br In addition to the liver
In addition to the liver-derived cancers, the trend of a decrease in tissue-specific SP gene expression generally held true among the other cancers (Figures 4 and S3), all of which exhibited either a significant coordinated decrease or no significant change in
the genes specific to their respective tissue of origin. Further-more, the same behavior was observed even when including only high-purity tumor samples (Figure S4A).
Consistent with the GSA results, LIHC and CHOL exhibited a significant coordinated decreased expression of liver-specific genes. None of the 176 liver-specific SP genes were significantly (padj < 0.05) increased in either LIHC or CHOL relative to paired-normal tissue, whereas 156 (89%) and 174 (99%) of these genes exhibited a significant decrease in expression for LIHC and CHOL, respectively. These genes encoded functions such as lipid and cholesterol transport and metabolism (apolipoproteins), the complement system, coagulation, and protease inhibition (serpins). Similar strong, coordinated decreases in the expres-sion of tissue-specific SP genes were observed in breast, colo-rectal, and lung cancers, in which only three or fewer genes in each set (<6%) were significantly increased in expression, while the majority were significantly decreased. The four cancer types that did not show a significant coordinated decreased expres-sion in SP genes specific to their corresponding tissue of origin were BLCA, esophageal carcinoma (ESCA), PRAD, and UCEC. However, ESCA, PRAD, and UCEC did exhibit a significant decrease in the expression of genes specific to a tissue near their tissues of origin (stomach, seminal vesicle, and ovary, respec-tively) (Figure S3), suggesting a similar phenomenon. The data cannot distinguish between tumor ABT-263 (Navitoclax) that have actively decreased their tissue-specific gene expression and those that originated from more stem-like cells from the start; however,
Figure 4. Tissue-Specific Expression Changes in SP Genes
The heatmap shows the significance and direction of coordinated expression changes in SP genes classified as specific to various tissue types. Cancer and tissue types are organized such that entries along the diagonal represent cancer types paired with their tissue of origin and are outlined in a solid box if there is a significant (padj < 0.05) coordinated expression decrease among the tissue-specific SP genes for that cancer type or in a dotted box otherwise. The log-transformed p values of cancer types sharing the same tissue of origin were averaged to facilitate this organization. The complete results for each individual tissue and cancer type are presented in Figure S3. The number of tissue-specific SP genes for each tissue type are indicated in the bar plot to the left of the heatmap. The distribution of tissue-specific SP gene expression changes across different cancer types is presented for two representative tissue types: prostate and liver. The log2FC values for each set of genes are represented by boxplots, with the individual gene values shown as gray points whose sizes indicate the significance (p value) of their FC. See also Figure S4.
the end state is the same in that (most) cancer types exhibit a lower expression of tissue-specific SP genes in tumor cells than in the corresponding normal tissue.
Evaluation of Secretory Pathway Stress Signatures
The common decrease in the expression of tissue-specific SP genes across many different cancer types suggests a general pattern in which tumor cells are relieving the burden on an already strained (Ma and Hendershot, 2004) secretory system. By limiting the production and secretion of tissue-specific com-ponents, tumor cells may be able to dedicate more resources to processing proteins that contribute to cell proliferation and other malignant processes. To investigate further, we evaluated the tu-mor versus paired normal DE data for signs of increased stress or burden on the secretory pathway.