Abstract
Innate inflammation promotes tumor development, although the role of innate inflammatory cytokines in established human tumors is unclear. Herein, we report clinical and translational results from a phase Ib trial testing whether IL1β blockade in human pancreatic cancer would alleviate myeloid immunosuppression and reveal antitumor T-cell responses to PD1 blockade. Patients with treatment-naïve advanced pancreatic ductal adenocarcinoma (n = 10) were treated with canakinumab, a high-affinity monoclonal human antiinterleukin-1β (IL1β), the PD1 blocking antibody spartalizumab, and gemcitabine/n(ab)paclitaxel. Analysis of paired peripheral blood from patients in the trial versus patients receiving multiagent chemotherapy showed a modest increase in HLA-DR+CD38+ activated CD8+ T cells and a decrease in circulating monocytic myeloid-derived suppressor cells (MDSC) by flow cytometry for patients in the trial but not in controls. Similarly, we used patient serum to differentiate monocytic MDSCs in vitro and showed that functional inhibition of T-cell proliferation was reduced when using on-treatment serum samples from patients in the trial but not when using serum from patients treated with chemotherapy alone. Within the tumor, we observed few changes in suppressive myeloid-cell populations or activated T cells as assessed by single-cell transcriptional profiling or multiplex immunofluorescence, although increases in CD8+ T cells suggest that improvements in the tumor immune microenvironment might be revealed by a larger study. Overall, the data indicate that exposure to PD1 and IL1β blockade induced a modest reactivation of peripheral CD8+ T cells and decreased circulating monocytic MDSCs; however, these changes did not lead to similarly uniform alterations in the tumor microenvironment.
Introduction
Tumor-promoting inflammation has long been recognized to be driven by myeloid cells producing innate inflammatory cytokines in mouse models and humans. Mice with chronic inflammation develop increased frequencies of spontaneous tumors that are preventable by either antibiotics or nonsteroidal antiinflammatory agents (1). Most chronic inflammatory diseases in humans are associated with an increased risk of cancer in that organ, such as pancreatitis and pancreatic ductal adenocarcinoma (PDAC; ref. 2). Furthermore, decreasing chronic inflammation can prevent cancer development as demonstrated by the link between aspirin use and decreased risk of colorectal cancer (3, 4).
Chronic inflammation is driven by three innate inflammatory cytokines: IL1, IL6, and TNFα. Increased in multiple autoimmune diseases, these three cytokines are coregulated such that an increase in one leads to increases in the others and blockade of one leads to decreased overall inflammation (5). Neoplastic cells express receptors for these cytokines, which can act as prosurvival growth signals, with IL1 and TNFα leading to NF-κB signaling and IL6 leading to STAT3 signaling. Myeloid cells in tissues may engage in positive feedback loops with neoplastic cells, supporting the early stages of oncogenesis, including in pancreatic cancer models (6, 7). Cytokine blockade seem to reverse this process and prevent cancer formation, as shown for IL6 blockade in mice and for IL1β blockade in humans (1, 8). The Canakinumab Antiinflammatory Thrombosis Outcome Study (CANTOS) trial, a phase III trial of the IL1β-blocking antibody canakinumab in patients with atherosclerosis, met its primary endpoint of reducing cardiovascular events, and simultaneously potentially decreased the risk of new diagnosis of lung cancer (8, 9). Further analysis of patients in the CANTOS trial who developed lung cancers revealed that 43% had detectable tumor ctDNA in baseline plasma samples, suggesting that IL1β blockade may have a role in curtailing incipient tumors, not just in preventing oncogenesis (10). This distinction between tumor prevention and treatment of established tumors is important, and thus far, clinical trials of canakinumab in nonsmall cell lung cancer have failed to demonstrate a therapeutic benefit even in the neoadjuvant or adjuvant setting in which disease burden is presumably low (11).
In mouse models of implantable tumors, blockade of innate inflammatory cytokines, including IL1β, leads to improved tumor control. IL1β blockade alone is usually insufficient as a single agent to regress established tumors; however, in combination with chemotherapy or immunotherapy, the inclusion of IL1β blockade improves outcomes in tumor-bearing mice (5, 7, 12). IL1β is a pleiotropic cytokine, and its roles in the tumor microenvironment (TME) include direct signaling on tumor cells, activation of endothelial cells, promotion of a chemokine-producing inflammatory fibroblast subset, activation of dendritic cells, and activation of immunosuppressive myeloid cells (13). IL1β directly induces and activates myeloid-derived suppressor cells (MDSC) in murine tumors (14–17).
PDAC is a systemic disease in the vast majority of patients. Tumor cells and activated fibroblasts in PDAC secrete G-CSF, CCL2, CXCL1/2/3, IL6, GM-CSF, and other cytokines and chemokines that induce production of monocytes and neutrophils in the bone marrow (18–24). This induction of myelopoiesis results in immature myeloid cells being released into circulation, leading to elevated numbers of monocytes and neutrophils in the blood (25). These myeloid cells are actively recruited into the tumor by local chemokine gradients. Once in the tumor, they differentiate into tumor-associated macrophages or MDSCs, contributing to local suppression of effector T cells and direct promotion of malignant cell growth.
NF-κB signaling in pancreatic tumor cells is essential for their survival in mouse models and may be driven by autocrine production of IL1α (26). Work from the Bar-Sagi lab suggests that IL1β is derived from PDAC tumor cells, which use TLR4 to activate the NLRP3 inflammasome and process pro-IL1β into secreted IL1β (27). Treatment with antibiotics or disruption of IL1β, NLRP3, or TLR4 all lead to decreased αSMA expression, decreased production of myeloid cell–recruiting chemokines from fibroblasts, and decreased tumor burden. Anti-IL1β–treated orthotopic tumors contain less extracellular matrix, fewer granulocytes, and an increased density of CD8+ T cells. Combination of anti-IL1β with anti-PD1 checkpoint blockade results in improved therapeutic efficacy compared with either therapy alone (27).
PD1 blockade in pancreatic cancer has thus far been efficacious in some of the 1% of patients with microsatellite unstable disease (28). As such, most patients with PDAC do not benefit from PD1 blockade or combination checkpoint blockade (29–31). Although the reasons for this are still unknown, recent data suggest that PD1 blockade does reinvigorate tumor-specific T cells in peripheral blood. However, these T cells have an NF-κB signaling signature, rather than the IFN response predominant signature seen in reinvigorated T cells in patients with other cancer types responding to PD1 blockade (32).
We hypothesized that blockade of IL1β and PD1 would result in alterations in myeloid, lymphoid, and fibroblast subsets within the microenvironment in pancreatic cancer to promote tumor-cell killing. We further predicted that circulating monocytes and neutrophils, in which production and mobilization from the bone marrow are augmented in patients with PDAC (25), would decline upon treatment with IL1β and PD1 blockade. A phase I trial offers a unique window to collect rich datasets on a small number of patients before and after perturbation of a biologic pathway of interest. To this end, we collected pre- and on-treatment blood and tumor tissue from patients enrolled in SR1, a phase I trial that included four drugs [canakinumab, spartalizumab (PD1-blocking antibody), gemcitabine, and n(ab)-paclitaxel]. To deconvolute the contribution of the immunotherapy agents versus chemotherapy, we also obtained longitudinal peripheral blood samples from patients with PDAC receiving standard-of-care gemcitabine/n(ab)-paclitaxel (GnP). We found that PD1 and IL1β blockade induced a modest reactivation of peripheral CD8+ T cells and decreased circulating monocytic MDSCs. Although these effects were observed in all patients exposed to treatment, relief of peripheral immune suppression did not lead to similarly uniform alterations in the TME.
Materials and Methods
Human samples
Whole blood, serum, and tumor tissues were collected from 10 patients enrolled in the phase I clinical trial SR1 (NCT04581343) at Dana-Farber Cancer Institute and New York University. This single-arm study was neither randomized nor blinded. Whole blood and serum were collected from patients with PDAC receiving gemcitabine/n(ab)-paclitaxel at DFCI under protocol #03-189. Blood and serum were used for immune cells and cytokine analysis.
Healthy donor serum was obtained from consenting patients at Massachusetts General Hospital under protocol #21-590. Participants were females between 30 and 75 years of age (inclusive) and had no symptoms of acute illness, no history of immune-mediated disease, and no use of immunomodulating medications within 1 month prior to enrollment; received no vaccines within 5 weeks prior to enrollment; and had never received an organ transplant. Blood was collected at two study visits 4 weeks apart. During each visit, physical well-being and eligibility were verified by the study team and a detailed medical history (including relevant comorbidities, current and recent medications, and vaccination history) was reviewed. Serum from healthy donors were used for cytokine analysis controls.
Healthy donor monocytes and T cells were obtained from deidentified leukapheresis cones from the Kraft Blood Donor Center.
Ethics statement
This study involves human participants and was conducted in accordance with the Declaration of Helsinki and the US Common Rule. Approval for #03-189 and #21-590 was granted by the DFCI IRB (IORG #0000035, FWA #00001121). Written informed consent was obtained from participants enrolled in #03-189 and #21-590 prior to sample collection.
Human blood processing
Whole blood was processed within 24 hours of collection for peripheral blood mononuclear cell (PBMC) isolation by gradient centrifugation using Ficoll-paqueplus (Cytiva). Cells were frozen in 90% FBS (Gibco)/10% DMSO (Sigma-Aldrich) at a maximum concentration of 5 × 106 cells/mL and stored in liquid nitrogen.
Flow cytometry
Frozen PBMCs were thawed in a warm bath and then placed in 10 mL of a solution of RPMI complete [RPMIc: RPMI 1640 media (Life Technologies 11875119) supplemented with 10% Fetal Bovine Serum (Gibco cat. 16000044), 1-mmol/L sodium pyruvate (Gibco cat.11360070), 2-mmol/L GlutaMAX (Gibco cat. 35050061), 0.1-mmol/L nonessential amino acids (Gibco cat. 11140050), 100-units/mL penicillin/streptomycin (Gibco cat. 15140122), and 0.11-mmols/L β-mercaptoethanol (Sigma-Aldrich M3148)] with DNase I (Sigma-Aldrich). PBMCs were washed once in PBS with EDTA 2 mmol/L and stained using a Zombie NIR fixable viability kit (BioLegend) prior to staining for flow cytometry. Stained cells were fixed with 1% formalin (Sigma) and analyzed with an SP6800 Spectral cell analyzer (SONY). FCS files were exported and analyzed using FlowJo version 10.8.1.
The following antibodies were used for flow cytometry. From BioLegend: CD3-PE clone OKT3 (RRID:AB_571913), CD19-PE clone HIB19 (RRID:AB_2616905), CD33-PE-Cy7 clone WM53 (RRID:AB_2904477), CD15-PB clone MMA (RRID:AB_2750480), CD14-FITC clone M5E2 (RRID:AB_2616906), CD16-BV785 clone 3G8 (RRID:AB_2563803), HLA-DR-BV570 clone L243 (RRID:AB_2650882), PD1-PE clone EH12.2H7 (RRID:AB_940483), CD137-PE-Cy7 clone 4B4-1 (RRID:AB_2287731), CD8-PB clone HIT8a (RRID:AB_10612929), CX3CR1-BV421 clone 2A9-1 (RRID:AB_2687147), CD14-BV605 clone M5E2 (RRID:AB_2563798), CD4-PB clone OKT4 (RRID:AB_571953), CCR2-BV510 clone K036C2 (RRID:AB_2566503), CD45RO-BV650 clone UCHL1 (RRID:AB_2563462), CD38-BV711 clone HIT2 (RRID:AB_11218990), CD45RA-FITC clone HI100 (RRID:AB_2650650), and CD8-PE/Cy7 clone SK1 (RRID:AB_2876774). From BD Biosciences: CD3-PE-CF594 clone UCHT1 (RRID:AB_11153674), CD45-PE-CF594 clone HI30 (RRID:AB_11154577), and CD4-BV510 clone SK3 (RRID:AB_2744424).
Human CD14+ MDSC differentiation
Monocytes were isolated from healthy donor leukopacs using a human CD14 isolation kit (RRID:AB_2665482; Miltenyi) as per the manufacturer’s protocol. Monocytes were resuspended in RPMIc without FBS and plated into a thermo-sensitive 6-well plate (Thermo Scientific 174901); 200 µL of human serum (either from healthy donors or patients) was added for a final concentration of 20% serum; 75 pg/mL of M-CSF (Peprotech) was added to each condition. Some monocytes were also cultured with 10 ng/mL of human IL6 (Peprotech), 10-ng/mL GM-CSF (Peprotech), 20-ng/mL TNFa (Peprotech), 20-ng/mL IL18 (Peprotech), or 20-ng/mL FLT3L (Peprotech). From days 3 to 5, the medium was changed, 400 µL of fresh RPMIc without FBS was added and replenishment of serum and/or recombinant cytokines as indicated. At day 7, the cells were harvested using cold PBS aided by a cell scraper.
Human T-cell proliferation assay
PBMCs were isolated from healthy donor fresh blood using Ficoll-paqueplus (Cytiva). The PBMCs were then washed, and T cells were isolated using the human Pan T-cell isolation kit (Miltenyi) as per the manufacturer’s protocol. After isolation, the T cells were stained with CFSE (Invitrogen), washed in FBS, and then cultured in RPMIc with CD3/CD28 activation beads (Gibco) in a U-bottom 96-well plate. After 3 hours, 50,000 MDSCs were added to each well containing preactivated T cells. After 3 days of incubation at 37°C, the cells were analyzed by flow cytometry.
Human serum cytokine analysis
On the day of collection from the patients, serum was centrifuged at 450 g for 15 minutes, aliquoted, and stored at −80°C. Serum samples were thawed, and cytokines were analyzed by cytokine bead array (HD71-CLIN, Eve Technologies, Canada). To determine whether canakinumab interfered with the detection of IL1β, serum from three randomly selected patients receiving standard-of-care therapy was aliquoted into four tubes containing canakinumab at 0, 1, 10, or 100 μg/mL. These serum samples were then analyzed using the cytokine bead array. Measured IL1β values for serum from the same patient did not considerably differ with the presence of canakinumab.
Canakinumab pharmacodynamics
For all patients (n = 10) enrolled in the SR1 clinical trial, serum was collected at the following time points: Predose Cycle 1; 168 hours postdose (±8 hours); 336 hours postdose (±8 hours); Predose Cycle 2 (672 hours postfirst dose). Canakinumab was measured by ELISA by Tempus Labs.
Single-cell RNA sequencing
Single-cell RNA sequencing (scRNA-seq) was performed as previously described (32). Briefly, fresh tumor tissue chunks of approximately 2 mm3 were frozen in slow-freeze media (50% FBS, 40% RPMI, 10% DMSO). Upon thawing, tumors were minced with sterile scalpels and digested with a tumor dissociation kit (Miltenyi Biotech #130–095–929) for 30 minutes at 37°C with periodic vortexing. Single-cell suspensions were then stained with Zombie NIR live/dead stain (BioLegend) and sorted on a BD FACS Aria II SORP machine. Up to 20,000 live cells were loaded into a 10× Chromium controller instrument along with Chromium Next GEM Single-Cell 5′ beads (10× Genomics PN1000263). No multiplexing was performed. After RT-PCR, cDNA was purified, and a library was constructed from each sample using a 10× Library Construction Kit (10× Genomics PN1000190) as per the manufacturer’s protocol. Libraries were sequenced on an Illumina NovaSeq system to obtain paired-end 150-bp reads. Tumor samples from 8 patients in the SR1 trial were analyzed. Technical replicates were not performed.
Analysis of scRNA-seq data
We used the 10x CellRanger “multi” pipeline (v6.0.1; RRID:SCR_023221) to align reads to the GRCh38 reference genome. Single-cell feature count matrices were generated for each library using default parameters. Downstream analysis of count matrices in R was performed using the “Seurat” package (v4.0.3; RRID:SCR_016341). Empty droplets were removed using the “DropletUtils” package (v1.8.0) by first defining them as droplets with fewer than 100 unique RNA reads and then by selecting only droplets with RNA content notably dissimilar to empty droplets with an FDR-adjusted P value threshold of 0.01. Genes with zero expression across all cells were discarded from further analysis. Data from each sample were log-normalized and combined into a batch-corrected expression matrix using Canonical Correlation Analysis (RRID:SCR_021745). Counts were scaled and subjected to dimensionality reduction using principal component analysis (RRID:SCR_014676). Uniform Manifold Approximation and Projection for Dimension Reduction (UMAP; RRID:SCR_018217) embedding was generated from the top 25 dimensions of the principal component analysis. Clusters were identified initially by constructing a Shared Nearest Neighbor graph based on each cell’s k-nearest neighbors (k = 15 for PBMCs, k = 10 for tumor-infiltrating lymphocytes) and then applying the Smart Local Moving algorithm to the graph. Cluster-defining markers were then identified by comparing gene expression using Model-based Analysis of Single-cell Transcriptomics (RRID:SCR_016340).
Multiplex immunofluorescence
Formalin-fixed, paraffin-embedded tissue from tumor biopsies was analyzed using multiplexed immunofluorescence (mIF), digital image analysis, and machine learning to identify immune cell subsets in the TME. Biopsies were analyzed from all 10 patients in the SR1 trial, with paired biopsies analyzed from 6 out of the 10 patients. Three mIF antibody panels optimized for PDAC were used for analysis of T-cell subsets [CD3 (clone F7.2.38, Dako), CD4 (clone EP204, Cell Marque), CD8 (clone C8/144B, Dako), FOXP3 (clone 206D, Biolegend), CD45RO (clone UCHL1, Dako)], myeloid-cell subsets and ARG1 immunosuppressive potential [CD15 (clone Carb3, Dako), CD14 (clone SP192, Abcam), ARG1 (clone EPR6672, Abcam), HLA-DR (clone TAL1B5, Santa Cruz Biotechnology), CD33 (PWS44, Leica Biosystems)], and macrophage polarization status [IRF5 (EPR17067, Abcam), CD163 (clone 10D6, Leica Biosystems), CD68 (clone PG-M1, Dako), CD86 (E2G8P, Cell Signaling), MRC1 (CD206, clone E2L9N, Cell Signaling)]. Additionally, a separate panel was employed to assess PD1/PDL1 [CD8 (clone 4B11, Leica), FOXP3 (clone D608R, Cell Signaling), PD1 (clone EPR4877, Abcam), and PDL1 (clone E1L3N, Cell Signaling)]. All panels included 2-(4-amidinophenyl)-1H-indole-6-carboxamidine as a nuclear marker and anticytokeratin (clone AE1/AE3, Dako) to identify epithelial cells. Methodologic details for the immune cell mIF panels have been previously published (33, 34). In brief, staining was performed using Leica BOND RX Research Stainer (Leica Biosystems, Buffalo, IL). Slides stained for the PD1/PDL1 panel were digitized using a Mantra 2 multispectral imaging system (Akoya Biosciences, Hopkinton, MA), whereas T-cell subset, myeloid-cell subset, and macrophage polarization slides were digitized using a Vectra Polaris (PhenoImager HT) multispectral imaging system (Akoya Biosciences). For each slide, an average of 4 (range: 2–10) representative regions of interest containing residual, morphologically viable tumor epithelium were scanned at 200× magnification. Multispectral images were unmixed and processed to generate single-cell phenotypes using the inForm software package (Akoya Biosciences). The resultant single-cell level data were further analyzed using R v.4.0 (R Foundation for Statistical Computing, Vienna, Austria) to generate cell density measurements. Region-level density measurements from each panel were averaged to produce patient-level immune cell density measurements.
Statistical analysis
Pairwise group comparisons and correlation analyses were performed using the Wilcoxon matched-pair signed rank test, one-way ANOVA test, and Pearson coefficient calculation. GraphPad PRISM software version 10 was used for statistical analysis.
Data availability
The single-cell data generated in this study are publicly available in Gene Expression Omnibus under accession number GSE267814. Other data generated in this study are available in the manuscript and its supplementary files or upon request from the corresponding authors.
Results
Study design and clinical outcomes
We conducted an open-label multicenter phase Ib study evaluating gemcitabine, n(ab)-paclitaxel, canakinumab (ACZ885), and spartalizumab (PDR001). The trial is referred to herein as the SR1 trial (NCT04581343). Eligible patients had previously untreated metastatic PDAC and RECIST measurable disease. The primary objective of this phase Ib study was to establish the recommended phase II/III dose of canakinumab in combination with spartalizumab, gemcitabine, and n(ab)-paclitaxel. Ten patients were enrolled to evaluate the incidence of dose-limiting toxicities in the first 56 days of dosing in at least 6 evaluable patients utilizing a Bayesian logistic regression model. Treatment was continued until disease progression or unacceptable toxicity (Fig. 1A). Between November 2020 and May 2021, all patients were enrolled and received study treatment. The final patient completed the study in February 2023. Patient characteristics are shown in Table 1. There were no dose-limiting toxicities among the first 6 evaluable patients, and the recommended phase II/III dose was established as follows: canakinumab 250 mg subcutaneously on day 1; spartalizumab 400 mg intravenously (IV) on day 1; gemcitabine 1,000 mg/m2 IV on days 1, 8, and 15; and n(ab)-paclitaxel 125 mg/m2 IV on days 1, 8, and 15 of each 28-day cycle.
Patient characteristics . | SR1 (n = 10) . | Gemcitabine n(ab)paclitaxel (n = 11) . | Healthy donors (n = 6) . |
---|---|---|---|
Age | |||
Median | 65 | 72 | 62 |
Min, Max | 49–81 | 57–86 | 42–64 |
≥65 | 6 (60%) | 9 (82%) | 0 (0%) |
Sex | |||
Female | 5 (50%) | 6 (55%) | 6 (100%) |
Race/ethnicity | |||
Black | 1 (10%) | 1 (9%) | |
White non-Hispanic | 9 (90%) | 10 (91%) | |
Stage at initial diagnosis | |||
I | 1 (10%) | 3 (27%) | |
IV | 9 (90%) | 8 (73%) | |
Site of the primary tumor | |||
Head | 5 (50%) | 6 (55%) | |
Neck | 2 (20%) | 0 (0%) | |
Body | 1 (10%) | 1 (9%) | |
Tail | 2 (20%) | 4 (36%) | |
Presence of liver mets | |||
Yes | 7 (70%) | 6 (55%) | |
No | 3 (30%) | 5 (45%) | |
Baseline CA19-9 | |||
Mean | 750 | 1,108 | |
Median | 70 | 49 | |
Prior exposure to FOLFIRINOX | |||
Yes | 1 (10%) | 2 (18%) | |
No | 9 (90%) | 9 (82%) | |
Prior surgical removal of the primary tumor | |||
Yes | 1 (10%) | 1 (9%) | |
No | 9 (90%) | 10 (91%) |
Patient characteristics . | SR1 (n = 10) . | Gemcitabine n(ab)paclitaxel (n = 11) . | Healthy donors (n = 6) . |
---|---|---|---|
Age | |||
Median | 65 | 72 | 62 |
Min, Max | 49–81 | 57–86 | 42–64 |
≥65 | 6 (60%) | 9 (82%) | 0 (0%) |
Sex | |||
Female | 5 (50%) | 6 (55%) | 6 (100%) |
Race/ethnicity | |||
Black | 1 (10%) | 1 (9%) | |
White non-Hispanic | 9 (90%) | 10 (91%) | |
Stage at initial diagnosis | |||
I | 1 (10%) | 3 (27%) | |
IV | 9 (90%) | 8 (73%) | |
Site of the primary tumor | |||
Head | 5 (50%) | 6 (55%) | |
Neck | 2 (20%) | 0 (0%) | |
Body | 1 (10%) | 1 (9%) | |
Tail | 2 (20%) | 4 (36%) | |
Presence of liver mets | |||
Yes | 7 (70%) | 6 (55%) | |
No | 3 (30%) | 5 (45%) | |
Baseline CA19-9 | |||
Mean | 750 | 1,108 | |
Median | 70 | 49 | |
Prior exposure to FOLFIRINOX | |||
Yes | 1 (10%) | 2 (18%) | |
No | 9 (90%) | 9 (82%) | |
Prior surgical removal of the primary tumor | |||
Yes | 1 (10%) | 1 (9%) | |
No | 9 (90%) | 10 (91%) |
Treatment-related adverse events were noted in all patients, including adverse events of grade 3 or higher in 7 patients (70%). These were generally consistent with those seen with chemotherapy and were predominately hematologic (Supplementary Table S1). One patient experienced pneumonitis leading to discontinuation of spartalizumab, and 1 patient experienced GI bleeding that led to discontinuation of all study agents. No patients experienced fatal (grade 5) adverse events.
Efficacy was evaluated in all 10 patients. The median progression-free survival (PFS) was 4.6 months (95% CI, 1.9–7.6), and the median overall survival (OS) was 12.9 months (95% CI, 3.7–21.7), with 60% of patients alive at 12 months (Supplementary Fig. S1). Objective response rate was evaluated based on RECIST v1.1. Two patients had a confirmed partial response (20%) with one additional unconfirmed partial response and 5 patients with stable disease as the best response for a disease control rate of 80% (8 of 10; Fig. 1B). There were no complete responses. Although the sample size in this phase I trial was limited, PFS and OS are comparable with that seen with gemcitabine/n(ab)paclitaxel (PFS = 5.5 months, OS = 8.5 months; ref. 35). Representative cross-sectional images of a responding patient are shown in Fig. 1C. Patients underwent baseline and on-study blood collection for multiple correlative translational research objectives, including pharmacokinetic analysis (Supplementary Table S2; Supplementary Fig. S2). Paired biopsies were also obtained with tissue collection at baseline and at 8 weeks (end of cycle 2).
IL1β and PD1 blockade induce modest reactivation of peripheral blood CD8+ T cells
Patients in the SR1 trial received two immunotherapy agents, canakinumab and spartalizumab. We also collected peripheral blood from 11 patients with PDAC who were receiving standard-of-care GnP (Table 1). PFS in our GnP cohort was 3.9 months (95% CI, 0.76–NA), consistent with expected outcomes (Supplementary Fig. 1C; ref. 35). We first wanted to verify that the targets of the immunotherapy agents were expressed in patients in SR1 and successfully blocked by treatment. All patients in the study expressed detectable serum IL1β levels at baseline and on-treatment that averaged sevenfold higher than age-matched healthy controls, indicated by the dashed line in Fig. 2A (Fig. 2A; Supplementary Table S3). Canakinumab itself did not interfere with detection of serum IL1β due to nonoverlapping epitopes with the antibody used for detection of serum IL1β (Supplementary Fig. S3). PD1 was expressed on a fraction of peripheral blood CD8+ T cells in all SR1 patients at baseline, as assessed by flow cytometry (Fig. 2B). Detection of surface PD1 expression decreased in on-treatment samples from patients in SR1 due to occlusion of the flow cytometry antibody epitope by spartalizumab; this effect was not observed in samples from patients receiving standard-of-care GnP (Fig. 2B). These results confirm on-target binding of spartalizumab.
PD1 blockade in other tumor types induces reinvigoration of CD8+ T cells, which can be observed in peripheral blood as a proliferative burst of preexisting T-cell clonotypes as well as newly expanding clonotypes that enter into the cell cycle (36). This induction of proliferation occurs in the peripheral blood of patients with pancreatic cancer as well, although the reinvigorated T cells show a gene signature of NF-κB expression consistent with nonproductive antitumor immunity (32). Herein, we used the flow cytometry markers CD38 and HLA-DR, which have been reported to overlap with Ki67+ cycling CD8+ T cells (37). We found that peripheral blood antigen-experienced CD8+ T cells contained a higher fraction of CD38+HLA-DR+ reinvigorated cells on treatment in patients in SR1 (Fig. 2C and D; Supplementary Fig. S4A); however, the magnitude of this increase is less than that seen in checkpoint blockade–responsive tumor types such as melanoma, lung cancer, or head and neck cancers (37, 38).
Consistent with the modest degree of CD8+ T-cell reinvigoration seen by flow cytometry, serum levels of the T cell–produced cytokines IL2 and IFNγ were not substantially changed by treatment in either the SR1 or GnP cohorts (Fig. 2E and F; Supplementary Table S3). The IFN-inducible chemokine CXCL9 was notably increased by treatment in the SR1 but not the GnP control cohorts, suggesting a slight induction of a systemic IFN response in patients receiving the PD1 blockade containing regimen (Fig. 2G; Supplementary Table S3).
IL1β blockade decreases peripheral monocytic MDSCs
Immature myeloid cells of granulocytic and monocytic origin can arise in the peripheral blood of patients with pancreatic cancer, contributing to systemic immune suppression and supplying the tumor with immunosuppressive cells (39, 40). We used flow cytometry to quantify MDSCs from PMBCs (Supplementary Fig. S4B). Monocytic MDSCs (MoMDSCs, CD14+CD33+HLA-DRlow) represented approximately 5% of total CD45+ cells at baseline across patients with pancreatic cancer (Fig. 3A). In our GnP cohort, this fraction remained similarly high; however, treatment with immunotherapy considerably reduced MoMDSCs in the patients in SR1, as a fraction of total CD45+ cells (Fig. 3A) and as a fraction of total monocytes (Fig. 3B and C).
To investigate which factors in serum might affect the differentiation of MoMDSCs, we adapted a serum suppression assay previously used to demonstrate MoMDSC induction in patients with sepsis, COVID-19, or pancreatic cancer (39, 41). Healthy donor monocytes were differentiated for 7 days in media containing serum from patients with pancreatic cancer. The resulting macrophages/MDSCs were cocultured with CFSE-labeled healthy donor T cells and provided with anti-CD3/CD28 stimulation to determine the degree to which T-cell proliferation was suppressed (Fig. 3D). As a positive control, monocytes differentiated into MDSCs in vitro by addition of recombinant GM-CSF and IL6 (iMDSCs) were cultured with CFSE-labeled T cells, and the extent of suppression of T-cell proliferation was determined. Paired analysis of serum from baseline and on-treatment samples showed that MDSC differentiation potential decreased (i.e., T-cell proliferation increased) in patients in SR1 but not in patients in GnP control (Fig. 3E). These results suggest that serum factors influencing MDSC differentiation may be altered by immunotherapy treatment in patients with PDAC. To test whether IL1β itself could affect MDSC differentiation, we added recombinant IL1β to healthy serum and differentiated monocytes for 7 days. Unlike the positive control iMDSCs, our IL1β-differentiated monocytes had high expression of MHC class II and did not suppress the proliferation of activated T cells (Supplementary Fig. S5A and S5B). We therefore conclude that IL1β alone is not sufficient for MoMDSC differentiation in this reductionist in vitro assay.
To determine which serum factors might change with immunotherapy treatment, we analyzed patient serum for 71 analytes by cytokine bead arrays (Supplementary Table S3). GM-CSF is known to induce MDSCs, but it was below the limit of detection in SR1 patient serum. IL6 was another likely candidate, as IL6 in combination with GM-CSF robustly induces MDSCs and IL6 has been reported to decrease in other trials of canakinumab, or in mouse models in which IL1β is blocked. However, serum IL6 was not substantially changed with treatment in patients in SR1 (Fig. 3F; Supplementary Table S3), suggesting that unlike in other cancer types, IL6 in pancreatic cancer is not IL1β dependent. Of the cytokines and chemokines tested, the three most likely responsible for the observed differences in MDSC differentiation were IL18, FLT3L, and TNFα, all of which were increased with treatment in patients in SR1 but not increased by treatment with GnP alone (Fig. 3G–I; Supplementary Table S3). Addition of recombinant FLT3L or TNFα to iMDSC differentiation cultures in vitro prevented the development of MDSCs, whereas IL18 had negligible effects (Supplementary Fig. S5C and S5D). This demonstrates that two cytokines that increase with treatment in the SR1 cohort are capable of preventing MDSC differentiation in vitro, and we hypothesize that FLT3L and/or TNFα may plausibly contribute to the decreased development of MoMDSCs in patients receiving blockade of IL1β and PD1. Further analysis of these two cytokines and their role in cancer-induced myelopoiesis is warranted.
Blockade of IL1β and PD1 shifts the transcriptional profile of intratumoral myeloid cells
Baseline and on-treatment freshly dissociated biopsies were obtained from patients in the SR1 trial; 4 patients had paired biopsies, and 4 had single biopsies (Supplementary Table S2). Samples were analyzed by scRNA-seq using the 10x Genomics platform. Global analysis of cell populations revealed tumor cells, myeloid cells, T cells, NK cells, and fibroblasts (Supplementary Fig. S6A–S6C). IL1B transcripts were robustly found in myeloid cells, with the highest expression coming from MDSCs (Supplementary Fig. S6D). IL1R was expressed by most cell types, as expected given the pleiotropic nature of this cytokine (Supplementary Fig. S6D). PDCD1 was expressed predominantly by T and NK cells, whereas PDCD1LG1 and PDCD1LG2 were detected at low levels in myeloid cells and infrequently in tumor cells (Supplementary Fig. S6D). Global myeloid-cell frequencies were not considerably altered when comparing baseline and on-treatment samples (Supplementary Fig. S6E).
We next subclustered the myeloid-cell populations to determine if transcriptional differences were induced by treatment. Consistent with prior reports, PDAC contained predominantly neutrophilic MDSCs and macrophages, with smaller populations of monocytes, dendritic cells, and cycling cells (Fig. 4A–C; refs. 42, 43). Although the fractional representations of each of the myeloid-cell clusters were not remarkably changed with treatment, we observed a trend toward increased macrophages and decreased MDSCs in on-treatment samples, particularly in patients who achieved partial response by RECIST criteria (Fig. 4D). Comparison of gene expression changes revealed considerable increases in predominantly macrophage-related genes such as APOE, C1QC, SPP1, CTSL, and MHC class I and class II genes (Fig. 4E). Genes related to MDSC function were substantially decreased with treatment, including genes related to inflammasome activation and inflammatory cytokine production (IL1B, NLRP3, TNF, S100A9, and S100A8), reflecting either a phenotypic shift or a change in cell-type frequencies (Fig. 4E). Overall, single-cell analysis of paired biopsies revealed a trend toward a shift in myeloid cells away from an MDSC phenotype, although too few patients were available to perform a confirmatory analysis.
Multiplex immunofluorescence reveals increase in CD8+ T-cell to regulatory T-cell ratio
We successfully obtained formalin-fixed, paraffin-embedded paired biopsies from 6 of the 10 patients enrolled in SR1 and either baseline or on-treatment biopsies were available from the remaining 4 patients. To characterize alterations in the immune microenvironment, we employed customized, spatially resolved mIF using previously designed panels for evaluating PD1 and PDL1 expression, T-cell subsets, macrophage polarization status, and MDSCs (33, 34, 44). PDL1 protein expression on tumor or stromal cells was low to undetectable in all patients at baseline and did not increase with treatment (Fig. 5A–C). CD8+ T cells were detected in all patient samples, although PD1 was expressed by a minority of tumor-infiltrating cells and did not increase with treatment (Fig. 5D–G). Previous groups have shown that CD8+ T cells in PDAC are predominantly located outside of tumor-cell areas (45, 46). We also observed a higher density of CD8+ cell staining in stromal regions as compared with intraepithelial regions (Fig. 5F and G). In 4 out of 6 patients with paired biopsies, intraepithelial CD8+ cell staining increased with treatment, and, in at least 1 patient who achieved a partial response by RECISTv1.1 criteria, we observed activated CD8+ T cells (CD8+CXCL13+) directly touching tumor cells (Fig. 5E–G). Although no considerable increases were observed in total T cells, total CD4+ T cells, total CD8+ T cells, or total regulatory T cells (Treg; Fig. 5H–M), likely due to the small sample size, we did observe an increase in the CD8+ T-cell to Treg ratio in all 6 patients for whom paired biopsies were available (Fig. 5N). These results suggest that although PD1 and its ligands are not highly expressed in pancreatic cancer, the combination of IL1β blockade and PD1 blockade modestly improved the T-cell landscape in PDAC.
Intratumoral myeloid cells are not strongly correlated with peripheral blood frequencies
We next analyzed macrophage polarization in the tumors of patients in SR1. Consistent with our prior findings, CD206+ M2-like macrophages and CD86+IRF5+ M1-like macrophages were both detectable (Supplementary Fig. S7A; refs. 33, 34). We did not observe any considerable changes in total macrophages or in macrophage polarization state when comparing baseline and on-treatment samples (Supplementary Fig. S7B–S7D). We also did not observe considerable shifts in total CD14+ monocyte-derived cells or CD15+ neutrophil-derived cells (Fig. 6A–C). Immunosuppressive granulocytic MDSCs (GrMDSC) can be defined as CD15+ cells that express arginase (ARG1). Staining for ARG1 revealed that most CD15+ cells expressed arginase, but that treatment did not lead to a considerable difference in tumor GrMDSC frequencies (Fig. 6D).
MoMDSCs, defined as CD14+CD33+HLA-DRlow cells (40), were decreased with treatment in the peripheral blood of patients in SR1, as assessed by flow cytometry (Fig. 3A–C). Using these same markers in the mIF analysis, we did not observe a substantial change in MoMDSCs in patient tumors upon treatment (Fig. 6E). To test whether peripheral blood MoMDSC frequencies were associated with intratumoral MoMDSC frequencies, we performed a linear regression analysis of peripheral blood MoMDSCs from pre- and on-treatment time points with intratumoral MoMDSC values from the same time points (Fig. 6F). These two parameters were poorly correlated, with an R2 value of 0.095 and a P value of 0.28. We next tried defining intratumoral MoMDSCs using ARG1 expression. Tumor CD14+ARG1+ cells were not remarkably changed by treatment, although the 2 patients with the highest MoMDSCs in baseline samples decreased more than twofold (Fig. 6G). Peripheral blood MoMDSCs and tumor CD14+ARG1+ cells were better correlated (R2 = 0.21), yet the positive trending correlation did not reach statistical significance (P = 0.098; Fig. 6H). Thus, although all but 1 patient in SR1 showed a decrease in peripheral blood MoMDSCs (Fig. 3A–B), this pattern was not fully reflected in the TME.
Discussion
Here, we report clinical and translational results from a phase I trial testing the hypothesis that IL1β blockade in human PDAC would alleviate myeloid immunosuppression and reveal antitumor T-cell responses to PD1 blockade. Indeed, we demonstrate that systemic MoMDSC frequencies were reduced in patients receiving combination therapy, thus showing that elevated IL1β in patients with PDAC is causally linked to MoMDSCs. Unfortunately, this decrease in systemic myelosuppression did not notably change suppressive myeloid-cell populations or activated T cells in the tumor. However, the trend toward an increase in CD8+ T cells suggests that improvements in the tumor immune microenvironment might be revealed by a larger study or by more potent therapeutic combinations addressing mechanisms of myeloid suppression of antitumor T-cell responses.
Given that the tested treatment program successfully decreased MoMDSCs in circulation, the lack of change in the TME is puzzling. Tumor-associated macrophages in pancreatic cancer are derived from infiltrating monocytes as well as from local proliferation of tissue-resident macrophages (47). Macrophages may be reprogrammed for antitumor function, and the presence of iNOS-expressing macrophages in proximity to tumor nests confers a favorable prognosis in PDAC (33, 48). MDSCs, however, are a relatively short-lived population sustained by an influx of monocytes and neutrophils from circulation (49–51). Mice deficient in CCR2 or treated with antibodies or small molecule inhibitors of CCR2 or its ligand CCL2 have dramatically reduced tumor MDSCs (49–52). In humans, blockade of monocyte trafficking using a CCR2 inhibitor in patients with locally advanced pancreatic cancer led to consistent reductions in tumor burden, a decrease in circulating monocytes, and a decrease in intratumoral CD11b+ cells (51). In our study, GrMDSCs were difficult to quantify due to variable loss of granulocytes during the PBMC separation and freezing process (40). We therefore focused on MoMDSCs in peripheral blood and tissue, although we also noted that CD15+ neutrophils were not considerably altered in tumors as assessed by mIF. In patients in SR1 trial, IL1β blockade led to similar reductions in circulating monocytes and MoMDSCs, but the intratumoral MoMDSC frequencies were not decreased. This apparent discrepancy could be due to several factors. First, decreasing MoMDSCs systemically can lead to decreases in the tumor, but it may require sustained inhibition of trafficking. The half-life of MDSCs in tumors is 2 to 4 days in mice although it is difficult to measure in humans (53, 54). It is possible that transient decreases in systemic MoMDSCs may not be adequate to affect tumor MoMDSC frequencies. The correlation between circulating and intratumoral MoMDSCs was weak in our study, suggesting that other factors contribute to recruiting and sustaining intratumoral MoMDSCs. Second, the blockade of CCR2 affects trafficking, not just cell numbers (51). Therefore, CCR2 inhibition would be expected to better curtail tissue influx than drug therapies such as IL1β blockade, which affect cell numbers and phenotype but have no direct effect on extravasation into tumors. Third, it is possible that the effects of IL1β blockade are transient, and that reexposure to IL1β in the TME may restore MoMDSC survival. Although serum concentrations of canakinumab remained above 8,000 ng/mL for the duration of the 28-day treatment cycle, we were unable to assess intratumoral levels of canakinumab and thus have no direct evidence that the drug adequately penetrates the TME.
Fibroblasts in PDAC interact closely with cancer and immune cells, in addition to producing the desmoplastic reaction characteristic of PDAC. At least two major subsets of cancer-associated fibroblasts exist in PDAC: myofibroblasts (myCAFs), which are juxtaposed to cancer cells, respond to TGFβ signaling and express αSMA and inflammatory fibroblasts (iCAFs), which are further away from cancer cells, respond to IL1β signaling and secrete IL6 (55–57). In mouse models, iCAFs and the IL6 they produce are tumor-promoting, leading to the hypothesis that selective disruption of iCAFs or the mediators they produce would be beneficial (55, 56). In our scRNA-seq analysis, we obtained very few fibroblasts. These fibroblasts did not neatly map to either iCAF or myCAF subsets, making it difficult to assess whether IL1β blockade in human PDAC affected fibroblast heterogeneity. A larger cohort study of canakinumab, spartalizumab, gemcitabine, and n(ab)paclitaxel (NCT04229004) has been initiated and will provide more data to address this question.
In the CANTOS trial, IL1β blockade seemed to control incipient primary lung cancers (8), but subsequent prospective trials of canakinumab in established tumors have been unable to show a clinical benefit (58–60). Although our 10-patient trial was not powered for efficacy, we note that the median OS time is only modestly improved compared with that reported with gemcitabine/n(ab)paclitaxel alone (35). The CANTOS trial enrolled patients with elevated innate inflammation, defined as having serum C-reactive protein greater than 2 mg/L (9). Only 3 of the 10 patients in our study met this criteria, and concentrations of IL1β itself were similarly variable. Of the 3 patients in SR1 with C-reactive protein greater than 2 mg/L, 1 patient achieved an unconfirmed partial response, whereas the other two did not derive clinical benefit. It is possible that blockade of IL1β or other innate inflammatory cytokines may be warranted in a subset of patients with established tumors and high chronic inflammation.
The quadruple therapy of canakinumab, spartalizumab, gemcitabine, and n(ab)paclitaxel is being further tested in an expansion cohort as part of the Precision Promise adaptive trial platform for metastatic pancreatic cancer (NCT04229004). Clinical outcomes and analysis of paired blood and tissue biopsies from these additional patients will further clarify the definitive role of IL1β as part of a combinatorial regimen in human PDAC. Our current study leads us to hypothesize that circulating MoMDSCs will consistently decrease in all patients receiving the drug combination, and clinical response will depend on these systemic changes translating to intratumoral decreases in MoMDSCs and increases in activated CD8+ T cells through as-of-yet undefined mechanisms.
Authors’ Disclosures
P.E. Oberstein reports other support from Novartis during the conduct of the study, as well as personal fees from Ipsen and Lilly outside the submitted work. O.E. Rahma reports personal fees from Merck, Celgene, Five Prime, GSK, Bayer, Roche/Genentech, Puretech, Imvax, Sobi, Boehringer Ingelheim and hold pending patent for “Methods of using pembrolizumab and trebananib” outside the submitted work. O.E. Rahma is a full-time associate of AstraZeneca not during the time period of the submitted work. H. Singh reports grants from AstraZeneca and other support from Dewpoint Therapeutics, Merck Sharpe & Dohme, Dava Oncology, and UpToDate outside the submitted work. T.A. Abrams reports grants and personal fees from AstraZeneca and personal fees from Eisai, Elevar, HistoSonics, and Sirtex outside the submitted work. J.M. Cleary reports personal fees from Blueprint Medicines, Incyte, and AstraZeneca and grants from Amgen, Merus, Roche, Servier, BMS, Esperas Pharma, AstraZeneca, Merck, Pyxis Oncology, Arcus, and GSK outside the submitted work. B.M. Huffman reports personal fees from Lilly outside the submitted work. K.J. Perez reports personal fees from Ipsen, Novartis, and Lantheus outside the submitted work. D.A. Rubinson reports personal fees from Taiho, Sirtex, and Instylla outside the submitted work. B.L. Schlechter reports Agemis and Janssen outside the submitted work. R. Surana reports personal fees from Grupo Oncoclinicas and other support from DAVA Oncology outside the submitted work. M.B. Yurgelun reports grants from Janssen and personal fees from Nouscom and UpToDate outside the submitted work. K.M. Sullivan reports personal fees from Eli Lilly outside the submitted work. M. Dougan reports personal fees from Neoleukin, Genentech, Partner Therapeutics, SQZ Biotech, AzurRx, Eli Lilly, Mallinckrodt Pharmaceuticals, Aditum, Foghorn Therapeutics, Palleon, Sorriso Pharmaceuticals, Generate Biomedicines, Asher Bio, Alloy Therapeutics, Third Rock Ventures, DE Shaw Research, Agenus, Astellas, and Curie Bio and other support from Veravas, Monod Bio, Axxis Bio, and Cerberus Therapeutics outside the submitted work. M. Dajee is a full-time associate of Novartis Institute for Biomedical Research and reports other support from Novartis Cooperation outside the submitted work. M.R. Pelletier is a full-time associate of Novartis Institute for Biomedical Research. S. Nazeer is a full-time associate of Novartis Pharma AG and a member of the clinical study team. M. Squires is a full-time associate of Novartis Pharma AG. B.M. Wolpin reports personal fees from Mirati; grants from Novartis; grants and personal fees from Revolution Medicines and Harbinger; personal fees from GRAIL, Ipsen, EcoR1 Capital, Third Rock Ventures, and Agenus; and grants from AstraZeneca and Eli Lilly outside the submitted work. J.A. Nowak reports grants from Natera and personal fees from Leica Biosystems outside the submitted work. D.M. Simeone reports grants from Novartis during the conduct of the study; grants from Micronoma, Immunovia, ClearNote Health, Biological Dynamics, and Tempus; personal fees from ClearNote Health, Interpace, Merck & Co., Bayer, and FibroGen outside of the submitted work. S.K. Dougan reports grants from Novartis during the conduct of the study; grants from BMS and Takeda; personal fees and other support from Kojin Therapeutics; and other support from Axxis Bio outside the submitted work. No disclosures were reported by the other authors.
Authors’ Contributions
P.E. Oberstein: Conceptualization, formal analysis, supervision, investigation, writing–original draft. A. Dias Costa: Formal analysis, investigation, methodology, writing–review and editing. E.A. Kawaler: Formal analysis, investigation, visualization, methodology, writing–review and editing. V. Cardot-Ruffino: Formal analysis, investigation, methodology, writing–review and editing. O.E. Rahma: Supervision, investigation, writing–review and editing. N. Beri: Resources, investigation, writing–review and editing. H. Singh: Resources, data curation, supervision, investigation, writing–review and editing. T.A. Abrams: Resources, writing–review and editing. L.H. Biller: Resources, writing–review and editing. J.M. Cleary: Resources, writing–review and editing. P. Enzinger: Resources, writing–review and editing. B.M. Huffman: Resources, writing–review and editing. N.J. McCleary: Resources, writing–review and editing. K.J. Perez: Resources, writing–review and editing. D.A. Rubinson: Resources, writing–review and editing. B.L. Schlechter: Resources, writing–review and editing. R. Surana: Resources, writing–review and editing. M.B. Yurgelun: Resources, writing–review and editing. S.J. Wang: Investigation, writing–review and editing. J. Remland: Data curation, project administration, writing–review and editing. L.K. Brais: Data curation, project administration, writing–review and editing. N. Bollenrucher: Investigation, writing–review and editing. E. Chang: Investigation, writing–review and editing. L.R. Ali: Data curation, formal analysis, writing–review and editing. P.J. Lenehan: Data curation, investigation, writing–review and editing. I. Dolgalev: Data curation, investigation, writing–review and editing. G. Werba: Investigation, writing–review and editing. C. Lima: Investigation, writing–review and editing. C.E. Keheler: Resources, data curation, writing–review and editing. K.M. Sullivan: Resources, data curation, writing–review and editing. M. Dougan: Resources, data curation, supervision, writing–review and editing. C. Hajdu: Resources, writing–review and editing. M. Dajee: Formal analysis, project administration, writing–review and editing. M.R. Pelletier: Formal analysis, project administration, writing–review and editing. S. Nazeer: Formal analysis, project administration, writing–review and editing. M. Squires: Supervision, project administration, writing–review and editing. D. Bar-Sagi: Conceptualization, supervision, writing–review and editing. B.M. Wolpin: Conceptualization, supervision, funding acquisition, writing–review and editing. J.A. Nowak: Conceptualization, supervision, writing–review and editing. D.M. Simeone: Conceptualization, supervision, funding acquisition, writing–review and editing. S.K. Dougan: Conceptualization, formal analysis, supervision, funding acquisition, investigation, visualization, writing–original draft.
Acknowledgments
This clinical trial and accompanying translational research were supported by the Pancreatic Cancer Action Network and Novartis Pharmaceuticals. V. Cardot-Ruffino was funded by a fellowship from the Pancreatic Cancer Action Network and the Francois Wallace Monahan Fund in loving memory of Michael Insel. S.K. Dougan, H. Singh, M.B. Yurgelun, J.M. Cleary, R.Surana, K.J. Perez, L.K. Brais, J.Remland, S.J. Wang, J.A. Nowak, and B.M. Wolpin were funded by the DFCI Hale Family Center for Pancreatic Cancer Research. S.K. Dougan was also funded by research grants from Novartis, the Ludwig Center at Harvard, Break Though Cancer, NIH R01AI158488, R01AI169188, U01 CA224146, and U01 CA274276 and is a member of the Parker Institute for Cancer Immunotherapy. P.E. Oberstein, E.A. Kawaler, N. Beri, I. Dolgalev, G. Werba, C. Hajdu, and D.M. Simeone were supported by the Perlmutter Cancer Center at NYU Langone Health. D.M. Simeone was also funded by research grants from Novartis, the Pancreatic Cancer Action Network, NIH R01CA245005, and NIH R01CA206105-01A1. We acknowledge Raja Narayan for statistical analysis. We acknowledge James Dougan for data analysis and Katherine Dougan for artistic support.
Note: Supplementary data for this article are available at Cancer Immunology Research Online (http://cancerimmunolres.aacrjournals.org/).