Neutralizing Murine TGFβR2 Promotes a Differentiated Tumor Cell Phenotype and Inhibits Pancreatic Cancer Metastasis

Pancreatic cancer, the fourth leading cause of cancer-related mortality, presents a formidable challenge for treatment (1). Early metastasis, aggressive tumor biology, and a stromal rich microenvironment provide potential mechanisms for the resistance of pancreatic cancer to conventional chemotherapy. Recent work suggests that stromal cells within the tumor microenvironment are critical determinants of tumor development, progression, and metastasis (2, 3). Therefore, there is heightened interest in strategies that target stromal cells, including cancer-associated fibroblasts, immune cells, and vascular cells.

Numerous cytokines participate in the progression of pancreatic malignancies. In particular, TGFβ has a complex function in pancreatic ductal adenocarcinoma (PDA). TGFβ is known to inhibit tumor progression in early stages of PDA development yet, at later stages, TGFβ functions as a tumor promoter (4). The features that underlie the switch of TGFβ from a tumor suppressor to a tumor promoter in PDA are unclear. Mutations in the TGFβ signaling pathway occur in a large percentage (>50%) of human PDA (5) and likely contribute to the TGFβ functional switch. This has been modeled in mice where alterations of the TGFβ pathway (e.g., deletion of Smad4 or Tgfβr2) cooperated with activated Kras to promote disease progression (6–8). Furthermore, elevated expression of TGFβ, which is frequent in PDA can promote tumor development, tumor cell epithelial-to-mesenchymal transition (EMT), and tumor cell survival and motility (9–12). TGFβ also induces immunosuppression, activation of fibroblasts, angiogenesis, and collagen deposition (13, 14). Therefore, specifically targeting the protumoral aspects of TGFβ might provide therapeutic efficacy. Pharmacologic strategies that block TGFβ activity have been explored in preclinical models of pancreatic cancer (reviewed in ref. 15). As discussed by Achyut and Yang (15), targeting the TGFβ pathway alters the tumor microenvironment and the outcome of therapy might be more dependent upon microenvironmental actions than on direct tumor cell effects.

Previously, 2G8 (aka MT1), a monoclonal antibody against mouse TGFβ receptor 2 (Tgfβr2), reduced primary tumor growth and metastasis in several syngenic models of breast cancer, primarily through its effects on tumor cells and tumor-associated immune cells (16). Given the clinical importance of TGFβ dysregulation in PDA (15), we hypothesized that stromal Tgfβr2 inhibition could be effective in mouse models of PDA. Given that 2G8 is mouse-specific, we implemented this antibody in human xenograft models of PDA to specifically target stromal Tgfβr2 without interrupting TGFβ signaling in the xenografted human tumor cells. We found that inhibition of the stromal TGFβ signaling promoted epithelial differentiation in tumor cells and inhibited metastasis, results that were recapitulated in immunocompetent models of PDA. The findings in xenografts highlight that TGFβ induction of stromal cell function is critical for its impact on tumor cell behavior and PDA progression.

Further information can be found in Supplementary Materials and Methods.

Murine pancreatic cancer cell line Pan02 (Panc02) was obtained from the Developmental Therapeutics Program at the NCI. The development of primary murine PDA cell lines (mPLR) is discussed in the Supplementary Data. Human pancreatic cancer cell lines (Capan-1 and MiaPaCa-2) were obtained from ATCC. Colo357 cells were obtained from Dr. Jason Fleming (MD Anderson Cancer Center, Houston, TX). C5LM2 is a cell line derived from liver metastasis from a Panc-1 tumor bearing mouse isolated by our laboratory. RAW 264.7 (TIB-71) cells and NIH 3T3 cells were obtained from ATCC. Cell lines were confirmed to be pathogen-free and human cell lines were authenticated to confirm origin before use. Cells were cultured in DMEM (Invitrogen) or RPMI (Invitrogen) containing 10% FBS and maintained at 37°C in a humidified incubator with 5% CO₂ and 95% air.

Animals were housed in pathogen-free facility and all animal studies were performed on a protocol approved by the Institutional Animal Care and Use Committee at University of Texas Southwestern Medical Center (Dallas, TX). Animals were treated with 2G8 (provided by Imclone Systems), a rat IgG2a anti-mouse Tgfβr2 that does not bind human TGFβR2 (Supplementary Fig. S1), AFRC Mac 48 (Mac 48, a rat IgG2a specific for phytochrome; European Collection of Animal Cell Cultures), or gemcitabine (Eli Lilly and Company) purchased from the clinical pharmacy at University of Texas Southwestern Medical Center (Dallas, TX).

Pan02 cells (5 × 10⁵) were injected orthotopically in 6 to 8-week-old C57BL/6 mice. Tumor implantation was confirmed using ultrasound measurements. Ten days after tumor cell implantation, mice were randomized to receive saline, gemcitabine (25 mg/kg/wk), 2G8 (60 mg/kg/wk), or 2G8 + gemcitabine. After 4 weeks of therapy, mice were sacrificed. Gross metastatic burden was assessed at necropsy.

LSL-Kras^G12D; Cdkn2a^lox/lox; p48^Cre (KIC) mice were generated as described previously (17). Mice (28–30-day-old) were randomized to receive saline, gemcitabine (12.5 mg/kg, three times a week), 2G8 (120 mg/kg once weekly), or 2G8 + gemcitabine. After 4 weeks of therapy, mice were sacrificed and organs harvested for tissue analysis. Liver micrometastasis was assessed by hematoxylin and eosin (H&E) on the anterior lobes of the liver.

Pan02 cells (2.5 × 10⁵) were injected using splenic parenchyma of 6- to 8-week-old C57BL/6 mice. Groups were randomized to receive 2G8 (30 mg/kg/wk) 1 day before splenic injection, postinjection day 1, postinjection day 7, postinjection day 14, or Mac48 (30 mg/kg/wk). Mice were sacrificed after 5 weeks after tumor cell injection.

Eight-week-old SCID mice were injected orthotopically with 1 × 10⁶ cells (Capan-1, Colo357, MiaPaCa-2, and C5LM2). One week after tumor cell injection, mice were randomized to receive 2G8 (30 mg/kg/wk) or saline. Mice were sacrificed after 8 to 10 weeks of treatment. At the time of sacrifice, gross metastases were counted and primary tumors were snap frozen in liquid nitrogen or fixed in 10% formalin.

Formalin-fixed tissues were embedded in paraffin, sectioned, and stained with Masson Trichrome and PAS-Alcian Blue by the Molecular Pathology Core (University of Texas Southwestern Medical Center, Dallas, TX). Immunohistochemistry was performed with the following antibodies: phospho-Histone H3 (Millipore, 06–570), cleaved caspase-3 (Cell Signaling Technology, 9664), E-cadherin (Santa Cruz Biotechnology, sc-7870), vimentin (PhosphoSolution, 2105-VIM), β-catenin (Cell Signaling Technology, 9582), Zeb1 (Santa Cruz Biotechnology, sc 25388), α-smooth muscle actin (Neomarkers, RB9010-P), S100A4 (Abcam, ab27957), F4/80 (Santa Cruz Biotechnology, sc-26642), MCP-1 (Santa Cruz Biotechnology, sc1304), CD206-FITC conjugated (Biolegend, 123006), CD11b-FITC conjugated (Biolegend, 101206), Gr1-PE conjugated (Biolegend, 108408), and NK 1.1 (Wako, 986–10001). DeadEnd Fluorometric terminal deoxynucleotidyltransferase–mediated dUTP nick end labeling (TUNEL) staining was performed according to the manufacturer's instructions (Promega, G3250). Fluorescent images were captured with Photometric Coolsnap HQ camera using NIS Elements AR 2.3 Software (Nikon). Color images were obtained Nikon Eclipse E600 microscope using Nikon Digital Dx1200me camera and ACT1 software (Universal Imaging Corporation). Pictures were analyzed using NIS Elements (Nikon).

Data were analyzed using GraphPad software (GraphPad Prism version 5 for Windows). All values are expressed as mean ± SEM. For all statistical analyses, ANOVA or, where appropriate, unpaired t test was performed and results were considered significant at P < 0.05.

To explore the relationship between stromal TGFβ signaling and the phenotype of primary tumor cells in vivo, we established human xenograft models of PDA in which we specifically inhibited stromal Tgfβr2 with the mouse-specific antibody 2G8 (Supplementary Fig. S1). Mice bearing established human pancreatic tumor orthotopic xenografts (Capan-1, Colo357, MiaPaCa-2, and C5LM2) were treated systemically with 2G8. In this setting, 2G8 specifically targets stromal Tgfβr2 where it blocks ligation by TGFβ, inhibits canonical Smad signaling, and induces internalization of the receptor (16). Mice were randomized to receive control or 2G8 weekly 7 to 10 days after tumor cell injection. We found that inhibition of stromal Tgfβr2 significantly reduced surface metastases on the liver and other visceral organs (Fig. 1A), reduced cell proliferation (Fig. 1B), and elevated apoptosis (Fig. 1C–E) in all four tumor models.

To investigate whether 2G8-treated stromal cells secrete paracrine factors that affect tumor cell viability, migration, and colony formation, conditioned media (CM) from cultured mouse stromal cells (RAW 264.7 and NIH 3T3 cells) treated with 2G8 was collected and incubated with human pancreatic tumor cells. Supplementary Figure S2 demonstrates that CM from 2G8-treated stromal cells did not alter tumor cell viability; however, 2G8 blunted the ability of stromal-derived media to promote tumor cell migration toward macrophages (Fig. 2A) and 3T3 cells (Fig. 2B). Furthermore, when we evaluated anchorage-independent growth CM from 2G8-treated RAW 264.7 cells (Fig. 2C) and 3T3 cells (Fig. 2D), they showed significantly reduced colony formation compared with other conditions.

To further define the effect of Tgfβr2 inhibition on the development of liver metastases, we performed a splenic injection model with Pan02 cells. Mice were randomized to receive a control antibody (Mac48) or 2G8 the day before splenic tumor cell injection (−1 day), or 1, 7, or 14 days after injection. Liver weight at experiment completion was used as a surrogate for metastatic burden. Tgfβr2 inhibition significantly reduced metastatic burden, with livers from 2G8-treated mice appearing normal while livers from Mac48-treated mice were replaced with tumor (Fig. 2E). Unlike agents that target TGFβR1 and TGFβR2 (12), timing of the treatment did not influence the tumor burden. Treatment of mice with 2G8 the day before injection or the day after injection did not alter its ability to limit the metastatic burden in the liver. Similarly, there was no significant difference between mice treated 1 week or 2 weeks after tumor cell inoculation. These data support the functional importance of stromal Tgfβr2 in the metastasis of pancreatic cancer.

These results suggest that stromal TGFβ signaling is critical for acute tumor development and metastasis of established tumor cell lines. To determine whether the therapeutic efficacy of 2G8 extended to syngenic immunocompetent models, we explored its activity in orthotopic Pan02 tumors and LSL-Kras^G12D; Cdkn2a^lox/lox; p48^Cre (KIC) mice (18). Importantly, Pan02 and KIC cells express Tgfβr2, active TGFβ, and are sensitive to 2G8 in vitro (Supplementary Fig. S3). To test 2G8 in vivo, animals with established primary tumor burden were randomized to receive saline, gemcitabine, 2G8, or a combination of 2G8 and gemcitabine. Inhibition of Tgfβr2 alone (2G8 treatment) or in combination with gemcitabine modestly attenuated the weight of Pan02 (Fig. 3A) and KIC (Fig. 3B) tumors. However, consistent with the human xenograft results, 2G8 alone or in combination with gemcitabine significantly decreased tumor cell viability as evidenced by the changes in cell proliferation and apoptosis shown in Fig. 3C and D and Supplementary Fig. S4A–S4C.

Strikingly, 2G8, as a single agent, was very effective at reducing metastasis (3- to 5-fold; Fig. 3E and F). In fact, inhibition of TGFβ signaling was more effective than gemcitabine at reducing metastases in mice bearing Pan02 tumors (Fig. 3E) and in KIC mice (Fig. 3F). However, cotreatment with gemcitabine was not additive with 2G8. Interestingly, 2G8 and gemcitabine reduced perfusion and permeability in KIC tumors (Supplementary Fig. S5), partially explaining the lack of additivity in the combination-treated groups. The results in syngenic, immunocompetent models implicate stromal Tgfβr2 as a critical driver of PDA dissemination.

Stromal cells are key participants in the construction and remodeling of the tumor microenvironment, activities that are regulated in part by TGFβ (14, 19–21). PDA is a desmoplastic disease that consists of high levels of collagen (19, 22), which facilitates tumor cell survival and may impede the delivery of chemotherapy to tumor cells (23–25). We assessed collagen deposition by Masson trichrome staining and found that human xenografts (Fig. 4A and B) and syngenic murine tumors (Supplementary Fig. S6) from mice treated with 2G8 had significantly reduced collagen deposition. We also found a concordant and significant 2G8-mediated reduction in mature fibroblasts as evidenced by α-smooth muscle actin (Fig. 4C and D) and S100A4 (Fig. 4E) immunoreactivity in Capan-1 and MiaPaCa-2 xenografts and Pan02 tumors (Fig. 4D and Supplementary Fig. S6C). These findings implicate Tgfβr2 regulation of ECM deposition and fibroblast phenotype as critical features of the PDA microenvironment.

Metastasis is facilitated by an anti-inflammatory (M2) immune cell phenotype, which TGFβ is known to drive (4, 26–29). In support of this, we found that blocking Tgfβr2 in RAW 264.7 cells in the presence or absence of TGFβ stimulated an M1 (proinflammatory) phenotype in vitro (Supplementary Fig. S7). We also evaluated the immune status of xenografts treated with 2G8. 2G8 reduced the number of F4/80⁺ (Fig. 5A and B) and CD68⁺ (data not shown) macrophages, increased the number of macrophages positive for MCP-1 (a marker of M1 macrophages; Fig. 5C), and decreased the number of macrophages expressing MMR (a marker of M2 macrophages; Fig. 5D) in Capan-1 and MiaPaCa-2 tumors. Furthermore, 2G8 therapy decreased myeloid-derived suppressor cells (MDSC, Gr1⁺CD11b⁺ cells; Fig. 5F), while significantly increasing NK cells (NK 1.1⁺ cells; Fig. 5G) in Capan-1 and MiaPaCa-2 tumors.

We also assessed the effect of Tgfβr2 inhibition on the immune landscape in the immunocompetent models. As displayed in Supplementary Fig. S8, inhibition of Tgfβr2 with 2G8 dramatically altered the immune cell phenotype in Pan02 tumors. These changes included a reduction in total macrophage number (F4/80, Supplementary Fig. S8A), an increase in the ratio of M1:M2 macrophages (Supplementary Fig. S8B and S8C), a reduction in MDSCs (Gr1⁺CD11b⁺, Supplementary Fig. S8D) and T regulatory cells (Treg, CD4⁺FoxP3⁺, Supplementary Fig. S8E), and an increase in NK cell recruitment (NK 1.1, Supplementary Fig. S8F). These results indicate that stromal Tgfβr2 functions to promote an immunosuppressive environment while blockade of Tgfβr2 function with 2G8 stimulates recruitment and retention of immune cells that can combat the tumor.

TGFβ can drive tumor cells to adopt a mesenchymal-like phenotype that promotes tumor cell invasion and metastasis (30–34). We hypothesized that targeted inhibition of Tgfβr2 would prevent or reverse the induction of EMT in vivo. Fig. 6A displays general histology (H&E) of spontaneous KIC tumors from each treatment group at the end of therapy. Mice receiving 2G8 alone or in combination with gemcitabine showed an increase in a ductal histologic phenotype that resembles PanIN lesions. This was confirmed by staining with PAS-Alcian Blue, which revealed a significant increase in mucin-secreting cells in 2G8-treated tumors (Fig. 6B). Primary tumors and metastases in the KIC model exhibited pathologic evidence of ductal differentiation and were classified as PDAs (Supplementary Fig. S9). These features included distinct gland formation and ductal-type mucinous epithelium. Furthermore, tumors from mice treated with 2G8 displayed a significant increase in E-cadherin expression (Fig. 6C and E) and a significant decrease in vimentin expression (Fig. 6D and E).

To extend these observations, we evaluated epithelial and mesenchymal markers in the highly mesenchymal Pan02 model (35). At baseline, Pan02 cells express nuclear β-catenin and high levels of Zeb1 (35). β-catenin is expressed on the membrane of epithelial cells and translocates into the nucleus during the process of EMT. We observed that while tumors from control- and gemcitabine-treated animals expressed nuclear β-catenin, tumors from mice receiving 2G8 expressed membranous β-catenin with increased prevalence of pseudoducts (Fig. 6F and G). Furthermore, we found that ECAD was seen most prominently in 2G8-treated tumors (Fig. 6G). This indicates that TgfβR2 inhibition drives an epithelial, differentiated phenotype in Pan02 tumors in vivo.

Given the significant changes seen in stromal and primary tumor cells in vivo, we evaluated the effect of 2G8 on TGFβ-induced changes in mPLR and Pan02 cells in vitro. Cells were plated on collagen (mPLR) or plastic (Pan02) and treated with serum-free media (SFM), TGFβ, 2G8, or TGFβ + 2G8 for 24 to 72 hours and analyzed by immunocytochemistry. mPLR2D cells treated with SFM expressed ECAD and NCAD, a mesenchymal marker, at detectable levels. Treatment with 2G8 reduced NCAD and increased ECAD expression levels, whereas TGFβ had the opposite effect (Supplementary Fig. S10). Pan02 cells treated in a similar fashion had a mild decrease in zeb1 and vimentin expression after exposure to 2G8 (Supplementary Fig. S10). However, Pan02 did not express ECAD in vitro at any time point under any condition (data not shown). Overall, our in vitro studies did not fully recapitulate the dramatic phenotypic changes induced by 2G8 in vivo. This led us to hypothesize that stromal cells participate in the 2G8-driven changes in tumor cell phenotype and reduction in metastases seen in vivo.

TGFβ, collagen, fibroblasts, and immune cells all contribute to tumor cell phenotype. Tumor cells that adopt a mesenchymal phenotype are more aggressive and metastatic (36). Capan-1 and Colo357 are epithelial, whereas MiaPaCa-2 and C5LM2 cells have a mesenchymal-like phenotype in vitro (37, 38). 2G8 treatment in vivo promoted an epithelial phenotype in each xenograft analyzed (Fig. 7 and Supplementary Fig. S11). Inhibition of Tgfβr2 induced a shift in β-catenin from the nucleus to the membrane (Fig. 7A–C), increased tumor cell expression of ECAD (ECAD; Fig. 7D), and decreased Zeb1 levels (Zeb1; Fig. 7E) in Capan-1 and MiaPaCa-2 tumors. Similar results were found in Colo357 and C5LM2 models (Supplementary Fig. S11). These data indicate that activation of stromal Tgfβr2 is critical for tumor cell adoption of a mesenchymal-like phenotype.

We have shown that pharmacologic blockade of stromal Tgfβr2 can slow primary tumor growth, reduce metastasis, and promote epithelial differentiation in mouse models of pancreatic cancer. The changes in tumor cell phenotype and behavior occurred in the context of microenvironmental changes that resulted in a proinflammatory immune cell phenotype and a reduced presence of mature/activated fibroblasts. These alterations in cellular and extracellular components of the tumor after Tgfβr2 blockade resulted in striking reductions in metastatic spread and help to functionally define the importance of stromal Tgfβr2 for primary pancreatic tumor growth and metastasis.

We employed human PDA xenografts that metastasize robustly but are often criticized for not recapitulating the progression of human PDA as well as genetically engineered mouse models (GEMM). Yet, we found the same proepithelial and antimetastatic effects after Tgfβr2 inhibition in a well-established GEMM of PDA. Furthermore, we used established human pancreatic cell lines rather than direct human xenografts. Given that we specifically wanted to target mouse stromal cells, direct xenografts would not have been appropriate because they are a mixed human tumor and stromal cell population (39). Our xenograft models utilized NOD-SCID mice that lack B- and T-cell immunity; therefore, our studies do not reflect the effect of Tgfβr2 inhibition on B and T cells that can participate in antitumor effects. However, we demonstrate that inhibition of Tgfβr2 on stromal cells promotes a proinflammatory macrophage phenotype that limits tumor cell migration and colony formation in vitro and reduces metastasis in vivo. Furthermore, we observed a potent antimetastatic effect in immunocompetent models with concordant changes in immune cells.

Our analyses focused on the effect of TGFβ on macrophage and fibroblast phenotype. However, we did identify significant changes after Tgfβr2 blockade in other cell types, including NK cells. Our results indicate that 2G8 therapy increased the level of tumor-associated NK cells. These finding are consistent with the study by Zhong and colleagues (16) who demonstrated that 2G8 increased NK cell–mediated killing and secretion of IFNγ. The effect of Tgfβr2 inhibition on other cells within the tumor microenvironment such as pancreatic stellate cells, mesenchymal stem cells, or endothelial cells was not explored in depth. These cells are known to respond to TGFβ and thus inhibition of TGFβ signaling in these cell types could contribute to the reduction in tumor progression observed.

Overall, our results emphasize the impact of stromal cell function on tumor cell phenotype and metastasis. In particular, our in vitro studies suggest interruption of Tgfβr2 signaling in stromal cells alters the expression of soluble factors that impact tumor cell behavior. This is exemplified by the concordant results of our anchorage-independent growth assays in vitro where CM from mouse stromal cells enhanced colony formation in a Tgfβr2-depedent manner and the observed reduction in metastasis in vivo after blockade of stromal Tgfβr2. Whether 2G8 treatment prevents the expression of protumor factors or induces the expression of factors that inhibit tumor cell colony formation and metastasis is not clear. We evaluated the expression of twelve cytokines by macrophages (Raw 264.7) and fibroblasts (3T3) after treatment with 2G8 (Supplementary Fig. S12) and found only modest changes in a small subset of cytokines. Whether the 2G8-induced elevation in MIP1α, MIP1β, TNFα, and IFNγ expressed by macrophages are sufficient to induce the dramatic alterations in tumor cell phenotype and progression remains to be validated.

In a GEMM of PDA, we found that suppression of Tgfβr2 resulted in an increase in PanIN lesions and an increased epithelial tumor cell phenotype compared with control-treated mice. In contrast, genetic deletion of Tgfβr2 in pancreatic tumor cells (Ptf1a^Cre/+; Kras^G12D; Tgfβr2^lox/lox) resulted in elevated CTGF expression, abundant stroma, and a worse overall survival compared with Tgfβr2^wt/wt animals (7). Given that TGFβ functions as a tumor suppressor early in PDA development, it is not surprising that targeted ablation of Tgfβr2 during or before tumor development drives tumor progression. We chose to target Tgfβr2 at a time point when the mice are known to have established invasive lesions (18). Furthermore, in Ptf1a^Cre/+; Kras^G12D; Tgfβr2^lox/lox mice, Tgfβr2 is ablated in tumor cells, whereas, in our study, we targeted tumor and stromal cell Tgfβr2 within the tumor microenvironment. Overall, our studies document the critical nature of stromal TGFβ signaling to stromal cell recruitment and activity and strengthen the argument that the microenvironment is a major participant in tumor progression. The results presented here support that blockade of stromal Tgfβr2 inhibits tumor cell EMT but do not rule out that inhibition of stromal Tgfβr2 promotes tumor cell mesenchymal-to-epithelial transition (MET). Given the prevalence of poorly differentiated (e.g., mesenchymal phenotype) tumor cells in control-treated tumors in each model system employed and the demonstration that EMT occurs early in PDA (36), it is feasible that 2G8 promotes MET in PDA cell lines in vivo.

Although the focus of our work was TGFβ-dependent tumor–stromal cell interactions in PDA, several studies targeting tumor cell TGFβ pathways have observed antitumor effects. In xenograft models, inhibition of TGFβR1 and TGFβR2 resulted in decreased metastasis after splenic injection of tumor cells (12). In addition, inhibition of tumor cell TGFβR2 in subcutaneous models of human PDA resulted in reduced primary tumor growth and reduced microvessel density (40). However, neither of these studies explored stromal cell or tumor cell phenotype after TGFβ receptor inhibition.

Recently, Hezel and colleagues (41) found that direct inhibition of TGFβ or the integrin αvβ6, which is critical for activation of latent TGFβ (42), accelerated the progression of pancreatic cancer in p48^CreKras^G12Dp53^lox/+ animals. The tumors in the p48^CreKras^G12Dp53^lox/+ animals do not exhibit the pronounced EMT observed in the GEMM used in the current study. Furthermore, the study primarily evaluated survival and documented that inhibition of TGFβ or αvβ6 at early or later stages accelerated disease progression in a Smad4-dependent fashion. However, differences in stromal or tumor cell phenotype after therapy were not evaluated. In addition, the number of animals in any treatment group with distinct metastatic lesions was too few to draw meaningful conclusions regarding the effect of these strategies on tumor dissemination. These studies taken in the context of our results suggest that there is a substantial mechanistic difference between targeting TGFβ ligand and inhibiting downstream signaling by increasing Tgfβr2 internalization.

Studies targeting stromal cells in pancreatic cancer have shown that decreased stroma results in increased drug delivery and a reduction in tumor growth (24, 25). In fact, several ongoing clinical trials in pancreatic cancer aim at stromal depletion by inhibiting the hedgehog pathway (43). We found in all six of our mouse models that targeting stromal Tgfβr2 with 2G8 resulted in a significant reduction in tumor-associated collagen deposition. However, we found that blockade of Tgfβr2 significantly reduced perfusion and permeability of high and low molecular weight dextran (Supplementary Fig. S5), thus collagen deposition is not the only factor contributing to poor drug distribution in PDA.

Results in murine models of cancer document the complexity of targeting TGFβ pathways in vivo. Our data suggest that specific blockade of stromal Tgfβr2 has a profound antimetastatic effect and demonstrate that stromal cell phenotype is more critical for tumor cell phenotype and metastasis than previously appreciated. TGFβ acts in a paracrine manner in virtually every cell type and dissecting which stromal cell type is primarily responsible for the antimetastatic effect of 2G8 remains a challenge. However, our results strongly implicate Tgfβr2 on stromal cells as a critical participant in pancreatic cancer metastasis and underscore the need for an improved understanding of TGFβ biology in this challenging disease.

R.A. Brekken reports receiving a commercial research grant from Imclone/Eli Lilly & Company. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.T. Ostapoff, B.K. Cenik, R. Ye, K.D. Carroll, R.A. Brekken

Development of methodology: K.T. Ostapoff, M. Wang, L.B. Rivera, K.D. Carroll, R.A. Brekken

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.T. Ostapoff, B.K. Cenik, M. Wang, R. Ye, M. Topalovski, K.D. Carroll

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.T. Ostapoff, B.K. Cenik, M. Wang, R. Ye, D. Nugent, R.A. Brekken

Writing, review, and/or revision of the manuscript: K.T. Ostapoff, M.M. Hagopian, M. Topalovski, K.D. Carroll, R.A. Brekken

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.T. Ostapoff, X. Xu, D. Nugent, R.A. Brekken

Study supervision: K.T. Ostapoff, R.A. Brekken

The authors acknowledge the assistance of Jason E. Toombs and our collaborators at Imclone Systems. The authors thank Drs. Elisabeth Martinez, James Kim, Andries Zijlstra, Jerry Neiderkorn, and Nabeel Bardeesy for advice and Dr. Diego Castrillon for pathologic consultation.

This work is supported in part by a sponsored research agreement from Imclone Systems (R.A. Brekken), the NIH (CA118240; R.A. Brekken), the Department of Surgery, University of Texas Southwestern Medical Center (K.T. Ostapoff), and the Effie Marie Cain Scholarship for Angiogenesis Research (R.A. Brekken).

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.