Pancreatic Epithelial IL17/IL17RA Signaling Drives B7-H4 Expression to Promote Tumorigenesis

Pancreatic cancer (PDAC) is predicted to become the second leading cause of cancer-related deaths (1). PDAC is usually detected at late stages and is highly aggressive from its inception, with cancer cells found in metastatic sites even before primary tumor macroscopic development (2). One of the hallmarks of PDAC is the early development of a dense immunosuppressive tumor microenvironment (TME), which is known to promote tumor initiation, maintenance, progression, and resistance to therapies (3, 4). The PDAC TME is established when the lowest grade premalignant lesions are formed (5, 6) and is characterized by a prominent fibro-cellular stroma composed of fibroblasts, extracellular matrix, immune cells, and vascular and lymphatic vessels, all surrounding the tumor epithelium (7). Understanding the early events circumventing immunosurveillance and allowing premalignant lesions to begin and progress may allow the design of novel immune interception strategies to prevent PDAC formation.

IL17 is a proinflammatory cytokine, mostly secreted by T Helper 17 (Th17) cells (8, 9), but also produced by γδ T cells (10), NK cells (ref. 11), lymphoid-tissue inducer (LTi)-like cells (12), invariant NK T (iNKT; ref. 13) and Group 3 innate lymphoid cells (ILC3; ref. 14). The IL17 family includes IL17A (IL17), IL17B, IL17C, IL17D, IL17E (also known as IL25) and IL17F (15), among which IL17 and IL17F can signal as homodimers or heterodimers through a heterodimeric receptor formed by IL17 receptor A (IL17RA) and IL17 receptor C (IL17RC) subunits. IL17RC is mainly expressed in nonhematopoietic cells, whereas IL17RA is ubiquitously expressed in both the epithelial and hematopoietic cell compartments (9, 16). IL17/IL17RA signaling plays an important physiological role in regulating the mucosal host defense against invading microorganisms (17), but it can also play pathogenic roles in several inflammatory or autoimmune conditions (18); furthermore, it has been described as pro-tumorigenic in several cancer types (19–27). Our group and other authors have reported that IL17-producing cells infiltrate the pancreas in the context of Kras activation and chronic inflammation, both of which are required drivers of premalignant lesions in PDAC (19, 28). Moreover, under such conditions, we reported that IL17 induces stemness by interacting with the oncogenic pancreatic epithelium, suggesting this may be the primary compartment mediating IL17 induction of tumorigenesis (20). Additionally, we found that IL17 signaling participates in NETosis, enhancing TME immunosuppression in established PDAC (29).

Transcriptomic analysis of sorted pancreatic intraepithelial neoplasia (PanIN) cells from mice exposed to IL17 blockade revealed that B7-H4 is regulated by IL17(19). B7-H4, also known as B7S1, B7x, or Vtcn1, is a member of the B7 family of coregulatory molecules that can inhibit the growth, cytokine secretion, and cytotoxicity of T cells, thereby modulating the immune response (30–32). B7-H4 is overexpressed in a broad spectrum of cancers and associated with worse prognosis (33). In pancreatic cancer, Qian and colleagues showed that B7-H4 knockdown by short interfering RNA (siRNA) in human cell lines inhibits cell proliferation, colony formation, and cell migration and increases apoptosis in vitro. Moreover, B7-H4 siRNA intratumoral injections in vivo inhibited the growth of L3.6p1 subcutaneous tumors (34).

Although it is known that IL17 drives tumorigenesis, particularly in the context of inflammation, the definitive cellular compartment utilized by IL17 and its molecular downstream mediators have not been fully identified. A better knowledge of the IL17 signaling pathway and its downstream mechanisms involved in pancreatic tumorigenesis, including potential downstream immunoregulatory targets, would ultimately help define novel strategies for PDAC immune interception. In this study, we used multiple transgenic and bone marrow chimeric animal models to reveal the cellular and molecular mechanisms triggered by IL17 to mediate pancreatic tumorigenesis. This knowledge provides the foundation and new targets for early pancreatic cancer interception.

All animal experiments were conducted in compliance with the NIH guidelines for animal research and were approved by the Institutional Animal Care and Use Committee of the University of Texas MD Anderson Cancer Center (MDACC). We used conditional P48:CRE;LSL-KRAS^G12D (KC) mice and P48:CreERT2;LSL-Kras^G12D (KCi) mice at 20, 30, and 40 weeks of age, both of which were previously described (35). The KC mice were bred with IL17RA^fl/fland B7-H4^KO mice to produce KC;IL17RA^fl/fl and KC;B7-H4^−/−, resulting in a triple transgenic mouse strain that does not express IL17RA (P48:CRE;LSL-KRASG12D;IL17RA^fl/fl) and B7-H4 (P48:CRE; LSL-KRASG12D; B7-H4^KO) in all the pancreas compartments. Mice of both sexes were utilized. IL17RA^fl/fl mice were purchased from The Jackson Laboratory, and B7-H4^KO mice were gifted to us by Dr. James P. Allison (University of Texas MDACC).

Adult mice (10 weeks of age) were given intraperitoneal injections of caerulein (Sigma-Aldrich; individual dose: 75 g/kg). Each week, mice received four injections, given 4 hours apart 3 days per week, for a total of 8 weeks.

Femurs and tibias were harvested from sacrificed donor mice [IL17RA^−/− or wild-type (WT)]. IL17RA^−/−8 were purchased from Amgen inc., and WT C57BL/6NTac mice were purchased from Taconic Bioscience. The bone marrow was flushed with PBS from bone canals. The flushed marrow was then filtered through a 70 μm cell strainer. Next, cells were centrifuged at 1,300 RPM for 5 minutes at 4°C. Next, cells were resuspended in sterile PBS at a concentration of 10,000,000 cells/150 µL. Recipient mice were irradiated with a total of 900 cGy and rested for 6 hours. Then, recipient mice were injected via tail-vain with 1 × 10⁷ bone marrow cells using a 27G syringe.

Mouse pancreases were fixed in 4% paraformaldehyde, embedded in paraffin and cut into 5 μm sections. Then, the tissue sections were stained with hematoxylin and eosin, Alcian Blue (Vector Laboratories) and Masson trichrome (Sigma-Aldrich) following the manufacturer protocols. To detect total tissue and quantify individual pancreas lesions, slides were scanned on Aperio CS2 (Leica Biosystems), and images were obtained at 10× magnification. The Image J software (36) was used to select and quantify individual areas. Tissues were quantified for the presence of normal tissue and/or epithelial lesions in a blinded manner by two independent investigators, and they were reidentified following quantification of all tissues. Epithelial lesions were identified as acinar-to-ductal metaplasia (ADM), early pancreatic intraepithelial neoplasias (PanIN) defined as PanIN-1A, or PanIN-1B or advanced PanIN defined as PanIN-2 and PanIN-3, as outlined previously (37, 38). Fibro-inflammatory stroma was defined as an abnormal tissue area devoid of epithelial lesions (Supplementary Fig. S1). The percentage of total tissue surface area occupied by each lesion was calculated by the Image J software for 5 to 10 random fields in each of the slides, and data were shown as mean % Total Surface Area ± SEM.

Paraformaldehyde-fixed paraffin-embedded mice pancreas tissue sections were deparaffinized and rehydrated, and then boiled for antigen retrieval with 1× citrate buffer, pH 6.0 (Sigma) for 15 minutes. Next to block endogenous peroxidases, we put the slides in 3% H₂O₂ for 15 minutes. Nonspecific epitopes were blocked with HyClone Bovine Serum Albumin (GE Healthcare Life Sciences) for 15 minutes. The sections were incubated overnight at 4°C with antibody (Supplementary Table S1). This was followed by an HRP-conjugated secondary antibody (Vector Laboratories, MP-7451) for 1 hour. Next, we used SignalStain DAB Substrate Kit (Cell Signaling) following the manufacturer’s instructions. Then, the slides were counterstained with hematoxylin (Dako) or methyl green (Vector Laboratories) and mounted in Acrytol Mounting Medium (Electron Microscopy Sciences). Slides were scanned on Aperio CS2 digital pathology slide scanner (Leica Biosystems) and visualized at 20× magnification.

Tissue sections were prepared as described in the Immunohistochemistry Section. We used anti-IL17-RA (clone H-168, Rabbit, Santa Cruz), anti-B7-H4 (R&D Systems Catalog # AF2154), anti-SMA (clone 1A4, Thermo Fisher Scientific), and anti-E-cadherin (Alexa Fluor 488, clone 36, BD Biosciences) for staining the tissue. The slides were mounted using ProLongTM Diamond Antifade Mountant with DAPI (Thermo Fisher, P36962) and visualized on Olympus IX73 Inverted microscope (Olympus Life Science) at 10× and 20× magnifications.

Sequential multiplex IHC and image acquisition were conducted as described by Tsujikawa and colleagues (39). Paraformaldehyde-fixed paraffin-embedded tissue sections were stained by hematoxylin, followed by whole-tissue scanning using Aperio ImageScope AT (Leica Biosystems) at 20×. After blocking peroxidase activity with 3% H₂O₂ for 15 minutes, antigen retrieval was performed in microwave BIOGENEX with one cycle (Cycle 1, 99°C × 15 minutes). Sequential IHC consisting of iterative cycles of staining, scanning, antibody, and AEC (horseradish peroxidase conjugated polymer and 3-amino-9-ethylcarbazole), and antibody stripping was performed (the sequence, information about, and concentration of antibodies used are in Supplementary Table S2). Image processing and analysis were done as previously described (39).

We got the cell locations given as coordinates by using the FCS express7software. We used the R package Spatstat to compute the L-values from Ripley’s K-function over various radii surrounding CK19⁺ cells to determine if secondary cells (CD8⁺, CD8⁺Granzb⁺, CD4⁺, CD4⁺FOXP3⁺, Ly6G⁺, or F4/80⁺ cells) are spatially clumped, dispersed, or randomly distributed around the CK19⁺ cells. Before running Spatstat, we generated matrices that contained the coordinates data for each cell type. Next, we used the ppp function to turn coordinate matrices into point patterns and combined them, with the Superimpose function required by Spatstat. The obtained L-values were utilized to establish whether secondary cells were grouped, dispersed, or dispersed randomly around neighboring CK19⁺ cells (40).

Total RNA was extracted from murine pancreatic tissue with RNeasy RNA isolation kit (Qiagen) following the manufacturer’s instructions. RNA (500 ng/sample) was converted to cDNA using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, 4368814). Quantitative RT-PCR was performed with Fast SYBR Green Master Mix on ViiA 7 Real-Time PCR System (Thermo Fisher) using PCR primers (Supplementary Table S3). Samples were run in triplicate, and the expression of the genes of interest was normalized to the expression of Gapdh in each sample and normalized to WT mouse pancreas of the same age and reported as 2^ΔΔCT.

LSL-KrasG12D/+, Trp53fl/+, and Pdx1-Cre (KPC) pancreatic cancer cells, which were derived from a spontaneous tumor from LSL-KrasG12D/+, Trp53fl/+, and Pdx1-Cre mice as previously described (41), were gifted to us by Dr. Anirban Maitra (University of Texas MDACC). Cell lines were authenticated by morphology. Cells were confirmed to be mycoplasma negative before each experiment by the Cytogenetics and cell authentication core at UT MDACC, and passage number did not exceed two to five post thawing. KPC cells were cultivated in Dulbecco’s Modified Eagle Medium (DMEM) with 4.5 g/L glucose (Corning) supplemented with 10% FBS (Sigma-Aldrich, St. Louis, MO) and 1% penicillin/streptomycin (HyClone Laboratories, Logan, UT) at 37°C and 5% CO₂ in a humidified atmosphere. We seeded 1 × 10⁵ KPC cells in a six-well plate in media supplemented with or without 10 ng/mL IL17A (R&D system, Minneapolis, MD) or IL6 (R&D system, Minneapolis, MD) for 6 days and collected the cells every day in triplicate for quantitative RT-PCRU0126 (Cell Signaling Technology, Danvers, MA) and BAY11-7082 (InvivoGen, San Diego, CA) were added to cell culture medium (10 and 5 μmol/L) to suppress cell signaling pathways after stimulation with recombinant IL17A and IL6 at concentration of 10 ng/mL and were collected in triplicate after 3 days for quantitative RT-PCR.

KPC cells were cultured in DMEM with L-glutamine (Gibco), antibiotics, and 10% FBS and treated with 10 ng/mL of murine recombinant IL17 for 3 days. For transcription inhibition, actinomycin D (Sigma-Aldrich) at 5 μg/mL was added at Day 3. The cells were collected for triplicated at 0, 2, 4, 6 hours after addition of actinomycin D for quantitative RT-PCR to plot the RNA decay.

Single-cell suspensions were prepared by mincing murine pancreas followed by digestion with Collagenase P (4 mg/mL, Roche Diagnostics), Dispase (4 mg/mL, Sigma-Aldrich), and DNAse (0.5 mg/mL, Sigma-Aldrich), and incubation for 5 to 10 minutes at 37°C in a shaker. Samples were then filtered through a 70 μm and then 40 μm strainer and stained with L/D aqua (Thermo Fisher) at room temperature for 10 minutes. Samples were resuspended in PBS with 0.004% bovine serum albumin and sorted for live cells using an Aria II sorter. Cell viability was measured using Trypan Blue. For single-cell RNA sequencing library preparation, samples were processed using the Chromium system (10× Genomics) to form Gel Beads in Emulsion. After capturing and barcoding the mRNAs, the reaction mixture was extracted from the Chromium instrument for reverse transcription. The resulting cDNA was then fragmented and amplified according to the 10× Genomics protocol. Finally, the libraries were purified, quantified, and sequenced on an Illumina NextSeq 550 with paired end reads.

The raw data were processed using cell ranger count (Cell Ranger v2.1.1; 10× Genomics) based on the mm10 mouse reference genome. Subsequent data analysis was done in R using the Seurat package v3.1 with default parameters. Dead cells were excluded by retaining cells with less than 20% mitochondrial reads. We assigned a cell cycle score using the CellCycleScoring function, and a cell cycle difference was calculated by subtracting the S phase score from the G2M score. We then used linear regression to scaled data on the counts and the cell cycle score difference. Principal component analysis was run with the RunPCA function and violin plots were then used to filter the data according to user-defined criteria. All tissue samples were batch-corrected through the R package Harmony. Cell clusters were identified via the FindNeighbors and FindClusters functions, using a resolution of 0.6 for all samples, and uniform manifold approximation and projection (UMAP) clustering algorithms were performed. Clusters were defined by user-defined criteria. Single-cell data are accessible through Gene Expression Omnibus repository under accession number GSE264054. We used four biological replicates, each for KC and KC; IL17RA^fl/fl at 30 weeks of age without technical replicates.

PDAC RNA sequencing data (FPKM_UQ) from the Cancer Genome Atlas Program were used. After excluding PNETs and non-PDAC samples, high and low gene expression levels were stratified based on average expression. Statistical analysis was performed using simple linear regression in Prism software (GraphPad Software, Inc., version 10).

We analyzed several previously published datasets for this study. The dataset by Carpenter and colleagues (GSE229413) includes 30 pancreata from healthy adult organ donors, where 12 had no premalignant lesions, and 18 had PanINs. We followed the analysis of the author and conducted a comparative expression analysis of Vtcn1 and Eomes. Additionally, the dataset by Steele and colleagues (GSE155698) included 16 PDAC samples and three adjacent normal pancreases. Similarly, following the author’s analysis, we proceeded to conduct a comparative expression analysis focusing on Vtcn1 and Eomes.

GraphPad Prism (GraphPad Software, Inc., version 10) was used to analyze the data. When comparing only two value sets. Statistical analysis was done using the Student t-test. When comparing data from three or more groups, statistical analysis was done using the one-way ANOVA analysis of variance. P values ≤ 0.05 (^∗), 0.01 (^∗∗), 0.001 (^∗∗∗), and 0.0001 (^∗∗∗∗) were regarded as statistically significant.

Single-cell data are accessible through Gene Expression Omnibus repository under accession number GSE264054. All other data generated in this study are available in the manuscript and its supplementary files or from the corresponding author upon reasonable request.

To determine the cellular compartment required for the IL17/IL17RA signaling mediating pancreatic tumorigenesis, we crossed P48-Cre;LSL-KRAS^G12D(KC) mice, which slowly develop pancreatic premalignant lesions (42), with IL17RA floxed mice (IL17RA^fl/fl) to generate transgenic mice with embryonic Kras^G12D activation and IL17RA deletion in the pancreatic epithelial compartment in the context of embryonic Kras activation (KC;IL17RA^fl/fl; Fig. 1A). We first confirmed epithelial P48-Cre mediated deletion of IL17RA through immunofluorescence, observing absence of IL17RA expression in pancreatic epithelial lesions from KC;IL17RA^fl/fl mice (Fig. 1B). Histopathological analysis revealed that pancreatic normal area was 5 times increased in KC;IL17RA^fl/fl mice compared with KC mice (48.14% ± 9.41% vs. 8.89% ± 3.32%; P = 0.0010), with significant reduction in the area occupied by fibro-inflammatory stroma (37.07% ± 7.42% vs. 55.89% ± 3.91%; P = 0.0377), early PanINs (8.44% vs. 19.19%; P = 0.0032), and advanced PanINs (3.80% ± 1.03% vs. 12.83% ± 1.43%; P < 0.001; Fig. 1C–E; Supplementary Fig. S2A). Although KC;IL17RAfl/fl mice developed fewer lesions, there was no evidence that proliferation was altered in the lesions that did develop based on Ki67 expression (Supplementary Fig. S3A). However, there was an increase in apoptosis in the lesions of the KC;IL17RAfl/fl mice at 30 weeks of age (Supplementary Fig. S3B). These results suggest IL17/IL17RA is required in the oncogenic epithelium compartment for the induction of early pancreatic tumorigenesis.

We then examined the role of IL17RA in the immune compartment by irradiating and transplanting KC mice with IL17RA deficient (IL17RA^−/−) hematopoietic cells obtained from the bone marrow of IL17RA global knock-out mice. The resulting animals had pancreatic Kras^G12D activation along with IL17RA deletion only in the hematopoietic component [KC+IL17-RA^−/− bone marrow (BM)], and we compared them with irradiated KC mice reconstituted with bone marrow from WT mice (KC+IL17-RA^+/+BM; Fig. 1F). IL17RA deletion was confirmed by sorting CD45⁺ cells from circulating blood and quantifying IL17RA expression by quantitative PCR (Fig. 1G). We found no histopathological differences between groups (Fig. 1H–J), suggesting IL17/IL17RA signaling in the immune compartment is not required during pancreatic tumorigenesis. We also found no difference in various markers of macrophages and T cells (Supplementary Fig. S4). Next, as Kras was activated embryonically in this model, to ensure that the phenotype was not influenced by the presence of lesions that had already developed at the time of irradiation, we performed another experiment in which the same donor BM cells (IL17-RA^−/− or IL17-RA^+/+) were transplanted into an inducible mouse model in which Kras is activated in the adult pancreas by cre-recombination induced upon tamoxifen injections (KC^P48;CreERT2; Supplementary Fig. S2B). When we euthanized animals at 20 weeks post-Kras activation, there were no differences found in oncogenic development between the two animal cohorts (Supplementary Fig. S2C–E). Altogether, our data suggests that IL17RA signaling in the pancreatic epithelium, and not in immune cells, is required during pancreatic tumorigenesis.

We determined the effect of pancreatic IL17RA depletion on the TME using multiplex immunohistochemistry, which allowed for basic pancreatic immune-profiling (Fig. 2A; ref. 39). KC;IL17RA^fl/fl mice presented significantly higher numbers of CD45⁺KI67⁺ cells, total CD8⁺ T cells (CD45⁺CD3⁺CD8⁺), as well as cytotoxic CD8⁺Gzmb⁺ T cells when compared with KC mice (Fig. 2B–D). Conversely, Tregs (CD45⁺CD3⁺CD4⁺FOXP3⁺), macrophages (CD45⁺F4/80⁺), and neutrophils (CD45⁺LY6G⁺) did not present significant differences between groups (Supplementary Fig. S5A–D). Next, using L-function, we spatially analyzed the distribution of immune cells around CK19⁺ premalignant cells (40). We found that the absence of epithelial IL17RA signaling resulted in redistribution of cytotoxic CD8⁺T cells (CD45⁺CD3⁺CD8⁺Gzmb⁺), which were more densely populated around epithelium (Fig. 2E) than in KC control mice. Taken together, these data reveal that IL17/IL17RA signaling in the pancreatic epithelium participates in reshaping the composition and spatial distribution of the immune cells infiltrating the pancreas during early tumorigenesis.

To gain insights into IL17RA-dependent transcriptional changes in the pancreatic compartment, we performed 10× Genomics Chromium droplet single-cell RNA sequencing (scRNA-seq) on pancreas from KC (four mice, 36,897 cells) and KC;IL17RA^fl/lf (four mice, 16,705 cells) animals. After UMAP dimensionality reduction, graph-based cell clustering revealed 31 clusters that were allocated to 13 cell populations that included both immune and epithelial cells based on previously published gene signatures (Fig. 3A; Supplementary Fig. S6A; refs. 43–45). Using these data, we identified that the primary cell types responsible for the production on IL17 were natural killer T (NKT) cells and CD4⁺ T cells. Within the CD4⁺ T cells, Th1 cells had the highest expression levels, followed by regulatory T cells (Tregs; Supplementary Fig. S7A and S7B). Gene Ontology (GO) analysis in the epithelial clusters revealed IL17RA-dependent changes in immune pathways, among which top pathways included regulation of immune responses, T-cell activation, adaptive immune responses, and regulation of T-cell activation (Fig. 3B). When we looked at individual epithelial clusters, we found that although Epithelial Cluster 2 expressed only Krt19, Epithelial Cluster 3 expressed both Cp1 and Krt19 markers. Cp1 acts as a marker for acinar cells, whereas Krt19 indicates the presence of PanINs. Their presence in Epithelial Cluster 3 suggests this represents a transitional state to premalignant lesions. In accordance with this, cells from Epithelial Cluster 3 had increased signaling of most RAS signature genes, including HRas, NRas, and MAPK pathway genes (Fig. 3C). These lesions undergo modulation of several immune pathways, such as antigen receptor-mediated signaling pathway, immune response-activation cell surface receptor signaling pathway, T-cell receptor signaling pathway, positive regulation of lymphocyte activation, and others as analyzed by Gene Set Enrichment Analysis (GSEA; Fig. 3D). To further investigate effects of IL17RA depletion in Epithelial Cluster 3, we used a stemness bioinformatics tool (46) to examine the downregulated genes and found a significant number of enriched genes belonging to the murine embryonic stem cell signature, as well as neural, hematopoietic, and intestinal stem cell types, similar to data previously reported by our group in sorted PanINs (Fig. 3E; Supplementary Fig. S6B; ref. 47). All together, these findings suggest that IL17/IL17RA signaling in the pancreatic epithelium, mostly cells expressing both Cp1 and Krt19, negatively affects immune surveillance by decreasing recruitment of activated CD8⁺ T cells.

To dissect the genes driving immune activation pathways in the pancreatic epithelium of KC;IL17RA^fl/fl mice, we examined the genes specifically downregulated within Epithelial Cluster 3 and compared these data against two gene datasets from our laboratory, one containing genes that were upregulated in KPC cells treated in vitro with IL17 and the other containing genes downregulated in premalignant lesions upon in vivo IL17 inhibition in KC^iMist1 mice (19). We found four unique genes (Ceacam1, GabrP, Liph, Vtcn1) overlapping between all three IL17-dependent datasets, but only one was involved in the immune regulated pathways: Vtcn1 (Fig. 4A). Vtcn1 encodes B7-H4. To validate that IL17 regulates Vtcn1, we performed quantitative RT-PCR for Vtcn1 (protein: B7-H4) and its related family member Cd276 (protein: B7-H4) in both KC and KC; IL17RA^fl/fl mice and found Vtcn1 was more highly expressed at 30 weeks in KC versus KC; IL17RA^fl/fl mice, whereas no differences were found in Cd276 expression (Fig. 4B). B7-H4 expression was confirmed by IHC in PanINs of KC mice as early as 20 weeks, and significantly decreased in PanINs from KC;IL17RA^fl/fl mice (12.33% ± 1.52% vs. 7.69% ± 0.82%, P = 0.0218) vs control (Fig. 4C–E). This difference in B7-H4 expression was not seen in KC mice receiving BM transplantation from IL17RA^−/− mice versus IL17RA^+/+mice following irradiation (Fig. 4F), further suggesting the importance of the pancreatic epithelial cells, rather than the immune compartment, in the effect of IL17 signaling on tumorigenesis.

It is well known that B7-H4 can inhibit CD8⁺ T cells and stimulate T-cell exhaustion by upregulating the transcription factor Eomesodermin (Eomes; refs. 48, 49). In the context of CD8⁺ T cells, increased expression of Eomes drives T-cell exhaustion rather than favoring the development of effector functions seen under conditions of basal expression levels. Li and colleagues (49) have proposed that this is achieved by direct binding of Eomes to regulatory genomic loci or indirect regulation to induce an exhausted fate in chronically stimulated T cells. Based on the transcriptional differences found in CD8⁺ T cells after deletion of IL17RA in the pancreatic compartment, we explored changes in the T-cell population upon IL17RA depletion in the pancreatic epithelium by further characterizing our scRNA-seq data. We identified eight T-cell clusters and observed a significantly decreased percentage of CD8⁺ T cells (28.57% vs. 47.88%) and exhausted CD8⁺ T cells (6.29% vs. 2.82%) in KC compared with KC; IL17RA^fl/fl mice. Moreover, we found mild nonsignificant increases in Th17 cells (9.99% vs. 4.71%) and Th1 cells (4.50% vs. 1.88%) infiltrating pancreas of KC mice, and other cell subtypes were of similar relative abundance between both groups (Supplementary Fig. S8A–C). Using our scRNA-seq data, we analyzed expression of various CD8⁺ T-cell exhaustion genes in both CD8⁺ T-cell clusters and found significantly lower expression of Eomes, transcription factor 7 (TCF-7), granzyme K, and CD28 in CD8⁺ T cells from KC;IL17RA^fl/f mice versus KC mice (Fig. 4G and H; Supplementary Fig. S9A). This decrease in Eomes expression was also seen by IHC staining in pancreatic tissue, whereas other markers of T-cell exhaustion remained unchanged (Supplementary Fig. S9B–I). We next analyzed CD8⁺ T cells from mice lacking epithelial IL17RA signaling by GO and observed changes in pathways including lymphocyte activation, T-cell activation, adaptive immune response, lymphocyte differentiation, and T-cell differentiation and cell killing, among others (Fig. 4I). We also analyzed other lymphoid- and myeloid-cell populations and found no significant differences between KC and KC; IL17RAfl/fl mice (Supplementary Fig. S8D and S8E).

Next, we explored a previously published human scRNA-seq dataset, which includes 30 pancreata from healthy adult organ donors where 12 of them have no premaligant lesions and 18 have PanINS (50), as revealed from histopathologic analysis. We found that Vtcn1 expression was significantly higher in the Epithelial Cluster from PanINs lessions versus normal (Fig. 4J). Moreover, we used another human scRNA-seq dataset (51), which includes 16 PDAC samples and three adjacent normal pancreases, and found epithelial Vtcn1 and CD8⁺ T-cell Eomes expression is significantly enhanced in PDAC compared with adjacent normal pancreas samples (Fig. 4K). Furthermore, we examined 170 PDAC samples from TCGA database and found that Eomes expression is positively correlated with the expression of IL17RA and CD8a in patients with PDAC (Fig. 4L and M).

The KPC cell line is a murine pancreatic adenocarcinoma cell line frequently used as a preclinical model for studying pancreatic cancer because of its origins from the Kras^G12D, Trp53^R172H, Pdx-1-Cre (KPC) mouse (52). Its genetic resemblance to human pancreatic cancer makes it a relevant tool for studying various aspects of the disease. In the next experiments, we used the KPC cell line to test how IL17 can induce B7-H4 expression in an in vitro setting. We turned our attention to IL6, which is considered an intermediate factor for IL17 (53), and explored whether the regulation of B7-H4 was directly mediated by IL17 or indirectly via IL6. We exposed KPC cells to media supplemented with IL17 or IL6 for 6 days and found Vtcn1 upregulation only after IL17 treatment, with a peak of expression at Day 6 (Supplementary Fig. S10A). These results suggest that IL17 directly regulates B7-H4. IL17 is a known activator of the transcription factor NF-κB and the MAPK/ERK pathway (9, 54). Thus, to explore the possible mechanisms by which IL17 modulates B7-H4 expression in KPC cells, we blocked the NF-κB pathway with Bay 11-7082 and found no changes in Vtcn1 expression by quantitative RT-PCR after a 3-day treatment with the inhibitor (Supplementary Fig. S10B). Similarly, inhibition of the MAPK/ERK pathway by U0126 revealed no changes in Vtcn1 expression (Supplementary Fig. S10C). Furthermore, as mRNA stabilization is another way in which IL17 can activate posttranscriptional pathways (55), we tested the ability of IL17 to prolong the half-life of Vtcn1 transcript. We stimulated KPC cells for 3 days with recombinant IL17 to induce Vtcn1 expression, added actinomycin D to inhibit transcription at Day 3 and monitored Vtcn1 mRNA levels at various timepoints for up to 6 hours. Results were compared with actinomycin D and IL17 alone treatments. Although actinomycin D reduced the mRNA expression of Vtcn1 compared with IL17 alone, we found that co-treatment of actinomycin D in combination with IL17 delayed Vtcn1 mRNA decay (Supplementary Fig. S10D). These findings suggest that IL17 directly mediates the stability of the Vtcn1 mRNA. Furthermore, our findings demonstrate the impact of B7-H4 on CD8⁺ T-cell exhaustion, mediated by the induction of Eomes expression. We also observed Eomes expression in CD8⁺ T cells in tumors from patients with PDAC, emphasizing the clinical relevance of our research.

To define the role of B7-H4 in PDAC tumor initiation and progression, we crossed KC mice with B7-H4 global knock-out mice (B7-H4^−/−; Fig. 5A). B7-H4 deletion was verified in KC;B7-H4^−/− mice by immunohistochemistry (Supplementary Fig. S11A). Histological examination revealed a significant decrease in pancreatic surface area occupied by advanced PanIN in KC;B7-H4^−/− when compared with KC mice (P = 0.0429), whereas no differences were seen in other parameters (Fig. 5B–E; Supplementary Fig. S11B). Furthermore, we found KC; B7-H4^−/− mice had significantly higher pancreatic infiltration with CD8⁺ T cells and Granzyme B⁺ cells (Fig. 5F–I). Taken together, these results suggest B7-H4 plays a role in the development of pancreatic cancer, specifically in the transition between ADMs/early PanINs to advanced PanINs.

In this work, we show that epithelial IL17/IL17RA signaling promotes PDAC development. Our data show depletion of IL17RA in the oncogenic pancreas limits the formation of premalignant lesions, whereas its absence in the hematopoietic cell compartment is dispensable for tumor formation. These findings are in accordance with studies reporting the role of IL17RA in epithelial cells in other diseases, as it has been shown that, although epithelial IL17RA signaling is required in mice to protect them from oral candidiasis (56), it can also worsen antibody-mediated glomerulonephritis when acting through renal tubular epithelial cells (57). Furthermore, protumorigenic roles have been described for epithelial IL17/IL17RA signaling in both colon tumors and skin papilloma (23, 58).

In addition, our results indicate that IL17/IL17RA signaling in the pancreatic epithelium reshapes the immune TME and impacts CD8⁺ T cells during early tumorigenesis. Previous data has shown that loss of IL17/IL17RA signaling increases efficacy of checkpoint inhibitors in mice orthotopically implanted with KPC cells and that this effect is CD8⁺ T-cell dependent (29). In addition to affecting CD8⁺ T-cell populations, our data suggests that alterations to IL17/IL17RA signaling may also disrupt Th1 and Th17 cell frequencies, although the results did not reach significance. Nevertheless, we cannot completely rule out relative contributions of these cell populations in the observed phenotype. When we examined the transcriptional changes in the pancreatic compartment caused by the absence of IL17RA, we discovered IL17RA deficiency associates with several alterations in immunological pathways, among which we identified an increase in the recruitment of cytotoxic CD8⁺ T cells and neutrophils in the vicinity of premalignant CK19⁺ cells. These results are consistent with our previously reported increase in activated CD8⁺ T cells after IL17 inhibition in PDAC tumors (29) but differ from studies reporting IL17 directly affects NK cells during fungal infections in systemic candidiasis (59); however, this inconsistency may be because of the difference in mouse models.

Moreover, we report B7-H4 promotes the development of advanced PanINs and is more abundant in KC than in KC;IL17RA^fl/fl premalignant lesions, indicating B7-H4 expression in the pancreatic epithelium is regulated by IL17/IL17RA signaling. This molecular mediator has been proposed to initiate dysfunction of activated tumor-infiltrating CD8⁺ T cells through the upregulation of Eomes, previously linked to increased exhausted fate in chronically stimulated T cells via direct binding to regulatory genomic loci or from indirect regulation (49). Our findings reveal increased expression of B7-H4 in the epithelial cluster of KC mice correlates with the expression of Eomes, which is also elevated on CD8⁺ T cells from KC when compared with KC;IL17RA^fl/fl mice. These results highlight both pancreatic tumor development and immunosuppression may be aided by B7-H4. However, it is worth mentioning that a recent study has shown that B7-H4 not only adversely affects CD8⁺ T cells, but also modulates regulatory CD4⁺ T cells (60). As we have not specifically studied CD4⁺ T cells in the TME in the absence of B7-H4, it is important to recognize that B7-H4 may also act through other immune cells besides cytotoxic T cells.

In addition to studying B7-H4 in the context of mouse models, we confirmed an upregulated expression of B7-H4 in the pancreas epithelium of human pancreatic lesions, along with the presence of Eomes expression in CD8⁺ T cells. Moreover, we found cytotoxic T cells from patients with PDAC exhibited an exhausted gene expression pattern, which becomes more abundant as disease progresses (61).

This study described an IL17-mediated regulation of B7-H4 and revealed its impact in pancreatic tumorigenesis. Further research should be undertaken to investigate the downstream molecular mechanisms of B7-H4 in PDAC tumors.

J.K. Kolls reports grants from NHLBI and grants from NIAID during the conduct of the study. J.P. Allison reports other support from Achelois, other support from Adaptive Biotechnologies, other support from Apricity, other support from BioAlta, other support from BioNTech, other support from Candel Therapeutics, other support from Dragonfly, other support from Earli, other support from Enable Medicine, other support from Hummingbird, other support from ImaginAb, other support from Lava Therapeutics, other support from Lytix, other support from Marker, other support from PBM Capital, other support from Phenomic AI, other support from Polaris Pharma, other support from Time Bioventures, other support from Trained Therapeutix, other support from Two Bear Capital, and other support from Venn Biosciences during the conduct of the study; other support from Achelois, other support from Adaptive Biotechnologies, other support from Apricity, other support from BioAtla, other support from BioNTech, other support from Candel Therapeutics, other support from Dragonfly, other support from Earli, other support from Enable Medcine, other support from Hummingbird, other support from ImaginAb, other support from Lava Therapeutics, other support from Lytix, other support from Marker, other support from PBM Capital, other support from Phenomic AI, other support from Polaris Pharma, other support from Time Bioventures, other support from Trained Therapeutix, other support from Two Bear Captial, and other support from Venn Biosciences outside the submitted work. G. Lozano reports grants from The University of Texas MDACC during the conduct of the study. F. McAllister reports personal fees from Neologics Bio outside the submitted work. No disclosures were reported by the other authors.

S. Castro-Pando: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. R.M. Howell: Investigation, visualization, methodology. L. Li: Investigation, visualization, methodology. M. Mascaro: Investigation, visualization, methodology. E.Y. Faraoni: Visualization, writing–original draft, writing–review and editing. O. Le Roux: Investigation, methodology. D. Romanin: Investigation, visualization, methodology, writing–original draft. V. Tahan: Investigation, methodology. E. Riquelme: Investigation, visualization, methodology. Y. Zhang: Investigation, visualization, methodology. J.K. Kolls: Conceptualization, resources. J.P. Allison: Conceptualization, resources. G. Lozano: Supervision, writing–original draft, writing–review and editing. S.J. Moghaddam: Conceptualization, investigation, visualization. F. McAllister: Conceptualization, resources, data curation, supervision, funding acquisition, validation, visualization, methodology, writing–original draft, project administration, writing–review and editing.