Seven Autophagy-Related lncRNAs Associated With Clinical Prognosis in Hepatocellular Carcinoma.


 Background: LncRNA may be involved in the occurrence, metastasis, and chemical reaction of hepatocellular carcinoma (HCC) through various pathways associated with autophagy. Therefore, it is urgent to reveal more autophagy-related lncRNAs, explore these lncRNAs' clinical significance, and find new targeted treatment strategies. Methods: In our study, RNA-seq and clinical data of normal and HCC patients were obtained from the TCGA database, and autophagy genes were obtained from the human autophagy database. Results: The risk prediction model containing seven autophagy-related lncRNAs was constructed. Overall survival (OS) curves show that the high-risk group patients significantly shorter than the low-risk group (P=2.292e-10), and the five years survival rate of the high-risk group (HR 0.286, 95%CI 0.199-0.411) is less than half of the low-risk group (HR 0.694, 95%CI 0.547-0.77). Univariate Cox regression indicated that risk score of the risk prediction model (P<0.001, 95%CI 1.210-1.389 ), T (P<0.001, 95%CI 1.443-2.287), and stage (P<0.001 ,95%CI 1.466-2.408 ) were independent prognostic indicators. However, only the risk score remains the independent prognostic indicator(P<0.001, 95%CI 1.197-1.400 ) based on the multivariate analysis. This risk model's prediction efficiency is significantly higher than other clinicopathological factors for 1-, 3- and 5-year survival rate prediction (AUC are 0.853, 0.794, and 0.764, respectively). Remarkably, the 7 autophagy-related lncRNAs may participate in Spliceosome, Cell cycle, RNA transport, DNA replication, and mRNA surveillance pathway and be related to the biological process of RNA splicing and mRNA splicing. Conclusion: In conclusion, the 7 autophagy-related lncRNAs might be promising prognostic and therapeutic targets for HCC.


Introduction
Hepatocellular carcinoma (HCC), as the most common type of liver cancer [1], with high cancer-related mortality and poor prognosis worldwide [2], makes a threat to public health. Alcohol, a atoxin, and hepatitis can increase HCC risk, and genetic and epigenetic changes can promote malignancy [3.4]. At present, surgical resection remains the most common treatment option for patients with HCC. Because of tumor metastasis and relapse [5,6], the prognosis is dissatisfying. Most patients with advanced HCC often have low 5-year survival [7]. Currently, molecular alterations of HCC have been reported in some studies [8], and those molecular mechanisms can be explored further in the diagnosis and treatment of HCC [9][10][11]. However, these efforts have not made a signi cant improvement in patient survival. Therefore, it is necessary to identify more potential biomarkers that may be related to HCC carcinogenesis.
As a highly selective quality control, mechanism autophagy is the key to maintaining homeostasis in physiological and pathological conditions [12], such as adapting to metabolic stress, removing dangerous cargo, and renovating during differentiation and development, genomic prevention damage Page 3/17 [13]. However, more and more evidence indicated that hat autophagy could also help tumor occurrence, maintenance, and development [14]. In pancreatic cancer [15], autophagy is a metabolic requirement for cancer cells' immune evasion, allowing tumors to achieve optimal proliferation in vitro and vivo. Besides, autophagy is necessary for tumor cell migration and metastasis because its inhibition will block cell migrates and metastasis [16]. In HCC, autophagy can promote tumor cells' metastasis by upregulating the expression of MCT1 and the activation of the Wnt/β-catenin signaling pathway [17]. The epithelialmesenchymal can also be activated by autophagy to promote cancer cell invasion [18]. Thus some researchers have tried to nd new targeted treatment strategies for HCC by studying autophagy pathways.
Long non-coding RNAs (lncRNAs) are a class of RNA transcripts that consist of more than 200 nucleotides in length and exhibit limited protein-coding capacity [19]. However, With in-depth exploration, lncRNA has been found to perform essential functions in various biological processes such as posttranscriptional regulation, transcriptional regulation, and chromatin modi cation [20]. Some studies have also reported that lncRNAs regulate many aspects of cancer progression and affect different malignant behaviors, such as cancer cell proliferation, apoptosis, and metastasis [21,22]. Remarkably, owing to the complexity and diversity, lncRNA's abnormal expression will promote the occurrence of a variety of tumors, such as cervical carcer [23], esophageal squamous cell carcinoma [24], and lung adenocarcinoma [25]. Autophagy has been considered to play a dual and contradictory role in carcinogenesis. The exact mechanisms that result in autophagy in cancer still need to be further explored [26]. lncRNA participates in the occurrence, invasion, metastasis, prognosis, and chemoresistance of HCC by regulating various pathways related to autophagy [27][28][29]. However, these studies only concentrated on single or a few lncRNAs for HCC. Therefore, we used the lncRNA expression pro les of the TCGA database to explore new biomarkers of autophagy-related lncRNAs that are only closely related to and predict patients' prognosis with HCC.

Data obtaining and processing
The entire sequencing pro le data of patients with liver hepatocellular carcinoma (HCC) were obtained from The Cancer Genome Atlas (TCGA, https://cancergenome.nih.gov/) database. According to the gene annotations from the GENCODE project (https://www.gencodegenes.org/) [30], the lncRNA and proteincoding gene pro le data were further classi ed. Besides, the patient's corresponding clinical information was downloaded from the TCGA database, such as survival time, survival status, age, gender, tumor grade, TNM stage. And then, the patients with incomplete clinical information, survival time that less than 30 days and non-hepatocellular carcinoma were removed. Since all of these data involved in this study were publicly available, the ethics committee has no speci c ethical approval.

Screening of Autophagy-Related lncRNAs
A list of the autophagy-related genes was downloaded from the Human Autophagy Database (HADb, http://www.autophagy.lu/). The autophagy gene expression pro le was extracted from the fore mentioned protein-coding gene pro le data. To identify the potential lncRNA related to autophagy-related genes, we performed a Pearson correlation analysis in the lncRNA and autophagy-related gene expression pro le. The thresholds were set as follows: | R | >0.4 and P < 0.001 were considered a strong correlation.
2.3 Identify prognosis-related autophagy lncRNAs and Calculate the Risk Score To con rm the potential prognostic value of autophagy-related lncRNAs, univariate Cox regression analysis and product-limit method (Kaplan-Meier method) were used to assess the association between autophagy-related lncRNA expression and survival data. Those autophagy-related lncRNAs are signi cantly related to survival (Both Kaplan-Meier method and univariate Cox regression satisfy P value<0.01) were obtained as prognosis-related lncRNAs. Those autophagy-related lncRNAs were then used into multivariate Cox regression analysis to obtain regression coe cients (β) with the lowest Akaike information criterion (AIC) values and then establish a risk score. The risk score calculation based as follows = βlncRNA1 × ExpressionlncRNA1 +βlncRNA2 × ExpressionlncRNA2 +…+βlncRNA1n × ExpressionlncRNAn. According to the median risk score, the patients were classi ed into high-risk and low-risk groups. The two groups were assessed using the log-rank test.

Analysis of Risk Score Model
The clinical data and risk score utilized univariate and multivariate Cox regression analysis for evaluating whether the risk score of the autophagy-Related lncRNAs can be an independent indicator for the prognosis. And then, the receiver operating characteristic (ROC) curve and area under the ROC curve (AUC value) were performed to evaluate diagnostic e cacies. Moreover, it assessed the correlation between risk score and clinical data.

Functional Analysis
To further explore the Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, these prognostic autophagy-related lncRNAs may participate. We performed Pearson correlation analysis in the nal prognostic autophagy-related lncRNA and protein-coding gene pro le data. The thresholds were set as follows: | R | >0.4 and P < 0.001. Then we analyzed those co-expressed genes with prognostic autophagy-related lncRNAs through the clusterPro ler package to speculate the pathways and biological processes that lncRNA may participate in.

Acquisition of Autophagy-related lncRNAs
A total of 223 patients with complete clinical information, 14,748 lncRNAs, and 19,767 mRNA were screened from the TCGA-HCC, 232 autophagy genes were obtained from HADb, among which 213 genes were expressed in TCGA-HCC. And then, 557 autophagy-related lncRNAs were identi ed according to the Pearson correlation analysis (Table S1).

Clinical Value of the Prognostic Autophagy-Related lncRNA
To evaluate whether the seven autophagy-related lncRNAs could be used as the independent prognosis biomarkers of patients in HCC, Univariate cox regression analysis, and Multivariate cox regression analysis were utilized to evaluate the relationship between the clinical data and risk score. Univariate Cox regression indicated that risk score (P<0.001, 95%CI 1.210-1.389), T (P<0.001, 95%CI 1.443-2.287), and stage (P<0.001, 95%CI 1.466-2.408) were independent prognostic indicators ( Figure 4A and Table 2). However, after implying multivariate analysis, only the risk score remains the independent prognostic indicator (P<0.001, 95%CI 1.197-1.400) ( Figure 4B and Table 2). Next, the receiver operating characteristic (ROC) curves were utilized to evaluate the risk score's predictive performance. The area under the ROC (AUC) curve of 1-, 3-, and 5-year were 0.853, 0.794, and 0.764, respectively ( Figure 4C). These ndings signi ed that the risk score could be a good factor in predicting HCC patients' prognosis. The risk score increased with stage, and T showed that those autophagy-related lncRNAs might be associated with the progression of HCC (Table 3).

Functional Analysis
Under the inclusion criteria of |R| >0.4 and P <0.001, a total of 3580 genes that have a co-expression relationship with 7 prognostic autophagy-related lncRNAs were obtained (Table S3). KEGG analysis shows that the 7 lncRNAs are directly or indirectly involved in Spliceosome, Cell cycle, RNA transport, DNA replication, Ribosome, mRNA surveillance pathway, and Endocytosis ( Figure 5B). GO results indicate that these lncRNAs may be related to the biological process of RNA splicing, mRNA splicing, RNA localization, covalent chromatin modi cation, and histone modi cation ( Figure 5A).

Discussion
Autophagy, an evolutionary and conservative multistage lysosomal degradation process that promotes metabolism and healthy circulation, plays a complex and contradictory role in tumor formation and cancer treatment [12]. As a subclass of the ncRNAs family, lncRNAs play an indispensable role in various biological processes of tumorigenesis, which are considered a new type of biomarker for cancer diagnosis and prognosis widely concerned [31]. The current researches were mainly focused on the function of single or a few lncRNAs involved in autophagy in HCC patients [32][33][34]. Therefore, it is necessary to explore more autophagy-related lncRNAs to predict HCC patients' prognosis.
On the one hand, as the only bene cial prognostic lncRNA in the prognostic prediction model. ,the gene alias of CTC-297N7.9 is lnc-TMEM220-1, which is an intergenic ncRNA. An HCC study showed that CTC-297N7.9 might be related to cofactor/chromatin/NAD binding and oxidoreductase/DNA-dependent ATPase activity [35]. Besides, due to the speci c low expression and high methylation of TMEM220 in gastric cancer tissues [36], some scholars speculate that CTC-297N7.9 that located upstream of the protein-coding gene TMEM220, may be able to regulate the methylation of TMEM220 or participate in autophagy through its functional proteins, which in turn affects the prognosis of HCC patients [37]. In our research, we have speculated that the highly expressed CTC-297N7.9 may be an inhibitory factor in the progression of HCC. This speculation was con rmed in another study on liver cancer, and the high expression of CTC-297N7.9 often predicts better overall survival and disease-free survival [38] and indicates that CTC-297N7.9 may be one of the critical molecules to improve HCC patients' survival, and it can be further explored in subsequent studies on HCC.
On the other hand, the 6 unfavorable prognostic lncRNAs in the prognostic prediction model have also been attached to various cancers. The o cial full name of PRRT3-AS1 is PRRT3 antisense RNA 1, as a non-protein-coding RNA, which is mainly expressed in liver tissue (RPKM 0.15), fat (RPKM 4.4), prostate (RPKM 3.3), and brain tissue (RPKM 3.0) [39]. In prostate cancer, Fan et al. con rmed that PRRT3-AS1 has a targeting relationship with PPARγ. Its silence can promote apoptosis autophagy and inhibit the proliferation, migration, and invasion of tumor cells through the mTOR signaling pathway [40]. Besides, PRRT3-AS1 is also considered to be related to GBM patients' prognosis [41]. RP11-479G22.8 is also known as lnc-ITGB1-1 in the LNCipedia database [42], and its transcription size is 2051 bp. Through the lncRNA disease prediction module of the lncRNASNP2 database [43], RP11-479G22.8 is closely related to HCC (P < 0.001). Therefore, RP11-479G22.8 is expected to be one of the potential indicators for prognostic prediction in HCC patients [35]. RP11-73M18.8 is a sense-intronic lncRNA with a transcript size of 811 bp, also known as lnc-ZFYVE21-3. Sense-intronic lncRNA is a sequence in the intron of the coding gene on the sense strand. It might harbor different histone modi cation at the transcription start site (TSS) than other ncRNAs [44], which indicated that these intronic lncRNAs maybe the novel biomarkers, such as type 2 diabetes mellitus [45]. LINC01138 is also a member of the sense intron ncRNA, located on chr1. Its abnormal expression has an important in uence on the occurrence and development of several cancers. In prostate cancer (PCa), as a lncRNA that directly target AR, the high expression of LINC01138 can promote the proliferation of tumor cells and inhibit their apoptosis, which indicated that LINC01138 could be a diagnostic and prognostic marker for PCa [46]. Besides, LINC01138 can increase the arginine methylation and protein stability of sterol regulatory element-binding protein one by interacting with PRMT5, thereby promoting lipid desaturation and cell proliferation in clear cell renal cell carcinoma and being associated with poor prognosis [47]. However. The knockdown of LINC01138 can inhibit the viability, proliferation, invasion, and migration and promotes apoptosis of gastric cancer cells through the LINC01138/miR-1273e/MAPK axis [48].In some studies related to HCC, high-expressed LINC01138 is not only signi cantly associated with poor survival [49] but also can interact with arginine methyltransferase 5 to promote cell proliferation, tumorigenicity, tumor invasion, and metastasis [50]. CTD-2510F5.4 is a 321 bp antisense lncRNA, also known as lnc-SKA2-1 in the LNCipedia database. Through the NPInter v4.0 database [51], we found that CTD-2510F5.4 mainly interacts with genes in the mRNA surveillance pathway and RNA transport pathway. The high expressed CTD-2510F5.4 also has a signi cant coexpression relationship with mRNAs of the cell cycle, DNA replication, and p53 signaling pathway [52]. It is closely related to the poor prognosis of patients with lung adenocarcinoma [53]. In gastric cancer, the highly expressed CTD-2510F5.4 may be an independent risk factor for tumors with pathological grade < III and no vascular or nerve in ltration [54]. RP11-324I22.4 is an antisense lncRNA; the gene alias is lnc-CUL2-3. As cancer or tumor suppressor genes, antisense lncRNAs play an essential role in the occurrence and development of human cancer [55][56][57]. Although there is currently no disease research related to RP11-324I22.4, antisense lncRNAs may certainly be the promising tumor biomarker and therapeutic target in future research.
The risk scoring model we identi ed ( AUC of 1-, 3-, and 5-year were 0.853, 0.794, and 0.764, respectively ) is more reliable than a similar study in HCC (AUC of the 1-, 3-, and 5-year survival are 0.764, 0.738, and 0.717, respectively) [58], but it still has its limitations. This study is only based on the TCGA database, and there are no suitable datasets in other databases to verify the risk prediction model. Furthermore, the research is only conducted at the level of bioinformatics; a comprehensive in vitro experiment is needed further to explore the regulatory mechanism of these autophagy lncRNAs.

Funding
The design, the collection, analysis, interpretation of data, and the writing of the manuscript were supported in part by the special scienti c research project of health young medical science and technology talents in Xinjiang Uygur Autonomous Region (WYWY-202010).

Availability of data and materials
All of these data involved in this study were publicly available.
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.

Competing interests
The authors declare that they have no competing interests.
Yaping Zhou, Email: 1462755959@qq.com Table 1 The HR, 95% CI of HR, and P-value of the 7 autophagyrelated lncRNA based on the multivariate Cox regression analysis. HR, hazard ratio.