Site-Dependent Immune Escape Due to Impaired Dendritic Cell Cross-Priming

Cancer immunotherapies augmenting endogenous antitumor T cells have demonstrated remarkable clinical benefit in certain cancers, yet the majority of patients remain refractory to current approaches (1, 2). Understanding the determinants of tumor immunogenicity is therefore essential, particularly for solid tumors that often exhibit sparse T-cell infiltration and primary resistance to immune checkpoint inhibitors (ICI) targeting the PD-1/PD-L1 and/or CTLA-4 pathways. Whereas tumor antigens are clear requisites for immune recognition, the contribution of the microenvironment on cancer immune surveillance is less well understood.

Formative studies using murine sarcoma models have established the capacity of innate and adaptive immunity to modulate primary tumor development and sculpt the immunogenicity of developing tumors (3, 4). Neoantigens generated by cancer-specific mutations and recognizable as foreign by T cells are central targets of immune surveillance and immunoediting in such models (5, 6). Neoantigen-specific T cells are also critical for ICI-mediated responses against less immunogenic tumors (7, 8). In patients with cancer, tumor mutational burden similarly correlates with ICI response in several malignancies (9, 10), and loss of neoantigen expression or deficits in antigen presentation, including the loss of IFN sensitivity, have been detected in tumors with primary or acquired immunotherapy resistance (11–13).

Tumor antigenicity alone, however, is insufficient to predict ICI response, and important biomarkers related to the tumor microenvironment (TME) have also been identified, including CD8⁺ T-cell infiltration, expression of T-cell–inflamed genes, PD-L1 staining, and other markers of ongoing immune reactivity (14, 15). However, the correlation is poor across cancer types between an inflamed TME and tumor mutational burden or the density of immunogenic antigens (10, 16, 17). This dissociation between tumor antigenicity and immunogenicity is likely multifactorial and consistent with the influence of both tumor cell–intrinsic and –extrinsic factors actively modulating the immune landscape. Various relevant pathways have indeed emerged, including checkpoint molecules or immunosuppressive cytokines downstream of oncogenic signaling, and host-related factors such as immune status or the microbiome (18–21). Interestingly, mixed clinical responses to ICI are sometimes observed within the same patient, and reduced ICI efficacy for tumors at certain anatomic sites has been reported (22, 23). These observations have largely been understood in the context of tumor heterogeneity or differences in tumor-intrinsic immunosuppression (24–26), though preclinical studies also raise the possibility of differential cancer surveillance based on tissue site of origin (27–29).

Pancreatic ductal adenocarcinoma (PDAC) is an extremely lethal cancer with a particularly immunosuppressive TME, characterized by poor CD8⁺ T-cell infiltration (immunologically “cold”) and ICI unresponsiveness (30–32). Yet, human PDAC contains mutational and genomic complexity (33), increased T-cell infiltration is associated with better prognosis (34), and T-cell reactivity to high-quality neoantigens was discovered in rare long-term PDAC survivors (35), suggesting promise for immunotherapy if barriers can be surmounted. Genomic and transcriptional analyses of human PDAC, however, have suggested that cytolytic activity does not correlate with increased mutational or neoepitope load (36) as noted in some cancers (37), and various tumor-intrinsic immunosuppressive mechanisms have been characterized in preclinical models of this disease (38–40).

Using the “KPC” genetically engineered murine model (GEMM) of PDAC, driven by pancreatic expression of mutant Kras and p53, we previously demonstrated that T-cell depletion does not affect the natural history of tumor development (41)—a finding in contrast to carcinogen-induced tumor models or more immunogenic GEMMs (4, 6). KPC-derived tumors had a low mutational burden and few predicted neoepitopes, and introduction of a single neoantigen could reestablish protective immunity and elicit immunoediting (41), leading to the hypothesis that immune quiescence in PDAC arises from low tumor antigenicity. In this study, we explored T-cell–mediated surveillance of neoantigen-expressing PDAC, finding that high tumor antigenicity is not sufficient within certain host tissue microenvironments, including the pancreas. Instead, we observed differential immune surveillance and escape of antigenic tumors without immunoediting, a consequence of poor immune priming, which could be restored by enhancing dendritic cell (DC) function.

Wild-type (WT) C57BL/6J mice were purchased from The Jackson Laboratory. Rag2^−/−, μMT, Act-mOVA, Batf3^−/−, CD40^−/−, Lta^−/−, and Cxcr3^−/− mice on a C57BL/6 background were obtained from The Jackson Laboratory and/or bred at the University of Pennsylvania (Philadelphia, PA). For tumor transplantation experiments, generally female mice at approximately 8 to 12 weeks of age were used. Mice were maintained in a specific pathogen–free barrier facility in accordance with American Association for Laboratory Animal Science guidelines. All study protocols involving mice were approved by the University of Pennsylvania Institutional Animal Care and Use Committee.

The 4662 PDAC cell line was derived from a spontaneous pancreatic adenocarcinoma in a female KPC (Kras^LSL-G12D/+Trp53^LSL-R172H/+Pdx-Cre^+/−) mouse on the C57BL/6J background (42), and the presence of rearranged Kras and Trp53 alleles were verified by RT-PCR (41). Cells were tested by the Research Animal Diagnostic Laboratory (RADIL) at the University of Missouri (St. Louis, MO) using the Infectious Microbe PCR Amplification Test (IMPACT) and regularly screened for Mycoplasma. Cells from a low-passage stock were used, and no reauthentication was performed. The 4662 tumor line (referred to herein as simply parental PDAC) was cloned by limiting dilution to generate the parental clones C9 and C10, or transduced with a retroviral construct containing full-length ovalbumin (Ova) fused to Td-tomato and cloned to generate V6.Ova and G10.Ova as described previously (41). The V6.Ova clone was used for subsequent studies unless otherwise indicated, and is referred to as PDAC.Ova. Tumor cell lines were generated from primary tumors by digesting with collagenase type IV (1 mg/mL; Sigma-Aldrich) in DMEM for 45 minutes at 37ºC and frozen down as low-passage cell lines after several in vitro passages to eliminate contaminating immune cells and fibroblasts. Cells were grown in D10 media containing DMEM (Gibco) plus 10% FBS (Gemini), 2 mmol/L l-glutamine (BioWhittaker), and gentamicin (80 μg/mL; Gibco). For in vitro growth assays, cells were plated at 1.25 × 10⁵ cells per T25 flask, and triplicate samples were harvested by trypsinization at the indicated time points and counted using a coulter counter.

Tumor cells were harvested by trypsinization at approximately 60% to 90% confluence, washed with DMEM, and counted using a hemocytometer. The viability of injected cells was >90% as assessed by Trypan blue exclusion. For subcutaneous implant studies, tumor cells were injected in 100 μL DMEM using a 27-gauge 1.0 cc TB syringe into the shaved flanks of mice. Tumor size was measured in two orthogonal diameters using a ruler every 4 days and plotted as mean tumor diameter. Mice were euthanized for tumor diameters approaching 20 mm or significant tumor ulceration. For orthotopic implant studies, tumor cells were injected into the tail of the pancreas in 25 μL DMEM using a 29-gauge 0.3 cc insulin syringe, with visualization of a small fluid bubble to indicate successful injection without leakage. Mice were anesthetized with 1% to 3% isoflurane, laparotomy performed using sterile surgical technique, and the spleen and pancreatic tissue externalized prior to injection. The peritoneum was closed using absorbable Polymend MT 5/0 suture (Penn Veterinary Supply) and the overlying skin with wound clips. Tumor size was quantitated weekly, or as indicated, by three-dimensional ultrasonography using the Vevo 2100 imaging system with a 55-MHz MicroScan transducer (Visual Sonics). Mice were euthanized for tumor volumes approaching 1,200 mm³ or moribund condition. For intraperitoneal implants, tumor cells were injected in 250 μL total volume using a 27-gauge 1.0 cc TB syringe. For intravenous implants, tumor cells were injected via retro-orbital injection in 100 μL PBS containing 2 mmol/L EDTA (to minimize cell clumping) using a 29-gauge 0.5 cc insulin syringe. Oral tolerance to ovalbumin was induced in WT mice by oral gavage with 50 mg ovalbumin protein (Sigma-Aldrich) in 250 μL PBS weekly for three doses as described (41), followed 1 week later by implantation of tumor cells. For retransplant experiments, “escape” or control “passage” tumor lines were injected subcutaneously into WT mice as above, and mice with progressively growing tumors >5 mm diameter by day 40 were considered tumor positive. For tumor rechallenge experiments, mice that had completely rejected subcutaneous PDAC.Ova tumors (>50 days from injection) were challenged with orthotopic or intraperitoneal tumor cells. For Ova⁺ and Ova⁻ tumor cell mixing experiments, a total of 1 × 10⁶ cells at the indicated ratio were injected. Mice challenged with parental PDAC after PDAC.Ova priming had been injected subcutaneously with PDAC.Ova on the opposite flank 40 days prior and treated at day 7 with control Ig or CD40 agonist.

For T-cell depletion, monoclonal anti-CD8 (2.43) and/or monoclonal anti-CD4 (GK1.5) were injected intraperitoneally at a dose of 0.2 mg every 4 days, starting 1 to 3 days prior to tumor challenge or CD40 agonist treatment as indicated. Natural killer (NK)–cell depletion was achieved by injection of 0.2 mg monoclonal anti-NK1.1 (PK136) on days −2, 0, 2, and then every 5 days relative to tumor injection. CD40 agonist (FGK45) was administered as a single dose of 0.1 mg, intraperitoneally, on day 7 or day 12 relative to tumor challenge. Control Igs (LTF-2 or 2A3) were used at comparable dosing schedules. Anti–PD-1 (RMP1–14) and anti–CTLA-4 (9H10) blocking antibodies were administered intraperitoneally at 0.2 mg every 3 days for three doses as indicated. All antibodies were obtained from BioXCell as low endotoxin preparations. Cell depletions using the indicated treatment regimens were confirmed by flow cytometry.

Single-cell suspensions were prepared from tumors, lymph nodes, or spleens by finely mincing tissue and digesting with 1 mg/mL collagenase type IV (Sigma-Aldrich) for 45 to 60 minutes at 37°C. Tumor-draining lymph nodes were isolated using a dissecting microscope. Cells grown in vitro were harvested by trypsinization, and in some cases treated with recombinant murine IFNγ (R&D Systems) at 100 U/mL for 48 hours prior to analysis. Live/dead cell discrimination was achieved using Live/Dead Fixable Aqua stain (Invitrogen). For H-2K^b-SIINFEKL tetramer staining, cells were incubated with PE-conjugated H-2K^b Ova (SIINFEKL) tetramer or negative control tetramer (MBL International) for 10 to 15 minutes at 37°C prior to surface marker staining, which was performed for 30 minutes at 4°C. For intracellular staining, fixation and permeabilization were performed using the eBioscience Foxp3/Transcription Factor Staining Buffer Kit. The following reagents were obtained from BioLegend: H-2K^b-FITC or -PerCP/Cy5.5 (AF6–88.5), H-2D^b-biotin (KH95), H-2K^b-SIINFEKL-APC (25-D1.16), CD31-APC (MEC13.3), CD3-FITC (145–2C11), CD4-APC (RM4–5), CD8α-PE/Cy7 (53–6.7), TCRγ/δ-FITC or -PE/Cy7 (GL3), NK1.1-PE/Cy7 or -BV711 (PK136), CD49b-FITC (DX5), NKp46-PE (29A1.4), CD62L-PE/Cy7 or -APC/Cy7 (MEL-14), F4/80-FITC or -APC/Cy7 (BM8), Gr-1-APC/Cy7 or -BV711 (RB6–8C5), CD103-APC or -BV421 (2E7), PD-1-PE/Cy7 or -APC/Cy7 (29F.1A12), granzyme B-AF647 or -APC (GB11), and T-bet-BV421 (4B10), LAG-3-PerCP/Cy5.5 (C9B7W), CTLA-4-PE-CF594 (UC10–4B9), CD45-AF700 (30–F11), Ki-67-BV605 (16A8), CD3-BV711 (17A2), CD44-BV785 (IM7), XRC1-FITC (ZET), I-A/I-E-APC (M5/114.15.2), CD11c-BV605 (N418), CD80-BV650 (16–10A1), CD19-BV711 (6D5), B220-BV711 (RA3–6B2), CD86-BV785 (GL-1), CD40-PE (3/23), CD64-PE-CF594 (X54–5/7.1), CD11b-PE/Cy5 (M1/70), and SIRPα-PE/Cy7 (P84). The following were from BD Biosciences: streptavidin-V450, CD45-PerCP (30–F11), CD4-V450 (RM4–5), CD4-BUV395 (GK1.5), CD8α-APC/Cy7 or -BUV805 (53–6.7), CD3ϵ-V450 (500A2), CD44-FITC (IM7), CD11b-V450 (M1/70), CD11c-PE (HL3), CD90.2-BV650 (30–H12), and CD274-BUV395 (MIH5). The following were from eBioscience: Foxp3-PE or -APC (FJK-16s), TIM-3-FITC (RMT3–23), and Eomes-PE/Cy5 (Dan11mag). Data were acquired on a FACSCanto or LSR Fortessa flow cytometer (BD Biosciences) and analyzed using FlowJo software (FlowJo, LLC). Analysis of cDC subsets was performed using a gating strategy as described previously (43), with markers including CD45, CD64, F4/80, MHC-II, lineage markers (CD3, CD19, B220, NK1.1, and Gr-1), CD11c, XCR1, and SIRPα, as well as CD103, CD11b, and CD8α where applicable.

Tumor and lung tissues were harvested and immediately fixed in zinc formalin (Sigma-Aldrich) by intratracheal instillation of fixative. Sectioning and standard hematoxylin and eosin (H&E) staining was performed by the Pathology Core Lab at the Children's Hospital of Philadelphia Research Institute. Images were acquired using a Nikon DS-Fi2 digital microscope camera.

Significance of Kaplan–Meier survival curves was determined by log-rank (Mantel–Cox) test. Flow cytometry data were compared by one-way ANOVA with Tukey post hoc test, unpaired t test, or Mann–Whitney test. Statistical analyses were performed using GraphPad Prism.

To evaluate T-cell immune surveillance within different microenvironments, we employed clonal KPC-derived murine PDAC cells engineered to have high antigenicity by enforced expression of the model antigen Ova fused to td-Tomato (PDAC.Ova; Supplementary Fig. S1A). Parental PDAC and PDAC.Ova lines displayed similar growth kinetics in vitro and in vivo in immunodeficient Rag2^−/− mice, neoantigen expression was stable over time, and tumors exhibited similar morphologies in vivo (Supplementary Fig. S1B–S1E). Although parental PDAC grew progressively in syngeneic WT hosts, PDAC.Ova was readily eliminated after subcutaneous transplant into immunocompetent mice (Fig. 1A–C; Supplementary Fig. S1F), consistent with our prior results (41). Rejection of PDAC.Ova required CD8⁺ T lymphocytes, but not B cells or NK cells, and tumor rejection was abrogated in mice with specific immunologic tolerance to Ova (Act-mOVA transgenic mice; Fig. 1C; ref. 41). In contrast to the uniform rejection of subcutaneous PDAC.Ova cells, orthotopic intrapancreatic challenge revealed reproducible outgrowth in a subset of WT hosts (Fig. 1D), with approximately 30% (7/24 mice) tumor growth over multiple experiments compared with no tumor outgrowth (0/25 mice) with subcutaneous transplantation. Rejection of orthotopic PDAC.Ova similarly required CD8⁺ T cells and was not observed in Act-mOVA mice (Fig. 1D) or upon induction of Ova tolerance in WT mice by serial oral gavage (Supplementary Fig. S2A–S2C). Control of orthotopic PDAC.Ova tumors was potentially enhanced in μMT mice (Fig. 1D), consistent with a protumorigenic role of B cells in PDAC (44, 45), although larger group sizes are needed for more conclusive assessment. At higher cell inoculums, PDAC.Ova grew progressively in the majority of WT mice following orthotopic implantation (75%, 9/12 mice), but only rarely following subcutaneous injection (12%, 3/25 mice; Fig. 1E and F).

When we tested tumor growth following intraperitoneal injection, we found that PDAC.Ova was lethal in WT mice, even at low cell inoculums, with kinetics similar to the parental line (Fig. 1G and H). Alternatively, intravenous injection of parental PDAC cells induced multiple lung metastases in WT mice, whereas PDAC.Ova cells did not (Fig. 1G and H; Supplementary Fig. S2D), unless injected into tolerant Act-mOVA hosts (Supplementary Fig. S2D and S2E). Thus, the outgrowth of PDAC.Ova cells in immunocompetent WT mice depended on the site of tumor implantation (Fig. 1H), but parental PDAC cells were generally lethal regardless of route (Fig. 1G). Similar results were also obtained with two clonal cell lines derived from the parental tumor and an additional PDAC.Ova tumor clone (Supplementary Fig. S1A and S1F). These data reveal that the immunogenicity of clonal PDAC.Ova cells is shaped by the host tissue microenvironment, with selective immune escape of an antigenic tumor in the pancreas and peritoneum.

To determine whether outgrowth of PDAC.Ova tumors reflected loss of neoantigen, we harvested orthotopic or intraperitoneal tumors from WT mice and generated low-passage cell lines (termed orthotopic or intraperitoneal “escapes”). As a control, PDAC.Ova lines were established from orthotopic tumors in WT mice treated with depleting anti-CD8 (“αCD8-passage”). By flow cytometry, orthotopic and intraperitoneal “escape” tumors retained expression of td-Tomato, presented the Ova-derived SIINFEKL peptide in the context of H-2K^b, and upregulated MHC class I molecules (H-2K^b and H-2D^b) upon treatment with IFNγ (Fig. 2A–C). When retransplanted subcutaneously into naïve WT mice, orthotopic “escape” and “αCD8-passage” tumors were still rejected in the majority of WT mice, but grew in T-cell–depleted hosts, and growth kinetics were comparable in “escape” and “αCD8-passage” tumors (Fig. 2D and E), although both groups were slightly more tumorigenic than the original PDAC.Ova at an equivalent cell dose (likely attributable to in vivo passage). These data are therefore consistent with a failure in antitumor immunity leading to escape, rather than tumor-intrinsic immunoediting.

To examine the immune response against PDAC.Ova, we analyzed the priming of tumor-specific SIINFEKL-H-2K^b tetramer⁺ (Ova tet⁺) T cells and tumor-infiltrating leukocyte (TIL) populations by flow cytometry. Compared with early subcutaneous PDAC tumors at day 9, PDAC.Ova tumors in WT mice had increased total T-cell infiltration, particularly CD8⁺ T cells, which was not observed in Act-mOVA mice or WT mice with induced oral tolerance to Ova (Fig. 3A; Supplementary Fig. S3A–S3C). There was an associated decrease in CD11b⁺ myeloid cells, including both CD11b⁺Gr-1^hi and F4/80⁺ subsets (Supplementary Fig. S3D). As expected, we found a significant expansion of Ova tet⁺ T cells in the spleens of WT mice following subcutaneous challenge with PDAC.Ova, but this was largely absent in Act-mOVA mice or WT mice treated with Ova gavage (Fig. 3B). When we assessed T-cell priming following intraperitoneal or orthotopic implantation of PDAC.Ova, we found diminished expansion of Ova tet⁺ T cells in the spleen compared with subcutaneous challenge (Fig. 3B). More robust CD8⁺ T-cell priming was observed in a subset of mice following orthotopic injection, which correlated with the absence of detectable tumor upon harvest at days 12 to 14 (Fig. 3C). In intraperitoneal challenged mice, initial expansion of Ova tet⁺ CD8⁺ T cells in the spleen was attenuated at multiple early time points (Fig. 3D).

Analysis of progressively growing intraperitoneal PDAC.Ova tumors harvested from WT mice showed that CD8⁺ and CD4⁺ T-cell infiltration was comparable with intraperitoneal parental PDAC tumors (Fig. 3E; Supplementary Fig. S3E and S3F), and relatively few infiltrating Ova tet⁺ CD8⁺ T cells were detected in PDAC.Ova tumors (Fig. 3B). NK-cell and myeloid cell infiltration within PDAC and PDAC.Ova tumors was also similar (Supplementary Fig. S3G). Harvest of orthotopic PDAC.Ova tumors from WT mice also revealed T-cell infiltration comparable with orthotopic parental PDAC tumors and low relative abundance of Ova tet⁺ T cells in the spleen (Supplementary Fig. S4A–S4C). In contrast to intraperitoneal PDAC.Ova tumors, higher levels of Ova-specific CD8⁺ T cells accumulated within established orthotopic PDAC.Ova tumors, at least in a subset of mice (Fig. 3B). Collectively, these findings demonstrate diminished CD8⁺ T-cell priming and reduced immunogenicity of PDAC.Ova in the peritoneal and pancreatic microenvironments.

We next performed phenotypic studies to examine the CD8⁺ T cells infiltrating subcutaneous PDAC.Ova compared with progressively growing intraperitoneally and orthotopic PDAC.Ova. CD8⁺ T cells in subcutaneous PDAC.Ova tumors largely displayed an effector phenotype, with increased expression of activation markers and effector molecules such as CD44, PD-1, T-bet, and granzyme B, in comparison with CD8⁺ T cells from subcutaneous parental PDAC (Fig. 3F; Supplementary Fig. S4D). In contrast, CD8⁺ TILs from intraperitoneal and orthotopic PDAC.Ova tumors lacked expression of effector molecules and exhibited elevated PD-1 and TIM-3 (Fig. 3F; Supplementary Fig. S4E and S4F), raising the possibility of T-cell exhaustion, although ex vivo functional assessment would be required to evaluate this possibility.

Finally, we hypothesized that low abundance or impaired maturation of cDCs within intraperitoneal or orthotopic PDAC.Ova compared with subcutaneous tumors might result in poor T-cell priming. We thus examined cDC subsets in the tumor and tumor-draining lymph node (dLN) by flow cytometry, enumerating XCR1⁺ cDC1s and SIRPα⁺ cDC2s, among cDCs defined as CD45⁺CD64⁻MHCII⁺CD11c⁺ cells lacking expression of lineage markers CD3, CD19, B220, NK1.1, and Gr-1, as described previously (43). Such analyses, however, did not reveal a reduction in total cDCs or cDC1s in the tumor or dLN correlating with impaired antitumor immunity (Fig. 3G and H). Maturation markers, including CD40, CD80, CD86, PD-L1, and MHC-II, on cDC1s were also not substantially altered on the basis of site of PDAC.Ova implantation or between parental PDAC and PDAC.Ova tumors growing at the same anatomic site (Supplementary Fig. S5).

Given the central function for Batf3-dependent cDC1s in antigen cross-presentation (46) and the CD40 pathway for CD4⁺ T cell–mediated DC licensing (28, 47), we evaluated T-cell priming in Batf3^−/− and CD40^−/− mice. Compared with robust expansion of Ova tet⁺ T cells in WT mice bearing subcutaneous PDAC.Ova, the absence of cDC1s nearly completely blocked early CD8⁺ T-cell priming (Fig. 4A). In CD40^−/− hosts, tumor infiltration by Ova-specific CD8⁺ T cells was significantly reduced but not eliminated (Fig. 4A). Similar experiments also confirmed the importance of cDC1s and CD40 in T-cell priming against tumors following intraperitoneal challenge (Fig. 4B). These data correlated with tumor outgrowth following subcutaneous or orthotopic challenge, as PDAC.Ova tumors grew in Batf3^−/− mice with kinetics similar to CD8⁺ T-cell–depleted mice (Fig. 4C–E). PDAC.Ova tumors also grew progressively in subsets of CD40^−/− mice, although the phenotype appeared more substantial following orthotopic compared with subcutaneous implantation (Fig. 4C–E). Following orthotopic PDAC.Ova injection, there was minimal Ova tet⁺ T-cell expansion or tumor infiltration at later time points in CD40^−/− mice with progressively growing tumors, comparable with Ova-tolerant hosts (Supplementary Fig. S4B). Accumulation of Ova tet⁺ T cells did occur, however, in the spleen and late orthotopic PDAC.Ova tumors in a subset of Batf3^−/− mice (Supplementary Fig. S4B and S4C), indicating that other APCs could prime CD8⁺ T cells in the absence of cDC1s; however, T cells lacked an effector phenotype (Supplementary Fig. S4E and S4F).

Prior work has suggested that cDC production of CXCR3 chemokines is required for optimal T-cell infiltration (48, 49) and that NK cells promote cDC recruitment to the TME (50). We therefore investigated the growth of subcutaneous and orthotopic PDAC.Ova in NK-cell–depleted WT mice or in CXCR3^−/− hosts. In both cases, we found that tumor rejection and survival was similar to WT mice (Fig. 4F–H) and that Ova tet⁺ CD8⁺ T-cell expansion or tumor infiltration was unchanged in the absence of NK cells or host CXCR3 (Fig. 4A). We confirmed effective depletion of NK cells in the TME and observed TIL profiles similar to WT mice, with increased CD8⁺ T-cell infiltration and reduced myeloid populations (Supplementary Fig. S3A–S3D). Growth of PDAC.Ova in lymphotoxin alpha-deficient (Lta^−/−) mice, which lack secondary lymphoid tissues (51), confirmed the importance of lymph nodes for immune priming and PDAC.Ova rejection (Fig. 4C–E). These data therefore indicate that the essential function of cDC1s relates largely to T-cell priming rather than lymphocyte recruitment and that antitumor immunity in this model occurs independently of NK cells.

We also asked whether CD4⁺ T-cell help was required for immune responses to subcutaneous and orthotopic PDAC.Ova tumors. Cohorts of WT mice were challenged with low-dose PDAC.Ova via subcutaneous or orthotopic injection and treated with control Ig or depleting anti-CD4 or anti-CD8. Rejection of subcutaneous PDAC.Ova tumors was dependent on CD8⁺ T cells but independent of CD4⁺ cells (Supplementary Fig. S6A and S6B), consistent with our prior results that also included high-dose subcutaneous challenge (41). In contrast, rejection of orthotopic PDAC.Ova tumors after low-dose challenge was completely dependent on CD4⁺ cells (Supplementary Fig. S6C and S6D), and in a subset of animals, delayed tumor escape was observed in the absence of CD4⁺ T-cell help. These findings reveal distinct immunologic requirements for tumor control at different anatomic sites and suggest that a more coordinated immune response is required within the pancreatic TME, although the precise contribution of CD4⁺ T cells in the setting of orthotopic PDAC.Ova tumors remains unclear.

We next investigated whether prior T-cell priming could protect mice against subsequent orthotopic or intraperitoneal PDAC.Ova implantation. Mice that had previously rejected PDAC.Ova cells via subcutaneous injection displayed complete protection against high-dose orthotopic or intraperitoneal tumor rechallenge (Fig. 5A–C), whereas PDAC.Ova grew in the majority of naïve WT mice. The timing of subcutaneous priming relative to intraperitoneal injection was important, as neither concurrent subcutaneous and intraperitoneal injection nor delayed subcutaneous injection (1 week after intraperitoneal injection) were protective (Fig. 5D); yet, subcutaneous injection a week prior to intraperitoneal challenge prevented lethality (Fig. 5D). In concurrently challenged mice, subcutaneous tumors were eliminated in most mice, despite progressive growth of PDAC.Ova within the peritoneum (Fig. 5E). When we assessed Ova-specific CD8⁺ T-cell expansion in the spleen at the height of the response on day 9, we found that priming in the subcutaneous and intraperitoneal concurrently injected mice was slightly increased over intraperitoneal alone but not significantly (Fig. 5F). Similarly, there was a trend toward reduced tumor infiltration by Ova tet⁺ T cells into the subcutaneous tumors of concurrently injected mice (Fig. 5G). These data indicate that the presence of intraperitoneal PDAC.Ova tumors inhibited protective antitumor priming with subcutaneous injection, whereas in concurrently injected animals, perhaps a lower threshold for immune reactivity still permitted tumor control at the subcutaneous site.

We hypothesized that augmenting APC function through CD40 agonist administration might enhance T-cell priming and restore antitumor immunity against intrapancreatic and intraperitoneal PDAC.Ova. We, therefore, challenged mice with orthotopic PDAC.Ova, followed by treatment on day 7 with a single dose of either control Ig or anti-CD40 agonist. Although tumors grew progressively in most control-treated mice, CD40 agonist–treated mice uniformly rejected this tumor challenge (Fig. 6A and B). We confirmed that the CD40 agonist was acting on host cells because its effect was lost in CD40^−/− mice (Fig. 6A and B). We established that T cells were required for the efficacy of anti-CD40, as T-cell depletion abrogated its protective effect (Fig. 6A and B). Similar results were also obtained when mice were challenged with PDAC.Ova tumor cells via intraperitoneal implantation (Fig. 6C). The efficacy of CD40 agonist monotherapy required the presence of neoantigen, as anti-CD40 treatment of WT mice injected with PDAC cells either had no effect (intraperitoneal challenge) or led to a modest but significant delay in tumor growth but no regressions (orthotopic challenge; Supplementary Fig. S7).

We subsequently explored whether enhancing T-cell function through blockade of PD-1 or CTLA-4 inhibitory receptors could prevent PDAC.Ova escape. Mice challenged with orthotopic PDAC.Ova were treated at day 12 with control Ig, single-dose CD40 agonist, or serial anti–PD-1 (three doses, every 3 days). Whereas the CD40 agonist was protective and induced rejection of intrapancreatic tumors, we observed poor efficacy of PD-1 blockade (Fig. 6D and E). We obtained similar results upon treatment of intraperitoneal PDAC.Ova tumors with control Ig, anti-CD40, or anti–PD-1 (Fig. 6F). Orthotopic PDAC.Ova tumors harvested from anti–PD-1–treated mice showed no evidence of neoantigen or MHC class I loss as a mechanism of escape, and PD-L1 expression was similar (Supplementary Fig. S8). If mice challenged with intraperitoneal PDAC.Ova were instead treated with serial anti–CTLA-4 starting at days 6 to 7 (three doses, every 3 days), we observed partial protection (Fig. 6G), consistent with the capacity of CTLA-4 blockade to also promote T-cell expansion (7). However, anti–CTLA-4 treatment displayed significantly less efficacy than the single-dose of CD40 agonist, underscoring the importance of APC activation through the CD40 pathway. These findings demonstrate that the immunosuppressive TME within the pancreas and peritoneum can be overcome by enhancing CD40 signaling, yet blockade of T-cell inhibitory receptors alone was not sufficient to prevent PDAC.Ova escape.

To understand the mechanism of action of CD40 agonism, we measured tumor-specific CD8⁺ T cells in the spleens of mice challenged with intraperitoneal PDAC.Ova tumor cells. Compared with control Ig–treated mice in which T-cell priming remained poor, we observed a significant expansion of Ova tet⁺ T cells following CD40 agonist treatment (Fig. 6H), consistent with enhanced CD8⁺ T-cell priming. Although CD40 agonist can stimulate various APC populations, we observed that cDC1s were required for this CD40 agonist–induced priming (Fig. 6H). Protection by CD40 agonist monotherapy following either intraperitoneal or orthotopic PDAC.Ova tumor challenge was abrogated in Batf3^−/− mice lacking cDC1s (Fig. 6A–C). These results demonstrate that CD40 agonist treatment acts through cDC1s to promote T-cell immunity and prevent escape of antigenic tumors within the pancreas and peritoneum.

To test whether the actions of CD40 agonism on cDC1s might also broaden the specificity of antitumor T-cell responses, we utilized mixing experiments with Ova⁺ and Ova⁻ tumor cells. When WT mice were challenged subcutaneously with an 80:20 mixture of PDAC.Ova and parental PDAC C10, respectively, and treated on day 7 with control Ig, we observed progressive tumor growth (Fig. 7A and B), although the kinetics of growth were mildly delayed compared with T-cell–depleted mice, which we previously showed is associated with a T-cell–dependent elimination of the Ova⁺ tumor fraction (41). Yet, if WT mice were instead treated with the CD40 agonist on day 7, we found that mixed tumors were controlled, and in many cases rejected completely, an effect which required the actions of both T cells and cDC1s (Fig. 7A and B). Immune elimination of Ova⁻ tumor cells was antigen-specific and not explained by a bystander effect within the TME because WT mice challenged with Ova⁺ and Ova⁻ tumor cells on opposite flanks displayed a similar phenotype (Fig. 7C). These results thus directly establish the ability of CD40 agonism to broaden the antitumor T-cell repertoire by inducing epitope spread within antigenically heterogeneous tumors. When we challenged mice with subcutaneous parental PDAC following an initial priming with PDAC.Ova plus either control Ig or CD40 agonist treatment on day 7, we observed substantial differences in tumor outgrowth (Fig. 7D), underscoring the development of immune memory to shared (non-Ova) antigens and the enhancement of T-cell memory generation by CD40 agonist treatment.

Identifying key determinants of tumor immunogenicity is critical to understanding immune escape and immunotherapy resistance. Although the TME of pancreatic cancer is presumed to be highly immunosuppressive, the extent to which such properties are orchestrated primarily by tumor cell–intrinsic biology versus the tissue microenvironment is difficult to ascertain. In this study, we used a clonal PDAC cell line with high antigenicity via enforced neoantigen expression to explore T-cell–mediated immune surveillance. PDAC.Ova tumor cells were immunogenic and readily eliminated following subcutaneous or intravenous challenge, yet we unexpectedly observed immune escape within specific tissues such as the pancreas and peritoneum. Tumor outgrowth was not due to neoantigen loss or tumor-intrinsic editing, but rather reflected diminished CD8⁺ T-cell priming and reduced immunogenicity of intraperitoneal and orthotopic PDAC.Ova tumors. Thus, there is differential immune surveillance at distinct host tissue sites. We additionally found qualitative differences in the antitumor response, including a disparate requirement for CD4⁺ T-cell help for control of orthotopic versus subcutaneous tumors and divergent growth of intraperitoneal and subcutaneous tumor cells after concurrent challenge.

Our findings add to emerging evidence that antitumor immunity is shaped by host tissue determinants in addition to tumor cell–intrinsic factors (29). Prior studies in autochthonous models of cancer formation have suggested differential immune surveillance of neoantigen-expressing tumors based on the tissue of origin. For example, in sarcomas initiated by oncogenic Kras and p53 loss, T-cell surveillance inhibits tumor development and leads to immune escape with neoantigen loss (6), but lung adenocarcinomas, initiated by similar driver mutations and harboring the same neoantigens, develop with only delayed kinetics and retain neoantigen expression (52). A study in the KPC model of pancreatic cancer shows that neoantigen expression in pancreatic epithelium accelerates premalignant lesion formation, PDAC development, and liver metastases, which was associated with the induction of a pathogenic Th17 CD4⁺ T-cell response (53). In contrast, expression of the same neoantigens during lung adenocarcinoma development elicits protective immunity with partial neoantigen loss (53).

Although autochthonous GEMMs have advantages over transplantable models, additional poorly defined genetic alterations are required for oncogenesis, and the activation (or inactivation) of divergent signaling pathways may be required for tumor formation at various tissue sites. Such differences could, in turn, contribute to the distinct immunobiology of Kras-driven tumors within different tissues. In the present study, we have eliminated such confounders by employing a clonal PDAC cell line to specifically interrogate the influence of the tissue microenvironment on T-cell surveillance of the same antigenic tumor. Limitations of this model system include reliance on a single dominant neoantigen, relative to the poorly characterized and likely variable number and strength of tumor antigens in humans, although in some cases responses appear to be dominated by a limited set of neoantigen-specific T cells (54). Additional limitations include those inherent in transplantable models, which require implantation site trauma and involve more accelerated tumor progression. Recognizing such limitations, our findings nevertheless reveal an important contribution of the host tissue microenvironment on immune surveillance, independent of tumor-intrinsic pathways. Potential relevance to mixed immunotherapy responses in human patients with cancer warrants further study, particularly given divergent tumor–immune profiles identified in different metastatic lesions within the same patient, independent of genetic heterogeneity (25, 55).

Mechanistically, we found that escape of PDAC.Ova tumors was associated with diminished priming of tumor-specific T cells, which required cDC1s, as well as CD40 signaling, but was independent of NK cells and the CXCR3–chemokine axis, in contrast to other tumor models (48–50). Consistent with a defect in T-cell priming, we demonstrated that either antecedent challenge with subcutaneous PDAC.Ova or a single dose of an anti-CD40 agonist prevented escape of intraperitoneal and orthotopic PDAC.Ova tumors. The CD40 agonist acted through cDC1s to augment cross-priming of CD8⁺ T cells, and through mixing experiments with Ova⁺ and Ova⁻ tumor cells, we directly established that CD40 agonism could also promote epitope spread. In contrast, enhancing T-cell function alone via blockade of inhibitory receptors was not sufficient to restore protective immunity. PDAC.Ova tumors developing after orthotopic or intraperitoneal injection, including tumors emerging with PD-1 blockade, retained neoantigen and MHC expression and showed no evidence of immunoediting.

Our findings differ from a prior study showing efficacy of PD-1/PD-L1 pathway blockade in neoantigen-expressing transplantable PDAC, with emergence of escape variants harboring MHC loss (56). Yet, they are consistent with other work finding no efficacy for checkpoint blockade in control of autochthonous Ova-expressing PDAC (53). This latter study also demonstrated decreased cDC infiltration and function in developing pancreatic tumors compared with lung tumors, which was unaffected by neoantigen expression and present in premalignant lesions, thus likely related to the tissue microenvironment (53). We have also previously shown that cDC dysfunction occurs early during the premalignant stages of pancreatic tumorigenesis in KPC mice, which is linked to increased cDC1 apoptosis in the setting of elevated IL6 (43). In this study, we did not find evidence of diminished cDC1 infiltration into intraperitoneal or orthotopic PDAC.Ova tumors compared with subcutaneous tumors; however, poor CD8⁺ T-cell priming and its rescue with CD40 agonist treatment was consistent with cDC1 dysfunction. Maturation marker expression on cDC1s was also largely unchanged, yet complex molecular programs (potentially dissociated from surface marker expression) likely distinguish immunogenic versus tolerogenic cDC1s, and characterizing functional subsets remains challenging (57, 58). We hypothesize that microenvironmental cues at different anatomic sites influence cDC1 functionality and further investigation is warranted.

In conclusion, our study provides several important, novel insights about the immunobiology of pancreatic cancer. We showed that immune escape of antigenic tumors can occur without immunoediting and depends on the tissue site, with outgrowth of neoantigen-expressing PDAC occurring in the pancreas and peritoneum, despite high immunogenicity after subcutaneous or intravenous challenge. Tumor outgrowth at such sites neither required tumor-intrinsic editing nor could be prevented by augmenting T-cell effector function through PD-1 or CTLA-4 checkpoint blockade. Instead, immune escape was associated with diminished CD8⁺ T-cell priming by cDC1s, without evidence of decreased DC infiltration or trafficking. Immunologic control at otherwise refractory sites was restored by enhancing cDC1 cross-presentation via CD40 agonist treatment, which promoted T-cell priming and broadened T-cell responses through epitope spread. Further understanding the host tissue determinants of cancer immunogenicity will thus be required to design improved combination immunotherapies for refractory solid tumors.

M.S. Diamond reports grants from NIH during the conduct of the study. R.H. Vonderheide reports a patent for cellular immunotherapy issued, licensed, and with royalties paid from Novartis, as well as royalties from Children's Hospital Boston for a licensed research-only mAb. No disclosures were reported by the other author.

M.S. Diamond: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. J.H. Lin: Formal analysis, investigation, methodology. R.H. Vonderheide: Conceptualization, resources, formal analysis, supervision, funding acquisition, writing–review and editing.

This work was supported by NIH grants K12 CA076931 (to M.S. Diamond), K08 CA241084 (to M.S. Diamond), and R01 CA229803 (to R.H. Vonderheide) and the Parker Institute for Cancer Immunotherapy (to R.H. Vonderheide). The authors wish to thank members of the laboratories of Robert Vonderheide and Dr. Ben Stanger for helpful discussions.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.