- HSP expression profiles in pan-cancer
1.1 HSP mRNA expressions in pan-cancer
On the basis of a literature review, 22 HSPs among HSP110, HSP90, HSP70 and HSP60 families were selected for analysis (Table 1) (5). We examined the mRNA expression of these 22 HSPs in cancer and non-tumor tissues in TCGA and found that mRNA levels of 10 HSPs were differentially expressed in 11 tumors (p < 0.05) (Figure 1A). HSPA2 mRNA was down-regulated in stomach adenocarcinoma, colon adenocarcinoma, bladder urothelial carcinoma, kidney renal clear cell carcinoma, kidney renal papillary cell carcinoma and kidney chromophobe (Figure 1B). The mRNA expression level of HSPA7 was decreased in colon adenocarcinoma but increased in lung adenocarcinoma, kidney renal clear cell carcinoma and kidney renal papillary cell carcinoma. HSPA1A mRNA was down-regulated in colon adenocarcinoma. HSPA6 mRNA was up-regulated in breast cancer, kidney renal clear cell carcinoma and kidney renal papillary cell carcinoma. HSPA4L mRNA expression was increased in lung squamous cell carcinoma and breast cancer and decreased in prostate adenocarcinoma. HSPA1L mRNA was down-regulated in uterine corpus endometrial carcinoma. HSPA12A mRNA was down-regulated in bladder urothelial carcinoma, breast cancer, uterine corpus endometrial carcinoma and kidney renal papillary cell carcinoma. HSPA12B mRNA was down-regulated in lung squamous cell carcinoma, lung adenocarcinoma, breast cancer, uterine corpus endometrial carcinoma and kidney renal papillary cell carcinoma. The mRNA expression levels of TRAP1 and HSPD1 were increased in lung squamous cell carcinoma.
We used the Oncomine database to validate the above findings. The results suggested that HSPs show varying degrees of differential expression in multiple tumors (Figure 1C).
1.2 HSP protein expressions in pan-cancer
We next used the THPA site to examine the expression profiles of HSP proteins in 16 tumors. Immunohistochemistry results indicated that HSP proteins showed different expression intensities in 16 tumors. The expression intensities of HSPs in 10 common tumors including lung cancer, liver cancer, colorectal cancer, pancreatic cancer, renal cancer, prostate cancer, breast cancer, endometrial cancer, ovarian cancer and melanoma were shown in Figure 2A. For example, HSP90B1, HSPA9, TRAP1, HSPH1 and HSPD1 showed high levels of immunostaining in 10 common tumors. HSPA6 showed moderate expression in lung cancer, breast cancer, endometrial cancer, ovarian cancer and melanoma and negative expression in endometrial cancer, renal cancer and liver cancer. HSPA8 showed moderate expression in endometrial cancer and breast cancer and negative expression in other cancers. The immunostaining of HSPA9 is shown in Figure 2B.
1.3 HSP expressions in pan-cancer cell lines
CCLE data revealed different expression levels of HSPs in 425 cell lines of 8 tumors, including breast cancer, endometrium cancer, kidney cancer, large intestine cancer, lung cancer, prostate cancer, stomach cancer and urinary tract cancer (Figure 2C). The following HSPs were expressed in cell lines in all eight tumors: HSP90AA1, HSP90AB1, HSPA8, HSP90B1, HSPD1, HSPA5, HSPA9, HSPA1B, HSPA4, HSPH1, HYOU1 and TRAP1. HSPA2 and HSPA1A were mainly expressed in lung cancer and breast cancer cell lines. HSPA12A was mainly expressed in lung cancer and kidney cancer cell lines. HSPA13 was mainly expressed in lung cancer, stomach cancer and endometrium cancer cell lines. HSPA14 and HSPA4L were mainly expressed in lung cancer and large intestine cancer cell lines. HSPA7, HSPA1L, HSPA12B and HSPA6 were rarely expressed in the cell lines examined.
- Relationships of HSP expressions and interactions in pan-cancer
Previous studies showed that HSPs cooperate with each other to exert critical cellular functions. Therefore, we used TCGA data to analyze the correlations of HSPs expressions in 9 tumors including stomach adenocarcinoma, breast cancer, lung adenocarcinoma, lung squamous cell carcinoma, bladder urothelial carcinoma, colon adenocarcinoma, uterine corpus endometrial carcinoma, kidney renal clear cell carcinoma and kidney renal papillary cell carcinoma. The results revealed mostly positive correlations among the expressions of HSP families in different cancers (Figure 3A). Further analysis results demonstrated that the strongest correlation was observed between HSPA4 and HSPA9 in colon adenocarcinoma (r = 0.94, p < 0.05).
We next used the STRING site to predict the potential PPI network of HSP families. The results showed that there were interactions among HSPs. HSPA4, HSPA5, HSPH1, HSPA12A and HSPD1 showed the highest degree score (Figure 3B).
- Relationships between HSP expressions and cancer-related pathways
The association of HSP expressions with cancer-related pathways was analyzed. The results demonstrated that HSP family proteins mainly participate in the fatty acid metabolism pathway, oxidative phosphorylation pathway, G2M checkpoint pathway, MTORC1 signaling pathway, mitotic spindle pathway, unfolded protein response pathway, protein secretion, reactive oxygen species pathway, E2F target pathway, MYC target pathway, UV response pathway and xenobiotic metabolism pathway (Figure 4A). We counted the numbers of pathways affected by each HSP; the results suggested that HSPA1L, HSPA12B, HSPA6, HSPA7 and HSPA12A may influence the activity of most cancer-related pathways and thus potently affect tumor development (Supplementary Table 1). In addition, the numbers of activated pathways were more than inactivated pathways affected by HSPs, indicating a cancer promotion role of these genes (Figure 4B).
- Relationships between HSP expressions and immune cell infiltration in pan-cancer
Previous studies reported that HSPs participate in immune cell reactions(23, 24). We next used Spearman correlation analysis to calculate the association of HSP expressions with infiltration of 22 immune cell types. HSP expressions were closely related to immune cell infiltration in the examined cancers (| r | ≥ 0.3, p < 0.05), except for uterine corpus endometrial carcinoma, cervical squamous cell carcinoma and endocervical adenocarcinoma, bladder urothelial carcinoma, lung squamous cell carcinoma and ovarian serous cystadenocarcinoma. The immune cell types included M1 macrophages, resting mast cells, M0 macrophages, M2 macrophages and CD4 memory activated T cells (Table 2 and Supplementary Table 1). For example, HSPD1 was positively associated with CD4 memory activated T cells and follicular helper T cells but negatively associated with monocytes in stomach adenocarcinoma, while HSPD1 was positively associated with CD4 memory activated T cells and negatively associated with resting mast cells in lung adenocarcinoma (Figure 4C).
- Relationships between HSP expressions and prognosis in pan-cancer
We analyzed the correlation of HSP expressions with prognosis using gene expression and clinical information in TCGA. All specimens were divided into two groups according to the median expression of HSPs, and log-rank test was performed to evaluate the correlation of expression with prognosis. The expressions of HSP families showed different effects on prognosis in 25 tumors, except for rectum adenocarcinoma, stomach adenocarcinoma, thymoma, prostate adenocarcinoma, pancreatic adenocarcinoma, ovarian serous cystadenocarcinoma and lymphoid neoplasm diffuse large B-cell lymphoma (Figure 5). HSP expressions were related to poor prognosis in sarcoma, acute myeloid leukemia, lung squamous cell carcinoma, uterine corpus endometrial carcinoma, uterine carcinosarcoma, esophageal carcinoma, colon adenocarcinoma, lung adenocarcinoma, bladder urothelial carcinoma, glioblastoma multiforme and lung adenocarcinoma but associated with good prognosis in cholangiocarcinoma, pheochromocytoma and paraganglioma. HSP expressions have different influenced on prognosis in the remaining tumors, such as liver hepatocellular carcinoma, adrenocortical carcinoma, kidney renal clear cell carcinoma, head and neck squamous carcinoma, lower grade glioma, mesothelioma and skin cutaneous melanoma.
- Mutation profiles of HSPs in pan-cancer tissues and cell lines
We next examined the mutation profiles of HSPs using TCGA data. The results showed that mutations of HSPs were mainly present in uterine corpus endometrial carcinoma, colon adenocarcinoma, stomach adenocarcinoma, rectum adenocarcinoma, lung squamous cell carcinoma and lung adenocarcinoma, with a mutation frequency of 0%–23% (Figure 6A). A waterfall plot for the mutation details of HSPs in uterine corpus endometrial carcinoma is shown in Figure 6B. Furthermore, CCLE analysis indicated that the mutation frequency of HSPs in 828 cell lines of 15 tumors was 0%–17% (Figure 6C).
Given that the HSPs often interact synergistically in regulating cellular functions, the mutation correlation among HSP genes were analyzed in uterine corpus endometrial carcinoma of the most common mutations. The results suggested that the mutation could be co-occurred or mutually exclusive among HSPs, but the co-occurrence mutation was more obvious (Figure 6D). The co-occurrence of mutations was frequently observed in HSPA1L-HSP90B1, HSPA1L-HSPA5, HSPA4L-HSPA4 and HSPA4L-HSPA6 pairs in uterine corpus endometrial carcinoma.
CNV analysis demonstrated that HSP family members showed different degrees of gene amplification and deletion in pan-cancer (Figure 7A). For instance, the gene amplification frequency of HSPA6 was 24% in bladder urothelial carcinoma and the frequency of HSP90AA1 gene deletion was 22% in cholangiocarcinoma.
- Relationships between HSP gene variations and expression in pan-cancer
We then examined the correlations of HSP gene variations with expression in cancer. Our results showed that the mutation of 15 HSPs, not including HSPA7, HSPD1, HSPA4, HSPA1B, HYOU1, HSP90B1 and HSPA12B, affected their protein expression in different tumors (p < 0.05, Figure 7B). Notably, the mutation of six HSPs, including HSPA4L, TRAP1, HSPH1, HSP90AA1, HSPA8 and HSPA9, was associated with their protein expression in uterine corpus endometrial carcinoma (p < 0.05).
We further found that CNVs of HSP families impacted their expression levels in different tumors (p < 0.05) (Table 3 and Supplementary Table 2). For example, copy number amplification of HSPA2 was associated with its increased expression in breast cancer, glioblastoma multiforme, head and neck squamous carcinoma, lung squamous cell carcinoma, uterine corpus endometrial carcinoma and ovarian serous cystadenocarcinoma (Table 3). Increased copy number of HSPD1 was associated with increased expression in stomach adenocarcinoma and breast cancer, while its up-regulation was observed in pancreatic adenocarcinoma regardless of whether HSPD1 gene was amplified or deleted (Table 3).
- Relationships between HSP gene variations and prognosis in pan-cancer
We next examined the relationships of HSP gene variations with prognosis in pan-cancer using TCGA data. We found that mutations of HSP family genes were generally associated with good prognosis in uterine corpus endometrial carcinoma, while mutations in these genes were a poor prognostic factor in other tumors (Figure 7C). Furthermore, CNVs of HSPs, except for HSPA6, HSPA7, HSPD1, HSPA4L, HSPA14, HSPA12A, HSP90B1 and TRAP1, correlated with survival period in different cancers. HSPA1A CNVs were correlated with shorter survival time in stomach adenocarcinoma (HR=1.317, OR=1.044–1.661, p=0.02) and esophageal carcinoma (HR=1.386, OR=1.007–1.908, p=0.045). HSP90AA1 CNVs was correlated with shorter survival rate in prostate adenocarcinoma (HR=3.391, OR=1.07–10.75, p=0.038) (Table 4).