PD-1 blockade induces durable responses in patients with metastatic melanoma and prolongs relapse-free survival in patients with resected melanoma; however, current biomarkers do not consistently associate with patient responses. In this study, we investigated the impact of nivolumab therapy on peripheral blood regulatory T cells (Treg) and its relation to patient outcomes.
Peripheral blood Tregs and conventional CD4+ T cells from patients with resected high-risk melanoma treated with adjuvant nivolumab were assessed for gene expression changes by RNA-seq. Percentages of circulating Tregs and phosphorylated-STAT3 (pSTAT3) expression levels were assessed by flow cytometry and validated in an independent cohort of active disease patients. Suppressive function of Tregs was assessed in allogeneic mixed lymphocyte reactions.
Tregs from non-relapse patients had increased expression of proliferation associated genes. An increase in the proportion of circulating Tregs and pSTAT3 expression and a reduction in Treg-suppressive capacity were observed in non-relapsing, but not relapsing patient samples 13 weeks after starting treatment. In vitro blockade of PD-1 increased Treg percentages and pSTAT3 expression, and reduced Treg-suppressive function. PD-1 blockade also led to IL10 production by T cells, resulting in higher Treg proliferation. The addition of a STAT3 inhibitor ameliorated the increase in Tregs, enhanced suppressive function, and decreased T-cell IL10 production in vitro.
These results demonstrate that induction of pSTAT3, reduced suppressive function, and a paradoxical increase in Treg proliferation are novel correlates of patient benefit from PD-1 blockade.
PD-1 blockade induces durable responses in patients with metastatic melanoma and prolongs relapse-free survival in patients with resected melanoma; however, biomarkers of response and clinical benefit remain unknown. The data presented herein support a role of regulatory T cells (Treg) in melanoma patient outcome after nivolumab therapy, demonstrating that induction of pSTAT3 in T cells and an accompanying reduction in Treg-suppressive function with a paradoxical increase in Treg percentages are novel correlates of patient benefit with PD-1 blockade.
PD-1–blocking antibodies (e.g., nivolumab) have been effective in the treatment of a variety of malignancies. Nivolumab treatment of unresectable metastatic melanoma has demonstrated response rates up to approximately 40% (1) and efficacy as adjuvant therapy for surgically resected metastatic melanoma (2). However, not all patients benefit from PD-1 blockade and biomarkers associated with patient outcome are unclear.
Ligation of PD-1–expressing T cells to PDL1 or PDL2 results in phosphorylation of SHP1/2 phosphatases, leading to ZAP70 dephosphorylation, RAS and PI3K signaling inhibition, BAFT activation, disruption of TCR and CD28 signaling, and, consequently, inhibition of T-cell activation, proliferation, and effector functions (3–6). PDL1 expression by melanoma cells is associated with poor prognosis (7). Tumor PDL1 expression has been shown to be a biomarker of response to PD-1 blockade (8), but studies have shown that patient benefit may still occur for PDL1-negative tumors (9).
Regulatory T cells (Treg) are a heterogeneous population of T cells responsible for maintaining self-tolerance (10), and have been shown to inhibit antitumor immune responses. Tregs function through a variety of mechanisms, including production of immunosuppressive cytokines, downregulation of co-stimulation by antigen-presenting cells (APC), and direct killing of effector T cells (11, 12). Naturally occurring Tregs (nTregs) develop in the thymus and are characterized as CD4+CD127low/−CD25+/FOXP3+ (13). Conventional CD4+ T cells (Tcons) can also acquire induced regulatory function that are termed iTregs or adaptive Tregs (13). TGFβ and IL10 produced in the tumor microenvironment promote the expansion of nTregs and conversion of Tcons into iTregs (14, 15). Tregs are overrepresented in the peripheral blood of metastatic melanoma patients (16), and increased ratios of Tregs to CD8+ T-cell tumor infiltrate correlate with reduced survival (17).
In ipilimumab-treated patients, higher baseline levels of circulating Tregs are associated with improved survival (18). However, the importance of Tregs in patient response to PD-1 blockade remains unclear. Previous work from our group showed that PD-1 blockade increased resistance of effector T cells to Treg suppression and directly reduced the suppressive function of Tregs in vitro (19). Herein, we sought to address the role of Tregs in resected metastatic melanoma patients after adjuvant treatment with the PD-1–blocking antibody nivolumab.
Materials and Methods
PBMC were harvested from leukapheresis of metastatic or resected melanoma patients enrolled in clinical trials evaluating nivolumab therapy (clinicaltrials.gov NCT01176474; ref. 2), NCT01176461 (ref. 9), NCT01783938 (ref. 20)], purified through Ficoll Histopaque density gradient (GE Healthcare), and cryopreserved. All patients provided written informed consent, and studies were conducted in accordance with good clinical practice and the Declaration of Helsinki. All protocols were approved by the Institutional Review Board at Moffitt Cancer Center and New York University Langone Medical Center. Samples were coded with an anonymized number and ex vivo analyses of patient samples were performed blinded. For all in vitro assays, baseline metastatic melanoma patient samples were used, except in nTreg suppression assays.
Recombinant human IL2 (Aldesleukin, Prometheus Laboratories) was commercially obtained. Recombinant human IL10 and TGFβ were purchased from R&D Systems. Stattic and rapamycin were obtained from Selleckchem. Reagents were prepared per the manufacture, stored at −80°C and thawed immediately before use. αPD-1, αCD3, αCD28, and IgG antibodies were obtained from Biolegend. αPDL1 antibody was purchased from BioXCell. Nivolumab was provided by Bristol-Myers Squibb. Antibodies were stored at 4°C.
Treg suppression assays
In experiments assessing Treg suppression, Tregs (CD3+CD4+CD127−/lowCD25+) from baseline and week 13 post-nivolumab therapy were flow sorted to >98% purity from patient PBMC and co-cultured with allogeneic negatively isolated CD8+ T cells (EasySep kit, StemCell Technologies) and irradiated PBMC. After 5 days, cells were pulsed with 1 μCi/well of tritiated thymidine for 18 hours and assessed for radioactive incorporation. Suppression was normalized against control cultures containing no Tregs (100% proliferation). Technical control wells contained only PBMC and Treg.
To assess PD-1/PDL1 blockade and STAT3 inhibition in vitro, nTregs were purified from healthy human donor PBMC by magnetic bead separation based on CD4 and CD25 expression to approximately 90% purity (Miltenyi). Isolated nTregs were expanded with Dynabeads Human T-Activator CD3/CD28 for T-Cell Expansion and Activation (ThermoFisher) at a nTreg to bead ratio of 1:1. RPMI culture medium was supplemented with 10% heat inactivated pooled human serum and recombinant human IL-2. Stattic (200 nmol/L), αPD-1 antibody (5 μg/mL), αPDL1 (10 μg/mL), DMSO (<0.01%), or IgG were added once on day 0 of expansion. After 3 days, nTregs were washed and plated with autologous CFSE labeled T cells at 1:30 ratio (Treg:T-cell) plus CD3/CD28 beads at 1:30 ratio (beads:T-cell). T-cell proliferation was assessed by CFSE dilution after 3 days of culture using a FACSCANTO II flow cytometer (BD Biosciences).
iTregs were generated by isolating naïve CD4+ T cells from baseline patient PBMC using an EasySep kit, and culturing in a 10 μg/mL OKT3 αCD3 antibody coated-plate plus 100 nmol/L rapamycin, 10 ng/mL TGFβ, 300IU/mL IL-2, and 2 μg/mL CD28.2 αCD28 antibodies. IgG, αPD-1, αPDL1 antibodies or stattic treatments were added as indicated. After 7 days, cells were washed, counted and co-cultured at indicated ratios with autologous CD8+ T cells (EasySep negative isolation) and 1:10 CD3/CD28 Dynabeads (bead:CD8+). After 3 days, Ki67 expression was assessed by flow cytometry. Values relative to Ki67 expression in CD8+ only groups (i.e., no Tregs) were graphed. In some experiments, supernatants were frozen for further cytokine evaluation by Luminex.
RNA-sequencing and gene set enrichment analysis
RNA-Seq was performed using the NuGen Ovation Encore Complete Kit. Briefly, RNA was extracted from sorted T cells using the Qiagen RNEasy Mini Kit, and 100 ng of RNA was used to generate cDNA and a strand-specific library following the manufacturer's protocol (NuGEN Technologies, Inc.). The libraries were then sequenced on the Illumina NextSeq 500 v2 sequencer with a 75-base paired-end run to generate 30 to 40 million read pairs per sample. Sequence reads were aligned to the human genome in a splice-aware fashion using Tophat2 (21). Aligned reads were then condensed into transcripts for calculation of differential expression at both the gene and transcript levels with the Cufflinks package (22). Transcript mapped to a single unique known gene, q < 0.05 and a known log2-fold change was considered significant when comparing two groups.
Surface staining was performed in FACS buffer (PBS, 2 mmol/L EDTA, 2% FBS), for 30 minutes at 4°C. Antibodies were obtained from BD Biosciences, Biolegend, eBioscience, or Cell Signaling Technology. Cell viability was assessed by DAPI (50 ng/mL). The FOXP3 transcription factor staining kit (eBioscience) was used for intracellular staining, and viability determined by Live Dead amine reactive dyes (ThermoFisher). LSR II (BD Biosciences) or Attune NxT (ThermoFisher) flow cytometers, and FlowJo v10 software (Ashland, Oregon) were used for data acquisition and analyses. Flow sorting of Tregs was performed in a FACSAria flow sorter (BD Biosciences). Gating strategy was based on viable CD4+CD25+CD127− cells (see Supplementary Fig. S1) with a dump channel excluding CD8+, CD19+, and CD56+ cells.
For survival correlation data, the percentage change in pSTAT3 expression was calculated as: (week 13 gMFI—baseline gMFI)/baseline gMFI * 100%.
In vitro assays
CD3+ T cells were negatively isolated using an EasySep kit, and cultured in X-VIVO media, supplemented with 10% AB human serum, l-glutamine, beta-mercaptoethanol, non-essential amino acids, HEPES, penicillin, streptomycin, gentamicin and 100 IU/mL IL2.
Supernatant cytokine levels were assessed by Luminex assays from R&D Systems.
Statistical analyses were performed using GraphPad Prism 7 software. Type-two, two tailed t tests were used for comparisons of two groups. One-way ANOVA were used for comparisons of three or more groups. Paired, two-tailed, t tests were used to determine statistical significance of pre- versus post-treatment. Correlation significance and R2 was assessed using linear regression. P values of ≤0.05 were considered significant.
Changes in gene expression after nivolumab therapy differ based on patient outcome and T-cell subset
To gain insight into the effects of nivolumab treatment on Tregs and conventional T cells, CD14-CD56-CD19-CD3+CD4+ CD127low/−CD25+ (Tregs; refs. 25, 26) and the remainder of the CD4+ population (Tcons) were flow sorted from pre- and post- adjuvant nivolumab treatment samples. Paired samples from peripheral blood mononuclear cells (PBMC; baseline and week 13 of treatment) of seven patients free of disease (no evidence of disease [NED] at four years post-resection) and seven relapsing patients were assessed by RNA-Seq (Supplementary Fig. S1A). Patient demographics and experimental values are reported in Supplementary Table S1. Changes in gene expression (post-treatment compared with baseline) were assessed for each patient and significantly changed genes (FDR corrected q ≤ 0.05) determined for Tregs and Tcons. As staining for the intracellular marker FOXP3 is not compatible with retaining viable cells for downstream assessment, potential differences in expression of FOXP3 in NED and relapsing patient Tregs were evaluated in a separate experiment. No significant differences in FOXP3 expression were noted in baseline (Supplementary Fig. S1B, P = 0.385) or post-treatment (Supplementary Fig. S1C, P = 0.289) between NED and relapsing patients.
In Tcons, 189 genes were significantly changed in NED samples and 227 in relapsing patient samples (Supplementary Fig. S1D). Seven genes overlapped between NED and relapse Tcons with an additional nine genes changed in both groups, but in opposing directions (Supplementary Table S2). In Tregs, 382 genes were significantly changed in NED samples and 256 in relapsing patients. Twenty-two genes overlapped between NED and relapse Tregs with an additional five genes changing in both groups in opposing directions (Supplementary Fig. S1D, Supplementary Table S3). We next compared the overlap in significantly changed genes between Tcons and Tregs. In NED patient samples, four genes overlapped between Tcons and Tregs with an additional 18 overlapping but in opposing directions (Supplementary Fig. S1D, Supplementary Table S4). In relapsing patient samples, 46 genes overlapped with an additional two overlapping but in opposing directions (Supplementary Fig. S1D, Supplementary Table S5).
Non-relapsing patient Tregs have increased expression of proliferation-associated genes post-nivolumab
The biological processes associated with these genes were assessed using gene-set enrichment analysis (GSEA) referenced to hallmark gene sets (ref. 24; Supplementary Fig. S1E). Analysis of Tcons showed upregulation of IFNγ response genes in both NED (q = 1.74e−02) and relapsed (q = 1.51e−07) patients (Supplementary Fig. S1E). In Tregs, upregulation of IFNγ response hallmark was observed in relapsed (q = 3.42e−09), but not NED patients (Supplementary Fig. S1F). Oxidative phosphorylation (q = 6.51e−05) was uniquely downregulated in NED Tregs, in contrast with the upregulation seen in NED Tcons (q = 2.60e−02). Heme metabolism associated genes were downregulated in relapse Tregs (q = 1.78e−03) and Tcons (q = 3.93e−02), in contrast to being upregulated in NED Tcons (q = 2.60e−02) but not in NED Tregs. Proliferation associated hallmarks, including G2–M checkpoint (q = 2.62e−59), E2F targets (q = 1.34e−43) and mitotic spindle (q = 7.21e−19) were only upregulated in NED Tregs. GSEA hallmark analysis of NED Tcon downregulated genes returned no results. These results suggest that the effects of nivolumab therapy on Tcons and Tregs gene expression are distinct.
Patients benefiting from adjuvant nivolumab therapy display an increase in circulating Tregs post-treatment
Given the observation that proliferation associated genes were upregulated in NED Tregs and that increased levels of circulating Tregs are associated with response to ipilimumab (18), we next assessed percentages of circulating Tregs by flow cytometry. The proportion of Tregs (CD4+CD25+CD127low/−FOXP3+) in relapsed compared with NED patients at baseline was not significantly different (P = 0.098; Fig. 1A; NED n = 18, relapsed n = 6).
PBMC samples from patients with unresectable metastatic melanoma treated with nivolumab were likewise assessed for Tregs as a percentage of live cells. Unresectable disease patient demographics and experimental values are reported in Supplementary Table S6. As shown in Fig. 1B, a significantly increased representation of Tregs was found in non-responding relative to responding patients (P = 0.045).
The proportion of Tregs at week 13 of treatment with nivolumab was next assessed. No significant differences were found in the post-treatment levels of Tregs at week 13 for adjuvant treated (P = 0.396), or active disease patients (P = 0.324). However, in paired analyses NED patients displayed increased percentage of Tregs post-nivolumab (ratio paired t test, n = 18, P = 0.016, Δ = 1.24%), but relapsing patients did not show a significant change (n = 6, P = 0.476; Fig. 1C). PBMC samples from active disease patients were likewise assessed for changes in the percentage of circulating Tregs. As shown in Fig. 1D, responding patients had a borderline-significant increase in Treg percentages at week 13 after initiation of nivolumab treatment (n = 6, P = 0.065), while non-responding patients did not (n = 11, P = 0.596).
Tregs from non-relapsing patients have reduced suppressive function
We next evaluated whether Treg suppression was associated with patient outcomes. Tregs (CD4+CD127low/−CD25+) from resected disease patient samples were assessed in an allogeneic mixed lymphocyte reaction. Tregs were flow sorted (gating strategy shown in Supplementary Fig. S2) from pre- and post-treatment (week 13) patient PBMC, then cultured with allogeneic APCs and CD8+ T cells. Tritiated thymidine incorporation as an indicator of T-cell proliferation was assessed at different ratios of Tregs:CD8+:APCs. Tritiated thymidine counts were normalized to controls containing no Tregs (i.e., 0:1:2 ratio) for baseline and week 13 treatment independently. Suppressive function was assessed in seven adjuvant NED patients and six patients who had relapsed (representative assay shown in Fig. 1E). Percent suppression, relative to CD8+ only cultures, at a ratio of 1:2:2 is shown in Fig. 1F. Although not significant (P = 0.076), NED patients trended toward decreased levels of suppression at baseline relative to relapsed patients. Significantly reduced suppression in NED relative to relapsed was observed in week 13 samples (P = 0.002). Paired analyses also showed a significant decrease in suppression from post-nivolumab Tregs of NED patients (P = 0.010), but no changes in relapsed patient samples (P = 0.683).
Tregs from patients benefiting from nivolumab have increased pSTAT3 expression post-treatment
Previous studies have shown that Treg proliferation and function are regulated in part by STAT3 signaling (27, 28). Therefore, we evaluated expression of pSTAT3 in PBMC samples from surgically resected, stage III/IV melanoma patients receiving adjuvant nivolumab. Circulating Tregs (CD4+CD25+CD127low/−FOXP3+) from NED patients displayed an increased percentage of pSTAT3-expressing cells at week 13 relative to baseline (Supplementary Fig. S3C, 12/14 patients; P = 0.004). This was not seen in relapsing patient samples (n = 6, P = 0.736). To determine whether this effect was general to T cells, we likewise evaluated Tcons and CD8+ T cells from NED patients. Tcons had significant upregulation of pSTAT3 after nivolumab treatment (Supplementary Fig. S3B, P = 0.015) as did CD8+ T cells (P = 0.029). Of these populations, Tregs displaying the highest average change from baseline (Δ=11.7%). pSTAT3 geometric mean fluorescence intensity was similarly elevated post-treatment in Tregs (P = 0.005), CD4+ Tcon (P = 0.008) and CD8+ T cells (P = 0.014) from NED patients (Fig. 2A). No significant changes were seen in any of these populations in relapsed patients.
pSTAT3 expression was also assessed in circulating Tregs from unresectable, stage III/IV metastatic melanoma patients treated with nivolumab. Responding patients had an increase in Treg pSTAT3 levels (9/11 patients increased, P = 0.007), whereas Tregs from patients with progressive disease did not display changes in pSTAT3 expression (n = 17, P = 0.934; Fig. 2B). Similar increases in pSTAT3 were seen in responding patients’ Tcons (P = 0.047) and CD8+ T cells (P = 0.021), but was not observed in non-responders. We next evaluated the association between changes in Treg pSTAT3 expression and overall survival in metastatic patients receiving nivolumab therapy. Increased survival in patients with metastatic disease was positively correlated with increases in Treg (R2 = 0.228, P = 0.045) and CD8+ T-cell (R2 = 0.229, P = 0.045) pSTAT3 expression (Fig. 2C).
PD-1 blockade enhances pSTAT3 expression in T cells in vitro
To determine whether the observed increases in T-cell pSTAT3 expression were a direct effect of PD-1 blockade, T cells isolated from baseline PBMC samples of resected disease patients were cultured for 48 hours with the addition of PD-1 blocking antibody or IgG control. Tregs, Tcons, and CD8+ T cells from PD-1 blockade cultures had enhanced pSTAT3 expression relative to IgG (Fig. 2D; Supplementary Fig. S6A). These results suggest that blocking PD-1 increases T-cell pSTAT3 expression independent of interactions with an intermediate cell population (e.g., APC or tumor).
Patient T cells express PD-L1
Given the ability of αPD-1 to upregulate T-cell pSTAT3 in purified T-cell cultures, we hypothesized that T cells may express PDL1. Baseline patient T cells were cultured in vitro for 72 hours with or without the addition of CD3/CD28 activation. PDL1 expression was detected in activated T cells with 28% of Tregs, 13% of Tcons and 10% of CD8+ expressing PDL1 in the representative flow plots shown in Supplementary Fig. S4A.
We next evaluated Tregs (CD4+CD127−CD25+) from a small number of resected patient pre- and post-treatment PBMC samples to investigate potential differences in inhibitory molecule expression. A decrease in PD-1 expression was seen in all patient samples assessed (NED: n = 5, P = 0.008; Relapse: n = 3, P = 0.130; Supplementary Fig. S4B). Supplementary Fig. S4C shows that Tregs expressed appreciable levels of PDL1 (15%–41% positive). NED patient Tregs had a reduction in PDL1 expression (P = 0.034), whereas no significant changes were seen in relapsing patients. No significant changes were seen in CTLA4 expression (Supplementary Fig. S4D).
PD-1 blockade increases T-cell IL10 production in vitro.
STAT3 positively regulates IL10 expression, a cytokine with known roles in Treg expansion, differentiation and function (29, 30). Therefore, we investigated the effects of PD-1 blockade on IL10 production. Total T cells isolated from baseline patient PBMC samples were cultured for two hours with PD-1–blocking antibody or IgG, CD3/CD28 stimulated for 72 hours, and cytokine production assessed. PD-1 blockade augmented IL10 levels relative to IgG (P = 0.023; Fig. 3A). Paired analysis showed an increase in IL10 production (Supplementary Fig. S6B, P = 0.0289). To confirm the increase in IL10 expression, T cells were cultured overnight with IgG or PD-1 blocking antibodies and subsequently activated for 6 hours via CD3/CD28. Assessment of IL10 mRNA demonstrated upregulation of IL10 expression (Supplementary Fig. S5).
We next investigated whether IL10 production was dependent on pSTAT3. Addition of the STAT3 inhibitor static (31) to cultures reversed the PD-1 blocking antibody mediated increase in IL10 (P = 0.0094; Fig. 3B). STAT3 inhibition did not significantly alter IL10 production in IgG–treated samples (P = 0.2993; Supplementary Fig. S6C). IL10 levels were reduced in all samples when STAT3 inhibitor was added to PD-1 blocking antibody (P = 0.0023; Supplementary Fig. S6D).
To determine the cellular source of the increased IL10, total T cells isolated from baseline patient PBMC samples were cultured for 72 hours with PD-1–blocking antibody or IgG and assessed for IL10 expression by intracellular flow staining. Indicated T-cell subsets (e.g., CD8+, Treg) were gated on and the percentage of IL10 expressing cells assessed for each subset. Figure 3C shows an increase in CD8+ T-cell IL10 production without (P = 0.0429), and with CD3/CD28 activation (Fig. 3D, P = 0.0087) in PD-1 blocking antibody treated cultures relative to IgG. Paired analysis is shown in Supplementary Fig. S6E (P = 0.0446) and 6F (P = 0.0192). CD4+ T-cell populations did not show differences in IL10 production.
IL10 production enhanced by PD-1 blockade is associated with increased Treg percentages in vitro
As IL10 has been demonstrated to induce differentiation and proliferation of Tregs (30), we investigated whether the observed increase in pSTAT3 and IL10 production mediated by PD-1 blockade influenced Treg percentages in vitro. T cells isolated from baseline patient PBMC were cultured with 50 ng/mL recombinant IL10 for 24 or 48 hours, and Tregs evaluated as a percentage of the CD4+ population. Addition of IL10 increased the proportion of CD4+CD25+ CD127low/−FOXP3+ Tregs as a percentage of the CD4+ population at 24 (P = 0.0005) and 48 hours (P < 0.0001; Fig. 4A). Paired analysis displayed upregulation of Treg percentages at 48 hours (P = 0.0006; Supplementary Fig. S6I). In a similar experiment, T cells were cultured with PD-1–blocking antibody or IgG and evaluated for proportion of Tregs as a percentage of the CD4+ population at 48 hours. PD-1 blockade increased Treg percentages relative to IgG (P = 0.0011, Fig. 4B). Paired analysis of samples showed upregulation (P = 0.0007; Supplementary Fig. S6H). Figure 4C demonstrates increases in Tregs evaluating CD4+CD25+ CD127low/−FOXP3+ cell counts after culturing with αPD-1 (P = 0.0001) or IL10 (P = 0.0001). PD-1–blocking antibody also increased expression of Ki67 in Tregs (P = 0.0002; Fig. 4D). Paired analysis showed upregulation in all samples assessed (P = 0.0059; Supplementary Fig. S6G).
To determine whether the higher Treg percentages observed in cultures with PD-1–blocking antibodies were mediated by increased IL10 production, IL10–neutralizing antibodies were added to cultures of T cells treated with PD-1 antibody or IgG. Neutralization reduced Treg percentages relative to PD-1 blockade alone (P = 0.0009) but had no impact on IgG cultures (P = 0.3218; Fig. 4E). Paired analysis showed that neutralizing IL10 did not change Treg percentages relative to IgG (Supplementary Fig. S6J), but reduced Treg percentages when added to PD-1 blockade (P = 0.0149; Supplementary Fig. S6K).
To determine whether the observed increases in Treg proportions were dependent on pSTAT3 induction, T cells were cultured with IgG or PD-1 blocking antibodies with or without STAT3 inhibitor (stattic). Addition of stattic to PD-1 blocking antibody-treated cultures reduced Treg percentages (P = 0.0205) but not when added to IgG-treated cultures (P = 0.0887; Fig. 4F). Paired analyses showed that stattic decreased Treg percentages when added to PD-1 antibody-treated cultures (P = 0.0068; Supplementary Fig. S6M), but not when added to IgG-treated controls (P = 0.3290; Supplementary Fig. S6L).
To address if the observed increases in Tregs resulted from inducible Tregs (iTregs), baseline patient samples were isolated for CD3+ T cells (Fig. 4G, top) and a portion of those cells depleted for CD25 (Fig. 4G, bottom). Cells were then cultured in the presence of PD-1–blocking antibody or recombinant IL10 for 72 hours and assessed for CD4+CD25+CD127low/−FOXP3+ Treg counts. Exogenous IL10 increased Treg numbers in both total T-cell (P = 0.0002) and CD25-depleted cultures (P < 0.0001; Fig. 4H). An increase over control was seen in the PD-1–blocking antibody-treated CD25-depleted culture (P = 0.0129). PD-1 blockade also increased Ki67 expression in CD25-depleted (P < 0.05) and total (P < 0.01) T cells (Fig. 4I).
Treg suppression is reduced in vitro by PD-1/PDL1 blockade and enhanced by STAT3 inhibition
To determine the effects of PD-1 blockade on nTreg-suppressive function in vitro, isolated Tregs were pre-treated with PD-1–blocking antibodies or IgG, and equivalent numbers of Tregs subsequently co-cultured with CFSE-labeled autologous T cells plus CD3/CD28 activation. After 3 days, cultures were assessed by flow cytometry for proliferation of target T cells. PD-1 blockade decreased Treg suppression, as evidenced by an increased percentage of proliferating Tcons (no Treg control: 93.3%, IgG: 45.5%, PD-1 antibody treated: 66.6% proliferating Tcons; Fig. 5A). To address the role of STAT3 in the suppressive function of Tregs, a similar in vitro nTreg assay was performed using DMSO or STAT3 inhibitor (stattic) pre-treated Tregs. Tregs treated with STAT3 inhibitor had increased suppressive function relative to DMSO (no Treg control: 41.3%, DMSO: 15.4%, STAT3 inhibitor: 10.4% proliferating Tcons; Fig. 5B).
As the data suggest that blocking PD-1 increases iTregs, we also evaluated the effects of PD-1 blockade and STAT3 inhibition on the suppressive function of iTregs. Naïve CD4+ T cells from baseline patient PBMC samples were polarized with or without the addition of PD-1–blocking antibodies or other indicated treatments. Polarized cells were washed before addition to autologous CD8+ T cells as outlined in Fig. 5C. To determine efficacy of polarization, polarized CD4+ T cells were evaluated by flow cytometry for expression of the transcription factors FOXP3 and HELIOS (Fig. 5D) as well as PD-1 and PDL1 (Fig. 5E). Polarized CD4+ T cells expressed high levels of HELIOS (90.8% positive) and FOXP3 (42.2% positive). These cells also expressed high levels of PD-1 (93% positive) and PDL1 (33.7% positive). No differences were observed between IgG and PD-1–cultured iTregs in expression of these markers or TGFβ production (data not shown).
Culture of naïve CD4+ T cells with PD-1–blocking antibodies decreased suppressive function at 1:10 (Treg:CD8+; P < 0.0001) and 1:5 (P = 0.0004; Fig. 5F). Given the observation that patient T cells and polarized iTregs express PDL1, we hypothesized that similar reduction in suppressive function could be achieved with PDL1 blockade. As seen in Fig. 5G, the addition of PDL1-blocking antibodies reduced suppression at 1:8 (P = 0.0179) and 1:4 (P = 0.0263) iTreg:CD8+ ratios. Production of inflammatory cytokines was also evaluated in iTreg suppression assays. Cultures containing iTregs displayed reduction in TNF, IFNγ, and Granzyme B relative to CD8+ only cultures (0:1 ratio; Fig. 5H). However, iTregs polarized in the presence of PD-1 (P < 0.01) or PDL1-blocking antibodies (P < 0.05) had an increase in all three cytokines, at all ratios relative to IgG control iTregs.
To determine the effects of STAT3 inhibition in iTreg function, naïve CD4+ T cells were polarized as before, but with the addition stattic. STAT3 inhibition augmented iTreg suppression at 1:8 (P = 0.0004) and 1:4 ratios (P = 0.0397; Fig. 5I).
Our data showed that after nivolumab treatment, circulating Tregs from patients with clinical benefit had reduced suppressive function, increased proliferation, and increased pSTAT3 expression. These changes were absent in patients without benefit. Increased pSTAT3 expression was also observed in CD4+ Tcons and CD8+ T cells in benefiting patients. PD-1 blockade in vitro of purified T cells recapitulated these effects, supporting a direct effect on T cells rather than mediated through tumor or APC interactions. In vitro assays demonstrated that PD-1 blockade increased T-cell IL10 production in a STAT3 dependent manner, which mediated the increase in Treg percentages. However, no differences were noted in the in vitro effects of PD-1 blockade between NED and relapsing patient samples (data not shown), limiting the use of these assays as predictive of patient outcomes to PD-1 blockade. These data support a model in which PD-1 blockade induces STAT3 signaling in Tregs, reducing suppressive function while increasing the production of IL10 by CD8+ T cells, resulting in Treg proliferation. Increases in Treg proliferation were seen in NED patient RNA-Seq gene signatures, circulating Treg proportions in NED patients, and in vitro assays. Although an accompanying increase in nTreg proliferation cannot be ruled out, these data suggest that induction of iTregs by enhanced IL10 production mediates the increase in Tregs.
How the impact on peripheral Tregs by PD-1 blockade correlates to changes in tumor-infiltrating Tregs in patients with unresectable disease remains to be investigated. Studies have demonstrated that the phenotype and function of intra-tumor Tregs is similar but enhanced relative to that of circulating Tregs (e.g., increased suppressive function in the tumor and enhanced inhibitory molecule expression; ref. 32). The major patient population in this study was surgically resected and free of tumor, limiting the ability to address this hypothesis. Given the scarcity of tumor biopsies from patients, particularly paired (pre- and post-treatment), future work should determine if the effects observed here are also observed in murine pre-clinical models.
PD-1 blockade exerted effects on T-cell isolated cultures in the absence of tumor or APCs. PDL1 was robustly expressed in patient T cells, suggesting that interaction of PD-1 with PDL1 expressed on T cells may play an important role in T-cell function and efficacy of immunotherapies. In support of this hypothesis, a recent study has demonstrated that high levels of PDL1 expression by CD8+ T cells are associated with progression of disease after ipilimumab treatment (33). Our data show that peripheral blood Tregs from resected patients express high levels of PD-1, CTLA4, and PDL1. Interpretations are limited by the low sample numbers available for assessment, but a significant reduction in PDL1 expression was observed in NED patients. In contrast, 2/3 relapsed patient samples assessed had an increase in PDL1 expression. These results highlight a potential important role of Treg PDL1 expression in patient outcome, which should be investigated in future studies.
Post-treatment peripheral blood nTregs had reduced suppressive function in NED patients, and in vitro assays showed similar decreases in the suppressive function of PD-1–exposed iTregs. Although not significant, pre-treatment Treg suppression was notably lower in NED patients. Future validation of alterations in suppressive function, including in active disease patients, may offer larger sample sizes and statistical power to determine whether baseline differences influence patient outcomes.
In some patients, Tregs assessed for suppressive activity resulted in enhanced CD8+ T-cell proliferation over no Treg controls (i.e., negative suppressive values). The Tregs used in these assays were isolated using the marker set CD4+CD127low/−CD25+, as intracellular staining of FOXP3 is incompatible with downstream use in functional assays. In assessing FOXP3 expression in circulating Tregs, no significant differences were found between NED and relapsing patients, suggesting that differences in FOXP3 expression could not explain differences in suppression. These limitations illustrate the need for further research identifying surface markers that uniquely distinguish suppressive T cells.
STAT3 has important roles in Treg biology, including acting as a co-transcription factor with FOXP3 and regulating IL10 gene expression among other STAT3 targets (34). In agreement with our data, STAT3 expression in Tregs was previously shown to be associated with decreased suppressive function (35). Increased amounts of pSTAT3+ Tcons are associated with immune activation in systemic lupus erythematosus (29) and graft-versus-host disease after allogeneic hematopoietic cell transplantation (36), suggesting enhanced reactivity. However, the role of Tcon pSTAT3 expression in the context of malignancy and patient response to therapies still remains to be fully addressed. Our results show that pSTAT3 induction in all populations of T cells is associated with patient outcome, and future work will specifically address the consequences of pSTAT3 induction in effector T cells.
The duality of STAT3 and IL10 in the tumor-immune context is becoming increasingly clear. STAT3 signaling and IL10 production by melanomas are associated with worse prognosis (37, 38), and our data, along with that of other groups (15), show that IL10 production results in the formation of iTregs. However, STAT3 and IL10 are also necessary for CTL effector function and memory formation, requisites for a productive antitumor immune response (39–41). Likewise, Tregs are associated with poor patient outcome in a variety of malignancies (42), but are also critical in inhibiting carcinogenesis through suppression of inflammation (43). Our results suggest more complex roles of IL10 in the context of checkpoint blockade of melanoma. Studies have shown increased IL10 expression in immune cells after PD-1 blockade (44), association of IL10 levels with patient response (45), and a potential antitumor efficacy of pegylated IL10 therapy (46), supporting the need for evaluating the contextual effects and clinical impact of IL10.
Our data demonstrate an important role of Tregs in the outcome of metastatic melanoma patients after nivolumab therapy and reveal three potential pharmacodynamic biomarkers of melanoma patient response to nivolumab: increased T-cell pSTAT3 expression, decreased Treg-suppressive function, and increased Treg proliferation. As pharmacodynamic biomarkers, these are not useful in stratifying patient treatments, but instead give further insight into potential novel mechanism of patient response and generate rationale for combination therapies. Data from in vitro experiments link these three biomarkers in a model where nivolumab directly induces Treg pSTAT3, reducing suppression and resulting in IL10 production leading to proliferation.
Disclosure of Potential Conflicts of Interest
D.M. Woods holds ownership interest (including patents) in Bristol Myers-Squibb. B.C. Betts is listed as an inventor on a provisional patent application pending regarding STAT3 inhibition in GVHD prevention and treatment. J.S. Weber holds ownership interest (including patents) in Altor and CytoMx, and is a consultant/advisory board member for Bristol Myers-Squibb, Merck, Incyte, GlaxoSmithKline, Genentech, Pfizer, AstraZeneca, Altor, and Celldex. No potential conflicts of interest were disclosed by the other authors.
Conception and design: D.M. Woods, R. Ramakrishnan, A.S. Laino, J.S. Weber
Development of methodology: D.M. Woods, R. Ramakrishnan
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.M. Woods, R. Ramakrishnan, A.S. Laino, K. Walton, B.C. Betts, J.S. Weber
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.M. Woods, R. Ramakrishnan, A.S. Laino, A. Berglund, B.C. Betts, J.S. Weber
Writing, review, and/or revision of the manuscript: D.M. Woods, A.S. Laino, A. Berglund, K. Walton, B.C. Betts, J.S. Weber
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.M. Woods
Study supervision: J.S. Weber
This work was supported by grant funding from the National Cancer Institute (NIH NCI R01 CA175732-01). We extend our appreciation to Christine Horak, Don Jackson and Bristol-Myers Squibb for their insight and assistance as well as for providing nivolumab. We also extended our appreciation to the Genomics Core Facility, Sean Yoder, the Flow Cytometry Core Facility, and Jodie Kroger at Moffitt Cancer Center for their assistance.