Comprehensive microRNA profiling of prostate cancer cells after ionizing radiation treatment

MicroRNAs (miRNAs) are small, non-coding RNAs that negatively regulate gene expression and have emerged as potential biomarkers in radiation response to human cancer. Only a few miRNAs have been identified in radiation response to prostate cancer and the involvement of the radiation-associated miRNA machinery in the response of prostate cancer cells to radiation is not thoroughly understood. Therefore, the purpose of the present study was to comprehensively investigate the expression levels, arm selection preference and isomiRs of radiation-response miRNAs in radiation-treated PC3 cells using a next-generation sequencing (NGS) approach. Our data revealed that the arm selection preference and 3′ modification of miRNAs may be altered in prostate cancer after radiation exposure. In addition, the proportion of AA dinucleotide modifications at the end of the read gradually increased in a time-dependent manner after PC3 radiation treatment. We also identified 6 miRNAs whose expression increased and 16 miRNAs whose expression decreased after exposure to 10 Gy of radiation. A pathway enrichment analysis revealed that the target genes of these radiation-induced miRNAs significantly co-modulated the radiation response pathway, including the mitogen-activated protein kinase (MAPK), Wnt, transforming growth factor-β (TGF-β) and ErbB signaling pathways. Furthermore, analysis of The Cancer Genome Atlas (TCGA) database revealed that the expression of these radiation-induced miRNAs was frequently dysregulated in prostate cancer. Our study identified radiation-induced miRNA candidates which may contribute to radiosensitivity and can be used as biomarkers for radiotherapy.


Introduction
Prostate carcinoma is the most frequently diagnosed visceral cancer in men worldwide. An increasing prevalence has been reported in recent decades (1). Radiation therapy is one of the primary modalities in prostate cancer treatment. Ionizing radiation damages cells through free radicals from the radiolysis of water that cause DNA double-strand breaks. However, the efficacy of the radiotherapy may be affected by the cellular response to radiation. Radiotherapy is highly effective in treating radiosensitive tumors and enhancing the therapeutic efficacy can increase the overall survival rate. However, the presence of radioresistant tumors leads to cancer relapse and metastasis. Understanding the tumor-radiation-related genes to predict the tumor response to radiotherapy may potentially modulate the treatment outcome for prostate cancer patients.
MicroRNAs (miRNAs) are a family of small, non-coding, single-stranded RNAs composed of ~22 nucleotides (nt) that negatively regulate protein expression at the post-transcriptional level (2). They function as gene regulators by binding to partially complementary sites of mRNAs and cause translation inhibition or direct degradation of the target mRNA. It has been suggested that miRNAs are responsible for controlling ~50% of all protein-coding genes (3). The widespread regulation of protein levels has been studied in cellular models (4). Previous studies have demonstrated that the expression of miRNAs is clearly involved in cancer development, and the deregulation of several miRNAs has been observed in various types of cancer, including prostate cancer. Porkka et al (5) was the first to identify a miRNA signature specific for prostate cancer by systematically profiling prostate cancer cell lines. Numerous studies have identified many dysfunctional miRNAs by using a high-throughput approach, which contributed to prostate cancer progression, including the let-7 family, miR-1, -20a, -21,-34a, -106b, -125b, -205 and -521 (6)(7)(8)(9)(10)(11)(12)(13). Although several studies have investigated the role of these dysfunctional miRNAs to develop prostate cancer therapy, few studies have determined the roles of miRNAs in radiation response in prostate cancer. The upregulation of miR-521 reduces the response to radiation damage by specifically targeting a DNA repair protein, the Cockayne syndrome protein A (13). Li et al (14) found that miR-106b was dysregulated after radiation treatment and suppressed radiation-induced p21 activation, suggesting it may override radiation-induced cell cycle arrest and cell growth inhibition. Radiation delivered in daily fractions altered a greater number of miRNAs compared with single-dose radiation, and involved the upregulation of miR-34a and let-7 miRNAs (15).
Next-generation sequencing (NGS) is a high-throughput screening technology, and NGS data can be applied in investigating miRNA expression, miRNA isoforms (isomiRs) and the arm selection preferences of miRNAs. Therefore, the purpose of the present study was to comprehensively investigate the distribution of miRNAs after radiation treatment in PC3 cells by using an NGS approach. Furthermore, we explored the function of radiation-associated miRNA by conducting an in silico analysis.

Materials and methods
Cell culture and radiation treatment. A PC3 cell line was obtained from the American Type Culture Collection and was maintained in RPMI-1640 and supplemented with 10% inactivated fetal bovine serum (FBS; Invitrogen, Carlsbad, CA, USA). The cells were exposed to various radiation dosages (0, 2, 6, 10, 14 and 18 Gy) and were subsequently cultured in fresh medium. The total RNA was obtained at various time points (0, 5, 15 and 40 h after treatment) by using TRIzol (Invitrogen) according to the manufacturer's instructions. The concentration, purity and amount of total RNA were determined using a NanoDrop 1000 spectrophotometer (NanoDrop Technologies, Inc., USA).
Collection and preprocessing of sequence reads. PC3 cells were exposed to 10 Gy of radiation. After radiation treatment, the cells were lysed at various time points (0, 5, 15 and 40 h) for RNA extraction. The RNA samples were prepared using an Illumina small RNA preparation kit, and were subsequently sequenced using the Illumina HiSeq platform. The generated sequence reads were first subjected to quality control to remove low-quality reads. The sequence reads were then subjected to 3' adaptor trimming to generate clean reads, as previously described (15,16). To attain a high confidence level, only the clean reads with a read count ≥2 and with a length ranging from 15 to 27 nt were included in further analyses.
Mapping clean reads to pre-miRNAs. To investigate miRNA expression profiles in different libraries, we mapped the qualified clean reads back to human pre-miRNAs (miRBase 19). To eliminate ambiguous multiple hits during the mapping procedure, no mismatch was allowed. Previous studies reported that, when mapped back to pre-miRNAs, sequence reads usually carried mismatches preferentially located at their terminal 3' ends (17)(18)(19)(20). This mismatch was named the 3' end modification. To determine whether the 3' end modification patterns differed among libraries, as described in our previous studies (21), we trimmed and collected the terminal 3' end mismatches one by one. In addition, the remaining perfect match reads had to be at least 18 nt in length. As a result, we kept reads with no less than 18-nt perfect alignment and 3' end modification patterns.
Classifying non-miRNA reads into different data sets. The sequence reads that may not be mapped back to pre-miRNAs were classified into classes by mapping to acquire different data sets with Bowtie (22) and allowing a single nucleotide variation. The sequences of mRNAs and other ncRNAs were derived from the NCBI RefSeq 47 (23). The tRNA sequences were downloaded from the Genomic tRNA database (24) and the rRNA sequences were downloaded from the SILVA database (25). The snoRNA, scaRNA and snRNA sequences were all downloaded from NONCODE (26). The sequence reads not belonging to any of the described RNA classes were uploaded to the RepeatMasker to identify repeat elements, which were classified as unknown.
miRNA expression level according to The Cancer Genome Atlas (TCGA) data. TCGA project collects both cancer and corresponding normal tissues from hundreds of prostate cancer patients. We downloaded all level-3 miRNA expression data of prostate adenocarcinoma from the TCGA Data Portal (https://tcga-data.nci.nih.gov/tcga/dataAccessMatrix. htm). These level-3 data included calculated expressions for each miRNA derived from the Illumina HiSeq sequencing results. A total of 198 tumor samples and 50 normal samples were found at the time the data were downloaded. We kept only the expression data of 50 participants who had both miRNA expression levels from both tumor and normal tissues. Normalized quantification expression levels for these 50 participants were further examined for each investigated miRNA.
Pathway enrichment analysis. We attempted to determine the functions of the miRNA target genes by investigating the pathways with which the miRNA target genes were involved. Therefore, we first downloaded the target genes of differentially expressed miRNAs from TargetScan 6.0, and then mapped the target genes onto the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways based on the Enzyme Commission (EC) numbers by using the R package SubPathwayMiner v.3.1 (27). Subsequently, the hypergeometric test was performed to identify significantly enriched pathways and calculate the false positive discovery rate in the FDR-corrected q-value.

Results
miRNA profiling of radiation-treated prostate cancer cells.
To characterize the mechanism involved in the radiation response of prostate cancer, we used NGS to comprehensively analyze the distribution of miRNAs after radiation treatment in PC3 cells. As indicated in Fig. 1A, PC3 cells were exposed to various dosages of radiation (0, 2, 4, 6, 8, 10, 12, 14, 16 and 18 Gy) and then subjected to a fresh culture medium for an additional 4 days. We found that the growth of the PC3 cells obviously decreased when exposed to 10 Gy of radiation. Therefore, we collected cell RNA at various times (0, 5, 15 and 40 h) following the 10-Gy radiation treatment. We confirmed the expression levels of Cox-2 and p21, which may be induced by radiation at 24 h according to previous studies (16,28). The expression levels of Cox-2 and p21 may be upregulated by radiation treatment in PC3 cells (Fig. 1B). We then performed the comprehensive miRNA profile at various time-points in radiation-treated PC3 cells by using the Illumina HiSeq platform.
Analysis of miRNA sequence reads. Once the samples were sequenced, we collected >9 million clean reads in all libraries (Table I). In addition to miRNA, we also determined After culturing for an additional 4 days, the morphology was observed using light microscopy (x40 magnification). (B) The expression pattern of COX-2 and p21 in radiation-treated PC3 cells was examined using a real-time PCR method. S26 was used as an internal control. (C) The distribution of small RNA reads in 11 categories was classified.  which molecules were the remaining non-miRNA reads. By mapping the non-miRNA reads back to a different data set, we classified the reads into 11 categories. Fig. 1C demonstrates that miRNA accounted for 80% of all clean reads in the prostate cell libraries. Other categories accounted for relatively low proportions, which indicated the high performance of the sample preparation protocol. In addition, the proportions of the categories were considerably similar among libraries, indicating that radiation treatment did not alter the composition of RNA samples in the prostate cell libraries.
After mapping the clean reads to the genome, most of the miRNA reads tend to exist as isomiR. As demonstrated in Fig. 2A, hsa-miR-2110-5p had 5 isomers, whereas the opposite-arm miRNA-3p had 15 isomiRs, which demonstrated that abundant miRNAs tend to have more isomiRs. Our data revealed that the isomiR quantity was highly correlated with miRNA abundance (Pearson's correlation coefficient, 0.91). In addition, we observed that the modified nucleotides were preferentially located at the 3' end of the sequence read (presented in lower case in Fig. 2A). The data indicated that one A nucleotide or one U nucleotide was frequently added at the end of the read. Notably, we found that the proportion of AA dinucleotides modified at the end of the read was gradually increased in a time-dependent manner after the PC3 cells were treated with radiation, which indicated that the 3' end modification may be altered by radiation treatment in PC3 cells. Our previous studies indicated that the use of miR-5p and -3p may be altered in human cancer (29)(30)(31). In the present study, our data indicated that arm selection preference was consistent across nearly all libraries. Only a few cases were observed in which the use of -5p and -3p arm selection had different preferences at various time-points after radiation  (8 to 29). The nucleotides in lowercase type denote the sequence fragments originating from the 3' modification event. (B) The proportion of 3' modifications at end of reads at different time-points after PC3 radiation treatment as observed from NGS data. (C) Fold-change of the -5p/-3p arm of miRNA at different time points after PC3 radiation treatment as observed from NGS data. isomiRs, miRNA isoforms; miRNA, microRNA; NGS, next-generation sequencing. treatment (Fig. 2C). Further research is required to support these findings.
Radiation-response miRNAs in prostate cancer. By summarizing the read count of all the isomiRs that belonged to the same mature miRNAs, we quantified the miRNA expression abundances, and presented the result in transcript per million (TPM). Twenty-two miRNAs were selected and are presented in Table II, demonstrating that their expression levels were altered >2-fold after being subjected to radiation exposure (expression of 6 miRNAs increased, and expression of 16 miRNAs decreased). To explore the putative role contributing to prostate cancer progression, we examined the effect of expression levels of radiation-associated miRNAs on prostate cancer from the available TCGA dataset by using an in silico analysis. We downloaded 100 miRNA expression profiles from 50 prostate cancer patients, including 50 cancer lesion and 50 corresponding normal tissues. As demonstrated in Fig. 3A, the expression levels of radiation-induced miRNAs, miR-25, miR-30a and miR-550a, were significantly upregulated in the prostate cancer cells compared with the corresponding normal tissue cells. Twelve radiation-suppressed miRNAs were identified, i.e. let-7d, miR-15a, miR-17, miR-30d, miR-92a, miR-197, miR-221, miR-320b, miR-342, miR-361, miR-501 and miR-671, and a significantly different expression between prostate cancer and the corresponding adjacent part was found, including 11 upregulated and 1 downregulated (Fig. 3B). Overall, the data indicated that most of the radiation-response miRNAs were identified as dysregulated in prostate cancer according to an in silico analysis (15/22; 1 downregulated, 14 upregulated and the rest demonstrated no change in expression in prostate cancer).
Pathway enrichment analysis of miRNAs. miRNAs can function as either oncogenes or tumor suppressors depending on their target genes. Therefore, identifying a target can facilitate elucidating the role of miRNAs in prostate cancer treatment radiation (32,33). Typically, one miRNA tends to have hundreds of target genes and a group of miRNAs co-modulated as a biological function involved in the regulation of a signaling pathway. Therefore, we further explored the biological function of radiation-response miRNAs by conducting a pathwayenrichment analysis. The putative target genes of miRNAs were obtained from TargetScan 6.0; subsequently, these target genes of the individual miRNAs were mapped onto KEGG pathways. Our data indicated that the target genes of radiationresponse miRNAs were frequently significantly enriched in several cancer-or radiation-related pathways, including the mitogen-activated protein kinase (MAPK), ErbB, p53, Wnt, transforming growth factor-β (TGF-β) and mTOR signaling pathways with an FDR <0.05 (Table III). We also subjected the target genes of the 2-gene set, upregulated miRNA and downregulated miRNAs, to pathway-enrichment analysis. Similar results were observed; their targets were significantly enriched in the prostate cancer pathway (Fig. 4) and radiationrelated pathways, including the MAPK, ErbB, Wnt and TGF-β signaling pathways (Tables IV and V).

Discussion
Our previous studies indicated that the distributions of 3' end modifications and the arm selection preference of miRNAs were different between normal and tumor tissues (29)(30)(31). The -5p and -3p of miRNA play a distant role by suppressing the different target genes. It was previously reported that, in contrast to the oncogenic effect of miR-17 (-5p), miR-17 * (-3P) plays a tumor suppressive role in prostate cancer (9,34,35). The miR-28-5p and miR-28-3p also play opposite roles in colon cancer cell proliferation and migration (36). In the present study, our data showed that the -5p and -3p of particular miRNAs were differently regulated by radiation (shown in Fig. 2C). Several studies have demonstrated that miRNAs contain various ends, which were caused by either RNA editing or non-template nucleotide additions (17,18,20). These miRNA isoforms (isomiRs) contribute to increased miRNA stability or strengthened miRNA-target gene interaction and are differentially expressed in different cellular conditions, including cancer (16,37,38). Our data revealed that the proportion of AA dinucleotide modifications at the end of the read gradually increased in a time-dependent manner after the PC3 cells were treated with radiation, suggesting that radiation may influence the particular miRNA stability or efficiency of silencing targets by regulating the 3' end modifications, which warrants further research. miRNAs are known to function as gene silencers and are involved in modulating biological functions, including cell growth, apoptosis, the cell cycle and the metastasis of cancer (39). Comprehensive miRNA profiling of prostate cancer has indicated that several miRNAs are differentially expressed between prostate cancer and the adjacent normal, which contributes to prostate cancer progression (40)(41)(42). In the present study, we analyzed miRNA expression from   TCGA database and found that the expression levels of radiation-induced miRNAs were frequently dysregulated in prostate cancer (Fig. 3). Our results are consistent with those of previous studies and demonstrated that miR-25, miR-17, miR-30d and miR-92a are overexpressed, and miR-221 is downregulated in prostate cancer (9,(42)(43)(44). However, the expression levels of let-7d and miR-15a decreased according to TCGA, which contradicted the results of previous studies (45)(46)(47). These dysfunctional miRNAs have potential to be used as biomarkers for prostate cancer prognosis or diagnosis. Therefore, understanding the function of miRNAs may provide practical benefits for clinical applications. Predicting the outcome of cancer treatment is the most promising application of miRNAs. Gonzales et al found miR-141 to be consistent with changes in other conventional biomarkers and to the clinical outcomes, suggesting that miR-141 can be used as a marker for monitoring therapeutic response in prostate cancer patients (48). The prognostic value of miRNA expression profiling in prostate cancer has also been demonstrated by Hulf et al. They demonstrated that DNA methylation and histone H3K9-deacetylation of the miR-205 locus is associated with miRNA silencing and deregulation of MED1, which is predictive of a poor prognosis in localized prostate cancer (49). Ionizing radiation is one of the 3 primary modalities used in cancer therapy. Radiation induces considerable DNA damages, which, if not repaired, cause cancer cells to progress to apoptosis and cell cycle arrest. Some cancer cells are resistant to radiation treatment due to activation of complex signaling pathways that counteract these damages, including ErbB, nuclear factor κB (NFκB), MAPK, PI3K/AKT and transforming growth factor-β (TGF-β) signaling pathways (50)(51)(52). Several radiation-related miRNAs have been identified that contribute to the radiosensitivity of cancer cells by modulating the radiation-response signaling pathway (51,52). Since miRNAs are generally slightly repressed by their target genes, the alteration of an individual miRNA is insufficient for accomplishing a biological function. Previous studies have introduced the concept of miRNA regulatory modules (MRMs), which potentially serve as a model for understanding the detailed influences of miRNAs in cellular biological functions (53)(54)(55). Therefore, in the present study, we were particularly interested in the consequences of changes in a group of radiation-induced miRNAs in prostate cancer. Our data indicated that targets of the co-expressed miRNAs were enriched in a radiation-related signaling pathway, suggesting that they co-modulated an abundance of target genes in the same pathway (Table III). miRNAs regulate various factors in radiation-related biological pathways and may affect the radiosensitivity of tumor cells (51). Radiation-response miRNAs have been identified in prostate cancer by using a microarray approach (13)(14)(15). By comparing these data, we identified known and unknown radiation-response miRNAs in prostate cancer by using an NGS approach. Li et al reported that the expression levels of miR-9, miR-22 and miR-30a decreased in radiation-treated PC3 cells (14). Radiation reduced the expression level of an miR-17-92a cluster and the let-7 family in prostate cancer (15). In the present study, we also identified radiation-response miRNAs that had been reported in other types of cancer but not in prostate cancer, such as miR-25, miR-15a, miR-30d, miR-125a, miR-221 and miR-342 (21,(56)(57)(58)(59)(60)(61)(62)(63). In addition, we identified a group of radiation-response miRNAs that have not been reported in any type of cancer. The pathway-enrichment analysis revealed that their targets are frequently enriched in the radiation-response signaling pathway. In summary, in the present study, we thoroughly investigated radiation-response miRNAs, which may be involved in the radiosensitivity of prostate cancer, by modulating radiationrelated signaling pathways using an NGS approach. These miRNA candidates may be effective targets for improving the efficacy of radiation treatment in future prostate cancer therapy. In addition, we observed that 3' end modifications and the -5p/-3p arm selection of miRNAs were altered in prostate cancer after radiation treatment. These finding require further research.