Regulatory T Cells in Ovarian Cancer Are Characterized by a Highly Activated Phenotype Distinct from That in Melanoma Aras Toker1, Linh T

Published OnlineFirst July 31, 2018; DOI: 10.1158/1078-0432.CCR-18-0554

Biology of Human Tumors Clinical Cancer Research Regulatory T Cells in Ovarian Cancer Are Characterized by a Highly Activated Phenotype Distinct from that in Melanoma Aras Toker1, Linh T. Nguyen1, Simone C. Stone1, S.Y. Cindy Yang2, Sarah Rachel Katz3, Patricia A. Shaw4, Blaise A. Clarke4, Danny Ghazarian4, Ayman Al-Habeeb4, Alexandra Easson5, Wey L. Leong5, David R. McCready5, Michael Reedijk1,2,5, Cynthia J. Guidos6,7, Trevor J. Pugh2,8, Marcus Q. Bernardini3, and Pamela S. Ohashi1,2,6

Abstract

Purpose: Regulatory T (Treg) cells expressing the transcrip- number of receptors associated with TCR engagement, includ- tion factor FOXP3 are essential for the maintenance of immu- ing PD-1, 4-1BB, and ICOS. Higher PD-1 and 4-1BB expression nologic self-tolerance but play a detrimental role in most was associated with increased responsiveness to further TCR cancers due to their ability to suppress antitumor immunity. stimulation and increased suppressive capacity, respectively. The phenotype of human circulating Treg cells has been Transcriptomic and mass cytometry analyses revealed the extensively studied, but less is known about tumor-infiltrating presence of Treg cell subpopulations and further supported Treg cells. We studied the phenotype and function of tumor- a highly activated state specifically in ovarian tumors. In infiltrating Treg cells in ovarian cancer and melanoma to comparison, Treg cells infiltrating melanomas displayed identify potential Treg cell–associated molecules that can be lower FOXP3, PD-1, 4-1BB, and ICOS expression and were targeted by tumor immunotherapies. less potent suppressors of CD8 T-cell proliferation. Experimental Design: The phenotype of intratumoral Conclusions: The highly activated phenotype of ovarian and circulating Treg cells was analyzed by multicolor tumor-infiltrating Treg cells may be a key component of an flow cytometry, mass cytometry, RNA-seq, and functional immunosuppressive tumor microenvironment. Receptors assays. that are expressed by tumor-infiltrating Treg cells could be Results: Treg cells isolated from ovarian tumors displayed exploited for the design of novel combination tumor immu- a distinct cell surface phenotype with increased expression of a notherapies. Clin Cancer Res; 1–12. 2018 AACR.

Introduction to uncover the mechanism of action and to identify biomarkers associated with response or resistance in order to establish The unprecedented success of immune checkpoint inhibitors guidelines for patient selection and provide a rationale for the has fueled efforts to further explore, expand, and develop novel development of combination immunotherapies (3, 4). cancer immunotherapy approaches. However, the mechanism Mechanisms of immune suppression within the tumor micro- of action of checkpoint inhibitors is not clear and response environment have been of particular focus, including the upre- rates are highly variable, and typically lower than 50%, across gulation of inhibitory molecules and the recruitment of sup- different cancers (1, 2). Therefore, research has been intensified pressive cell populations such as myeloid-derived suppressor cells (MDSC), tumor-associated macrophages (TAM), and regulatory T (Treg) cells (5–7). 1The Campbell Family Institute for Breast Cancer Research, Princess Margaret Treg cells are essential for the maintenance of immunologic Cancer Centre, University Health Network, Toronto, Ontario, Canada. self-tolerance and the prevention of autoimmunity (8). The 2Department of Medical Biophysics, University of Toronto, Toronto, Ontario, 3 transcription factor FoxP3 is a master regulator of Treg cell Canada. Division of Gynecologic Oncology, University Health Network, fi Toronto, Ontario, Canada. 4Department of Laboratory Medicine and Pathobi- function, and its de ciency leads to devastating autoimmune ology, University Health Network, University of Toronto, Toronto, Ontario, pathology in mice (8) and humans (9). In mice expression of Canada. 5Department of Surgical Oncology, University Health Network, Toronto, FoxP3 protein is strictly restricted to bona fide suppressive Treg Ontario, Canada. 6Department of Immunology, University of Toronto, Toronto, cells, whereas in humans, effector T cells without suppressive Ontario, Canada. 7Program in Developmental and Stem Cell Biology, Hospital for 8 function can also transiently express low levels of FoxP3 upon Sick Children Research Institute, Toronto, Ontario, Canada. Princess Margaret activation (10). Among many mechanisms Treg cells deploy to Genomics Centre, University Health Network, Toronto, Ontario, Canada. exhibit their immunosuppressive function are direct lysis of Note: Supplementary data for this article are available at Clinical Cancer immune effector cells, depletion of growth factors and metabo- Research Online (http://clincancerres.aacrjournals.org/). lites, inhibition of antigen-presenting cell function, and secretion Corresponding Author: Pamela S. Ohashi, Princess Margaret Cancer Centre, 610 of regulatory cytokines (9). University Avenue, Suite 8-407, Toronto, Ontario, Canada M5G 2C9. Phone: 416- The majority of circulating Treg cells belong to a distinct 946-4501, ext. 3689; Fax: 416-204-2276; E-mail: [email protected] þ and phenotypically stable lineage of CD4 T cells in mice and doi: 10.1158/1078-0432.CCR-18-0554 humans, as the expression of the master transcription factor 2018 American Association for Cancer Research. FoxP3 is initiated and epigenetically imprinted during thymic

www.aacrjournals.org OF1

Toker et al.

Translational Relevance Materials and Methods Study participants Rates of response to immunotherapy in ovarian cancer are All human tissue and blood were obtained through protocols low despite pronounced immune infiltration. Regulatory approved by the Institutional Review Board in concordance with T (Treg) cells infiltrate many human tumors and are in most the Declaration of Helsinki. Written informed consent was cases associated with poor prognosis. In this study, we obtained from all donors who provided the samples. found that tumor-infiltrating Treg cells express a distinct array of targetable immunoreceptors, and are highly acti- Tumor specimens vated and potently suppressive. Melanoma-infiltrating Treg Tumor specimens were obtained from the UHN Biospecimen cells share many features of this Treg cell phenotype, but Program, through surgical resection from patients diagnosed possess distinct differences that could require reconsidera- with epithelial ovarian cancer (mostly high-grade serous) or tion and adjustment of immunotherapeutic approaches melanoma. Patient characteristics are summarized in Supplemen- according to cancer type. A highly suppressive tumor micro- tary Table S1. Fresh tumor specimens were enzymatically digested environment could also explain the low response rate in (RPMI1640 containing 2 mmol/L L-glutamine, 1 mg/mL colla- ovarian cancer and suggests that combination therapies that genase, 10 mg/mL Pulmozyme, 100 U/mL penicillin, 100 mg/mL boost antitumor immunity and diminish immunosuppres- streptomycin, 10 mg/mL gentamicin sulfate, and 1.25 mg/mL sion could be more promising. The Treg cell expression Amphotericin B) using the GentleMACS system (Miltenyi Biotec). pattern of inhibitory receptors could become an important Tumor digests were cryopreserved in human serum containing factor to consider in treatment decisions to avoid unwanted 10% DMSO prior to analysis. A total of 2.2 107 (1.7 107) increases in Treg cell–mediated immunosuppression by and 1.1 107 (1.3 107) cells were processed and analyzed per cytotoxic T-cell–targeted immunotherapy regimens. ovarian tumor and melanoma specimen, respectively. Cell via- bility for the ovarian and melanoma specimens was 61.7% (22.4%) and 43.0% (29.1%), respectively (Supplementary Table S1). development (11, 12). However, a subset of Treg cells also retains plasticity and can revert to a phenotype with diminished Flow cytometry and cell sorting suppressive function and produce IFNg or IL17 under inflam- Cells were stained with anti-CD3 (clone OKT3), anti-CD4 þ matory conditions (13, 14). FOXP3 -suppressive Treg cells can (RPA-T4), anti-CD8 (RPA-T8), anti–PD-1 (eBioJ105), anti–4-1BB also be induced in vivo from FOXP3 conventional T cells by (4B4-1), anti–TIM-3 (344823), anti-OX40 (Ber-ACT35), anti- stimulation in the presence of TGFb (15). Hence, the pool of TIGIT (MBSA43), anti–DNAM-1 (DX11), anti-CD27 (O323), circulating Treg cells consists of thymus-derived and peripherally anti-CD28 (CD28.2), anti-CD25 (M-A251), anti-ICOS (C398.4A), induced subsets, and the relative contribution of each subset to anti-CD45RA (HI100), and anti-CD127 (A019D5). Cells were stable and plastic Treg cell subpopulations has not been exten- fixed and permeabilized with the Foxp3 transcription factor sively studied. staining buffer set followed by intracellular staining with anti- Treg cells accumulate at the tumor site in many cancers FOXP3 (236A/E7) and anti–CTLA-4 (BNI3). Samples were including head and neck, colorectal, liver, lung, breast, ovarian, acquired on a BD LSR Fortessa SORP flow cytometer and data skin, and pancreas (5). A high prevalence of Treg cells within were analyzed using the FlowJo software. tumor-infiltrating lymphocytes (TIL) is associated with a poor Cell sorting was performed on a BD FACSAriaII or BD FACSAria outcome, in particular when combined with poor infiltration Fusion SORP cell sorter by The Flow and Mass Cytometry Facility, þ by CD8 cytotoxic T cells, in most cancers including ovarian The Hospital for Sick Children, Toronto, Canada. Intracellular (5, 16). In line with this, experimental depletion of Treg cells FOXP3 staining was performed on an aliquot of sorted cells as fosters immune responses against established tumors in murine above. models (17, 18). Therefore, the idea of interfering with Treg cell function has gained much interest for the treatment of Mass cytometry human cancers. Purified mAbs were conjugated with heavy metals by the Flow The cell surface expression of various immunoglobulin and and Mass Cytometry Facility, The Hospital for Sick Children. TNFR superfamily members by human circulating Treg cells has Whole tumor digests were counted and 2 106 cells for each been well defined (19), but potential changes to the surface sample were stained for cell surface markers (Supplementary phenotype following recruitment to tumors has not been carefully Table S3) in staining media (PBS containing 1% BSA and characterized across cancers. Because many receptors such as 0.02% NaN3) for 30 minutes at 4 C. Cells were washed with PD-1, 4-1BB, and ICOS represent potential targets for immuno- protein-free PBS, stained with 1 mmol/L cisplatin for 5 minutes at therapy, we investigated their expression by intratumoral Treg room temperature, fixed using the transcription factor buffer set cells from patients with ovarian cancer and melanoma. We found (BD Biosciences) followed by intracellular staining for 60 minutes that Treg cells infiltrating ovarian tumors have a highly activated at 4C. Cells were washed with staining media and stained with phenotype with potent suppressive activity and enhanced expres- 100 nmol/L iridium-labeled DNA-intercalator (Fluidigm) in PBS sion of both costimulatory and inhibitory receptors, which could containing 0.3% saponin and 1.6% formaldehyde at 4C for up to potentially be exploited for targeted Treg cell modulation. In 48 hours. Cells were washed twice with deionized water prior to contrast, the phenotype of melanoma-infiltrating Treg cells was adding EQ normalization beads containing Ce140, Eu151, indicative of a lower activation state, which could contribute to Eu153, Ho165, and Lu175 (Fluidigm) and acquiring on a Helios the higher response rates to immunotherapy in melanoma. mass cytometer by The Flow and Mass Cytometry Facility, The

OF2 Clin Cancer Res; 2018 Clinical Cancer Research

Treg Cells in Ovarian Cancer Are Highly Activated

Hospital for Sick Children. After normalizing and randomizing DNA methylation analysis values near zero using the Helios software, FCS files were Bisulfite conversion of genomic DNA was performed with uploaded to Cytobank for analysis. Data were initially visualized the EZ DNA Methylation Kit (Zymo). Converted DNA was ampli- on bivariate plots with hyperbolic arcsine axes using scale argu- fied with the primers 50-TGATTTGTTTGGGGGTAGAGGATT-30 ments between 2 and 5. Pregating was performed as illustrated in and 50-ACACCCATATCACCCCACCTAA-30, cloned into the Supplementary Fig. S4A. ViSNE and SPADE analyses were per- pCR4-TOPO vector (Invitrogen) and sequenced by the SickKids þ þ þ formed on pregated CD3 or CD3 CD4 CD8 cells using 23 to TCAG DNA sequencing facility. 24 mapping parameters (Supplementary Table S3). ViSNE anal- þ ysis was performed with equal sampling (6,000–20,000 CD3 Statistical analysis events from each sample), and perplexity was set to 30 and theta Statistical significance was determined by two-tailed unpaired to 0.5. Final KL divergence values of 2.60 to 3.37 were achieved Mann–Whitney U test or unpaired t test. The n values used to after a total of 1,000 iterations. The SPADE algorithm was run calculate statistics are defined and indicated in figure legends. using the same mapping parameters (Supplementary Table S3) Significance is indicated within figures, and differences that were without downsampling and target nodes were set to 30 to 50. not significant (P > 0.05) are denoted ns. Multiple hypothesis testing was performed using Sidak multiple comparisons test. Suppression assay Peripheral blood mononuclear cells (PBMC) were labeled with þ the eFluor450 Cell proliferation dye and CD8 responder cells Results þ were purified by FACS. Suppressor cells (Treg or control) from Ovarian tumors are infiltrated by suppressive FOXP3 Treg cells tumor specimens were FACS-purified. PBMCs were depleted of that express a distinct array of immunoreceptors T cells using a PE-conjugated anti-CD3 (clone UCHT1) antibody Tumor-infiltrating Treg cells are potent immune suppressors, and the EasySep Human PE Positive Selection Kit (Stemcell and a high frequency of intratumoral Treg cells is associated with a Technologies). Responder cells and Treg cells were combined to poor outcome in epithelial ovarian cancers (16). We therefore þ a total of 2.5 104 T cells at the indicated ratio of CD8 T cells aimed to expand the knowledge on intratumoral Treg cells by and Treg or control cells and activated with 5 mg/mL immobilized analyzing their expression of costimulatory and inhibitory recep- anti-CD3 (clone OKT3) in the presence of 7.5 104 T cell– tors in ovarian cancer. We initially observed that intratumoral þ þ þ depleted PBMCs. Cells were harvested 4 days later and dilution CD3 CD4 FOXP3 CTLA-4hi cells expressed PD-1 (Fig. 1A). We þ of the proliferation dye on live CD8 cells was analyzed by further observed that intermediate levels of PD-1 expression þ flow cytometry. Suppression was calculated using the following defined a distinct subset of FOXP3-expressing CD4 T cells þ formula: % suppression ¼ 1 [division index (responder þ (Fig. 1B). Importantly, circulating FOXP3 cells from patients suppressor cells)/division index (responder cells alone)] 100. with ovarian cancer did not express PD-1 (Fig. 1C). ICOS was also þ upregulated by tumor-infiltrating FOXP3 cells (Fig. 1D and E). Treg cell activation assay Therefore, Treg cells from ovarian cancer have upregulated Treg cells were FACS-purified as above and cultured at 2.0 104 TCR-induced genes, including PD-1 and ICOS. þ cells per well in a 96-well round-bottom plate. Tosyl-activated To assess whether PD-1intFOXP3 cells had regulatory func- Dynabeads (Thermo Fisher Scientific) were coated with anti-CD3 tion, we evaluated strategies to purify this population. The com- (clone OKT3) and anti-CD28 (CD28.2) either at a high concen- monly used combination of CD25 and CD127 (to isolate þ tration (10 mg each per 108 beads) or at a low concentration CD25 CD127lo cells) clearly had its limitations owing to the þ (1 mg each) according to the manufacturer's instructions, and Treg heterogeneity of CD25 expression by intratumoral FOXP3 cells cells were stimulated by the addition of 2 104 Dynabeads and (Supplementary Fig. S1A and S1B). Using PD-1 and ICOS þ þ 10 IU/mL rhIL2. One day after activation, cell surface expression we could distinguish three main subsets of CD3 CD4 TIL: of 4-1BB and CD3 was analyzed by flow cytometry. PD-1hiICOSint, PD-1intICOShi and PD-1 ICOS . Among these subsets FOXP3-expressing cells were contained within the RNA-seq PD-1intICOShi population (Supplementary Fig. S1C). In addition, þ RNA libraries were prepared using SMARTer Stranded Total the PD-1intICOShi population overlapped with FOXP3 cells to a þ RNA-seq Kit-Pico Input Mammalian (Clontech Laboratories). The similar extent as CD25 CD127lo (Fig. 1F–H). Importantly, gating paired-end libraries were sequenced on NextSeq 500 (Illumina) on PD1intICOShi did not introduce bias in favor of Treg cell for 75 cycles. RNA sequencing (RNA-seq) was performed by the subpopulations expressing CD25 or 4-1BB, whereas the þ Princess Margaret Genomics Centre (Toronto, Canada). Gene- CD25 CD127lo gate was enriched for cells that expressed CD25 level transcript expression and read counts were quantified using and 4-1BB (Supplementary Fig. S1D and S1E). Therefore, RSEM (v1.2.29) with Gencode transept annotation (v26). Differ- the use of CD25 for cell sorting could bias the analysis of þ entially expressed genes were identified using DESeq2 with intratumoral FOXP3 cells and this could be overcome by using default parameters to normalize and compare between gene-wise PD-1 and ICOS. read counts of defined sample groups. Genes with a minimum of Because human effector T cells can upregulate FOXP3 2-fold expression difference and FDR-corrected P-value <0.1 are following TCR engagement(10),wesoughttoconfirm the þ reported as statistically significant. Log-transformed gene expres- Treg cell identity of intratumoral FOXP3 cells directly through sion values [log2(TPMþ1)] were used to generate heat maps of functional assays as well as epigenetic and transcriptomic selected signature genes or differentially expressed genes. Princi- analyses. Tumor-derived PD-1intICOShi cells could suppress þ pal component analysis (PCA) was performed using library-size proliferation of CD8 T cells in a dose-dependent manner, normalized gene-wise read counts from the top 500 genes with whereas PD-1hiICOSint cells were devoid of suppressive capac- þ greatest variance. ity (Fig. 1I). FOXP3 Treg cells show DNA demethylation at the

www.aacrjournals.org Clin Cancer Res; 2018 OF3

Toker et al.

Figure 1. Tumor-infiltrating Treg cells in ovarian cancer express PD-1 and ICOS. A, Expression of PD-1 by CD3þCD4þFOXP3þCTLA-4hi TILs. B and C, Expression of PD-1 on FOXP3þ TILs was analyzed by flow cytometry. Representative staining on gated CD3þCD4þ TILs (B) and data summarizing PD-1 expression levels on FOXP3þ cells from PBMCs and TILs (C) are shown (n ¼ 10). D and E, Expression of ICOS on FOXP3þ TILs was analyzed by flow cytometry. Representative staining on gated CD3þCD4þ TILs (D) and data summarizing percentage of ICOS-expressing cells within the FOXP3þ gate in PBMCs and TILs (E) are shown (n ¼ 10). Error bars, SD. , P < 0.0001, Mann–Whitney U test. F, Gating strategies for purifying intratumoral FOXP3-expressing cells using PD-1 and ICOS (top) or CD25 and CD127 (bottom). Plots are gated on CD3þCD4þ. G, Percentage of FOXP3þ cells within the Treg cell gates as in F. H, ViSNE analysis on flow cytometry data of TILs gated on live singlets from 3 patients with ovarian cancer. Mapping by PD-1 and ICOS (top) or CD25 and CD127 (bottom), in addition to CD3, CD4, and CD8 but not FOXP3 (both). Colors indicate expression level of FOXP3. I, In vitro suppression by intratumoral PD-1intICOShi Treg cells. Pooled results from 3 patients are shown. J, TSDR methylation of the indicated intratumoral CD4þ T-cell subsets from 3 patients. Each box represents an individual CpG motif; each row represents one patient. Percentage of DNA methylation at each motif is color-coded according to the scale. K, Hierarchical clustering of genes differentially expressed by þ intratumoral CD4 T-cell subsets as determined by RNA-seq. Colors show row-normalized Z-score of gene expression values (log2 TPM þ 1).

OF4 Clin Cancer Res; 2018 Clinical Cancer Research

Treg Cells in Ovarian Cancer Are Highly Activated

Figure 2. Intratumoral Treg cells display a distinct expression pattern of costimulatory and inhibitory receptors and a distinct transcriptional signature. A–C, The expression of intracellular FOXP3 in PBMCs and TILs from patients with ovarian cancer was analyzed by ﬂow cytometry. Representative dot plots on gated CD3þ cells (A), summary data depicting the percentage of FOXP3þ cells within the CD3þ gate (B), and levels of FOXP3 expression (C) are shown (n ¼ 10). D–H, The expression of cell surface 4-1BB, TIM-3, OX40, and TIGIT as well as intracellular CTLA-4 by Treg cells in ovarian cancer specimens and peripheral blood was analyzed by ﬂow cytometry. Representative contour plots on CD3þCD4þ gated TILs (left) and the percentage of intratumoral and circulating FOXP3þ cells þ expressing the indicated markers (right) are shown (n ¼ 10). Expression levels of TIGIT on gated FOXP3 cells are shown in H, right. Error bars, SD. , P < 0.0001, Mann–Whitney U test. I, Hierarchical clustering of differentially expressed genes in intratumoral Treg cells from patients with ovarian cancer and circulating

Treg cells from healthy donors as determined by RNA-seq. Colors show row-normalized Z-score of gene expression values (log2 TPM þ 1).

Treg cell–specific demethylated region (TSDR) within the We next addressed the question whether expression of addi- FOXP3 locus, whereas activated effector cells that transiently tional TCR-induced receptors was different in intratumoral Treg þ upregulate FOXP3 remain methylated (20). Intratumoral cells compared with peripheral blood. The proportion of FOXP3 PD-1intICOShi cells displayed a demethylated TSDR (approxi- Treg cells within tumor-infiltrating T cells and FOXP3 expression mately 50% methylation due to X-chromosome hypermethyla- on a per cell basis was markedly increased compared with periph- tion in female patients), thus confirming that this subset con- eral blood (Fig. 2A–C). Moreover, expression of 4-1BB, CTLA-4, stitutes bona fide Treg cells with epigenetically imprinted stable and OX40 was higher on intratumoral Treg cells compared with FOXP3 expression, whereas the TSDR of PD-1hiICOSint and Treg cells from peripheral blood (Fig. 2D–F). TIM-3 was expressed PD-1-ICOS cells was methylated (Fig. 1J). Transcriptomic at elevated levels on intratumoral Treg cells in some patients but profiling by RNA-seq revealed that PD-1intICOShi cells had a showed high interpatient variability (39.5% 21.9% for intra- distinct gene expression profile (Fig. 1K; Supplementary tumoral Treg cells and 21.8% 12.8% for peripheral blood Table S2). We also analyzed Treg cells isolated from PBMC by Treg cells, Fig. 2G). The majority of circulating Treg cells were þ RNA-seq and found that although the overall gene expression TIGIT , but the proportion of TIGIT-expressing cells and TIGIT pattern was quite distinct from all intratumoral subsets, expres- expression levels were further enhanced in intratumoral Treg sion of canonical Treg cell signature genes was shared between cells (Fig. 2H). In line with these phenotypic changes, the gene circulating and intratumoral Treg cells (Supplementary Fig. S2). expression pattern of intratumoral Treg cells was distinct from Taken together, these results showed that tumor-infiltrating circulating Treg cells (Fig. 2I). Taken together, these results þ CD4 PD-1intICOShi T cells were bona fide Treg cells with potent demonstrate that Treg cells from tumors have substantial changes suppressive activity and a unique transcriptomic program. to their transcriptomic program and an altered immunoreceptor

www.aacrjournals.org Clin Cancer Res; 2018 OF5

Toker et al.

Figure 3. Expression of costimulatory and inhibitory receptors is most pronounced in FOXP3hi effector Treg cells. A, Categorization of CD4þ T cells into CD45RAþFOXP3lo na€ve Treg (nTreg) cells, CD45RAFOXP3hi effector Treg (eTreg) cells, and CD45RAFOXP3lo activated effector T (aTeff) cells. Plots show CD3þCD4þ gate, and boxed numbers indicate percentage of cells within the corresponding gate. B, Summary of results shown in A from 10 patients. C, Percentage of cells within the subsets deﬁned in A, which express the indicated receptors within TILs and PBMCs. A summary of 10 patients is shown. D, GMFI of PD-1 on the indicated subsets within TILs and PBMCs. E, Representative histogram showing the expression of PD-1 on intratumoral FOXP3hi Treg cells, FOXP3lo activated effector cells, and FOXP3 effector cells. Error bars, SD. , P < 0.01; , P < 0.001; , P < 0.0001, Mann–Whitney U test.

expression pattern. However, receptor upregulation occurred in a subsets, although overall PD-1 expression was lower by mass selective manner, as the percentage expressing CD27, CD28, and cytometry (Supplementary Fig. S4B). PD-1 ICOS resting T cells DNAM-1 was either reduced or not drastically changed (Supple- (Trest) did not express TIM-3 or TIGIT, whereas PD-1hiICOSint– mentary Fig. S3). activated T cells (Tact) and PD-1intICOShi Treg cells were TIM- We next examined the prevalence of na€ve and effector Treg 3loTIGIThi (Fig. 4A). Trest cells also expressed the lowest levels of cells and the surface phenotype of these Treg cell subpopulations CD27, CD28, and Ki-67 among the three subpopulations. In within ovarian tumors. Miyara and colleagues have categorized contrast, Tact and Treg cells expressed higher levels of these human FOXP3-expressing regulatory and nonregulatory cells molecules and expression of CD27 and CD28 was further elevated based on the expression levels of FOXP3 and CD45RA (21). We on Treg compared with Tact cells (Fig. 4A). In addition, 4-1BB was have found that CD45RA FOXP3hi effector Treg cells were the detectable only on Treg cells but none of the other subsets. In most prevalent subset within ovarian tumors, whereas contrast, expression levels of the chemokine receptors CXCR3 and þ CD45RA FOXP3lo na€ve Treg cells were essentially absent CCR5, the costimulatory molecule DNAM-1, the checkpoint (Fig. 3A and B). Within tumors both CD45RA FOXP3hi effector molecule PD-L1, and the transcription factor T-bet were similar Treg cells and CD45RA FOXP3lo effector T cells expressed 4-1BB, in all subsets. Moreover, all subsets lacked the expression of the ICOS, OX40, and CTLA-4, but expression of these markers on the inhibitory receptor BTLA and CD160, but were uniformly þ þ former subset was higher (Fig. 3C). Strikingly, intratumoral CD45RA CD45RO CD69 (Fig. 4A and data not shown). These FOXP3hi cells expressed intermediate levels of PD-1, whereas the findings demonstrate that different subsets of T cells found FOXP3lo subset was either PD-1hi or PD-1 , akin to FOXP3 cells within the tumor microenvironment can be distinguished by the (Fig. 3D and E), suggesting that bona fide FOXP3hi effector expression of markers associated with T-cell activation or Treg cells display a characteristic cell surface phenotype within exhaustion. ovarian tumors that is distinct from FOXP3lo-activated effector To further explore the activated Treg cell phenotype, we ana- cells. Therefore, the gating strategy using FOXP3 and CD45RA lyzed our mass cytometry data using dimensionality reduction shows that FOXP3hiCD45RA Treg cells in the tumor express algorithms. In line with an activation-dependent tuning of the þ higher levels of 41BB, ICOS, OX40, CTLA4, and intermediate tumor-infiltrating CD4 T-cell phenotype, ViSNE analysis levels of PD-1 compared with the FOXP3lo-activated effector revealed that Tact and Treg cells exhibited more similarities on population. a global level than either subset did with Trest cells (Fig. 4B). þ Notably, prominent populations of actively proliferating Ki-67 Mass cytometry analysis reveals a partially overlapping cells fell within the regions of the maps that coexpressed phenotype of Treg cells and conventional T-cell subsets Tact- and Treg cell-associated markers (Fig. 4B); hence, actively þ The distinct expression pattern of costimulatory and inhibitory proliferating CD4 T-cell subsets were characterized by the receptors by tumor-infiltrating Treg cells prompted us to further coexpression of several costimulatory and inhibitory receptors explore their phenotype. To this end, we performed in-depth regardless of suppressive function. In summary, these findings þ analyses of CD4 T-cell subsets in ovarian cancer specimens by further substantiated the high prevalence of T cells with an 35-parameter mass cytometry (Supplementary Fig. S4A; Supple- activated phenotype within the tumor, and the phenotype of mentary Table S3). Similar to our flow cytometry analyses, we intratumoral Treg cells in particular was indicative of a highly could detect PD-1hiICOSint, PD-1intICOShi, and PD-1 ICOS activated state.

OF6 Clin Cancer Res; 2018 Clinical Cancer Research

Treg Cells in Ovarian Cancer Are Highly Activated

Figure 4. Mass cytometry analysis reveals partially overlapping expression of immune markers by intratumoral regulatory and nonregulatory T cells. A, Thirty- ﬁve–parameter mass cytometry was performed on ovarian cancer specimens. Heatmaps show expression of the indicated markers by CD4þ T-cell subsets from 4 patients. Similar results were obtained in two independent experiments (n ¼ 8). B, ViSNE analysis of mass cytometry data from A. Data from 3 patients were used to run the algorithm on CD3þCD4þ cells; ViSNE maps are shown for one representative patient. Expression of markers shown above each individual map are highlighted according to the color scheme. Gates indicate Treg, Tact, and Trest cell areas of the maps as depicted in the top left map.

Melanoma-infiltrating Treg cells exhibit distinct phenotypic ovarian cancer and melanoma specimens and discovered two þ differences from ovarian tumor–infiltrating Treg cells subpopulations of intratumoral FOXP3 Treg cells. One of these We next examined whether this highly activated Treg cell subpopulations expressed higher levels of FOXP3, 4-1BB, ICOS, phenotype could also be found in cancers that respond better to TIGIT, PD-1, CD25, TIM-3, and CD28 and is therefore referred to þ immunotherapy and thus extended our CD4 Treg cell analyses to as Treg-hi, whereas the other subpopulation with lower expres- melanoma specimens. ViSNE analysis of mass cytometry data sion of all of these markers is referred to as Treg-lo (Fig. 5C; revealed that melanoma-infiltrating Treg cells formed a more Supplementary Fig. S5A and S5B and data not shown). In line heterogeneous population (Fig. 5A and B). This heterogeneity with this observation, expression of select markers including became evident from the relatively diffuse distribution pattern of 4-1BB, PD-1, and ICOS was increased in ovarian cancer Treg cells, þ melanoma-infiltrating Treg cells over the entire CD4 T-cell although the frequency of tumor infiltration by Treg cells among region on the ViSNE map, whereas ovarian cancer Treg cells total T cells was not different from melanoma (Fig. 5D–G; formed more distinct clusters. Among the dominant markers that Supplementary Fig. S5C and S5D). Accordingly, the Treg-hi defined this Treg cell cluster in both cancers were ICOS, TIGIT, and subpopulation comprised a higher percentage of T cells and the KI-67 (Fig. 5A). To further elucidate the Treg cell heterogeneity, we ratio of Treg-hi to Treg-lo was significantly increased within performed SPADE analysis on the same mass cytometry dataset of ovarian tumors compared with melanoma (Fig. 5H and I). Taken

www.aacrjournals.org Clin Cancer Res; 2018 OF7

Toker et al.

Figure 5. Mass cytometry analysis reveals two subpopulations of intratumoral Treg cells. A and B, ViSNE analysis of mass cytometry data on ovarian cancer and melanoma specimens. ViSNE was run on CD3þ cells either excluding (A) or including FOXP3 (B) from mapping parameters as described in the Materials and Methods section. Expressions of markers shown above each individual map are highlighted according to the color scheme. Maps are gated on CD3þ cells (left panels showing CD4 expression) or FOXP3þ cells (remaining panels). Clustering was performed on data from 3 patients, and ViSNE maps from one patient are shown. C, Mass cytometry data from ovarian cancer and melanoma specimens were pregated on CD3þ cells and analyzed by SPADE. Expression of ICOS is highlighted according to the rainbow scale in each individual tree. Data from 4 patients were used to generate SPADE trees; data from one representative patient per group are shown. Similar results were obtained in two independent experiments (n ¼ 4). D, Percentage of intratumoral FOXP3þ Treg cells among CD3þ cells in ovarian cancer and melanoma. E, GMFI of FOXP3 expression by Treg cells (n ¼ 10). F, Percentage of 4-1BBþ cells among Treg cells in ovarian cancer and melanoma. G, GMFI of the indicated markers on gated Treg cells in ovarian cancer and melanoma was analyzed by ﬂow cytometry (n ¼ 10). H, The relative abundance of cells within the Treg-hi and Treg-lo clusters as deﬁned by SPADE analysis in ovarian cancer and melanoma is depicted as frequency within CD3þ cells. I, The ratio of cells within the Treg-hi and Treg-lo clusters in ovarian cancer and melanoma (n ¼ 4). Error bars, SD. , P < 0.0001; , P < 0.01; , P < 0.05; Mann–Whitney U test.

together, these findings showed that although ovarian tumors and The Treg-hi phenotype defines highly responsive Treg cells þ melanomas are infiltrated by similar proportions of FOXP3 Treg with superior suppressive capacity cells, these infiltrates were qualitatively different. Importantly, We next sought to characterize the implications of the Treg-hi these differences did not reflect global phenotypic differences of and Treg-lo phenotype on the immune regulatory function of Treg þ the T-cell infiltrate as PD-1 expression by CD8 T cells was not cells. To this end, we selected two receptors that showed marked different (Supplementary Fig. S6). differential expression between the two Treg cell subpopulations.

OF8 Clin Cancer Res; 2018 Clinical Cancer Research

Treg Cells in Ovarian Cancer Are Highly Activated

their functional capacity in an in vitro suppression assay. As shown þ in Fig. 6C, 4-1BB Treg cells had superior suppressive capacity compared with patient-matched 4-1BB Treg cells. In agreement with these findings, bulk Treg cells purified from ovarian tumors had enhanced suppressive capacity when compared with melanoma-infiltrating Treg cells (Fig. 6D). These results showed that the immunologic receptor expression pattern including increased 4-1BB, PD-1, and ICOS expression defines highly activated and suppressive Treg cells that infiltrate ovarian tumors.

Discussion Tumor-infiltrating Treg cells have an activated phenotype Peripheral blood Treg cells are a heterogeneous population that can be subdivided into several partially overlapping subsets by the differential expression CD45RA, CCR4, CD39, CD25, and others (19, 21). Circulating Treg cells also express some costimulatory and inhibitory receptors such as CD27 and CD28, but many others like ICOS, OX40, 4-1BB, TIM3, and PD-1 are absent or expressed at low levels ex vivo (19, 22). In contrast, tumor- infiltrating Treg cells express CD45RO and increased levels of the effector molecules CD39 and CD73, indicating previous antigen experience and an elevated activation state (23). However, other aspects of the intratumoral Treg cell phenotype with potential clinical relevance, such as expression of costimulatory and inhibitory receptors, have not been comprehensively studied. We report that the array of such immunologic receptors expressed by intratumoral Treg cells is strikingly different from Figure 6. circulating Treg cells, suggesting that substantial phenotypical Treg cells expressing PD-1 and 4-1BB are highly responsive to stimulation changes occur concomitantly with tumor infiltration. Our find- þ int hi and have superior suppressive capacity. A, CD4 PD-1 ICOS Treg cells ings are supported by two recent studies that showed that Treg fi were puri ed from ovarian tumors and stimulated with a low cells isolated from breast cancer, non–small cell lung cancer, and concentration or high concentration of immobilized anti-CD3/anti-CD28 antibodies or left unstimulated. Expression of CD3 and 4-1BB was evaluated colorectal cancer have distinct transcriptomic signatures com- after 24 hours by flow cytometry. B, Linear regression analysis (n ¼ 10) pared with peripheral blood Treg cells (24, 25). Genes upregu- of Treg cells stimulated as in A. The difference in 4-1BB expression lated in the tumor included costimulatory and cytokine receptors between high- and low-dose CD3/CD28 stimulation is shown in relation as well as trafficking and effector molecules, suggesting that to ex vivo PD-1 expression levels. C, The suppressive capacity of tumor-infiltrating Treg cells share common features across can- þ int hi þ CD4 PD-1 ICOS 4-1BB and 4-1BB Treg cells from ovarian tumors was cers. Our transcriptomic and mass cytometry analyses show that assessed in vitro (n ¼ 4). , P < 0.01; , P < 0.05, unpaired t test. D, The many of these features are in fact shared between Treg cells and suppressive capacity of Treg cells purified from ovarian cancer and þ melanoma tissues was assessed in vitro. The ratio of responder T cells to CD4 effector T cells, suggesting that a large proportion of tumor- þ Treg cells was 1:2 (n ¼ 7–8). , P < 0.05, Mann–Whitney U test. infiltrating CD4 T cells are activated and potentially involved in antitumor immunity. These observations are in line with a recent study that showed that tissue-specific transcriptomic signatures þ First, we correlated the responsiveness of Treg cells to TCR stim- from CD4 T-cell subsets are dominant over lineage-specific (i.e., ulation with their PD-1 expression levels. We purified Treg cells regulatory vs. effector) signatures in liver cancer (26). However, in from ovarian tumors and cultured them for one day either spite of the common tumor-associated gene expression profile, without TCR stimulation or with high or low concentration of key cell subset-specific features such as suppressive function immobilized anti-CD3/anti-CD28 antibodies. A minor fraction remain intact and may even be enhanced within the tumor of unstimulated Treg cells expressed 4-1BB after one day, whereas microenvironment. Indeed, we show that elevated expression of 4-1BB expression increased after stimulation in a concentration- PD-1 and 4-1BB by intratumoral Treg cells correlated with respon- dependent manner along with downregulation of the TCR com- siveness to TCR stimulation and suppressive capacity. In agree- plex (Fig. 6A). Importantly, Treg cells expressing relatively high ment with our findings, previous reports have shown that the levels of PD-1 were more readily responsive to a low dose of TCR expression of PD-1 and TIM-3 correlated with increased IL10 þ þ stimulation, such that upregulation of 4-1BB was not substan- production by Treg cells (27). In addition, TIGIT TIM-3 Treg tially more pronounced at the high-dose TCR stimulation cells exhibit enhanced immunosuppression in a murine tumor (Fig. 6B). These results suggested that PD-1 expression levels model (28), and the presence of a TIM-3–expressing subset of could be used as a correlate of activation state in tumor-infiltrating tumor-infiltrating Treg cells has been associated with progression Treg cells. in non–small cell lung cancer (22). Furthermore, infiltration of þ We next investigated the significance of ex vivo 4-1BB expression primary cutaneous squamous cell carcinomas by OX40 Treg cells þ on the functional capacity of Treg cells. We purified 4-1BB and was associated with a higher incidence of metastasis (29), 4-1BB Treg cells from ovarian cancer specimens and assessed suggesting that a highly activated intratumoral Treg cell

www.aacrjournals.org Clin Cancer Res; 2018 OF9

Toker et al.

phenotype is defined by both activating and inhibitory receptors Treg cells. It is desirable that Treg cell depleting or modulating that are induced by TCR stimulation. These findings are all immunotherapies spare circulating Treg cells with broad consistent with the interpretation that TCR-triggered Treg cells self-antigen specificity, thereby avoiding a systemic breach as defined by expressing one or more of these markers including in self-tolerance and the occurrence of autoimmunity (5). PD-1, 41BB, TIM3, TIGIT, or OX-40 have increased immunosup- Depletion of Treg cells by targeting the chemokine receptor pressive function. CCR4 can induce melanoma antigen-specificimmune responses in vitro (37), but the mechanism of action as well Treg cells exhibit distinct features in ovarian cancer and as the specificity of this approach to intratumoral Treg cells melanoma remain incompletely understood. Several studies in animal Although many studies investigating intratumoral Treg cell models have shown that intratumoral Treg cells can be selec- phenotype were focused on the expression of individual receptors tively depleted by antibodies that target CTLA-4, GITR, and in different cancers, a high degree of overlap in transcriptomic OX40, owing to the elevated expression of these markers by signatures from tumor-infiltrating Treg cells across cancers sug- tumor-infiltrating Treg cells (38). It has been suggested that the gests that some factors that shape the Treg cell phenotype are mechanism of action of ipilimumab in patients with melano- shared between malignancies (24–26). However, comparative ma may involve depletion of CTLA-4 expressing Treg cells transcriptomic analyses also showed that Treg cells in hepatocel- through antibody-dependent cell-mediated cytotoxicity by lular carcinoma, non–small cell lung cancer, and breast cancer nonclassical monocytes (39). Our observation that intratu- exhibit partially distinct gene expression programs (26), suggest- moral Treg cells express elevated levels of CTLA-4 provides a ing that substantial heterogeneity among Treg cells infiltrating potential mechanism for their selective depletion after ipili- different cancers may exist beyond a common tumor-associated mumab treatment. core gene expression signature. We now extend on these studies by Our studies suggest that immunotherapies targeting other directly showing that a variety of costimulatory and coinhibitory molecules expressed by intratumoral Treg cells such as 4-1BB, molecules are differentially expressed in ovarian cancer and OX40, and ICOS may act by depleting or modulating Treg melanoma. cells. For instance, triggering ICOS on Treg cells through Treg cells infiltrating ovarian cancers displayed a Treg-hi ICOSL-expressing tumor cells or plasmacytoid dendritic cells phenotype that was characterized by higher expression of FOXP3, boosts suppressive capacity (40, 41); hence, blockade of this PD-1, 4-1BB, ICOS, CD25, and other markers in comparison pathway could attenuate Treg cell function. In contrast, stud- with the Treg-lo phenotype, which was enriched in melanoma. ies with murine Treg cells have shown that signaling through Individual features of the Treg-lo phenotype have previously the inhibitory receptors TIGIT and PD-1 can attenuate their been associated with increased plasticity and effector T-cell func- suppressive function (28, 42). Therefore, blockade of inhib- þ tion. For instance, FoxP3 CD25 murine Treg cells are prone itory receptors may have the unwanted side effect of increased to conversion to effector T cells in response to inflammatory Treg cell activity. These considerations are particularly impor- cues (30). Indeed, a substantial fraction of melanoma antigen- tant because many mAbs that target checkpoint receptors that þ specific FOXP3 cells can secrete Th1-type cytokines following are coexpressed by effector T cells and Treg cells are currently immunotherapy in patients with melanoma (31). Moreover, approved or undergoing clinical evaluation for the treatment þ CD45RO FOXP3lo effector T cells with increased capacity for of cancer (3, 4). Such immunomodulating agents often act IFNg and TNFa production accumulate in the peripheral blood of through multiple mechanisms, and in most cases the þ patients with melanoma (32). Infiltration by CD4 FOXP3lo cells relative contribution of each mechanism to the therapeutic is associated with improved antitumor immunity and survival in effect is unknown (43). Development of immune therapeutic colorectal cancer (33). Therefore, a higher prevalence of the Treg- approaches will therefore require careful consideration of the lo phenotype and reduced in vitro suppressive capacity suggests effect of the manipulating agent on both effector T cells and þ that FOXP3 cells in melanoma are heterogeneous, and include Treg cells. nonregulatory cells and plastic Treg cells with effector T-cell Conventional chemotherapies can have profound effects on potential. Importantly, melanoma-infiltrating Treg cells dis- the immune system and alter the composition of immune cells þ þ played a T-bet CXCR3 tumor migratory phenotype and con- within the tumor microenvironment (44). Chemotherapy fre- þ tained a Ki-67 subset, suggesting that they were specifically quently results in immunogenic cancer cell death that is accom- recruited to the tumor and recognizing tumor antigen in situ. panied by release of antigens and antigen-presenting cell mat- Such tumor-specific cells with reduced suppressive function and uration signals (44). Moreover, neoadjuvant chemotherapy proinflammatory potential could play a beneficial role in an induces NF-kB–driven upregulation of MHC class I and PD-L1 immunotherapeutic setting and contribute to the comparatively by cancer cells and increased T-cell infiltration of ovarian high response rates in melanoma (2, 34). tumors (45–47); hence, the idea of reinforcing chemotherapy-induced tumor-specific immune responses through check- The potential of Treg cell–targeted immunotherapies point blockade has gained much interest (48). Some conven- Extensive immunosuppression owing to the Treg-hi tional chemotherapeutic agents such as cyclophosphamide and phenotype in ovarian cancer described here as well as TAM docetaxel can also selectively reduce Treg cells (49, 50), and (7) or innate lymphoid cells (35) could explain the high reduction of intratumoral Treg cells after neoadjuvant chemo- resistance of ovarian cancer to immunotherapy despite pro- therapy is associated with improved responses (47), suggesting þ nounced CD8 T-cell infiltration (2, 36). Targeting Treg cell– that a window of opportunity may exist where CD8-targeted suppressive function has gained much interest as an avenue immunotherapies could exploit a temporary reduction in to promote tumor immunotherapy, but challenges remain immunosuppressive Treg cells following chemotherapy. Alter- such as the need to selectively manipulate tumor-infiltrating natively, combining chemotherapies that deplete MDCSs (44)

OF10 Clin Cancer Res; 2018 Clinical Cancer Research

Treg Cells in Ovarian Cancer Are Highly Activated

with Treg cell–targeted immunotherapies may cause a more Acquisition of data (provided animals, acquired and managed patients, profound ablation in immunosuppressive cell populations and provided facilities, etc.): P.A. Shaw, D. Ghazarian, A. Al-Habeeb, A. Easson, create a tumor microenvironment that is highly favorable for W.L. Leong, M. Reedijk, C.J. Guidos, T.J. Pugh, M.Q. Bernardini Analysis and interpretation of data (e.g., statistical analysis, biostatistics, com- antitumor immunity. putational analysis): A. Toker, S.Y.C. Yang, B.A. Clarke, C.J. Guidos, T.J. Pugh The search for rational combination (immuno)-therapies Writing, review, and/or revision of the manuscript: A. Toker, S.Y.C. Yang, from an ever-growing pool of available options needs to take S.R. Katz, A. Al-Habeeb, A. Easson, D.R. McCready, M. Reedijk, T.J. Pugh, into account several factors including the expression pattern of P.S. Ohashi the target and its potential mechanism of action. We now report Administrative, technical, or material support (i.e., reporting or organizing the expression of various costimulatory and coinhibitory recep- data, constructing databases): S.R. Katz, M.Q. Bernardini Study supervision: T.J. Pugh, P.S. Ohashi tors by Treg cells is highly relevant to therapeutic approaches Other (conducted experiments): S.C. Stone currently in use or under development, adding another facet to the potential mechanism of action of these agents. The possi- bility of Treg cell modulation is therefore an additional factor Acknowledgments that needs to be taken into consideration in the design of This work was supported by a CIHR Foundation award to P.S. Ohashi and a TRI award sponsored by OICR to P.S. Ohashi and the Garron Family Cancer combination therapies. Centre to C.J. Guidos. We thank the Flow and Mass Cytometry Facility (FMCF), The Hospital for Sick Children, Toronto, Canada, for assistance with cell sorting Disclosure of Potential Conflicts of Interest and mass cytometry data acquisition. The authors thank Diana Gray for T.J. Pugh reports receiving other commercial research support from providing patient data. They also thank the patients and their families for Boehringer Ingelheim and speakers bureau honoraria from Merck, and is agreeing to provide the tissue for research. a consultant/advisory board member for Chrysalis Biomedical Advisors and DynaCare/Impact Genetics. M.Q. Bernardini is a consultant/advisory board The costs of publication of this article were defrayed in part by the fl member for Astra Zeneca. No potential con icts of interest were disclosed by payment of page charges. This article must therefore be hereby marked the other authors. advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Authors' Contributions Conception and design: A. Toker, L.T. Nguyen, P.S. Ohashi Received February 15, 2018; revised June 18, 2018; accepted July 26, 2018; Development of methodology: A. Toker, C.J. Guidos, T.J. Pugh published first July 31, 2018.

References 1. Ott PA, Hodi FS, Kaufman HL, Wigginton JM, Wolchok JD. Combination 15. Adeegbe DO, Nishikawa H. Natural and induced T regulatory cells in immunotherapy: a road map. J Immunother Cancer 2017;5:16. cancer. Front Immunol 2013;4:190. 2. Gaillard L S, Secord AA, Monk B. The role of immune checkpoint 16. Curiel TJ, Coukos G, Zou L, Alvarez X, Cheng P, Mottram P, et al. inhibition in the treatment of ovarian cancer. Gynecol Oncol Res Pract Specific recruitment of regulatory T cells in ovarian carcinoma fosters 2016;3:11. immune privilege and predicts reduced survival. Nat Med 2004; 3. Pardoll DM. The blockade of immune checkpoints in cancer immuno- 10:942–9. therapy. Nat Rev Cancer 2012;12:252–64. 17. Klages K, Mayer CT, Lahl K, Loddenkemper C, Teng MWL, Ngiow SF, et al. 4. Swart M, Verbrugge I, Beltman JB. Combination approaches with immune- Selective depletion of Foxp3þ regulatory T cells improves effective ther- checkpoint blockade in cancer therapy. Front Oncol 2016;6:233. apeutic vaccination against established melanoma. Cancer Res 2010; 5. Tanaka A, Sakaguchi S. Regulatory T cells in cancer immunotherapy. Cell 70:7788–99. Res 2017;27:109–18. 18. Teng MWL, Ngiow SF, Von Scheidt B, McLaughlin N, Sparwasser T, Smyth 6. Gabrilovich DI. Myeloid-derived suppressor cells. Cancer Immunol Res MJ. Conditional regulatory T-cell depletion releases adaptive immunity 2017;5:3–8. preventing carcinogenesis and suppressing established tumor growth. 7. Worzfeld T, Pogge von Strandmann E, Huber M, Adhikary T, Wagner U, Cancer Res 2010;70:7800–9. Reinartz S, et al. The unique molecular and cellular microenvironment of 19. Mason GM, Lowe K, Melchiotti R, Ellis R, de Rinaldis E, Peakman M, et al. ovarian cancer. Front Oncol 2017;7:24. Phenotypic complexity of the human regulatory T cell compartment 8. Ohkura N, Kitagawa Y, Sakaguchi S. Development and maintenance of revealed by mass cytometry. J Immunol 2015;195:2030–7. regulatory T cells. Immunity 2013;38:414–23. 20. Wieczorek G, Asemissen A, Model F, Turbachova I, Floess S, Liebenberg V, 9. Miyara M, Sakaguchi S. Human FoxP3(þ)CD4(þ) regulatory T cells: their et al. Quantitative DNA methylation analysis of FOXP3 as a new method knowns and unknowns. Immunol Cell Biol 2011;89:346–51. for counting regulatory T cells in peripheral blood and solid tissue. Cancer 10. Gavin MA, Torgerson TR, Houston E, DeRoos P, Ho WY, Stray-Pedersen A, Res 2009;69:599–608. et al. Single-cell analysis of normal and FOXP3-mutant human T cells: 21. Miyara M, Yoshioka Y, Kitoh A, Shima T, Wing K, Niwa A, et al. FOXP3 expression without regulatory T cell development. Proc Natl Acad Functional delineation and differentiation dynamics of human CD4þ Sci U S A 2006;103:6659–64. T cells expressing the FoxP3 transcription factor. Immunity 2009;30: 11. Fontenot JD, Dooley JL, Farr AG, Rudensky AY. Developmental 899–911. regulation of Foxp3 expression during ontogeny. J Exp Med 2005; 22. Gao X, Zhu Y, Li G, Huang H, Zhang G, Wang F, et al. TIM-3 expression 202:901–6. characterizes regulatory T cells in tumor tissues and is associated with lung 12. Toker A, Engelbert D, Garg G, Polansky JK, Floess S, Miyao T, et al. cancer progression. PLoS One 2012;7:e30676. Active demethylation of the Foxp3 locus leads to the generation 23. Liu C, Workman CJ, Vignali DAA. Targeting regulatory T cells in tumors. of stable regulatory T cells within the thymus. J Immunol 2013; FEBS J 2016;283:2731–48. 190:3180–8. 24. Plitas G, Konopacki C, Wu K, Bos PD, Morrow M, Putintseva EV, et al. 13. Dominguez-Villar M, Baecher-Allan CM, Hafler DA. Identification of T Regulatory T cells exhibit distinct features in human breast cancer. helper type 1-like, Foxp3þ regulatory T cells in human autoimmune Immunity 2016;45:1122–34. disease. Nat Med 2011;17:673–5. 25. De Simone M, Arrigoni A, Rossetti G, Gruarin P, Ranzani V, Politano C, 14. Beriou G, Costantino CM, Ashley CW, Yang L, Kuchroo VK, Baecher-allan et al. Transcriptional landscape of human tissue lymphocytes unveils C, et al. IL-17 producing human peripheral regulatory T cells retain uniqueness of tumor-infiltrating T regulatory cells. Immunity 2016;45: suppressive function. Blood 2009;113:4240–50. 1135–47.

www.aacrjournals.org Clin Cancer Res; 2018 OF11

Toker et al.

26. Zheng C, Zheng L, Yoo JK, Guo H, Zhang Y, Guo X, et al. Landscape 38. Chao JL, Savage PA. Unlocking the complexities of tumor-associated of infiltrating T cells in liver cancer revealed by single-cell sequencing. regulatory T cells. J Immunol 2018;200:415–21. Cell 2017;169:1342–56. 39. Romano E, Kusio-Kobialka M, Foukas PG, Baumgaertner P, Meyer C, 27. Chew V, Lai L, Pan L, Lim CJ, Li J, Ong R, et al. Delineation of an Ballabeni P, et al. Ipilimumab-dependent cell-mediated cytotoxicity of immunosuppressive gradient in hepatocellular carcinoma using high- regulatory T cells ex vivo by nonclassical monocytes in melanoma patients. dimensional proteomic and transcriptomic analyses. Proc Natl Acad Sci Proc Natl Acad Sci U S A 2015;112:6140–5. U S A 2017;114:E5900–9. 40. Martin-Orozco N, Li Y, Wang Y, Liu S, Hwu P, Liu Y. Melanoma cells express 28. Kurtulus S, Sakuishi K, Ngiow SF, Joller N, Tan DJ, Teng MWL, et al. TIGIT ICOS ligand to promote the activation and expansion of T-regulatory cells. predominantly regulates the immune response via regulatory T cells. J Clin Cancer Res 2010;70:9581–90. Invest 2015;125:4053–62. 41. Faget J, Bendriss-Vermare N, Gobert M, Durand I, Olive D, Biota C, et al. 29. Lai C, August S, Albibas A, Behar R, Cho SY, Polak ME, et al. Ox40þ ICOS-ligand expression on plasmacytoid dendritic cells supports breast regulatory T cells in cutaneous squamous cell carcinoma suppress effector cancer progression by promoting the accumulation of immunosuppressive T-cell responses and associate with metastatic potential. Clin Cancer Res CD4þ T cells. Cancer Res 2012;72:6130–41. 2016;22:4236–48. 42. Sage PT, Francisco LM, Carman CV, Sharpe AH. The receptor PD-1 controls 30. Komatsu N, Mariotti-Ferrandiz ME, Wang Y, Malissen B, Waldmann H, follicular regulatory T cells in the lymph nodes and blood. Nat Immunol Hori S. Heterogeneity of natural Foxp3þ T cells: a committed regulatory 2013;14:152–62. T-cell lineage and an uncommitted minor population retaining plasticity. 43. Nguyen LT, Ohashi PS. Clinical blockade of PD1 and LAG3 — potential Proc Natl Acad Sci U S A 2009;106:1903–8. mechanisms of action. Nat Rev Immunol 2015;15:45–56. 31. Jandus C, Bioley G, Dojcinovic D, Derre L, Baitsch L, Wieckowski S, et al. 44. Galluzzi L, Buque A, Kepp O, Zitvogel L, Kroemer G. Immunological effects Tumor antigen-specific FOXP3þ CD4 T cells identified in human meta- of conventional chemotherapy and targeted anticancer agents. Cancer Cell static melanoma: Peptide vaccination results in selective expansion of 2015;28:690–714. Th1-like counterparts. Cancer Res 2009;69:8085–93. 45. Peng J, Hamanishi J, Matsumura N, Abiko K, Murat K, Baba T, et al. 32.FujiiH,ArakawaA,KitohA,MiyaraM,KatoM,Kore-EdaS,etal. Chemotherapy induces programmed cell death-ligand 1 overexpression Perturbations of both nonregulatory and regulatory FOXP3þ T cells via the nuclear factor-kBto foster an immunosuppressive tumor microen- in patients with malignant melanoma. Br J Dermatol 2011;164: vironment in Ovarian Cancer. Cancer Res 2015;75:5034–45. 1052–60. 46. Lo CS, Sanii S, Kroeger DR, Milne K, Talhouk A, Chiu DS, et al. 33. Saito T, Nishikawa H, Wada H, Nagano Y, Sugiyama D, Atarashi K, et al. Neoadjuvant chemotherapy of ovarian cancer results in three patterns Two FOXP3þCD4þ T cell subpopulations distinctly control the prognosis of tumor-infiltrating lymphocyte response with distinct implications for of colorectal cancers. Nat Med 2016;22:679–84. immunotherapy. Clin Cancer Res 2017;23:925–34. 34. Ott PA, Hodi FS, Robert C. CTLA-4 and PD-1/PD-L1 blockade: new 47. Bohm€ S, Montfort A, Pearce OMT, Topping J, Chakravarty P, Everitt GLA, immunotherapeutic modalities with durable clinical benefit in melanoma et al. Neoadjuvant chemotherapy modulates the immune microenviron- patients. Clin Cancer Res 2013;19:5300–9. ment in metastases of tubo-ovarian high-grade serous carcinoma. 35. Crome SQ, Nguyen LT, Lopez-Verges S, Yang SYC, Martin B, Yam JY, et al. Clin Cancer Res 2016;22:3025–36. A distinct innate lymphoid cell population regulates tumor-associated T 48. Mathew M, Enzler T, Shu CA, Rizvi NA. Combining chemotherapy with cells. Nat Med 2017;23:368–75. PD-1 blockade in NSCLC. Pharmacol Ther 2018;186:130–7. 36. Webb JR, Milne K, Kroeger DR, Nelson BH. PD-L1 expression is associated 49. Ghiringhelli F, Menard C, Puig PE, Ladoire S, Roux S, Martin F, et al. with tumor-infiltrating T cells and favorable prognosis in high-grade serous Metronomic cyclophosphamide regimen selectively depletes CD4þCD25þ ovarian cancer. Gynecol Oncol 2016;141:293–302. regulatory T cells and restores T and NK effector functions in end stage 37. Sugiyama D, Nishikawa H, Maeda Y, Nishioka M, Tanemura A, Katayama I, cancer patients. Cancer Immunol Immunother 2007;56:641–8. et al. Anti-CCR4 mAb selectively depletes effector-type FoxP3þCD4þ 50. Roselli M, Cereda V, Giovanna di Bari M, Formica V, Spila A, Jochems C, regulatory T cells, evoking antitumor immune responses in humans. et al. Effects of conventional therapeutic interventions on the number and Proc Natl Acad Sci U S A 2013;110:17945–50. function of regulatory T cells. Oncoimmunology 2013;2:e27025.

OF12 Clin Cancer Res; 2018 Clinical Cancer Research

Regulatory T Cells in Ovarian Cancer Are Characterized by a Highly Activated Phenotype Distinct from that in Melanoma

Aras Toker, Linh T. Nguyen, Simone C. Stone, et al.

Clin Cancer Res Published OnlineFirst July 31, 2018.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-18-0554

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2018/07/31/1078-0432.CCR-18-0554.DC1

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/early/2018/09/14/1078-0432.CCR-18-0554. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.