|
|
 |
 |
|
|
|
|
For each node i in the above PPI network, we defined three measures of its topological property: (1) 'Degree' is defined as the number of links to node i; (2) 'Degree index' is defined as the proportion of the number of links between node i and ¡®known cancer proteins¡¯ to its Degree; (3) For a original network, a ¡®k-core¡¯ sub-network is defined as its maximum subgraph in which each node has no less than k links[1], it can be generated by recursively deleting vertices from the network. When k increases, a series of ¡®k-core¡¯ sub-networks gradually reveal the globally central region of the original network. The ¡®K value¡¯ of a node is defined as the maximum value of k before this node is deleted from the network and it represents the centrality of a node.
The ¡®Degree¡¯ and ¡®K value¡¯ of a node can be calculated by tools of k-core analysis, such as UCINET; The ¡®Degree index¡¯ can be calculated by a series of operations in Excel.
The Degree and K value distributions were analyzed to determine the topological features of known cancer genes. We found that known cancer genes had significantly higher Degrees(r=0.686, p<0.0001, Spearman's coefficient of rank correlation, Figure 1A. This positive correlation indicates that the known cancer genes are highly connected. Additionally, we studied whether known cancer genes were close to the topological center of the network using the k-core analysis method. As a result, when k increased, the proportion of remaining cancer genes in the k-core sub-network also increased. Known cancer genes has a significant positive correlation with their K value (r=0.651, P=0.0005, Spearman's coefficient of rank correlation, Figure 1B, indicating that known cancer genes tends to be centrally located in the network.
As described above, three topological features, 'Degree,' 'Degree index,' and 'K value' were chosen to identify HCC candidate markers. For each HCC-significant protein, the 'Degree', 'Degree index' and ¡®K value¡¯ were calculated, and the mean values of these measures were also computed. When the three measures were all higher than their corresponding mean values, the protein was identified as a HCC candidate marker.
Immunohistochemical analysis was performed to validate the associations of LTBR, CD55, BMP7, BMP4, INHBE, MYO6, and SP100 expression with the clinicopathological features of HCC patients. We observed significant differences in BMP7, BMP4, and MYO6 protein expression between HCC and PCLT tissues (all P<0.05; see Supplement, Table 1 and Table 2.
Bone morphogenetic proteins (BMPs) belong to the TGF-beta superfamily. In addition to their function in bone tissue, BMPs are also involved in the pathogenesis of several solid tumors[2]. Among BMPs, BMP7 overexpression has been detected in esophageal squamous cell carcinoma, lung cancer, colorectal cancer, and breast cancer[3-6]. In HCC, Midorikawa et al. reported that BMP7 regulated the growth of HCC cells and that glypican-3 overexpression modulated this cell proliferation by inhibiting BMP7 activity[7]. However, BMP7 expression in HCC and its association with the clinicopathologic features of HCC patients have not been evaluated. In the present study, we found that the positive staining for BMP7 was mostly observed in the cytoplasm (Figure 2A) of tumor cells in HCC tissues. In HCC cases, 10 (31.25%), 8 (25.00%), 10 (31.25%), and 4 (12.50%) tissue samples showed negative, low, medium, and high BMP7 protein expression, respectively. In PCLT cases, there were 4 (12.50%), 19 (59.38%), 8 (25.00%), and 1 (4.12%) tissue samples that showed negative, low, medium, and high BMP7 protein expression, respectively, indicating that BMP7 expression was significantly greater in HCC tissues than in PCLT tissues (x2=9.08, P>0.01,P<0.05, Table 3). Additionally, BMP7 expression levels significantly increased with the stage of tumor node metastasis (TNM) of HCC cases (x2=20.30, P<0.005, Table 3).
BMP4 is implicated in the tumorigenesis of serous ovarian carcinoma, colorectal, breast, and lung cancers[8-11]. In HCC, Maegdefrau et al. found increased BMP4 mRNA and proteinin HCC cell lines and tissue samples, compared to primary human hepatocytes and corresponding non-tumorous tissue[12]. They also observed that BMP4 suppression using siRNA strong reduced migratory and invasive potential, and anchorage-independent growth, of HCC cells, and that BMP4 may promote vasculogenesis. However, the association between BMP4 expression and the clinicopathologic features of HCC patients has not been evaluated. In the present study, we found that the positive staining for BMP4 was mostly observed in the cytoplasm (Figure 2C) of tumor cells in HCC tissues. With the similar results of Maegdefrau et al[12]. , we also detected that the expression levels of BMP4 in HCC tissues were significantly higher than those in PCLT tissues (x2=12.38, P<0.01, Table 3). Additionally, BMP4 protein expression in HCC tissues with higher TNM stage was significantly greater than in tissues with low TNM stage disease (x2=16.26, P>0.01,P<0.025, Table 3).
MYO6 (myosin VI) is involved in intracellular vesicle and organelle transport. MYO6 is a putative prostate cancer marker localized to the Golgi apparatus[13]. MYO6 up-regulation was detected in prostate cancer and seemed to be involved in the development of this disease[14].However, MYO6 expression in HCC and its association with the clinicopathologic features of HCC patients has not been evaluated. We found that positive staining for MYO6 was mainly observed in the cytoplasm and/or nucleus (Figure 2E) of tumor cells in HCC tissues. The MYO6 protein expression levelsin HCC tissues were significantly greater than in PCLT tissues (x2=12.29, P<0.01, Table 3). Additionally, MYO6 expression levels in HCC tissues with TNM stage III were significantly higher than those with TNM stages I-II (x2=19.33, P<0.005, Table 3).
Using immunohistochemical validation, our data offer convincing evidence that BMP7, BMP4, and MYO6 may have clinical relevance in HCC, and may be candidate markers for evaluating HCC advancement. This demonstrated application of the LiverAtlas database in identifying HCC candidate markers suggests that the new data resource in LiverAtlas may facilitate research on hepatic pathologies by contributing to novel disease biomarker discovery.
Please see detailed information on patients, tumor tissue specimens, immunohistochemical protocols and statistical analysis in Supplement.
