Abstract
Pancreatic ductal adenocarcinoma (PDAC) remains remarkably lethal with a 5-year survival rate of 8%. This is mainly attributed to the late stage of presentation, as well as widespread resistance to conventional therapy. In addition, PDAC tumors are largely nonimmunogenic, and most patients have displayed incomplete responses to cancer immunotherapies. Our group has previously identified TGFβ as a crucial repressor of antitumor immune function in PDAC, particularly with respect to cytotoxic T lymphocytes. However, pharmacologic inhibition of TGFβ signaling has had limited efficacy in clinical trials, failing to promote a significant antitumor immune response. Hence, in this work, we extend our analysis to identify and circumvent the mechanisms of resistance to TGFβ signal inhibition in PDAC. Consistent with our previous observations, adoptive transfer of TGFβ-insensitive CD8+ T cells led to the near complete regression of neoplastic disease in vivo. However, we demonstrate that this cannot be recapitulated via global reduction in TGFβ signaling, through either genetic ablation or pharmacologic inhibition of TGFBR1. In fact, tumors with TGFβ signal inhibition displayed increased PD-L1 expression and had no observable change in antitumor immunity. Using genetic models of advanced PDAC, we then determined that concomitant inhibition of both TGFβ and PD-L1 receptors led to a reduction in the neoplastic phenotype, improving survival and reducing disease-associated morbidity in vivo. Combined, these data strongly suggest that TGFβ and PD-L1 pathway inhibitors may synergize in PDAC, and this approach warrants clinical consideration.
This article is featured in Highlights of This Issue, p. 495
Introduction
TGFβ is a well-established modifier of the pancreatic tumor microenvironment (TME), though its overall contributions to carcinogenesis are poorly understood. Previous in vivo studies have found that, upon epithelial loss of TGFβ signaling, the P48-Cre/LSL-Kras (KC) mouse model of PanIN disease developed advanced pancreatic cancer with increased incidence and severity of developing adenocarcinoma from neoplastic lesions (1–3). In contrast, systemic deficiency of Tgfbr1 in the Elastase-KrasG12D model of cystic papillary neoplastic (CPN) disease (4) led to a reduction in the neoplastic phenotype (5). Combined, these data underscore the need for increased understanding of the multiple roles of TGFβ in pancreatic ductal adenocarcinoma (PDAC) pathobiology. This is especially important prior to the introduction of TGFβ inhibitors to the management of PDAC patients.
In our previous work, we determined that TGFβ represses the activation of cytotoxic T lymphocytes (CTL) in early PDAC development. Indeed, adoptive transfer of Tgfbr1 haplo-insufficient (Tgfbr1+/−) T cells into KC mice leads to a cytotoxic response toward developing lesions and a significant reduction in tumor burden (6). Clinically, TGFβ expression was inversely associated with that of GranzymeB (GrzB), a surrogate marker of cytotoxicity (6). These findings appear to support therapeutic targeting of TGFBRs in pancreatic cancers to promote T-cell–mediated cytotoxicity. In this work, we extend our analysis to determine the potential efficacy of systemic TGFβ signal inhibition, particularly with respect to antitumor immunity. We found that, although selective TGFβ inhibition in CD8+ T cells leads to regression of neoplastic disease, systemic blockade of TGFβ signaling fails to promote cytotoxicity due to compensatory upregulation of PD-L1 on the cancer epithelium. By targeting both TGFβ and PD-L1 receptors, we were able to promote the T-cell–mediated clearance of PDAC, and this approach warrants further exploration in the management of pancreatic cancer patients.
Materials and Methods
Chemicals and reagents
Galunisertib [LY2157299 (7), Supplementary Fig. S1A] was purchased from Selleck Chem, reconstituted as per the manufacturer's instructions, and used at 75 mg/kg in vivo. The anti–PD-1 antibody RMP1-14 was purchased from BioXCell and used at fixed dose of 200 μg per injection.
Flow cytometry
Mesenteric lymph nodes or spleens were isolated, ruptured, washed in PBS, and contents filtered. Two million cells were seeded into a round-bottom 96-well plate, washed in PBS, incubated with a golgi plug/protein transport inhibitor (BD biosciences), and stained with anti–CD4-APC-Cy7, CD25-PE (BD biosciences), CD69-APC, CD8-PerCP-Cy5.5 (Biolegend), and an Alive/Dead kit (Invitrogen) at 1:200 to 1:1,500 in PBS at room temperature for 20 minutes. Cells were then fixed with 1% PFA in PBS for 10 minutes at room temperature and stained with either anti-FoxP3-FITC (eBioSci) or anti–Perforin-FITC (Biolegend) at 1:150 in perm/stain buffer (BD biosciences) for 30 minutes over ice and washed 3 times with perm/wash buffer (BD biosciences). Cells were analyzed with a BD Fortessa Cytometer. All flow plots correspond to live, single cells and are representative of 100,000 events unless otherwise stated.
CD8 isolation and adoptive transfer
Cells were isolated form the spleen, and CD8+ cells isolated by a dynabead CD8-positive isolation kit were used as per the manufacturer's specification (Life Technologies). After isolation, 2 million CD8 cells were resuspended in PBS and injected into the anesthetized recipient animals via the retro-orbital vein. Recipient animals were euthanized after 1 month and tissues subject to pathologic analysis.
Mice
P48-Cre x LSL-KRASG12D (KC), Tgfbr1+/−, KC/Tgfbr1+/−, and Pdx1-Cre x LSL-KRASG12D x LSL-TP53R172H (KPC) mice were generated as described. KC mice were administered adoptive transfer of either naïve or Tgfbr1+/− CD8+ T cells and euthanized at 3 months. Similarly, 6-week KPC mice were administered intraperitoneal injection of either a PBS vehicle, Galunisertib, anti–PD-1, or Galunisertib/anti–PD-1 once every other day. This cohort was euthanized either after 2.5 months or when moribund. For euthanasia, animals were anesthetized with xylazine/ketamine until unresponsive to toe tap and/or agonal breathing. Thoracotomy served as the primary method of euthanasia and exsanguination the secondary method.
Histology, IHC, immunofluorescence, and slide scoring
Age-matched mice were euthanized, and the pancreas, colon, small bowel, liver, and spleen were subjected to pathologic examination. Tissues were fixed in 10% formalin, paraffin-embedded, and sections at 4 mm interval were cut from each tissue and stained with hematoxylin and eosin (H&E), trichrome (Sigma Aldrich), or via IHC or immunofluorescence (IF).
For IHC, slides were deparaffinized by xylenes and rehydrated by ethanol gradient, then heated in a pressure cooker using DAKO retrieval buffer. Endogenous peroxidases were quenched in 3% hydrogen peroxide in methanol for 30 minutes. Tissues were blocked with 0.5% BSA in PBS for 30 minutes and incubated with primary antibodies against pSMAD2 or PD-L1 (Cell Signaling Technology) at 1:50–1:200 overnight at 4°C. Slides were developed using either Streptavidin or secondary antibodies, followed by DAB substrate/buffer (DAKO).
For IF, slides were heated via pressure cooker in DAKO retrieval buffer and tissues blocked with 0.5% BSA in PBS for 1 hour at room temperature. Sections were exposed to primary antibodies against CK19 (University of Iowa Hybridoma Bank), PCNA, CD3 (Santa Cruz Biotechnology), or GrzB (Abcam) at 1:50–1:200 overnight at 4#C. Slides were developed using AlexaFluor 488– or 594–conjugated secondary antibodies (1:200–1:1,000, Abcam), mounted in DAPI-containing media (Santa Cruz Biotechnology), exposed to DAPI, FITC, and Texas Red filters, and images superimposed. All counts and scores were performed by two blinded investigators. For any contradicting scores, a third investigator was consulted.
Authentication of key resources
KC, Tgfbr1+/−, KC/Tgfbr1+/−, and KPC mice were bred in house and genotypes verified by PCR for all relevant transgenes (Pdx1-Cre, P48-Cre, Tgfbr1+/−, LSL-KRASG12D, LSL-TP53R172H). Galunisertib is routinely verified in our lab by the ability to neutralize TGFβ1-induced pSMAD2 phosphorylation in vitro by Western blot as seen in our previous publication (6). The effectiveness of Galunisertib in vivo is similarly validated by the absence or reduction of pSMAD2 by IHC. In addition, the anti–PD-1 antibody RMP1-14 is verified by the manufacturer for the ability to recognize purified mouse PD-1 by Western blot and has been successfully used to neutralize PD-1 signaling in murine PDAC (8).
Statistical analysis
Data were analyzed by two-way ANOVA and fit to a general linear model in Minitab16, the validity of which was tested by adherence to the normality assumption and the fitted plot of the residuals. Results were arranged by the Tukey method and were considered significant at P < 0.05. Results are presented as mean ± SEM unless otherwise noted.
Study approval
All experiments involving the use of mice were performed following protocols approved by the Institutional Animal Care and Use Committee at the University of Illinois at Chicago. Slides were obtained from patients who provided written informed consent. Tissues were deidentified and provided by the Northwestern University Pathcore following local Institutional Review Board approval.
Results
TGFBR1-deficient CTLs reduce tumor burden, but systemic reduction of TGFβ signals fails to promote cytotoxicity
To determine the clinical utility of strategies targeting the TGFβ pathway, we first generated KC mice and conducted adoptive transfer experiments as described in our previous study (6), but this time comparing them to animals with systemic ablation of TGFβ signaling akin to that seen using a TGFβ pathway inhibitor. In brief, KC mice were allowed to reach 2.5 months of age, at which time they develop significant neoplastic disease. We then administered retro-orbital injection of either 2 × 106 wild-type (WT) CD8+ T cells, or CD8+ T cells from Tgfbr1+/− mice, and euthanized the mice after 30 days. Consistent with our previous findings (6), WT CD8+ T cells failed to significantly alter lesion development, yet Tgfbr1+/− CD8+ T cells led to rapid clearance of the neoplastic phenotype and an increase in both GrzB producing T cells (Supplementary Fig. S1B and S1C), as well as GrzB deposition in the remaining neoplastic tissues (Fig. 1A and B, N = 3/group).
Although these findings indicate that TGFβ signal depletion enhances the cytotoxic immune response, they do not necessarily predict responsiveness to broad spectrum or globally administered TGFβ inhibitors. Therefore, we crossed KC mice into the Tgfbr1+/− background to generate KC/Tgfbr1+/−. Despite KC/Tgfbr1+/− mice harboring the protective Tgfbr1+/− transgene in T cells, these animals had no significant alteration in disease course, nor any observable increase in cell-mediated cytotoxicity or GrzB deposition (Fig. 1A and B, N = 3–6/group).
In addition to TGFβ, immune checkpoints such as PD-L1/PD-1 have been identified as a primary mediator of immune tolerance in the pancreas during islet cell transplantation (9). Although TGFβ signal inhibition has enhanced immune checkpoint inhibition in immunogenic cancer types (10), this combination has yet to be explored in the management of PDAC. We therefore first assessed mice for expression of PD-L1, which is often upregulated in the pancreatic TME (Supplementary Fig. S2). Interestingly, only KC/Tgfbr1+/− mice displayed an increase in PD-L1 expression on the surface of the neoplastic epithelium (Fig. 1A and B, N = 3–6/group). To determine whether pharmacologic inhibition of TGFβ signals would similarly cause the upregulation of PD-L1, we next used the Pdx1-Cre x LSL-KRASG12D x LSL-TP53R172H (KPC) model of advanced PDAC and administered the most widely used TGFβ pathway inhibitor Galunisertib (LY2157299). In accordance with our previous results, when 6-week-old KPC mice were treated with 75 mg/kg of the TGFBR1 inhibitor Galunisertib, mice also displayed increased PD-L1 expression and had no significant alteration in lesion development (Fig. 1C, N = 4/group).
Concomitant TGFBR1 and PD-1 inhibition significantly reduces tumor burden in vivo
In order to determine the physiologic importance of PD-L1 upregulation in response to Galunisertib, we generated KPC mice and after allowing PDAC to develop, administered an IP injection every other day of either a saline vehicle (N = 9), 75 mg/kg Galunisertib (N = 4), a fixed 200 μg dose of the PD-1–neutralizing antibody RMP1-14 (anti–PD-1, N = 4), or a combination of Galunisertib and anti–PD-1 (Fig. 2A, N = 7). Both Galunisertib and PD-1–neutralizing antibodies (e.g., Nivolumab or Pembrolizumab) have been used in clinical trials for PDAC (11, 12). However, as human and murine PD-1 proteins differ significantly in the extracellular domain (13), it was necessary to utilize a mouse-specific PD-1–neutralizing antibody for this study.
Although Galunisertib was highly effective at neutralizing SMAD2 phosphorylation (Fig. 2B and C), once again there was no significant effect on tumor development apart from the development of regions of cystic papillary neoplasms. Similarly, anti–PD-1 failed to significantly alter disease course as a monotherapy as previously reported (8). However, when the two agents were combined, the KPC mice displayed a significant reduction in the neoplastic phenotype despite the increased PD-L1 expression (Fig. 2A–C; Supplementary Figs. S3–S5). This was also associated with a modest survival benefit in dual-treated mice compared with untreated mice, which was not seen in single-agent controls (Fig. 2C; Supplementary Fig. S6, P = 0.0221). This was also accompanied by improvement in weight gain (Supplementary Fig. S3B), as well as reduction in both fibrosis and proliferation in remaining areas of disease (Fig. 2B and C). Further, although there was no observable change in pancreas length (Supplementary Fig. S4B), mice treated with both Galunisertib and anti–PD-1 had reductions in pancreas weight consistent with a reduction in the tumor stroma (Supplementary Fig. S4C). Despite the reduced tumor burden, there were no obvious signs of autoimmunity or drug toxicity in mice treated with the combination regimen (Supplementary Fig. S5).
Galunisertib cooperates with PD-1 inhibition to promote T-cell–mediated regression of advanced PDAC
Consistent with the observed reduction in tumor burden, on histology mice treated with the combination regimen displayed an overwhelming lymphocyte infiltrate, which localized strongly to remaining areas of disease (Fig. 3A and B). These infiltrating cells were uniformly positive for CD3, affirming their T-cell lineage. In addition, the T-cell infiltrate of Galunisertib and anti–PD-1 dual-treated mice had robust expression of GrzB, which also localized to any remaining areas of neoplastic disease (Fig. 3A and B; Supplementary Fig. S7, N = 3/group). These local changes in immune function were further verified by flow cytometry of the mesenteric lymph nodes, which revealed similar increase in T-helper activation (CD4+CD69+) cells exclusively in mice treated with the combination regimen (Fig. 3C and G; Supplementary Fig. S8). As macrophages are a primary target of Th1 cells and may have an important role in the response to immune checkpoint inhibitors (14), we also determined the presence of CD11b/GR1+ granulocytes. Although these cells were largely absent from the lymph nodes, we found a significant increase in CD11b/GR1+ cells in the spleen. Further, these CD11b/GR1+ cells were largely positive for the mouse macrophage marker F4/80 (Fig. 3D and G; Supplementary Fig. S8). Most importantly, mice treated with the combination regimen also had increased CTL activation in the mesenteric lymph nodes (CD8+CD69+ cells), as well as enhanced IFNγ, GrzB, and Perforin production (Fig. 3D–G; Supplementary Fig. S8), further affirming an increase in cell-mediated cytotoxicity.
Discussion
To date there are no curative strategies in the management of advanced pancreatic cancer patients. As PDAC often presents in late stages, most efforts focus on extending life when possible and providing supportive/palliative care. Immune checkpoint inhibition has been a major therapeutic breakthrough in many cancers. The PD-1–neutralizing antibody Pembrolizumab is now approved for non–small cell lung cancers, melanoma, some squamous cell carcinomas, and Hodgkin's lymphoma. Similarly, the anti–PD-1 antibody Nivolumab is first line in metastatic melanoma and has been approved for progressive squamous cell carcinomas of the head and neck, renal cell carcinoma, and non–small cell lung cancer (15).
As a result, investigators have been eager to determine whether similar strategies targeting PD-1/PD-L1 are applicable in the management of PDAC. Importantly, PD-L1 is often overexpressed in PDAC and is associated with poor outcomes (16). Though ineffective as a monotherapy, the combination of Nivolumab with Abraxane is showing early promise, as is the combination of Nivolumab with GVAX-based immunotherapy (17). Though encouraging, these trials appear to only modestly extend survival, and overall mortality remains high.
The mechanisms underlying the reduced efficacy of immune checkpoint inhibition in pancreatic cancer remain unclear. Therefore, it is essential to dissect alternate means through which tumor cells promote peripheral immune tolerance in order to combine with immune checkpoint inhibitors to improve their efficacy and further extend survival. As discussed and demonstrated, TGFβ is an important immunomodulator that plays a role both in physiologic and pathologic tolerance. In the absence of TGFβ signaling, T lymphocytes rapidly acquire high levels of FasL and GrzB, two main functional components for cytotoxicity. Similarly, T cells expressing a constitutively active form of TGFBR1 also remain refractory to full activation (18).
However, much like anti–PD-1 antibodies, drugs targeting the TGFβ pathway have also only shown partial efficacy in the clinic. In an ongoing phase II clinical trial, Galunisertib was added to the first-line agent Gemcitabine, which extended overall survival from 7.2 to 10.9 months, and substantially reduced serum levels of TGFβ1 (19). However, should therapies targeting the TGFβ pathway induce PD-L1 upregulation, this may substantially limit their therapeutic efficacy. Similarly, as TGFβ directly suppresses effector T-cell function independent of PD-L1, its near ubiquitous overexpression in pancreatic cancer patients may also explain the limited efficacy of PD-1/PD-L1 inhibitors in pancreatic cancer patients (Fig. 4).
The link between TGFβ signaling and immune checkpoint inhibition is now becoming clear, and their combination is showing early efficacy in other cancers types (10). For instance, in metastatic urothelial cancer patients, a poor response to the anti–PD-L1 antibody Atezolizumab was associated with a TGFβ signaling signature in fibroblasts. Similarly, the combination of TGFβ and PD-L1 inhibition greatly enhanced antitumor immune function (20). In addition, concomitant TGFβ/PD-L1 pathway inhibition greatly enhanced T-cell responses in esophageal squamous cell carcinoma (21), and blockade of greatly sensitized genetically reconstituted models of colon cancer metastasis (22). Further, TGFβ pathway inhibitors enhanced responsiveness to PD-1/PD-L1–neutralizing antibodies in mice bearing poorly immunogenic 4T1-LP breast tumors (23).
Despite this increasing body of evidence, this combination has not been fully explored in PDAC. As it is well established that TGFβ is almost ubiquitously overexpressed in the pancreatic TME and is itself a primary means for immune evasion, it is essential to further study the ability of TGFβ inhibitors to promote immune cell clearance of pancreatic tumors, as well as overcome the innate resistance of pancreatic cancers to immune checkpoint inhibitors and improve patient outcomes.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: D.R. Principe, H.G. Munshi, P.J. Grippo
Development of methodology: D.R. Principe, P.J. Grippo
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.R. Principe, A. Park, M.J. Dorman, S. Kumar, N. Viswakarma, J. Rubin, C. Torres, P.J. Grippo, A. Rana
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.R. Principe, A. Park, M.J. Dorman, S. Kumar, N. Viswakarma, H.G. Munshi, P.J. Grippo, A. Rana
Writing, review, and/or revision of the manuscript: D.R. Principe, H.G. Munshi, P.J. Grippo, A. Rana
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.R. Principe, A. Park, R. McKinney, A. Rana
Study supervision: D.R. Principe, P.J. Grippo, A. Rana
Other (partial funding of this study): P.J. Grippo
Acknowledgments
This work is dedicated to the memory of our friends Geraldine Leonardo and Rodney Hughes, both of whom recently lost their lives to pancreatic cancer. The authors also would like to thank our summer students Matt Narbutis and Shravya Chanamolu for all their hard work and dedication.
This work was supported by University of Illinois College of Medicine Hazel I. Craig Fellowships to D.R. Principe and A. Park, NIH R01CA186885 and Veterans Affairs Merit Award I01BX002922 to H.G Munshi, and by Veterans Affairs Merit Awards BX002703 and BX002355 to A. Rana. In addition, D.R. Principe is supported by NIH 1F30CA236031.
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.