Differential expression of SNRP members in patients with HCC
Firstly, it was determined that genes for SNRP members are located on definite genomic sites [25,26] (Table 1). We analyzed the transcript levels of the entire cohort of 1858 assays in various cancer types using the ONCOMINE database [16] (Fig. 1). The results showed that significant unique analyses were found in the SNRPB (26 tests), SNRPD1 (32 tests), SNRPD2 (22 tests), SNRPD3 (12 tests), SNRPE (39 tests), SNRPF (15 tests) and SNRPG (23 tests) groups, respectively. SNRPB was significantly elevated in tumor tissues, especially in bladder, cervical, colorectal, gastric, head and neck, kidney, liver and breast cancers, with 2,136 samples. In addition, the levels of each SNRP member were high in patients with gastric and colorectal cancers. Compared with other cancers, liver cancer had the large sample size and strong mRNA expression of SNRPB, SNRPD1, SNRPD2, and SNRPE. In contrast, high levels of SNRPD3 are rarely found in tumor tissue.
The results of ONCOMINE [16] described that the transcript levels of SNRPB, SNRPD1, and SNRPD2 were significantly elevated in HCC tissues [25,27]. But SNRPD3, SNRPF, and SNRPG had no data set to study. For SNRPE mRNA levels, it was upregulated in all three datasets of the TCGA database [24-27]. It was summarized in Figure 1 and Table 2. We also compared the transcript levels of SNRPs between HCC and normal tissues by using UALCAN [17] (Fig. 2a-2g). We found that SNRPB, D1, D2, D3, E, F, and G were all upregulated in tumor tissues. Besides, we analyzed the correlation among different genes in HCC tissues and determined that SNRPB and SNRPD1 had the highest correlation (Fig. 2h). In conclusion, our results showed that transcriptional expression of SNRPB, SNRPD1, SNRPD2, SNRPD3, SNRPE, SNRPF, and SNRPG was overexpressed in HCC patients. Table 3 shows multiple comparisons of significant unique analyses or median expression in HCC patients using ONCOMINE and UALCAN servers. SNRPB, SNRPD2, SNRPD3 and SNRPF; SNRPD1 and SNRPD2; SNRPD1, SNRPE and SNRPG were analyzed by best gene ranking percentile, with no differences. In addition, there was no significance between SNRPB and SNRPE; SNRPD1, SNRPD2, SNRPD3, SNRPF, and SNRPG by median expression.
Correlation between mRNA expression and tumor stages of SNRP members
We used GEPIA [18] to analyze the correlation between mRNA expression and cancer stage in patients treated with HCC. Although there were significant differences between SNRPB, D1, D2, D3, F and G groups at stages I, II, III and IV, there were no differences between SNRPE groups and tumor stage (Fig. 3). That is, the mRNA expression of SNRPB, D1, D2, D3, F and G had a significant relationship with the cancer stage of the patients and appeared higher in advanced stage cancers.
Prognostic value of SNRP members in patients undergoing HCC
Prognostic significance of mRNA expression, including OS, PFS, and RFS was under observation. It could be found that patients were classified as low (black) and high (red) risk based on their respective OS thresholds (Fig. 4). High levels of SNRPB, SNRPD1, SNRPE and SNRPG mRNA levels suggest a trend towards worse OS, but without significant differences. The cutoff values distinguishing between the high and low groups based on automatic selection of the best cutoff [20], can be seen in Fig. S1.
Increases in SNRPB, SNRPD1, SNRPD2, and SNRPG were associated with poor PFS (Fig. S2). The relevant cutoff values can be seen in Fig. S4. Furthermore, high mRNA levels of SNRPB SNRPD1, SNRPD2, SNRPE, and SNRPG led to shorter RFS (Fig. S5), while no similar findings were found for SNRPD3, and SNRPDF. The cutoff value for each member can be showed in Fig. S7.
There are several known prognostic factors for HCC, such as age, gender, WHO grade, p_T stage, and p_TNM stage. It was necessary to examine whether each member could independently predict prognosis. Univariate Cox analysis presented seven genes were positively associated with survival prognosis (Fig. 5, S3, S6). Moreover, p_T and p_TNM stages were also significantly related to OS (Fig. 5a), PFS (Fig. S3a), and RFS (Fig. S6a). Subsequent multivariate Cox regression analysis indicated that SNRPD1 was significantly correlated with OS (Fig. 5b), but failed to be an independent prognostic factor for PFS (Fig. S3b) or RFS (Fig. S6b). In conclusion, high levels of SNRPB, SNRPD1 and SNRPG were associated with prognosis of OS, PFS and RFS. Details are shown in Table 4.
Genetic alterations and correlations of SNRP members
We analyzed genatic mutations and interactions in HCC patients by cBioPortal [21] (INSERM Cancer Cell 2014 dataset, MSK Clin Cancer Res 2018 dataset, INSERM Nat Genet 2015 dataset, MSK PLOS One 2018, AMC Hepatology 2014 dataset, RIKEN Nat Genet 2012 dataset, TCGA Firehose Legacy dataset, TCGA PanCancer Atlas dataset). The results (TCGA PanCancer Atlas dataset) illustrated that the percentages of gene alterations were 0.27% mutations (1/372), 0.53% deep deletions (2/372), 7.53% amplifications (28/372) in all SNRP members, respectively (Fig. 6a). The frequency of alterations in SNRPs b was analyzed by using the AMC Hepatology 2014 dataset, TCGA Firehose Legacy dataset, and TCGA PanCancer Atlas dataset (SNRPB, 0.5%; SNRPD1, 0.3%; SNRPD2, 0.8%; SNRPD3, 0.3%; SNRPE, 5%; SNRPF, 0.2%; SNRPG, 0.2%) (Fig. 6b). Then, we further presented the survival results based on genetic alterations. Unfortunately, we did not find significance among genetic alterations, OS or PFS, respectively (p = 0.792, 0.662, Fig. 6c, 6e). However, disease-free survival (DFS) was significant in the genetically altered and unaltered groups (Fig. 6d). The reasons why genetic alterations in SNRPs and prognosis did not seem to be associated with prognosis might be related to the context of the study in the database, the material methodology and the small sample size. Table 5 summarizes the OS or DFS data for the genetically altered and unaltered groups, including the total number and number of events. Only the DFS data for SNRPE had a P value <0.05.
Functional enrichment analysis of SNRP members
To understand the function of SNRP members and their neighboring proteins, we used GO and KEGG pathways by Metascape [24]. The result indicated two major GO-biological processes (Fig. 7a), spliceosomal snRNP assembly (GO:0000387) and spliceosomal complex assembly (GO:0000245). U1 snRNP binding (GO:1990446) was only associated with the molecular function (Fig. 7b). The top 4 GO enrichments (Fig. 7c) were cellular components: methylosome, plCIn-Sm protein complex, U5 snRNP, and U7 snRNP. The top 2 KEGG enrichments (Fig. 7d) were structural complexes: Sm core complex; pathway: systemic lupus erythematosus.
Interaction of correlated genes and proteins of SNRP members
We briefly illustrated the correlation of SNRP members at the genatic level by Gene MANIA online tool [22] (Fig. S8a). The results demonstrated SNRP members share genetic thresholds very closely. Markedly, relationships were found among SNRP members regarding the PPI network. We used STRING [23] to determine the correlation of SNRP members at the protein expression level (Figure S8b).