Arginase 1 is a key driver of immune suppression in pancreatic cancer

1Pancreatic ductal adenocarcinoma (PDA) is currently the third leading cause of cancer-related deaths in the United States with a 5-y survival rate of 12%. This poor survival rate is due to late detection and ineffective treatments. The hallmark mutation in PDA is found in the KRAS gene, most commonly KRASG12D, whereas disease progression is accelerated by loss of tumor suppressor genes, including TP53, SMAD4, and INK4A.

2Standard of care for PDA is a combination of systemic chemotherapy that includes FOLFIRINOX or gemcitabine plus nab-paclitaxel, which provides limited improvement in patient survival. Oncogenic KRAS proteins were long thought to be ‘undruggable’. Recently, new approaches to drug discovery have led to the development of KRASG12C inhibitors, a common mutation in non-small cell lung cancer; initial clinical data show not only dramatic responses but also cases of primary or secondary resistance. While the applicability of G12C inhibitors is limited in PDA, which is rarely present with this specific allele, new KRASG12D inhibitors have recently been described and are entering the clinic. Based on the observations in lung cancer and on predictions based on mouse models that allow genetic inactivation of oncogenic KRAS, it is to be expected that resistance to KRAS G12D inhibitors will similarly arise; activation of anti-tumor immune responses remains our best hope for long-term tumor control. Although immune checkpoint inhibitors are effective in other cancers, this benefit has not translated to PDA due to the severely immunosuppressive tumor microenvironment (TME) that characterizes this disease. Approaches combining immune checkpoints and myeloid cell targeting might be more promising, based on preclinical studies. Here, we explore new approaches targeting macrophages to induce anti-tumor immunity in pancreatic cancer.

3The pancreatic TME includes cancer-associated fibroblasts and a heterogenous population of immune cells, the majority of which is myeloid cells. These myeloid cells include tumor-associated macrophages (TAMs), immature myeloid cells, also referred to as myeloid-derived suppressor cells, and granulocytes, such as neutrophils. In contrast, CD8+ T cells are rare in the pancreatic TME, although their prevalence is heterogeneous in different patients. Single-cell RNA sequencing (sc-RNA-seq) analysis revealed that most CD8+ T cells infiltrating PDA have an exhausted phenotype. An understanding of the mechanisms mediating immune suppression in pancreatic cancer is needed to design new therapeutic approaches for this disease. Of note, analysis of long-term survivors revealed persistence of tumor-specific memory CD8+ T cells, indicating that when an anti-tumor immune response does occur, it leads to effective tumor control. Conversely, evidence of loss of antigens over time due to immunoediting has also been described. Furthermore, myeloid cell depletion in mouse models of pancreatic cancer led to activation of anti-tumor T cell responses, spurring an effort to target myeloid cells in pancreatic cancer; yet, clinical efficacy has been low, and more effective approaches are needed.

4Myeloid cells infiltrating the neoplastic pancreas, and ultimately those in PDA, express high levels of Arginase 1 (Arg1). ARG1 is one of the enzymes that metabolize arginine, and it is prevalently located in the cytoplasm, while the ARG2 isoform is prevalently mitochondrial. Indeed, we previously showed that Arg1 expression in myeloid cells is driven by oncogenic Kras expression/signaling in epithelial cells, starting during early stages of carcinogenesis. Beyond our work, ARG1 is widely appreciated as a marker of alternatively polarized macrophages. An increase in tumor myeloid Arg1 expression has been reported in other cancers, including renal cell carcinoma, breast, colon, and lung cancer.

5In addition to its role as a marker of alternatively polarized macrophages, ARG1 is a metabolic enzyme that breaks down the amino acid L-arginine to urea and ornithine. Connecting this activity to myeloid Arg1 expression, several reports detail how arginine is necessary for the activation and proliferation of CD8 T cells. This indicates that myeloid cells may deplete arginine in the TME to dampen anti-tumor T cell activity. Based on these concepts, a small-molecule arginase inhibitor, CB-1158 (INCB001158) (Calithera Biosciences, Inc, South San Francisco, CA), was developed. CB-1158 treatment as monotherapy, or in combination with anti-PD-1 checkpoint inhibitor, decreased tumor growth in vivo in mice with Lewis lung carcinoma. CB-1158 also inhibits human arginase, and it is being tested in a phase I clinical trial in patients with advanced or metastatic solid tumors.

6Encouraged by these studies and the prominent disease-specific expression of Arg1 in PDA macrophages, we set forth to determine its functional role. Here, we used a dual-recombinase approach to delete Arg1 in myeloid cell lineages via Cre-loxP technology and at the same time induced oncogenic Kras in pancreatic epithelial cells using the orthogonal Flp-Frt recombination approach. We discovered that Arg1 deletion in myeloid cells profoundly reshaped the TME, increased the infiltration of CD8+ T cells, and reduced malignant disease progression. However, we also noticed a resistance mechanism whereby other cells in the TME, such as epithelial cells, upregulate Arg1 expression, potentially blunting the effect of its inactivation in myeloid cells. We thus employed a systemic approach, whereby we treated a syngeneic orthotopic model of pancreatic cancer with the arginase inhibitor CB-1158 and found it to sensitize PDA to anti-PD1 immune checkpoint blockade. These results illustrate a functional role of Arg1 in pancreatic tumor-derived myeloid cells, reveal novel aspects of intratumoral compensatory metabolism, and provide new inroads to increase the efficacy of checkpoint immune therapy for PDA.

7We have previously reported that expression of oncogenic Kras in pancreas epithelial cells drives expression of Arg1 in macrophages in vivo during early stages of carcinogenesis. We sought to determine whether Arg1 was also associated with late stages of carcinogenesis in mouse and human tumors. As there was no validated antibody for human ARG1 available, we performed ARG1 RNA in situ hybridization in human PDA, together with co-immunofluorescence staining for the immune cell marker CD45 and for the epithelial cell marker E-cadherin (ECAD) (Figure 1A). We observed prevalent ARG1 expression in CD45+ cells, and occasional low expression in ECAD+ cells (Figure 1A, Figure 1B). Analysis of a previously published sc-RNA-seq dataset, which includes 16 human PDA samples (; Figure 1C and Figure 1—figure supplement 1A), revealed highest expression of ARG1 in myeloid cells (Figure 1D, Figure 1—figure supplement 1B). In contrast, we observed minimal ARG1 expression in CD4+ T cells, CD8+ T cells, epithelial cells, and fibroblasts (Figure 1D). ARG2 was expressed at high levels in endocrine cells, and at lower, but detectable levels in epithelial cells (Figure 1—figure supplement 1B). Thus, we conclude that myeloid cells are the main source of ARG1 in the pancreatic cancer microenvironment.

8Infiltration of myeloid cells, specifically macrophages, in pancreatic cancer portends worse patient survival. We thus assessed whether ARG1 expression correlated with worse patient outcomes. Based on a publicly available human PDA microarray data set (GSE71729), we found that high ARG1 expression in pancreatic cancer correlated with worse survival (Figure 1E), suggesting that ARG1 in myeloid cells may play a functional role in human PDA.

9Next, to determine whether the increase in Arg1 expression was recapitulated in a mouse model of pancreatic cancer, we analyzed our mouse sc-RNA-seq data from healthy mice and tumor-bearing KPC (KrasLSL-G12D/+;p53LSL-R172H/+;Ptf1aCre/+) mice (Figure 1F and Figure 1—figure supplement 1C and D). We detected the highest level of Arg1 expression in macrophages (Figure 1G and Figure 1—figure supplement 1E). Importantly, Arg1 expression was enriched in PDA infiltrating macrophages compared to macrophages in the normal pancreas (Figure 1G and Figure 1—figure supplement 1E). Other cell types, such as epithelial cells and fibroblasts, only had sporadic Arg1 expression. Arg2 was also expressed in macrophages, albeit at low levels, and was the prevalent isoform in tumor-associated granulocytes (Figure 1—figure supplement 1E). To evaluate the protein expression of ARG1, we performed co-immunofluorescence staining. We included healthy pancreas samples, PanIN-bearing pancreata from KC mice (KrasLSL-G12D/+;Ptf1aCre/+), and PDA. We did not detect ARG1 protein in the healthy pancreas. In contrast, both PanIN and PDA presented with frequent co-localization of the macrophage marker F4/80 with ARG1, consistent with prevalent expression in this cell population (Figure 1H and I and Figure 1—figure supplement 1F). Consistent with human data, expression of ARG1 in other cell types was rare (Figure 1H and Figure 1—figure supplement 1F). Thus, both in mouse and human pancreatic cancer, ARG1 is highly expressed and largely confined to tumors, where it is predominantly expressed in macrophages.

10To examine the function of myeloid Arg1 in PDA, we generated mice lacking Arg1 expression in myeloid cells. Specifically, we crossed Arg1f/f mice with Lyz2Cre/+ mice to generate Lyz2Cre/+;Arg1f/f mice. Lyz2Cre/+ mice were generated by inserting the Cre cDNA in the endogenous M lysozyme (LysM) locus, which is broadly expressed in myeloid cells, including macrophages and neutrophils. To validate the deletion of Arg1 in macrophages, we harvested bone marrow (BM) cells from wild type (WT) or Lyz2Cre/+;Arg1f/f mice and cultured these directly in pancreatic cancer cell conditioned media (CM) for 6 d. This protocol yields BM-derived TAMs (Figure 2A), as previously described. ARG1 expression was readily detectable in WT TAMs; in contrast, Lyz2Cre/+;Arg1f/f TAMs had none or very low ARG1 protein, indicating efficient Cre recombination (Figure 2B). We then sought to determine whether ARG1 expression in TAMs affected the metabolite composition of the media. We thus performed extracellular metabolomics by liquid chromatography couple tandem mass spectrometry (LC-MS/MS) and observed elevated levels of L-Arginine in the Lyz2Cre/+;Arg1f/f TAM medium compared with WT TAM medium (Figure 2C). This finding is consistent with WT TAMs depleting arginine from their growth medium at an enhanced rate, relative to TAMs lacking Arg1 expression.

11To investigate the function of Arg1 in myeloid cells during PDA progression, we generated KrasFrt-STOP-Frt-G12D/+;Ptf1aFlpO/+ (KF); Lyz2Cre/+;Arg1f/f mice, hereafter referred to as KFCA (Figure 2D and Figure 2—figure supplement 1A). The dual-recombinase system integrates both Flippase-FRT (Flp-FRT) and Cre-loxP recombination technologies to independently modify epithelial cells and myeloid cells. We aged a cohort of KF and KFCA mice for 2 mo, a timepoint at which PanIN lesions are formed in KF mice (Figure 2—figure supplement 1A). To evaluate the efficiency of Arg1 deletion, we performed co-immunofluorescent staining for ARG1, F4/80, and CK19. As expected, we observed abundant expression of ARG1 in F4/80+ macrophages in KF, and minimal to no expression of ARG1 in macrophages from KFCA tissue (Figure 2—figure supplement 1B). We then performed histopathological evaluation of the tissue to determine the functional effect of loss of myeloid Arg1. Here, we observed a trending decrease in ADM and PanIN lesions (although not statistically significant), accompanied by a reduction of desmoplastic stroma in KFCA, compared to age-matched KF pancreata (Figure 2—figure supplement 1C).

12To investigate the causes underlying the trending reduction in PanIN formation, we performed an in-depth histological analysis. We previously showed that macrophages can directly promote epithelial cell proliferation during early stages in carcinogenesis. We and others have also demonstrated that macrophages also indirectly promote carcinogenesis by suppressing CD8+ T cell infiltration and activation. Immunostaining for the proliferation marker Ki67 revealed a reduction in total cell proliferation, and a reduction in cell death as determined by cleaved caspase-3 (CC3) staining (Figure 2—figure supplement 1D). Immunofluorescence staining for immune cells by CD45 did not reveal changes in total CD45+ cells (Figure 2—figure supplement 1E). Similarly, we did not find changes in infiltrating F4/80+ macrophages or CD8+ T cells by immunohistochemistry staining (Figure 2—figure supplement 1F). Thus, loss of Arg1 in macrophages appears to reduce the ability of macrophages to promote proliferation but, at this stage, does not appear to correlate with changes in the immune system. These findings are consistent with our previous observations on the role of macrophages during the onset of carcinogenesis.

13Given the trending reduction in PanIN formation, we next investigated the effects of myeloid Arg1 deletion on progression to PDA. For this purpose, we aged KF and KFCA mice to 9–16 mo (Figure 2D), an age where we have previously observed invasive cancer formation in KF mice. Accordingly, 16 out of 26 KF mice (62%) had invasive disease; in contrast, only 5 out of 24 KFCA mice had progressed (21%) (Figure 2E). We then stained the tissue for ARG1 together with F4/80, and CK19. As expected, we observed high ARG1 expression in macrophages in KF tissues and little to no ARG1 expression in macrophages in KFCA pancreata (Figure 2F). Surprisingly, we observed substantial ARG1 expression in epithelial cells from the KFCA pancreata, while epithelial cells in KF mice had little to no ARG1 expression (Figure 2F and Figure 2—figure supplement 2A). So, while deletion of Arg1 in myeloid cells impairs malignant progression, it unleashes a compensatory upregulation of Arg1 in epithelial cells.

14We noted that the epithelia that expressed ARG1 in KFCA mice had an elongated appearance, consistent with tuft cells, a cell type that is not present in the healthy pancreas but common in low-grade PanINs. By co-immunostaining for the tuft cell marker COX1, we confirmed that tuft cells are the exclusive source of epithelial ARG1 in these samples (Figure 2—figure supplement 2A), suggesting a tuft cell response to accumulating arginine in the microenvironment.

15To comprehensively characterize the phenotype of myeloid Arg1 deleted mice, we performed sc-RNA-seq on KF (n=1) and KFCA (n=1) pancreata dissected from 11 mo old mice. Unsupervised clustering identified abundant stromal and immune cells, and a small population of epithelial cells, in both genotypes (Figure 2G and Figure 2—figure supplement 2B and C). The percentage of total macrophages was similar between KF and KFCA pancreas (Figure 2—figure supplement 2D and E, Figure 2—figure supplement 3A and B). We then subclustered the macrophages, from which we identified five different macrophage populations based on distinct gene profiles (Figure 2H and Figure 2—figure supplement 3C). The macrophage 1 population was defined by Apoe, C1qa, and C1qc markers of TAMs in mouse and human pancreatic cancer, while the macrophage 5 population expressed Ear2 and Retnla (Figure 2H). Since macrophage APOE (Apolipoprotein E) is tumor promoting, we performed co-immunofluorescent staining for APOE, F4/80, and ECAD and detected a reduction in APOE expression (Figure 2—figure supplement 3D and E). scRNA-seq identified Apoe expression mainly in the KF macrophage 1 population with reduced expression in the KFCA population (Figure 2—figure supplement 3F).

16Intriguingly, sc-RNA-seq revealed that deletion of Arg1 in myeloid cells led to an upregulation of Arg2 expression, mainly in the macrophage 4 population expressing Chil3, Ly6c2, and Clec4e genes (Figure 2I), suggesting the existence of a compensatory mechanism. We then performed reactome pathway enrichment analysis on the macrophage sc-RNA-seq data to determine both up- and down-regulated pathways (Figure 2J and Figure 2—figure supplement 3G). Interestingly, KFCA macrophages, mainly the macrophage 3 population, exhibited upregulation of signaling pathways involved in MHC I antigen processing and cross presentation (Figure 2J and Figure 2—figure supplement 3H), suggesting an improvement in antigen-specific CD8+ T cell activation upon Arg1 deletion. Among the differentially expressed genes in MHC I-mediated antigen processing and presentation were H2-K1, Psmb8, H2-Q7, Ubb, H2-T23, Psmb9, H2-T22, Uba52, H2-Q6, Mrc1, Rbx1, Psmd8, Psmb5, and Znrf2 (Figure 2—figure supplement 3H).

17Macrophages in pancreatic cancer are highly immunosuppressive and have defective antigen presentation. As sc-RNA-seq suggested that Arg1 null macrophages have higher antigen presenting activity, we stained KF and KFCA pancreata for CD8 and observed increased CD8+ T cell infiltration in the latter (Figure 3A). We then sub-clustered T and natural killer (NK) cells in the sc-RNA-seq datasets from KF and KFCA pancreata and classified these into the following subclusters: naïve CD8+ T cells, cytotoxic CD8+ T cells, exhausted CD8+ T cells, CD4+ T cells, regulatory T cells (Treg), γδ T cells, NKT, and NK cells (Figure 3B and C). From this analysis, we observed an increase in the percentage of both cytotoxic and exhausted CD8+ T cells in the KFCA model compared with KF (Figure 3D and E). Correspondingly, expression of genes related to CD8+ T cell cytotoxicity, including Granzyme b (Gzmb), Perforin 1 (Prfn1), and Interferon gamma (Ifng) was upregulated in CD8+ T cells from the KFCA model (Figure 3F). Accordingly, co-immunofluorescent staining for CD8, GZMB, and ECAD showed an increase in GZMB-expressing CD8+ T cells in KFCA pancreata (Figure 3G and H). We also found that the expression of genes involved in T cell exhaustion such as cytotoxic T-lymphocyte associated protein 4 (Ctla4), Furin, lymphocyte activating 3 (Lag3), and programmed cell death 1 (Pdcd1) was upregulated in the KFCA model, compared with the KF (Figure 3I). Taken together, deletion of Arg1 in myeloid cells resulted in an increase in CD8+ T cell infiltration and activation, counterbalanced by an increase in exhaustion as well.

18Our genetic model revealed that deletion of Arg1 in myeloid cells reduced PanIN/PDA progression and was accompanied by an increased infiltration of CD8+ T cells. However, we also observed several compensatory mechanisms that may have blunted the effect of Arg1 loss. These included expression of Arg2 in myeloid cells and expression of Arg1 in epithelial cells. We reasoned that systemic inhibition of arginase, using a pharmacologic approach might bypass these compensatory mechanisms.

19We implanted a KPC pancreatic cancer cell line (7940B) orthotopically into the pancreas of syngeneic C57BL6/J mice. Upon tumor detection (by palpation or by ultrasound), we randomly divided the mice into two groups to receive either vehicle or Arginase inhibitor (CB-1158, Calithera Biosciences) (Figure 4A). We pharmacologically treated the mice for 10 d and then harvested the tumors 20 d post-implantation. First, we investigated the infiltration of CD8+ T cells by staining the vehicle and CB-1158 mouse tissue for CD8 by immunohistochemistry. Interestingly, we observed that systemic inhibition of Arginase by CB-1158 increased the infiltration of CD8+ T cells (Figure 4B and C), recapitulating our findings from the genetically engineered model. To identify whether there was an increase in CD8+ T cell activation, we stained the tissue samples for CD8 and the activation marker GZMB, as well as the epithelial marker ECAD. While in the control tissue, GZMB expression was rare, in the CB-1158 treatment group, GZMB was common in CD8+ T cells (Figure 4D and E).

20Data from the genetic model showed an increase in T cell activation but also an increase in exhaustion. We thus repeated the syngeneic orthotopic transplantation experiment described above, but this time we added blockade of the PD1 immune checkpoint to circumvent T cell exhaustion. Tumor bearing mice were divided into four different groups to receive: (1) vehicle + IgG control, (2) arginase inhibitor (CB-1158) + IgG, (3) vehicle + anti-PD1, and (4) CB-1158 + anti-PD1 (Figure 4F). We observed a trending decrease in tumor weight in the CB-1158 + IgG group and a significant decrease in tumor weight in the combination group of CB-1158 + anti-PD1 immune checkpoint blockade compared with our vehicle control group (Figure 4G). These findings recapitulated the decrease in tumor formation observed in the KFCA mouse model.

21We then proceeded to characterize the tumor tissue. We examined ARG1 expression in macrophages (F4/80+) and epithelial (ECAD+) cells by co-immunofluorescence staining and observed no changes in ARG1 distribution or expression among treatment groups (Figure 4—figure supplement 1A), consistent with the notion that the CB-1158 Arginase inhibitor does not affect enzyme production but suppresses its activity. Hematoxylin and eosin (H&E) staining showed large necrotic areas in both CB-1158 and CB-1158 + anti-PD1 groups (Figure 4—figure supplement 1B). To further characterize the tumor tissue, we stained for cell proliferation (Ki67) and apoptotic cell death (CC3) by immunohistochemistry staining (Figure 4—figure supplement 1C and D). The combination treatment group showed a significant decrease in cell proliferation compared to the vehicle control group (Figure 4—figure supplement 1C and E). We also observed either a significant or trending increase in CC3 upon CB-1158 + IgG, vehicle + anti-PD1, or the combination treatment (Figure 4—figure supplement 1D and F). Additionally, metabolomics analysis of these orthotopic tumors revealed an increase in arginine levels in the tumors from mice treated with CB-1158 and anti-PD1 (Figure 4H).

22Arginase breaks down arginine into urea and ornithine. Thus, it is equally conceivable that lack of arginine (Arg) or excess of ornithine (Orn) and/or urea might impair the ability of CD8+ T cells to proliferate and/or become activated. To determine how loss of Arginase increases CD8+ T cell proliferation and activation, we performed an in vitro CD8+ T cell proliferation and activation assay. We isolated naïve CD8+ T cells from spleen and lymph nodes of C57BL/6 mice and cultured them for 72 hr with complete RPMI media or with media lacking arginine (−Arg RPMI). In both media conditions, we added groups where the media was supplemented with either ornithine or urea, or with a combination of both (Figure 4I). We found that arginine deprivation results in reduced CD8+ T cell proliferation and activation (Figure 4J and K). Additionally, we found that culturing cells in RPMI media containing urea sightly enhanced CD8+ T cell proliferation and activation, compared to RPMI media control (Figure 4J and K, gray bars), while no enhancement was observed with ornithine supplementation.

23When cells were cultured in RPMI media lacking arginine, supplementation with ornithine, but not urea, partially rescued the proliferation defect caused by lack of arginine. No further rescue occurred adding urea; neither ornithine nor urea rescued the activation defect caused by lack of arginine. We thus concluded that CD8+ T cells require arginine for proliferation and activation, and that the arginine breakdown byproducts are not harmful to CD8+ T cell proliferation nor are they able to fully rescue lack of arginine.

24Together, these findings indicate that Arginase inhibition in combination with anti-PD1 reduces immune suppression and decreases pancreatic cancer tumor growth (see working model in Figure 5).

25Abundant myeloid cells in the TME are associated with poor prognosis in multiple types of cancer, including PDA. In contrast, high levels of T cells correlate with longer survival. We and others have shown that myeloid cells promote pancreatic cancer growth both directly and by inhibiting CD8+ T cell anti-tumor immunity. The specific mechanisms by which macrophages drive immune suppression, and how to best target macrophages to improve outcomes in pancreatic cancer remain poorly understood.

26Macrophages are plastic cell types, traditionally classified into pro-inflammatory ‘M1’ and anti-inflammatory ‘M2’ subtypes, based on their gene expression pattern. Recently, a wealth of evidence now supports the notion that TAMs are distinct from M1 and M2. Adding further complexity, TAMs also exhibit heterogeneous populations within an individual tumor. ARG1 is classically considered a ‘marker’ of the anti-inflammatory M2 state. In this study, we show that Arg1 is expressed in human and mouse TAMs, as well as in other myeloid populations. Furthermore, when we stratified human PDA based on ARG1 expression, we found inverse correlations between ARG1 and survival. However, prior to this work, a functional role of Arg1 in pancreatic TAMs and tumor immunity had not been evaluated.

27Arginases are enzymes that hydrolyze the amino acid L-arginine to urea and L-ornithine in the liver urea cycle. An analysis of plasma metabolites and tumor interstitial fluid in an autochthonous PDA mouse model revealed that L-arginine was drastically reduced in the tumor interstitial fluid. Additionally, ornithine levels increased in the PDA tumor interstitial fluid compared to plasma, suggesting arginase contribution to immunosuppression through depletion of arginine.

28There are two isoforms of arginase, ARG1 and ARG2, located in the cytoplasm and mitochondria, respectively. The arginase genes share a 58% sequence identity and are almost identical at the catalytic site. Arg1 is normally expressed by hepatocytes, but it is also expressed in TAMs in a variety of tumor types. Arg2 is located in various cell types, including renal cells, neurons, and macrophages. In a mouse model of Lewis lung carcinoma, myeloid cells express high levels of Arg1, resulting in impaired T cell function. In that study, 3LL tumor cells were implanted subcutaneously in the right flank of the mice, and at the same time, mice were treated with the pan-arginase inhibitor tool compound N-hydroxy-nor-L-Arg. Arginase inhibition reduced subcutaneous tumor growth in a dose-dependent manner. The anti-tumor effect of the arginase inhibitor was in part dependent on T cell function, as inhibition on tumor growth was not observed when mice lacking a functional immune system were treated with the arginase inhibitor. In another study, inhibition of arginase activity by the arginase inhibitor tool compound Cpd9 decreased T cell suppression and reduced tumor growth in a KrasG12D lung cancer mouse model. In a more recent study using a mouse model of neuroblastoma, myeloid Arg1 expression promoted tumor growth.

29To address the role of macrophage ARG1 in PDA, we used a dual-recombinase genetic approach to delete Arg1 in myeloid cells and aged the mice until most animals in the control group developed invasive PDA. We discovered that deletion of Arg1 in myeloid cells reduced progression to invasive disease, conversely resulting in accumulation of early lesions with prominent Tuft cells, a cell type that is protective toward tumor progression. Reduced tumor progression was accompanied by changes in the immune microenvironment, such as an increase in infiltration and activation of CD8+ T cells. However, a fraction of the mice developed invasive disease. Complete CD8+ T cell reactivation against tumors did not occur as we also observed an increase in exhausted CD8+ T cells. Additionally, from sc-RNA-seq, we observed increased Arg2 expression in myeloid cells. This finding is in line with a report in which patients with ARG1 deficiency had an increase in ARG2 levels, suggesting compensation of ARG2 for the loss of ARG1.

30Further analysis revealed that, in the absence of myeloid Arg1, the chemosensory tuft cells began to express ARG1. Tuft cells sense their surrounding environment and respond in a variety of ways, including modifying immune responses in a way that restrains PDA malignant progression. Their compensation for the loss of myeloid Arg1 expression in what is likely now an arginine-rich microenvironment is consistent with these functions and suggests that systemic inhibition of ARG1 may have a more profound effect than its ablation from myeloid cells alone.

31To systemically inhibit ARG1, we used CB-1158 (INCB001158), an orally bio-available small molecule inhibitor of arginase. CB-1158 is not cell permeable and thus does not inhibit liver ARG1, which would lead to immediate toxicity. In vitro, CB-1158 inhibits human recombinant ARG1 and ARG2. The in vivo action is attributed to inhibition of extracellular ARG1 released by TAMs and other myeloid cells. We used a syngeneic, orthotopic transplantation model based on KPC pancreatic cancer cells transplanted in C57Bl6/J mice. Treatment with CB-1158 led to an increase in infiltrating CD8+ T cells, recapitulating the findings in the spontaneous model. Immune checkpoint therapies such as PD-1/PD-L1 used to enhance the anti-tumor immune response of T cells are not effective in pancreatic cancer, in part due to resistance mechanisms induced by myeloid cells. We found that the combination treatment of CB-1158 with anti-PD1 immune checkpoint blockade reactivated exhausted CD8+ T cells and decreased tumor growth. CB-1158 is currently evaluated as a single agent and in combination with immune checkpoint therapy or chemotherapy in patients with solid tumors (https://www.clinicaltrials.gov/, NCT02903914 and NCT03314935). An open question is whether the inhibitor also has a direct effect on epithelial cells. In a model of obesity-driven pancreatic cancer, Arg2 is upregulated in epithelial cells, and its knockdown reduces tumor growth in a cell autonomous manner, independently from immune responses. While we did not observe Arg2 expression in epithelial cells in non-obese KF mice, analysis of human data revealed low, but detectable expression of Arg2, possibly supporting the possibility that, in human patients, CB-1158 might work through systemic inhibition of both arginase isoforms in multiple cell compartments, including epithelial cells.

32Because of the heterogeneity of myeloid cells acquired by their interaction with the TME, myeloid cells most likely utilize multiple mechanisms that affect their function. Our findings show that one of the mechanisms by which myeloid cells promote immune suppression and tumor growth in pancreatic cancer is through overexpression and activity of Arg1. Thus, arginase inhibition may be an effective therapeutic strategy to enhance anti-tumor immune responses.

33Arg1f/f mice (Stock # 008817) and C57BL/6 J WT mice (Stock# 000664) were obtained from the Jackson Laboratory and bred in-house. Arg1f/f mice were generated to have loxP sites flanking exons 7 and 8 in the Arg1 gene. Lyz2Cre/+ mice and KF (KrasFrt-STOP-Frt-G12D/+;Ptf1aFlpO/+) mice were donated by Dr. Howard Crawford. Lyz2Cre/+ mice express Cre in myeloid cells due to the insertion of the Cre cDNA into the endogenous M lysozyme (LysM) locus. Lyz2Cre/+;Arg1f/f mice were generated by crossing Lyz2Cre/+ mice with Arg1f/f mice. KFCA mice were generated by crossing KF mice with Lyz2Cre/+;Arg1f/f mice. Age matched male and female mice were used in these studies.

3450 μl of 50,000 7940B cells (C57BL/6 J) resuspended in a 1:1 ratio of RPMI medium 1640 (Gibco, 11875093) and Matrigel matrix basement membrane (Corning, 354234) were injected into the pancreas of C57BL/6 J mice. For surgery procedure, see. 7940B cells were derived from a male KPC (KrasLSL-G12D/+;p53LSL-R172H/+;Ptf1aCre/+) mouse tumor. Cells were tested for mycoplasma by MycoAlertTM Plus Mycoplasma Detection Kit (Lonza). Once mice had established tumors, around day 10 after implantation, mice were randomized into different treatment groups.

35Arginase inhibitor, CB-1158 was synthesized and provided by Calithera Biosciences, Inc, South San Francisco, CA (2017). For the mice studies, CB-1158 was dissolved in Milli-Q water and administered by oral gavage twice a day at 100 mg/kg. This treatment started 10 d after tumor implantation and lasted for 10 d. The control group received Milli-Q water by oral gavage, twice a day. Purified anti-mouse PD1 antibody (BioXcell #BE0033-2, clone J43) was used for the in-vivo anti-PD1 blockade experiments. Anti-PD1 was used at a dose of 200 μg/i.p. injection, every 3 d. The control group received Polyclonal Armenian hamster IgG (BioXcell, BE0091), and it was administered in parallel to anti-PD1.

36Human sc-RNA-seq data were previously published in (NIH dbGaP database accession #phs002071.v1.p1). Healthy mouse pancreas sc-RNA-seq data were previously published in (NIH dbGap database accession #GSM5011581), and mouse spontaneous PDA sc-RNA-seq data generated using the KPC (KrasLSL-G12D/+;p53LSL-R172H/+;Ptf1aCre/+) model was previously published (NIH dbGap databse accession GSE202651). To generate the KF and KFCA sc-RNA-seq data, pancreatic tissue was harvested from KF (n=1) and KFCA mice (n=1) at 11 mo of age. The tissue was mechanically minced, then digested with Collagenase V (Sigma C9263, 1 mg/ml in RPMI) for 30 min at 37°C with shaking. Digestions were filtered through 500 μm, 100 μm, and 40 μm mesh to obtain single cells. Dead cells were removed using the MACS Dead Cell Removal Kit (Miltenyi Biotec). Single-cell complementary DNA libraries were prepared and sequenced at the University of Michigan Advanced Genomics Core using the 10× Genomics Platform. Samples were run using 50-cycle paired-end reads on the NovaSeq 6000 (Illumina) to a depth of 100,000 reads. The raw data were processed and aligned by the University of Michigan Advanced Genomics Core. Cell Ranger count version 4.0.0 was used with default settings, with an initial expected cell count of 10,000. Downstream sc-RNA-seq analysis was performed using R version 4.0.3, R package Seurat version 4.0.2, and R package SeuratObject version 4.0.1 (RStudio Team RStudio: Integrated Development for R [RStudio, 2015]); http://www.rstudio.com/ R Core Development Team R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, 2017); https://www.R-project.org/. Data were filtered to only include cells with at least 100 genes and genes that appeared in more than 3 cells. Data were normalized using the NormalizeData function with a scale factor of 10,000 and the LogNormalize normalization method. Data were then manually filtered to exclude cells with <1000 or >60,000 transcripts and <15% mitochondrial genes. Variable genes were identified using the FindVariableFeatures function. Data were scaled and centered using linear regression of transcript counts. PCA was run with the RunPCA function using the previously defined variable genes. Cell clusters were identified via the FindNeighbors and FindClusters functions, using dimensions corresponding to approximately 90% variance as defined by PCA. UMAP clustering algorithms were performed with RunUMAP. Clusters were defined by user-defined criteria. The complete R script including figure-specific visualization methods is publicly available on GitHub (https://github.com/PascaDiMagliano-Lab/Arginase-1-is-a-key-driver-of-immune-suppression-in-pancreatic-cancer, copy archived at swh:1:rev:ddd8595b93230f76ebc219588953cd36f4941572; ).

37For Kaplan-Meier overall survival, we used the human dataset GSE71729 containing 125 primary PDA tumor samples. The samples were split into ARG1-low (n=62) and ARG1-high (n=63) groups. Survival analysis with log-ranked test was subsequently plotted in GraphPad Prism v9.

38Tissues were fixed in 10% neutral-buffered formalin overnight, dehydrated, paraffin-embedded, and sectioned into slides. H&E and Gomori’s Trichrome staining were performed according to the manufacturer’s guidelines.

39Paraffin sections were re-hydrated with two series of xylene, two series of 100% ethanol, and two series of 95% ethanol. Slides were rinsed with water to remove previous residues. CITRA Plus (BioGenex) was used for antigen retrieval and microwaved for 5 min and then 3 min. Once cool down, sections were blocked with 1% bovine serum albumin (BSA) in PBS for 30 min. Primary antibodies were used at their corresponding dilutions (Key resources table) and incubated at 4°C overnight. Biotinylated secondary antibodies were used at a 1:300 dilution and applied to sections for 45 minat room temperature (RT). Sections were then incubated for 30 min with ABC reagent from Vectastain Elite ABC Kit (Peroxidase), followed by DAB (Vector).

40Deparaffinized slides were blocked with 1% BSA in PBS for 1 hr at RT. Primary antibodies (Key resources table) were diluted in blocking buffer and incubated overnight at 4°C, followed by secondary antibody (Alexa Fluor secondaries, 1:300) for 45 min at RT. Slides were mounted with Prolong Diamond Antifade Mountant with DAPI (Invitrogen). TSA Plus Fluorescein (PerkinElmer) was also used in the Co-IF for primary antibodies.

41The RNA Scope Multiplex Fluorescent Detection Kit (Advanced Cell Diagnostics) was used according to the manufacturer’s protocol. The probe used for ARG1 was Hs-ARG1 (401581, Advanced Cell Diagnostics). Freshly cut human paraffin-embedded sections were baked for 1 hr at 60°C prior to staining. Slides were then deparaffinized and treated with hydrogen peroxide for 10 min at RT. Target retrieval was performed in a water steamer boiling for 15 min, and then slides were treated with the ProteasePlus Reagent (Advanced Cell Diagnostics) for 30 min. The RNA scope probe was hybridized for 2 hr at 40°C. The signal was amplified using the AMP (amplification) materials provided in the ACD Multiplex Kit (Advanced Cell Diagnostics). The signal was developed with horseradish peroxidase (HRP) channel. Once completed, the samples were washed in PBS, and then blocked for 1 hr with 5% donkey serum at RT. Primary antibody against CD45 (1:400) was incubated overnight at 4°C. Secondary antibodies (1:300 in blocking buffer) were incubated for 1 hr at RT, and samples were washed three times in PBS. Slides were counterstained with DAPI and mounted with ProLong Gold Antifade Mountant (Thermo Fisher Scientific).

42Images were taken either with an Olympus BX53 microscope, a Leica SP5 microscope, a Leica STELLARIS 8 FALCON Confocal Microscopy System, or scanned with a Pannoramic SCAN scanner (Perkin Elmer). Quantification of positive cell number or area was done using ImageJ, three to five images/slide (200× or 400× magnification) taken from three to four samples per group or using the Halo software (Indica Labs).

43The 7940B cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM, 11965–092) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin streptomycin. Tumor cell CM were collected from the 7940B cells that were cultured to confluency. Media were centrifuged at 300 g for 10 min at 4°C to remove contaminating tumor cells. These CM were used for the macrophage polarization assay. For macrophage polarization, BM cells were isolated from WT or Lyz2Cre/+;Arg1f/f mice femurs. Once isolated, BM cells were cultured with 7940B cell CM plus total DMEM media at a 1:1 ratio for 6 d. Fresh media were added during day 3. This process allowed the differentiation and polarization of BM cells to TAMs.

44Naïve CD8 T cells were isolated from C57BL/6 mouse spleens and lymph nodes by magnetic bead separation (Miltenyi Biotec, cat # 130-096-543), following the manufacturers’ protocols. T cells were activated with plate-bound anti-CD3e (2 μg/ml) and soluble anti-CD28 (1 μg/ml) in the presence of mIL-2 (10 ng/ml; R&D Systems). Cells were cultured in media constituted with RPMI powder (USBiological, R9010-01) according to the manufacturer’s instructions and supplemented with 10% FBS and 50 μM 2-Mercaptoethanol (Gibco) or in RPMI medium lacking arginine (−Arg RPMI). 1.15 mM arginine was added to the RPMI medium. Ornithine (Orn) and urea were added at 1.15 mM to the corresponding media. 2 × 105 cells in 200 μL of media were plated in a flat-bottom 96-well plate and cultured for 3 d. To assess proliferation, cells were labeled with 5 μM Cell Trace Violet in 0.1% BSA/PBS for 30 min at 37°C prior to activation.

45TAMs from WT or Lyz2Cre/+;Arg1f/f mice were lysed in RIPA buffer (Sigma-Aldrich) with protease and phosphatase inhibitors (Sigma-Aldrich). Protein samples were quantified, normalized, and then electrophoresed in a 4–15% SDS-PAGE gel (BioRad). Protein was transferred to a PVDF (polyvinylidene difluoride) membrane (BioRad), blocked with 5% milk for 1 hr at RT, and then incubated with primary antibodies overnight at 4°C (Key resources table). Membranes were then incubated with HRP-conjugated secondary antibody (1:5000) for 2 hr at RT. Membranes were washed, incubated in Western Lightning Plus-ECL (PerkinElmer), and then visualized with the ChemiDoc Imaging System (BioRad).

46CM (200 µL) was collected from each well of WT and Lyz2Cre/+;Arg1f/f TAMs in a six-well plate after 6 d of culture and used for extracellular metabolite profiling. Briefly, to the 200 µL of media, 800 µL of ice-cold 100% methanol was added. Cell lysates from parallel plates were used for protein quantification, and the protein amount was used to normalize the volume of samples collected for metabolomics. The samples were centrifuged at 12,000 g for 10 min after which the supernatant was collected, dried using SpeedVac Concentrator, reconstituted with 50% v/v methanol in water, and analyzed by targeted LC-MS/MS and processed as previously described. For metabolomics in orthotopic tumors, ~50 mg of tumor tissue was grinded into powder with mortar and pestle containing liquid nitrogen. Thereafter, the samples were suspended in 1 mL of 80% methanol in a 1.5-mL Eppendorf tube and stored overnight at –80°C. Next day, samples were centrifuged at top speed, and supernatant was collected into a new 1.5-mL Eppendorf tube, dried, and processed for LC-MS/MS as with the in vitro samples.

47GraphPad Prism 9 was used to perform statistical analysis. T-test or ANOVA was performed for group comparisons. R software 3.5.2 was used for the analysis of the microarray data set. A p-value was considered statistically significant when p<0.05.