|Home | About | Journals | Submit | Contact Us | Français|
Chronic pancreatitis is a well-known risk factor for pancreatic ductal adenocarcinoma (PDA) development in humans, and inflammation promotes PDA initiation and progression in mouse models of the disease. However, the mechanistic link between inflammatory damage and PDA initiation is unclear. Using a Kras-driven mouse model of PDA, we establish that the inflammatory mediator Stat3 is a critical component of spontaneous and pancreatitis-accelerated PDA precursor formation and supports cell proliferation, metaplasia associated inflammation, and MMP7 expression during neoplastic development. Furthermore, we show that Stat3 signaling enforces MMP7 expression in PDA cells and that MMP7 deletion limits tumor size and metastasis in mice. Finally, we demonstrate that serum MMP7 level in human PDA patients correlated with metastatic disease and survival.
It is generally believed that pancreatic ductal adenocarcinoma (PDA) initiation commonly occurs with metaplastic conversion of normal cells to non-invasive ductal precursors such as pancreatic intraepithelial neoplasia (PanIN). Gain-of-function mutations in the Kras gene are nearly universal in human PDA (Almoguera et al., 1988) and pancreas-specific expression of mutant Kras in mice recapitulates the human PanIN-to-PDA sequence (Hingorani et al., 2003), suggesting that Kras mutation may represent an initiating event. Although both PanINs and PDA express ductal markers, mouse models have recently revealed that mutant Kras can reprogram acinar cells to enter the PanIN/PDA lineage in a process termed acinar to ductal metaplasia (ADM) (De La et al., 2008; Guerra et al., 2007; Habbe et al., 2008; Ji et al., 2009; Morris et al., 2010).
Chronic pancreatitis has been identified as risk factor for PDA development in humans (Lowenfels et al., 1993) and has been shown to significantly accelerate PanIN and PDA development in Kras-driven mouse models of PDA (Guerra et al., 2007). In addition, acute pancreatitis can progress to chronic pancreatitis in human patients, often under conditions in which environmental insults such as alcohol and cigarette-smoking are present (Lankisch et al., 2009). Notably, recent reports have shown that acute pancreatitis also markedly accelerates PanIN and PDA development in Kras-driven mouse models of PDA (Carriere et al., 2009; Morris et al., 2010).
Mouse models of acute pancreatitis have revealed that the exocrine pancreas possesses remarkable regenerative capacity after inflammatory damage. In response to acute pancreatitis induced by the cholecystokinin analogue caerulein, acini transiently de-differentiate into duct-like structures that express embryonic factors characteristic of pancreatic embryonic progenitors. In the absence of further damage, re-differentiation into functional acinar cells occurs rapidly within one week of caerulein treatment (De La and Murtaugh, 2009; Fendrich et al., 2008; Jensen et al., 2005; Morris et al., 2010; Siveke et al., 2008). In contrast, mutant Kras compromises the ability of acinar cells to regenerate following acute pancreatitis and locks damaged cells in a persistently de-differentiated, ductal state that can rapidly give rise to PanINs (De La and Murtaugh, 2009; Morris et al., 2010). Thus, while pancreatitis provides a permissive environment for Kras-driven neoplasia, the molecular mechanisms supporting this state remain largely unknown.
Signal transducer and activator of transcription 3 (Stat3) integrates signals from cytokines and growth factors into transcriptional responses in target cells and may serve as a mediator of inflammation-associated processes, such as pancreatitis-driven PanIN development. Stat3 mediates a complex spectrum of cellular responses including inflammation, cell proliferation, and apoptosis (Levy and Lee, 2002). Stat3 also supports the inflammation-associated tumorigenesis in the colon (Bollrath et al., 2009; Grivennikov et al., 2009), stomach (Ernst et al., 2008), and lung (Li et al., 2007). In the pancreas, Stat3 is aberrantly activated in human PDA, controls proliferation in PDA cell lines (Scholz et al., 2003), and is implicated in a mouse model of acinar to ductal reprogramming driven by persistent exocrine expression of the key pancreatic transcription factor Pdx1 (Miyatsuka et al., 2006). However, the role of Stat3 in Kras-dependent initiation of the PanIN/PDA lineage has not been investigated.
Matrix metalloproteinase 7 (MMP7), also known as matrilysin, is a member of a family of zinc-dependent endopeptidases. MMP7 is expressed predominantly in epithelial cells of glandular tissue, unlike other MMPs, which are often stromally derived (Wilson et al., 1997). MMP7 is overexpressed in PanINs and PDA (Crawford et al., 2002; Yamamoto et al., 2001) and correlates with decreased survival and possibly tumor size, lymph node involvement, and distant metastasis (Fukushima et al., 2001; Jones et al., 2004; Yamamoto et al., 2001). Significantly, MMP7-deficient mice are resistant to acinar apoptosis and ADM in an obstructive-pancreatitis model (Crawford et al., 2002) and MMP7 is required for in vitro ADM driven by EGF signaling (Sawey et al., 2007). Other in vitro studies suggest that MMP7 is also involved in disease progression by dictating the invasive and metastatic capacity of PDA cells (Yamamoto et al., 2001). Interestingly, MMP7 is induced through a Stat3-dependent mechanism in prostate and breast cancer cell lines (Udayakumar et al., 2002; Yuan et al., 2008), but a role for MMP7 in Kras-driven PDA or whether the Stat3-MMP7 axis is maintained in PDA progression has not been investigated.
In this study, we investigated the in vivo role of Stat3 in initiating the PanIN/PDA lineage in the setting of pancreatitis, tested for the presence of the Stat3-MMP7 signaling in the pancreas, and determined the role of MMP7 in PDA initiation and progression.
To determine if Stat3 signaling is involved in the pancreatic response to inflammatory damage, we audited pancreatic Stat3 activation levels in response to acute pancreatitis. We induced acute pancreatitis with caerulein as described previously (Morris et al., 2010). Immunohistochemistry (IHC) revealed no active, phosphorylated Stat3 (p-Stat3) in control adult pancreas. In contrast, p-Stat3 was widely observed in the damaged exocrine compartment of control mice 1 day after caerulein injection. Corresponding with acinar regeneration, p-Stat3 staining was again generally absent in the exocrine pancreas by 7 days after caerulein treatment (Figures 1A–D). Western blotting for p-Stat3 in pancreas lysates and quantitative RT-PCR (qPCR) for Socs3, a Stat3 target gene, revealed a similar pattern of transient activation (Figure 1I and 1J), suggesting that Stat3 signaling is active during pancreatic regeneration but quiescent in the undamaged pancreas.
We have previously shown that mutant Kras blocks acinar regeneration in favor of ductal reprogramming and PanIN formation after acute pancreatitis (Morris et al., 2010). Therefore, we examined Stat3 signaling during pancreatitis-induced PanIN development in mice expressing a conditional, constitutively active form of Kras (KrasG12D) (Hingorani et al., 2003) targeted to pancreatic progenitors under the control of the Ptf1a promoter (Ptf1a-Cre) (Kawaguchi et al., 2002). In 6 week-old Ptf1a-Cre; KrasG12D mice treated with PBS, p-Stat3 expression was absent in most morphologically normal acini and ducts, but was readily detectable in spontaneously formed PanINs, directly associated stromal cells, and occasionally in acini surrounding the PanINs (Figure 1E). Similar to control mice, Stat3 activation was widely observed in epithelial cells in Ptf1a-Cre; KrasG12D mice 1 day after caerulein treatment (Figures 1F, 1I, and 1K). However, in contrast to control mice, Ptf1a-Cre; KrasG12D animals displayed persistent Stat3 activation 7 and 21 days post-injection (Figures 1G–1I and 1K) in the setting of metaplastic ducts and PanIN lesions, suggesting a role for Stat3 in reprogramming normal acinar cells into PDA precursors following inflammatory damage.
One possible explanation for persistent Stat3 activation in caerulein treated Ptf1a-Cre; KrasG12D mice is enforced expression of factors known to activate the JAK/STAT cascade. Previous studies have indicated an obligate role for IL-6 in Ras transformation, and that oncogenic Ras supports Stat3 activation by inducing the secretion of IL-6 (Ancrile et al., 2007). Similar to Stat3 signaling, we found that IL-6 expression was only transiently upregulated after caerulein injection in wild-type mice, while IL-6 was persistently expressed in Ptf1a-Cre; KrasG12D animals (Figure S1). Therefore, persistent expression of IL-6 may be involved in prolonged Stat3 activation in Ptf1a-Cre; KrasG12D mice following inflammatory insults.
To establish if the Stat3 pathway is functionally relevant during PDA precursor formation, we generated mice expressing KrasG12D and conditional Stat3 alleles (Ptf1a-Cre; KrasG12D; Stat3f/f) (Takeda et al., 1998). At both birth and adulthood (6 weeks) Ptf1a-Cre; Stat3f/f pancreata appeared normal, indicating that pancreatic Stat3 signaling was dispensable for pancreas development (data not shown). We analyzed 6 week old Ptf1a-Cre; KrasG12D and Ptf1a-Cre; KrasG12D; Stat3f/f mice treated with PBS. While we detected occasional PanINs in Ptf1a-Cre; KrasG12D; Stat3f/f mice, the pancreatic area replaced by lesions staining with Alcian blue, which is a marker of intestinal mucin accumulation that characterizes human PanINs, was significantly smaller when compared to Ptf1a-Cre; KrasG12D mice (Figure 2A–2C). These data suggest that Stat3 deletion interferes with Kras-driven spontaneous PanIN formation and that Stat3 signaling plays an intrinsic role in PanIN development in the absence of inflammatory insults.
Next, we investigated the role of epithelial Stat3 signaling on Kras-driven PanIN formation in the setting of acute pancreatitis. By 7 days after caerulein injection, the exocrine compartment of Ptf1a-Cre; KrasG12D mice was nearly completely replaced by fibrosis and ductal structures (Figures 2D), as evident by rare expression of the acinar marker amylase (Figures 2H), and widespread expression of the duct marker cytokeratin 19 (CK 19) (Figures 2L). Ductal structures observed in Ptf1a-Cre; KrasG12D mice 21 days after caerulein frequently stained positive with Alcian blue (Figure 2T). In contrast, pancreata of Stat3-deficient mice treated with caerulein did not display gross fibrosis at necropsy, maintained dramatically more amylase-positive parenchyma (Figures 2E, 2I, 2G and 2K), and displayed fewer CK 19-positive duct structures (Figures 2M and 2O). The area of pancreas replaced by Alcian blue-positive lesions was also significantly reduced in Stat3-deficient mice (Figures 2T-2V). Although the frequency of PanINs was significantly different between the genotypes, PanINs observed in the both genotypes were of equivalently low grades (PanIN 1A and 1B). As expected, Stat3 activation was maintained in metaplastic ducts in Ptf1a-Cre; KrasG12D mice at both 7 and 21 days post-injection but was only observed in stromal cells surrounding PanIN lesions and areas of ductal metaplasia, but not in epithelial cells of acini, metaplastic ducts, or PanINs in Ptf1a-Cre; KrasG12D; Stat3f/f mice (Figures 2P–S). Given that chronic pancreatitis is a significant risk factor for PDA in humans, we also tested the effects of chronic pancreatitis and found similarly protective effects of Stat3 deletion (Figure S2A–R). Summarily, these data establish Stat3 as a critical component of the permissive environment provided by pancreatitis to drive Kras dependent PanIN formation.
We next used a cell culture model of acinar to ductal de-differentiation to ask whether Stat3 mediates ductal reprogramming of acinar cells in a cell autonomous manner. Briefly, acinar clusters isolated from adult mice embedded in collagen I trans-differentiate into ductal structures in response to the EGF ligand, TGF-αα (Means et al., 2005; Miyamoto et al., 2003). In support of our in vivo data, we found that Ptf1a-Cre; Stat3f/f acinar explants formed significantly fewer CK19-positive ductal structures after TGF-α treatment as compared to control acinar explants, suggesting that Stat3 contributes to acinar to ductal de-differentiation in part through cell autonomous mechanisms in vitro (Figures S2S–W).
To determine the long-term effects of Stat3 deletion on Kras-driven neoplasia, we analyzed Ptf1a-Cre; KrasG12D and Ptf1a-Cre; KrasG12D; Stat3f/f eight months after acute pancreatitis. While no PDA was found in either group, Ptf1a-Cre; KrasG12D mice appeared to have marked pancreatic exocrine insufficiency as demonstrated by steatorrhea and significant loss of body weight (Figure S2X-F’). Histology revealed that the pancreatic epithelium was replaced almost entirely by PanINs. In contrast, Stat3-deficient mice maintained body weight comparable to controls, did not have steatorrhea, and retained significantly more normal exocrine parenchyma than Ptf1a-Cre; KrasG12D animals (Figure S2X–F’). Thus, the protective effect of Stat3 deletion persists even months after acute damage. Notably, PanIN 1B was the highest grade of lesion seen in either genotype, suggesting that additional molecular events critical to PanIN and PDA progression, such as the loss of tumor suppressor genes, had not yet occurred over the course of our aging experiments.
Activated Stat3 signaling has been shown to regulate cellular proliferation. To determine whether pancreatic Stat3 deletion affects cell proliferation during Kras-driven metaplasia and PanIN development after pancreatitis, we determined the proliferative index of ductal cells, acinar cells, and PanIN lesions. We found that there were significantly more amylase+/Ki67+ acinar cells 2 days after caerulein and significantly more CK19+/Ki67+ ductal cells 2 and 7 days after caerulein in Ptf1a-Cre; Kras mice compared to Ptf1a-Cre; KrasG12D; Stat3f/f animals (Figure 3A–P). Thus, loss of Stat3 impairs proliferation during Kras-driven pancreatic metaplasia. In contrast, the proliferative capacity of the few lesions positive for Alcian blue observed in Ptf1a-Cre; KrasG12D; Stat3f/f animals was not altered when compared to controls (Figure 3Q–X). Staining for cleaved-Caspase-3 showed that there were few apoptotic cells in all three genotypes at both timepoints (data not shown), suggesting that contribution of apoptosis is relatively minimal during pancreatitis-promoted ductal reprogramming. Collectively, this data demonstrates that Stat3 supports persistent cell proliferation that characterizes Kras-dependent acinar metaplasia and PanIN formation.
Since Stat3 is a critical mediator of inflammation during neoplasia (Yu et al., 2009), we used FACS analysis to evaluate whether Stat3 deletion altered the initial inflammatory response during Kras-driven metaplasia following injury. One day after caerulein injection, the percentage of total infiltrating immune cells (CD45+ cells) was higher in Ptf1a-Cre; KrasG12D pancreata compared to tissue from control and Ptf1a-Cre; KrasG12D; Stat3f/f mice (Figure 4A). Additionally, an increase in the CD11b+Gr1+ myeloid subset (known to contain myeloid derived suppressor cells and immature myeloid cells), as well as an increased trend in infiltrating B (B220+) and T (Thy1.2+) lymphocytes was seen in Ptf1a-Cre; KrasG12D mice compared to both control and Stat3-deficient mice (Figure S3A–G). While additional analysis is necessary to refine the identity and function of these immune subsets, these data indicate that the lack of Stat3 signaling during initiation of Kras-driven metaplasia decreases total inflammatory infiltration and alters immune cell composition.
To better understand how Stat3 might contribute to metaplasia associated inflammation, we determined the expression of cytokines in pancreata from Ptf1a-Cre; KrasG12D and Ptf1a-Cre; KrasG12D; Stat3f/f mice 1 day after caerulein. QPCR for Stat3 target genes, including cytokines (IL-6, IL-1β, IL-11), the inflammatory mediator ICAM1, and the prostaglandin-producing enzyme Cox-2 (Yu et al., 2009) as well as other inflammatory mediators associated with pancreatitis (the cytokine Tnfα, and the chemokine CCL2) (Malka et al., 2000; Marra, 2005) revealed significantly higher levels in the pancreata of Ptf1a-Cre; KrasG12D mice compared to those from Ptf1a-Cre; KrasG12D; Stat3f/f mice (Figure 4B). Since immune cells present in tissue inflammation produce cytokines (Shrivastava and Bhatia, 2010), these differences in cytokine expression may reflect the distinct inflammatory composition observed in each genotype. However, acini are known to represent an epithelial source of cytokines during caerulein pancreatitis (Bhatia et al., 2002; Blinman et al., 2000; Grady et al., 1997). To determine if the differences in cytokine expression observed in bulk tissue may in part be due to epithelial Stat3 dependent function, we measured the expression of the Stat3 dependent genes IL6, IL-11, and Cox-2 in primary cultured acini treated with caerulein isolated from control and Ptf1a-Cre; Stat3f/f mice. Interestingly, we found a significant decrease in expression of these factors in caerulein treated acini lacking Stat3 compared with control (Figure 4C). Taken together, our data demonstrate that epithelial Stat3 modulates the initial immune response, as well as contributes to the cytokine milieu present during initiation of Kras-driven metaplasia.
Since Stat3 appears to be critical for Kras dependent PanIN development, we searched for possible downstream signaling factors supported by Stat3. Crawford and colleagues have previously shown that MMP7 is expressed in the exocrine compartment in response to obstructive pancreatitis induced by pancreatic ductal ligation (Crawford et al., 2002). Furthermore, in vitro studies in prostate and breast cancer cell lines have implicated MMP7 as a downstream target of Stat3 signaling (Udayakumar et al., 2002; Yuan et al., 2008). To determine if there is an in vivo connection between Stat3 and MMP7 during regeneration following acute pancreatitis or in Kras-driven metaplasia and PanIN development, we stained for MMP7 in pancreas from control, Ptf1a-Cre; KrasG12D and Ptf1a-Cre; KrasG12D; Stat3f/f mice after caerulein treatment. In the absence of damage, MMP7 was not expressed in control mice or the normal parenchyma of Ptf1a-Cre; KrasG12D mice, but was seen in spontaneously formed PanINs (data not shown). In control mice, MMP7 was expressed 2 days post-treatment but absent 7 days following injection (Figure 5A, 5D). In contrast, MMP7 expression in Ptf1a-Cre; KrasG12D mice was detectable 2 days after caerulein treatment and remained prominent 7 days post injection (Figures 5B and E). This differential MMP7 expression pattern between the genotypes mirrored Stat3 activation profiles after pancreatitis. Intriguingly, there was significantly less MMP7 staining, as well as MMP7 expression by qPCR, in Stat3-deficient pancreata compared to Ptf1a-Cre; KrasG12D mice at both time points (Figures 5C, F, G, and H). No differences were detected in the expression of other MMPs including MMP2, MMP3, MMP9, and MMP14 between Ptf1a-Cre; KrasG12D and Stat3-deficient mice (Figure S4). These data indicate that Stat3 activity regulates MMP7 expression during ductal reprogramming in vivo, suggesting a potential role for MMP7 in PanIN development following pancreatitis.
To test the hypothesis that MMP7 is a critical downstream effector of Stat3 during pancreatitis driven PanIN development, we compared PanIN formation in Pfta1-Cre; KrasG12D mice containing heterozygous or homozygous germline deletion of MMP7 (Pfta1-Cre; KrasG12D; MMP7Het and Pfta1-Cre; KrasG12D; MMP7KO , respectively). Unlike the effect of Stat3 deletion, we observed no significant difference in PanIN frequency irrespective of MMP7 expression (Figure 5I). These results were validated using another cohort of mice based on the Pdx1 promoter (data not shown). Therefore, while Stat3 supports MMP7 expression during Kras-driven neoplasia in the pancreas, MMP7 does not appear to be a critical Stat3 effector for ductal metaplasia and PanIN development.
Although MMP7 signaling does not appear to be essential to pancreatitis-induced PanIN formation, previous in vitro work using various cancer cell lines suggested that MMP7 may play a role in cancer progression, including metastasis (Adachi et al., 1999; Yamamoto et al., 2001). Therefore, we examined if the positive relationship between Stat3 and MMP7 observed during preneoplastic initiation persists in PDA cells and whether MMP7 itself plays role in PDA progression and tumor biology. First, we screened seven human PDA cell lines for MMP7 expression and found that CFPAC-1 and Colo357 expressed MMP7 at readily detectable levels (arbitrarily determined as absolute Ct ≤ 35 by qPCR; Figure S5A). We treated Colo357 and CFPAC-1 cells with compound S3I-201, which inhibits Stat3 activity by binding to the Stat3 SH2 domain and preventing homodimerization and activation of Stat3 (Lin et al., 2009; Siddiquee et al., 2007). A concentration-dependent decrease in both p-Stat3 levels and MMP7 expression was observed Colo357 cells treated with the inhibitor (Figure 6A). CFPAC-1 cells also displayed significant reduction in both p-Stat3 and MMP7 expression at all concentrations tested (Figure S5B), indicating that Stat3 signaling dynamically controls MMP7 expression in PDA cells.
Since Stat3 is active and continues to support MMP7 expression in PDA cells we investigated if MMP7 itself contributes to PDA biology. We analyzed the effect of MMP7 deletion on a number of Kras-driven mouse models of PDA. First, we observed Pdx1-CreLate; KrasG12D; MMP7KO and Pdx1-CreLate; KrasG12D; MMP7Het animals until a censor date of 13 months (expression pattern of Pdx-CreLate mice previously described in (Heiser et al., 2006)). We found that all Pdx1-CreLate; KrasG12D; MMP7Het mice developed PDA (4 of 8 MMP7Het mice succumbed to PDA before censor and the other 4 animals had PDA at the time of sacrifice) while 4 of 11 Pdx1-CreLate; KrasG12D; MMP7KO mice (36.3%) were PDA free (Figure 6B). The highest lesion grades in these animals were PanIN 1 in 1 mouse and PanIN 2 in 3 mice. These results suggest that MMP7 impacts PDA progression in vivo.
Next we performed a more extensive analysis to test at what stage MMP7 deletion impacts PDA progression in transgenic mice fated to develop PDA with 100% penetrance due to pancreatic deletion of one allele of p53 (Bardeesy et al., 2006) (Pdx1-CreLate; KrasG12D; p53f/+; MMP7Het or MMP7KO). We found that tumors identified at the time of death in Pdx1-CreLate; KrasG12D; p53f/+; MMP7Het mice were significantly larger than tumors observed in Pdx1-CreLate; KrasG12D; p53f/+; MMP7KO mice (mean diameter of 1.6 cm compared to 1.0 cm, Figures 6C–E). Furthermore, there were clear differences in disease stage between the MMP7Het and MMP7KO groups. While Pdx1-CreLate; KrasG12D; p53f/+; MMP7Het mice demonstrated the usual stigmata of end-stage PDA in the form of ascites, regional disease, and distant metastasis in a majority of cases, metastasis was dramatically reduced in Pdx1-CreLate; KrasG12D; p53f/+; MMP7KO mice (Figures 6D, 6E, S5H, S5I, and Table 1). No MMP7KO mice had obvious metastasis to lymph nodes while 58% of MMP7Het mice had lymph node involvement. Similarly, whereas only 12.5% of MMP7KO mice had hepatic involvement, 66.7% of MMP7Het mice had liver metastasis. These striking differences suggest that MMP7 contributes to PDA progression by supporting tumor size and metastasis.
Despite the profound differences in disease staging between the MMP7Het and MMP7KO cohorts, there was surprisingly no difference in overall survival between mice the two groups (Figure S5C). While these results may, at first glance, imply that MMP7 loss results in increased malignancy (i.e. death despite lower tumor burden), a more detailed analysis of the cohorts suggested otherwise. First, despite the lack of difference in overall survival, 3 of 9 mice with complete MMP7 loss survived beyond 36 weeks, while no Pdx1-CreLate; KrasG12D; p53f/+; MMP7Het mice survived beyond 35 weeks and with the majority succumbing before 30 weeks (Figure S5C). Also, 5 of 9 Pdx1-CreLate; KrasG12D; p53f/+; MMP7KO mice appeared to have died from functional obstructive disease in the setting of relatively small primary tumors. Three mice, including one long-term survivor (Mouse 221, 37.9 weeks, Table S1), had a massively dilated stomach and duodenum with a relatively small pancreatic head tumor located at the transition point (Figures 6E and S5E). Another long-term survivor (Mouse 328, 36.1 weeks, Table S1) had biliary outlet disease that resulted in jaundice and grossly dilated gallbladder and common bile duct that measured 5mm (Figures S5F and S5G). Four of these 5 MMP7KO animals did not have any evidence of distant disease and in all likelihood would have survived longer if not for these functional obstructions of the gastrointestinal tract. In contrast, only 2 of 10 Pdx1-CreLate; KrasG12D; p53f/+; MMP7Het mice had significant obstruction but one of these also had disseminated disease (Mouse 344, Table S1). Finally, the potential for relatively longer survival in mice with MMP7 deletion was confirmed in a second cohort of mice utilizing Ptf1a-Cre to activate expression of mutant Kras in the context of one deleted allele of p53 (Ptf1a-Cre; KrasG12D; p53f/+; MMP7Het or MMP7KO). Two out of 4 of these Ptf1a-Cre; KrasG12D; p53f/+; MMP7KO mice survived beyond 34 weeks while no MMP7Het mice reached even 29 weeks of age (Figure S5D). This suggests that complete loss of MMP7 may extend life span beyond the time frame observed in mice with intact MMP7 activity.
While we observed significant differences in disease stage in our studies, we saw no differences in tumor histology in any of the cohorts we followed. All mice analyzed developed tumors that can be generally described as moderately to poorly differentiated (Figures 6F and 6G), irrespective of MMP7 status. Of interest, MMP7 staining in Pdx1-CreLate; KrasG12D; p53f/+; MMP7Het mice was seen in a patchy distribution throughout the tumor localizing to the luminal surfaces of glandular structures, consistent with staining patterns seen in human PDA (Figures 6H and S5J). As expected, no MMP7 expression was found in PDA observed in Pdx1-CreLate; KrasG12D; p53f/+; MMP7KO mice (Figure 6I).
Previous studies with human PDA samples have shown that MMP7 levels correlate with tumor stage (Fukushima et al., 2001; Yamamoto et al., 2001) and overall survival (Jones et al., 2004; Yamamoto et al., 2001). To study the potential utility of serum MMP7 to provide pre-operative clinical information, we accessed the University of Utah Huntsman Cancer Institute database and identified 101 PDA patients in whom serum MMP7 levels were measured.
When patients were grouped by tumor class, all patients without lymph node involvement (N0) had serum MMP7 levels below 20.2 ng/ml, while patients with metastatic disease (M1; Stage IV; Please see Table S2 for the TNM staging system for PDA) had significantly higher levels (Figures 7A and 7B). However, regional lymph node involvement (N1; Stage IIB) did not correlate to MMP7 levels even though some N1 patients did have MMP7 levels above 20.2 ng/ml.
Serum MMP7 levels were a significant predictor of survival, independent of whether MMP7 was used as a continuous or discrete variable, with 20 ng/ml as the threshold value (Table S3). With one exception, patients with MMP7 above this value had higher than expected risk of earlier death. PDA patients with MMP7 serum levels below 20 ng/ml had a median survival of 329 days, while those above the threshold had a median survival of 114.5 days (Table S3). Kaplan-Meier survival analysis confirmed significantly decreased survival in the high serum MMP7 group (Figure 7C and Table S4).
We found that tumor stage, distant metastasis, tumor resectability, and whether the patient received treatment were all favorable predictors of survival in univariate analyses (Table S3). Interestingly, serum MMP7 levels remained a significant predictor of survival even after stratifying for treatment type and tumor class in multivariate analysis (Table S5). These data suggest serum levels of MMP7 may be a useful clinical tool to predict survival in patients suffering from PDA.
The reprogramming of normal pancreatic cells into those with a neoplastic fate is a key step in PDA initiation and deserves closer study. Mouse models have revealed that mutant Kras significantly alters pancreatic plasticity, and can drive acinar cells towards a ductal fate capable of giving rise to PDA (De La et al., 2008; Guerra et al., 2007; Habbe et al., 2008; Ji et al., 2009; Morris et al., 2010). Chronic pancreatitis is a potent risk factor for PDA development in humans, and pancreatitis induction in Kras-driven mouse models of PDA have revealed that inflammatory damage can promote (Guerra et al., 2007; Morris et al., 2010) and/or accelerate (Carriere et al., 2009; Morris et al., 2010) PanIN and PDA formation. Kras appears to drive acinar-derived PanIN development following pancreatitis by blocking normal regeneration in favor of a persistently de-differentiated, ductal state that is permissive for PanIN formation. This finding suggests that assuming a de-differentiated state may be a rate-limiting step during Kras-driven neoplastic reprogramming in the pancreas. In this study, we induced both acute and chronic pancreatitis in transgenic mice expressing an oncogenic Kras allele in the context of epithelial Stat3 deficiency and found that these animals formed significantly fewer PanINs and retained more normal parenchyma than mice that had intact Stat3 signaling. These data implicate Stat3 as an important accomplice to Kras in diverting normal regeneration and maintaining the dedifferentiated ductal state initiated by pancreatitis.
Also, we found that in the presence of mutant Kras, Stat3 supports metaplasia associated inflammation in a manner that is distinct from what is observed during normal pancreatic regeneration. We found that pancreatic epithelial Stat3 deletion results in attenuated inflammatory cell infiltration and reduced expression of pro-inflammatory cytokines. We speculate that the reduced cytokine expression in Ptf1a-Cre; KrasG12D; Stat3f/f mice is due to a combination of cell-autonomous effects of Stat3 deletion in the pancreatic epithelium and a reduced number of infiltrating inflammatory cells. Therefore, our data suggest that Stat3 plays both a critical cell autonomous role in maintaining the proliferative, persistently de-differentiated epithelial state permissive for PDA precursor formation and also contributes to metaplasia associated inflammation. Further work is required to enhance our understanding in how Stat3 activation contributes to immune cell recruitment. Also, aberrant cytokine expression has been shown to enhance tumorigenesis (Ernst et al., 2008), and additional studies are required to determine how epithelial-derived, Stat3 dependent cytokines, and those produced by infiltrating immune cells contribute to PanIN development and PDA progression. Another unresolved question concerns the roles of the distinct immune cell types infiltrating during Kras-driven pancreatitis-induced PanIN formation. Summarily, these findings demonstrate that Stat3 activation in pancreatic epithelial cells plays a significant role linking pancreatitis and PDA initiation in vivo.
While we have focused our efforts on caerulein-induced pancreatitis, it would be interesting to investigate whether Stat3 signaling is also important in other models of pancreatic damage. These include hereditary pancreatitis caused by a missense mutation in the trypsin gene (PRSS1) (Archer et al., 2006) or cytokine-induced pancreatitis caused by ectopic expression of interleukin-1ββ (Marrache et al., 2008). Another potential avenue of investigation is whether Stat3 interacts with the primary cilia, which are the cellular appendages known to regulate Hedgehog signaling. The elimination of the primary cilia produces pancreatic lesions that resemble those found in patients with chronic pancreatitis or cystic fibrosis (Cano et al., 2006).
Similar to findings obtained from other organs, we confirmed that Stat3 controls MMP7 expression in PDA cell lines. In this study, we used several genetic models to demonstrate that MMP7 has a functional role in PDA progression by predominantly affecting tumor size and metastasis. Furthermore, we and others have found that MMP7 levels correlate with both tumor stage and survival (Jones et al., 2004; Yamamoto et al., 2001).
The mechanisms by which MMP7 affects PDA tumor size and metastasis need to be clarified in future studies. However, based on the number of substrates that MMP7 can act upon, multiple pathways may be involved. Epithelial basement membrane and extracellular components are MMP7 targets (Remy et al., 2006), suggesting that PDA cells lacking in MMP7 may not be able to escape the primary tumor site. Another possibility is the need for MMP7 activity for proper seeding at a potential metastatic site (Lynch et al., 2005). MMP7 can also influence the immune system by controlling chemokine gradients to affect neutrophil infiltration (Li et al., 2002). Such effects on immune cells may help propagate the feed-forward loop established by aberrant Stat3 signaling and perhaps contribute to the exuberant desmoplastic reaction seen in PDA. Finally, MMP7 may also affect the tumor angiogenic response as MMP7 knockout mice develop prostate cancers that have fewer tumor blood vessels that are also smaller in diameter (Littlepage et al., 2010). Future studies will have to address the exact mechanisms by which MMP7 functions in the context of PDA, a feat that is complicated by the observation that MMP7 is expressed only regionally throughout the tumors, suggestive of local differences in the malignant nature of individual cells throughout a tumor. The concept of the “invasive front” that possesses more malignant and metastatic potential was recently reviewed (Christofori, 2006) and it is conceivable that metastases first arose from these nests of MMP7-positive cells. This highly speculative concept is an intriguing idea that could be tested with the development of lineage tracing systems dependent on MMP7 expression.
Our study implicates both Stat3 and MMP7 as potential targets for adjuvant therapy development. The involvement of Stat3 in PDA initiation highlights the potential of Stat3 as a therapeutic target for preventing inflammation-associated PDA formation at the earliest stage. A recent study confirmed the potential utility of targeting the Stat3 pathway by showing that JAK2 kinase inhibition and Stat3 knockdown both abrogated tumor xenograft growth in variety of solid organ cancers, including pancreas (Hedvat et al., 2009). The role of Stat3 in pancreatic cancer progression and metastasis in vivo requires future studies, though there is in vitro data showing the importance of Stat3 to PDA cell growth (Scholz et al., 2003).
The significant reduction of metastatic disease seen in MMPKO mice has profound translational implications for PDA patients. Currently, only about 20% of patients undergoing PDA resections survive beyond five years (Yeo et al., 1995). Our study suggests that MMP7 inhibition may be a potential adjuvant therapy to block disease progression after resection. However, previous clinical trials involving MMP inhibitors have yielded disappointing results (Coussens et al., 2002). Part of the failure may have been because clinical studies enrolled patients with advanced stage cancers, while successful results in mice occurred with treatment starting at earlier disease stage. It is possible that the beneficial effect of MMP7 inhibition is greater in prevention of progression of subclinical disease as in the adjuvant setting than in reduction of size of gross metastatic disease. Furthermore, some MMPs may have tumor suppressive effects that could be compromised by non-specific pan-MMP inhibition (Kessenbrock et al., 2010), suggesting that targeting specific MMPs may be a more effective therapeutic strategy.
Using a large repository of serum samples from human PDA patients, we found that serum MMP7 levels significantly correlate with metastatic disease and survival, albeit with low sensitivity. This finding may be able to help guide patients and physicians with diagnostic and prognostic information to predict the survival benefit before patients undergo PDA resections. These considerations are particularly important as pancreatic resections entail complicated and lengthy operations that not only require months of recovery but carries significant risk for complications and death (Winter et al., 2006). Serum MMP7 levels might also be considered in decisions regarding postoperative adjuvant chemoradiation therapy after resection.
Recent sequencing studies have revealed that PDA cells possess a bewildering array of mutations (Jones et al., 2008). In order to develop improved therapeutics, it is essential to identify critical signaling nodes that PDA cells depend upon for growth and spread. Our study implicating Stat3 and MMP7 as key mediators of PDA initiation and progression not only points to potentially novel targets for future therapies but also demonstrates the utility of the Kras/inflammation system as an in vivo model to dissect complex signaling pathways important to pancreatic carcinogenesis.
Experimental animals were generated by crossing Ptf1a-Cre (Dr. C. Wright, Vanderbilt University) (Kawaguchi et al., 2002), Pdx1-CreLate (Dr. P. Herrera, University of Geneva Medical School, Geneva, Switzerland) to mice expressing LSL-KrasG12D (Dr. D. Tuveson, Cancer Research UK Cambridge Research Institute, Cambridge, United Kingdom), (Hingorani et al., 2003) and Stat3flox (Takeda et al., 1998), MMP7KO (Crawford et al., 2002), or p53flox/ (Bardeesy et al., 2006). Transgenic mice were in a mixed background. Littermates not expressing Cre as well as Ptf1a-Cre and Stat3f/+ mice of the same age were used as control. All mouse experiments were performed under the approval of the UCSF IACUC.
Acute pancreatitis was induced by intraperitoneal (IP) caerulein injections (50μg/kg; Sigma) - one injection per hour for 6 injections and repeated two days later. Chronic pancreatitis was induced by 3 or 4 weeks of caerulein injections as previously described (Martinez-Romero et al., 2009; Strobel et al., 2007) with the following modifications: The 3 week protocol: 6 hourly IP injections (50 μg/kg) per day, 3 days per week, for 3 weeks. The 4 week protocol: one IP injection per day (70 μg/kg), 5 days per week, for 4 weeks. Animals were sacrificed 3 days after the final injection in the both chronic pancreatitis models.
Primary acinar tissue cultures were prepared by modifying published protocols (Means et al., 2005). Transdifferentiation events were induced by addition of TGF-αα (50 ng/ml; Chemicon International). Cultures were maintained in a 37°C and 5% CO2 incubator for 5 days with medium replaced on days 1 and 3. Cultured acini were treated with 10nM caerulein for 20 hours.
Serum samples were obtained from 101 consecutive subjects with histologically or cytologically confirmed PDA for whom pretreatment serum samples were available. The cohort consisted of 46 women (median age 69 years, range 43–86 years) and 55 men (median age 64 years, range 42–86 years). Whole blood was collected prior to the patients receiving treatment. Serological MMP7 measurements were performed by ELISA (Quantikine Human MMP7 (total) Immunoassay, R&D Systems) according to the manufacturer recommendations except that serum samples were diluted 4-fold with sample diluent. All human subjects gave informed consent and the study was approved by the Institutional Review Board of the University of Utah.
Data presented as mean ± SD or SEM as noted and subjected to unpaired t-test, unless otherwise noted. P value < 0.05 was considered to be significant. Statistical analysis on clinical variables and serum MMP7 levels methods are described in Supplemental Experimental Procedures.
We thank all of our colleagues who shared the indicated mouse lines as detailed in the Methods section, Cecilia Austin for tissue processing and preparation, and all Hebrok lab members for helpful discussion. Imaging experiments were supported by the UCSF Diabetes and Endocrinology Research Center. Work in Hebrok Laboratory was supported by the Pancreatic Cancer Action Network and the NIH (CA112537). A. Fukuda was supported by a post-doctoral Research Fellowship of the Japan Society for the Promotion of Science, the Klein Family Foundation, and the National Pancreas Foundation. S. Wang was supported by the NIH under Ruth L. Kirschstein National Research Service Award F32 from the National Cancer Institute and the American College of Surgeons Resident Research Scholarship. Work at the University of Utah was supported by the NIH (R03 CA115225 to SJM and P30CA042014 to the Huntsman Cancer Institute for core facilities), the Huntsman Cancer Institute Pancreas Cancer Research Program and the Huntsman Cancer Foundation.
Conflict of interest statement: The authors declare that no conflicts of interest exist.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.