The impact of JAK1 pathogenic variants and MHC-I expression on response to immune checkpoint inhibition in endometrial cancer

Abstract

Purpose:

Immune checkpoint inhibitors (ICI) have increasing application in endometrial cancer, underscoring the need for robust biomarkers for patient selection. JAK1 pathogenic variants (PV) have previously been implicated in immune evasion. Here, we identify biomarkers predictive of ICI response in endometrial cancer and the implications of JAK1 PV in this context.

Experimental Design:

This is a translational study of tumors from 84 endometrial cancer patients treated with ICI. High-throughput proteomic-based profiling was used to quantify 193 phospho-/protein expression levels, including key JAK/STAT signaling pathway components. Associations with clinical outcomes were assessed using multivariate regression analysis. The functional consequences of JAK1 PV were explored through in vitro signaling assays and analyses of TCGA database.

Results:

MHC-I expression correlated with improved progression-free survival (p = 0.035), validated in orthogonal approaches. Notably, a subset of patients harboring JAK1 PVs demonstrated exceptional survival on ICI. In TCGA cohort of microsatellite instability-high (MSI-H) and DNA polymerase epsilon (POLE)-mutated tumors, homozygous loss of JAK1 (JAK1^Hom) trended toward decreased survival, whereas heterozygous loss of JAK1 (JAK1^Het) was associated with significantly improved survival (p = 0.026), suggesting partial retention of antigen presentation pathways. Among our ICI-treated MSI-H tumor samples, NK cell marker NCAM1 was associated with improved survival (p=0.02).

Conclusions:

These data support MHC-I as a potential predictive biomarker for ICI response in endometrial cancer. Additionally, we show that partial retention of JAK1 signaling in JAK1^Het tumors is associated with improved survival, potentially attributable to enhanced NK cell activity in tumors with low MHC-I expression.

Background

Endometrial cancer is currently the fourth most commonly diagnosed cancer among women in the United States and is one of the few cancers with rising incidence and mortality over the past two decades.^1,2 While the prognosis of early-stage endometrial cancer is relatively favorable– with 5-year survival rates approaching 95%– this dramatically declines in the setting of metastatic or recurrent disease, with 5-year survival rates ranging from 16–25%.³ Recent advances in the clinical management of endometrial cancer offer promise for improving outcomes, including the incorporation of molecular classification into treatment planning and prognostication⁴, as well as the expanding use of immune checkpoint inhibitors (ICIs) as part of the standard of care.⁵

The advent of ICIs has revolutionized cancer treatment across a growing number of clinical indications, including endometrial cancer. Pembrolizumab, an anti-Programed Death 1 (anti-PD-1) monoclonal antibody, was the first tumor-agnostic drug approved for patients with previously treated, advanced-stage microsatellite instability high (MSI-H) or mismatch repair deficient (dMMR) tumors based on findings from KEYNOTE-158.^6,7 The objective response rate (ORR) of endometrial cancer patients in this trial was approximately 50%. Subsequently, KEYNOTE-775 expanded the use of pembrolizumab to previously treated MMR proficient (pMMR) endometrial cancer in combination with lenvatinib, where the ORR was 30% in patients who received this combination, compared to 15% in patients who received traditional chemotherapy.^8,9 The role of ICIs was further expanded with trials like GY018 and RUBY, which investigated the addition of ICI to standard chemotherapy in primary advanced or recurrent endometrial cancer. In GY018, the addition of pembrolizumab resulted in an ORR of 50%, compared to 34% in the control group.¹⁰ RUBY similarly looked at the addition of dostarlimab, another anti-PD1 antibody, and demonstrated an overall survival benefit with a hazard ratio (HR) of 0.64 (95% CI, 0.46–0.87); P=0.0021.¹¹ The aforementioned studies show that ICIs have a clinically significant role in the treatment of endometrial cancer, however, they also show that a large proportion of patients still do not benefit from treatment, while potentially suffering from significant immune related adverse events. The current understanding of primary and acquired resistance to ICIs is limited and may vary by tumor type.

Major histocompatibility complex class I (MHC-I) is required for CD8+ T cell recognition and subsequent cytotoxic killing of tumor cells. While MHC class II (MHC-II) is traditionally expressed by professional antigen-presenting cells, its expression has also been observed in certain tumor cells. Additionally, MHC-II expression on tumor cells has been associated with improved responses to ICI therapy in certain cancers, such as breast cancer, melanoma, and Hodgkin’s lymphoma.^12–14 To our knowledge, the predictive value of MHC expression has not been evaluated in endometrial cancer. We hypothesize that expression levels of MHC molecules may be predictive of prolonged survival in ICI-treated endometrial cancer.

Given the tumor mutational landscape can influence ICI responsiveness – and consequently the predictive value of biomarkers – we sought to evaluate the prognostic and predictive relevance of potential biomarkers such as MHC in this context. Janus kinase-1 (JAK1) and the JAK-STAT proteins are important transducers of immune signals: disruption of the JAK/STAT pathway impedes the upregulation of antigen-presentation on tumor cells by the inflammatory cytokine interferon γ (IFNγ).^15,16 JAK1 pathogenic variants (PV) – most commonly truncating frameshift mutations at microsatellites in the coding sequence – have been implicated in immune evasion and ICI resistance across several cancer types including lung, melanoma, colorectal, and endometrial cancers.^17–21 JAK1 PVs are highly prevalent in endometrial cancer, being present in 14% of endometrial tumors based on data from The Cancer Genome Atlas (TCGA). Recent studies in colorectal and endometrial cancer have reported exceptional responses to ICIs in patients with JAK1 PV tumors, challenging our classic understanding that these mutations are implicated in ICI resistance.^22,23 A study of 24 MSI-H endometrial cancer tumors showed that pretreatment JAK1 mutations were not associated with resistance to pembrolizumab, and all three complete responders in this study harbored a JAK1 PV.²² These findings raise the possibility of an alternative biological explanation underlying ICI sensitivity in JAK1-mutant endometrial cancer.²²

To evaluate the predictive value of MHC and other potential markers of improved survival on ICIs for patients with endometrial cancer, we conducted a retrospective tumor tissue study in patients with known clinical outcomes. We contribute to the body of literature with the broader aim of improving patient outcomes in endometrial cancer by identifying tumor characteristics that will confer benefit from ICI through high-throughput proteomic analysis of 193 protein and phosphoprotein targets. We also investigate the mechanisms by which endometrial cancers with a dysfunctional JAK/STAT pathway can still have a robust response to PD-1 blockade. Together, these analyses provide insight into the complex intracellular relationships that drive ICI response, which can be harnessed to refine patient selection for immunotherapeutic agents and also identify novel therapeutic targets.

Patients and Methods

Study approval, patient tissue and clinical data attainment

This is a retrospective tissue analysis conducted across two institutions. Electronic medical records (EMRs) at Vanderbilt University Medical Center (Nashville, TN) and Cedars-Sinai Medical Center (Los Angeles, CA) were queried for patients with histologically confirmed endometrial cancer who received at least two cycles of pembrolizumab between May 2017 and July 2023 and had available pretreatment surgical specimens or tumor biopsies. Inclusion criteria were kept broad allowing for ICI use as a single agent or in conjunction with other therapies as well as in the upfront or recurrent setting. A total of 142 patients met initial inclusion criteria. Of these, 59 were excluded due to having unavailable or inadequate tissue, yielding 84 patients for analysis. Tumor samples used were pretreatment tissue obtained as part of standard clinical care. If multiple specimens were available, the closest to ICI initiation was selected. Clinical and demographic data, including outcomes through June 2024, were extracted from the EMR (Table 1).

Table 1. Patient characteristics.

Clinical and pathologic characteristics of the 84 endometrial cancer patients included in this study. Percentages may not total to 100 due to rounding. Continuous variables are reported as mean (standard deviation, SD).

	n = 84
Age (years (SD))	66.3 (11.1)
Race
White	58 (69.0%)
Black	12 (14.3%)
Hispanic	8 (9.5%)
Asian	5 (6.0%)
Other	1 (1.2%)
BMI (kg/m2, (SD))	29.9 (7.0)
Stage
I	32 (38.1%)
II	3 (3.6%)
III	13 (15.5%)
IV	36 (42.9%)
Grade
1	24 (30.4%)
2	15 (19.0%)
3	40 (50.6%)
Histology
Serous	19 (22.6%)
Endometrioid	50 (59.5%)
Carcinosarcoma	2 (2.4%)
Clear cell	5 (6.0%)
Other	8 (9.5%)
MMR status
MMR proficient	50 (59.5%)
MMR deficient	33 (39.3%)
Unknown	1 (0.01%)
No. Cycles (SD)	11.5 (13.4)
Concurrent therapy
None	26 (31.0%)
Cytotoxic chemotherapy	23 (27.4%)
Lenvatinib	35 (41.7%)
No. prior treatment lines (SD)	1.7 (1.4)
PFS (months, (SD))	13.8 (17.4)
Time from tissue collection to ICI start (months, (SD))	15.0 (17.7)

Abbreviations: MMR, mismatch repair; No., number; PFS, progression free survival; ICI, immune checkpoint inhibitor

The study was conducted in accordance with consensus ethical principles outlined in the Declaration of Helsinki, the Council for International Organizations of Medical Sciences International Ethical Guidelines, the International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use (ICH) good clinical practice guidelines, and applicable laws. The protocol was approved by the independent ethics committees and institutional review boards at each contributing institution (Vanderbilt University Medical Center IRB #220755, Cedars-Sinai Medical Center IRB #2470).

Laser capture microdissection and reverse-phase protein microarray (LCM-RPPA)

Enriched cancer cell subpopulations were isolated from 6 μM formalin-fixed, paraffin embedded (FFPE) sections using an Arcturus Pixcell IIe LCM system (Arcturus). Samples were lysed in 15–40uL of buffer using the QProteome FFPE Tissue Kit (Qiagen) according to manufacturer’s instructions. Cell lysates were printed in triplicate spots (approximately 10nL per spot) onto nitrocellulose-coated slides (Grace Biolabs) using an Quanterix 2470 Arrayer (Quanterix, Inc.) as described.²⁴ Antibodies used on the arrays were extensively validated before use and are listed in Supplementary Table S1. RPPA immunostaining was performed as previously described.²⁵

Multiplex immunofluorescence (mIF)

Immunofluorescence assays for CK +/− HLA-A/B/C and HLA-DR were developed and optimized at Vanderbilt University Medical Center using tyramide signal amplification (TSA) to enhance antigen sensitivity. Tumor-specific expression of MHC-I and MHC-II was localized by co-staining with pan-cytokeratin (panCK). Four micrometer FFPE tissue sections were cut, deparaffinized, and antigen retrieval performed with citrate buffer (pH 6; Agilent, S2369). Endogenous peroxidases were inhibited with 3% hydrogen peroxidase (Fisher Scientific, BP2633) and proteins blocked (Agilent, K3468). Sections were first incubated with the primary antibody against (panCK DAKO, Z0622 at 1:800) overnight at 4°C, followed by an HRP-conjugated secondary antibody and TSA reagent (Tyramide Superboost Kit; catalog no. B40912, Invitrogen) according to manufacturer’s recommendations. After washing, antigen retrieval, and protein blockade, incubation steps were repeated for the second primary antibodies (HLA-ABC EMR8–5 Abcam at 1:1000 and HLA-DR TAL1B5 Santa Cruz at 1:1000) as described above, followed by a fluorophore conjugated secondary antibody. Counterstaining with DAPI was performed for nuclei identification. Human tonsil tissue and placenta were used as a positive and negative control tissues, respectively, for assay optimization and on every run as an internal batch control.

Image analysis and quantification

Whole-slide images were digitally acquired using an AxioScan Z1 slide scanner (Carl Zeiss) at 20x. Automated quantitative scoring was performed by a pathologist blinded to sample characteristics using QuPath software v0.2.0 (RRID: SCR_018257).²⁶ Cell segmentation was performed with StarDist extension.²⁷ Object classifiers were trained on annotated training regions from control tissue and tumor samples to outline tumor and nontumoral compartments. Tumor cells were defined by panCK expression and subcellular characteristics. For cases with patchy or null CK expression, tumor areas were manually annotated. Out-of-focus areas, tissue folds, necrosis, and normal tissue were excluded from the analysis. The high and low intensity areas from control tissue and tumor samples were combined to create a bimodal distribution histogram of intensity measurements to set the threshold to define a positive cell for HLA-A/B/C and HLA-DR. Each tissue sample was visually assessed for correct performance of the quantification algorithm. Results were compared to standard single-color chromogenic immunohistochemistry (IHC).

Tissue culture and JAK1 transduction

Human endometrial cancer cell lines KLE (CRL-1622^™, RRID: CVCL_1329), RL95–2 (CRL-1671^™, RRID: CVCL_0505), and AN3 CA (HTB-111^™, RRID: CVCL_0028)) were obtained from the American Type Cancer Collection (ATCC). KLE was selected as a microsatellite stable JAK1 wild-type (WT) cell line and RL95–2 and AN3 CA were selected as MSI-H JAK1 PV cell lines. Cell lines were tested quarterly for mycoplasma contamination. All media components were purchased from commercial vendors and prepared and stored under sterile conditions. Cell lines were utilized within the early passage (<30 passages from acquisition from ATCC). KLE cells were grown in DMEM:F-12 (ATCC) supplemented with 10% fetal bovine serum (FBS; Life Technologies). RL95–2 cells were grown in DMEM:F-12 (ATCC) supplemented with 10% FBS and 0.005 mg/ml insulin. AN3 CA cells were maintained in EMEM (ATCC) supplemented with 10% FBS (FBS; Life Technologies).

Isogenic positive control cell lines were generated by expressing WT JAK1 in RL95–2 and AN3 CA cells. Gateway recombinant cloning technology (pShuttle^™) was used to construct a plasmid containing a functional JAK1 gene (GeneCopoeia, GC-X1026-CF), pLX-EF1α-JAK1, also containing a blasticidin resistance gene. RL95–2 and AN3 CA cell lines were transduced with pLX-EF1α-JAK1 and kept in selection with blasticidin for a week then passaged several times to establish stable cell lines prior to use experimental use. Quantitative PCR was used to confirm reintroduction of wild type JAK1 in parental JAK1 PV cell lines. RNA from each cell line was isolated with the Maxwell16 system using the Simply RNA Cell Cartridge (Promega). We generated cDNA from 1ug of total RNA with the Sensifast cDNA kit (Meridian Bioscience). cDNA was diluted with 80ul NFdH20 for 10ng RNA equivalents per uL. Each reaction was conducted with 50ng RNA equivalents per reaction using the BioRad SsoAdvanced Universal SYBR Green Supermix and the BioRad CFX96 thermocycler.

Flow cytometry

All cell lines were stimulated with IFNγ at 50ug/mL for 48 hours. For experiments involving JAK1/2 inhibition, cells were treated with 1 uM ruxolitinib (a JAK1/2 inhibitor; SelleckChem) or 1 uM NVP-BSK805 (a JAK2 selective inhibitor; Novartis). Cells were then surface-stained for PD-L1, MHC-I, and MHC-II expression, which was quantified by flow cytometry.

For flow cytometry, cells were washed in phosphate-buffered saline (PBS) and harvested with Accutase (EMD Millipore, #SCR005) for 1 min at 37 °C. Dissociated cells were washed once in flow staining buffer (PBS + 1% bovine serum albumin) and incubated with respective flow antibodies at 4 °C for 30 min in the dark. Flow cytometry was performed using the following antibodies: HLA-A,B,C/Alexa Fluor488 (Biolegend, clone W6/32, 1:200), HLA-DR/PE-Cy7 (Biolegend, clone L243, 1:200), and CD274(PD-L1)/APC (Biolegend, clone 29E.2A3, 1:400). DAPI was used as a viability dye to exclude dead cells. Samples were analyzed on an Attune NxT flow cytometer (Life Technologies).

RNA Sequencing and CIBERSORT

RNA was extracted from FFPE tissue sections using the Maxwell^® 16 LEV RNA FFPE Purification Kit (Promega) according to the manufacturer’s protocol. RNA quantity and integrity were assessed using NanoDrop. Bulk RNA sequencing was performed by our institution’s genomics laboratory core. Sequencing was performed at Paired-End 150 bp on the Illumina NovaSeq 6000 targeting an average of 50M reads per sample and a quality control analysis was performed. Reads were aligned to Gencode GRCh38.p13 genome using STAR (v2.7.11a). Gencode v38 gene annotations were provided to STAR to improve the accuracy of mapping. FeatureCounts (v2.0.6) was used to count the number of mapped reads to each gene which were input into the CIBERSORT algorithm (https://cibersort.stanford.edu) using the LM22 signature matrix to estimate the relative proportions of 22 immune cell types as previously described.²⁸

TCGA database analysis

Gene expression data, somatic mutation profiles, RPPA data, and associated clinical and survival data were obtained from the Uterine Corpus Endometrial Carcinoma (UCEC) project from TCGA database using the R package TCGAbiolinks.²⁹ Statistical significance of gene expression signatures were determined by ANOVA and Tukey’s HSD test for multiple comparisons. A p-value of < 0.05 was considered significant for all analyses. All statistical analyses were conducted using R statistical software (version 4.3.0, R Project for Statistical Computing, RRID: SCR_001905). Variant allele frequencies (VAFs) for each JAK1 mutation were compared to the median VAF for each sample determined by maftools to provide an estimate of JAK1 PV clonality. JAK1 PVs with VAFs less than or equal to the sample median were classified as heterozygous (JAK1^Het), while those with VAFs greater than the median were classified as homozygous loss-of-function (JAK1^Hom). In cases where there were multiple subclones within a tumor, the predominant subclone was used to define the median VAF. Using this approach, all tumors classified as JAK1^Het have heterogenous presence of JAK1 PV in the tumor as a whole—while most tumors classified as JAK1^Het are true JAK1 allelic heterozygotes, there is a subset in which it cannot be determined if there is actually a mixed population of JAK1^WT and JAK1^Hom leading to the intermediate VAF.

Statistical analysis

Statistical analyses were performed using R Project for Statistical computing (RRID:SCR_001905) or STATA BE 18.0 (RRID:SCR_012763). For mIF analysis, progression-free survival (PFS) was compared between cohorts with positive versus negative staining using Kaplan-Meier analysis and categorical Cox proportional hazard (PH) regression analysis. For RPPA analysis, p-values were calculated for each marker using a multivariate Cox PH regression analysis of PFS adjusting for MMR. A p-value of < 0.05 was considered significant. Given the exploratory nature of the study and sample size limitations, nominal P values are reported without correction for multiple comparisons. For targets where a numerical difference in survival was seen, samples were categorized into expression level tertiles for Kaplan-Meier survival analyses. For mIF, a cut-off of > 5% positive tumor cells was used to define positive staining for tissue-specific MHC-II (tsMHC-II), based on previous studies in other tumor types.^12,13 A cut-off >90% was used to define positive tissue-specific MHC-I (tsMHC-I). Serial cutoff thresholds for positive MHC-I and MHC-II immunostaining were additionally analyzed.

Data availability:

Data underlying the findings described in this manuscript are available in a computational capsule in Code Ocean, https://codeocean.com/capsule/9296641/tree/v1. Data is also available upon request from the corresponding author.

Results

Proteomic profiling of ICI-treated endometrial cancer samples identifies markers predictive of survival

Tumor samples were collected from endometrial cancer patients with known clinical outcomes who initiated pembrolizumab therapy between May 2017 and July 2023 at two institutions. Of the 84 patients included, 50 (59.5%) had pMMR tumors and 33 (39.3%) had dMMR tumors (Table 1). The majority of patients included received ICI after disease progression or recurrence. Pembrolizumab was given as part of front-line treatment in 15.5% (n=13), after one prior line of treatment in 41.7% (n=35) and after 2 or more prior treatment lines in 42.9% (n=36) of patients. Patients with dMMR tumors in front-line and recurrent settings demonstrated improved PFS compared to those with pMMR tumors (HR = 0.41 [95% CI 0.20 – 0.86], p= 0.020) (Supplemental Figure S1). Consequently, MMR status was used as a covariate when analyzing the cohort in aggregate. Most tumors were of endometrioid histology (n = 50, 59.5%), followed by serous histology (n = 19, 22.6%). The disease stage and treatment regimens of the patients included were heterogenous: 26 (31.0%) received pembrolizumab monotherapy, 23 (27.4%) received pembrolizumab with concurrent cytotoxic chemotherapy, and 35 (41.7%) received pembrolizumab in combination with lenvatinib (Table 1). Use of combination therapy correlated strongly with MMR status due to FDA approvals and standard of care practice patterns.

An exploratory analysis of 193 protein and phosphoproteins was performed to identify potential biomarkers and therapeutic targets in tumor epithelia isolated by LCM. Protein levels were quantified by RPPA, and association with PFS were evaluated using Cox PH regression (Figure 1A). Among the proteins analyzed, β2-microglobulin (B2M) – the light chain of the MHC-I complex – was the only protein positively associated with PFS at a nominal level of significance (p = 0.035) (Figure 1B). Analysis of PFS and MHC-I expression by treatment subgroup are show in Supplemental Figure S2. Conversely, phosphorylated cyclin D1 demonstrated the strongest inverse association with PFS (p <0.001), followed by phosphorylated p38 MAPK (p = 0.015) (Figure 1C– 1D). Additionally, patients with high expression of B7-H4, an immune checkpoint ligand in the family of PD-L1 (B7-H1), had numerically shorter survival with a median PFS of 9.7 months whereas the median PFS was not reached in the low B7-H4 expression cohort, but this was not statistically significant (p = 0.054) (Figure 1E).

Figure 1. Proteomic analysis of ICI-treated endometrial cancer identifies prognostic markers.

A. Summary of key findings from high throughput analysis of 193 protein and phosphoproteins targets using laser capture microdissection reverse phase protein array (LCM-RPPA) are shown. Cox proportional hazards regression was used to assess associations with progression-free survival (PFS); p < 0.05 was considered statistically significant. Protein expression levels were categorized into tertiles (“low,” “moderate,” and “high”) and plotted using Kaplan-Meier survival curves. B. MHC-I, as measured by β2-microglobulin, was the only positive predictor of improved PFS in this cohort, p = 0.035. C. Phosphorylated Cyclin D1 was the strongest negative predictor of PFS (p < 0.001), followed by D. phosphorylated p38 MAPK (p=0.015). E. High B7-H4 showed numerically shorter PFS with a median PFS of 9.7 months whereas the median PFS was not reached in the low expression cohort, but this did not reach statistical significance (p = 0.054). F. MHC-II expression was hypothesized to be associated with improved PFS but was not found to be predictive in this cohort (p = 0.77).

MHC-I is predictive of improved PFS in ICI-treated endometrial cancer

The above RPPA analysis included a subset of proteins involved in antigen presentation to evaluate our hypothesized predictive markers, MHC-I and MHC-II. We hypothesized that higher expression of tumor-specific antigen-presenting complexes would be correlated with prolonged survival in endometrial cancer patients treated with ICIs. As stated above, higher levels of MHC-I as measured by B2M were associated with longer PFS (p = 0.035, Figure 1A). However, MHC-II expression (HLA-DR/DP/DQ/DX) was not associated with PFS (p = 0.701, Figure 1E). This suggests that MHC-I-mediated antigen presentation may play a more critical role than MHC-II in driving T cell-mediated responses to ICI in endometrial cancer.

To validate these findings with greater specificity for tsMHC, mIF was performed for both class I and II, co-localized with panCK to identify cancer cells. No difference in MHC-I and MHC-II expression levels were observed across MMR or histologic subtypes (Supplemental Figure S3A – S3B). A cut-off of > 5% positive tumor cells was used to define a positive stain for MHC-II, based on previous studies in other tumor types.^12,13 Using this threshold, 44% of samples were classified as MHC-II positive (Supplemental Figure S3D). Consistent with the RPPA findings, MHC-II was not predictive of improved PFS (HR = 1.05 (95%CI 0.56 – 1.99), p = 0.87, Supplemental Figure S3F).

In this cohort, MHC-I expression was more robust than MHC-II. Due to the left-skewed distribution of tsMHC-I, a cutoff of > 90% was used to define positivity, resulting in 44% of samples being classified as tsMHC-I positive (Supplemental Figure S3C). At this threshold, there a numerically improved PFS in patients with MHC-I-positive tumors (HR 0.54, (95%CI 0.27–1.07), p = 0.078, Supplemental Figure S3E), and while statistical significance was not met by this method of MHC-I assessment, the relationship seen was generally consistent with the findings from the RPPA analysis. We additionally performed RNA sequencing on a subset of 22 patients with additional available tissue and performed CIBERSORT as a method of enumerating infiltrating lymphocytes as previously described.²⁸ Infiltrating immune cell types as quantified by CIBERSORT absolute proportions were compared across RPPA MHC-I expression tertiles (Supplemental Figure S4) and no significant correlations were found.

Tumors harboring pathogenic JAK1 variants exhibit reduced STAT3 phosphorylation and MHC-I expression

Of the 84 patients in the cohort, 42 (50%) had clinical next-generation sequencing (NGS) available, and six had tumors that were found to harbor JAK1 PVs. All six patients were dMMR and did not receive lenvatinib. This subgroup was analyzed in greater detail given the high frequency of JAK1 PVs in endometrial cancer^17,19,22 and the established role of the JAK/STAT pathway in upregulating MHC-I and PD-L1 expression.^30,31 It has been hypothesized that tumors with JAK1 PVs may have impaired response to ICIs. Among the six patients with tumors harboring JAK1 PVs, three patients experienced early progression, with worsening of disease reported on their first interval imaging following treatment initiation. In contrast, two patients achieved durable complete responses at 12 and 73 months, respectively. The final patient lacked sufficient tissue for RPPA analysis but mIF was performed. The tumor tissue of this patient was MHC-I positive but did not have MHC-II staining. This patient progressed nine months after treatment initiation and had an OS of 12 months.

To determine if JAK1 PVs are associated with altered antigen presentation and disrupted JAK/STAT signaling, we compared protein levels between JAK1 PV and JAK1 WT tumors using RPPA and mIF. Tumors with JAK1 PVs demonstrated significantly lower tsMHC-I expression than JAK1 WT tumors (p = 0.020). While tsMHC-II expression was numerically lower in JAK1 PV tumors, this was not statistically significant (p = 0.106) (Figure 2A). Notably, all JAK1 WT tumors exhibited some tsMHC-I expression, whereas JAK1 PV tumors either had absent or heterogenous expression (Figure 2B). We reaffirmed these effects with in vitro assays. Using JAK1 PV cell lines and isogenic JAK1 WT controls (Supplemental Figure S5), we measured differences in MHC-I and PD-L1 expression by flow cytometry following treatment with IFN-γ, demonstrating that JAK1 is essential for upregulation of IFN-γ response elements in endometrial cancer (Supplemental Figure S6A–C). We additionally show baseline MHC-I expression is maintained after inhibition of JAK1 in the WT cell line, KLE, as well as in JAK1 PV cell lines RL95–2 and AN3 CA (Supplemental Figure S6D).

Figure 2. Comparison of JAK1 WT and JAK1 PV tumors by LCM-RPPA and mIF.

A. Tumors with JAK1 PV (n=6) exhibited lower tsMHC-I expression compared to JAK1 WT tumors (n = 38) (p = 0.020, Mann-Whitney t-test), and numerically lower tsMHC-II expression (p = 0.106). B. Representative mIF images stained with anti-HLA-ABC (EMR8–5, Abcam, 1:1000) for MHC-I, panCK (Z0622, DAKO, 1:800) to identify tumor cells, and DAPI counterstaining for nuclei identification. The top row shows diffuse MHC-I staining in a JAK1 WT tumor; the middle row shows heterogenous MHC-I staining in a JAK1 PV tumor; and the bottom row shows absent MHC-I staining in a JAK1 PV tumor. C. LCM-RPPA analysis of JAK/STAT signaling components showed decreased STAT3 phosphorylation in JAK1 PV tumors compared to JAK1 WT tumors (p = 0.004, Mann-Whitney t-test). D. STAT1 and STAT3 phosphorylation levels are shown for the five patients with JAK1 PVs with available RPPA data. Mutation status and PFS denoted below the graph. E. Expression of NCAM1 (CD56), a marker highly expressed on NK cells, was stratified into tertiles and log-rank test was performed between NCAM1-high and NCAM1-low populations, showing a prognostic association. F. The same analysis was performed with expression of the NK cell inhibitory ligand NKG2A.

Activation levels of other proteins in the JAK/STAT signaling pathway were assessed by measuring protein phosphorylation using RPPA, including STAT1, a key mediator of anti-tumor immune upregulation, and STAT3, a canonically pro-tumor signaling protein.^32,33 While there was no significant difference in STAT1 phosphorylation, STAT3 phosphorylation was significantly decreased in JAK1 PV tumors compared to JAK1 WT tumors (Figure 2C). Individual analysis of the five patients with JAK1 PV tumors with available RPPA data revealed that the two exceptional responders exhibited minimal STAT3 phosphorylation, whereas two of the three rapid progressors had elevated STAT3 phosphorylation (Figure 2D).

We hypothesized that the presence of ICI exceptional responders with JAK1 PV tumors in our cohort, as well as in prior reports,^22,23 may reflect the contribution of MHC-I independent effector cells, such as natural killer (NK) cells. The decreased STAT3 phosphorylation observed in JAK1 PV tumors supports this hypothesis, as STAT3 is known to promote a pro-tumorigenic immunosuppressive microenvironment, partly through transcriptional repression of activating ligands for NK cells.^34,35 This finding in combination with lower MHC-I expression among JAK1 PV tumors led us to investigate if there was an association between tumor-infiltrating NK cells and survival in our cohort. Expression of NCAM1 (CD56), a canonical NK cell marker, was significantly associated with improved survival among patients with dMMR/MSI tumors (Figure 2E). Conversely, cases with high expression of the NK cell inhibitory ligand NKG2A had comparatively worse survival than those with low expression, but this was not statistically significant (Figure 2F). There was no clear association between either NCAM1 (CD56) or NKG2A and survival among patients with pMMR tumors. These findings suggest that in dMMR/MSI endometrial tumors – including all JAK1 PV tumors in this cohort – tumor-infiltrating NK cells may have an important role in anti-tumor immunity.

MSI-associated JAK1 alterations result in JAK1 variants with distinct clinical phenotypes

Given the paradoxical observation that tumors with JAK1 PVs lack inducible MHC-I expression yet include exceptional responders to ICI – both in our cohort and prior publications^22,23 – we sought to further investigate how JAK1 PV variants relate to anti-tumor immunity and clinical outcomes. To do so, we analyzed data from 509 patients in the UCEC cohort of TCGA PanCancer Atlas with available mutation data.⁴

Endometrial cancer exhibits the highest rate of JAK1 alterations among all tumor types in TCGA, with a prevalence of 14%.^19,22 Given our prior findings that JAK1 function is required for IFN-γ-mediated upregulation of MHC-I and that multiple JAK1 mutations were observed across tumors from different patients, we hypothesized that not all JAK1 PVs may be homozygous or clonal. If this were accurate, a subset of patients may retain JAK1 function either by retaining heterozygosity or by persistent polyclonal populations in the tumor microenvironment. To determine JAK1 PV zygosity, we compared the VAF for each JAK1 PV compared to overall median VAF for that sample to classify each patient tumor sample into one of four categories by their JAK1 mutation status: JAK1 wild-type (JAK1^WT), JAK1 homozygous loss-of-function (JAK1^Hom), JAK1 heterozygous loss-of-function (JAK1^Het), or JAK1 variants of uncertain significance (JAK1^VUS) (Supplemental Figure S7A–B). To evaluate whether heterozygous or homozygous mutations in JAK1 drive broad transcriptional changes, we performed principal component analysis (PCA) comparing JAK1^WT, JAK1^Hom, and JAK1^Het tumors (Supplemental Figure S7C). PCA yielded limited findings, other than groupings correlating with previously established genomic subtypes (Supplemental Figure S7D–E). Notably, JAK1 PVs were enriched in MSI-H and POLE tumors, with 29.0% of MSI-H tumors and 24.5% of POLE tumors harboring JAK1 mutations (Figure 3A).

Figure 3. Patients with JAK1^Het tumors have improved survival and intermediate MHC-I expression when compared to JAK1^Hom and JAK1^WT tumors.

A. Frequency of JAK1 mutation status is presented by endometrial cancer genomic subtype: copy number high (CN-high), copy number low (CN-low), microsatellite instability (MSI), polymerase epsilon mutated (POLE). JAK1 PVs, both JAK1^Het and JAK1^Hom, were almost exclusively found in the MSI and POLE subtypes. B. Kaplan-Meier survival analysis among MSI and POLE tumors shows that patients with JAK1^Het tumors showed improved survival compared to patients with JAK1^Hom tumors by log-rank test (p = 0.026). C. The survival advantage held when the analysis was restricted to the MSI subset of tumors, but in a smaller sample size, was no longer statistically significant by log- rank test. D. Integrated expression scores for MHC-I, MHC-II, IFN-γ, and immune checkpoints were calculated and compared across JAK1 mutation groups. Statistical significance of gene expression signatures was determined by ANOVA and Tukey’s HSD test for multiple comparisons. E. Heatmap showing RNA expression levels of the individual components of the integrated expression scores for MHC-I, MHC-II, and immune checkpoints.

Tumor mutational burden (TMB) for the majority of MSI-H and POLE tumors was above the clinical cut-off for high TMB (>10) (Supplemental Figure S8). Given the distinct clinical phenotypes and ICI responses associated with the different molecular subtypes of endometrial cancer,⁴ further analyses were limited to the 196 patients with MSI-H or POLE-mutated tumors. Each of these tumor samples harbored between one to five distinct JAK1 mutations, with several specimens having multiple pathogenic mutations including those occurring in hotspot regions such as K860fs and P430fs.¹⁹

JAK1^Het tumors are associated with improved survival and intermediate MHC-I expression

In the combined MSI-H and POLE cohorts, patients with JAK1^Het tumors demonstrated improved overall survival (OS) compared to patients with JAK1^Hom tumors (p = 0.026), and a numerically improved OS compared to patients with JAK1^WT tumors, but this did not meet statistical significance (p=0.081) (Figure 3B). When analyzing the MSI-H cohort alone, patients with JAK1^Het tumors had numerically improved OS compared to patients with JAK1^Hom tumors but this did not reach statistical significance by log-rank test (p = 0.067) (Figure 3C). This analysis was repeated using a multivariate Cox proportional hazard regression model with TMB as a co-variate. JAK1^Het again demonstrated improved survival with a very large effect size (regression coefficient (β) = −18.08), but a p-value could not be calculated due to lack of events in the JAK1^Het cohort. These findings suggest that partial retention of JAK1 function may enhance antitumor immunity and patient prognosis in endometrial cancer.

To identify components of the anti-tumor immune response that may contribute to these differing prognoses, we generated integrated expression scores for MHC-I, MHC-II, IFN-γ signaling, and immune checkpoints, and compared scores between JAK1 mutation status groups (Figure 3D – 3E). Tumors with JAK1 PVs had decreased expression of MHC-I-related genes. A decrease in MHC-II-related genes and components of the IFN-γ response pathway was observed among JAK1^Hom tumors compared to JAK1^WT tumors. Neither JAK1^Het nor JAK1^Hom tumors had decreased expression of immune checkpoint-related genes when compared to JAK1^WT.

Discussion

Immunotherapy has become a transformative addition to endometrial cancer treatment, leading to significant clinical benefit among select patients.^6,7,10,36,37 Molecular characterization of tumors that respond to ICI offers an opportunity to refine personalized treatment strategies while also deepening our understanding of how treatment response is shaped by the tumor immune microenvironment in endometrial cancer. Identifying better molecular biomarkers for ICI response could ultimately allow for the development of therapeutic strategies to enhance treatment efficacy and yield a better mechanistic understanding of resistance. Here, we present a translational analysis of tumor characteristics and clinical outcomes of a heterogenous group of endometrial cancer patients treated with ICI either as a single agent or in combination with chemotherapy or lenvatinib, reflecting real-world practices and current FDA approvals for ICI use in endometrial cancer.

We demonstrate that MHC-I expression may serve as a predictive biomarker of improved progression free survival on ICI therapy in endometrial cancer, with higher expression associated with improved clinical outcomes. This is supported by strong biological rationale, as MHC-I plays a central role in cytotoxic T cell-mediated tumor recognition, and its downregulation is a well-established mechanism of tumor immune evasion.³¹ Prior research has also shown that loss of MHC-I on IHC staining is prevalent in endometrial cancer, including in tumors with dMMR.³⁸

Currently, dMMR and MSI-H are the only guideline-endorsed biomarkers for predicting response to ICI in endometrial cancer.^6,37 However, meta-analyses demonstrate approximately half of patients with dMMR/MSI-H tumors do not respond to ICI therapy, indicating that these markers alone are insufficient to guarantee response.^39–42 Moreover, a subset of patients with pMMR tumors also derive meaningful benefit from ICIs, particularly in the setting of advanced or recurrent endometrial cancer.^10,11 Incorporating MHC-I into clinical-decision making may enhance patient selection for immunotherapy. Tumors can adapt various mechanisms to evade immune destruction including the downregulation of MHC-I.³¹ While we provide evidence of pretreatment MHC-I expression having predictive value for response to ICI, assessing MHC-I expression at different time points is an important area for future study. This could help in determining the role of ICI in the recurrent setting after patients have already been exposed to ICI as part of upfront treatment with chemo-immunotherapy regimens such as those used in GY018 and RUBY as this is currently not well defined. Furthermore, tumors exhibiting MHC-I loss may ultimately benefit from agents that restore MHC-I expression, such as histone deacetylase.⁴³

³¹In contrast, MHC-II expression was not associated with ICI response in our endometrial cancer cohort.³⁹ This differs from findings in other tumor types, where MHC-II expression has been reported as a predictive biomarker of response to immunotherapy.^12–14 Further research is needed to understand the context-dependent role of MHC-II in modulating ICI responsiveness across different malignancies.⁴⁴

Having established the potential predictive value of MHC-I expression, we looked at upstream regulators of MHC-I that may influence ICI responsiveness. Prior work has shown that MHC-I upregulation or inducibility often depends on functional JAK1 signaling⁴⁵, a relationship we confirmed in endometrial cancer in our in vitro studies. Loss-of-function mutations in JAK1 are therefore a plausible mechanism of immune escape.^17–20,46 Given this, it is surprising that many patients with dMMR/MSI-H tumors harboring JAK1 PVs have exceptional responses to ICIs. Here, we found that this discrepancy may be clarified by evaluating the zygosity and/or clonality of the various JAK1 mutations.

While it is important to acknowledge that not all mutations are driver mutations, especially in hypermutated and ultra-mutated tumor subtypes, regardless of the mechanism of JAK1 loss, disruption of the JAK/STAT signaling pathway does have substantial impacts on tumor biology. Prior studies in melanoma, lung, colorectal and endometrial cancer lend support JAK1 PV as driver mutations of tumor progression through immune escape.^17–20,46 We found JAK1^Het patients demonstrated exceptional survival. This held true when looking only within the MSI-H cohort and when accounting for tumor mutational burden in a multivariate analysis. The down regulation of MHC-I in this population may increase the importance of alternative effector immune cells aside from CD8+T cells on anti-tumor response.

Recent studies have identified a correlation between improved outcomes and higher relative natural killer (NK) cell abundance among MSI-H tumors, both in mouse models and in analyses of human TCGA data.^22,47,48 In our cohort of ICI-treated dMMR/MSI-H human endometrial cancer tumors, we similarly observed a positive correlation between tumor-infiltrating NK cell abundance and PFS. One proposed mechanism is that tumors that have developed resistance to cytotoxic T cell-mediated killing – through enrichment of MHC-I-deficient clones – may become more sensitive to NK cell effector clearance.⁴⁹ Supporting this, Dubot et. al. used a CRISPR-engineered JAK1-mutated tumor mouse model to show that immunity in JAK1 deficient tumors with decreased MHC-I expression was dependent on NK cell activity.⁴⁸ To further investigate this relationship in endometrial cancer, we analyzed the TGCA database to categorize JAK1 PV tumors as JAK1^Het or JAK1^Hom. We found that patients with JAK1^Het tumors had the best survival outcomes – even surpassing patients with JAK1 WT tumors. JAK1^Het tumors also retained intermediate MHC-I expression. This aligns with prior findings from our lab showing that MHC-I heterogeneity promotes increased NK cell infiltration into the tumor microenvironment.⁵⁰ Additionally, our in vitro data demonstrate that baseline MHC-I expression is not solely dependent on JAK1. Taken together, we hypothesize that NK cells play a key role in the exceptional survival seen in patients with JAK1 PV endometrial cancers, but a certain baseline level of MHC-I expression is required for ICI response. NK cells are likely prognostically important in dMMR/MSI tumors, particularly in those with JAK1 PVs, given the role of CD8+ T-cells is likely diminished with decreased MHC-I expression. One limitation of the study presented here is that we did not perform a detailed spatial analysis of the tumor microenvironment (TME). Furthermore, anatomic location can greatly impact the make-up of the TME with differential infiltrates expected in a primary endometrial tumor compared to various distant metastatic sites, and this study contained a heterogenous mix of primary and metastatic tumor samples. Future work should focus on examining JAK1 heterozygosity in an expanded ICI-treated endometrial cancer cohort, investigating therapeutic strategies to leverage NK cell activity to overcome ICI resistance, and refining patient stratification approaches to better identify those likely to derive benefit and exclude those who will likely not benefit or who will achieve durable responses without ICIs.

Additional insights from our exploratory RPPA analyses suggest several promising directions for future research. The protein showing the strongest inverse association with survival was phosphorylated cyclin D1 (p <0.001), followed by phosphorylated p38 MAPK (p = 0.015). Phosphorylation of cyclin D1 activates cyclin dependent kinases, CDK4 and CDK6, ultimately driving cell cycle transition from G1 to S-phase, promoting tumor growth. This finding supports increasing interest in the use of CDK4/6 inhibitors in endometrial cancer treatment.^51–53 Tumors with higher cyclin-D1 phosphorylation may be particularly susceptible to CDK4/6 inhibition, and the addition of CDK4/6 inhibitors could potentially help overcome ICI resistance in the subset of patients that have tumors with increased cyclin-D1 phosphorylation. We also found that patients with tumors that had increased B7-H4 levels had worse survival long-term, although this did not reach statistical significance (p = 0.054). B7-H4 is an immune checkpoint ligand that remains poorly characterized but is currently under investigation as a therapeutic target in early phase clinical trials.^54–56 Our findings provide preliminary clinical data to support its relevance in endometrial cancer, warranting further investigation into its role in immune modulation and potential as a targeted treatment in this disease.

In conclusion, our data suggest that MHC-I is a promising biomarker for ICI response in endometrial cancer. Paradoxically, JAK1 signaling– which is required for MHC-I upregulation– is reduced in JAK1^Het tumors yet associated with improved survival. This may be explained by increased NK cell activity, which can target MHC-I-low cells. Further studies are needed to validate these findings and elucidate underlying mechanisms.

Translational Relevance:

Immune checkpoint inhibitors (ICIs) are approved in endometrial cancer but lack reliable biomarkers for patient selection. JAK1 loss-of-function pathogenic variants (PV), which abrogate interferon-induced antigen presentation, are common in endometrial cancer and are classically associated with immune evasion and ICI resistance – although this has recently been questioned. Using high-throughput proteomic analyses of 193 protein and phosphoprotein targets, we identified candidate biomarkers for endometrial cancer ICI response and evaluated their utility in the context of JAK1 PVs. Tumor MHC-I expression was predictive of response, and notably, JAK1 PVs were present in some exceptional responders despite impaired MHC-I upregulation. We found that biallelic JAK1 mutations impair antigen presentation and are associated with poorer clinical outcomes, whereas patients with subclonal or heterozygous mutations had superior survival. Additionally, we identified proteins negatively associated with response, highlighting targets for future clinical investigation. These data provide insights that may help refine ICI treatment strategies and improve outcomes.