Activation of adenosine receptor A2a by gut symbionts promotes immune tolerance


 The gut microbiota is critical to immune homeostasis, but our understanding of the underlying molecular mechanisms is very limited. Here, we demonstrate a division of labor among members of the eight-membered “model microbiome” altered Schaedler flora in promoting distinct immunophenotypes. We report that Parabacteroides goldsteinii ASF519 induces immune tolerance by promoting interleukin (IL)-10 production in a variety of myeloid-derived immune cells. The IL-10 induction is dependent on the activation of adenosine receptor A2a by microbial enzymes of the methionine cycle. ASF519 colonization in mice increased the level of adenosine in ceca and induced IL-10 secreting dendritic cells in colonic lamina propria. These immunophenotypes were pharmacologically reversed by A2a blockage. In mouse models of human autoimmune diseases, ASF519 supplementation significantly ameliorated insulitis in type 1 diabetes and collagen-induced arthritis. This study unveils a novel paradigm of gut microbiota-adenosine receptor interactions in immune tolerance and potentially provides a new therapeutic strategy for immune disorders.


Abstract 23
The gut microbiota is critical to immune homeostasis [1][2][3][4] , but our understanding of the 24 underlying molecular mechanisms is very limited. Here, we demonstrate a division of 25 labor among members of the eight-membered "model microbiome" altered Schaedler 26 flora 5 in promoting distinct immunophenotypes. We report that Parabacteroides induced the production of the anti-inflammatory cytokine IL-10 in a bacterial growth-87 dependent manner ( Fig. 1a; Extended Data Fig. 1c). In contrast, Clostridium spp. 88 ASF356 only induced IL-10 at the exponential stage, while a weak IL-10 induction 89 was observed with Eubacterium plexicaudatum ASF492 and Pseudoflavonifractor 90 spp. ASF500. Measurement of another anti-inflammatory cytokine, TGF-β, revealed 91 that Lactobacillus intestinalis ASF360 and Lactobacillus murinis ASF361 induced 92 BMDCs to release this cytokine (Fig. 1a), which positively correlated with bacterial 93 growth (Extended Data Fig. 1c); ASF519 had a mild effect on TGF-β. For their 94 immunogenic potential, although the uninoculated bacterial culture medium induced 95 a high level of pro-inflammatory TNF-α, this effect was diminished by the mono-96 culture of 7 ASF members (Fig. 1a), with the exception of E. plexicaudatum ASF492 97 that displayed a recurring TNF-α release when the stationary-phase SSM was used. 98 These measurements revealed distinct cytokine inducing profiles of ASF members 99 coupled with growth phase-specific microbial physiology (Fig. 1a). Notably, ASF519 100 is the only member that induced two anti-inflammatory cytokines, IL-10 and TGF-β 101 (Fig. 1b). 102 exponential and stationary phases, strongly promoted Foxp3+ Treg proliferation 106 (80+6% by ASF519 SSM versus 20+9% by uninoculated bacterial culture medium) 107 ( Fig. 1a, Extended Data Fig. 1d). The Foxp3+ Treg subset increased 1.8-fold when 108 cocultured with the ASF519 SSM-stimulated BMDCs (Fig. 1c). A weak stimulatory 109 effect on Treg proliferation was observed with three other members (ASF356, 110 ASF361, and ASF502) but the effect peaked at distinct bacterial growth stages. Their 111 impacts on both cytokine profiles and Treg proliferation demonstrate a "division of 112 labor" among ASF members in regulating immunophenotypes of BMDCs (Fig. 1b). 113 In particular, ASF519 promotes tolerogenic phenotypes of BMDCs via inducing anti-114 inflammatory cytokines and Treg proliferation. Considering the high abundance of 115 ASF519 in the feces of ASF mice 5 and the relatively high prevalence (~23%) of P. 116 goldsteinii in the human gut microbiome according to The Human Microbiome 117 Project (Extended Data Fig. 2), we were compelled to further explore its 118 immunoregulatory activities. 119

Synergistic regulation of myeloid cell-specific IL-10 induction 120
Because IL-10 is produced by a variety of immune cell types and is involved in both 121 innate and adaptive immunity, we tested the IL-10 inducing specificity of ASF519 on 122 a panel of diverse immune cells. We found that stimulation of isolated DCs, bone 123 marrow-derived macrophages, and isolated splenic monocytes all yielded a high 124 level of IL-10 (Fig. 1d); however, neither splenic panB nor panT stimulation, nor 125 stimulation of immune-irrelevant BEND3 endothelial cells, showed a significant 126 impact on IL-10 production. These data suggest that ASF519-mediated IL-10 127 induction is specific to myeloid-derived antigen-presenting cells. We also observed 128 that IL-10 induction is independent of TNF-α signaling (Extended Data Fig. 3a), and 129 that it is a slow response, which ramps up after 24 hrs of monotonous stimulation of 130 BMDCs by ASF519 SSM (Extended Data Fig. 3b and 3c). 131 functions, cell communication, apoptosis, and other biological processes (Extended 139 Data Fig. 4e). Notably, negative regulation of antigen processing and presentation of 140 peptide or polysaccharide antigen via MHC class II, and negative regulation of 141 oligodendrocyte apoptotic process, are enriched most in upregulated and 142 downregulated gene sets, respectively. At the gene level (Fig. 1g, Extended Data 143 Fig. 4f), we observed a dramatic induction of tolerogenic Il10, lipocalin-2 (Lcn2), and 144 indoleamine 2, 3-dioxygenases (Ido1/2), as well as a reduction of the co-stimulatory 145 molecule Cd40 and inflammatory cytokines, including Il12b, Il23a, and Il1f9 (or Il-36 146 gamma). It is known that IL-36 gamma promotes inflammation and DC maturation; 147 IL-12B and IL-23A together build up the heterodimeric inflammatory IL-23 which is 148 associated with autoimmune diseases and cancer. Their decreased transcript 149 abundance may additively enhance the tolerogenic phenotypes of BMDCs. It is also 150 notable that this step-wise stimulation with ASF519 SSM added before zymosan 151 exerted a synergistic and antagonistic effect on certain transcripts (Fig. 1g,  152 Extended Data Fig.4g-4j). For instance, Il10, Ido2, and U90926 were synergistically 153 induced; whereas suppression of Il12b, Il1f9, and Cd40 by ASF519 SSM was 154 attenuated by subsequent zymosan stimulation. We validated this synergy on Il10 at 155 the protein level (Fig. 1h). Therefore, components of ASF519 SSM can induce 156 multifaceted tolerogenic phenotypes of BMDCs in vitro. 157

Microbial activation of adenosine receptor A2a induces IL-10 production 158
To further dissect this immunomodulatory mechanism, we sought to identify the 159 responsible molecules using bioanalytical techniques (Fig. 2a). The IL-10 inducing 160 function of ASF519 SSM was largely heat-labile (Extended Data Fig. 5a) and 161 immobilized proteinase K digestion of ASF519 SSM completely abolished IL-10 162 induction (Fig. 2b). These results, plus a dose-dependent IL-10 induction by 163 concentrated crude proteins (Fig. 2c), confirmed the proteinaceous nature of the 164 responsible molecules. We then performed gel filtration to fractionate crude proteins 165 in ASF519 SSM. BMDC assays revealed consecutive fractions of an early elution 166 peak exhibiting strong IL-10 induction (Fig. 2d). The elution chromatogram 167 demonstrated that these bioactive fractions were in the excluded volume of the gel 168 filtration column that was designed for separating 20-8,000 kDa globular proteins 169 (Extended Data Fig. 5b-5c), suggesting a large structure is formed. Native-and complex architecture that is susceptible to proteinase K digestion (Extended Data 172 To test whether ASF519-mediated IL-10 induction is mediated by the 186 activation of adenosine receptors, we applied adenosine receptor-specific 187 antagonists to BMDC assays. Intriguingly, supplementation of the A2a inhibitor 188 ZM241385 drastically dampened IL-10 induction in a dose-dependent manner in 189 response to ASF519 SSM (Fig. 2e). By contrast, antagonists of A1, A2b, and A3 at 190 the tested concentrations had no significant impact (Fig. 2e). The engagement of 191 A2a in IL-10 induction was pharmacologically reproduced using bioactive gel-192 filtration fractions as stimuli of BMDCs in inhibition assays (Extended Data Fig. 5e) 193 and further confirmed with the genetic knockout of A2a where A2a -/-BMDCs disabled 194 IL-10 induction when exposed to ASF519 SSM (Fig. 2f). A2a activation then 195 increased the intracellular level of cyclic AMP (Extended Data Fig. 5f), which is 196 known to be involved in the transcriptional regulation of Il10. Therefore, activation of 197 adenosine receptor A2a signaling in BMDCs is required for IL-10 induction by 198 microbial proteins. 199 AHCY can hydrolyze S-adenosylhomocysteine (SAH) to generate adenosine, 200 a natural ligand of adenosine receptors. To validate its role in IL-10 induction, we first 201 applied the AHCY inhibitor DZA to BMDC assays, which revealed that AHCY 202 inhibition decreased IL-10 induction in a dose-dependent manner (Fig. 2e). We then 203 heterologously expressed and purified recombinant enzymes (Extended Data Fig. To validate the role of adenosine receptor A2a in ASF519-mediated 239 immunoregulation in vivo, we blocked A2a signaling by i.p. injection of an A2a 240 antagonist, ZM 241485, into antibiotic-pretreated mice that received live or heat-241 killed ASF519 (Fig. 3f). Immunophenotyping of colonic lamina propria cells revealed 242 an increase in IL-10 producing DCs among CD103-CD11c+ and CD103+ CD11c+ 243 populations in live versus killed ASF519 recipients ( Fig. 3g and 3h). Importantly, 244 these effects were significantly decreased by the A2a antagonist. This IL-10 245 immunophenotype is consistent with our in vitro observation that blockage of A2a 246 rather than other adenosine receptors sufficiently interrupted IL-10 production by 247 BMDCs in response to ASF519 SSM. Other major immune cell types, such as 248 CD11b+ F4/80+ macrophages, CD4+ T cells, and CD19+ CD45+ B cells, did not 249 show significant changes in their IL-10 secretion (Extended Data Fig. 9 and 10). It 250 is also notable that ASF519-treated mice exhibited decreased lamina propria B cells 251 and CD11b+ CD45+ cells in the colon ( Fig. 3i and 3j), which was also restored by 252 A2a blockage. Together, these data demonstrate that the immunoregulatory function 253 of ASF519 in colonic lamina propria goes beyond IL-10 induction in DCs to include 254 more general homeostatic activities of leukocytes through A2a activation. we introduced ASF519 to T1D-prone non-obese diabetic mice (NOD) by weekly oral 261 gavage from four to thirteen weeks of age (Fig. 4a). At 14 weeks, ASF519 recipients 262 had a lower incidence of hyperglycemia (>250 mg/dL, 57%) compared to killed 263 ASF519 recipients (88%) (Fig. 4b), indicating an improved pancreatic function as a 264 result of ASF519 supplementation. To measure immunopathogenesis directly in the 265 pancreas, mice were sacrificed at 14 weeks old to perform histological analysis of 266 insulitis by scoring infiltration of insulin-producing beta cells by pathogenic 267 lymphocytes. In line with improved glucose control, the severity of insulitis was 268 reduced in ASF519 mice (Fig. 4c). Insulin immunostaining on adjacent sections of 269 histological staining also demonstrated that more insulin-producing beta cells were present in the islets of ASF519 mice. Therefore, ASF519 colonization in NOD mice 271 plays a protective role against T1D progression. 272 We then asked whether ASF519 could ameliorate other autoimmune 273 diseases. Rheumatoid arthritis is an autoimmune disease that mainly involves the 274 synovial tissues within joints. We introduced ASF519 to DBA mice before the 275 injection of collagen to induce arthritis (Fig. 4d). The qPCR analysis confirmed the 276 long-term colonization of ASF519 in the gut of DBA mice as after 3 weeks of the final 277 gavage, the abundance of ASF519 in feces was >11 fold higher in live versus killed 278 ASF519 recipient mice (Fig. 4e). Following 30 days post-collagen immunization, 279 mice that received killed ASF519 significantly lost body weight (Fig. 4f), whereas 280 ASF519 mice were able to maintain their body weight. In line with that, killed ASF519 281 recipient mice developed much more severe arthritis than ASF519 mice (Fig. 4g). 282 These data demonstrate that ASF519 colonization strongly protected DBA mice from 283 collagen-induced arthritis. 284

Discussion 285
Accumulating evidence from mouse and human studies demonstrates that the 286 microbiome is integral to immune regulation, and that alterations to the microbiome 287 may underlie many immune-mediated and metabolic diseases. Here, we leverage 288 the ASF, an important and long-studied "model microbiome" to demonstrate that 289 there exists a division of labor among members of the microbiome in modulating 290 immune regulatory phenotypes. This led us to identify that one specific member, P. 291 goldsteinii ASF519, is a potent inducer of IL-10 in myeloid cells in an adenosine 292 receptor A2a-dependent manner both in vitro and in vivo. By a combination of 293 activity-guided fractionation and genetic methods, we identified that enzymes 294 belonging to the methionine/adenosine cycling pathway were sufficient to induce IL-295 10 in dendritic cells. Furthermore, gnotobiotic mice mono-colonized with ASF519 296 exhibited a profound increase in cecal abundance of adenosine, adenine, and S-297 adenosylhomocysteine. This evidence combined with our data showing that the IL-298 10 induction is A2a-dependent demonstrates that the IL-10-inducing activity of 299 ASF519 is mediated by adenosine signaling. Finally, we demonstrated that ASF519 300 supplementation can reduce the severity of arthritis in the CIA model as well as 301 insulitis resulting from autoimmune diabetes in NOD mice.
The novel mechanism we here describe is distinct from known microbially 303 involved IL-10 regulation, including IL-10 induction in DCs, macrophages, or 304 monocytes by bacterial polysaccharide capsules 26-27 , and SCFA-mediated induction 305 through GPCR43 activation in microbiota antigen-specific Th1 cells 28 . Our study 306 broadens the previously described GPCR interactions with gut microbial 307 metabolites 29,30 . In regard to the critical role of DCs in both innate and adaptive 308 immunity, the increased IL-10 induction by ASF519 promotes mucosal immune 309 tolerance by reducing antigen presentation to lymphocytes and the immunogenic 310 potency of leukocytes in colonic lamina propria. As A2a signaling is not only involved 311 in immunomodulation, but also metabolism 31 and cell proliferation 32 , ASF519 312 mediated A2a activation may modulate multiple biological functions that together 313 manifest the observed benefits in disease models here described. We tested the engagement of adenosine receptor A2a in ASF519-mediated 344 immune regulation through pharmacological blockade in mice. 9-11-wk old C57BL/6J 345 mice were treated by an antibiotic cocktail (0.5 g/L vancomycin, 1.36 g/L neomycin, 346 plus 3.75 g/L sweeter aspartame) for 48 hrs. Then, the antibiotic water was 347 exchanged with a 10%(w/w) PEG4000 solution in water, and the mice were fasted 348 for 16 hrs. After that, we administered these prepared mice twice at the interval of 3 349 days by gavaging 200 μl of live or heat-killed P. goldsteinii in PBS at 1x10^9 CFU. We grew each ASF member in the brain heart infusion broth (BHI) under the 383 anaerobic condition at 37℃. The BHI medium consists of 37 g/L BHI (Sigma), 5 g/L 384 yeast extract, 0.5 g/L L-cysteine, 10 mg/L resazurin, and 5 mg/L hemin, dissolved in 385 1/2 (v/v) miniQ water and 1/2 (v/v) tap water. After autoclaving, 1 L of BHI was 386 supplemented with 0.2 ml 1% vitamin K, 50 ml newborn calf serum, and 50 ml horse 387 serum, and 10 ml of 20% (m/v) filter-sterilized yeast extract. The medium was placed 388 in the anaerobic chamber overnight to reduce oxygen before bacterial inoculation.

Cell isolation and cultivation 401
We developed bone marrow-derived dendritic cells (BMDCs) from mouse bone 402 marrow. Mice at 8-12 wk old were euthanized and sacrificed to collect hind legs. 403 After removing skin, flesh, and muscles from the legs, bone marrow was flushed out 404 from bones and collected for red blood cell lysis using the RBC lysis buffer 405

Metagenomic data analysis 501
The genomic DNA was extracted from feces or cecal contents using the 502 ZymoBIOMICS DNA kit (Zymo #D6300). Shotgun metagenomic sequencing was 503 performed using Illumina HiSeq2000 NextGen Sequencer. Data was cleaned by 504 removing mouse genome contamination. Microbial abundance was analyzed by 505 MetaPhlAn2 38 . The prevalence of Parabacteroides goldsteinii in the human gut 506 microbiome studies was analyzed by the curatedMetagenomicData package 39 in R. 507

Bioassays for treated or fractionated samples 508
To perform acetone precipitation, we mixed 4 times the sample volume of cold (-509 20°C) acetone to ASF519 SSMs or medium broth in an acetone-compatible bottle 510 and then incubated the mix for at least 60 min at -20°C. The precipitate was 511 collected by centrifugation at 13,000g for 10 min, then air-dried at room temperature 512 for 30 min before adding 1 x PBS buffer to dissolve the protein pellet on ice. Protein 513 concentration was measured by the Pierce BCA protein assay kit (Thermo Fisher 514 #23227). We then evaluated the IL-10 inducing capability of acetone precipitates at 515 varying working concentrations in BMDC assays. To verify the role of proteinaceous 516 components in IL-10 induction and Treg proliferation, we treated SSMs of ASF519 517 grown in either the defined medium or supplemented BHI with immobilized to reconstitute 10 U protease-agarose power (1U/200 μl, Sigma #P9290-10UN), 520 mixed 1 ml protein sample with 200 μl of reconstituted protease, killed-protease 521 (generated by heating at 95°C for 10 min), or 33 mM Tris-HCI pH7.5, for incubation 522 at 37°C for 1 hr, followed by centrifugation to remove protease-agarose beads. By 523 following the in vitro BMDC stimulation protocol described above, we evaluated the 524 IL-10 inducing capability of heat-and proteinase K treated samples. 525 For gel filtration to fractionate proteins by size, we loaded the concentrated 526 crude samples, following filtration using 0.2 μm syringe filters, to the AKTA pure 527 system with a 10 ml sample injection loop and a HiLoad 16/600 Superdex 200 pg 528 column. Proteins were eluted into 1x PBS at 1 ml/min and collected every 2 ml in 529 sterile microtubes. We then ran in vitro BMDC assays using 10 μl of each fraction as 530 the stimulus to determine bioactive fractions that can induce IL-10 production. The 531 bioactive fractions were pooled and concentrated for further analysis, including 532 protein quantification, native-/SDS-PAGE analysis, and protein identification by mass 533 spectrometry. The gel filtration standard (Bio-Rad #1511901) was analyzed with the 534 same gel filtration settings to indicate the molecular size. 535

Expression of His-tagged proteins 554
The The impact of pure enzymes on IL-10 induction was evaluated by adding pure 574 enzymes into BMDC assays with 1 x 10 5 cells in 200 μl in a 96-well microplate by 575 following the in vitro step-wise stimulation protocol. The level of IL-10 in cell culture 576 supernatants was determined by the IL-10 ELISA kit. 577

Histological analyses 578
We performed hematoxylin and Eosin (H&E) staining to examine infiltration in 579 pancreatic islets. The slides were deparaffinized in two washes of xylenes at 5 min 580 and then rehydrated in series of decreasing Ethanol concentration, 100%, 95%, 581 70%, and 50% and distilled water. The slides were stained with hematoxylin (Gill-3, destained (3-4 dips) with acid water and raised with distilled water then blued with 584 ammonia water. After washing with distilled water and a quick 95% ethanol wash, 585 the slides were stained with eosin (Thermo Scientific #6766007) and dehydrated with 586 two washes of 95% ethanol, two washes of 100% ethanol, and two washes of 587 xylene. The slides were mounted with permount (Fisher # SP15-500), cover with a 588 coverslip, and left to dry overnight. Images were taken with EVOS® FL Auto Imaging wash with distilled water and counterstained with hematoxylin (Gill-3, Fisher #3537-609 32) for 30 sec, and raised in running distilled water. The slides were then blued with 610 ammonia water for 3 min and raised in running distilled waters. The slides were 611 dehydrated with a series of washes with 70% ethanol for 5 minutes, 95% ethanol for 612 2 min, 100% ethanol for 2 min, and xylene for 4 min. The slides were mounted with 613 permount (Fisher #SP15-500), covered with a coverslip, and left to dry overnight. 614 Images were taken with EVOS® FL Auto Imaging System (Life Technologies) at 20x 615 in brightfield. We verified that the islets without insulitis still had functional beta cells 616 by immunostaining for insulin.