Background
Hepatocellular carcinoma (HCC), which accounts for 75–85% of liver cancer cases, is considered the sixth most common malignancy and the fourth with cancer-related death worldwide [
1]. The main causes of liver cancer are chronic infection with hepatitis B/C virus, exposure to aflatoxin, alcohol abuse, and obesity [
2]. HCC is usually associated with poor outcomes because the treatment of HCC could be effective only when diagnosed at early stages [
3]. The prognosis of HCC is currently dependent on histopathological parameters and the tumor staging system. However, such traditional approaches might not be adequate for the accurate prediction of clinical outcomes in HCC patients. Therefore, it is imperative to identify more robust and accurate prognostic indicators that can help clinicians optimize therapeutic strategies.
Autophagy is a natural regulatory mechanism by which cells remove nonessential and dysfunctional components. It is a dynamic process that includes the induction of autophagosomes, their nucleation, double membrane growth and closure, and finally, fusion with the lysosome, which leads to disintegration of the engulfed materials [
4]. Abnormal autophagy has been associated with the pathogenesis of a variety of diseases, including malignant tumors [
5]. In tumors, autophagy can exert opposite environment-dependent effects, which can lead to either suppression or promotion of tumor growth [
6]. Indeed, while autophagy is considered an essential gatekeeper for restricting early tumorigenesis in multiple tissues [
7], defective autophagy has been shown to promote tumor proliferation in several tissues [
8]. In fact, deficiency in autophagy could lead to the release of arginase I from the liver, which leads to the degradation of circulating arginine. Hence, autophagy might maintain cancer growth through circulating arginine [
9].
Recent studies have reported that autophagy plays a crucial role in the pathogenesis of HCC. Indeed, autophagy levels are noticeably higher in HCC tumor tissues, compared with adjacent normal tissues. In addition, the invasion of peripheral areas by HCC tumors has been associated with higher levels of autophagy in HCC cancer cells [
10]. Autophagy promotes HCC cell proliferation through the induction intracellular ATP via mitochondrial oxidative phosphorylation [
11]. Despite that several indexes have been proposed for HCC prognosis [
12‐
14], little studies have considered autophagy-related genes (ARGs) for the prediction of clinical outcomes in HCC patients. Lin et al. reported that an expression signature for ARGs related to survival prediction for HCC patients [
15]. Due to individual differences in HCC patients and the expression levels of relevant genes, additional predictors of HCC prognosis are needed that are not influenced by other clinical characteristics.
Methods
RNA-seq data and clinical information of HCC patients were obtained from The Cancer Genome Atlas (TCGA) database and The International Cancer Genome Consortium (ICGC) dataset. Genes associated with autophagy were extracted from the Human Autophagy Database (HADb), an autophagy-dedicated database that provides information on human genes involved in autophagy.
Functional annotation of differentially expressed ARGs
The R package EdgeR was used to perform differential gene expression analysis on ARGs in the TCGA data. ARGs exhibiting a log2 fold-change > 1 in HCC, compared with non-tumor tissues, and an adjusted P < 0.05 were considered to be significantly altered. Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis was performed using DAVID web-tool (The Database for Annotation, Visualization and Integrated Discovery) to unveil biological attributes and signaling pathways associated with the differentially expressed ARGs. The GOplot and ClusterProfiler R packages were used for visualization of the selected enriched ontologies and pathways.
Construction of the prognostic risk model
Univariate cox regression analysis was used to identify differentially expressed ARGs associated with overall survival (OS) in HCC patients from the TCGA-LIHC data set. The identified OS-related ARGs were then included in a multivariate cox regression analysis to identify potential independent prognostic ARGs in HCC patients. The obtained prognostic ARGs were used to construct a risk score model. The regression coefficients in the linear formula were used as relative weights of ARG genes in the multivariate model. A risk score was calculated for each patient, a median value was identified for all patients, and HCC patients were then divided into low risk (score below the median) and high risk (score above the median) groups. The high and low risk groups were stratified and visualized using Kaplan-Meier (K-M) survival curves and analyzed for statistical significance using the log-rank test. The ARG-based risk score was finally included in a multivariate cox regression of OS to identify its prognostic value in HCC patients.
Evaluation of the prognostic capacity of the model
The survivalROC package was used to analyze the prognostic value of the ARG-based risk model in R environment. The Receiver Operating Characteristic (ROC) curve was used to check the prognostic efficiency of the risk model in survival prediction. An area under the ROC curve (AUC) was used to measure the prognostic efficiency of the model.
Statistical analysis
Data management and statistical analysis were performed using the R software. Plots were created using the R software and GraphPad Prism v7. K-M curves were plotted, and a log-rank test was applied to check for statistical differences between survival curves. A P < 0.05 was used as a threshold for statistical significance.
Discussion
The role of autophagy in maintaining genome integrity and cellular metabolism and homeostasis has been well demonstrated; however, its prognostic significance in human malignant tumors has not been fully explored [
16,
17]. Autophagy can maintain the survival of tumor cells under stress, and hence, promote tumor progression. Despite that endogenous tumor factors and exogenous interventions to promote or suppress autophagy have been proposed as potential cancer treatments [
4], autophagy-targeting cancer therapies remain controversial. Previous studies have reported that differential translation of autophagy-related transcripts may lead to malfunctional autophagosome in HCC cells [
18]. Autophagy activation can promote the proliferation of HCC cells through JNK1/Bcl-2 signaling [
19]. In addition, autophagy can promote metastasis through Wnt/β-catenin signaling [
20] and via the induction of epithelial-mesenchymal transition [
21]. Autophagy is considered an important mechanism of drug resistance by supporting the survival of cancer cells under metabolic and therapeutic stress [
22]. In fact, sorafenib, the only drug approved for the treatment of advanced HCC, may promote autophagy in HCC cells through cellular protein networks. Luo et al. reported that the combination of PSMD10 and Atg7 could be used as a prognostic predictor in HCC patients receiving sorafenib therapy [
23]. In addition, the expression level of the autophagy-related marker LC3 has been associated with poor outcomes in HCC patients receiving surgical resection [
24].
In this study, the high-throughput transcriptomics data of HCC patients were analyzed to identify potential prognostic ARGs. A total of 62 ARGs were differentially expressed in HCC patient tumor samples, compared with normal tissues, including 58 up-regulated and 4 down-regulated genes. Univariate cox regression analysis was then performed on these genes to identify 34 ARGs that were correlated with OS of HCC patients. Of these, 5 risk-associated differentially expressed ARGs (HDAC1, RHEB, ATIC, SPNS1 and SQSTM1) were further identified using multivariate cox regression analysis and were used to construct a prognostic model for the risk-stratification of HCC patients based on a weighted risk score. Survival analysis showed that low-score groups exhibited better OS, compared with patients in high-score group. The multi-target ROC curve was then performed to validate the prognostic significance of the model, which was further analyzed for its correlation with clinical parameters of HCC patients. Previous work revealed that the 3 ARGs BIRC5, FOXO1 and SQSTM1 were associated with OS in HCC patients. HCC patients were stratified based on pathological stage [
15]. Furthermore, results suggest that the risk score could significantly stratify HCC patients based on their histological and T-based staging systems.
HDAC1, a member of the histone deacetylase (HDACs) family, has been shown to play a crucial role in the epigenetic regulation of key oncogenes through the form a closed chromatin structure via histone deacetylation. A growing line of evidence has shown that HDAC1 could affect various oncogenic processes, such as cell proliferation and invasion, in multiple malignant tumors. The down-regulation of homeobox A10 has been shown to inhibit the proliferation of HCC cells and induce cell cycle arrest through the regulation of HDAC1 expression [
25]. In addition, the transcription factor Yin-Yang 1 has been reported to reduce sensitivity of HCC cells to treatment by inducing HDAC1 expression [
26]. Furthermore, miR-34a was demonstrated to inhibit cellular proliferation and induce apoptosis by down-regulation of HDAC1 expression in HCC cells [
27]. A meta-analysis showed that high expression of HDAC1 is associated with poor OS in gastrointestinal and lung cancers, which indicates that HDAC1 may serve as a prognostic signature in these malignancies [
28,
29].
Our results showed that RHEB, a key regulator of mTOR signaling, exhibited a high expression level in cancer samples, compared with normal and adjacent normal samples. Previous analysis of cancer cytogenetic and transcriptomic databases indicated that RHEB mRNA expression was up-regulated in different carcinoma histotypes and was associated with poor outcomes in multiple types of malignancies [
30]. Besides, RHEB expression was associated with higher cancer stages, higher mortality, tumor differentiation and pathological satellites in patients with hepatitis B-related HCC [
31,
32].
Previous studies have reported that ATIC is a bifunctional protease that catalyzes the last two steps in the purine biosynthesis pathway. Depletion of ATIC or suppression of its transformylase activity significantly decreased the survival rate of cells in clonogenic survival assays, which indicates that ATIC may promote the proliferation and migration in cancer cell lines [
33]. Indeed, suppression of ATIC expression significantly inhibited the ability of HCC cells to proliferate and migrate through the regulation of the AMPK-mTOR-S6K1 signaling pathway. Therefore, in line with our results, the high expression of ATIC could be positively correlated with adverse prognosis in HCC patients [
34].
SQSTM1 has been reported as a potential oncogene in various cancers, including HCC. p62, the gene product of SQSTM1, is a versatile protein that acts as an adaptor that induces the degradation of specific active molecules through autophagy [
35]. Wei et al. reported that SQSTM1 contributes to the development of autophagy-deficient cancers via NF-kappaB pathway. Therefore, targeting autophagy and the autophagy-associated SQSTM1 gene expression could be exploited for developing more effective cancer treatments [
36]. Indeed, phosphorylated SQSTM1/p62 has been shown to accumulate in the HCC tumor region, while its inhibitor may inhibit cell proliferation and resistance to anticancer agents [
37]. Furthermore, multiple studies reported that SQSTM1 could serve as a novel prognostic biomarker in multiple cancers types, including nasopharyngeal carcinoma, lung cancer, oral squamous cell carcinoma, and HCC [
38‐
41].
SPNS1 (Spinster homolog 1) is a hypothetical lysosomal H
+-carbohydrate transporter that functions in late stage macroautophagy in vertebrates [
42]. In this study, SPNS1 showed the greatest contribution to outcome predictions compared to the other 4 genes analyzed. In addition to OS, K-M analysis for DFS showed that high levels of SPNS1 also correlated with shorter DFS time (
p = 0.013). Yanagisawa et al. reported that upregulation of SPNS1 regulates luminal solute compositions, thereby altering the subcellular distribution of lysosomes and the accumulation of p62 [
43]. Dysregulation of autophagy lysosomes may promote the invasion and migration of HCC [
44].
In the study presented here, we demonstrate the relationship between drug sensitivity of 17 HCC cell lines and the relative expression levels of risk-associated ARGs using the GDSC database. Even though many of these drugs are not in clinical use, identifying correlations between risk-associated ARGs and drug sensitivity may identify putative therapeutic biomarkers for further validation. Alterations in cancer genomes can influence clinical outcome to anticancer treatment. However, many cancer drugs already used and under development are not associated with specific genomic markers that can guide clinical application to maximize patient benefit [
15]. In present study, we postulate that HDAC1 is a potential therapeutic target for HCC patients since high HDAC1 expression was associated with elevated drug resistance. By molecularly stratifying patient populations, drug sensitivity information can optimize the design of clinical trials and ultimate success of anticancer treatment.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.