constitutive proteins that span the surface of LDs, were reported to regulate lipid mobilization and LDs’ morphology (27, 28). There are two plins in Drosophila, Plin1 and Plin2. To explore whether Plins are involved in the regulation of LDs’ reconfiguration in response to immune activation, their time-course expression was detected in the fat body by qRT-PCR. E. coli infection induced a significant downregulation of plin1 mRNA levels at 4 hpi, which was then gradually restored to basal levels at 24 hpi (Fig. 3A). The changing trend of plin1 expression seemed to be negatively correlated with the changes in LD size and IMD activity. However, the expression level of plin2 was only slightly turned down at 4 hpi and back to normal at 12 h after E. coli infection (Supplemental Fig. 2A). Previous studies have shown that deficiency of plin2 resulted in reduced rather than enlarged LDs (27). Therefore, these results indicate a potential role of plin1 in the regulation of LDs’ morphology in response to IMD pathway activation. Furthermore, either deficiency of plin1 by mutation (plin1³⁸) or specific knockdown of plin1 in the fat body [two independent UAS-plin1-RNAi (27) driven by ppl-GAL4] promoted the formation of large LDs (RNAi efficiency was shown in Supplemental Fig. 3), whereas ectopic expression of plin1 in the fat body led to an accumulation of much smaller LDs compared with controls (Fig. 3B, 3B1). These results were reminiscent of previous studies that Plin1 may function to enhance lipid mobilization and inhibit LD coalescence (52). Additionally, we found that V5-tagged Plin1 almost completely degraded within 2 h after CHX (a protein synthesis inhibitor) treatment of Drosophila S2 cells (Supplemental Fig. 4A). This fast turnover of Plin1 further supports our observation that the transcriptional regulation of plin1 is likely sufficient to modify LDs in quick response to transient innate immune activation.

Martik/Putzig complex, a chromosome remodeling complex, has been reported to suppress plin1 at the transcriptional level (52). The mRNA levels of both mrt and pzg were upregulated in the fat body after bacterial infection (Fig. 3C, 3D). Interestingly, homologous alignment showed that at least one conserved binding motif of Relish existed in the promoter region of both mrt and pzg genes across Drosophila species with different evolutionary ages (Supplemental Fig. 5A, 5B). This implies a potential regulation of these genes by IMD/Relish. PGN derived from Gram-negative bacteria can activate IMD signaling in Drosophila S2* cells in vitro (53). The activity of luciferase controlled by the promoter of mrt or pzg was enhanced after PGN treatment in S2* cells, which was blocked by knocking down Relish via dsRNA (54) (Fig. 3E, 3F). Additionally, two Relish binding motifs in the truncated mrt promoter region (T-mrt(Rel), −870 to +1 bp, in Supplemental Fig. 5C) were required for mrt transcription (Fig. 3G) because PGN treatment did not enhance T-mrt-Luc activity anymore when these two sites were removed (Fig. 3G). Thus, these results suggest that suppression of plin1 by IMD signaling might be through upregulation of mrt/pzg. Altogether, these results provide an explanation for LDs’ growth in the early stage of transient IMD pathway activation.

Naturally, whether Plin1 participated in the defense of bacterial infection was tested next. Because E. coli is nonpathogenic to flies, another Gram-negative bacterium, (S. typhimurium), which is a deadly pathogen for flies (55), was used to evaluate Plin1 function on immune defense. Compared with genetic controls, either plin1 deficiency (plin1³⁸) (Fig. 4A, 4B) or fat body–specific knockdown of plin1 (ppl-GAL4 > UAS-plin1-RNAi) (Fig. 4C, 4D) significantly prolonged the survival rate and slightly reduced bacterial loads (CFUs) after S. typhimurium injection, indicative of enhanced resistance against bacterial infection. Conversely, ectopic expression of plin1 in the fat body (ppl-GAL4 > UAS-plin1) led to a dramatic increase in mortality rates of flies infected with S. typhimurium (Fig. 4E, reducing infection OD because plin1 overexpressed flies died too quickly) or even by nonpathogenic E. coli (Fig. 4G), possibly because of uncontrolled bacterial growth (Fig. 4F, 4H). However, it is worthy to note that deficiency of plin1 did not affect AMPs (Dpt and AttacinA) response upon E. coli infection (Supplemental Fig. 6A) but specifically improved Dpt expression upon S. typhimurium infection (Supplemental Fig. 6B). Interestingly, overexpression of plin1 dampened AMPs response in both E. coli and S. typhimurium infections (Supplemental Fig. 6C and 6D). Taken together, these results suggest that adaptive downregulation of plin1 in response to IMD signaling activation protected the host against bacterial infections.

The next question is whether downregulated Plin1-induced LDs’ growth also benefits the host against bacterial infection. Sustained immune activation is a high energy–cost process, which requires active lipolysis and usually leads to excessive ROS accumulation because of the release and oxidation of free fatty acids, one hallmark for inflammatory damages (18, 56). However, LDs’ growth could efficiently reduce the accumulation of free fatty acids and probably relieve ROS-related tissue damages (57). As expected, plin1 deficiency (plin1³⁸) (Fig. 5A, 5A1) or knockdown (ppl-GAL4 > UAS-plin1-RNAi) (Fig. 5B, 5B1), which promoted LDs growth, was accompanied by an almost invisible level of ROS compared with control. In contrast, overexpression of plin1 in fat body cells (ppl-GAL4 > UAS-plin1), which increased the fraction of small LDs (Fig. 3B, 3B1), markedly increased ROS intensity (Fig. 5B, 5B1). These results suggest a correlation between the size of LDs and the intensity of ROS accumulation in fat body cells. To further support this notion, we skewed ROS metabolism in fat bodies through knockdown of superoxide dismutase genes (sod1 or sod2) or catalase gene, all of which encode enzymes for intracellular ROS clearance (58, 59). All these flies (ppl-GAL4 > UAS-sod1 RNAi, sod2 RNAi, or catalase gene RNAi) contained slightly elevated ROS levels (Fig. 5C, 5D) and commensurate LD growth in fat bodies (Fig. 5E, 5F). If blocking the large LDs’ formation by simultaneous overexpression of plin1 in these genetic backgrounds (Fig. 5E, 5F), much higher ROS accumulation was observed in fat bodies (Fig. 5C, 5D). Altogether, these results suggest that Plin1-controlled LDs’ reconfiguration takes part in antioxidative functions.

The activation of immune signaling, such as NF-κB or TNF signaling, often associates with ROS-induced inflammatory stress (56, 60). gstD-GFP reporter is a good sensor for ROS in vivo (47). Therefore, a transgenic allele with a gstD-GFP insertion was used to monitor ROS activity in flies by measuring GFP intensity. Indeed, in Drosophila, infection either by nonpathogenic E. coli or by strong pathogenic S. typhimurium induced an increase of intracellular ROS levels in the fat body (Fig. 6A, 6B). These results support the link between bacterial infection and the accumulation of intracellular oxidative stress. If we removed this infection-associated intracellular ROS by feeding flies with NAC, a widely used ROS scavenger, the survival of flies after pathogenic S. typhimurium infection was improved compared with noninfected controls (Fig. 6C). It is worthy to note that feeding flies with NAC at 12 h, not 0 h, post–S. typhimurium infection benefited the fitness of flies much better. It is likely that at the time point of 12 hpi, excessive ROS accumulation had already developed, and initiate ROS of the early infectious stage is useful for defending against bacteria (61, 62). These results suggest that excessive oxidative stress, which develops during bacterial infection, is harmful to the host.

Because LDs are major hubs for lipid metabolism in fat body cells and function on ROS clearance, it prompted us to investigate whether plin1-mediated LDs modification is involved against bacterial infection, probably through regulating intracellular ROS. At first, we found that elevated ROS levels in WT fat bodies were diminished in fat bodies of plin1 deficiency flies (Fig. 6D, 6E). Next, a RU486-induced fat body–specific GAL4 (GS106-GAL4) was used to modulate the expression of plin1 just before S. typhimurium infection, which excludes the possible side effects of plin1 on the development of flies. As expected, overexpression or knockdown of plin1 in the fat body resulted in a significant decrease or increase in LDs’ size, respectively (Fig. 6F). Downregulation of plin1 prolonged the survival rate after S. typhimurium infection (Fig. 6G), whereas ectopic expression of plin1 shortened the life span dramatically (Fig. 6H). Meanwhile, fluorescent probe DCFH-DA staining indicated an elevated intracellular ROS level in plin1 overexpression flies and a reduced ROS level in plin1 knockdown flies during infection (Fig. 6I, 6I1, and 6I2). Taken together, these results suggest that large LDs formation contributes to alleviating intracellular oxidative stress induced by bacterial infection, and Plin1 might serve as an important modulator in response to IMD activation.

Metabolic reprogramming of lipids has been widely reported to be associated with immune responses (1, 2, 63). As a major intracellular organelle for lipid metabolism and storage, LDs also seem to be involved in immune processes. Immune stimulation either by infection with bacteria (64, 65), virus (66–68), fungus (69), or protozoan parasites (70) or by cytokines inoculation (71, 72) may promote the biogenesis of LDs in mammalian leukocytes. Recently, Hash et al. (17) also reported that LDs are infection-inducible organelles in the gut of Drosophila at a certain time point postinfection. Bosch et al. (13) further indicated that LDs recruit antimicrobial proteins in response to LPS and function as innate immune hubs. However, the status and morphology of LDs change rapidly in vivo. This dynamic transformation of LDs in response to immune stimulation is seldom described. Whether adaptive morphological changes of LDs play active rather than passive roles in pathogenesis and key regulators linking LDs’ reconfiguration and infection still need further investigation.

In this study (Fig. 7), we carefully traced the time-course morphogenesis of LDs in the fat body along with the dynamic curve of IMD signaling activity. We found that transient IMD activation by bacterial infection promoted LDs’ growth in the fat body within 12 hpi. Both LDs’ size and fat levels in the fat body were maximum when IMD activity almost achieved its peak. Previous studies and our results (data not shown) show that transcriptional levels of most triglyceride synthesis genes are suppressed during the initial phase of infection (73, 74), suggesting that the substrates for LDs’ biogenesis in the fat body were probably imported lipids rather than de novo synthesized fatty acids, and IMD signaling activation is required for this process. Detailed analysis showed that plin1 downregulation is critical for LDs’ growth in response to transient IMD activation, considering its expression was suppressed by Relish-activated MRT/PZG complex. Although the immune response is an energy–cost process, the fat contents in specific tissues, such as fat bodies, are surprisingly increasing at the early stage of immune activation. These findings prompt us to imagine that LDs’ biogenesis is likely an active host adaptation to immune challenges. To further support this hypothesis, we found that enlarged LDs benefit the host against intracellular ROS–mediated oxidative stress induced by bacterial infection.

Excessive ROS accumulation is often the main cause of inflammation/infection-induced cellular damages. In fact, biological processes such as cancer, neural activity, and inflammation are all energy intensive and rely on robust fat metabolism, which releases large amounts of free fatty acids. The excess accumulation of free fatty acids in the cytoplasm promotes lipotoxicity and ROS-induced oxidative stress (75–77). The high levels of intracellular ROS can further promote lipolysis and free fatty acids release (78). This vicious circle finally drives the host to enter a severe metaflammatory state during chronic hyperinflammation and consequently shortens the lifespan. A recent study showed that renal purge of hemolymphatic lipids can efficiently prevent ROS-mediated tissue damage during inflammation (79). A similar antioxidant function of LDs was also reported in neuronal stem cell niche (80) and cancer cells (81). In our study, blocking the breakdown and promoting the growth of LDs by downregulated plin1 could efficiently eliminate ROS accumulation and prolong flies’ lifespan after bacterial infection. However, the detailed mechanisms of how large LDs prefer to prevent ROS accumulation needs further investigation. One possibility is that the formation of large LDs sequesters the release of excessive free lipids, and oxidation contributes to the main source of ROS generation. Another possibility is that the total contact areas between mitochondria and their surrounding LDs will be reduced along with LDs growing. As mitochondria provide a major place for the oxidation of lipids usually supplied by LDs (9), reduced contacts between LDs and mitochondria might be another not-bad way to cut down ROS generation. Thus, LDs’ growth is beneficial for the redox homeostasis of the host. The downregulation of Plin1 to promote the enlargement of LDs might be an effective host adaptation to resolve inflammation-associated stress in response to immune activation. It is worthy to note that ROS levels seem to provide feedback on the regulation of LDs’ size as well because the knockdown of ROS scavenging enzymes also promotes LDs’ growth in noninfection conditions (Fig. 5E and F). Although elevated ROS resulting in bigger LDs is independent with Mrt/Pzg–Plin1 axis in noninfection conditions (data not shown), this interesting observation suggests that there is at least another feedback mechanism from ROS to larger LDs’ formation, besides IMD–Plin1 axis in infection conditions. This evidence also strongly indicates that adaptive growth of LDs is an efficient and instinctive way to prevent from ROS damages in the host.

Moreover, large LDs’ formation can reduce the opportunity of pathogens to use free fatty acids for their own growth (82–84). This is possibly one reason why plin1-deficient flies, owning bigger LDs, had lower bacterial loads postinfection. In addition, larger LDs might contain more resident histones, a cationic protein, which has been reported to kill bacteria in a previous study (12). In mammals, IFN-γ treatment of Mycobacterium tuberculosis–infected bone marrow–derived macrophage can induce the formation of LDs, in which neutral lipids serve as a source to produce eicosanoids for enhancing host defense (85). A recent study made a detailed analysis that LDs recruit cathelicidin, a broad-spectrum AMP, in response to LPS stimulation (13). However, whether the change of LDs’ morphology alters the recruitment of LD-anchored proteins and underlying molecular and cellular mechanisms needs further investigation. In our study, the reduced expression of AMPs genes was also found when plin1 overexpression (Supplemental Fig. 6C, 6D) and Dpt was upregulated after plin1 mutant flies were infected with bacteria (Supplemental Fig. 6B). These results also suggest that a link between antimicrobial signaling with plin1 directly or Plin1 mediated LDs’ modification indirectly. However, the large LDs’ formation in response to bacterial infection could in turn benefit the host to combat pathogens actively.

Plin1 is an important protein factor on the surface of LDs. It has been reported to control the mobilization of lipids on LDs’ surface by recruiting kinds of enzymes (86, 87) and is sufficient to alter the morphology of LDs (27). In this study, we found that the expression of plin1 rather than plin2 preferred to be regulated by innate immune signaling. This provokes us to conceive that Plin1 may serve as a bridge to link immunity and lipid metabolism through the modification of LDs. In response to transient immune activation, adaptative enlarged LDs benefit the host against pathogens and inflammation-induced stress. The alternation of the levels of mammalian Plins protein on the LDs’ surface was once mentioned after LPS stimulation (13). Our study provides a possibility that perilipins might respond to immune signals and play an active role in infectious pathogenesis through transforming LDs. It is worthy in the future to trace and dissect the dynamic protein compositions on the surface of LDs along the different stages of inflammation, especially the proteins that interact with Plin1. In summary, we found that the Plin1-mediated LDs’ morphological alteration is not only an adaptive consequence after bacterial infection but also actively contributes to pathogenic regulation. Therefore, reconfiguration of LDs may provide a potential therapeutic target for the resolution of inflammation.

We thank Dr. Xun Huang (Institute of Genetics and Developmental Biology, CAS) for providing stocks of plin1³⁸, UAS-plin1-mcherry, and UAS-plin1-RNAi flies and valuable comments, Dr. Zhiwei Liu (Shanghai Ocean University) for providing gstD-GFP, Dr. Zhihua Liu (Institute of Biophysics, CAS) for providing S. typhimurium (SR-11) strain, and Dr. Song-qing Liu (Institute of Biophysics, CAS) for fly food preparation and stock maintenance. We thank Dr. Hui Xiao, Dr. Parag Kundu, Dr. Philippe Sansonetti (Institut Pasteur of Shanghai, CAS), and Dr. Chengshu Wang (Institute of Plant Physiology and Ecology, CAS) for comments and manuscript polishing.

The authors have no financial conflicts of interest.