The transcriptome sequencing data and clinical data were collected from TCGA database (https://cancergenome.nih.gov/). LncRNAs and PCGs were isolated from transcriptome sequencing data. A total of 500 PCa patients were found in TCGA database. We selected 31 PCa patients with complete data for this study. Based on the radioresponse, 31 patients were categorized into CR group (n = 15) and non-CR group (n = 16) which contained partial response, stable disease, and progressive disease. Their clinical characteristics were shown in Supplementary Table 1.

With normalized data, the differential expression of lncRNAs and PCGs between CR group and non-CR group was analyzed using the edgeR package in R software. The fold change (FC) and false discovery rate (FDR) for each lncRNA and PCG were calculated. The cut-off criteria was logFC > 1 and FDR < 0.05.

To validate the bioinformatics analysis results, we further collected the tumor tissues and clinical characteristics of 40 PCa patients who received radiotherapy from the First Affiliated Hospital, Anhui Medical University, which were categorized into CR group (n = 20) and non-CR group (n = 20) according to their radioresponse. Their clinical characteristics were shown in Supplementary Table 1. All of human studies in our study were in accordance with Declaration of Helsinki and the guidelines of the Committee for Ethical Review of Research at the First Affiliated Hospital, Anhui Medical University.

PCG-seq data was isolated from transcriptome sequencing data. A total of 17,964 PCGs were identified for each sample. The variance analysis was performed, and it was ranked from large to small. The top 25% of the PCGs (4,491 PCGs) with larger variance were selected for WGCNA analysis.

The expression profile of these 4,491 PCGs was constructed to a gene co-expression network using the WGCNA package in R software (21). The weighted co-expression relationship among all dataset subjects in an adjacency matrix was assessed by the Pearson correlation. In this study, the soft threshold was set as β = 9 (scale-free R² = 0.93) to ensure a scale-free network. The adjacency matrix was used to calculate the topological overlap measurement (TOM) representing the overlap in the shared neighbors to further identify functional modules in the co-expression network with these 4491 PCGs (22).

We performed principal component analysis (PCA) of all PCGs in each module, and defined the value of principal component one as module eigengenes (MEs). MEs were deemed to be representative of gene expression profiles in the module. Next, we performed a correlation analysis between the MEs of each PCG module and the clinical characteristics of PCa patients to identify the clinical significant module.

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of PCGs co-expressed with LINC01600 were performed using Database for Annotation, Visualization, and Integrated Discovery (DAVID) [(23); version 6.8; https://david.ncifcrf.gov/] online functional annotation tool. GO analysis includes three categories: biological processes (BP), cellular components (CC), and molecular functions (MF). Enriched GO terms and KEGG pathways were determined based on the cut-off criteria of adjusted P < 0.05.

The human PCa cell lines including PC-3 and DU145 were purchased from Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences (Shanghai, China), and all cells were cultured in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) at 37 °C with 5% CO₂ in a humidified incubator. Small interfering RNA (siRNA) for LINC01600 was designed and synthesized by GenePharma (Shanghai, China). The si-LINC01600 sequence was shown in Supplementary Table 2. Cells were cultured in six-well-plates until 70% confluent, and then transfected with siRNA using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. After incubating the cells for 48 h at 37°C in a humidified chamber supplemented with 5% CO₂, the cells were harvested for RT-qPCR.

Total RNA was isolated from PCa tissues using TRIzol reagent (Invitrogen, California, USA). The cDNA was synthesized using HiScript III RT SuperMix for qPCR Kit (Vazyme Biotech, Nanjing, China), according to the manufacturer's protocol at 37°C for 15 min and 85°C for 5 s. The cDNA was subsequently analyzed using ChamQ Universal SYBR qPCR Master Mix (Vazyme Biotech, Nanjing, China) and the ABI7500 system (Applied Biosystems, Foster City, USA). The amplification program was as follows: Initial denaturation step at 95°C for 30 s, followed by 40 cycles at 95°C for 10 s, 60°C for 30 s. The expression of LINC01600, JUND, ZFP36, and ATF3 was calculated relative to the internal reference gene, GAPDH, using the 2^−ΔΔCt method (24). The primer sequences were shown in Supplementary Table 2.

The potential target miRNAs of lncRNAs were predicted by miRcode (25) (http://www.mircode.org/) online database. The miRNAs that might regulate PCGs were predicted by TargetScan (26) (http://www.targetscan.org/vert_72/) and DIANA [(27) diana.imis.athena-innovation.gr/DianaTools/index.php] online databases. Venn diagram was used to confirm overlapping miRNAs. Cytoscape software (version 3.6.1; https://cytoscape.org/) was used to visual the ceRNA network.

Statistical analysis was constructed by SPSS (version 24.0) and R software (version 3.4.2; https://www.r-project.org/). Pearson correlation analysis was used to identify the PCGs co-expressed with lncRNAs. The chi-square test was used to identify the correlation between lncRNAs and radioresponse. P < 0.05 was considered statistically significant.

According to the cut-off criteria (logFC > 1, FDR < 0.05), 65 differentially expressed lncRNAs (Supplementary Table 3) and 468 differentially expressed PCGs (Supplementary Table 4) were identified between CR group and non-CR group. LINC01600 (logFC = 1.0046, FDR = 0.0155) and JUND (logFC = 1.0210, FDR = 0.0002) were upregulated in non-CR group.

To further identify lncRNAs which were highly correlated with radioresponse from 65 differentially expressed lncRNAs, we performed the chi-square test. According to the median value, the expression of lncRNAs was divided into low and high expression. LINC00919 and LINC02010 were excluded because more than half of the cases did not detect their expression. LINC01600 (P = 0.004), LINC01060 (P = 0.012), MIR137HG (P = 0.032), LINC02006 (P = 0.032), and LINC02531 (P = 0.032) were identified as highly correlated with radioresponse (Table 1 and Supplementary Table 5). We chose LINC01600 with the lowest P-value for subsequent analysis.

To identify PCGs associated with radioresponse in PCa, we performed a WGCNA analysis. The samples of 31 PCa patients receiving radiotherapy were clustered using average linkage method and Pearson correlation analysis. We excluded 2 outlier samples and finally included 29 samples for subsequent analysis (Figures 1A,B). Next, we performed a network topology analysis of various soft-thresholding powers to have relatively balanced scale independence and average connectivity of WGCNA. In our study, the power of β = 9 (scale free R² = 0.93) was selected as the soft-thresholding parameter to make sure a scale-free network (Figures 1C–E).

A total of 22 PCG modules were generated in a hierarchical clustering tree through dynamic tree cut and merged dynamics. After merging modules with dissimilarity less than 25%, 20 distinct PCG modules were identified (Figures 2A,B). Correlation analysis was performed between the MEs of each PCG module and radiotherapy outcome. According to P < 0.05 of correlation analysis, the darkred module (Cor = 0.37, P = 0.04) was identified as the PCG module highly correlated with radioresponse (Figure 2C). Subsequently, we calculated the gene significance for radioresponse and the module membership in darkred module. The scatter plot (Cor = 0.34, P = 0.018) further validated the high correlation between the darkred module and radioresponse (Figure 3). The list of PCGs for the darkred module (n = 48) was shown in Supplementary Table 6.

In order to screen the potential target PCGs for LINC01600, we performed Pearson correlation analysis between LINC01600 and 17964 PCGs. According to the cut-off criteria of | Pearson correlation coefficient | > 0.4 and P < 0.05, a total of 558 PCGs co-expressed with LINC01600 were screened (Supplementary Table 7).

After taking the intersection of the 558 PCGs co-expressed with LINC01600, the 48 PCGs in the drakred module obtained by WGCNA, and the 468 differentially expressed PCGs between CR group and non-CR group, we obtained the potential target PCGs of LINC01600, that is, JUND, ZFP36, and ATF3 (Figure 4).

We performed GO enrichment analysis on 558 PCGs co-expressed with LINC01600. The PCGs in BP group were mainly enriched in transcription, regulation of cell cycle, miRNA metabolic process, double-stand break repair via homologous recombination, DNA replication, and DNA repair. The PCGs in CC group were mainly enriched in trans-Golgi network, small subunit processome, nucleolus, nuclear chromosome, telemetric region, cytoplasm, and chromosome. The PCGs in MF group were mainly enriched in RNA binding, nucleic acid binding, double-stranded RNA binding, DNA binding, damaged DNA binding, and 3–5 exonuclease activity. KEGG pathway analysis revealed that these PCGs were mainly involved in RNA transport, ribosome biogenesis in eukaryotes, pyrimidine metabolism, homologous recombination, DNA replication, and cell cycle (Figure 5). These results indicated that the 558 PCGs co-expressed with LINC01600 were involved in various biological processes such as DNA damage repair, metabolism, cell cycle, and so on, which might be related to cell radioresponse.

RT-qPCR was used to detect the expression of LINC01600, JUND, ZFP36, and ATF3 in tumor tissues of 40 PCa patients collected in our hospital. The radioresponse of 40 PCa patients receiving radiotherapy included 20 CR and 20 non-CR. We found that LINC01600 and JUND were significantly upregulated in non-CR group compared with CR group (Figures 6A,B). However, there was no differential expression of ZFP36 and ATF3 between CR group and non-CR group (Figures 6C,D). These results initially verified that LINC01600 and JUND were associated with radioresponse in PCa. We chose JUND as the potential target PCGs of LINC01600 for subsequent analysis.

LINC01600 was downregulated by siRNA at two different sites in PCa cell line. The lncRNA expression of LINC01600 and the mRNA expression of JUND were detected by RT-qPCR. The results showed that after downregulating LINC01600 by si-LINC01600_1 and si-LINC01600_2, the mRNA expression of JUND was also downregulated (Figure 7). In summary, LINC01600 may regulate JUND.

To further explore the potential mechanisms of LINC01600, we constructed a ceRNA network of LINC01600—miRNAs—JUND. The miRcode database was used to predict the potential target miRNAs of LINC01600. The TargetScan and DIANA databases were used to predict the miRNAs that might regulate JUND. After taking the intersection of the above two results, four miRNAs, that is, miR-206, miR-490-3p, miR-216b-5p, and miR-613, were screened for constructing the ceRNA network (Figure 8).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.