Inhibition of Glycolysis in Pathogenic TH17 Cells Through Targeting a Mir-21–Peli1–C-Rel Pathway Prevents Autoimmunity

Published May 15, 2020, doi:10.4049/jimmunol.2000060 The Journal of Immunology

Inhibition of Glycolysis in Pathogenic TH17 Cells through Targeting a miR-21–Peli1–c-Rel Pathway Prevents Autoimmunity

Rong Qiu,*,†,1 Xiang Yu,*,1 Li Wang,* Zhijun Han,‡ Chao Yao,† Yange Cui,† Guojun Hou,* Dai Dai,* Wenfei Jin,‡ and Nan Shen*,x,{,||,#

It is well known that some pathogenic cells have enhanced glycolysis; the regulatory network leading to increased glycolysis are not

well characterized. In this study, we show that CNS-inﬁltrated pathogenic TH17 cells from diseased mice speciﬁcally upregulate glycolytic pathway genes compared with homeostatic intestinal TH17 cells. Bioenergetic assay and metabolomics analyses indicate that in vitro–derived pathogenic TH17 cells are highly glycolytic compared with nonpathogenic TH17 cells. Chromatin landscape analyses demonstrate TH17 cells in vivo that show distinct chromatin states, and pathogenic TH17 cells show enhanced chromatin accessibility at glycolytic genes with NF-kB binding sites. Mechanistic studies reveal that miR-21 targets the E3 ubiquitin ligase

Peli1–c-Rel pathway to promote glucose metabolism of pathogenic TH17 cells. Therapeutic targeting c-Rel–mediated glycolysis in pathogenic TH17 cells represses autoimmune diseases. These ﬁndings extend our understanding of the regulation TH17 cell glycolysis in vivo and provide insights for future therapeutic intervention to TH17 cell–mediated autoimmune diseases. The Journal of Immunology, 2020, 204: 000–000.

+ subset of effector CD4 TH cells that produce IL-17A lamina propria (3, 4). Under autoimmune or pathogen infection (1, 2), TH17 cells, mediates barrier tissue integrity and condition, TH17 cells acquire the capability to produce IFN-g A mucosal defense and contributes to the development and GM-CSF and are often seen at sites of inflamed loci (5). of multiple autoimmune diseases. At steady-state condition, TH17 cells also could be differentiated in vitro by a combi- TH17 cells producing IL-10 are mainly found at sites of intestinal nation of cytokines, either by TGF-b1plusIL-6(TH17 [b], nonpathogenic) or by IL-6 plus IL-1b and IL-23 (TH17 [23], pathogenic) (6, 7). *Shanghai Institute of Rheumatology, Renji Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai 200001, China; †Shanghai Institute of Nutrition and Upon Ag stimulation, lymphocytes undergo extensive clonal ex- Health, Shanghai Institutes for Biological Sciences, University of Chinese Academy pansion and differentiation for immune defense or tolerance (8–11). ‡ of Sciences, Chinese Academy of Sciences, Shanghai 200001, China; Department of Activated lymphocytes are highly glycolytic and demonstrate a Biology, Southern University of Science and Technology, Shenzhen 518000, China; xShenzhen Futian Hospital for Rheumatic Diseases, Shenzhen 518000, China; {State striking increase in cell content and nutrients uptake. c-Myc, c-Rel, Key Laboratory of Oncogenes and Related Genes, Shanghai Cancer Institute, Renji and mTORc1-dependent glycolytic activity is crucial for effective T Hospital, Shanghai 200001, China; ||Center for Autoimmune Genomics and Etiology, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH 45229; and #Department and B cell immune response (12–15), whereas Foxo1 and Foxp3 of Pediatrics, University of Cincinnati College of Medicine, Cincinnati, OH 45229 proteins function as key metabolic repressors that limit glycolytic 1R.Q. and X.Y. contributed equally. activity in regulatory T cells (16, 17). It was reported that by ORCIDs: 0000-0001-7395-0419 (L.W.); 0000-0001-9374-2510 (C.Y.); 0000-0002- restricting glucose availability, tumor cells limit aerobic glycolysis 0028-3739 (W.J.). and effector function of tumor-infiltrating T cells (18, 19). Recently, it Received for publication January 21, 2020. Accepted for publication April 1, 2020. was demonstrated that TH17 cells developed in the presence of seg- This work was supported by grants from the National Natural Science Foundation of mented filamentous bacteria or Citrobacter rodentium show func- China (31630021, 81421001, and 31930037), the Shanghai Municipal Key Medical Center tional and metabolic heterogeneity, with Citro rodentium–induced Construction Project (2017ZZ01024-002), the Shanghai Sailing Program (17YF1417900), the Shenzhen Science and Technology Project (JCYJ20180504170414637), and the pathogenic TH17 cells showing a global upregulation of multiple Sanming Project of Medicine in Shenzhen (SZSM201602087). This work was done metabolic pathways (20). by an innovative research team of high-level local universities in Shanghai led by N.S. Tight communication between epigenetic remodeling and meta- R.Q., X.Y., W.J., and N.S. designed the experiments and wrote the manuscript; R.Q. bolic regulation for cell plasticity is beginning to be understood and X.Y. did most of the experiments; L.W. helped with the mouse experiments; Z.H. and C.Y. helped with ATAC-seq and RNA-seq data analysis; Y.C., D.D., and G.H. (21, 22). Assay for transposase-accessible chromatin technol- helped with RT-PCR experiments. ogy sequencing (ATAC-seq) was developed to study global The sequences presented in this article have been submitted to the Gene Expression chromatin dynamics by using relatively low cell number (23). It Omnibus (https://www.ncbi.nlm.nih.gov/geo/) under accession number GSE127768. was successfully used to study chromatin dynamics of regula- Address correspondence and reprint requests to Prof. Nan Shen, Shanghai Institute of tory T cells under inflammatory condition (24), development of Rheumatology, Renji Hospital, Shanghai Jiao Tong University School of Medicine, 145 plasmacytoid dendritic cells (25), and tumor-specific CD8+ Shan Dong Road, Shanghai 200001, China. E-mail address: [email protected] T cells (26). However, chromatin architecture of T 17 cells The online version of this article contains supplemental material. H in vivo is unknown, and signaling pathways that induce TH17 Abbreviations used in this article: ATAC-seq, assay for transposase-accessible chromatin technology sequencing; EAE, experimental autoimmune encephalomyelitis; cell chromatin remodeling are also incompletely understood. ECAR, extracellular acidification rate; FC, fold change; GC-TOFMS, gas chromato- microRNAs are a class of noncoding RNAs that modulate gene graph time-of-flight mass spectrometry; PTXF, pentoxifylline; RIP, RNA-binding expression at the posttranscriptional level (27). Specific micro- protein immunoprecipitation; RNA-seq, RNA sequencing. RNAs have been reported to be essential for the metabolic func- Copyright Ó 2020 by The American Association of Immunologists, Inc. 0022-1767/20/$37.50 tion of certain cells such as miR-33a/b, which was shown to be

www.jimmunol.org/cgi/doi/10.4049/jimmunol.2000060 2 miR-21 CONTROLS GLYCOLYSIS OF PATHOGENIC TH17 CELLS crucial for cholesterol and lipid metabolism (28, 29), and let-7 Intracellular cytokine staining family members were recently shown to be essential for B cell Cultured or tissue isolated lymphocytes were washed and stimulated with metabolic function (30). The regulation of TH17 cell metabolic PMA (50 ng/ml) plus ionomycin (750 ng/ml) for 4 h at 37˚C. Cells were reprogramming in vivo to distinct microenvironment by micro- stained with LIVE/DEAD BV510 (catalog no. L34955; Thermo Fisher RNAs is not well understood. Scientific), anti-CD45 APC-Cy7 (catalog no. 557659; BD Biosciences), In this study, first by bioenergetic and metabolomics study, we anti-CD3 PerCP5.5 (catalog no. 35-0031-82; eBioscience), and anti-CD4 BV421 (catalog no. 562891; BD Biosciences) for 30 min at 4˚C. Cells demonstrate that in vitro–derived nonpathogenic TH17 (b) cells were then fixed, permeabilized with Perm/Wash buffer (catalog no. show reduced glycolytic activity compared with pathogenic TH17 555028; BD Biosciences), and stained with anti–IFN-g APC (catalog no. (23) cells. Further, by global transcriptional analyses, we dem- 554413; BD Biosciences), anti–GM-CSF PE (catalog no. 554406; BD onstrate that homeostatic T 17 cells derived in vivo show re- Biosciences), or anti–IL-17A PerCP5.5 (catalog no. 560666; BD Biosci- H ences) for 30 min at 4˚C, or cells were fixed, permeabilized with Foxp3 pressed glycolytic pathway gene expression. Chromatin landscape Transcription Factor Staining Buffer Set (catalog no. 00-5523-00; Thermo analyses indicate that TH17 cells in vivo show distinct chromatin Fisher Scientific), and stained with anti-Foxp3 APC (catalog no. 17-5773-82; states. Pathogenic TH17 cell–specific open chromatin region are eBioscience) and anti-RORc PE (catalog no. 562607; BD Biosciences). + enriched for motifs of NF-kB family transcription factors, and Cytokine and transcription factor profiles of CD4 cells were analyzed on an FACSCalibur with FlowJo software (Tree Star). pathogenic TH17 cells show enhanced chromatin accessibility at glycolytic genes. Mechanistic studies reveal that miR-21 targets Immunoblots the E3 ubiquitin ligase Peli1–c-Rel pathway to promote glucose metabolism of pathogenic T 17 cells. Pharmaceutical inhibition TH17 cells were lysed with RIPA Lysis and Extraction Buffer (catalog no. H 89900; Thermo Fisher Scientific) supplemented with Protease and Phos- c-Rel–mediated glycolysis in pathogenic TH17 cells represses phatase Inhibitor Cocktail (catalog no. 78441; Thermo Fisher Scientific). autoimmune diseases. Individual cell lysates (10 mg/lane) were separated by 10% SDS-PAGE and transferred to an Immobilon-P PDVF Membrane (MilliporeSigma, Materials and Methods Billerica, MA). After being blocked with SuperBlock T20 PBS Blocking Buffer (catalog no. 37516; Thermo Fisher Scientific), the membranes were Mice incubated with Abs listed below: Peli1 (catalog no. 31474S; Cell Signaling IL-17A–enhanced GFP reporter mice were from Biocytogen (Beijing, Technology), c-Rel (catalog no. 12707S; Cell Signaling Technology), China). miR-212/2 and miR-21f/+ mice on the C57B6 background were RelA (catalog no. 8242S; Cell Signaling Technology), Slc2a1 (catalog no. from Shanghai Model Organisms (Shanghai, China). CD4-cre, Lyz2-cre, ab115730; Abcam), HK2 (catalog no. 2867S; Cell Signaling Technology), CD11c-cre, and 2D2 TCR transgenic mice were from The Jackson Lab- and tubulin (catalog no. 2146S; Cell Signaling Technology). oratory. All mice were maintained under specific pathogen-free conditions. All animal experiments were performed in compliance with the guide for Ago2 immunoprecipitation the care and use of laboratory animals and were approved by the Institu- Ago2 immunoprecipitation was performed according to the protocol pro- tional Animal Care and Use Committee at Renji Hospital of Shanghai Jiao vided with the Magna RIP RNA-Binding Protein Immunoprecipitation Kit Tong University School of Medicine. (catalog no. 17-700; Merck Millipore). For the RNA-binding protein im- 2/2 Cell culture munoprecipitation (RIP) lysate preparation, control and miR-21 TH17 (23) cells were collected by centrifugation at 1500 rpm for 5 min and then Naive CD4+ T cells were isolated from lymph nodes or spleen of 6–8-wk-old lysed with RIP lysis buffer. For the magnetic bead/Ab preparation, purified mice using a Naive CD4+ T Cell Isolation Kit (catalog no. 130-104-453; Abs (5 mg) were incubated with beads for 30 min at room temperature Miltenyi Biotec) according to the manufacturer’s instruction. Naive CD4+ using anti-Ago2 (catalog no. 03-110; Merck Millipore), as the Ab of in- T cells were stimulated with plate-bound anti-CD3 mAb (catalog no. terest, and normal mouse IgG as negative control. Supernatant from the 16-0038-85, 5 mg/ml; Thermo Fisher Scientific) in the presence of anti- RIP lysate was obtained by centrifugation at 14,000 rpm for 10 min; CD28 mAb (catalog no. 16-0289-85, 2 mg/ml; Thermo Fisher Scientific) in 100 ml was incubated overnight at 4˚C with bead–Ab complex, and 10 ml a 48-well plate under neutral condition (catalog no. 16-7041-95, 10 mg/ml served as 10% input for generating a standard curve or for RT-PCR anti–IL-4 mAb; Thermo Fisher Scientific; and catalog no. 16-7311-38, comparison. Immunoprecipitated RNAs were purified and analyzed by 10 mg/ml anti–IFN-g mAb; Thermo Fisher Scientific), TH17 (b) condition quantitative RT-PCR for Peli1 mRNA quantification. (catalog no. 7666-MB-005, 2 ng/ml TGF-b1; R&D Systems; catalog no. 406-ML-025, 20 ng/ml IL-6; R&D Systems; catalog no. 16-7041-95, Induction of experimental autoimmune encephalomyelitis 10 mg/ml anti–IL-4 mAb; Thermo Fisher Scientific; and catalog no. IL-17A–enhanced GFP reporter mice were injected s.c. in the tail base with 16-7311-38, 10 mg/ml anti–IFN-g mAb; Thermo Fisher Scientific), and MOG peptide (catalog no. 163913-87-9, 200 mg/mouse; ChinaPeptides) T 17 (23) condition (catalog no. 406-ML-025, 20 ng/ml IL-6; R&D 35–55 H in CFA (catalog no. 263910, catalog no. 231141; BD Difco). Five minutes Systems; catalog no. 401-ML-025, 20 ng/ml IL-1b; R&D Systems; catalog and forty-eight hours after the injection of MOG peptide, the mice were no. 1887-ML-010, 20 ng/ml IL-23; R&D Systems; catalog no. 16-7041-95, 35–55 injected i.p. with pertussis toxin (catalog no. P7208, 200 ng/mouse; Sigma- 10 mg/ml anti–IL-4 mAb; Thermo Fisher Scientific; and catalog no. Aldrich). Status of the mice was monitored, and disease severity was scored 16-7311-38, 10 mg/ml anti–IFN-g mAb; Thermo Fisher Scientific). Cells three times a week as follows: 0, no clinical signs; 1, limp tail (tail paralysis); were cultured in complete IMDM medium (Life Technology) supple- 2, complete loss of tail tonicity or abnormal gait; 3, partial hind limb pa- mented with 10% FBS and 2-ME (1:1000). ralysis; 4, complete hind limb paralysis; and 5 = forelimb paralysis or In vitro pentoxifylline treatment moribund. For adoptive transfer experimental autoimmune encephalomyelitis (EAE), naive 2D2 TCR transgenic CD4+ T cells were differentiated + Naive CD4 T cells were differentiated under pathogenic TH17 (23) into TH17 (23) cells for 4 d and treated with H2O or 500 mg/ml PTXF for the condition for 2 d; cells were collected and preincubated with 500 mg/ml last 16 h. The cells were harvested and washed twice with PBS and 7 pentoxifylline (PTXF) (catalog no. P1784; Sigma-Aldrich) or H2O for resuspended at 1 3 10 cells/ml. Two hundred microliters of cell sus- 15 min at 37˚C. Total-cell suspensions were then activated with 5 mg/ml pension (2 3 106) cells were injected i.p. into each recipient C57BL/6 plate-coated anti-CD3, 2 mg/ml soluble anti-CD28, 20 ng/ml IL-6, 20 ng/ml mouse with sublethal irradiation. EAE is scored as same as shown above. IL-1b, 20 ng/ml IL-23, 10 mg/ml anti–IL-4 mAb, and 10 mg/ml anti–IFN-g Cells were isolated from whole brain and spinal cord at the peak of disease mAb overnight at 37˚C in the presence of PTXF. by Percoll gradient centrifugation (37%/70%) and subjected to flow cytometric analysis or flow cytometric sorting. Adenovirus transduction Quantitative RT-PCR For overexpression of GFP or c-Rel in nonpathogenic TH17 (b) cells, naive + CD4 T cells were differentiated under nonpathogenic TH17 (b) condition Total RNA was extracted from in vitro–differentiated TH17 (b), TH17 (23), for 2 d, then adenovirus supernatant containing GFP or c-Rel (Hanbio, H2O, or PTXF-treated TH17 (23) cells with TRIzol Reagent (Sigma- Shanghai, China) was added, and the cells were spun at 2000 rpm for 1.5 h Aldrich). The expression of the genes encoding mouse Slc2a1 (forward: at 30˚C. After spin infection, the cells were cultured in TH17 (b) condition 59-GAGTGCAGGGAGGAGAGGG-39, reverse: 59-GTGTCCGTGTCTT- and harvested on day 5 for sorting of GFP+ transduced cells. CAGCAGTT-39), Slc2a3 (forward: 59-CACATTTCAGACCCTTTAGGC-39, The Journal of Immunology 3 reverse: 59-ACAGGGAGTGGAGAGTTGAAT-39), Hk1 (forward: 59-GG- compound annotation, statistical analysis, and pathway analysis were TACGCTTAGGTGGGCAG-39, reverse: 59-ACACCACTCCGACTTTTG- processed using XploreMET software (v3.0), as described in a previous AAT-39), Hk2 (forward: 59-TGCTAGGTTGACAGCTCTCTCT-39, reverse: publication (34). Compound identification for GC-TOFMS was performed 59-ACAAAACGCTCACTAGACCGA-39), Hk3 (forward: 59-GCTTTGGT- by comparing the mass fragments with JiaLib (v1.0.0) mass spectral da- TACTGCTGTTGCT-39, reverse: 59-AGTTGTGGGTGGCTGAGGA-39), tabases. The principal component analysis and orthogonal partial least- Pfkfb3 (forward: 59-GACTCGCTTCTCAGGTTTTTG-39, reverse: 59- squares discriminant analysis, Student t test, fold change (FC), and the AGCACACACACACACAACCAC-39), Aldoa (forward: 59-GGGTGGG- heatmap of z-score were also performed with XploreMET software. Overall, AAGAAGGAGAAC-39, reverse: 59-CCACTTGGGGTATACTTTCCT-39), we detected ∼200 measurable and reproducible metabolite signals; of them, Tpi1 (forward: 59-CATCATCTGCCCCATCTTGT-39, reverse: 59-GTCA- 80 were annotated. The differential metabolites were obtained using uni- GTTCTACTCCCTCCCCT-39), Gpi1 (forward: 59-TGGGTAGGGTGGG- variate statistical analysis (Student t test), as shown in Supplemental Table I. GTCAG-39,reverse:59-GAGGAGGGGAAGGAGGTG-39), Pgam1 (forward: FC . 1(p , 0.05) indicates a relatively higher concentration present in 59-GGAGGAACTGTGCTGTGGTGT-39,reverse:59-CAGGGTCAGAGAG- TH17 (23) group, whereas FC , 1(p , 0.05) means a relatively lower AA ATTAGGG-39), Pgm2 (forward: 59-CACGCCACCTGATTGAAGA-39, concentration as compared with the TH17 (23) group. reverse: 59-GGGAAGTAAAAAATCAAGACCG-39), Eno1 (forward: 59-TTGAGGAAGAGCTGGGCAG-39,reverse:59-TTGAGCTGCTGGGAT- Metabolism assays 9 Pgk1 9 9 GACAT-3 ), (forward: 5 -ATTTTCTTTCCTGCCTTTGGT-3 ,reverse: Extracellular acidification rate (ECAR) was determined using Seahorse 59-GTTCCTGGTGCCACATC CA-39), Pkm (forward: 59-AGTAGG- Xfe24 Analyzer (Agilent Technologies). Briefly, TH17 (b), TH17 (23), GGCTGAGGGTGTG-39,reverse:59-AGGAAAGCAGGTCAGGGC- 2/2 6 control, and miR-21 T 17 (23) cells (0.5 3 10 per well) were plated 9 Ldha 9 9 H 3 ), (forward: 5 -TCGTCTGAGCTGTGGTTAGTA-3 ,reverse: on Cell-Tak-coated Seahorse Culture plate for 30 min. ECAR, a measure 9 9 Peli1 9 5 -TTGGCAGCAGGGCAGAG-3 ), (forward: 5 -CCAGCAG- of glycolysis, was measured under basal conditions and in response to 9 9 ACCAGCCTTAACT-3 ,reverse:5-GTCCTCACTTTCATCCGCTAA- glucose (10 mM), oligomycin (1.0 mM), and 2-deoxyglucose (50 mM) 9 Actb 9 9 3 ), and (forward: 5 -GGCTGTATTCCCCTCCATCG-3 , reverse: (Agilent Seahorse XF Glycolysis Stress Test Kit, catalog no. 103020-100). 59-CCAGTTGGTAACAATGCCATGT-39) were quantified by real-time PCR using PrimeScript RT Reagent Kit (Takara Bio). All gene expres- ATAC-seq data processing sion results were normalized to the basal expression of the housekeeping gene Actb. Amplification of cDNA was using an ABI PRISM 7900 HT ATAC-seq reads were trimmed using trim_galore 0.4.4_dev with the default cycler (Applied Biosystems). Quantitative RT-PCR was employed with settings and then mapped to mm10 reference by Bowtie 2 (35) with pa- SYBR Green methodology according to the following situations: 95˚C for rameters “–end-to-end–no-unal -X 2000” low-quality reads (mapping 1 min, then performed with 40 cycles at 95˚C for 15 s, and 60˚C for 30 s quality , 10); singletons and duplicated reads were removed using with 40 cycles. SAMtools (36). To call the differential accessible regions between sam- ples, filtered bam files were downsampled to the same fragment number microRNA array analyses and then processed by regulatory genomics toolbox–THOR with param- eters “–p value 0.05–binsize 100–step 50” (37). Differential peaks derived + ∼ GFP TH17 cells were sorted from in vivo suspensions, 200–500 ileum or from regulatory genomics toolbox–THOR were further filtered according CNS TH17 cells (200–500 TH17 cells from one mouse ileum lamia propria to their significant scores by cutting from the switch point in the rank plot. or inflamed spinal cord, three to four mice per group) were used for lysis, Enriched transcription factor binding motifs were searched by HOMER 39 ligation, 59 ligation, reverse transcription, and preamplification using findMotifsGenome.pl (38) using top 1000 differential accessible peaks miScript Single Cell qPCR Kit (catalog no. 331053; QIAGEN), and cDNA centered by the peak summits and extended to 6100 bp. was used for microRNA quantitative PCR array analyses. The data were normalized to a U6 small nuclear RNA control. microRNA quantitative RNA-seq data processing PCR primers are as follows: miR-21 forward: 59-CGTAGCTTATCAGA- CTGATGTTGA-39, miR-146a forward: 59-TGAGAACTGAATTCCATG- RNA-seq reads were mapped to mm10 and refGene annotation tran- GGTT-39, miR-155 forward: 59-TTAATGCTAATTGTGATAGGGGT-39, scriptome downloaded from University of California, Santa Cruz, Genome , miR-27a forward: 59-CACAGTGGCTAAGTTCCGC-39, miR-24 forward: Browser using TopHat2 (39), low-quality reads (mapping quality 10) 59-GCTCAGTTCAGCAGGAACAG-39, and U6 forward: 59-AGATTAG- were filtered by SAMtools, and then gene level expression was counted by CATGGCCCCTG-39, with the general reverse primer supplied at the htseq-count (40) using all refGenes. Differential expressed genes were miScript Single Cell qPCR Kit (QIAGEN). called by DESeq2 (41) using false discovery rate of 0.05 and FC $ 1.5. For differential metabolic pathway gene expression analyses, these key RNA sequencing metabolic pathway genes are selected: glycolysis pathway genes (Slc2a1, Sl2a3, Hk1, Hk2, Hk3, G6pdx, Pfkl, Pfkfb3, Pfkfb2, Pfkfb1, Aldoa, Aldob, + 2/2 Naive CD4 T cells from control and miR-21 mice were differentiated Aldoc, Tpi1, Gpi1, Pgam1, Eno1, Eno2, Eno3, Pgk1, Pkm, Ldha, Ldhb, and under pathogenic TH17 (23) differentiation condition. Total RNA was Pdk1); glutaminolysis pathway genes (Slc1a5, Slc3a2, Slc7a5, Slc38a1, prepared from these cells using TRIzol Reagent (Invitrogen). RNA se- Gls2, Ppat, Gls, Gfpt1, Cad, Glud1, Oat, Got1, and Got2); TCA cycle quencing (RNA-seq) libraries were prepared using a TruSeq Stranded Total (Acly, Aco1, Aco2, Cs, Dlat, Dld, Dlst, Fh1, Idh1, Idh2, Idh3a, Idh3b, + RNA Library Prep Kit (Illumina). For GFP TH17 cells sorted from in vivo Idh3g, Mdh1, Mdh1b, Mdh2, Ogdh, Pck1, Pck2, Pcx, Pdha1, Pdhb, Sdha, suspensions, 50,000 cells were resuspended in 700 ml TRIzol Reagent, total Sdhb, Sdhc, Sdhd, Sucla2, Suclg1, and Suclg2); fatty acids pathway genes RNA was prepared by miRNeasy Micro Kit (catalog no. 217084; QIAGEN), (Acaa1a, Acaa2, Acat1, Acat2, Acad9, Acad10, Acad11, Acadl, Acadm, and RNA-seq libraries were prepared using SMART-Seq v4 Ultra Low Input Acads, Acadsb, Acadvl, Ehhadh, Gcdh, Acox1, Acox2, Acox3, Acsbg1, RNA Kit (catalog no. 634894; Clontech Laboratories). Sequencing was Acsbg2, Acsl1, Acsl3, Acsl4, Acsl5, Acsl6, Acsm2, Acsm3, Acsm4, Acsm5, performed on an Illumina HiSeq X Ten System in a 150 bp/150 bp paired- Acot2, Acot3, Acot6, Acot7, Acot8, Acot9, Acot12, Aldh2, Decr1, Decr2, end mode. Echs1, Hadha, Mcee, Mut, Eci2, Pecr, Ppa1, Cpt1a, Cpt1b, Cpt1c, Cpt2, Crat, Crot, Fabp1, Fabp2, Fabp3, Fabp4, Fabp5, Fabp6, Slc27a1, Slc27a2, ATAC-seq Slc27a3, Slc27a4, Slc27a5, Slc27a6, Prkaa1, Prkaa2, Prkab1, Prkab2, ATAC-seq library preparations were performed as described (23). In brief, Prkaca, Prkacb, Prkag1, Prkag2, and Prkag3); amino acids pathway genes 50,000 cells were washed in cold PBS and lysed. Transposition was performed (Aoc1, Acy1, Agmat, Aldh18a1, Aldh4a1, Aldh9a1, Amd1, Arg2, Asl, Ass1, using Nextera DNA Sample Preparation Kit (catalog no. FC-121-1030; Illu- Ckb, Cps1, Gamt, Gatm, Glud1, Got2, Maob, Nags, Nos2, Oat, Otc, Sat1, mina) at 37˚C for 30 min. After purification of the DNA with the MinElute Srm, Cdo1, Cth, Ldha, Mpst, Sds, Acadm, Acat1, Aox1, Auh, Bcat2, PCR Purification Kit (QIAGEN), DNA was amplified for five cycles. Addi- Bckdha, Dbt, Dld, Ehhadh, Hmgcl, Hmgcs1, Ivd, Lars, Mccc2, Oxct2a, tional PCR cycles were evaluated by real-time PCR. Final product was cleaned Pdhb, Dao, Lap3, Odc1, P4ha1, Prodh, Prodh2, Pycr1, Pycrl, Aadat, by AMPure Beads at a 1.53 ratio. Libraries were sequenced on a HiSeq X Ten Aanat, Acmsd, Cat, Cyp1b1, Ddc, Gcdh, Haao, Ido1, Inmt, Kmo, Kynu, System in a 150 bp/150 bp paired-end mode. Ogdhl, Tdo2, Tph2, Wars, Adi1, Ahcy, Apip, Bhmt, Cbs, Dnmt1, Enoph1, Mat1a, Mtap, Mtr, and Tat). Metabolomics Statistical analysis The untargeted metabolomics profiling was performed on gas chromato- graph time-of-flight mass spectrometry (GC-TOFMS) system by Metab- For other quantitative data, Prism software was used for statistical analysis. oProfile, Shanghai, China. The sample preparation procedures were Differences between groups were compared with a two-tailed unpaired t according to the previously published methods with minor modifications test. A p value ,0.05 was considered statistically significant: *p , 0.05, (31–33). The GC-TOFMS raw data processing, peak deconvolution, **p , 0.01, and ***p , 0.001. 4 miR-21 CONTROLS GLYCOLYSIS OF PATHOGENIC TH17 CELLS

Results showed that amino acid, amino sugar metabolism, and glycolytic intermediates trended to be higher in T 17 (23) cells (Fig. 1D, TH17 cells derived in vitro show different metabolic states H Supplemental Table I). Taken together, these data suggest that To study the metabolic activity of TH17 cells, first we induced TH17 cells generated in vitro have distinct metabolic states; TH17 cells in vitro by a combination of cytokines TGF-b1, IL-6 pathogenic TH17 (23) cells polarized in the absence of TGF-b1 (TH17 [b]) or IL-6, IL-1b, and IL-23 (TH17 [23]). We tested the signaling are highly glycolytic compared with nonpathogenic glycolytic activity of TH17 (b) and TH17 (23) cells by analyzing TH17 (b) cells. key metabolic enzyme gene expression; we found TH17 (23) cells TH17 cells show differential metabolic pathway gene activation express higher glycolytic pathway genes than TH17 (b) cells, and in vivo TH17 (23) cells express great more Slc2a1/3, Hk2, Tpi1, Aldoa, Eno1, Pkm, and Ldha (Fig. 1A, Supplemental Fig. 1A). We further To gain insight of the metabolic states of TH17 cells in vivo, we + checked the glycolytic activity of TH17 (b) and TH17 (23) cells by sorted GFP TH17 cells from homeostatic ileum and inflamed glycolysis stress test; both basal and maximal glycolytic capability CNS of EAE mice (Supplemental Fig. 1C). Then, we sent ileum were much higher in pathogenic TH17 (23) cells (Fig. 1B). andCNS-infiltratedTH17 cells for RNA-seq analyses. We We next subjected TH17 (b)andTH17 (23) cells for high- found 1415 upregulated genes and 1249 downregulated genes resolution metabolomics analyses (31, 32). Pathogenic TH17 in CNS-infiltrated TH17 cells compared with ileum TH17 cells (23) cells were metabolically distinct from TH17 (b) cells (Supplemental Fig. 1D, 1E). Gene Ontology and pathway analyses (Fig. 1C, Supplemental Fig. 1B) and showed higher level in- indicate that several immune pathways were activated in CNS- termediates of carbohydrates, amino acids, and organic acids, infiltrated TH17 cells as reported (7); interestingly, we found whereas TH17 (b) cells showed higher level intermediates of multiple metabolic pathways were enriched (Fig. 2E, 2F). We nucleotides and fatty acids. Pathway analyses of altered metabolites found homeostatic ileum TH17 cells express greatly more Foxp1

FIGURE 1. TH17 cells derived in vitro show different metabolic states. (A) RT-PCR analysis of key glycolytic pathway genes in TH17 (b) and TH17 (23) cells differentiated in vitro for 48 h; gene expression was normalized to Actb.(B) ECAR of TH17 (b) and TH17 (23) cells differentiated for 96 h assessed by a glycolytic stress test. ECAR was measured under basal conditions and in response to glucose (10 mM), oligomycin (1.0 mM), and 2-deoxyglucose (2-DG)

(50 mM). (C) Principal component analysis (PCA) analysis of identified metabolites of TH17 (b) and TH17 (23) cells (n = 5) differentiated in vitro for 48 h by metabolomics. (D) Metabolomics analysis of TH17 (b) and TH17 (23) cells differentiated in vitro for 48 h; the top differentially observed metabolites are shown in heat map. Data are representative of three independent experiments (A and B). Error bars represent SEM. *p , 0.05, **p , 0.01, ***p , 0.001, determined by two-tailed unpaired t test. The Journal of Immunology 5 gene (Fig. 2A), which is critical for maintenance of the metabolic which is consistent with the reports that TGF-b1 signaling represses quiescence of naive T cells (42, 43). Ileum homeostatic TH17 cells c-Rel activity during lymphocytes activation (44). A group of genes also express several transcription factors with repressor activity related to glycolytic pathway were also with decreased chromatin c-Maf and Foxo1, and Foxo1 was a repressor of glucose metabolism accessibility in TH17 (b) cells, which was also consistent with their (16) (Fig. 2A), which is consistent with our finding that CNS- relative expression level (Fig. 3D). Taken together, these data suggest infiltrated TH17 cells express higher level of glycolysis pathway that TH17 (b) cells differentiated in the presence of TGF-b1 signaling genes (Fig. 2B–D), whereas ileum TH17 cells show higher enrich- show repressed chromatin accessibility at glycolytic genes. ment of glutaminolysis pathway genes. Taken together, these data TH17 cells show discrete chromatin states in vivo suggest that TH17 cells in vivo have differential expression of key metabolic pathway genes, with ileum homeostatic TH17 cells To further study possible regulation mechanism for TH17 cell showing repressed glycolysis pathway gene activation. metabolic reprogramming, we performed ATAC-seq to assess chromatin landscape of TH17 cells derived in vivo (Supplemental TH17 cells derived in vitro show discrete chromatin states Fig. 2B, 2E, 2F). We found that TH17 cells had discrete chromatin To dissect regulation mechanism for TH17 cell metabolic hetero- states in vivo (Fig. 4A); CNS-infiltrated TH17 cells showed great geneity, we differentiated TH17 cells in vitro either in the presence differential global chromatin accessibility compared with ileum (TH17 [b]) or absence (TH17 [23]) of TGF-b1 and sorted out homeostatic TH17 cells (Fig. 4B). Transcriptional factor motif + GFP TH17 cells; then, we performed ATAC-seq to assess chro- analysis of top 1000 enriched peaks indicated that NF-kB family matin landscape genome wide (Supplemental Fig. 2A, 2C, 2D). members (RelA and c-Rel) were repressed in homeostatic TH17 We found TH17 cells derived in vitro had discrete chromatin states cells (Fig. 4C). Interestingly, the regulatory region of key TH17 (Fig. 3A); TH17 (23) cells showed great differential global chro- cell–related transcription factors c-Maf and Ahr was with enhanced matin accessibility compared with TH17 (b) cells (Fig. 3B). chromatin accessibility in homeostatic TH17 cells; both c-Maf and Transcriptional factor motif analysis of enriched peaks indicated Ahr were shown to be critical regulators of TH17 cells (Supplemental that NF-kB family members were repressed in nonpathogenic Fig. 2G). The regulatory region of metabolic repressor Foxo1 was TH17 (b) cells derived in the presence of TGF-b1 signaling (Fig. 3C), with reduced chromatin accessibility in CNS-infiltrated TH17 cells,

FIGURE 2. TH17 cells show differential metabolic pathway gene activation in vivo. (A) Scatter plot shows differential expressed genes between CNS TH17 (red) and ileum TH17 (blue) cells with differential TH17 signature genes highlighted (black dot). differentially-expressed genes are filtered as false discovery rate (FDR) , 0.05 and FC $ 1.5 from DESeq2. (B) Scatter plot shows differential expressed genes between CNS TH17 (red) and ileum TH17 (blue) cells with glycolysis pathway genes highlighted (black dot) and differential genes labeled. DEGs are filtered as in (A). (C) Scatter plot shows differential expressed genes between CNS TH17 (red) and ileum TH17 (blue) cells with glutaminolysis pathway genes highlighted (black dot) and differential genes labeled. DEGs are filtered as in (A). (D)CNSTH17 and ileum TH17 cells depend on distinct metabolic pathways. Numbers indicate the differentially expressed genes and total genes for each selected pathway. (E) KEGG pathway enrichment for CNS TH17 cell highly expressed genes. (F) KEGG pathway enrichment for ileum TH17 cell highly expressed genes. Color represents log-transferred FDR, and dot size represents the number of differentially expressed genes observed for this pathway (E and F). 6 miR-21 CONTROLS GLYCOLYSIS OF PATHOGENIC TH17 CELLS

FIGURE 3. TH17 cells derived in vitro show discrete chromatin states. (A) Scatter plot for differentially opened/closed regions (OCRs) between TH17 (23) (red) and TH17 (b) cells (blue) with differential numbers labeled. (B) ATAC-seq reads density heatmap around the OCRs (shows the peak summit 6 1kbregion)between TH17 (23) and TH17 (b)cells.(C) The enriched transcription factor motifs identiﬁed from top 1000 differentially OCRs between TH17 (23) cells and TH17 (b) cells; color bar represents log-transferred p values. Right panel shows the identiﬁed motif sequences for key regulators. (D) WashU Epigenome Browser visualization of differentially OCR signals for key glycolytic genes in TH17 (23) and TH17 (b) cells. Arrows represent the NF-kB binding motif sites.

which suggests that CNS-infiltrated TH17 cells were relieved from great more chromatin accessibility in CNS pathogenic TH17 cells, Foxo1-mediated metabolic repression (Supplemental Fig. 2G). A which include miR-21, miR-23a cluster, and miR-146a (Fig. 5A), group of genes key to metabolic pathways and solute carrier family whereas miR-21 and miR-146a were the two with the highest members were also with repressed chromatin accessibility in steady- abundance in CNS TH17 cells (Fig. 5B). state ileum TH17 cells, which was consistent with the gene expression Our group previously demonstrated that miR-21 was highly data (Fig. 4D). Taken together, our data suggest that metabolic upregulated in inflammatory lesions of multiple human auto- reprogramming of TH17 cells in vivo are controlled at the epigenetic immune diseases (45). And, by breeding conditional knockout level, and homeostatic TH17 cells show repressed chromatin acces- mice, we showed that T cell–intrinsic miR-21 was essential for sibility at glycolytic genes. autoimmune encephalomyelitis (Supplemental Fig. 3A–G). To better understand the general functional features of miR-21– Identification of a miR-21–Peli1–c-Rel pathway controlling deficient TH17 cells, we differentiated naive T cells from pathogenic TH17 cell glycolysis 2/2 control and miR-21 mice in vitro under pathogenic TH17 Specific microRNAs are appreciated as key metabolic regulators (23) condition; RNA was extracted and subjected to RNA-seq. for immune cell function; to identify microRNAs that are essential In the absence of miR-21,TH17 cells showed decreased expression for glycolysis of pathogenic TH17 cells, we checked chromatin of TH17 cell signature genes, which include lineage-specific accessibility of the regulatory region of microRNAs with NF-kB transcription factors, cytokines, and cytokine receptors (Fig. 5C, binding sites and their expression in TH17 cells derived in vivo. 5D). Remarkably, many of the genes involved in glycolytic or We found the regulatory region of a group of microRNAs show related metabolic pathways were highly downregulated, including The Journal of Immunology 7

FIGURE 4. TH17 cells show discrete chromatin states in vivo. (A) Scatter plot for differentially opened/closed regions (OCRs) between CNS TH17 (red) and ileum TH17 cells (blue) with differential numbers labeled. (B) ATAC-seq reads density heatmap around the OCRs (shows the peak summit 6 1kb region) between CNS TH17 and ileum TH17 cells. (C) The enriched transcription factor motifs identified from top 1000 differential OCRs between CNS TH17 cells and ileum TH17 cells; color bar represents log-transferred p values. Right panel shows the identified motif sequences for key regulators. (D) WashU Epigenome Browser visualization of differentially OCR signals with corresponding gene expression levels for key TH17 signature genes and metabolic genes in CNS TH17 and ileum TH17 cells. Arrows represent the NF-kB binding motif sites. key transporters for glucose intake, Slc2a1/3, and rate-limiting screened direct targets of miR-21 by combining target prediction enzymes, Hk1/2, Pgm2, and Ldha (Fig. 5D). Interestingly, genes with Ago2 high-throughput RNA-seq isolated by crosslinking of the glycolytic pathway that are highly expressed in CNS- immunoprecipitation (46), TargetScan, and RNA-seq data to infiltrated pathogenic TH17 cells are generally downregulated in identify overlapping gene transcripts. Using a combination of 2/2 miR-21 TH17 cells (Fig. 5E). We further checked the glyco- these three approaches, we identified 13 common putative targets 2/2 lytic activity of control and miR-21 TH17 cells by glycolysis (Fig. 5G). Among them, Peli1 has been reported to function as a stress test; both basal and maximal glycolytic capability were negative regulator of CD4+ T cell activation by ubiquitination of 2/2 decreased in miR-21 TH17 cells (Fig. 5F). Taken together, c-Rel (47). c-Rel was reported to induce c-Myc expression in these data suggest that miR-21 controls glucose metabolism of activated T cells (48), a metabolic regulator of germinal center pathogenic TH17 cells. B cell function (14). More importantly, c-Rel was reported as a To further investigate the precise mechanism underlying how critical regulator of TH17 cell function (49, 50); therefore, we chose miR-21 promotes pathogenic TH17 cell glucose metabolism, we Peli1 for further analysis. By in vitro RNA immunoprecipitation 8 miR-21 CONTROLS GLYCOLYSIS OF PATHOGENIC TH17 CELLS

FIGURE 5. Identification of a miR-21–Peli1–c-Rel pathway controlling pathogenic TH17 cell glycolysis. (A) WashU Epigenome Browser visualization of ATAC-seq signals across selected microRNA genome loci in ileum homeostatic and CNS-infiltrated pathogenic TH17 cells. (B) Relative expression of 2/2 microRNA in ileum homeostatic and CNS-infiltrated pathogenic TH17 cells (n = 3). (C) KEGG pathway enrichment for genes downregulated in miR-21 TH17 cells. Color represents the enriched p values, and dot size represents the number of differentially expressed genes observed for this pathway. (D) 2/2 Normalized gene expression level in control and miR-21 TH17 (23) cells (n = 2) of selected genes of interest in TH17 cell biology (left) and in glycolytic 2/2 pathway (right). (E) Scatter plot of gene expression FCs for glycolytic pathway genes from miR-21 /control TH17 cells and CNS/ileum TH17 cells. (F) 2/2 ECAR of TH17 (23) cells differentiated from control and miR-21 mice for 96 h assessed by a glycolysis stress test. (G) Venn diagram of upregulated 2/2 genes in miR-21 TH17 (23) cells from RNA-seq, genes targeted by miR-21 predicted by TargetScan, and miR-21 seed sequence–enriched genes from crosslinking immunoprecipitation–seq data. The right panel shows the 13 overlapped genes as miR-21 potentially targeted genes. (H) Normalized RT-PCR 2/2 result for enriched Peli1 mRNA level in Ago2 immunoprecipitated total RNA from in vitro–differentiated control and miR-21 TH17 (23) cells. (I) Western blot results for c-Rel and Peli1 from TH17 cells differentiated under indicated conditions. Data shown are one experiment representative of three independent experiments (I). Data are representative of three independent experiments (B, F, and H). Error bars represent SEM. *p , 0.05, **p , 0.01, determined by two-tailed unpaired t test. assay, we found Ago2-immunoprecipitated RNAs had significantly together, these data indicate that a miR-21–Peli1–c-Rel pathway 2/2 reduced Peli1 mRNA level in miR-21 TH17 (23) cells (Fig. 5H), promotes glycolysis of pathogenic TH17 cells. which suggests that Peli1 was a functional target of miR-21 in Pharmaceutical inhibition c-Rel–mediated glycolysis in pathogenic TH17 cells. We further validated these at the protein + 2/2 pathogenic TH17 cells prevents autoimmunity level in purified GFP TH17 cells; we found that miR-21 TH17 cells have elevated level of Peli1 protein, whereas the protein level Our above data suggested that genes with NF-kB family tran- 2/2 of c-Rel was reduced in miR-21 TH17 cells (Fig. 5I). Taken scription factors motif were repressed in homeostatic ileum TH17 The Journal of Immunology 9

cells and in vitro–derived nonpathogenic TH17 cells at the chro- nonpathogenic TH17 (b) cells (Fig. 6B). To study whether c-Rel matin level. To further explore the role of TGF-b1 signaling on was essential for glycolysis pathway gene activation of pathogenic NF-kB family transcription factors and its effect on TH17 cell TH17 cells, we took advantage of a c-Rel speciﬁc inhibitor PTXF glycolysis, we differentiated TH17 cells in vitro under nonpatho- (51), which was approved by U.S. Food and Drug Administration genic TH17 (b) cell condition with variable concentration of for clinical use for a variety of diseases. Inhibition of c-Rel re- mouse rTGF-b1; we found that TGF-b1 signaling inhibited c-Rel duced the expression key glycolytic pathway gene expression in expression in a dose-dependent manner, whereas RelA expression activated pathogenic TH17 (23) cells (Fig. 6C, 6D), including was unchanged at the protein level (Fig. 6A). We further checked Slc2a3, Hk2, Tpi1, Eno1, and Ldha. Next, we asked whether c-Rel and RelA protein level at in vitro–derived nonpathogenic overexpression of c-Rel in nonpathogenic TH17 (b) cells could TH17 (b) and pathogenic TH17 (23) cells; we found pathogenic promote pathogenic TH17 cell signature genes and glycolytic TH17 (23) cells expressed speciﬁcally more c-Rel protein than pathway gene expression; indeed, enforced expression of c-Rel in

FIGURE 6. Pharmaceutical inhibition c-Rel–mediated glycolysis in pathogenic TH17 cells prevents autoimmunity. (A) Western blot results for c-Rel and RelA from TH17 cells differentiated under indicated conditions. (B) Western blot results for c-Rel and RelA from TH17 cells differentiated under indicated conditions. (C) Western blot results for c-Rel and RelA from TH17 cells differentiated under indicated conditions. (D) Normalized RT-PCR results for key glycolytic pathway genes in TH17 (23) cells differentiated in vitro for 48 h, treated with H2O or 500 mg/ml PTXF for 16 h. (E) ECAR of TH17 (23) cells differentiated for 48 h and then were treated with H2O or 500 mg/ml PTXF for 16 h, assessed by a glycolysis stress test. (F) Basal and maximal glycolytic capability of TH17 (23) cells differentiated for 48 h and then were treated with H2O or PTXF for 16 h. (G) EAE development in C57BL/6 mice, i.p. PBS or 100 mg/kg PTXF per mouse (n = 6) from day 7 to day 14. (H) EAE development in recipient C57BL/6 mice (sublethal irradiation; n = 6) i.p. with + pathogenic TH17 (23) cells differentiated from naive 2D2 CD4 T cells for 96 h and then treated with H2O or 500 mg/ml PTXF for 16 h. Data shown are one experiment representative of three independent experiments (A–C, G, and H). Data are representative of three independent experiments (D–F). Error bars represent SEM. *p , 0.05, **p , 0.01, ***p , 0.001, determined by two-tailed unpaired t test. 10 miR-21 CONTROLS GLYCOLYSIS OF PATHOGENIC TH17 CELLS

TH17 (b) cells promoted key pathogenic TH17 cell signature gene and key metabolism pathway genes. These data suggest that with expression (Supplemental Fig. 3H), including IL-17a, IL-17f, Csf2, increased chromatin openness of its regulatory region, miR-21 IL-2,andIL-23r. And, we found enforced expression of c-Rel pro- controls glycolysis of pathogenic TH17 cells under autoimmune moted selected glycolysis pathway gene expression with a level of inflammation. 20–50% upregulation (Hk1, Hk2, Slc2a1,andSlc2a3) (Supplemental TH17 cells mediate pathogenic role in multiple autoimmune Fig. 3I), although less pronounced to the level of upregulation of diseases. Although anti–IL-17A treatment received great efficacy pathogenic TH17 signature genes. We further checked key glycolytic in psoriasis, it badly failed for inflammatory bowel disease pa- enzymes at the protein level, and we found overexpression of c-Rel tients. Understanding metabolic reprogramming of TH17 cells elevated protein level of Slc2a1 (Glut1) and Hk2 (Supplemental in vivo may therefore provide more defined therapeutic interven- Fig. 3J) in nonpathogenic TH17 cells. Together, these data suggest tion to TH17 cell–mediated autoimmune diseases and insights into that TGF-b1 signaling represses c-Rel–mediated glycolysis of TH17 cell–mediated host defense. The microenvironment of ho- pathogenic TH17 cells. Next, we checked glycolytic activity of meostatic mucosal surfaces and various inflammatory sites could control and c-Rel inhibitor PTXF-treated pathogenic TH17 cells by have great differential effects on chromatin states of TH17 cells glycolysis stress test; both basal and maximal glycolytic capability in vivo, which, in turn, could facilitate their discrete metabolic were decreased in pathogenic TH17 cells by c-Rel inhibitor PTXF reprogramming. Our study so far dissected the metabolic states treatment (Fig. 6E, 6F). To study c-Rel–mediated glycolysis in and chromatin states of TH17 cells in vivo and identified a miR- pathogenic TH17 cells under disease condition, we immunized mice 21–Peli1–c-Rel axis as a critical metabolic regulator for patho- with CFA plus MOG35–55 to induce autoimmune encephalomyelitis; genic TH17 cells. systemic administration of c-Rel inhibitor PTXF prevents mice from disease initiation (Fig. 6G). We further differentiated patho- Acknowledgments genic TH17 cells from 2D2 mice, and treatment of c-Rel inhibitor We thank Youcun Qian for helpful discussions, technical expertise, and PTXF blocked myelin-specific TH17 cell–mediated autoimmune review of the manuscript. encephalomyelitis (Fig. 6H). Taken together, our data suggest that targeting c-Rel–mediated glycolysis in pathogenic TH17 cells pro- tects mice from autoimmunity. Disclosures The authors have no financial conflicts of interest. Discussion TH17 cells exhibit great heterogeneity in vivo, the basic energy References requirement of TH17 cells under homeostatic or autoimmune 1. Littman, D. R., and A. Y. Rudensky. 2010. Th17 and regulatory T cells in me- condition is beginning to be understood. In the current study, by diating and restraining inflammation. Cell 140: 845–858. 2. Stockinger, B., and S. Omenetti. 2017. The dichotomous nature of T helper 17 metabolomics analyses of in vitro–derived TH17 cells, we found cells. Nat. Rev. Immunol. 17: 535–544. that nonpathogenic and pathogenic TH17 cells exhibit distinct 3. Ivanov, I. I., K. Atarashi, N. Manel, E. L. Brodie, T. Shima, U. Karaoz, D. Wei, metabolic activity. And, by global transcriptome analyses of K. C. Goldfarb, C. A. Santee, S. V. Lynch, et al. 2009. Induction of intestinal Th17 cells by segmented filamentous bacteria. Cell 139: 485–498. in vivo–derived TH17 cells, we found CNS-infiltrated pathogenic 4. Sano, T., W. Huang, J. A. Hall, Y. Yang, A. Chen, S. J. Gavzy, J. Y. Lee, TH17 cells show discrete metabolic pathway gene activation J. W. Ziel, E. R. Miraldi, A. I. Domingos, et al. 2015. An IL-23R/IL-22 circuit compared with homeostatic T 17 cells. Interestingly, a recently regulates epithelial serum amyloid A to promote local effector Th17 responses. H [Published erratum appears in 2016 Cell 164: 324.] Cell 163: 381–393. published report showed that colon C. rodentium–induced patho- 5. Atarashi, K., T. Tanoue, M. Ando, N. Kamada, Y. Nagano, S. Narushima, W. Suda, A. Imaoka, H. Setoyama, T. Nagamori, et al. 2015. Th17 cell induction genic TH17 cells exhibit a global activation of multiple metabolic by adhesion of microbes to intestinal epithelial cells. Cell 163: 367–380. pathways compared with segmented filamentous bacteria–induced 6. Ghoreschi, K., A. Laurence, X. P. Yang, C. M. Tato, M. J. McGeachy, homeostatic TH17 cells (20). J. E. Konkel, H. L. Ramos, L. Wei, T. S. Davidson, N. Bouladoux, et al. 2010. To dissect the regulation of T 17 cell metabolic states in vivo, Generation of pathogenic T(H)17 cells in the absence of TGF-b signalling. H Nature 467: 967–971. we performed genome-wide chromatin landscape profiling. We 7. Gaublomme, J. T., N. Yosef, Y. Lee, R. S. Gertner, L. V. Yang, C. Wu, found that TH17 cells had discrete chromatin states in vivo. The P. P. Pandolfi, T. Mak, R. Satija, A. K. Shalek, et al. 2015. Single-cell genomics regulatory region of key T 17 cell–related transcription factors, unveils critical regulators of Th17 cell pathogenicity. Cell 163: 1400–1412. H 8. Man, K., and A. Kallies. 2015. Synchronizing transcriptional control of T cell metabolic regulators, metabolism-related genes, and microRNAs metabolism and function. Nat. Rev. Immunol. 15: 574–584. were with differential chromatin accessibility between homeo- 9. Pollizzi, K. N., and J. D. Powell. 2014. Integrating canonical and metabolic signalling programmes in the regulation of T cell responses. Nat. Rev. Immunol. static TH17 cells and pathogenic TH17 cells. Our data suggest that 14: 435–446. TH17 cells in vivo adapt to distinct microenvironment with dis- 10. Buck, M. D., D. O’Sullivan, and E. L. Pearce. 2015. T cell metabolism drives crete epigenetic and metabolic states. Metabolic regulators and immunity. J. Exp. Med. 212: 1345–1360. 11. Boothby, M., and R. C. Rickert. 2017. Metabolic regulation of the immune metabolism-related genes were regulated at the epigenetic level humoral response. Immunity 46: 743–755. for their functional requirements. TGF-b1 signaling was further 12. Wang, R., C. P. Dillon, L. Z. Shi, S. Milasta, R. Carter, D. Finkelstein, shown to be crucial for remodeling of T 17 cell chromatin states L. L. McCormick, P. Fitzgerald, H. Chi, J. Munger, and D. R. Green. 2011. The H transcription factor Myc controls metabolic reprogramming upon T lymphocyte in vitro; c-Rel was identified as an essential metabolic regulator activation. Immunity 35: 871–882. 13. Verbist, K. C., C. S. Guy, S. Milasta, S. Liedmann, M. M. Kaminski, R. Wang, for pathogenic TH17 cells. and D. R. Green. 2016. Metabolic maintenance of cell asymmetry following Interestingly, we found the conserved regulatory region of a division in activated T lymphocytes. Nature 532: 389–393. group of microRNAs were with differential chromatin accessibility 14. Heise, N., N. S. De Silva, K. Silva, A. Carette, G. Simonetti, M. Pasparakis, and of T 17 cells derived in vivo, which could have contributed to the U. Klein. 2014. Germinal center B cell maintenance and differentiation are H controlled by distinct NF-kB transcription factor subunits. J. Exp. Med. 211: regulation of discrete metabolic states of TH17 cells in vivo for 2103–2118. their fine-tune property. The regulatory region of miR-21 was with 15. Karmaus, P. W. F., X. Chen, S. A. Lim, A. A. Herrada, T. M. Nguyen, B. Xu, Y. Dhungana, S. Rankin, W. Chen, C. Rosencrance, et al. 2019. Metabolic increased chromatin accessibility in CNS-infiltrated TH17 cells. heterogeneity underlies reciprocal fates of TH17 cell stemness and plasticity. Although miR-21 was shown previously to be important in TH17 Nature 565: 101–105. cell generation (52), its exact action has not been well under- 16. Wilhelm, K., K. Happel, G. Eelen, S. Schoors, M. F. Oellerich, R. Lim, B. Zimmermann, I. M. Aspalter, C. A. Franco, T. Boettger, et al. 2016. FOXO1 stood. By global transcriptomics analyses, we demonstrate that couples metabolic activity and growth state in the vascular endothelium. Nature 2/2 miR-21 TH17 cells express fewer TH17 cell signature genes 529: 216–220. The Journal of Immunology 11

17. Angelin, A., L. Gil-de-Go´mez, S. Dahiya, J. Jiao, L. Guo, M. H. Levine, 35. Langmead, B., and S. L. Salzberg. 2012. Fast gapped-read alignment with Z. Wang, W. J. Quinn, III, P. K. Kopinski, L. Wang, et al. 2017. Foxp3 repro- Bowtie 2. Nat. Methods 9: 357–359. grams T cell metabolism to function in low-glucose, high-lactate environments. 36. Li, H., B. Handsaker, A. Wysoker, T. Fennell, J. Ruan, N. Homer, G. Marth, Cell Metab. 25: 1282–1293.e7. G. Abecasis, and R. Durbin; 1000 Genome Project Data Processing Subgroup. 18. Chang, C. H., J. Qiu, D. O’Sullivan, M. D. Buck, T. Noguchi, J. D. Curtis, 2009. The sequence alignment/map format and SAMtools. Bioinformatics 25: Q. Chen, M. Gindin, M. M. Gubin, G. J. van der Windt, et al. 2015. Metabolic 2078–2079. competition in the tumor microenvironment is a driver of cancer progression. 37. Allhoff, M., K. Sere´, J. F Pires, M. Zenke, and I. G Costa. 2016. Differential Cell 162: 1229–1241. peak calling of ChIP-seq signals with replicates with THOR. Nucleic Acids Res. 19. Ho, P. C., J. D. Bihuniak, A. N. Macintyre, M. Staron, X. Liu, R. Amezquita, 44: e153. Y. C. Tsui, G. Cui, G. Micevic, J. C. Perales, et al. 2015. Phosphoenolpyruvate is 38. Heinz, S., C. Benner, N. Spann, E. Bertolino, Y. C. Lin, P. Laslo, J. X. Cheng, a metabolic checkpoint of anti-tumor T cell responses. Cell 162: 1217–1228. C. Murre, H. Singh, and C. K. Glass. 2010. Simple combinations of lineage- 20. Omenetti, S., C. Bussi, A. Metidji, A. Iseppon, S. Lee, M. Tolaini, Y. Li, determining transcription factors prime cis-regulatory elements required for G. Kelly, P. Chakravarty, S. Shoaie, et al. 2019. The intestine harbors func- macrophage and B cell identities. Mol. Cell 38: 576–589. tionally distinct homeostatic tissue-resident and inflammatory Th17 cells. 39. Kim, D., G. Pertea, C. Trapnell, H. Pimentel, R. Kelley, and S. L. Salzberg. 2013. Immunity 51: 77–89.e6. TopHat2: accurate alignment of transcriptomes in the presence of insertions, 21. Etchegaray, J. P., and R. Mostoslavsky. 2016. Interplay between metabolism and deletions and gene fusions. Genome Biol. 14: R36. epigenetics: a nuclear adaptation to environmental changes. Mol. Cell 62: 695–711. 40. Anders, S., P. T. Pyl, and W. Huber. 2015. HTSeq--a Python framework to work 22. Rajbhandari, P., B. J. Thomas, A. C. Feng, C. Hong, J. Wang, L. Vergnes, with high-throughput sequencing data. Bioinformatics 31: 166–169. T. Sallam, B. Wang, J. Sandhu, M. M. Seldin, et al. 2018. IL-10 signaling re- 41. Love, M. I., W. Huber, and S. Anders. 2014. Moderated estimation of fold models adipose chromatin architecture to limit thermogenesis and energy ex- change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15: 550. penditure. Cell 172: 218–233.e17. 42. Feng, X., H. Wang, H. Takata, T. J. Day, J. Willen, and H. Hu. 2011. Tran- 23. Buenrostro, J. D., P. G. Giresi, L. C. Zaba, H. Y. Chang, and W. J. Greenleaf. scription factor Foxp1 exerts essential cell-intrinsic regulation of the quiescence 2013. Transposition of native chromatin for fast and sensitive epigenomic pro- of naive T cells. Nat. Immunol. 12: 544–550. filing of open chromatin, DNA-binding proteins and nucleosome position. Nat. 43. Stephen, T. L., M. R. Rutkowski, M. J. Allegrezza, A. Perales-Puchalt, Methods 10: 1213–1218. A. J. Tesone, N. Svoronos, J. M. Nguyen, F. Sarmin, M. E. Borowsky, J. Tchou, 24. van der Veeken, J., A. J. Gonzalez, H. Cho, A. Arvey, S. Hemmers, C. S. Leslie, and J. R. Conejo-Garcia. 2014. Transforming growth factor b-mediated sup- and A. Y. Rudensky. 2016. Memory of inflammation in regulatory T cells. Cell pression of antitumor T cells requires FoxP1 transcription factor expression. 166: 977–990. Immunity 41: 427–439. 25. Ceribelli, M., Z. E. Hou, P. N. Kelly, D. W. Huang, G. Wright, K. Ganapathi, 44. Arsura, M., M. Wu, and G. E. Sonenshein. 1996. TGF beta 1 inhibits NF-kappa M. O. Evbuomwan, S. Pittaluga, A. L. Shaffer, G. Marcucci, et al. 2016. A B/Rel activity inducing apoptosis of B cells: transcriptional activation of I kappa druggable TCF4- and BRD4-dependent transcriptional network sustains malig- B alpha. Immunity 5: 31–40. nancy in blastic plasmacytoid dendritic cell neoplasm. Cancer Cell 30: 764–778. 45. Zhu, S., W. Pan, X. Song, Y. Liu, X. Shao, Y. Tang, D. Liang, D. He, H. Wang, 26. Philip, M., L. Fairchild, L. Sun, E. L. Horste, S. Camara, M. Shakiba, W. Liu, et al. 2012. The microRNA miR-23b suppresses IL-17-associated au- A. C. Scott, A. Viale, P. Lauer, T. Merghoub, et al. 2017. Chromatin states define toimmune inflammation by targeting TAB2, TAB3 and IKK-a. Nat. Med. 18: tumour-specific T cell dysfunction and reprogramming. Nature 545: 452–456. 1077–1086. 27. Mehta, A., and D. Baltimore. 2016. MicroRNAs as regulatory elements in im- 46. Loeb, G. B., A. A. Khan, D. Canner, J. B. Hiatt, J. Shendure, R. B. Darnell, mune system logic. Nat. Rev. Immunol. 16: 279–294. C. S. Leslie, and A. Y. Rudensky. 2012. Transcriptome-wide miR-155 binding map 28. Najafi-Shoushtari, S. H., F. Kristo, Y. Li, T. Shioda, D. E. Cohen, R. E. Gerszten, reveals widespread noncanonical microRNA targeting. Mol. Cell 48: 760–770. and A. M. Naä¨r. 2010. MicroRNA-33 and the SREBP host genes cooperate to 47.Chang,M.,W.Jin,J.H.Chang,Y.Xiao,G.C.Brittain,J.Yu,X.Zhou,Y.H.Wang, control cholesterol homeostasis. Science 328: 1566–1569. X. Cheng, P. Li, et al. 2011. The ubiquitin ligase Peli1 negatively regulates T cell 29. Rayner, K. J., Y. Sua´rez, A. Da´valos, S. Parathath, M. L. Fitzgerald, N. Tamehiro, activation and prevents autoimmunity. Nat. Immunol. 12: 1002–1009. E. A. Fisher, K. J. Moore, and C. Fernańdez-Hernando. 2010. MiR-33 contributes 48. Grumont, R., P. Lock, M. Mollinari, F. M. Shannon, A. Moore, and to the regulation of cholesterol homeostasis. Science 328: 1570–1573. S. Gerondakis. 2004. The mitogen-induced increase in T cell size involves PKC 30. Zhu, H., N. Shyh-Chang, A. V. Segre`, G. Shinoda, S. P. Shah, W. S. Einhorn, and NFAT activation of Rel/NF-kappaB-dependent c-myc expression. Immunity A. Takeuchi, J. M. Engreitz, J. P. Hagan, M. G. Kharas, et al; MAGIC Investigators. 21: 19–30. 2011. The Lin28/let-7 axis regulates glucose metabolism. Cell 147: 81–94. 49. Ruan, Q., V. Kameswaran, Y. Zhang, S. Zheng, J. Sun, J. Wang, J. DeVirgiliis, 31. Chen, W. L., Y. Y. Wang, A. Zhao, L. Xia, G. Xie, M. Su, L. Zhao, J. Liu, C. Qu, H. C. Liou, A. A. Beg, and Y. H. Chen. 2011. The Th17 immune response is R. Wei, et al. 2016. Enhanced fructose utilization mediated by SLC2A5 is a controlled by the Rel-RORg-RORg T transcriptional axis. J. Exp. Med. 208: unique metabolic feature of acute myeloid leukemia with therapeutic potential. 2321–2333. Cancer Cell 30: 779–791. 50. Chen, G., K. Hardy, E. Pagler, L. Ma, S. Lee, S. Gerondakis, S. Daley, and 32. Chen, W. L., J. H. Wang, A. H. Zhao, X. Xu, Y. H. Wang, T. L. Chen, J. M. Li, M. F. Shannon. 2011. The NF-kB transcription factor c-Rel is required for Th17 J. Q. Mi, Y. M. Zhu, Y. F. Liu, et al. 2014. A distinct glucose metabolism sig- effector cell development in experimental autoimmune encephalomyelitis. nature of acute myeloid leukemia with prognostic value. [Published erratum J Immunol. 187: 4483–4491. appears in 2014 Blood 124: 2893.] Blood 124: 1645–1654. 51. Grinberg-Bleyer, Y., H. Oh, A. Desrichard, D. M. Bhatt, R. Caron, T. A. Chan, 33. Qiu, Y., G. Cai, M. Su, T. Chen, X. Zheng, Y. Xu, Y. Ni, A. Zhao, L. X. Xu, R. M. Schmid, U. Klein, M. S. Hayden, and S. Ghosh. 2017. NF-kB c-Rel is crucial S. Cai, and W. Jia. 2009. Serum metabolite profiling of human colorectal cancer for the regulatory T cell immune checkpoint in cancer. Cell 170: 1096–1108.e13. using GC-TOFMS and UPLC-QTOFMS. J. Proteome Res. 8: 4844–4850. 52. Murugaiyan, G., A. P. da Cunha, A. K. Ajay, N. Joller, L. P. Garo, 34. Ni, Y., M. Su, Y. Qiu, W. Jia, and X. Du. 2016. ADAP-GC 3.0: improved peak S. Kumaradevan, N. Yosef, V. S. Vaidya, and H. L. Weiner. 2015. MicroRNA-21 detection and deconvolution of co-eluting metabolites from GC/TOF-MS data promotes Th17 differentiation and mediates experimental autoimmune enceph- for metabolomics studies. Anal. Chem. 88: 8802–8811. alomyelitis. J. Clin. Invest. 125: 1069–1080.