The expression of HMGCS2 is repressed in not only inflamed tissues but also remission tissues from patients with UC
Previous study indicated that HMGCS2 repression takes place in inflamed tissues from UC patients(Martín-Adrados et al., 2023). However, few studies focused on the expression of HMGCS2 in UC patients with remission phase. Here, the microarray data analysis of GSE38713 and GSE59071 datasets showed that the expression of HMGCS2 was downregulated in inflamed as well as remission tissues of UC patients (Fig. 1A and B), suggesting that although UC patients have been clinically diagnosed as in remission phase, persistent mucosal damage remains uncorrected.
HMGCS2 knockdown downregulated fatty acid metabolism pathway and its leading edges
Intestinal epithelial cells provide the first line of defense in the colonic mucosa. There is evidence showing that HMGCS2 knockdown induced endoplasmic reticulum stress in Caco-2 cells(Martín-Adrados et al., 2023). To determine whether HMGCS2 is significant for epithelial cell damage, we generated Caco-2 and HT-29 cells lacking HMGCS2 by lentivirus infection. Western blot analysis validated efficient reduction of HMGCS2 protein level in Caco-2 and HT-29 cells (Fig. 1C and D), indicating the successful establishment of HMGCS2 knockdown cell lines. The RNA of Caco-2 and HT-29 cells lacking HMGCS2 was then extracted and subjected to RNA sequencing analysis. Differentially expressed genes (DEGs) analysis in each group were performed by applying the criteria p-value < 0.05 using DESeq2 package in R software. Volcano maps in Fig. 1E visualized DEGs in Caco-2 and HT-29 cells with HMGCS2 knockdown as compared to control groups. A total of 559 DEGs (199 down-regulated and 360 up-regulated) and 11030 DEGs (5357 down-regulated and 5673 up-regulated) DEGs were identified in Caco-2 and HT-29 cells with HMGCS2 knockdown, respectively.
To reveal the pathway dysregulation affected by HMGCS2 knockdown of Caco-2 and HT-29 cells, we employed the R-based fgsea package for Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis. All the significant KEGG pathways affected in Caco-2 and HT-29 cells were selected using a cut-off threshold of p-value < 0.05 (Fig. 1F). The results showed that cytokine cytokine receptor interaction pathway was upregulated in both Caco-2 and HT-29 cells after KEGG knockdown. More interestingly, fatty acid metabolism pathway was significantly downregulated in both cell lines (Fig. 1G), indicating that HMGCS2 might play a critical role in the damage of epithelial cells via influencing the fatty acid metabolism pathway.
The Hub HMGCS2−FA gene panel was established in vitro and was validated in human data.
Encouraged by above findings, we further identified potential hub genes for epithelial cell damage associated with HMGCS2 within the leading-edge genes from fatty acid metabolism pathway (Fig. 2A, heatmap). By overlapping the leading-edge genes between Caco-2 and HT-29 cells, we identified 10 candidate genes associated with HMGCS2 (HubHMGCS2−FA) (Fig. 2A). To further investigate the role of the HubHMGCS2−FA gene panel in UC patients, we analyzed the gene expression pattern as well as the alteration of fatty acid metabolism pathway in GSE38713 and GSE59071 datasets.
The dataset GSE38713 comprises a total of 36 colonic mucosal samples, including 15 samples from the inflamed regions during the active phase of UC, 8 samples from the remission phase of UC and 13 samples from healthy controls. The gene expression level of HubHMGCS2−FA panel was shown in Fig. 2B. KEGG pathway analysis using fgsea package in R software revealed that fatty acid metabolism pathway was significantly downregulated in not only the inflamed tissues (Fig. 2C) but also the remission tissues (Fig. 2D) of UC patients in GSE38713. GSVA analysis was also performed to estimate fatty acid metabolism pathway of every single sample of UC patients. Then, we created a correlation heatmap using the corrplot package to display the connections between the ten genes of HubHMGCS2−FA panel and the fatty acid metabolism pathway, and the correlation strength was indicated as a score. The results showed that HMGCS2 was positively correlated with fatty acid metabolism pathway and multiple pairs of genes from HubHMGCS2−FA panel in inflamed tissues of UC patients (Fig. 2E). What`s more, the expression of HMGCS2 in remission tissues was also significantly positively correlated with fatty acid metabolism pathway and multiple pairs of genes from HubHMGCS2−FA panel (Fig. 2F).
We then included dataset GSE59071 to validate the above observations. GSE59071 comprises a total of 108 samples with 74 samples from inflamed regions during the active phase of UC, 23 samples from the remission phase of UC and 11 samples from healthy controls. The gene expression level of HubHMGCS2−FA panel was shown in Fig. 3A. KEGG pathway analysis revealed that fatty acid metabolism pathway was significantly downregulated in not only the inflamed tissues (Fig. 3B) but also the remission tissues (Fig. 3C) of UC patients in GSE59071. The correlation heatmap showed that HMGCS2 was also positively correlated with fatty acid metabolism pathway and multiple pairs of genes from HubHMGCS2−FA panel in inflamed tissues of UC patients in GSE59071 (Fig. 3D). Also, the expression of HMGCS2 in remission tissues was significantly positively correlated with fatty acid metabolism pathway and multiple pairs of genes from HubHMGCS2−FA panel in GSE59071 (Fig. 2E).
The risk score model HMGCS2−FA showed good performance in predicting UC patients in remission phase from healthy controls.
We established a predictive model with a scoring system that represents the linear combination gene expression values of the HubHMGCS2−FA panel, named risk score modelHMGCS2−FA. The assessment of performance metrics for the risk score modelHMGCS2−FA revealed significantly higher risk for UC patients in not only active phase but also remission phase as compared to that from healthy individuals in both GSE38713 and GSE59071 datasets (Fig. 3F). These data indicated that during the remission phase, UC patients might still have persistent colonic injury and have not achieved fully healing. The specificity and sensitivity of risk score modelHMGCS2−FA in predicting UC patients were verified using ROC curve analysis, demonstrating unbiased prediction of UC patients in active phase as well as remission phase in the microarray datasets using the HubHMGCS2−FA-guided machine learning bioinformatic approach. Based on the risk score modelHMGCS2−FA analysis, the AUC values were 1 and 0.99 for the inflamed tissues from UC as compared to controls in GSE38713 and GSE59071, respectively (data was not shown). More interestingly, the AUC values were 1 and 0.85 in remission tissues of UC patients as compared to controls in GSE38713 and GSE59071, respectively (Fig. 3G). These results collectively indicate that the risk score modelHMGCS2−FA can be used for the prediction of UC in not only active phase but also remission phase from healthy controls. The HubHMGCS2−FA gene panel could be a potential biomarker or therapeutic target for ulcerative colitis patients in active phase or even remission phase.