Elevated T Cell Immunoreceptor with Ig and ITIM Domains (TIGIT) Expression and Immune Cell Dysfunction Characterize Complex Proteins Associated With SET1 (COMPASS)—Like Complex Gene—Mutated Pancreatic Ductal Adenocarcinoma (PDAC)

Abstract

Pancreatic ductal adenocarcinoma (PDAC) is highly resistant to immune therapies. Limited biomarkers, such as mismatch repair proteins, have been used to identify those who may respond to immunotherapy. We identified a subset of aggressive PDACs (~25%) carrying mutations in the complex of proteins associated with SET1-like complex genes (CLCGs), which can be used as new biomarkers for targeted immunotherapy. In this study, we compared the immune microenvironment of PDACs harboring CLCG mutations with matched wild-type PDACs using multiplex fluorescent immunohistochemistry and computational imaging techniques. We observed that CLCG-mutant PDACs were infiltrated with fewer CD4⁺ T cells and antigen-presenting cells (APCs) but elevated immune checkpoint T cell immunoreceptor with Ig and ITIM domains (TIGIT) expression on CD4⁺ T cells and APCs. There was no difference in the expressions of other immune checkpoints, such as programmed death-1 receptor ligand and T-cell immunoglobulin and mucin domain—containing protein 3. More CD4⁺ T cells near epithelial cells (tumor cells) and APCs expressed TIGIT in CLCG-mutant PDACs. Additionally, CLCG-mutant PDACs displayed a malfunctional immune cell crosstalk. Single-cell RNA-sequencing data confirmed the elevated TIGIT expression on CD4⁺ T cells and increased exhausted CD4⁺ T cells in CLCG-low PDACs. These findings uncovered the unique underlying mechanisms of immune suppression in CLCG-deficient PDACs and identified CLCG as potential biomarkers to identify those who may benefit from TIGITtargeting immunotherapies.

Introduction

Pancreatic ductal adenocarcinoma (PDAC) is the third leading cause of cancer death in the United States.¹ Despite substantial treatment efforts, most patients develop advanced and surgically unresectable diseases with a 5-year survival rate of 13%.¹ Recent improvements in large scale sequencing techniques have significantly enhanced our understanding of the complexity and molecular heterogeneity of PDAC. These advances have facilitated the identification of available biomarkers and targeted therapeutic approaches.² Recent progress on the well-known mutated genes, including KRAS and TP53, has revealed that gene-mediated epigenetic reprogramming directs the initiation of pancreatic cancer tumorigenesis.^3,4 However, the role of the regulators associated with epigenetic remodeling was not fully understood. We have focused on the genes involved in the complex of proteins associated with SET1 (COMPASS)—like complex (CLC), a protein complex that plays an essential role in histone methylation.⁵ Alterations in the components of COMPASS-like complex, including KMT2A, KMT2B, KMT2C, KMT2D, and KDM6A, were identified in a significant portion of PDACs and were correlated with aggressive behavior.^6–8 We previously studied a cohort of patients with PDAC harboring alterations in the CLC genes (CLCGs) and characterized their clinicopathologic and molecular features.⁹ We have demonstrated that CLCG-mutant PDACs represent a subset of more aggressive PDACs with poor or squamous differentiation and are associated with poor patient survival.⁹ Mechanistically, we found that loss of KDM6A or KMT2D in tumor cells promotes epithelial-mesenchymal transition through the noncanonical activin A pathway.^6,8 However, whether and how CLCG alterations affect the tumor immune microenvironment is unknown.

Tumor cell evasion from immune surveillance is accompanied by reprogramming of the tumor immune microenvironment. PDAC is well-known for being unresponsive to currently available immunotherapy, mostly due to an immunosuppressive tumor microenvironment (TME).² The presence of tumor-associated macrophages (TAMs), myeloid-derived suppressor cells, and regulatory T cells (Tregs) contributes to an environment that supports tumor growth and suppresses effective antitumor immune response.^10,11 Subsets of TAMs can have M2-like features associated with anti-inflammatory or protumor effects by maintaining a functional Treg population¹² or secreting immunosuppressive cytokines that inhibit CD8⁺ T cells.¹³ Tregs expressing FOXP3 have been shown to suppress the immune response of effector T cells, including CD4⁺ T helper (Th) cells and CD8⁺ cytotoxic T cells.^14,15 Increased expression of immune checkpoint markers, such as T cell immunoreceptor with Ig and ITIM domains (TIGIT) and T-cell immunoglobulin and mucin domain—containing protein 3 (TIM3), on T cells is another immune evasion mechanism in exhausted T cells.^16–18 Additionally, other immune checkpoint mediators, such as the programmed death-1 receptor ligand (PD-L1), are expressed on tumor cells and TAMs.^19,20 A recent study examining the association between KMT2 family mutation and immune checkpoint inhibitor therapy showed that patients with KMT2 family gene mutations benefited more from immune checkpoint inhibitor therapy.²¹ However, the underlying mechanisms of immunomodulation by CLCG alterations in tumor cells are unknown.

In this study, we compared the immune microenvironment of PDACs harboring CLCG mutations with CLCG-wild-type (WT) PDACs using multiplex fluorescent immunohistochemistry (mfIHC), which allowed us to simultaneously label up to 7 different color markers. By analyzing the cellular composition, immune checkpoint expression, and spatial interactions between cells, we unveiled the critical role of TIGIT in shaping the immunosuppressive microenvironment in CLCG-mutant PDACs. We further validated the elevated TIGIT expression in T cells using human PDAC single-cell RNA-sequencing (scRNA-seq) data sets.

Materials and Methods

Sex as a Biological Variable

Our study examined specimens from both male and female patients, and similar findings are reported for both sexes.

Human Specimens

This study included 16 patients diagnosed with PDAC who received oncologic care at Michigan Medicine (Ann Arbor, MI). All patient tumor samples underwent comprehensive genomic profiling as previously described⁹ to identify tumors with (n = 8) or without (n = 8) alterations in genes encoding elements of the COMPASS-like complex, including KMT2A, KMT2B, KMT2C, KMT2D, and KDM6A. The groups were matched for age, sex, and presence of concomitant driver mutations (KRAS, TP53, SMAD4, and CDKN2A). A detailed comparison of clinicopathologic and molecular features between CLCG-mutant and WT PDACs is provided in Supplementary Table S1. Hematoxylin and eosin slides were reviewed, and specific tumor-rich blocks were selected by trained pathologists (E.D. and J.S.) for mfIHC analysis.

Multiplex Fluorescent Immunohistochemistry Staining

Tissue sections (5 μm) were cut from the formalin-fixed and paraffin-embedded tissue blocks onto charged slides, and the slides were processed as previously reported.²² Briefly, after optimization of the staining parameters for each individual primary antibody, the slides from all 16 cases were sequentially stained with 2 panels of antibodies (Table) using a Ventana Discovery Ultra stainer (Ventana Medical Systems). After the primary antibody incubation, the slide was washed with Tris-buffered saline with Tween-20 buffer, and Ventana OmniMap anti-rabbit horseradish peroxidase was applied, followed by the application of Opal Tyramide Signal Amplification, which creates a covalent bond between the fluorophore and the tissue at the horseradish peroxidase site. Subsequent heat-induced epitope retrieval using CC2 buffer removed the primary antibody/secondary antibody complex, and the next antibody was applied. A panel of 6 fluorochromes was sequentially applied to the tissue. The slides were then mounted with ProLong Gold containing DAPI and scanned using an Akoya Polaris IF scanner at 20× magnification.

Table.

Antibodies and corresponding fluorescent color used for mfIHC panels

Color	Panel 1 antibodies	Panel 2 antibodies
Cyan (Opal 480)	Pan cytokeratin/PanCK	Pan cytokeratin/PanCK
White (Opal 780)	CD3	CD3
Red (Opal 690)	CD8	CD8
Orange (Opal 620)	CD163	CD163
Yellow (Opal 570)	TIM3	FoxP3
Green (Opal 520)	TIGIT	PD-L1

mflIHC, multiplex fluorescent immunohistochemistry.

Multiplex Fluorescent Immunohistochemistry Image Analysis

Multiple tumor cell—rich areas (panel 1, WT n = 59 and CLCG mutant n = 39; panel 2, WT n = 57 and CLCG mutant n = 41) were selected from each slide, and all the images were analyzed using inForm Tissue Analysis Software (Akoya Biosciences). For each slide, an average of 5 to 8 regions was selected based on tissue size, ensuring comprehensive coverage of tumor cell—rich zones. Each region measured 1845 μm × 1385 μm and encompassed 10,000 to 13,000 cells. A library using single antigen staining for each fluorophore was created for unmixing and validation of multiplex fluorescent composite staining to confirm that there was no spectral overlapping. With the inForm training process, the tissue was identified and segmented into stroma and epithelial compartments, and cells were segmented into nucleus, cytoplasm, and membrane. DAPI counterstain was used to determine the size and shape of each nucleus, and x and y coordinates were assigned to each cell in each section. The shape and thickness of cytoplasm and its distance from the nucleus were calculated automatically by the system using the fluorescent signal of CD3 and CD8. The cell phenotypes were assigned using the trainable inForm Tissue Analysis Software, as previously described,²² after manual selection based on single staining criteria. Briefly, cell phenotypes were initially identified and manually assigned based on antibody fluorescence signals. For each marker, 30 to 100 representative cells were selected and used as reference input. This training set enables the inForm software to automatically classify and assign phenotypes to the remaining cells in the image. The cells were assigned as epithelial cell (EC)/tumor cell (pancytokeratin⁺), antigen-presenting cell (APC; CD163⁺), T cell (CD3⁺), CD8⁺ T cell (CD3⁺CD8⁺), CD4⁺ T cell (CD3⁺CD8^–), Treg (CD3⁺CD8^–FoxP3⁺), and other cells (pancytokeratin⁻CD163⁻CD3⁻). Fluorescent intensity scores of immune checkpoints PD-L1, TIGIT, and TIM3 were also determined, which allowed for the formulation of complex phenotypes in combination with the original cell types (EC, APC, and T cells). The cellular interaction was manifested by calculating mean distance and cell engagement using the inForm Tissue Analysis Software as previously described by our group.²² Briefly, each cell was assigned to a spatial coordinate within the tissue image, and the software computed the average distance between each cell. Cell engagement was calculated as the percentage of engaged cells relative to total cells within an engagement zone. The engagement zones were established by defining a radius of 15 μm around each T cell and 40 μm around each APC and EC. The central cell was classified as “engaged” with other immune cells or ECs present within the circle. The proportion of engaged cells relative to total cells present in each region was calculated. This analysis is performed at the single-cell level using automated image analysis software, enabling highly precise and reproducible detection of cell engagements.

G-Cross Spatial Analysis

We performed an analysis to quantify the intermixing of 2 specific cell types, that is, to find cell-cell interactions across multiple samples using the G-cross function,²³ a second-order spatial point pattern statistic that quantifies the spatial relationship between 2 cell types. The G-cross function is a spatial point pattern statistic that calculates the cumulative distribution function of cells’ nearest neighbor distances. Specifically, it quantifies the probability that a reference cell type, i, will find a cell type, j, within a given radius, r. We computed both the empirical G-cross function (km), which represents the observed cumulative distribution of intercellular distances, and the theoretical G-cross function (theo), which models the expected distribution under the assumption of complete spatial randomness. For each cell-type pair, areas under the curve (AUCs) were calculated at predefined distances of 60 microns using the trapezoidal integration method. Then, we computed normalized G-cross scores by dividing empirical AUCs by their theoretical counterparts at each radius. These normalized scores were used as a quantitative measure of spatial association between cell types. The resulting AUC metrics were merged with sample-level metadata, including CLCG mutation status, and used for statistical testing. We applied the Wilcoxon rank-sum test to assess differences in spatial association between mutant and WT groups at the 60-micron radius. P values were adjusted for multiple comparisons using the false discovery rate, and phenotype pairs showing significant differences were selected for further interpretation.

Single-Cell RNA-Sequencing Analysis

scRNA-seq data sets were downloaded from publicly available sources (GSE155698, GSE154778, and GSE156405). Data sets for 30 PDAC patient samples were collected and processed for analysis using Seurat, version 3.0 (Satija Lab, New York Genome Center).²⁴ Briefly, data were filtered to include all cells with >250 genes and with mitochondrial percentage <25%. Filtered data were normalized using the NormalizeData function and scaled using ScaleData. Principal component analysis was performed using the RunPCA function, followed by batch correction using harmony.²⁵ SoupX was used for ambient RNA correction to remove background contamination, followed by filtering out low-quality cells based on standard metrics: number of detected genes (nFeature_RNA), total unique molecular identifier counts (nCount_RNA), and percentage of mitochondrial reads. Doublets were removed using Doublet-Finder. After quality control, key quality control metrics (violin plots of nFeature, nCount, and percent.mt) were inspected across samples and data sets to identify and exclude any sample-level outliers due to technical issues. No data sets or individual samples were identified as systematic outliers requiring complete exclusion. In addition, sample-level outliers and technical batch effects within each data set were examined prior to integration by visualizing data set—specific embeddings and cell-type distributions. Clusters were identified via the FindNeighbors and FindClusters functions. FindMarkers was performed to obtain signature genes for each cluster, and we identified broad cell types using published markers. For each broad cell type (eg, CD4⁺ T cell and CD8⁺ T cell), we again performed normalization, scaling, batch correction, and clustering to identify fine markers. The k-nearest neighbor batch effect test was used to quantitatively assess batch mixing.²⁶ We also performed cell-type—restricted renormalization and batch correction, a strategy commonly used in fine-grained single-cell analyses to improve resolution within transcriptionally similar populations.²⁷ Differential gene expression analysis was performed for comparisons between clusters, and the Wilcoxon rank-sum test was performed to detect significance. Genes with a false discovery rate—adjusted P value <.05 were considered differentially expressed. The samples were then stratified as previously reported.²⁸ Briefly, the average expression of KDM6A and KMT2D in the annotated EC/tumor cell population was quantified, and the samples were divided into the CLCG-high group (n = 8) and the CLCG-low group (n = 10).

Statistics

Statistical analyses were performed using the R program (version 4.3.2; R Foundation for Statistical Computing) and GraphPad Prism version 10.4.1 (GraphPad Software). For scRNA-seq data analysis, RStudio, version 2023.12.0, and R package Seurat, version 3.0, were used. For comparisons between cell phenotypes, distance, and cell engagement, the 2-tailed Student’s t test was used. For data not normally distributed, the nonparametric Wilcoxon rank-sum test was used. Pearson correlation coefficients were used to measure R and R². P < .05 was considered statistically significant.

Results

Complex of Proteins Associated with SET1-Like Complex Gene—Mutant Pancreatic Ductal Adenocarcinomas Have Decreased CD4⁺ T Cells and Antigen-Presenting Cells

CLCG-mutant PDACs accounted for 24.3% of the PDAC cases based on The Cancer Genome Atlas PanCancer Atlas data set (Supplementary Fig. S1, n = 175). We previously reported that CLCG-mutant PDACs were more often poorly differentiated or undifferentiated with squamous/basal type differentiation and TP53 mutation.⁹ Patients with CLCG-mutant PDACs also tended to have decreased patient survival compared with CLCG-WT PDACs.⁹ Immunosuppressive TME is well known to contribute to the poor prognosis in PDAC. However, whether and how CLCG mutations in the tumor cells affect the immune TME in PDAC and if there is a unique therapeutic target in these tumors is not known. We performed mfIHC to compare the immune TME of CLCG-mutant PDACs with CLCG-WT PDACs (Fig. 1A). A total of 98 randomly selected tumor cell—enriched areas from 16 clinically and genetically matched patients’ PDAC surgical samples (8 CLCG mutant and 8 CLCG WT) were annotated and analyzed. Overall, the number of APCs (CD163⁺), CD4⁺ T cells (CD3⁺/CD8⁻), and Tregs (CD3⁺/CD8⁻/FoxP3⁺) was significantly decreased in CLCG-mutant PDACs compared with CLCG-WT PDACs (Fig. 1A, B). The fractions of APCs and CD4⁺ T cells were also reduced in CLCG-mutant PDACs compared with CLCG-WT PDACs; however, these differences did not reach statistical significance (Supplementary Fig. S2A–E). Because Tregs can inhibit tumor immune responses by inactivating with Th cells and CD8⁺ T cells,¹⁵ we further investigated their distribution in the TME. Although the proportion of Tregs within the CD4⁺ T-cell population did not differ significantly between WT and CLCG-mutant PDACs (Fig. 1C), a significant spatial proximity was observed between Tregs and Th cells in CLCG-mutant PDACs compared with WT PDACs (Fig. 1D, E). No significant differences were observed in the spatial proximity between Tregs and either CD8⁺ T cells or APCs when comparing WT and CLCG-mutant PDACs (Supplementary Fig. S2F, G). Furthermore, in CLCG-mutant PDACs, the fraction of EC (tumor) cells was positively correlated with the fraction of Tregs, a relationship that was not observed in CLCG-WT PDACs (Fig. 1F). These findings imply that the TME in CLCG-mutant PDACs is more immunosuppressive compared with CLCG-WT PDACs.

Figure 1.

CLCG-mutant pancreatic ductal adenocarcinomas (PDACs) have decreased CD4⁺ T cells and APCs. (A) Representative mfIHC images of wild-type (WT) and CLCG-mutant (CLCG) PDACs. Cyan = epithelial cell (EC)/tumor cell; green= PD-L1; yellow=Treg; orange=APC; pink=CD8⁺ T cell; white = CD4⁺ T cell. Orange arrow, APC; white arrow, CD4⁺ T cell. Scale bar=30μm. (B) Quantification of average cell count of EC, APC, total T cell, CD4⁺ and CD8⁺ T cell, and Treg in WT (n = 59) and CLCG-mutant (n = 39) PDACs. (C) Quantification of average ratio of Treg/Total CD4⁺ T cells in WT (n = 57) vs. CLCG-mutant PDACs (n = 41). (D) Mean distance from Treg to helper T cell in WT (n = 57) vs. CLCG-mutant PDACs (n = 41). (E) Representative mfIHC images showing the spatial proximity between Tregs and Th cells in CLCG-mutant PDACs compared with WT. Yellow arrows, Treg; white arrows, helper T cell. Scale bar = 30 μm. (F) Correlation analysis between ratio of Tregs/total CD4⁺ T cells and ratio of ECs/total cells in WT (n = 57) and CLCG-mutant (n = 41) PDACs. *P < .05. mflIHC, multiplex fluorescent immunohistochemistry; ns, no significance.

Complex of Proteins Associated with SET1-Like Complex Gene—Mutant Pancreatic Ductal Adenocarcinomas Have Elevated TIGIT Expression on CD4⁺ T Cells and Antigen-Presenting Cells

Cancer cells use enhanced immune checkpoints to evade immune response.²⁹ To compare the immune checkpoint protein expressions between CLCG-mutant and WT PDACs, we first investigated PD-L1 expression levels on ECs and APCs. No difference was observed in PD-L1 expression on ECs/tumor cells and APCs (Fig. 2A, B). TIGIT is an essential immune checkpoint that inhibits immune cell responses to cancers.^16–18 Although TIGIT is mainly expressed on exhausted T cells and natural killer cells, a subset of APCs/macrophages with anti-inflammatory and more tumor-promoting properties can also express TIGIT, which is often associated with functional inhibition.^30,31 We compared TIGIT expressions on T cells and APCs between WT and CLCG-mutant PDACs. Interestingly, CLCG-mutant PDACs had increased TIGIT expression on CD4⁺ T cells and APCs, but not CD8⁺ T cells, compared with WT PDACs (Fig. 2C–H). Interestingly, the expression of TIM3, another immune checkpoint protein, was similar on T cells and APCs between WT and CLCG-mutant PDACs (Supplementary Fig. S3), suggesting that the observation of TIGIT alterations is specific. To assess variability within and between each patient sample, we compared the ratio of TIGIT⁺ CD4⁺ T cells with total CD4⁺ T cells across different areas from the same sample and among samples within the same study group. Our analysis revealed substantial differences in this ratio, both within individual samples and across the cohort (data not shown), highlighting considerable internal heterogeneity and diverse patient representation. These findings are supportive of exhausted CD4⁺ T cells and APC phenotype in CLCG-mutant PDACs and the increased immune checkpoint TIGIT expression as an underlying mechanism.

Figure 2.

CLCG-mutant pancreatic ductal adenocarcinomas (PDACs) have elevated TIGIT expression on CD4⁺ T cells and antigen-presenting cells (APCs). Quantification of average ratio of (A) PD-L1⁺ epithelial cells (ECs) of total ECs and (B) PD-L1⁺ APCs of total APCs in wild-type (WT) (n = 57) and CLCG-mutant (n = 41) PDACs. Quantification of average ratio of (C) TIGIT⁺ T cells to total T cells; (D) TIGIT⁺ CD4⁺ T cells to total CD4⁺ T cells; and (E) TIGIT⁺ CD8⁺ T cells to total CD8⁺ T cells in WT (n = 59) and CLCG-mutant (n = 39) PDACs. (F) Quantification of average ratio of TIGIT⁺ APCs to total APCs in WT (n = 59) and CLCG-mutant PDACs (n = 39). (G) Representative multiplex fluorescent immunohistochemistry images of TIGIT⁺ CD4⁺ T cells (green and white, white arrows) in the tumor microenvironment of CLCG-mutant PDAC (CLCG) compared with WT PDAC. (H) Representative multiplex fluorescent immunohistochemistry images of TIGIT⁺ APCs (orange and green, orange arrows) in the tumor microenvironment of CLCG-mutant PDAC compared with WT PDAC. Scale bar = 30 μm. *P < .05; **P < .01; ns, no significance.

More T Cells in Proximity to Epithelial Cells/Tumor Cells Express TIGIT in Complex of Proteins Associated with SET1—Like Complex Gene—Mutant Pancreatic Ductal Adenocarcinomas

Analyzing the spatial relationships between tumor cells and immune cells within the TME enhances our understanding of immune interactions and the specific spatial distributions of immune cells. Such analyses can also assist in identifying potential targets for immuno-oncology therapies. To characterize the unique spatial immune TME in CLCG-mutant PDACs compared with WT PDACs, we first quantified the mean distance from ECs/tumor cells to T cells, given the important function of T cells in tumor immunity. Although there was no difference in the mean distance between ECs/tumor cells and CD4⁺ T cells or total T cells (Fig. 3A, B), we found that more T cells, specifically CD4⁺ T cells, in proximity to ECs/tumor cells express TIGIT in CLCG-mutant PDACs compared with WT PDACs (Fig. 3D, E, and M). On the contrary, the mean distance from ECs to TIGIT⁺ or total CD8⁺ T cells was similar between WT and CLCG-mutant PDACs (Fig. 3C, F). We used the G-function for a spatial distribution analysis to determine if 2 cell types were randomly mixed or exhibited clustering patterns.³² The AUC of the G-function allowed quantification of the degree of mixing of a cell type around another.³³ Using G-function analysis, we found that ECs/tumor cells in CLCG-mutant PDACs were much more likely to associate with TIGIT⁺ CD4⁺ T cells, but not TIGIT⁺ CD8⁺ T cells or total CD4⁺ or CD8⁺ T cells, compared with WT PDACs (Fig. 3G–L). These findings further support an immunosuppressive TME characterized by CD4⁺ T-cell exhaustion in proximity to tumor cells with increased TIGIT expression in CLCG-mutant PDACs.

Figure 3.

More CD4⁺ T cells in proximity to epithelial cells (ECs) express TIGIT in CLCG-mutant PDACs. Quantification of the mean distance from ECs to (A) T cells; (B) CD4⁺ T cells; (C) CD8⁺ T cells; (D) TIGIT⁺ T cells; (E) TIGIT⁺ CD4⁺ T cells; and (F) TIGIT⁺ CD8⁺ T cells in WT (n = 59) and CLCG-mutant (n = 39) pancreatic ductal adenocarcinomas (PDACs). (G-L) Box plots showing the area under the curve of G-function analysis between EC and T-cell interactions in WT (n = 59) and CLCG-mutant (n = 39) PDACs. (M) Representative mfIHC images comparing TIGIT expression on CD4⁺ T cells between WT PDAC (WT) and CLCG-mutant PDAC (CLCG). White arrows, TIGIT⁺ CD4⁺ T cells; yellow arrows, TIGIT⁻ CD4⁺ T cells; cyan, ECs/tumor cells. Scale bar = 30 μm. *P < .05. ns, no significance.

More T Cells in Proximity to Antigen-Presenting Cells Express TIGIT in Complex of Proteins Associated with SET1—Like Complex Gene—Mutant Pancreatic Ductal Adenocarcinomas

TIGIT on T cells can bind CD155 on APCs to modulate cytokine production and induce immunosuppressive APCs.³⁴ Therefore, we examined the spatial relationships between APCs and T cells. Interestingly, we observed significantly more T cells, including CD4⁺ and CD8⁺ T cells, in proximity to APCs expressing TIGIT in CLCG-mutant PDACs compared with WT PDACs (Fig. 4A–C). G-function analyses were consistent with increased cellular mixing between APCs and TIGIT⁺ T cells in CLCG-mutant PDACs compared with WT PDACs (Fig. 4D–G). There was no significant difference in the spatial distribution between APCs and T cells (total, CD4⁺, and CD8⁺ T cells) (Supplementary Fig. S4). These findings are supportive of a TIGIT—mediated immunosuppressive APC mechanism in CLCG-mutant PDACs.

Figure 4.

More T cells in proximity to antigen-presenting cells (APCs) express TIGIT in CLCG—mutant pancreatic ductal adenocarcinomas (PDACs). Quantification of the mean distance (A-C), and G-function area under the curve (D-F) between (A, D) APCs and TIGIT⁺ T cells, (B, E) APCs and TIGIT⁺ CD4⁺ T cells, and (C, F) APCs and TIGIT⁺ CD8⁺ T cells in wild-type (WT) (n = 59) and CLCG-mutant (n = 39) PDACs. (G) Representative mfIHC images demonstrating spatial relationships between APCs (orange) and TIGIT⁺ T cells (green and white or pink) in CLCG-mutant (CLCG) or WT PDAC. White arrows, TIGIT⁺ T cells; Orange arrows, APCs. Scale bar = 30 μm. *P < .05; **P < .01.

Complex of Proteins Associated with SET1—Like Complex Gene—Mutant Pancreatic Ductal Adenocarcinomas Are Characterized by Malfunctional Immune Cell Crosstalk

Spatial cellular distribution patterns and cell-cell engagement between ECs, APCs, and T cells have been used to imply cell interactions and immune function.^22,35,36 In a typical immune response, the interaction between APCs and CD4⁺ T cells is expected to positively correlate with the interactions between CD4⁺ and CD8⁺ T cells as well as between ECs and CD8⁺ T cells.³⁶ Therefore, we compared the spatial relationships between the 3 cell types and their correlations with each other to identify potential immune dysfunctions in CLCG-mutant PDACs. We quantified cell-cell interactions using both distance and engagement assays, where shorter distance and higher engagement indicated more interaction between cells. As expected, in WT PDACs, when APCs were more engaged with CD4⁺ T cells, more interaction was present between CD4⁺ and CD8⁺ T cells (Fig. 5A, B, left panels). However, this correlation was lost in CLCG-mutant PDACs (Fig. 5A, B, right panels). We also found that there was more interaction between ECs/tumor cells and CD8⁺ T cells when APCs were engaged with CD4⁺ T cells in WT PDACs (Fig. 5C, D, left panels). However, this correlation was again lost in CLCG-mutant PDACs (Fig. 5C, D, right panels). We then investigated the relationship between CD4⁺ -CD8⁺ T-cell interaction and EC/tumor cell-CD8⁺ T-cell interactions. In WT PDACs, the more CD4⁺ T cells engaged with CD8⁺ T cells, the more CD8⁺ T cells interacted with ECs/tumor cells (Fig. 5E, F, left panels). This correlation disappeared in CLCG-mutant PDACs (Fig. 5E, F, right panels). These findings are supportive of an immunosuppressive and tumor—permissive immune microenvironment in CLCG-mutant PDACs, resulting from effector T-cell malfunction mediated by TIGIT.

Figure 5.

CLCG—mutant pancreatic ductal adenocarcinomas (PDACs) are characterized by malfunctional immune cell crosstalk. Correlation analysis in wild-type (WT) (n = 59) and CLCG-mutant PDACs (n=39) between (A) mean distance from antigen-presenting cells (APCs) to CD4⁺ T cells and from CD4⁺ to CD8⁺ T cells; (B) fraction of APCs engaged with CD4⁺ T cells and mean distance from CD4⁺ to CD8⁺ T cells; (C) mean distance from APCs to CD4⁺ T cells and from epithelial cells (ECs) to CD8⁺ T cells; (D) fraction of APCs engaged with CD4⁺ T cells and mean distance from ECs to CD8⁺ T cells; (E) fraction of CD8⁺ T cells engaged with CD4⁺ T cells and mean distance from ECs to CD8⁺ T cells; (F) fraction of CD8⁺ T cells engaged with CD4⁺ T cells and fraction of ECs engaged with CD8⁺ T cells.

Single-Cell RNA-Sequencing Reveals Elevated TIGIT Expression in CD4⁺ T Cells in Complex of Proteins Associated with SET1—Like Complex Gene—Low Pancreatic Ductal Adenocarcinomas

To validate our findings of increased TIGIT expression on CD4⁺ T cells in CLCG-deficient PDACs, we analyzed scRNA-seq data sets from 30 human PDAC samples. A total of 52,512 cells were analyzed and clustered based on published markers.³⁷ The Uniform Manifold Approximation and Projection plots, signature marker genes, and fractions of each cell type are shown in Supplementary Figure S5. We compared PDACs with tumor cells expressing higher CLCG to those expressing lower CLCG. Eight samples (containing 17,910 cells) were identified as the CLCG-high group, and 10 samples (containing 13,391 cells) were identified as the CLCG-low group (Supplementary Table S2). We observed no difference in the CD4⁺ and CD8⁺ T-cell percentages between the CLCG-high and CLCG-low groups (Fig. 6A, B). We next compared TIGIT expressions in total T cells, CD4⁺ T cells, and CD8⁺ T cells. Consistent with our mfIHC data, CLCG-low PDACs had increased TIGIT expression in total T cells and CD4⁺ T cells, but not CD8⁺ T cells, compared with CLCG-high PDACs (Fig. 6C). To investigate the transcriptional profile of CD4⁺ T cells, we subclustered the CD4⁺ T cells and identified 4 populations based on published markers^37–39 (Fig. 6D): the naïve CD4⁺ T cells (Th0) expressing markers CCR7 and SELL; the Tregs expressing high levels of Foxp3, TIGIT, and CTLA4; the exhausted CD4⁺ T cells (Tex) expressing checkpoint markers TIGIT and CTLA4 although lacking Foxp3; and other clusters not expressing genes for known subsets of CD4⁺ T cells. Quantifications of each CD4⁺ T-cell subset demonstrated a significant increase in the percentage of exhausted CD4⁺ T cells in CLCG-low PDACs (Fig. 6E, F). We then performed unsupervised subclustering of CD8⁺ T cells and identified 6 subsets based on published markers^37–41 (Fig. 6G): naïve CD8⁺ T cells (LEF1, CCR7, and SELL); effector CD8⁺ T cells (PRF1, GZMB, and CX3CR1); mucosal-associated invariant CD8⁺ T cells (SLC4A10); exhausted CD8⁺ T cells (EOMES, GZMK, and TIGIT); memory CD8⁺ T cells (GZMH, GZMK, and TIGIT); and tissue-resident memory CD8⁺ T cells (ZNF683 and ITGA1). Interestingly, the only difference observed in CD8⁺ T-cell subsets between CLCG-high and CLCG-low PDACs was an increase in TIGIT⁺ tissue-resident memory CD8⁺ T cells in the CLCG-low group (Fig. 6H, I). This suggests a more suppressive cytotoxic function of tissue-resident memory CD8⁺ T cells in CLCG-low PDACs compared with CLCG-high PDACs. These findings are consistent with our mfIHC results and support a model of TIGIT—mediated T-cell immune inhibition in CLCG-deficient PDACs (Fig. 7).

Figure 6.

scRNA-seq data reveal elevated TIGIT expression in CD4⁺ T cells in CLCG-low pancreatic ductal adenocarcinomas (PDACs). (A) UMAPs of T-cell subsets in CLCG-high and CLCG-low PDACs. (B) Quantification of cell fractions of CD4⁺ and CD8⁺ T cells in total T cells in CLCG-high and low PDACs. (C) Violin plots of normalized gene expression of TIGIT in total T cells, CD4⁺ T cells, and CD8⁺ T cells in CLCG-high and low PDACs. (D) Violin plots for selected marker genes in CD4⁺ T-cell subsets. (E) UMAPs of CD4⁺ T-cell subsets from CLCG-high and CLCG-low PDACs. (F) Quantification of cell fractions of CD4⁺ T-cell subsets in total CD4⁺ T cells from CLCG-high and low PDACs. (G) Violin plots for selected marker genes in CD8⁺ T-cell subsets. (H) UMAPs of CD8⁺ T-cell subsets from CLCG-high and CLCG-low PDACs. (I) Quantification of cell fractions of CD8⁺ T-cell subsets in total CD8⁺ T cells from CLCG-high and low PDACs. n = 8 each group. *P < .05, ****P < .0001. ns, no significance.

Figure 7.

Illustration of CLCG mutation-related reprogramming of pancreatic tumor immune microenvironment. Compared with CLCG-wild-type (WT) pancreatic ductal adenocarcinomas (A), CLCG—mutant pancreatic ductal adenocarcinomas (B) are characterized by increased TIGIT expression in CD4⁺ T cells and antigen-presenting cells (APCs), disrupted immune cell crosstalk, and malfunctional immune responses.

Discussion

Immunosuppressive TME is a hallmark of PDAC, and characterizing TME is crucial for understanding the underlying mechanisms of PDAC progression and identifying novel targets to improve therapeutic efficacy. In this study, we compared the immune composition of the TME in tumor cell—enriched areas of human PDAC samples, specifically focusing on cases with alterations in CLCG to those without such alterations. We have previously demonstrated that CLCG-mutant PDACs are a more aggressive subset of PDAC with poor patient survival.⁹ It is well established that tumor-infiltrating effector CD4⁺ and CD8⁺ T cells exert antitumor effects and correlate with improved patient outcomes.⁴² In this study, we demonstrated that PDACs harboring CLCG mutations are characterized by a reduced number of CD4⁺ T cells and APCs. Tregs comprise a significant fraction of CD4⁺ T cells and are known to be immunosuppressive in PDAC.¹⁵ Our observation revealed a reduced distance between Tregs and Th cells and a positive correlation between the fractions of ECs/tumor cells and Tregs only in CLCG-mutant PDACs. These findings suggest a more immunosuppressive TME in CLCG-mutant PDACs, which may account for the aggressive behavior of these tumors.

TIGIT is an immune inhibitory receptor on both CD4⁺ and CD8⁺ T cells that suppresses antitumor functions of T cells.^43,44 Studies on TIGIT primarily focused on CD8⁺ T cells.¹⁸ Here, we found that in CLCG-mutant PDACs, CD4⁺ T cells express higher levels of TIGIT, whereas CD8⁺ T cells do not. Expression of TIGIT has been observed in Tregs and non—Treg CD4⁺ T cells in patients with PDAC.⁴⁵ Due to limitations in our mfIHC antibody panel, we were unable to differentiate between TIGIT⁺ Tregs and TIGIT⁺ non—Treg CD4⁺ T cells. However, as we did not observe significant changes in the percentage of Tregs, it suggests that the increased TIGIT expression in CD4⁺ T cells is primarily because of non—Treg CD4⁺ T cells. Heiduk et al⁴⁵ previously reported that in PDACs, higher TIGIT expression in intratumoral conventional CD4⁺ T cells is associated with reduced proinflammatory cytokine production and an exhausted phenotype. These findings led us to explore the interactions between TIGIT⁺ CD4⁺ T cells and tumor cells or APCs in PDAC. TIGIT binds to multiple ligands—including CD155, CD112, CD113, and Nectin-4—expressed on APCs and tumor cells, thereby mediating T-cell—intrinsic inhibitory signaling.^34,46–48 Our observation of an increased number of TIGIT⁺ CD4⁺ T cells near tumor cells and APCs in CLCG-mutant PDACs supports the presence of an immunosuppressive TME in this subset of PDACs. In contrast, although other immune checkpoint molecules, such as PD-L1 and TIM3, are also linked to immunosuppressive activity,^20,49–52 we found no difference in their expression levels. This underscores the specific role of TIGIT in immune suppression and its potential as a therapeutic target in CLCG-mutant PDACs.

CD163, a scavenger receptor, is found on dendritic cells⁵³ and macrophages in tissue.³¹ Previous studies have shown that TIGIT can influence macrophage function by activating the TIGIT/CD155 pathway, leading macrophages to adopt an anti-inflammatory M2 cytokine profile.³⁰ In patients with acute myeloid leukemia, an increase in TIGIT⁺ M2 leukemia—associated macrophages has been linked to a poorer prognosis.³¹ Notably, inhibiting TIGIT shifts macrophages toward a proinflammatory M1 phenotype and enhances their phagocytic activity.³¹ Therefore, our observation of increased TIGIT expression on APCs/macrophages in CLCG-mutant PDACs indicates a tumor-promoting environment driven by TIGIT signaling in macrophages. We acknowledge that our mfIHC antibody panels have limitations in accurately identifying specific APC or macrophage subsets, highlighting the need for further investigation in future studies.

A deficient antitumor immune response can be characterized by reduced interactions between tumor cells and CD8⁺ T cells or between APCs and CD4⁺ T cells.^22,35,36 In PDAC, a higher number of APCs near CD4⁺ T cells is positively correlated with the number of tumor cells near CD8⁺ T cells.³⁵ We observed similar patterns of cellular interaction in CLCG-WT PDACs. However, in CLCG-mutant PDACs, the interaction between APCs and CD4⁺ T cells did not correlate with the interaction between tumor cells and CD8⁺ T cells or between CD4⁺ and CD8⁺ T cells, indicating a significant disruption in antigen presentation and immunogenic activity in CLCG-mutant PDACs.

We validated our key mfIHC findings on TIGIT expression in T cells with publicly available human PDAC scRNA-seq data sets. Differential gene expression analysis revealed that TIGIT expression was elevated in total T cells and CD4⁺ T cells, but not in CD8⁺ T cells, from CLCG-low PDACs compared with CLCG-high PDACs. Meng et al³⁸ previously identified an exhausted CD4⁺ T-cell population in PDAC, marked by TIGIT and other signature genes similar to Tregs, but lacking FoxP3 expression. Our scRNA-seq analyses showed a significant increase in these exhausted CD4⁺ T cells, rather than Tregs, in CLCG-low PDACs, likely accounting for the heightened TIGIT expression in CD4⁺ T cells. These results corroborated our mfIHC analysis and suggested that CD4⁺ T-cell inhibition is a primary cause of immune suppression in CLCG-mutant PDACs. Although we did not see overall differences in TIGIT expressions on CD8⁺ T cells, we identified 2 distinct TIGIT-expressing CD8⁺ T-cell populations using published gene signatures.^37–40 One group consisted of exhausted CD8⁺ T cells expressing TIGIT and other exhaustion markers like CTLA4, LAG3, and EOMES.⁵⁴ The other group comprised memory T cells displaying TIGIT and tissue-resident memory cell signatures, such as ITGA1.^18,55,56 Interestingly, although we did not observe any differences in the exhausted CD8⁺ T-cell population, there was an increase in TIGIT⁺ tissue-resident memory CD8⁺ T cells in CLCG-low PDACs. These findings suggest a suppressed cytotoxic function of tissue-resident memory CD8⁺ T cells, alongside the inhibition of CD4⁺ T-cell function in CLCG-low PDACs.

Although the scRNA-seq data set results validated our key mfIHC findings of elevated TIGIT expression on CD4⁺ T cells, this study has several limitations that warrant further investigation. Although we successfully captured tumor cell—enriched areas for mfIHC image analysis, the overall patient sample size remains small due to the limited availability of tumors with genomic sequencing data that are identified as CLCG-mutant variants. Additionally, it is not feasible to differentiate the roles of individual mutations in KMT2A, KMT2B, KMT2C, KMT2D, and KDM6A, which are part of the COMPASS-like complex, due to the small sample size. We observed a higher number of metastatic tumors in the study cohort compared with the control cohort, likely reflecting the aggressive nature of the tumor. This imbalance could potentially confound our results. However, when we compared TIGIT expressions on CD4⁺ T cells between primary and metastatic tumors, we found no significant difference (data not shown). This suggests that the increased prevalence of metastatic tumors does not account for the observed findings. We also observed that CLCG-mutant cases exhibit a higher frequency of squamous differentiation (50% vs 13%) and are more likely to be poorly differentiated (50% vs 12%), as shown in Supplementary Table S1. These findings are consistent with our previous observations and reflect inherent differences between the cohorts. However, there were no well-differentiated PDAC cases in the CLCG-mutant cohort and only a single poorly differentiated PDAC case in the control cohort, limiting the ability to make meaningful direct comparisons. Furthermore, a comparison of important clinicopathologic and molecular features between CLCG-mutant and WT PDACs showed no significant difference (Supplementary Table S1). This suggests that clinicopathologic and molecular features do not account for the observed findings. Another limitation is the absence of specific genetic information about CLCGs in the scRNA-seq samples. Because CLCG mutations are typically loss-of-function mutations, we assume that CLCG-low PDACs behave similarly to CLCG-mutant PDACs. More comprehensive genetic profiling and scRNA-seq analyses will be necessary to further elucidate the molecular mechanisms underlying immunosuppression and to identify novel key factors influencing the immune microenvironment in PDAC.

In summary, this research revealed the specific immunosuppressive microenvironment characteristic of a significant and aggressive subset of PDACs with CLCG mutations, identifying TIGIT as a key mechanism of immune evasion in CD4⁺ T cells and APCs. Additionally, it lays the groundwork for future exploration of the interactions between CLCG-deficient tumor cells and immune cells. We anticipate that this study will aid in establishing TIGIT as a potential therapeutic target to enhance immunotherapy for CLCG-deficient PDACs.

Supplementary Material

The online version contains supplementary material available at https://doi.org/10.1016/j.modpat.2026.100972

Data Availability

The data used or analyzed during the study are available from the corresponding author upon request.