Pancreatic cancer is characterized by an extensive and complex microenvironment, and is resistant to both chemotherapy and immune checkpoint blockade. The study by Principe and colleagues in this issue of Cancer Research proposes a combinatorial approach based on targeting the very mechanisms of resistance to gemcitabine, a commonly used chemotherapeutic agent. The authors show that gemcitabine treatment causes profound changes in the pancreatic cancer microenvironment, including elevated TGFβ signaling and immune checkpoint expression, as well as increased antigen presentation in tumor cells. Accordingly, they show that the combination of chemotherapy, TGFβ signaling inhibition, and immune checkpoint blockade effectively restores antitumor immunity and results in a significant survival benefit.
See related article by Principe et al., p. 3101
Pancreatic cancer is a deadly malignancy, and limited progress has been achieved in patient outcomes. While surgery offers the best chance at a cure, most surgical and nonsurgical patients will ultimately succumb to disease progression within a year of diagnosis. Treatment is most often palliative with FOLFIRINOX and gemcitabine/nab-paclitaxel as the first- and second-line therapies. Pancreatic cancer is characterized by local and systemic immune suppression, and no benefit has been derived from immune checkpoint blockade in this disease (for review, see ref. 1).
Tumor cells and their corresponding microenvironment have different but complementary roles in contributing to immune suppression in this disease. Fibroblasts regulate immune infiltration of suppressive cell types (2). Myeloid cells accumulate in this disease and acquire a suppressive phenotype (3). As a result, clinical trials of immune checkpoint–blocking agents have not benefited patients with pancreatic cancer (1). Therefore, efforts by multiple laboratories are seeking to devise strategies to extend the benefits of immunotherapy to pancreatic cancer, in the belief that it might provide the best chance at long-term remission in patients.
The current study by Principe and colleagues investigates the long-term effect of gemcitabine treatment in vitro on cancer cells and in vivo in a genetically engineered mouse model of pancreatic cancer expressing oncogenic Kras in the pancreas epithelium, known as KC mouse (4). Interestingly, the authors show that long-term gemcitabine treatment remodels the tumor microenvironment (TME). They observed altered expression of immune checkpoint ligands (PD-L1 and PD-L2), MHC class I, and multiple CCL and CXCL cytokines. The authors show that the effects of long-term gemcitabine treatment in KC mice include immune suppression, but also reveal potential vulnerabilities that can be exploited therapeutically. Human tumor cells resistant to five first-line FDA-approved drugs for pancreatic ductal adenocarcinoma (PDAC; gemcitabine, 5-fluorouracil, paclitaxel, irinotecan, or oxaliplatin) generally had enhanced expression of PDL-1/2, CTLA-4, and HLA-A, B, and C. Using a high-throughput assay to perform differential analysis of tumor cells upon transient treatment of gemcitabine versus long-term treatment with gemcitabine, the authors show that perturbations in expression of proinflammatory cytokines were found only with long-term treatment. Importantly, TGFβ expression was dramatically increased upon long-term gemcitabine treatment, both in tumor cells and in associated fibroblasts. It is likely that the differentially secreted cytokines from tumor cells in response to gemcitabine treatment alter the composition of the tumor stroma, enhancing TGFβ expression. Their findings of enhanced antigen presentation and immune checkpoint expression following gemcitabine treatment also suggest that chemotherapies may prime the TME for immune intervention. Furthermore, the authors observed that only under the setting of genetic or pharmacologic inhibition of TGFβ did combinatorial treatment with anti-PD-1 and gemcitabine effectively reactivate the antitumor cytotoxic T-cell response.
Previous studies had established that the cross-talk between tumor cells and the surrounding microenvironment is fundamental in conferring resistance to gemcitabine. For instance, myeloid cell–derived pyrimidines protect cancer cells from gemcitabine, likely competing for its uptake in cancer cells (5). Similarly, deoxycytidine release from fibroblasts promotes gemcitabine resistance by inhibiting the processing of gemcitabine in tumor cells, thus reducing its effectiveness (6). Macrophages, which are an important component of the microenvironment, suppress the ability of cytotoxic T cells to infiltrate the tumor and become activated against it (7). The concept of reprogramming macrophages with agents such as CD40 agonists as a component of combination therapy approaches is entering the clinic.
The mechanistic focus of the article by Principe and colleagues is on TGFβ signaling, a much studied and yet not fully understood signaling pathway in pancreatic cancer. In particular, the authors address the role of TGFβ in the context of the pancreatic cancer microenvironment. The TGFβ ligand is secreted by multiple cell types, including epithelial cells, fibroblasts, and T cells. Conversely, different cell types express the receptors. Principe and colleagues show that TGFβ biosynthesis is increased in cancer-associated fibroblasts in response to gemcitabine treatment, subsequently leading to a suppressive immune environment. The authors observed increased expression of MHC class I in tumor epithelial cells from gemcitabine treatment, which in theory should allow T cells to recognize antigens and mount an effective immune response. However, the parallel increase in TGFβ signaling may dampen that response by altering T-cell phenotypes. TGFβ blocks naïve T-cell differentiation toward a Th1 effector phenotype, while promoting conversion toward regulatory T cells and reducing antigen-presenting abilities of dendritic cells (8). Principe and colleagues also observed an increase in PD-L1 and PD-L2 expression both on epithelial cells and in the surrounding stroma, which might constitute another mechanism of T-cell inactivation, even as T cells enter the microenvironment. Mice lacking one copy of the TGFβ receptor Tgfbr1 in the context of pancreatic epithelium–specific expression of oncogenic Kras (denoted KCT mice) show an increase in MHC class I expression in the epithelium, as well as PD-L1 and PD-L2. Interestingly, while KCT animals at baseline have a similar progression as KC animals, they are sensitized to the combination of gemcitabine and anti-PD-1 treatment. Even more impressive is the treatment of the aggressive KPC model of pancreatic cancer. KPC mice express oncogenic Kras and mutant p53 in the pancreatic cancer epithelium. These animals progress rapidly to invasive disease and are largely resistant to treatment. Treatment of these animals with gemcitabine, similarly to the less aggressive KC model, resulted in upregulation of MHC class I and PD-L1 and PD-L2, but no increase in infiltrating lymphocytes. The authors devised a combination therapy approach, where KPC mice were treated with gemcitabine, PD-1 immune checkpoint blockade, and galunisertib, a TGFβ signaling inhibitor. With this combination, the authors observed an increase in infiltration and activation of CD8+ T cells, resulting in a dramatic survival benefit to an extent that is rarely observed in this model.
One of the strengths of the current study is the use of multiple, complementary model systems. The authors also employed the use of a primary patient-derived xenograft model to confirm upregulation in MHC class I and PD-L1/PD-L2 and increase in CCl/CXCL1 cytokines after long-term treatment with gemcitabine.
In addition to providing a path to translation, this article also opens new areas of research. Tgfbr1, the receptor for TGFβ, is expressed by multiple cell types within the pancreatic cancer microenvironment. As the authors use a full-body inactivation approach, the effect of TGFβ inhibition might be on tumor cells, immune cells, or on fibroblasts. The potential effect of TGFβ signaling inhibition on fibroblasts is particularly intriguing. Long considered a relatively homogeneous population, fibroblasts in pancreatic cancer have been shown to be highly heterogeneous, with different populations and potentially different functional roles. According to one classification system, fibroblasts include a smooth muscle actin (SMA)-high myCAF population and SMA-low iCAFs. Relevant to the current study, myCAF differentiation is driven by TGFβ signaling (for review, see ref. 2). Determining whether the balance across different fibroblast types was altered by genetic or pharmacologic targeting of TGFβ is beyond the scope of the current study, but might be an interesting avenue for future work, in addition to using a dual recombinase approach and cell type–specific inactivation of Tgfbr1. The effect on fibroblasts is also interesting because attempts at targeting fibroblasts, or the extracellular matrix they secrete, in clinical trials have been largely ineffective. Yet, given the role of fibroblasts in shaping the immune response, exploring other possibilities to include them in combination targeting approaches would be interesting.
The authors present a novel strategy to target both the tumor epithelium, as well as the TME through blockade of PD-1 and TGFβ in concert with gemcitabine treatment in the setting of gemcitabine resistance. This approach bears significance for unresectable patients who progress on gemcitabine therapy, as there is no standard third-line regimen. Of note, a clinical trial investigating the effects of the TGFβ signaling inhibitor, galunisertib, plus gemcitabine versus gemcitabine alone did show a modest benefit in overall survival of about 2 months (9). This benefit may be further augmented by the addition of PD-1 immune checkpoint blockade. Finally, while the authors focus mainly on gemcitabine as their cytotoxic agent of choice, the question rises whether a similar synergistic mechanism exists for FOLFIRINOX-based regimens, which are the first-line therapy of choice for patients with PDAC.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.