Our study is a unique attempt at identifying and investigating the role of early EV- related biomarkers that can be quickly adapted for use in the detection of early stage pancreatic cancer, a disease that is notorious for its propenisty to metastasize quickly with barely any symptoms in the initial stages.
EVs or exosomes have been identified as promising candidates for the identification of biomarkers given their robustness in circulation and in being resistant to degradation. Translating in vitro findings to preclinical studies and eventually for validation into the clinic has a high failure rate. For our studies, we adopted a novel strategy of developing a mouse-human hybrid TMA that can be easily used to test biological significance of biomarkers. The TMA was developed by taking different developmental stages of tumors in the KPC GEMM mouse model. Additionally, it included cancers and adjacent normal tissue as well. Clinical normal and diseased tissues were also cored onto the TMA. Bi-species reactive antibodies were used to investigate the presence of Loxl2, Mmp3, Kif5b and Sfrp2.
Lysyl oxidase related protein-1 (LOR-1) (currently termed LOXL2) promotes tumor fibrosis and tumor invasiveness by invading blood vessels, nerves, and muscles adjacent to tumors [46]. High concentrations of LOXL2 significantly correlated with increased tumor malignancy and cells of LOXL2-expressing tumors are surrounded by high concentrations of dense collagen fibers and contained many fibrotic foci. Additionally, LOXL2 has gained attention due to its critical role in tumor progression and metastatic potential in pancreatic cancer [47]. The protein mediates the crosstalk of cancer cells [48, 49] and fibroblasts [50, 51], driving EMT [52–56], and regulating ECM remodeling and crosslinking [57, 58]. Inhibition of expression reduces tumor volume and metastasis in PDAC pre-clinical trials [57]. Thus, LOXL2 seems to be the key communicator between tumor cells and the tumor microenvironment and is known to mediate ECM remodeling and CAF activation [59]. MMPs are involved in not only cancer initiation and progression, but also functional promotion of angiogenesis, invasion, metastasis, and avoidance of immune surveillance. MMPs are primarily produced by stromal cells in response to paracrine cytokine signaling in the tumor environment. While analyzing the expression of MMP3 molecules in a large panel of human pancreatic adenocarcinomas, evidence of the involvement of this pathway in all stages of pancreatic cancer progression. Further evidence that pancreatic acinar cell MMP3 interacts with KRAS to initiate pre-malignant alterations in the surrounding stroma was found using transgenic mouse models. Selective MMP-activation occurs in the CAM secretome even when cultured cells have no exposure to cancer cells. Hence, this shows the possibility a cell-autonomous mechanism where CAMs, with increased selective MMP activation and production, can contribute to the remodeling of the cancer microenvironment [60]. SFRP2 secreted from aged fibroblasts augment angiogenesis and metastasis of melanoma cells contributing to tumor progression [61]. Secreted plasma levels of SFPP2 is a non-invasive tool for the detection of gastric cancer and is a valuable serum tumor marker [62]. It has been a fibroblast-secreted marker and in colorectal cancer, Sox2-dependent CAF population secrete Sfrp2 and inactivation of Sox2 or Sfrp1/2 in CAFs impairs the induction of migration and invasion of colon cancer cells, including their tumorigenicity in vivo [63]. In pancreatic cancer, SFRP2 cell-free DNA promoter hypermethylation in plasma is a biomarker for pancreatic adenocarcinoma [64]. In addition, the hypermethylation and aberrant expression of frizzled-related protein genes are common in pancreatic cancer and have been shown to be involved in pancreatic carcinogenesis[65]. TET1 protein binds to SFPR2 and inhibits canonical and non-canonical Wnt signaling pathways; this process prevents EMT transition in pancreatic tumors [66]. All biomarkers showed similar patterns of staining across mouse and human samples. Cancer cells strongly expressed these proteins in both primary and distant metastases. It is difficult to validate the presence of the protein in human samples in an age dependent manner due to lack of clinical samples. In KPC GEMM mice, the expression of Kif5b, Sfrp2, and Loxl2 was high in mice as old as 35 days. Mmp3 however, stained weakly, thereby forcing us to disregard it as a putative early biomarker. Thus, our study provides compelling evidence for the utility of Sfrp2 and Kif5b as putative markers for early pancreatic cancer diagnosis.
Additionally, our experimental findings suggest that the secreted EVs contain several hundred signaling moieties that can modulate the TME. Studies by several groups have shown that exosomes secreted from cancer cells can modulate the pre metastatic niche and influence tumor metastases in a liver metastasis animal model of pancreatic cancer [24]. If TME remodulation is different, we hypothesized that the rate of metastases would also be different. To test this assertion, we injected EVs that were purified from different experimental conditions such as those obtained from KPC cells, stellate cells and the cancer cells and stellate cells cultured together. There was no significant change in the liver weights or increase in the levels of metastasis in comparison to the control. This could partly be explained by our experimental setup. Short exposure times of 24 hours would not induce potent changes in the proteome to increase metastasis in the co-culture condition. However, the experimental limitation is that in vitro, cancer cells soon outnumber stellate cells even though they are nine-fold more abundant. Cancer cells are more aggressive and have rapid doubling times in comparison to stellate cells. The TME is composed of multiple cell types and other acellular factors such as cytokines and chemokines. By including additional cell types in the co-culture condition, we could have seen a more aggressive metastatic phenotype resulting in significant increase in liver metastasis. The development of an aggressive PMN was not visible. It was imperative to investigate if the EVs that were injected into the mice could influence any biological activity. Several studies [67–70] have shown that there is a distinct pancreatic microbiome which could decisively influence the development of pancreatic cancer. To investigate this phenomenon, fecal pellets were collected from mice who were injected with PBS control or with EVs from KPC, PSC and co-culture conditions two weeks after education. The microbiome was analyzed and we observed significant changes in the α and β-diversity profiles within two weeks. Each of the experimental conditions induced changes in the microbial flora. Compared to the PBS control injections, EVs from stellate cells, specifically induce changes in the populations of bacteria from the family coriobacteriaceae, genus Adlercreutzia. While, EVs from the cancer cells modulate levels of bacteria from the phylum Tenericutes, family Mogibacteriaceae, genus Adlercreutzia. There was a significant enrichment of bacteria fro the phylum Firmicutes, genus coprococcus, when mice were educated with EVs from the co-culture educated samples These findings provide ample evidence for the potent bioactivity of EVs to induce changes in the murine microbiome.
The most important finding in our study is the very local effect in the stroma. In both mouse and human tissue, the stroma adjacent to the tumor showed high expression of the protein while in adjacent normal tissue, though acinii show elevated levels of the protein, the surrounding tissues however, do not show expression of these biomarkers. Earlier studies have shown that there are different populations of fibroblasts within the stroma indicative of the heterogeneity [71–74]. Perhaps, these distinct subpopulations of cells respond differently to the secretion of proteins in the immediate vicinity. Additional studies are needed to characterize these stromal differences in pancreatic cancer. Thus, we propose that there are distinct cell populations in the immediate TME that are modulated when EVs are secreted from cancer cells in both the primary and metastatic site allowing for selective gene regulation. Additionally, since we see the expression of proteins except Mmp3 in very early stages of the KPC pancreatic cancer development, one could hypothsize that these factors could profuce a local effect first and distant effect later. However, our studies indicate that both Kif5b and Sfrp2 seem to be good robust candidates for early detection of pancreatic cancer.
The major limitation of the immunohistochemical approch is the availability of effective single antibodies that are reactive to both the mouse and human tissues. In our study, most markers that were selected for validation were discarded as they were mono species reactive. This situation can be easily remedied by using other non-antibody based strategies such as using nanobodies, DARPins, monobodies or affibodies or MALDI-based visualization techniques. Overall, the goal of our study was to identify early cancer-stromal specific changes in our study system and identify the biomarker that would be useful in identifying these early stage interactions between cancer cells- TME for PDAC detection. We have provides a novel mechanistic insight into the ways by which the EVs potentiate changes in the gut microbiome in our mice studies thereby shifting the microbiome towards a more ‘cancerous’ phenotype.