Bladder urothelial carcinoma is the fourth most common and eighth deadliest cancer among men, with four times as many males diagnosed than females among the approximately 80,000 annual cases in the United States [1
]. One quarter to one third of bladder carcinomas invade bladder muscles and become muscle invasive bladder cancer (MIBC) [2
]. This more aggressive form of bladder cancer has a greater than 10% lower survival rate than other forms and a 20–40% chance of recurrence [2
]. However, the complicated molecular mechanism and diverse etiology of bladder cancer development hinder improvements in the detection and treatment of MIBC, and there has been no new treatment methods approved for over 30 years [3
]. Therefore, there is a pressing need for identifying a functional biomarker for the early identification and treatment of MIBC, as well as for differentiating MIBC from non-muscle invasive bladder cancer. This study will comprehensively investigate the role of one such promising biomarker candidate, long non-coding RNAs (lncRNAs), in MIBC, using large-scale bioinformatics analyses.
Long non-coding RNAs (lncRNAs) are a category of non-coding RNAs with a length longer than 200 nucleotides [4
]. In the past few years, studies of lncRNAs have revealed their roles in regulating a diverse number of critical cellular processes, including transcription and interaction with the immune system [4
]. LncRNAs are implicated in both tumor suppressing and oncogenic functions in carcinogenesis, including that of bladder cancer [5
]. They are attractive targets for investigation in cancer, because of their potential to serve both as diagnostic markers and as therapeutic targets. They may be excellent diagnostic markers, because there are often dramatic changes in their levels in cancer samples compared to normal samples, and they can be detected noninvasively in peripheral blood, often as part of exosomes [6
]. They may also serve as useful therapeutic targets, because they may be used to regulate proteins that are not targetable, such as those that need to be upregulated or activated [6
Extensive research has been performed on the manipulation of one or more lncRNAs in vitro, and the resulting changes in bladder carcinoma cells have been observed [7
], which has led to the identification of multiple lncRNAs that are upregulated and downregulated in bladder carcinoma. However, there is still a relative lack of comprehensive analyses examining the landscape of all lncRNA expression in MIBC. Such analyses would not be possible in vitro, but could be done through computational profiling. Previous bioinformatics analyses have examined lncRNAs in MIBC using GENCODE annotations [10
], but no study has performed a more comprehensive assessment of lncRNAs using LNCipedia annotations. Finally, most of the previous lncRNA studies examined known pathways involved in tumor initiation and proliferation and are limited by the number of samples or patients they investigated [8
]. In this study, we utilized RNA-sequencing data of MIBC samples from 406 patients, downloaded from The Cancer Genome Atlas (TCGA) database. We discovered a panel of 20 lncRNAs significantly dysregulated in MIBC that are also significantly associated with patient survival and clinical variables. We correlated lncRNA expression with genomic alterations to investigate possible mechanisms for lncRNA dysregulations, and we analyzed the lncRNAs’ association with cancer and immune-associated pathways and cancer-related signatures. Collectively, our data suggest that these 20 lncRNAs may serve as a potentially useful biomarker panel for predicting patient prognosis, clinical characteristics, and genomic profiles based on their expressions. As these lncRNAs are clinically relevant, they should also be investigated as potential treatment targets.
Long non-coding RNAs are known to play important roles in cancer pathogenesis pathways, including promoting cancer initiation and maintaining cancer development, illustrating their great potential as diagnostic biomarkers or therapeutic targets for cancer [17
]. Many lncRNAs have been implicated in the regulation of proteins through direct interaction [3
]. In this study, we comprehensively examined lncRNA dysregulation in MIBC in the context of clinical relevance, potential interaction with proteins, and correlation with genomic alterations in MIBC, in order to identify the most significant lncRNAs implicated in bladder cancer.
We identified 3191 lncRNAs, among 107,039 known lncRNAs, to be differentially expressed in a patient cohort of 406 MIBC samples vs. adjacent normal samples (FDR < 0.05). Out of these lncRNA candidates, we found 20 key lncRNAs to be the most critically involved in MIBC, based on correlation with patient survival rates (univariate Cox regression analysis, p
< 0.05) and association with selected clinical variables (Kruskal–Wallis test, p
< 0.05). To the best of our knowledge, none of these 20 lncRNAs have been found to be implicated in bladder cancer. The lack of overlap between the lncRNAs we discovered and the lncRNAs reported in the previous manuscript is partly due to many studies documenting only one or a few lncRNAs implicated in bladder cancer [10
]. Because we are examining a larger number of lncRNAs than most papers, the lncRNAs discovered by other studies could be insignificant in our study after multiple comparison correction. However, we did discover that several lncRNAs found to be significant in Wen et al. [18
] are also significantly dysregulated in our analysis, although they were not part of the 20 key lncRNAs in our study. We have investigated a significantly larger set of lncRNAs than similar studies profiling lncRNAs in bladder cancer by using lncRNA annotations from LNCipedia, which seeks to capture all known lncRNA transcripts by compiling and standardizing transcripts discovered from different literature sources [19
]. Studies using GENCODE annotations would only examine 19,812 transcripts, compared to the 107,039 transcripts in LNCipedia [10
]. Previous studies have identified that lncRNAs like LINC00460
, and DUXAP8
are implicated in bladder cancer, but did not examine the entire landscape of lncRNA transcripts [11
We then used GSEA to determine which canonical pathways and oncogenic signatures were most associated with the 20 clinically significant lncRNAs. When performing GSEA for canonical pathways using the C2 gene set, we found that some lncRNAs, including lnc-BOD1-1:7, -1:8, and -1:9, as well as lnc-GCH1-2:1, -2:2, and -2:3 were associated with many immune- and cancer-associated pathways. Other lncRNAs were mainly correlated with either immune-associated pathways, with lnc-CGRRF1-3:1 as an example, or cancer-associated pathways, with lnc-ACSBG2-1:1 and lnc-ANKRD54-1:1 as examples. Finally, lncRNAs such as lnc-EIF2AK4-1:1, -1:4, and -1:5, as well as lnc-SERF1B-1:4 were not correlated with many of the immune-associated or cancer-associated pathways. For determining which oncogenic signatures were most associated with the 20 key lncRNAs, we performed GSEA using the C6 gene set, and found that lnc-CGRRF-3:1, lnc-ACSBG2-1:1, and lnc-BOD1-1:7, -1:8, and -1:9 correlated with the most oncogenic signatures. Notably, lnc-BOD1-1:7, -1:8, and -1:9 were highly correlated with both pathways’ gene sets and oncogenic signatures. In terms of pathways, those that are associated with the greatest number of lncRNAs include pathways involved in apoptosis and the downregulation of genes SMAD2/3 and SMAD4, complement pathways, and the Wnt/Ca2+ pathway. With regards to signatures, KRAS oncogenic signatures were found to have the most significant association with lncRNAs.
Finally, to understand possible causes of lncRNA dysregulation, we used REVEALER to correlate lncRNA expression with genomic alterations in MIBC. Interestingly, we found that none of the 20 significant lncRNAs were correlated with any mutations, and instead were only correlated with genomic amplifications and deletions. More upregulated lncRNAs are associated with genomic alterations than are downregulated lncRNAs, and several CNA hotspots, including 13q12-14, 14q24-32, and 3p22-24, are correlated with the expression of multiple upregulated lncRNAs. All the above loci have been previously implicated in bladder cancer risk [22