Oncogenic KIT Modulates Type I IFN–Mediated Antitumor Immunity in GIST

There is renewed interest in harnessing type I IFN and their downstream effects for antitumor immunity (1–3). However, IFNs and oncogenes often share signaling pathways. For example, mutant EGFR and mutant KRAS in colorectal cancer, and mutant BRAF in melanoma can affect type I IFN signaling (4–6). Therefore, targeting oncogenes may influence the type I IFN immune response and vice versa.

Type I IFNs include several classes of IFNα and IFNβ1, which bind to a heterodimer composed of IFN receptors IFNAR1 and IFNAR2 (7, 8). The activated IFNAR heterodimer phosphorylates the JAK/STAT pathway and induces signal transduction. Although all cells can produce type I IFNs, macrophages, dendritic cells (DC), and neutrophils are the predominant producers of these cytokines in the tumor microenvironment (1, 8). Activation of intracellular innate immune sensors, such as stimulator of IFN genes (STING), via danger-associated molecular patterns (DAMP), rapidly initiates transcription of genes encoding type I IFNs (2, 9, 10).

Type I IFNs have pleiotropic roles in the host immune response to cancer. Cytotoxic CD8⁺ T cells rely on type I IFN signaling during cross-presentation, priming, and subsequent targeting of tumor cells via antigen loaded on MHC class I. MHC class I gene expression is under the transcriptional regulation of the MHC class I transactivator (CITA), which is an IFN-inducible nuclear protein complex that is epigenetically or genetically altered in many cancers (11). Tumor disruption of the JAK/STAT pathway downstream of IFN is another mechanism of MHC class I downregulation and can result in immune escape during checkpoint blockade (12). In addition, tumor-intrinsic STAT signaling is implicated in the expression of chemokines, such as CXCL10, which recruit cytotoxic T cells and can contribute to the efficacy of chemotherapy (13).

KIT is an oncogenic protein that can directly activate STAT signaling (14). Mutations in KIT are the most frequent cause of gastrointestinal stromal tumor (GIST). GIST is the most common human sarcoma and serves as a paradigm for a single mutation in an oncogene driving cancer development and progression (15, 16). Human GIST is recapitulated in Kit^V558Δ/+ mice, in which a single germline heterozygous knock-in deletion in Kit exon 11 gives rise to a cecal GIST with 100% penetrance (17). Activation of KIT stimulates multiple signal transduction pathways, including PI3K/AKT, MAPK/ERK signaling, and STAT1, -3, and -5 signaling pathways (18). The small molecule imatinib is highly effective in blocking the ATP-binding pocket of the mutant KIT tyrosine kinase and demonstrates the success of targeted therapy (19).

In this study, we hypothesized that because KIT activates STAT1 in GIST, KIT inhibition with imatinib would reduce tumor type I IFN signaling. We found that GIST mice lacking global type I IFN signaling had larger tumors with reduced MHC class I expression and an altered immune infiltrate. Imatinib also reduced tumor MHC class I antigen presentation, thereby modifying tumor-specific immunity. Nevertheless, a STING agonist could counter some of the negative immune effects of imatinib by enhancing type I IFN signaling. Last, we correlated these findings in mice to human GIST, showing that imatinib therapy may limit tumor immunogenicity through decreased tumor human lymphocyte antigen (HLA) class I expression. Our results highlight the complexity of the effects of targeted therapy on the interplay between tumor-cell growth and immunogenicity.

Six- to 12-week-old Kit^V558Δ/+ and Kit^{V558Δ;T669I/+} mice (17, 20) and Ifnar^−/− mice (ref. 21; purchased from The Jackson Laboratory) were maintained in a specific pathogen–free animal facility. All the mice were on a C57BL/6J background and were age- and sex-matched for experiments. All mice were used with the approval of and in accordance with an Institutional Animal Care and Use Committee protocol at the University of Pennsylvania.

GIST cell lines T1, HG129, and T1R have been previously described (22–24). T1 cells were a gift from Dr. Taguchi (Kochi Medical School, Japan). HG129 and T1R cells were established in our laboratory. Cell lines were first authenticated with sequencing for their respective KIT mutations and have not been reauthenticated as they have retained their phosphorylated KIT expression. Cells were maintained in serum-complete RPMI medium: RPMI modified with glutamine, HEPES, and phenol red (Gibco, cat. #22400089), and supplemented with 10% FCS (Lab Force, cat. #FB5002-H), 0.05 mmol/L β-mercaptoethanol (Sigma-Aldrich, cat. #M3148), and 1× penicillin–streptomycin (Gibco, cat. #15140-122). Cells were passaged 2 to 3 times prior to use. Mycoplasma testing was last performed in January 2018.

Cells were treated for 6 to 24 hours in serum-complete RPMI medium with 1,000 U/mL IFNα (PBL Assay Science, cat. #11100-1), 100 ng/mL IFNγ (R&D Systems, cat. #285-IF), or 100 ng/mL TNF (R&D Systems, cat. #210-TA-005). Imatinib was obtained from Novartis and dissolved in sterile H₂0. JAK inhibitors ruxolitinib (100 nmol/L) and PF-06700841 (100 nmol/L) were purchased from Selleckchem (cat. #S1378, S8804, respectively) and dissolved in DMSO. siRNA knockdown of STAT1 (Invitrogen, cat. #4390824) and KIT (Invitrogen, cat. #4392420) was achieved using Lipofectamine (Invitrogen, cat. #11668027) transfection with nontargeting negative controls (Invitrogen, cat. #4390843) according to the manufacturer’s protocol.

For mouse experiments, imatinib was obtained from Novartis and dissolved in the drinking water at 600 mg/L. For antibody depletion, 8- to 10-week-old age- and sex-matched mice were given 200 μg/mouse of anti-IFNAR (clone MAR1-5A3, Bio X Cell), 200 μg/mouse of anti-CD8 (clone YTS169.4, Bio X Cell), 400 μg/mouse of anti-CSF1R (clone AFS98, Bio X Cell), or their corresponding isotype control antibodies every other day for 2 weeks. The STING agonist 5,6-dimethylxanthenone-40 acetic acid (DMXAA; Selleckchem, cat. #1537) was dissolved in 7.5% NaHCO₂ and given every third to fourth day at 200 μg/mouse to 6-week-old mice for 21 days. Control mice were given vehicle.

Harvested tumors were minced and digested in Liberase TL (Roche, ref# 5401020001) for 30 minutes at 37°C. After quenching with FBS, tumor fragments were passed through 100 μm nylon cell strainers, washed in PBS with 1% FBS, and passed through a 40 μm strainer. After processing into cell suspensions, all samples were immediately analyzed by flow cytometry or used to isolate specific cell populations, as described below.

Flow cytometry was performed using an LSRFortessa (BD) and analyzed on FlowJo v10 (BD). Fc receptors were blocked using anti-CD16/32 (clone 2.4G2, Bio X Cell). Antibodies specific for mouse and human antigens (Supplementary Table S1) conjugated to various fluorochromes were used to stain single cells. If necessary, cells were fixed using the eBioscience Fixation and Permeabilization Buffer Kit (cat. #88-8824-00) as per the manufacturer’s protocol. Intracellular cytokine staining was performed with the BD Cytofix/Cytoperm Kit (cat. #554714), as per the manufacturer’s protocol. For intracellular IFNγ and TNF staining, cells were stimulated with 50 ng/mL phorbol 12-myristate 13-acetate (PMA; Sigma-Aldrich, cat. #79346), 1 μmol/L ionomycin (Life Technologies, cat. #I24222), and 1 μg/mL lipopolysaccharide (LPS; Sigma-Aldrich, cat. #L6529) and incubated at 37°C with 5% CO₂ in the presence of 10 μg/mL brefeldin A (Affymetrix, cat. #00-4506-51) for 4 hours.

Tumor single-cell suspensions were incubated with mouse CD45 microbeads (Miltenyi Biotec, cat. #130-052-301), as per the manufacturer’s protocol, and run through two sequential LS columns per 3 × 10⁷ cells, saving the positive fraction. Kit⁺ tumor cell selection was performed by incubation of the CD45⁻ fraction with CD117 microbeads (Miltenyi Biotec, cat. #130-091-224), as per the manufacturer’s protocol, and collecting the positive fraction. CD45⁻Kit⁻ cells were collected from the remaining negative fraction. Macrophages were isolated using F480 microbeads (Miltenyi Biotec, cat. #130-110-443), as per the manufacturer’s protocol, and two sequential positive selections from CD45⁺. Purity of all collected cell populations was >90% by flow cytometry.

Total RNA was extracted from tumor tissues or cells using RNAeasy Plus Mini kit (QIAGEN, cat. #74134). One microgram RNA was reverse transcribed with the Taqman Reverse Transcription Kit (Applied Biosystems, cat. #N8080234) and amplified with PCR Taqman probes (Applied Biosystems; Supplementary Table S2), repeated in triplicates. Quantitative PCR was performed using a ViiA7 real-time PCR system (Applied Biosystems). Data were calculated by the 2^−ΔΔCt method as described in the manufacturer’s instructions and expressed as fold increase over GAPDH control.

Specimens were fixed in 4% paraformaldehyde and embedded in paraffin. Next the specimens were sectioned at 5-μm thickness and mounted on glass slides. Routine hematoxylin and eosin staining and IHC staining for Ki67 (Thermo Fisher Scientific, cat. #RM-9106-S) were then performed as described previously (17, 25). Percentage of Ki67⁺ cells was quantified by counting stained and unstained nuclei in five 100 × 100 μm high-power fields (HPF) of at least three different tumors per treatment group. Whole slides were scanned on the Aperio VERSA 200 platform.

Cell lysates were made by resuspending cell pellets in NP40 cell lysis buffer (Invitrogen, cat. #FNN0021) and incubating on ice for 30 minutes while vortexing every 10 minutes. Tissue lysates were made by homogenizing frozen tissue in tissue lysis buffer. Protein lysates were run on 4% to 12% gradient gel (Bio-Rad) and then transferred to nitrocellulose membrane (Millipore). Blots were blocked in 2.5% milk for 1 hour and then incubated in primary antibody (Supplementary Table S3) with 2.5% BSA at 4°C overnight. The next day, blots were washed and probed with secondary antibody [anti-rabbit IgG horseradish peroxidase (HRP)–linked, Cell Signaling Technology, cat. #7074, and anti-mouse IgG HRP-linked, Cell Signaling Technology, cat. #7076] in 2.5% milk. Chemiluminescence was detected after adding HRP substrate (Thermo Fisher Scientific, cat. #34075). For immunoprecipitation, cell lysate was incubated overnight with the primary antibody to phosphorylated Kit (Cell Signaling Technology, antibody #3391) and then pulled down with Protein A/G agarose beads (Cell Signaling Technology, cat. #9863P).

Tumor specimens were obtained from patients with GIST who underwent surgical resection of their tumors after approval by an Institutional Review Board protocol. Patients provided written informed consent prior to resection, and studies were conducted in accordance with the Declaration of Helsinki. Patient information remained deidentified and encrypted in compliance with Health Insurance Portability and Accountability Act regulations. Tissue was banked by flash freezing in liquid nitrogen and stored at −80°C until use.

RNA sequencing (RNA-seq) and bioinformatics were performed previously on mouse and human GIST treated with imatinib (26, 27). Raw counts of the indicated genes were normalized and analyzed using the R software package DESeq2. Raw datasets have been previously deposited through the Sequencing Read Archive under the accession number PRJNA521803.

Unpaired two-sample, two-tailed Student t test was performed on datasets using GraphPad Prism 7.0 (GraphPad Software) unless otherwise stated. A P value < 0.05 was considered significant. Data are shown as mean ± SEM. Spearman correlation was performed where applicable.

To investigate the interaction between oncogenic KIT signaling and type I IFN signaling, we treated Kit^V558Δ/+ mice with the targeted KIT inhibitor imatinib. Previously, we published RNA-seq of bulk tumor treated with 3 weeks of imatinib (27). Analysis of this dataset for type I IFN signaling genes showed a significant decrease in expression of Ifnar1, Ifnar2, Ifit1–3, and Stat1, as well as MHC class I–related genes (e.g., B2m and H2-K1; Fig. 1A). By gene set enrichment analysis, antigen presentation and JAK/STAT signaling were among the pathways significantly downregulated by imatinib therapy (Fig. 1B). To validate these findings, we examined MHC class I expression on tumor cells (CD45⁻Kit⁺) by flow cytometry (Supplementary Fig. S1A). After 1 and 4 weeks of imatinib, tumor cell MHC class I (H-2K^b) was 50% lower (Fig. 1C), but expression was restored after imatinib discontinuation (Supplementary Fig. S1B). In contrast, immune cell (CD45⁺) MHC class I expression was unchanged by imatinib (Fig. 1D). Ifnb1 mRNA levels in bulk Kit^V558Δ/+ tumors decreased by 90% after 1 week of imatinib and remained diminished at 3 weeks (Fig. 1E). Tumor-associated macrophages (TAM) proved to be the dominant source of IFNβ1 (Fig. 1F). Accordingly, depletion of intratumoral macrophages with an antibody to CSF1R, as we have done before (28), significantly reduced Ifnb1 mRNA levels (Fig. 1G). To exclude the possibility of off-target effects of imatinib (i.e., effects not mediated via Kit inhibition), we treated Kit^{V558Δ/+;T669/+} mice, which are resistant to imatinib due to a secondary Kit mutation (20). We found that imatinib did not alter tumor cell MHC class I expression or Ifnb1 mRNA levels (Fig. 1H). Thus, Kit signaling is linked to type I IFN and its downstream signaling.

To determine the clinical relevance of our findings from the murine GIST model, we used our published human GIST RNA-seq database (26) to examine 35 GIST surgical specimens that each contained an imatinib-sensitive mutation in KIT exon 11. Imatinib-treated human GISTs had lower expression of type I IFN signaling components and HLA class I genes (Fig. 2A). Similar results were obtained with RNA-seq of KIT⁺ cells isolated from human GISTs (Supplementary Fig. S2A; ref. 23). Consistent with the mouse data, human tumors that had responded to imatinib expressed significantly less IFNB1 mRNA than those that had not responded to imatinib therapy (Fig. 2B). Meanwhile, KIT⁺ cells did not express IFNB1 mRNA, complementary to our mouse findings. Consistent with the fact that STAT1 is a mediator and transcription factor for the type I IFN response, high STAT1 gene expression by immunoblot in untreated GISTs corresponded with detection of phosphorylated STAT1 as well as high expression of HLA proteins (Fig. 2C). STAT1 expression in human GISTs significantly correlated with gene expression of B2M and HLA A, B, and C (Fig. 2D). There were no discernible patient or tumor differences between tumors above or below the median level of STAT1 expression (Supplementary Table S4). Importantly, STAT1 expression correlated significantly with GIST immunogenicity, as characterized by tumor cytolytic activity (Cyt) score and immune score (Fig. 2D; refs. 26, 29–31). Collectively, our mouse findings appeared to be applicable to human GISTs.

The type I IFN receptor IFNAR was expressed by a variety of intratumoral cells in Kit^V558Δ/+ mice, but tumor cells expressed the most (Supplementary Fig. S3A). To assess whether type I IFN regulated MHC class I expression in vivo (7, 32), we administered a blocking antibody specific for IFNAR to Kit^V558Δ/+ mice. Tumor cell MHC class I expression was reduced by 40% (Fig. 3A). IFNAR blockade also diminished tumor STAT1 activation, as did imatinib after 2 weeks (Fig. 3B).

To further establish the role of type I IFN in GIST, we crossed Kit^V558Δ/+ mice to Ifnar^−/− mice (21). Tumor cell expression of IFNAR was reduced in heterozygous Ifnar^+/−Kit^V558Δ/+ mice and nearly abrogated in homozygous Ifnar^−/−Kit^V558Δ/+ mice (Supplementary Fig. S3B). Tumor size was almost 4 times larger in homozygous Ifnar^−/−Kit^V558Δ/+ mice compared with Kit^V558Δ/+ mice expressing IFNAR (Fig. 3C). Tumors were also larger in heterozygous Ifnar^+/−Kit^V558Δ/+ mice, suggesting that even a reduction in type I IFN signaling impaired tumor control. In concert with our results using a blocking IFNAR antibody in Kit^V558Δ/+ mice, tumor cells in Ifnar^−/−Kit^V558Δ/+ mice had significantly less MHC class I expression (Fig. 3D).

As type I IFN signaling is important in the recruitment of antigen-presenting cells (APC) to tumors and their activation once there (1), we analyzed intratumoral macrophages and DCs. Ifnar^−/−Kit^V558Δ/+ tumors had a similar percentage of intratumoral macrophages as defined by CD11b⁺F480⁺ expression (Fig. 3E). Macrophage activation, represented by MHC class II expression, was also unchanged (Supplementary Fig. S3C). Furthermore, there was no significant difference in the percentage of DCs (defined as CD11c⁺MHCII⁺) or the proportion of Batf3 DCs (defined as CD103⁺CD11b⁻; Fig. 3E; Supplementary Fig. S3C; ref. 33).

To elucidate the mechanism of reduced tumor control in Ifnar^−/−Kit^V558Δ/+ mice, we examined tumor-infiltrating lymphocytes (TIL). There was no change in the percentage of CD8⁺ T cells (Fig. 3F) or other lymphoid cells (Supplementary Fig. S3D). However, intratumoral CD8⁺ T cells from Ifnar^−/−Kit^V558Δ/+ mice were phenotypically different from those in Kit^V558Δ/+ mice. There was a dramatic reduction in PD1⁺CD103⁻CD8⁺ T cells (Fig. 3F), which we previously showed to be the major effectors of antitumor immunity in our GIST model (34). Meanwhile, tumor cell expression of PDL1 was unchanged (Supplementary Fig. S3E; ref. 25). Furthermore, Ifnar^−/−Kit^V558Δ/+ TILs produced less TNF and IFNγ after in vitro stimulation (Fig. 3G). To investigate the importance of CD8⁺ T cells in Ifnar^−/−Kit^V558Δ/+ mice, we depleted them with an antibody and found no difference in tumor growth (Fig. 3H). In contrast, CD8⁺ T-cell depletion in control Kit^V558Δ/+ mice resulted in substantial tumor growth, implying that CD8⁺ T cells in Ifnar^−/−Kit^V558Δ/+ mice had little antitumoral function. Ifnar^−/−Kit^V558Δ/+ tumors had intact Kit activation, but similar to treatment with the IFNAR-blocking antibody, STAT1 activation was attenuated (Fig. 3I). Collectively, type I IFN is important in the antitumor immune response in GIST.

To investigate whether type I IFN directly induced MHC class I expression on tumor cells, we utilized the human GIST cells lines T1 and HG129 because there are no murine GIST cell lines that retain KIT expression. IFNα alone stimulated HLA class I expression on T1 cells, whereas IFNγ and TNF did not (Fig. 4A). IFNα, IFNγ, and TNF upregulated similar IFN-stimulated genes, although only IFNα and TNF increased IFNB1 transcription (Fig. 4B). To identify the impact of KIT signaling, we treated human T1 and HG129 cells with both IFNα and imatinib. Imatinib impaired IFNα-mediated upregulation of HLA class I expression (Fig. 4C). The effect was Kit-specific as KIT siRNA was sufficient to blunt the increase in HLA class I expression by IFNα (Supplementary Fig. S4A) and imatinib had no effect on IFNα-induced HLA class I expression in T1R cells (Supplementary Fig. S4B), an imatinib-resistant cell line (24).

Consistent with the principal role of STAT1 in IFN signal transduction, STAT1 siRNA abolished the individual effects of IFNα and imatinib on HLA class I expression (Fig. 4D). By immunoblot, IFNα increased phosphorylated STAT1 and total STAT1, and this effect of IFNα was inhibited by imatinib (Fig. 4E). We next investigated which kinase was responsible for activating STAT1. In short culture (6 hours), IFNα stimulation did not alter levels of phosphorylated JAK1 (Fig. 4F), the canonical kinase in type I IFN signaling (8). JAK inhibitors also did not inhibit STAT1 activation in the presence of IFNα, whereas imatinib did. Meanwhile, STAT1 coimmunoprecipitated with phosphorylated KIT, and this association increased after IFNα stimulation (Supplementary Fig. S4C). These data suggest that KIT signaling is required for activation of STAT1 in GIST during type I IFN responses.

Given its importance in tumor control, we tested whether further enhancing type I IFN signaling could be exploited for therapeutic benefit. We tested the murine STING agonist DMXAA, which initiates a cytosolic innate immune signaling cascade that raises transcription of genes encoding type I IFNs (2, 10, 35). Previously, STING has been delivered intratumorally in subcutaneous tumor models, which is not practical for an intra-abdominal tumor like GIST (2). After optimizing a tolerable dosing regimen, we found that 3 weeks of DMXAA alone reduced tumor weight in Kit^V558Δ/+ mice by one third (Fig. 5A). Consistent with our in vitro studies, this was associated with a significant increase in tumor Ifnb1 mRNA (Fig. 5B), STAT1 activation (Fig. 5C), and MHC class I expression (Fig. 5D). Furthermore, although DMXAA treatment did not alter the number or activation of intratumoral APCs (Fig. 5E), there was a significant increase in total intratumoral CD8⁺ T cells as well as PD1⁺CD103⁻CD8⁺ T cells (Fig. 5F), complementing our findings in IFNAR-deleted GIST. In addition, the downstream IFN gene Cxcl10 was upregulated, and there was a concomitant increase in TNF production (Fig. 5G), implicating both tumoral recruitment and activation of CD8⁺ T cells after DMXAA treatment. The effect of DMXAA on GIST tumor weight was dependent on CD8⁺ T cells, as their depletion abrogated the antitumor effects of DMXAA (Fig. 5H).

Given that imatinib reduced tumor type I IFN signaling and MHC class I expression, we investigated whether STING activation could overcome these immunosuppressive effects. Neither imatinib followed by DMXAA nor simultaneous therapy altered the effect of imatinib on tumor weight in Kit^V558Δ/+ mice. We discovered that STING (TMEM173), as well as other innate immune sensors, were downregulated after imatinib treatment in new analysis of our published 3-week RNA-seq data (ref. 26; Fig. 6A). Therefore, we tested sequential therapy (Fig. 6B). We pretreated GIST mice with 10 days of DMXAA followed by 11 days of imatinib. To determine the effect of DMXAA pretreatment, we compared this combination with imatinib-only mice that received drug starting on day 10. The addition of DMXAA further reduced tumor weights by 30% compared with imatinib alone (Fig. 6C), and these tumors exhibited significantly less Ki67 (Fig. 6D). DMXAA blunted MHC class I downregulation by imatinib (Fig. 6E) and significantly increased the percentage of PD1⁺CD103⁻CD8⁺ T cells (Fig. 6F), suggesting that the additional antitumor effect was driven in part by the immune response. DMXAA partially rescued the inhibition of STAT1 signaling by imatinib without altering KIT signaling (Fig. 6G), showing that repeated innate stimulation could boost tumor type I IFN signaling.

Although the central importance of KIT signaling in GIST viability is established, we provide evidence that it also shapes the antitumor immune response. The driver oncogene KIT is paramount in orchestrating the immune landscape through downstream signaling, which is not surprising because it is sufficient for GIST tumorigenesis, at least in mice. Previously, we showed that KIT signaling modulates polarization of intratumoral macrophages, activation of T cells, and maturation of Baf3 DCs (24, 28, 34). Here, we have demonstrated that oncogenic KIT also controls intratumoral type I IFN signaling, directly and indirectly. Inhibition of KIT activation in tumor cells reduced intrinsic type I IFN signaling, resulting in less MHC class I expression in both murine and human GIST. A prior immunohistochemistry study also showed lower HLA class I expression in patients treated with imatinib (36, 37). Decreased MHC class I expression on tumor cells should make them targets of natural killer (NK) cells, but we did not find evidence of increased NK-cell number or activation in the IFNAR-deleted GIST. We also found that interruption of KIT signaling indirectly decreased type I IFN production, which may be explained by a reduction in the number of intratumoral macrophages, the predominant source of type I IFNs in the tumor. This complements our previous finding that imatinib polarizes TAMs to an M2-like phenotype (28). Although the composition and activation of DCs and macrophages were unchanged in IFNAR-deleted GIST, it is likely that their antigen-presenting function was reduced, as evidenced by the reduced recruitment and function of intratumoral T cells.

The molecular basis of this type I IFN signaling modulation is through STAT1. It is known that ligand-induced phosphorylated KIT can bind directly to and activate STAT1 (38). Although mast cells with mutated KIT (D816V) have constitutively active STAT1 (14), there is relatively little constitutively active STAT1 in GIST cell lines (39), which is consistent with our findings of low levels of phosphorylated STAT1 in the unstimulated GIST cell lines. We found that type I IFN stimulation did not further activate the canonical JAK1, but it did potently increase Stat1 phosphorylation and transcription in a manner that depended on oncogenic KIT. Thus, oncogenic KIT and type I IFN converge on STAT1 in GIST. Although there is evidence that IFNs can trigger other non-JAK activation of STATs, such as via PI3K or p38 (40), our data suggest that type I IFN may potentiate oncogenic KIT activation of STAT1 in GIST. In fact, phosphorylated KIT coimmunoprecipitated with STAT1, and this was increased in the presence of type I IFN. STAT1 proteins can be found on cell surface plasma membrane rafts (41), where they are described to preassociate with IFNAR (42). IFN-induced receptor heterodimerization may bring STAT1 in close proximity to mutant KIT for cross-phosphorylation and would be another example of cytokine and growth factor receptor cross-talk (43, 44).

Our human GIST data highlight the importance of STAT1 in tumor immunogenicity, as STAT1 gene expression was significantly correlated with Immune score and Cyt score. Although imatinib decreased intratumoral expression of IFNB1 and can partly explain the decreased STAT1 expression, there was also a subset of untreated human GISTs with low STAT1 expression. Thus, there may be extrinsic factors in the microenvironment or additional signaling pathways that contribute to STAT1 expression in human GIST. Although type II IFN (IFNγ) can also activate JAK/STAT and share similar downstream functions, we found that IFNγ did not induce HLA I expression in GIST cell lines. Furthermore, we previously produced an Ifng^−/−Kit^V558Δ/+ GIST (28), which did not have appreciable changes in immune infiltrate or T-cell function.

Traditional chemotherapy and radiotherapy increase type I IFN production and signaling in the tumor environment, as apoptotic cells and DNA damage induce DAMPs and innate immune sensors such as STING (13, 35, 45). Our GIST mouse model and human GIST correlates reveal that targeted therapy has the opposite effect, because imatinib decreased type I IFN production and tumor MHC class I expression. To compensate for this seemingly undesirable effect, we used a STING agonist to augment type I IFN signaling. Some of the immunosuppressive effects of imatinib were blunted, but only when DMXAA was given prior to imatinib. Imatinib severely attenuated STAT1 activation after type I IFN stimulation in vitro and decreased the expression of nucleic acid sensors like STING in vivo, which may explain the loss of efficacy with simultaneous administration of STING agonist and imatinib. Similarly, we found previously that a CD40 agonist was only effective in Kit^V558Δ/+ mice when delivered prior to imatinib (46). This argues for optimizing the immune environment before imatinib therapy or dosing imatinib intermittently to allow for recovery of type I IFN signaling.

Our study demonstrates an immunomodulatory role for type I IFN in GIST and may have translational applications. Previously, a small, single-arm trial suggested benefit of administering IFNα during imatinib therapy (47). Intriguingly, treatment also induced the intratumoral infiltration of IFNγ-producing CD8⁺ T cells. Given that imatinib reduces type I IFN signaling, this innate immune pathway may be exploited clinically to enhance the effects of tyrosine kinase inhibition. Our findings may also be applicable to other tumors that rely on oncogene activation.

T.G. Bowler is an employee of Regeneron Pharmaceuticals. R.P. DeMatteo reports grants from Blueprint Medicines during the conduct of the study. No disclosures were reported by the other authors.

M. Liu: Conceptualization, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. M.S. Etherington: Conceptualization, data curation, formal analysis, validation, investigation, writing–review and editing. A. Hanna: Data curation, formal analysis, validation, writing–review and editing. B.D. Medina: Data curation, formal analysis, methodology, writing–review and editing. G.A. Vitiello: Conceptualization, formal analysis, writing–review and editing. T.G. Bowler: Formal analysis, validation. N.J. Param: Data curation, validation. L. Levin: Data curation, validation. F. Rossi: Supervision, funding acquisition. R.P. DeMatteo: Conceptualization, resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

The investigators were supported by NIH grants R01 CA102613, the David Foundation, the GIST Cancer Research Fund, and Betsy Levine-Brown and Marc Brown (R.P. DeMatteo).

The authors are grateful to the assistance of the Office of Laboratory Animal Welfare, University Laboratory Animal Resources, and flow cytometry staff at the University of Pennsylvania. They also thank Brent Buford for his logistical help and Steve Mervis for his administrative support.

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.