Olaparib and radiotherapy induce type I interferon and CD8⁺ T cell-dependent sensitization to immunotherapy in pancreatic cancer

Abstract

PARP inhibitors sensitize pancreatic ductal adenocarcinoma (PDAC) to radiation by inducing DNA damage and replication stress. These mechanisms also have the potential to enhance radiation-induced type I interferon (T1IFN) mediated anti-tumoral immune responses. We hypothesized that the PARP inhibitor olaparib would also potentiate radiation-induced T1IFN to promote anti-tumor immune responses and sensitization of otherwise resistant PDAC to immunotherapy. To test this hypothesis, we assessed the effects of olaparib and radiation on T1IFN production and sensitivity to αPD-L1 immunotherapy, as well as on the tumor microenvironment by single-cell RNA sequencing (scRNA-seq). We found that olaparib enhanced T1IFN production following radiation and had superior therapeutic efficacy in immune competent models. Olaparib and radiation treatment sensitized PDAC tumors to αPD-L1 resulting in decreased tumor burden and a 33% complete response rate. Combination treatment provided durable immune responses as shown by tumor rejection upon tumor rechallenge of previously cured mice. Furthermore, scRNA-seq analysis revealed that combination treatment induced an immunogenic tumor microenvironment, characterized by interferon responses in both PDAC and myeloid cell populations, macrophage polarization, and increased CD8⁺ terminal effector T cell frequency and activity, findings confirmed by IHC and flow cytometry. Furthermore, CD8⁺ T cells and T1IFN signaling were required for therapeutic efficacy as host depletion of CD8⁺ T cells or the T1IFN receptor diminished treatment responses. Overall, our results indicate that olaparib enhances radiation-induced T1IFN-mediated immune signaling and subsequently an adaptive immune response thus sensitizing pancreatic cancer to αPD-L1 therapy, supporting an ongoing clinical trial of this therapy in patients with PDAC.

INTRODUCTION

In an effort to further improve radiotherapy efficacy in pancreatic cancer, recent trials have combined immunotherapy with radiotherapy in resectable, borderline resectable, and metastatic pancreatic cancer patients. These trials have demonstrated safety but limited benefit to overall and progression free survival (1–4). Thus, there is a critical need to develop novel therapeutic combinations that when combined with immunotherapy improve patient outcomes.

Radiation induces DNA damage that, if unrepaired, leads to lethal DNA double strand breaks and tumor cell death. In addition, radiation-induced DNA damage promotes indirect tumor cell killing via the activation of a type I interferon (T1IFN) anti-tumor immune response (5). Radiation induces cytosolic DNA arising from the release of damaged nuclear DNA as small double stranded DNA fragments or acentric chromosome fragments known as micronuclei (5,6). Free cytosolic DNA in turn activates pattern recognition receptor pathways such as cGAS/STING and POLIII/RIG-I/MAVS that stimulate innate immune signaling and consequent T1IFN expression (7,8). Once secreted T1IFN interacts with its receptor, IFNAR1, to activate JAK/STAT1 signaling resulting in an interferon stimulated gene response (9). This pathway is commonly suppressed in cancers to avoid immune surveillance and may contribute to immunotherapy resistance (10). T1IFN is a key regulator of the immune system, having roles in both innate and adaptive immunity. Immune signaling mediated by T1IFN has broad immune consequences including antigen presentation, chemokine production, activation of cytotoxic CD8⁺ T cells, and PD-L1 expression (11). In some cancers, synergy between radiation and immunotherapy is dependent on T1IFN-mediated immune signaling (12). Thus, this signaling pathway is a crucial connection between radiation therapy and tumor immune surveillance.

Our prior work demonstrated the preclinical and clinical benefit of DNA damage response inhibition in combination with chemoradiation in locally advanced pancreatic cancer (13,14). Taken together these studies motivate combination approaches of immunotherapy and DNA damage response inhibition to maximize radiotherapy efficacy and overcome the treatment resistance of pancreatic cancer. Recent studies from our group and others have demonstrated that inhibitors of the DNA damage response such as those targeting DNA-PK and ATM potentiate the radiation-induced T1IFN response and sensitize otherwise resistant pancreatic cancer to immunotherapy (15–18). PARP is a critical mediator of the DNA damage response to radiation (19). PARP inhibitors, such as olaparib, have unique biology in their ability to prevent repair of radiation-induced DNA damage and induce DNA damage by the formation of PARP1-DNA complexes (known as PARP1 trapping) (20,21). Our prior study demonstrated that PARP inhibition by olaparib synergized with radiotherapy through the formation of a PARP1-DNA complex leading to replication stress, and increased persistent DNA double-strand breaks independent of BRCA2 and homologous recombination repair status (22).

Therefore, in this study we sought to test the hypothesis that olaparib enhances radiation-induced T1IFN production and immune signaling resulting in sensitization of otherwise resistant pancreatic cancer to immunotherapy. We began by assessingT1IFN production and its downstream immune signaling consequences including interferon stimulated gene expression, PD-L1 surface expression, and MHC-I surface expression following treatment with olaparib and radiation. These studies led us to test a novel combination treatment regimen consisting of olaparib, radiation, and αPD-L1 for the treatment of pancreatic cancer in two independent syngeneic immunocompetent mouse models. Using scRNA-seq, we characterized the tumor microenvironment in pancreatic tumors undergoing treatment and identified important immune compartments potentially responsible for therapeutic efficacy. The necessity of these immune compartments was then confirmed with functional in vivo experiments in CD8⁺ T cell-depleted or Ifnar1^−/− mice. Taken together our findings show that olaparib improves the efficacy of radiation and immunotherapy warranting further investigation into the clinical benefit of combination therapy with olaparib, radiation, and immunotherapy.

MATERIALS AND METHODS

Cell lines and drug solutions

Panc-1 (RRID:CVCL_0480), MiaPaCa2 (RRID:CVCL_0428), and AsPC-1 (RRID:CVCL_0152) cells were obtained from and authenticated by the ATCC (RRID:SCR_001672). mT4 and KPC2 (mouse pancreatic cell lines, were generously provided by Drs. D. Tuveson (Cold Spring Harbor, NY) and M. Pasca di Magliano (University of Michigan, Ann Arbor, MI), respectively (23,24). Cells were grown in either RPMI-1640 (MiaPaCa2, AsPC-1, and KPC2; Invitrogen, RRID:SCR_008452) or DMEM (mT4; Invitrogen, RRID:SCR_008452) medium supplemented with 10% FBS and 1% Gibco Penicillin/Streptomycin (Invitrogen, RRID:SCR_008452). Olaparib was synthesized and provided by AstraZeneca. For in vitro studies, olaparib was dissolved in dimethyl sulfoxide (Sigma, RRID:SCR_008988) and stored in aliquots at −20°C. For in vivo experiments olaparib was dissolved in dimethyl sulfoxide (final concentration 10%) and then diluted in 10% 2-hydroxypropyl-β-cyclodextrin (Cayman Chemicals, RRID:SCR_008945) and stored at 4°C for up to 3 days. The mouse PD-L1 blocking antibody (10F.9G2, Cat#BE0101, RRID: AB_10949073) and IgG2b isotype control (LFT-2, Cat#BE0090, RRID: AB_1107780) were purchased from BioXCell. The mouse CD8a blocking antibody (2.43, Cat#BP0061, RRID: AB_1125541) and IgG2b isotype control (LFT-2, Cat#BP0090, RRID: AB_1107780) were purchased from BioXCell. For in vitro experiments using Ifnar1 blocking antibody the InVivoMab anti-mouse IFNAR-1 antibody (clone MAR1–5A3, Cat#BE0241, RRID: AB_2687723) and the relevant isotype control was purchased from BioXcell. Cell lines were frequently authenticated and were tested monthly for mycoplasma contamination. Cell lines were used for experiments between passages 4–10 after thawing.

Irradiation

Irradiations were performed using a Philips RT250 (Kimtron Medical, Woodbury, CT) at a dose rate of approximately 2Gy/minute at the University of Michigan Comprehensive Rogel Cancer Center Experimental Irradiation Shared Resource (RRID:SCR_025766). Dosimetry was performed using an ionization chamber connected to an electrometer system that is directly traceable to a National Institute of Standards and Technology calibration. When animal tumors were irradiated, animals were anesthetized with isoflurane and positioned under the collimator such that each subcutaneous tumor was at the center of a 2.4cm aperture, while the rest of the mouse was shielded from radiation.

IFNβ-GFP reporter assay

The Panc-1 IFNβ1-GFP reporter cells were established by transfecting Panc-1 cells with pLKO.1-hygro-IFNβ1-GFP reporter construct generously provided by Dr. R. Greenburg (University of Pennsylvania, Philadelphia, PA)(6) and were selected for with hygromycin (50 μg/mL). The established Panc-1 IFNβ1-GFP reporter cells were treated with olaparib (3 μM or 5 μM) one hour before radiation treatment (8 Gy) and were collected 72 hours after radiation treatment. GFP expression levels were quantified by fluorescence intensity which was analyzed via flow cytometry (BD Biosciences, BD LSRFortessa Cell Analyzer, RRID:SCR_018655). Data are shown as change in mean fluorescence intensity (MFI) relative to the control treatment group.

Quantitative RT-PCR

Flowing indicated treatments, cells were washed with PBS, trypsinized, and then collected. Cell pellets were washed once more with PBS before beginning RNA extraction. RNA extraction was conducted using the RNeasy Mini Kit (Qiagen, RRID:SCR_008539) and RNase-free DNase digestion (Qiagen, RRID:SCR_008539). Following extraction, RNA concentration and purity was measured using a Nanodrop 2000 spectrophotometer (Thermo Fisher Scientific, RRID:SCR_018042). RNA was converted into cDNA using the High-Capacity RNA-to-cDNA Kit (Applied Biosystems, RRID:SCR_005039). Relative gene expression levels were quantified by quantitative PCR (qPCR) using PowerUp SYBR Green Master Mix (Invitrogen, RRID:SCR_008452) run on a StepOnePlus^™ Real-Time PCR System (Thermo Fisher Scientific, RRID:SCR_008452). The fold change (ΔΔCt) method was used and normalized to Gapdh. The following qPCR primers were used: mouse Cxcl9 (forward): 5’-CCTAGTGATAAGGAATGCACGATG-3’, mouse Cxcl9 (reverse): 5’-CTAGGCAGGTTTGATCTCCGTTC-3’, mouse Cxcl10 (forward): 5’-CCTGCCCACGTGTTGAGAT-3’, mouse Cxcl10 (reverse): 5’-TGATGGTCTTAGATTCCGGATTC-3’, mouse Mx1 (forward): 5’-GGGGAGGAAATAGAGAAAATGAT-3’, mouse Mx1 (reverse): 5’-GTTTACAAAGGGCTTGCTTGCT-3’, mouse Ifnβ (forward): 5’-CCCTATGGAGATGACGGAGA-3’, mouse Ifnβ (reverse): 5’-CTGTCTGCTGGTGGAGTTCA-3’, mouse Gapdh (forward): 5’-TGGAGAAACCTGCCAAGTATGA-3’, mouse Gapdh (reverse): 5’-CTGTTGAAGTCGCAGGACACAA-3’, human CXCL9 (forward): 5’-GTGGTGTTCTTTTCCTCTTGGG-3’, human CXCL9 (reverse): 5’-ACAGCGACCCTTTCTCACTAC-3’, human CXCL10 (forward): 5’-CTCCAGTCTCAGCACCATGA-3’, human CXCL10 (reverse): 5’-GCTCCCCTCTGGTTTTAAGG-3’, human IFNβ (forward): 5’-CATTACCTGAAGGCCAAGGA-3’, human IFNβ (reverse): 5’-CAATTGTCCAGTCCCAGAGG-3’, human PDL1 (forward): 5’-TGGCATTTGCTGAACGCATTT-3’, human PDL1 (reverse): 5’-TGCAGCCAGGTCTAATTGTTTT-3’, human GAPDH (forward): 5’-ATGACATCAAGAAGGTGGTG-3’, human GAPDH (reverse): 5’-CATACCAGGAAATGAGCTTG-3’.

ELISA

Cells were plated in 12 well tissue culture plates 24 hours prior to treatment with olaparib and radiation. Following the indicated wait time media was collected from wells and centrifuged for 5 minutes at 1000rpm. Media was then plated in 150 μL doublets in 96 well plates. Cytokine quantification by ELISA was performed in the Rogel Cancer Center Immune Monitoring Core (RRID:SCR_025765) using DuoSets (R&D Systems, Minneapolis, MN, RRID:SCR_006140) using the manufacturer’s recommended protocol with minor variation. The following ELISA DuoSets were used: Mouse CXCL9/MIG DuoSet ELISA (Cat#DY492), Mouse CXCL10/IP-10/CRG-2 DuoSet ELISA (Cat#DY466), Mouse IFN-beta DuoSet ELISA (Cat#DY8234–05).

Flow cytometry analysis

For cell surface markers which were analyzed in in vitro experiments (PD-L1 and MHC-I) cells were treated with indicated treatment and collected 72 hours later via trypsinization. Cell pellets were washed with PBS, and then resuspended in 50–200 μL of PBS and incubated with the corresponding antibody for one hour at room temperature in the dark. The stained samples were then washed with PBS and resuspended in 300 μL PBS for flow cytometry (BD Bioscience, BD LSRFortessa Cell Analyzer, RRID:SCR_018655) analysis. Mean expression levels of either PD-L1 or MHC-I were quantified using FlowJo version 10.10 software (RRID:SCR_008520). Data are represented as change in MFI relative to control treatment for that experiment. The PE-labeled anti-human PD-L1 (clone 29E.2A3, Cat#329705, RRID:AB_940366) and anti-mouse PD-L1 (clone 10F.9G2, Cat#124307, RRID:AB_2073557) antibodies and their relative isotype controls were obtained from Biolegend (RRID:SCR_001134). The Alexa Flour 488-labeled anti-human MHC-I (clone IVA26, Cat#MA5–44137, RRID:AB_2913069) and FITC-labeled anti-mouse MHC-I (clone AF6–88.5.5.3, Cat#11–5958-82, RRID:AB_11149502) antibodies and their relative isotype controls were obtained from Thermo Fisher Scientific (RRID:SCR_008452).

Flow cytometry analysis of tumor samples from in vivo experiments was conducted on tumor tissues (approximately 0.5 g). Tumor tissues were minced and collected into 50 mL tubes containing 5–10 mL digestion buffer (1 mg/mL Collagenase I, 1 mg/mL collagenase IV and 0.15 mg/mL Dnase I in RPMI-1640 medium) and incubated at 37°C on a shaker (180 rpm) for 20–30 minutes. Digested samples were filtered through 70 μM cell strainers, washed with FACS buffer (2% FBS in PBS), and collected into 50 mL tubes. Samples were spun down (300 g for 5 minutes) and resuspended in 10 mL FACS buffer and then loaded on Ficoll buffer and spun at 800 g for 20 minutes without brake and with mild acceleration. The middle layer was collected, washed, spun down (300 g for 5 minutes) and the cell pellet was used for flow analysis. Cells were treated with Cell Stimulation Cocktail containing PMA and ionomycin (Biolegend, Cat#423302, RRID:SCR_001134), GolgiPlug^™ (1:1000; BD Biosciences, Cat#555029, RRID:SCR_013311) and GolgiStop^™ (1:1000; BD Biosciences, Cat#554724, RRID:SCR_013311) and incubated at 37°C for 4 hours. Cells were resuspended in FACS buffer and stained for surface markers and a cell viability stain for 1 hour in the dark. Following washing, cells were fixed and permeabilized using a fixation/permeabilization kit (Thermo Fisher Scientific, eBioscience^™, Cat#00–5223-56, RRID:SCR_008452) and incubated overnight at 4°C. Following washing, cells were stained for intercellular markers for one hour in the dark, mixed with counting beads, and analyzed by flow cytometry (BD Fortessa, BD LSRFortessa Cell Analyzer, RRID:SCR_018655) using FlowJo version 10.10 software (RRID:SCR_008520). The antibodies used for flow analysis are: Brilliant Violet 605^™ anti-mouse CD45 Antibody (1:100; Biolegend, clone 30-F11, Cat#103139, RRID:AB_2562341), FITC anti-mouse CD90.2 (Thy1.2) Antibody (1:500; Biolegend, clone 20-H12, Cat#105306, RRID:AB_313177), PE anti-mouse CD8a Antibody (1:100; Biolegend, clone 53–6.7, Cat#100708, RRID:AB_312747), IFN gamma Monoclonal Antibody, PerCP-Cyanine5.5 (1:50; Thermo Fisher, clone XMG1.2, Cat#45–7311-82, RRID:AB_1107020), TNF alpha Monocolonal Antibody, PE-Cyanine7 (1:100; Thermo Fisher, clone MP6-XT22, Cat#12–7321-82, RRID:AB_466199), FITC anti-mouse Ly-6G Antibody (1:300; Biolegend, clone 1A8, Cat#127606, RRID:AB_1236494), APC anti-mouse/human CD11b Antibody (1:500; Biolegend, clone M1/70, Cat#101212, RRID:AB_312795), Alexa Fluor^® 700 anti-mouse F4/80 Antibody (1:200; Biolegend, clone BM8, Cat#123130, RRID:AB_2293450), PE anti-mouse CD80 Antibody (1:50; Biolegend, clone 16–10A1, Cat#104708, RRID:AB_313129), Brilliant Violet 785^™ anti-mouse CD86 Antibody (1:50; Biolegend, clone GL-1, Cat#105043, RRID:AB_2566722), Brilliant Violet 421^™ anti-mouse I-A/I-E Antibody (1:100; Biolegend, clone M5/114.15.2, Cat#107632, RRID:AB_2650896), PE/Cyanine7 anti-mouse CD11c Antibody (1:50; Biolegend, clone N418, Cat#117318, RRID:AB_493568).

In vivo mouse models

In vivo experiments followed an approved protocol by the University of Michigan Committee for Use and Care of Animals (RRID: SCR_025789). KPC2 or mT4 cells were collected and suspended into PBS (10 million cells/mL). 100ul of cell suspension (1 million cells) were injected subcutaneously, bilaterally into the flanks of 3–5 week old, male mice. KPC2 cells were injected into either FVB (Envigo, RRID:MGI:3609372) or athymic nude-Foxn1^nu (Envigo, RRID:IMSR_ENV:HSD-070) mice and mT4 cells were injected into C57/Bl6 (Envigo, RRID:MGI:7264769) mice. Ifnar1^−/− breeding mice were provided by Dr. Y Lei at the University of Michigan (25). These mice were originally purchased Ifnar1^−/− strain 32045 from Jackson Laboratories and were backcrossed to C57/Bl6 for over 9 generations and contain a homozygous mutation for Ifnar1. All breeding followed University of Michigan Unit for Laboratory Animal Medicine and Committee for Use and Care of Animals (SCR_025789) regulations. Genotyping was conducted on all mice using the following primers: mouse Ifnar1 (forward): 5’-CGAGGCGAAGTGGTTAAAAG-3’; mouse Ifnar1 (reverse wildtype): 5’-ACGGATCAACCTCATTCCAC-3’; mouse Ifnar1 (reverse mutant): 5’-AATTCGCCAATGACAAGACG-3’. Olaparib (50 mg/kg) was given via oral gavage one hour before radiation and continued for 7–11 days. PD-L1 blocking antibody or IgG2b isotype control was administered intraperitoneally 100 μg/mouse every 3 days starting at day −1 (day 0 represents the start of radiation treatment) for a total of 3 doses. For experiments using CD8a blocking antibody, Cd8a blocking antibody or IgG2b isotype control were administered intraperitoneally 250 μg/mouse every 3 days starting at day −1 (day 0 represents the start of radiation treatment) for a total of 4 doses. Tumor length (“a”) and width (“b”) were measured and recorded twice a week using calipers. Tumor volume (TV) was calculated using the following equation: TV=π/6(a x b²). Mice were randomized to treatment groups based on tumor volume size. 10 mice per treatment arm will provide at least 80% power to detect an increase in median time to tumor doubling from 7 to between 16 and 23 days, depending on the extent of within mouse, between tumor, correlation. This calculation is based on a one-sided 0.05 level log-rank test power calculation. TV was plotted as tumor growth curves and tumor volume doubling time. Measurements were taken until day 80 or until humane endpoints were reached (tumor length equal to or greater than 20 cm or lesion of greater than 50 percent of tumor surface). Mouse weights were also monitored throughout experiments.

Immunohistochemistry

mT4 tumors harvested from C57/Bl6 mice on day 4 of treatment from each indicated treatment group. Tumors were fixed in 10% formalin and then embedded in paraffin. Five μm thick sections were cut and baked for 60 minutes at 60°C before deparaffinization in xylene. Slides were then rehydrated in water by decreasing strengths of alcohol. Slides then underwent antigen retrieval in 1X AR6 buffer (PerkinElmer) using the microwave treatment. Slides were stained for the indicated marker using the EnVision G|2 DoubleStain System (Agilent) as previously described (15). Slides then airdried and were mounted with permanent mounting medium. Bright field images were taken on an Olympus BX-51 microscope (RRID:SCR_023069), Olympus digital camera, and DP controller software. The number of CD8a positive or Granzyme B positive cells per high powered field were calculated manually and plotted. The CD8a (clone D4W2Z, Cat#98941, RRID:AB_2756376) and Granzyme B (clone D6E9W, Cat#46890, RRID:AB_2799313) antibodies used for immunohistochemical staining were purchased from Cell Signaling Technology (RRID:SCR_002071).

Single-cell sequencing using 10x Genomics

Subcutaneous mT4 tumors from C57/Bl6 mice were harvested (6–8 tumors per group), minced and digested in collagenase digest buffer (RPMI-1640 medium, Type I collagenase 0.1 mg/mL, Type IV collagenase 0.1 mg/mL, Dnase 0.15 mg/mL) for 30 minutes at 37°C. Samples were filtered through a 40 μm mesh filter and washed with 2%FBS in PBS. Cells were counted and depleted of dying cells twice using the Dead Cell Removal kit (Miltenyi, RRID:SCR_008984). Immune cells were enriched using a magnetic CD45⁺ isolation kit (Miltenyi, RRID:SCR_008984). CD45⁺ and CD45⁻ cells were counted and resuspended at ~1000 cells/μL. Equal amounts CD45⁺ and CD45⁻ cells from each sample were mixed together. Single cell suspensions were counted once more on the Luna Fx7 Automated Cell Counter (LogosBio) and diluted to a concentration of 700–1000 cells/μL. Samples with greater than 85% viability were processed for sequencing. Single cell sequencing was conducted at the University of Michigan Advanced Genomics Core Research Facility (RRID: SCR_025788). Single-cell 3’ library generation was done on the 10X Genomics Chromium Controller (RRID:SCR_019326) following the manufacturers protocol for the 3’v3.1 chemistry with NextGEM Chip G reagents (10X Genomics, Pleasanton, CA, USA). Final library quality was checked using the LabChip GXII HT (PerkinElmer, RRID:SCR_012163) and were quantified by Qubit (Thermo Fisher Scientific, RRID:SCR_020553). Pooled libraries were subjected to 150 base pair, paired-end sequencing (Illumina NovaSeq 6000, RRID:SCR_016387). To create de-multiplexed Fastq files the software Bcl2fastq2 Conversion Software (Illumina, RRID:SCR_015058) was used. The CellRanger v7 Pipeline (10X Genomics, RRID:SCR_023221) was used to align reads and create count matrices.

Bioinformatics analysis of single-cell sequencing data

A total of ~100 million reads were generated from the 10X Genomics sequencing analysis for each of the samples. The sequencing data was pre-processed using the CellRanger v7 Pipeline (10X Genomics Inc., Pleasanton, CA, USA, RRID:SCR_023221), including alignment against mm10 genome. The output from Cell Ranger (Cell Ranger Summary) indicated that 94% of the input reads aligned with ~2600 median genes/cell aligning to the reference. We filtered out cells with less than 200 genes per cell and with more than 10% mitochondrial read content. Use of the Seurat R package (Satija lab; v4, RRID:SCR_007322)(26) was utilized for downstream analysis. These include normalization, identification of variable gene expression across cells, scaling based on number of UMI, dimensionality reduction (PCA and UMAP), un-supervised clustering, and identification of differentially expressed of cell-type specific markers. Downstream analysis of cell populations includes gene signature score analysis, differential gene expression analysis and gene set enrichment analysis (GSEA, RRID:SCR_003199).

Cluster identification

Following unsupervised clustering, our cells were clustered into 21 different groups which were labeled as 9 different cell types based on characterized lineage markers. Key markers used for cell type identification include (27,28): T cell specific markers – Il7r, Cd8a, Nkg7, Cd3e, and Cd2; PDAC specific markers – Onecut2, Clu, and Krt8; neutrophil specific markers – S100a9, S100a8, and Cxcl2; macrophage specific markers – Lyz2, Ctss, and Apoe; fibroblast specific markers – Col1a1 and Dcn; B cell specific markers – Cd79a and Ms4a1; endothelial cell specific markers – Vwf and Cdh5; cycling T cell specific markers – Cd8a, Nkg7, Cd3e, Cd2, Mki67, Kif11 and Kif15; and dendritic cell specific markers – Flt3, Cox6a2, and Ccr9.

Specific cell populations were isolated for further downstream analysis including the PDAC cells, myeloid cells (dendritic cells, macrophages, and neutrophils) and the CD8⁺ T cells. To isolate the CD8⁺ T cells, first all T cell clusters were isolated (excluding cycling T cells) and unsupervised clustering was conducted on this population. Marker expression was used to identify CD4⁺, CD8⁺ and NK T cells. The CD8⁺ T cell clusters were isolated and unsupervised clustering was conducted once more. This final clustered population of CD8⁺ T cells was used for downstream analysis. CD8⁺ T cell clusters were named based on differential gene expression of cell type markers including: naïve CD8⁺ T cell specific markers – Ccr7, Dapl1, Lef1, Klf2, S1pr1, Tcf7, Sell, and Il7r; terminal effector memory CD8⁺ T cell specific markers – Itga4, Cd28, Gzmk, Cxcr3, Gzma, Cd5, Itgb1, and Ccr2 (29,30); dysfunctional CD8⁺ T cell specific markers – Irf4, Xcl1, Cd4, Cd3, Ifng, Cx3cr1, Rgs16, Tnfrsf9, Cd160, Hif1a, Tgfb1, and Nr4a2; terminal effector CD8⁺ T cell specific markers – Gzmb, Prf1, S100a4, Id2, Cxcr6, Pdcd1, Lag3, and Havcr2; and interferon stimulated associated gene (ISAG) CD8⁺ T cell specific markers – Isg15, Ifit1, Ifit3, Isg20, Bst2, and Cxcl10 (31).

Interactome analysis

Interactome plots were created using all detected receptor-ligand interactions after interactome analysis was conducted on all cells following code previously published for this methodology (32). Packages used for this analysis are Seurat (RRID:SCR_007322) and dplyr (RRID:SCR_016708) (26,33).

Gene signature score analysis

Gene signature score analysis was conducted by using the AddModuleScore function within the Seurat R package (RRID:SCR_007322) (26). This function calculates the average feature expression level of clusters based on single cell expression level, subtracted by the aggregated expression of control feature sets.

Gene set enrichment analysis (GSEA)

First, fold change values were calculated for each treatment group comparison using the FoldChange function in Seurat RRID:SCR_007322) (26). Genes which had an average log2 fold change of less than two were filtered out. The Hallmark pathways reference from the msigdbr package (RRID:SCR_022870) was used, which has pathways from the Molecular Signatures Database (MsigDB, RRID:SCR_016863) (34). The Hallmark pathways was used in the clusterProfiler package (RRID:SCR_016884) in order to run GSEA (RRID:SCR_003199) (35). Heatmaps displaying the pathway comparisons display the normalized enrichment scores (NES) and were created with the ComplexHeatmap (RRID:SCR_017270) package (36). GSEA enrichment plots were created with the clusterProfiler package (RRID:SCR_016884). All GSEA analysis (RRID:SCR_003199) was done in R using the RStudio interface (RRID:SCR_000432).

Statistical analyses

Data are presented as mean ± SEM unless otherwise specified. For statistical tests comparing two groups, two-tailed, unpaired t tests were conducted in GraphPad PRISM version 10 (GraphPad Software, RRID:SCR_002798). When comparing more than two treatment groups one-way ANOVA tests were done in PRISM. For tumor growth experiments one-way ANOVA statistical tests were done on indicated days for tumor volume averages. For tumor volume doubling time experiments, the time required for tumor doubling was determined for each tumor as the earliest day on which the tumor was at least two times as large as that same tumor on the first day of treatment. The Kaplan-Meier method was used in PRISM to analyze doubling time and determine if there was a statistical significance between doubling time of treatment groups. P-values of <0.05 were considered statistically significant and are denoted in the figures as follows: *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.

Data Availability

All raw single-cell RNA sequencing files can be found on GEO with the accession number GSE276238. Code will be provided upon request.

RESULTS

Olaparib enhances radiation-induced T1IFN production and signaling

Olaparib induces persistent DNA double-strand breaks in response to radiation (22). Given that radiation-induced DNA damage stimulates T1IFN production (6), we hypothesized that olaparib would enhance radiation-induced T1IFN production and signaling. To test this, Panc-1 cells stably expressing a human IFNβ1 promoter-driven GFP reporter construct (6) were treated with olaparib and radiation and assessed for GFP fluorescence at 72 hours following treatment. We found that olaparib alone or radiation alone induced a modest increase in T1IFN assessed by the IFNβ1-GFP reporter (Fig. 1A). In combination, olaparib significantly enhanced radiation-induced T1IFN expression in a clinically achievable (37), concentration-dependent manner compared to radiation alone. These results are further supported by a similar induction of T1IFN expression in response to radiation and PARP inhibition by talazoparib (Fig. S1A). We next tested the effects of combined treatment with olaparib and radiation on T1IFN production in murine pancreatic cancer cell lines (KPC2 and mT4) which are compatible with immunocompetent syngeneic in vivo models. We found that Ifnβ mRNA expression was induced beginning at 48 hours post-treatment and reached maximal levels at 72 hours (Fig. S1B). This T1IFN effect was found to be independent of the canonical cGAS/STING pathway, as shown by persistent T1IFN signaling in cGAS knockout Panc-1 cells following treatment with olaparib and radiation (Fig. S1C), which is consistent with our prior studies (15,17). Furthermore, IFNβ was significantly increased by the addition of olaparib to radiation when compared to radiation alone in both murine cell lines as well as in the human PDAC cell line Panc-1 (Fig. 1B). Additionally, we found increased protein levels of IFNβ1 in the supernatant of mT4 cells following treatment with olaparib and radiation (Fig. 1C). Consistent with increased expression of T1IFN, the combination of olaparib and radiation induced an interferon stimulated gene response marked by significantly increased mRNA levels of both Cxcl9 and Cxcl10 chemokines in KPC2, mT4, Panc-1, and MiaPaCa2 cells (Fig. 1D, E). This response was maximal at 72 hours (Fig. S1D–F) and when olaparib was combined with 8 Gy radiation (Fig. S1G–I). Therefore for following experiments in the present manuscript we use 8 Gy radiation as this provides the most immunogenic response in PDAC cells. This is further supported by increased expression of Mx1 in KPC2 and mT4 cells (Fig. S1J, K). This was consistent with increased CXCL9 and CXCL10 protein levels in the supernatant of KPC2 cells following treatment with olaparib and radiation (Fig. S1L, M)

Figure 1. Olaparib promotes radiation-induced T1IFN and interferon stimulated gene expression.

(A) Pancreatic cancer cells were treated with olaparib (1 μM, 3 μM or 5 μM) one hour before radiation treatment (8 Gy) and collected 72 hours after treatment for analysis. Flow cytometry of treated Panc-1 cells with stable IFNβ1 promoter-GFP reporter to determine the mean fluorescence intensity (MFI) of GFP. Data are the mean of 3 independent experiments ± SEM, each performed in technical triplicate. Note denoted on graph: radiation alone treatment was statistically significant when compared to control with a **P<0.01. (B-H) Pancreatic cancer cells were treated with 5 μM olaparib one hour before radiation treatment (8 Gy) and collected 72 hours after treatment for analysis. (B) qRT-PCR of KPC2, mT4, and Panc-1 cells with primers targeting Ifnβ. Data are the mean fold change ± SEM relative to control treatment and represent 3 independent experiments run in technical triplicate. (C) Protein abundance quantification via ELISA for IFNβ1 in the media from mT4 cells. Herring testis DNA was transfected into cells (2μg) as a positive control and had an value of 44.83±14.76 pg/mL (mean ± SEM). Data are the average abundance ± SEM and represent 3 independent experiments run in technical triplicate. (D, E) qRT-PCR of KPC2, mT4, Panc-1, and MiaPaCa2 cells with primers targeting Cxcl9 (D), and Cxcl10 (E). Data are the mean fold change ± SEM relative to control treatment and represent 3 (D (mt4), E (Panc-1), E (MiaPaCa2)) or 4 (D (KPC2), E (KPC2), E (mT4)) independent experiments performed in technical triplicate. (F) Flow cytometry analysis of cell surface PD-L1 in KPC2, mT4, and Panc-1 cells. Data are the MFI relative to control treatment and represent 3 independent experiments performed in technical triplicate. (G) qRT-PCR of Panc-1 and MiaPaCa2 cells with primers targeting pd-l1. Data are the mean fold change ± SEM relative to control treatment and represent 3 independent experiments run in technical triplicate. (H) Flow cytometry analysis of cell surface MHC-I expression in mT4 cells. Data are the MFI relative to control treatment and represent 3 independent experiments performed in technical triplicate. Statistical significance was determined using a one-way ANOVA test (A, B (mT4), B (KPC2), D, E, F (KPC2)) or a two-tailed, unpaired t test (B (Panc-1), C, F (Panc-1), H). *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.

T1IFN-mediated immune signaling has additional consequences on the immune system including PD-L1 expression and antigen presentation (11) thus, we sought to investigate whether olaparib enhanced radiation-induced PD-L1 and MHC-I cell surface expression. We found that the combination of olaparib and radiation significantly increased PD-L1 protein cell surface expression and mRNA expression in pancreatic cancer cells (Fig. 1F, G). This effect was dependent on interferon receptor signaling as IFNAR1 blocking antibody diminished the effects of olaparib and radiation on PD-L1 expression (Fig. S1N). Enhanced radiation-induced PD-L1 surface expression by olaparib was confirmed in two additional human pancreatic cancer cell lines (MiaPaCa2 and AsPC-1) (Fig. S1O, P). Finally, to further assess functional endpoints in the T1IFN response, we examined MHC-I cell surface expression levels and found that the combination of olaparib with radiation significantly enhanced MHC-I surface protein levels in human and murine pancreatic cancer cells (Fig. 1H; Fig. S1Q–S). Taken together, these data demonstrate that PARP inhibition by olaparib enhances both T1IFN production and T1IFN stimulated gene responses following treatment with radiation.

Combined therapy efficacy of olaparib and radiation involves an intact immune system and sensitizes pancreatic cancer to αPD-L1 immunotherapy

To begin to investigate a role for an anti-tumoral immune response in the therapeutic efficacy of olaparib and radiation, we began by comparing tumor responses to combined treatment in immunodeficient versus immunocompetent mouse models. Immunodeficient NSG mice or immunocompetent, syngeneic FVB mice with subcutaneous KPC2 tumors were treated with olaparib and radiotherapy (1.6 Gy x 5) and tumor growth was monitored. In immunodeficient mice, we found that olaparib and radiation had only a modest effect on tumor growth that was consistent with our prior studies using similar conditions (Fig. 2A) (14,22). In immunocompetent mice, while treatment with olaparib alone had no effect on tumor growth, the combination of olaparib with radiation resulted in significant tumor growth inhibition relative to radiation alone (Fig. 2B). These data suggest a contribution of the immune system to the therapeutic efficacy of olaparib and radiation.

Figure 2. Combined treatment with olaparib and radiation inhibits tumor growth in immune competent mice and sensitizes to immunotherapy.

(A, B) Tumor volume measurements of KPC2 tumors over time during treatment with olaparib, radiation (1.6Gy X 5 fractions), or the combination of olaparib and radiation in immunodeficient athymic nude mice (n= 8–10 per group) (A) and immunocompetent FVB mice (n= 8–10 per group) (B). Data represent mean tumor volumes ± SEM. (C) Schematic showing treatment schedule of olaparib, targeted radiation therapy, and αPD-L1 therapy. (D) FVB mice with subcutaneous KPC2 tumors were treated as illustrated with olaparib, radiation, and αPD-L1 (C). Data represent mean tumor volumes ± SEM. Data are the mean ± SEM of n= 6 (IgG), 6 (αPD-L1), 6 (olaparib+IgG), 6 (olaparib+αPD-L1), 6 (RT+IgG), 6 (RT+αPD-L1), 10 (olaparib+RT+IgG), and 10 (olaparib+RT+αPD-L1) tumors per treatment group. Statistical significance was determined by a one-way ANOVA test at day 31. (E) Schematic showing treatment schedule of olaparib, targeted radiation therapy, and αPD-L1 therapy. (F) C57BL/6 mice with subcutaneous mT4 tumors were treated as illustrated with olaparib, radiation, and αPD-L1 (E). Data represent mean tumor volumes ± SEM. Data are from n= 6 (IgG), 7 (αPD-L1), 6 (olaparib+IgG), 7 (olaparib+αPD-L1), 8 (RT+IgG), 10 (RT+αPD-L1), 7 (olaparib+RT+IgG) and 9 (olaparib+RT+αPD-L1) tumors per treatment group. Statistical significance was determined by one-way ANOVA test at day 44. (G, H) Mice with complete responses to olaparib+RT+αPD-L1 from panel F were rechallenged with mT4 (10⁶) cells 7 days after complete response. Naïve C57BL/6 mice (mice who have never received treatment or tumor injections previously) were also rechallenged with mT4 (10⁶) cells. Data are tumor establishment time (approximately 50 mm³ tumor volume) (G) or mean tumor volumes ± SEM (H) from naïve mice (n=10 tumors) and previously treated C57BL/6 mice (n=8 tumors). Statistical significance was determined by Kaplan-Meier test (G) or two-tailed, unpaired t test at day 28 after rechallenge (H).

Expression of PD-L1 on tumor cells restrains CD8⁺ T activity leading to tumor immune evasion (38). To investigate whether blocking PD-L1, which is enhanced by olaparib and radiation treatment (Fig. 1F, G), could further promote anti-tumoral immune responses, immunocompetent FVB mice with KPC2 tumors were treated with the concurrent combination of olaparib, radiation, and αPD-L1 (Fig. 2C). Olaparib or αPD-L1 monotherapies were ineffective at inhibiting tumor growth (Fig. 2D; S2A). Radiation alone had a modest effect on tumor growth that was enhanced by treatment with olaparib but not αPD-L1. In contrast, the addition of olaparib to radiation and αPD-L1 treatment significantly enhanced tumor growth inhibition relative to radiation with olaparib or αPD-L1 alone (Fig. 2D; S2A).

To further evaluate the efficacy of combination therapy with olaparib, radiation, and αPD-L1 we compared treatment schedules in which αPD-L1 was given pre-, concurrent, or post- olaparib and radiation. While pre-treatment with αPD-L1 produced maximal tumor growth response, it was also accompanied by significant reductions in blood cell counts (Fig. S2B, C). We therefore elected to continue to use a concurrent treatment schedule for evaluation of efficacy in immunocompetent mice with mT4 tumors treated as illustrated (Fig. 2E). While radiation alone (8 Gy x 2) inhibited tumor growth, olaparib or αPD-L1 alone were ineffective (Fig. 2F; S2D). Furthermore, the combination of olaparib or αPD-L1 with radiation was also ineffective in significantly delaying tumor growth. The combination of olaparib with radiation and αPD-L1, however, induced significant tumor growth inhibition (relative to RT+αPD-L1) in the absence of significant weight loss (Fig. S2E), as well as a 33% complete tumor regression rate (Fig. 2F; S2D, F). To determine whether combination therapy produced durable tumor responses, mice with complete responses were rechallenged with mT4 tumors. When compared to naïve mice in which engraftment of mT4 tumors was 100%, mice previously cured by olaparib, radiation, and αPD-L1 therapy had significantly reduced engraftment rates (50%) and impaired growth in the subset of tumors which did engraft (Fig. 2G, H; S2G, H). These data show that concurrent treatment with olaparib, radiation, and αPD-L1 is effective and tolerable and induces long-term immune memory in vivo.

Combination treatment with olaparib, radiation, and αPD-L1 reshapes the tumor microenvironment

To begin to evaluate the effects of therapy on the tumor microenvironment, we conducted single-cell RNA sequencing on mT4 tumors treated with olaparib, radiation, and αPD-L1 as illustrated (Fig. 2C) but with analysis mid-way in treatment at day 4. Using an un-supervised approach cells were clustered into 9 different tumoral, stromal, and immune cell types based on lineage markers (Fig. S3A, B). All cell types were present in each of the treatment groups (Fig. 3A). To assess the effects of therapy on the in the tumor microenvironment as a whole, we first examined the predicted interactions between cell types via Interactome analysis, which revealed that many of the significant cell to cell interactions occur between T cells and other cell compartments, including PDAC cells, macrophages and fibroblasts (Fig. 3B). Additionally, we found via gene signature analysis that globally the only immune signaling response upregulated by treatment with olaparib, radiation, and αPD-L1 was T1IFN response (Fig. S3C). We next wanted to investigate how treatment impacted individual cellular compartments and began by assessing the PDAC cells, which were distributed across all treatment groups (Fig. S3D). Gene signature analysis revealed that PDAC cells following olaparib, radiation, and αPD-L1 treatment had significantly higher T1IFN response and interferon gamma response gene signatures as compared to PDAC cells treated with radiation and αPD-L1, while induction of other common immune signaling pathways remained constant across treatment groups (Fig. 3C; S3E). GSEA (gene set enrichment analysis) demonstrated that combination treatment with olaparib, radiation, and αPD-L1 increased expression of multiple immune response pathways including interferon alpha, interferon gamma, and inflammatory responses, as well as allograft rejection (Fig. 3D), changes consistent with immune activation (39,40). Furthermore, specific comparison of PDAC cells treated with radiation and αPD-L1 versus olaparib, radiation, and αPD-L1 revealed that olaparib treatment induced enrichment of key immune response pathways including interferon alpha and interferon gamma response as well as the G2M checkpoint (Fig. 3E; Fig. S3F), which is consistent with our current (Fig. 1) and prior studies (41), respectively. The addition of olaparib to radiation and αPD-L1 treatment also resulted in the suppression of pathways that have been shown to enhance PDAC cell proliferation and survival including TNFα signaling and heme metabolism (42,43). Taken together, these data indicate that treatment with olaparib, radiation, and αPD-L1 potentiates interferon signaling within pancreatic tumors.

Figure 3. Olaparib, radiation, and αPD-L1 therapy augments the anti-tumor immune microenvironment.

(A) UMAP projections of all cell clusters split by treatment group. Mice were treated with 1 dose (8 Gy) radiation on day 0, five doses olaparib (50 mg/kg) on days 0–4, and two doses of αPD-L1 (100 μg) on days −1, and 2. Tumors (6–8 tumors per group) were harvested on day 4. (B) Interactome plots displaying all detected receptor ligand interactions for each treatment group when compared to control. Each circle represents a cell population present in the tumor microenvironment and the color of each line represents the cell population that is the source of the detected interaction. Each line represents a single detected interaction. The plots represent all (top row) detected interactions and only significant (bottom row) interactions. (C-H) PDAC tumor cells (C-E), myeloid cells (F, G), or macrophages (H) were isolated for additional bioinformatics analysis. Gene signature analysis conducted on PDAC cells (C) and macrophages (H). Violin plots are the results for T1IFN response and interferon gamma response in PDAC cells (C) and for M1 polarization phenotype in macrophages (H). Statistical significance between groups was determined by Wilcoxon test (C, H). Heatmaps (D, F) and bar plots (E, G) display the results of GSEA analysis. Heatmaps for PDAC (D) and myeloid (F) cells display the top 10 most significantly enriched pathways (adjusted p value ≤ 0.05) selected from each comparison. Normalized enrichment scores for these pathways were combined to create heatmaps. (D, F) Direct comparison between cells in the olaparib+radiation+αPD-L1 treatment group and the radiation+αPD-L1 treatment group using GSEA for PDAC (E) and myeloid (G) cells. Normalized enrichment scores are shown for all pathways with an adjusted p value ≤ 0.05 (E) or a p value ≤ 0.05 (G). *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.

Given the importance of the myeloid cell compartment to the generation of an immunosuppressive microenvironment in PDAC (44), we next analyzed the myeloid cell compartment, which was composed of neutrophils, macrophages and dendritic cells (Fig. S3G). GSEA revealed that treatment with olaparib, radiation, and αPD-L1 enriched important immune response pathways within the myeloid cell compartment, including interferon gamma response, interferon alpha response, inflammatory response, and allograft rejection (Fig. 3F). Direct comparison of the myeloid cells treated with radiation and αPD-L1 versus treated with olaparib, radiation, and αPD-L1 demonstrated both increases and decreases in protumor signaling within the myeloid cell compartment (Fig. 3G; S3H) including increased coagulation, glycolysis, and mTORC1 signaling (45,46) and decreased TGFβ signaling (47). The addition of olaparib to radiation and αPD-L1 treatment contributes to both the protumor and antitumor activity of myeloid cells. Macrophages are the most prevalent myeloid cell type in pancreatic tumors (48). Interferon signaling can promote antitumor immune function in macrophages (49). Indeed, we observed that the combination of olaparib, radiation, and αPD-L1 increased M1 macrophage polarization (Fig. 3H) but did not alter M2 polarization or phagocytosis function (Fig. S3I). Taken together, these data support that treatment with olaparib, radiation, and αPD-L1 promotes immune signaling in both tumor and myeloid cells.

Treatment with olaparib, radiation, and αPD-L1 augments CD8⁺ T cell populations within the tumor microenvironment

Given the involvement of an adaptive immune response to combined therapy efficacy (Fig. 2), we next evaluated CD8⁺ T cell clusters (Fig. S4A–E). We identified six CD8⁺ T cell clusters and found that the frequency of CD8⁺ T cell populations varied based on treatment group (Fig. 4A, B). Specifically, we found that the proportion of CD8⁺ terminal effector cells (identified by expression Gzmb, Prf1, Havcr2, and Pdcd1 expression) increased with treatment, the largest proportion occurring in response to combined treatment with olaparib, radiation, and αPD-L1. Consistent with this increase in terminal effector cells, we observed a corresponding decrease in naïve T cells (characterized by Ccr7, Dapl1, Lef1, and Klf2 expression). These changes in response to treatment were accompanied by a modest increase in the dysfunctional cells (marked by Irf4, Xcl1, Ccl4, and Ccl3 expression) but increases in the tissue effector memory (TEM; Itga4, Cd28, Gzmk, and Cxcr3 expression) and interferon stimulated associated genes (ISAG) T cells (expressing Isg15, Ifit1, Ifit3, and Isg20) (Fig. 4C). TEM CD8⁺ T cells have been previously reported as memory T cells which display immediate effector function at sites of inflammation (29,30) while ISAG CD8⁺ T cells are likely involved in a rapid immune response (31). Further, these effects were associated with a decrease in B cells following treatment suggesting that T cells are the key adaptive immune cell type contributing to olaparib, radiation, and αPD-L1 efficacy (Fig. S4F). Taken together, these data suggest that treatment with olaparib, radiation, and αPD-L1 modulates the tumor microenvironment to support a CD8⁺ T cell-anti-tumor immune response.

Figure 4. Treatment with olaparib, radiation, and αPD-L1 transforms CD8⁺ T cell populations within the tumor microenvironment.

(A) UMAP projection of isolated CD8⁺ T cell populations. (TEM – tissue effector memory, ISAG – interferon stimulated associated genes) (B) Frequency of CD8⁺ T cell populations split by treatment group. (TEM – tissue effector memory, ISAG – interferon stimulated associated genes (C) Dotplot showing an incomplete list of markers used to name CD8⁺ T cell subclusters. The size of the circle represents the percentage of cells within that cluster that express the given gene. The color represents the average fold change expression of that gene within that cluster. (D) Violin plots displaying CD8⁺ T cells gene signature scores for T1IFN response, interferon gamma response, IL2/STAT5 response, and T cell effector signature are shown and split by treatment group. Statistical significance was determined by Wilcoxon test. (E) GSEA was conducted using differentially expressed genes between the control group and each treatment individually. The top 10 most significant enriched pathways (with an adjusted p value ≤ 0.05) were selected for each comparison. Normalized enrichment scores for these pathways were combined to create the resultant heatmap. (F) Direct comparison between the CD8⁺ T cells in the radiation+αPD-L1 group and the olaparib+radiation+αPD-L1 group using and GSEA analysis. Normalized enrichment scores are shown for pathways with an adjusted p value ≤ 0.05. *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.

To further investigate whether treatment influenced CD8⁺ T cell functions, we next performed gene signature and GSEA analyses. Gene signature analysis revealed that the addition of olaparib to radiation and αPD-L1 treatment significantly increased the T1IFN, interferon gamma, and IL2/STAT5 responses, as well as the T cell effector phenotype gene signature scores when compared to radiation and αPD-L1 treatment (Fig. 4D). We then evaluated differentially expressed genes between each treatment group compared to the control treatment group. Treatment with olaparib, radiation, and αPD-L1 resulted in the upregulation of metabolic pathways such as oxidative phosphorylation (Fig. 4E) that support the metabolic reprogramming of activated CD8⁺ T (50,51). Additionally, combined therapy enriched many pathways that directly signify CD8⁺ T cell activation, including interferon gamma response, IL2/STAT5 signaling, and allograft rejection (Fig. 4E). These pathways are essential for CD8⁺ T cell differentiation, infiltration, and activation allowing for immune responses to be carried out by the CD8⁺ T cells (39,40,52). To further define the CD8⁺ T cell changes directly affected by olaparib treatment, we compared differentially expressed genes in the CD8⁺ T cells in response to radiation and αPD-L1 in the presence or absence of olaparib (Fig. 4F, S4G). We found that olaparib specifically contributed to upregulation of interferon alpha, allograft rejection, and oxidative phosphorylation pathways (Fig. 4F) consistent with olaparib enhancing the CD8⁺ T cell mediated anti-tumor immune response. Taken together these data suggest that treatment with olaparib, radiation, and αPD-L1 enhances CD8⁺ T cell anti-tumor responses through modulation of the number and function of CD8⁺ T cells.

Treatment with olaparib, radiation, and αPD-L1 increases CD8⁺ T cells in the tumor microenvironment

To confirm that treatment with olaparib, radiation, and αPD-L1 increases CD8⁺ T cell infiltration and activation into pancreatic tumors in vivo, we examined the number and functionality of intratumoral CD8⁺ T cells via immunohistochemical staining and flow cytometry. CD8a IHC staining showed that olaparib alone or in combination with αPD-L1 did not significantly increase the number of intratumoral CD8⁺ T cells (Fig. 5A, B). The combination of olaparib and/or αPD-L1 with radiation, however, did cause a significant increase CD8a⁺ T cells. Furthermore, this increase in intratumoral CD8⁺ T cells was accompanied by an increase in Granzyme B staining suggesting enhanced CD8⁺ T cell effector function specifically in response to combined treatment with olaparib, radiation, and αPD-L1 (Fig. 5C, D). Flow cytometry analysis of whole tumor lysates showed increased TNFα⁺ CD8⁺ T cells following treatment with olaparib, radiation, and αPD-L1 when compared to radiation and αPD-L1 suggesting that the addition of olaparib is responsible for increased CD8⁺ T cell activation intratumorally (Fig. 5E). We also found a significant increase in IFNγ⁺TNFα⁺ CD8⁺ T cells in the tumor draining lymph nodes following treatment with olaparib, radiation, and αPD-L1 compared to radiation and αPD-L1, demonstrating that treatment induces activation of both tumoral and peripheral CD8⁺ T cells (Fig. 5F). Furthermore, combined therapy increased proliferative intratumoral CD8⁺ T cells marked by a significant increase in the percentage of Ki67⁺ CD8⁺ T cells (Fig. S5A). Consistent with our single cell data (Fig. 3H), M1 anti-tumor macrophages were increased in tumors following treatment with olaparib, radiation, and αPD-L1 when compared to all other treatments (Fig. 5G). Overall, these data confirm that combined treatment with olaparib, radiation, and αPD-L1 reshapes both the intratumoral microenvironment and host immune system shown by increased M1 macrophage polarization and increased CD8⁺ T cell frequency, proliferation, and activity.

Figure 5. Olaparib, radiation, and αPD-L1 treatment enhances the frequency and activity of intratumoral CD8⁺ T cells that together with T1IFN signaling are required for treatment efficacy.

(A-D) C57BL/6 mice with subcutaneous mT4 tumors were treated as illustrated in Figure 2E, tumors were harvested at day 4, and stained for CD8a (A, B) or Granzyme B (C, D) by immunohistochemistry. Data are representative images (A, C) or the mean ± SEM of CD8a⁺ cell number per high power field (n= 9–12 fields) (B) or mean ± SEM of Granzyme B positive cells per high power field (n=9–12 fields) (D) from 3–4 tumors per treatment condition. (E-G) C57BL/6 mice with subcutaneous mT4 tumors were treated as illustrated in Figure 2E and tumors and tumor draining lymph nodes were harvested on day 7 of treatment for analysis via flow cytometry. Data are from n=10 (IgG), 10 (RT), 10 (RT+αPD-L1), 10 (olaparib+RT), and 10 (olaparib+RT+αPD-L1) tumors and tumor draining lymph nodes. (E) Percentage of CD90⁺CD8⁺ cells in the tumor that were TNFα⁺. (F) Percentage of CD90⁺CD8⁺ cells in the tumor draining lymph node that were IFNγ⁺TNFα⁺. (G) Percentage of F4/80⁺ cells in the tumor that were MHC-II^highCD86^high. Statistical significance was determined by two-tailed, unpaired t test (B, E, F) or one-way ANOVA test (D, G). *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.

Treatment efficacy of olaparib, radiation, and αPD-L1 is dependent upon CD8⁺ T cells and T1IFN signaling

To further evaluate the contribution of CD8⁺ T cells to the combination therapy efficacy we next evaluated mT4 tumor growth following CD8⁺ T cell depletion (via CD8a blocking antibody). We found that the ability of olaparib, radiation, and αPD-L1 therapy to inhibit tumor growth was significantly diminished by CD8⁺ T cell depletion as compared to controls (Fig. 6A; S6A). Additionally, mice treated with olaparib, radiation, αPD-L1, and αCD8a showed no difference in time to tumor volume doubling when compared to control mice. While mice treated with only triple combination showed significantly delayed time to tumor volume doubling when compared to all other treatment groups including mice who received triple treatment and αCD8a (Fig. 6B). Taken together, these data demonstrate a requirement for CD8⁺ T cells in the therapeutic efficacy of olaparib, radiation, and αPD-L1.

Figure 6. Treatment efficacy with olaparib, radiation, and αPD-L1 is dependent on CD8⁺ T cell function and T1IFN signaling.

(A, B) C57BL/6 mice with subcutaneous mT4 tumors were treated with or without combination treatment of olaparib, radiation, and αPD-L1 (as illustrated in Figure 2E) in the absence or presence of mouse αCD8 antibody (250 μg, administered via intraperitoneal injection on days −1, 2, 5, and 8). Data are the mean tumor volumes ± SEM (A) or tumor volume doubling time (B). Data are from n= 10 (IgG), 10 (αCD8), 15 (olaparib+RT+αPD-L1), and 20 (olaparib+RT+αPD-L1+αCD8). Statistical significance was determined by one-way ANOVA test at day 42 (A) and Kaplan-Meier test (B). (C, D) Wildtype and Ifnar1^−/− mice bearing subcutaneous mT4 tumors were treated with or without combination treatment of olaparib, radiation, and αPD-L1 (as illustrated in Figure 2E). Data are the mean tumor volumes ± SEM (C) or tumor volume doubling time (D). Data are from n= 20 (WT, IgG treatment), 14 (WT, olaparib+RT+αPD-L1 treatment), 22 (Ifnar1^−/−, IgG treatment), and 20 (Ifnar1^−/−, olaparib+RT+αPD-L1 treatment). Statistical significance was determined by one-way ANOVA test at day 32 (C) or Kaplan-Meier test (D). *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. (E) Radiation and PARP inhibitor (olaparib) induce nuclear DNA damage (22) that upon translocation to the cytosol stimulates innate immune signaling and consequently T1IFN expression. T1IFN modulates the immune system in both an autocrine and paracrine manner resulting in an interferon stimulated gene response in PDAC cells including expression of chemokines (CXCL9/CXCL10), antigen presentation proteins (MHC-I), and PD-L1, as well as effects on host immune cells resulting in increased frequency and activation of CD8⁺ T cells, increased M1 macrophage polarization, immune checkpoint activation, and sensitization of pancreatic cancer to αPD-L1 immunotherapy. Abbreviations used: RT – radiation, T1IFN – Type 1 interferon, PDAC – pancreatic ductal adenocarcinoma, PARP – poly(ADP-ribose) polymerase, PD-L1 – programmed death ligand 1.

Our data show that olaparib and radiation induce T1IFN production (Fig. 1). Thus, we hypothesized that the anti-tumoral immune response to olaparib, radiation, and αPD-L1 is mediated by T1IFN. To test this hypothesis, we compared mT4 tumor responses to therapy in Ifnar1^−/− mice (deficient for T1IFN receptor) or wildtype C57/Bl6 mice. We found that treatment efficacy was significantly diminished in Ifnar1^−/− mice treated with triple combination therapy when compared to wildtype mice (Fig. 6C, D; S6B). These data demonstrate the importance of host T1IFN signaling in the efficacy of combined treatment with olaparib, radiation, and αPD-L1 and support a connection between tumor-derived T1IFN and the host tumor immune microenvironment.

DISCUSSION

In this study, we have found that the PARP inhibitor olaparib enhances radiation-induced T1IFN-mediated immune signaling, consequently sensitizing pancreatic cancer to αPD-L1 immunotherapy (Fig. 6E). We show that olaparib enhances both the production of T1IFN and its downstream signaling consequences following radiation treatment, including induction of interferon stimulated gene expression, PD-L1 surface expression and MHC-I surface expression. Therapeutically, we show in two independent syngeneic immunocompetent pancreatic cancer mouse models that olaparib and radiation treatment sensitizes pancreatic tumors to PD-L1 immune checkpoint blockade leading to substantial tumor burden control and increased time to tumor volume doubling. Single cell RNA-sequencing revealed that treatment with olaparib, radiation, and αPD-L1 modulates the tumor microenvironment, specifically through increasing the proportion of CD8⁺ terminal effector T cells. This was further confirmed by the increased infiltration and activation of CD8⁺ T cells present in tumors as well as the dependency of therapeutic efficacy on CD8⁺ T cell function. Finally, we show that T1IFN-mediated immune signaling is partially responsible for therapeutic efficacy with olaparib, radiation, and αPD-L1. Overall, our results indicate that olaparib enhances radiation-induced T1IFN-mediated immune signaling and subsequently an adaptive immune response thus sensitizing pancreatic cancer to αPD-L1 therapy.

In contrast to other DNA damage response therapeutics such as those targeting ATM, ATR, or DNAPK (53), PARP inhibition by olaparib has unique biology in that it prevents the repair of single-strand breaks, the most frequent DNA lesion induced by radiation, and causes PARP1-DNA complexes (also known as PARP trapping) both of which contribute to the formation of lethal double-strand breaks in replicating cells (22,54). In this study, we demonstrate the ability of olaparib to enhance the radiation induced innate immune response and T1IFN expression. While our findings are overall consistent with the ability of other DNA damage response inhibitors to enhance radiation-induced T1IFN signaling (15,17,18,55,56), the potential contribution of the unique DNA damage caused by PARP inhibition/trapping to anti-tumor innate immune responses is important for future study (57). The effects of PARP inhibitor and radiation on T1IFN are likely independent of the canonical cGAS/STING pathway as supported by our prior studies and the observed independence on cGAS in the present study (15,18) (Fig. S1C). Furthermore, the FDA approval of PARP inhibitors including olaparib warrants their further development in rational therapeutic combinations that can be translated to early phase clinical trials.

Scheduling of radiation and immunotherapy is an important consideration in the design of rational clinical trials. Preclinically, αPD-1 given prior to radiotherapy causes enhanced radiation sensitivity of intratumoral CD8⁺ T cells leading to a diminished adaptive immune response (58). We also investigated the effects of administration schedule on anti-tumor immune responses to olaparib, radiation, and immunotherapy using radiation doses (8 Gy) previously shown to induce an optimal anti-tumor immune response (5). Interestingly, we found that administration of immunotherapy prior to olaparib and radiotherapy was effective in inhibiting tumor growth but accompanied by increased hematopoietic toxicity (Fig. S2B, C) therefore warranting use of a concurrent treatment schedule in this study.

The pancreatic cancer tumor microenvironment is abundant in immunosuppressive myeloid-derived immune cells (M2 macrophages, myeloid derived suppressor cells, and neutrophils) as well as dense stroma, both of which contribute to immunotherapy resistance (59). The modest clinical benefit of combined treatment with radiation and immune checkpoint inhibitors (1) suggest the need to overcome resistance mechanisms beyond PD-1/PD-L1 and CTLA-4 immune checkpoints. We show in this study that combined treatment with olaparib, radiation, and αPD-L1 alters the myeloid compartment and, specifically, increases anti-tumorigenic M1 macrophage signatures (Fig. 3H, Fig. 5G). While the direct contribution of these changes in the myeloid compartment to the adaptive anti-tumoral immune responses observed in this study are unclear, the potential importance of the myeloid population to immunotherapy resistance in pancreatic cancer is significant and warrants future investigation.

Radiation dose and fractionation are important considerations for preclinical research intended to translate to patient clinical trials. Early preclinical studies supported that hypofractionated radiation regimens (ie, 8 Gy X 3) were optimal for inducing anti-tumoral immune responses in combination with immunotherapy (5). While the use of hypofractionated regimens may not be the best approach for all cancer types or therapeutic combinations (60), the 8 Gy fractions used in the present study were superior relative to other dose/fractionation schedules in inducing a T1IFN response and efficacious in combination with PARP inhibitor and immunotherapy (Fig. S1G–I; Fig. 2). Furthermore, the hypofractionated regimen used in this study has translational relevance to the clinical management of pancreatic cancer given the increasing use of hypofractionated SBRT (stereotactic body radiotherapy) as well as the safety and modest efficacy in combination with immunotherapy in patients (1,61,62). The determination of the optimal radiation dosage and schedule will require prospective clinical trials.

Pancreatic cancer is characterized by aggressive local and systemic disease and an immune suppressed tumor microenvironment. While effective local control is essential for clinical management, the importance of systemic control is also crucial. Our study demonstrates that olaparib can improve local tumor responses to radiation and immunotherapy via activation of a T1IFN-driven, CD8⁺ T cell adaptive immune response. While this infers that there may be systemic benefit beyond the local tumor, effects on tumors outside of the radiation field were not addressed in this study and will be a future focus. Further, spontaneous mouse models of pancreatic cancer are likely superior to injectable tumor models in terms of recapitulating the tumor immune microenvironment of human pancreatic cancer thus, underscoring the importance of testing novel therapeutic strategies in autochthonous mouse models (63). On the basis of our prior work demonstrating the ability of PARP inhibitors to synergize with radiation independent of BRCA1/2 status, we focused the current study on BRCA1/2 wildtype, homologous recombination repair proficient pancreatic cancer. While this represents the vast majority of pancreatic cancers, it does not preclude potential efficacy in BRCA1/2 mutant pancreatic cancer as our prior work showed synergy between olaparib and radiotherapy in both BRCA1/2 wildtype and mutant pancreatic cancer (22). Furthermore, clinical data suggesting a benefit of PARP inhibition with immunotherapy in platinum sensitive, homologous recombination repair deficient pancreatic cancer patients supports extension of the current therapeutic regimen to homologous recombination repair defective tumors (64).

Overall, this study supports the use of olaparib in combination with radiation to sensitize to PD-L1 immune checkpoint blockade in pancreatic cancer with potential for application to other cancers which use PD-L1 as an immune evasion strategy. This study is the foundation for a current phase 1 clinical trial which will test the safety and efficacy of dose-escalated olaparib in combination with radiotherapy and durvalumab in patients with locally advanced pancreatic cancer (NCT05411094). Evaluation of molecular correlates in this trial will be an important future direction to validate the importance of T1IFN and CD8⁺ T cells to therapeutic outcomes in patients.