TNF Activates a NF-κB−Regulated Cellular Program in Human CD45RA− Regulatory T Cells that Modulates Their Suppressive Function This information is current as of September 25, 2021. Meital Nagar, Jasmine Jacob-Hirsch, Helly Vernitsky, Yackov Berkun, Shomron Ben-Horin, Ninette Amariglio, Ilan Bank, Yoel Kloog, Gideon Rechavi and Itamar Goldstein

J Immunol 2010; 184:3570-3581; Prepublished online 24 Downloaded from February 2010; doi: 10.4049/jimmunol.0902070 http://www.jimmunol.org/content/184/7/3570 http://www.jimmunol.org/

Supplementary http://www.jimmunol.org/content/suppl/2010/02/22/jimmunol.090207 Material 0.DC1 References This article cites 46 articles, 20 of which you can access for free at: http://www.jimmunol.org/content/184/7/3570.full#ref-list-1 by guest on September 25, 2021 Why The JI? Submit online.

• Rapid Reviews! 30 days* from submission to initial decision

• No Triage! Every submission reviewed by practicing scientists

• Fast Publication! 4 weeks from acceptance to publication

*average

Subscription Information about subscribing to The Journal of Immunology is online at: http://jimmunol.org/subscription Permissions Submit copyright permission requests at: http://www.aai.org/About/Publications/JI/copyright.html Email Alerts Receive free email-alerts when new articles cite this article. Sign up at: http://jimmunol.org/alerts

The Journal of Immunology is published twice each month by The American Association of Immunologists, Inc., 1451 Rockville Pike, Suite 650, Rockville, MD 20852 Copyright © 2010 by The American Association of Immunologists, Inc. All rights reserved. Print ISSN: 0022-1767 Online ISSN: 1550-6606. The Journal of Immunology

TNF Activates a NF-kB–Regulated Cellular Program in Human CD45RA– Regulatory T Cells that Modulates Their Suppressive Function

Meital Nagar,*,† Jasmine Jacob-Hirsch,* Helly Vernitsky,*,† Yackov Berkun,‡ Shomron Ben-Horin,x Ninette Amariglio,* Ilan Bank,x Yoel Kloog,† Gideon Rechavi,* and Itamar Goldstein*,x

Emerging data suggest that (Treg) dysfunction and consequent breakdown of immunological self-tolerance in autoimmunity can be mediated by factors that are not Treg-intrinsic (e.g., cytokines). Indeed, recent studies show that in rheumatoid arthritis the proinflammatory cytokine TNF reduces the suppressive function of Tregs, whereas in vivo TNF blockade restores this function and accordingly self-tolerance. However, until now a coherent mechanism by which TNF regulates the Treg has not been Downloaded from described. In this paper, we show that TNF induces preferential and significant activation of the canonical NF-kB pathway in human Tregs as compared with CD25– conventional T cells. Furthermore, TNF induced primarily in CD45RA– Tregs a transcription program highly enriched for typical NF-kB target , such as the cytokines lymphotoxin-a and TNF, the TNFR superfamily members FAS, 4-1BB, and OX-40, various antiapoptotic genes, and other important immune-response genes. FACS analysis revealed that TNF also induced upregulation of cell surface expression of 4-1BB and OX40 specifically in CD45RA–FOXP3+ Tregs. In contrast, TNF had only a minimal effect on the Treg’s core transcriptional signature or on the intracellular levels of the http://www.jimmunol.org/ FOXP3 in Tregs. Importantly, TNF treatment modulated the capacity of Tregs to suppress the proliferation and IFN-g secretion by conventional T cells, an effect that was fully reversed by cotreatment with anti-TNFR2 mAbs. Our findings thus provide new mechanistic insight into the role of TNF and TNFR2 in the pathogenesis of autoimmunity. The Journal of Immu- nology, 2010, 184: 3570–3581.

umor necrosis factor (TNF-a) is a prototype member of iopathic arthritis (JIA), ankylosing spondylitis, and psoriatic ar- the TNF superfamily (TNFSF) of ligands. It binds the thritis (3). Nevertheless, little is known at the molecular level

T corresponding TNFR superfamily (TNFRSF) members about the specific effects of TNF on the various T cell subsets, and by guest on September 25, 2021 TNFR1 (TNFRSF1A; p55) and TNFR2 (TNFRSF1B; p75). TNF is consequently the fine details of the mechanism of action of TNF generated and expressed by many types of cells, including lym- blockade, in vivo, remain somewhat obscure. phocytes, and can induce a plethora of immune responses (1). VariousstudiesinbothmiceandhumanssuggestthatTNFblockade Several lines of evidence suggest that TNF has a central role in the has the potential to augment regulatory T cell (Treg) numbers and pathogenesis of a variety of human inflammatory disorders (2). function. For example, in RA patients treated with anti-TNF mAbs Moreover, TNF blockade has a notable therapeutic efficacy in there is an increase in the quantity of circulating CD25HI Tregs as well a number of T cell-dependent human autoimmune disorders, such as restoration of their partly impaired suppressive function (4, 5). as inflammatory bowel disease, rheumatoid arthritis, juvenile id- Moreover, TNF blockade during the in vitro activation and expansion of unselected human T cells results in augmented expansion of x + + *Sheba Research Center, ‡Division of Pediatrics, and Department of Med- FOXP3 CD4 T cells with a regulatory phenotype (6, 7). icine, The Chaim Sheba Medical Center, Tel Aviv University, Sackler Faculty of It should be noted that studies in patients with various auto- Medicine, Tel Hashomer; and †The George S. Wise Faculty of Life Sciences, Tel Aviv University, Tel Aviv, Israel immune disorders and mice models show that Treg-type cells are Received for publication June 29, 2009. Accepted for publication January 26, 2010. present at considerable numbers at the actual site of inflammation This work was supported in part by an award from the Flight Attendant Medical (8, 9). Thus, a major challenge is to define the Treg extrinsic Research Institute. factors that facilitate tissue inflammation even when Tregs are The microarray data presented in this article have been submitted to the National present in abundance. In this regard, there is controversy as to Center for Biotechnology Information Expression Omnibus under accession whether TNF—a cytokine abundant at sites of inflammation—can number GSE18893. interfere with the suppressive function of the Treg (4, 8–13). Address correspondence and reprint requests to Dr. Itamar Goldstein, The Sheba – Cancer Research Center, Chaim Sheba Medical Center, Tel Hashomer 52621, Israel. Tregs as opposed to CD25 conventional T cells (Tcons) consti- E-mail address: [email protected] tutively express high levels of TNFR2, but both subsets do not ex- The online version of this article contains supplemental material. press TNFR1 (4, 12). Whereas the downstream molecular events Abbreviations used in this paper: acTreg, activated regulatory T cell; DI, division initiated following TNFR1 ligation have been studied in great detail, index; EASE, Expression Analysis Systematic Explorer; IPA, Ingenuity Pathways there is significantly less information about signal transduction via Analysis, JIA, juvenile idiopathic arthritis; MB, microbead; MFI, mean fluorescence TNFR2 that, as opposed to TNFR1, does not contain a death domain intensity; mTNF, membrane-bound TNF; nrTreg, naı¨ve/resting regulatory T cell; PMA/Ion, PMA and ionomycin; rhIL-2, recombinant human IL-2; RTQPCR, real- (14, 15). Binding of TNF to TNFR1 activates at least three distinct time quantitative PCR; SF, synovial fluid; Tcon, conventional T cell; TNFRSF, TNF cellular pathways: IKK/NF-kB, JNK/c-Jun, and the proapoptotic superfamily; TNFSF, TNF superfamily; Treg, regulatory T cell. caspases 8 and 3 (1). At present, the specific cellular program in- Copyright Ó 2010 by The American Association of Immunologists, Inc. 0022-1767/10/$16.00 duced by TNFR2 stimulation in Tregs is practically unknown. www.jimmunol.org/cgi/doi/10.4049/jimmunol.0902070 The Journal of Immunology 3571

In this study, using various complementary methods, such as gene Briefly, CFSE-labeled CD25– T cells at 1 3 105 cells per well were plated into 96-well plates (Costar) and activated by plate-bound immobilized expression microarrays, PCR-based assays, analysis of protein ex- ++ pression and/or phosphorylation by FACS, and other relevant assays OKT3 (2.5 mg/ml) in the presence of graded amounts of CD25 Tcells. Selected cultures, as indicated in the text, were supplemented with TNF (50 to test suppressive functions, we now identify the cellular program ng/ml), rhIL-2 (100 IU/ml), anti-TNFR2 mAbs (10 mg/ml), or Infliximab that mediates the inhibitory effect of TNF on the Treg suppressive (50 mg/ml) at time 0 and 72 h later. As indicated, we also activated the Treg/ function. Our data show a central role for activation of the canonical Tcon cocultures with K562 artificial APCs, engineered to stably express the NF-kB pathway—primarily in CD45RA– Tregs—that consequently FcR CD32 alone (KT32), or with 4.1BBL (KT32/4.1BBL), kindly provided by Carl H. June (University of Pennsylvania Cancer Center, Philadelphia, induces a distinctive proinflammatory cellular program. PA). To induce potent TCR activation, the KT32 cells were preincubated with OKT3 and anti-CD28 mAbs at a concentration of 1 mg/ml for 10 min Materials and Methods and then irradiated with 10,000 cGy. Regardless of the activation protocol, 5 d later the T cells were harvested and analyzed by FACS. CD4+ Tcon Human subjects proliferation was determined using the proliferation platform of FlowJo Blood samples were obtained from healthy adult blood donors. In- 7.2.5 to calculate the division index (DI) and draw a “fit model” of gen- flammatory synovial fluid (SF) samples were obtained from JIA patients. erations onto the CFSE dilution histograms. The DI reflects the average Signed written informed consent or parental/guardian permission was number of cell divisions that the cells underwent and is considered a good obtained from all of the study participants as appropriate. The Institutional objective value to compare the rate of proliferation from sample to sample. Ethics Committee at the Chaim Sheba Medical Center approved the study. IFN-g analysis Reagents To assess the capacity of Tregs to suppress the production of IFN-g by Tcons, 100 ml supernatant samples were collected (on day 3) from the Recombinant human IL-2 (rhIL-2) was obtained from Boeheringer-Man- Downloaded from various Treg/Tcon cocultures. The supernatants were then analyzed by the nheim (Mannheim, Germany). Recombinant human TNF-a was purchased human IFN-g cytoset ELISA (Biosource) according to the manu- from Biosource Invitrogen (Carlsbad, CA). The humanized chimeric anti- facturer’s instructions. For intracellular cytokine detection, the T cells TNF mAb, Infliximab, was from Centocor (Horsham, PA). The mouse anti- were activated with 20 ng/ml PMA and 0.8 mM ionomycin (Sigma-Al- hFOXP3 mAb 236A/E7 and Treg staining kit were obtained from eBio- drich) in the presence of monensin (GolgiStop from BD Biosciences) at science (San Diego, CA). The fluorochrome-conjugated mouse mAbs against a concentration of 2 mg/ml for 5 h. Subsequently, the cells were fixed/ human CD4, CD134, CD137, CD95, CD69, CD25, CD45RA, CD120A, permeabilized using the eBioscience FOXP3 staining buffer set and then CD27, and CD127 were all from BD Pharmingen (San Jose, CA). The mouse stained with anti–IFN-g–PE, CD4–FITC, and FOXP3–allophycocyanin http://www.jimmunol.org/ mAbs against TNFR2 (CD120B) were from R&D Systems (Minneapolis, mAbs (all from eBioscience), as previously described (17). MN). The fluorochrome-conjugated rabbit mAbs to phospho-IkB were from Cell Signaling Technology (Danvers, MA), and the mouse mAbs to phospho- Isolation of RNA and real-time quantitative PCR NF–kB-p65 were from BD Biosciences (San Jose, CA). Total RNAwas isolated using the RNeasy mini kit (Qiagen, Valencia, CA), and FACS analysis cDNAwas synthesized using the RETROscript kit (Ambion, Austin, TX). The real-time quantitative PCR (RTQPCR) for mRNA levels of FOXP3, STAT1, Cell samples were analyzed either on a FACSCalibur using the Cellquest NFKB2, TRAF3, LTA, and the housekeeping gene HPRT1 was carried out software or on a digital flow cytometer, FACSCanto II or FACSAria, using using TaqMan kits supplied with predesigned optimized the FACSDiva 6.1.2 software (both from BD Biosciences). FACS data primers and probes (Applied Biosystems, Foster City, CA). The samples were analysis was done with the FlowJo 7.2.5 software (Tree Star, Ashland, OR). run according to the manufacturer’s instructions, in triplicates, using an ABI by guest on September 25, 2021 Immunostaining of cell surface markers was carried out as previously de- Prism 7900HT Sequence Detector (Applied Biosystems). The relative gene scribed (16). For detection of FOXP3, the cells were fixed/permeabilized expression was normalized to HPRT1 expression, and fold change was cal- using the eBioscience FOXP3 staining buffer set, as previously described culated using the 22DDCT comparative method. (17). Gene expression arrays, gene enrichment, and functional Treg and Tcon cell isolation analyses PBMCs and SF lymphocytes were obtained by density gradient centrifu- All of the experiments were performed using Affymetrix HU GENE 1.0 ST gation on Histopaque 1077 (Sigma-Aldrich, St. Louis, MO). Highly pure Array (Affymetrix, Santa Clara, CA). Sample processing was performed CD4+CD25HI Tregs were obtained as previously described (18). Briefly, + according to the Affymetrix WT protocol (www.affymetrix.com/support). CD4 T cells were isolated from PBMCs by positive selection with CD4 The gene-level log-scale robust multiarray analysis sketch algorithm microbeads (MBs) from Miltenyi Biotec (Bergisch Gladbach, Germany) (Affymetrix Expression Console and Partek Genomics Suite 6.2) was used and further purified into CD4+CD25HI and CD25– T cells using a FACS- ++ – for crude data analysis. Genes were filtered and analyzed using fold change Aria cell sorter system (Fig. 2A). Less pure CD25 and CD25 T cell calculations and unsupervised hierarchical cluster analysis (Spotfire De- populations were isolated using positive selection with CD25 MBs (Mil- ++ cisionSite for Functional Genomics). Further processing including func- tenyi Biotec). To obtain a population enriched for CD25 T cells, we tional gene networks analysis and gene set enrichment analysis were modified the protocol recommended by the manufacturer and used 7.5 ml performed using DAVID Bioinformatics Resources (http://david.abcc. 3 7 (instead of 10 ml) CD25 MBs per 1 10 PBMCs. ncifcrf.gov/tools.jsp) or the Ingenuity Pathways Analysis (IPA) platform or T cell cultures and activation based on publication data sources as indicated. The primary microarrays data from this research have been deposited in the NCBI Gene Expression The cells were cultured in RPMI 1640 supplemented with 10% FBS, 2 mM L- Omnibus data repository under accession number GSE18893 (www.ncbi. glutamine, 100 U/ml penicillin, and 100 mg/ml streptomycin (all from In- nlm.nih.gov/geo/query/acc.cgi?acc=GSE18893). vitrogen Life Technologies, Carlsbad, CA) and maintained at 37˚C in a hu- midified 5% CO2 incubator. Unless otherwise specified, the cultures were Statistical analysis supplemented with rhIL-2 (100 IU/ml). Purified T cells were typically plated into 24-well plates (Costar; Corning, Lowell, MA) at 2 3 106 cells per well. The DAVID Expression Analysis Systematic Explorer (EASE) score (a p For TNF treatment, the various purified T cell populations were incubated modified Fisher exact value) was used to determine gene-cluster over- p , p for the indicated time with medium supplemented with 50 ng/ml TNF (or representation, and 0.01 was considered significant. The values for t control PBS). T cell activation and expansion were generally induced by im- the data in Fig. 5 were calculated by the paired Student test with loga- p , mobilized plastic-bound OKT3 (2.5 mg/ml), soluble anti-hCD28 mAbs (1 mg/ rithmic transformation, and 0.05 was considered significant. ml), and rhIL-2 (100 IU/ml). Cell cultures were supplemented with fresh complete medium plus 100 IU/ml rhIL-2 every 72 h until the end of the ex- Results periment. CD45RA–FOXP3+ Tregs express higher levels of TNFR2 Suppression assay and CFSE dilution compared with other T cell subsets + Tregs, isolated from PBMCs, were tested for their ability to suppress the Previous studies (4, 19) show that at homeostasis blood-derived CD4 proliferation of autologous CD4+CD25– Tcons, as previously described (6). FOXP3+ T cells express as a whole higher levels of TNFR2/CD120B 3572 TNF-DEPENDENT NF-kB ACTIVATION MODULATES TREG FUNCTION compared with CD25–FOXP3– T cells. It is also postulated that the To analyze the expression of CD120B on in vivo activated human human FOXP3+ Treg lineage is subdivided into two subsets based on Tregs versus Tcons, we isolated T cells from inflamed SF of JIA CD45RA expression (20, 21). A recent study further suggests sub- patients (Fig. 1D, left panel). Of note, in the patient samples tested dividing circulating human T cells into six subsets (22), based on (n . 5), nearly all of the synovial FOXP3+ T cells were CD45RA– CD45RA, CD25, and/or FOXP3 expression. As shown in Fig. 1A (left and thus were akin to acTregs. The expression of CD120B on the panel), this classification includes CD25+++/FOXP3HI/CD45RA– ac- in vivo activated synovial CD4+CD25– Tcons was indeed upre- tivated Tregs (acTregs) that are immediate effectors of suppression and gulated compared with that on the isotype staining and circulating CD25++/FOXP3lo/CD45RA+ naı¨ve/resting Tregs (nrTregs) (R2 and Tcons. However, CD120B expression on synovial CD45RA– R1, respectively). We now show that the highest expression of FOXP3+ was still significantly higher compared with that on their CD120B is detected on CD45RA–CD25+++ acTregs (R2), followed by FOXP3– Tcon counterparts (Fig. 1D, right panel). CD25++CD45RA+ nrTregs (R1) and CD4+CD25++CD45RA– (FOX- Thus, TNFR2—the main TNFR in human T cells—is generally P3lo) recently activated Tcons (R3). By contrast, FOXP3–CD45RA– expressed at higher levels in FOXP3+ acTregs compared with Tcons (R4 and R5) express much lower levels of CD120B, and naive those in Tcons, in vivo at homeostasis and during inflammatory CD25– CD45RA+ Tcons (R6) do not show meaningful expression of responses as well as upon ex vivo activation. this receptor (Fig. 1A, right panel, numbers correspond to mean fluorescence intensity [MFI] of CD120B staining). TNF induces preferential upregulation of multiple gene Next, we purified by a FACSAria cell sorter only the CD4+ transcripts in acTregs T cells with the top ∼5% of CD120B expression and found that Because CD45RA– acTregs distinctively express high levels of the purified TNFR2HI population mostly contained FOXP3+ TNFR2 and, moreover, they are postulated to mediate the im- Downloaded from T cells with both low and high FOXP3 expression (Fig. 1B). We mediate effector–suppressor function (22), it was of interest to also found that freshly isolated circulating CD8+ T cells, regard- study the unique effects of TNF on acTregs. To obtain pure ac- less of CD25 coexpression, rarely express CD120B. As previously Tregs, we isolated by FACSAria the CD4+CD25+++ T cells with published (4), we also did not detect TNFR1/CD120A expression the top 3% of CD25 expression. This latter CD4+CD25HI pop- on human CD4+ T cells (data not shown). ulation was exceedingly enriched for FOXP3+ Tregs (.90%) that – HI To address the activation-dependent upregulation of CD120B, were also largely CD45RA /CD120B (Fig. 2A). http://www.jimmunol.org/ we activated in vitro CD25+++ Tregs and naive CD25–CD4+ As an initial approach to examine the cellular program induced T cells. As shown in Fig. 1C, ∼24 h after ex vivo activation the by TNF in CD25HI acTregs versus Tcons, we employed micro- Tcons started to upregulate TNFR2 expression. Yet, throughout array gene expression analysis. Thus, the latter purified Treg and the culture period, CD120B levels remained significantly lower in Tcon populations were either treated with 50 ng/ml TNF or left activated Tcons compared with those in the acTregs. No mean- untreated for 2 and 24 h to detect both early and late events. This ingful cell surface expression of CD120A was detected in both relatively high concentration of TNF was used because TNFR2 T cell fractions following activation (data not shown). has low affinity for soluble TNF, and indeed previous observations by guest on September 25, 2021

FIGURE 1. TNFR2 is preferentially upregulated on CD45RA– Tregs at homeostasis and upon immune activation. A, CD4+ T cells isolated from PBMCs were analyzed for surface expression of CD25, CD45RA, and CD120B. Left panel, Contour plot shows the subdivision of the CD4+ population into six subsets, as defined by the expression of CD45RA and CD25 (R1 to R6). Right panel, The relative expression of CD120B in each region is shown in the histogram overlays; numbers correspond to MFI of CD120B staining. B, CD4+ T cells were sorted by FACSAria into CD4+TNFR2HI (black line) and CD4+ TNFR2– (gray line) populations and immediately stained for FOXP3 (left panel) and CD120B (right panel) expression. C, Purified CD25– and CD25+ T cells were activated and cultured for 7 d as described in Materials and Methods. Plot (logarithmic scale) depicts the fold change in the MFI of CD120B staining normalized to isotype staining at the indicated time points. D, Freshly isolated T cells from inflamed SF of a JIA patient were stained for cell surface CD4 and intracellular FOXP3 (left panel). The histogram overlays show the cell surface CD120B expression for the FOXP3– gate (gray line) and FOXP3+ gate (black line), and numbers indicate the MFI of the indicated population (right panel). Data were acquired with a FACSCalibur and FACSCanto II instruments and analyzed with FlowJo 7.2.5 software. Data shown are representative of $3 experiments performed. The Journal of Immunology 3573 imply that in vitro only high concentrations of soluble TNF our microarray gene expression data, we performed gene set modulate the human Treg suppressive capacity (4). enrichment analysis using the online available DAVID Bio- The microarray-based analysis of the mRNA transcripts induced informatics Resources (http://david.abcc.ncifcrf.gov/tools.jsp) that in acTregs by TNF showed that the transcription of 74 genes was use multiple heterogeneous functional gene annotation database upregulated by $2-fold at 2 and/or 24 h compared with that in sources (24). untreated acTregs (Supplemental Table I). Importantly, the number We first performed the analysis for the list of gene transcripts of gene transcripts induced by TNF in Tcons was much lower (total induced by TNF in Tregs at both time points (Supplemental Table 30 genes; Supplemental Table II), and this was evident at both time I). In the enrichment analysis report (Supplemental Table III), we points as shown in the Venn diagram (Fig. 2B). Furthermore, more included only annotation source terms enriched .2-fold with a gene transcripts were upregulated in acTregs at 2 h compared with threshold of minimum five counts and an EASE score (a modified those at 24 h. The transcription of five genes (NFKB2, NFKBIA, Fisher exact p value) ,0.01. This analysis revealed significant RELB, BIRC3, and SGK1) was consistently induced by TNF in the enrichment for three main clusters of functional gene networks, as two T cell populations and at both time points. follows: 1) IKK/NF-kB cascade and its related TNFR2 signaling Of note, the number of gene transcripts downregulated by TNF in pathway (NFKB1, NFKB2, RELB, NFKBIA, TNFAIP3, TANK, both populations was very minimal (data not shown), an obser- IKBKE, LTA, etc.); 2) immune response activation and response vation that correlates with previous findings in other cell types (23). to stimulus (TNFRSF9, FAS, LTA, IL27RA, CD74, CD69, STAT1, Thus, TNF treatment induces a more pronounced gene-transcrip- IRAK2, CD83, CCR8, etc.); and 3) negative and positive regula- tion program in acTregs compared with that in Tcons. tion of apoptosis (NFKB1, TNFAIP3, BIRC3, IFIH1, CD74,

TNFRSF9, FAS, etc.). Notably, the highest-ranking terms in this Downloaded from TNF induces preferential upregulation of typical NF-kB target DAVID-based enrichment analysis were the TNFR2 signaling genes in Tregs pathway (BIOCARTA) and I-kB kinase/NF-kB cascade (GO), Studies analyzing the gene network induced by pure TNFR2 indicating that TNF induces in acTregs robust transcription of stimulation in T cells that express TNFR2 as their sole TNFR have genes that are directly involved in feedback regulation of signal not been published. To identify such functional networks within transduction downstream of TNFR2.

Next, we performed gene set enrichment analysis against a list of http://www.jimmunol.org/ TNF-regulated genes published by Tian et al. (23). In a series of works, these investigators generated a high-quality list of 50 unique annotated genes regulated by TNF in HeLa cells. Using a Tet-regulated system for the expression of a nondegradable NF- kB inhibitor (FLAG-I–kB-Mut), they further identified within the former list a subgroup of 28 genes that were both TNF-regulated and NF-kB–dependent, and importantly all contained at least one functional NF-kB binding site within their promoter. The en- richment analysis revealed (Supplemental Table III, last row) 10 by guest on September 25, 2021 hits with a high fold enrichment score (59.8) and an extremely significant p value of 5.74 3 10217 (EASE score). Of note, no genes were common between our gene list and the related list of TNF-regulated but NF-kB–independent genes also reported in this study (e.g., JUN, JUNB, FZD2, GATA2, etc.). Given the lower number of gene transcripts induced by TNF treatment in Tcons (Supplemental Table II), the gene set enrich- ment analysis for the latter gene list was of a lower yield. The two most relevant GO terms significantly enriched were as follows: immune system process (nine hits, p = 2.62 3 1024) and apoptosis (six hits, p = 6.46 3 1023). Moreover, a much smaller percentage within the genes induced by TNF treatment in Tcons, were also common to the gene list reported by Tian et al. (23) (five hits, p = 1.39 3 1026), and notably these genes were all essential members of the IKK/NF-kB cascade itself (NFKB1, NFKB2, RELB, NFKBIA, and TNFAIP3). To further illustrate the intricate gene–gene interactions within the unique gene network induced by TNF in Tregs, we used the IPA platform’s neighborhood explorer feature, filtered to include only transcription- or expression-related gene–gene interactions in FIGURE 2. TNF induces distinct gene transcripts in CD25HI Tregs the context of the NF-kB pathway and the list of transcripts induced + versus Tcons. A, CD4 T cells were isolated by a FACSAria instrument $1.5-fold by TNF in acTregs (Supplemental Table III). This anal- into CD4+CD25HI T cells with the top 3% of CD25 expression and CD25– + HI – ysis (Fig. 3) illustrates an elaborate NF-kB–dependent genetic Tcons. The purified CD4 CD25 (right panels) and CD25 (left panels) network regulating the transcription or expression of multiple genes, subsets were then analyzed for cell surface CD25, CD45RA, CD120B, and including genes involved in inflammation (e.g., LTA, TNF, ICAM1, intracellular FOXP3. B, Venn diagrams of TNF-induced gene transcription in CD25HI Tregs and CD25– Tcons. Data were obtained after either 2 (left TNFRSF4, TNFRSF9, and IRF1) and the inhibition of apoptosis panel)or24h(right panel) of TNF treatment. Only genes with increased (e.g., TNFAIP3, BIRC3, NFKB1, and BCL2L1/Bcl-xL). Importantly, expression $2-fold in TNF treated versus control media cultures were within this genetic network, several gene products (marked by red included. Data shown are representative of two experiments using T cells rectangles) have been previously associated with the down- from two different healthy blood donors. modulation of Treg-mediated suppression (4, 10, 13, 25–30). 3574 TNF-DEPENDENT NF-kB ACTIVATION MODULATES TREG FUNCTION

TNF preferentially activates the canonical NF-kB pathway in min) revealed that the area under the curve for TNF treatment Tregs (Fig. 4B) was ∼3-fold larger in Tregs (upper panel) compared Previous studies have yielded conflicting results as to whether pure with that in Tcons (lower panel). This measure for PMA/Ion TNFR2 signaling activates the canonical NF-kB pathway (14). The treatment was, in contrast, slightly larger in Tcons compared with hallmark of activation of this pathway is phosphorylation and that in Tregs. In parallel experiments, we found that TNF also induced within 10–15 min a moderate, but reproducible, increase depletion of I-kB followed by phosphorylation and nuclear + HI translocation of NF-kB p65 (31). in the intracellular level of phospho-I–kB within CD4 CD25 To detect changes in p65 and I-kB phosphorylation in Tregs Tregs as compared with that in untreated cultures or TNF-treated versus Tcons, we used phospho-protein–specific multiparametric Tcons (data not shown). flow cytometric analysis. Thus, by costaining for surface CD4 and Thus, taken together the data presented above imply that at homeostasis TNF induces activation of the canonical NF-kB CD25, we could specifically follow the changes in phosphoryla- HI HI – + pathway in CD25 Tregs and that the activation was as a rule less tion (at single-cell resolution) in CD25 and CD25 CD4 T cells – dim within a rather heterogeneous population of T cells. marked in unactivated CD25 (CD120B ) Tcons. Freshly isolated CD25+ and CD25– T cells were rested over- night in the presence of rhIL-2 (100 IU) and then treated for 15 Analysis of TNF effect on functional pathways involved in min with TNF, PMA and ionomycin (PMA/Ion) or vehicle only. inflammation As shown in Fig. 4A, TNF treatment (red heavy line) induced a 2- To facilitate comprehensive inflammation-focused association fold increase in the MFI of phospho-p65 in CD4+CD25HI Tregs analyses, Loza et al. (32) assembled a list of key inflammation- compared with that in untreated cells (green line), whereas the related genes and further divided them into 17 functionally defined Downloaded from same TNF treatment induced a smaller (1.3-fold) increase in subpathways. Thus, we sorted the genes that were induced $1.5- phospho-p65 levels in CD25– Tcons (Fig. 4A, lower panel). As fold by TNF in Tregs (Supplemental Data Set 1) according to this predicted, treating the cells with the potent mitogen PMA/Ion functional classification. (blue line) led to a robust and rapid increase in phospho-p65 levels We found that our gene list was significantly enriched for two of a similar magnitude in both cell types. Furthermore, plotting the subpathways: NF-kB and TNFSF signaling (Tables I, II, re- changes in the MFI of phospho-p65 against time (5, 10, 15, and 30 spectively). Careful inspection of the TNF-induced changes in the http://www.jimmunol.org/ by guest on September 25, 2021

FIGURE 3. TNF induces a prominent NF-kB–dependent gene network in Tregs. The IPA Neighborhood Explorer feature was used to depict known molecular interactions between the NF-kB complex and the different annotated genes upregulated $1.5-fold by TNF. The gene–gene interactions were filtered to include only transcription- or expression-associated connections. The upregulated genes are depicted in shades of red, with red indicating the highest and pink the lowest fold change. The arrows indicate acts on [i.e., induces directly (solid line) or indirectly (dashed line) transcription and/or expression], and each gene or protein is shown in its most likely subcellular location. Gene products within red rectangles have been previously shown to downmodulate Treg function (4, 10, 13, 25–30). The Journal of Immunology 3575

subpathways: NFKB2, STAT1, LTA, and TRAF3. In agreement with the microarray data, we found that TNF treatment primarily upregulated the transcription of these genes in Tregs (Fig. 5A, fold change in parentheses). For instance, the transcription of NFKB2 and LTA was particularly increased in Tregs compared with that in Tcons (19.2- and 4.8-fold versus 4.6- and 1.6-fold, respectively). Next, we analyzed in highly purified naive CD45RA+ Tcons, CD45RA– acTregs, and CD45RA+ nrTregs (sorted by FACSAria) the different effects of TNF treatment on the transcription of relevant genes. As shown in a typical experiment (Table IV), the induction of transcription of the proinflammatory genes NFKB2, STAT1, and LTA by TNF was significantly higher within the ac- Treg population as compared with those within the other two T cell subsets. Next, we asked whether on top of inducing transcription of the TNF gene its translation into protein product was also enhanced in Tregs following exogenous TNF treatment. To address this issue, we isolated CD25+ and CD25– T cells from PBMCs and treated

part of them with TNF for 3 d. The various cultures were then Downloaded from activated with PMA/Ion in the presence of a Golgi transport in- hibitor and immediately analyzed for FOXP3 expression and in- tracellular TNF production. Our data show that the CD4+FOXP3+ T cells within the CD25++ cultures were more strongly induced to produce intracellular TNF following TNF treatment as compared – ++ – with FOXP3 CD25 or CD25 T cells (Supplemental Fig. 1). To http://www.jimmunol.org/ verify that the latter increase in TNF signal detected in Tregs did not originate from exogenous TNF bound to the membrane (via TNFR2, for example), we also stained separately unpermeabilized cells, where indeed no meaningful staining for membrane-bound TNF (mTNF) was detected. Minimal effect of TNF treatment on the Treg transcriptional signature

To identify the effect of TNF on the major genes that distinguish the by guest on September 25, 2021 human Treg lineage, we first scanned our data for transcripts that were .3-fold increased in freshly isolated acTregs compared with those in Tcons. The gene list thus generated (Table III) indeed included many of the canonical Treg genes that shape the Treg transcriptional signature (33, 34). Importantly, we found that overall TNF did not significantly change (.1.5- or ,0.5-fold change) the transcription of this gene list, including FOXP3, FIGURE 4. TNF induces significant phosphorylation of p65 in Tregs. CD25++ and CD25– T cells obtained from healthy blood donors were CTLA4, and IL2RA. Likewise, the transcription of TNFR2 treated with TNF (red line), PMA/Ion (blue), or vehicle only (green). CD4+ (TNFRSF1B) that ranked fifth among the Treg signature genes was T cells were analyzed for cell surface CD25 and intracellular phospho-p65 only minimally affected. levels. A, Histogram overlays show an example of phospho-p65 staining in Because FOXP3 is the master regulator of the Treg lineage (35, the CD25HI (upper panel) and CD25– (lower panel) subsets. B, Plots depict 36) and influences its transcriptional signature, we measured the the fold change in the MFI of intracellular phospho-p65 staining in re- effect of TNF treatment on its mRNA and protein expression. By sponse to TNF or PMA/Ion treatment at the indicated time points (from 5– RTQPCR, we did not detect significant inhibition of FOXP3 30 min), as normalized to the appropriate isotype staining. The plots show + HI + transcription by TNF treatment (Fig. 5B). In parallel, by flow phospho-p65 kinetics within the CD4 CD25 Treg (upper panel) or CD4 cytometry, we also analyzed the changes in FOXP3 protein levels. CD25– Tcon populations (lower panel). Data shown are representative of We found that TNF in the context of IL-2 cotreatment did reduce three independent experiments. by ∼30% the IL-2–dependent upregulation of FOXP3, but im- portantly TNF treatment per se did not significantly change transcription of genes associated with the two latter subpathways baseline FOXP3 expression (Fig. 5C). underlines that their upregulation generally peaked at 2 h and was more pronounced in acTregs. Moreover, TNF induced the tran- TNF specifically induces cell surface expression of 4-1BB, OX40, scription of LTA and the TNF gene itself (both specific TNFR2 and FAS in acTregs ligands), signifying the induction of a positive autocrine feedback As shown in preceding experiments, mRNA levels of three loop. Interestingly, within the large list of TNFRSF genes (Table TNFRSF members were preferentially induced by TNF in acTregs. II), TNF selectively induced the transcription of only three Thus, we tested whether this effect was also translated into an members: 4-1BB (CD137, TNFRSF9), FAS (CD95, TNFRSF6), increase at the protein-product level at 24 h after TNF treatment. By and OX40 (CD134, TNFRSF4). immunostaining and subsequent FACS, analysis we found that both Next, we confirmed by RTQPCR that TNF indeed upregulated 4-1BB and OX40 were robustly induced by TNF in FOXP3+ Tregs the mRNA abundance for four selected genes within these two but not in Tcons. As seen in a typical healthy donor (Fig. 6A), in 3576 TNF-DEPENDENT NF-kB ACTIVATION MODULATES TREG FUNCTION

Table I. List of NF-kB signaling subpathway genes induced by TNF FOXP3+ to FOXP3– events). FAS surface expression was also $1.5-fold increased on circulating Tregs compared with that on Tcons and further upregulated by 6.8-fold on synovial acTregs. Gene Symbol Treg 2 h Treg 24 h Tcon 2 h Tcon 24 h TNF/TNFR2 signaling modulates the Treg suppressive function NFKB2 8.3 4.9 4.1 2.9 NFKBIA 4.8 4.3 3.6 2.8 Previous studies using the in vitro coculture suppression assay RELB 4.6 4.7 2.7 2.5 suggest that exogenous soluble TNF can inhibit the human Treg TANK 2.7 2.1 2.3 1.8 suppressive function (4, 19). To further investigate the functional IRAK2 2.6 1.8 1.8 1.6 a consequences of TNF treatment on Tregs in our system, we co- NFKB1 2.3 NS 2.7 NS + – NFKBIE 2.2 1.8 NS NS cultured freshly isolated CD25 Tregs with CFSE-labeled CD25 NFKBIZ 2.0 NS NS NS Tcons at various ratios. The cocultures were activated by immo- REL 2.0 1.5 1.9 1.6 bilized OKT3 mAbs and treated, as indicated, with TNF and/or BCL3 1.8 NS NS NS rhIL-2, TNF and anti-TNFR2 mAbs, or anti-TNF mAb (In- STAT1 1.6 2.1 NS NS BCL6 1.6 1.7 1.7 1.6 fliximab), or untreated. Tcon proliferation in the various co- cultures was assessed on day 5 by the proliferation analysis tool of Numbers represent fold change in gene expression induced by TNF treatment in either Tregs or Tcons at the indicated time point. the FlowJo 7.2.5 software to calculate the DI and draw a fit model aA ,1.5-fold change. of generations onto the relevant CFSE dilution histograms (Fig. 7A). We found that excess of either exogenous rhIL-2 (100 IU/ml),

Tregs treated with TNF the cell surface levels of 4-1BB and OX- TNF (50 ng/ml), or both modulated the Treg suppressive function, Downloaded from 40 were upregulated 3.4- and 3.2-fold, respectively, as compared which was otherwise very evident and Treg-ratio–dependent (Fig. with untreated cells. This increase was abrogated when the cells 7B, left panel). Importantly, the TNF-dependent inhibition of were pretreated with anti-TNFR2 mAbs (data not shown). The suppression was fully reversed by cotreatment with anti-TNFR2 expression of FAS that was initially much higher in Tregs was blocking mAbs. Infliximab treatment resulted in augmented sup- further enhanced by TNF, whereas FAS expression was only pression, and as previously shown (6, 7), it moreover directly marginally increased by TNF in Tcons. In contrast, the expression reduced Tcon proliferation independent of Treg function. http://www.jimmunol.org/ of CD27, another TNFRSF member, used as a negative control, Next, we analyzed by ELISA the effect of TNF treatment on the was practically unaffected. Inconsistent with the gene microarray Treg capacity to suppress IFN-g secretion by Tcons (Fig. 7B, right data, the cell surface expression of CD69 and CD127 was only panel). Using a similar suppression assay scheme, as above, we marginally increased in Tregs or Tcons following TNF treatment found that TNF or rhIL-2 treatment alone had a moderate in- (Fig. 6A, lower panels). hibitory effect on the robust Treg-dependent suppression of IFN-g To further answer whether these three TNFRSF members are secretion. Importantly, cotreatment with both rhIL-2 and TNF regulated differently by TNF in acTregs versus nrTregs, we treated almost completely inhibited this Treg-dependent suppression. By with TNF blood-derived CD25++/HI T cells, a cell population that contrast, the blockade of TNFR2 by mAbs restored suppression to contains a variable (age- and donor-dependent) percentage of baseline. Infliximab was very potent in reducing overall IFN-g by guest on September 25, 2021 acTregs and nrTregs. The subsequent FACS analysis of a typical secretion in a manner that was quite Treg-independent. experiment revealed that 4-1BB and OX40 were mostly induced Given that TNF consistently induced significant upregulation of by TNF in CD45RA– FOXP3+ Tregs but to a much lesser extent in 4-1BB on acTregs and signaling via the latter receptor has been CD45RA+ Tregs (Fig. 6B, Table IV). strongly implicated in downmodulation of Treg-dependent sup- Because increased TNF levels normally characterize the pression in mice, we next asked whether 4-1BB signaling by ex- inflamed synovium, we also compared FOXP3+ and FOXP32 ogenous 4-1BB ligand has a similar effect in human T cells. To T cells isolated from SF with their counterparts isolated from the address this issue, we performed in vitro suppression assays where blood of the same JIA patient for the cell-surface expression of 4- we activated the Treg/Tcon cocultures with artificial APCs either 1BB, OX40, and FAS (Fig. 6C). Consistent with the data obtained KT32 or KT32/4.1BBL, which were preincubated with OKT3/anti- from in vitro TNF treatment, 4-1BB and OX40 were particularly CD28 mAbs as previously detailed (37). In addition, we also tested upregulated on synovial FOXP3+ acTregs compared with their the effect of cotreatment with TNF on suppression in this system. circulating counterparts (5.4- and 4.6-fold versus 2.1- and 1.7- The data show, as predicted, that either TNF or 4-1BBL alone fold, respectively; values correspond to the ratio of MFI of significantly inhibited suppression but more importantly that TNF treatment—a strong inducer of 4-1BB in acTregs—in the presence of KT32/4.1BBL almost completely abolished suppression (Fig. 8). Table II. List of TNF(R)SF signaling subpathway genes induced by TNF $1.5-fold Discussion This study reveals in detail the mechanism by which TNF mod- Gene Symbol Treg 2 h Treg 24 h Tcon 2 h Tcon 24 h ulates the function of human CD45RA– acTregs and consequently TNFAIP3 3.4 3.1 1.8 2.2 sheds new light on how TNF blockade restores immunologic self- TNFRSF9 3.2 2.1 NS NS tolerance. Our data show that TNF via TNFR2, constitutively LTA 2.4 1.6 1.9 NS expressed in Tregs and particularly at higher levels in the FAS 2.5 2.3 NS 1.7 – TRAF3 2.0 1.9 1.7 1.7 CD45RA subset, induces the activation of the canonical NF-kB TNFRSF4 1.8 NSa NS 1.7 cascade. Moreover, we describe for the first time the distinctive TRAF4 1.6 1.7 NS NS proinflammatory NF-kB–regulated transcription program induced, TNF 1.6 NS NS NS preferentially, in acTregs by TNF that functions to modulate their TRAF1 1.6 NS NS NS TRAF2 1.5 1.7 NS NS suppressive capacity. It should be pointed out that the various experiments analyzing gene expression patterns within CD45RA– Numbers represent fold change in gene expression induced by TNF treatment in + HI either Tregs or Tcons at the indicated time point. Tregs are based on a somewhat heterogeneous CD4 CD25 + aA ,1.5-fold change. T cell population usually containing ∼90% FOXP3 cells. The Journal of Immunology 3577 Downloaded from http://www.jimmunol.org/

FIGURE 5. TNF induces NFKB2, LTA, TRAF3, and STAT1 but does not change FOXP3 transcription. A and B, RTQPCR analysis of indicated gene transcription in human CD4+CD25++ and CD25– T cells treated for 24 h with TNF (50 ng/ml) or left untreated, relative to expression in untreated Tregs and normalized to HPRT1. Numbers in parentheses show the fold change induced by TNF. The y-axis is in log scale, and data (mean 6 1 SD of duplicates) are representative of three independent experiments in different donors. pp , 0.05; ppp , 0.01. C, CD25+ and CD25– T cells were treated for 24 h with IL-2, TNF, both, or media alone and then stained for cell surface CD4 and intracellular FOXP3 (dot plot). The histogram overlays show FOXP3 expression (CD4+ gate) in the various cultures, and numbers represent their respective MFIs. The data represent three independent experiments. by guest on September 25, 2021

Previous studies offer conflicting results regarding the effect of 38, 39). Interestingly, another recent study proposes that the sol- TNF on Treg functions. Two reports based on human data agree that uble (cleaved) form of TNFR2 is of prime importance in medi- TNF directly downmodulates Treg suppressive function (4, 5). In ating the Treg suppressive function, by quenching surplus TNF at contrast, a study in mice implies that TNF may actually augment sites of inflammation (19). Treg suppressive function in vitro and probably in vivo as well Our findings that TNF/TNFR2 signaling inhibits the Treg sup- (12). This discrepancy may represent an example of cross-species pressive function via activation of the NF-kB cascade are the first differences between human and mice T cells. Regarding this it to experimentally demonstrate such a function for TNF. Whereas should be noted that multiple studies in both humans and mice there is ample evidence that TNFR1 signaling in many cell types imply that in vivo TNF blockade augments Treg function (2, 4, 5, other than T cells activates the NF-kB cascade, studies in Tregs showing a link between TNFR2 and NF-kB activation have not been reported (1, 14). A previous study by Valencia et al. (4) that Table III. TNF has a minimal effect on the highest-ranking Treg showed an inhibitory effect of TNF/TNFR2 signaling in Tregs signature genes also concluded that TNF inhibited of FOXP3 transcription, though NF-kB activation was not addressed in this study. By contrast, our Treg versus Gene Symbol Tcona Treg 2 h Treg 24 h Tcon 2 h Tcon 24 h data based on various complementary modalities do not show significant inhibition of FOXP3 transcription during the first 24 h IL2RA 27.2 1.2 1.0 0.9 1.4 after TNF treatment, while demonstrating a central role for the ZNFN1A2 13.5 1.1 1.1 1.2 0.9 HLA-DRA 12.9 0.7 0.7 1.5 1.1 activation of the canonical NF-kB pathway. Moreover, previous FOXP3 9.3 0.9 1.0 1.0 1.0 studies (6, 11) and our present data show that CD4+ T cells iso- TNFRSF1B 7.2 1.0 1.0 1.2 0.8 lated from inflamed SFs of JIA patients—a milieu where TNF is VAV3 5.6 1.1 1.1 0.9 1.0 abundant—are actually enriched for T cells expressing high levels BCL2 4.4 0.8 0.9 1.1 1.2 HLA-DRB1 4.4 0.8 0.9 1.2 0.9 of FOXP3 (Fig. 1D). Studies that have addressed the transcrip- CCR4 4.4 1.0 0.8 0.8 0.8 tional regulation of FOXP3 also do not report that NF-kB/REL CTLA4 4.2 1.0 0.9 0.5 0.8 family members bind conserved DNA motifs within the FOXP3 ZNFN1A4 4.1 1.1 1.0 1.2 1.1 promoter or its other regulatory regions (40, 41). IL2RB 3.8 1.1 1.0 1.0 0.8 Data from a number of studies imply, even if indirectly, that TNF aFold change in gene transcript abundance (numbers in boldface) in untreated via TNFR2 activates a NF-kB pathway-dependent transcription Tregs versus Tcons at 2 h ranked in descending order. Numbers in nonboldface + represent fold change in the indicated gene transcript induced by TNF treatment in program in T cells. In mice lacking TNFR2, stimulation of CD8 either Tregs or Tcons at the specified time point. T cells via TCR and CD28 results in reduction in the time span of 3578 TNF-DEPENDENT NF-kB ACTIVATION MODULATES TREG FUNCTION Downloaded from http://www.jimmunol.org/ by guest on September 25, 2021

FIGURE 6. TNF selectively upregulates expression of OX40, 4-1BB, and FAS on CD45RA– Tregs. A, CD25++ and CD25– T cells isolated from a healthy blood donor were treated with TNF (50 ng/ml) or untreated. Twenty-four hours later, the CD4+ T cells were analyzed for CD25 and one of the following markers: CD95/FAS, CD134/OX40, CD27/TNFRSF7, CD137/4-1BB, CD127/IL7R, or CD69. Numbers represents the MFI of untreated (green) or TNF- treated (red) cells for CD4+ gate. B, In parallel, the CD4+CD25++ T cell subset was analyzed for cell surface coexpression of CD45RA and one of CD137, CD134, or CD95. The contour plots depict either TNF-treated (right panels) or untreated (left panels) cultures, and numbers in parentheses correspond to MFI of CD45RA– events. C, Freshly isolated CD4+ T cells from PBMCs and inflamed SF of a typical JIA patient were analyzed for expression of in- tracellular FOXP3 and cell surface CD137, CD134, or CD95. Left panels, FOXP3+ gate. Right panels, FOXP3– gate. Color-coded numbers correspond to the MFIs. All data represent $3 independent experiments. activation of the I-kB/NF-kB cascade (42). Transgenic mice in- CD4+FOXP3– effectors. These data support a view that the up- duced to express human p75-TNFR2 develop severe multiorgan regulation of TNFR2 on acTregs is not merely a reflection of their inflammatory syndrome associated with constitutive NF-kBcas- activated state but is more likely a subset-specific characteristic. cade activation in the PBMC compartment. Moreover, studies in This latter point is further underlined by the observation that primary cortical neurons from TNFR12/2 mice show that pure TNFRSF1B (TNFR2) ranked among the five top genes that define TNFR2 stimulation induces significant phosphorylation and de- the human Treg transcriptional signature (Table III), a list that also pletion of I-kB followed by nuclear translocation of p65/RelA (43). includes IL2RA, ZNFN1A2, and FOXP3. Consequently, it may be Our data further establish the link between TNF and the human envisaged that TNFR2 is important for Treg homeostasis in vivo, Treg lineage by showing that at homeostasis TNFR2 expression is as it is indeed postulated for the other three genes (44). particularly increased on circulating acTregs and moreover that It can be hypothesized that at homeostasis low physiological FOXP3+ T cells can be sorted out with good purity based on the levels of mTNF expressed by various cells may support Treg expression of TNFR2. In addition, in vivo induced FOXP3+ Tregs maintenance by, for example, delivering NF-kB–dependent anti- isolated from inflamed SFs of arthritis patients express signifi- apoptotic signals. However, persistently high levels of mTNF—as cantly higher levels of TNFR2 compared with their counterpart occurs during autoimmune joint inflammation—directly inhibit The Journal of Immunology 3579

Table IV. TNF primarily induces NF-kB–dependent gene transcription transmembrane TNF transgenic mice invariantly develop aggres- – in CD45RA acTregs sive inflammatory arthritis that indeed can be ameliorated by genetic deletion of p75 or treatment with recombinant soluble Gene Fold Fold human TNFR2 (45, 46). T Cell Subset Symbol Changea CD No. Changeb Among the proinflammatory immune response genes induced by STAT1 CD95 TNF specifically in acTregs, we have identified two members of the CD25++ RA+ 2.1 0.9 2 TNFSF (LTA and TNF) and three members of the TNFRSF (FAS, CD25+++ RA 4.6 1.3 CD252 RA+ 1.6 1.3 TNFRSF4, and TNFRSF9). Evidence is moreover provided that + NFKB2 CD134 this may occur in situ, because FOXP3 synovial T cells most CD25++ RA+ 2.8 1.3 likely exposed in vivo to elevated TNF levels significantly upre- +++ 2 CD25 RA 14.1 2.8 gulate the expression of these TNFRSF members as compared CD252 RA+ 2.4 1.0 LTA CD137 with their blood-circulating counterparts. Regarding the role of CD25++ RA+ 1.1 1.3 TNFRSF4 and TNFRSF9 in Treg physiology, although it seems CD25+++ RA2 19.2 3.4 complex- and context-dependent, the prevailing view [as recently 2 CD25 RA+ 1.9 1.1 reviewed by Croft et al. (30)] is that signaling via OX40 or 4-1BB Boldface numbers indicate acTreg population. during immune activation can reversibly inhibit Treg suppressive a Fold change in gene transcription, as measured by RTQPCR, induced by TNF functions. This occurs by both a direct effect on Tregs and an treatment (relative to untreated cells and normalized to HPRT1) in the indicated Treg subset or Tcons. indirect effect on Tcons boosting their resistance to suppression. b Fold change in MFI staining of the specified CD marker induced by TNF treat- Indeed, our data from suppression assays using KT32 transfected Downloaded from ment (relative to untreated cells). to express human 4-1BBL suggest that 4-1BB signaling further enhances the inhibitory effects of TNF on acTregs. Thus, we Treg function as shown in this study and very likely also induce propose that paracrine TNF (soluble or membrane-bound) from some level of resistance in activated TNFR2+ effector Tcons to- Tcons and/or APCs induces a NF-kB–dependent positive feed- ward suppression. This scenario is supported by observations back loop in acTregs that includes expression of cell surface 4-

made in the transmembrane TNF transgenic mouse model en- 1BB plus OX-40. The latter TNFRSF members subsequently http://www.jimmunol.org/ gineered to overexpress transmembrane forms of TNF. The (upon ligation) further activate the NF-kB cascade to promote by guest on September 25, 2021

FIGURE 7. TNF and/or IL-2 modulate the Treg suppressive function. CD252 Tcons were labeled with CFSE, plated into 96-well plates, and activated by plastic-bound OKT3 alone or in the presence of graded amounts of CD25++ Tregs. A, Five days later, the CD4+ Tcons were ana- lyzed for CFSE dilution by FACS. We used the Proliferation Platform (FlowJo 7.2.5) to calculate the DI and draw a fit model of generations onto FIGURE 8. TNF and 4-1BB signaling cooperate to abrogate in vitro the CFSE dilution histograms. Shown are examples of proliferation anal- suppression. In this suppression assay KT32 or KT32/4.1BBL preincubated ysis results for untreated dividing Tcons (right panel) or at a 1:1 ratio of with OKT/CD28.2 mAbs were used as artificial APCs to activate the co- Tcons to Tregs (left panel). B, The plot depicts the DI values of Tcons at cultures. Where indicated the cocultures were also treated with TNF-a different Treg ratios. The different cocultures were treated (as indicated in (50 ng/ml). Five days later, CD4+ Tcons were analyzed for CFSE dilution, as the embedded legend) with TNF, IL-2, TNF and IL-2, TNF and anti- detailed in Fig. 7. A, Examples of the analysis of activated dividing Tcons TNFR2 mAbs, or Infliximab or left untreated. C, The plot depicts IFN-g alone (right panel) or at 1:1 ratio with Tregs (left panel). B, The plot depicts levels in the supernatants of the same cocultures collected 3 d after acti- DI values of Tcons at different Tcon to Treg ratios. The type of APC and vation (logarithmic scale; values are mean of duplicates). Data represent treatment used are indicated in the embedded legend. Data represent three three independent experiments. independent experiments. 3580 TNF-DEPENDENT NF-kB ACTIVATION MODULATES TREG FUNCTION transient Treg dysfunction. Preliminary in vitro experiments sug- tion limits activation-induced expression of FOXP3 in conventional human CD252CD4+ T cells. Int. Immunol. 20: 1041–1055. gest that TNF pretreatment, apart from inducing FAS upregulation, 18. Earle, K. E., Q. Tang, X. Zhou, W. Liu, S. Zhu, M. L. Bonyhadi, and specifically increases the susceptibility of acTregs to undergo ap- J. A. Bluestone. 2005. In vitro expanded human CD4+CD25+ regulatory T cells optosis upon exposure to agonistic mAbs to FAS, yet because this suppress effector T cell proliferation. Clin. Immunol. 115: 3–9. 19. van Mierlo, G. J., H. U. Scherer, M. Hameetman, M. E. Morgan, R. Flierman, system is rather artificial, it is difficult to determine whether TNF T. W. Huizinga, and R. E. Toes. 2008. Cutting edge: TNFR-shedding by CD4+ actually renders acTregs susceptible to apoptosis in vivo. CD25+ regulatory T cells inhibits the induction of inflammatory mediators. J. In summary, we show that TNF induces a TNFR2/NF-kB– Immunol. 180: 2747–2751. – 20. Hoffmann, P., R. Eder, T. J. Boeld, K. Doser, B. Piseshka, R. Andreesen, and dependent proinflammatory program predominantly in CD45RA M. Edinger. 2006. Only the CD45RA+ subpopulation of CD4+CD25high T cells acTregs that counters their innate suppressive capacity but without gives rise to homogeneous regulatory T-cell lines upon in vitro expansion. Blood drastically altering their core transcriptional signature. It can be 108: 4260–4267. 21. Valmori, D., A. Merlo, N. E. Souleimanian, C. S. Hesdorffer, and M. Ayyoub. envisioned that when TNF levels drop, for example, during anti- 2005. A peripheral circulating compartment of natural naive CD4+ Tregs. J. Clin. TNF therapy, the Treg suppressive function can be reinstated Invest. 115: 1953–1962. 22. Miyara, M., Y. Yoshioka, A. Kitoh, T. Shima, K. Wing, A. Niwa, C. Parizot, rather rapidly, a mechanism that may explain the remarkable C. Taflin, T. Heike, D. Valeyre, et al. 2009. Functional delineation and differ- clinical efficacy of TNF blockade in restoring immunological self- entiation dynamics of human CD4+ T cells expressing the FoxP3 transcription tolerance in many treated arthritis patients (47). factor. Immunity 30: 899–911. 23. Tian, B., D. E. Nowak, M. Jamaluddin, S. Wang, and A. R. Brasier. 2005. Identification of direct genomic targets downstream of the nuclear factor-kB Acknowledgments mediating tumor necrosis factor signaling. J. Biol. Chem. 280: 17435–17448. This work was performed in partial fulfillment of the requirements for 24. Sherman, B. T., D. W. Huang, Q. Tan, Y. Guo, S. Bour, D. Liu, R. Stephens, a Ph.D. degree of M.N. and H.V. (The George S. Wise Faculty of Life Sci- M. W. Baseler, H. C. Lane, and R. A. Lempicki. 2007. DAVID Knowledgebase: Downloaded from ences, Tel Aviv University, Israel). G.R. holds the Djerassi Chair in Oncol- a gene-centered database integrating heterogeneous gene annotation resources to ogy at the Tel Aviv University Sackler Faculty of Medicine, Israel. facilitate high-throughput gene functional analysis. BMC Bioinformatics 8: 426. 25. Valzasina, B., C. Guiducci, H. Dislich, N. Killeen, A. D. Weinberg, and M. P. Colombo. 2005. Triggering of OX40 (CD134) on CD4+CD25+ T cells Disclosures blocks their inhibitory activity: a novel regulatory role for OX40 and its com- parison with GITR. Blood 105: 2845–2851. The authors have no financial conflicts of interest. 26. Choi, B. K., J. S. Bae, E. M. Choi, W. J. Kang, S. Sakaguchi, D. S. Vinay, and

B. S. Kwon. 2004. 4-1BB-dependent inhibition of immunosuppression by acti- http://www.jimmunol.org/ vated CD4+CD25+ T cells. J. Leukoc. Biol. 75: 785–791. References 27. Fragale, A., L. Gabriele, E. Stellacci, P. Borghi, E. Perrotti, R. Ilari, A. Lanciotti, A. L. Remoli, M. Venditti, F. Belardelli, and A. Battistini. 2008. IFN regulatory 1. Baud, V., and M. Karin. 2001. Signal transduction by tumor necrosis factor and + + its relatives. Trends Cell Biol. 11: 372–377. factor-1 negatively regulates CD4 CD25 regulatory T cell differentiation by 2. McDevitt, H., S. Munson, R. Ettinger, and A. Wu. 2002. Multiple roles for tumor repressing Foxp3 expression. J. Immunol. 181: 1673–1682. 28. Liu, H., M. Komai-Koma, D. Xu, and F. Y. Liew. 2006. Toll-like receptor 2 necrosis factor-a and lymphotoxin a/b in immunity and autoimmunity. Arthritis + + Res. 4(Suppl 3): S141–S152. signaling modulates the functions of CD4 CD25 regulatory T cells. Proc. Natl. 3. Lin, J., D. Ziring, S. Desai, S. Kim, M. Wong, Y. Korin, J. Braun, E. Reed, Acad. Sci. USA 103: 7048–7053. D. Gjertson, and R. R. Singh. 2008. TNFa blockade in human diseases: an 29. Jana, S., P. Jailwala, D. Haribhai, J. Waukau, S. Glisic, W. Grossman, M. Mishra, overview of efficacy and safety. Clin. Immunol. 126: 13–30. R. Wen, D. Wang, C. B. Williams, and S. Ghosh. 2009. The role of NF-kBand Smad3 in TGF-b-mediated Foxp3 expression. Eur. J. Immunol. 39: 2571–2583.

4. Valencia, X., G. Stephens, R. Goldbach-Mansky, M. Wilson, E. M. Shevach, and by guest on September 25, 2021 30. Croft, M. 2009. The role of TNF superfamily members in T-cell function and P. E. Lipsky. 2006. TNF downmodulates the function of human CD4+CD25hi T- diseases. Nat. Rev. Immunol. 9: 271–285. regulatory cells. Blood 108: 253–261. 31. Karin, M., and A. Lin. 2002. NF-kB at the crossroads of life and death. Nat. 5. Ehrenstein, M. R., J. G. Evans, A. Singh, S. Moore, G. Warnes, D. A. Isenberg, Immunol. 3: 221–227. and C. Mauri. 2004. Compromised function of regulatory T cells in rheumatoid 32. Loza, M. J., C. E. McCall, L. Li, W. B. Isaacs, J. Xu, and B. L. Chang. 2007. arthritis and reversal by anti-TNFa therapy. J. Exp. Med. 200: 277–285. Assembly of inflammation-related genes for pathway-focused genetic analysis. 6. Goldstein, I., S. Ben-Horin, A. Koltakov, H. Chermoshnuk, V. Polevoy, PLoS One 2: e1035. Y. Berkun, N. Amariglio, and I. Bank. 2007. a b Integrin+ and regulatory 1 1 33. Zheng, Y., S. Z. Josefowicz, A. Kas, T. T. Chu, M. A. Gavin, and Foxp3+ T cells constitute two functionally distinct human CD4+ T cell subsets A. Y. Rudensky. 2007. Genome-wide analysis of Foxp3 target genes in de- oppositely modulated by TNFa blockade. J. Immunol. 178: 201–210. veloping and mature regulatory T cells. Nature 445: 936–940. 7. Nadkarni, S., C. Mauri, and M. R. Ehrenstein. 2007.Anti-TNF-a therapy induces 34. Hill, J. A., M. Feuerer, K. Tash, S. Haxhinasto, J. Perez, R. Melamed, D. Mathis, a distinct regulatory T cell population in patients with rheumatoid arthritis via and C. Benoist. 2007. Foxp3 transcription-factor-dependent and -independent TGF-b. J. Exp. Med. 204: 33–39. regulation of the regulatory T cell transcriptional signature. Immunity 27: 786– 8. Korn, T., J. Reddy, W. Gao, E. Bettelli, A. Awasthi, T. R. Petersen, 800. B. T. Ba¨ckstro¨m, R. A. Sobel, K. W. Wucherpfennig, T. B. Strom, et al. 2007. 35. Hori, S., T. Nomura, and S. Sakaguchi. 2003. Control of regulatory T cell de- Myelin-specific regulatory T cells accumulate in the CNS but fail to control velopment by the transcription factor Foxp3. Science 299: 1057–1061. autoimmune inflammation. Nat. Med. 13: 423–431. 36. Fontenot, J. D., M. A. Gavin, and A. Y. Rudensky. 2003. Foxp3 programs the 9. Mellanby, R. J., D. C. Thomas, and J. Lamb. 2009. Role of regulatory T-cells in development and function of CD4+CD25+ regulatory T cells. Nat. Immunol. 4: autoimmunity. Clin. Sci. (Lond.) 116: 639–649. 330–336. 10. van Amelsfort, J. M., J. A. van Roon, M. Noordegraaf, K. M. Jacobs, 37. Suhoski, M. M., T. N. Golovina, N. A. Aqui, V. C. Tai, A. Varela-Rohena, J. W. Bijlsma, F. P. Lafeber, and L. S. Taams. 2007. Proinflammatory mediator- M. C. Milone, R. G. Carroll, J. L. Riley, and C. H. June. 2007. Engineering + + induced reversal of CD4 ,CD25 regulatory T cell-mediated suppression in artificial antigen-presenting cells to express a diverse array of co-stimulatory rheumatoid arthritis. Arthritis Rheum. 56: 732–742. molecules. Mol. Ther. 15: 981–988. 11. Ruprecht, C. R., M. Gattorno, F. Ferlito, A. Gregorio, A. Martini, 38. Lee, L. F., B. Xu, S. A. Michie, G. F. Beilhack, T. Warganich, S. Turley, and A. Lanzavecchia, and F. Sallusto. 2005. Coexpression of CD25 and CD27 H. O. McDevitt. 2005. The role of TNF-a in the pathogenesis of type 1 diabetes + identifies FoxP3 regulatory T cells in inflamed synovia. J. Exp. Med. 201: in the nonobese diabetic mouse: analysis of dendritic cell maturation. Proc. Natl. 1793–1803. Acad. Sci. USA 102: 15995–16000. 12. Chen, X., M. Ba¨umel, D. N. Ma¨nnel, O. M. Howard, and J. J. Oppenheim. 2007. 39. Wu, A. J., H. Hua, S. H. Munson, and H. O. McDevitt. 2002. Tumor necrosis Interaction of TNF with TNF receptor type 2 promotes expansion and function factor-a regulation of CD4+CD25+ T cell levels in NOD mice. Proc. Natl. Acad. of mouse CD4+CD25+ T regulatory cells. J. Immunol. 179: 154–161. Sci. USA 99: 12287–12292. 13. Walker, L. S. 2009. Regulatory T cells overturned: the effectors fight back. 40. Mantel, P. Y., N. Ouaked, B. Ru¨ckert, C. Karagiannidis, R. Welz, K. Blaser, and Immunology 126: 466–474. C. B. Schmidt-Weber. 2006. Molecular mechanisms underlying FOXP3 in- 14. Carpentier, I., B. Coornaert, and R. Beyaert. 2004. Function and regulation of duction in human T cells. J. Immunol. 176: 3593–3602. tumor necrosis factor type 2. Curr. Med. Chem. 11: 2205–2212. 41. Huehn, J., J. K. Polansky, and A. Hamann. 2009. Epigenetic control of FOXP3 15. Holtmann, M. H., M. Schuchmann, G. Zeller, P. R. Galle, and M. F. Neurath. expression: the key to a stable regulatory T-cell lineage? Nat. Rev. Immunol. 9: 2002. The emerging distinct role of TNF-receptor 2 (p80) signaling in chronic 83–89. inflammatory disorders. Arch. Immunol. Ther. Exp. (Warsz.) 50: 279–288. 42. Kim, E. Y., and H. S. Teh. 2004. Critical role of TNF receptor type-2 (p75) as 16. Goldstein, I., S. Ben-Horin, J. Li, I. Bank, H. Jiang, and L. Chess. 2003. Ex- a costimulator for IL-2 induction and T cell survival: a functional link to CD28. pression of the a1b1 integrin, VLA-1, marks a distinct subset of human CD4+ J. Immunol. 173: 4500–4509. memory T cells. J. Clin. Invest. 112: 1444–1454. 43. Marchetti, L., M. Klein, K. Schlett, K. Pfizenmaier, and U. L. Eisel. 2004. Tumor 17. Nagar, M., H. Vernitsky, Y. Cohen, D. Dominissini, Y. Berkun, G. Rechavi, necrosis factor (TNF)-mediated neuroprotection against glutamate-induced ex- N. Amariglio, and I. Goldstein. 2008. Epigenetic inheritance of DNA methyla- citotoxicity is enhanced by N-methyl-D-aspartate receptor activation. Essential The Journal of Immunology 3581

role of a TNF receptor 2-mediated phosphatidylinositol 3-kinase-dependent NF- in transgenic mice expressing transmembrane tumor necrosis factor alpha. Ar- kB pathway. J. Biol. Chem. 279: 32869–32881. thritis Rheum. 54: 2872–2885. 44. Setoguchi, R., S. Hori, T. Takahashi, and S. Sakaguchi. 2005. Homeostatic main- 46. Alexopoulou, L., M. Pasparakis, and G. Kollias. 1997. A murine transmembrane tenance of natural Foxp3+ CD25+ CD4+ regulatory T cells by interleukin (IL)-2 and tumor necrosis factor (TNF) transgene induces arthritis by cooperative p55/p75 induction of autoimmune disease by IL-2 neutralization. J. Exp. Med. 201: 723–735. TNF receptor signaling. Eur. J. Immunol. 27: 2588–2592. 45. Edwards, C. K., III, A. M. Bendele, L. I. Reznikov, G. Fantuzzi, E. S. Chlipala, 47. Feldmann, M., and S. R. Maini. 2008. Role of cytokines in rheumatoid L. Li, L. L. Moldawer, J. D. Mountz, Y. Y. Li, and C. A. Dinarello. 2006. Soluble arthritis: an education in pathophysiology and therapeutics. Immunol. Rev. 223: human p55 and p75 tumor necrosis factor receptors reverse spontaneous arthritis 7–19. Downloaded from http://www.jimmunol.org/ by guest on September 25, 2021

Nagar et al. Suppl. Table I List of genes induced by TNF 2-fold in Treg at 2 or 24 hours symbol name Treg 2hrs Treg 24hrs Tcon 2hrs Tcon 24hrs nuclear factor of kappa light poly- NFKB2 8.34 4.92 4.11 2.94 peptide gene enhancer in B-cells 2 RGS1 regulator of G-protein signalling 1 5.54 7.05 1.25 2.37 nuclear factor of kappa light poly- NFKBIA peptide gene enhancer in B-cells 4.85 4.28 3.62 2.78 inhibitor, alpha v- reticuloendotheliosis viral oncogene homolog B, nuclear factor RELB 4.56 4.74 2.73 2.47 of kappa light polypeptide gene enhancer in B-cells 3 X open reading frame CXorf21 4.29 3.64 1.53 2.89 21 CDGAP Cdc42 GTPase-activating protein 4.01 1.38 2.14 1.13 serum/glucocorticoid regulated SGK 3.78 3.54 2.36 2.22 kinase tumor necrosis factor, alpha-induced TNFAIP3 3.42 3.07 1.77 2.22 protein 3 tumor necrosis factor receptor TNFRSF9 3.23 2.11 1.03 1.34 superfamily, member 9 inhibitor of kappa light polypeptide IKBKE gene enhancer in B-cells, kinase 3.00 2.27 1.91 2.12 epsilon SDC4 syndecan 4 (amphiglycan, ryudocan) 3.00 1.30 1.29 1.43 CD274 CD274 molecule 2.88 1.57 1.54 1.70 ZNF267 protein 267 2.81 1.92 1.16 0.96 CFL2 cofilin 2 (muscle) 2.80 0.91 0.95 0.96 TTC7A tetratricopeptide repeat domain 7A 2.78 2.02 2.27 1.72 ymphotoxin alpha (TNF superfamily, LTA 2.77 1.78 1.97 1.52 member 1) BIRC3 baculoviral IAP repeat-containing 3 2.70 2.00 3.06 2.76 TRAF family member-associated TANK 2.68 2.07 2.25 1.76 NFKB activator PASK PAS domain containing serine 2.63 2.96 1.15 1.75 SNORD25 small nucleolar RNA, C 2.62 0.77 0.48 0.79 UDP-Gal:betaGlcNAc beta 1,4- B4GALT5 2.59 1.63 1.47 1.15 galactosyltransferase, polypeptide 5 interleukin-1 receptor-associated IRAK2 2.57 1.79 1.83 1.56 kinase 2 human immunodeficiency virus type I HIVEP1 2.55 1.85 1.98 1.10 enhancer binding protein 1 CD83 CD83 molecule 2.53 2.05 1.42 1.40 basic transcription BATF 2.52 2.23 1.47 1.38 factor, ATF-like Fas (TNF receptor superfamily, FAS 2.50 2.27 1.46 1.73 member 6) TRAF-interacting protein with a TIFA 2.48 2.12 1.78 2.12 forkhead-associated domain

1

LOC158830 similar to Ab2-183 2.45 1.39 1.48 1.57 interferon induced with helicase IFIH1 2.41 1.85 2.35 1.54 Cdomain 1 PELI1 pellino homolog 1 2.40 1.97 3.50 1.97 nuclear factor of kappa NFKB1 lightpolypeptide gene enhancer in B- 2.35 1.33 2.73 1.46 cells 1 IL27RA interleukin 27 receptor, alpha 2.33 1.98 2.61 1.47 MGC22014 hypothetical protein MGC22014 2.28 1.35 1.38 1.40 HEG1 HEG homolog 1 (zebrafish) 2.26 1.28 1.64 1.04 TFRC (p90, CD71) 2.23 1.31 0.81 0.93 KLHL5 kelch-like 5 (Drosophila) 2.22 2.10 1.48 1.54 GPR174 G protein-coupled receptor 174 2.22 1.59 1.34 1.45 SPPL2A signal peptide peptidase-like 2A 2.21 1.35 1.76 1.23 Notch homolog 2 (Drosophila) N- NOTCH2NL 2.19 1.38 1.13 1.61 terminal like nuclear factor of kappa light NFKBIE polypeptide gene enhancer in B-cells 2.19 1.75 1.20 1.45 inhibitor, epsilon CCND2 cyclin D2 2.17 1.39 1.05 1.05 zinc finger CCCH-type containing ZC3H12A 2.17 1.93 1.30 1.29 12A v- myeloblastosis viral oncogene MYB 2.16 2.64 1.54 1.38 homolog (avian) CCR8 chemokine (C-C motif) receptor 8 2.14 1.63 1.01 0.96 SRG survival-related gene 2.13 1.16 1.16 1.19 CYB5A cytochrome b5 type A (microsomal) 2.13 1.61 2.30 1.26 LOC730273 hypothetical protein LOC730273 2.12 0.82 0.95 1.08 sphingosine-1-phosphate SGPP2 2.08 2.12 1.59 1.66 phosphotase 2 membrane-associated ring finger MARCH3 2.07 2.28 1.56 1.59 (C3HC4) 3 nuclear factor of kappa light NFKBIZ polypeptide gene enhancer in B-cells 2.03 1.31 1.16 1.17 inhibitor, zeta CD69 CD69 molecule 2.00 1.33 1.47 1.25 DHRS3 dehydrogenase 1.99 2.75 2.70 1.47 GK glycerol kinase 1.97 2.60 1.15 1.62 similar to DNA segment, Chr 11, MGC71993 Brigham & Womens Genetics 0434 1.96 3.93 2.26 2.91 expressed ST8 alpha-N-acetyl-neuraminide ST8SIA1 1.92 2.54 1.14 2.20 alpha-2,8-sialyltransferase 1 TNIP1 TNFAIP3 interacting protein 1 1.90 2.07 1.69 1.36 CD82 CD82 molecule 1.84 2.25 2.10 1.54 STARD10 START domain containing 10 1.72 2.19 1.26 1.72 signal transducer and activator of STAT1 1.64 2.07 1.27 1.23 transcription 1, 91kDa GZMK granzyme K (granzyme 3; tryptase II) 1.63 2.14 1.30 1.37 ZMIZ2 zinc finger, MIZ-type containing 2 1.52 2.50 1.32 1.87

2

bromodomain and WD repeat BRWD2 1.51 3.50 1.00 2.11 domain containing 2 SLC41A2 solute carrier family 41, member 2 1.46 2.51 1.33 1.23 cholinergic receptor, nicotinic, alpha CHRNA6 1.44 9.73 0.94 1.34 6 CMT1A duplicated region transcript CDRT1 1.38 2.52 0.89 0.50 1 transmembrane protein induced by TMPIT 1.31 2.20 1.77 1.21 tumor necrosis factor alpha ectonucleoside ENTPD1 1.31 2.62 1.25 1.06 triphosphatediphosphohydrolase 1 NELL2 NEL-like 2 1.18 2.47 0.98 1.58 FCGBP Fc fragment of IgG binding protein 1.12 5.39 1.35 3.25 CD74 molecule, major CD74 histocompatibility complex, class II 1.11 2.49 1.45 2.49 invariant chain family with sequence similarity 112, FAM112A 1.10 2.25 1.16 1.37 member A LOC654346 similar to galectin 9 short isoform 1.07 3.06 0.76 1.46 hypothetical gene supported by LOC441233 0.96 2.05 1.02 1.64 AK128010 similar to Keratin, type I cytoskeletal LOC392516 0.74 2.18 1.01 1.49 18 (Cytokeratin-18)

Numbers represent fold change in gene expression induced by TNF treatment in either Treg or Tcon cells at the indicated time point; ranked in descending order according to Treg 2hrs.

3 Nagar et al. Suppl. Table II List of genes induced by TNF 2-fold in Tcon at either time point symbol name Tcon 2hrs Tcon 24hrs Treg 2hrs Treg 24hrs nuclear factor of kappa light NFKB2 polypeptide gene enhancer in B- 4.11 2.94 8.34 4.92 cells 2 nuclear factor of kappa light NFKBIA polypeptide gene enhancer in B- 3.62 2.78 4.85 4.28 cells inhibitor, alpha PELI1 pellino homolog 1 3.50 1.97 2.40 1.97 BIRC3 baculoviral IAP repeat-containing 3 3.06 2.76 2.70 2.00 nuclear factor of kappa light NFKB1 polypeptide gene enhancer in B- 2.73 1.46 2.35 1.33 cells 1 (p105) v-rel reticuloendotheliosis viral oncogene homolog B, nuclear RELB 2.73 2.47 4.56 4.74 factor of kappa light polypeptide gene enhancer in B-cells 3 DHRS3 dehydrogenase 2.70 1.47 1.99 2.75 IL27RA interleukin 27 receptor, alpha 2.61 1.47 2.33 1.98 SGK serum 2.36 2.22 3.78 3.54 interferon induced with helicase C IFIH1 2.35 1.54 2.41 1.85 domain 1 cytochrome b5 type A CYB5A 2.30 1.26 2.13 1.61 (microsomal) TTC7A tetratricopeptide repeat domain 7A 2.27 1.72 2.78 2.02 IER5 immediate early response 5 2.26 1.54 1.51 1.53 similar to DNA segment, Chr 11, MGC71993 Brigham & Womens Genetics 0434 2.26 2.91 1.96 3.93 expressed TRAF family member-associated TANK 2.25 1.76 2.68 2.07 NFKB activator CDGAP Cdc42 GTPase-.14 1.13 4.01 1.38 CD82 CD82 molecule 2.10 1.54 1.84 2.25 phosphatidylinositol glycan anchor PIGV 2.02 1.23 1.91 1.11 biosynthesis, class V inhibitor of kappa light polypeptide IKBKE gene enhancer in B-cells, kinase 1.91 2.12 3.00 2.27 epsilon TRAF-interacting protein with a TIFA 1.78 2.12 2.48 2.12 forkhead-associated domain tumor necrosis factor, alpha- TNFAIP3 1.77 2.22 3.42 3.07 induced protein 3 chromosome X open reading CXorf21 1.53 2.89 4.29 3.64 frame 21 CD74 molecule, major CD74 histocompatibility complex, class II 1.45 2.49 1.11 2.49 invariant chain FCGBP Fc fragment of IgG binding protein 1.35 3.25 1.12 5.39

1

RGS1 regulator of G-protein signalling 1 1.25 2.37 5.54 7.05 ST8 alpha-N-acetyl-neuraminide ST8SIA1 1.14 2.20 1.92 2.54 alpha-2,8-sialyltransferase 1 LOC647633 hypothetical protein LOC647633 1.13 2.06 1.57 1.06 LOC647633 hypothetical protein LOC647633 1.13 2.06 1.57 1.06 bromodomain and WD repeat BRWD2 1.00 2.11 1.51 3.50 domain containing 2 LOC730658 similar to ankyrin repeat domain 26 0.91 2.01 1.61 0.86

Numbers represent fold change in gene expression induced by TNF treatment in either Treg or Tcon cells at the indicated time point; ranked in descending order according to Tcon 2hrs.

2 Suppl. Table III Nagar et al. Gene set enrichment analysis for TNF-induced genes (2-fold) in Treg cells

Source Term A Fold Enrich. Count (%) P-valueB Gene symbols

BIOCARTA h_TNFR2 signaling pathway 36.0 5 (7.6%) 5.86E-06 TNFAIP3, TANK, NFKBIA, LTA, NFKB1

GO_BP I-B kinase / NF-B cascade 16.0 7 (10.6%) 4.07E-06 NFKBIA, IKBKE, TNFAIP3, TANK, IRAK2, STAT1, LTA

GO_BP immune system development 10.1 7 (10.6%) 5.69E-05 RELB, NFKB2, NFKBIA, LTA, IL27RA, CD74, TTC7A

PANTHER ankyrin repeat-containing 9.6 6 (9.1%) 4.38E-04 RELB, NFKB1, NFKB2, NFKBIZ, NFKBIA, NFKBIE negative regulation of GO_BP 6.8 5 (7.61%) 0.005691 NFKB1, TNFAIP3, BIRC3, FAS, CD74 apoptosis GO_BP cell activation 7.9 6 (9.7%) 8.52 E-04 RELB, NFKB2, LTA, CD274, IL27RA, CD74

GO_BP protein kinase cascade 5.6 7 (10.6%) 0.001336 STAT1, TNFAIP3, TANK, IRAK2, NFKBIA, LTA, IKBKE,

RELB, LTA, IL27RA, IFIH1, CD74, RGS1, FAS, TNFRSF9, GO_BP immune response 4.3 13 (19.7%) 2.26E-05 NFKB2, CD83, CCR8, CD274, IKBKE RELB, LTA, NFKBIA, IL27RA, IFIH1, CD74, RGS1, FAS, GO_BP immune system process 4.05 15 (22.7%) 8.03E-06 TNFRSF9, NFKB2, CD83, CCR8, CD274, IKBKE, TTC7A STAT1, TNFAIP3, FAS, BIRC3, TNFRSF9, NFKBIA, LTA, GO_BP apoptosis 4.0 10 (15.2%) 6.26E-04 NFKB1, IFIH1, CD74 TNIP1, RELB, LTA, NFKBIA, NFKB1, IL27RA, IFIH1, GO_BP response to stimulus 2.05 20 (30.3%) 0.001378 ENTPD1, CD74, CD69, RGS1, STAT1, FAS, IRAK2, NFKB2, TNFRSF9, CD83, CCR8, CD274, IKBKE TNF-regulated NF-B- NFKBIA, NFKBIE, NFKB1, NFKB2, RELB, TNFAIP3, Tian et al.C 59.8 10 (15.2%) 5.74E-17 dependent TNIP1(Naf1), TNFRSF9, SDC4, CD83 Gene set enrichment analysis was performed for annotated genes up-regulated 2-fold by TNF in Tregs at 2hrs and/or 24hrs. ATerms are in descending order according to their fold enrichment score. The analysis was performed using the DAVID Bioinformatics Resources and only includes source terms enriched >2-fold with a threshold of 5 genes count. BP-value (DAVID EASE score) of <0.01 was considered significant. CGene set enrichment analysis against a TNF-regulated NF-B-dependent 28-genes list reported by Tian et al. (19).