Identification of Novel CD4+ T Cell Subsets in the Target Tissue Of

Identification of Novel CD4+ T Cell Subsets in the Target Tissue of Sjögren's Syndrome and Their Differential Regulation by the Lymphotoxin/LIGHT Signaling Axis This information is current as of October 2, 2021. Scott Haskett, Jian Ding, Wei Zhang, Alice Thai, Patrick Cullen, Shanqin Xu, Britta Petersen, Galina Kuznetsov, Luke Jandreski, Stefan Hamann, Taylor L. Reynolds, Norm Allaire, Timothy S. Zheng and Michael Mingueneau

J Immunol 2016; 197:3806-3819; Prepublished online 7 Downloaded from October 2016; doi: 10.4049/jimmunol.1600407 http://www.jimmunol.org/content/197/10/3806 http://www.jimmunol.org/ Supplementary http://www.jimmunol.org/content/suppl/2016/10/06/jimmunol.160040 Material 7.DCSupplemental References This article cites 46 articles, 8 of which you can access for free at: http://www.jimmunol.org/content/197/10/3806.full#ref-list-1

Why The JI? Submit online. by guest on October 2, 2021

• 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 © 2016 by The American Association of Immunologists, Inc. All rights reserved. Print ISSN: 0022-1767 Online ISSN: 1550-6606. The Journal of Immunology

Identiﬁcation of Novel CD4+ T Cell Subsets in the Target Tissue of Sjo¨gren’s Syndrome and Their Differential Regulation by the Lymphotoxin/LIGHT Signaling Axis

Scott Haskett, Jian Ding,1 Wei Zhang,1 Alice Thai, Patrick Cullen, Shanqin Xu, Britta Petersen, Galina Kuznetsov, Luke Jandreski, Stefan Hamann, Taylor L. Reynolds, Norm Allaire, Timothy S. Zheng, and Michael Mingueneau

Despite being one of the most common rheumatologic diseases, there is still no disease-modifying drug for primary Sjo¨gren’s syndrome (pSS). Advancing our knowledge of the target tissue has been limited by the low dimensionality of histology techniques and the small size of human salivary gland biopsies. In this study, we took advantage of a molecularly validated mouse model of pSS to characterize tissue-infiltrating CD4+ T cells and their regulation by the lymphotoxin/LIGHT signaling axis. Novel cell Downloaded from subsets were identified by combining highly dimensional flow and mass cytometry with transcriptomic analyses. Pharmacologic modulation of the LTbR signaling pathway was achieved by treating mice with LTbR-Ig, a therapeutic intervention currently being tested in pSS patients (Baminercept trial NCT01552681). Using these approaches, we identified two novel CD4+ T cell subsets characterized by high levels of PD1: Prdm1+ effector regulatory T cells expressing immunoregulatory factors, such as Il10, Areg, Fgl2, and Itgb8, and Il21+ effector conventional T cells expressing a pathogenic transcriptional signature. Mirroring these observations in mice, large numbers of CD4+PD1+ T cells were detected in salivary glands from Sjo¨gren’s patients but not in http://www.jimmunol.org/ normal salivary glands or kidney biopsies from lupus nephritis patients. Unexpectedly, LTbR-Ig selectively halted the recruitment of PD12 naive, but not PD1+, effector T cells to the target tissue, leaving the cells with pathogenic potential unaffected. Altogether, this study revealed new cellular players in pSS pathogenesis, their transcriptional signatures, and differential dependency on the lymphotoxin/LIGHT signaling axis that help to interpret the negative results of the Baminercept trial and will guide future therapeutic interventions. The Journal of Immunology, 2016, 197: 3806–3819.

rimary Sjo¨gren’s syndrome (pSS) is characterized by a frequent extraglandular involvement and increased risk for B cell lymphocytic infiltration of the lacrimal and salivary lymphoma (reviewed in Refs. 1, 2). Th1 cells are thought to be the by guest on October 2, 2021 P glands, leading to xerophtalmia and xerostomia, as well as predominant CD4+ T cell subset infiltrating human biopsies (3, 4), although cytokine profiling identified secretion patterns, such as those characterized by IL-10 and TGF-b production, which are Immunology Research, Biogen, Cambridge, MA 02142 compatible with other effector or regulatory T cell (Treg) subsets 1J.D. and W.Z. contributed equally to this work. (4–6). Accordingly, markers of Th17 cells (7, 8), Tregs (9–11), ORCIDs: 0000-0001-9853-5289 (S. Haskett); 0000-0002-5689-1784 (L.J.); 0000- and T follicular helper (Tfh) cells (6, 12) were detected in pSS 0002-6941-0900 (T.L.R.). biopsies, but the relative importance of each of these subsets in Received for publication March 9, 2016. Accepted for publication September 11, pSS pathogenesis and their relationships remain unknown. 2016. Blocking the LTbR signaling pathway showed clear efficacy in S. Haskett, J.D., and M.M. designed, executed, and analyzed all experiments; W.Z. several preclinical models of pSS (13–15). LTbR signaling is established and optimized the hydrodynamic injection technique; immunohistochemistry method development was directed by G.K. and T.L.R.; immunohistochemistry important for secondary and ectopic/tertiary lymphoid tissue or- experiments were performed by S.X. and B.P., and immunohistochemistry quantifica- ganization (reviewed in Ref. 16). Specifically, LTbR engagement tion methods were designed and executed by L.J. and S. Hamann; histological analyses by LIGHT or LTab regulates the expression of peripheral lymph and interpretation were conducted by T.L.R.; A.T., P.C., and N.A. performed all transcriptomic experiments; S. Haskett and T.L.R. contributed to figure and manuscript node addressin (PNAd) and mucosal vascular addressin cell ad- preparation; and M.M. designed the research study and wrote the manuscript. All hesion molecule 1 on high endothelial venules, which are required authors participated in data and project discussion and reviewed and approved the final version of the manuscript. for efficient immune cell recruitment into tissues (17). Novel technological approaches, such as digital immunohisto- The gene expression data presented in this article have been submitted to the National Center for Biotechnology Information Gene Expression Omnibus under accession chemistry techniques, single-cell repertoire sequencing, and highly number GSE81621. multiparametric mass cytometry, have recently improved our Address correspondence and reprint requests to Dr. Michael Mingueneau, Immunol- knowledge of the target tissue in pSS (18–20). Despite these ad- ogy Research, Biogen, 115 Broadway, Cambridge, MA 02142. E-mail address: vances, our knowledge of the cellular composition of the immune [email protected] infiltrate in pSS patients is still incomplete, partly because the The online version of this article contains supplemental material. number of cells that can be recovered from human lip biopsies Abbreviations used in this article: B6, C57BL/6; CyTOF, cytometry by time-of-flight; eTreg, effector Treg; FC, fold change; FDR, false discovery rate; IEG, intermediate- precludes further dissection of cellular subsets and phenotypes early gene; LGT, lacrimal gland tissue; PNAd, peripheral lymph node addressin; pSS, contributing to each cell lineage and the mechanisms by which primary Sjo¨gren’s syndrome; SLE, systemic lupus erythematosus; Tconv, conventional they infiltrate the target tissue. T cell; Tfh, T follicular helper; Treg, regulatory T cell; VAT, visceral adipose tissue. To circumvent these technical limitations, we took advantage of Copyright Ó 2016 by The American Association of Immunologists, Inc. 0022-1767/16/$30.00 the NOD mouse, which is one of the best-studied and validated www.jimmunol.org/cgi/doi/10.4049/jimmunol.1600407 The Journal of Immunology 3807 models of pSS (reviewed in Ref. 21). Using a combination of mass nonetheless sufficient to characterize enzyme-sensitive surface marker cytometry, highly multiparametric flow cytometry, and tran- expression. scriptomics, we characterized the kinetics of immune infiltration Mass and flow cytometry in lacrimal gland tissue (LGT) and further dissected the subsets of LGT-infiltrating CD4+ T cells in the target tissue. Interestingly, Single-cell suspensions were stained with 30 metal-conjugated Abs (Table I) and prepared for cytometry by time-of-flight (CyTOF) analysis, LTbR signaling blockade using LTbR-Ig revealed that not all as described (23). For flow cytometry, cells were preincubated with anti- CD4+ T cell subsets are affected equally by this treatment, an CD16/32 Ab (2.4G2; BD Biosciences) and stained with fluorochrome- observation that helps us to understand the negative results of the conjugated Abs (described below). All samples were analyzed using an primary end point in the Baminercept clinical trial and will guide LSR II instrument (BD Biosciences). the design of future therapeutic strategies. Fluorochrome-conjugated Abs The following mAbs specific for murine Ags were purchased from BD Materials and Methods Biosciences, BioLegend, and eBioscience: BUV395-conjugated CD4 (SK3), Mice BV785-conjugated PD1 (29F.1A12), V510-conjugated CD45.2 (104), FITC- conjugated CD19 (1D3), PE-conjugated CD69 PE (H1.2F3), PE/Dazzle Mice were purchased from the Jackson Laboratory and maintained under 594–conjugated Helios (ICFC), FITC-conjugated CD200 (OX-90), PerCP pathogen-free conditions. Only male mice were used in this study. NOD Cy5.5-conjugated CD73 (TY/11.8), PE-Cy7–conjugated ICOS (C398.4), mice were screened for diabetes prior to use. All experiments were approved allophycocyanin-conjugated FoxP3 (FJK-16s), Alexa Fluor 700–conjugated by the Biogen Institutional and Animal Care and Use Committee (IACUC CD45.1 (A20), PE-conjugated CD30L (RM153), PE-conjugated LAG3 0466-2013). (C9B7W), BV421-conjugated Ki67 (16A8), PE-conjugated ICAM1 (HCD54), PE-conjugated CD119 (GIR-208), PE-conjugated GITR (DTA-1), Patient biopsies PE-conjugated CD106 (429), PE-conjugated Ly108 (330-AJ), PE-conjugated Downloaded from DR3 (4C12), PE-conjugated CD62L (MEL-14), PE-conjugated CXCR5 Blocks of formalin-fixed tissue biopsies were acquired from Folio Bio- (L138D7), PE-conjugated CD103 (2E7), PE-conjugated BTLA (8F4), sciences (Powell, OH). Salivary tissues included submandibular glands PE-conjugated Eos (ESB7C2), PE-conjugated Aiolos (8B2), PE-conjugated (n = 2), parotid glands (n = 1), and labial glands (n = 3) from Sjo¨gren’s T-bet (4B10), PE-conjugated BLIMP-1 (5E7), and PE-conjugated Bcl-6 patients. Diagnoses were consistent with primary or secondary Sjo¨gren’s (7D1). For intracellular Ags, cells were fixed and permeabilized using an syndrome. Tonsils and one major salivary gland from normal donor sub- Anti-Mouse/Rat Foxp3 Staining Set PE (eBioscience). jects were also analyzed in parallel as control tissues. Renal biopsies from patients with systemic lupus erythematosus (SLE) were diagnosed as class Cell sorting, RNA isolation, and gene-expression microarrays http://www.jimmunol.org/ III (n = 1), class IV (n = 1), or class IV/V (n = 2). Cell subsets were sorted directly into lysis buffer on a FACSAria cytometer DNA vectors (BD Biosciences). Lysates were stored at 280˚C until processing. Total RNA was isolated using an RNAqueous Total RNA Isolation Kit (Ambion). The LTbR-Ig plasmid was generated by cloning the PCR-amplified 1.4-kb Gene-expression microarrays and analysis were performed as described (24). sequence encoding mouse LTbR–mouse IgG1 [murine receptor fused to Microarray data were submitted to the National Center for Biotechnology mouse IgG1 described previously (15)] into the pLIVE vector (Mirus). Information Gene Expression Omnibus under accession number GSE81621 Plasmids were purified using Endotoxin-free Maxi-prep Kits (QIAGEN). (http://www.ncbi.nlm.nih.gov/geo/). Control mice were injected with the unmodified pLIVE vector. Histology and immunohistochemistry

LTbR-Ig treatment and serum detection of the fusion protein by guest on October 2, 2021 Histological examination was performed on H&E-stained sections. Whole All injections were done in 16-wk-old mice. Hydrodynamic injections of excised mouse lacrimal glands were fixed in formalin (10%) and embed- DNA vectors were performed as described (22). Mouse LTbR–mouse IgG1 ded in paraffin prior to sectioning at 5 mm. Human biopsy specimens and fusion protein was prepared at Biogen and injected i.p. Serum levels of the specimens of normal control tissues were sectioned at 3 mm. Anti-mouse fusion protein were detected using a previously described ELISA (15). CD3 (1:100, clone SP7; Thermo Scientific), anti-mouse B220 (0.039 mg/ml, clone RA3-6B2; BD Pharmingen), anti-human CD4 (1.0 mg/ml, clone Preparation of single-cell suspensions EPR6855; Abcam), anti-human FOXP3 (3.1 mg/ml, clone D2W8E; Cell Lacrimal glands were minced using scissors to dissociate the tissue and Signaling Technology), and anti-human PD1 (2.84 mg/ml, clone NAT105; enzymatically digested at 37˚C under rotating agitation (100 rpm) in Ventana) immunohistochemistry and immunofluorescence were performed DMEM containing 232 U/ml collagenase II (Worthington) and 8 U/ml on the Ventana platform (Roche). Whole-slide images were acquired on a DNase I (Sigma). After two washes in calcium- and magnesium-free Pannoramic 250 digital scanner (PerkinElmer). To quantify immunoreac- PBS containing 1 mM EDTA, cellular aggregates were resuspended in tivity per area and number of positive cells per area in all bright-field 0.5 ml of TrypLE Express Enzyme (Life Technologies) and incubated for analyses, customized algorithms were developed using Visiopharm image- 2 min at 37˚C. Enzyme inactivation was achieved by dilution with 4 ml of analysis software. Inflammatory infiltrates on H&E images were detected DMEM and immediately followed by gentle up-and-down pipetting with a with a texture-based algorithm using Fourier Transform. Fourier Transform P1000 to facilitate cell dissociation. The resulting cell suspension was inertia values were calculated based on the RGB-Green component of the washed twice with medium supplemented with 0.8 U/ml DNase I before original 24-bit color image, and a threshold was applied to distinguish areas preparation for flow or mass cytometry. Spleens and lymph nodes were of inflammation from nonaffected tissue. To measure CD4 and PD1 immu- disaggregated into single-cell suspensions, and RBCs were removed using noreactivity and colocalized CD4/PD1 immunofluorescent signal in human RBC lysis buffer (eBioscience). Because lacrimal gland single-cell sus- tissues, images were acquired with a Vectra 3 imaging system and analyzed pensions cannot be accurately counted (because of the presence of epi- with inForm 2.2 software (both from Perkin Elmer). A threshold above thelial cells and debris), glands from NOD mice were systematically mixed background was set for CD4 and PD1 to identify areas of immunoreactivity with glands from age-matched C57BL/6 (B6) control mice in all immuno- and quantify regions of colocalized expression per total glandular area. phenotyping studies. For LTbR-Ig studies, 20,000 B6 splenocytes were spiked in all single-cell suspensions from NOD lacrimal glands. In both cases, cells from NOD and control mice were identified postacquisition Results using CD45.1 and CD45.2 congenic markers, respectively. In immuno- Dynamics of immune infiltration in LGT from NOD mice phenotyping studies, the NOD/B6 ratio was used to express the number of tissue-infiltrating NOD cells relative to the number of tissue-resident im- Integrated analysis of publicly available mouse and human pSS mune cells in age-matched control B6 mice. For LTbR-Ig studies, the datasets showed that 58% of the human pSS transcriptional gene NOD/B6 ratio was used to calculate absolute numbers of tissue-infiltrating signature was upregulated in LGT from 16-wk-old male NOD mice cells in LTbR-Ig–treated versus control-treated NOD mice. Note that (Supplemental Fig. 1, Supplemental Table IA, IB), indicating that CXCR5 detection is sensitive to the enzymatic digestion required for the preparation of LGT single-cell suspensions. Thus, this staining was done the NOD mouse model was a relevant model to study the human on mechanically dissociated glands. This protocol results in low (15–20% disease from a molecular standpoint. In this model, immune infil- instead of 50–60%) frequencies of PD1highICOShigh cells, which are tration was not detectable in 4-wk-old animals, but small periductal 3808 TISSUE-INFILTRATING T CELL SUBSETS IN SJO¨ GREN’S andperivascularfociofinfiltrationweredetectedataround8wk tissue-resident immune cells were detectable in LGT from 8-wk- of age and gradually became more extensive between 16 and 24 wk old B6 control mice (Fig. 1B). Most of them were myeloid cells of age, with 100% penetrance (Fig. 1A). Mass cytometric anal- expressing CD11b and F4/80, but small numbers of NK and ab T ysis using a 30-Ab panel (Table I) revealed that small numbers of cells were also observed. Although there was a slight age-associated Downloaded from http://www.jimmunol.org/ by guest on October 2, 2021

FIGURE 1. Dynamics of immune cell inﬁltration in the NOD mouse model of Sjo¨gren’s syndrome. (A) H&E-stained lacrimal gland sections from 4–24-wk-old NOD mice. Inset, higher magniﬁcation. (B) viSNE clustering maps of CD45+ events detected by CyTOF in representative lacrimal glands from indicated mice. Each dot represents an individual cell, and its color shows the level of expression of the indicated marker. (C) Numbers of CD45.1+ immune cells detected in lacrimal glands from 4–32-wk-old NOD animals expressed as a ratio relative to the number of CD45.2+ immune cells in age-matched B6 mice. (D) Pie charts showing the relative abundance of each immune subset in indicated animals. Results are the average 6 SD for n = 6 animals for each genotype and time point. The Journal of Immunology 3809

Table I. CyTOF immunophenotyping Ab panel CD4+ T cells, with both effector genes (in red) and immunoregulatory genes (in blue). Altogether, these results indicated Ag Ab Clone Symbol Mass (Da) that LGT-infiltrating CD4+ T cells contained a mix of multiple CD8 53-6.7 Cd 114 subsets with effector and regulatory functions. Ly-6G RB6-8C5 Pr 141 high high CD11c N418 Nd 142 A unique PD1 ICOS T cell phenotype in LGT from NOD CD115 AFS98 Nd 144 mice CD4 RM4-5 Nd 145 We used some of the top upregulated markers displayed in Fig. 2C F4/80 BM8 Nd 146 + CD45.2 104 Sm 147 to identify the subsets contributing to the LGT CD4 T cell pool. CD19 6D5 Sm 149 Among them, PD1 and ICOS were particularly useful because IgD 11-26c.2a Nd 150 they identified a unique population of PD1highICOShighCD4+ IgM RMM-1 Eu 151 T cells in NOD LGT (Fig. 3A). These cells were already detected CD3ε 145-2C11 Sm 152 CD45.1 A20 Eu 153 in glands from 6-wk-old NOD mice and were also present in CD11b M1/70 Sm 154 cervical lymph nodes, albeit at low frequencies. In contrast, Thy1.2 30-H12 Gd 156 PD1highICOShigh cells were barely detectable in control B6 mice CD93 AA4.1 Gd 158 (Fig. 3A), suggesting that they might be involved in disease CD23 B3B4 Tb 159 CD5 53-7.3 Gd 160 pathogenesis. Ly-6C HK1.4 Dy 162 To further characterize this population, we used CD73 (Nt5e) CD25 PC61 Dy 164 and CD200 markers, also identified in Fig. 2C. These markers CD317 927 Ho 165 separated the PD1highICOShigh population into two clearly distinct Downloaded from CD326 G8.8 Er 166 subsets (Fig. 3B). CD73lowCD200low cells contained 90–95% of NKp46 29A1.4 Er 167 + + high CD21 7G6 Er 168 thymic-derived FOXP3 HELIOS Tregs. In contrast, CD73 high + TCRb H57-597 Tm 169 CD200 cells contained only a minor fraction of FOXP3 CD49b HMa2 Er 170 T cells. These two PD1highICOShigh subsets shared high levels of CD44 IM7 Yb 171 expression of CD69, LAG3, and Ki67, whereas CD30L was only

CD86 GL1 Yb 172 http://www.jimmunol.org/ expressed on the CD73highCD200high T cell subset. In contrast, IA-IE M5/144 Yb 174 2 2 CD38 90 Lu 175 PD1 ICOS T cells were negative for all of these markers B220 RA3-6B2 Yb 176 (Fig. 3B). Kinetic analysis of these cellular phenotypes in B6 and Cisplatin Pt 195 NOD LGT (Fig. 3C, 3D) showed that small numbers of tissue- List of Ags, Ab clones, and metals used to immunophenotype cells infiltrating resident Tregs were present in LGT in steady-state, whereas up to inflamed lacrimal glands in the NOD mouse model of pSS. 50% of Tregs were detected in NOD LGT. The proportion of PD1high ICOShigh T cells mirrored the frequencies of FOXP3+ cells and peaked between 16 and 24 wk of age and then declined (Fig. 3D). + Taken together, these results indicated that three main subsets increase in the number of glandular T cells in B6 mice, CD19 by guest on October 2, 2021 + B cells were never detected. In contrast, glands from NOD mice were contributing to the LGT-infiltrating CD4 T cell population. high high low showed massive recruitment of immune cells (Fig. 1B, 1C). Large Two PD1 ICOS effector subsets displayed a CD73 low high high 2 2 numbers of CD4+ T cells were detected as early as 8 wk of age, CD200 or CD73 CD200 phenotype, whereas a PD1 ICOS along with B cells, which became the dominant cell subset at later subset defined a third subset that likely corresponded to a pop- time points (Fig. 1B, 1D). ulation of naive conventional T cells (Tconvs). Transcriptomic analysis reveals cellular heterogeneity within Prdm1+ effector Tregs, Il21+ Th1 cells, and naive T cells are glandular CD4+ T cells the three main subsets in inflamed LGT To better define the phenotype of infiltrating cells, we next We next purified LGT-infiltrating T cell subsets identified in Fig. 3B isolated LGT CD4+ T cells, as well as their splenic counter- and performed gene-expression analyses to characterize their cel- parts, from 16-wk-old NOD mice. To account for any tran- lular and molecular identity. The transcriptomes from the two scriptional changes caused by LGT processing that would subsets of PD1highICOShigh cells (CD73lowCD200low and CD73high affect all cells, irrespective of their cellular identity, we also CD200high) were clearly distinguishable from the PD12ICOS2 isolated B cells from the same tissue. As expected, a set of transcriptome, as evidenced by the large number of genes located intermediate-early genes (IEGs), such as Fos, Nr4a2,andJun, on or close to the diagonal of the FC plot in Fig. 4A (Supplemental was similarly upregulated in LGT B and T cells compared with Table IC). These genes were similarly expressed in the two subsets splenic populations (Fig. 2A) and likely resulted from the ad- of PD1highICOShigh cells and included Tbx21, Ifng, Cd44, Cxcr3, ditional isolation steps required to purify glandular cell subsets. and Ctla4 (Supplemental Fig. 2). A number of genes was found More interestingly, there was a defined set of genes uniquely farther away from the diagonal in Fig. 4A (areas colored in blue and upregulated in LGT CD4+ T cells (dashed box in Fig. 2A). green), indicating that the two subsets of PD1highICOShigh cells Filtering this set of genes on cytokines, chemokines, and their were themselves clearly distinct from each other at the transcrip- receptors (Fig. 2B) revealed a complex phenotype comprising tional level. Filtering genes on cytokines, chemokines, and their proinflammatory and anti-inflammatory factors. This observa- receptors (Fig. 4B, Supplemental Table ID) revealed that the tion was confirmed by the analysis of the top upregulated genes CD73lowCD200low subset uniquely expressed Ifngr1, Il2ra,and in LGT CD4+ T cells (Fig. 2C) defined by the following cri- Cxcr6 and preferentially expressed Il10 and Areg, whereas the teria: fold change (FC) in LGT versus spleen CD4+ T cells . 3 CD73highCD200high subset uniquely expressed Lta, Spp1, Il2, with p value (false discovery rate [FDR]) , 0.05; FC in LGT Tnfsf13b,andXcl1 and preferentially expressed Tnf, Il21,andIfng. versus spleen B cells , 2 (to exclude IEGs); and mean ex- None of these genes were expressed in PD12ICOS2 cells, which pression . 6 in LGT-resident CD4+ T cells. This set of genes expressed high levels of Ccr7, Il7r, Sell (CD62L), and S1pr1,sug- clearly revealed the heterogeneous nature of LGT-resident gestive of a recent migration into the target tissue. This conclusion 3810 TISSUE-INFILTRATING T CELL SUBSETS IN SJO¨ GREN’S Downloaded from http://www.jimmunol.org/ by guest on October 2, 2021

FIGURE 2. Transcriptomic analysis reveals cellular heterogeneity within glandular CD4+ T cells. Splenic and LGT-infiltrating CD4+ Tcellsand CD19+ B cells were sorted by FACS from 16-wk-old NOD mice, and RNA was processed for microarray gene-expression analysis. (A)FC-FCplot comparing the transcriptomes of the indicated cell subsets. Data were filtered on genes with p value (FDR) , 0.05 for at least one of the two contrasts andmeanexpression. 6inLGT-residentCD4+ T cells. Blue lines, FC = 3. Note that genes on the diagonal correspond to intermediate early genes likely resulting from LGT processing. Genes within the dashed box are displayed in (C). (B) Heat map of the genes in (A) that encode proteins with cytokine or chemokine activities and their receptors (based on gene ontology annotations) and that are upregulated $2-fold in LGT versus spleen CD4+ T cells. Samples are hierarchically clustered according to gene expression. (C) MA plot showing mean gene expression in LGT-infiltrating CD4+ Tcells (x-axis) versus the ratio of expression in LGT versus spleen populations (y-axis). Data were filtered on the genes found within the dashed box in (A) corresponding to the top differentially expressed genes in LGT CD4+ T cells (i.e., genes satisfying the following expression criteria: FC in LGT versus spleen CD4+ Tcells. 3withp value [FDR] , 0.05; FC in LGT versus spleen B cells , 2 [to exclude IEGs]; and mean expression . 6inLGT-resident CD4+ T cells). Selected genes of interest are color-coded. Data show integrated results from n = 4 biological replicates for LGT-infiltrating cell types and n = 3 replicates for splenic counterparts. was further supported by their transcriptional signature (Supplemental Il21+ Th1 phenotype, but, surprisingly, did not show evidence of Fig. 2) and transcription factor profile (Supplemental Fig. 3), which expression of the canonical Tfh signature genes Cxcr5 or Bcl6 at were largely shared with splenic T cells. the mRNA and protein levels (Fig. 4D). The CD73lowCD200low The transcriptional signatures (Fig. 4C, Supplemental Table IE) subset had an effector Treg (eTreg) phenotype, as evidenced by and transcription factor profiles (Supplemental Fig. 3) of PD1high high expression levels of Il10 and Prdm1 (BLIMP1) (Fig. 4C). ICOShigh T cell subsets helped to further refine the transcriptional Expression of Tbx21 (Supplemental Fig. 3) further indicated that identity of these subsets. The CD73highCD200high subset had an this subset of eTregs was Th1 polarized. This regulatory subset also The Journal of Immunology 3811 Downloaded from http://www.jimmunol.org/

FIGURE 3. PD1 and ICOS identify a unique T cell phenotype in LGT from NOD mice. (A) Expression of PD1 and ICOS on CD4+ T cells in LGT, cervical lymph nodes (Cerv. LN), mesenteric lymph nodes (Mes. LN), and spleen from the indicated mice. Numbers correspond to the frequency of PD1high ICOShigh cells. (B) CD4+ T cell subsets are deﬁned as described in the key and dot plots (left panels). Line graphs showing expression of the indicated markers on CD4+ T cell subsets in lacrimal glands from 20-wk-old NOD mice (right panels). (C) Expression of HELIOS and FOXP3 on CD4+ T cells in lacrimal glands from the indicated mice. Numbers indicate the percentage of cells in the boxes. (D) Frequency of CD4+ T cell subsets in NOD mice at the by guest on October 2, 2021 indicated ages. Representative and integrated results from four independent experiments with n = 3–6 animals for each age and genotype.

expressed all three IKAROS family member genes at the transcript injection was a very efficient method to deliver genes in vivo. (Fig. 4C) and protein (Fig. 4D) levels, as well as the immunoreg- Accordingly, a single hydrodynamic injection of the LTbR-Ig– ulatory secreted protein Fgl2. In addition to IKAROS transcription encoding plasmid induced similar or higher levels of expression factors, we confirmed the expression of many other genes identified in serum than did weekly injection of the fusion protein, and as differential by microarray analyses (Fig. 4D). expression was sustained over the 8-wk experimental period The transcriptomes from effector and regulatory T cell subsets (Fig. 5A). Using this experimental approach, we validated the identified in NOD LGT were not unique to this tissue or disease previously described efficacy of LTbR-Ig treatment when ad- model. In fact, the Il21+ Th1 effector cell subset shared striking ministered prophylactically in preclinical models of pSS (14, 15) transcriptional similarities with a Slamf6-dependent pathogenic (data not shown). We next performed therapeutic administration of CD4+ T cell population recently characterized in the B6.Sle1b and LTbR-Ig in 16-wk-old mice (i.e., in mice with well-established BWF1 lupus mouse models (25, 26) (Fig. 4E). Similarly, LGT immune infiltration) and observed that this regimen was also able eTregs showed a clear expression of the canonical Treg signature to efficiently decrease immune infiltration, as evidenced by the (27), as well as shared the expression of many genes specific to 2.8-fold reduction in the area of the immune infiltrate in H&E- other tissue-resident Tregs, such as visceral adipose tissue (VAT) stained sections, the 1.9-fold reduction in the area immunoreactive (28) and muscle Tregs (29) (Fig. 4F). In conclusion, this tran- for CD3 or B220, and the 3–4-fold reduced numbers of T and scriptional analysis revealed the cellular identity of the three novel B cells quantified by flow cytometry (Fig. 5B, 5C). Importantly, subsets of LGT-infiltrating CD4+ T cells as Prdm1+ eTregs, Il21+ however, the glandular epithelium still showed foci of immune Th1 cells, and naive T cells. infiltration, even though they were smaller and more diffuse than in control animals (Fig. 5B, insets). Altogether, these re- b LT R-Ig selectively reduces glandular recruitment of naive, sults indicated that, although LTbR-Ig efficiently decreased but not effector, T cell subsets immune infiltration in animals with established disease, it only We next analyzed the effect of LTbR-Ig treatment on the re- partially blocked the immune-infiltration process, likely indi- cruitment and accumulation of the novel CD4+ T cell subsets that cating that not all immune subsets were similarly sensitive to we identified in inflamed LGT. Using hydrodynamic injection of the treatment. an LTbR-Ig–encoding plasmid, a method used for systemic gene Accordingly, we observed that CD4+ T cell subsets were not delivery (reviewed in Ref. 22), we first confirmed that hydrodynamic equally affected by LTbR-Ig treatment. The treatment dramatically 3812 TISSUE-INFILTRATING T CELL SUBSETS IN SJO¨ GREN’S Downloaded from http://www.jimmunol.org/ by guest on October 2, 2021

FIGURE 4. Prdm1+ eTregs, Il21+ Th1 cells, and naive T cells are the three dominant T cell subsets in inflamed LGT. The LGT-infiltrating CD4+ T cell subsets described in Fig. 3B were sorted by FACS from 16-wk-old NOD mice, and RNA was processed for microarray gene-expression analysis. (A) FC-FC plot comparing PD1highICOShigh T cell subsets (CD73lowCD200low on the y-axis and CD73highCD200high on the x-axis) with the PD12ICOS2CD4+ T cell subset. Data were filtered on genes with p value (FDR) , 0.05 for at least one of the two contrasts and mean expression . 6 in at least one LGT-infiltrating T cell subset (see complete list of genes in Supplemental Table IC). Blue lines, FC = 2. Areas colored in blue and green include the genes that are ex- clusively or preferentially expressed in CD73lowCD200low and CD73highCD200high subsets, respectively. (B) Expression heat map of the genes displayed in (A) that encode proteins with cytokine or chemokine activities and their receptors and that show $2-fold significant differential expression for at least one of the two contrasts in (A). Samples are color-coded and hierarchically clustered according to gene expression (see complete list of genes in Supplemental Table ID). (C) Expression heat maps displaying transcriptional signatures for PD1highICOShigh T cell subsets: CD73lowCD200low (left panel) and CD73high CD200high (right panel). Signatures were generated by filtering genes displayed in (A) on genes with mean expression . 6 and (Figure legend continues) The Journal of Immunology 3813 decreased the PD12ICOS2 subset, whereas the PD1highICOShigh and T cells, there were no or very rare CD4+PD1+ cells detected in the PD1lowICOSlow subsets were much less affected (Fig. 6A, 6B). Im- kidney biopsies from the four SLE patients analyzed, indicating portantly, separating PD1highICOShigh (effector) and PD12ICOS2 that this CD4+PD1+ T cell subset could be a cellular feature (naive) populations according to FOXP3 expression revealed that specific to the immune infiltrate in Sjo¨gren’s syndrome. Diffusely the impact of the treatment largely correlated with the naive infiltrating CD4+ T cells in normal salivary glands did not express (PD12ICOS2) versus effector (PD1highICOShigh) status of T cells PD1 (Fig. 7C), confirming that this T cell phenotype was indeed and, to a much lesser extent, with their regulatory (FOXP3+)or associated with the disease state. As expected, large numbers conventional (FOXP32) phenotype (Fig. 6C). Tregs and Tconvs of PD1+CD4+ T cells were also detected in the inflamed tonsils with an effector phenotype were only slightly affected by the from healthy subjects, validating the specificity of the staining treatment (1.3- and 1.9-fold, respectively), whereas Tregs and (Fig. 7C). Automated quantification of the percentage of CD4 area Tconvs with a naive phenotype were strongly affected (5.8- and positive for PD1 confirmed the colocalization of CD4 and PD1 20-fold, respectively). Because only a small fraction of Tregs (9 6 markers in a large subset of CD4+ T cells in Sjo¨gren’s patients 6%) in the target tissue had a naive phenotype, the overall Treg (Fig. 7D). In contrast, this colocalization metric was, on average, population was decreased to a lesser extent (1.6-fold, Fig. 6B) .20 times lower in lupus nephritis samples. FOXP3 staining than the Tconv population (5-fold, Fig. 6B), which contained a revealed large numbers of Tregs in minor and major salivary large fraction of cells with a naive phenotype (54 6 18%). glands from Sjo¨gren’s patients (Fig. 8A, 8B). However, the ma- Accordingly, the overall Treg/Tconv ratio increased from 0.7 6 jority of PD1+ cells were FOXP32, suggesting that most tissue- 0.3 to 1.5 6 0.4 (p value = 1 3 1025), which resulted almost resident PD1+ cells corresponded to an effector, and potentially entirely from the depletion of naive Tconvs (20-fold decrease, pathogenic, T cell subset in humans. Interestingly, a minor pop- p , 0.0001, Fig. 6C). In contrast, the Treg/effector Tconv ratio ulation of PD1+ cells immunoreactive for FOXP3 was clearly Downloaded from was not significantly changed in treated versus untreated ani- detectable in specimens from Sjo¨gren’s patients (Fig. 8A, 8B), but mals (3.0 6 1.1 versus 2.4 6 0.7, respectively; p =0.12). not in tonsils (Fig. 8A, bottom panels), normal salivary glands In conclusion, the primary effect of LTbR signaling blockade (Fig. 8C), or infiltrated kidneys from lupus patients (Fig. 8D). was to dramatically decrease the number of naive Tconvs infil- Altogether, these results indicated that the effector and regulatory trating the target tissue with very little effect on Tregs, but, im- CD4+PD1+ T cell subsets identified in the target tissue from NOD portantly, with also very little effect on effector Tconvs, which are mice were also present in patients with Sjo¨gren’s syndrome, albeit http://www.jimmunol.org/ the cells displaying the pathogenic transcriptional signature in the in varying proportions. target tissue (Fig. 4). Thus, the lack of effect of LTbR signaling blockade on the T cell subset with pathogenic potential could Discussion potentially contribute to the modest clinical benefits seen in the Advancing our knowledge of the pathogenic cell subsets infiltrating recently disclosed Baminercept clinical trial (30). exocrine glands in pSS patients has been limited by the low di-

+ + mensionality of traditional histology techniques and the limited High numbers of PD1 CD4 T cells are detected in major and number of cells available for single-cell analyses in lip biopsy minor salivary glands from Sjo¨gren’s patients samples. In this study, we took advantage of the NOD model of by guest on October 2, 2021 To validate the significance of these findings for human disease, we pSS, which displays extensive immune infiltration in LGT that is extended this analysis to human specimens. Immunofluorescent accompanied by the upregulation of 58% of the human pSS analysis of PD1 and CD4 expression in lip biopsies and major transcriptional signature. In inflamed LGT from this model, we salivary glands from Sjo¨gren’s patients identified PD1+ cells in all identified a population of Tregs as the most abundant CD4+ T cell specimens analyzed (Fig. 7A). PD1+ cells were abundant and subset, accounting for up to 50% of the entire population at the easily identifiable in all samples, with the exception of one sample peak of immune infiltration. These cells are characterized by a that contained only a few PD1+ cells, despite the presence of PD1highICOShighCD73lowCD200lowFOXP3+ phenotype, as well as numerous CD4+ T cells. The vast majority of PD1+ cells were also expression of Ikzf2 (HELIOS), indicating their thymic origin, CD4+, indicating that that these cells belonged to the CD4+ expression of Prdm1 (BLIMP1) and Il10, which are hallmarks of T lymphocyte lineage. Importantly, CD4+PD1+ cells were de- eTregs (reviewed in Ref. 31), and expression of Tbx21, which tected in major salivary glands typically biopsied in patients with a indicates that this population is geared to control Th1 responses. suspicion of lymphoma and more severe forms of the disease, as Even though it is detected at low frequency in steady state, this well as in lip biopsy samples, which are routinely performed in population of LGT-resident eTregs accumulates upon inflamma- clinical practice in patients with active, but potentially more tion in pSS. LGT-resident Tregs express 80% of the typical Treg muted forms, of the disease. To determine whether the presence of signature but also share similarities with other tissue-resident this T cell subset in the target tissue from Sjo¨gren’s patients was a Tregs (reviewed in Ref. 32, 33), such as VAT and muscle Tregs. commonality among rheumatologic diseases, we performed the The presence of Tregs was reported in salivary glands from pSS same analyses on kidney needle biopsies from lupus nephritis patients (9–11). Available phenotyping data from patients point to patients (Fig. 7B). Despite the large numbers of infiltrated CD4+ similarities in phenotype with LGT-resident NOD eTregs, such as

$3-fold significant upregulation in the subset of interest relative to the two other LGT CD4+ T cell subsets. Samples are color-coded and hierarchically- clustered according to gene expression (see complete list of genes in Supplemental Table IE). (D) Line graphs showing protein expression of indicated markers in T cell subsets from 16-wk-old NOD mice. Corresponding microarray mean expression values (log2) are indicated below each plot for LGT subsets. Note that CXCR5 and BCL6 were the only proteins yielding a negative staining. Positive controls for these stainings were obtained from LGT-infiltrating B cells and germinal center B cells from Peyer’s patches, respectively (data not shown). (E) Bar graphs showing log2(mean expression) for the genes characterizing the pathogenic CD4+ T cell population recently characterized in the B6.Sle1b and BWF1 lupus models. (F) Volcano plots comparing the LGT-eTreg (CD73low CD200low) and PD12ICOS2 subsets. Upregulated (red) and downregulated (blue) genes from the indicated transcriptional signatures are highlighted with the corresponding gene numbers. The total number of genes in each signature is also indicated in parentheses after the signature name (upregulated-downregulated). Integrated results from n = 3 biological replicates for each LGT-infiltrating cell subset and n = 2 replicates for splenic T cells. 3814 TISSUE-INFILTRATING T CELL SUBSETS IN SJO¨ GREN’S Downloaded from http://www.jimmunol.org/ by guest on October 2, 2021

FIGURE 5. LTbR-Ig efficiently reduces established immune infiltration. (A)LTbR-Ig protein concentration in the serum from NOD mice at the indicated time points following a single hydrodynamic injection (HDI) of LTbR-Ig–encoding or control plasmids or a weekly injection of LTbR-Ig fusion protein. (B) Representative photomicrographs of H&E-stained and CD3/B220-immunoreacted lacrimal gland sections from 24-wk-old NOD mice hydrodynamically injected with control (Ctrl, three mice) or LTbR-Ig–encoding plasmids (LTbR-Ig, three mice) at 16 wk of age (left and middle panels). Insets, higher magnification of inflammatory foci (original magnification 38). Quantification of the area of inflammation and the area of CD3/B220 signal over total tissue area (Ctrl, 10 mice; LTbR-Ig, 17 mice) (right panels). (C) Quantification of the indicated immune subsets in lacrimal glands from control and treated mice by flow cytometry. Average FC values are indicated. Integrated results from three independent experiments. Significant p values are shown (Mann–Whitney U test). n.s., nonsignificant.

PD1, GITR, and IL-10 expression (34, 35). Accordingly, our study models of pSS. They demonstrated that Tregs play an impor- confirms that the autoimmune target tissue from Sjo¨gren’s patients tant role in the control of exocrine gland inflammation and that contains large numbers of FOXP3+ Tregs, as well as identifies a new hormonal factors modulate Treg activity in the draining lymph subset of tissue-resident Tregs expressing PD1 that was not present in node of the autoimmune target tissue (36, 37). However, these normal salivary glands or in the infiltrated kidneys from SLE patients. studies did not explore the impact of Treg depletion/transfer on The discovery of this eTreg subset in the target tissue of tissue-resident T cell subsets nor did they explore the contribution Sjo¨gren’s syndrome in mice and humans warrants further char- of LGT-resident eTregs to the observed phenotypes. As reviewed acterization. Future studies should evaluate, in particular, the recently (31, 33, 38), our knowledge of eTreg biology is still in- suppressive functions of these tissue-resident eTregs using ex vivo complete. Thus, understanding how LGT-resident eTregs suppress suppression assays, neonatal/SCID transfer models, and/or de- self-reactive T cell clones in the target tissue will be an important pletion studies. Such approaches were recently reported in mouse question to address. The transcriptional signatures derived in this The Journal of Immunology 3815 Downloaded from FIGURE 6. LTbR-Ig selectively reduces glandular recruitment of naive, but not effector, CD4+ T cells. Flow cytometry analysis of CD4+ Tcellsubsetsin lacrimal glands and spleens from 24-wk-old NOD mice treated with control (Ctrl) or LTbR-Ig–encoding plasmids at 16 wk of age, as described in Fig. 5. (A)

Expression of PD1 and ICOS (left panels) and CD4 http://www.jimmunol.org/ and FOXP3 (right panels) in CD4+ T cells from representative NOD mice in each treatment group. Numbers indicate the percentage cells in the outlined areas. (B and C) Number of cells detected per gland in control (Ctrl) and treated (LTbR-Ig) animals for each indicated cell subset. Integrated results from three independent experiments. Average FC values are indicated. Signiﬁcant p values are shown (Mann– Whitney U test). n.s., nonsigniﬁcant. by guest on October 2, 2021

study point to new molecular candidates: Areg, a ligand for EGFR Il2, with no evidence of other Th phenotypes, such as Th2 or that stimulates Treg-suppressive functions (39) and was recently Th17. The numerous effector molecules detected in this subset demonstrated to have a key role in Treg-mediated tissue repair indicate that this cellular phenotype identifies the pathogenic (40); Fgl2, a secreted factor with potent immunoregulatory CD4+ T cell subset associated with pSS. Among these effector properties (41); or Itgb8, which is required for the activation of molecules, Il21, Tnfsf13b (BAFF), and Lta, as well as chemotactic latent TGF-b in Tregs (42). factors, such as Xcl1 and Ccl1, likely contribute to the formation In addition to Prdm1+ eTregs, this study identified a popula- of T–B aggregates (43). Importantly, however, we did not detect tion of Il21+ Th1 cells defined by a PD1highICOShighCD73high expression of Cxcr5 or Bcl6, indicating that this Il21-expressing CD200high phenotype and high expression of Ifng, Tbx21, Tnf, and Th1 cell subset does not have a fully differentiated Tfh phenotype. 3816 TISSUE-INFILTRATING T CELL SUBSETS IN SJO¨ GREN’S Downloaded from http://www.jimmunol.org/ by guest on October 2, 2021

FIGURE 7. Abundant PD1+CD4+ T cells are detected in infiltrated salivary glands from Sjo¨gren’s patients but not in infiltrated kidneys from SLE patients. Immunofluorescent analysis of minor (n = 3) and major (n = 3) salivary (SAL.) glands from patients with Sjo¨gren’s syndrome (A), renal biopsies from patients with SLE (n =4)(B), and control tonsil and normal salivary gland tissues (C). Low-magnification images show the pattern and distribution of CD4 (green) and PD1 (red) immunoreactivity (top row). CD4, PD1, and their overlap in higher-magnification areas for each subject (second, third, and fourth rows, respectively). (D) Quantification of (A) and (B) showing the percentage of CD4+ area that is also PD1+. Note that this metric is not directly related to the frequency of CD4+PD1+ cells because it measures areas (pixels) of fluorescence and, thus, is specific for colocalized signal but underestimates total area occupied by double-positive cells. The p value was obtained using the Student unpaired t test with Welch’s correction. The Journal of Immunology 3817 Downloaded from http://www.jimmunol.org/ by guest on October 2, 2021

2 FIGURE 8. Most PD1+CD4+ T cells in Sjo¨gren’s salivary glands are FOXP3 effector cells, with a minor PD1+FOXP3+ population uniquely detected in Sjo¨gren’s patients but not in SLE patients. Immunohistochemical analysis of PD1 and FOXP3 was performed on the samples described in Fig. 7. (A)Low- magnification images of salivary glands from Sjo¨gren’s patients showing PD1 (purple) and FOXP3 (brown) immunoreactive cells (left panels). Higher-magnification images (of the outlined areas in the left panels) for each subject, revealing the presence of a subset of PD1+ cells also positive for FOXP3 (right panels, arrows). (B) Quantification of the number of FOXP3+,PD1+, and FOXP3/PD1 double-positive cells per mm2 of tissue. The p values were obtained using the Student unpaired t test with Welch’s correction. Representative images of a normal salivary gland (C) and SLE renal biopsies (D) stained for PD1 (purple) and FOXP3 (brown).

This observation is consistent with the absence of ectopic ger- Mirroring these observations in the NOD mouse model of pSS, minal centers in this model (43) and the fact that only 10–20% of large numbers of CD4+PD1+FOXP32 cells were detected in sali- pSS patients present ectopic germinal centers. vary gland biopsies from Sjo¨gren’s patients. This cellular phenotype 3818 TISSUE-INFILTRATING T CELL SUBSETS IN SJO¨ GREN’S was largely absent from normal salivary glands and kidney infil- together, these observations help to interpret the recently disclosed trates from SLE patients, indicating that it was associated with the results of the baminercept clinical trial (30). In this trial, bami- disease state and corresponded to a cellular feature likely specific to nercept was no more effective than placebo on salivary gland Sjo¨gren’s disease. The fact that this subset was present in minor and function, despite a modest effect on disease activity. Although the major salivary glands and in the presence and absence of organizing absence of an effect on salivary production could be due to patient ectopic lymphoid structures additionally suggests that this pheno- enrollment criteria and prior irreversible structural damage to the type occurs in patients with severe disease manifestations, as well as glands of the patients, the results of this clinical trial are in striking in patients with more muted forms of the disease or at an earlier contrast with the efficacy of LTbR-Ig in preclinical models (13– stage of the pathogenic process. Based on the transcriptomic anal- 15). The present study might reconcile some of these paradoxical ysis of the nonregulatory PD1high T cell subset in NOD target tissue, observations; by having only minimal effect on the pathogenic one can infer that this CD4+PD1+FOXP32 subset in human salivary effector Th1 cells, LTbR-Ig treatment could potentially result in glands also corresponds to the pathogenic T cell subset in Sjo¨gren’s continual epithelial tissue damage and/or inhibition of glandular patients. Future studies should further explore and validate the functions, despite the clear reduction in the size of the immune pathogenic phenotype of this T cell subset in humans, using the infiltrate observed in preclinical studies. markers that we identified in mice, such as ICOS, CD200, and CD73, In conclusion, this study describes two previously unrecognized as well as the corresponding transcriptional signature (Supple- cellular players in pSS pathogenesis that are defined as Prdm1+ mental Table I). eTregs and Il21+ Th1 cells. The characterization of the tran- The pathogenicity of the effector PD1high T cell subset identified scriptional signatures of these dominant LGT-infiltrating CD4+ in NOD mice was further suggested by its transcriptional similarity T cell subsets opens many new avenues of investigation and will with another CD4+ T cell subset described in lupus-prone B6.Sle1b guide the design of innovative therapeutics that selectively elim- Downloaded from and BWF1 mouse models (25, 26). In these strains, CD4+ Tcells inate or inhibit effector subsets while sparing tissue-resident reg- with a PD1highCD44highCD62Llow phenotype were shown to secrete ulatory activities. large amounts of osteopontin in spleen and kidney and were characterized as pathogenic. The pathogenic role of osteopontin remains Acknowledgments to be explored in pSS, but it is noteworthy that Spp1-transgenic We thank Biogen colleagues Bob Dunstan, Jo Viney, and Linda Burkly for mice develop a pSS-like phenotype (44). Importantly, the patho- discussion and input on this project, Nels Pederson and Chioma Nwankwo http://www.jimmunol.org/ genicity of this subset in lupus-prone strains is Slamf6 dependent, for generating vectors for hydrodynamic injections, Doug Drager and Sarah and Slamf6 was strongly upregulated in LGT-resident Th1 cells in Smith for help and guidance with hydrodynamic injections, and Attila NOD mice. Altogether, these observations identify PD1highICOShigh Fabian and Grigoriy Losyev for excellent cytometry technical support. CD73highCD200high cells as a new cellular target in pSS. Future studies are needed to validate the pathogenicity of this tissue- Disclosures infiltrating T cell subset. This will require the development of The authors have no financial conflicts of interest. new tools and reagents that selectively modulate and/or deplete Tconvs while leaving the tissue-resident Tregs untouched. The References by guest on October 2, 2021 transcriptional signatures that we identified in this study should 1. Fisher, B. A., R. M. Brown, S. J. Bowman, and F. Barone. 2015. A review of facilitate the identification of appropriate markers showing selec- salivary gland histopathology in primary Sjo¨gren’s syndrome with a focus on its tivity of expression in the pathogenic Tconv subset. potential as a clinical trials biomarker. Ann. Rheum. Dis. 74: 1645–1650. LGT-resident Th1 effector cells express high levels of Lta. In- 2. Nocturne, G., and X. Mariette. 2013. Advances in understanding the pathogenesis of primary Sjo¨gren’s syndrome. Nat. Rev. Rheumatol. 9: 544–556. terestingly, Ltb is upregulated 8-fold in salivary glands from pSS 3. Mitsias, D. I., A. G. Tzioufas, C. Veiopoulou, E. Zintzaras, I. K. Tassios, patients (Supplemental Table IA), and LTA and PDCD1 genes are O. Kogopoulou, H. M. Moutsopoulos, and G. Thyphronitis. 2002. The Th1/Th2 + cytokine balance changes with the progress of the immunopathological lesion of hypomethylated in peripheral CD4 T cells from pSS patients Sjogren’s syndrome. Clin. Exp. Immunol. 128: 562–568. (45). Altogether, these data suggest that lymphotoxin-expressing 4. Kolkowski, E. C., P. Reth, F. Pelusa, J. Bosch, R. Pujol-Borrell, J. Coll, and PD1+CD4+ T cells could play a pathogenic role in human pSS. D. Jaraquemada. 1999. Th1 predominance and perforin expression in minor salivary glands from patients with primary Sjo¨gren’s syndrome. J. Autoimmun. Accordingly, the importance of LTbR signaling in pSS was clearly 13: 155–162. demonstrated in preclinical models (13–15). The present study 5. Ohyama, Y., S. Nakamura, G. Matsuzaki, M. Shinohara, A. Hiroki, T. Fujimura, confirms these earlier observations but also reveals that this A. Yamada, K. Itoh, and K. Nomoto. 1996. Cytokine messenger RNA expression in the labial salivary glands of patients with Sjo¨gren’s syndrome. Arthritis treatment selectively impacts recently infiltrated naive T cells and Rheum. 39: 1376–1384. not PD1highICOShigh effector T cells. Accumulation of the latter 6. Maehara, T., M. Moriyama, J. N. Hayashida, A. Tanaka, S. Shinozaki, Y. Kubo, cells in the tissue could result from in situ expansion of resident K. Matsumura, and S. Nakamura. 2012. Selective localization of T helper subsets in labial salivary glands from primary Sjo¨gren’s syndrome patients. Clin. T cell clones, a mechanism that was recently reported for VAT Exp. Immunol. 169: 89–99. Tregs (46) and is supported by the high levels of Ki67 expressed 7. Sakai, A., Y. Sugawara, T. Kuroishi, T. Sasano, and S. Sugawara. 2008. Iden- tification of IL-18 and Th17 cells in salivary glands of patients with Sjo¨gren’s by these cells. Alternatively, eTregs and Th1 cells might migrate syndrome, and amplification of IL-17-mediated secretion of inflammatory cy- into the tissue using LTbR-independent mechanisms. In support of tokines from salivary gland cells by IL-18. J. Immunol. 181: 2898–2906. this model, the expression of Itgae (CD103) is 40-fold higher on 8. Mieliauskaite, D., I. Dumalakiene, R. Rugiene, and Z. Mackiewicz. 2012. Ex- pression of IL-17, IL-23 and their receptors in minor salivary glands of patients eTregs than on naive T cells (Supplemental Table IC) and could with primary Sjo¨gren’s syndrome. Clin. Dev. Immunol. 2012: 187258. promote their accumulation in the lacrimal epithelium, which 9. Christodoulou, M. I., E. K. Kapsogeorgou, N. M. Moutsopoulos, and expresses the CD103 ligand E-cadherin. Similarly, resident Th1 H. M. Moutsopoulos. 2008. Foxp3+ T-regulatory cells in Sjogren’s syndrome: correlation with the grade of the autoimmune lesion and certain adverse prog- cells express 7-fold higher levels of Itgb1 than do naive T cells nostic factors. Am. J. Pathol. 173: 1389–1396. (Supplemental Table IC). ITGB1 can bind VCAM-1 when asso- 10. Furuzawa-Carballeda, J., J. Sańchez-Guerrero, J. L. Betanzos, A. B. Enriquez, C. Avila-Casado, L. Llorente, and G. Hernańdez-Molina. 2014. Differential ciated with ITGA4. Interestingly, vascular endothelia in inflamed cytokine expression and regulatory cells in patients with primary and secondary areas of the LGT upregulate VCAM-1 and the PNAd epitope (47), Sjo¨gren’s syndrome. Scand. J. Immunol. 80: 432–440. but only PNAd (and not VCAM-1) is decreased by LTbR-Ig (14). 11. Sarigul, M., V. Yazisiz, C. I. Bassorgun, M. Ulker, A. B. Avci, F. Erbasan, + 2 T. Gelen, R. M. Gorczynski, and E. Terzioglu. 2010. The numbers of Foxp3 + As a result, the recruitment of CD62L naive, but not CD62L , Treg cells are positively correlated with higher grade of infiltration at the salivary eTregs and Th1 T cells would be predicted to be decreased. Taken glands in primary Sjogren’s syndrome. Lupus 19: 138–145. The Journal of Immunology 3819

12. Maehara, T., M. Moriyama, H. Nakashima, K. Miyake, J. N. Hayashida, 30. St.Clair, E. W., A. N. Baer, G. Noaiseh, A. Parke, A. Coca, T. Utset, A. Tanaka, S. Shinozaki, Y. Kubo, and S. Nakamura. 2012. Interleukin-21 M. C. Genovese, D. J. Wallace, J. McNamara, K. Boyle, et al. 2015. The clinical contributes to germinal centre formation and immunoglobulin G4 production efficacy and safety of baminercept, a lymphotoxin-beta receptor fusion protein, in IgG4-related dacryoadenitis and sialoadenitis, so-called Mikulicz’s disease. in primary Sjo¨gren’s syndrome: results from a randomized, double-blind, Ann. Rheum. Dis. 71: 2011–2019. placebo-controlled phase II trial. Arthritis Rheumatol. 67 (Suppl. 10). Available at: 13. Shen, L., L. Suresh, J. Wu, J. Xuan, H. Li, C. Zhang, O. Pankewycz, and http://acrabstracts.org/abstract/the-clinical-efficacy-and-safety-of-baminercept- J. L. Ambrus, Jr. 2010. A role for lymphotoxin in primary Sjogren’s disease. J. a-lymphotoxin-beta-receptor-fusion-protein-in-primary-sjogrens-syndrome-results- Immunol. 185: 6355–6363. from-a-randomized-double-blind-placebo-controlled-phase-ii-trial/. 14. Fava, R. A., S. M. Kennedy, S. G. Wood, A. I. Bolstad, J. Bienkowska, 31. Cretney, E., A. Kallies, and S. L. Nutt. 2013. Differentiation and function of A. Papandile, J. A. Kelly, C. P. Mavragani, M. Gatumu, K. Skarstein, and Foxp3(+) effector regulatory T cells. Trends Immunol. 34: 74–80. J. L. Browning. 2011. Lymphotoxin-beta receptor blockade reduces CXCL13 in 32. Burzyn, D., C. Benoist, and D. Mathis. 2013. Regulatory T cells in nonlymphoid lacrimal glands and improves corneal integrity in the NOD model of Sjo¨gren’s tissues. Nat. Immunol. 14: 1007–1013. syndrome. Arthritis Res. Ther. 13: R182. 33. Richards, D. M., M. Delacher, Y. Goldfarb, D. Ka¨gebein, A. C. Hofer, 15. Gatumu, M. K., K. Skarstein, A. Papandile, J. L. Browning, R. A. Fava, and J. Abramson, and M. Feuerer. 2015. Treg cell differentiation: from thymus to A. I. Bolstad. 2009. Blockade of lymphotoxin-beta receptor signaling reduces peripheral tissue. Prog. Mol. Biol. Transl. Sci. 136: 175–205. aspects of Sjo¨gren’s syndrome in salivary glands of non-obese diabetic mice. 34. Alunno, A., M. G. Petrillo, G. Nocentini, O. Bistoni, E. Bartoloni, S. Caterbi, Arthritis Res. Ther. 11: R24. R. Bianchini, C. Baldini, I. Nicoletti, C. Riccardi, and R. Gerli. 2013. Charac- 16. Schneider, K., K. G. Potter, and C. F. Ware. 2004. Lymphotoxin and LIGHT terization of a new regulatory CD4+ T cell subset in primary Sjo¨gren’s syn- signaling pathways and target genes. Immunol. Rev. 202: 49–66. drome. Rheumatology (Oxford) 52: 1387–1396. 17. Browning, J. L., N. Allaire, A. Ngam-Ek, E. Notidis, J. Hunt, S. Perrin, and 35. Kobayashi, M., S. Kawano, S. Hatachi, C. Kurimoto, T. Okazaki, Y. Iwai, R. A. Fava. 2005. Lymphotoxin-beta receptor signaling is required for the ho- T. Honjo, Y. Tanaka, N. Minato, T. Komori, et al. 2005. Enhanced expression of meostatic control of HEV differentiation and function. Immunity 23: 539–550. programmed death-1 (PD-1)/PD-L1 in salivary glands of patients with Sjo¨gren’s 18. Szyszko, E. A., K. A. Brokstad, G. Oijordsbakken, M. V. Jonsson, R. Jonsson, syndrome. J. Rheumatol. 32: 2156–2163. and K. Skarstein. 2011. Salivary glands of primary Sjo¨gren’s syndrome patients 36. Ellis, J. S., X. Wan, and H. Braley-Mullen. 2013. Transient depletion of CD4+ express factors vital for plasma cell survival. Arthritis Res. Ther. 13: R2. CD25+ regulatory T cells results in multiple autoimmune diseases in wild-type 19. Maier-Moore, J. S., K. A. Koelsch, K. Smith, C. J. Lessard, L. Radfar, D. Lewis, and B-cell-deficient NOD mice. Immunology 139: 179–186. B. T. Kurien, N. Wolska, U. Deshmukh, A. Rasmussen, et al. 2014. Antibody- 37. Lieberman, S. M., P. A. Kreiger, and G. A. Koretzky. 2015. Reversible lacrimal Downloaded from secreting cell specificity in labial salivary glands reflects the clinical presentation gland-protective regulatory T-cell dysfunction underlies male-specific autoim- and serology in patients with Sjo¨gren’s syndrome. Arthritis Rheumatol. 66: mune dacryoadenitis in the non-obese diabetic mouse model of Sjo¨gren syn- 3445–3456. drome. Immunology 145: 232–241. 20. Mingueneau, M., S. Boudaoud, S. Haskett, T. L. Reynolds, G. Nocturne, 38. Campbell, D. J. 2015. Control of regulatory T cell migration, function, and E. Norton, X. Zhang, M. Constant, D. Park, W. Wang, T. Lazure, C. LePajolec, homeostasis. J. Immunol. 195: 2507–2513. A. Ergun, and X. Mariette. 2016. CyTOF immunophenotyping identifies a blood 39. Zaiss, D. M., J. van Loosdregt, A. Gorlani, C. P. Bekker, A. Gro¨ne, M. Sibilia, Sjo¨gren’s signature correlating with disease activity and glandular inflammation. P. M. van Bergen en Henegouwen, R. C. Roovers, P. J. Coffer, and A. J. Sijts.

J. Allergy Clin. Immunol. 137:1809–1821.e12. 2013. Amphiregulin enhances regulatory T cell-suppressive function via the http://www.jimmunol.org/ 21. Delaleu, N., C. Q. Nguyen, A. B. Peck, and R. Jonsson. 2011. Sjo¨gren’s syn- epidermal growth factor receptor. Immunity 38: 275–284. drome: studying the disease in mice. Arthritis Res. Ther. 13: 217. 40. Arpaia, N., J. A. Green, B. Moltedo, A. Arvey, S. Hemmers, S. Yuan, 22. Liu, F., Y. Song, and D. Liu. 1999. Hydrodynamics-based transfection in animals P. M. Treuting, and A. Y. Rudensky. 2015. A distinct function of regulatory by systemic administration of plasmid DNA. Gene Ther. 6: 1258–1266. T cells in tissue protection. Cell 162: 1078–1089. 23. Mingueneau, M., S. Krishnaswamy, M. H. Spitzer, S. C. Bendall, E. L. Stone, 41. Chruscinski, A., H. Sadozai, V. Rojas-Luengas, A. Bartczak, R. Khattar, S. M. Hedrick, D. Pe’er, D. Mathis, G. P. Nolan, and C. Benoist. 2014. Single- N. Selzner, and G. A. Levy. 2015. Role of regulatory T cells (Treg) and the Treg cell mass cytometry of TCR signaling: amplification of small initial differences effector molecule fibrinogen-like protein 2 in alloimmunity and autoimmunity. results in low ERK activation in NOD mice. Proc. Natl. Acad. Sci. USA 111: Rambam Maimonides Med. J. 6: 6. 16466–16471. 42. Worthington, J. J., A. Kelly, C. Smedley, D. Bauche´, S. Campbell, J. C. Marie, 24. Allaire, N. E., S. E. Bushnell, J. Bienkowska, G. Brock, and J. Carulli. 2013. and M. A. Travis. 2015. Integrin avb8-mediated TGF-b activation by effector Optimization of a high-throughput whole blood expression profiling methodol- regulatory T cells is essential for suppression of T-cell-mediated inflammation.

ogy and its application to assess the pharmacodynamics of interferon (IFN) Immunity 42: 903–915. by guest on October 2, 2021 beta-1a or polyethylene glycol-conjugated IFN beta-1a in healthy clinical trial 43. Ding, J., W. Zhang, S. Haskett, A. Pellerin, S. Xu, B. Petersen, L. Jandreski, subjects. BMC Res. Notes 6: 8. S. Hamann, T. L. Reynolds, T. S. Zheng, and M. Mingueneau. 2016. BAFF 25. Keszei, M., C. Detre, W. Castro, E. Magelky, M. O’Keeffe, K. Kis-Toth, overexpression increases lymphocytic infiltration in Sjo¨gren’s target tissue, but G. C. Tsokos, N. Wang, and C. Terhorst. 2013. Expansion of an osteopontin- only inefficiently promotes ectopic B-cell differentiation. Clin. Immunol. 169: expressing T follicular helper cell subset correlates with autoimmunity in B6. 69–79. Sle1b mice and is suppressed by the H1-isoform of the Slamf6 receptor. FASEB 44. Husain-Krautter, S., J. M. Kramer, W. Li, B. Guo, and T. L. Rothstein. 2015. The J. 27: 3123–3131. osteopontin transgenic mouse is a new model for Sjo¨gren’s syndrome. Clin. 26. Tahir, S., Y. Fukushima, K. Sakamoto, K. Sato, H. Fujita, J. Inoue, T. Uede, Immunol. 157: 30–42. Y. Hamazaki, M. Hattori, and N. Minato. 2015. A CD153+CD4+ T follicular cell 45. Altorok, N., P. Coit, T. Hughes, K. A. Koelsch, D. U. Stone, A. Rasmussen, population with cell-senescence features plays a crucial role in lupus patho- L. Radfar, R. H. Scofield, K. L. Sivils, A. D. Farris, and A. H. Sawalha. 2014. genesis via osteopontin production. J. Immunol. 194: 5725–5735. Genome-wide DNA methylation patterns in naive CD4+ T cells from patients 27. Hill, J. A., M. Feuerer, K. Tash, S. Haxhinasto, J. Perez, R. Melamed, with primary Sjo¨gren’s syndrome. Arthritis Rheumatol. 66: 731–739. D. Mathis, and C. Benoist. 2007. Foxp3 transcription-factor-dependent and 46. Kolodin, D., N. van Panhuys, C. Li, A. M. Magnuson, D. Cipolletta, -independent regulation of the regulatory T cell transcriptional signature. Im- C. M. Miller, A. Wagers, R. N. Germain, C. Benoist, and D. Mathis. 2015. munity 27: 786–800. Antigen- and cytokine-driven accumulation of regulatory T cells in visceral 28. Cipolletta, D., M. Feuerer, A. Li, N. Kamei, J. Lee, S. E. Shoelson, C. Benoist, adipose tissue of lean mice. Cell Metab. 21: 543–557. and D. Mathis. 2012. PPAR-g is a major driver of the accumulation and phe- 47. Mikulowska-Mennis, A., B. Xu, J. M. Berberian, and S. A. Michie. 2001. notype of adipose tissue Treg cells. Nature 486: 549–553. Lymphocyte migration to inflamed lacrimal glands is mediated by vascular cell 29. Burzyn, D., W. Kuswanto, D. Kolodin, J. L. Shadrach, M. Cerletti, Y. Jang, adhesion molecule-1/alpha(4)beta(1) integrin, peripheral node addressin/l-selectin, E. Sefik, T. G. Tan, A. J. Wagers, C. Benoist, and D. Mathis. 2013. A special and lymphocyte function-associated antigen-1 adhesion pathways. Am.J.Pathol. population of regulatory T cells potentiates muscle repair. Cell 155: 1282–1295. 159: 671–681. A pSS human transcriptional signature (1255 probes; 933 genes)

100

10 1 (human parotid glands)

1 GSE406 1 Fold change pSS versus non-pSS

-10 -10 1 10 100 Fold change pSS versus non-pSS GSE23117 (human minor salivary glands)

B human pSS signature (1740 probes)

[1013 probes with FC>1;p-value(FDR)<0.05, i.e. 58%]

0 118 1013

-3

-6

-9

-12

log10(p-value) -15

-18

-21

-24

-27 -100 -10 1 10 100 Fold change 16-week versus 8-week old NOD mice GSE32681 (mouse lacrimal glands) Supplementary Figure 1. 58% of the human pSS transcriptional signature is upregulated in inflamed NOD lacrimal glands. (A) Publically-available microarray datasets were used to generate a FC-FC plot comparing gene expression in minor salivary glands (GSE23117, x-axis) and parotid glands (GSE40611, y-axis) from pSS and non-pSS subjects. Filtered on genes with p-value(FDR)<0.05 in at least one of the two datasets and mean expression >6 in at least one group of subjects. 933 genes (1255 probes) showed significant FC>2 in the two datasets and were defined as the pSS transcriptional signature (black square). Blue lines, FC=2. See complete list of genes included in the human pSS signature in Supplementary Table 1A. (B) Volcano plot comparing whole lacrimal gland tissue transcriptomes from 16-week old (i.e., at the peak of immune infiltration) and 8-week old NOD mice (i.e., at the initiation of immune infiltration) (GSE32681). The human pSS signature defined in panel A (933 genes corresponding to 1740 mouse probes) is highlighted in red on the mouse dataset with the corresponding number of gene probes significantly upregulated and downregulated in inflamed mouse lacrimal glands, i.e., in 16-week old NOD mice (see complete list of genes in Supplementary Table 1B).

1 CD73high CD200high CD73low CD200low - - Hspa4l Kif1b PD1 ICOS Tnfrsf9 Bambi-ps1 Lbh Spleen T cells Adora2a Psap Casp1 Fmnl2 Lclat1 Pltp Casp3 Gbp2 Pdcd1 S100a6 Dusp10 Endod1 Csf2rb2 Zdhhc2 Trim13 Mcm6 Mdfic Csf2rb Hivep3 Cd244 Casp4 Gimap6 Acot11 Ahr Gpr183 E2f3 Klra17 Fam129a Rell1 Il2rb Trim36 Tbx21 Cxcr3 Tubb2a Pear1 Dnajb4 Cdca8 Rasl11b Cst7 Uhrf1 Slc41a2 Anxa2 Ctsl Akap12 Dirc2 S100a4 Gramd3 Acot7 Gng2 Mgat4a Comt Plac8 Nfatc1 Pde3b Prkca Cd55 Dusp16 Nt5e Ctsh Tmem2 Gramd4 Cd200 Gpr18 Plscr1 Mrps6 Ampd1 Tjp2 Sgk3 Spock2 Ly86 Gfod1 Rapgef4 Ptprv Myo1e Sema4a Tox H2-Aa Bcl2a1a Bc H2-Eb1 Ppap2c Zan Filip1l Serpina3g Pdlim1 Lag3 Pik3ip1 Kdm2b Als2cl Slc5a3 Axl Klhdc1 Rgs16 Rreb1 Twsg1 Sell Ccr8 Cyfip1 Eef2k Irak1bp1 Acss2 Syt11 Ndrg3 Rbpj Klhdc2 Penk Wisp1 Aqr Gdpd5 Commd8 Spry1 Atp1b1 Ifng Zc3h12d Slpi Tpi1 Dnah8 Hnrpll Klf2 Maf Treml2 Tnfsf11 Farp1 Hs3st3b1 Raph1 Lair1 Prnp Tyrobp Sh3rf1 Bmpr2 Basp1 H1f0 Acvr1b Abi2 Ugcg Ubash3b Zbtb18 Ptgfrn Arhgap31 Cnn3 Tox2 Zdhhc14 Nrn1 Fam69a Gpm6b Tiam1 Ccr9 Dsel Klrd1 Socs2 Dapl1 Dst Amigo2 Cyfip1 Tnfrsf4 Actn1 Itgb8 Igfbp4 Galm S1pr1 Slc22a15 Spib Angptl2 Cd44 Mctp2 Ptger2 Scml4 Eea1 Dtx1 Stx11 Trps1 Xkrx Tnfrsf1b Rras2 Icos BC018473 Cish Trat1 Folr4 Ikzf2 Itga6 Rora Satb1 Ehd1 Emb Rhbdl3 Dab2ip Rasgrp2 Atxn1 Pik3r5 Ahi1 Prkd3 Gna13 Arhgap29 Msi2 Alcam Pdlim4 Impa2 Itgb3 Bhlhe40 St8sia6 Dusp4 Ypel2 Tbc1d8 Nfil3 Adh1 Cxcl10 Myo10 Gem Klf3 Nr4a2 Ctla4 Mpeg1 Ramp1 Tcf4 Mapre2 Plaur Rrm2 Lyrm2 Bach2 Gpr68 Spice1 Il18rap Grcc10 Nkg7 Supplementary Figure 2. Transcriptional signatures from PD1highICOShigh and PD1−ICOS− LGT-infiltrating T cell subsets. Expression heatmaps displaying transcriptional signatures for PD1highICOShigh (left) and PD1−ICOS− (right) LGT-infiltrating T cell subsets. The PD1highICOShigh signature corresponds to the genes which are similarly regulated in CD73lowCD200low and CD73highCD200high subsets relative to the PD1−ICOS− subset. Signatures were generated by filtering genes displayed in Figure 4 panel A on genes showing mean expression >6 and at least 3-fold significant upregulation (p-value(FDR)<0.05) in the subset(s) of interest relative to the other LGT-infiltrating CD4+ T cell subset(s). Samples are color-coded and hierarchically-clustered according to gene expression (see complete list of genes in Supplementary Table 1E).

2 CD73high CD200high CD73low CD200low PD1- ICOS- Ikzf3 Spleen T cells Prdm1 Foxp3 Mxi1 Cebpb Maff Jun Crem Nfkbia Rel Nfil3 Klf4 Pawr Nr4a2 Mxd4 Maf Tbx21 Pbx3 Tox Nfatc1 Rbpj Hif1a Jarid2 Hivep3 Elk3 E2f3 Ahr Trps1 Gfi1 Id2 Ikzf4 Ikzf2 Rora Nfat5 Plagl1 Gata3 Pou2af1 Ets1 Bptf Smad7 Irf8 Stat1 Irf7 Tsc22d3 Mafk Tsc22d1 Arid5a Trim13 Mef2a Ncoa4 Runx3 Bach1 Pbx2 Rreb1 Klf2 Klf7 Smad1 Klf3 Foxp1 Tcf4 Spib Rere Gtf2i Bach2 Tcf7 Nfe2l2 Supplementary Figure 3. Transcription factor profiles expressed by LGT-infiltrating CD4+ T cell subsets. Expression heatmap of the genes displayed in Figure 4 panel A that encode proteins with transcription factor activities (based on gene ontology annotations) and that show at least 2-fold significant differential expression (p-value(FDR)<0.05) in one of the two PD1highICOShigh T cell subsets relative to the PD1−ICOS− subset. Samples are color-coded (see key) and hierarchically-clustered according to gene expression.