C57BL/6 female mice aged 3 months were obtained from Charles River. These mice were used to generate the Panc-02 model. Male and female KPC-wild type mice were maintained at Einstein, which have the same genetic background as KPC mice (LSL^KrasG12D+/- LSL-p53^R17+/- Pdx1 Cre)25 but are negative for Kras^G12D and Tp53^R172H mutations and for Cre recombinase. These mice were used to generate the orthotopic KPC mice. In addition, a nude mouse model was used, that is, Fox n1nu/J (cat#000819), Jackson Laboratories, lacking T cells.

The Panc-02 cell line was derived from a methylcholanthrene-induced ductal adenocarcinoma growing in a C57BL/6 female mouse.26 Panc-02 cells were cultured in McCoy’s5A medium supplemented with 10% fetal bovine serum, 2 mM glutamine, non-essential amino acids, 1 mM sodium pyruvate and 100 U/mL Pen/Strep. The KPC tumor cell line was derived from a pancreatic KPC tumor (LSL^KrasG12D+/- LSL-p53^R172-/-; Pdx1-Cre) in our lab, which was kindly provided by Jacco van Rheenen, Cancer Genomics, Utrecht, The Netherlands.27

Immune cells from spleens of mice were isolated as described previously.28 Immune cells were also isolated from pancreatic tumors using Dispase (Roche cat#049422078001) and Collagenase (Sigma-Aldrich cat #C0130) as we described previously.29 Anti-CD3 (BD Bioscience, cat # 560527) and anti-CD8 antibodies (BD Bioscience, cat # 552877) were used to identify CD8 T cells, anti-CD3 and anti-CD4 (BD Bioscience, cat # 552051) to identify CD4 T cells, anti-CD11b (BD Bioscience, cat # 553312) and anti-Gr1 (BD Bioscience, cat # 553127) to identify MDSC, and anti-CD11b and anti-F4/80 (eBioscience, cat # 17-4801-82) to identify TAM. Appropriate isotype controls were included for each sample. A total of 50 000–1 00 000 cells were acquired by scanning using a LSR-II Fluorescence Activated Cell Sorter system special order (Beckton and Dickinson), and analyzed using FlowJo V.7.6 software. Cell debris and dead cells were excluded from the analysis based on scatter signals and use of Fixable Blue or Green Live/Dead Cell Stain Kit (BD Bioscience, cat # 564406).

Tumors were dissected from pancreas and immediately fixed with buffered formalin, and the tissue was embedded in paraffin. Sections (5 µm) were sliced and placed on slides, then deparaffinized at 60°C for 1 hour, followed by xylene, an ethanol gradient (100%–70%), water, and PBS. Slides were then incubated for 30 min in 3% hydrogen peroxide, followed by boiling in citrate buffer for 20 min. Once the slides were cooled, washed, and blocked with 5% goat serum, the sections were incubated with primary antibodies such as anti-CD4 (Cell Signaling Technology, cat # 25229) (1:100 dilution), anti-CD8α (1:400 dilution)(Cell Signaling Technology, cat # 98941), anti-Perforin (Cell Signaling Technology, cat # 31647)(1:300 dilution), anti-Granzyme (Cell Signaling Technology, cat # 44153) (1:200 dilution), anti-CD31 (Cell Signaling Technology, cat # 77699) (1:100 dilution), or anti-α-Smooth muscle actin (αSMA) antibodies (Cell Signaling Technology, cat # 19245)(1:400), followed by incubation with secondary antibody (mouse anti-goat IgG-HRP) (Cell Signaling Technology, cat# 8114S), and SignalStain Boost immunohistochemistry (IHC) Detection Reagent (Cell Signaling Technology, cat# 8114S). Subsequently, the slides were incubated with 3,3'-diaminobenzidine (DAB) (Vector Laboratories, cat# SK-4100), counterstained with hematoxylin, dehydrated through an ethanol gradient (70%–100%) and xylene, and mounted with Permount. The slides were scanned with a 3D Histech P250 High Capacity Slide Scanner to acquire images and quantification data. Secondary antibodies without primary antibodies were used as negative control.

CD4 and CD8 T cells were depleted in C57BL/6 mice with orthotopic Panc-02 tumors during GEM+NAM treatment. Briefly, 10 days after tumor cell injection (when tumors were approximately 5 mm²), mice were treated with NAM+GEM and with 300 µg of anti-CD4 (Clone GK1.5; Cat#BE0003-1: BioXCell) or anti-CD8 (Clone YTS169.4; Cat$BE0117: BioXCell) antibodies (five injections every third day), as outlined in online supplemental figure S1D. All mice were euthanized 2 days after the last anti-CD4 or anti-CD8 treatment and analyzed for tumor weight and number of metastases. As control, isotype-matched rat antibodies against HRPN were used (Clone LTF-2; Cat#BE0090: BioXCell).

Pieces of 3 mm³ tumors were submerged in 5 vol of RNAlater (Invitrogen) (n=5 samples/group). Total RNA was isolated from these tissues using TRIzol (Life Technologies) as we described previously.30 The RNA concentrations in the samples were measured using a NanoDrop 2000 instrument (Thermo Fisher Scientific) and checked for quality on agarose gels. All RNA samples used in this study exhibited optical density (OD) 260/280 ratios greater than 1.9 and RNA integrity numbers greater than 8.5.

A total of 770 immune-related mouse genes were analyzed using the nCounter Mouse PanCancer Immune Profiling Panel (NanoString), which are categorized in various pathways (online supplemental table S1).

RNA in situ hybridization (RNAscope, ACD Bio-Techne, Minneapolis, Minnesota, USA) was performed on formalin-fixed paraffin-embedded sections using the RNAScope 2.5 HD Detection Kit RED according to manufacturer’s instructions. Probes used were directed against mouse Ccl21a (cat no 489921) and mouse Fcer1a (cat no 886381). Scans of whole tissues were acquired with Aperio CS2 (Leica Biosystems). Images analyzes and quantifications were performed on whole slides with HALO Image Analysis (Indica Labs). Images were acquired using EVOS M7000 Imaging System (Thermo Fisher Scientific).

Differential expression analyzes of mRNA expression data in 20 samples were performed by using the DESeq2 R package V.1.20.0. A total of 770 immunology-related mouse genes, created from the nCounter Mouse PanCancer Immune Profiling Panel (NanoString), were then implemented as candidate genes in this study. GSVA was employed to detect the variation values of the GO term pathways in each group using the R package GSVA.

All experiments were repeated in triplicate unless otherwise stated. Statistical analyzes were performed using GraphPad Prism software V.7 (GraphPad Software), and statistical significance was defined as p<0.05. Two-way analysis of variance was performed on the experimental data for tumor volumes and mouse weights. The results are presented as the mean±SE of the mean.

Differentially expressed genes were compared with Student’s t-test, and adjusted p values of less than 0.01 and fold changes of greater than 2 were considered to indicate significant dysregulation. p values were adjusted by the Benjamini-Hochberg (BH) method. Adjusted p values are also called q values. Data were analyzed using R (V.3.5.1).

To statistically compare the effects NAM+GEM on the growth of metastases and tumors or on immune responses in the mouse models, the Mann-Whitney U test was applied using Prism. *p<0.05, **<0.01, ***<0.001, ****<0.0001. Values of p<0.05 were considered statistically significant. Survival studies were analyzed using Mantel-Cox test. p<0.05 is significant.

The effect of low doses of NAM (20 mg/dose) and GEM (1.2 mg/dose) on pancreatic cancer was analyzed in two different tumor models, that is, an immune competent orthotopic Panc-02 model, and an immune competent orthotopic KPC model. First, we tested NAM+GEM in the immune competent orthotopic Panc-02 model. Briefly, 10⁶ tumor cells were orthotopically injected into the pancreas and 10 days later the treatment with NAM+GEM started as outlined in online supplemental figure S1A (at this time point tumors are palpable in the pancreas) and the effect of treatment on tumors and metastases was analyzed. As shown in figure 1AB, a significant reduction in tumor weight (p<0.0001) and metastases (p<0.01) was observed in the NAM+GEM group compared with the saline group, while ascites production was relatively high in all control groups (grade 0–5) but relatively minor in the NAM+GEM group (grade 0–2) (figure 1C). The tumor weight and number of metastases in mice treated with the vehicle (negative control: Apple Juice+Pepper), were not significantly different from the saline-treated mice (online supplemental figure S2). Tumors and metastases are depicted in online supplemental figure S3AB.

Second, we tested the NAM+GEM treatment in the immune competent orthotopic KPC model. Briefly, 10⁵ KPC tumor cells were injected into the pancreas (the tumors in this model are more aggressive than in the orthotopic Panc-02 model because of the homozygous p53^R172-/- mutation), and 10 days later the treatment with NAM+GEM was started (at this time point tumors are palpable in the pancreas) and continued for 14 days as outlined in online supplemental figure S1B. As shown in figure 1DE, the tumor weight in the NAM+GEM group was significantly reduced (p<0.05) as well as the number of metastases (p<0.01), while the ascites production was high in all control groups (grade 0–5), and minor in the NAM+GEM group (grade 0–2) (figure 1F). Tumors and metastases are depicted in online supplemental figure S4AB. Several mice died in the saline, NAM, or GEM group because of the aggressive cancer, but none in the NAM+GEM group.

To analyze the clinical relevance of the combination therapy, we tested the effect of NAM+GEM on the survival rate of orthotopic Panc-02 mice. Briefly, 10⁶ tumor cells were injected into the pancreas and 10 days later the NAM+GEM treatment was started and continued for 14 days as described above. Mice were monitored for the following weeks without any further treatment. Here, we show that NAM+GEM significantly improved the survival time of the orthotopic Panc-02 mice not only compared with the saline group but also compared with all other control groups ( figure 1G ).

In skin cancer prevention, it has been shown that NAM significantly increased peritumoral and infiltrating CD4 and CD8 T cells compared with a placebo.22 Here, we tested the effect of the combination of GEM+NAM on immune cells in the orthotopic Panc-02 tumors. At the end of treatment, whole tumors were digested with collagenase and dispase to generate single cell suspensions as we described previously.14 Subsequently, the cell population was analyzed by flow cytometry. We found that the CD45⁺ cells (leucocytes) (p<0.0001), CD4⁺ (p<0.05) and CD8⁺ T cells (p<0.01) significantly increased in the NAM+GEM group compared with the saline group (figure 2A). We also analyzed the number of TAMs in the Panc-02 tumors and found a significant reduction in the NAM+GEM group, but not in the GEM only group, compared with saline (p<0.05) (figure 2B). The MDSC population was also reduced in tumors of the NAM+GEM group, although this was not statistically significant because of high variability in the saline treated mice (figure 2C). However, the MDSC population in blood was reduced by GEM alone (p<0.05) (figure 2D), supporting our results in a previous study.14

To confirm the flow cytometry data, we analyzed which T cells infiltrated the pancreatic tumors by IHC, an approach we described previously.14 As shown in figure 2E–H and online supplemental figure S5AB, the number of CD4 and CD8 T cells was significantly higher in the tumors of orthotopic Panc-02 mice treated with NAM+GEM (p<0.01 compared with the saline group).

We also analyzed the production of perforin and granzyme B in the pancreatic tumors through IHC. We found a significant increase in the production of granzyme B (p<0.01) (figure 2IJ and online supplemental figure S5C) but less robust in perforin in the pancreatic tumors of mice treated with NAM+GEM compared with saline (figure 2KL and online supplemental figure S5D).

Since a strong decrease was observed in the size of the pancreatic tumors accompanied with a strong influx of CD4 and CD8 T cells, we analyzed whether these T cells exhibited antitumor reactivity. For this purpose, spleen cells of NAM+GEM-treated and control mice were restimulated with survivin, a TAA expressed by the Panc-02 tumors,24 and analyzed by ELISPOT and magnetic bead technology. It appeared that CD4 and CD8 T cells were highly activated in the NAM+GEM group but not in the other groups (figure 2M).

Since the number and the activity of both CD4 and CD8 T cells was significantly increased in the orthotopic Panc-02 tumors of NAM+GEM-treated mice in correlation with a significant reduction of tumors and metastases, we analyzed in further detail whether this NAM+GEM effect could be reduced by CD4 or CD8 T cells depletions in vivo, as outlined in online supplemental figure S1D. As shown in figure 3A, the average tumor weight in mice treated with NAM+GEM plus CD4 antibodies was significantly increased by 58% compared with the NAM+GEM alone group (p=0.0397, Mann-Whitney U test), while the tumors in mice depleted for CD8 T cells were increased by 22% compared with the NAM+GEM alone group but this was not significant (p=0.3016, Mann-Whitney U test). As expected, the tumors in mice treated with NAM+GEM plus the isotype control (negative control) were not significantly different from the tumor in mice treated with NAM+GEM only (p=0.6111, Mann-Whitney U test). Similarly, the number of metastases in mice treated with NAM+GEM plus CD4 antibodies significantly increased by 97% compared with mice treated with NAM+GEM alone (p=0.0238, Mann-Whitney U test), while the number of metastases in mice treated with NAM+GEM plus antibodies to CD8 T cells was increased by 66% but this was statistically not significant (p=0.4048, Mann-Whitney U test) (figure 3B). Also, here the number of metastases in mice treated with NAM+GEM plus the isotype control was not significantly different compared with mice treated with NAM+GEM only (p=0.6825, Mann-Whitney U test). Flow cytometry analysis of the spleen confirmed that CD4 and CD8 T cells were efficiently depleted in vivo (figure 3C). Pictures of tumors of all mice are shown in online supplemental figure S6.

After demonstrating that in NAM+GEM treatment CD4 T cells significantly contributed to eradication of the pancreatic cancer, we also determined the contribution of NAM+GEM to eradication of tumors and metastases in nude mice (lacking T cells). For this purpose, we generated in addition to the mouse models in figure 1, a third mouse model, that is, orthotopic Panc-02 tumors in nude mice (Fox n1nu/J (000819)) in a C57BL/6 background, followed by NAM+GEM treatment and compared with Saline treatment. As shown in figure 3D,E, the average tumor weight decreased by 40% and number of metastases by 45%. However, this was statistically not significant. Pictures of tumors of all mice are shown in (online supplemental figure S7).

The immunomodulatory effects observed in the above experiments demonstrates that T cell infiltration is improved in the NAM+GEM group. This compelled us to look at altered architecture of the tumor surrounding stroma such as the ECM. For instance, collagen I, produced by CAFs, is known to prevent penetration of drugs and immune cells into the pancreatic tumors, particularly when cross-linked by hyaluronic acid fibrils. Here we showed a significant decrease (p<0.05) in collagen I by Trichrome staining in the tumor area of NAM+GEM-treated mice compared with the saline group (figure 4A–C and online supplemental figure S8A). Moreover, we found a reduction in the expression of the hyaluronic acid binding protein (HABP) through RT-PCR in the NAM+GEM group compared with the saline group (figure 4D). Also, NAM or GEM separately reduced the expression of HABP.

αSMA, a protein expressed by a myofibroblastic CAF subpopulation, was significantly increased (p<0.05) in the tumor area of NAM+GEM-treated mice compared with the saline group (figure 4E–G and online supplemental figure S8B). Also, GEM-treated mice showed a significant increase in the αSMA compared with the saline group. Various reports have been published about the function of αSMA, which will be discussed later.

To allow the influx of immune cells into the tumors, blood vessels will be required. Therefore, we analyzed the tumors for CD31 expression. We found that CD31 was most abundantly expressed in the tumor areas of NAM+GEM treated mice (figure 4H–J and online supplemental figure S8CD).

LNS are often observed in patients with PDAC. Some reported a correlation with improved outcome and others a correlation with a worse outcome.31 32 Here, we found peritumoral LNS in the NAM+GEM group (4/4) with CD4 and CD8 T cells (figure 5A), and less frequently intratumoral LNS in NAM (3/5) and GEM (2/5) groups but not in the saline group (online supplemental figure S9). IHC showed CD31-positive vessels in both peritumoral and intratumoral LNS (figure 5B). More detail about peritumoral and intratumoral LNS in the different treatment groups is shown in (online supplemental figure S9).

To obtain more insight in the pathways potentially involved in the above-described changes, we analyzed tumors of all treatment groups of the orthotopic Panc-02 model by Nanostring Technology and assessed specifically an immune-oncology gene panel (online supplemental table S1). We found significant changes in expression of genes relevant for immune responses and the progression/regression of the pancreatic cancer in the NAM+GEM group compared the saline group. This includes, a 16-fold increase in the expression of Ccl21a, which is involved in T cell migration and recruitment of LNS, a fourfold increase in Interleukin-1 Receptor-Like 2 (IL1RL2), which is involved in activating T cells and increasing DC immunogenicity33, a 10-fold increase in Fcer-1 which is involved in epitope spreading34, a fourfold increase in Ccl9 which is involved in attracting DC and activating T cells35. Chemokine (C-C) ligand 21a (Ccl21a) was also increased in the NAM or GEM only groups but less robust than in NAM+GEM combination. In another study, we also found that GEM improved the migration of T cells to pancreatic tumors.14 Finally, we found a fourfold reduction in Lipocalin-2 (Lcn-2)36 and a less robust but still significant reduction in Mucin-1 (Muc-1)37 and Epithelial cell adhesion molecule (Epcam)38. These genes are involved in the promotion of invasive angiogenic drug resistant tumor cells, invasiveness and tumor progression. A heatmap of the most relevant genes is shown in figure 6A and an explanation of the function of genes in figure 6B and refs 33–42.

To validate the nanostring data, we tested the expression of Ccl21 (involved in the formation of LNS and T cell migration) by RNA hybridization (RNAscope) in the tumors of the orthothopic Panc-02 model. We found a significantly higher number of cells expressing Ccl21 in the NAM+GEM group compared with the saline group, but also NAM or GEM showed increased numbers of Ccl21-positive cells compared with the saline group, although this was not significant (figure 6C,E). A similar pattern was observed for Fcer. The highest number of cells expressing Fcer was found in the NAM+GEM group compared with the saline group, and less in the NAM or GEM groups (figure 6D,F).