br Fig Expression of key cholesterol homeostasis genes in

Fig. 3. Expression of key cholesterol homeostasis genes in tumor and adjacent normal tissues in local cohort. The Echinomycin of A HMGCR, B SREBF2, C LDLR, D SCARB1, E NR1H3, F NR1H2 and G ABCA1 was investigated using qPCR.
Fig. 4. Expression of key cholesterol homeostasis genes in TCGA COAD &READ datasets. The expression of A HMGCR, B SREBF2, C LDLR, D SCARB1, E NR1H3, F NR1H2 and G ABCA1 in tumor and normal tissues.
normal tissues only in the early stage patients suggesting the potential role of LDLR in the early stages of CRC (Fig. 5I). In addition, there was no correlation of tumoral LDLR levels with any other clin-icopathological features.
On the other hand, SREBP2 expression was significantly increased in the early stage (Fig. 5N) and decreased in late stage patients in tu-mors in comparison to the adjacent normal tissues (Fig. 5O). In fact, there was a significant difference in the intertumoral SREBP2 levels between early stage and late stage patients (Fig. 5P). These findings highlight the differential regulation of SREBP2 in the disease progres-sion.
3.6. Correlation of cholesterol homeostasis genes and clinicopathological parameters
As HMGCR, SREBF2, NR1H3 and NR1H2 exhibited similar trend in the local as well as TCGA cohort, these genes were selected for their association with clinicopathological features of CRC. The patients were divided into low and high expression groups according to the median expression value of the genes in cancerous tissues. The analysis of TCGA data revealed a significant correlation of low NR1H2 expression with a number of different pathological features including early pathological TNM stage (*p = 0.011), early pathology T stage (*p = 0.0105), ab-sence of metastasis (*p = 0.033) and absence of lymph node metastasis (**p = 0.0124) (Table 3). These observations suggested that low NR1H2 expression associated with the earlier events of CRC develop-ment. In addition, expression of NR1H3 and NR1H2 associated 
significantly with tumor site (Table 3). Low HMGCR expression corre-lated with advanced pathological TNM staging (*p = 0.0394) thus, implying the association of HMGCR with disease progression (Table 3). There was no significant correlation between SREBF2 expression and any clinicopathological feature (Supplementary Table I). No significant relationship was found between the clinical features and mRNA ex-pression in local cohort.
3.7. Prediction of diagnostic value of cholesterol homeostasis genes
Fig. 5. Protein expression of LXR, LDLR and SREBP2 in CRC tissues: (A-B) Representative Western Blots for LXRs in early stage and late stage patients. (C) Densitometric analysis of LXR levels in adjacent normal and tumor tissues in local cohort. (D) Densitometric analysis of LXR levels in adjacent normal and tumor tissues in the early stage patients; (E) in the late stage patients. (F) Comparison of LXR levels in the tumor tissues of early stage (n = 14) and late stage patients (n = 6) by densitometry. (G) Representative Western Blots of LDLR in tumor and adjacent normal tissues in CRC patients. (H) Densitometric analysis of LDLR expression in adjacent normal and tumor tissues in local cohort (I) Densitometric analysis of LDLR expression in adjacent normal and tumor tissues in early stage patients; (J) in late stage patients. (K-L): Representative Western Blots of SREBP2 in early stage and late stage patients. (M) Densitometric analysis of SREBP2 levels in adjacent normal and tumor tissues in local cohort. (N) Densitometric analysis of SREBP2 levels in adjacent normal and tumor tissues in early stage patients; (O) in the late stage patients. (P) Comparison of SREBP2 levels in the tumor tissues of early n = 4) and late (n = 10) pathology T stage by densitometry. Results are expressed as mean ± SEM.
3.8. Correlation of cholesterol homeostasis genes with overall survival in CRC
Kaplan Meier curves were plotted to find the association between the expression of HMGCR, SREBF2, NR1H3, NR1H2 and the overall survival (OS) in CRC patients. Total 374 CRC patients from TCGA da-tabase were divided into low expression and high expression group based on the median expression. Among all the genes studied, only HMGCR low expression correlated with poorer overall survival in CRC patients (Fig. 7A). However, multivariate analysis revealed that HMGCR could not be used as the independent prognostic marker in CRC (Supplementary Table II). In order to completely understand the re-lationship between the genes and survival, the patients were further divided into various subgroups. It was observed that low HMGCR ex-pression could be a potential prognostic marker in patients with the age ≤ 50 years (Fig. 7B) and only colon as the tumor site (Fig. 7C). Low SREBF2 and NR1H3 expression was correlated to poor OS in patients with age ≤ 50 years (Fig. 7D) and early pathology T stage (Fig. 7E), respectively. However, high tumoral expression of NR1H2 in the late pathology T stage was associated with shorter OS (Fig. 7F). No other significant correlations were found among other subgroups and the gene expression.