Identification and analysis of genes related to osteoporosis metabolism
To clarify the metabolism-related DEGs in osteoporotic tissue samples, we applied the GEO-GSE35956 dataset to 944 metabolism-related genes extracted from the MSigDB website (Table S1) for the analysis of metabolism-related DEGs. It was shown that 115 DEGs were screened in this study compared with bone tissue samples from non-osteoporotic patients (Fig. 1A), of which 54 genes were upregulated and 61 genes were downregulated.
Subsequently, KEGG and GO analysis of these 115 DEGs in this study revealed that these genes were mainly involved in nucleotide biosynthetic process, nucleoside phosphate biosynthetic process, mitochondrial matrix, nuclear DNA-directed RNA polymerase complex, in addition to metabolism-related pathways including coenzyme binding, drug metabolism - other enzymes, biosynthesis of cofactors, and biosynthesis of nucleotide sugars (Fig. 1B-E).
Construction Of Wgcna And Identification Of Core Modules
Next, WGCNA construction was performed on the GEO-GSE35956 dataset. The WGCNA program package in the R toolkit was used to construct the weighted co-expression network. In this study, the soft threshold was set to 5 and the scale independence was 0.9, and the average connectivity was relatively high, so that the network was constructed closer to the scale-free network (Fig. 2B, C). Based on this, a hierarchical clustering tree and co-expression module of the WGCNA network were constructed, and 51 gene modules were finally obtained in this study (Fig. 2A), among which the violet module (Fig. 2D), darkmagenta module (Fig. 2E), green module (Fig. 2F) and greenyellow module (Fig. 2G) had the highest correlation with osteoporosis.
In order to screen for metabolic molecules with clinical translation potential, this study intersected key genes from the core modules (violet, darkmagenta, green and greenyellow) with metabolism-related DEGs (osteoporotic tissue vs. non-osteoporotic tissue, GEO-GSE35956) to obtain nine key genes, namely CKM, LIPC, PLCD4, POLR2A, SPHK1, ADI1, IMPA2, NNMT and CYP26A1 (Fig. 2H), of which IMPA2, SPHK1, POLR2A, CYP26A1 and CKM were up-regulated, and NNMT, PLCD4, LIPC, ADI1 were down-regulated in osteoporotic tissues (Fig. 2I).
Identification And Analysis Of Metabolism-related Genes In Osteoarthritis
To elucidate the regulatory mechanism of osteoporosis on osteoarthritis, the obtained key genes of osteoporosis were substituted into the GEO-GSE55235 dataset for analysis in this study. It was shown that CKM and POLR2A were down-regulated in arthritic tissues (Fig. 3A), while LIPC was up-regulated in arthritic tissues (Fig. 3A). Compared with normal synovium tissues, the expression of CKM, POLR2A and LIPC in osteoporotic tissue samples was contradicted (Fig. 2I). Therefore, the 115 metabolism-related DEGs obtained previously inside osteoporosis were substituted into the GEO-GSE55235 dataset for further analysis in this study. We obtained 30 metabolism-related DGEs, and further analysis revealed that the expression trends of UXS1, PAFAH1B2, CERK, PMM1, NME4, GAMT, GGT5, GLUL, ACSL4 and LDHB were consistent in osteoporosis (Fig. 3D) and osteoarthritis (Fig. 3C), i.e., UXS1, PAFAH1B2, CERK, PMM1, NME4 and GAMT were up-regulated both in osteoporosis and osteoarthritis, while GGT5, GLUL, ACSL4 and LDHB were down-regulated both in osteoporosis and osteoarthritis.
Next, we performed protein binding predictions for UXS1, PAFAH1B2, CERK, PMM1, NME4, GAMT, GGT5, GLUL, ACSL4, and LDHB using the PPI website. The study revealed the possibility of protein binding of GGT5, LDHB and GLUL (Fig. 3E). Further molecular docking also indicated that GGT5 may bind to LDHB (Fig. 3F) and LDHB may bind to GLUL (Fig. 3G). These results suggest that these genes may have an impact on the course of osteoarthritis in the context of osteoporosis.
Screening Of Metabolic Signature Genes In Osteoarthritis
To further focus on factors with clinical translational potential, the LASSO algorithm and SVM algorithm were combined to analyze these 10 metabolism-related genes separately. The LASSO algorithm obtained 8 candidate genes (UXS1, PAFAH1B2, CERK, PMM1, NME4, GAMT, GGT5 and LDHB) (Fig. 4A). The SVM algorithm obtained 10 candidate genes (Fig. 4B). The intersection of the two algorithms yielded 8 candidate genes (Fig. 4C), which were UXS1, PAFAH1B2, CERK, PMM1, NME4, GAMT, GGT5 and LDHB. Further AUC analysis of these 8 signature factors showed that all of these genes had AUC values greater than 0.7 or higher (Fig. 4D).
Next, using the GEO-GSE55457 dataset (osteoarthritis) as a validation dataset, the study showed that the expression of UXS1 and GAMT was upregulated in osteoarthritic synovial tissues compared to controls (Fig. 5A). Meanwhile, AUC analysis of these 8 alternative signature factors in the GEO-GSE55457 dataset showed that the AUC values of UXS1 and GAMT were 0.950 and 0.780, respectively. Therefore, we focused on UXS1 and GAMT from 8 metabolic signature genes in the further research.
Uxs1 And Gamt Genes Affect Pyroptosis In Osteoarthritis Tissues
52 pyroptosis related genes were obtained from the MSigDB website and literature reports[20] (Table S2). GEO-GSE55235 dataset was used to compare the pyroptosis related genes in synovial tissues of healthy people and patients with osteoarthritis. Ten pyroptosis-related genes showed differential expression in synovial tissues from patients with osteoarthritis compared to controls (Fig. 6A), with CASP1, CASP8, GPX4 and PYCARD showing up-regulation in osteoarthritic tissues (Fig. 6B), while IL-1β, PLCG1, IRF1, NOD1, CYCS and IL-6 showed down-regulation in osteoarthritic tissues (Fig. 6B), suggesting that these genes may be involved in the development of osteoarthritis.
To clarify whether UXS1 and GAMT were involved in regulating cellular scorching in osteoarthritis, this study analyzed the correlation of UXS1 and GAMT with these 10 above-mentioned genes. Our study showed a moderate negative correlation (R < -0.4) between UXS1 and IL-1β (Fig. 6C); a moderate negative correlation (R < -0.4) between GAMT and IL-1β (Fig. 6D), and a moderate positive correlation (R > 0.4) between GAMT and PYCARD (Fig. 6D). In conclusion, our study suggested that UXS1 and GAMT may affect pyroptosis in osteoarthritis by regulating IL-1β and PYCARD expressions. Further regulatory mechanisms need to be experimentally verified.
Uxs1 And Gamt Genes Affect Necroptosis In Osteoarthritis
We obtained 67 necroptosis-related genes from the MSigDB website (Table S3) and applied the GEO-GSE55235 dataset to compare the differential expression of necroptosis-related genes in synovial tissues from normal healthy people and patients with osteoarthritis. Compared with controls, FADD, TLR3, CASP8, MAP3K7, MYC, TNFRSF1B, PANX1, MAP3K7, CFLAR, SPATA2, IDH2, KLF9, HSP90AA1, BNIP3, BCL2L11, EGFR, TARDBP, and TNFRSF21 genes in osteoarthritis patients showed differential expression in synovial tissues (Fig. 7A, B).
To explore the correlation of UXS1 and GAMT with above-mentioned 18 genes, we found that UXS1 was moderately negatively correlated with KLF9, MYC, BNIP3 and BCL2L11 (R <-0.4), and moderately positively correlated with MAP3K7 (R > 0.4) (Fig. 7C). Meanwhile, GAMT was moderately negatively correlated with KLF9, MYC, BNIP3, TARDBP and BCL2L11 (R <-0.4), and moderately positively correlated with MAP3K7 and IDH2 (R > 0.4) (Fig. 7D). Among them, KLF9, MYC, BNIP3, TARDBP, BCL2L11 and MAP3K7 were correlated with UXS1 and GAMT (Fig. 7C, D). These results suggest that UXS1 and GAMT may affect necroptosis in osteoarthritis, and further regulatory mechanisms need to be experimentally verified.
Identification And Validation Of Regulatory Mechanisms Of Uxs1 And Gamt
In this study, the upstream miRNAs and lncRNAs of UXS1 and GAMT genes were analyzed by TargetScan, miRanda and miRDB databases. A total of 36 UXS1-related miRNAs and 1 GAMT-related miRNA were intersected by these 3 databases (Fig. 8A). Then, prediction of lncRNAs upstream of UXS1 and GAMT genes were performed based on known miRNAs using the spongeScan database, and a total of 43 lncRNAs were obtained. In addition, ceRNAs were constructed for these predicted miRNAs and lncRNAs. The study showed that hsa-miR-18a-3p, hsa-miR-342-3p, hsa-miR-145-5p and hsa-miR-146a-3p were the core node genes of UXS1 (Fig. 8A, blue).
To explore the downstream regulatory mechanisms of UXS1 and GAMT genes, we applied the GEO-GSE55235 dataset for GSEA analysis. The GSEA enrichment analysis showed that UXS1 gene was correlated with cytokine-cytokine receptor interaction, gap junction, oxidative phosphorylation, p53 signaling pathway, Parkinson disease and proximal tubule bicarbonate reclamation (Fig. 8B). Meanwhile, GAMT genes was associated with other enzymes of drug metabolism, glycolysis gluconeogenesis, lysosome, natural killer cell mediated cytotoxicity, PPAR signaling pathway and toll like receptor signaling pathway (Fig. 8C).