Pancreatic adenocarcinoma induces bone marrow mobilization of myeloid-derived suppressor cells which promote primary tumor growth

Introduction

Tumor-induced alterations in bone marrow myelopoiesis are driven by growth factors and cytokines secreted by the primary tumor resulting in myeloid-derived suppressor cell (MDSC) expansion and mobilization [1–3]. MDSC are a heterogeneous population of immature and undifferentiated myeloid cells found in increased numbers in numerous cancers including breast, prostate, renal cell, melanoma, and lung [4–9]. The prevalence of these cells in cancer patients correlates with disease stage, and normalization in MDSC numbers occurs after successful treatment [10]. Once mobilized to the tumor stroma, MDSC promote primary tumor growth and invasion through a variety of mechanisms: arginase production, iNOS and ROS upregulation, L-selectin down regulation, cysteine depletion, promotion of neoangiogenesis, and regulatory T-cell recruitment [6, 7, 11–26]. MDSC also promote tumor-induced immunosuppression and host immune evasion by inhibiting lymphocyte activation and antigen recognition [27, 28].

The identification and classification of MDSC in human cancers remain a difficult problem due to the considerable heterogeneity of this group of myelomonocytic cells. Although MDSC in murine models of cancers reliably express Gr-1⁺CD11b⁺, there is no direct analogs cell surface marker for Gr-1 in humans. The heterogeneity of MDSC in patients with malignant disease has made characterization of these cells difficult clinically due to the lack of a singular marker. Multiple reports including several reviews have been published that established a clear link between CD15⁺ MDSC in cancer patients and Gr-1⁺ MDSC in mice [1, 8–11]. Recent studies demonstrate that induction of various subpopulation of MDSC is directly related to the specific cytokine profile of the inducing tumor [29]. Numerous subpopulations have been described in various tumors and include CD15⁺CD11b⁺, CD33⁺CD11b⁺, and CD14⁺CD11b⁺, as well as others [6, 12, 20, 30–34]. In the course of our prior studies, we observed that human pancreatic adenocarcinoma avidly recruits CD11b⁺CD15⁺ myeloid cells to the primary tumor.

Pancreatic adenocarcinoma, henceforth called pancreatic cancer, is a treatment refractory malignancy that represents the fourth leading cause of cancer death in the United States [35]. We have previously observed that pancreatic cancer creates a highly immunosuppressive tumor microenvironment though the production of inhibitory cytokines and the recruitment of immunosuppressive cells [33, 36–38], including T regulatory cells. While several reports have implicated MDSC in growth of pancreatic cancer in mouse models [39, 40], few reports have suggested that MDSC play an important role in patients with pancreatic adenocarcinoma [7].

Treatments targeting MDSC have demonstrated impaired tumor growth, increased survival, and improved response to vaccine therapies in murine models of breast cancer, glioma, and melanoma [41–45]. Zoledronic acid, a potent aminobisphosphonate and osteoclast inhibitor initially developed to treat osteoporotic bone disease, has demonstrated efficacy as an anticancer agent used to treat bony metastasis and prolong disease-free survival in patients with breast cancer [46, 47]. The anti-neoplastic mechanism of zoledronic acid remains poorly understood. Direct apoptotic effects have been reported [48, 49] as well as indirect mechanisms including inhibition of neoangiogenesis [50–52], modulation of T-cell immune response [53], and prevention of myeloid cell differentiation into MDSC [45, 54].

Few studies have targeted MDSC in pancreatic cancer [55]. In this study, we demonstrate that MDSC represent a clinically relevant immunosuppressive cell subset in human pancreatic cancer and that targeting MDSC in a murine model of pancreatic cancer using zoledronic acid reduces MDSC prevalence and tumor growth.

Materials and methods

Results

Discussion

Pancreatic cancer remains one of the most aggressive malignancies, with a mortality rate that approaches its incidence [35]. Despite many innovative treatments, including immunomodulatory therapies, pancreatic cancer remains highly resistant to treatment [57–59]. One manner in which pancreatic cancer evades therapeutic response is by the induction of an immunosuppressive microenvironment. We, and others, have previously described the role of Treg in pancreatic cancer as an efficient means of preventing adequate host antitumor responses [36, 37, 60]. Recently, myeloid-derived suppressor cells (MDSC) have been accepted as important mediators in tumor-induced immunosuppression.

MDSC are now recognized as a key immunosuppressive cell type in the development and progression of malignancy [33]. The prevalence of these cells and their immunosuppressive capacity has been reported in multiple different malignancies including head and neck, prostate, renal cell, and breast [4–9]; however, relatively little research about these cells has been carried out in pancreatic cancer. In this publication, we quantify the frequency of MDSC in human pancreatic cancer, and investigate the role of MDSC depletion in a mouse model of pancreatic adenocarcinoma as a therapeutic modality.

Early research by Schmielau and Finn suggested an increased number of “low density granulocytes” present in patients with pancreatic cancer that possessed an immunosuppressive phenotype [7]. This study focused exclusively on patients with metastatic disease and demonstrated that circulating CD15⁺ cells isolated from peripheral blood were able to block T-cell function. Although this study laid essential groundwork for the study of MDSC in human cancer patients, it did not thoroughly characterize the cells of interest. We have expanded on this research and demonstrated that the CD15⁺ cells previously observed represent CD15⁺CD11b⁺ MDSC. Other authors have demonstrated a similar increase in granulocytic MDSC in human cancers including bladder cancer, renal cell cancer, colon cancer, and lung cancer [6, 12, 20, 30]. In the current study, we demonstrate that CD15⁺CD11b⁺ cells are not only upregulated in the peripheral blood of patients with pancreatic cancer, but they are also observed in increased prevalence in the bone marrow; most importantly, these cells infiltrated tumors in large numbers. In addition, increasing prevalence of the CD11b⁺CD15⁺ cells in peripheral blood correlates with more advanced disease state. Patients with biopsy proven pancreatic adenocarcinoma have increased numbers of phenotypic MDSC when compared to normal, age-matched controls, and patients with metastatic disease had significantly increased numbers over patients with local disease (Fig. 1). In addition, these cells were found to coexpress Arginase-1, a key method of MDSC immunosuppression that has repeatedly been demonstrated throughout the literature [6, 12, 20, 30, 33, 76]. These findings are similar to other studies, which correlate MDSC prevalence with clinical disease stage in patients with advanced cancer including pancreatic cancer [10].

Through soluble mediators, the primary tumor has been implicated in the generation of MDSC in the bone marrow, mobilization into the peripheral circulation, homing to the primary tumor, and imprinting of the immunosuppressive phenotype [2, 3, 27, 29, 32, 45]. This is the first publication demonstrating increased numbers of MDSC in the bone marrow of patients with solid tumors. Numerous factors are associated with the mobilization of these cells from the bone marrow including IL-6, IL-10, GM-CSF, G-CSF, SCF, and CCL2 [2, 27, 29, 32, 42, 61–66]. The variation in MDSC phenotype between different cancer types is likely related to differential production of chemokines by the primary tumor that results in a heterogeneous myeloid cell population [29, 32, 67].

The above-mentioned heterogeneity has made the study of these cells difficult in human patient populations. We observed that human pancreatic cancer tumors avidly recruit CD15⁺CD33⁺CD11b⁺ MDSC into the peripheral circulation and to the primary tumor. In human pancreatic adenocarcinoma, these cells represent the majority of tumor-infiltrating leukocytes and greatly outnumber tumor-infiltrating lymphocytes, including Treg (Fig. 2). We recently observed that a subset of CD11b⁺ cells with a monocytic phenotype (CD14⁺CD33⁺) also infiltrate human pancreatic cancers, and ongoing efforts focus on functional and phenotypic characterization of this subset, as monocytic MDSC have been described in a variety of human cancers as well as animal models [33].

Interestingly, comparison of the circulating MDSC to those found within the tumor microenvironment demonstrates notable differences in cell surface marker phenotype. CD15 expression was variable between tumor-associated and circulating MDSC. In addition, CD33 expression on tumor-derived MDSC segregated into two populations: CD33^hi and CD33^lo. Although phenotypic analyses were unable to be carried out on the MDSC subgroups because of small quantities of human specimen, it can be postulated that the differential expression of cell surface markers was the result of imprinting on the MDSC mobilized from the bone marrow by tumor-derived factors. Recently, Rodrigues et al. [68] described that monocytes from normal donors exposed to malignant glioma cells adopted an immunosuppressive phenotype and exhibited increased CD33 expression. Importantly, it was also noted that direct cell contact was necessary to “educate” normal donor monocytes to develop maximal immunosuppressive phenotype, a finding that is supported by our data as we observed maximal suppression induced in tumor-confined MDSC. Additionally, a recent published report suggests that peripheral MDSC are skewed toward suppressive function but require activation in the tumor microenvironment to develop full suppressive potential [69], also supported by our data.

Variable “imprinting” or “education” of MDSC was observed in our Pan02 mouse model of pancreatic cancer. In tumor-bearing mice, Gr-1⁺CD11b⁺ cells were isolated from various compartments and their individual suppressive potential was measured on T-cell proliferation, with tumor-associated MDSC having the highest suppressive potential. Although all MDSC shared the same cell surface markers, their ability to suppress antigen-mediated T-cell response varied. MDSC isolated from the bone marrow were essentially inert; splenocyte-derived MDSC were immunosuppressive but by a much smaller degree than MDSC isolated from tumor. The immunosuppressive variability suggests that MDSC mobilization from the bone marrow and the development of the immunosuppressive phenotype may be isolated processes, separated both temporally and anatomically. Current cell surface markers used to identify MDSC in both animal models and humans are not sufficient given the phenotypic and functional variability of these heterogeneous cells. This is especially true in humans where cell surface markers are less specific than in murine model and the actual MDSC cell population is likely a subset of CD15⁺CD11b⁺ cells.

Given the abundance of MDSC in the primary tumor and circulation of patients with pancreatic cancer, we hypothesized that these cells play an instrumental role in the engraftment and support of the primary tumor. We selectively targeted MDSC using zoledronic acid, a potent and bone avid aminobisphosphonate that has previously been shown to target MDSC in breast and cervical cancer [44, 45, 70].

In mice inoculated with Pan02, zoledronic acid effectively decreased the rate of accumulation of MDSC in the tumor microenvironment resulting in impaired tumor growth and prolonged survival; however, the exact mechanism of impaired MDSC accumulation by zoledronic acid is unclear [44, 48–52, 70, 71]. Regardless of the mechanism, zoledronic acid effectively decreased accumulation of MDSC to the tumor and phenotypically altered the tumor microenvironment. This allowed for the host to mount a more robust T-cell response, as evidenced by a shift in the cytokine profile from immunosuppressive to immunostimulatory and increased prevalence of effector CD4 and CD8 T cells. The resulting decreased tumor growth and improved survival suggests that MDSC play an important role in the progression of pancreatic cancer.

It is possible that zoledronic acid has a direct effect on the primary tumor, as in vitro studies have shown anti-tumor activity [48, 49]. However, bioavailability studies looking at soft tissue and bone concentrations of zoledronic acid demonstrate that even with supratherapeutic daily dosing greater than used in this study (150 μg/kg/day), zoledronic acid did not accumulate at clinically appreciable levels in the pancreas or soft tissues [72]. Concentrations of zoledronic acid detected in bony tissues were up to 1,500-fold greater than the non-mineralized tissues and at least 4,000-fold higher than in plasma [72]. Therefore, it is unlikely that zoledronic acid would have direct anti-tumor effects in vivo [48, 49].

Although we observed a significant reduction in rate of tumor growth and an improvement in survival in mice treated with zoledronic acid, all animals eventually died from progressive disease. There are several possible reasons why zoledronic acid was not able to cure mice inoculated with Pan02. First, tumors promote immunosuppression and immune evasion through multiple redundant mechanisms, including Treg. Zhang et al. [73] elegantly demonstrated that although elimination of these cells is sufficient to abrogate tumor growth, it is unable to effect complete tumor eradication. This suggests that local tumor-induced immunosuppression is multifactorial, and single arm therapies are unlikely to be sufficient.

Second, Treg also play an important role in immune invasion in pancreatic cancer [36, 37, 60, 74]. Treatment with zoledronic acid resulted in a decreased percentage of CD4⁺ Foxp3⁺ cells. While the concomitant increase in tumor-associated CD4 T cells resulted in an unchanged absolute number of Treg, the increased prevalence of CD4+ Foxp3⁻ cells altered the ratio of immune effectors to immune suppressors in favor of immune effectors. However, the continued presence of Treg likely plays an important role in immune evasion in the setting of decreased MDSC.

Our above data indicate that MDSC are an important immunosuppressive cell active in the immune evasion elicited by pancreatic adenocarcinoma and that the depletion of these cells may provide a novel therapeutic modality for the treatment of this aggressive malignancy. Zoledronic acid is effective in depleting tumor-associated MDSC; however, its mechanism of action is unknown and leaves room for improvement, as all mice in our studies eventually progressed to death from disease. Combination therapy with chemotherapeutic modalities may provide improved outcomes [75]. In this vein, we are currently conducting a clinical trial to investigate the immune effects and clinical efficacy of zoledronic acid in patients with resectable pancreatic cancer (ClinicalTrials.gov identifier: NCT00892242). Information gained from such trials will be useful to dissect the complex immunologic role of MDSC and therapeutic benefit of MDSC targeting strategies. Novel therapeutic modalities targeting tumor-induced immunosuppression with immunostimulatory treatments and chemotherapy are urgently needed to overcome this aggressive malignancy.