Pancreatic ductal adenocarcinoma (PDAC) is a lethal cancer that has proven refractory to immunotherapy. Previously, treatment with the DNA hypomethylating drug decitabine (5-aza-dC; DAC) extended survival in the KPC-Brca1 mouse model of PDAC. Here we investigated the effects of DAC in the original KPC model and tested combination therapy with DAC followed by immune checkpoint inhibitors (ICI). Four protocols were tested: PBS vehicle, DAC, ICI (anti-PD-1 or anti-VISTA), and DAC followed by ICI. For each single-agent and combination treatment, tumor growth was measured by serial ultrasound, tumor-infiltrating lymphoid and myeloid cells were characterized, and overall survival was assessed. Single-agent DAC led to increased CD4+ and CD8+ tumor-infiltrating lymphocytes (TIL), PD1 expression, and tumor necrosis while slowing tumor growth and modestly increasing mouse survival without systemic toxicity. RNA-sequencing of DAC-treated tumors revealed increased expression of Chi3l3 (Ym1), reflecting an increase in a subset of tumor-infiltrating M2-polarized macrophages. While ICI alone had modest effects, DAC followed by either of ICI therapies additively inhibited tumor growth and prolonged mouse survival. The best results were obtained using DAC followed by anti-PD-1, which extended mean survival from 26 to 54 days (P < 0.0001). In summary, low-dose DAC inhibits tumor growth and increases both TILs and a subset of tumor-infiltrating M2-polarized macrophages in the KPC model of PDAC, and DAC followed by anti-PD-1 substantially prolongs survival. Because M2-polarized macrophages are predicted to antagonize antitumor effects, targeting these cells may be important to enhance the efficacy of combination therapy with DAC plus ICI.

Significance:

In a pancreatic cancer model, a DNA hypomethylating drug increases tumor-infiltrating effector T cells, increases a subset of M2 macrophages, and significantly prolongs survival in combination with immune checkpoint inhibitors.

See related commentary by Nephew, p. 4610

Pancreatic ductal adenocarcinoma (PDAC) is a lethal malignancy that has proven refractory to both conventional cytotoxic and targeted therapies. Notably, despite recent remarkable results with immunotherapy against other cancers, responses in PDAC have been elusive (1–3). Thus, developing more effective therapies using combinations of immune checkpoint inhibitors (ICI) with other agents is being actively pursued (e.g., refs. 4–8 and references therein). Epigenetic alterations are prominent in PDACs (9), and epigenetically acting drugs are receiving increasing interest for possible utility against treatment refractory tumors. In a rapidly progressive mouse model of PDAC, KPC-Brca1, we previously found that treatment with decitabine (5-aza-dC; DAC), a well-tolerated DNA hypomethylating drug, produced changes in DNA methylation in the epithelial and stromal cells of the tumors, inhibited tumor cell proliferation, and extended the survival of the animals (10). Given this background, we have now extended our approach to combination therapy, using the well characterized and highly chemoresistant KPC mouse model (KrasLSL.G12D/+; P53LSL.R172H/+; Pdx1Cretg/+), treated with single-agent DAC and with mAbs against two ICIs: PD-1 and V-domain Ig suppressor of T-cell activation (VISTA; also known as PD-1H). The slower progressing KPC model is more widely used than KPC-Brca1, which facilitates comparisons of results across studies. Our results show that single-agent DAC induces tumor necrosis and inhibits tumor growth, without systemic toxicities, and that this hypomethylating agent alters specific subpopulations of tumor-infiltrating lymphoid and myeloid cells in the KPC model. We find that sequential combinations of DAC plus the ICIs are more effective than either type of agent alone in slowing tumor growth, and that DAC followed by anti-PD-1 substantially prolongs mouse survival.

Mouse strains

Procedures relating to the care and use of mice in this study were conducted in accordance with NIH guidelines, with the animals housed in an Association for Assessment and Accreditation of Laboratory Animal Care-accredited facility. The KrasG12D/+;Trp53R172H/+;Pdx1-Cre (KPC) mice have been described previously (11). All animal studies were carried out in accordance with approved Institutional Care and Use Committee protocols for Columbia University Irving Medical Center (New York, NY; AC-AAAU2462).

Ultrasound imaging of tumors

Ultrasound imaging of normal and tumor-bearing mouse pancreata using the Vevo 770 System with a 35 MHz RMV scanhead (Visual Sonics, Inc.) was performed in the Small Animal Imaging Shared Resource of the Herbert Irving Comprehensive Cancer Center (HICCC, New York, NY), and serial images at 0.25-mm intervals were collected as described previously (12). Tumor boundaries were marked on the ultrasound images and Vevo 770 software calculated a collective tumor volume based on close approximation of tumor topography. The tumor doubling time was calculated using the formula {T_d}\ = {\ ( {{t_2} - {t_1}} ) \times {\frac{{\log ( 2 )}}{{\log ( {\frac{{{q_2}}}{{{q_1}}}} )}}$⁠, where q1 is the tumor volume at treatment enrollment and q2 is the tumor volume at death or euthanasia, at their respective times, t1 and t2.

Treatment paradigms evaluating autochthonous tumors

We utilized three different types of treatment protocols (Supplementary Fig. S1A–S1C). In Paradigm I (n = 65), designed with a survival endpoint, we enrolled KPC mice after they developed 3–5 mm tumors, as measured by high-resolution ultrasound. Because of potential differences in mouse survival depending on the location of the tumors, in this treatment paradigm we only enrolled mice with tumors localized in the tail of the pancreas (TOP). Mice received 1 μg/g body weight of DAC via intraeritoneal injection every 72 hours for 3 doses, or PBS as vehicle control, then either PBS or 100 μg anti-mouse PD-1H/VISTA (BioLegend) or anti-mouse PD-1 (Merck) IP every 72 hours for 3 doses. In the combination treatments (DAC + anti-VISTA or anti-PD-1), the mice received the first dose of the immune checkpoint inhibitor at 72 hours after the last DAC dose. Mice were sacrificed when moribund as defined by a scoring system of physical and behavioral symptoms. In Paradigm II (n = 8) we assessed the effects of treatment with DAC on head-of-pancreas (HOP) tumors, using a timed endpoint. Mice received 1 μg/g body weight of DAC, or PBS as control, via intraperitoneal injection every 72 hours for 3 doses. Mice were euthanized and necropsied at 72 hours after completion of treatment. In Paradigm III (n = 12) we assessed the effects of DAC when the drug was started earlier, when most mice have only pancreatic situ neoplasia (PanIN). Mice at 4 months of age, without ultrasound-detectable tumors, were given 1 μg/g body weight of DAC intraperitoneally, or PBS vehicle in the controls, every 72 hours for 5 doses, and were euthanized and necropsied at 72 hours after the last injection.

Tumor cell allografting

Tumor tissue was resected from KPC mice within our breeding colony and was trimmed and washed in sterile PBS and minced. Tumor fragments were incubated with collagenase (1 mg/mL in DMEM) at 37°C for 45 minutes, filtered to remove nondissociated fragments and debris, and the resulting single-cell suspension plated and cultured in DMEM containing 10% FCS to obtain a purified tumor cell population. After no more than 10 passages, the tumor cells were injected subcutaneously at 2 × 106 cells in each flank of litter-mate recipient mice, which did not carry Pdx1-Cre and did not have autochthonous tumors. Starting at 6 days after the injection of the tumor cells, when small tumors were palpable, the mice received 1 μg/g body weight of DAC or the equivalent volume of PBS intraperitoneally every 72 hours for 3 doses and were sacrificed 48 hours after the last dose, to obtain the tumors for size measurements and histologic studies.

In vitro DAC response

Primary KPC tumor cells, isolated as above, were seeded at 5 × 105 cells per 100-mm plate and cultured for 24 hours in DMEM containing 10% FCS. DAC was then added to achieve a range of final concentrations (see Results), and the medium, with the drug, replaced every 24, 48, or 96 hours before harvesting the cells for DNA and RNA extraction.

Methylation-sensitive pyrosequencing

Methylation-sensitive pyrosequencing (MS-Pyroseq) for determining CpG methylation levels in B1 repetitive elements was performed at EpigenDx using genomic DNA samples from control and DAC-treated KPC cells maintained in tissue culture, and from KPC tumors obtained from the in vivo experiments.

IHC and Immunofluorescence

IHC and immunofluorescence (IF) were performed on formalin-fixed paraffin-embedded (FFPE) tissue sections using primary antibodies against mouse CD4 (eBioscience EPR19514; 1:50), CD8 (Thermo Fisher Scientific; MA1-70041; 1:50), FoxP3 (Santa Cruz Biotechnology; sc-53876; 1:500), CD163 (Abcam; ab182422; 1:500), CD103 (BD Biosciences M290 FitC; 1:200), F4/80 (eBioscience; 14-4801-85; 1:1,000), Ki67 (Abcam; ab15580; 1:500), Arginase-1 (Santa Cruz Biotechnology; sc-18351; 1:200), p53 (Vector Laboratories; VP-P956; 1:200), PD-1H/VISTA (BioLegend; 143704; 1:100), Chi3l3/Chi3l4 (Abcam; ab192029; 1:500), p16 (Abcam; ab211542; 1:400), p21 (Abcam; ab188224; 1:400), PD1 (Abcam; ab214421; 1:400), dsRNA/9D5 (Absolute Antibody; Ab00458-23.0; 1:200), CD3 (Abcam; ab16669; 1:100), RELMα (Abcam; ab39626; 1:50), and CD11b (Thermo Fisher Scientific; 14-0112-81; 1:50). FITC anti-goat, FITC anti-mouse, FITC anti-rabbit, TRITC anti-goat, TRITC anti-mouse, TRITC anti-rat, TRITC anti-armenian hamster, and Alexa 647 anti-rabbit fluorescent (Jackson ImmunoResearch) secondary antibodies were used, all at 1:300 dilutions, for the IF imaging. Controls using secondary antibodies alone were negative, and all IHC and IF results were reproducible in at least two repeated experiments.

Flow cytometry

Cell suspensions from KPC tumors were prepared using the Mouse Tumor Dissociation Kit (Miltenyi Biotec) according to the manufacturer's recommendations. Cells were Fc-blocked with purified rat anti-mouse CD16/CD32 (Clone: 2.4 G2, Becton Dickinson) for 30 minutes at 4°C. Dead cells were discriminated using the LIVE/DEAD (LD) Fixable Near-IR Dead Cell Stain Kit (Thermo Fisher Scientific) and samples were stained for the extracellular markers using directly labeled primary antibodies against Chi3l3/4 (Abcam; ab192029), CX3CR1 (BioLegend SA011F11 BV711), CD103 (BD Biosciences M290 FitC), CD45 (BioLegend 30-F11 BV605), CD11b (BioLegend M1/70 AF700), CD11c (BioLegend N418 BV650), and F4/80 (BioLegend BM8 PE/Cy7). Staining was visualized by FACS analysis using a BD FACSCelesta Multicolor Cell Analyzer (BD Biosciences) and analyzed using FlowJo v10 software (FlowJo LLC).

Histologic analysis and counting of tumor-infiltrating cells

We measured the areas of zonal necrosis in each tumor using low power (4×) images of two hematoxylin and eosin (H&E)-stained tissue sections of each tumor, with pixel counting using ImageJ. The two levels were chosen to be at least 6 sections or 30-μm apart. Pixel counts were converted into mm2 by capturing a 1 mm2 grid at ×4 magnification.

For quantitating proliferating (Ki67) and tumor-infiltrating cells identified by IHC with cell type–specific antibodies, tumor area and cell counts were calculated over two entire tissue sections, obtained from the top and bottom of each paraffin-embedded tissue block (i.e., separated by a thickness of approximately 3 mm). Tumor area was calculated using a collection of ×20 magnification images. Marker-positive cell counts were obtained from 10 random ×40 fields per histologic section using the Cell Counter function in NIH ImageJ (13) to tabulate manual counts. To minimize possible effects of an acute response to necrotic cell debris, cells immediately adjacent to necrotic areas (within 2 mm) were not included in the final counts. However, recounts including these areas did not significantly alter the results. To count lymphocytes and macrophages around normal pancreatic ducts and PanIN lesions, we selected a 500-μm perimeter around the ducts and premalignant lesions in each field, counted the marker-positive cells in these areas, and averaged the results over the number of counted fields. All cell counts were visually quantified in a case-blinded fashion by two individuals. The assessment between the two readers showed high concordance in all cases and therefore they were averaged to produce the reported results.

RNA-sequencing, gene-set enrichment, and deconvolution analyses

RNA was isolated using TRIzol (Invitrogen) and integrity was confirmed as RIN > 7 on a Bio Analyzer (Agilent Technologies). Poly-A pull-down was used to enrich for mRNAs, and libraries were prepared using the Illumina TruSeq RNA kit. Libraries were sequenced on an Illumina HiSeq 4000 machine with 100-bp paired end reads, in the Genomics Shared Resource of the HICCC. After base calling and adaptor trimming, the reads were mapped to the mouse reference genome (UCSC/mm9) using Tophat, and the relative expression of genes was estimated by FPKM (fragments per kilobase of transcript per million mapped reads) using Cufflinks (14). Candidate in vivo genes were defined using Student t test with unequal variance assumption and defined using nominal P < 0.05, fold change > 1.5 or <0.67 and FPKM > 1 in at least one group. Candidate genes induced in KPC tumor cells by DAC exposure in tissue culture after 48 or 96 hours, using a range of DAC concentrations, were identified using linear regressions (Gene expression ∼ DAC concentration) and defined using nominal P < 0.05, R2 > 0.5, FPKM > 1 in at least one DAC concentration, and the same dose effect trend for both time points (48 and 96 hours). All performed tests were two-sided. Multiple testing corrected P values (false discovery rate; FDR) using the Benjamini–Hochberg method are reported in Supplementary Tables S1 and S2.

To test for enrichment of genes associated with immune and stromal cell lineages in our RNA-seq data, we downloaded the most recent expression data generated by the Immunological Genome Project (15) from primary myeloid, lymphoid, and stromal cells from C57BL/6J mice using RNA-seq (NCBI GEO accession number GSE109125). Data were normalized using TMM method with edgeR R package. Using the ImmGen annotations, we categorized their samples according to cell type. In this dataset, we used expression in stromal cells as a baseline to identify genes that were relatively overexpressed in myeloid and lymphoid lineages (FDR < 0.05 and fold change > 2). Tests for enrichment of these myeloid and lymphoid signatures among the genes differentially expressed in our RNA-seq data from the KPC tumors were performed using logistic regression. Using statistical criteria (FDR < 0.05 and fold change > 2), we identified genes specifically overexpressed in monocyte, macrophage, and neutrophils, and tested for enrichment of each of these genes among the genes that we had found to be differentially expressed in our RNA-seq data from the KPC tumors.

Enrichment for biological pathways among our in vitro cultured KPC cell DAC-induced genes was performed using the Broad institute's GSEA tool and the hallmark gene sets (MSigDB, Broad Institute), which contained expressed signatures derived by aggregating many MSigDB gene sets to represent well-defined biological pathways, including IFN responses. For the deconvolution analysis, we used the current Immgen RNA-seq dataset (GEO accession GSE109125) and our tissue culture-based KPC tumor cell data to derive immune cell and cancer cell expression profiles, respectively. Because the Immgen RNA-seq data was aligned using STAR aligner with read counts, we realigned our data using the same pipeline. Read counts were normalized using the TMM method with edgeR R package (16) and data deconvolution was performed using the DeconRNASeq R package (17).

To assess the expression of ERV transcripts, we used two complementary approaches. First, we used the repeatmasker annotation of the mouse reference genome for ERV elements to calculate FPKM counts for known ERV elements, as described by Kazachenka, and colleagues (18). Second, to be more comprehensive and assess ERV elements that were not present in the genome of the reference mouse, we generated a faux genome using 15 mouse provirus sequences, kindly provided by Dr. R. Gifford (references and NCBI accession numbers in Supplementary Table S4). RNA-seq data were then aligned against this genome to calculate FPKM of ERV transcripts. For repetitive elements, we focused on the SINE/LINE elements using the repeatmasker annotation to calculate FPKM. Finally, to ask whether DAC-induced Irf7 expression might be ERV induction dependent, we tested the correlation of Irf7 expression with all significant ERVs using our in vitro cultured KPC tumor cell data and Spearman correlations (FDR < 0.05).

Additional statistical methods

Statistical approaches for RNA-seq data are described above. The P values for differences in cell counts by IHC are based on two-sided Student t tests under the assumption of unequal variance. The statistical analyses of mouse survival and comparisons of survival curves were based on the Mantel–Cox (log-rank) test, performed using GraphPad Prism 8 software. A value of P < 0.05 was considered as statistically significant for both methods of analysis.

Data repository URL and accession numbers

RNA-seq data have been deposited at NCBI GEO with accession GSE101336.

Low dose DAC inhibits tumor growth and reduces global DNA methylation in the KPC model of PDAC

Cancer formation in KPC mice progresses from preinvasive lesions, akin to human PanIN, to stroma-rich invasive adenocarcinomas, analogous to human PDACs. In this study, we tested treatment with single-agent DAC and single-agent ICIs, and sequential combinations of DAC followed by ICIs, using survival and timed endpoint paradigms (Fig. 1A; Supplementary Fig. S1). Paradigm I was a survival experiment in which we performed ultrasound imaging on the mice starting at 2 months of age and enrolled for treatment all mice with tumors that were 3–5 mm in diameter. Because our prior experience has suggested different survival times and differences in accuracy of tumor detection by ultrasound for animals with tumors located in the head versus tail anatomic regions of the pancreas, we restricted this experiment to tail of pancreas (TOP) tumors (most reliable early tumor detection) and utilized the head of pancreas (HOP) tumors for biological assays but not survival, in a timed endpoint protocol that we refer to as Paradigm II. In Paradigm III, which we also used for biological assays but not survival data, we started treatment of the mice earlier, when the pancreata contained PanIN lesions but no overt tumors.

Figure 1.

Effects of DAC on tumor growth, histology, and cell proliferation-related markers in KPC mice. A, Timing of treatment initiation and drug administration in treatment Paradigm I. B, There was a significant increase in tumor doubling time in single-agent DAC-treated mice compared with the mock-treated (PBS) control mice. C, H&E-stained tissue sections showing little or no necrosis in a tumor from a mock-treated control tumor, contrasting with large areas of zonal necrosis (asterisk) in a tumor from a DAC-treated mouse. D, Quantitative image analysis shows that DAC treatment significantly increases tumor necrosis. E, IHC revealed a decrease in Ki67+ cells following DAC treatment. The fields are chosen to be representative. F, Decreased numbers of Ki67+ cells per high power field (HPF) indicated a decrease in tumor cell proliferation with DAC treatment. G, IHC revealed an increase in p21+ cells following DAC treatment. The fields are chosen to be representative. H, The increased numbers of p21+ cells per HPF are consistent with the decrease in tumor cell proliferation induced by DAC treatment. Numbers of cases for histology and cell counting reflect the number of histologic blocks available. Original magnification, ×20. Scale bars, 100 μm.

Figure 1.

Effects of DAC on tumor growth, histology, and cell proliferation-related markers in KPC mice. A, Timing of treatment initiation and drug administration in treatment Paradigm I. B, There was a significant increase in tumor doubling time in single-agent DAC-treated mice compared with the mock-treated (PBS) control mice. C, H&E-stained tissue sections showing little or no necrosis in a tumor from a mock-treated control tumor, contrasting with large areas of zonal necrosis (asterisk) in a tumor from a DAC-treated mouse. D, Quantitative image analysis shows that DAC treatment significantly increases tumor necrosis. E, IHC revealed a decrease in Ki67+ cells following DAC treatment. The fields are chosen to be representative. F, Decreased numbers of Ki67+ cells per high power field (HPF) indicated a decrease in tumor cell proliferation with DAC treatment. G, IHC revealed an increase in p21+ cells following DAC treatment. The fields are chosen to be representative. H, The increased numbers of p21+ cells per HPF are consistent with the decrease in tumor cell proliferation induced by DAC treatment. Numbers of cases for histology and cell counting reflect the number of histologic blocks available. Original magnification, ×20. Scale bars, 100 μm.

Close modal

In treatment Paradigm I, we found a significant slowing of tumor growth (40 % increase in tumor doubling time) in the mice treated with single-agent DAC, compared with mock-treatment with PBS vehicle (Fig. 1B). Importantly, consistent with our prior findings in the KPC-Brca1 model (10), with the low DAC dosage utilized in the current study (1 μg/g body weight via intraperitoneal injection every 72 hours; total of 3 doses) we found no evidence of systemic or hematologic toxicity (as evidenced by lack of anemia or neutropenia) during the period of drug. Circulating white blood cell counts were only mildly reduced in the DAC-treated KPC mice compared with mock-treated animals (Supplementary Table S3) and no other side effects were noted. In addition, we observed no increase in animal mortality during the first 3 weeks of treatment (when tumor sizes were small) in the DAC treatment arm compared with the PBS control arm, further indicating no systemic toxicity with this dosage of the drug.

To ask whether this low dosage of DAC led to measurable effects on global DNA methylation, we carried out MS-Pyroseq of B1 repetitive elements in tumor DNA samples, 72 hours after the last dose of DAC (Paradigm II), or at the survival endpoint (Paradigm I). The results from Paradigm II indicated that a modest (6%) but statistically significant drug-induced DNA global hypomethylation had occurred in vivo and persisted for at least 3 days after the last DAC dose (Supplementary Fig. S2). As would be expected, direct exposure of explanted KPC tumor cells to DAC in tissue culture produced greater reductions in B1 element methylation (Supplementary Fig. S2).

Treatment with DAC leads to increased tumor necrosis, decreased cell proliferation, and increased numbers of CD8-positive tumor-infiltrating lymphocytes

In tumors from Paradigm I, the most striking finding in H&E-stained histologic sections was an effect on zonal necrosis: measurements by digital image analysis revealed a significant (>3-fold) increase in the areas of necrosis in the tumors from the DAC-treated compared with mock-treated mice (Fig. 1C and D). Furthermore, IHC for Ki67, a marker for cells that are actively engaged in the cell cycle, revealed 2.5-fold fewer cycling tumor epithelial cells, at the survival endpoint, with DAC compared with PBS (Fig. 1E and F). Opposite to this reduction in Ki67-expressing cells, the DAC treatment produced a significant increase in the numbers of cells expressing cyclin-dependent kinase inhibitors p21 (Fig. 1G and H), and p16 (Supplementary Fig. S3), which inhibit cell-cycle progression and are often reduced or silenced in the development of human pancreatic cancers (19).

On the basis of IHC followed by cell counting, we found a significant 5-fold increase in CD8+ tumor-infiltrating lymphocytes (TIL) with single-agent DAC treatment. We observed these cells both diffusely infiltrating the tumors and concentrated in lymphoid aggregates. Both the total number of CD8+ cells per tumor area and the number of CD8+ cells per lymphoid aggregate were increased by DAC (Fig. 2A and B; Supplementary Fig. S4A), and we found a similar increase in these cells in tissue sections of the HOP tumors from the timed endpoint Paradigm II (Supplementary Fig. S4C). Numbers of CD4+ TILS were also increased by DAC (Fig. 2C and D). We found a significantly higher number of PD-1 expressing cells in DAC-treated tumors (Fig. 2E and F), with most of the PD-1+ cells being CD8+ (Supplementary Fig. S4D). The number of FoxP3+ suppressor T cells was variable among the tumors but was not significantly altered by DAC, and the ratio of FoxP3+/CD8+ cells showed a nonsignificant decline (Supplementary Fig. S4B).

Figure 2.

DAC treatment increases tumor-infiltrating CD8+ and CD4+ lymphocytes and PD-1–expressing cells. A, IHC showing an increase in CD8+ TILs with DAC treatment compared with PBS vehicle. B, Cell counting revealed a significant increase in CD8+ TILs following DAC treatment. C, IHC images of CD4+ TILs with DAC compared with PBS. D, Cell counting revealed an increase in the number of CD4+ cells between DAC- and PBS-treated tumors. E, IHC showed that the number of PD-1–positive cells is increased by DAC. F, The increase in PD-1+ cells was confirmed by cell counting. Data are from treatment Paradigm I. The number of cases for cell counting are indicated in parentheses and reflect the number of histologic blocks available. Original magnification, ×20. Scale bars, 100 μm.

Figure 2.

DAC treatment increases tumor-infiltrating CD8+ and CD4+ lymphocytes and PD-1–expressing cells. A, IHC showing an increase in CD8+ TILs with DAC treatment compared with PBS vehicle. B, Cell counting revealed a significant increase in CD8+ TILs following DAC treatment. C, IHC images of CD4+ TILs with DAC compared with PBS. D, Cell counting revealed an increase in the number of CD4+ cells between DAC- and PBS-treated tumors. E, IHC showed that the number of PD-1–positive cells is increased by DAC. F, The increase in PD-1+ cells was confirmed by cell counting. Data are from treatment Paradigm I. The number of cases for cell counting are indicated in parentheses and reflect the number of histologic blocks available. Original magnification, ×20. Scale bars, 100 μm.

Close modal

In Paradigm III, we tested the effects of DAC starting the treatment at 4 months of age, which is before the formation of tumors in most of the mice, but late enough to have allowed the formation of preinvasive PanIN lesions. At three days after the last DAC dose, IHC of tumor sections revealed CD8+ lymphocytes within and adjacent to the PanIN lesions, with very few such cells around normal-appearing ducts in the same sections. However, despite this tropism of the CD8+ lymphocytes to the PanIN lesions, in this treatment paradigm we found no significant increase in the number of TILs with DAC compared with the PBS (Supplementary Fig. S5).

Gene expression profiling implicates effects of DAC on the cellular composition of the tumor microenvironment

To further understand the effects of DAC treatment on the tumor microenvironment, we profiled gene expression using RNA-seq in whole tumors from treatment Paradigm II, following three cycles of DAC treatment or PBS vehicle. The RNA-seq data revealed only small sets of treatment-responsive genes, with 195 genes upregulated and 20 downregulated by the in vivo DAC treatment (top-ranked over-expressed genes in Table 1; complete list in Supplementary Table S1). Among the strongly upregulated mRNAs were transcripts from genes encoding a group of myeloid cell markers, including Chi3l3 (also known as Ym1, Fig. 3A) and Chi3l4 (also known as Ym2), Gpnmb, Ear6, Ear11, and Ms4a8a, several of which (Chi3l3, Chi3l4, Gpnmb, Ms4a8a) are associated with alternatively polarized “M2” macrophages (20–23).

Table 1.

Gene expression profiling of DAC-treated tumors.

Gene namePBSDACFold increaseP
Chi3l3 44.1 754.8 17.1 0.032 
Ear6 0.5 4.8 9.0 0.029 
Chi3l4 1.3 9.6 7.6 0.041 
Ear11 1.3 8.2 6.3 0.018 
Zic3 0.3 1.6 5.4 0.006 
Spock3 1.1 5.2 0.037 
2810047C21Rik1 0.5 2.5 4.6 0.011 
Wnt4 5.3 23.9 4.5 0.004 
Prss30 0.5 1.8 3.8 0.026 
Slc10a2 1.1 3.9 3.7 0.04 
Ms4a8a 17.9 58.3 3.3 0.03 
Irx3 1.7 5.3 3.1 0.039 
Apol9a 3.5 10.5 3.0 0.032 
Fam180a 3.7 11 3.0 0.035 
Apol9b 5.3 15.1 2.9 0.023 
Ephx4 1.4 4.2 2.9 0.009 
Hoxc6 0.6 1.8 2.9 0.031 
Atp6v0d2 5.2 14.3 2.7 0.003 
Muc6 21.6 2.7 0.039 
Gpnmb 8.7 22.7 2.6 0.047 
Pamr1 10.3 26.8 2.6 0.038 
Sfi1 13 33.5 2.6 0.03 
Tnip3 2.5 6.5 2.6 0.013 
Wdr72 6.3 16.6 2.6 0.011 
Gene namePBSDACFold increaseP
Chi3l3 44.1 754.8 17.1 0.032 
Ear6 0.5 4.8 9.0 0.029 
Chi3l4 1.3 9.6 7.6 0.041 
Ear11 1.3 8.2 6.3 0.018 
Zic3 0.3 1.6 5.4 0.006 
Spock3 1.1 5.2 0.037 
2810047C21Rik1 0.5 2.5 4.6 0.011 
Wnt4 5.3 23.9 4.5 0.004 
Prss30 0.5 1.8 3.8 0.026 
Slc10a2 1.1 3.9 3.7 0.04 
Ms4a8a 17.9 58.3 3.3 0.03 
Irx3 1.7 5.3 3.1 0.039 
Apol9a 3.5 10.5 3.0 0.032 
Fam180a 3.7 11 3.0 0.035 
Apol9b 5.3 15.1 2.9 0.023 
Ephx4 1.4 4.2 2.9 0.009 
Hoxc6 0.6 1.8 2.9 0.031 
Atp6v0d2 5.2 14.3 2.7 0.003 
Muc6 21.6 2.7 0.039 
Gpnmb 8.7 22.7 2.6 0.047 
Pamr1 10.3 26.8 2.6 0.038 
Sfi1 13 33.5 2.6 0.03 
Tnip3 2.5 6.5 2.6 0.013 
Wdr72 6.3 16.6 2.6 0.011 

Note: Genes with >2.5-fold increased expression in tumors from the DAC-treated mice. Expression values are in FPKM. Data are from treatment Paradigm II.

Figure 3.

Gene expression profiling and data deconvolution confirms an abundant tumor stromal component and implicates an increase in a specific myeloid cell subpopulation in the DAC-treated tumors. A, Expression of Chi3l3 mRNA was significantly increased by DAC treatment in the KPC tumors in this study. B,Chi3l3 mRNA was induced in CD11b+ splenic macrophages from mice hosting C26-GM colon cancer cells engineered to secrete GM-CSF (NCBI GSE5455), compared with CD11b+ cells isolated from tumor-free mice. C, Results of deconvolution analysis for proportions of cell populations performed using the ImmGen RNA-seq reference dataset (see Materials and Methods). This analysis suggests that the tumors in vivo contain approximately equal (20% each) populations of malignant KPC cells, stromal cells, macrophage-lineage cells, and other cell types.

Figure 3.

Gene expression profiling and data deconvolution confirms an abundant tumor stromal component and implicates an increase in a specific myeloid cell subpopulation in the DAC-treated tumors. A, Expression of Chi3l3 mRNA was significantly increased by DAC treatment in the KPC tumors in this study. B,Chi3l3 mRNA was induced in CD11b+ splenic macrophages from mice hosting C26-GM colon cancer cells engineered to secrete GM-CSF (NCBI GSE5455), compared with CD11b+ cells isolated from tumor-free mice. C, Results of deconvolution analysis for proportions of cell populations performed using the ImmGen RNA-seq reference dataset (see Materials and Methods). This analysis suggests that the tumors in vivo contain approximately equal (20% each) populations of malignant KPC cells, stromal cells, macrophage-lineage cells, and other cell types.

Close modal

These findings led us to consider whether many of the gene expression changes induced by DAC in the KPC tumors might be due to changes in tumor-infiltrating myeloid lineage cells. To test this hypothesis, we adopted a targeted gene-set enrichment approach, using an RNA-seq dataset (NCBI GEO accession GSE109125) from the ImmGen project (10, 24) as a reference for expression levels in specific types of immune system cells and mesenchymal stromal cells. This analysis revealed an enrichment in myeloid lineage signature genes, with OR = 1.9 and P = 8.6 × 10−5, with a 24% absolute overlap with the upregulated genes in our dataset (including Chi3l3, Gpnmb, and Ear11). This enrichment was greater for signature genes of the subthreshold for monocytes (OR = 1.9, P = 0.06). The strongest expression of Chi3l3 in the ImmGen RNA-seq dataset is found in macrophages and granulocytes, and in several other publicly available expression datasets Chi3l3 mRNA is found to be inducible by tumor-derived cytokines or other stimuli to become highly expressed in CD11b+ macrophages (example in Fig. 3B). For completeness, it should be noted that Chi3l3 expression can be found either in CD11b+ macrophages or, in some studies alternatively in CD11b/Cd11c+ myeloid dendritic cells, depending on tissue context, consistent with the known cytokine-dependent developmental fluidity of myeloid cells (25–28).

Using our RNA-seq data, with the ImmGen RNA-seq dataset as a reference, we performed a deconvolution analysis (Materials and Methods) to assess which cell types are predominantly represented in the expression profiling of the autochthonous KPC tumors. The results suggest that malignant KPC tumor cells represented 22%, cancer associated fibroblast-like stromal cells 30%, and macrophage/myeloid lineage cells 20% (Fig. 3C). This gross distribution of cell types did not change in response to DAC, suggesting that the significant shift in myeloid markers implied by our RNA-seq data might reflect a DAC induced increase in a specific subpopulation of myeloid lineage cells, a presumption that is confirmed by immunostaining and flow cytometry data in the next section.

Overall, these results imply that many of the expression changes in the whole KPC tumors might be due to DAC-induced changes in a subpopulation of tumor-infiltrating cells, thus reflecting non–cell autonomous effects of the drug treatment. Nonetheless, we also sought to ask whether we could detect evidence for induction of classical DAC target genes in the treated tumors. We first performed a gene-set enrichment analysis among upregulated genes that we observed using RNA-seq applied to KPC tumor cells in tissue culture exposed to a range of DAC concentrations and harvested at two time points (Supplementary Table S2). We found a significant induction of multiple immune pathways, and like our previously published observations in DAC-treated KPC-Brca1 mouse tumors (10), a significant upregulation of IFN-responsive genes (Supplementary Fig. S6A). Induction of endogenous retroviruses (ERV) and double-stranded RNA (dsRNA) has been shown to contribute to these pathways, and in the RNA-seq data from in vitro KPC cell DAC exposure we found ERV, SINE, and LINE inductions (Supplementary Tables S4–S8). While the strongest findings came from the experiments using DAC exposure in vitro, one ERV transcript class was also significantly upregulated in the autochthonous KPC tumors by DAC treatment in vivo (treatment Paradigm II; Supplementary Fig. S6B and S6C). Interestingly, among 59 ERV transcript classes induced by DAC treatment, 40 showed significant correlations with expression of the IFN-responsive Irf7 gene (Spearman correlation FDR < 0.05; Supplementary Table S9; Supplementary Fig. S6D and S6E). In addition, IF with a mAb against dsRNA revealed an increase in dsRNA in DAC-treated tumors in vivo (Supplementary Fig. S6F).

DAC leads to increased tumor-infiltrating Chi3l3/4+ macrophages

Two of the top-ranked overexpressed transcripts from expression profiling of the in vivo DAC-treated tumors, Chi3l3 and Chi3l4, are from chitinase-like genes that are activated during myeloid-lineage cell differentiation in inflammatory conditions (20, 29–32). These genes are known to be markers for “alternatively activated” or “M2-polarized” macrophages (20, 29), although in some situations, Chi3l3 is expressed by antigen-presenting DCs (28). To ask whether the increased expression of these genes in the DAC-treated tumors reflected an increased number of tumor-infiltrating Chi3l3/4+ cells, we performed IHC on DAC-treated compared with PBS control tumors using a mAb that recognizes both the Chi3l3 and Chi3l4 proteins. The results showed a highly significant 4-fold increase in the number of Chi3l3/4+ cells in the tumors from the DAC-treated mice in both the timed endpoint and survival paradigms (Fig. 4AC), indicating a non–cell autonomous effect of DAC through recruitment and/or local expansion of this cell population. A combination of this strongly increased number of Chi3l3/4+ cells, presumably together with an increase in Chi3l3/4 mRNA expression per cell, likely accounts for the approximately 17-fold net overexpression of Chi3l3 mRNA that we observed by RNA-seq.

Figure 4.

Increased Chi3l3/4+ tumor-associated macrophages in KPC tumors following DAC and DAC+ICI treatment. A, IHC revealed a marked increase in tumor-infiltrating Chi3l3/4+ cells after DAC treatment. Fields are chosen to be representative. B, Counting of Chi3l3/4+ cells per tumor area revealed a 3-fold increase in the numbers of these cells following DAC treatment compared with PBS vehicle control in treatment Paradigm I. In the combination treatment arms using DAC followed by ICIs, there was a further increase in tumor-infiltrating Chi3l3/4+ cells. C, Counting of Chi3l3/4+ cells per tumor area revealed an a three-fold increase in the number of these cells in DCA-treated tumors compared with PBD in treatment Paradigm II. D, Chi3l3/4 and Arginase-1 show substantial overlap, but there are also many Arginase-1+; Chi3l3/4 cells. E, Double IF for Chi3l3/4 and RELMα shows substantial overlap of these two markers. F, Double IF revealed partial overlap of Chi3l3/4 with CD11b. Original magnification, ×20. Scale bars, 100 μm.

Figure 4.

Increased Chi3l3/4+ tumor-associated macrophages in KPC tumors following DAC and DAC+ICI treatment. A, IHC revealed a marked increase in tumor-infiltrating Chi3l3/4+ cells after DAC treatment. Fields are chosen to be representative. B, Counting of Chi3l3/4+ cells per tumor area revealed a 3-fold increase in the numbers of these cells following DAC treatment compared with PBS vehicle control in treatment Paradigm I. In the combination treatment arms using DAC followed by ICIs, there was a further increase in tumor-infiltrating Chi3l3/4+ cells. C, Counting of Chi3l3/4+ cells per tumor area revealed an a three-fold increase in the number of these cells in DCA-treated tumors compared with PBD in treatment Paradigm II. D, Chi3l3/4 and Arginase-1 show substantial overlap, but there are also many Arginase-1+; Chi3l3/4 cells. E, Double IF for Chi3l3/4 and RELMα shows substantial overlap of these two markers. F, Double IF revealed partial overlap of Chi3l3/4 with CD11b. Original magnification, ×20. Scale bars, 100 μm.

Close modal

We next used a panel of markers in IHC/IF and flow cytometry to assign the tumor-infiltrating Chi3l3/4+ cells to a specific cell type in the myeloid lineage. As expected, by two-color IF we found a lack of coimmunostaining with markers for malignant epithelial cells (mutant p53), stromal myofibroblasts (αSMA), and lymphocytes (CD3, CD103; Supplementary Fig. S7A–S7D). Dual IF with known macrophage markers showed that the majority of Chi3l3/4+ cells are also positive for the M2 macrophage markers Arginase-1 (Arg-1) and RELM- α, and a subset of these cells are positive for CD163, F4/80, and CD11b, with only a small percentage of the Chi3l3/4+ cells being positive for the dendritic cell marker CD103 (Fig. 4DF; Supplementary Fig. S7C). The subset of Chi3l3/4+ cells that were positive for Cd11b had a variable and relatively low intensity of staining for this antigen; whereas those cells strongly positive for CD11b were mostly Chi3l3/4. Although nearly all Chi3l3/4+ cells were positive for Arg-1, we also observed many Arg-1+;Chi3l3/4 and CD11b+;Chi3l3/4 cells, as well as some CD163+;Chi3l3/4 cells (Fig. 4D; Supplementary Fig. S7D). These results indicate the Chi3l3/4+ cells are a specific subset of M2 macrophages and are consistent with our finding that there was no difference in the mean number of Arg-1+ cells in the tumors from the PBS versus DAC treatment arms (Fig. 5A). Finally, relevant to a potential role in immune checkpoints, by two-color IF, we also found that most Chi3l3/4+ cells also express PD-L1 (Fig. 5B).

Figure 5.

The tumor-infiltrating Chi3l3+ cells are a subset of Arginase-1 and DAC affects TILs. A, The total number of Arg1+ cells is not increased by DAC treatment. B, Double IF revealed partial overlap of Chi3l3/4 with PD-L1, although the two antigens have different subcellular distributions (punctate pattern for PD-L1). Data are from treatment Paradigm I. C, Sequential combination therapy with DAC followed by anti-PD-1 leads to an increase in CD8+ TILs, compared with treatment with DAC alone. The number of cases for cell counting reflects the number of histologic blocks available. N.S., nonsignificant.

Figure 5.

The tumor-infiltrating Chi3l3+ cells are a subset of Arginase-1 and DAC affects TILs. A, The total number of Arg1+ cells is not increased by DAC treatment. B, Double IF revealed partial overlap of Chi3l3/4 with PD-L1, although the two antigens have different subcellular distributions (punctate pattern for PD-L1). Data are from treatment Paradigm I. C, Sequential combination therapy with DAC followed by anti-PD-1 leads to an increase in CD8+ TILs, compared with treatment with DAC alone. The number of cases for cell counting reflects the number of histologic blocks available. N.S., nonsignificant.

Close modal

We supplemented these IHC/IF results with flow cytometry, which showed that most Chi3l3/4+ cells are CD11c; CD103; CX3CR1+, indicative of macrophages, rather than dendritic cells. The majority of Chi3l3/4+ cells were positive for CD11b by flow cytometry; however, the level of CD11b expression in most of these cells was low. Also, by flow cytometry, among CD11b+Chi3l3/4+ cells, approximately 40% were F4/80+ (Supplementary Fig. S8A–S8C). Although a subset of Chi3l3/4-positive cells did not express high levels of the classical macrophage marker CD11b, these also did not express the dendritic cell marker CD103 (Supplementary Fig. S7C). Taken together, these results suggest that the majority of tumor-infiltrating Chi3l3/4+ cells are a unique subset of Arginase-1 and PD-L1–expressing M2-polarized macrophages that are recruited or polarized by paracrine signals.

When we examined tissue sections from treatment Paradigm III pancreata to assess PanIN lesions, we observed that Chi3l3/4+ macrophages are specifically present around these premalignant lesions, with very few such cells seen in the normal pancreatic parenchyma and around nonneoplastic ducts. Although these cells are clearly homing to the PanIN lesions, they were seen around PanIN both in the DAC-treated and in the PBS control mice, and we did not find a significant DAC-dependent increase in their numbers (Supplementary Fig. S9A and S9B). Similarly, we did not find a significant increase in Arg-1 or CD163-positive cells in DAC-treated mice with PanIN lesions (Supplementary Fig. S9C–S9E).

DAC effects can be studied in allograft tumors

While the autochthonous KPC tumors are the closest mimic of human PDACs, it is desirable to have alternative experimental systems that can give faster results without requiring serial ultrasound monitoring of large numbers of mice. To characterize DAC effects in an immunocompetent allograft system, we created syngeneic subcutaneous allografts using a KPC tumor cell explant line developed from a tumor occurring in our mouse breeding colony, with subcutaneous implantation of these cells into littermate host mice. Using an identical dose regimen to the in vivo experiments, we found a significant effect of DAC in reducing tumor size at a defined endpoint (similar to in vivo treatment Paradigm II), and we observed increased numbers of Chi3l3/4+ cells in the DAC treated allograft tumors (Supplementary Fig. S10A and S10B).

Combination therapy with DAC plus ICIs leads to further increases in tumor-infiltrating CD8+ T cells and Chi3l3/4+ macrophages

Given this encouraging data from single-agent DAC, we next evaluated sequential DAC and immune checkpoint blockade, starting with an assessment of effects on tumor-infiltrating lymphoid and myeloid cells. We chose to evaluate two ICIs, anti-PD-1, and anti-VISTA. VISTA is an immune checkpoint family member (33–35) that has been less studied as a therapeutic target than the related molecule PD-1. However, it is expressed in more broadly in lymphoid cells and has recently been shown to be one of the more abundant immune checkpoint mediators in PDAC (36, 37). We observed a significant increase in VISTA+ TILs after DAC treatment (Supplementary Fig. S11A). By analyzing tumors from treatment Paradigm I, we found that both of the combination therapies (DAC followed by anti-PD-1 and DAC followed by anti-VISTA) led to a further increase in CD8+ TILs (Fig. 5C). Although this observation supports a potential benefit, we also noted that the putatively immunosuppressive Chi3l3/4+ myeloid cells were increased by the DAC plus ICI combinations (Fig. 4B; Supplementary Fig. S11B).

Combination therapy with DAC plus ICIs decreases tumor growth and increases overall survival

We next evaluated the effects of sequential DAC plus immune checkpoint blockade on tumor growth and mouse survival, using treatment Paradigm I. We observed a significant slowing of tumor growth (prolongation of doubling time) with single-agent DAC, anti-PD1, and anti-VISTA compared with PBS controls, with further tumor growth inhibition in the DAC + anti-VISTA and DAC + anti-PD-1 treatment arms (P < 0.01 compared with PBS controls; Fig. 6A and B). As shown in Fig. 6C and D, mouse survival was minimally prolonged by the single-agent ICIs, and DAC alone modestly prolonged median survival (32 vs. 26 days; P < 0.05). Both DAC + anti-VISTA and DAC + anti-PD-1 prolonged survival significantly compared with the PBS control arm (41 vs. 26 days; P < 0.0001 and 54 vs. 26 days; P < 0.0001, respectively) and there was a significant increase in survival in DAC + anti-PD-1 compared with DAC alone (54 vs. 32 days; P < 0.01) or anti-PD-1 alone (54 vs. 30 days; P < 0.01). Although the best combination of agents (DAC followed by anti-PD-1) doubled survival in the mice, the treatment was not curative, and few animals survived beyond 80 days (Fig. 6D).

Figure 6.

Effect on tumor growth and mouse survival of single-agent and combination therapy with DAC and immune checkpoint inhibitors. A, There is a significant increase in tumor doubling time (slowing of tumor growth) with single-agent DAC, single-agent anti-VISTA, and sequential DAC followed by anti-VISTA, compared with PBS vehicle controls. The greatest inhibition of tumor growth is seen with the sequential combination therapy. Error bars, SD. B, Tumor doubling time is increased with single-agent DAC and sequential DAC followed by anti-PD-1 treatment, compared with PBS vehicle controls. C, Kaplan–Meier analyses show that combination therapy with DAC followed by anti-VISTA produces a modest but significant increase in mean survival compared with PBS (log rank; ***, P < 0.0001) and that single-agent DAC confers a weaker benefit (log rank; *, P < 0.05). D, Kaplan–Meier analyses show modest benefits with single agent DAC and single-agent anti-PD-1 (both log rank; *, P < 0.05) and a major increase in mean survival with the DAC followed by anti-PD-1 combination treatment (log rank; ***, P < 0.0001) compared with PBS vehicle controls. The combination therapy was also significantly more effective than anti-PD-1 alone or DAC alone (log rank, P < 0.01; log rank, P < 0.05). Data are from treatment Paradigm I. #, The PBS vehicle control arm includes 10 concurrent controls plus three historical controls from an identical tumor detection and mock treatment initiation procedure. The differences in mean survival between treatment arms remained statistically significant when only the concurrent controls were considered.

Figure 6.

Effect on tumor growth and mouse survival of single-agent and combination therapy with DAC and immune checkpoint inhibitors. A, There is a significant increase in tumor doubling time (slowing of tumor growth) with single-agent DAC, single-agent anti-VISTA, and sequential DAC followed by anti-VISTA, compared with PBS vehicle controls. The greatest inhibition of tumor growth is seen with the sequential combination therapy. Error bars, SD. B, Tumor doubling time is increased with single-agent DAC and sequential DAC followed by anti-PD-1 treatment, compared with PBS vehicle controls. C, Kaplan–Meier analyses show that combination therapy with DAC followed by anti-VISTA produces a modest but significant increase in mean survival compared with PBS (log rank; ***, P < 0.0001) and that single-agent DAC confers a weaker benefit (log rank; *, P < 0.05). D, Kaplan–Meier analyses show modest benefits with single agent DAC and single-agent anti-PD-1 (both log rank; *, P < 0.05) and a major increase in mean survival with the DAC followed by anti-PD-1 combination treatment (log rank; ***, P < 0.0001) compared with PBS vehicle controls. The combination therapy was also significantly more effective than anti-PD-1 alone or DAC alone (log rank, P < 0.01; log rank, P < 0.05). Data are from treatment Paradigm I. #, The PBS vehicle control arm includes 10 concurrent controls plus three historical controls from an identical tumor detection and mock treatment initiation procedure. The differences in mean survival between treatment arms remained statistically significant when only the concurrent controls were considered.

Close modal

Despite the efficacy of ICIs against some cancers, (38, 39), such benefits have not yet been seen in human PDACs. Explanations could include high expression of immune checkpoint receptors, and a paucity of infiltrating cytotoxic T cells or antigen-presenting cells (40), which may be due to low antigenicity of the cancer cells or immune-suppressive properties of the tumor stroma (41–43). Strategies to overcome these roadblocks in preclinical models have included combinations of immune modulator agents and cytokines (44, 45), and combining immune modulators with epigenetically acting drugs (46–49). We previously found that low-dose DAC inhibits tumor growth and prolongs survival in the rapid-onset KPC-Brca1 mouse model of PDAC (10). Here we have built on these findings to study the effects of DAC, and DAC plus immune-modulating therapy, against PDACs in the more commonly used slower-onset KPC model, which is genetically similar to most human PDACs and which facilitates treatment protocols starting at an ultrasound-defined tumor size, which is more analogous to the human clinical situation.

Our most significant findings using single-agent DAC in this model were increased tumor necrosis, slowing of tumor growth, increased numbers of CD4+, CD8+, PD1+, and VISTA+ TILs, and an influx and/or intratumoral expansion of Chi3l3/4+ tumor-associated macrophages. While hypomethylating drugs are sometimes criticized for theoretically producing widespread changes in gene expression, in fact our RNA-seq data from the in vivo system revealed only small sets of differentially expressed genes in the whole KPC tumors, with the DAC-induced expansion and/or recruitment of the Chi3l3/4+ myeloid cell subpopulation reflected both in IHC findings and in the gene expression profiling, in which Chi3l3 was the most highly induced gene.

Our evidence based on IF and flow cytometry suggests that the tumor-infiltrating Chi3l3/4+ cells are a subtype of alternatively polarized (M2) macrophages. Prior studies have defined a population of M2-polarized macrophages that express Arginase-1, Chi3l3/4, and the resistin-like molecule (RELM)-alpha/FIZZ1 (50, 51)–a marker profile that matches the cell population that expanded in response to DAC in our experiments. M2-polarized macrophages have been associated with immunosuppressive functions and unfavorable cancer outcomes (51–53). Among possible mediators, Arginase-1, which we have shown here is highly expressed by Chi3l3/4+ macrophages in the KPC tumors, has been proposed as a key regulator via its effects on reactive oxygen species (32, 53, 54). Recent data further suggest that myeloid cells in the pancreatic tumor stroma directly inhibit CD8 cell function (55).

Paracrine effects likely contribute to the recruitment or polarization of this myeloid cell population. Regarding possible direct effects of DAC on myeloid cells, the complex tumor microenvironment, as known from prior studies and highlighted here by our RNA-seq–based deconvolution analysis, makes identification of methylation changes in subpopulations of cells difficult. Indeed, when we performed bisulfite sequencing on whole tumor DNAs we were unable to document detectable changes in Chi3l3 promoter methylation, despite the clear induction of this gene visualized at the level of protein products in subpopulations of cells. Nonetheless, we were able to identify direct effects of DAC on induction of ERVs and expression of dsRNA in tumor cells, which may have downstream consequences for immunogenicity of these cells.

Several different ICI molecules are expressed on the surface of T lymphocytes. We found that not only did the total number of effector T cells (CD8+ and CD4+ TILs) increase with DAC treatment, but that T cells expressing two checkpoint molecules, PD1 and VISTA, were also increased. VISTA/PD1-H is one of the more recently recognized family members, which has similarities with both PD-1 and PD-L1 (34, 56). Antibodies against VISTA have not been used in humans, but genetic deletion of this molecule produced T-cell activation and autoimmunity in mice (33, 57), and in one study a greater accumulation of leukocytes was seen in the pancreas after inhibition of VISTA compared with PD-1 (35). Our results showing recruitment of these T cells and increased expression of both PD1 and VISTA provided a rationale for targeting both of these molecules in combination with DAC.

On the basis of these considerations, we tested the in vivo efficacy of sequential combination therapy using DAC followed by two different ICIs, anti-VISTA, or anti-PD-1. We observed inhibition of tumor growth and a major extension of mean survival with DAC followed by anti-PD-1. Compared with DAC + anti-PD-1, there was slightly more inhibition of tumor growth with DAC followed by anti-VISTA, but this difference in growth inhibition was not a statistically significant, and we found less extension of survival with the DAC + anti-VISTA combination. Given the complicating considerations of tumor necrosis affecting the ultrasound interpretation of net tumor growth, and the role of locally invasive disease followed by sepsis, and potentially metastatic disease (11, 58), as causes of death, a positive but somewhat variable correlation of tumor growth inhibition with extension of survival is not unexpected.

Overall, these findings have clear and promising implications for the design of future studies to improve the effectiveness of epigenetic modulation, with DAC or other DNA methyltransferase inhibitors, or with drugs targeting other epigenetically acting enzymes, in combination with immune checkpoint inhibition. Despite the clear beneficial effects of adding DAC to ICI therapy, our observations using both single-agent DAC and its combination with the ICIs raise the possibility that the influx and/or local expansion of M2-polarized tumor-associated macrophages might be limiting the efficacy of this approach. Tumor-promoting functions of M2-polarized TAMs are well documented both generally and in PDACs (53, 59). Even without drug treatment, cancer-associated myofibroblasts, which are abundant in KPC tumors, are known to produce cytokines that affect macrophages (60, 61), and tumor epithelial cells in this model produce GM-CSF, which affects myeloid cell populations (62). In this study, recruitment of the Chi3l3/Ym1-positive sub-class of M2 macrophages further increased when DAC was combined with ICIs. It is reasonable to suggest that the initial effect of DAC, perhaps in part driven by ERV induction and dsRNA expression, is beneficial and leads to accumulation of effector T cells. However, subsequently a more suppressive myeloid population accumulates in the tumor and this may restrict efficacy of the immune response. This observation suggests that targeting of myeloid cells might be a useful strategy to further enhance efficacy of epigenetic therapy. Encouragingly, approaches for such targeting in preclinical models are now becoming available (5, 63). Finally, given the low toxicity of DAC that we have observed both in the KPC model and previously in KPC-Brca1 mice, follow-up studies should also test more frequent DAC and anti-PD-1 dosing and longer treatment, as well as combining hypomethylating drugs with dual immune checkpoint inhibition. Encouragingly in this regard, the current dosing schedule for DAC was without systemic toxicities, and in two tumor-bearing KPC mice we have also tested a more aggressive 5-dose regimen of anti-PD1 given every 48 hours and found no evidence of adverse systemic effects during the period of ICI administration.

J. Fang reports grants from NIH during the conduct of the study. C. Do reports grants from Merck Medical Innovation Science Program and grants from NIH during the conduct of the study. C.G. Drake reports personal fees from AZ Medimmune (consulting), Bayer (consulting), BMS (consulting), Compugen (consulting), Ferring (consulting), F-Star (consulting), Genocea (consulting), Jannsen (consulting), Kleo (consulting), Merck (consulting), Merck-Serono (consulting), Pfizer (consulting), Pierre Fabre (consulting), Roche/Genentech (consulting), Shattuck Labs (consulting), Tizona (consulting), Urogen (consulting), and Werewolf (consulting) during the conduct of the study; other compensation from Compugen (stock options), Harpoon (stock options), Kleo (stock options), Tizona (stock options), Urogen (stock options), and Werewolf (stock options) outside the submitted work; in addition, C.G. Drake has a patent for BMS licensed to Johns Hopkins University. G.A. Manji reports other compensation from Roche/Genentech (investigator-initiated clinical trial support), Merck (investigator-initiated clinical trial support), BioLineRx (investigator-initiated clinical trial support), and Regeneron Pharm (investigator-initiated clinical trial support) outside the submitted work. B. Tycko reports grants from Merck MISP Program Grant (Merck did not have any input into the data analysis or any other aspect of the manuscript) during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.

T.A. Gonda: Conceptualization, data curation, supervision, validation, investigation. J. Fang: Data curation, software, investigation, visualization, methodology, writing-original draft. M. Salas: Data curation, investigation. C. Do: Formal analysis, investigation, visualization, methodology, writing-original draft. E. Hsu: Data curation. A. Zhukovskaya: Data curation, visualization. A. Siegel: Data curation, investigation. R. Takahashi: Data curation, formal analysis. Z.A. Lopez-Bujanda: Formal analysis, validation. C.G. Drake: Supervision, methodology, writing-original draft, writing-review and editing. G.A. Manji: Formal analysis, funding acquisition, validation. T.C. Wang: Supervision, writing-original draft. K.P. Olive: Resources, writing-original draft. B. Tycko: Conceptualization, resources, data curation, software, supervision, funding acquisition, visualization, methodology, writing-original draft, writing-review and editing.

The authors thank Stephen Goff, PhD (Columbia University) and Robert Gifford, PhD (University of Glasgow) for their advice and assistance with retroviral analyses. This work was supported by grants from the NIH U54 CA126513 (to T.C. Wang and B. Tycko), K08DK100544 (to T.A. Gonda), from the Merck Medical Innovation Science Program (to B. Tycko and T.A. Gonda), AACR-Landon Innovator Award for Cancer Prevention Research (to T.A. Gonda), and the Leokadia and Robert Lynn Research Fund (to T.A. Gonda).

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

1.
Royal
RE
,
Levy
C
,
Turner
K
,
Mathur
A
,
Hughes
M
,
Kammula
US
, et al
Phase 2 trial of single agent Ipilimumab (anti-CTLA-4) for locally advanced or metastatic pancreatic adenocarcinoma
.
J Immunother
2010
;
33
:
828
33
.
2.
Bailey
P
,
Chang
DK
,
Forget
MA
,
Lucas
FA
,
Alvarez
HA
,
Haymaker
C
, et al
Exploiting the neoantigen landscape for immunotherapy of pancreatic ductal adenocarcinoma
.
Sci Rep
2016
;
6
:
35848
.
3.
Brahmer
JR
,
Tykodi
SS
,
Chow
LQ
,
Hwu
WJ
,
Topalian
SL
,
Hwu
P
, et al
Safety and activity of anti-PD-L1 antibody in patients with advanced cancer
.
N Engl J Med
2012
;
366
:
2455
65
.
4.
Tan
E
,
El-Rayes
B
. 
Pancreatic cancer and immunotherapy: Resistance mechanisms and proposed solutions
.
J Gastrointest Cancer
2019
;
50
:
1
8
.
5.
Panni
RZ
,
Herndon
JM
,
Zuo
C
,
Hegde
S
,
Hogg
GD
,
Knolhoff
BL
, et al
Agonism of CD11b reprograms innate immunity to sensitize pancreatic cancer to immunotherapies
.
Sci Transl Med
2019
;
11
:
eaau9240
.
6.
Stromnes
IM
,
Brockenbrough
JS
,
Izeradjene
K
,
Carlson
MA
,
Cuevas
C
,
Simmons
RM
, et al
Targeted depletion of an MDSC subset unmasks pancreatic ductal adenocarcinoma to adaptive immunity
.
Gut
2014
;
63
:
1769
81
.
7.
Zhu
Y
,
Knolhoff
BL
,
Meyer
MA
,
Nywening
TM
,
West
BL
,
Luo
J
, et al
CSF1/CSF1R blockade reprograms tumor-infiltrating macrophages and improves response to T-cell checkpoint immunotherapy in pancreatic cancer models
.
Cancer Res
2014
;
74
:
5057
69
.
8.
Jiang
H
,
Hegde
S
,
Knolhoff
BL
,
Zhu
Y
,
Herndon
JM
,
Meyer
MA
, et al
Targeting focal adhesion kinase renders pancreatic cancers responsive to checkpoint immunotherapy
.
Nat Med
2016
;
22
:
851
60
.
9.
Lomberk
G
,
Blum
Y
,
Nicolle
R
,
Nair
A
,
Gaonkar
KS
,
Marisa
L
, et al
Distinct epigenetic landscapes underlie the pathobiology of pancreatic cancer subtypes
.
Nat Commun
2018
;
9
:
1978
.
10.
Shakya
R
,
Gonda
T
,
Quante
M
,
Salas
M
,
Kim
S
,
Brooks
J
, et al
Hypomethylating therapy in an aggressive stroma-rich model of pancreatic carcinoma
.
Cancer Res
2013
;
73
:
885
96
.
11.
Hingorani
SR
,
Wang
L
,
Multani
AS
,
Combs
C
,
Deramaudt
TB
,
Hruban
RH
, et al
Trp53R172H and KrasG12D cooperate to promote chromosomal instability and widely metastatic pancreatic ductal adenocarcinoma in mice
.
Cancer Cell
2005
;
7
:
469
83
.
12.
Olive
KP
,
Jacobetz
MA
,
Davidson
CJ
,
Gopinathan
A
,
McIntyre
D
,
Honess
D
, et al
Inhibition of Hedgehog signaling enhances delivery of chemotherapy in a mouse model of pancreatic cancer
.
Science
2009
;
324
:
1457
61
.
13.
Schneider
CA
,
Rasband
WS
,
Eliceiri
KW
. 
NIH Image to ImageJ: 25 years of image analysis
.
Nat Methods
2012
;
9
:
671
5
.
14.
Trapnell
C
,
Williams
BA
,
Pertea
G
,
Mortazavi
A
,
Kwan
G
,
van Baren
MJ
, et al
Transcript assembly and quantification by RNA-Seq reveals unannotated transcripts and isoform switching during cell differentiation
.
Nat Biotechnol
2010
;
28
:
511
5
.
15.
Kim
CC
,
Lanier
LL
. 
Beyond the transcriptome: completion of act one of the Immunological Genome Project
.
Curr Opin Immunol
2013
;
25
:
593
7
.
16.
Robinson
MD
,
Oshlack
A
. 
A scaling normalization method for differential expression analysis of RNA-seq data
.
Genome Biol
2010
;
11
:
R25
.
17.
Gong
T
,
Szustakowski
JD
. 
DeconRNASeq: a statistical framework for deconvolution of heterogeneous tissue samples based on mRNA-Seq data
.
Bioinformatics
2013
;
29
:
1083
5
.
18.
Kazachenka
A
,
Bertozzi
TM
,
Sjoberg-Herrera
MK
,
Walker
N
,
Gardner
J
,
Gunning
R
, et al
Identification, characterization, and heritability of murine metastable epialleles: Implications for non-genetic inheritance
.
Cell
2018
;
175
:
1259
71
.
19.
Li
G
,
Ji
Y
,
Liu
C
,
Li
J
,
Zhou
Y
. 
Reduced levels of p15INK4b, p16INK4a, p21cip1 and p27kip1 in pancreatic carcinoma
.
Mol Med Rep
2012
;
5
:
1106
10
.
20.
Gundra
UM
,
Girgis
NM
,
Ruckerl
D
,
Jenkins
S
,
Ward
LN
,
Kurtz
ZD
, et al
Alternatively activated macrophages derived from monocytes and tissue macrophages are phenotypically and functionally distinct
.
Blood
2014
;
123
:
e110
22
.
21.
Schmieder
A
,
Schledzewski
K
,
Michel
J
,
Schonhaar
K
,
Morias
Y
,
Bosschaerts
T
, et al
The CD20 homolog Ms4a8a integrates pro- and anti-inflammatory signals in novel M2-like macrophages and is expressed in parasite infection
.
Eur J Immunol
2012
;
42
:
2971
82
.
22.
Yu
B
,
Sondag
GR
,
Malcuit
C
,
Kim
MH
,
Safadi
FF
. 
Macrophage-associated Osteoactivin/GPNMB mediates mesenchymal stem cell survival, proliferation, and migration via a CD44-dependent mechanism
.
J Cell Biochem
2016
;
117
:
1511
21
.
23.
Ripoll
VM
,
Irvine
KM
,
Ravasi
T
,
Sweet
MJ
,
Hume
DA
. 
Gpnmb is induced in macrophages by IFN-gamma and lipopolysaccharide and acts as a feedback regulator of proinflammatory responses
.
J Immunol
2007
;
178
:
6557
66
.
24.
Ericson
JA
,
Duffau
P
,
Yasuda
K
,
Ortiz-Lopez
A
,
Rothamel
K
,
Rifkin
IR
, et al
Gene expression during the generation and activation of mouse neutrophils: implication of novel functional and regulatory pathways
.
PLoS One
2014
;
9
:
e108553
.
25.
Gallina
G
,
Dolcetti
L
,
Serafini
P
,
De Santo
C
,
Marigo
I
,
Colombo
MP
, et al
Tumors induce a subset of inflammatory monocytes with immunosuppressive activity on CD8+ T cells
.
J Clin Invest
2006
;
116
:
2777
90
.
26.
Schmieder
A
,
Michel
J
,
Schonhaar
K
,
Goerdt
S
,
Schledzewski
K
. 
Differentiation and gene expression profile of tumor-associated macrophages
.
Semin Cancer Biol
2012
;
22
:
289
97
.
27.
Gordon
S
,
Pluddemann
A
. 
Tissue macrophage heterogeneity: issues and prospects
.
Semin Immunopathol
2013
;
35
:
533
40
.
28.
van Rijt
LS
,
Vos
N
,
Willart
M
,
Muskens
F
,
Tak
PP
,
van der Horst
C
, et al
Persistent activation of dendritic cells after resolution of allergic airway inflammation breaks tolerance to inhaled allergens in mice
.
Am J Respir Crit Care Med
2011
;
184
:
303
11
.
29.
Nair
MG
,
Cochrane
DW
,
Allen
JE
. 
Macrophages in chronic type 2 inflammation have a novel phenotype characterized by the abundant expression of Ym1 and Fizz1 that can be partly replicated in vitro
.
Immunol Lett
2003
;
85
:
173
80
.
30.
Welch
JS
,
Escoubet-Lozach
L
,
Sykes
DB
,
Liddiard
K
,
Greaves
DR
,
Glass
CK
. 
TH2 cytokines and allergic challenge induce Ym1 expression in macrophages by a STAT6-dependent mechanism
.
J Biol Chem
2002
;
277
:
42821
9
.
31.
Webb
DC
,
McKenzie
AN
,
Foster
PS
. 
Expression of the Ym2 lectin-binding protein is dependent on interleukin (IL)-4 and IL-13 signal transduction: identification of a novel allergy-associated protein
.
J Biol Chem
2001
;
276
:
41969
76
.
32.
Chang
CI
,
Liao
JC
,
Kuo
L
. 
Macrophage arginase promotes tumor cell growth and suppresses nitric oxide-mediated tumor cytotoxicity
.
Cancer Res
2001
;
61
:
1100
6
.
33.
Wang
L
,
Le Mercier
I
,
Putra
J
,
Chen
W
,
Liu
J
,
Schenk
AD
, et al
Disruption of the immune-checkpoint VISTA gene imparts a proinflammatory phenotype with predisposition to the development of autoimmunity
.
Proc Natl Acad Sci U S A
2014
;
111
:
14846
51
.
34.
Wang
L
,
Rubinstein
R
,
Lines
JL
,
Wasiuk
A
,
Ahonen
C
,
Guo
Y
, et al
VISTA, a novel mouse Ig superfamily ligand that negatively regulates T cell responses
.
J Exp Med
2011
;
208
:
577
92
.
35.
Liu
J
,
Yuan
Y
,
Chen
W
,
Putra
J
,
Suriawinata
AA
,
Schenk
AD
, et al
Immune-checkpoint proteins VISTA and PD-1 nonredundantly regulate murine T-cell responses
.
Proc Natl Acad Sci U S A
2015
;
112
:
6682
7
.
36.
Balli
D
,
Rech
AJ
,
Stanger
BZ
,
Vonderheide
RH
. 
Immune Cytolytic Activity Stratifies Molecular Subsets of Human Pancreatic Cancer
.
Clin Cancer Res
2017
;
23
:
3129
38
.
37.
Blando
J
,
Sharma
A
,
Higa
MG
,
Zhao
H
,
Vence
L
,
Yadav
SS
, et al
Comparison of immune infiltrates in melanoma and pancreatic cancer highlights VISTA as a potential target in pancreatic cancer
.
Proc Natl Acad Sci U S A
2019
;
116
:
1692
7
.
38.
Reck
M
,
Rodriguez-Abreu
D
,
Robinson
AG
,
Hui
R
,
Csoszi
T
,
Fulop
A
, et al
Pembrolizumab versus Chemotherapy for PD-L1-positive non-small-cell lung cancer
.
N Engl J Med
2016
;
375
:
1823
33
.
39.
Robert
C
,
Schachter
J
,
Long
GV
,
Arance
A
,
Grob
JJ
,
Mortier
L
, et al
Pembrolizumab versus Ipilimumab in advanced melanoma
.
N Engl J Med
2015
;
372
:
2521
32
.
40.
Nomi
T
,
Sho
M
,
Akahori
T
,
Hamada
K
,
Kubo
A
,
Kanehiro
H
, et al
Clinical significance and therapeutic potential of the programmed death-1 ligand/programmed death-1 pathway in human pancreatic cancer
.
Clin Cancer Res
2007
;
13
:
2151
7
.
41.
Fearon
DT
. 
The carcinoma-associated fibroblast expressing fibroblast activation protein and escape from immune surveillance
.
Cancer Immunol Res
2014
;
2
:
187
93
.
42.
von Ahrens
D
,
Bhagat
TD
,
Nagrath
D
,
Maitra
A
,
Verma
A
. 
The role of stromal cancer-associated fibroblasts in pancreatic cancer
.
J Hematol Oncol
2017
;
10
:
76
.
43.
Kumar
V
,
Donthireddy
L
,
Marvel
D
,
Condamine
T
,
Wang
F
,
Lavilla-Alonso
S
, et al
Cancer-associated fibroblasts neutralize the anti-tumor effect of CSF1 receptor blockade by inducing PMN-MDSC infiltration of tumors
.
Cancer Cell
2017
;
32
:
654
68
.
44.
Feig
C
,
Jones
JO
,
Kraman
M
,
Wells
RJ
,
Deonarine
A
,
Chan
DS
, et al
Targeting CXCL12 from FAP-expressing carcinoma-associated fibroblasts synergizes with anti-PD-L1 immunotherapy in pancreatic cancer
.
Proc Natl Acad Sci U S A
2013
;
110
:
20212
7
.
45.
Mace
TA
,
Shakya
R
,
Pitarresi
JR
,
Swanson
B
,
McQuinn
CW
,
Loftus
S
, et al
IL-6 and PD-L1 antibody blockade combination therapy reduces tumour progression in murine models of pancreatic cancer
.
Gut
2018
;
67
:
320
32
.
46.
Dear
AE
. 
Epigenetic modulators and the new immunotherapies
.
N Engl J Med
2016
;
374
:
684
6
.
47.
Hessmann
E
,
Johnsen
SA
,
Siveke
JT
,
Ellenrieder
V
. 
Epigenetic treatment of pancreatic cancer: is there a therapeutic perspective on the horizon?
Gut
2017
;
66
:
168
79
.
48.
Topper
MJ
,
Vaz
M
,
Chiappinelli
KB
,
DeStefano Shields
CE
,
Niknafs
N
,
Yen
RC
, et al
Epigenetic therapy ties MYC depletion to reversing immune evasion and treating lung cancer
.
Cell
2017
;
171
:
1284
300
.
49.
Stone
ML
,
Chiappinelli
KB
,
Li
H
,
Murphy
LM
,
Travers
ME
,
Topper
MJ
, et al
Epigenetic therapy activates type I interferon signaling in murine ovarian cancer to reduce immunosuppression and tumor burden
.
Proc Natl Acad Sci U S A
2017
;
114
:
E10981
e90
.
50.
Nair
MG
,
Du
Y
,
Perrigoue
JG
,
Zaph
C
,
Taylor
JJ
,
Goldschmidt
M
, et al
Alternatively activated macrophage-derived RELM-alpha is a negative regulator of type 2 inflammation in the lung
.
J Exp Med
2009
;
206
:
937
52
.
51.
Seifert
L
,
Werba
G
,
Tiwari
S
,
Giao Ly
NN
,
Nguy
S
,
Alothman
S
, et al
Radiation therapy induces macrophages to suppress T-cell responses against pancreatic tumors in mice
.
Gastroenterology
2016
;
150
:
1659
72
.
52.
Wan
G
,
Xiang
L
,
Sun
X
,
Wang
X
,
Li
H
,
Ge
W
, et al
Elevated YKL-40 expression is associated with a poor prognosis in breast cancer patients
.
Oncotarget
2017
;
8
:
5382
91
.
53.
Partecke
LI
,
Gunther
C
,
Hagemann
S
,
Jacobi
C
,
Merkel
M
,
Sendler
M
, et al
Induction of M2-macrophages by tumour cells and tumour growth promotion by M2-macrophages: a quid pro quo in pancreatic cancer
.
Pancreatology
2013
;
13
:
508
16
.
54.
Singer
K
,
Gottfried
E
,
Kreutz
M
,
Mackensen
A
. 
Suppression of T-cell responses by tumor metabolites
.
Cancer Immunol Immunother
2011
;
60
:
425
31
.
55.
Zhang
Y
,
Velez-Delgado
A
,
Mathew
E
,
Li
D
,
Mendez
FM
,
Flannagan
K
, et al
Myeloid cells are required for PD-1/PD-L1 checkpoint activation and the establishment of an immunosuppressive environment in pancreatic cancer
.
Gut
2017
;
66
:
124
36
.
56.
Flies
DB
,
Wang
S
,
Xu
H
,
Chen
L
. 
Cutting edge: a monoclonal antibody specific for the programmed death-1 homolog prevents graft-versus-host disease in mouse models
.
J Immunol
2011
;
187
:
1537
41
.
57.
Flies
DB
,
Han
X
,
Higuchi
T
,
Zheng
L
,
Sun
J
,
Ye
JJ
, et al
Coinhibitory receptor PD-1H preferentially suppresses CD4(+) T cell-mediated immunity
.
J Clin Invest
2014
;
124
:
1966
75
.
58.
Steele
CW
,
Karim
SA
,
Leach
JDG
,
Bailey
P
,
Upstill-Goddard
R
,
Rishi
L
, et al
CXCR2 inhibition profoundly suppresses metastases and augments immunotherapy in pancreatic ductal adenocarcinoma
.
Cancer Cell
2016
;
29
:
832
45
.
59.
Habtezion
A
,
Edderkaoui
M
,
Pandol
SJ
. 
Macrophages and pancreatic ductal adenocarcinoma
.
Cancer Lett
2016
;
381
:
211
6
.
60.
Takahashi
H
,
Sakakura
K
,
Kudo
T
,
Toyoda
M
,
Kaira
K
,
Oyama
T
, et al
Cancer-associated fibroblasts promote an immunosuppressive microenvironment through the induction and accumulation of protumoral macrophages
.
Oncotarget
2017
;
8
:
8633
47
.
61.
Pitarresi
JR
,
Liu
X
,
Sharma
SM
,
Cuitino
MC
,
Kladney
RD
,
Mace
TA
, et al
Stromal ETS2 regulates chemokine production and immune cell recruitment during acinar-to-ductal metaplasia
.
Neoplasia
2016
;
18
:
541
52
.
62.
Bayne
LJ
,
Beatty
GL
,
Jhala
N
,
Clark
CE
,
Rhim
AD
,
Stanger
BZ
, et al
Tumor-derived granulocyte-macrophage colony-stimulating factor regulates myeloid inflammation and T cell immunity in pancreatic cancer
.
Cancer Cell
2012
;
21
:
822
35
.
63.
Ekiz
HA
,
Lai
SA
,
Gundlapalli
H
,
Haroun
F
,
Williams
MA
,
Welm
AL
. 
Inhibition of RON kinase potentiates anti-CTLA-4 immunotherapy to shrink breast tumors and prevent metastatic outgrowth
.
Oncoimmunology
2018
;
7
:
e1480286
.