Oncogenic Kras targeting with MRTX1133 or Daraxonrasib specifically synergize with anti-CTLA4 to promote anti-tumor immunity in pancreatic cancer Authors: Krishnan K Mahadevan, Ana S Maldonado, Bingrui Li, Aaron A Bickert, Adrian Kacperczyk-Perdyan, Shreyasee V Kumbhar, Sujan Piya, Amari M Sockwell, Sami J Morse, Kent Arian, Hikaru Sugimoto, Shabnam Shalapour, David S Hong, Timothy P Heffernan, Anirban Maitra, Raghu Kalluri Published: 2026 ABSTRACT AbstractLack of sustained response to oncogenic Kras (Kras*) inhibition in pancreatic ductal adenocarcinoma (PDAC) underscores the need to identify effective combination therapies. Here, we demonstrate that Kras* targeting using MRTX1133 or Daraxonrasib recruits diverse T-cell infiltrates, including regulatory (Tregs), effector and exhausted T cells into the PDAC microenvironment. Kras* inhibition induces T-cell influx and offers a therapeutic window to specifically prime PDAC to anti-CTLA4 immune checkpoint blockade efficacy, in contrast to anti-PD1, anti-Tim3, anti-Lag3, anti-Vista, and anti-4-1BB agonist combination therapy. Mechanistically, anti-CTLA4 combination therapy transcriptionally reprograms effector Tregs to a naive phenotype, reverses CD8+ T-cell exhaustion, and promotes recruitment of functional tertiary lymphoid structures to mediate anti-tumor immunity. Single-cell ATAC sequencing reveals that Treg reprogramming by anti-CTLA4 is epigenetically regulated by downregulation of AP-1 family transcription factors in the IL-35 promoter region. This study reveals an actionable vulnerability in the adaptive immune response in Kras* targeted PDAC with immediate clinical implications.Subject terms: Cancer microenvironment, Pancreatic cancer, Pancreas Introduction Pharmacological targeting of oncogenic Kras (Kras*), a mutation that was considered ‘undruggable’ until recently represents a significant breakthrough in the treatment of several malignancies and has the potential to benefit numerous cancer patients. Strategies to inhibit oncogenic Kras include small molecule inhibitors of KrasG12D (MRTX1133, RMC-9805, ASP3082 and KRB-456)1–5 and KrasG12C mutations (Sotorasib/AMG-5106–8, Adagrasib/MRTX8499–11), pan-Ras inhibitors (RMC-6236, RMC-7977)12,13, pan Kras inhibitors (BI-2493, BI-2865)14, iExosomes inhibiting KrasG12D 15,16 and drugs inhibiting downstream targets of oncogenic Kras such as MEK and ERK17,18. Among the different oncogenic Kras mutations, G12C mutations are frequent in non-small cell lung cancers (NSCLCs) and colorectal cancers, whereas G12D mutations are frequent in PDAC and cholangiocarcinoma19–21. Other frequently mutated Kras genes include G12R and G12V mutations in PDAC22. KrasG12D represents the most frequently mutated gene in pancreatic ductal adenocarcinoma (PDAC)22,23. Targeting KrasG12D with small molecule inhibitors such as MRTX1133, BI-2493, RMC-6236 (Daraxonrasib), and iExosomes resulted in robust increase in survival in mouse models of PDAC1,3,7,10,14–16,24. However, prolonged treatment with small molecule inhibitors of Kras* eventually resulted in development of resistance to Kras* targeting and tumor relapse in preclinical models of PDAC18,24–27. Further, Sotorasib and Adagrasib (G12C inhibitors) have resulted in close to 30-50% response rates in non-small cell lung cancers (NSCLCs), pancreatic and biliary tract carcinomas but disease progression/relapse was observed in clinical trials11,28–30. Several strategies to address cancer cell intrinsic mechanisms of resistance to genetic or pharmacological Kras* suppression include targeting CDK4/6, YAP, CDK8, TEAD, MYC and SOS, in addition to targeting EMT, proteostasis networks and EGFR eventually lead to relapse despite longer survival outcomes in preclinical models of PDAC18,25–27,31–37. Therefore, identification and leveraging tumor microenvironment-dependent vulnerabilities resulting from Kras* targeting is critical to achieve meaningful treatment outcomes in PDAC patients. Prior studies with genetically engineered mouse models using an inducible Kras* allele demonstrated that Kras* is a dominant driver of T cell suppression in colorectal cancers and PDAC38–40. Kras* targeting across different cancer models led to a global increase in T cell infiltrates including CD4+ T cells, Tregs and CD8+ T cells in the tumor microenvironment (TME) indicating that the downstream effects of Kras* targeting could result in targetable vulnerabilities in the adaptive immune response38–46. Kras* targeting has shown divergent vulnerabilities in the immune response across cancer models in an organ specific context38,41,44,46. Therefore, we hypothesized that the immune microenvironment remodeling following Kras* targeting may offer new vulnerabilities critical to inform specific combination therapies in PDAC. In this work, we examine the adaptive immune response following targeting KrasG12D with MRTX1133 and pan-RAS inhibitor, Daraxonrasib (RMC6236) in PDAC. Our study reveals that distinct Kras* targeting strategies result in similar recruitment of T cell infiltrates. We identify a synergy between Kras* targeting and anti-CTLA4 therapy, driven by transcriptional regulation of Tregs and enhanced CD8+ T cell responses alongside the recruitment of tertiary lymphoid structures (TLS), which suppresses PDAC and prolongs survival. This study informs rational combination strategies for clinical testing. Results Pharmacological inhibition of Kras* results in parallel recruitment of effector and suppressor T cells in pre-clinical mouse models To assess alterations in the PDAC TME following Kras* targeting, we performed single cell RNA sequencing (scRNA seq) analysis in orthotopic KPC1 PDAC (primary PDAC line derived from 343P KPC- Pdx1-Cre; LSL-KrasG12D/+; Trp53R172H/+; Rosa-LSL-EYFP tumor bearing mice) treated with KrasG12D specific small molecule inhibitor, MRTX1133 (Fig. 1A). MRTX1133 treatment was started once the KPC1 mice developed palpable, large PDAC (Fig. 1A). UMAP analysis of the scRNA sequence (scRNA seq) captured fibroblasts, cancer/ductal cells, acinar cells, T cells, innate lymphoid cells (ILCs), endothelial cells, dendritic cells (DCs), myeloid cells, B-/plasma cells and granulocytes in the PDAC tissue (Supplementary Fig. S1A–C). MRTX1133 treatment led to a significant increase in the proportion of T and B cells, accompanied by a decrease in myeloid cells, cancer/ductal cells and fibroblasts40,41 (Fig. 1B–E, Supplementary Fig. S1A, B). Kras* targeting led to a global increase in T cell infiltrates in PDAC. A Schematic representation of orthotopic injection and MRTX1133 treatment of KPC1 mice. B UMAP of all cell types with T cell subsets in Vehicle/MRTX1133 treated KPC1 PDAC determined by scRNA-seq analysis (n = 3 tumors combined group). C Dot plots of genes used to define T cell subtypes in (B). D, E Relative abundance of all cell types (D) and T cells as a percentage of CD45+ cells (E) in KPC1 PDAC determined by scRNA seq. F, G Violin plots for B2m, Muc1 (F) and Ido1 (G) in ductal/cancer cells and fibroblasts of KPC1 mice determined by scRNA seq. H Schematic representation of orthotopic injection and Daraxonrasib treatment of KPC2 mice. I UMAP of all cell types with T cell subsets in Veh/Iso (n = 3 tumors combined/group), Daraxonrasib (n = 4 tumors combined/group) treated KPC2 PDAC determined by scRNA seq analysis. J Relative abundance of T cells as a percentage of CD45+ cells in KPC2 PDAC determined by scRNA seq. Data are presented as mean in (D), (E) and (J) and as violin plots in (F) and (G). Significance was determined by two-sided Wilcoxon rank-sum test in (F) and (G). See related Supplementary Fig. S1. Analysis of tumor infiltrating T cells revealed a global increase in the number of CD4+, CD8+ T cells, NK T cells and innate lymphoid cells (Fig. 1B–D, Supplementary Fig. S1D). Further sub-clustering of the T cells demonstrated an increase in number and frequencies of CD8 effector cells (Prf1, Tbx21, Ccl5, Gzmb high and Pdcd1, Lag3 low), exhausted CD8 cells (Pdcd1, Lag3 high), Tregs (CD4 cells with Ikzf2, Foxp3 high and Tbx21 low), CD4 naive (Lef1, Tcf7, and Sell) and memory cells (Cd28 and Cd40lg) (Fig. 1B–E, Supplementary Fig. S1D, E). Exhausted CD8+ T cells can retain cytotoxic programs47, but demonstrate higher expression of exhaustion/inhibitory markers such as Pdcd1, Lag3. The CD8 effector cluster shows very low expression of these exhaustion markers in addition to an effector/activation profile (e.g., Tbx21, Nkg7, and Prf1). Analysis of cancer/ductal cells and fibroblast populations demonstrated an increase in interferon responsive gene, B2m (Fig. 1F), a subunit of MHC-1 which aids in the recruitment of CD8+ T cell response in PDAC and colorectal cancers (CRCs)38,48. Furthermore, Kras* targeting resulted in upregulation of Muc1 and its downstream mediator Ido1 in the cancer cells and fibroblasts which are primary meditators of Treg recruitment (Fig. 1G)49. Consistent with an increase in Treg recruitment following MRTX1133 treatment, we observed suppression of Il6 and Tnf in the cancer cells and fibroblasts (Supplementary Fig. S1F). Tregs can indirectly suppress the secretion of pro-inflammatory cytokines, including TNF-α, IL-6, and IL-1β, through IL-10 mediated mechanisms50–52. Specifically, Treg depletion has been shown to increase TNF-α levels in fibroblasts53. Further, analyses of the cytokines involved in Treg recruitment demonstrated an increase in Nfkb1 (NF-κB) and Cxcr4/Ackr3 (Cxcr7) in cancer cells and fibroblasts following Kras* targeting (Supplementary Fig. S1G)54. Next, we use an orthotopic KPC2 (primary PDAC line derived from HY19636-PKC- P48-Cre, LSL-KrasG12D/+, Trp53L/+ mice) tumor-bearing mice treated with the pan-RAS inhibitor, Daraxonrasib to determine impact on T cell infiltrates (Fig. 1H). We sorted for CD45+ lymphocytes to enrich for the immune population in the TME. UMAP analysis of the scRNA seq identified T cells, B cells, granulocytes, mast cells and myeloid cells (Fig. 1I). Analysis of the tumor infiltrating T cells revealed an increase in the frequencies of CD4 memory, CD4 naive cells, Tregs, effector and exhausted CD8+ T cells (Fig. 1I, J), as observed with MRTX1133 (vide supra). Kras* targeting synergizes with anti-CTLA4 but not anti-PD1 therapy The recruitment of distinct T cell subtypes following Kras* inhibition presented an opportunity to target therapeutic vulnerabilities in the adaptive immune responses. Although checkpoint immunotherapy confers survival benefit in cancers such as melanoma, lymphomas and subsets of lung cancers, PDAC remains refractory to T cell targeted therapies in preclinical models and clinical trials55–58. First, we assessed the efficacy of anti-CTLA4 and anti-PD1 treatment in combination with MRTX1133 and Daraxonrasib in KPC2 tumor-bearing mice (Fig. 2A). Although EYFP is optimized for lower cytotoxicity and reduced oxidative stress compared to traditional GFP59, it remains a non-mammalian protein that can provide exogenous MHC-I epitopes. To address potential immunogenicity, we use the KPC2 mouse model for subsequent experiments to avoid the fluorescent reporter confounding therapeutic immune responses (Supplementary Fig. S2A, B)60. MRTX1133 and Daraxonrasib significantly prolonged survival and inhibited PDAC progression (Fig. 2B–K). Further, analysis for CK19 (a marker of ductal/cancer cells to assess tumor burden) confirmed significant tumor inhibition upon MRTX-1133 and Daraxonrasib treatments (Supplementary Fig. S3A–I). Analysis of pERK showed downregulation of pERK demonstrating similar levels of target engagement with MRTX1133 and Daraxonrasib (Supplementary Fig. S3A–F). Despite initial suppression of PDAC growth and prolonged survival, MRTX1133 and Daraxonrasib treated mice eventually succumbed to PDAC, suggesting resistance to Kras* targeting (Fig. 2B, C, Supplementary Fig. S3G–J). Addition of anti-CTLA4 immunotherapy to MRTX1133 and Daraxonrasib demonstrated prolonged tumor growth inhibition, survival and in the timeframe studied, the combination prevented the emergence of resistance to Kras* targeting (Fig. 2B–K, Supplementary Figs. S3A–J and S4A–C). Endpoint tumor-weight analysis demonstrated enhanced activity of the MRTX1133+anti-CTLA4 combination. Mean tumor weight was 1.222 g in the vehicle/isotype group, 1.021 g with anti-CTLA4 alone, 0.432 g with MRTX1133 alone, and 0.265 g with MRTX1133+anti-CTLA4 (Supplementary Fig. S4D). In a Bliss-type log-linear factorial model, the MRTX1133 × anti-CTLA4 interaction was significant, with an interaction ratio of 0.753 (95% CI, 0.583–0.974; P = 0.032), indicating that the observed combination tumor weight was approximately 25% lower than expected under independent treatment effects. Histologic quantification further supported a greater-than-additive effect of MRTX1133+anti-CTLA4. Mean PDAC burden was 90.2% in vehicle-treated mice, 94.6% with anti-CTLA4 alone, 54.8% with MRTX1133 alone, and 14.3% with MRTX1133+anti-CTLA4. Under an additive model, the expected PDAC burden for the combination was 59.2%, whereas the observed PDAC burden was 14.3%. The MRTX1133×anti-CTLA4 interaction was significant, corresponding to an additional 44.9 percentage-point reduction in PDAC burden beyond additivity (95% CI, 29.1–60.7 percentage points; P = 1.4 × 10−5) (Supplementary Fig. S4D). Kras* targeting increased T cell infiltrates and synergized with anti-CTLA4 therapy in PDAC. A Schematic representation of orthotopic injection and Daraxonrasib, MRTX1133, αCTLA4, and αPD1 treatments of KPC2 mice. B, C Kaplan–Meier survival curves of Veh/Iso (n = 17), αCTLA4 (n = 8), MRTX1133 (n = 11), MRTX1133 + αCTLA4 (n = 11), MRTX1133 + αPD1 (n = 9) treated KPC2 mice (B), Veh/Iso (n = 6), Daraxonrasib (n = 8), αCTLA4 (n = 7), Daraxonrasib + αCTLA4 (n = 10), Daraxonrasib + αPD1 (n = 7) treated KPC2 mice (C). D, E Representative H&E (top) and CD8-CD19-DAPI immunofluorescence images (bottom) from MRTX1133 treated age-matched KPC2 mice with αCTLA4/αPD1 (D), with quantification of tissue histology (E). Veh/Iso (n = 3), MRTX1133 (n = 5), MRTX1133 + αCTLA4 (n =6 ), MRTX1133 + αPD1 (n = 7). F Pancreas/tumor weights of age-matched KPC2 mice treated with Veh/Iso (n = 11), MRTX1133 (n = 13), MRTX1133 + αCTLA4 (n = 15), MRTX1133 + αPD1 (n = 15). G Quantification of CD8+, CD19+ cells from D (Veh/Iso (n = 4), MRTX1133 (n = 5), MRTX1133 + αCTLA4 (n = 5), MRTX1133 + αPD1 (n = 6)). H, I Representative H&E (top) and CD8-CD19-DAPI immunofluorescence images (bottom) from Daraxonrasib treated age-matched KPC2 mice with αCTLA4/αPD1 (H), with quantification of tissue histology (I) (n = 3/group). J Pancreas/tumor weights of age-matched KPC2 mice treated with Veh/Iso (n = 8), Daraxonrasib (n = 7), Daraxonrasib + αCTLA4 (n = 9), Daraxonrasib + αPD1 (n = 6). K Quantification of CD8+, CD19+ cells from (H) (Veh/Iso (n = 3), Daraxonrasib (n = 3), Daraxonrasib + αCTLA4 (n = 4), Daraxonrasib + αPD1 (n = 3)). L Schematic representation of MRTX1133, and αCTLA4 treatments of autochthonous KPPC mice. M–O Representative H&E (top) and CD8-CD19-DAPI immunofluorescence images (bottom) from MRTX1133 treated age-matched KPPC mice with αCTLA4 (M), quantification of tissue histology (N), and CD8+, CD19+ cells (O) from (M) (KPPC Veh/Iso: n = 3, MRTX1133: n = 2, MRTX1133 + αCTLA4: n = 3). Note: Same αCTLA4 mice are used in (B) and (C). One Veh/Iso mice in (M) was used earlier in another study41. The Veh/Iso controls that reached tumor-related moribundity endpoints in Figs. 4 and 5B are included in Fig. 2B. Data are presented as mean ± SD in (E), (F), (G), (I), (J), (K), (N), and (O). Log-rank test was used in (B) and (C), two-way ANOVA in (E), (I) and (N), Kruskal–Wallis with Dunn’s multiple comparisons test in (F) and CD19 comparisons in (G), one-way ANOVA with Dunnett’s multiple comparisons test in (J), (K), (O) and CD8 comparisons in (G). Scale bars 100 μm. ns not significant. See related Supplementary Figs. S2–S6. In contrast, addition of anti-PD1 with Kras* targeting did not result in an improvement when compared to Kras* targeting alone (Fig. 2B–K, Supplementary Fig. S3A–J, Supplementary Fig. S4A–C). Analysis of the immune infiltrates demonstrated an increase in CD8+ T cells in MRTX1133 or Daraxonrasib treated tumors compared to controls. Addition of anti-CTLA4 further expanded CD8+ and CD19+ cells in the pancreatic TME likely resulting in sustained PDAC inhibition response following Kras* targeting (Fig. 2D–K). Further, Kras* inhibition with either MRTX1133 or Daraxonrasib leads to a significant depletion of myeloid cells in our orthotopic PDAC models. Interestingly, we observed no further additive reduction in myeloid proportions when combining these inhibitors with anti-CTLA4 or anti-PD1 (Supplementary Fig. 5A–D). Body weight comparisons as a surrogate for systemic toxicity did not demonstrate significant differences between mice that received various small molecule inhibitor treatments (Supplementary Fig. S5E, F). We next probed the efficacy of Kras* inhibitors and anti-CTLA4 combination therapy in eliciting robust anti-tumor responses in autochthonous genetically engineered mouse models (GEMs) (Fig. 2L). To assess the stromal component in the orthotopic models, we performed Collagen1 and Podoplanin IHC in the genetically engineered mouse models (GEMMs) (KPPC and KPC GEMMs) and orthotopic models (KPC2 and KPC3) of PDAC. The orthotopic models did demonstrate a substantial stromal component albeit at a lower level compared to the GEMMs (Supplementary Fig. S6A, B). We treated autochthonous KPPC (P48-Cre, LSL-KrasG12D/+, Trp53L/L) mice with MRTX1133 and anti-CTLA4 therapy following establishment of PDAC starting at 6 weeks of age (Fig. 2L–N). Mice were sacrificed for age matched analysis ~3 weeks after indicated treatment when the vehicle treated mice reached moribundancy (Fig. 2L–N). MRTX1133 treated KPPC mice demonstrated significant suppression of PDAC compared to the vehicle treated controls and addition of anti-CTLA4 further resulted in suppression of PDAC with an increase in the proportion of uninvolved pancreatic tissue (Fig. 2M, N). Analysis of tumor infiltrating lymphocytes demonstrated an increase in CD8+ T cells and an increasing trend in CD19+ B cells with MRTX1133 and anti-CTLA4 combination therapy compared to MRTX1133 monotherapy (Fig. 2M–O). Next, we used autochthonous KPC (Pdx1-Cre, LSL-KrasG12D/+, Trp53R172H/+) GEM mice to determine if the combination therapy demonstrates synergy in KPC mice (Supplementary Fig. S6C). All vehicle (4 out of 4) treated KPC mice demonstrated invasive PDAC 10 weeks post-treatment (Supplementary Fig. S6D, E). Age matched analysis of mice treated with MRTX1133 demonstrated suppression in 3 out of 4 tumors and 1 mouse demonstrated resistance to Kras* inhibition (Supplementary Fig. S6D, E). Mice that received combination therapy demonstrated minimal PDAC and suppression in 3 out of 3 mice (Supplementary Fig. S6D, E). To further validate our findings, we used a different KPC3 orthotopic tumor model (primary PDAC line derived from KPC T- Pdx1-Cre, LSL-KrasG12D/+, Trp53R172H/+ derived cell line) for long-term follow-up after treatment with MRTX1133 and anti-CTLA4 (Supplementary Fig. S6F, G). Anti-CTLA4 as a monotherapy did not increase survival of PDAC tumor-bearing mice, as shown before57,whereas the combination of MRTX1133 and anti-CTLA4 demonstrated sustained tumor suppression in majority of the PDAC mice (Supplementary Fig. S6F, G). Our results indicate that MRTX1133 and Daraxonrasib demonstrated therapeutic synergy with anti-CTLA4 irrespective of the Kras* targeting strategy. Kras* targeting in combination with anti-CTLA4 drives Tregs towards a naive state, attenuates CD8 exhaustion and increases effector CD8+ T cells Single cell RNA seq. analysis of the T cells in the KPC3 mice treated with MRTX1133 and anti-CTLA4 showed a decrease in frequencies and number of exhausted CD8+ T cells, expansion of memory/stem-like, effector CD8+ T cells, naive and memory CD4+ T cells (Fig. 3A, B, Supplementary Fig. S7A–C). Analysis of scRNA seq data in a second, KPC2 model treated with Daraxonrasib in combination with anti-CTLA4 therapy validated the increase in total T cell infiltrates, expansion of memory/stem-like and effector CD8+ T cells, naive and memory CD4+ T cells (Fig. 3C, D, Supplementary Fig. S7D–F). Notably, the anti-CTLA4 combination with Kras* targeting did not decrease total Treg numbers (Fig. 3A–D). We used the 9H10 clone of the anti-CTLA4 antibody that blocks CTLA4 binding to the B7 co-receptor in this study (similar to mechanisms of Ipilimumab and Tremelimumab that do not deplete Tregs61). Rather, unsupervised analysis of differentially regulated genes (DRGs)62,63 in the tumors that received MRTX1133 and anti-CTLA4 combination therapy revealed a distinct transcriptional profile in Tregs compared to those in anti-CTLA4 or MRTX1133 treated tumors (Fig. 3E, F). The Tregs from MRTX1133 and anti-CTLA4 treated tumors demonstrated downregulation of AP-1-related genes including Rgs2, Fos, Fosb, Junb, and Jund (Fig. 3E, F, Supplementary Fig. S8A)64–66. Among these, Rgs2, Fos, Fosb have been shown to be involved in differentiation of Tregs. Junb promoted effector T reg development through facilitation of IRF4 and Jund protected Tregs from apoptosis64–66. Further, we observed a downregulation of Dyrk1a, Dusp4, Samsn1, and Ccr5 in the Tregs of mice that received MRTX1133 and anti-CTLA4 combination therapy (Fig. 3E, F, Supplementary Fig. S8A). Dyrk1a expression has been shown to drive CD4+ T cells into Tregs differentiation lineages64–66, Dusp4 is known to stabilize Foxp3 expression in Tregs67, Samsn1 and Ccr5 enabled migration and homing68 (Fig. 3E, F, Supplementary Fig. S8A). Taken together, our analysis indicates that genes involved in Treg differentiation, effector function, homing and migration were all downregulated in the tumor infiltrating Tregs of mice that received MRTX1133 and anti-CTLA4 combination therapy. Analysis of the Tregs in Daraxonrasib and anti-CTLA4 treated tumors also revealed downregulation several of the genes that promote effector T reg phenotype such as Rgs2, Fos, Fosb, Junb, Dusp4, and Ccr5 validating the results with MRTX1133 and anti-CTLA4 treated tumors (Fig. 3G, Supplementary Fig. S8A–C). Further, the Tregs from MRTX1133/Daraxonrasib and anti-CTLA4 treated tumors demonstrated upregulation of Ifi213 which promotes IFN responsive genes (Fig. 3E–G, Supplementary Fig. S8C). Next, we analyzed expression of effector Treg cytokines, IL-10, TGF-β1, and IL-35 in the Tregs of mice that received Kras* targeting and immunotherapy69–71. We identified a decrease in Il10, Tgfb1, and the IL-35 subunit Ebi3 expression in the Tregs of mice treated with MRTX1133 or Daraxonrasib in combination with anti-CTLA4 indicating suppression of the effector Treg function (Fig. 3H, I). Immunostaining analysis for IL-10 and Tgf-β validated that Kras* inhibition increased the IL-10+ infiltrates. Further, addition of anti-CTLA4 reversed the IL-10+ cell infiltration, whereas the addition of anti-PD1 had no impact compared to monotherapy (Supplementary Fig. S9A–D). Similarly, analysis of Tgf-β-positive cells revealed an increase in expression following Kras* inhibition and further suppression was observed with addition of anti-CTLA4 therapy (Supplementary Fig. S9E–H). Assessment of morphology of the IL-10 and Tgf-β+ infiltrates revealed that IL-10+ cells were likely T cells, whereas Tgf-β was expressed by diverse cell populations in the PDAC TME. Overall, these findings further validate the suppression of Il10 and Tgfb observed at a transcriptional level in combination with anti-CTLA4 therapy. Collectively, our data indicates that Kras* targeting in combination with anti-CTLA4 therapy drives Tregs to a naive state and inhibits effector T reg pathways critical for suppression of CD8+ T cells. scRNA seq analysis of T cell infiltrates in MRTX1133, Daraxonrasib and checkpoint immunotherapy treated PDAC. A, B UMAP of all cells (left) and T cells in αCTLA4 (n = 2 tumors combined/group), MRTX1133 (n = 3 tumors combined/group), MRTX1133 + αCTLA4 (n = 3 combined/group) treated (right) KPC3 PDAC (A), with relative frequencies of T cell subsets (B). C, D UMAP of all cells (left) and T cells in Veh/Iso (n = 3 tumors combined/group), Daraxonrasib (n = 4 tumors combined/group), Daraxonrasib + αCTLA4 (n = 5 tumors combined/group), Daraxonrasib + αPD1 (n = 3 tumors combined/group) treated (right) KPC2 PDAC (C), with relative frequencies of T cell subsets (D). E, F Heatmap of differentially regulated genes in the tumor-infiltrating Tregs of αCTLA4, MRTX1133, MRTX1133 + αCTLA4 treated KPC3 PDAC (Red arrow—upregulated genes, blue arrow—downregulated genes in the MRTX1133 + αCTLA4 group compared to other groups) (E), with summary table of gene functions (F). G Violin plots for Rgs2, Fos, Fosb, Junb, Dyrk1a, Dusp4, Jund, Ccr5, and Samsn1 in tumor-infiltrating Tregs of Daraxonrasib treated KPC2 mice with αCTLA4/αPD1 (P values: 1: 0.029, 2: p < 2.22E-16, 3: 0.036, 4: 0.00058, 5: 7.2E-06, 6: 0.0044, 7: 1.1E-09, 8: 0.027, 9: 4.5E-12, 10: 0.032). H, I Violin plots for Il2, Il10, Tgfb1 and Ebi3 in tumor-infiltrating Tregs of MRTX1133 (H), Daraxonrasib (I) treated mice with αCTLA4/αPD1 (P values: 1: 1.3E-07, 2: 8.3E-16, 3: <2.22E-16, 4: 5.8E-05, 5: 2.6E-16, 6: 0.0057, 7: 1.9E-13, 8: 0.033, 9: 0.0022, 10: 4.8E-15). J Violin plots for immune checkpoint genes Adora2a, Adora2b, Vtcn1, Btla, Ctla4, Ido1, Lag3, Cybb, Pdcd1 Havcr2, Vsir, and Tnfrsf9 in tumor-infiltrating T, NKT and γδ T cells of KPC3 (MRTX1133 experiments) and KPC2 (Daraxonrasib experiments) PDAC. K Schematic representation of orthotopic injection and MRTX1133, αLag3, αTim3, αVista and α41-BB agonist antibody treatments of KPC2 mice. L Kaplan–Meier survival curves of Veh/Iso (n = 17), MRTX1133 (n = 7), MRTX1133 + αLag3 (n = 5), MRTX1133 + αTim3 (n = 5), MRTX1133 + αVista (n = 5), MRTX1133 + α41-BB agonist (n = 5) treated KPC2 mice. Note: The Veh/Iso and MRTX1133 treated mice are repeated from Fig. 2B. Data are presented as mean in (B) and (D), and as violin plots in (G), (H), (I), and (J). Significance was determined by log-rank test in (L), and by two-sided Wilcoxon rank-sum test in (G), (H), and (I). ns not significant. See related Supplementary Figs. S6–S11. Next, we analyzed the transcriptional profiles of the CD8+ T cell infiltrates in PDACs of mice treated with Kras* targeting in combination with anti-CTLA4 therapy. Analysis of the DRGs identified downregulation of H3f3b, Nr4a3 and Nr4a2 in the CD8+ T cells (Supplementary Fig. S10A). H3f3b has been shown to promote deletional tolerance of CD8+ T cells72. Nr4a3 has been shown to limit the generation of memory response73 and Nr4a2 limits stem-like phenotype of the CD8+ T cells74, all indicating that the combination therapy reversed tolerance in the CD8s. In addition, we found upregulation of 4-1BB (Tnfrsf9) following MRTX1133 treatment and suppression following addition of anti-CTLA4 therapy (Supplementary Fig. S10A), suggesting that it could be a potential target for immunotherapy. Consistent with the effector CD8 phenotype, we identified upregulation of markers of activated state including Sp100, Prkcq, Cyclin D3 (Ccnd3)75–77 and interferon responsive genes Mndal, Ifi209, Ifi206, Ifi203, and Ifi2712a78 (Supplementary Fig. S10A) in the MRTX1133 and anti-CTLA4 treated CD8+ T cells. Analysis of the CD8+ T cells infiltrates with the Daraxonrasib treated tumors further validated suppression of H3f3b, Nr4a3, Nr4a2 and upregulation of interferon responsive genes in combination with anti-CTLA4 treatment (Supplementary Fig. S10B). Further, we evaluated tumor-specific T cell markers (Cxcl13, Tnfrsf9, Xcl1) alongside key cytolytic and effector functional markers (Gzmb, Prf1, and Ifng) (Supplementary Fig. S10C, D) on the CD8+ effector and memory/stem-like T cells. Our data demonstrates an increase in the tumor-specific and cytolytic/effector markers in the intra-tumoral CD8s of mice that received the combination of MRTX1133 + αCTLA4 and Daraxonrasib + αCTLA4 compared to their respective controls (Supplementary Fig. S10C, D). Overall, Kras* targeting in combination with anti-CTLA4 decreased negative regulators of CD8+ T cells and promoted interferon responsive genes indicating an activated CD8+ T cell state. Next, we probed the reasons for lack of response to anti-PD1 therapy in combination with Kras* targeting despite the recruitment of exhausted CD8+ T cells. Analysis of T cell populations from the MRTX1133 and Daraxonrasib experiments indicated high levels of expression of multiple immune checkpoint genes including Ctla4, Lag3, Tim-3 (Havcr2), Vista (Vsir), and Tnfsf9 (4-1BB) in addition to PD1 (Pdcd1) in the exhausted CD8+ T cells (Fig. 3J). We hypothesized that targeting the other immune checkpoints may reverse the exhaustion of the tumor infiltrating CD8+ T cells following Kras* targeting. Therefore, we individually treated KPC2 PDAC bearing mice with anti-Lag3, anti-Tim3, anti-Vista and anti-4-1BB (agonist antibody) in combination with MRTX1133 (Fig. 3K). MRTX1133 prolonged the survival of the KPC2 tumor-bearing mice, but the combination of the above-mentioned checkpoint immunotherapies did not result in additional survival benefit or superior tumor inhibition in KPC2 mice (Fig. 3K, L, Supplementary Fig. S11A–C). Failure of other checkpoint inhibitors to reverse the exhaustion of CD8+ T cells indicates that these cells are likely terminally exhausted and other methods of priming are required. Our results indicate that anti-CTLA4 specifically synergizes with Kras* targeting therapies with expansion of memory/stem-like and effector CD8+ T cells, whereas addition of anti-PD1, anti-Tim3, anti-Lag3, anti-Vista and anti-4-1BB (agonist antibody) failed to reverse CD8 exhaustion in mice with PDAC. Kras* targeting in combination with anti-CTLA4 drives epigenetic reprogramming of Tregs We next analyze the chromatin regulatory elements of sorted CD45+ immune cells from KPC2 PDAC mice treated with MRTX1133 in combination with anti-CTLA4 or anti-PD1 (Fig. 4A) by single cell ATAC sequencing (scATAC seq). scATAC seq of tumor infiltrating CD45+ cells identified 9 immune cell clusters including CD4+ T cells, Treg, CD8+ T cells, DCs, 3 myeloid clusters including two macrophage populations, and B cells in PDAC (Fig. 4A). Further analysis of the T and NK cells identified 9 T and NK cell subclusters including NK cells, CD8+ effector T cells, memory/stem-like CD8+ T cells, CD8+ exhausted T cells, CD8+ tissue-resident memory T cells (CD8 TRM), Tregs, Th2 cells, naive CD8+ and CD4+ T cells (Fig. 4A, Supplementary Fig. S12A). Consistent with our results from scRNA seq experiments, Kras* targeting increased T cell infiltrates and the anti-CTLA4 combination further increased frequencies of CD8 effectors, memory/stem-like CD8+ T cells, naive CD8+ and CD4+ T cells with a decrease in CD8+ exhausted T cell frequencies (Fig. 4A, Supplementary Fig. S12A). The scATAC seq method enabled identification of transcription factor (TF) activity and their fate/cell-specific binding programs in the Tregs. The suppression of cytokines IL-10, IL-35, and TGF-β1 is critical for limiting CD8 cytotoxicity in the tumor by Tregs, as observed in mice treated with Kras* inhibitor and anti-CTLA4 (Fig. 3E–I). We further evaluated TF motifs that specifically bind to the promoter/enhancer regions to regulate the effector Treg response (Fig. 4B, Supplementary Fig. S12B). Analysis of the IL-10, IL-35, and Tgf-β genes revealed significant suppression of the peaks in the Ebi3 (IL-35 component) gene promoter region and in the enhancer region of Il10 (Fig. 4B, Supplementary Fig. S12C). Next, we probed for TFs that can bind to the motifs in the promoter and enhancer regions of IL-10 and IL-35 of Tregs. Comparison of motif enrichment of the TFs between different treatment groups in the Tregs revealed significant downregulation of Bach2, Nfatc1, Nr4a2, Maf and AP-1 related TFs, Fos, Fosb, Junb, Jund, in the Kras* inhibitor and anti-CTLA4 arm (Fig. 4C, Supplementary Fig. S12D). The AP-1 TFs specifically bind to the promoter region of IL-35 gene to promote effector Treg function79. Footprint analysis of Fos, Fob, Junb and Jund binding chromatin regions revealed preferential binding of the chromatin regions in Tregs all groups compared to the MRTX1133 and anti-CTLA4 treated mice (Fig. 4D). Treg motifs, including Foxp, AP-1, Runx, Maf, Nr4a, and Eomes-associated motifs were significantly enriched in the Veh, MRTX1133, or MRTX1133 + αPD1 groups but were not significantly enriched in the MRTX1133 + αCTLA4 group (Supplementary Fig. S12E). Motif enrichment and footprint analysis in the Tregs revealed downregulation of the TFs at the promoter/enhancer regions of IL-10 and IL-35 genes involved in effector Treg differentiation in mice treated with MRTX1133 and anti-CTLA4 immunotherapy in PDAC. Taken together, our findings are consistent with earlier studies that demonstrate the role of IL-10 in limiting effector cytokine production and IL-35 in regulating inhibitory receptors on memory T cells. Both the cytokines cooperatively drive CD8 exhaustion through upregulation of BLIMP1 and other checkpoint receptors71,79. The combination of Kras* inhibition and anti-CTLA4 epigenetically reprograms Tregs to a naive state, attenuating the effects of IL-10 and IL-35 which releases the breaks on the effector CD8 response. scATAC seq. analysis of tumor infiltrating Tregs in MRTX1133 and checkpoint immunotherapy treated PDAC. A UMAP of all CD45+ cell types (top left) and T cell subsets in Veh/Iso (n = 3 tumors combined group), MRTX1133 (n = 4 tumors combined/group), MRTX1133 + αCTLA4 (n = 5 tumors combined/group), MRTX1133 + αPD1 (n = 3 tumors combined/group) (top right, bottom) treated KPC2 PDAC. B Genome browser tracks of Tregs around the Ebi3 gene locus in Veh/Iso, MRTX1133, MRTX1133 + αCTLA4, MRTX1133 + αPD1 treated KPC2 PDAC. C Heatmap representation of ATAC-seq motif enrichment of TFs across Tregs of Veh/Iso, MRTX1133, MRTX1133 + αCTLA4, MRTX1133 + αPD1 treated KPC2 PDAC. D ATAC seq. motif footprint aggregate plot around the predicted Fos, Fosb, Jund, and Junb motifs in Tregs of Veh/Iso, MRTX1133, MRTX1133 + αCTLA4, MRTX1133 + αPD1 treated KPC2 PDAC. Data represents average footprint profile for each experimental group in (D). See related Supplementary Fig. S12. Depletion of CD4+ T cells or CD25+ Tregs synergize with Kras* targeting in PDAC The transcriptional changes observed in tumor-infiltrating Tregs, CD4⁺, and CD8⁺ T cells from mice treated with Kras* targeting combined with anti-CTLA-4 therapy prompted us to investigate the functional roles of these T cell subsets in the immune regulation of PDAC (Figs. 3 and 4). Anti-CTLA4 combination with MRTX1133 and Daraxonrasib did not result in a significant Th1 response (Fig. 3A–D) (as marked by expansion of CD4+ Tbet+ cells in response to αCTLA4 therapies47) despite robust PDAC inhibition and effector CD8 response. To determine the functional contribution of the Tregs, CD4+ T cells, and to assess the contribution of ‘CD4 help’ in generating effector CD8 responses, we treated the KPC2 tumor bearing mice with anti-CD25 and anti-CD4 depletion antibodies in combination with MRTX1133 therapy (Fig. 5A). Analysis of peripheral blood confirmed systemic depletion of CD4+ T cells and CD25+ T cells in the anti-CD4 and anti-CD25 antibodies treated mice (Supplementary Fig. S13A, B). Although depletion of CD4+ T cells did not result in significant changes in other T cell populations, depletion of CD25+ T cells resulted in an increased frequencies of CD4+ CD25- Foxp3+ population in the circulation (Supplementary Fig. S13A, B). MRTX1133 treated KPC2 mice succumbed to PDAC despite prolonged survival (Fig. 5B). Depletion of CD25+ cells in combination with MRTX1133 further improved the survival of the KPC2 mice (Fig. 5B). Complete depletion of the CD4+ T cells resulted in robust inhibition of PDAC in combination with MRTX1133 with similar survival kinetics as the MRTX1133 + anti-CTLA4 treated mice (Fig. 5B). Depletion of CD4+ T cells or Tregs synergize with Kras* targeting in PDAC. A Schematic representation of orthotopic injection and MRTX1133, αCD4 and αCD25 treatments of KPC2 mice. B Kaplan–Meier survival curves of Veh/Iso (n = 17), MRTX1133 (n = 11), MRTX1133 + αCD4 (n = 8), MRTX1133 + αCD25 (n = 10), MRTX1133+ αCTLA4 (n = 11), MRTX1133 + αPD1 (n = 9) treated KPC2 mice. Note: The Iso/Veh, MRTX1133, MRTX1133+ αCTLA4, and MRTX1133 + αPD1 curves are repeated from Fig. 2B to allow for direct comparisons. C–G Immunophenotyping analysis of KPC2 PDAC of indicated groups. CD3+(T cells); CD3+CD4+ (CD4+ T cells); CD3+CD8+ (CD8+ T cells) were measured as a percentage of CD45+ cells (C); CD4+Foxp3+ Tregs were measured as a percentage of CD3+ cells (D); CD4+T-bet+ (Th1) measured as a percentage of CD45+ cells (E), and CD19+ B cells measured as a percentage of CD45+ cells (F). CD8/Treg ratio and Th1/Treg ratio (G) (Veh/Iso (n = 8), MRTX1133 (n = 9), MRTX1133 + αCD4 (n = 4), MRTX1133+ αCTLA4 (n = 7), MRTX1133 + αPD1 (n = 5) in C, D, E, G), (Veh/Iso (n=5), MRTX1133 (n=4), MRTX1133+αCTLA4 (n=7), MRTX1133+αPD1 (n=5) in F). Data are presented as mean ± SD in (C–G). Significance was determined by log-rank test in (B), Kruskal–Wallis with Dunn’s multiple comparisons test in (E), one-way ANOVA with Dunnett’s multiple comparisons test in (C), (D), (F) and (G). ns not significant. See related Supplementary Fig. S13. Next, to evaluate mechanisms underpinning the alterations in tumor infiltrating T cells as a result of immune checkpoint blockade therapies/T cell depletions, we performed flow cytometry analysis in the age-matched KPC2 tumors treated with MRTX1133 in combination with anti-PD1, anti-CTLA4 and anti-CD4 depletion antibodies. Consistent with the findings from scRNA seq analysis, MRTX1133 treatment resulted in an increase in CD3+ T, CD3+ CD4+ T, CD3+ CD8+ T, and CD4+ Foxp3+ T cell frequencies in the PDAC TME (Fig. 5C, Supplementary Fig. S13C). The addition of anti-CTLA4 antibody to MRTX1133 treated KPC-2 tumors demonstrated further increase in total T cells, CD4+ T cells and CD8+ T cell frequencies (Fig. 5C). The anti-CD4 antibody treated mice in combination with MRTX1133 demonstrated depletion of intra-tumoral CD4+ T cells, Tregs and enhanced CD8+ T cell frequencies further validating that the CD4+ T cells are inhibitory to the CD8+ T cell function in PDAC (Fig. 5C, D). Analysis of the Th1 response in PDAC revealed that the Th1 cells were a rare population in the TILs (~1%), which increased following Kras* targeting with MRTX1133 (~ 2%). However, no further changes in the Th1 population were observed following addition of anti-CTLA4 therapy (Fig. 5E). Further analysis of the B cell frequencies demonstrated a significant increase in B cell frequencies in mice that received MRTX1133 and anti-CTLA4 therapy (Fig. 5F). In addition, analysis of Th1/Treg ratios did not demonstrate any differences between the treatment groups (Fig. 5G). Analysis of the CD8/Treg ratio demonstrated an increase with MRTX1133 therapy and further increase with addition of anti-CTLA4 therapy indicating a suppression of Tregs with enrichment of CD8 response (Fig. 5G). Collectively, robust tumor inhibition response and flow cytometry analysis following CD4+ T cell depletion indicates that CD4+ T cells are collectively inhibitory to CD8+ T cells and that the ‘CD4 help’ is dispensable in the immune regulation of PDAC. Kras* targeting in combination with αCTLA4 induces formation of tertiary lymphoid structures In-depth histological analysis of PDAC tissues from KPC2 mice demonstrated an increase in the presence of tertiary lymphoid structures (TLS) in mice treated with MRTX1133 or Daraxonrasib in combination with anti-CTLA4 (Fig. 6A, B, Supplementary Figs. S14A and S15A, B). Although occasional TLS were found in PDAC tissues of mice treated with MRTX1133/Daraxonrasib monotherapy or in combination with αPD1 antibody, the combination of Kras* targeting with anti-CTLA4 demonstrated consistent increase TLS with 100% penetrance in the KPC2 tumors (Fig. 6A, B, Supplementary Fig. S14A). Further, histological assessment of age-matched KPC and KPPC GEMs treated with MRTX1133 and anti-CTLA4 therapy revealed TLS with 100% penetrance in mice that received combination therapy (Fig. 6C–F, Supplementary Fig. S16A, B). TLS refers to the de novo occurrence of organized aggregates of immune cells in non-lymphoid organs80,81. In the context of cancer and immunotherapy, the presence of TLS is frequently associated with favorable responses to immunotherapy and better prognosis80,81. The composition of TLS may vary depending on the host tissue but are frequently characterized by an inner zone of B cells, surrounded by T cells and accompanied by the presence of follicular dendritic cells (FDCs) which enable formation of germinal center B cells, follicular T helper (TFH) cells and mature dendritic cells80–84. In addition, the TLS are accompanied by endothelial venules that provide the associated vasculature that likely mediate lymphocyte recruitment84. To confirm the histopathological identification of TLS, we performed immunostaining analysis for markers of T cells, B cells, endothelial vasculature, FDCs and TFH cells (Fig. 6G–L). Immunofluorescence staining with Tyramide signal amplification (TSA) multiplex analysis revealed presence of CD19+ B cells, CD8+ T cells, CD31+ endothelial cells in vascular structures and CD21+ and CD23+ cells (expressed by TFH, FDCs and B cell populations) confirming that these immune aggregates seen with Kras* targeting and αCTLA4 combination therapy are TLS at varying stages of maturation (Fig. 6G–L). We ascertain that each lymphoid structure that was identified as TLS demonstrated neo-vascularization with formation of endothelium identified by CD31 staining with co-existence of T and B cell regions in orthotopic KPC2 mice and KPC, KPPC GEMs that received combination therapy (Supplementary Figs. S17A, B and S18A, B). Kras* targeting in combination with αCTLA4 induces formation of tertiary lymphoid structures in PDAC. A, B Representative H&E sections of tertiary lymphoid structures (TLS) in MRTX1133, Daraxonrasib treated KPC2 mice with αCTLA4/αPD1 (A), with quantification of number of TLS (Left: Veh/Iso (n = 3), MRTX1133 (n = 5), MRTX1133+ αCTLA4 (n = 6), MRTX1133 + αPD1 (n = 7), Right: Veh/Iso (n = 5), Daraxonrasib (n = 3), Daraxonrasib + αCTLA4 (n = 4), Daraxonrasib + αPD1 (n = 3)) (B). C–F Representative H&E sections of TLS in MRTX1133, αCTLA4 treated mice. H&E of KPPC mice (C), and quantification (n = 3 tumors/group) (D). H&E of TLS in KPC mice with MRTX1133+ αCTLA4 (E), with quantification (Veh/Iso (n = 4), MRTX1133 (n = 4), MRTX1133+ αCTLA4 (n = 3)) (F). G, H Representative CD8-CD19-CD31-DAPI (left) and CD21-CD23-DAPI (right) immunofluorescence images of TLS (G), with quantification of MRTX1133 treated KPC2 mice with αCTLA4/αPD1 (H) (Left: Veh/Iso (n = 4), MRTX1133 (n = 5), MRTX1133+ αCTLA4 (n = 4), MRTX1133 + αPD1 (n = 6), Right: Veh/Iso (n = 4), MRTX1133 (n = 5), MRTX1133+ αCTLA4 (n = 6), MRTX1133 + αPD1 (n = 6)). I, J Representative CD8-CD19-CD31-DAPI immunofluorescence images of TLS (I), with quantification of Daraxonrasib treated KPC2 mice with αCTLA4/αPD1 (J) (Veh/Iso (n = 3), Daraxonrasib (n = 3), Daraxonrasib + αCTLA4 (n = 4), Daraxonrasib + αPD1 (n = 3)). K, L Representative CD8-CD19-CD31-DAPI (left) and CD21-CD23-DAPI (right) immunofluorescence images of TLS (K), with quantification of MRTX1133 treated KPC2 mice with αCTLA4/αPD1 (L) (Veh/Iso (n = 3), MRTX1133 (n = 2), MRTX1133 + αCTLA4 (n = 3)). M, N UMAP of all cells (left) and B cell subsets in αCTLA4, MRTX1133, MRTX1133 + αCTLA4 treated (right) KPC3 PDAC (M), with frequency of B cell subsets (N) (n = 3 mice/group). O, P UMAP of all cells (left) and B cell subsets in Veh/Iso, Daraxonrasib, Daraxonrasib + αCTLA4, Daraxonrasib + αPD1 treated (right) KPC2 PDAC (O), with frequency of B cell subsets (P) (n = 3 tumors/group). Note: B cells from UMAPs are analyzed in Fig. 6M and Fig. 6O and T cells from the scRNA seq data are analyzed in Fig. 3A and C, respectively. Data are presented as mean in (N) and (P), mean ± SD in (B), (D), (F), (H), (J) and (L). Significance was determined by Kruskal–Wallis with Dunn’s multiple comparisons test in (B), (D), (F), (H) and (L), one-way ANOVA with Sidak’s multiple comparisons test in (J). ns not significant. Scale bars 100 μm. See related Supplementary Figs. S14–S19. Next, we performed single-cell RNA sequencing (scRNA seq) analysis of the sorted live cells from KPC3 PDAC bearing mice treated with MRTX1133 and/or anti-CTLA4 antibody (Fig. 6M). UMAP analysis of the scRNA seq dataset captured 8 clusters including fibroblasts, cancer/ductal cells, acinar cells, myeloid cells, DCs, T/NKT cells, endothelial cells, and B/plasma cells (Fig. 6M). Further sub-clustering of the B and plasma cell populations revealed an increase in the frequencies and number of naive (follicular B cells) (expressing Ighd, Ms4a1, Pax5, Cd21, and Cd23), plasma cells (expressing Igkc, Xbp1, Jchain, Tnfrsf13b, Sdc1, Tnfrsf17, Cd320, Irf4, and Prdm1), germinal center B cells (Aicda, Gesam, S1pr2, Bcl6, and Fas) and IFN-stimulated/activated B cell populations (Ifit3, Slfn5, and Herc6) (Fig. 6M, N, Supplementary Fig. S19A–C). To validate our findings from the MRTX1133 and anti-CTLA4 combination therapy, we used a second model with the Daraxonrasib compound in combination with anti-CTLA4 and anti-PD1 therapy. We performed scRNA seq analysis on sorted live, CD45+ cells to enrich for the immune populations from PDAC of age-matched Daraxonrasib mice treated KPC2 mice in combination with anti-CTLA4 and anti-PD1 therapy (Fig. 6O, P). UMAP analysis of the scRNA seq dataset captured 4 major clusters including T/NK cells, granulocytes, myeloid cells and B/plasma cells and 3 clusters with residual populations of acinar cells, fibroblasts and ductal/cancer cells (Fig. 6O). Further sub-clustering of the B/plasma cell populations revealed increase in naive (follicular) B cells, germinal center B cells (3 clusters) and IFN-stimulated/activated B cell populations in the Daraxonrasib and anti-CTLA4 treated tumors (Fig. 6O, P, Supplementary Fig. S19D–F). The plasma cell populations weren’t captured in the Daraxonrasib treated tumors as they are negative/low for CD45 (Fig. 6M–P). Interestingly, minor populations of regulatory B cells and marginal zone B cells were observed in the CD45+ sorted samples that weren’t captured in the earlier analysis of B cells. However, treatment with Daraxonrasib, anti-CTLA4/anti-PD1 did not affect these cell numbers in the KPC2 tumors (Fig. 6O, P). Further analysis of the transcriptional profiles of B cells in the Kras* targeting and anti-CTLA4 therapy indicated that the IFN stimulated/activated B cells demonstrate an increase in cytokines including hallmark interferon-stimulated genes viz. Ifit3, Slfn5, and Isg15 and germinal center B cell transcription factor Bcl6 (Supplementary Fig. S19A, D). Our data establishes the recruitment of TLS in mice treated with Kras* targeting MRTX1133/Daraxonrasib in combination with anti-CTLA4 in multiple orthotopic and autochthonous GEMs. The presence of neo-vascularization, specific T and B cell recruitment, expression of CD21 and CD23, expansion of follicular B cells, plasma cells and IFN-stimulated/activated B cells validate that these structures are TLS and not lymphoid aggregates. Tertiary lymphoid structures functionally contribute to synergy of Kras* targeting with anti-CTLA4 therapy Next, to determine whether the TLS play a functional role in mediating synergy between MRTX1133/Daraxonrasib and anti-CTLA4 in PDAC, we treat mice that received combination therapy with lymphotoxin-β inhibitor (LTBi) (Baminercept) as it was the most robust cytokine expressed by the tumor infiltrating B cell (Fig. 7A, Supplementary Fig. S20A, B). In addition, we use an antibody to deplete CD19+ B cells in mice that received combination therapy (Fig. 7A, Supplementary Fig. S20C, D). Analysis of peripheral blood from mice that received anti-CD19 depletion antibody revealed systemic depletion of CD19+ B cells (Supplementary Fig. S20C, D). LTBi and B cell depletion reversed the survival benefit conferred by the combination, despite marginal increase in survival compared to mice that received MRTX1133/Daraxonrasib monotherapy indicating that the TLS are functionally critical for synergy of Kras* inhibitors with anti-CTLA4 therapy (Fig. 7B–F, Supplementary Fig. S21A–D). We performed histological analysis of the pancreatic tissues when the mice that received MRTX1133/Daraxonrasib monotherapy, MRTX1133/Daraxonrasib in combination with anti-CTLA4 and LTBi, and MRTX1133/Daraxonrasib in combination with anti-CTLA4 and anti-CD19 succumbed to PDAC (Fig. 7D–F). The mice that received MRTX1133/Daraxonrasib and anti-CTLA4 demonstrated robust tumor inhibition at a time point when the tumors with inhibition of TLS succumbed to PDAC (Fig. 7D–F, Supplementary Fig. S21A–D). Our data indicates the functional contribution of TLS in PDAC inhibition with combination therapy (Fig. 7D–F). Our data establishes that Kras* targeting resulted in recruitment of T cell infiltrates and synergizes with anti-CTLA4 therapy but not with other checkpoint inhibitors including anti-PD1, anti-Tim3, anti-Lag3, anti-Vista or anti-4-1BB agonist antibody therapy. The combination of Kras* targeting and anti-CTLA4 therapy resulted in recruitment of TLS in the pancreatic TME with expansion of naive (follicular) B cells, plasma cells, germinal center B cells and IFN-stimulated/activated B cells (Fig. 7G). Inhibition of tertiary lymphoid structures negates the effects of Kras* targeting and anti-CTLA4 combination therapy. A Schematic representation of orthotopic injection and MRTX1133, Daraxonrasib, αCD19, αCTLA4, and LTBi (Baminercept) treatments of KPC2 mice. B, C Kaplan–Meier survival curves of Iso (n = 5), MRTX1133 (n = 5), MRTX1133 + αCTLA4 (n = 8), MRTX1133 + αCTLA4 + αCD19 (n = 4), MRTX1133 + αCTLA4+LTBi (n = 8) (B), and Daraxonrasib (n = 5), Daraxonrasib + αCTLA4 (n = 8), Daraxonrasib + αCTLA4 + αCD19 (n = 5), Daraxonrasib + αCTLA4+LTBi (n = 8) (C) treated KPC2 mice. Note: Same Iso treated mice were repeated in Figs. 7B and 7C to allow for direct comparisons. D, E Representative H&E sections in MRTX1133, Daraxonrasib, αCD19, αCTLA4 and LTBi treated KPC2 mice (D), with quantification of relative percentage of tissue phenotype (Iso (n = 4), MRTX1133 (n = 4), MRTX1133 + αCTLA4 (n = 6), MRTX1133 + αCTLA4 + αCD19 (n = 3), MRTX1133 + αCTLA4+LTBi (n = 7), Daraxonrasib (n = 4), Daraxonrasib + αCTLA4 (n = 3), Daraxonrasib + αCTLA4 + αCD19 (n = 5), Daraxonrasib + αCTLA4+LTBi (n = 6)) (E). F Endpoint tumor weights of mice from the indicated treatment groups (Iso (n = 5), MRTX1133 (n = 5), MRTX1133 + αCTLA4 (n = 7), MRTX1133 + αCTLA4 + αCD19 (n = 3), MRTX1133 + αCTLA4+LTBi (n = 7), Daraxonrasib (n = 5), Daraxonrasib + αCTLA4 (n = 8), Daraxonrasib + αCTLA4 + αCD19 (n = 5), Daraxonrasib + αCTLA4+LTBi (n = 6)). G Schematic representation summarizing the synergy between Kras* targeting and checkpoint immunotherapy. Created in BioRender. Mahadevan, K. (2026) https://BioRender.com/fj8ldvh. Data are presented as mean ± SD in (E) and (F). Significance was determined by two-way ANOVA in (E), log-rank test in (B) and (C), and by one-way ANOVA with Sidak’s multiple comparisons test in (F). Scale bars: 100 μm. See related Supplementary Figs. S20 and S21. Discussion Pancreatic cancer remains refractory to checkpoint immunotherapies in pre-clinical models and clinical studies45,46,85,86. Genetic silencing and pharmacological inhibition of Kras* with both MRTX1133 and Daraxonrasib results in an increase in tumor infiltrating CD4+ T cells, CD8+ T cells, Tregs and B cells in the acute phase with complete regression of advanced PDAC in genetic models40,41. In contrast, combining standard-of-care chemotherapy with checkpoint immunotherapies does not yield additional survival benefit in PDAC55,85,87, indicating that the changes in the immune response is specific to Kras* inhibition. We are intrigued by the fact that pan-RAS inhibition with Daraxonrasib led to expansion of global T cell infiltrates, since wild-type RAS signaling would be expected to result in a decrease in T cell proliferation, based on decades of research on Ras regulation of cellular proliferation88–91. The concurrent increase in T cell populations indicates that, at a given therapeutic dose, Daraxonrasib predominantly targeted oncogenic Kras*, with minimal impact on expanding immune population. Further, reduction of immunosuppressive Cd11b+ myeloid cells may contribute to the enhancement of anti-tumor T cell responses. This observation aligns with established literature showing that KrasG12D-driven signaling (e.g., via GM-CSF upregulation) promotes a myeloid-rich, T-cell-excluded PDAC TME92,93. Our findings suggest that myeloid depletion is an immediate consequence of KrasG12D inhibition and likely contributes to enhanced T cell infiltration in PDAC resulting in a therapeutic window to target T cell responses. The initial phenotypic analysis of the tumor infiltrating lymphocytes following Kras* inhibition was performed with KPC1 tumors. However, there was a confounding variable as the KPC1 cancer cells demonstrated expression of EYFP which could be potentially immunogenic. Therefore, for subsequent experiments and characterization with both MRTX1133 and Daraxonrasib and immune checkpoint blockade therapy, we used the KPC2 syngeneic orthotopic model with no confounding reporters. In addition, we use a third KPC3 model to validate that the findings from the KPC2 model with long-term follow-up and characterization. Further, we use two independent KPC and KPPC GEMMs to further validate the synergy between Kras* inhibition and anti-CTLA4 therapy. The fact that similar phenotypes in the TILs with both pan-RAS and KrasG12D allele-specific inhibitors across multiple PDAC orthotopic and GEMMs strengthen the robustness of the observed phenotype. Despite a robust increase in T cell infiltration, they did not result in a long-term memory/activated CD8+ T cell response and therefore led to escape from pharmacological Kras* targeting with mice succumbing to PDAC. Our results demonstrate that irrespective of the Kras* targeting strategy, specific addition of anti-CTLA4 combination led to long-term PDAC suppression, inhibition of effector Tregs, switch to a naive Treg phenotype, enhanced CD8 effector response in combination with development of TLS within the PDAC. The suppression of the effector Treg cytokines Il10, Tgfb1, and Il35 contributed to the naive Treg phenotype and expansion of the effector CD8 response in the tumors that received anti-CTLA4 combination therapy. Our data indicates that anti-CTLA4 combination epigenetically suppresses II35 and Il10 in Tregs likely through downregulation of AP-1 family of transcription factors Fos, Fosb, Junb, and Jund at the promoter and enhancer regions of these genes. Alternatively, the observed phenomenon may be explained by newly recruited Tregs remaining in a naive state rather than maturing into an effector Treg phenotype in the presence of Kras* inhibition and anti-CTLA4 combination therapy. It is interesting to note that targeting Kras* results in reprogramming of the tumor microenvironment reflecting all its constituents including the immune cells93. This observation highlights that Kras* mediated oncogenesis significantly reprograms the microenvironment towards tumor immune responses with causal impact. Deletion of CD4+ T cells or CD25+ T cells enhances the impact of Kras* inhibitors, directly implicating the need for productive immune responses in realizing the full efficacy of Kras* inhibitors. Targeting Kras* results in cancer cells alters many signaling pathway axis and likely blocks secretome that inhibits recruitment of T cells and B cells. Consequently, Kras* targeting favors secretome that can help their recruitment with functional impact. Our data suggest that while the resulting immune responses is of benefit, it is not optimal and therefore leads to PDAC relapse. But if relevant immune checkpoint inhibitor is combined with Kras* targeting, long-term benefits are observed without emergence in majority of the mice. Analysis of exhausted CD8+ T cells following Kras* targeting revealed high expression of multiple immune checkpoint genes including Tim3, Lag3, 4-1BB, PD1, and Vista. Interestingly, treatment with individual checkpoint therapies such as anti-Tim3, anti-Lag3, anti-41-BB agonist antibody, anti-Vista, and anti-PD1 did not reverse CD8+ T cell exhaustion in combination with the Kras* inhibitors. A likely explanation could be that these CD8+ T cells are terminally exhausted and unable to reverse with individual checkpoint targeting (vide supra), and additional methods of priming are required to recruit new CD8+ T cells58. Our results indicate that addition of CTLA4 blockade likely results in an expansion of tumor reactive CD8+ T cells and broadens the T cell repertoire in addition to reversing existing exhausted CD8 populations94–96. Surprisingly, another study using subcutaneous PDAC model demonstrates marginal improvement in survival with Kras* targeting in combination with anti-PD1 therapy97. However, this study represents a PDAC setting with intrinsically high T cell infiltration and sensitivity to checkpoint blockade as a monotherapy55–57,97. Further, subcutaneous tumor models lack the tumor microenvironment and associated lymphoid tissue unique to organs in the gastrointestinal tract. Our work determines the role of immune checkpoint inhibitors in a cancer context with the appropriate PDAC tumor microenvironment, lymphatic drainage and stromal components in pancreatic cancer. Another study using a genetically engineered iKPC PDAC model with long-term follow-up demonstrated that KrasG12D inhibition does not synergize with anti-PD1 therapy98, consistent with our findings indicating that combination therapy with anti-PD1 did not reverse terminally exhausted CD8s in PDAC. Further analysis of the Tregs/CD4+ T cells with depletion of specific cell populations revealed that the CD4s are inhibitory overall and impede the anti-tumor CD8 response. Surprisingly, our data with multiple PDAC models demonstrate a lack of Th1 response with Kras* targeting monotherapy or with anti-CTLA4 combination therapy indicating that ‘CD4 help’ could be dispensable for the generation of effector CD8+ T cell response in PDAC. Comparative analysis of melanoma and PDAC infiltrates from patient samples further confirms that Th1 cells were nearly non-existent in the PDAC TME99. In the context of melanoma, Th1 and Tregs comprised the majority of the CD4+ infiltrates. Anti-CTLA4 therapy resulted in an expansion of the Th1 helper response accompanied by Treg suppression. Our prior work in this area which probed the role of KrasG12D targeting on the immune microenvironment using the inducible KPC GEM models indicate that the tumor infiltrating CD4+ T cells differentiate into Th2, Th17 and Treg lineages and not the Th1 lineage in PDAC40. Depletion of CD4+ T cells in the context of KrasG12D targeting heightened the CD8+ cytotoxic lymphocyte frequencies further indicating that CD4s overall are inhibitory to CD8+ T cell response40. Other studies that probe the functional role of CD4+ T cells, in particular Th2 and Th17 in different PDAC contexts with pre-clinical depletion models conclude that the CD4+ T cells are inhibitory to the cytotoxic CD8+ T cell response100–102. Taken together, our work highlights the profile of the tumor infiltrating lymphocytes and the distinct anti-tumor immune responses in PDAC. Combination therapy (Kras* targeting and CTLA4 blockade) leads to emergence of TLS in PDAC. The emergence of TLS as a response to immunotherapies represents a favorable prognostic indicator for control of cancer80,81. TLS occurring in melanoma patients was first identified as a marker of good prognosis and response to immunotherapies. Subsequently, the presence of TLS across multiple cancer models have been associated with longer survival rates with prolonged tumor inhibition103–105. TLS differ from the secondary lymphoid organs such as lymph nodes as they lack encapsulation, which enables direct access of the cellular components (CD8+ T cell, B cell and plasma cells) to facilitate PDAC inhibition80,81,106. It is unclear what the lymphoid tissue stimulatory components are in inducing TLS in cancer. Earlier studies have implicated several immune populations including Th17 cells107, some subsets of innate lymphoid cells (ILCs)104,108, effector CD8+ T cells and B cells109–111. In the PDAC models here, Kras* inhibition resulted in an increase in CD4+ T cell infiltrates including CD4-Foxp3- T cells and Th17 cells as shown in prior studies40,100, CD8+ T cells (including exhausted and effector CD8s) and B cell infiltrates. However, the presence of TLS in the context of Kras* targeting occurs only in combination with anti-CTLA4 which likely indicates that an enhanced CD8 effector response with suppression of exhausted CD8s and effector Tregs drives the formation of TLS. TLS are frequently seen in cancer at different stages of maturation including immature (consisting of dense lymphocyte aggregates without markers of FDCs and B/T cell zones), primary follicle like TLS (presence of FDCs but lacking germinal centers) and secondary follicle like TLS (mature with active germinal centers)80,112,113. The combination of Kras* targeting and anti-CTLA4 resulted in expansion of germinal center B cells and plasma cells with expression of FDC/TFH markers CD21 and CD23 likely indicating that the TLS were maturing with functional capacity. Another critical aspect of anti-tumor immunity associated with TLS is whether they can re-educate the T cells to accentuate an effector CD8 response. Although an increase in IFN stimulated/activated B cells and effector CD8+ T cells was observed with the Kras* targeting combination with anti-CTLA4 therapy, lack of expansion of the TFH cells or Th17 cells in the scRNA seq analysis of combination therapy tumors indicates that the TLS associated with PDAC aren’t fully mature, as also seen in our immunostaining analysis. To determine whether the TLS contribute functionally to PDAC inhibition with the combination therapy, we used them in combination with LTBi and B cell depletion experiments. Our results further validate that although the TLS are at varying stages of maturation, they are critical in realizing the full potential of Kras* targeting in combination with anti-CTLA4 therapy. Overall, this study provides a detailed assessment of the adaptive immune response and T cell regulatory networks upon Kras* inhibition and provides compelling evidence for combination with ipilimumab in the clinic for long-term benefit. Methods The study complies with all relevant ethical regulations; all animal experiments and procedures were reviewed and approved by the Institutional Animal Care and use Committee (IACUC) at MD Anderson Cancer Center (MDACC). Mouse studies Several primary PDAC cells lines, 343 P KPC (Pdx1-cre, LSL-KrasG12D/+, Trp53R172H/+; Rosa-LSL-EYFP cell line, referred to as KPC1), Hy19636-PKC (P48-Cre, LSL-KrasG12D/+, Trp53F/+ cell line, referred to as KPC2), KPC-T (Pdx1-cre, LSL-KrasG12D/+, Trp53R172H/+, referred to as KPC3) were used for orthotopic implantation into the tail of the pancreas of 8-16w old C57BL6/J mice (concentration 0.5 × 106 cells in 20 μL PBS). The autochthonous KPC (Pdx1-cre, LSL-KrasG12D/+, Trp53R172H/+) and KPPC ((P48-Cre, LSL-KrasG12D/+, Trp53F/F) mice have been characterized earlier114,115. For experiments with immune checkpoint inhibitors, anti-mouse CTLA4 (BioXcell, clone 9H10, catalog no. BE0131) or anti-mouse PD-1 (BioXcell, clone 29F.1A12, catalog no. BE0273) were injected three times in 1 week, intraperitoneally in a final volume of 100 μL as indicated. A starting dose of 200 μg was followed by two doses of 100 μg each (1 mg/mL concentration). Isotype treatments in this group received a combination of Rat IgG2a (BioXcell, clone 2A3, BE0089) and Syrian hamster IgG (BioXcell, BE0087) in the same route, time, and dosing as the checkpoint inhibitor antibodies. For anti-mouse Lag3 (BioXcell, clone C9B7W, catalog no. BE0174), anti-mouse Tim-3 (CD366) (BioXcell, clone, catalog no. BE0115), anti-mouse Vista (BioXcell, clone RMT3-23, catalog no. BE0310), anti-mouse 4-1BB (CD137) (BioXcell, clone LOB12.3, catalog no. BE0169) were injected intraperitoneally, 200 μg starting dose, 100 μg (2 doses) in a final volume of 100 μL PBS in a span of 1 week. The control mice received a cocktail of corresponding isotype antibodies include Rat IgG1 (BioXcell, Catalog no. BE0088), Rat IgGa2 (BioXcell, clone 2A3, catalog no. BE0089), Armenian hamster IgG (BioXcell, catalog no. BE0091) in the same route, time, and dosing as the checkpoint inhibitor antibodies. For CD4+ and CD25+ T cell depletion experiments, anti-mouse CD4 (BioXcell, clone GK1.5, catalog no. BE0003-1) or anti-mouse CD25 (IL-2Rα) (BioXcell, clone PC-61.5.3, catalog no. BE0012) and their isotype antibodies (Rat IgG2b, BioXCell, cloneLTF2, catalog no. BE0090 and Rat IgG1, BioXcell, clone HRPN, Catalog no. BE0088) were injected intraperitoneally, twice per week, 200 μg of antibody for the course of the experiments in a final volume of 100 μL PBS. For CD19+ B cell depletion experiments, anti-mouse CD19 (BioXcell, clone1D3, catalog no. BE0150) or the isotype Rat IgG2a (BioXcell, clone 2A3, BE0089) were injected intraperitoneally, 200μg of antibody in a final volume of 100 μL PBS, twice per week as indicated. For lymphotoxin-β inhibiton, 100 μg of baminercept was injected in a final volume of 100μL PBS intraperitoneally, every 5 days for the entire course of the experiments. For MRTX1133 experiments, mice were treated with 30 mg/kg MRTX1133 in 100 μL of 10% captisol (SelleckChem, Catalog No. S4592) in 50 mM sodium citrate (pH 5) (Tekanova) vehicle BID i.p. or the corresponding vehicle in control mice as described earlier1,41. For Daraxonrasib experiments, mice were treated with Daraxonrasib, prepared using formulation of 10% DMSO (MP Biomedicals, 0219605590), 20% PEG-400 (Possible Missions Inc, NC0572735), 10% Solutol HS15 (SelleckChem, Catalog No. E1708) and 60% milliQ water as described in the prior study12 and administered by oral gavage, 25 mg/kg, once per day in a final volume of 100 μL. Biosimilars with the same structure was synthesized and provided for both MRTX1133 and Daraxonrasib by TRACTION, the therapeutics discovery division of MD Anderson. The control mice received the vehicle by oral gavage in the same dose and frequency. The mice were housed in modified barrier conditions, fed ad libitum (LabDiet 5053) in 12-h light/dark cycle at 21 °C, 40–60% humidity. The mice were euthanized before reaching moribund state defined by poor grooming, weight decrease (20% of total BW), abnormal posture, and enlargement of the abdomen (ascites), skin discoloration (jaundice) upon physical exam. Both male and female mice were used for all the experiments described in this manuscript. For histological assessments of H&E stained sections, five high power fields or the entire tumor area (for smaller tumors) were imaged for analysis. All experiments were approved by MD Anderson Cancer Center IACUC. Cell culture The KPC1 cells were cultured in 10% fetal bovine serum (FBS) RPMI (Corning, 10-040-CV) with 1% Penicillin-Streptomycin (PS) (Corning, 30-002-CI). The KPC2 and KPC3 cells were cultured in 10% FBS DMEM (Corning, 10-017-CV) with 1%PS. All cells were tested and confirmed to be negative for mycoplasma using LookOut Mycoplasma PCR Detection Kit (Sigma-Aldrich, MP0035) prior to use in in vivo experiments or analysis. Flow cytometry PDAC tissues/pancreas were minced with razor blades and digested in 1.5 mg/mL of Type-1 Collagenase (Gibco, 17100017) in 10% Bovine serum albumin (Sigma-Aldrich, A9418) in 10 mL of PBS at 37 °C for 6–10 min in an incubator at 150 rpm. The samples were subsequently washed in cRPMI (10% heat-inactivated FBS (HI-FBS) (Atlanta Biologicals, Atlanta, GA, USA), 1% PS, 1 mM sodium pyruvate (ThermoFischerScientific, 11360070), and 50 mM β-mercaptoethanol (Life Technologies, 31350010)) and centrifuged at 800 g for 3 min at room temperature 4 °C. Cells were filtered using a 70 μm strainer, washed and resuspended in FACS buffer (2% HI-FBS in PBS). Cells were stained in a final volume of 100 μL with antibodies listed in Supplementary Table1 diluted in 20% brilliant stain buffer (BD Bioscience, 566349), Live/dead stain, 1:1000 (eBioscience, 65-0865-14), and anti-mouse CD16/32, 1:200 (TONBObiosciences, 40-0161) for 30 min on ice. Subsequently, cells were permeabilized (Foxp3/Transcription Factor Staining Buffer Set (eBioscience, 00-5523-00) as per manufacturer’s instructions and incubated with intra cellular antibodies for 30 min following which the cells were fixed (BD Bioscience, 554655) and data were acquired using a Cytek Aurora and analyzed with FlowJo v10.7.1. The gating strategy used for analysis and sorting in this manuscript are included in Supplementary Fig. S22. Single cell RNA/ATAC sequencing PDAC tissues/pancreas were minced and digested in 1.5 mg/mL of Type-1 Collagenase in 10% Bovine serum albumin in 10 mL at 37 °C for 6–10 min in an incubator at 150 rpm. To mitigate the risk of digestion artifacts arising from differences in tumor weights, we utilized similar weights of tumor tissue for each sample during the dissociation process, ensuring that the ratio of tissue to digestion buffer was similar across all groups. For the UMAP analyses, we downsampled the datasets to achieve comparable cell numbers across all treatment conditions. As expected, mice treated with Daraxonrasib or MRTX1133 monotherapy exhibited significant tumor inhibition resulting in lower tumor weights, with further reductions observed in the anti-CTLA4 combination groups (Supplementary Fig. S23A–E). The samples were subsequently washed in cRPMI and stained with Live/dead (Fixable-viability dye (FVD)) 1:1000 and/or CD45 (1:100), CD16/32 (1:200) for 20 min and sorted for Live cells or CD45+ cells and submitted to the ATGC core for single-cell processing. Chromium Next GEM Single Cell 3ʹ Reagent Kit v3.1 was used for scRNA seq and Chromium Next GEM Single Cell Multiome ATAC + Gene Expression kit was used for scATAC seq. downstream processing as per the manufacturer’s protocol. scRNA-seq data processing and analysis Raw FASTQ files were processed with Cell Ranger version 7.0.1 to generate count matrices. For aligning reads from various mouse models, we used the mm10 reference genome provided on the 10X Genomics website. Downstream analysis was performed in R (version 4.1.3) using the Seurat package (version 4.4.0) (Butler et al., 2018)62. We applied several Seurat functions to refine and interpret the data. First, to minimize the influence of noise and eliminate low-quality cells, we filtered out cells expressing an unusually low or high number of genes. Specifically, we determined filtering thresholds using R’s quantile function, retaining cells with gene counts between the 2.5th and 97.5th percentiles and discarding those outside this range. Additionally, cells exhibiting more than 5% mitochondrial counts were excluded from further analysis. Normalization and variance stabilization were achieved with the “Sctransform” function. We then reduced dimensionality via principal component analysis (PCA). The ‘FindNeighbors’ function identified nearest neighbors within the PCA space, and the optimal number of principal components was selected by examining outputs from both ‘JackStrawPlot’ and ‘ElbowPlot’. Cells were subsequently clustered using the ‘FindClusters’ function with an appropriate resolution setting, and the ‘RunUMAP’ function was employed to generate UMAP visualizations of the clusters. Finally, the ‘DoHeatmap’ function showcased the top 10 genes for each meta-cluster, while the ‘VinPlot’ function illustrated the distribution of gene expression probabilities across clusters. Gene imputation was carried out using the MAGIC (Markov Affinity-based Graph Imputation of Cells) algorithm116. Cell-cycle analysis was performed using the CellCycleScoring function in the Seurat R package. Canonical gene sets for the S and G2/M phases were obtained from the cc.genes.updated.2019 dataset and converted to the corresponding mouse gene symbols. For each cell, S-phase and G2/M-phase scores were computed based on these reference gene sets, and cells were subsequently classified into one of three phases, G1, S, or G2/M, according to the relative enrichment of phase-specific marker genes. scATAC-seq data processing and analysis For scATAC-seq, raw data were initially processed with Cell Ranger ATAC version 2.1.0, which mapped reads to the mm10 genome and generated fragment files. These fragment files were then analyzed in R using the ArchR package (version 1.0.2). Quality control steps included removing cells with fewer than 4000 unique fragments or a TSS enrichment score below three, and doublets were filtered out using default parameters. Cells that met these quality criteria were clustered using ArchR’s integration of the Seurat method at a resolution of 0.5. Dimensionality reduction and visualization were performed using UMAP, configured with 30 neighbors. Peak calling was conducted on aggregated coverage profiles for each identified cell population using MACS2 with its default settings. Additional analyses, such as motif enrichment and footprinting, were executed using the corresponding functions within ArchR under recommended settings. Statistical synergy analysis Synergy between MRTX1133 and anti-CTLA4 was assessed using the four-arm factorial design consisting of vehicle/isotype control, MRTX1133, anti-CTLA4, and MRTX1133+anti-CTLA4. For tumor weight, we used a Bliss-type log-linear interaction model: Log(tumor weight) = β0 + β1(MRTX1133) + β2(αCTLA4) + β3(MRTX1133 × αCTLA4). A significant negative interaction term was interpreted as evidence that the observed combination effect was greater than expected under independent/multiplicative treatment effects. For histologic tumor burden, PDAC percentage was analyzed using a linear factorial model: PDAC% = β0 + β1(MRTX1133) + β2(αCTLA4) + β3(MRTX1133 × αCTLA4). Because uninvolved pancreas and PDAC percentages add to 100%, PDAC percentage was used as the primary histologic endpoint. A significant negative interaction term indicated a greater-than-additive reduction in PDAC burden. Heteroskedasticity-robust HC3 standard errors were used for interaction testing. Immunostaining Formalin-fixed paraffin-embedded (FFPE) tissues were sectioned (5 µm thickness), melted at 65 °C for 1 h and deparaffinized. Antigen retrieval was performed in Tris-EDTA buffer (pH 8.6) at 95 °C for a total of 45 min (20 min at 95 °C) using a steamer and cooled in an ice bath for 30 min. Subsequently, using a Pap pen, a hydrophobic chamber was created surrounding the tissue, followed by washing in PBST (0.1% Tween 20) for 3 min. The sections were then treated with 3% H2O2 for approximately 10–15 min, followed by blocking with 1.5% bovine serum albumin in PBST for 1 h at room temperature. Primary antibodies for phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) (Cell Signaling Technology, #4376S, 1:100), CK19 (EP1580Y) (Abcam, #52625, 1:400), Cd11b (Abcam, #133357, 1:500), or Tgf-β (Abcam, #ab215715, 1:500) were applied to the sections and incubated for 3 h at RT. Sections were then incubated with rabbit-on-rodent horseradish peroxidase polymer (BioCare Medical) for 15 min, followed by DAB (Life Technologies) incubation for 5–10 min or until positive expression was seen. For IL-10 staining (R&D Systems, AF519, 1:50), were applied to the sections and incubated for 3 h at RT. Sections were then incubated in goat-on-rodent polymer (GHP516) for 15 min and subsequently visualized using DAB. The sections were then counterstained with hematoxylin, dehydrated, and finally cover slipped. Representative histological images were assessed and collected using a Leica DM1000 LED microscope equipped with a DFC295 camera (Leica) and LAS version 4.4 software (Leica), with quantification analysis performed using ImageJ. For CD8, CD19 and CD31, co-expression analysis, a Tyramide signal amplification (TSA) staining was conducted. 5 µm thickness sections were processed as described above until the blocking step. After blocking for 1 h at RT using blocking buffer, the sections were further incubated with CD8 (D8A8Y) (Cell Signaling Technology, #85336S, 1:200) primary antibody for 3 h at RT. The slides were then washed and further incubated with rabbit-on-rodent HRP polymer (BioCare Medical) for 15 min, followed by a 30-min incubation with Opal 690 (Akoya Biosciences, 1:100) followed by a second antigen retrieval step. Next, for CD19 staining, sections incubated in blocking buffer, followed by incubation in CD19(D4V4B) (Cell Signaling Technology, #90176S, 1:100) for 3 h at RT. The sections were then incubated in rabbit on rodent HRP polymer for 15 min, followed by 30 min incubation in Opal 520 (Akoya Biosciences, 1:100). Next, following antigen retrieval, the sections were incubated in CD31 primary antibody (D8V9E) (Cell Signaling Technology, #77699S, 1:100), subsequent incubation in rabbit-on-rodent HRP polymer and Opal 570 (Akoya Biosciences, 1:100) followed by a final antigen retrieval step and cover slipping with Fluoroshield with DAPI (Sigma-Aldrich, F6057). Similarly, a TSA for CD21 and CD21 expression was performed (Reagent details in Supplementary Table2). All TSA images were observed and imaged using Zeiss LSM800 confocal microscope and quantified using Image J. Five high power fields per mouse tumor or the entire tumor area (for smaller tumors) were imaged for all immunostaining analysis throughout the manuscript. Statistical analysis Figures in the manuscript were generated using Graphpad prism 9 and R-studio. Normality of distribution of datasets was assessed using Shapiro–Wilk test. For comparisons involving two groups, unpaired-T test (datasets with normal distribution) or Mann–Whitney test (datasets with non-normal distribution) was used. For comparisons of multiple groups, one-way ANOVA (datasets with normal distribution) or Kruskal–Wallis test (datasets with non-normal distribution) was used. Log-rank test was used for survival analysis and two-way ANOVA was used for histological comparisons. P values throughout the manuscript represent *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; ns, not significant. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting Summary Transparent Peer Review file Source data Source Data