In this study, we first utilized ChRO-seq to identify RNA polymerase activity in hemangiosarcoma tumor tissue and normal splenic tissue. Analysis of the correlation matrix for our ChRO-seq dataset finds that 14 of the HSA tumor samples clustered together, while 3 of the HSA samples appeared to be more similar to normal splenic tissue (Figure 1). While speculative, it is possible that these 3 samples represent a different subtype of HSA and, as such, displayed a gene expression profile that is more similar to normal tissue. Alternatively, it is also possible that these 3 presumptive HSA tissues were actually tumor-adjacent normal tissue or a mix of normal and tumor tissue. While grading of canine hemangiosarcoma is not often utilized due to the aggressive nature of the neoplasm, a grading scheme does exist and differences in gene expression between the samples may correlate with progression and eventual outcome [34].
Previous HSA transcriptomic profiling studies have implicated a number of signaling pathways in HSA pathogenesis. Tamburini et al, for example, found that, when compared to cell lines derived from splenic hematomas, HSA cell lines exhibited several different distinct gene expression profiles, including signatures for angiogenesis, inflammation, adhesion, invasion metabolism, cell cycle and signaling. In another study, Gorden et al performed microarray and RNA-seq analyses on visceral HSA tumor tissues (12 spleen, 7 heart, 4 liver, 1 lung) and identified three distinct tumor subtypes that were associated with either angiogenesis, inflammation, and adipogenesis. Our finding that ECM remodeling appears to be a major gene expression signature in HSA differs from these previous studies. One possible reason for the difference in outcomes between our study and the Tamburini study is that our report compared gene expression patterns in HSA splenic tissue with normal splenic tissue while the Tamburini study was comparing the gene expression profiles of HSA cell lines with splenic hematoma cell lines[16]. Thus, the differences may be due to the fact that the expression profile of normal splenic samples differs significantly from hematoma samples. Additionally, these differences may also be because the expression profile of cell lines may differ significantly from primary tissue samples due to decreased cellular complexity and prolonged cell culture. One potential reason why outcomes from our study differed from the Gorden study is that we evaluated differences in gene expression patterns between HSA and normal splenic tissue while the other study analyzed gene expression patterns within HSA tissues. Lastly, another reason why outcomes from our study may have differed from both of these previous studies is that ChRO-seq analysis detects nascently synthesized RNA as opposed to mature transcripts, whose levels can be affected by a variety of factors including transcript stability.
Regarding the specific types of ECM factors that were upregulated in our dataset (Table 2), we identified genes encoding fibrous components of the ECM, including 12 collagen genes, 3 laminin genes, and fibronectin. Several ECM binding proteins were also identified in this dataset including, Lumican and Biglycan, which are small leucine-rich proteoglycans that regulate collagen fibril and matrix assembly [35–38]. Integrin alpha 2 is directly associated with fibril-forming collagens (1,2,3,5,6,14, 18) [39] while DDR2 functions as a cell surface receptor for fibrillar collagen and regulates cell differentiation by remodeling ECM[40, 41]. Additionally, integrin alpha 5 binds directly to fibronectin while integrin alpha 6 binds to laminin [42][43]. Several ECM-related enzymes were also found in this dataset including P4HA2 and PLOD1 which catalyze collagen biosynthesis [44, 45] along with ADAMTS14, ADAMTS4, ADAM12 and TLL1 which process procollagen to collagen by cleaving N-propeptide and C- propeptide [46, 47]. Lastly, two molecules directly involved in ECM turnover, tissue inhibitor of metalloproteinase 1 (TIMP1) and matrix metalloproteinase 14 (MMP14), were also found in this dataset [48, 49][50].
The extent to which ECM genes were upregulated in HSA tumors in our study is highlighted by our GO analysis which found that 9 of the top 10 biological process categories were ECM-related (Figure 2). Interestingly, 5 of these 9 categories related to collagen function. We more directly tested for the abundance of collagen fibers in the tumor samples using Masson’s trichrome stain and found extensive collagen deposition throughout the tumor tissue (Figure 3). Collagen deposition was observed both in the more solid areas of the tumor and, in particular, in the tumor regions filled with malformed vascular channels. In these vascular regions, neoplastic endothelial cells are often found to encircle the collagen bundles forming “gumball”-shaped structures that look like inverted blood vessels. While stromal fibroblasts are presumably primarily responsible for the synthesis of these collagen fibers, it is also possible that the tumor cells may also be partly responsible for synthesis of these collagen fibers.
In addition to collagen, we also found that PDPN and LAMA4 were highly overexpressed in HSA tumor tissue when compared to normal tissue. These molecules were of particular interest to us given their close association with cancer progression in other types of cancer. Podoplanin is a mucin-type protein that contains an extracellular region, transmembrane domain, and intracellular tail. It is widely known as a marker for lymphatic endothelial cells and also to play a critical role in heart and lung development and in development of the lymphatic endothelial system [51–53]. PDPN appears to play several roles in cancer progression. A number of studies have shown that PDPN expression in cancer cells promotes tumor cell proliferation and invasion [28–30]. In addition to cell-intrinsic effects, PDPN is also believed to promote tumor metastasis by interacting with its receptor, CLEC2, on the platelets. This interaction then promotes the coating of tumor cells by platelets, thereby protecting tumor cells from the immune system [54]. In addition to its role in human cancers, PDPN is also overexpressed in canine squamous cell carcinomas and melanomas [55] and PDPN mAbs were recently found to have potent anti-tumor activity in mouse xenograft models of canine melanoma [56]. Interestingly, overexpression of PDPN in mice leads to disseminated intravascular coagulation [57], a condition that is strikingly similar to that seen in many dogs with HSA [58]. Our ChRO-seq data demonstrated that PDPN expression varies significantly between HSA samples (Figure 1, 4). IHC analysis found that PDPN was robustly expressed in transformed endothelial cells in certain HSA tumor samples. However, similar to our ChRO-seq finding, we did find that PDPN appears to only be expressed in a subset of HSA tumors. In future studies, we plan to test whether PDPN expression in HSA tumors correlates with disease severity.
Laminins form large heterotrimeric abg protein complexes and are a prominent component of basement membranes. LAMA4 is distinct from other laminin isoforms in that it lacks a polymerization domain with the loss of this domain potentially facilitating tumor cell migration [31]. Interestingly, LAMA4 was recently described as an “oncolaminin” due to its strong association with cancer cell migration and tumor progression in a range of cancers [32]. These links to cancer include recent studies which found that LAMA4 and MCAM (melanoma cell adhesion molecule) are highly enriched in tumor blood vessels in renal cell carcinoma and colorectal carcinoma. Additionally another study found that expression of these molecules is enhanced in locally invasive and metastatic clear cell renal cell carcinoma [33]. Further, antisense oligonucleotides against laminin-8 (LAMA4 and LAMB1) were found to block the invasion of glioma cells and neovascularization in vitro [59]. Taken together, these published studies indicate that LAMA4 plays a key role for vascular development, tumor progression and metastasis. In our study, deseq2 analysis found that LAMA4 is highly transcribed in HSA tumor tissue. This observation was supported by PCR analysis of mRNA isolated from HSA tumor tissue (Figure 4). Additionally, our IHC analysis found that LAMA4 protein expression appears to be primarily limited to malignant endothelial cells (Figure 5). Given the previously defined roles for LAMA4 in cancer progression and tumor metastasis, our findings raise the possibility that LAMA4 is an important mediator of canine HSA pathobiology.