Heterogeneous Stromal Signaling within the Tumor Microenvironment Controls the Metastasis of Pancreatic Cancer

One of the hallmarks of pancreatic ductal adenocarcinoma (PDAC) is the presence of dense desmoplastic stroma, which makes up approximately 60%–90% of the tumor volume (1). Emerging evidence suggests that the stroma develops concurrently with the neoplasm and likely plays a role in regulating the initiation and progression of PDAC (2, 3). Hence, understanding how the stromal signals regulate PDAC growth, invasion, and metastases could potentially lead to the development of novel targeted and more effective therapies for pancreatic cancer.

The hedgehog signaling (Hh) pathway has been shown to be a key regulator of PDAC stromal cell signaling (4, 5) and a facilitator of PDAC metastases formation (6, 7). In addition, recent reports have shown that Sonic Hedgehog (Shh) is overexpressed in PDAC tumor cells, while its downstream signaling is confined to the stromal compartment, forming a paracrine signaling axis between neoplastic and stromal cells (8, 9). However, the role of Hh signaling and stromal depletion in PDAC is debatable (10, 11). Thus, this dispute has suggested the necessity for the molecular dissection of the complexity of stromal signals in PDAC.

The recent characterization of pancreatic-associated protein, Annexin A2(AnxA2), depicted a prometastatic mechanism in PDAC (12, 13) for dissecting the stromal signaling in the tumor microenvironment. It was demonstrated that tyrosine 23 (Y-23) phosphorylation on AnxA2 is required for the cell surface translocation of AnxA2 and subsequent PDAC invasion and metastases formation (12, 13). Insulin-like growth factor 1 receptor (IGF-1R) tyrosine kinase has been implicated in the phosphorylation of AnxA2 (14). The IGF-1 ligand was postulated by others (8) to be secreted by stromal cells under the regulation of Hh signaling. More recently, it was shown that the oncogenic signaling is transmitted from tumor cells to the stromal cells through this paracrine Shh axis and subsequently receives the reciprocal signals from the stromal cells via an IGF-1R–mediated axis (9).

In addition, AnxA2 has been characterized as an in vitro and in vivo substrate of the Src family of non-receptor tyrosine kinases (15, 16). c-Met was shown to be upstream of Src and indirectly responsible for AnxA2 phosphorylation during cell scattering and branching morphogenesis (17). The role of the hepatocyte growth factor (HGF)/c-Met is also well documented in PDAC (18). In tumor cells, c-Met is often overexpressed, rather than mutated (19), which activates the pathway and results in excessive proliferation and tumor cell motility (20). HGF signals through its receptor c-Met, and is required for Src-induced AnxA2 phosphorylation in the epithelial-to-mesenchymal transition (EMT) during branching morphogenesis in early development (17).

Both IGF-1 and HGF are extracellular molecules that are present in PDAC stroma (21, 22). We therefore hypothesized that stromal signals activate intracellular effectors in PDAC such as IGF-1R and Src that culminate in Y-23-AnxA2 phosphorylation. We specifically investigated two paracrine stromal-to-epithelial pathways: the Hh/IGF-1/IGF-1R pathway and the HGF/c-Met/Src pathway.

The human Panc10.05, Panc2.8, and the murine KPCA (A) and KPCA (Y23A) pancreatic tumor cell lines and human cancer-associated fibroblasts (hCAF) were established in accordance with the Johns Hopkins Medical Institution Institutional Review Board–approved protocols, and obtained between 2012 and 2014, and authenticated by DNA and gene expression profiling and cultured as described previously (13, 23, 24). hCAFs used were at passage 2 and cultured for 3–4 additional passages if necessary. The Panc02 cells were authenticated by DNA and gene expression profiling and cultured as described previously (25). Dasatinib, tyrphostin AG490, and INC280 were obtained from Selleck Chemicals, Sigma Aldrich, and AbMole, respectively. NVP-AEW541 and NVP- LDE225 were provided by Novartis.

All animal experiments conformed to the guidelines of the Animal Care and Use Committee of the Johns Hopkins University. The animals were maintained in accordance with the guidelines of the American Association of Laboratory Animal Care. Procedures for the orthotopic model were described previously (12, 13).

Elution of AnxA2 (26) and Western blot analysis with rabbit anti-AnxA2, mouse anti-P-Y23-AnxA2, mouse anti-β-actin, rabbit anti-HGF (all from Santa Cruz Biotechnology), rabbit anti-IGF-1(Abcam), or goat anti-Shh (R&D Systems) antibodies were described previously (12). Immunohistochemical staining for E-Cadherin and HGF was performed with rabbit anti-E-Cadherin (Abcam) and rabbit anti-HGF(Santa Cruz Biotechnology) antibodies using a standard protocol on an automated stainer from Leica Microsystems. IHC for SMA was performed as described previously (27). IHC for AnxA2 was conducted with mouse anti-AnxA2(Invitrogen), or mouse anti-P-Y23-AnxA2 antibodies, IHC for Shh with goat-anti-Shh (R&D Systems), and IHC for IGF-1 with rabbit-anti-IGF-1(Abcam) antibodies as described previously (13). All IHC slides were analyzed and scored by a pathologist (A. Liu). The levels of secreted IGF-1 were measured by ELISA (R&D Systems) following the manufacturer's instructions.

Invasion assays were performed using the Trevigen invasion assay kit according to the manufacturer's instructions (Trevigen) with modifications as described previously (13). Cell invasion was measured using the CCK8 assay (Sigma).

For mass spectrometry, multiple reaction monitoring (MRM) was carried out on a platform of mass spectrometer 5500 QTRAP (AB SCIEX) online coupled with Prominence HPLC system (SHIMADZU).

The RNeasy Micro Kit (Qiagen) was used to extract total RNA. Quantitative real-time RT-PCR (qPCR) was performed on the StepOnePlus Real Time PCR System (Life Technologies) and analyzed by the StepOne software V2.1.

The lentiviral plasmid encoding shRNAs were obtained from Dharmacon and used by following a previously described procedure (13).

Statistical analysis was performed using GraphPad Prism v6.0 (GraphPad Software). The data are presented as the means ± SEM. For all analyses, P value equal or less than 0.05 was considered statistically significant.

Src and IGF-1R kinases have been implicated in AnxA2 phosphorylation in a number of benign and malignant biologic processes (14, 28–32). The role of IGF-1R and Src in AnxA2 phosphorylation, however, has not yet been established in PDAC; thus, were first examined. We used IGF-1R (NVP-AE541; ref. 33) and Src (dasatinib; ref. 34) inhibitors to examine whether IGF-1R and Src are responsible for phosphorylating AnxA2 at Y-23 in the Panc10.05 human pancreatic tumor cell line known to lack mutations in those pathways and express high levels of surface AnxA2 (12, 19). Dasatinib was shown to inhibit Src kinase activity but it was also found to be active against other tyrosine kinases (33). Nevertheless, we found that both Src and IGF-1R inhibitors decrease the amount of cell surface AnxA2 and most notably decrease the levels of phosphorylated (P)-Y23-AnxA2 in the human PDAC cell line in vitro (Fig. 1A and B; Supplementary Fig. S1A and S1B). A similar reduction in P-Y23-AnxA2 levels was observed in other PDAC cell lines including the Panc2.8 human PDAC cell line and the KPC tumor cell line derived from the Kras/p53 mutation conditional knock-in mice (KPC mice; refs. 13, 35) following the same inhibitor treatments (Supplementary Fig. S1C and S1D). In contrast, a decrease in AnxA2 phosphorylation was not observed with inhibition of other tyrosine kinases such as the EGFR kinase by Tyrphostin AG490 (Supplementary Fig. S2; ref. 36). Moreover, the treatment of PDAC cells with Src and/or IGF-1R inhibitors decreased invasion when analyzed by the Boyden chamber invasion assay (Fig. 1C). Taken together, these data support IGF-1R and Src as two independent kinases that phosphorylate AnxA2 at Y-23 in PDAC.

Next, we evaluated the upstream signals leading to IGF-1R and Src- mediated phosphorylation of AnxA2, respectively. IGF-1 was postulated (8) to be secreted by stromal cells under the regulation of Hh signaling. We therefore assessed the expression and secretion of IGF-1 in hCAF, which are the major component of PDAC stroma. Four different human fibroblasts and three different human PDAC cell lines were tested for the expression and/or secretion of IGF-1 and the data shown below are representative of them. NVP-LDE225 (37) was used to target the Hh signaling pathway and to determine the effect of Hh inhibition on IGF-1 expression and secretion. Our data show that Hh inhibition in hCAFs led to a decrease in IGF-1 expression as well as an anticipated decrease in Gli-1 expression (Fig. 1D). Moreover, inhibition of Hh pathway in hCAFs resulted in significant reduction of IGF-1 secretion from those cells. Conversely, tumor cells secrete minimal amounts of IGF-1 when compared with hCAFs and secretion of this protein from tumor cells is not affected by Hh pathway inhibition (Fig. 1E). Taken together, these results suggested that the stromal production of IGF-1 appears to be regulated by the Hh signaling.

In addition, as described above, the Src-mediated AnxA2 phosphorylation appears to be under the regulation of the HGF/c-Met pathway. It is known that HGF is also specifically secreted by the stromal fibroblast in PDAC (22) likely under autocrine regulation (38, 39). However, whether tumor cells may also play a role in regulating the stromal production of HGF remains to be explored.

As the Y-23 phosphorylation of AnxA2 was shown to be critical for PDAC invasion, we sought to determine whether IGF-1 and HGF, upstream of IGF-1R and c-Met/Src kinases, respectively, affect the invasion of tumor cells. We utilized AnxA2 knockout KPC murine PDAC cell lines, in which the wild-type AnxA2 cDNA [designated “KPCA(A)”] or the mutated AnxA2 cDNA expressing an alanine instead of tyrosine at position 23 [designated “KPCA(Y23A)”] were reintroduced. The change of tyrosine to alanine renders the protein unable to be phosphorylated at position 23 (12, 13). As shown in Fig. 1F and G, addition of exogenous recombinant murine IGF-1 (rIGF-1) and HGF (rHGF) proteins into the culture media increases the invasive potential of murine PDAC cells expressing the wild-type AnxA2 cDNA in a dose-dependent manner. Notably, the addition of rIGF-1 or rHGF showed no significant effect on KPCA (Y23A) invasion ability. These data support the hypothesis that stromal factors, IGF-1 and HGF, are upstream of IGF-1R and c-Met/Src kinases that are responsible for Y-23 phosphorylation of AnxA2 and enhance invasion of PDAC cells likely through Y-23 phosphorylation.

We next sought to determine whether hCAFs as a source of HGF and IGF-1 could enhance the invasion of PDAC cells. Utilizing invasion assay, we showed that coculture of Panc10.05 with hCAFs significantly increases the invasive potential of the tumor cells compared with the single culture of Panc10.05 (Fig. 2A). It should be noted that hCAFs are not invasive. However, when AnxA2 was knocked down in the Panc10.05 cells by the shRNA of AnxA2 (shAnxA2; Supplementary Fig. S3A), Panc10.05 cells lost their invasive potential and their invasion was also not enhanced by hCAFs (Fig. 2A). Knockdown of AnxA2 did not affect the proliferation capability of the cells (12).

We then performed an invasion assay using coculture of Panc10.05 and hCAF cells and inhibited HGF/c-Met and/or Hh/IGF-1 signaling using pharmacologic agents under clinical development for intervening with these two signaling pathways by targeting HGF/c-Met (INC280; ref. 40) and Hh (NVP-LDE225), respectively. Hh inhibitor did not inhibit the invasion of singly cultured Panc10.05 cells. c-Met inhibitor modestly, but not significantly, inhibited the invasion of singly cultured Panc10.05 cells (Supplementary Fig. S4A), likely because c-Met is expressed on PDAC cells and can be activated by HGF in the media. In contrast, as shown in Fig. 2B, inhibition of either HGF/c-Met or Hh/IGF-1 results in a significant decrease in invasion of Panc10.05 cells cocultured with hCAFs when compared with control. Moreover, there is a strong trend toward greater decrease in invasion upon dual signaling inhibition compared with single inhibitions; although, it is not statistically significant, likely due to the fact that single inhibitions have already reduced the invasion of PDAC cells cocultured with hCAFs close to the level of single PDAC cell culture, making an additional reduction in invasion difficult to be discerned by this assay.

To further confirm that the observed decrease in invasion in Fig. 2B is specifically mediated by HGF and IGF-1, we used lentiviral-shRNA to knockdown HGF and IGF-1 in hCAFs, respectively (Supplementary Fig. S3B and S3C). A significant decrease in invasion was observed when Panc10.05 cells were cocultured with hCAFs following HGF knockdown (Fig. 2C) or IGF-1 knockdown (Fig. 2E), but not with control shRNA (shCtrl)-transfected hCAFs. Furthermore, the invasion capacity of cocultured Panc10.05 cells returned to their baseline as single Panc10.05 culture when either HGF or IGF-1 was knocked down from hCAFs (Fig. 2C and E). Moreover, HGF/c-Met inhibitor treatment had no effect on the invasion of Panc10.05 cells cocultured with hCAFs that had HGF knockdown (Fig. 2D), nor did the Hh inhibitor treatment when hCAFs had IGF-1 knockdown (Fig. 2F). Treatment of Panc10.05 cells with the above mentioned inhibitors nor knockdown of HGF or IGF-1 from hCAFs affected the proliferation of those cells (Supplementary Fig. S4B and S4C). These results confirmed that HGF acts through the c-Met pathway and Hh acts through IGF-1 to enhance invasion.

Next, we explored how the signaling mediated by IGF-1 and HGF is transmitted to the neoplastic cells in PDAC. We examined whether inhibiting the Hh/IGF-1/IGF-1R and the HGF/c-Met/Src signaling pathways via pharmacologic agents can modulate AnxA2 tyrosine phosphorylation in PDAC cells cocultured with stromal cells. After treating the Panc10.05/hCAF cocultures with NVP-LDE225 and INC280, AnxA2 was eluted from the surface of the tumor cells into the EGTA-containing buffer (26) and P-Y23-AnxA2 was evaluated by Western blot analysis. As a control, single Panc10.05 cell cultures were also treated with the inhibitors to confirm that the signals that lead to IGF-1R and c-Met/Src activation originate from the stroma. Inhibition of Hh and HGF/c-Met in the 10% FBS media led to a modest decrease in P-Y23-AnxA levels in single Panc10.05 culture (Supplementary Fig. S5A–S5D), suggesting that factors present in serum (Supplementary Fig. S6) have a minor effect in activating the pathways under study leading to AnxA2 phosphorylation. To attenuate the activation of the pathways under study by factors present in the serum, we used serum-reduced media (5% FBS), which contains decreased concentrations of HGF and IGF-1 (Supplementary Fig. S6). As expected, neither the Hh inhibitor nor the HGF/c-Met inhibitor affected the levels of total AnxA2 or P-Y23-AnxA2 in Panc10.05 cells in the singly cultured serum-reduced media (Fig. 3A and B; Supplementary Fig. S5E and S5F). It is likely the concentration of HGF and IGF-1 in serum-reduced media is lower than the effective thresholds. However, because the treatment of Hh inhibitors had to be done in the medium with 10% FBS (Fig. 1), we continued to use the same medium with 10% FBS for the remaining experiments even through there is a minor effect from the HGF and IGF-1 in the serum. In Panc10.05 cells cocultured with hCAFs in a no-contact fashion, only the HGF/c-Met inhibitor decreased the total cell surface AnxA2 on Panc10.05 cells (Supplementary Fig. S5G). Importantly, each treatment significantly decreased the amount of P-Y23-AnxA2 on the cell surface of Panc10.05 cells cocultured with hCAFs while dual inhibition of HGF/c-Met and Hh led to a significant decrease of P-Y23-AnxA2 when compared with single inhibitions (Fig. 3C and D; Supplementary Fig. S5H). When HGF/c-Met and Hh are inhibited in vitro, the degree of inhibition of P-Y23-AnxA2 on tumor cells is demonstrated more potently when the tumor cells are cocultured with hCAFs compared with those grown in single culture in 10% FBS media. Thus, the effect of HGF/c-Met and Hh inhibition on P-Y23-AnxA in tumor cells cocultured with hCAFs must have exerted primarily through hCAFs, but not the existing serum factors.

Because two independent signaling pathways that originate from the stroma (IGF-1 and HGF) play a role in AnxA2 phosphorylation, we next attempted to determine whether these two signaling pathways result in phosphorylation of AnxA2 at different sites, making it necessary for the dual stromal regulation of AnxA2. By using the mass spectrometry approach, we identified 4 phospho-tyrosine sites in addition to Y-23 (Supplementary Figs. S7 and S8). We concentrated our studies on Y-333 as this site showed a significant decrease in phosphorylation after both Src and IGF-1R inhibitor treatments (Supplementary Fig. S8E and S8F). With this mass spectrometry approach, we confirmed that the inhibition of Src and IGF-1R resulted in a significant decrease in Y-23 phosphorylation (Fig. 3E; Supplementary Fig. S8A).

We next determined whether the same stromal factors that play a role in Y-23 phosphorylation (Fig. 3C and D) would also regulate Y-333 phosphorylation by quantifying the absolute levels of phosphorylation on Y-333 and Y-23 with a targeted LC/MS-MS MRM method. We found that both stromal pathways affect the phosphorylation of Y-333 and Y-23 in a similar way as the fold changes in each of the treatment groups when normalized to vehicle control are similar (Fig. 3F; Supplementary Fig. S8G). Of note, the fold of decrease in Y-23 phosphorylation in Fig. 3F is not as significant as that seen in Fig. 3C and D, likely due to the differences in detection between Western blot analysis and mass spectrometry (41). Although the existence of multiple phosphorylation sites does not explain why two stromal signals are needed for AnxA2 phosphorylation, these data on different tyrosine sites provide further confirmation that dual stromal signaling regulates the phosphorylation of AnxA2 in PDAC and have also suggested investigating other potential mechanisms at the stroma level.

Next, we used two PDAC mouse models to evaluate the role of the Hh/IGF-1 and HGF/c-Met pathways on the regulation of AnxA2 in primary PDAC tumors in vivo. The first is KPC mice, which exhibits a multistage tumorigenesis that progresses from normal, through PanIN lesions, to invasive and metastatic PDAC (35). The second is an orthotopic implant model where tumors are grown subcutaneously from a cell line (the KPC cells) derived from KPC mice and are then implanted orthotopically into the pancreas of syngeneic mice (13). We chose to implant the subcutaneously grown tumor into the pancreas instead of injecting the tumor cell suspensions into the pancreas to avoid the spilling of tumor cells to the peritoneum. In addition, the tumors would be immediately established in the pancreas following the orthotopic implantation. First, we evaluated the effects of the Hh and HGF/c-Met inhibitor treatments on the activation of stroma of the PDAC tumors in these mice using immunohistochemical analysis to determine the expression of α-smooth muscle actin (SMA), a marker of activated pancreatic stellate (fibroblast) cells (42). As shown in both mouse models, dual targeting of Hh and HGF/c-Met resulted in a statistically significant and strong inhibition of SMA expression (Fig. 4A and B). In contrast, the Hh inhibitor only moderately suppressed the expression of SMA in KPC mice (Fig. 4B; Supplementary Fig. S9A) and had no effect on SMA in the orthotopic mouse model (Fig. 4B). As the HGF/c-Met inhibitor targets the activation of c-Met on the tumor cell surface, it showed no effect on the expression of SMA.

Next, we semiquantified the expression level for the total cellular AnxA2 and the cell surface P-Y23-AnxA2 by IHC. The relative expression of P-Y23-AnxA2 to the total cellular AnxA2 as a ratio of P-Y23-AnxA2 to AnxA2 was calculated. The results show that in both the KPC and orthotropic implant models, dual inhibition of stromal signaling leads to a statistically significant decrease in the cell surface levels of P-Y23-AnxA2 on PDAC cells (Fig. 4C and D; Supplementary Fig. S9B). These data suggest that targeting both Hh and HGF signaling is required for strong inhibition of P-Y23-AnxA2 on the PDAC tumor cell surface.

We then assessed the effect of the dual inhibition of Hh and HGF/c-Met on EMT because P-Y23-AnxA2 in PDAC was previously shown to be responsible for inducing EMT (12). We performed immunohistochemical staining for E-cadherin, an epithelial marker, and semiquantified the expression of E-cadherin on PDAC. As shown, the combination of HGF/c-Met and Hh inhibitors significantly and strongly increased the levels of E-cadherin in both mouse models as compared with the vehicle treatments (Fig. 4E and F; Supplementary Fig. S9C). In contrast, single inhibition showed a moderate increase in the level of E-cadherin in both models. These results suggest that dual inhibition of Hh and HGF/c-Met signaling has a stronger inhibitory effect on EMT in primary PDAC tumors.

We next tested whether the inhibition of the stromal signaling that leads to AnxA2 phosphorylation has an effect on primary tumor growth and metastases formation. We chose to examine this using the orthotropic implant model, which allowed us to establish baseline primary tumors almost identical in size in all mice. Tumor-implanted mice develop metastases spontaneously at about the same rate. We first determined whether Hh and HGF/c-Met inhibition could suppress primary PDAC growth. Tumors from each treatment group were examined by ultrasound at baseline and then again on the last day of treatment to assess the tumor volume. Single inhibition of Hh or HGF/c-Met signaling modestly decreased primary tumor volume (Fig. 5A). Importantly, only dual inhibition of Hh and HGF/c-Met signaling significantly decreased primary tumor volume (Fig. 5A). We previously reported that AnxA2 does not affect primary PDAC growth but only metastasis formation (12, 13); therefore, the effect of dual stromal signaling inhibition may be mediated by other mechanisms, which remain to be explored.

We next tested whether inhibition of the stromal signaling that leads to AnxA2 phosphorylation has an effect on metastases formation in vivo. No obvious toxicity was observed during the entire course of treatment with either single or dual inhibitors. No weight difference was observed among different treatment groups. Mice began dying or became morbid on day 17 following orthotopic pancreatic tumor implantation; therefore, all mice were euthanized and their pancreata, livers, lungs, peritoneum, and gut were harvested to examine metastases formation grossly and histologically by the hematoxylin and eosin staining. In addition, AnxA2 IHC staining was utilized to aid the detection of micrometastases. The results showed that single inhibition of Hh or HGF/c-Met shows a trend toward decrease in the numbers of metastases formed in comparison with the vehicle treatment (Fig. 5B). More importantly, our results showed that dual blockade of stromal signals resulted in statistically significant reduction of the total numbers of metastases formed compared with vehicle, suggesting that both Hh signaling and HGF/c-Met signaling in the stroma contribute to the development of PDAC metastases (Fig. 5B–E).

Next, we evaluated the biological significance of having dual stromal signaling in regulating the same intratumoral invasion pathway. First, we analyzed the spatial expression of the activating ligands under study, HGF, Shh, and IGF-1 by IHC on primary tumors from KPC and orthotopic PDAC mice (Fig. 6) and investigated the intertumoral and intratumoral expression of these proteins in tumor tissues using sequential tissue slices. When the expression of either protein, Shh or HGF (Fig. 6A and B), is generally high in a tumor from one mouse, the expression of the other protein tends to be low in the tumor from the same mouse, suggesting that there is an intertumoral heterogeneity between different tumors regardless of whether they develop spontaneously as in the KPC mice or are orthotopically implanted. These data also suggest that stromal cells provided by the host mice may contribute to the heterogeneity of tumors. In addition, we determined the correlation in expression of Shh and IGF-1 in the primary tumors. In summary, immunohistochemical analysis revealed a significant positive correlation between Shh and IGF-1 and a negative correlation between Shh and HGF or between IGF-1 and HGF in the primary PDAC tumors (Table 1; Supplementary Fig. S10).

As further support, we also found that there is intratumoral heterogeneity of both Shh and HGF. Specifically, whether the tumors are considered overall high Shh/low HGF or overall high HGF/low Shh, Shh and HGF have opposing expression patterns between different areas within the same tumor (Fig. 6C). Interestingly, the heterogeneity of Shh/IGF-1 and HGF does not correlate with the overall distribution of P-Y23-AnxA2 on epithelial tumor cells, which tends to be homogenous. These findings are consistent with the aforementioned results showing that both Hh/IGF-1 and HGF/c-Met signaling regulate tyrosine phosphorylation of AnxA2. Taken together, our data suggest the need for targeting both pathways to overcome the heterogeneity in expression of the stromal signals that lead to phosphorylation of AnxA2 and its consequent activation during the course of invasion and metastases.

The role of IGF-1 and HGF in the invasion/metastases process in cancer has been suggested (43). Other intracellular pathways such as PRL-3, TRP, and RasGRF1 were reported to be regulated by both IGF-1 and HGF signaling, respectively (44–46). We chose to dissect the stromal regulation of the prometastatic protein AnxA2, because of the promising preclinical data describing the role of AnxA2 in invasion/metastases in PDAC as well as the established methodology to study this protein (12, 13, 47). It remains to be explored whether other intracellular pathways are also subjected to the dual regulation of IGF-1 and HGF signaling in PDAC.

This study demonstrates for the first time the role of phosphorylation of AnxA2 in regulating invasion and metastases through stromal signaling within the pancreatic TME. Our results point to a dual process within the stroma that mediates AnxA2 phosphorylation, thereby facilitating an EMT tumor cell phenotype, tumor cell invasion in vitro, and metastases formation in vivo (Fig. 5E). Importantly, we show that strong inhibition of invasion and metastasis requires inhibition of both signaling pathways. After we identified a second phosphorylation site on AnxA2, we demonstrated that stromal signals interact similarly with both phosphorylation sites, suggesting that, despite the complexity of the stromal signaling, tumor cells are not plastic at this level of regulation. However, we found that the presence of intertumoral and intratumoral differential spatial expression of stromal signals likely contributes to tumor heterogeneity.

Dual regulation of a protein is common in other biologic processes (48, 49). Phosphorylation regulation of proteins by two independent kinases has also been reported in nondiseased cells (49) and in cancer (50). The mechanism by which the two stromal signaling pathways in this study complement each other in vivo remains to be explored. We showed that both of these pathways regulate AnxA2 phosphorylation at different tyrosine sites. We cannot exclude that, in addition to Src and IGF-1R, other kinases may also participate in AnxA2 phosphorylation, nor can we rule out additional phosphorylation sites (51) that may be differentially regulated when compared with Y-23 and Y-333. Nevertheless, the similarity in the regulation between Y-23 and Y-333 suggested that tumor cells are more homogeneously regulated than we initially conceived. Moreover, the cellular heterogeneity of expression of stromal signals in the TME appears to better explain the need for dual stromal signaling culminating in AnxA2 phosphorylation regulation and subsequent invasion and metastases. Both IGF-1 and HGF are primarily secreted by cells of the mesenchymal origin, such as fibroblasts (52) although, we cannot exclude that other cell types present in the PDAC TME can contribute to the amount of those ligands present. Importantly, we showed that blockade of a single pathway is not effective in suppression of primary PDAC growth and metastases formation, as it most likely targets only tumor areas where the protein is expressed. Conversely, dual inhibition of these stromal signals is more effective in targeting heterogeneous tumors and larger tumor areas, which results in significant decrease in primary tumor volume and metastases.

Thus, our finding may provide an explanation for the ineffectiveness of single signal inhibition of the Hh and c-Met pathways (53). Controversial data have been reported on the role of Hh signaling and stromal depletion. Olive and colleagues and Provenzano and colleagues reported a beneficial effect of stromal targeting via Hh signaling inhibition or via enzymatic degradation of hyaluronan on the sensitivity to gemcitabine in transgenic mouse model of PDAC, respectively (54, 55). Others reported that depletion of stromal fibroblasts in PDAC via Hh-dependent or -independent mechanisms accelerates progression of cancer and decreases survival (10, 11). Our study suggests a possible mechanism for the observed discrepancy between the published findings by demonstrating the complexity and heterogeneity of the PDAC stromal cells and their signaling. Further investigation is warranted to uncover the complexity of stromal signals in PDAC, as well as determining whether inhibition of prometastatic stromal signals culminates in AnxA2 phosphorylation, resulting in a survival benefit.

E.M. Jaffee reports receiving a commercial research grant from Aduro Biotech, Roche, and Bristol-Meyer Squib. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A.A. Rucki, K. Foley, L. Zheng

Development of methodology: A.A. Rucki, K. Foley, Q. Xiao, R. Sharma

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Foley, Q. Xiao, J. Kleponis, A. Wu, G. Mo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.A. Rucki, K. Foley, P. Zhang, Q. Xiao, A. Wu, A. Liu, L. Zheng

Writing, review, and/or revision of the manuscript: A.A. Rucki, K. Foley, P. Zhang, E.M. Jaffee, L. Zheng

Study supervision: E.M. Jaffee, L. Zheng

We thank Dr. Goggins for providing us with the hCAF cells.

This work was supported in part by NIHR01 CA169702 to L. Zheng, NIHK23 CA148964-01 to L. Zheng, Viragh Foundation and the Skip Viragh Pancreatic Cancer Center at Johns Hopkins to E.M. Jaffee and L. Zheng, the NCI SPORE in Gastrointestinal CancersP50 CA062924 to E.M. Jaffee and L. Zheng, and a Lustgarten Foundation grant to L. Zheng.

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.