Identification of DEGs
We acquired mRNA expression profiles of plasma cell samples from 101 patients of new or relapse multiple myeloma and 15 NDs from the GSE6477. Gene expression distribution of profiles was matched via boxplot analysis (Figure 1A). Statistical analysis was required to validate the comparison between NDs and MM patients. We applied the GEO2R to identify DEGs associated with MM. 1383 DEGs (538 upregulated genes and 845 downregulated genes) were identified in the MM samples compared to NDs’ samples. We set the cut‐off criterion as p < 0.05 and [logFC] ≥ 1 (Figure 1B). The top 100 genes were clustered in the heatmap between patients with MM patients and NDs (Figure 1C).
Functional and pathway enrichment analysis
We performed GO and KEGG enrichment analysis to investigate the functions of DEGs using DAVID. The top GO (Figure 2A-C, Table 1) and KEGG (Figure 2D, Table 2) terms for DEGs were shown. For biological process (BP), DEGs were mainly enriched in immune response, inflammatory response, ECM organization, leukocyte migration, cell adhesion (Figure 2A). DEGs in molecular function (MF) were significantly associated with protein binding, ECM structural constituent, serine type endopeptidase activity, protease binding and receptor activity (Figure 2B). The cellular components (CC) analysis indicated that proteins encoded by DEGs were mostly located in the extracellular exosome, extracellular space, extracellular region and ECM (Figure 2C). KEGG enrichment analysis showed that CAMs, proteoglycans in cancer, ECM-receptor interaction and PI3K/Akt signaling pathway were significantly enriched in DEGs. The top KEGG pathways for DEGs were shown (Figure 2D). Subsequently, we analyzed a pathway interrelation and related genes by examining KEGG enrichment analysis results in ClueGO and CluePedia (Figure 2E). These results suggested that cell adhesion might play a crucial part in the malignant progression of MM.
PPI network and modular analysis of selected genes
To identify key genes in the cell adhesion processes, 180 overlapped genes in DEGs and genes involved with the cell adhesion were screened, including 123 mRNAs and 57 non-coding RNAs (Figure 3A). Then, we constructed a PPI network containing 123 protein, with 123 nodes and 429 edges based on the STRING database and Cytoscape (Figure 3C). 12 hub genes whose degree values ≥ 15 for further analysis were chosen, including ITGB1, FN1, ITGB3, ITGAM, PTPRC, ITGB2, ITGA5, ITGB5, CDH1, IL4, ITGA9 and LAMB1 (Figure 3B). Additionally, Cytoscape displayed a total of eight modules in the default MCODE settings for modular analysis (Figure 3D-K, Table 3).
Diagnostic and prognostic values of hub genes in MM
First of all, we performed the ROC curve analysis among 12 hub genes based on the GSE2658. The results showed that ITGAM, ITGB2, ITGA5, ITGB5, CDH1, IL4, ITGA9, and LAMB1 achieved an AUC value of > 0.7, demonstrating that these eight genes have high sensitivity and specificity for MM, suggesting they can be served as biomarkers for the diagnosis of MM (Figure 4A-H).
Furthermore, we evaluated the influence of all aforementioned 8 hub genes on clinical prognosis using PrognoScan database based on the GSE2658 through K-M curve and log-rank test to identify whether these genes were concerned with the survival of MM patients. ITGA9 (Figure 4I) and LAMB1 (Figure 4J) were significantly associated with DSS in MM by evaluating the correlation between these gene expressions and survival rates.
The relationship between ITGA9/LAMB1 mRNA expression and clinical characters of patients with MM
First of all, the relationship between ITGA9 or LAMB1 mRNA expression and the clinical parameters was analyzed based on both clinical data and mRNA expression data of GSE136324 (Table 4, n=426). The results showed that LAMB1 expression was negatively correlated with β2-MG (Figure 5A, p=0.024, spearman correlation coefficient r = -0.110). Patients with low LAMB1 expression had a tendency of being in high R-ISS stage (Figure 5C, p=0.001). No significant relationship was found between LAMB1 mRNA expression and malignant plasma cells (PCs) in BM (Figure 5B, p=0.9427) and GEP groups (Figure 5D, p=0.1785).
Furthermore, there was a negative correlation between ITGA9 expression and malignant PCs in BM (Figure 5F, p < 0.001, spearman correlation coefficient r = -0.165). Compared with in the standard group, MM patients in the GEP high group had a lower ITGA9 expression (Figure 5H, p = 0.0039). ITGA9 mRNA expression had no correlation with β-2MG (Figure 5E, p=0.323) and R-ISS stage (Figure 5G, p=0.281).
Function and signaling pathways analysis of ITGA9 and LAMB1
To investigate the function of ITGA9 and LAMB1 on MM progression, we performed a comprehensive analysis including GSEA, TF prediction and Spearman correlation analysis.
Firstly, an enrichment map was constructed using genesets which related to cell adhesion with p‐value <0.05 (Figure 6A). Genesets including GO calcium dependent cell cell adhesion via plasma membrane cell adhesion molecules, GO cell adhesion via plasma membrane adhesion molecules, KEGG ECM receptor interaction, NABA basement membranes, NABA core matrisome were enriched in patients with low LAMB1 expression in the GSE2658, suggesting that LAMB1 was involved in the cell adhesion in myeloma cells (Figure 6B-F). Furthermore, cancer microenvironment dn, MANALO hypoxia dn, ELVIDGE hypoxia dn genesets were significantly enriched in patients with low LAMB1 expression in the GSE2658 (Figure 6G-I). For futher study, we found the positive correlation between LAMB1 and HIF-1 mRNA expression in a varity of cancers using GEPIA database (Figure 6J, p = 0.00018, spearman correlation coefficient r = 0.038). The result of TF prediction by the online site PROMO (http://alggen.lsi.upc.es/cgi-bin/promo_v3/promo/promoinit.cgi?dirDB=TF_8.3/) strengthened the correlation that HIF-1 may bind to the promoter of LAMB1 (Figure 6K). Therefore, LAMB1 which may be regulated by HIF-1 played a vital role in myeloma cell adhesion.
ITGA9 also was a cricial cell adhesion molecule since GSEA results showed that genesets including KEGG ECM receptor interaction, GO cell adhesion via plasma membrane adhesion molecules, NABA ECM glycoproteins, GO calcium dependent cell cell adhesion via plasma membrane cell adhesion molecules, GO protein complex involved in cell adhesion, NABA core matrisome, cell-cell adhesion were enriched in patients with low ITGA9 expression in the GSE2658 (Figure 7B-H). An enrichment map was constructed using genesets which related to cell adhesion with p‐value < 0.05 (Figure 7A). Results demonstrated that SCHLOSSER myc targets repressed by serum and hallmark myc targets v1 genesets were enriched in patients with low ITGA9 expression in the GSE2658 (Figure 7I-J). There was a negative correlation between the mRNA expression of ITGA9 and MYC in multiple cancers based on the GEPIA database (Figure 7K, p = 7.7e-19, spearman correlation coefficient r = -0.09). We predicted that MYC may bind to the promoter of ITGA9 and repressed its expression based on TF prediction by PROMO (Figure 7L). In conclusion, ITGA9 was a vital cell adhesion molecule in myeloma which may be negtive regulated by MYC.
ITGA9 and LAMB1 expression and clinical outcome association in different cancers
We explored the expression and prognosis of ITGA9 and LAMB1 mRNA in different cancers based on the Oncomine database and the PrognoScan database respectively.
LAMB1 is highly expressed in most cancers while low in some cancers including breast cancer, leukemia, ovarian cancer and prostate cancer (Figure 8A). Furthermore, LAMB1 was significantly associated with OS and DSS respectively in AML, DLCBL, glioma, NSCLC, colorectal cancer (Fig 8B-G) and colorectal cancer, breast cancer (Fig 8H-I).
Compared with ND, the transcript levels of ITGA9 indicated significant low expression in bladder cancer, brain and central nervous system cancer, breast cancer, leukemia, liver cancer, lung cancer and etc., suggesting that the down-regulation of ITGA9 was common in various types of cancer (Fig 9A). ITGA9 was significantly associated with OS and DSS respectively in AML, B cell lymphoma, lung cancer, colorectal cancer and esophagus cancer (Fig 9B-F) and colorectal cancer, breast cancer (Fig 9G-H). The details including hazard ratio (HR) with 95% confidence intervals(CI) and p-values were shown in the Table 5.