Here we report the identification of novel tumor stellate cell specific genes and proteins. In addition, hepatic vs. pancreatic stellate cell specific transcripts were discovered. The mRNA and protein expression levels of candidate genes identified by genome-wide transcriptional analysis were confirmed by qRT-PCR, ELISA and Immunoblot analyses. The specific expression pattern of the candidate proteins was further assured in-vitro by immunocytochemistry of isolated stellate cells and ex-vivo by immunohistochemistry of formalin fixed paraffin embedded tissues. The identified molecular fingerprint of stellate cells might be instrumental in development of novel biomarkers and rational design of therapeutic strategies aiming to selectively target cancer or inflammation associated stellate cells. One drawback of this study is the absence of a third comparator, namely stellate cells from the normal pancreas and liver. However, stellate cells from the normal pancreas do not grow efficiently when propagated by the outgrowth method. Neither could stellate cells from fibrotic tissues like chronic pancreatitis and pancreatic cancer be efficiently propagated by collagenase digestion and centrifugation. The yield of the latter method is very low in comparison to the outgrowth method. Therefore the authors have decided to use only one stellate cell propagation method (outgrowth) to prevent bias that may result from differences in methodology. As a trade-off, a comparator (normal stellate cells) could not be used.
Among the here identified tumor stellate cell specific genes, JAK2 and CELSR3 pose interesting targets for developing therapeutic strategies. Aberrant JAK2 signaling has been linked to myeloproliferative disorders such as polycythemia vera and chronic myelogenous leukemia [31, 32]. Small molecular inhibitors of JAK2 signaling have already entered clinical trials. Therefore the role of JAK2 overexpression in PSC of PDAC tumors remains to be functionally elucidated. Upregulation of CELSR3 in tumor PSC could also provide a potential druggable target since the protein encoded by this gene is located at the plasma membrane and has intriguing signaling capabilities . CELSR3 is a member of the flamingo protein subfamily which is part of the cadherin superfamily. The flamingo cadherins have nine cadherin domains, seven epidermal growth factor-like repeats and two laminin A G-type repeats in their ectodomain . It is postulated that these proteins are receptors involved in contact-mediated communication, with cadherin domains acting as homophilic binding regions and the EGF-like domains involved in cell adhesion and receptor-ligand interactions. Together, these data suggest an important role for CELSR3 in tumor stellate cells that warrants further investigation.
Recent studies have reported differential regulation of genes in murine and human stellate cells throughout their activation process and under certain drug therapies. Although these studies provide valuable information, a careful interpretation of their data is warranted. First, murine stellate cells may not reflect the situation in humans due to interspecies differences. Second, stellate cells show a great variance depending on the donor, therefore a weakness of the past studies might be the limited number of patients' stellate cells investigated to cover the inter-individual heterogeneity. Our report constitute one of the most comprehensive studies on stellate cell transcriptome using 22 different human donors, thus diminishing potential patient specific biases. This might be a plausible explanation for the success of our approach to identify organ- and disease-specific stellate cell transcriptome. In contrast to the studies comparing quiescent vs. activated stellate cells, we compared human stellate cells that were activated in-vivo either by chronic inflammation or cancer. Currently there is no information on how the "ancestral microenvironmental activity" of stellate cells affects their transcriptome after in-vitro cultivation. Therefore, the identified tumor stellate cell specific genes e.g. CELSR3 might provide a favorable therapeutic profile to selectively target tumor stroma while sparing the stellate cell activity under physiological conditions.
It has been reported that experimental ablation of tumor associated fibroblasts or down-regulation of the hedgehog-signalling pathway in the pancreatic tumor stroma decreased cancer cell growth and greatly increased intra-tumoral uptake of chemotherapeutic drugs in murine models of colon, breast cancer an pancreatic cancers [5, 35]. It is likely that targeting the stroma in order to uncouple stromal-cancer cell interactions may interrupt multiple aberrant autocrine and paracrine pathways that promote pancreatic cancer cell growth, invasion and metastasis [4, 21]. Currently, there is no stellate cell specific promoter known to specifically target the stromal cells in the liver or pancreas. Moreover, the similarities between PSC and HSC make it very difficult to target one population without creating side-effects on the other one. For example, the usage of retinoids seems promising in deactivating PSC in-vitro. In contrast, treatment with retinoids were shown to provoke fibrogenic effects in HSC, thus, limiting its efficient systemic use in the clinical setting .
In this context, our data provide valuable information focusing on the subtle but important differences of liver vs. pancreatic stellate cells that exist rather than their similarities. We found collagen type 11a1 expression to be highly specific for the PSC and almost absent in HSC. Type XI collagen is a minor fibril-forming collagen that assists in proper type II collagen fibril formation . Col11a1 is mainly expressed in articular cartilage and the vitreous fluid of the eye [39–43]. During embryogenesis, Col11a1 expression is also detectable in many other human fetal tissues including the bone . Marshall and type II Stickler syndromes are genetically transmitted diseases that are caused by mutations in Col11a1 gene manifested with sensory and skeletal abnormalities [44, 45]. In Col11a1 knockout mice, chondrocytes fail to fully differentiate causing a chondrodystrophic phenotype with skeletal abnormalities. These data show that Col11a1 is essential for skeletal morphogenesis because it controls type II collagen fibrillogenesis, chondrocyte maturation and bone mineralization [46, 47]. Importantly, in adult life these tissues are not undergoing dynamic changes. Therefore Coll11a1 may provide an interesting target for conditional knockout to target PSC in order to assess PSC contribution in conditions like chronic pancreatitis or pancreatic cancer.
Buchholz et al. compared gene expression of human HSC, PSC and skin fibroblasts . Their data show that, compared to dermal fibroblasts, HSC and PSC exhibit great similarities in their transcriptional phenotypes and possibly share a common origin [3, 9]. Indeed, HSC and PSC are activated by common cytokines, growth factors and alcohol metabolites during development of tissue fibrosis [3, 7, 30, 48–50]. However, the fundamental differences in their microenvironments may condition these cells to differentially respond to organ injury and cancer [1, 3, 30]. For example, in contrast to hepatocellular carcinoma, desmoplasia is a typical feature of pancreatic ductal adenocarcinoma. This difference may result in part due to the release of potent fibrogenic mediators by pancreatic cancer cells [1, 19, 30, 51]. We have previously shown that, compared to the primary tumor, liver metastasis of PDAC elicit a weaker fibrogenic response in the HSC that surround them. This observation can be due to the differences of HSC and PSC, or it could, considering the temporal sequence of events, merely reflect a shorter cancer-HSC interaction . Interestingly, we found almost twice as much differentially regulated genes with an organ specific expression pattern as compared to chronic inflammation or tumor related genes. Moreover, differences in gene expression levels between HSC and PCS were more pronounced compared to disease specific stellate cells. These differences can be due to various factors. It may reflect that PSC and HSC do not share a common origin, or it may also suggest that chronic inflammation and cancer develop in similar microenvironments created by the activated stroma of the respective organs.
In the clinical setting, chronic inflammatory conditions of the liver (cirrhosis) and pancreas (chronic pancreatitis) not only increase the incidence of cancer, but also constitute diagnostic hurdles. Biopsies that do not show tumor structures are not helpful in diagnosis since they are useless in exclusion of the tumor. However, identification of tumor versus inflammation specific signatures of the stroma may increase diagnostic accuracy of biopsies even where tumor tissue is missed.
Conclusively, identification of inherent and acquired differences in stellate cell transcriptome will improve current understanding of stellate cell pathophysiology and may facilitate the discovery of novel selective therapeutic targets. Our data provide a new platform to understand and possibly interfere with organ or disease specific responses of stellate cells. Further research is warranted to analyze the importance of these findings at functional levels.