Expression status and prognostic value of autophagy-related lncRNAs in prostate cancer

ABSTRACT

LncRNAs involve in the autophagy to regulate prostate cancer (PCA) initiation and progression. Therefore, it urges to explore more significant AR-lncRNAs in PCa. mRNA data and clinical information of PCa were achieved from TCGA database, and ARGs were obtained from the HADb. AR-lncRNAs were identified by correlation analysis of DE ARGs and lncRNAs. Univariate Cox regression, LASSO regression, and multivariate Cox regression were used to identify the prognostic AR-lncRNA signature and constructed a risk model. GESA was used to biological function analysis between high- and low-risk score group. A nomogram was constructed and used to predicate the survival of PCa patients. A calibration curve was used to determine the accuracy of the predication model. AR-related ceRNA network was constructed by correlation analysis. Expression of six AR-related lncRNAs were detected by qRT-PCR. 222 ARGs and 385 AR-lncRNAs were screened from PCa and normal tissues, and 17 AR-lncRNAs were identified as prognostic signature for PCa. Based on the expression of prognostic signature, a risk score was calculated, and PCa samples were distributed into high- and low-risk score groups. The biological function and predicated value of the prognostic signature were also examined. Finally, based on the correlation between each ARG and its prognostic signature, three modules of AR-lncRNA-miRNA-mRNA regulatory networks were constructed based on 6 AR-lncRNAs, 17 miRNAs, and 12 ARGs. And we found that AC012085.2, UBXN10-AS1, LINC00261 downregulated, whereas AP004608.1, AC104667.2, AC008610.1 upregulated in PCa compared with BPH tissues. Our finding supplied the potential AR-lncRNAs prognostic signature for PCa.

Introduction

Prostate cancer (PCa) is the second most diagnosed cancer and fifth-leading cause of cancer-related death among males worldwide. Approximately 1.3 million new cases and 359,000 deaths of PCa in 2018, and the incidence of PCa has progressively increased over time [1,2]. With the development of docetaxel chemotherapy and androgen deprivation therapy, the survival rate of patients with localized PCa patient has improved [3,4]. However, 30% of PCa patients experience recurrence after primary therapy, and the currently available treatment options for those patients showed the limited efficacy [5]. Therefore, it is urge to investigate the pathogenic mechanism of PCa and identify novel therapeutic targets, which could be used to develop new strategies for diagnosing and treating PCa.

Autophagy, a lysosomal degradation process, is essential for cell survival, differentiation, development, and homeostasis [6]. Autophagy has been reported to play a dual role in cancer, on the one hand, autophagy promotes cell survival in response to starvation and stress, on the other hand, autophagy induces cell death by suppressing mitochondrial damage [7,8]. The autophagy degradation pathway can be activated under different pathological conditions. Under physiological conditions, autophagy prevents cancer development by protecting against the cellular accumulation of damaged substances and inhibiting tumorigenesis [9]. Conversely, defective autophagy enhances the autophagic response and promotes tumor growth, invasion, and metastasis [10,11]. Previous studies have revealed that several autophagy-related processes involve in PCa [12]. For example, inhibition of autophagy induces PCa cell resistance to docetaxel, and promotion of autophagy inhibits PCa cells sensitive to cisplatin [13]. However, it has been reported that activation of autophagy increases the radio-sensitivity of PCa [14]. Hence, autophagy acts as a key role in regulation of tumorigenesis, metastasis, and progression in PCa.

Long non-coding RNAs (lncRNAs) is one group of non-coding RNA (ncRNA), longer than 200 nucleotides and without protein-coding potential [15,16]. Although lncRNAs cannot be translated into proteins, numerous studies have described that lncRNAs contribute to cell proliferation, differentiation, and the transformation of malignancies by regulating multiple mechanisms in tumors [17,18]. Increasing evidence suggests that lncRNAs regulate tumor initiation, malignant transformation, and metastasis by modulating autophagic pathways [19–23]. Of particular interest are lncRNAs, which have been found to affect tumor growth, chemoresistance, and radio-resistance by regulating autophagy-related mechanisms [24–26]. However, the functions of many autophagy-related genes and lncRNAs remain unknown.

In this study, based on TCGA RNA-seq and clinical data, AR-lncRNA signature has been identified, and the AR-lncRNAs related prognostic models have been constructed and used to predict the disease-free survival (DFS) for PCa.

Methods

Data collection and processing

Transcriptome RNA sequencing data for 495 PCa patient samples and 52 normal samples, as well as clinical information corresponding to all 547 samples were obtained from The Cancer Genome Atlas (TCGA, http://cancergenome.nih.gov/), the clinical data included age, clinical stages, overall survival (OS) and disease-free survival (DFS). In addition, a total of 222 ARGs were identified from the Human Autophagy Database (HADb, http://autophagy.lu/clusteting/index.html).

Screening of the differentially expressed genes (DEGs) and differentially expressed lncRNAs (DE-lncRNAs)

The DEGs and DE-lncRNAs, which are between PCa samples and normal samples, were identified using the Limma R package with a threshold of |log (fold change, FC) | > 1 and P-value < 0.05.

Identification of ARGs and AR-lncRNAs

An ARG matrix was generated by integrating 222 ARGs with an mRNA expression matrix. The AR-lncRNAs were identified according to the Pearson’s correlation analysis of the correlations between 222 ARGs and lncRNAs, the threshold as follows: |correlation Coefficient| > 0.3 and a P-value < 0.05.

Construction of AR-lncRNAs signature and risk model

All PCa samples were divided into a training set (70%) and a testing set (30%). The prognostic significances of AR-lncRNAs in PCa were identified by univariate Cox regression analysis with a cutoff value of P < 0.05. The prognostic AR-lncRNAs were included into a least absolute shrinkage and selection operator (LASSO) regression to obtain a DFS-related AR-lncRNAs signature in PCa. Then, multivariate Cox regression was used to construct an AR-lncRNAs signature to predict prognosis in PCa. The risk score was obtained according to the following formula: risk score =$\sum _{i=1}^{n}coef\left(i\right)\times x\left(i\right)$, where coef ($i$) and $x\left(i\right)$ represented the estimated regression coefficient and the expression value of each AR-lncRNA, respectively. Risk-related lncRNAs were defined as those with a hazard ratio (HR) >1, and protective lncRNAs were defined as those with a HR <1. Subsequently, PCa patients were stratified into a high-risk group and low-risk group according to the median risk score. The DFS between the high-risk group and low-risk group was evaluated by Kaplan–Meier survival analysis using the term “survival” and the “survminer” R package. The risk curves for the high-risk group and low-risk group were drawn using the “pheatmap” R package. Time-dependent receiver operating characteristic (ROC) curves generated using the ROC R package and were used to calculate the predicative accuracy of the clinical pathological factors and risk scores for survival time.

Gene set enrichment analysis (GSEA)

The GSEA was performed to explore the pathway enrichment by DE AR-lncRNAs in the high-risk group and low-risk group. The enriched gene sets were acquired by using a threshold of P < 0.05 and a false discovery rate (FDR) <0.25. The enriched functional pathways were identified based on a normalized enrichment score (NES) >1 and FDR<0.01. A bubble plot was created to show the top 20 pathways according to P-value.

Establishment and validation of the nomogram

In order to identify the independent prognostic risk factors for PCa patients, the risk score and clinical characteristics (age, T stage, N stage, and C stage) of each patient were incorporated in a multivariate Cox regression model and the independent prognostic risk factors were identified with P-value<0.05. Furthermore, a nomogram was constructed based on risk score and the clinical variables (age, N stage, and C stage), and used to predict the 1-, 3-, and 5-year DFS of PCa patients in the entire set. The concordance index (C-index, 0.833) was used to determine the predictive ability of the nomogram, with a higher C-index value indicating a greater predictive ability of the nomogram model. The calibration curves were used to examine the concordance between predicated survival and observed survival after a bias correction.

Construction of the autophagy-related lncRNA-miRNA-mRNA (ceRNA) network

The ARGs and AR-lncRNAs were filtered according to the previous analysis results with a threshold of |log (FC) | > 1 and a P-value < 0.05. Correlation between ARGs and AR-lncRNAs were determined with the criteria of |correlation coefficient| > 0.3 and P-value < 0.05, and the correlated ARGs and AR-lncRNAs were used to construct the co-expression network.

Moreover, the correlation between lncRNA and miRNA was predicated using the miRcode database (http://www.mircode.org/), and ARGs targeted toward specific miRNAs were predicated using StarBase (http://starbase.sysu.edu.cn/). MiRNAs possessed binding sites for both ARGs and AR-lncRNA prognostic signatures, and the hypergeometric distribution test was used to screen the common miRNAs of ARGs and AR-lncRNAs with a threshold of P < 0.05. The lncRNA-miRNA pairs were identified following criteria of |Correlation coefficient| > 0.3 and P-value < 0.05. Finally, the lncRNA-miRNA-mRNA network was visualized using Cytoscape software.

Clinical specimens

In the present study, total 10 PCa tissues and 7 benign prostatic hyperplasia (BPH) tissues were harvested from patients who underwent surgical resection in the first Affiliated Hospital of Jinan University, and have not accepted the preoperative radiotherapy and chemotherapy. All patients involved in this research were known and signed the informed consent. This study was approved by the Ethics Committee of the first Affiliated Hospital of Jinan University (KY-2020-095). All clinical samples were resected and stored in the liquid nitrogen at −80°C.

RNA extraction and qRT-PCR analysis

Total RNA was isolated from specimens using Trizol reagent (Takara, Dalian, China) and reverse transcribed into cDNA using Bestar qPCR RT Kit (DBI Bioscience, Shanghai, China) according to the manufacturer’s instruction. Then, real-time PCR was performed by SYBR Green Bestar® one-step RT qPCR kit (DBI Bioscience, Shanghai, China) according to the manufacturer’s protocol. The reaction condition as following, warm start at 95°C for 2 min, followed by40 cycles of 95°C for 30s, 58°C for 20 s, 72°C for 20 s. The primers used in this study as Table 3. GAPDH was used to normalize the expression levels of lncRNAs. The relative expression was quantified by the 2^‑ΔΔCt method.

Table 3.

The sequences of primers used in this study.

Gene		Sequences (5’-3’)
AC012085.2	F	5’-CAACATCCGAGGTTCCATTTCA-3’
AC012085.2	R	5’-GGGAGACCAAGGCGAGAAGAT-3’
UBXN10-AS1	F	5’-AGTCTGTGACCCGTGGGAGTT-3’
UBXN10-AS1	R	5’-AACAAGAAATAAGCCGTCAGCAC-3’
AP004608.1	F	5’-CTGCGTTGAGAACCAGATGC-3’
AP004608.1	R	5’-ATTTGTATTAGGAGGAGGTAGGG-3’
AC104667.2	F	5’-CGCAGAATGGGAGGAAGATA-3’
AC104667.2	R	5’-GGGTCACCTGGTCACATCTC-3’
AC104667.2	F	5’-GCAGCAGGGATGGTGTAGGT-3’
AC104667.2	R	5’-CAGGCACCTCCCAGTAAACG-3’
LINC00261	F	5’-ACAAACAGCCACAGCACCCTC-3’
LINC00261	R	5’-TTCCATTCCTTGTATGCCTTTCT-3’
GAPDH	F	5’-TGTTCGTCATGGGTGTGAAC-3’
GAPDH	R	5’-ATGGCATGGACTGTGGTCAT-3’

Results

1. Differentially expressed AR-lncRNAs in PCa

The process of this study has been illustrated in workflow chart (Figure 1). A total 959 DEGs, including 202 upregulated and 757 downregulated DEGs (Figure 2(a, b)), and 134 DE-lncRNAs, including 52 upregulated and 82 downregulated DE-lncRNAs (Figure 2)c, d)), were identified from 495 PCa samples and 52 normal samples with the threshold of |log (FC) |>1 and P-value<0.05 (Supplementary file 1–2). Of these, 222 DE-ARGs were identified from above 959 DEGs (Supplementary file 1). One hundred and ten AR-lncRNAs were screened according to the criteria of |Correlation coefficient| > 0.3 and P-value < 0.05 (Supplementary file 3). Top 20 DE AR-lncRNAs are shown in Table 1.

Figure 1.

The workflow of this study.

Figure 2.

Differentially expressed AR-lncRNAs in PCa.

(a-c) The volcano plots revealed DEGs andDE-lncRNAsbetween495 PCa samples and 52 normal samples. Red dots and blue dots represent significantly downregulated and upregulated DEGs or DE-lncRNAs in PCa. (b-d) The heatmap showed the DEGs and DE-lncRNAs between 495 PCa samples and 52 normal samples.

Table 1.

The top 10 positive and negative correlation between ARGs and lncRNAs.

ARG ID	LncRNA ID	Cor.	P value
ENSG00000152137	ENSG00000267505	0.94684508	9.2992E-190
ENSG00000152137	ENSG00000267405	0.908755397	1.1348E-146
ENSG00000152137	ENSG00000254510	0.908579331	1.6101E-146
ENSG00000152137	ENSG00000174403	0.895393743	6.1147E-136
ENSG00000073282	ENSG00000230937	0.887381136	3.5152E-130
ENSG00000152137	ENSG00000268388	0.87648195	5.1863E-123
ENSG00000152137	ENSG00000241158	0.858760737	1.0728E-112
ENSG00000152137	ENSG00000224958	0.855456006	6.287E-111
ENSG00000073282	ENSG00000261116	0.849408284	8.3655E-108
ENSG00000158458	ENSG00000226237	0.838541563	1.6031E-102
ENSG00000110931	ENSG00000233901	−0.606051261	9.01952E-40
ENSG00000089685	ENSG00000224958	−0.597613628	1.89962E-38
ENSG00000158458	ENSG00000234949	−0.587757636	5.96812E-37
ENSG00000150995	ENSG00000266402	−0.565300858	1.00549E-33
ENSG00000150995	ENSG00000255198	−0.562849298	2.18694E-33
ENSG00000089685	ENSG00000241158	−0.554567377	2.87949E-32
ENSG00000110931	ENSG00000225431	−0.543484251	8.11585E-31
ENSG00000110931	ENSG00000237989	−0.531780759	2.41639E-29
ENSG00000089685	ENSG00000226237	−0.524239439	2.00841E-28
ENSG00000089685	ENSG00000254510	−0.521557291	4.21147E-28

Abbreviations: Cor.,Correlation coefficient.

2. Identification of the AR-lncRNA signature in PCa

Previous analyses had identified 222 ARGs and 110 AR-lncRNAs in PCa. Then, 39 AR-lncRNAs, associated to DFS in PCa, were identified by univariate Cox regression and Kaplan–Meier analysis with a criterion of P < 0.05 (Table 2 and Supplementary file 4). Next, the 39AR-lncRNAs were incorporated into a LASSO regression analysis, and 17 AR-lncRNAs were identified following a minimum lambda value of 0.028 (Figure 3(a, b), Supplementary file 5). Subsequently, the 17 AR-lncRNAs were included in a multivariate Cox regression and identified as prognostic signature in PCa. Each sample in the training set or testing set was assigned to a high-risk group or low-risk group, based on the median risk score (Figure 3(c, d), top). Survival plots indicated that patients in the high-risk score group had a higher dead rate than patients in the low-risk score group (Figure 3(c, d), bottom). In addition, DFS curves indicated that patients in high-risk score group had shorter survival times (Figure 3(e, f)). Moreover, the AUC value of the ROC curve confirmed the accuracy of the risk model, the AUC value at 1-, 2-, 3-, 4-, 5-year predication were 0.829, 0.791, 0.814, 0.787, 0.756 in training set, and the AUC value at 1-, 2-, 3-, 4-, 5-year predication were 0.654, 0.695, 0.695, 0.712, 0.715 in testing set (Figure 3(g, h)). Finally, a clinical correlation analysis revealed that the risk score and 17 AR-lncRNAs significantly associated with the C-stage and T/N stages of prostate cancer (Figure 3(i)).

Table 2.

Univariable Cox proportion regression identified 39 AR-lncRNAs in PCa.

Gene symbol	Coef	HR	HR.95 L	HR.95 H	P value
AC008610.1	0.46590286	1.59345221	1.25022021	2.03091417	0.00016702
UBXN10-AS1	−0.46028591	0.63110318	0.49483139	0.80490292	0.00020835
PGM5-AS1	−0.256809	0.77351594	0.67135034	0.89122903	0.0003805
AC084018.1	0.44939524	1.56736401	1.21795991	2.01700395	0.00047904
AC015845.2	−0.49005076	0.6125953	0.4586912	0.81813865	0.00090112
AC012085.2	−0.68920852	0.50197321	0.33375453	0.75497734	0.00093396
PCAT14	−0.18458845	0.8314464	0.74523454	0.9276316	0.0009499
AL135999.1	0.44221726	1.55615379	1.17944666	2.05317858	0.00176571
AC005180.2	−0.31956032	0.72646838	0.59248545	0.89074982	0.00212482
AP006748.1	−0.23210638	0.79286178	0.68369513	0.91945923	0.00213382
SNHG25	0.30390201	1.35513626	1.09108372	1.68309199	0.00599082
AP001107.5	−0.40472711	0.66715885	0.49665327	0.89620053	0.00719362
AC092535.4	0.18244155	1.200144	1.04956892	1.37232114	0.00764715
AC005180.1	−0.26605363	0.76639802	0.62791619	0.93542091	0.0088836
FGF14-AS2	−0.41811124	0.658289	0.48114986	0.9006433	0.00894182
ADAMTS9-AS1	−0.48049076	0.6184798	0.42951139	0.89058699	0.00979914
FENDRR	−0.34938543	0.7051213	0.53866383	0.92301733	0.01098953
AL138881.1	−0.43186516	0.64929692	0.46496777	0.90670047	0.01124975
AL365181.4	−0.47200193	0.62375231	0.43146485	0.90173497	0.01207285
LINC00908	−0.64319705	0.52560934	0.31568703	0.87512361	0.01340646
PCA3	−0.12861811	0.8793097	0.79396489	0.97382838	0.01354649
AC104137.1	−0.43643366	0.64633737	0.45466484	0.91881306	0.01502616
LINC01509	−0.35839832	0.69879467	0.52142208	0.93650425	0.01643569
AC010719.1	0.24313194	1.27523686	1.04010657	1.56352157	0.0193809
AP001610.2	−0.19003924	0.82692668	0.70042932	0.97626944	0.0248652
AC144450.1	−0.22893273	0.79538204	0.64949848	0.97403244	0.02679665
LINC02489	−0.30183375	0.73946099	0.56494827	0.96788075	0.02797308
AC116614.1	0.15647517	1.16938173	1.01557007	1.34648871	0.02965342
AC036108.3	−0.41239646	0.66206174	0.45134568	0.97115308	0.03488358
CTBP1-AS	0.2255696	1.25303624	1.01070408	1.55347134	0.03968045
AL353622.1	0.29995654	1.34980015	1.00967205	1.80450715	0.04287309
AL391335.1	−0.53298689	0.5868495	0.34916368	0.98633492	0.04423025
AC104667.2	−0.29620883	0.74363212	0.55714583	0.99253857	0.044346
LINC00261	0.21532823	1.24026893	1.00473048	1.53102453	0.04508261
C20orf166-AS1	−0.37396762	0.68799919	0.47686518	0.99261364	0.04554288
SNHG3	0.31529199	1.37065947	1.00410788	1.87102145	0.04705738
LINC01018	−0.37136099	0.68979489	0.47756196	0.99634608	0.04776241
AC011523.1	0.2127276	1.23704763	1.00208618	1.52710104	0.04777606
AC141930.1	−0.17366961	0.84057457	0.70752835	0.99863928	0.04821683

Abbreviations: Coef, Coefficient; HR, Hazard ratio; HR.95 L/H, 95%confidence interval of the hazard ratio.

Figure 3.

Identification of the AR-lncRNA signature in PCa.

(a, b) The Lasso coefficient values of 39 AR-lncRNAs in PCa; the optimal log (lambda) value is indicated by the vertical dashed lines. The profiles of Lasso coefficient spectra for 17 AR-lncRNAs.(c, d) Top: The risk score distributed between high-risk group and low-risk group by the median risk score both in training set and testing set, respectively. Bottom: The scatter plots showed the correlations between PCa patients and risk scores, survival status, and survival time.(e, f) The DFS curves for high-risk group and low-risk group in the training set and testing set, respectively.(g, h) The ROC curves for 1-, 2-, 3-, 4-, and 5-year DFS indicated the efficiency of the 17AR-lncRNA prognostic signature in PCa.(i) A heatmap showing the differently expressed 17 prognostic AR-lncRNAs, and the correlation between their expression and patient clinical characteristics.

3. GSEA of the AR-lncRNAs prognostic signature in high-risk score group and low-risk score group

A biological function analysis was performed by GSEA. The AR-lncRNAs prognostic signature significantly enriched in multiple tumor-related pathways between the high-risk group and low-risk group (Figure 4(a), Supplementary file 6). GSEA results indicated that “E2F target,” “interferon gamma response,” mitotic spindle,” “allograft rejection,” “MYC target V1,” “G2M checkpoint,” “DNA repair,” “interferon alpha response,” and “MYC target V2” were enriched in the high-risk group with a cutoff of FDR < 0.01 (Figure 4(b)). “Cholesterol homeostasis,” “androgen response,” “bile acid metabolism,” “protein secretion,” “UV response UP,” “apoptosis,” “fatty acid metabolism,” “oxidative phosphorylation,” “peroxisome,” “UV response DN,” and “HEME metabolism” were enriched in the low-risk group with a threshold of FDR < 0.01 (Figure 4(c)). Taken together, the GSEA results suggested that above mentioned signaling pathways involved in aggressive PCa progression, and implied the prognostic values of AR-lncRNAs for PCa, those AR-lncRNAs acted important role in regulating tumor progression and poor prognosis for PCa patients.

Figure 4.

GSEA of the AR-lncRNAs prognostic signature in high-risk score group and low-risk score group.

(a) Bubble charts of the top 20 enriched pathways as determined by GSEA analysis.(b, c) GSEA showing the enrichment of Hallmarks and the KEGG pathway in the high-risk score group and low-risk score group.

4. Identification of the independent predictive factors in PCa

A multivariate Cox regression was used to explore the predictive value of AR-lncRNAs that acted as the prognostic signature in PCa. The results showed that risk score, based on AR-lncRNAs prognostic signature, could serve as independent prognostic factors for PCa (Figure 5(a). Based on the multivariate Cox regression analysis, a nomogram, including age, N stage, C stage, and risk score was constructed (Figure 5(b). The calibration curves for 1-, 3-, and 5-year DFS confirmed the predictive ability of the prognostic model (Figure 5(c)).

Figure 5.

Identification of the independent predictive factors in PCa.

(a) A forest plot for the multivariate Cox regression analysis in PCa.(b) The nomogram of 1-, 3- or 5-year of DFS based on the risk score, age, C stage, and N stage.(c) Calibration plots for assessing the accuracy of the prognostic model for DFS.

5. Construction of the autophagy-related ceRNA network

To investigate the underlying mechanism of AR-lncRNAs in PCa, a lncRNA-miRNA-mRNA regulatory network was constructed. The relationships between lncRNAs and miRNAs, miRNAs and mRNAs were predicated by using miRcode, StarBase, Targetscan, and miRTarBase online software. The results showed that there were three distinct modules of ceRNA networks that possessed 6 lncRNAs (AC012085.2, UBXN10-AS1, AP004608.1, AC104667.2, AC008610.1, and LINC00261), 17 miRNAs (has-miR-106a-5p, has-miR-17-5p, has-miR-7-5p, has-miR-20b-5p, has-miR-20a-5p, has-miR-93-5p, has-miR-23c, has-miR-204-5p, has-miR-144-3p, has-miR-4770, has-miR-19b-3p, has-miR-19a-3p, has-miR-143-3p, has-miR-148a-3p, has-miR-152-3p, has-miR-27b-3p, and has-miR-1200c-3p), and 12 target genes (BNIP3L, DLC1, PTEN, GABARPL1, VAMP3, ITPR1, ULK3, ATG4D, MTMR14, RAB24, MAP2K7, and NRG1) (Figure 6). 6 lncRNAs (AC012085.2, UBXN10-AS1, AP004608.1, AC104667.2, AC008610.1, and LINC00261) exerted the regulatory roles in PCa by acting as competing endogenous RNAs (ceRNAs) bind miRNAs to modulate the target genes. And we demonstrated that AC012085.2, UBXN10-AS1, LINC00261 downregulated, whereas AP004608.1, AC104667.2, AC008610.1 upregulated in PCa compared with BPH samples by qRT-PCR (Figure 7).

Figure 6.

Construction of the autophagy-related ceRNA network. Green rhombuses represent AR-lncRNAs, red quadrangles represent paired miRNAs, and blue circles represent ARGs.

Figure 7.

QRT-PCR detected the expression of six lncRNAs (AC012085.2, UBXN10-AS1, AP004608.1, AC104667.2, AC008610.1, and LINC00261) between 10 PCa and 7 BPH samples.

Discussion

In recent years, assays for prostate-specific antigen (PSA) have been performed to enable the early diagnosis of asymptomatic PCa. However, prostatectomy surgery often fails to remove latent cancer cells in prostate tissue [27]. Although the death rates for PCa have been reduced, it remains difficult to reduce the incidence of PCa through early diagnosis or to change the risk for developing PCa [28,29]. Therefore, there remains an urgent need to identify high-performance biomarkers for use in diagnosing and treating PCa. Previous studies have shown that AR-lncRNAs associate with a tumor’s prognosis. For example, a study by Huang et al. have indicated that lncRNA PVT1 initiates cytoprotective autophagy and accelerates tumor development in pancreatic ductal adenocarcinoma, and a high level of PVT1 expression predicts a poor prognosis for patients [31]. Luan et al. have screened 10 differentially expressed AR-lncRNAs in gliomas, and have declared they act biomarkers for predicting the prognosis of glioma patients [32]. However, little is understood about the biological function and mechanism of AR-lncRNAs in PCa.

In the present study, we screened 222 ARGs and 17 AR-lncRNAs based on information gathered from a bioinformatics analysis. A risk signature and prognostic model were constructed based on 17 AR-lncRNAs and the clinical characteristics of PCa patients. Our results demonstrated that the risk signature was an important factor affecting the progression of PCa. In addition, three autophagy-related lncRNA-miRNA-ARGs tribasic networks were established via the co-expression of lncRNAs and ARGs, and 6 AR-lncRNAs (AC012085.2, UBXN10-AS1, AP004608.1, AC104667.2, AC008610.1, and LINC00261) involved in ceRNA interaction networks. Interestingly, emerging evidence indicates that LINC00261 acts as a tumor suppressor in various types of cancer, including lung cancer, liver cancer, breast cancer, gastric cancer, and pancreatic cancer, PCa [30–33]. Here, LINC00261 downregulated in PCa, is consistent with the results of previous studies. Nevertheless, AC012085.2, UBXN10-AS1, AP004608.1, AC104667.2, and AC008610.1 hardly ever reported in cancers. We first demonstrated the roles of the five AR-lncRNAs in PCa, AC012085.2 and UBXN10-AS1 also acted as a tumor suppressor, whereas AC104667.2, AC008610.1, and AP004608.1 played as oncogenes in PCa.

As described in previous studies, autophagy plays a dual role in tumorigenesis and development [8]. In tumors, the autophagy process removes nonessential and dysfunctional substances to sustain cell homeostasis and survival. Moreover, autophagy activates the apoptosis pathway and helps to eradicate tumor cells [18]. Several studies have shown that lncRNAs are novel regulators of autophagy, and affect the autophagy process in a variety of ways. First, lncRNAs act as competitive endogenous RNA (ceRNA) bind miRNAs to affect miRNA expression, and thereby directly or indirectly modulate the autophagy process [34]. Second, lncRNAs also regulate the expression of ATG genes. Third, lncRNAs mediate autophagy-induced apoptosis to promote tumor progression by modulating several signaling pathways, such as the PI3K/AKT/mTOR pathway [35]. Our GSEA analysis has demonstrated that distinct difference sex is ted between the autophagy-related signaling pathways in the high- and low-risk groups. Several cancer-related pathways and autophagy-related pathways enriched in the high-risk group, whereas immunomodulatory- and metabolism-pathways enriched in the low-risk group. These results are concordant with the current understanding that increasing immunity correlates with a better prognosis. Harri et. al, have reported that lipid degradation promotes prostate cancer cell survival [36], which is consistent with our data, an elevated level of fatty acid metabolism associated with increased patient survival times.

In the present study, we validated the expression of expression of six AR-related lncRNAs who involved in ceRNA network. And our experimental results supported the findings of bioinformatics analysis, suggesting that bioinformatics analysis is the reliable approach to identify potential molecular markers for diagnosis and therapy. Although we verified the primary findings in this study, the functional experiments have not been conducted in this study, our further research will revolve around investigation of the potential molecular mechanisms underlying the predictive characteristics of AR-lncRNAs.

Conclusion

Based on TCGA and HADb online databases, we identified ARGs and AR-lncRNAs, then discovered the independent prognostic signature and developed AR-lncRNA prognostic predicated model for PCa. Our findings provide novel insights into the biochemical mechanisms of PCa, and might help to identify new diagnostic and therapeutic biomarkers for PCa patients.

Funding Statement

This work was supported by the National Natural Science Foundation of China [81802567]; Leading Specialist Construction Project-Department of Urology, the First Affiliated Hospital, Jinan University [711006]; Science and Technology Program of Guangzhou [202102080043].

Authors’ contribution

Conceived and designed the experiments: Yutong Li and Bin Pan; Performed the experiments: Xiaoping Qin; Analyzed the data: Yu wang; Contributed reagents/ materials/analysis tools: Biyun Gao, Muan Ling, Wenjun Yin; Writing–original draft: Guo Chen; Writing–review & editing: Bin Pan.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Ethical approval

All procedures involving humans were approved by the Ethics Committee of the first Affiliated Hospital of Jinan University (KY-2020-095).

Data availability statement

The datasets generated during and/or analyzed during the current study are available from the corresponding author (Bin Pan) on a reasonable request. The data generated or used during the study are available online as follows:

http://cancergenome.nih.gov/

http://autophagy.lu/clusteting/index.html

http://www.mircode.org/

http://starbase.sysu.edu.cn/

Supplementary material

Supplemental data for this article can be accessed here