Expression analysis of BRIP1
In the current work, we aimed to investigate the multimolecular characteristics and role of BRIP1 (Genome location: chr7(q32.1), consensus CDS: CDS11631.1, Figure S2a). As presented in Figure S2b, conserved domain of DEAD_2 (pfam06733) was commonly consisted in BRIP1 protein structure among different species. The evolutionary relationship of the BRIP1 protein among various species was presented in phylogenetic tree (Figure S3).
We first used the ONCOMINE database to compare BRIP1 mRNA expression between tumors and the adjacent normal tissues (Fig. 2a). Higher expression of BRIP1 were noticed in Brain and CNS cancer, Breast cancer, Cervical cancer, Colorectal cancer, Gastric cancer, Head and Neck cancer, Pancreatic cancer, and Sarcoma than in the corresponding normal tissues. However, significantly downregulated expression of BRIP1 were also noticed in several tumors, the details were shown in Table 1. Collectively, 22 datasets showed a higher mRNA expression of BRIP1 in different tumors than that in normal samples, while 5 datasets showed controversy results, higher expression of BRIP1 were found in the normal tissues. We then assessed the levels of BRIP1 protein expression across different cancers in datasets from Human Protein Atlas (HPA). As shown in Figure S4a, most tumor tissues showed moderate to strong nuclear or nuclear membranous staining, especially in Colorectal cancer, Head and neck cancer, Carcinoid, Urothelial cancer, Prostate cancer, and Melanoma. The expression of BRIP1 in different tissues, blood cells, and brain tissues under the normal physiological state were also evaluated. As presented in Figure S4b, highest expression of BRIP1 was identified in Thymus, followed by Testis, and Bone marrow. Combined the three datasets (FANTOM5 (Function annotation of the mammalian genome 5), GTEx, and HPA) together, we found a BRIP1 expression in all included tissues; however, RNA tissue enhanced (lymphoid tissue) phenomenon was noticed (Figure S4b). Similarly, BRIP1 expression was detected in all blood cells, and phenomenon of low RNA blood cell type specificity was noticed, with the consensus datasets of HPA, Monaco, and Schmiedel (Figure S4c). While in brain tissues, highest expression of BRIP1 was found in Basal ganglia, followed by cerebral cortex, and olfactory region (Figure S4d).
Since the cancer types that can be used to explore the association of BRIP1 expression between tumors and normal tissues are limited on ONCOMINE, we also utilized TIMER2 database to analyze the differential expression of BRIP1 across all TCGA tumors. Figure 2b showed higher BRIP1 expressions in cancer tissues of BLCA (Bladder Urothelial Carcinoma), BRCA (Breast invasive carcinoma), CHOL (Cholangiocarcinoma), COAD (Colon adenocarcinoma), ESCA (Esophageal carcinoma), GBM (Glioblastoma multiforme), HNSC (Head and Neck squamous cell carcinoma), KIRC (Kidney renal clear cell carcinoma), KIRP (Kidney renal papillary cell carcinoma), LIHC (Liver hepatocellular carcinoma), LUAD (Lung adenocarcinoma), LUSC (Lung squamous cell carcinoma), STAD (Stomach adenocarcinoma), THCA (Thyroid carcinoma), UCEC (Uterine Corpus Endometrial Carcinoma) (P < 0.001), CESC (Cervical squamous cell carcinoma), and READ (Rectum adenocarcinoma) (P < 0.01) than in the corresponding normal tissues. As there are several cancer types in TIMER2 lacking the data of normal tissues, we take the data of normal samples from the GTEx dataset to further explore the differential BRIP1 expression between adjacent normal tissues and cancer tissues of DLBC (Lymphoid neoplasm diffuse large B-cell lymphoma), SARC (Sarcoma), THYM (Thymoma), and UCS (Uterine Carcinosarcoma), and higher BRIP1 expressions were noticed in the cancer tissues (Fig. 2c, P < 0.05). However, no statistical significance was detected for other cancers (Figure S5a). Besides, the results of pooling analysis from various reports in the ONCOMINE database also verified that BRIP1 is highly expressed in breast cancer, sarcoma, colorectal cancer, and head & neck cancer (Figure S6). Additionally, the correlation between BRIP1 expression and cancers with different pathological stages was investigated using the “Pathological Stage Plot” module of HEPIA2. Significantly differences were found in ACC (Adrenocortical carcinoma), BRCA, SKCM (Skin Cutaneous Melanoma), KIRP, LIHC, LUSC, KIRC, THCA, UCS, KICH (Kidney Chromophobe), and OV (Ovarian serous cystadenocarcinoma) (Fig. 2e). Cancers without significance were shown in Figure S5b-S5d.
Moreover, we analyzed BRIP1 expression in different molecular and immune subtypes. As shown in Figure S7a, significantly different BRIP1 expression was observed in various molecular subtypes of BRCA, LGG, PCPG (Pheochromocytoma and Paraganglioma), COAD, LUSC, STAD, HNSC, KIRP, OV, UCEC (P < 0.001 for all), LIHC (P < 0.01), SKCM, and ESCA (P < 0.05 for all). Figure S7b showed that significant difference of BRIP1 expression exist across immune subtypes of C1 to C6 (represent would healing, IFN-γ dominant, inflammatory, lymphocyte deplete, immunologically quiet, and TGF-β dominant, respectively) in BLCA, LGG, SARC, BRCA, LUAD, SKCM, COAD, LUSC, STAD, ESCA, OV, THCA, KICH, PCPG, UCEC, KIRC, READ, and KIRP. Of interest, lowest BRIP1 expression was noticed in subtype C3 in most cancers, except for LGG and KIRC (BRIP1 expression in C5 is the lowest). For cancer types with no significant difference were presented in Figure S8. Jointly, the differential expression of BRIP1 in various molecular and immune subtypes may contribute to the differing role of BRIP1 in the prognosis of different tumors.
After exploring the differential expression patterns of BRIP1 between tumors and normal tissues, we also used the HPA dataset to examine the protein expression patterns of BRIP1 in breast cancer, lung cancer, colorectal cancer, liver cancer, and prostate cancer. As shown in Fig. 2d, high expression of BRIP1 was found in the tumor tissues, whereas low to medium expression of BRIP1 was noticed in the adjacent normal tissues.
Survival analysis of BRIP1
According to the expressional levels of BRIP1, we divided the cases into two groups (low and high expression groups) to explore the correlation of gene expression and the survival status of patients across different tumors. As performed in Fig. 3a, high expression of BRIP1 was associated with worse OS (Overall Survival) prognosis for ACC, KIRP, LGG, LUAD, MESO (Mesothelioma), and PAAD (P < 0.05 for all). While high BRIP1 expression was linked to better OS prognosis of COAD, READ, STAD, and THYM (P < 0.05 for all). Data of Disease-free Survival (DFS) analysis indicated that high expression of BRIP1 was correlated to poor DFS prognosis of ACC, LGG, LIHC, PAAD (Pancreatic adenocarcinoma), and THCA (P < 0.05 for all) (Fig. 3b).
Besides, evidence from Kaplan-Meier plotter tool showed that high BRIP1 expression was associated with poor OS (P = 0.026), PPS (post-progression survival) (P < 0.01), and DMFS (distant metastasis-free survival) (P < 0.001) prognosis for breast cancer (Figure S9a). For ovarian cancer, high expression of BRIP1 was correlated to worse OS (P = 0.027) and PFS (progress-free survival) (P = 0.036) of patients (Figure S9b). Similarly, high BRIP1 expression was linked to poor OS, FP (first progression), and PPS (P < 0.001 for all) prognosis for lung cancer (Figure S9c). Conversely, high expression of BRIP1 was correlated to better OS (P = 0.024), FP (P < 0.01) and PPS (P < 0.001) prognosis for gastric cancer (Figure S9d). In addition, positive correlation was found between high BRIP1 expression and worse OS (P = 0.021), PFS (P < 0.01), RFS (relapse-free survival) (P = 0.047), and FP (P < 0.001) prognosis for liver cancer (Figure S9e).
We also used the Sangerbox tool to evaluate the independent prognostic role of BRIP1 across all TCGA tumors. As shown in Figure S10, BRIP1 could serve as independent prognostic biomarker to predict the OS of patients for PCPG, ACC, KICH, LGG, READ, MESO, LIHC, KIRP, PAAD, UCEC, PRAD (Prostate adenocarcinoma), and LUAD; to predict the DSS (disease-specific survival) of patients for PCPG, ACC, KICH, LGG, KIRC, COAD, MESO, LIHC, KIRP, PAAD, PRAD, and LUAD; to predict the DFI (disease-free interval) of patients for THCA, LIHC, KIRP, and PAAD; and to predict the PFI (progress-free interval) of patients for UVM (Uveal Melanoma), PCPG, ACC, KICH, LGG, THCA, MESO, LIHC, KIRP, PAAD, PRAD, and LUAD (P < 0.05 for all). Survival analysis data of Sangerbox showed that the AUC (area under curve) for 1-year, 3-year, and 5-year was moderate to high in predicting the OS of patients for ACC, COAD, KIRP, LGG, LUAD, MESO, PAAD, READ, STAD, and THYM (Figure S11).
Moreover, we utilized the the univariate and multivariate Cox regression analysis to calculate the prognostic factors of OS for KIRP, ACC, LGG, COAD, READ, STAD, THYM, LUAD, MESO, and PAAD (Table S1). Clinical characteristics and BRIP1 expression were calculated. In addition, we combined the parameters with statistical significance from univariate analysis to construct the prognostic nomograms for predicting the 1-year, 3-year, and 5-year survival probability for the above cancers (Figure S12). Collectively, we can conclude that in most cancers, the high expression of BRIP1 is significantly correlated to the worse prognosis of patients.
Enrichment analysis of BRIP1
Based on the online tool of String and GEIP12, 50 targeting BRIP1-binding proteins and 100 BRIP1-correlated genes were selected to further explore the multimolecular characteristics and role of BRIP1 in the carcinogenesis, progress, and prognosis of various tumors. Figure 4a showed the PPI network of the 50 proteins. The top 100 most correlated gene of BRIP1 were identified using GEPIA2. As shown in Fig. 4b, BRIP1 was mostly correlated to CLSPN, FANCI, DTL, BRCA1, and TMPO, with correlation coefficient as 0.74, 0.74, 0.72, 0.71, and 0.71, respectively (P < 0.001 for all). In addition, positively correlation between the BRIP1-correlated five genes and BRIP1 was noticed in numerous of cancers (Fig. 4c). Combine the two sets of genes together, we found three commonly members, TOPB1, BRCA1, and BARD1 (Fig. 4d).
Enrichment analysis of KEGG and GO were conducted among the combination of the two sets of genes. As shown in Fig. 4e, pathways of “Basal transcription factors”, “Homologous recombination”, and “Nucleotide excision repair” might partially explain the role of BRIP1 on tumorigenesis of different cancers. Results from GO enrichment analysis further verified that most BRIP1-related genes are associated with DNA metabolism cellular biology or pathways (e.g., DNA replication, DNA repair, chromosomal region, single-stranded DNA binding, ATPase activity, etc.) (Fig. 4e).
Mutation, Methylation and Genome-wide association of BRIP1 analysis
We utilized multiple online databases to obtain the genetic alteration information of BRIP1 across different tumors. Patients with UCEC presented with the highest alteration frequency of BRIP1 (near 10%), and the major type for BRCA patients is the “amplification” of CAN, with an alteration frequency of 8% (Fig. 5a). The 3D structure of BRIP1 was performed in Fig. 5b. Figure 5c showed the case number, sites, and types of BRIP1 alteration, and the primary genetic alteration type of “missense” mutation was observed, with the number of 182. Besides, A745T/V alteration in the Helicase_C_2 domain was found in two cases of UCEC and one case of HNSC. Moreover, the correlation between BRIP1 alteration and clinical prognosis for BRCA was investigated. As shown in Fig. 5d, the alteration of BRIP1 was associated with poor prognosis in DSS (P = 0.0252), but not in PFS, OS, and DFS (P > 0.05 for all).
Using database of MEXPRESS, the potential correlation between BRIP1 expression and methylation level was explored. As shown in Table S2, significantly correlation was observed between BRIP1 expression and methylation for 24 TCGA tumors, including BLCA, BRCA, CESC, COAD, DLBC, GBM, HNSC, LAML (Acute Myeloid Leukemia), LGG, LIHC, LUAD, LUSC, OV, PCPG, PRAD, READ, SARC, SKCM, STAD, TGCT, THCA, THYM, UCEC, and UCS (P < 0.05 for all). The correlation between methylation and expression of BRIP1 among different cancers was also evaluated by GSCALite database. Generally, negatively correlations between gene expression and methylation across cancers were observed (Fig. 6a).
In addition, mutation frequency of SNV across TCGA tumors was also analyzed on GSCALite tool. As shown in Fig. 6b, we found a high SNV in many cancers, including UCEC, SKCM, COAD, BLCA, READ, CESC, LUAD, ESCA, and HNSC. The constitute of Heterozygous/Homozygous CNV of BRIP1 in 33 TCGA tumors were presented in Fig. 6c. It is worth noticing that heterozygous amplification is the major type of CNV in KIRP, while the homozygous deletion type of CNV is the primary type in KICH. Besides, the correlation between CNV and mRNA expression of BRIP1 were also explored, and the results were performed in Fig. 6d. Notably, highest correlation was observed for BRCA and LUSC, which indicated the BRIP1 expression is significantly regulated by CNV in BRCA and LUSC.
Besides, we further explored the genome-wide association of BRIP1 mRNA in cancer by analyzing the correlation between BRIP1 and other genes using the Regulome Explorer. As shown in Fig. 7, positively correlation was observed in multiple cancers, and the detailed information was presented in Table S3-S17. To further understand the multimolecular characteristics and role of BRIP1 in the tumorigenesis of cancers, we further analyzed the association between BRIP1 expression and inhibition or activation of ten major signaling pathways, the detailed information for calculating the pathway score was described at length elsewhere(22). As shown in Figure S13, BRIP1 was highly correlated to the activation of apoptosis, cell cycle, and DNA damage response, and inhibition of hormone ER and RAS/MAPK signaling pathways.
Immune analysis of BRIP1
As a prominent part of tumor microenvironment, tumor-infiltrating immune cells are complexly interacted with the carcinogenesis, development, and prognosis of cancer(23–25). As one of the most abundant stromal cells populated in the tumor microenvironment, Cancer- associated fibroblasts (CAFs) are essentially involved in the progression of cancers(26, 27). In the current study, we utilized multiple algorithms to explore the association between immune cells’ infiltration level and the expression of BRIP1 across different cancers. Generally, we observed a series of positively correlation between the expression of BRIP1 and the CAFs estimated infiltration value for CESC, ESCA, HNSC, HNSC-HPV-, KICH, KIRP, LGG, LIHC, LUAD, MESO, OV, PAAD, PRAD,THCA, and UCS (Fig. 8a). While negatively correlation for PRAD was also noticed with different algorithm. Additionally, significantly positive correlation was noticed between the BRIP1 expression and CD8+T-cells immune infiltration for HNSC-HPV+, KIRC, LUAD, and THYM based on most algorithms (Figure S14a). Figure 8b and Figure S14b showed the scatter plots of the above cancers generated by one algorithm. As an example, BRIP1 expression in CESC is positively associated with the infiltration level of CAFs (Rho = 0.231, P = 1.02e-04) based on the EPIC algorithm (Fig. 8b). Besides, the relationship between BRIP1 expression and MSI (Microsatellite instability)/TMB (Tumor mutational burden)/neoantigen of all TCGA cancers were investigated. We observed positive correlations between BRIP1 expression and MSI for GBM, LUSC, UCEC, COAD, STAD, KIRC, READ, and KICH but noticed a negative correlation for DLBC (P < 0.05 for all) (Figure S15). BRIP1 expression is also positively correlated to TMB for ACC, LUAD, PRAD, UCEC, COAD, STAD, SKCM, KIRC, and KICH, but is negatively correlated to KIRP (Figure S16, P < 0.05 for all). Besides, only positive correlation was found between neoantigen and BRIP1 expression for PRAD, LUAD, BRCA, UCEC, and STAD (Figure S17, P < 0.05 for all). Additionally, statistically significant correlation of BRIP1 expression and immune checkpoints and pathways across most of TCGA tumors was observed, and the heatmaps was presented in Figure S18. The above findings are worthy of further in-depth research to explore its clinical value.