The prognostic value of long noncoding RNAs in prostate cancer: a systematic review and meta-analysis

The abnormally expressed LncRNAs played irreplaceable roles in the prognosis of prostate cancer (PCa). Therefore, we conducted this systematic review and meta-analysis to summarize the association between the expression of LncRNAs, prognosis and clinicopathology of PCa. 18 eligible studies were recruited into our analysis, including 18 on prognosis and 9 on clinicopathological features. Results indicated that aberrant expression of LncRNAs was significantly associated with biochemical recurrence-free survival (BCR-FS) (HR = 1.55, 95%CI: 1.01–2.37, P < 0.05), recurrence free survival (RSF) (HR = 3.07, 95%CI: 1.07–8.86, P < 0.05) and progression free survival (PFS) (HR = 2.34, 95%CI: 1.94–2.83, P < 0.001) in PCa patients. LncRNAs expression level was correlated with several vital clinical features, like tumor size (HR = 0.52, 95%CI: 0.28–0.95, P = 0.03), distance metastasis (HR = 4.55, 95%CI: 2.26–9.15, P < 0.0001) and histological grade (HR = 6.23, 95% CI: 3.29–11.82, P < 0.00001). Besides, down-regulation of PCAT14 was associated with the prognosis of PCa [over survival (HR = 0.77, 95%CI: 0.63–0.95, P = 0.01), BCR-FS (HR = 0.61, 95%CI: 0.48–0.79, P = 0.0001), prostate cancer-specific survival (HR = 0.64, 95%CI: 0.48–0.85, P = 0.002) and metastasis-free survival (HR = 0.61, 95%CI: 0.50-0.74, P < 0.00001)]. And, the increased SChLAP1 expression could imply the worse BCR-FS (HR = 2.54, 95%CI: 1.82-3.56, P < 0.00001) and correlate with Gleason score (< 7 vs ≥ 7) (OR = 4.11, 95% CI: 1.94-8.70, P = 0.0002). Conclusively, our present work demonstrated that LncRNAs transcription level might be potential prognostic markers in PCa.


INTRODUCTION
Prostate cancer (PCa) is the most commonly diagnosed cancer and the third leading cause of cancerrelated death in men [1]. Histopathological evaluation of biopsy has been set as the golden standard for the diagnosis of PCa, while the drawbacks like infection and bleeding restrained the clinical use [2]. The surveillance for biochemical recurrence (BCR) is one of the vital parameter throughout the treatment of PCa. The low specificity of the widespread diagnostic marker, prostatespecific antigen (PSA), makes it difficult to distinguish Review Oncotarget 57756 www.impactjournals.com/oncotarget indolent or aggressive cancer stages [3]. Without other valuable predictive parameters for early prostate cancer screening, most diagnoses are made in the terminal stage due to the lack of specific and sensitive methods for early prostate cancer screening [4]. Since the high degree of intra-cancer and inter-patient heterogeneity at the molecular level [5], it is an effective to profile the expression of multiple genes to establish the molecular processes occurring in the prostate cancer.
Long non-coding RNAs (LncRNAs) are a class of RNA with transcripts longer than 200 nucleotides and lack functional open reading frames [6]. They can be polyadenylated and may operate in nuclear and/ or cytoplasmic fractions. The lack of opening reading frames can either be intergenic, that is located between protein-coding genes, or intragenic, located within an intron of a host protein-coding gene or on the antisense strand [7]. Owing to their biological properties and clinical value in diagnosis, prognosis, and treatment, LncRNAs have been widely investigated. LncRNAs involve in various cell biological processes, like cellular differentiation, proliferation, DNA damage responses and chromosomal imprinting. The abnormal expression of LncRNAs has been reported in various human diseases, including tumors [8]. The lncRNA metastasis associated lung adenocarcinoma transcript 1 (MALAT1) played a vital role in metastasis formation in lung cancer and was a potential therapeutic target [9]. LncRNA-activated by TGF-β (lncRNA-ATB) was significantly up-regulated in hepatocellular carcinoma (HCC) metastases and associated with poor prognosis [10].
In prostate cancer, a well-known example of LncRNAs is the prostate cancer antigen 3 (PCA3; also known as DD3), which overexpresses and promotes invasion and migration in prostate cancer cells by miR-1261 sponging [11]. The level of PCA3 in urine has been used as a diagnostic biomarker for PCa with a sensitivity of 58-82% and a specificity of 56-76% [12][13][14]. The urinary PCA3 is now widely used for prostate cancer detection and has been approved by the US Food and Drug Administration (FDA) [15]. The expression pattern of lncRNAs also along with coding genes could serve as a prognostic marker. Sun et.al found that MALAT1 was dramatically elevated in human prostate cancer tissues, and its expression was highly associated with Gleason score, tumor stage, PSA level and castration resistance [16]. Besides, decreased expression level of prostate cancer associated transcript-14 (PCAT-14) was prognostic for the metastatic disease and poor survival for patients with prostate cancer [17].
The abnormal expression of lncRNAs could be of prognostic significance. The prognostic value of LncRNAs in PCa has been explored by many studies. The most commonly used methods for detecing prognostic significance include microarray, qRT-PCR, in situ hybridization assay (ISH) and available database. However, the inaccuracy and insufficiency caused by the small size and single experiment program might interfere with revealing the real profiles of LncRNAs in PCa. We assumed that he true prognostic value of lncRNAs in PCa could be unravelled through multiple sensitive and reliable detection methods in large scale, multicenter studies. Therefore, we performed the meta-analysis to estimate systematically to explore the potential value of LncRNAs in the prognosis and clinical outcomes in PCa among a relatively larger amount of PCa patients.

Publication bias and sensitivity analysis
We applied Begg's test to estimate the publication bias among these studies. All the Begg's tests in our analysis showed no publication bias, due to the value of P > 0.05, respectively. The sensitivity analysis which was performed by Stata11.0 software evaluated the stability of our results. We found that no individual study significantly interfered with the overall results which demonstrated the credibility of the present meta-analysis ( Supplementary Figures 1-4).

DISCUSSION
Long non-coding RNA contained more than 200 nucleotides constitutes a great proportion of non-coding transcripts [35]. Many LncRNAs exhibited cell-type specific expression and located in specific subcellular compartments [36,37]. LncRNAs could function as a role of molecular scaffolds for targeting gene regulatory proteins/complexes to specific genomic loci [7]. So, they could influence the expression of target proteins of neighboring protein-coding genes, regulate the distal transcriptional elements and modulate the activity of protein-binding partners [38][39][40]. Furthermore, LncRNAs could act as a suppressor or activator of gene expression. The increase or decrease of a number of LncRNAs contribute to oncogenesis by influencing many cellular processes [41].
The aberrant expression of LncRNAs is related to the development and progression of prostate cancer through affecting tumor cell proliferation, metastasis, selfrenewal, survival, and apoptosis by either transcriptional or post-transcriptional regulation [42]. Several PCa-specific LncRNAs have been reported, and some are associated with distinct subtypes of the disease. In prostate cancer, the up-regulated prostate cancer antigen 3 (PCA3; also known as DD3), is already available as a diagnostic test in urine [43,44]. It has indicated that the overexpressed PCA3 could modulate prostate cancer cells survival by altering androgen receptor (AR) signaling [45]. Besides, the lately study elaborated that PCGEM1 and PRNCR1, bound successively to the androgen receptor and strongly enhanced both ligand-dependent and ligand-independent androgen-receptor-mediated gene activation programs and proliferation in prostate cancer cells [46]. Apart from Gleason score, the increased expression of SChLAP1 was validated as a significantly prognostic biomarker for metastatic prostate cancer increased with prostate cancer progression and predicted the poor clinical outcome in patients with localized prostate cancer following radical prostatectomy and patients with lethal prostate cancer [27,31,47] . The upregulation of SChLAP1 in PCa patients could lead to poor outcomes, including metastasis and prostate cancer-specific mortality, by antagonizing the tumor-suppressive functions of the SWI/SNF complex [32]. While, a novel prostate cancer and lineage-specific LncRNA PCAT14, which is transcriptionally regulated by AR, is overexpressed in low grade disease and lack of PCAT14 predicts for disease aggressiveness and recurrence in PCa [30].
On the purpose of detecting the prognostic value of LncRNAs in PCa, we performed this comprehensive systematic review and meta-analysis of the current literature which is the first systematical analysis of the relationship between LncRNAs expression level with    Our results demonstrated that the high expression of 11 LncRNAs was related with poor prognosis, so was the low expression of 6 LncRNAs. PCAT14 and SChLAP1 were reported by no less than two studies, thus, subsequently, we conducted meta-analysis for prognostic value of these two LncRNAs in PCa, respectively. We found that the decreased PCAT14 expression could predict poor OS, BCR-FS, PSS and MFS in PCa patients. While the overexpressed SChLAP1 among PCa patients had worse BCR-FS. Regarding the relationship with clinicopathological features, the increased expression level of CCAT2 and LOC400891 could be the identifiers of an existence of distance metastasis, tumor diameter (≤ 2.5 vs > 2.5 cm) and histological grade (II vs III + IV) for PCa. The level of SChLAP1 existed a significant difference between the group with Gleason score < 7 and ≥ 7. The non-significant correlation between LncRNAs and other characters might be caused by the insufficient studies for each LncRNA.
However, several limitations existed in our analysis should be considered. The included studies in our metaanalysis weren't sufficient with limited sample size and all were English researches. No study with negative results was included in our analysis which could amplify the relation between LncRNAs and clinical values of PCa. Studies contained diverse LncRNAs used different followup endpoints. Besides, the cut-off value distinguished high or low levels of LncRNAs differed among these studies.
In conclusion, our study was the first meta-analysis to evaluate the clinical value of expression level of LncRNAs in prostate cancer. Despite the limitation, we demonstrated that transcription level was correlated with prognosis of PCa and several vital clinical characters. However, further comprehensive and large-scale research should be performed to confirm our findings.

Literature search strategy and study eligibility criteria
We searched databases like PubMed and Web of Science for studies published in English up to February 17, 2017. The following keywords were used "Long noncoding RNA" or "Long intergenic non-coding RNA" or "lncRNA" or "LincRNA" and "prostate cancer" or "PCa" with the limit to human.

Criteria of eligibility
The inclusion criteria for our meta-analysis were: (1) articles published as a full paper in English; (2) all patients were diagnosed with PCa; (3) LncRNAs expression levels were measured in PCa tissues; (4) the association of LncRNAs with survivals (OS/ BCR-FS/ RFS/ DFS/ MFS/ PSS/ PFS) was detected; (5) correlation between LncRNAs and clinicopathological features was performed at least two parameters; (6) studies provided sufficient information to estimate hazard ratios (HR) and 95% confidence interval (95% CI). Studies which failed to provide enough data were excluded from this metaanalysis. Only the latest or most complete data were chosen when we dealt with duplicated publications.

Data extraction
The usable data were extracted independently by two reviewers (Ma WJ and Jing W). Any disagreements between the three reviewers were resolved by consensus involving other two reviewers (Chen X, Ding L and Ma JH). The reviewers screened the name of first author, year of publication, country, the type of LncRNAs, a method for detection of LncRNAs, cut-off value and the follow-up time, clinicopathological parameters and the HRs with 95% CIs for survival analysis.

Statistical analysis
The HRs and 95% CI were used to evaluate the association between lncRNAs and prognosis. A provided HR > 1 implied a poor survival for the high expressed lncRNAs group. On the contrary, HR < 1 meant a worse survival for the group with decreased lncRNAs expression level. We extracted HR according to the following two methods: (1) The HRs and 95% CI were obtained directly from the publication; (2) We calculated the HRs and 95%CI by extracting several survival rates from the Kaplan-Meier survival curves using Engauge Digitizer version 4.1. The second method may generate errors by variation. Meanwhile, Aiming to investigate the relationship between the expression of lncRNAs and clinicopathologic characteristics, the ORs and 95% CI were used.
All analyses were performed using the STATA software version 11.0 and Cochrane Collaboration Review Manager Version 5.2. To investigate the heterogeneity among studies, I 2 statistics and chi-square Q test were used. When I 2 value more than 50% or a P-value less than 0.05 for Q test, the heterogeneity was regarded as significant. Fixed-effects model was used when there was no significant heterogeneity between studies. Otherwise, the randomeffects model was used. We also performed sensitivity analyses to test the effect of each study on pooled results. Begg's test was applied for assessing publication bias. Statistical significance was defined when a P < 0.05.

ACKNOWLEDGMENTS AND FUNDING
This work was financially supported by the Hubei Province health and family planning scientific research project (WJ2017Z007).