HCC is the second most common cause of cancer-related deaths.[21] Surgical resection is the most effective first-line treatment for specific patients. The 5-year recurrence-free survival rate after partial hepatectomy is only 48.4%.[22, 23] For HCC, most studies that have published lncRNAs signatures associated with prognosis have focused on primary HCC, while few have focused on relapse HCC and thus, finding effective biomarkers for HCC is crucial.
Data mining strategies can be used to explore significant biological phenotypes associated with high-dimensional datasets. TCGA and GEO databases, with large-scale genomic analyses, can evaluate the molecular features associated with cancer outcomes. Recent developments in next-generation sequencing technologies have greatly expanded our knowledge on non-coding RNAs, and these non-coding RNAs are considerably more abundant than mRNAs.[24] Many studies have revealed the role of lncRNAs in cancer development, indicating their potential as novel biomarkers for cancer diagnosis and prognosis.[25–27] To better understand the molecular markers of relapse HCC, we comprehensively analyzed the database and identified lncRNAs that can be used to predict OS and RFS. Herein, we first retrieved GSE101432 dataset containing data pertaining to the HCC primary and relapse tumors, and matched non-tumor tissues from the GEO database, following which we analyzed the differential expression patterns of lncRNA and mRNA.
We identified 1218 lncRNAs and 1676 mRNAs upregulated in the primary tumor samples and 1540 lncRNAs and 1787 mRNAs upregulated in the relapse tumor samples. The results of PCA revealed a significant difference for recurrent HCC between lncRNA and mRNA. Additionally, we found that 1540 of the 2261 (68.1%) lncRNAs were upregulated and 508 genes were co-upregulated between the two groups, suggesting upregulated lncRNAs play a more important role in relapse HCC than downregulated lncRNAs.
LncRNA-mRNA co-expression combined with GO enrichment analysis showed that lncRNAs were related to cell adhesion, cell-cell junction organization, adherens junction organization, cell junction assembly, and pattern specification process, suggesting that lncRNAs are related to the development of cancer. In the relapse tumor group, the co-expression network showed that upregulated lncRNA and co-expressed DE mRNA are involved in processes related to cancer metastasis; this creates conditions to promote the relapse of HCC. CTD-2369P2.5 co-expressed protein coding gene ICAM1 that is associated with HCC has a high level of expression, and this expression is regulated by lncRNAs. In HCC, ICR specifically affects cancer stem cell properties of ICAM-1(+) HCC cells and lncRNA ICR contributes to portal vein tumor thrombus development. [28]
WGCNA technology has been leveraged to identify upregulated gene modules associated with relapse of HCC. Additionally, WGCNA has been used to analyze lncRNA and mRNA by clustering genes with similar expression patterns. [29] It considers the expression of all genes evaluated in the experiment to reveal co-expressed gene clusters (modules), which are likely also co-regulated. If some genes are co-expressed in the control group but not expressed in the pathological samples, it can be assumed that the regulatory mechanism has changed, which may be the cause or result of the disease. Therefore, the genes in these modules can play a role in cancer and are therefore considered potential therapeutic targets or diagnostic/prognostic biomarkers. [30] Herein, we employed WGCNA by assessing co-expressed gene networks in the GSE101432 dataset, leading to the identification of 15 gene co-expression modules. Based on the p- and R-values, we chose four modules namely: black, green-yellow, pink, and yellow, and performed GO enrichment analysis on them. Of these four modules, the yellow and black modules were most closely linked to cell proliferation, differentiation, and survival, as well as some transcription-related biological processes; this indicates that there are multiple abnormal cell activities in relapse HCC. These two modules were thus suspected to be crucial regulators of relapse HCC, and were therefore analyzed further.
Given that the co-expression network and WGCNA were found to be closely associated with the relapse of HCC, we next screened for hub genes that showed higher expression and correlation in relapse tumors. We found that the expression of RP11-334A14.8, RP4-738P11.4, TRBV6-6, LINC00668, and LINC00941 was higher in the relapse tumor group than in the primary tumor group. We then assessed the expression of these genes via qRT-PCR in a separate set of primary-HCC patient and relapse-HCC patient clinical samples, which confirmed that LINC00668 and LINC00941 expression levels were increased in both HCC tumor groups, whereas PHACTR2 expression was reduced in these samples relative to control tissue levels. LINC00668 and LINC00941 and their co-expression mRNA were selected using correlation analysis with R > 0.1 and p < 0.05 as the selection threshold values. Therefore, we selected BAIAP2L2, MDFI, ZNF750, and SLC44A5 as LINC00668 co-expression mRNAs. Besides, GPR160, LOX, NXPH4, OTX1, GMNN, and MICB were LINC00941 co-expression mRNAs. As a result, BAIAP2L2, MDFI, LOX, OTX1, and MICB were highly expressed in both the groups. This means that LINC00941 is co-expressed with LOX, OTX1, and MICB and is related to the relapse of HCC. Furthermore, LINC00668 was found to be co-expressed with BAIA02L2, KCTD17, and NDUFA4L2, which has biological significance for the relapse of HCC.
To the best of our knowledge, LINC00941 has been reported to exhibit protumorigenic and prometastatic behaviors during tumorigenesis, such as colorectal cancer (CRC) and gastric cancer. [31, 32] Wu et al. found that LINC00941 activates the TGF-β/SMAD2/3 axis in metastatic CRC, which provided new insight into the mechanism of metastatic CRC and a novel potential therapeutic target for advanced CRC. This is a potential marker for recurrent HCC. As with LINC00941 co-expression mRNAs, expression of OTX1 has been found to be positively correlated with nodal metastasis status (p = 0.009) and TNM staging (p = 0.001) in HCC tissues.[33] Moreover, overexpression of OTX1 promotes the proliferation, migration, invasion, and tumor angiogenesis of HCC cells. [34] MICB is highly expressed in HCC, and its expression level is significantly and negatively associated with TNM stages. [35] Among the patients with different stages of hepatitis, asymptomatic individuals have higher MICB serum levels, while liver cirrhosis patients have lower MICB serum levels (p < 0.0001) compared to those in other patient groups. [36] The lysyl oxidase (LOX) family members are secreted copper-dependent amine oxidases, which are characterized by catalytic activity that contributes to the remodeling of the cross-linking of the structural extracellular matrix. [37] Umezaki et al. found that high LOX expression was associated with epithelial-mesenchymal transition (EMT) markers and predicted early recurrence and poor survival in patients with HCC. This is supported by our findings, which indicate thatLINC00941 and co-expression mRNA are potential biomarkers and therapeutic targets for HCC relapse.
However, another lncRNA, located at ch18p11.31, may play a pivotal role in the relapse of HCC. In HCC, the molecular mechanism indicated that LINC00668 affects cell division, cell cycle, mitotic nuclear division, and drug metabolizing cytochrome P450 enzymes (all p ≤ 0.05). [38] However, the co-expression mRNA, BAIAP2L2, was found to be highly expressed in gastric cancer tumors and its expression significantly correlated with tumor diameter, TNM stage, and lymph node metastasis, respectively. [39] KCTD17 is upregulated in the liver tissues of obese mice and nonalcoholic fatty liver disease patients; however, few studies have been published on its role in cancer. [40] NDUFA4L2 can promote cell migration, invasion, proliferation, and EMT of cancer cells under hypoxic conditions.[41]
In terms of survival analysis, whether in OS or RFS, both LINC00941 and LINC00668, as well as the co-expression mRNAs, may be representative survival markers. In conclusion, our present research developed two-lncRNA signatures for predicting the prognosis of patients with HCC relapse. Our study indicates that the LINC00941 and LIN00668 signatures may involve HCC tumorigenesis, prognostic, metastasis and relapse. These results provide a foundation for future basic research on the mechanism of relapse of HCC.
While the results of this research are interesting and meaningful, they have some limitations. First, some lncRNAs, such as XLOC and TRBV6-6, lacked relevant information in TCGA or other databases; thus, we were unable to conduct further research on them. Second, the WCGNA method is inherently limited by the criteria used for module selection and the network rejection threshold, which may affect the final research results.