High expression of SGOL2 in HCC
We found that SGOL2 mRNA expression was upregulated in different tumors, including hepatocellular carcinoma, colorectal cancer, and breast cancer, in the Oncomine database (Figure S1A). In addition, we further searched the GEPIA database to systematically assess the expression profile of SGOL2 in a variety of carcinomas (Figure S1B). To better understand its expression levels in different diseases, including HCC, cirrhosis, and dysplasia, we selected and analyzed data from the Wurmbach liver dataset (Figure 1A). The results showed that SGOL2 expression was dramatically upregulated in HCC tissues compared with normal tissues (P<0.05), whereas there were no significant differences among the cirrhosis, dysplasia, and normal liver tissue groups. Moreover, SGOL2 protein levels were analyzed using the HPA database. As shown in Figure 1B, SGOL2 was weakly expressed in the tumor tissue (Patient ID: 3402) and negatively expressed in normal liver tissue (Patient IDs: 2429/2279). Furthermore, we analyzed the protein levels of SGOL2 in HCC and matched adjacent non-tumor tissues in cohort 2 by Western blots and found that SGOL2 expression was extremely upregulated in HCC (Figure 1C), and immunohistochemistry (IHC) staining from the two independent HCC cohorts confirmed the results (Figure 1D-F). In addition, we divided the patients in cohort 2 into 2 groups according to differentiation levels: a well-differentiated group and a poorly differentiated group. Interestingly, the expression levels of SGOL2 in the less differentiated group were markedly higher than those in the well-differentiated group, indicating that the expression level of SGOL2 is directly proportional to tumor progression (Figure 1F). Similarly, we found the same result: SGOL2 expression in the Grade 3 group was dramatically higher than that in the Grade 1/2 group in cohort 1 (Table 1, p=0.001).
To expand the number of patients included in the analysis, we confirmed the overexpression of SGOL2 in the TCGA-HCC database. We found that the expression level of SGOL2 showed a positive association with grade levels (Figure S3G), consistent with our previous results in cohorts 1 and 2. Moreover, SGOL2 was found to be roughly proportional to the stage levels (Figure S3B). The mRNA level of SGOL2 was higher in the HCC than in the non-tumor tissues (Number of T vs N=371 vs 50) (Figure S3A). We also conducted subgroup analysis in various subgroups (race, sex, age, weight), which showed significantly elevated SGOL2 expression levels (Figure S3C-F). Furthermore, SGOL2 expression was significantly elevated in TP53-mutant patients (Figure S3I). Interestingly, no difference was shown between the HCC patients with and without lymph node metastasis (Figure S3H). Thus, these results indicated that SGOL2 overexpression was related to the development of HCC.
Upregulation of SGOL2 expression indicated poor prognosis in HCC patients
To determine whether SGOL2 could be a novel prognostic marker in HCC, we analyzed its prognostic significance in HCC patients. SGOL2 overexpression was closely related to poor overall survival (HR=2.29 (1.6–3.28), P=3.5e-06), relapse-free survival (HR=1.96 (1.38–2.78), P=0.00013), progression-free survival (HR=2.1 (1.55–2.84), P=9.2e-07) and disease-specific survival (HR=2.84 (1.81–4.47), P=2.3e-06) in HCC patients (Figure 1G, Figure S4A-C). To better identify the negative relationship between SGOL2 expression and the prognosis of hepatocellular carcinoma, we performed survival analysis in verification cohort 1. According to the immunohistochemical score, we divided the patients (n=97) in cohort 1 into a low expression group (n=44) and a high expression group (n=53), and then performed survival analysis. As shown in Figure 1H, we found that the overall survival of the low expression group was markedly higher than that of the high expression group (p=0.001). Subsequently, we conducted a SGOL2-based prognostic model. All HCC patients in cohort 1 were randomly divided into two groups: the training group (n=67) and the validation group (n=30). Next, we identified four variables that were closely related to survival: expression, grade, AJCC TNM, and AFP based on the lasso regression model (Figure 1I-J). Based on the multivariable Cox proportional hazards model, we also predicted the 3-year survival of HCC patients using a nomogram (Figure 1K). We also assessed the discrimination power of this nomogram by receiver operating characteristic (ROCs) curves. The area under the ROC curve for the 3-year survival probability of the training group and the validation group were 0.898 and 0.687, respectively (Figure 1L, Figure S4D-E). The calibration curves of the nomogram showed good probability consistencies between the two groups (Figure S4F-G). In conclusion, high SGOL2 expression indicated a poor prognosis in HCC patients.
The association between SGOL2 and immune cells and mutations in HCC
Next, we assessed the mutation frequency of SGOL2 in liver cancer. The mutation frequency of SGOL2 in HCC was 0.7% (Figure S5A). Thus, we further analyzed the mutation types of SGOL2 in COSMIC. We found that 45.80% of the samples showed missense substitutions, and 13.19% of the samples showed synonymous substitutions (Figure S5B). As shown in Figure S5C, the substitution mutations mostly occurred at C>T (74 samples, 17.96%). Then, we further identified the association between SGOL2 expression and immune infiltration in TIMER. We evaluated the association between SGOL2 expression and immune cell abundances. As shown in Figure S5D, SGOL2 expression showed a weakly positive correlation with tumor purity (Cor=0.148, P=5.68E-03). In addition, SGOL2 expression showed strong positive correlations with various immune cells, particularly macrophages (Cor=0.483, P=2.69E-21) and dendritic cells (Cor=0.47, P=3.99E-20).
SGOL2 promoted malignancy of HCC cells in both vitro and vivo
SK-HEP-1 and HEP3B cells were chosen to evaluate the function of SGOL2 according to the mRNA levels of SGOL2 in different HCC cell lines based on the data from CCLE (Figure 2A). After transfection with lentivirus, we confirmed that SGOL2 was significantly decreased at both the mRNA and protein levels (Figure 2B-C). Transwell assays indicated that low SGOL2 expression suppressed the migration and invasion of HCC cells (Figure 2D-E). Moreover, we tested the expression alterations of key EMT-related proteins responding to the downregulation of SGOL2 expression (Figure 2K). Interestingly, the results showed that shSGOL2 resulted in increased expression of E-cadherin and reduced expression of N-cadherin, fibronectin, vimentin, β-catenin, and MMP9. Thus, downregulation of SGOL2 expression could inhibit cell metastasis by repressing migration, invasion, and EMT in HCC. Next, a sphere formation assay was performed to determine whether stemness could be influenced by downregulating SGOL2 expression. Consistently, the spheres in the shSGOL2 group were dramatically fewer and smaller than those in the shNC group (Figure 2F). Furthermore, the shSGOL2 group developed fewer cell colonies than the NC group (Figure 2G), and we also observed that low SGOL2 expression suppressed the proliferation of HCC cells by CCK-8 assays (Figure 2J). In addition, flow cytometry-based assays demonstrated that apoptotic indices in the shSGOL2 group were dramatically higher than those in the NC group (Figure 2H), and the cell cycle was strongly influenced by the downregulation of SGOL2 expression (Figure 2I). To further verify the role of SGOL2 in HCC in vivo, we constructed xenograft tumor models by SK-HEP-1 shNC and SK-HEP-1 shSGOL2. The mice were sacrificed on Day 21 after inoculation, and the formed tumors, lung, and liver were statistically analyzed (Figure 3A-B). Both the volumes and weights of the formed tumors were dramatically decreased in the shSGOL2 group compared with the shNC group (Figure 3B). We further analyzed angiogenic markers (CD34), proliferative markers (Ki-67 and proliferating cell nuclear antigen [PCNA]), and EMT-related markers (E-cadherin, N-cadherin, vimentin, Snail, and Slug) by IHC. Downregulation of SGOL2 expression resulted in the suppression of both proliferation and metastasis (Figure 3C-D), which was consistent with the above in vitro results. We also found that the apoptotic area in the shSGOL2 group was much larger than that in the shNC group (Figure 3E). Thus, these data indicated that SGOL2 promoted tumor growth and metastasis. To further validate our results, we also tested both the mRNA and protein levels of SGOL1 after the knockdown of SGOL2 in HCC cell lines by PCR and Western blots, respectively. As shown in Figure S2 A-B, we found that the reduction in SGOL2 did not alter the expression of SGOL1. Thus, SGOL2 KD repressed the development of HCC by knocking down SGOL2 but not SGOL1.
MAD2 overexpression was closely related to SGOL2 and indicated a poor prognosis in HCC patients
Next, we tried to clarify the signal transduction pathway of SGOL2 in HCC cells. SGOL2 and MAD2 were reported to be involved in the separation of eukaryotic sister chromatids during the cell cycle[21]. Thus, we hypothesized that SGOL2 promotes tumors by influencing the expression of MAD2. To explore the role of MAD2, a factor closely related to SGOL2, in liver cancer, we used the UALCAN database to analyze its expression profile, clinical value, and prognostic significance. As shown in Figure 4A, MAD2 expression was also markedly upregulated in HCC. Moreover, the expression of MAD2 in HCC showed a positive correlation with SGOL2 in the TCGA database (R=0.78, P=0) (Figure 4B). Interestingly, we observed that high MAD2 expression was also related to unfavorable OS in HCC patients (Figure 4C).
Differentially expressed hub genes associated with SGOL2 and MAD2 and their prognostic significance in HCC patients
We used the data from LinkedOmics for analysis to identify differentially expressed genes related to both SGOL2 and MAD2 in HCC by Spearman’s test (Figure S6A, D). The top 50 positively or negatively correlated markers were represented in Figure S6B-F. Then, the positively correlated genes with coefficients > 0.8 were selected for further analysis. In total, we identified 85 genes positively associated with SGOL2 and 51 genes positively related to MAD2. Among these, 47 genes were positively related to both SGOL2 and MAD2 (Figure 4D). Then, we constructed a PPI network based on the 47 differentially expressed genes using STRING and Cytoscape (Figure 4E) and used it for KEGG enrichment analysis (Figure 4F). The top 15 hub genes of the network were chosen for further analysis using cytoHubba based on the clusters identified in the PPI network using MCODE (Figure 4G). Biological processes, such as cell division, cell proliferation, and apoptotic processes, were significantly affected and enriched based on GO analysis results (Figure 4H). The co-expressed genes were mainly involved in the cell cycle, progesterone-mediated oocyte maturation, oocyte meiosis, and the p53 signaling pathway based on KEGG results (Figure 4I). Then, we tried to assess whether these identified hub genes were related to prognosis. All 15 genes were significantly related to poor OS (BUB1B, NUSAP1, TTK, CCNB2, TOP2A, KIF2C, CCNB1, KIF23, TPX2, KIF11, KIF4A, CDK1, BUB1, CENPE, CDCA8) (Figure S7).
SGOL2 dysregulated the cell cycle process by regulating the MAD2 protein
To validate these data through the bioinformatics analysis above, we further demonstrated the role of SGOL2 in HCC cells, especially in the cell cycle process, based on the above results. First, the protein level of MAD2 was highly declined in HCC cells through downregulation of SGOL2 expression, while overexpression of SGOL2 increased the expression of MAD2 (Figure 5A-B). After the knockdown of SGOL2, the protein levels of PCNA, cyclin D1, and cyclin E1 were significantly decreased (Figure 5A), whereas upregulation of SGOL2 expression strongly increased the expression of PCNA, cyclin D1, and cyclin E1 (Figure 5B). Furthermore, when MAD2 was blocked by its specific inhibitor M2I-1, highly aggressive malignant behaviors of HCC cells caused by overexpression of SGOL2 were significantly reversed (Figure 5C-D). A rescue assay was performed to confirm that the knockdown effect of SGOL2 shRNA could be reversed by the overexpression of MAD2. As shown in Figure 6A-B, the number of migrated or invaded cells in the lower chamber of the shSGOL2+MAD2 group was much more than that of the shSGOL2 group. Moreover, the number of spheres or colonies in the shSGOL2+MAD2 group rose sharply compared to that in the shSGOL2 group, as shown in Figure 6C-D. As shown in Figure 6E, CCK-8 assays demonstrated that the upregulation of MAD2 expression could reverse the inhibitory effect of SGOL2 knockdown on the viability of HCC cells. From the above results, we can conclude that overexpressed MAD2 could reverse the knockdown effect of SGOL2 shRNA in HCC. Altogether, these data indicated that SGOL2 dysregulated the cell cycle and promoted the development of HCC by regulating the MAD2 protein.
SGOL2 exerted its effect by directly binding with MAD2
Next, we examined whether SGOL2 could directly interact with MAD2. Immunofluorescence (IF) staining showed that SGOL2 colocalized with MAD2 in both SK-HEP-1 and HEP3B cell lines by Confocal microscopy (Figure 7A), and the coimmunoprecipitation (Co-IP) assay further verified that SGOL2 could bind with MAD2 (Figure 7B). Altogether, these data collectively verified that SGOL2, binding with MAD2 and forming a SGOL2-MAD2 complex, regulated MAD2 and then fueled tumor cell growth by dysregulating the cell cycle process, which finally promoted the malignant behaviors of HCC cells, including proliferation, migration, invasion, stemness and EMT (Figure 7C).
The predicted transcription factors binding to SGOL2 or MAD2 were obtained from JASPAR
To better understand how SGOL2-MAD2 exerts its effects on HCC, we first searched the JASPAR database to identify the potential transcription factors that might bind with the promoter sequences of both SGOL2 and MAD2. As a result, 22 transcription factors and 16 transcription factors were identified for SGOL2 and MAD2, respectively. Among them, 3 transcription factors were identified to bind not only SGOL2 but also MAD2: ZNF148, PPARG::RXRA, and ETV6 (Figure S8A-B). As shown in Figure S8E-J, we also found that ZNF148 showed a parallel correlation with SGOL2 and MAD2 by TIMER database (SGOL2-ZNF148: rho=0.599, P=1.48e-37; MAD2-ZNF148: rho=0.512, P=3.79e-26). Similar to ZNF148, ETV6 was also positively associated with both SGOL2 and MAD2 (SGOL2-ETV6: rho=0.458, P=1.18e-20; MAD2-ETV6: rho=0.41, P=1.68e-16). However, we failed to find a close correlation between PPARG::RXRA and SGOL2 or MAD2 (SGOL2-RXRA: rho=0.163, P=1.64e-03; MAD2-RXRA: rho=0.071, P=1.71e-01). Thus, ZNF148 and ETV6 may be potential transcription factors that regulate both SGOL2 and MAD2. We also tried to predict the sites in the promoter region of SGOL2 or MAD2 that the two transcription factors may bind with by JASPAR database detection (Figure S8C-D). In conclusion, we demonstrated that the transcription factors ZNF148 and ETV6 may regulate both SGOL2 and MAD2, and their effects on HCC may follow the modulation of ZNF148 or ETV6 expression levels.