WGCNA construction of a co-expression network
Using the microarray data from the GSE6477 dataset for pre-processing, we removed a sample (GSM148937) determined to be an outlier and clustered the remaining 161 samples (Fig. S1). Based on the soft-thresholding power β = 7 (R2 > 0.85), a scale-free network a scale-free network was generated to analyze co-expression (Fig. 1a-b). Based on an average hierarchical clustering algorithm, seven co-expression modules were identified (Fig. 1c). In these co-expression modules, 1098 genes (turquoise), 1199 genes (blue), 273 genes (green), 173 genes (red), 87 genes (black), and 43 genes (magenta) were grouped together. Moreover, co-expression modules were not found for 727 genes (gray).
Identification of specific modules related to MM development
Turquoise module eigengene (869 genes) correlated strongly with NDMM (Pearson’s R2 = 0.35, P = 5e-6). Genes within this module were upregulated in NDMM and relapsed MM samples, downregulated in normal and MGUS samples, while their expression remained unchanged in smoldering MM samples. Conversely, genes the green module (213) were downregulated in NDMM (Pearson’s R2 = -0.34, P = 9e-6) and relapsed MM samples (Pearson’s R2 = -0.2, P = 0.01), significantly upregulated in normal (Pearson’s R2 = 0.67, P = 1e-22), while their expression remained unchanged in MGUS and smoldering MM samples. Based on the P-value for each module, we concluded that the development of MM was significantly influenced by the green and turquoise modules (Fig. 1d).
In addition, scatter plots of module membership versus gene significance (turquoise and green modules) also revealed a high correlation between genes within these modules and disease stages (Fig. 1e-f). Heat maps depicting all genes' expression levels and eigengene values for the turquoise and green modules also illustrate the trend related to different stages of the disease (Fig. 1g-h). Therefore, the turquoise and green modules were considered to be the key modules.
Analysis of GO and KEGG enrichment of genes in the turquoise and green modules
Utilizing the WGCNA package's GO enrichment function, we examined the biological significance of genes in both turquoise and green modules. Genes in the turquoise module were primarily participated in regulating biological processes, including mitochondrial and organelle inner membrane, ribosome biogenesis, structural constituent of ribosome, and mitochondrion (Fig. S2a). The Gene in the green module were largely involved in regulating immune system process, immune response, extracellular space, defense response, and leukocyte activation (Fig. S3a).
Using clusterProfiler, we analyzed the pathways significantly enriched in the turquoise and green modules. Genes in the turquoise module were related to Huntington’s disease, oxidative phosphorylation, ribosome, Parkinson’s disease pathways (Fig. S2b). The green module was mainly enriched for genes regulating phagosome, tuberculosis, Staphylococcus aureus infection (Fig. S3b).
Hub genes identified in the turquoise and green modules
A CytoHubba plug-in was used in Cytoscape to examine each gene's hub score, so that hub genes could be identified with high confidence (Supplementary Table S1). The top 10 hub genes of the turquoise module were UQCR11, ATP5O, NPM1, COX6B1, ETFA, FBL, GPI, SNRPE, SSBP1, and AHCY. The top 10 hub genes of the green module were HMOX1, MMD, CSTA, FABP4, FRMD4B, FTL, HLA-DPB1, ITGB2, LILRB5, and MS4A6A.
Survival analysis
Based on another independent dataset GSE24080, our aim was to determine whether the identified hub genes predict survival time for MM patients based on their expression levels. Fig. S4-6 showed the survival analysis results for the 20 hub genes. Four hub genes (ATP5O, NPM1, SNRPE, and SSBP1) with high expression in the turquoise module and four hub genes (HMOX1, FABP4, ITGB2, and MS4A6A) with low expression in the green module were associated with worse overall survival (OS). Furthermore, three hub genes (NPM1, SNRPE, and SSBP1) with high expression in the turquoise module and two hub genes (HMOX1 and FTL) with low expression in the green module were associated with poor event-free survival (EFS). These results suggest that the hub genes NPM1, SNRPE, SSBP1, and HMOX1 are significantly related to both OS and EFS of patients with MM (Fig. S4). Among these hub genes, SSBP1 was identified as a critical regulator for further study due to its highest intramodular connectivity (kWithin = 44.001).
Validating the expression of SSBP1 in MM cell lines and patients on basis of the Oncomine data
Utilizing the Oncomine database, we determined SSBP1 mRNA expression in MM. As a result of the cancer versus cancer analysis, SSBP1 mRNA expression data were found in three datasets (Barretina CellLine, Wooster CellLine, and Garnett CellLine) derived from different tumor cell lines. The SSBP1 mRNA levels in MM were significantly higher than those in other common cancers across all three datasets (Barretina CellLine, Wooster CellLine, and Garnett CellLine) (Fig. 2a-c). The meta-analysis of these three datasets also revealed excessive expression of SSBP1 in MM (Fig. 2d). Based on the Agnelli Myeloma 3 and Zhan Myeloma 2 datasets, the meta-analysis of cancer versus normal also indicated that patients with MM or plasma cell leukemia had higher SSBP1 levels than healthy individuals (Fig. 2e). Kaplan-Meier analysis based on the Mulligan dataset and Carrasco dataset found that myeloma patients with high SSBP1 expression exhibited shorter OS than those with low SSBP1 expression (both P < 0.01; Fig. 2f-g).
Relationship between bone marrow SSBP1 protein expression and clinical features of MM
We assessed the bone marrow histopathological specimens from patients with NDMM admitted to our hospital between October 2013 and August 2018. Our cohort included 54 patients with NDMM, of whom 6 had relapsed. Among them, 24 patients were women and 30 were men. In total, 3 cases were classified as International Staging System (ISS) stage I, 19 as stage II, and 32 as stage III. Follow-up lasted an average of 38 months.
Immunohistochemical analysis indicated that SSBP1 was heterogeneously expressed in patients with NDMM (Fig. 2h). A higher level of SSBP1 expression was observed in RMM than in NDMM (Fig. 2i; P = 0.017). SSBP1 expression was significantly higher at relapse than at the beginning of the treatment (Fig. 2j, P = 0.041). The median H-score of 34 was chosen as the threshold value to separate patients with NDMM into high (n = 27) and low (n = 27) expression groups. As indicated in Table 1, increased SSBP1 expression is correlated with elevated serum calcium and lactate dehydrogenase levels. As shown in Fig. 2k, kaplan-Meier analysis revealed that patients with high SSBP1 expression had a shorter OS compared to those with low expression (HR = 2.384, 95% CI: 1.001-5.679; P = 0.044) . The results suggest that elevated SSBP1 expression contributes to the malignant growth of MM.
Effects of SSBP1 inhibition on MM cell proliferation and apoptosis
By using CRISPR/Cas9, the expression of SSBP1 protein was disrupted in MM cells. Western blotting results (Fig. 3a) showed that the lentivirus SSBP1-sgRNA2 was most effective in suppressing SSBP1 expression, and hence, was used in subsequent experiments. Results showed that suppressing SSBP1 expression inhibited proliferation in two MM cell lines, RPMI8226 and MM1.1S (Fig. 3b). According to flow cytometry analysis, the SSBP1-sgRNA2 group had significantly more apoptotic cells and cells had been arrested in the G1 phase (Fig. 3c-d).
Furthermore, comparatively to either the GFP or control group, the SSBP1-sgRNA2 group expressed significantly more Bax, cleaved caspase 3 and caspase 9, as well as PARP, while Bcl-2 expression was significantly lower (Fig. 3e). According to these results, the targeted disruption of SSBP1 expression promotes apoptosis and inhibits proliferation, thereby impairing the growth of MM cells.
Suppression of SSBP1 expression activates the p38MAPK pathway by promoting mtROS accumulation and inducing mitochondrial damage
Next, we tried to explore the mechanism by which SSBP1 regulates MM cell apoptosis. As SSBP1 regulates mitochondrial function, its dysregulation may lead to mitochondrial damage and accumulation of mtROS. After SSBP1 was downregulated in MM cells, mtROS levels increased significantly (Fig. 4a-b). Analysis of western blots revealed that the SSBP1 downregulation resulted in a decrease in Cyt c levels in mitochondria accompanied by an increase in Cyt c levels in cytoplasm (Fig. 4c), indicating that the mitochondrial membrane was damaged.
KEGG-related GSEA analysis showed that SSBP1 regulated downstream pathways, including MAPK signaling pathways and protein processing in the endoplasmic reticulum, which were associated with cancer development (Fig. 4d). We also examined whether SSBP1 downregulation affects MAPK pathway activation in MM cells. SSBP1 disruption cells showed a significant increase in phosphorylated p38 levels, but this had no effect on JNK and ERK phosphorylation (Fig. 4e). These results suggest that suppression of SSBP1 expression induces mtROS accumulation and mitochondrial damage in MM cells, resulting in p38MAPK pathway activation.
Mito-TEMPO and p38MAPK inhibitor treatment ameliorates MM cell apoptosis caused by SSBP1 downregulation
To validate the involvement of the mtROS/p38MAPK signaling pathway in MM cell apoptosis induced by SSBP1 downregulation, we used Mito-TEMPO (a mitochondria-specific antioxidant) and SB203580 (a p38MAPK specific inhibitor). As shown in Fig. 5a, both Mito-TEMPO and SB203580 reversed the effects of SSBP1 downregulation on MM cell apoptosis. In addition, Mito-TEMPO reduced the mtROS levels in SSBP1 knockout MM cells; however, SB203580 had no effect on mtROS (Fig. 5b). As expected, both Mito-TEMPO and SB203580 partly reversed p-p38MAPK and apoptosis-related proteins (Bcl-2, Bax, cleaved caspase 3, cleaved caspase 9, and PARP) in cells with downregulated SSBP1 expression (Fig. 5c-d). Taken together, our results confirm that SSBP1 downregulation induces apoptosis by promoting mtROS accumulation and activating the p38MAPK signaling pathway.
SSBP1 downregulation inhibits tumor progression in vivo
To determine the effects of SSBP1 inhibition on the tumorgenesis of MM, we created a human myeloma xenograft model. As shown in Fig. 6a SSBP1-sgRNA2 mice displayed significantly less tumor growth during days 14–21 compared to GFP mice (P < 0.001). In the GFP group, mean tumor volume at day 21 was approximately one-third that of the SSBP1-sgRNA2 group (328.5 mm3 ± 79.5 mm3 in GFP group vs. 89.7mm3 ± 15.7 mm3 in SSBP1- sgRNA2 group; P < 0.01). As visualized by immunohistochemistry examination, the accumulation of Ki67 was consistent with the changes of tumor volume and SSBP1 level (Fig. 6b). Then we performed western blots and founded that SSBP1 disruption enhanced a significant increase in proteins (Bax, cleaved caspase 3, and cleaved PARP), while decreasing Bcl-2 expression (Fig. 6c-d). In vivo, these findings indicate that disrupting SSBP1 expression in myeloma cells could inhibit tumor progression.