Prognostic significance of long non-coding RNAs in clear cell renal cell carcinoma

Abstract Background: Clear cell renal cell carcinoma (ccRCC) is the most common form of kidney cancer in adults, and patients with advanced ccRCC have a 5-year survival rate of <30%. The poor prognosis of ccRCC is closely related to its lacking of potential therapeutic and prognostic biomarkers. This meta-analysis aimed to elucidate the precise prognostic value of long non-coding RNAs (lncRNAs) in patients with ccRCC. Methods: A literature search was performed in related databases up to January 31, 2019. Hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) were calculated to explore the relationship between special lncRNAs expression and survival in patients with ccRCC. Results: After literature researching, a total of 16 studies, including 13 lncRNAs were identified. The data from studies that investigated the association between lncRNA expression and survival outcomes in patients with ccRCC were extracted. Results revealed that lncRNAs expression was significantly associated with poor overall survival (OS) outcome in patients with ccRCC (HR = 1.71, 95%CI = 1.40–2.01 in up-regulated subgroup; HR = 0.53, 95% CI = 0.25–0.80 in down-regulated subgroup). The overexpression of PVT1 was significantly associated with poor OS in ccRCC (HR = 1.51, 95% CI = 1.02–2.00). Meanwhile, up-regulation of LUCAT1 was significantly related to worse OS in ccRCC patients (HR = 1.51, 95% CI = 1.01–2.00). Conclusions: These results suggest that lncRNAs could be used to predict unfavorable prognosis and function as potential prognostic biomarkers in ccRCC.


Introduction
Renal cell carcinoma (RCC) is one of the most common malignant tumors in urinary system, accounting for about 5% of adult malignant tumors. [1] Clear cell renal cell carcinoma (ccRCC) is the commonest subtype of RCC, which makes up approximately 3-fourths of RCC. [2] Due to lacking of early diagnostic markers and the strong resistance of ccRCC to radiotherapy and chemotherapy, the overall prognosis of patients with ccRCC is still poor. [3] In order to improve the prognosis, new markers need to be found for early diagnosis and decisionmaking related to the selection of appropriate treatment.
Long non-coding RNA (lncRNA) is a newly discovered class of non-coding RNA with >200 nucleotides in length. Accumulating studies have shown that lncRNAs are involved in many physiological and pathological processes, such as cell growth, apoptosis, stem cell pluripotency and development, by acting as transcriptional, post-transcriptional or epigenetic regulators. [4,5] Recently, mounting evidences reveal that lncRNAs play important roles in human diseases, especially in tumorigenesis and tumor prognosis, which may be potential biomarkers and therapeutic targets. [6,7] Currently, the special expression profiles of lncRNAs have been observed in ccRCC and dysregulated lncRNAs regulate oncogenesis and metastasis-associated genes. [8,9] Therefore, lncRNAs might be feasible as diagnostic biomarkers and prognostic factors. In order to explore the correlation between dysregulated lncRNAs and prognosis of patients with ccRCC, we conducted this quantitative meta-analysis.

Trial selection
All published controlled clinical trials involving the relationship between lncRNAs and clinical values in ccRCC patients were collected. The analyzed data were clinicopathology and prognosis of ccRCC patients during the follow-up. Studies without original data related to survival outcome were excluded. All analyses were performed based on previously published studies. Thus, no ethic approval and informed consent were required in this study.

Search strategy
A literature search was performed in Pubmed and Web of Science (up to January 31, 2019), EMBASE (up to 2018), and Cochrane Controlled Trials Register (up to 2018) to identify relevant clinical controlled trials in English, which was conducted with the terms "lncRNA" and "renal cancer". The detail searching strategy in Pubmed was ((long non conding RNA) OR lncRNA) AND (cancer OR tumor OR carcinoma) AND (renal OR kidney OR (renal cancer) OR (renal cell cancer) OR (renal cell carcinoma)). The studies involving the prognosis of patients with ccRCC were adopted for further screening. And studies from the bibliographies of retrieved trials were also scanned.
The inclusion criteria were formulated as follows: (1) English literature; (2) follow-up completed in ccRCC patients after surgery; (3) studies that detected the expression level of lncRNA in tissue and presented the clinicopathological features, such as age, tumor diameter, tumor stage and so on; (4) studies that investigated the association between the expression level of lncRNA and survival outcome; (5) studies that provided a Hazard ratio (HR) or relative risk (RR), 95% confidence interval (CI) and P value, and Kaplan-Meier curves or required data obtained by contacting corresponding authors.
This meta-analysis adopted the following criteria to exclude irrelevant studies: (1) non-human trials; (2) non-English literature; (3) studies performed only in cellular level; (4) ccRCC patients 40 cases; (5) studies included other type RCC patients, for example papillary renal cell carcinoma and chromophobic renal cell carcinoma; (6) studies without follow-up information; (7) studies focusing on lncRNA genetic alterations, including methylation patterns or polymorphisms; (8) the analysis of HRs based on multiple lncRNAs; (9) the duplicate data from The Cancer Genome Atlas; and (10) studies that could not obtained original and sufficient data for HR and 95% CI estimation.

Data extraction
Two investigators (Wang Yan and Li Zhan) were blinded to each other when reviewing the trials and independently selected studies using the following steps: (1) examining titles and abstracts to remove obviously irrelevant reports; (2) retrieving the full text of potentially relevant reports; (3) examining full-text reports for compliance of studies with eligibility criteria; (4) making final decisions on study inclusion. Any discrepancies were resolved by consensus. The quality assessment of these included studies was performed in accor-dance with the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) criteria. If a consensus could not be reached, the senior author (Li Wei) made the final decision for trial eligibility and data extraction. For each study, the following information was recorded: first author's name, year of publication, country, number of patients, median or mean age of patients, sample size and type, detection method, cutoff definition, follow-up time, and HRs associated with dysregulated lncRNAs expression for over survival (OS), disease-free survival (DFS) or progression free survival (PFS), along with their 95% CIs and P values.

Statistical analysis
The HRs and 95% CI of survival outcome were calculated for each study. In lncRNAs up-regulated group, HR > 1 implied a worse survival. Conversely, an observed HR < 1 suggested poor survival for lncRNA down-regulated group. A pooled effect was calculated using a random effect model to take into account within-study and between-study variance. Heterogeneity was assessed with the Q statistic (the result of a statistical test based on the Q statistic (P)) and the I 2 statistic. For the Q test, a P value less than .05 indicated significant heterogeneity; for the I 2 statistics, an I 2 value greater than 50% was considered significant heterogeneity. If I 2 value less than 50%, a fixed effect model was performed; otherwise, a random effect model was applied for the metaanalysis. While statistically significant heterogeneity was found, the source would be explored by subgroup analysis and/or meta-regression.
The sensitivity analysis was performed to examine the consistency of the overall effect estimate respectively and publication bias was examined using funnel plot. All statistical procedures in this meta-analysis were performed using STATA 14.0 and Revman 5.2. The P < .05 was considered as statistically significant.

Search result
As shown in Figure 1, 338 records relating to lncRNA expression and ccRCC were identified in PubMed, Embase, Web of Science, and Cochrane databases. After screening titles, 233 records were excluded. After the abstracts were reviewed, 39 full-text articles were assessed for eligibility. Finally, after excluding 23 articles, a total of 16 articles were included in the current metaanalysis.
3.4. Correlation between lncRNAs expression and prognosis 3.4.1. Association between lncRNAs and OS. These 16 articles evaluating OS were divided into up-regulated subgroup and down-regulated subgroup according to the lncRNAs expression level in patients with ccRCC. Significant heterogeneity between included studies was observed in down-regulated subgroup data (I 2 = 91.2%). Therefore, the random effect model was used in the meta-analysis of down-regulated subgroup to calculate the pooled HR and 95% CI. While the fixed effect model was used for up-regulated subgroup (I 2 = 6.2%). The results showed that expression of lncRNAs was significantly associated with poor OS outcome in patients with ccRCC, with a pooled HR of 1.71 (95% CI = 1.40-2.01; Fig. 3) in up-regulated subgroup and a pooled HR of 0.53 (95% CI = 0.25-0.80; Fig. 3) in down-regulated subgroup.

Association between lncRNAs and DFS/PFS.
Among the 4 articles evaluating DFS/PFS, no statistically significant heterogeneity was evident (I 2 = 28.1%, P = .238 in up-regulated subgroup; I 2 = 0.0%, P = .586 in down-regulated subgroup). Therefore, a fixed effect model was used to calculate the pooled HR and 95% CI. The overexpression of PVT1 was found to be not significantly correlated with poor DFS /PFS in patients with ccRCC, with a pooled HR of 1.54 (95% CI = 0.93-2.14; Fig. 5). Moreover, the low expressions of NONHSAT123350 and lnc-ZNF180-2 were significantly associated with poor DFS/PFS outcome in patients with ccRCC, with a pooled HR of 0.79 (95% CI = 0.69-0.88; Fig. 5).

Heterogeneity analysis and publication bias
Sensitivity analysis was performed by meta-based Influence Analysis using the fixed effect model in OS analysis of upregulated subgroup. This test suggested that the Li2017 study had influenced the overall result of the OS analysis (Fig. 6A). When the Galbraith plot was assessed, Li2017 study was identified as an outlier causing heterogeneity in the OS analysis (Fig. 6B). By excluding this study, a similar and significant pooled HR was obtained (HR = 1.63, 95% CI = 1.30-1.96). Moreover, heterogeneity was found to be absent (I 2 = 0.0%, P = .453; Fig. 6C). Begg's funnel plots were used to evaluate publication bias (Fig. 6D). For the pooled analyses of OS in up-regulated subgroup, the Begg test showed a P value of .283 after continuity corrected. Therefore, this meta-analysis was free of notable publication bias. No conclusive graph could be generated because of the small size of down-regulated subgroup and DFS/PFS analysis. So, the heterogeneity and publication bias analyses were not evaluated.

Discussion
Non-coding RNAs profiling has been found to be useful in the prediction of cancer patient's clinical outcome, metastatic recurrence and survival time. [27,28] During carcinogenesis, genetic alterations could drive tumor evolution toward higher grades of malignancy, however, the extent to how the lncRNAs alterations influence this process remains incompletely understood. The up-regulation or down-regulation of many lncRNAs, such as MALAT1 and H19, contributed to oncogenesis by affecting proliferation, energetic metabolism and other cellular processes. [29][30][31] Many long non-coding RNAs have been identified in various cancer genomes, and can be used as novel biomarkers for cancer. [32] As multiple reports suggest, lncRNAs play an important role in urothelial carcinogenesis, such as bladder cancer. [33] Some lncRNAs, for example, MAGI2-AS3 and ADAMTS9-AS2 may serve as candidate diagnostic biomarkers or therapeutic targets for bladder cancer. [34] The pathogenesis and progression of renal cancer involve multistep changes in the gene profiling, and a variety of approaches have been used to uncover the molecular profiles www.md-journal.com that contribute to ccRCC development and tumor progression. Many metabolic pathways in ccRCC are reprogrammed. [35] The metabolomic profile demonstrated that mitochondrial dysfunction and oxidative stress potentially associated with tumorigenesis and tumor progression. [36] The specific molecular alterations occur in ccRCC under metabolic reprogramming, but the role of lncRNAs is still unclear. [37] The potential value of lncRNAs for ccRCC remains further investigation. In order to find some biomarkers for ccRCC, we conducted this meta-analysis to systematically analyze the association between lncRNA expression and prognosis of patients with ccRCC. After literature searching and screening, a total of 16 studies were included. In the classification of clinicopathological features, most lncRNAs have little association with the tumor size. The up-regulated PANDAR, H19, SPRY4-IT1 could be potentially considered as novel biomarkers to detect distant metastasis in ccRCC patients in early stages. The dysregulated PANDAR, NBAT-1, H19, LUCAT1, lnc-ZNF180-2, MALAT1, SPRY4-IT1 might be used as biomarkers of lymph node metastasis. Except MALAT1, PVT1, and LUCAT1, the other lncRNAs were detected by single study. The characteristics of studies on MALAT1 have been performed. [26] And the studies on LUCAT1 did not included the relationship between LUCAT1 expression and clinicopathological features in patients with ccRCC. We then combined the 2 studies on PVT1 with 4 groups to clarify the correlation between clinicopathological data and PVT1 level. The expression of this lncRNA was related to the gender of patients with ccRCC but not age. PVT1 was significantly associated with the Fuhrman grade and TNM stage of patients with ccRCC after we pooled HR and P value. Because of the limitations of included studies, further studies should be conducted to verify these conclusions.
Among these 14 lncRNAs, the overexpression of PANDAR, lncRNA H19, PVT1, linc00152, LUCAT1, MRCCAT1, MALAT1, SPRY4-IT1, and lncRNA MIAT were associated with poor prognosis; meanwhile the low expression of CADM1-AS1, ENSG241684, lnc-ZNF180-2, NONHSAT123350, NBAT1 were related to poor prognosis. Although these dysregulated lncRNAs were associated with the prognosis of patients with ccRCC, merely MALAT1, PVT1, and LUCAT1 were not reported solely. The association between the expression level of MALAT1 and OS has been performed. [26] We then In this study, PVT1 is among those lncRNAs, which is overexpressed and associated with tumorigenesis and poor prognosis in a range of cancers. [38,39] Chen et al found that overexpression of PVT1 could promote the proliferation, cell cycle progression, and migration of melanoma cells. [40] Moreover, Gao et al clarified that after transfected with PVT1 siRNA, the proliferation, migration, and invasion of cervical cancer cells were greatly decreased. [41] The overexpressed PVT1 regulated miR-195 pathway to influence lipid metabolism and the metabolic reprogramming was verified to be potential mechanism to promote proliferation and invasion of tumor cells. [42] The    [43] Li et al revealed that the expression of PVT1 was up-regulated in ccRCC tissues, and knockdown of PVT1 induced apoptosis by increasing the expression of poly ADP ribose polymerase and Bcl-2-associated X protein, and promoted cell cycle arrest at the G1 phase by decreasing the expression of cyclin D1. This study indicated that PVT1 promoted the progression of ccRCC partly through activation of the epidermal growth factor receptor pathway. [44] These findings suggest that PVT1 may be an oncogenic biomarker for ccRCC, and the PVT1 associated pathway may serve as a novel therapeutic target for treating ccRCC.
Our study revealed that LUCAT1 was correlated not only with clinicopathological features but also with prognosis of ccRCC. Xiao et al revealed that the up-regulated expression of LUCAT1 in ccRCC tissues was associated with tumor grade, clinical pathological stage and survival time; LUCAT1 could bind to polycomb PRC2 complex and suppress p57 expression to promote renal cancer cell proliferation. [45] Zheng et al clarified that the expression level of LUCAT1 was positively correlated with malignant stage and poor prognosis of ccRCC, and LUCAT1 promoted proliferation and invasion in ccRCC cells through AKT/GSK-3b signaling pathway. [11] Wang found that LUCAT1 was critical for proliferation and invasion of ccRCC cells by regulating micoRNA-495-3p and special adeninethymine-rich DNA-binding protein 1. [10] These findings indicate that the LUCAT1 related axis is a potential therapeutic target and molecular biomarker for ccRCC.
This meta-analysis summarized current researches on the relationship between aberrant lncRNAs expression and prognosis of patients with ccRCC. However, several limitations should be considered. First, the number of studies included in our metaanalysis was insufficient and the sample size was limited. Second, most included studies were from China, merely one from Japan, and one from Germany; therefore, our data might not be globally applicable. Thirdly, the standard for dysregulated expression differed among these studies, and it was difficult to obtain the same value. Consequently, further studies and more rigorous criteria will be needed to confirm the function of lncRNAs in ccRCC.

Conclusion
In conclusion, our meta-analyses indicate that the expression levels of some special lncRNAs are correlated with poor survival in patients with ccRCC. The expression of lncRNAs could be used to predict unfavorable prognosis and function as potential prognostic biomarkers in ccRCC.