ZEB1 promotes chemoimmunotherapy resistance in pancreatic cancer models by downregulating chromatin acetylation of CXCL16

Pancreatic cancer (PC) is one of the most lethal cancers and was predicted to become the second leading cause of cancer-related death within this decade (1). Only a small proportion of patients with PC would benefit from targeted therapy and immunotherapy (2–4). Intratumor heterogeneity, driven by the unique genomic alterations and the immunosuppressive subpopulation of immune cells and stromal cells in the tumor microenvironment, leads to immune escape and treatment resistance in PC (5–7). Combination therapies may hold promise for improving treatment outcomes. Unfortunately, chemotherapy did not increase the effectiveness of immunotherapy in PC. Emerging evidence indicates that chemotherapy resistance has the potential to facilitate immune evasion via upregulation of immunosuppressive molecules such as CD47, PD-L1, and PGE₂ through metabolic or oncogenic pathway reprogramming (8, 9). It fostered an immunosuppressive microenvironment in PC (10, 11). The tumor microenvironment is critical in driving the malignant phenotypes and treatment resistance (12–16). Dissecting the mechanisms through which the reprogrammed microenvironment grants PC cells the ability to escape the cytotoxic effect of chemotherapy and immunotherapy is key to fostering potential therapeutic strategies, especially combination therapy.

Cellular pyroptosis is a form of inflammatory cell death triggered by pore-forming amino-terminal fragments generated through cleavage of gasdermin family proteins. It is characterized by cell membrane perforation, activation of inflammasomes, and release of proinflammatory cytokines such as IL-1β and IL-18. In recent years, CD8⁺ T cells, as key effector cells of the cytotoxic immune response, have been shown to exert antitumor effects by inducing tumor cell pyroptosis, primarily through release of granzyme A (GZMA) and GZMB. Mechanistically, GZMA induces pyroptosis in target cells by cleaving gasdermin B (GSDMB) (17), while GZMB not only activates caspase-3 in target cells (18), but also directly cleaves GSDME at the same site as caspase-3, thereby triggering pyroptosis (19). Meanwhile, gemcitabine, a nucleoside analog, has been found to indirectly promote pyroptosis by modulating mitochondrial ROS and activating the caspase-3/GSDME pathway, in addition to its direct inhibitory effects on tumor cell proliferation (20). These findings provide a potential theoretical basis for combining immunotherapy with chemotherapy.

Epigenetic modification, such as DNA and histone modification, profoundly affects the tumor immune microenvironment (TIME) by dynamically modifying gene expression in the tumor microenvironment. Through inhibition of DNA methylation, suppression of immune-related genes can be reversed, leading to an increase in the number and function of tumor-infiltrating CD8⁺ T cells and thereby restoring immune function (21). Epigenetic modification induced cancer immune evasion by decreasing tumor immunogenicity, a critical factor associated with neoantigen quality and its presentation (22, 23). Histone acetylation modulates chromatin accessibility, which plays a pivotal role in cancer immune evasion (24–29).

Histone deacetylases (HDACs) are a group of enzymes that remove acetyl groups from histones. HDAC inhibition increases the sensitivity of chemotherapy and suppresses PC progression by blocking phenotypic transformation of fibroblasts in preclinical models (30, 31). HDAC1 is identified as a marker of poor immune checkpoint blockade (ICB) response in hepatocellular carcinoma (32); its inhibition enhances CD8⁺ T cell activity and improves immunotherapy efficacy in lung and colorectal cancers (33). However, its role in PC remains unclear. A phase II clinical trial showed that HDAC inhibitor had a synergistic effect when combined with anti–PD-1 immunotherapy and anti-VEGF antibody in patients with proficient mismatch-repair/microsatellite-stable (pMMR/MSS) colorectal cancer, who are deemed resistant to immunotherapy (34). However, whether combining immunotherapy, chemotherapy, and HDAC inhibitors would provide synergistic efficacy in PC remains to be determined. While zinc finger E-box binding homeobox 1 (ZEB1) is known to play critical roles in chemoresistance and cellular plasticity (35, 36), its contribution to chemoimmunotherapy resistance, particularly through regulation of HDAC-associated chromatin accessibility and immune microenvironment reprogramming, remains poorly understood.

In this study, we identified that knocking down ZEB1 substantially inhibited PC progression and enhanced chemoimmunotherapy response in vivo through enhancing CD8⁺ T cell–induced pyroptosis and inhibiting crosstalk between CD8⁺ T cells and neutrophils in PC. Treatment with mocetinostat (an epigenetic reprogramming inhibitor of ZEB1) synergizes with gemcitabine and anti–PD-1, enhancing the efficacy of chemotherapy and immunotherapy in an allograft mouse model, patient-derived organoid models, and patient-derived xenograft mouse models.

ZEB1 promotes chemoimmunotherapy resistance in PC. Given that patients with PC are resistant to chemoimmunotherapy, we established 2 human PC stable cell lines (AsPC-1-R and MIA PaCa-2-R) that are resistant to gemcitabine and inactivate CD8⁺ T cells (Supplemental Figure 1, A–F; supplemental material available online with this article; https://doi.org/10.1172/JCI195970DS1). To investigate the underlying mechanism, we performed RNA sequencing in WT and chemoimmunotherapy-resistant PC cell lines. The upregulated transcription factors (TFs) in chemoimmunotherapy-resistant PC cells were merged with TFs that were upregulated in PC tissue and HDAC-interacting TFs (Supplemental Figure 1G). We identified 6 genes and finally focused on ZEB1, which has been reported to promote tumor progression and migration. We found that ZEB1 was upregulated upon gemcitabine treatment as well as in gemcitabine-resistant stable cell lines (Supplemental Figure 1, H and I). Knockdown (KD) of ZEB1 increased the sensitivity of PC to gemcitabine and activated CD8⁺ T cells (Supplemental Figure 1, J–M).

Targeting ZEB1 activates TIME and sensitizes PC to chemoimmunotherapy. Next, we assessed the impact of ZEB1 KD on chemoimmunotherapy sensitivity in vivo using an allograft PC mouse model (Supplemental Figure 1N). To investigate whether knocking down ZEB1 synergizes with chemoimmunotherapy through TIME, we orthotopically inoculated control (KPC-shV) and ZEB1-KD KPC (KPC-shZEB1) cells into immunocompetent and immunodeficient mice under treatment with gemcitabine. The results showed that ZEB1 inhibition induced more dramatic tumor regression in immunocompetent mice (Figure 1, A and B). Inhibition of ZEB1 notably enhanced the tumor-suppressive effect of gemcitabine and prolonged overall survival (OS) in immunocompetent mice (Figure 1, C–E, and Supplemental Figure 1, O and P). Further experiments confirmed that ZEB1 KD in combination with gemcitabine and anti–PD-1 therapy resulted in superior tumor suppression (Figure 1, F and G). To further investigate the function of ZEB1 on TIME, we performed single-cell RNA-Seq (scRNA-Seq) using the tumor tissue collected from mice allografted with KPC-shV or KPC-shZEB1 cells treated with gemcitabine shown in Figure 1, C, H, and I. Compared with control, ZEB1 KD substantially increased the proportion of total T cells and CD8⁺ T cells (Figure 1J, Supplemental Figure 1Q, and Supplemental Figure 2, A and B). These findings were validated by IHC staining and flow cytometry analysis (Supplemental Figure 1, R and S). Ligand-receptor pair communication analysis revealed enhanced interaction between PC cells and T cells following ZEB1 KD (Figure 1K). These findings indicated that blocking ZEB1 enhanced chemoimmunotherapy through activation of CD8⁺ T cells in vivo.

Figure 1

Blocking of Zeb1 enhances gemcitabine efficacy through activation of the PC immune microenvironment. (A and B) Tumor images and weight of orthotopic allograft mouse model (immunocompetent and immunodeficient) established from KPC-shV and KPC-shZeb1 cells and treated with gemcitabine (GEM; 50 mg/kg) 3 times a week (n = 3). (C and D) Tumor images and weight of orthotopic allograft mouse model established from KPC-shV and KPC-shZeb1 cells in each treatment condition (n = 5). (E) Survival of orthotopic allograft mouse model established from KPC-shV and KPC-shZeb1 cells in each treatment condition (n = 6–10). (F and G) Tumor images and weight of orthotopic allograft mouse model established from KPC-shV and KPC-shZeb1 cells and treated with gemcitabine (50 mg/kg) and anti–PD-1 (10 mg/kg) 3 times a week (n = 6). (H) Uniform manifold approximation and projection (UMAP) plot of scRNA-Seq data derived from orthotopic allograft mouse model reveals the presence of 10 distinct cell types. Cell types are distinguished by color. (I) UMAP plot displays the distribution and subclustering of T and NK cell subsets. TSNE, t-distributed stochastic neighbor embedding. (J) Stacked histogram shows the proportion of each T/NK cluster in KPC-shV and KPC-shZeb1 mouse tumor tissues. (K) Circle plots depict the strength of cell-cell interactions between subclusters of T/NK cells and tumor cells, as identified through CellChat analysis. The edge weights and numerical values indicate the strength score of these interactions, while the direction of the arrows denotes the cell clusters responsible for signaling release and reception. *P < 0.05, **P < 0.01, ***P < 0.001, by unpaired, 2-tailed Student’s t test (B), 1-way ANOVA with Tukey’s multiple-comparison test (D and G), and log-rank test (E). Data are presented as mean ± SD in B, D, and G.

GZMA⁺CD8⁺ T cells are enriched in ZEB1-KD tumors with gemcitabine treatment. To identify the specific functional subtype of CD8⁺ T cells that was associated with ZEB1, we further clustered CD8⁺ T cells into LEF1–naive, LY6C2–naive, GZMA–effector, and DSCAM–effector T cells based on gene signatures (Figure 2A). Of these 4 subsets of CD8⁺ T cells, GZMA–effector T cells (GZMA⁺CD8⁺ T cells), which constituted the largest group of cytotoxic effector T cells (cytotoxic T lymphocytes [CTLs]), were increased by 3.8-fold after ZEB1 KD (Figure 2, B–D). This subset was the only one that prominently expressed Gzma (Figure 2B and Supplemental Figure 2, C–E), whose role as a cytotoxic mediator in killing tumor cells has been widely reported (17, 37, 38). To further evaluate the role of GZMA⁺CD8⁺ T cells in PC, we performed spatial transcriptomics and multiplex IHC (mIHC) on the same tumor tissues as used for scRNA-Seq. We determined that the percentage of GZMA⁺CD8⁺ T cells was remarkably increased, especially in the core region of tumor tissues, with ZEB1 KD (Figure 2, E and F). Subsequent intercellular communication analysis revealed that compared with the other 3 subtypes of CD8⁺ T cells, GZMA⁺CD8⁺ T cells had the strongest interaction with tumor cells upon ZEB1 inhibition (Figure 2, G–I). These findings demonstrated that GZMA⁺CD8⁺ T cells were the main mediators of the immune response triggered by ZEB1 inhibition in PC.

Figure 2

Increased infiltration of Gzma⁺ CD8⁺ T cells in tumor tissue with Zeb1 KD. (A) UMAP reveals that CD8⁺ T cells can be classified into 4 distinct major subtypes. Tef, T effector. (B) Density plot shows the expression of the Gzma gene, with brighter colors indicating higher expression. Gzma is mainly expressed in the Gzma⁺ effector CD8⁺ T cell subset. (C) Percentage of Gzma⁺ effector CD8⁺ T cells within the total Cd45⁺ population in the KPC-shV and KPC-shZeb1 groups. (D) Number of inferred significant ligand-receptor (LR) pairs between any 2 cell types based on single-cell analysis data. (E) Top panels: MIF of mouse tumor tissues. Scale bars: 2 mm (left) and 200 μm (right). Bottom panels: Marker gene set scores for CD8⁺ T cells based on spatial transcriptomics data. Brighter colors indicate higher scores, suggesting a greater abundance of CD8⁺ T cells in those regions. (F) Spatial transcriptome sequencing displays the distribution of 4 major annotated cell types in the control and experimental groups. (G and H) Circle plots show the number and the strength score of LR among 4 cell types across 2 groups, based on spatial transcriptome data. (I) Specifically, the interaction strength of GZMA-related LR pairs is dramatically increased in the KPC-shZeb1 group.

ZEB1 inhibition enhances the anticancer response of CD8⁺ T cells and CAR T cell therapy. PC is characterized by a suppressive immune microenvironment that severely limits the CTL response (39–41). In concordance with the scRNA-Seq analysis, the results of in vitro experiments illustrated that ZEB1 inhibition enhanced recruitment and activation of CD8⁺ T cells while reducing their apoptosis (Figure 3, A–D, and Supplemental Figure 3A). We evaluated whether the activated CD8⁺ T cells decreased cell viability and potentiated the gemcitabine sensitivity of PC cells, and found that CD8⁺ T cell treatment augmented gemcitabine sensitivity, which was enhanced by ZEB1 inhibition (Supplemental Figure 3, B–D). Moreover, expression of MHC-I was upregulated in ZEB1-KD PC cells (Supplemental Figure 3E). MHC-I signaling analyzed by scRNA-Seq data confirmed the strengthened interaction between CD8⁺ T cells and ZEB1-KD PC cells (Supplemental Figure 3F). Besides, the CAR T cell model was established to evaluate tumor lysis activity both in vitro and in vivo. Under different effector-to-target (E/T) ratios, we found that CAR T cells encountering ZEB1-KD AsPC-R cells had elevated lysis ability, which was also validated in the KPC-OVA/OT1-CD8⁺ T cell model (Figure 3, E and F). Furthermore, OT1-CD8⁺ T cells showed a dramatic antitumor effect in vivo when ZEB1 was knocked down in tumor tissue, highlighting the key function of ZEB1 in regulating the sensitivity of PC to CAR T cell therapy (Figure 3, G and H). Additionally, we wondered whether the recruited CD8⁺ T cells increased gemcitabine sensitivity by regulating expression of genes that are associated with gemcitabine sensitivity. Equilibrative nucleoside transporter 1 (ENT1) is a therapeutic response marker for gemcitabine. Our prior study revealed that ZEB1 induces PC gemcitabine resistance by inhibiting ENT1 transcription (42), and thus we wondered whether CD8⁺ T cells could also modulate ENT1 expression. Treatment with conditioned medium (CM) of CD8⁺ T cells upregulated ENT1 expression and enhanced Cy5-gemcitabine accumulation in PC cells (Figure 3I and Supplemental Figure 3, G–I). Thus, in addition to their conventional cytotoxic effects, cytotoxic lymphocytes also upregulated ENT1 expression in PC cells, increasing their sensitivity to gemcitabine. Besides, a recent study showed that gemcitabine could foster pyroptosis by activating the caspase-1/GSDMD pathway in PC, and pyroptosis activation by VbP, an enzymatic activator of caspase-1, confers PC gemcitabine sensitivity (43). Meanwhile, cytotoxic lymphocytes can trigger pyroptosis in target cells (17). Accordingly, we wondered whether the recruitment of cytotoxic CD8⁺ T cells induced by ZEB1 KD could boost gemcitabine-related pyroptosis. Gene set enrichment analysis (GSEA) of differentially expressed genes in PC cells after ZEB1 KD revealed pyroptosis as a statistically significant pathway (Supplemental Figure 3J). We then performed a classical calcium release assay to evaluate tumor pyroptosis and found that the combination of gemcitabine and CM from CD8⁺ T cells increased calcium influx, enhancing the lethal lysis of PC cells (Figure 3J). Furthermore, KD of ZEB1 acted synergistically with gemcitabine and CD8⁺ T cells to promote pyroptosis in PC cells (Figure 3, J and K, and Supplemental Figure 3, K and L). Collectively, these results indicate that targeting ZEB1 in PC cells synergized with gemcitabine by activating CD8⁺ T cells, thereby enhancing anticancer response and cytotoxicity.

Figure 3

Blocking of ZEB1 enhances the antitumor activity of CD8⁺ T cells. (A) CD8⁺ T cell migration assay. (B) Detection of the level of activation markers of mouse CD8⁺ T cells by qPCR after coculturing with KPC-shV and -shZeb1 cells for 48 hours. (C) Flow cytometry analysis of the apoptotic rate of human CD8⁺ T cells after coculturing with AsPC-R-shV and -shZEB1 cells. (D) Western blot detection of apoptotic markers in human CD8⁺ T cells after coculturing with AsPC-R-shV, shZEB1 cells. (E) Specific lysis of AsPC-R-shV–luciferase and AsPC-R-shZEB1–luciferase cells after coculturing with CAR T cells for 48 hours. (F) Detection of specific lysis of KPC-shV-Ova–luciferase and KPC-shZeb1-Ova–luciferase after coculturing with mouse Ot1-CD8⁺ T cells for 24 hours. (G and H) Tumor images and weight of orthotopic allograft mouse model established from KPC-shV-Ova and KPC-shZeb1-Ova cells and treated with mouse Ot1-CD8⁺ T cells (n = 3–5). (I) Detection of ENT1 expression in AsPC-R-shV and -shZEB1 cells after treatment with CD8⁺ T cell CM for 48 hours. (J) Representative images of AsPC-R-shV and -shZEB1 cell after treatment with gemcitabine (1,000 nM) and conditioned medium of CD8⁺ T for 48 hours (n = 3). Cells were labeled using the calcium ion probe Calbryte 590 (AAT Bioquest, 20700), and the red fluorescence signal represents pyroptotic cells. Scale bars: 50 μm. (K) Detection of pyroptotic proteins in AsPC-R-shV and -shZEB1 cells after treatment with gemcitabine (1,000 nM) and coculture with CD8⁺ T cells. Data are representative of at least 3 (A–F, I, and K) independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001, by unpaired, 2-tailed Student’s t test (A–C), 2-way ANOVA (E and F), and 1-way ANOVA with Tukey’s multiple-comparison test (H). Data are presented as mean ± SD in A–C and H and mean ± SEM in E and F.

Blocking of ZEB1 activates CD8⁺ T cells partially by inhibiting the function of neutrophils. Our scRNA-Seq atlas analysis of TIME in vivo revealed that inhibition of ZEB1 not only increased the CD8⁺ T cell population but also decreased the proportion of granulocytes (neutrophils) and their interaction with tumor cells (Figure 4A and Supplemental Figure 1). The spatial transcriptomics and mIHC analyses indicated that there were fewer neutrophils infiltrated within the tumor’s core region when ZEB1 was knocked down, aligning with our scRNA-Seq findings (Figure 2F). Additionally, scRNA-Seq analysis and in vitro experiments demonstrated that ZEB1 KD inhibited neutrophil activities, including migration and polarization, thereby promoting N1-polarized neutrophil differentiation (Figure 4, B–D). Considering that neutrophils constitute a prominent immunosuppressive cell population within the tumor microenvironment (TME), leading to T cell exclusion and unresponsiveness to antigen-specific stimulation (44), we investigated whether ZEB1 KD affects T cell response via neutrophils. Initially, we verified the suppressive effects of neutrophils on CD8⁺ T cell migration in vitro (Figure 4E). To elucidate how ZEB1 KD in tumor cells could impair the function of CD8⁺ T cells through neutrophils, we established a coculture system with these 3 cell types and then collected them for analysis (Figure 4F). Assessment of CD8⁺ T cell and neutrophil markers revealed pronounced activation of CD8⁺ T cells and inhibition of neutrophils following ZEB1 KD (Figure 4, G and H). Moreover, ZEB1 KD notably augmented the therapeutic efficacy of gemcitabine and anti-Ly6G combination therapies by decreasing infiltration of neutrophils, while increasing infiltration of CD8⁺ T cells (Figure 4, I and J, and Supplemental Figure 4, A–D). To pinpoint the crucial factors mediating the interaction between PC cells and neutrophils, we analyzed intercellular communications involving ligand-receptor pairs and found that the SPP1 (tumor)–CD44 (neutrophil) signal was dramatically inhibited in the ZEB1 KD group (Supplemental Figure 4, E–G). SPP1-CD44 is critical in neutrophil recruitment and the formation of neutrophil extracellular traps (NETs). We further confirmed that KD of ZEB1 decreased SPP1 expression both in vitro and in vivo (Supplemental Figure 4, H and I). Furthermore, SPP1 recombinant protein treatment induced a dose-dependent inhibition of neutrophil cytotoxicity (Supplemental Figure 4J). And neutrophils inhibited CD8⁺ T cell function by downregulating PD-L1 (Supplemental Figure 4, K and L). Taken together, the results indicate that targeting ZEB1 in tumor cells effectively inhibited recruitment and polarization of neutrophils, leading to activation of CD8⁺ T cells and a synergistic antitumor effect with chemoimmunotherapy in PC.

Figure 4

Zeb1 promotes recruitment of neutrophil and drives their polarization toward an immunosuppressive phenotype. (A) Circle plots compare the strengths of cell-cell interactions between granulocytes and other cell types. (B) Neutrophil migration assay. Relative migration of mouse neutrophils after coculture with KPC-shV and -shZeb1 cells for 12 hours. (C) Violin plot showing AUCell scores of the N1 and N2 gene sets in neutrophils derived from the shV and shZeb1 models. (D) Neutrophil activation. Detection of N1 polarization markers (Icam, Cxcl10, Tnfa) and the N2 polarization marker Cxcr2 in neutrophils by qPCR after coculturing with KPC-shV or -shZeb1 cells for 12 hours. (E) Relative migration of mouse CD8⁺ T cells after coculturing with neutrophils. (F) Schematic of a 3-cell coculture system. (G) CD8⁺ T cells were isolated from the 3-cell coculture systems, and levels of activation markers were detected by qPCR. (H) Neutrophils were isolated from the 3-cell cocultured system, and levels of N1 and N2 polarization markers were detected by qPCR. (I and J) Tumor images and weight of orthotopic allograft mouse model established from KPC-shV and -shZeb1 cells and treated with gemcitabine (50 mg/kg) and anti-Ly6g (25 μg) 3 times a week (n = 5). Data are representative of at least 3 (B, D, E, G, and H) independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001, by unpaired, 2-tailed Student’s t test (B, D, E, G, and H), Wilcoxon’s rank-sum test (C), and 1-way ANOVA with Tukey’s multiple-comparison test (J). Data are presented as mean ± SD in B, D, E, G, H, and J.

Inhibition of ZEB1 synergizes with chemoimmunotherapy through activation of CXCL16. To further elucidate the mechanism of ZEB1-regulated chemoimmunotherapy, we performed intercellular communication analysis using our scRNA-Seq data. We identified 64 signaling pathways that were notably upregulated in ZEB1-KD tumors. Among various cytokines and chemokines in KPC cells, CXCL16 was the most markedly upregulated upon ZEB1 KD (Figure 5A and Supplemental Figure 5A). Given that CXCR6, the specific receptor for CXCL16, is reported to be highly expressed in intratumoral CD8⁺ T cells, and that CXCR6⁺CD8⁺ T cells are critical for checkpoint blockade therapy (45–47), we hypothesized that the increased sensitivity to chemoimmunotherapy following ZEB1 KD is attributable to the enhanced chemotaxis and activity of CD8⁺ T cells driven by elevated CXCL16. We confirmed the reversed correlation between ZEB1 and CXCL16 (Supplemental Figure 5, B–E). Exogenous recombinant CXCL16 increased migration and activation of CD8⁺ T cells, leading to the enhanced antitumor effect (Supplemental Figure 5, F–K). The enhancement of tumor cell recognition and CD8⁺ T cell activation in OT-1 T cells suggested a direct response to CXCL16 stimulation. To further delineate the role of the ZEB1/CXCL16 axis in CD8⁺ T cell activity and the sensitivity of PC tumors to chemoimmunotherapy or T cell therapy, we investigated CD8⁺ T cell functions. We found that CXCL16 KD notably attenuated the migration, activation, and cytotoxicity of CD8⁺ T cells enhanced by ZEB1 KD (Figure 5, B–D, and Supplemental Figure 5, L–N). Notably, the therapeutic benefit of chemoimmunotherapy or T cell therapy induced by ZEB1 inhibition was notably reversed by CXCL16 KD (Figure 5, E–H, and Supplemental Figure 5, O–Q). Collectively, these results indicate that CXCL16/CXCR6 signaling, which is activated when ZEB1 is knocked down, mediated recruitment and activation of CD8⁺ T cells, rendering PC tumors highly vulnerable to chemoimmunotherapy and T cell therapy.

Figure 5

ZEB1/HDAC1 inhibits the recruitment and function of CD8⁺ T cells by epigenetically regulating CXCL16. (A) Circle plot showing the inferred Cxcl16-Cxcr6 signaling network between each CD8⁺ T cell subcluster and tumor cells. Edge weights represent the strength of the interactions. (B) Relative migration of mouse CD8⁺ T cells, which were cocultured with KPC-shV-shV, shZeb1-shV, and KPC-shZeb1-shCxcl16 cells for 48 hours. (C) Detection of activation markers in mouse CD8⁺ T cells, which were cocultured with tumor cells for 48 hours. (D) Detection of specific lysis of tumor cells after coculturing with mouse Ot1-CD8⁺ T cells for 24 hours. (E and F) Tumor images and weight of orthotopic allograft mouse model established from the indicated cell lines and treated with gemcitabine (50 mg/kg) 3 times a week (n = 5). (G and H) Tumor images and weight of orthotopic allograft mouse model established from the indicated cell lines and treated with mouse Ot1-CD8⁺ T cells (n = 3). (I) Relative mRNA level of ZEB1 and CXCL16 in AsPC-R-shV and -shZEB1 cells. (J) ATAC-Seq, CUT&Tag-Seq of H3K27ac, CUT&Tag-Seq of H3K4me, and RNA-Seq showed changes in chromatin openness, transcriptional activity, and apparent modification levels in the CXCL16 promoter region and gene body region. (K and L) CUT&Tag-qPCR assay of the CXCL16 promoter region in AsPC-WT-siNC, AsPC-R-siNC, and AsPC-R-siZEB1 cells with antibodies against H3K27ac and HDAC1 (n = 3). Data are representative of at least 3 (B–D, I, K, and L) independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001, by 1-way ANOVA with Tukey’s multiple-comparison test (B, C, F, H, I, K, and L) and 2-way ANOVA (D). Data are presented as mean ± SD in B, C, F, H, I, K, and L, and mean ± SEM in D.

ZEB1/HDAC1 complex suppressed CD8⁺ T cells activity through epigenetic inhibition of CXCL16. To elucidate the specific mechanism through which ZEB1 negatively regulated CXCL16 expression to decrease response to chemoimmunotherapy, we conducted a luciferase reporter assay to assess the role of ZEB1 in CXCL16 transcriptional regulation. The results showed that ZEB1 KD increased CXCL16 mRNA level but did not affect CXCL16 promoter activity (Figure 5I and Supplemental Figure 5, R and S), suggesting that ZEB1 may regulate CXCL16 in an epigenetically dependent manner. We further investigated the modification of the CXCL16 promoter using CUT&Tag sequencing (CUT&Tag-Seq). AsPC-1-R cells showed a clearly reduced level of H3K27ac in the CXCL16 promoter region, while the H3K4me level only showed a slight reduction (Figure 5J). These findings imply that histone acetylation predominantly regulated CXCL16 expression in PC. CUT&Tag qPCR further confirmed a lower enrichment of H3K27ac signal in AsPC-1-R cells compared with the parental cells, while ZEB1 KD partially restored the H3K27ac enrichment (Figure 5K). Next, we sought to elucidate the mechanism by which ZEB1 modulates H3K27 acetylation at the CXCL16 promoter. Given that HDAC1 is a well-characterized corepressor of ZEB1 and facilitates ZEB1-mediated deacetylation of downstream targets, we performed HDAC1 CUT&Tag-qPCR in parental AsPC-1 cells and gemcitabine-resistant AsPC-1-R cells, with or without ZEB1 KD. Strikingly, HDAC1 enrichment at the CXCL16 promoter region was markedly elevated in gemcitabine-resistant cells, and this effect was almost completely abrogated upon ZEB1 depletion (Figure 5L). Collectively, these findings revealed that the epigenetic modification of the CXCL16 promoter by the HDAC1/ZEB1 complex contributed to CXCL16 silencing in PC.

HDAC inhibitor synergizes with chemoimmunotherapy and CAR T cell therapy in PC. To evaluate the translational potential of ZEB1 in PC chemoimmunotherapy, we selected mocetinostat, an epigenetic inhibitor of ZEB1, to assess its synergistic effect with chemoimmunotherapy in PC. We established an orthotopic allograft mouse model and treated the mice with gemcitabine, gemcitabine+anti–PD-1 (G+P), gemcitabine + mocetinostat (G+M), and gemcitabine + anti–PD-1 + mocetinostat (G+P+M), respectively. We found G+P plus 60 mg/kg mocetinostat treatment (G+P+M) significantly inhibited the tumor growth; however, this regimen didn’t significantly prolong the OS compared with G+M treatment (Supplemental Figure 6, A–C). Since previous clinical trials evaluating the efficacy of gemcitabine in combination with mocetinostat in PC patients did not meet the primary end point due to severe side effects, we decided to explore whether a lower dosage of mocetinostat (30 mg/kg) might improve efficacy. As expected, this treatment strategy significantly reduced tumor volume and improved OS (Figure 6, A and B, and Supplemental Figure 6D), while having markedly less severe side effects, as evidenced by tissue morphology and blood parameters associated with liver and kidney function (Supplemental Figure 6, E and F). To investigate the mechanism by which mocetinostat enhances PC chemoimmunotherapy efficacy, we conducted flow cytometry and IHC to evaluate tumor-infiltrated immune cell profiling. The results showed that the triple-drug treatment led to a dramatic increase in CD8⁺ T cell infiltration (Figure 6C and Supplemental Figure 6G). Meanwhile, levels of neutrophils, often implicated in suppressing anticancer T cell activity across various cancer types, were significantly reduced following G+P+M treatment (Figure 6C and Supplemental Figure 6G). These results indicated that mocetinostat increased chemoimmunotherapy response by remodeling TIME. CAR T cell therapy has shown promising outcomes in hematological malignancies (48). However, the efficacy of CAR T therapy in solid tumors remains limited, particularly in highly desmoplastic PC. To elucidate the impact of mocetinostat on CAR T cell therapy, we constructed patient-derived PC organoids (PDOs) and established CAR T–infiltrated and real-time killing models. High-content confocal laser scanning microscopic images and videos showed that mocetinostat facilitated directional migration and augmented infiltration of CAR T cells into the PDOs within the coculture environment (Figure 6D). Notably, the synergistic effect of mocetinostat and CAR T cells induced dramatic PDO deformation, extensive cell lysis, and cell apoptosis, but the effect was not observed with CAR T cells alone (Supplemental Figure 6, H–L). Given the substantial efficacy of G+P+M in TIME activation, a patient-derived organoid xenograft mouse model (PDOX) was established to investigate whether mocetinostat enhances CAR T cell therapy response in vivo. As shown in Figure 6, E–I, treatment with mocetinostat dramatically improved the antitumor efficacy of CAR T therapy. Thus, mocetinostat reinforced the antitumor immunity and enhanced the efficacy of chemoimmunotherapy and CAR T cell therapy in PC.

Figure 6

Mocetinostat enhances chemoimmunotherapy and CAR T efficacy in PC. (A) Tumor images of orthotopic allograft mouse model established from KPC cells in each treatment condition: gemcitabine (50 mg/kg); gemcitabine+Moce (30 mg/kg); gemcitabine+Moce (30 mg/kg) + anti–PD-1 (10 mg/kg), 3 times a week (n = 6). (B) Survival of orthotopic allograft mouse model established from KPC cells in each treatment condition (n = 10). (C) Flow cytometry analysis of the proportion of all T cells (Cd45⁺, Cd3⁺), CD8⁺ T cells (Cd3⁺, Cd8⁺), and neutrophils (Cd11b⁺, Ly6g⁺) to total Cd45⁺ cells in tumor tissue (n = 3). (D) CAR T–infiltrated PDO model: CAR T was used to infect PDOs for 24 hours after the 24 hours of Moce (500 nM) treatment of PDOs (n = 3). Left: 3D model synthesized by the algorithm. Green represents the PDOs, and red represents CAR T cells. On the right is the 2D image of CAR T–infiltrating PDOs; CAR T is shown in red with living cell dye. Scale bars: 20 μm. (E) Tumor images of the PDOX mouse model treated with CAR T and Moce. (F and G) Tumor weight and volume of the PDOX mouse model (n = 3–5). (H) Representative H&E and Ki-67 IHC staining in tumor tissues of the PDOX mouse model established from PC patients’ organoids and treated with CAR T and Moce (n = 3). Scale bars: 50 μm. (I) Flow cytometry analysis of the proportion of CAR T cells (human CD3⁺CD8⁺) divided into total cells in mouse tumor tissues of PDOX mice after treatment with CAR T and Moce (n = 3–5). *P < 0.05, **P < 0.01, ***P < 0.001, by log-rank test (B), 1-way ANOVA with Tukey’s multiple-comparison test (C, F, and I) and 2-way ANOVA (G). Data are presented as mean ± SD in C, F, and I; mean ± SEM in G.

Mocetinostat enhances chemoimmunotherapy sensitivity by targeting HDAC1/2-ZEB1 complex. Next, we treated resistant cells with mocetinostat and found that it increased the sensitivity of AsPC-R cells to gemcitabine and activated CD8⁺ T cells in vitro (Supplemental Figure 7, A–F). To determine whether the efficacy of mocetinostat in the PC response to chemoimmunotherapy depended on inhibiting the HDAC1/2-ZEB1 functional complex, we performed a co-IP assay. We found that ZEB1 could interact with HDAC1 and HDAC2 (Supplemental Figure 7G). Intriguingly, mocetinostat reduced the stability of ZEB1 and HDAC1 but not HDAC2 (Supplemental Figure 7H). These results indicated that mocetinostat promoted the response to chemoimmunotherapy by disrupting the HDAC1-ZEB1 complex in PC.

ZEB1 and CXCL16 expression are positively correlated with gemcitabine resistance and associated with poor clinical outcomes. We explored the correlation between ZEB1 and CXCL16 expression, as well as CD8⁺ T cell infiltration, and the sensitivity of gemcitabine in patients with PC. ZEB1 expression was positively correlated with gemcitabine resistance, whereas CXCL16 or CD8 expression was negatively correlated (Figure 7A). We further validated these findings using scRNA-Seq of tumor tissues from patients with PC, demonstrating that those with higher CXCL16 expression were more sensitive to gemcitabine treatment (Figure 7, B–E, and Supplemental Figure 7, I and J). Collectively, our data highlight ZEB1 as a central regulator modulating the efficacy of immunotherapy and gemcitabine in PC through its epigenetic regulation of CXCL16 expression and the intratumoral balance of CD8⁺ T cells and neutrophils (Figure 7F).

Figure 7

ZEB1 and CXCL16 are associated with chemotherapy resistance, immunosuppression, and prognosis in PC patients. (A) Multiple immunofluorescence of ZEB1, CXCL16, and CD8 in tumor tissues of chemosensitive and chemoinsensitive PC patients. Scale bars: 50 μm. (B) Based on CXCL16 expression, tumor cells were categorized into CXCL16-high and CXCL16-low groups. (C) Compared with the chemoresistant group, tumor cells with high CXCL16 expression were predominantly found in the chemosensitive group. (D) Stacked histogram indicates a dramatic increase in the proportion of CD8⁺ T cells in the chemosensitive group. (E) In the TCGA dataset, patients receiving adjuvant gemcitabine chemotherapy were stratified into high-CXCL16-expression (n = 21) and low-expression groups (n = 44) based on the optimal cutoff value for CXCL16 gene expression. Kaplan-Meier survival curves indicate that patients with high CXCL16 expression exhibited a significantly better prognosis. (F) Schematic diagram. Crosstalk between PC cells, CD8⁺ T cells, and neutrophils contributes to tumor immune invasion and gemcitabine resistance through the ZEB1/HDAC1-CXCL16 signaling axis.

Chemotherapy inadvertently promotes tumor immune escape, ultimately leading to treatment failure, recurrence, and metastasis. It has been demonstrated that in gemcitabine-resistant PDAC cells, CMTM6 stabilizes PD-L1 expression and inhibits T cell activity (49, 50). In addition, gemcitabine induces the DNA damage response, activating APOBEC3C/3D enzymes, which enhance DNA repair and upregulate the immune checkpoint molecule PD-L1, consequently suppressing T cell function and facilitating immune evasion (51). And EVs secreted by chemotherapy-resistant cells transport miR-21-5p (known to target tumor suppressors such as PDCD4) or PVT1, further inhibiting T cell activity (52).

ZEB1 is one of the key TFs that promote cellular plasticity and tumor metastasis in PC (53, 54). Previous studies showed that ZEB1 induced gemcitabine resistance in PC by activating ITGA3/JNK signaling and downregulating ENT1 (42). However, the role of ZEB1 in driving chemoimmunotherapy resistance remains elusive. We found that blocking ZEB1 enhanced the efficacy of chemotherapy and immunotherapy (anti–PD-1 therapy and CAR T therapy) by reprogramming the immune microenvironment of PC. Specifically, ZEB1 inhibition increased infiltration of CD8⁺ T cells while decreasing infiltration of neutrophils in vivo. Mechanistically, ZEB1 bound with HDAC1 to regulate the chromatin accessibility of CXCL16 through histone acetylation, which induced the imbalance of CD8⁺ T cells and neutrophils. Furthermore, our study showed that CD8⁺ T cells reversed chemoresistance by increasing ENT1 expression, echoing the feed-forward loop between chemotherapy and immunotherapy. This work delineates the central role of ZEB1 in reprogramming TIME through epigenetic mechanisms, thereby identifying potential therapeutic targets for enhancing chemotherapy and immunotherapy in PC.

Epigenetic modification, such as acetylation, plays a critical role in driving treatment resistance (55, 56). HDAC1 is a cotranscriptional repressor of ZEB1. However, previous clinical trials evaluating HDAC inhibitors, including mocetinostat — either alone or in combination with chemotherapy or immune checkpoint inhibitors — have failed to demonstrate obvious efficacy in PC and were frequently associated with dose-limiting toxicities, particularly at standard or high doses (57–59). In mouse models, we found that HDAC inhibitor led to a better treatment response when combined with chemotherapy and immunotherapy. We proposed a potential treatment strategy by combining HDAC inhibitor with chemotherapy and immunotherapy, which achieved promising efficacy in PC. Chemokines are critical in regulating immune evasion by facilitating the communication between tumor cells and other cell types in the microenvironment (60–64). We found that CXCL16 recruited and activated CD8⁺ T cells, especially GZMA⁺CD8⁺ T cells, a subpopulation of CD8⁺ T cells that has potent cytotoxicity to tumor cells. GZMA⁺CD8⁺ T cells promote pyroptosis of tumor cells via GSDMD. This is consistent with previous reports showing that cytotoxic lymphocytes can induce pyroptosis in target cells (17). Epigenetic modifications regulate the efficacy of immunotherapy by remodeling TIME (65). We delved into the role of epigenetic modification in regulating CXCL16 expression and delineated that the HDAC1-ZEB1 complex promotes deacetylation of Cxcl16, resulting in decreased transcription and expression of CXCL16. This evidence supports the rationale for combining HDAC inhibitors, chemotherapy, and immunotherapy for the treatment of PC.

Furthermore, we noticed decreased infiltration of neutrophils in the tumor microenvironment when HDAC1 or ZEB1 was blocked. Neutrophils induce cancer immune evasion by secreting immune-modulating cytokines, leading to the decreased treatment response to immunotherapy (66). Studies showed that suppressing infiltration of neutrophils and myeloid-derived suppressor cells (MDSCs) by inhibiting CXCR2 resulted in the synergistic antitumor immunity when combined with immunotherapy (61, 67). Intriguingly, senescence-like neutrophils are more potent in driving immunosuppression than their canonical counterparts (68). Moreover, HDAC inhibitor showed a synergistic antitumor effect with CXCR2 inhibitor by eliminating infiltration of senescence-like neutrophils in prostate cancer (68). Future studies may evaluate whether blocking HDAC and neutrophils would increase sensitivity to chemotherapy and immunotherapy in PC.

This study also has limitations. The mechanism by which ZEB1 mediated upregulation of SPP1 remains unclear. Additionally, the potential of other HDAC inhibitors to enhance the efficacy of chemoimmunotherapy in PC warrants further investigation. Currently, no inhibitors are available that specifically target ZEB1. The development of a ZEB1-specific inhibitor would provide a valuable tool to further validate its role in mediating resistance to chemoimmunotherapy.

In conclusion, this study identified a ZEB1-driven reprogramming of the tumor microenvironment that contributed to resistance to both immunotherapy and chemotherapy in PC. While chemotherapy and immunotherapy primarily target tumor cells directly, HDAC inhibitors offer a promising synergistic strategy by modulating key components of TIME. These findings underscore the therapeutic potential of targeting epigenetic modifications, particularly histone acetylation, to overcome treatment resistance and improve outcomes in PC.

Sex as a biological variable. Our study examined male and female animals, and similar findings are reported for both sexes.

PDOX mouse model establishment. PDOs were inoculated onto the back of NSG mice to establish the PDOX F₁ generation, and subsequently, the F₁ tumors were chopped into small pieces and inoculated onto the back of nude mice to generate the F₂ generation. The F₃ generation was subsequently obtained with the same operation. The PDOX mouse model used in the experiments was the F₃ generation.

Construction of CAR T cells. Construction of EGFR-targeted CAR T cells was performed using a standard protocol. Initial validation of EGFR surface expression was conducted via flow cytometric analysis using APC-conjugated anti-EGFR monoclonal antibody (ABclonal, AB_3662630), demonstrating >90% positivity in both the AsPC-1 cell line and patient-derived organoid models. A second-generation CAR construct was engineered, comprising an anti-EGFR single-chain variable fragment derived from cetuximab; CD8α extracellular hinge and transmembrane domains; 4-1BB (CD137) costimulatory domain; CD3ζ signaling domain; and T2A-linked GFP reporter. The construct was cloned into pLVX-EF1α lentiviral vector and sequence-verified. Lentiviral particles were produced by triple transfection of 293T cells with the CAR transfer vector, psPAX2 packaging plasmid, and pMD2.G envelope plasmid, followed by ultracentrifugation to achieve final titers of 1 × 10⁸ transducing units/mL. For CAR T cell generation, CD8⁺ T lymphocytes were isolated from healthy donor PBMCs via negative selection (EasySep Human CD8⁺ T Cell Isolation Kit, STEMCELL Technologies) and activated with anti-CD3/CD28 Dynabeads at a 1:1 bead-to-cell ratio in the presence of 100 IU/mL recombinant human IL-2; and lentivirus was added to infect the CD8⁺ T cells. One week later, infection efficiency was assessed by measuring the percentage of cells exhibiting GFP using flow cytometry. The positive rate of EGFR–CAR T cells was 50%.

10x scRNA sequencing. According to the user’s manual (CG00315) for the 10x Genomics Chromium Next GEM Single Cell 3′ Kit v3.1 (catalog 1000268), the single-cell suspension was immediately loaded onto a chip to generate GEMs (Gel Bead-in-Emulsion) droplets using the 10x Chromium Controller. Reverse transcription, cDNA amplification, and DNA library construction were performed sequentially according to the protocol. The concentration and fragment size of the libraries were measured using Invitrogen Qubit 4.0 and Agilent 4150 TapeStation. High-throughput sequencing was conducted using high-throughput paired-end 150 bp (PE-150) mode. This work is assisted by OE Biotech Co.

ATAC-Seq, CUT&Tag-Seq, RNA-Seq, and joint analysis. ATAC-Seq: AsPC-1 and AsPC GEM cells were collected for preparation of cell suspensions and to obtain cell nuclei. Then, Tn5 transposase was added to cleave DNA into fragments. PCR amplification of DNA fragments and sequencing were performed on the illumina NovaSeq platform. CUT&Tag-Seq: CUT&Tag was performed in AsPC-1 and AsPC-GEM cells by using anti-H3K27ac (Cell Signaling Technology 8173) anti-H3K4me (Cell Signaling Technology 9751) antibodies. The experimental process is as described in Supplemental Methods. Finally, DNA libraries were sequenced on the illumina NovaSeq platform. RNA-Seq: Total RNA was extracted using the Trizol (Invitrogen) method, and RNA purity was detected using NanoDrop One (Thermo Fisher Scientific). Next, mRNA was enriched (T-oligo) and purified, and a library was constructed. Subsequently, sequencing was performed on the MGISEQ-2000 platform. Joint analysis of ATAC-Seq, CUT&Tag-Seq, and RNA-Seq data was performed by Frasergen Bioinformatics Co.

Human data. Deidentified clinical information was provided by Peking Union Medical College (PUMC) Hospital.

Statistics. Statistical analyses were performed using GraphPad Prism 9 and R 4.1.2. Data from all animal experiments are presented as mean ± SD and analyzed by 1-way ANOVA with Tukey’s multiple-comparison test or unpaired, 2-tailed Student’s t test. P value less than 0.05 was considered statistically significant. Animal survival analysis was analyzed by log-rank test. Further details about statistical analysis are indicated in Methods and the figure legends. Error bars indicate SEM or SD.

Study approval. This study was approved by the IRB at PUMC Hospital. Banked, deidentified tissues were used. Written consent was obtained from all participants. All animal experiments were approved by the IACUC at PUMC Hospital.

Data availability. Values for data points in the figures are available in the Supporting Data Values file. scRNA-Seq data generated in this study (GEO GSE281084) and spatial transcriptomics sequencing data (GEO GSE281083) are available in the Gene Expression Omnibus database. FASTQ files of the scRNA-Seq data from human PDACs were obtained from Peng et al. (69) (Genome Sequence Archive [GSA] CRA001160). The data on adjuvant gemcitabine chemotherapy in patients are available in The Cancer Genome Atlas database (TCGA-PAAD, file: “Clinical”). This study did not generate new unique codes. Any additional information required to reanalyze the data reported in this article will be provided upon reasonable request.