PD-1/PD-L1 Blockade Enhances T-cell Activity and Antitumor Efficacy of Imatinib in Gastrointestinal Stromal Tumors

This article is featured in Highlights of This Issue, p. 325

The advent of targeted molecular therapy has revolutionized the treatment of many cancers, including gastrointestinal stromal tumors (GISTs). The majority of GISTs contain an activating mutation in either the KIT or PDGFRA oncogene (1, 2). Imatinib mesylate (Gleevec) is a tyrosine kinase inhibitor that specifically targets KIT and PDGFRA (3). The tumor response to imatinib in GISTs is impressive, but most often transient. Resistance commonly develops within 2 years, often due to a secondary KIT mutation (4, 5). Imatinib acts primarily via direct effects on tumor cells. However, we previously showed that imatinib also inhibits tumor cell production of the immunosuppressive enzyme indoleamine 2,3-dioxygenase (IDO; ref. 6).

Many tumors express ligands that engage inhibitory receptors on T cells and decrease T-cell activation and function within the tumor microenvironment. There is growing evidence that tumor cells commonly exploit the programmed death 1 (PD-1, PDCD1) and programmed death ligand 1 (PD-L1, PDCD1LG1) axis to evade the immune system (7). PD-1 is an inhibitory receptor that is upregulated after T-cell activation and remains elevated with antigen persistence and therefore is often increased on tumor-infiltrating T cells (8–10). The ligands for PD-1 are PD-L1 (B7-H1, CD274) and PD-L2 (B7-DC, CD273, PDCD1LG2). PD-L1 is expressed widely on immune cells and can be upregulated by proinflammatory stimuli, such as interferons, but is also expressed by tumor cells in a variety of cancers (11). PD-L1 expression on tumors can lead to impaired T-cell proliferation and effector function, leading to apoptosis of tumor-specific T cells (12). PD-L2 expression is restricted mostly to hematopoietic cells. PD-1/PD-L1 blockade has demonstrated encouraging antitumor effects in several solid tumors, including kidney, bladder, and lung cancer, as well as melanoma (13–19).

Despite the efficacy of tyrosine kinase inhibition, nearly all patients with a metastatic GIST develop tumor progression and eventually succumb to their disease. There has not been any improvement in the first-line therapy for GISTs since imatinib was approved in 2002. In this study, we analyzed freshly isolated T cells from the tumor and peripheral blood of patients with GISTs for the expression of inhibitory receptors. We determined the effects of imatinib on IFNγ-related genes and the PD-1/PD-L1 axis in GISTs. Furthermore, in a genetically engineered mouse model of GISTs, we tested the combination of imatinib with PD-1/PD-L1 blockade.

Tumor specimens and matched peripheral blood were obtained from 85 patients with GISTs who underwent surgery at our institution and were consented to a protocol approved by the Institutional Review Board. Blood was drawn before surgical incision, and peripheral blood mononuclear cells were isolated by density centrifugation over Ficoll–Plaque PLUS (GE Healthcare). Tumor tissue was subjected to mechanical dissociation to obtain single-cell suspensions, as described previously (6). After procurement, all specimens were processed, and cells were immediately analyzed with flow cytometry. Tumor (KIT⁺) and stromal cells (KIT⁻) were isolated using human CD117 microbeads (Miltenyi Biotec). The purity of isolated cells was greater than 90% by flow cytometry.

The human GIST cell lines GIST-T1 (KIT exon 11 mutant; ref. 20), HG129 (also KIT exon 11 mutant; ref. 21), and GIST882 (KIT exon 13 mutant; provided by Jonathan Fletcher [Department of Pathology, Brigham and Women's Hospital, Harvard Medical School, Boston, MA]) were maintained at 37°C in 5% CO₂ in RPMI1640 medium supplemented with 10% \FBS, 2 mmol/L l-glutamine, 50 U/mL penicillin–streptomycin, 0.1% 2-mercaptoethanol, and 10 mmol/L Hepes. Cells were incubated with recombinant human IFNγ (100 ng/mL; R&D Systems), imatinib (100 nmol/L; Novartis), or the pan-JAK inhibitor tetracyclic pyridine 6 [P6; 1 μmol/L; provided by Jacqueline Bromberg from the Department of Medicine, Memorial Sloan Kettering Cancer Center (MSKCC), New York, NY (22)].

Kit^V558Δ/+ mice (23) that were 6 to 12 weeks old were maintained in a specific pathogen-free animal facility, age- and sex-matched for experiments, and used in accordance with an institution-approved protocol. Tumors were minced, incubated in 5 mg/mL collagenase IV (Sigma-Aldrich) plus 0.5 mg/mL DNAse I (Roche Diagnostics) in HBSS for 30 minutes at 37°C, then quenched with FBS, and washed through 100- and 40-μm nylon cell strainers (Falcon; BD Biosciences) in PBS with 1% FBS. Tumor-draining lymph nodes and spleens were mechanically dissociated as described previously (6). Cells were immediately analyzed with flow cytometry. Imatinib was administered at 90 mg/kg per day in the drinking water. 1-Methyl-D-tryptophan (1-MT; Sigma Aldrich) was prepared and administered twice daily by oral gavage at 400 mg/kg per day as before (6). Anti-PD-1 (RMP1-14; Bio X cell) or anti-PD-L1 (10F.9G2; Bio X cell) blocking antibodies or isotypes (rat IgG2a or rat IgG2b) were administered intraperitoneally at a dose of 250 and 200 μg per mouse, respectively, at the indicated time points. Murine T cells were purified from mesenteric lymph nodes from C57BL/6J mice (B6, The Jackson Laboratory) and tumors from Kit^V558Δ/+ mice using CD3-biotin and anti-biotin MicroBeads (Miltenyi Biotec). Positive selection was performed using two sequential MACS columns. CD3⁺ T cells were then cultured in anti-CD3–coated 96-well plates (BD Biosciences) with 10 μg/mL of either anti-PD-1 antibody or isotype, and supernatant was harvested at 48 hours.

Human-specific antibodies were purchased from BD Biosciences (CD45, 2D1; CD3, SK7; CD4, RPA-T4; PD-1, MIH4), eBioscience (CD8, RPA-T8), BioLegend (PD-L1, 29E.2A3), R&D Systems (TIM-3, 344823), and Enzo Life Sciences (LAG-3, 17B4). Mouse-specific antibodies were purchased from BD Biosciences (CD3, 125-2C11; CD4, GK1.5; CD44, IM7; CD62L, MEL-14; PD-L1, MIH5; CD69, H1.2F3; IFN-γ, XMG1.2; TNF, MP6-XT22), eBioscience (CD8, 53-6.7; PD-1, J43; PD-L2, 122; Foxp3, FJK-16s; Ki67, SolA15; Granzyme B, NGZB), and BioLegend (CD45, 30-F11; PD-L1, 10F.9G2; CD25, PC61). Appropriate isotype controls were used where applicable. A viability dye was typically used to exclude dead cells. Intracellular staining was performed using the eBioscience Fixation and Permeabilization Buffer Kit. For intracellular cytokine staining, cells were stimulated with phorbol 12-myristate 13-acetate (PMA, 50 ng/mL) and ionomycin (750 ng/mL) for 4 hours at 37°C, 5% CO₂ in the presence of 1 μg/mL brefeldin A (BD Biosciences). Surface staining was performed and cells were fixed and permeabilized with the BD Cytofix/Cytoperm Kit and stained for IFNγ and TNFα. Supernatant cytokines were measured by cytometric bead array according to the manufacturer's instructions (Mouse Inflammation Kit; BD Biosciences). Data were acquired using a BD FACSAria or Fortessa LSR flow cytometer and analyzed using FlowJo software (Tree Star).

Formalin-fixed and paraffin-embedded specimens were sectioned at 5-μm thickness and mounted on glass slides. The PD-L1 (clone 5H1) antibody was obtained from Lieping Chen (Department of Immunobiology, Yale University School of Medicine, New Haven, CT) (12). In brief, antigen retrieval was achieved with citrate buffer. Anti-PD-L1/B7-H1 murine IgG (1:1,000) was applied on tissue sections overnight at 4°C. After washing with TBS (0.05 mol/L Tris base, 0.9% NaCl, pH 8.4), tissue sections were incubated with biotinylated anti-mouse IgG (1:100; BA-2001, Vector Laboratories), followed by incubation with Elite ABC Kit for 30 minutes at room temperature. Antibody binding was detected with the TSA Biotin System (NEL700A001KT, PerkinElmer) and DAB (DAKO, K3468). Murine PD-L1 staining was performed using the polyclonal antibody against mouse B7-H1/PD-L1 antibody (1:100, R&D Systems). Terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) and KIT immunostaining (1:200, D13A2; Cell Signaling Technology) were performed as described previously (24). Slides were analyzed and imaged on the Axio Imager 2 wide-field microscope (Zeiss).

Total RNA was extracted from each human GIST specimen, cell line, or murine GIST specimen, reverse transcribed, and amplified with PCR TaqMan probes for human PDCD1LG1 (i.e., PD-L1, Hs01125301_m1), IFNGR1 (Hs00988303_m1), IRF1 (Hs00971960_m1), GAPDH (Hs02758991_g1), murine Pdl1 (Mm00452054_m1), Ido1 (Mm00492590_m1), and Gapdh (Mm99999915_g1, all Applied Biosystems). Quantitative PCR was performed using a ViiATM7 real-time PCR system (Applied Biosystems). Relative expression was calculated by the 2^−ΔΔC_t method according to the manufacturer's protocol and expressed as the fold increase over the indicated control.

For transient STAT1 knockdown, GIST-T1 cells were transfected with 30 nmol/L of ON-TARGET plus SMARTpool siRNA specific for human STAT1 (L-003543-00) or a nontarget control siRNA (D-001810-10-05; Thermo Scientific) using Lipofectamine RNAiMAX (Invitrogen) for 48 hours. GIST-T1 cells were treated with PBS or IFNγ (100 ng/mL) for 6 hours.

Western blot analysis of whole-protein lysates from frozen tumor tissues or cells was performed as described previously (6). Antibodies for PD-L1 (E1L3N), phosphorylated SRC (Tyr416), phosphorylated STAT1 (Tyr701), phosphorylated STAT3 (Ser727 or Tyr705), total and phosphorylated KIT (Tyr719), phosphorylated ERK1/2 (Thr202/Tyr204), phosphorylated SMAD2 (Ser465/467), TGFβ, and GAPDH were purchased from Cell Signaling Technology. Anti-IDO antibody (clone 10.1) was purchased from EMD Millipore. Whole protein from frozen tumor tissues was lysed, and total protein was tested for cytokine/chemokine expression using a Proteome Profiler Array (R&D Systems). Densitometry was conducted on blots using ImageJ (NIH, Bethesda, MD).

Data are expressed as mean ± SEM or median. Unpaired two-tailed Student t test or one-way ANOVA comparisons were performed as applicable using GraphPad Prism 6.0 (GraphPad Software). P < 0.05 was considered significant.

The inhibitory receptors PD-1, LAG-3, and TIM-3 (HAVCR2) are typically expressed on antigen-specific and dysfunctional T cells within tumors and during chronic viral infection (10, 25). To identify the inhibitory receptors that are relevant in GISTs, we performed flow cytometry on T cells from 106 tumor specimens and matched blood, freshly obtained from 85 patients with GISTs undergoing surgery (Supplementary Table S1). The percentages of CD4⁺ and CD8⁺ T cells among leukocytes were lower in tumors than in matched blood. CD8⁺ T cells were less frequent than CD4⁺ T cells in both compartments (Fig. 1A). Intratumoral T cells displayed greater cell-surface expression of PD-1, LAG-3, and TIM-3 compared with T cells from matched blood (Fig. 1B–D). PD-1 was expressed at the highest levels on both intratumoral CD4⁺ and CD8⁺ T cells (48% and 39%, respectively). Moreover, the percentage of PD-1 on CD4⁺ and CD8⁺ T cells correlated within the same tumor specimen (Fig. 1E).

The expression of inhibitory receptors on tumor-infiltrating T cells suggested that T-cell immune evasion occurs in human GISTs. PD-1⁺ T cells tended to also express LAG-3 and less frequently TIM-3 (Fig. 2A). In contrast, PD-1⁻ T cells had little expression of LAG-3 or TIM-3. The majority of T cells from matched blood did not express any inhibitory receptors (Fig. 2B). Previously, we showed that tumors sensitive to tyrosine kinase inhibition (based on serial radiologic assessment prior to surgery) contained more CD3⁺ and CD8⁺ T cells, but a lower percentage of CD4⁺ T cells and T regulatory cells (Treg) compared with untreated and resistant tumors (6). Here, we found that compared with untreated tumors, there was higher PD-1 expression on CD4⁺ T cells in sensitive tumors and CD8⁺ T cells in tumors with acquired resistance to imatinib (i.e., increasing tumor size or vascularity after a period of stable or shrinking disease by serial radiologic assessment; Fig. 2C). There was no association between PD-1 expression and type of oncogene mutation (KIT, PDGFRA, or wild-type), tumor location, size, or mitotic rate (data not shown). In 10 patients who had multiple tumors removed during surgery, we found a similar expression pattern of inhibitory receptors within individuals, despite differences in tumor location, size, and mitotic rate (Fig. 2D).

To assess the expression of PD-L1 in human GISTs, we evaluated human GIST specimens by IHC. PD-L1 expression was low in most cases and tended to be focal (Fig. 3A). PD-L1 was present on tumor cells, but also on some intratumoral leukocytes. To ascertain the relative location of PD-L1 within the tumor, we measured PD-L1 mRNA levels in freshly isolated tumor (CD45⁻KIT⁺) and stroma (CD45⁻KIT⁻) cells from two resistant human GISTs. PD-L1 mRNA was equally distributed between these two compartments (Fig. 3B). Furthermore, we performed real-time PCR on 41 bulk human GIST samples. The PD-L1 mRNA level was variable but did not appear to correlate with treatment response to tyrosine kinase inhibition (Fig. 3C), mutation (Fig. 3D), or tumor size (Fig. 3E). There was a weak inverse relationship between mitotic rate and PD-L1 expression (Fig. 3F). Overall, intratumoral PD-L1 expression was variable and heterogeneous.

Next, we found that PD-L1 expression was low on three imatinib-sensitive human GIST cell lines (Fig. 4A). Because IFNγ is known to induce PD-L1 expression (12), we tested its effect on GIST cell lines. IFNγ upregulated PD-L1 mRNA in GIST-T1 cells over time as measured by real-time PCR (Fig. 4B). Similar results were observed in GIST882 and HG129 cells (data not shown). IFNγ-induced upregulation of PD-L1 transcripts was abolished with either the JAK inhibitor pyridine 6 or the KIT inhibitor imatinib. IFNγ treatment also induced PD-L1 protein expression, but imatinib abrogated the effect as measured by flow cytometry (data not shown) and Western blot analysis (Fig. 4C). The mechanism appeared to be partly through the inhibition of STAT-1 activation, as JAK inhibition by pyridine 6 also reduced PD-L1 upregulation and, like imatinib, was associated with decreased phosphorylated STAT1 (Fig. 4C) and STAT1 mRNA levels (data not shown). Furthermore, STAT1 knockdown by siRNA in GIST-T1 cells blocked IFNγ-induced upregulation of PD-L1 protein and mRNA (Fig. 4D and E). STAT1 knockdown was confirmed by Western blot analysis (Fig. 4D). We next measured the mRNA levels of IFNγ receptor (IFNGR1) and its major downstream signaling component IFN response factor 1 (IRF1). Notably, imatinib reduced the IFNγ-mediated increase in IFNGR1 and IRF1 mRNA (Fig. 4F). Collectively, IFNγ played a significant role in the regulation of PD-L1 expression in human GIST cells, but its effect was inhibited by imatinib.

To determine whether imatinib also alters tumor IFNγ signaling in Kit^V558Δ/+ mice, we assessed IFNγ-related genes in our previously published mRNA microarray data (6). After 1 week of imatinib therapy, the expression of multiple IFNγ-related genes was reduced in bulk tumor (Fig. 5A). Pdl1 was not present on the array. Notably, PD-L1 expression was markedly reduced after 4 weeks of imatinib by IHC (Fig. 5B). To further investigate the extent to which imatinib treatment modulates PD-1 and PD-L1 expression in GISTs, we performed flow cytometry on intratumoral T cells and tumor cells (Fig. 5C). PD-L1 was present on tumor cells and decreased with 1 week of imatinib treatment. PD-L2 expression was minimal (data not shown). T-cell subsets within the tumor expressed both PD-1 and PD-L1 at baseline, but imatinib had little to no effect on expression levels. CD8⁺ T cells from the tumor-draining lymph node and spleen generally had very low PD-1 expression (Fig. 5D). Compared with intratumoral PD-1⁻CD8⁺ T cells, PD-1⁺CD8⁺ T cells from the tumor produced only minimal amounts of IFNγ after stimulation with PMA and ionomycin, suggesting that these cells are dysfunctional (Fig. 5D). However, in vitro treatment of bulk intratumoral T cells from untreated Kit^V558Δ/+ mice with anti-PD-1 antibody increased their production of both TNF and IFNγ (Fig. 5E). In contrast, T cells from the mesenteric lymph node of wild-type mice had little cytokine production at baseline or after PD-1 blockade. Thus, PD-1 blockade preferentially affected tumor-infiltrating T cells compared with T cells from the mesenteric lymph node, which essentially lacked PD-1 expression.

Given the presence of PD-1 and PD-L1 in both human and murine GISTs, we hypothesized that blockade of the PD-1/PD-L1 axis would have an antitumor effect. We treated Kit^V558Δ/+ mice with imatinib or vehicle combined with PD-1 or PD-L1 antibodies or isotype control for 4 weeks. The combination of imatinib and PD-1/PD-L1 blockade was not associated with noticeable toxicity in our model, and mice did not show signs of autoimmunity. Notably, anti-PD-1 and anti-PD-L1 altered tumor weight only when combined with imatinib (Fig. 6A). The combination treatment decreased phosphorylated KIT, phosphorylated STAT1, IDO, and TGFβ and its upstream mediator phosphorylated SMAD2 (Fig. 6B). Furthermore, Pdl1 and Ido1 mRNA levels were downregulated after 4 weeks of treatment (Fig. 6C). Importantly, the enhanced antitumor effect of imatinib and anti-PD-1 persisted at 3 months of treatment (Fig. 6D) and was detectable as early as 1 week (Fig. 6E). After 1 week of combined treatment, we noticed decreased KIT staining, indicating a reduction of tumor cells, especially in tumors from mice treated with imatinib plus anti-PD-L1, and increased TUNEL staining, indicating increased tumor cell apoptosis (Fig. 6F). There was no change in the frequency of CD4⁺ or CD8⁺ T cells or Tregs after 1 or 4 weeks of combination treatment (data not shown). Remarkably, intratumoral CD8⁺ T cells showed increased proliferation and inflammatory cytokine production upon combination treatment for 1 week compared with imatinib alone (Fig. 6G). In accordance with our observations from the mRNA microarray, IFNγ protein and two of the chemokines it induces, CXCL9 and CXCL10, were downregulated after 1 week of treatment (Fig. 6H). Furthermore, PD-1 blockade combined with the IDO inhibitor 1-MT was more effective after 1 week of treatment in our GIST mouse model than IDO inhibition alone (Fig. 6I). Thus, IDO inhibition (which also occurs after imatinib therapy alone) is sufficient to enable the antitumor effect of anti-PD-1. Taken together, we showed that KIT inhibition combined with PD-1/PD-L1 blockade reduced tumor weight in Kit^V558Δ/+ mice, which was associated with increased tumor cell apoptosis and enhanced frequency of cytokine-producing CD8⁺ T cells.

In this study, we found that the PD-1/PD-L1 axis contributes to tumor immune evasion in GISTs. Among the inhibitory receptors analyzed on intratumoral T cells in human GISTs, PD-1 was expressed at the highest frequency. Notably, the PD-1 expression on CD4⁺ and CD8⁺ T cells within sensitive and resistant human GISTs had a bimodal distribution, with a distinct population having very high expression, suggesting that a subset of patients might particularly benefit from PD-1/PD-L1 blockade. The level of PD-1 expression on intratumoral CD4⁺ T cells correlated with that on CD8⁺ T cells within a patient. Furthermore, in patients with multiple tumors, the different tumors had similar amounts of inhibitory receptors on intratumoral T cells, despite differences in tumor location, size, and mitotic rate. This observation does not resolve whether the antitumor immune response is either tumor driven, despite the known heterogeneity among tumor subclones, or patient driven, as systemic host factors may be responsible for a consistent T-cell response. PD-1 expression has been demonstrated to identify T cells that recognize tumor-specific proteins (26). PD-1⁺CD8⁺ T cells accumulated in murine GISTs but produced less IFNγ upon in vitro stimulation compared with PD-1⁻CD8⁺ T cells, suggesting that PD-1 marks functionally impaired intratumoral T cells in our model. Nevertheless, bulk intratumoral T cells made inflammatory cytokines after treatment with anti-PD-1 in vitro, as did intratumoral CD8⁺ T cells from Kit^V558Δ/+ mice that had been treated with anti-PD-1 in vivo.

PD-L1 expression in human GISTs was variable and did not correlate with treatment status, tumor mutation, and tumor size. Tumors with a lower mitotic rate had a very modest correlation with higher PD-L1 expression, which has been suggested by prior data (27). We showed that PD-L1 expression in human GISTs was similar between tumor and stromal cells. It is unclear whether PD-L1 expression by either tumor or immune cells is required for response to anti-PD-1/PD-L1 therapy. In a previous report, a subset of melanoma patients with PD-L1⁻ tumors responded to PD-1 blockade, suggesting that PD-L1 expression may not be necessary for response (28). Investigation of this question is hampered by the suboptimal antibodies for PD-L1 IHC. In fact, in several human GISTs, we found only focal immunostaining despite relatively high PD-L1 mRNA levels.

Multiple mechanisms have been shown to drive PD-L1 expression. First, PD-L1 can be induced by IFNγ (12, 29). Although human GIST cell lines had low PD-L1 expression, IFNγ treatment markedly increased it, likely through STAT1. IFNγ generally enhances the inflammatory response but may also promote immunosuppression by reducing tumor recognition and T cell–mediated lysis (30). PD-L1 expression has also been linked to EGFR mutations, PI3K/AKT signaling, and, in BRAF-resistant melanoma cell lines, reactivation of the MAPK pathway (31–33). In GISTs, PD-L1 expression may conceivably be induced by other cytokines, signaling pathways, and immune cells within the tumor microenvironment, as well as by IFNγ. Oncogene inhibition downregulates PD-L1 expression in melanoma cell lines (33, 34). Similarly, in our mouse model, imatinib decreased PD-L1 mRNA, protein, and cell-surface expression. In addition, imatinib treatment of GIST cell lines abrogated the effect of IFNγ on PD-L1 expression. The effects of imatinib were likely mediated by suppression of JAK/STAT and PI3K/AKT, which are both downstream of KIT, and by repression of IFNγ and IFNγ-related genes. The reduction in IFNγ is consistent with the previous finding that imatinib reduced MHC class I expression in human GISTs (35) and our demonstration that imatinib lowered class II in murine and human tumor-associated macrophages in GISTs (36).

Tumor cells treated with targeted therapy were recently shown to induce a reactive secretome that promotes the survival of sensitive cells, and the expansion and dissemination of drug-resistant clones (37). This therapeutic obstacle might be overcome by combining targeted therapy with mobilization of the immune system. Indeed, our data demonstrated that anti-PD-1 and anti-PD-L1 in Kit^V558Δ/+ mice with established tumors increased the effects of imatinib by enhancing CD8⁺ T-cell function, resulting in substantial tumor cell apoptosis. Administered alone, anti-PD-1 and anti-PD-L1 were ineffective. It seems unlikely that pretreatment with the blocking antibodies would further improve the combination of PD-1/PD-L1 blockade and imatinib. Imatinib-induced tumor cell death could sensitize GISTs to PD-1/PD-L1 blockade through the release of endogenous tumor antigens that subsequently activate the immune system. However, the rapid tumor response by 1 week makes it more likely that a preexisting immune response was amplified. It does not seem that PD-1 inhibition in our model reduced tumor growth independent of the adaptive immune system, a mechanism recently shown in human melanoma tumor cells, which frequently express PD-1 (38).

The other contributing factor that explains the efficacy of combining imatinib with anti-PD-1 or anti-PD-L1 is the inhibition of IDO. IDO is an enzyme that catalyzes the degradation of the essential amino acid tryptophan to kynurenine, whose metabolites suppress T cells. Previously, we showed that imatinib reduces tumor cell production of IDO in our model by decreasing the levels of the transcription factor ETV4, which regulates IDO transcription (6). Recent studies have suggested that elevated expression of metabolic enzymes (e.g., IDO) and inhibitory molecules (e.g., PD-1) is induced by IFNγ as an adaptive response by the tumor (29, 39). Our data imply that IDO inhibition by imatinib partially accounts for the antitumor efficacy of concomitant PD-1/PD-L1 blockade. This supposition is consistent with a previous report showing that IDO inhibition augmented the efficacy of T-cell immunotherapy in B16 melanoma (40). Furthermore, tumor regression after therapeutic PD-1 blockade has been demonstrated to require preexisting CD8⁺ T cells that are negatively regulated by PD-1/PD-L1–mediated adaptive immune resistance. Response to pembrolizumab (anti-PD-1) was associated with higher numbers of CD8⁺, PD-1⁺, and PD-L1⁺ cells in the tumor microenvironment before treatment (41). The combination of imatinib and PD-1/PD-L1 blockade may be most effective in treatment-naïve tumors, which are the most sensitive to oncogene inhibition. Melanomas with high levels of somatic mutations have been shown to be more likely to respond to checkpoint immune blockade (42). However, our data show that murine GISTs, which have only one mutation, respond to PD-1/PD-L1 blockade in the setting of tyrosine kinase inhibition.

In conclusion, PD-1 was expressed at high levels on tumor-infiltrating T cells in human GISTs, while PD-L1 levels were variable. In a mouse model of GIST, anti-PD-1 and anti-PD-L1 had no antitumor effect when used alone but did increase the efficacy of imatinib. The mechanism involved a reduction of IDO levels and an increase in CD8⁺ T-cell function. PD-1/PD-L1 blockade is a promising strategy to improve the effects of targeted therapy in GISTs.

No potential conflicts of interest were disclosed.

Conception and design: A.M. Seifert, S. Zeng, T.S. Kim, N.A. Cohen, R.P. DeMatteo

Development of methodology: A.M. Seifert, S. Zeng, T.S. Kim, P. Besmer, R.P. DeMatteo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.M. Seifert, S. Zeng, T.S. Kim, M.J. Beckman, J.H. Maltbaek, J.K. Loo, M.H. Crawley, C.R. Antonescu, R.P. DeMatteo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.M. Seifert, S. Zeng, T.S. Kim, N.A. Cohen, M.J. Beckman, M.H. Crawley, F. Rossi, P. Besmer, C.R. Antonescu, R.P. DeMatteo

Writing, review, and/or revision of the manuscript: A.M. Seifert, S. Zeng, J.Q. Zhang, N.A. Cohen, M.J. Beckman, B.D. Medina, J.H. Maltbaek, J.K. Loo, F. Rossi, C.R. Antonescu, R.P. DeMatteo

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.M. Seifert, S. Zeng, T.S. Kim, R.P. DeMatteo

Study supervision: A.M. Seifert, R.P. DeMatteo

We are grateful to the Tissue Procurement Service for assistance in the acquisition of human tumor specimens. We thank members of the Sloan Kettering Institute Laboratory of Comparative Pathology, Molecular Cytology, Flow Cytometry, and Pathology Core Facilities, Colony Management Group, and Research Animal Resource Center. We thank Lieping Chen for giving us the murine anti-human PD-L1 monoclonal antibody (clone 5H1) and Jacqueline Bromberg for providing the JAK inhibitor pyridine 6. We thank Russell Holmes for logistical and administrative support.

The investigators were supported by NIH grants R01 CA102613 and T32 CA09501, Cycle for Survival, Swim Across America, Stephanie and Fred Shuman through the Windmill Lane Foundation, and David and Monica Gorin (to R.P. DeMatteo), GIST Cancer Research Fund (to R.P. DeMatteo and C.R. Antonescu), F32 CA162721 and the Claude E. Welch Fellowship from the Massachusetts General Hospital (to T.S. Kim), F32 CA186534 (to J.Q. Zhang), P01 CA47179 (to C.R. Antonescu), and P50 CA140146-01 (to C.R. Antonescu and P. Besmer). The Molecular Cytology and Flow Cytometry Core Facilities were supported by Cancer Center Support Grant P30 CA008748.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.