br Activation of the UPR br Disruption
Activation of the UPR
Disruption of the secretory pathway results in the accumulation of misfolded proteins, which in turn activates a series of adaptive processes collectively known as the UPR to restore ER homeostasis (Ron and Walter, 2007). Coordinated expression in-creases in UPR-associated genes would therefore be indicative of Acetylcysteine undergoing secretory pathway stress and UPR activa-tion. For each cancer type, we evaluated the enrichment of expression changes in genes affiliated with the UPR (all affiliated genes, not only secreted or SP genes). The results revealed a significant coordinated increase in UPR-related gene expression in nearly all cancer types (Figures 5 and S5A), consistent with
previous reports regarding the prevalence of UPR activation among many cancers (Dejeans et al., 2014; Ma and Hendershot, 2004). CHOL and papillary thyroid carcinoma (THCA), however, exhibited a negligible coordinated expression increase in UPR-associated genes. The same results were observed when considering only high-purity tumor samples (Figure S4B), although CHOL was excluded due to the absence of purity scores for this cancer type.
Given that CHOL and THCA were among the cancer types ex-hibiting a strong coordinated expression decrease in tissue-spe-cific SP genes (Figure 4), the data are supportive of the observed pattern whereby tumor cells alleviate secretory pathway stress by reducing the expression of SP genes specific to sustaining the function of their tissue of origin. Likewise, cancer types with an insignificant decrease in the expression of their respec-tive tissue-specific SP genes (BLCA, ESCA, PRAD, and UCEC) exhibited coordinated expression increases associated with ER stress and the UPR (Figures 5 and S5A). Estimation of Secretory Burden
Proteins traversing the secretory pathway undergo a number of maturation processes such as folding and post-translational modifications (PTMs). Larger proteins with a greater number of PTMs will require more cellular resources than shorter, less-modified proteins, and thus may impart a greater burden on the secretory pathway (Feizi et al., 2017; Gutierrez et al., 2018). We reasoned that a shift in expression toward lower-cost pro-teins may constitute another potential strategy to alleviate
Figure 5. Coordinated Expression Increases Associated with the UPR
Shown are the log-transformed directional p values representing the signifi-cance of coordinated expression changes in genes associated with the UPR, defined as those included in the unfolded protein response gene set in the Hallmark gene set collection from MSigDB. Bars are colored blue if there is a significant (padj < 0.05) expression increase among the genes for that cancer type; if not, they are colored yellow. See also Figures S5 and S6.
secretory pathway stress in tumor cells. To quantify this cost, we formulated a secretory burden (SB) score for each SP gene i as a function of its encoded protein length L (i.e., number of amino acids) and number of disulfide (NDS) and glycosylation (Ngly) sites: SBi = Li + NDS;i + Ngly;i (Equation 1)
where each property is normalized by the median (med) value among all of the SP genes.
For each cancer type, the Spearman correlation between gene SB scores and expression fold changes was calculated (Figures 6 and S5B). Although the correlation coefficients were low, the trend was consistent with our observations regarding UPR acti-vation and decreased expression of tissue-specific SP genes, which is best illustrated by the two extremes, BLCA and CHOL. BLCA yielded the strongest negative correlation between SB score and log2FC, suggesting that expression increases tend to be associated with low-burden SP genes, whereas the oppo-site was true for CHOL. Given that BLCA showed evidence of UPR activation and exhibited the least significant expression decrease in tissue-specific SP genes, it suggests that the inability of BLCA cells to relieve secretory pathway stress via reduction in tissue-specific SP gene expression may constrain their ability to process proteins with a high secretory burden. Conversely, CHOL exhibited the strongest expression decrease in tissue-specific SP genes and showed little evidence of UPR activation, which is indicative of lower secretory pathway stress, thus relaxing the constraint on which proteins the secretory pathway can accommodate. Cancer types mirroring the trend of BLCA included ESCA, PRAD, and UCEC, whereas THCA followed that of CHOL.