Adipocyte autophagy limits gut inflammation by controlling oxylipin levels

Lipids play a major role in inflammatory diseases by altering inflammatory cell functions, through their use as energy substrates or as lipid mediators such as oxylipins. Autophagy, a lysosomal degradation pathway that limits inflammation, is known to impact on lipid availability, however whether this controls inflammation remains unexplored. We found that upon intestinal inflammation visceral adipocytes upregulate autophagy and that adipocyte-specific loss of the autophagy gene Atg7 exacerbates inflammation. While autophagy decreased lipolytic release of free fatty acids, loss of the major lipolytic enzyme Pnpla2/Atgl in adipocytes did not alter intestinal inflammation, ruling out free fatty acids as anti- inflammatory energy substrates. Instead, Atg7-deficient adipose tissues exhibited an altered oxylipin balance, driven through an NRF2-mediated upregulation of Ephx1. This was accompanied by a shift in adipose tissue macrophage polarization with reduced secretion of IL-10, leading to lower circulating levels of IL-10. These results suggest an underappreciated fat-gut crosstalk through an autophagy- dependent regulation of anti-inflammatory oxylipins, indicating a protective effect of adipose tissues for distant inflammation.


Introduction
Autophagy is an essential cellular recycling pathway that engulfs cellular contents, including organelles and macromolecules, in a double membraned autophagosome and directs them towards lysosomal ablate the essential autophagy gene Atg7 specifically in mature adipocytes (Atg7 Ad ) in adult mice ( Figure 2A). Tamoxifen administration led to the significant reduction of Atg7 transcript levels in visceral (Tregs) were decreased in adipocyte autophagy-deficient animals compared to wild-type animals ( Figure S2D), despite not affecting disease recovery. Intestinal FOXP3 + Tregs are classified into three 140 distinct subsets based on co-expression of TH2 and TH17 transcription factors GATA3 + and RORgt + , 141 respectively 22 . While all populations tended to be diminished in Atg7 Ad mice, only RORgt -FOXP3 + Tregs were significantly reduced ( Figure S2E). These data suggest that adipocyte autophagy is dispensable

149
To test this, visceral adipocytes were collected from wild-type and Atg7 Ad mice treated with water or 150 DSS and subjected to RNA sequencing. Since we anticipated sex-specific differences in adipocyte  19,25 . To confirm the importance of adipocyte autophagy for 179 optimal lipolytic output, adipose tissue explants were stimulated with the β-adrenergic receptor agonist 180 isoproterenol and FFA levels were quantified. As expected, FFA secretion was reduced upon lipolysis 181 stimulation in autophagy-deficient as compared to autophagy-proficient adipocytes ( Figure 4A). TNFa 182 is a crucial cytokine for human and murine IBD pathologies 26 and can affect adipose tissue through 183 inhibition of lipogenesis and by promoting FFA secretion 27 . Since circulating TNFa levels were elevated 184 during DSS colitis ( Figure S1D), we investigated its effects on adipocyte lipid metabolism. Expression 185 of the gene encoding for TNF receptor 1, Tnfrsf1a, was upregulated during DSS-induced inflammation 186 in both genotypes, suggesting that TNFa-sensing was unaffected by the loss of adipocyte autophagy 187 ( Figure 4B). Next, we assessed the impact TNFa on FFA release from adipose tissue explants from 188 Atg7 Ad mice or littermate controls. In the presence of TNFa, adipocytes increase FFA secretion, 189 however strikingly, this was significantly blunted in autophagy-deficient adipocytes ( Figure 4C).

190
Consistent with the decreased lipolytic activity of autophagy-deficient adipocytes, Atg7 Ad mice exhibit 191 reduced serum FFA levels compared to wild-type littermates upon DSS colitis ( Figure 4D). While we 192 established that autophagy could modulate overall FFA release, we next tested whether autophagy 193 affects the production and secretion of specific FFA species. To investigate this, serum samples from water and DSS-treated animals were analysed by GC-FID. Confirming our initial findings, the serum concentration of many FFA species was reduced upon adipocyte autophagy loss, indicating that adipocyte autophagy controls overall FFA levels rather than specific FFAs ( Figure 4E). Interestingly, loss of adipose tissue mass was comparable between both genotypes upon DSS-induced colitis ( Figure   198 S4A).

199
It has previously been described that the adipokines leptin and adiponectin can influence intestinal

203
Taken together, our data suggests that adipocyte autophagy blunts the release of FFA both upon β-204 adrenergic receptor-and TNFa-mediated lipolysis.

206
Adipocyte lipolysis is dispensable for DSS-induced colitis severity 207 Based on our data, we hypothesized that differences in FFA availability may be responsible for a 208 differential intestinal immune response. We therefore sought to determine the importance of adipocyte 209 lipolysis during DSS-induced colitis ( Figure 5A) by deleting the cytoplasmic lipase ATGL, a rate-limiting 210 enzyme in the lipolytic pathway 30 . Using adipocyte-specific Pnpla2/Atgl (Atgl Ad ) knockout mice, we first 211 confirmed that Pnpla2/Atgl was efficiently deleted in purified visceral adipocytes ( Figure 5B). Strikingly,

212
Atgl Ad mice lost comparable amounts of body weight upon DSS-induced colitis ( Figure 5C), although 213 adipose tissue loss was completely prevented ( Figure 5D). This data underlines that Atgl-driven lipolysis unlikely to be the mechanism by which autophagy in adipocytes exerts its anti-inflammatory role.

221
Adipocyte autophagy loss promotes NRF2-mediated stress response and alters tissue oxylipin levels

222
To get a better understanding of pathways that may be affected by the loss of autophagy in adipocytes.
an increase in estrogen receptor 1 (Esr1) expression, due to the Cre transgene expression ( Figure 6A- B). Across both treatment groups, we found a total of 32 genes being differentially regulated between 227 WT and Atg7 Ad visceral adipocytes. Six genes were differentially expressed under both water and DSS 228 treatment conditions ( Figure 6C). Due to the limited number of differentially expressed genes between 229 Atg7 Ad and wild-type adipocytes, we opted to look for altered pathway using ranked gene set enrichment 230 analysis (GSEA) 31 , which includes genome-wide alterations to given gene sets of major cellular 231 pathways. Upon DSS-induced colitis, we found that the xenobiotic pathway, was significantly enriched 232 in Atg7 Ad adipocytes ( Figure 6D+S5A). Enzymes which are known for their role in xenobiotic metabolism 233 such as the large family of cytochromes P450 monooxygenases and epoxide hydrolases (EPHX) 234 metabolize and detoxify exogenous substrates and mediate the production of oxylipins from 235 endogenous polyunsaturated fatty acids. The expression of many of the key genes involved in these 236 processes are regulated by NRF2, a major transcription factor of the xenobiotic and oxidative stress 237 responses. We found that Ephx1 was consistently upregulated upon Atg7 loss in adipocytes ( Figure   238 6C). Remarkably, Ephx1 expression was also increased in datasets obtained from other studies in 239 which autophagy genes such as Atg3 and Beclin-1 were specifically deleted in adipocytes ( Figure S5B-240 C) 19,25 . Among the genes that were enriched in Atg7-deficient adipocytes were several other NRF2-241 target genes ( Figure 6E). In agreement with an activation of the NRF2 pathway, NRF2 protein 242 abundance was increased in Atg7 Ad visceral adipose tissues ( Figure 6F). Specificity for NRF2 activation 243 was further confirmed since only NRF2 target gene Ephx1 was transcriptionally upregulated, whereas 244 Ephx2 which is not controlled by NRF2 remained transcriptionally unchanged in autophagy-deficient 245 adipocytes ( Figure 6G). However, both EPHX1 and EPHX2 protein expression were increased in Atg7 Ad 246 adipose tissues ( Figure 6H) suggesting that EPHX2 may be affected by autophagy deletion on a post-   We therefore tested whether the increased expression of EPHX enzymes would shift the balance of adipose tissues during DSS colitis ( Figure 6J). This was consistently observed for all analyzed DHA-259 derived EpFAs which are important substrates for EPHX1 ( Figure 6K) 15 . Strikingly, these effects appear 260 to be locally restricted to the adipose tissues since no changes in these plasma oxylipin levels were 261 observed ( Figure 6L). In summary, these data suggest that loss of adipocyte autophagy activates NRF2

298
Immune cells reside within distinct tissue environments, however, the impact of local metabolic cues on 299 inflammatory processes remains incompletely understood 36 . Our results indicate that autophagy in 300 mature adipocytes contributes to the balance of intra-tissual oxylipin levels. Further, we demonstrate 301 that adipocytes autophagy is part of the anti-inflammatory immune response by promoting the release 302 of IL-10 from adipose tissues. Autophagy-dependent secretion from adipose tissues contributes to 303 systemic IL-10 levels, and limits inflammation at a distant tissue site, the colon. Therefore, our study 304 provides novel insights into a cross-tissue anti-inflammatory mechanism, enabling the development of 305 therapeutic approaches to target this crosstalk.

306
While polymorphisms in autophagy genes are well established as genetic risk factors for IBD, little is 307 known about autophagy's role in adipocytes in this disease. We found that autophagy is induced in 308 visceral adipocytes upon DSS-induced colitis, which was marked, among others, by a transcriptional

317
Early studies found that autophagy is crucial for the normal differentiation of adipose tissues in vivo 18,20 .

334
Loss of adipocyte autophagy increased NRF2 stability, likely through the sequestration of its regulator 335 KEAP1 as shown by Cai et al. 19 . Here we demonstrate for the first time that this antioxidant/xenobiotic 336 pathway exacerbates an inflammatory disease. Increased expression of EPHX1 was paralleled by a 337 dysbalance in oxylipins shifted towards decreased levels of EpFA and increased DiolFA. Similar to our 338 findings, EPHX1 was recently found to convert in particularly omega-3 DHA substrates in adipocytes 339 and liver 15,44 . Since our data suggest a broader dysregulation of EpFA:DiolFA, it is likely that EPHX2, 340 which was accumulated on protein level in Atg7 Ad adipose tissues, may also contribute to the conversion of oxylipin substrates. Increasing evidence suggests that macrophages are regulated by oxylipins in their environment. Indeed, increased presence of omega-3 derived EpFA achieved either through these two hallmarks of anti-inflammatory macrophages were equally decreased.

Mice
Adipoq-Cre ERT2 mice 58 were purchased from Charles River, UK (JAX stock number: 025124) and were crossed to Atg7 floxed mice 59 . Experimental cages were sex-and age-matched and balanced for genotypes. Genetic recombination was induced at 8-10 weeks of age by oral gavage of 4mg tamoxifen per mouse for five consecutive days. All experimental procedures were conducted two weeks after last tamoxifen administration. DSS-induced colitis was induced by 1.

Histopathology assessment
Distal, mid and proximal colon pieces were fixed in 10% neutral buffered formalin for 24 hours before washed and transferred into 70% ethanol. Tissue pieces from each sample were embedded in the same paraffin block and 5µm sections were subsequently stained with haematoxylin and eosin (H&E). Scoring of histology sections was executed in a blinded fashion according to a previously reported scoring system 21 . In brief, each section was assessed for the degree inflammation, the depth of tissue damage, possible crypt damages, with high scores signifying increased tissue damage. In addition, signs of regeneration (epithelial closure, crypt regeneration) were assessed, with high scores indicating delayed regeneration. Changes were multiplied with a factor classifying the involvement tissue area. Total score was calculated and presented.

Adipose tissue and colon digestion
We collected mesenteric adipose tissue separate from a collective set of visceral adipose tissue depots (including omental, gonadal and retroperitoneal adipose tissue) to distinguish proximal versus distal effects of intestinal inflammation on adipose tissues. Adipose tissues were collected and digested in DMEM containing 1% fatty acid-free BSA (Sigma, 126609), 5% HEPES (Gibco, 15630-056), 0.2mg/mL Liberase TL (Roche, 5401020001) and 20µg/mL DNaseI (Roche, 11284932001). Tissues were minced in digestion medium and incubated for 25-30min at 37°C at 180rpm. Tissues were further broken down by pipetting using wide-bore tips and filtered through a 70µm mesh. Digestion was quenched by adding medium containing 2mM EDTA. Adipocyte and stromal vascular fraction were separated by centrifugation (700g, 10min) and collected for further downstream analysis.
Colon digestions were performed as previously described 62 . Colons were opened longitudinally and faecal content was removed by washing with PBS. Then colons were washed twice in RPMI containing 5% FBS and 5mM EDTA at 37°C under agitation. Tissues were minced and digested in RPMI supplemented with 5% FBS, 1mg/mL collagenase type VIII (Sigma) and 40μg/mL DNaseI (Roche). Cell suspension was strained through 40µm mesh and cells were subjected to downstream analysis.

Flow Cytometry
Flow cytometry staining was performed as previously described 63

Quantitative PCR
Adipocytes and adipose tissue RNA were extracted using TRI reagent (T9424, Sigma). Colon tissue RNA were extracted in RLT buffer containing 1,4-Dithiothreitol. Tissues were homogenised by lysis in 2mL tubes containing ceramic beads (KT03961-1-003.2, Bertin Instruments) using a Precellys 24 homogenizer (Bertin Instruments). RNA was purified following RNeasy Mini Kit (74104, Qiagen) manufacturer instructions. cDNA was synthesized following the High-Capacity RNA-to-cDNA™ kit protocol (4388950, ThermoFischer). Gene expression was assessed using validated TaqMan probes and run on a ViiA7 real-time PCR system. All data were collected by comparative Ct method either represented as relative expression (2 -ΔCt ) or fold change (2 -ΔΔCt ). Data were normalized to the two most stable housekeeping genes; for adipose tissues Tbp and Rn18s and for colon Actb and Hprt.

Bulk RNA sequencing
Visceral adipocytes were isolated as floating fraction upon digestion. RNA was extracted and converted to cDNA as described above. PolyA libraries were prepared through end reparation, A-tailing and adapter ligation. Samples were then size-selected, multiplexed and sequenced using a NovaSeq6000.
Raw read quality control was performed using pipeline readqc.py (https://github.com/cgatdevelopers/cgat-flow). Resulting reads were aligned to GRCm38/Mm10 reference genome using the pseudoalignment method kallisto 64 . Differential gene expression analysis was performed using DEseq2 v1.30.1 65 . Pathway enrichment analysis was performed on differentially expressed genes for "Biological Pathways" using clusterProfiler (v4.0) R package 66 . DESeq2 median of ratios were used for visualisation of expression levels. Heatmaps of selected gene sets were presented as z-scores using R package pheatmap. Gene enrichment analysis was performed using GSEA software using Hallmark gene sets 31 . R code is available under https://github.com/cleete/IBD-Adipocyte-Autophagy

Lipolysis assays
Adipose tissues were collected and washed in PBS before subjected to lipolysis assays. For isoproterenol stimulation, adipose tissues were cut into small tissue pieces and incubated in serum-free DMEM -High Glucose (Sigma, D5796) with 2% fatty acid-free BSA (Sigma, 126579) in the absence or presence of 10µM isoproterenol (Sigma, I6504) for the indicated time. TNFα-induced lipolysis was induced as previously described 41 . In brief, small adipose tissue pieces were cultured in DMEM -High Glucose for 24 hours in the absence or presence of 100ng/mL recombinant TNFα (Peprotech, 315-01A) and then transferred into serum-free DMEM containing 2% fatty acid free BSA for 3 hours.
Supernatants were collected and FFA concentration normalized to adipose tissue input.

Adipose tissue explant cultures
Gonadal or mesenteric adipose tissue explants were collected from mice at indicated time points. For autophagic flux measurements, explants (~50-100mg) were cultured for DMEM supplemented with 10% FBS (Sigma, F9665) and 100U/ml Pen-Strep for 4h in the absence or presence of lysosomal inhibitors 100nM Bafilomycin A1 and 20mM ammonium chloride. Explants were washed in PBS before collection and then frozen at -80°C until proteins were extracted for immunoblotting. For measurement of cytokine secretion, adipose tissue explants were cultured for 6h in DMEM/High Modified (D6429, Sigma) with 100U/ml Pen-Strep in the absence of FBS. Supernatant was collected, spun down (400g, 5min) to remove cell debris and then frozen until further analysis.

Free fatty acid analysis
Total supernatant and serum FFA levels were measured using Free Fatty Acid Assay Quantification Kit (ab65341, Abcam). For detailed analysis of FFA species, lipids were extracted by Folch's method 67 and subsequently run on a one-dimensional thin layer chromatography (TLC) using a 10x10cm silica gel G plate in a hexane/diethyl ether/acetic acid (80:20:1, by vol.) solvent system. Separated FFA were used for fatty acid methyl esters (FAMEs) preparation through addition of 2.5% H2SO4 solution in dry methanol/toluene (2:1 (v/v)) at 70°C for 2h. A known amount of C17:0 was added as an internal standard for quantification. FAMEs were extracted with HPLC grade hexane. A Clarus 500 gas chromatograph with a flame ionizing detector (FID) (Perkin-Elmer) and fitted with a 30m x 0.25mm i.d. capillary column (Elite 225, Perkin Elmer) was used for separation and analysis of FAs. The oven temperature was programmed as follows: 170°C for 3min, increased to 220°C at 4°C/min), and then held at 220°C for 15min. FAMEs were identified routinely by comparing retention times of peaks with those of G411 FA standards (Nu-Chek Prep Inc). TotalChrom software (Perkin-Elmer) was used for data acquisition and quantification.

Oxylipin analysis
Oxylipins were analyzed by means of liquid chromatography mass spectrometry 68,69 . The plasma samples were analyzed following protein precipitation and solid-phase extraction on reversed phase/anion exchange cartridges 68,69 . The adipose tissue was homogenized in a ball mill and oxylipins were extracted with a mixture of chloroform and iso-propanol following solid-phase extraction on an amino propyl SPE cartridge 70,71 . Oxylipin concentrations were determined by external calibration with internal standards 68,69 .

Immunoblotting
Autophagic flux in adipose tissues was measured by incubating adipose tissue explants from experimental animals in RPMI in the absence or presence of lysosomal inhibitors 100nM Bafilomycin A1 and 20mM NH4Cl for 4 hours. DMSO was used as 'vehicle' control. Adipose tissues were collected and snap frozen. Protein extraction was performed as previously described 72

Transmission electron microscopy
Mice were sacrificed by increasing concentrations of CO2. Adipose tissues were excised, cut into small 1-2mm pieces and immediately fixed in pre-warmed (37 ˚C) primary fixative containing 2.5% glutaraldehyde and 4% formaldehyde in 0.1M sodium cacodylate buffer, pH7.2 for 2 hours at room temperature and then stored in the fixative at 4 ˚C until further processing. Samples were then washed for 2x 45 min in 0.1M sodium cacodylate buffer (pH 7.2) at room temperature with rotation, transferred to carrier baskets and processed for EM using a Leica AMW automated microwave processing unit.
Briefly, this included three washes with 0.1M sodium cacodylate buffer, pH 7.2, one wash with 50mM glycine in 0.1M sodium cacodylate buffer to quench free aldehydes, secondary fixation with 1% osmium tetroxide + 1.5% potassium ferricyanide in 0.1M sodium cacodylate buffer, six water washes, tertiary fixation with 2% uranyl acetate, two water washes, then dehydration with ethanol from 30%, 50%, 70%, 90%, 95% to 100% (repeated twice). All of these steps were performed at 37 ˚C and 15-20W for 1-2 mins each, with the exception of the osmium and uranyl acetate steps, which were for 12 min and 9 min respectively. Samples were infiltrated with TAAB Hard Plus epoxy resin to 100% resin in the AMW and then processed manually at room temperature for the remaining steps. Samples were transferred to 2ml tubes filled with fresh resin, centrifuged for ~2mins at 2000g (to help improve resin infiltration), then incubated at room temperature overnight with rotation. The following day, the resin was removed and replaced with fresh resin, then the samples were centrifuged as above and incubated at room temperature with rotation for ~3 hrs. This step was repeated and then tissue pieces were transferred to individual Beem capsules filled with fresh resin and polymerised for 48 hrs at 60 ˚C. Once polymerised, blocks were sectioned using a Diatome diamond knife on a Leica UC7 Ultramicrotome. Ultrathin (90nm) sections were transferred onto 200 mesh copper grids and then post-stained with lead citrate for 5 mins, washed and air dried. Grids were imaged with a Thermo Fisher Tecnai 12 TEM (operated at 120 kV) using a Gatan OneView camera.

Statistical Analysis
Data were tested for normality before applying parametric or non-parametric testing. For two normallydistributed groups, unpaired Student's tests were applied. Comparisons across more than two experimental groups were performed using One-Way or Two-Way ANOVA with Šídák multiple testing correction. Data were considered statistically significant when p<0.05 (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001). Typically, data were pooled from at least two experiments, if not otherwise indicated, and presented as mean. Data was visualized and statistics calculated in either GraphPad Prism 9 or R software.