Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Diagnostic and prognostic value of long noncoding RNAs as biomarkers in urothelial carcinoma

  • Johanna Droop,

    Affiliation Department of Urology, Medical Faculty, Heinrich-Heine-University Duesseldorf, Duesseldorf, Germany

  • Tibor Szarvas,

    Affiliations Department of Urology, Medical Faculty, University of Duisburg-Essen, Essen, Germany, Department of Urology, Semmelweis University Budapest, Budapest, Hungary

  • Wolfgang A. Schulz,

    Affiliation Department of Urology, Medical Faculty, Heinrich-Heine-University Duesseldorf, Duesseldorf, Germany

  • Christian Niedworok,

    Affiliation Department of Urology, Medical Faculty, University of Duisburg-Essen, Essen, Germany

  • Günter Niegisch,

    Affiliation Department of Urology, Medical Faculty, Heinrich-Heine-University Duesseldorf, Duesseldorf, Germany

  • Kathrin Scheckenbach,

    Affiliation Department of Otolaryngology, Medical Faculty, Heinrich-Heine-University Duesseldorf, Duesseldorf, Germany

  • Michèle J. Hoffmann

    michele.hoffmann@hhu.de

    Affiliation Department of Urology, Medical Faculty, Heinrich-Heine-University Duesseldorf, Duesseldorf, Germany

Diagnostic and prognostic value of long noncoding RNAs as biomarkers in urothelial carcinoma

  • Johanna Droop, 
  • Tibor Szarvas, 
  • Wolfgang A. Schulz, 
  • Christian Niedworok, 
  • Günter Niegisch, 
  • Kathrin Scheckenbach, 
  • Michèle J. Hoffmann
PLOS
x

Abstract

Many long noncoding RNAs (lncRNAs) are deregulated in cancer and contribute to oncogenesis. In urothelial carcinoma (UC), several lncRNAs have been reported to be overexpressed and proposed as biomarkers. As most reports have not been confirmed independently in large tissue sets, we aimed to validate the diagnostic and prognostic value of lncRNA upregulation in independent cohorts of UC patients. Thus, expression of seven lncRNA candidates (GAS5, H19, linc-UBC1, MALAT1, ncRAN, TUG1, UCA1) was measured by RT-qPCR in cell lines and tissues and correlated to clinicopathological parameters including follow-up data (set 1: N n = 10; T n = 106). Additionally, publicly available TCGA data was investigated for differential expression in UC tissues (set 2: N n = 19; T n = 252,) and correlation to overall survival (OS). All proposed candidates tended to be upregulated in tumour tissues, with the exception of MALAT1, which was rather diminished in cancer tissues of both data sets. However, strong overexpression was generally limited to individual tumour tissues and statistically significant overexpression was only observed for UCA1, TUG1, ncRAN and linc-UBC1 in tissue set 2, but for no candidate in set 1. Altered expression of individual lncRNAs was associated with overall survival, but not consistently between both patient cohorts. Interestingly, lower expression of TUG1 in a subset of UC patients with muscle-invasive tumours was significantly correlated with worse OS in both cohorts. Further analysis revealed that tumours with low TUG1 expression are characterized by a basal-squamous-like subtype signature accounting for the association with poor outcome. In conclusion, our study demonstrates that overexpression of the candidate lncRNAs is found in many UC cases, but does not occur consistently and strongly enough to provide reliable diagnostic or prognostic value as an individual biomarker. Subtype-dependent expression patterns of lncRNAs like TUG1 could become useful to stratify patients by molecular subtype, thus aiding personalized treatments.

Introduction

Urothelial carcinoma (UC) represents the most common histological subtype of urinary bladder cancer. It is a heterogeneous disease, ranging from low-malignant tumours restricted to the urothelial tissue layer (stage pTa) to highly malignant and often lethal muscle-invasive carcinomas (≥pT2). Notably, less malignant variants of UC tend to recur quite frequently and high-grade lower stage tumours may progress to invasive cancers. For these reasons, improved diagnostic and prognostic biomarkers are desirable, especially to monitor for recurrences, to identify patients with a poor prognosis and to select appropriate treatments for individual patients [1].

Long noncoding RNAs (lncRNAs) are a diverse set of transcripts, which are defined as encompassing more than 200 nucleotides, but do not contain any substantial open reading frame. Many lncRNAs are expressed in a tissue-specific fashion and have been reported to undergo expression changes during tumour development or progression [2]. These properties render them valid candidates as diagnostic or prognostic cancer biomarkers. In addition, lncRNAs appear to be functionally involved in oncogenesis exerting tumour-promoting or tumour-suppressive functions [2,3].

Several lncRNAs have been reported to be frequently or consistently overexpressed in UC and have been proposed as individual diagnostic or prognostic biomarkers; some were moreover demonstrated to influence proliferation, survival, migration and other cancer-relevant properties of UC cell lines [4]. These lncRNAs include UCA1 [5], MALAT1 [6], H19 [7], GAS5 [8], TUG1 [9], ncRAN (SNHG16) [10] and linc-UBC1 (BLACAT) [11]. However, only few lncRNAs have been validated independently for their suitability as biomarkers in different populations or for their functional involvement in UC.

For instance, some studies have concurrently reported overexpression of HOXC11-AS, also known as HOTAIR, in many UCs, which appears to be associated with higher grade and worse prognosis. In cell culture experiments, HOTAIR likewise often confers a more aggressive phenotype [12,13]. In contrast, the lncRNA MEG3 encoded in an imprinted cluster at 14q32 is strongly down-regulated in most UCs according to two independent reports [14,15]. The requirement for replication studies in this field is further illustrated by the case of the lncRNA CDKN2B-AS, also known as ANRIL, which was proposed to be upregulated in UC in one study, but reported as essentially unchanged in another one [16,17]. In addition to differences in the study populations, this discrepancy may relate to the investigation of different ANRIL isoforms between the two studies, since a third investigation found frequent upregulation of specific splice variants only [18].

In the present study, we have therefore investigated the expression of the seven candidate lncRNAs mentioned above, for which insufficient validation data has been published so far. We used RT-qPCR on total RNA from 12 UC cell lines compared to a benign cell line and a large tissue set with complete clinical and histopathological data. Where appropriate, we initially used different primer pairs to check whether transcript variants of the respective lncRNA differ in their expression across UC cell lines. Expression data from tissues was compared with results obtained from the TANRIC database [19], which provides convenient access to the RNA-seq data and clinical information from the comprehensive investigation of UC genomics by the TCGA consortium. These data were analysed to determine for each lncRNA whether it was more strongly expressed in cancer tissues than in benign tissues, whether its expression correlated with important histopathological parameters, and especially whether any association with clinical outcome could be discerned.

Material and methods

Patients and tissues

The set of tissue samples (set 1) used for quantitative real time RT-PCR analysis (RT-qPCR) consisted of 106 tumour tissues and 10 benign tissues. The majority of the patients was diagnosed with muscle-invasive and high-grade bladder cancer (13 pTa, 13 pT1, 15 pT2, 44 pT3, 19 pT4; 39 low-grade, 67 high-grade). This tissue set was collected according to the principles expressed in the Declaration of Helsinki and with written patient informed consent as approved by the ethics committee of the medical faculty of the University Duisburg-Essen, Study Number 07–3537. A further amendment was approved by the committee allowing the extensive characterization of such tissues including the analysis of tissue-RNA by RT-PCR. Median follow up time for the complete cohort was 22.7 months (range 0.2–198).

Sample preparation and expression analyses were performed and described in accordance with the MIQE guidelines (minimum information for publication of quantitative real time PCR experiments) as detailed in the supplementary methods section (S1 File) and supplementary S1S3 Tables.

Expression data provided for lncRNAs by the TANRIC database is based on the TCGA bladder urothelial carcinoma (BLCA) dataset (set 2) consisting of 252 tumour tissue samples and 19 benign tissue samples, which were with only few exceptions obtained from muscle-invasive tumours (one pT1, 72 pT2, 118 pT3, 38 pT4, 23 without information about staging). Median follow-up time for this cohort was 15.9 months (0.49–163).

Analyses of molecular subtype gene expression signatures were performed by means of the cBioPortal tool based on the provisional TCGA bladder urothelial carcinoma set (“extended” data set: 413 tumors in total; 408 tumor samples with RNA Sequencing mRNA expression (3 pTa, 121 pT2, 195 pT3 and 59 pT4) and Genesis 1.0 software.

Cell lines

UC cell lines VM-CUB1, SW-1710, HT-1376, 5637, and BFTC-905 were obtained from the DSMZ (Braunschweig, Germany), the cell lines T-24, RT-4, RT-112, 639-V, J82, UM-UC-3 and UM-UC-6 were kindly made available by Dr. J. Fogh (New York, NY), Dr. M. A. Knowles (Leeds, UK) and Dr. B. Grossman (Houston, USA) [20]. Cell lines were regularly verified by DNA fingerprint analysis and checked for mycoplasm contamination. The TERT-immortalized normal human urothelial cell line TERT-NHUC was obtained from Dr. M. A. Knowles [21] and cultured in keratinocyte serum-free medium (Gibco, life technologies; Carlsbad, CA, USA) supplemented with 0.25 ng/ml epidermal growth factor, 12.5 μg/ml bovine pituitary extract and 1:100 ITS (Gibco), 0.35 μg/ml (-)N-epinephrine and 0.33 mg/ml hydrocortisone (Sigma Aldrich; St. Louis, MO, USA).

RNA extraction and reverse transcription

RNA was extracted using Qiazol reagent (Qiagen; Hilden, Germany) and chloroform and precipitated by isopropanol. It was further purified by means of the RNeasy Mini Kit (Qiagen) including on-column DNAse digestion. Quality and integrity of the RNA were assessed on 1.5% agarose gels. Tissue RNA was reverse transcribed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems; Foster City, CA, USA). RNA from cell lines was reverse transcribed by means of the Quantitect reverse transcription Kit (Qiagen) according to the manufacturer.

Quantitative real time PCR

RT-qPCR analyses in set 1 for the seven lncRNAs and the proliferation marker gene MKI67 as an additional control were performed on a Roche Lightycler 96 (Roche; Risch, Switzerland). Reactions were carried out in 20 μl volume using the QuantiTect SYBR Green RT-PCR Kit (Qiagen) with 10 pmol of each primer. Information on primer sequences, annealing temperatures and standard curves can be obtained from S1 Table. Reactions comprised an initial activation step for 15 minutes at 95°C and 40 cycles each with denaturation at 94°C for 20 s, annealing at an individual temperature for 20 s and DNA synthesis at 72°C for 30 s, followed by a final melting curve analysis. Concentration values are calculated based on standard curves carried out for each gene and each run using the Roche Lightcycler 96 software (Version 1.1). PCR efficiency and melting peak integrity was checked for each run. Each sample was assayed in duplicates. Specific run information on slope, efficiency, coefficient of determination (R^2) and mean of melting temperature across all samples (Tm) is given in S2 and S3 Tables. TBP and SDHA were measured as reference genes [22] and a normalization factor was calculated for each sample using their geometric mean [23]. Expression values of the assessed lncRNAs and MKI67 are given relative to this normalization factor. See supplementary methods (S1 File) for further information about RT-qPCR conditions according to MIQE guidelines.

Use of the TANRIC and cBioPortal database analysis tools

The TANRIC database provides easy access to publicly available RNA-Seq data for various tumour entities, especially for lncRNA expression. In the TANRIC database lncRNA expression was analyzed for the TCGA bladder urothelial carcinoma (BLCA) dataset and queried in the “My lnRNA”-tool by either lncRNA annotation or, for linc-UBC1 [chr1:205404014–205425214] and [chr17:74553846–74561430] for ncRAN, in hg19 genomic coordinates. LncRNA expression values were obtained as log2 RPMK (reads per kilo base per million mapped reads) from the TANRIC database as indicated in the “My lncRNA”-tool and median expression was calculated for benign and tumour samples and visualized in boxplots using R. Additional boxplots and Kaplan-Meier curves for each lncRNA were obtained from the database tool.

The cBioPortal database tools (Oncoprint, download of mRNA Expression z-Scores (RNA Seq V2 RSEM)) were employed to explore correlations between expression of lncRNA and marker genes defining molecular subtypes of UC.

Statistical methods

Mann–Whitney U test was performed for paired group comparisons. Overall survival and disease-specific survival analysis was done using univariate Cox regression analysis. In addition Kaplan–Meier log-rank test was applied for overall survival. For multivariate analysis, the Cox proportional hazards regression model was applied and variables with effect on survival in univariate analysis (p ≤ 0.05) were included in the Cox proportional hazards regression models. A p-value of at most 0.05 was considered to be statistically significant. All statistical analyses were performed with the SPSS software package (version 21; SPSS).

Results

Expression changes of candidate lncRNAs in UC and their correlation with clinicopathological parameters

Expression of the seven lncRNAs UCA1, MALAT1, H19, GAS5, TUG1, ncRAN (SNHG16) and linc-UBC1 (BLACAT) was measured by RT-qPCR in UC tissue set 1 consisting of 106 tumor tissues and 10 benign tissues obtained from radical cystectomies. The results are summarized as boxplots depicting relative expression in tumour and benign tissues for each lncRNA (Fig 1, left). For the same lncRNAs, expression data from the TCGA RNA-sequencing study of 252 UC tissues (set 2) was evaluated via the TANRIC database. Boxplots for comparison between tumour and benign tissues are likewise presented in Fig 1 (right). For sample set 1 we additionally measured the expression of the proliferation marker gene MKI67 (encoding Ki67). As expected, MKI67 was weakly expressed in most benign tissues and highly significantly overexpressed in the tumour samples (S1 Fig).

thumbnail
Fig 1. lncRNA expression data for tumour and benign tissues from tissue set 1 and 2.

Boxplot representations of lncRNA expression in set 1 (RT-qPCR, relative expression to geometric mean of reference genes SDHA and TBP) and set 2 (RNA-Seq in the TCGA bladder cancer cohort, expression as log2 RPMK, data obtained from the TANRIC database). P-values for difference between control (N) and tumour (T) samples were calculated by Mann-Whitney U-test.

https://doi.org/10.1371/journal.pone.0176287.g001

The RT-qPCR results on set 1 indicated no difference in the median expression of UCA1 between normal and tumour tissues (Fig 1 left, median 1.19 vs. 1.2). Likewise, only 2/12 investigated UC cell lines displayed UCA1 overexpression compared to control cells (S2 Fig). In the TCGA data (set 2, Fig 1 right), a slight (1.5-fold in median expression), but statistically significant (p = 0.026) increase in UCA1 expression in tumour tissues was observed (Fig 1). Although a previous report on suitability of UCA1 as a urine biomarker stated that high expression of UCA1 is associated with high grade and high stage of UC [3], the TCGA data rather indicates that increased UCA1 expression originated mainly from higher UCA1 expression levels in non-muscle invasive UC (NMIBC) compared to muscle-invasive disease (MIBC; S3 Fig right). Similarly, UCA1 expression tended to be higher in non-invasive and in low-grade tumours rather than muscle-invasive tumours of set 1 (S3 Fig left), but these differences were not statistically significant (Table 1).

In the univariate analysis patients with high UCA1 expression from tissue set 1 (above median) had a considerably better overall survival (HR 0.567; 95% CI, 0.367–0.876; p = 0.011; Table 2), which also remained significant in both multivariate analysis (p = 0.015; Table 3) and Kaplan-Meier survival curve (Fig 2, left). Concurring results, albeit with lower significance levels, were obtained in both uni-and multivariate analyses of muscle-invasive tumours only (S3 Fig, S4 and S5 Tables). In the TCGA dataset, neither high nor low expression of UCA1 was associated with clinical outcome.

thumbnail
Fig 2. Impact of lncRNA expression levels on patient overall survival in tissue set 1 and 2.

Kaplan-Meier curves are shown for lncRNA candidates displaying significant association with overall survival in set 1 (stratified by median expression; p-values for Cox regression analysis, time in months). Kaplan-Meier curves for set 2 were obtained from the TANRIC database (time in days).

https://doi.org/10.1371/journal.pone.0176287.g002

thumbnail
Table 2. Univariate analyses of the impact of lncRNA expression on patient survival.

https://doi.org/10.1371/journal.pone.0176287.t002

thumbnail
Table 3. Multivariate analyses of the impact of lncRNA expression on patient survival.

https://doi.org/10.1371/journal.pone.0176287.t003

The major isoform of MALAT1 has no intronic sequences, but many different splice variants are annotated in the ensembl database. Therefore, we first compared results for MALAT1 expression in UC cell lines obtained with primer pairs covering splice variants in the 5’-region or 3’-region. Expression measured by both primer pairs correlated very well, revealing only two UC cell lines with high expression compared to benign cells (S2 Fig). Thus, we performed RT-qPCR analyses for the tissue samples only with the primer pair for the 3’-region, which has also been applied in reports by Diederichs and colleagues who have studied MALAT1 and other lncRNAs extensively [24]. The median expression of MALAT1 tended to be lower in cancer tissues compared to normal controls in both sample sets, but the differences were not significant (Fig 1). However, comparing MALAT1 expression between NMIBC and MIBC revealed that reduced lncRNA expression in UC originated mainly from muscle-invasive tumours; this effect was even significant in our own sample set 1 (p = 0.05; Table 1; S3 Fig left). These findings do not fit with previous reports that MALAT1 expression is increased in UC and associated with distant metastasis [25,26]. In fact, our uni- and multivariate analyses revealed a poor prognosis for patients with reduced MALAT1 expression (set 1 univariate Table 2: HR 0.547, 95% CI, 0.353–0.848, p = 0.007; set 1 multivariate Table 3: HR 0.625, 95% CI, 0.40–0.976, p = 0.039; TCGA set 2 univariate S6 Table: HR 0.67, 95% CI, 0.558–0.99, p = 0.044). Kaplan-Meier analysis suggested a shorter overall survival time for patients with low MALAT1 expression (p = 0.06), especially in the lowest quartile (Fig 2 and S4 Fig).

For the lncRNA H19 no significant expression changes overall were observed in either patient cohort (Fig 1), although individual NMIBC or MIBC tumour samples showed a clear upregulation. Similarly, only two of 12 UC cell lines displayed highly increased expression compared to benign cells (S2 Fig).

Likewise, GAS5 expression was not significantly altered in both cohorts (Fig 1). In all UC cell lines, expression was reduced compared to the normal control (S2 Fig).

A long and a short isoform of ncRAN have been described [10]. Expression analysis by RT-qPCR in tumour cell lines revealed a strong correlation between both isoforms (S2 Fig). Therefore, we chose to measure the long isoform in the tissue sets. Concordantly between datasets 1 and 2, median expression of ncRAN was clearly elevated in tumour tissues by approximately 1.7-fold, but the difference was only statistically significant in set 2 (p<0.001, Fig 1). Accordingly, increased ncRAN expression was associated with overall survival of TCGA patients in univariate analysis (HR 1.53, 95% CI, 1.036–2.249, p = 0.032, S6 Table), but not in multivariate analysis. However, expression was neither significantly associated with stage, grade nor outcome in set 1 (Tables 1 and 2, S3 Fig, S4 Table).

Several tumour tissues analysed by RT-qPCR in set 1 clearly overexpressed linc-UBC1, but overall median expression was only slightly augmented and the difference between tumour and benign tissues was accordingly not significant (Fig 1). Generally, linc-UBC1 expression was very low, in four UC cell lines too, expression was below the detection limit (S2 Fig). Inconsistent with a previous report [11], no significant association was observed with lymph node metastasis or survival (Table 1, S5 Fig). Instead, linc-UBC1 was much stronger expressed (4-fold in the median) in low grade compared to high grade tumour tissues. In set 2, linc-UBC1 was significantly overexpressed (p<0.001, Fig 1), correlating with worse survival (p = 0.028, S5 Fig).

While we observed neither significant differential expression of TUG1 in UC cell lines (S2 Fig), nor in sample set 1 (Fig 1 left),nor a significant correlation between TUG1 expression and either tumour grade or stage (S3 Fig, Tables 1 and 2), TUG1 was significantly increased in the set 2 TCGA cohort (p = 0.001, Fig 1 right). Intriguingly, TUG1 was the only candidate that displayed significant results for both data sets in Kaplan-Meier survival analyses (Fig 2). Surprisingly, despite a general tendency towards overexpression, Kaplan-Meier analysis revealed that in both sample sets a subgroup of patients with low TUG1 expression faced a poor prognosis with worse overall survival (Fig 2, set 1 p = 0.012; set 2 p = 0.044). This association between low TUG1 expression and poor overall survival remained significant for uni- and multivariate analysis of set 1 (univariate Table 2: HR 0.579, 95% CI, 0.375–0.895, p = 0.014; multivariate Table 3: HR 0.616, 95% CI, 0.392–0.967, p = 0.035).

Differential expression of lncRNA candidates among molecular subtypes of UC

To further elucidate the association between low TUG1 expression and poor prognosis in a subgroup of patients, we investigated an extended set of TCGA tumour samples (provisional set, n = 408) that also contains a higher number of non-muscle-invasive cases. As illustrated in Fig 3A, TUG1 was again highly expressed in a subset of tumours (n = 166, 40.7%), but 59.3% of tumour samples (n = 242) displayed low TUG1 expression. As in set 1 (Fig 3B left), in the extended TCGA dataset more cases with low TUG1 expression were found among the muscle-invasive tumours (Fig 3B right).

thumbnail
Fig 3. Correlation between low TUG1 expression in UC and Basal-Squamous-like subtype.

(a) Waterfall plot representing TUG1 expression in tumour samples of the extended TCGA tumor tissue set (n = 408). TUG1 expression is given as z-scores. The reference for calculating z-scores of RSEM data in TCGA studies are diploid samples. Samples with high expression had a z-score above 0, specimen with low expression had negative values. (b) Boxplot representations comparing TUG1 expression in muscle-invasive and non muscle-invasive tumours of set 1 and extended set 2. (c) Heat map clustering for KRT5, KRT6, KRT14, FOXA1 and GATA3 with TUG1 expression in the TCGA cohort (expression levels are given as a colour gradient between dark blue (low expression) and red (high expression)). The tumour cluster with low TUG1 expression is shown in the left panel, that with a high expression in the centre panel and tumours with an intermediate expression pattern cluster in the right panel. (d) Expression levels of KRT5, KRT14, FOXA1 and GATA3 in tumours from the TCGA cohort with a TUG1 expression above median (TUG1 high) and below median (TUG1 low). P-values for difference between high and low TUG1 expressing group were calculated by Mann-Whitney U-test (*p≤0.05, ***p≤0.001).

https://doi.org/10.1371/journal.pone.0176287.g003

We then investigated whether TUG1 expression might relate to any molecular subtype of UC. While the discussion about the number of different molecular subtypes in UC and their according gene signatures is ongoing, a consensus has been reached on the existence of basal-squamous-like subtype (BASQ) [27], which is consistently associated with poor prognosis across all studies. According to the consensus its minimal subset of expression markers includes increased expression of cytokeratins (KRT) 5, 6 and 14 and reduced expression of FOXA1 and GATA3. Indeed, in the extended TCGA data set most tumours with low TUG1 could be assigned to the BASQ type with significant upregulated KRT5 and KRT14 expression and significantly lower expression of luminal markers FOXA1 and GATA3 (Fig 3C and 3D, Fig 4A). On the contrary, tumours with significantly increased TUG1 expression, which encompass 8% of TCGA cases, were usually positive for markers of a luminal subtype, e.g. FOXA1, GATA3, but not for basal KRTs (Fig 3C). Notably, about one third of cases with variable TUG1 expression could not be straightforwardly assigned to a molecular subtype (intermediate pattern).

thumbnail
Fig 4. Association of differential lncRNA expression with molecular subtypes of UC.

(a) TCGA expression data (z-scores) of the extended dataset (n = 408) was downloaded for genes defining molecular subtypes of MIBC according to Dadhania et al [28] and samples were assigned to the indicated subtypes by hierarchical clustering analysis using Genesis 1.0. Expression data of lncRNA candidates was downloaded accordingly, data was not available for linc-UBC1 and ncRAN. (b) Illustration of differential expression (log2 fold change) of lncRNA candidates between molecular subclasses of NMIBC based on data published by Hedegaard et al (see reference for S3 Table) [29].

https://doi.org/10.1371/journal.pone.0176287.g004

To substantiate the association of low TUG1 expression with the BASQ subtype, we included further marker genes reported to define molecular subtypes of UC. Since no definitive consensus signatures for the other molecular subtypes have been agreed on, we applied a marker signature published in a recent meta-analysis by Dadhania et al [28]. Hierarchical clustering analysis of the extended TCGA data set for a panel of 30 coding genes distinguishing between subtypes and the lncRNA candidates demonstrated again an association between low TUG1 expression and the basal-like subtype (Fig 4A). Similar expression patterns across these tumour samples became evident for UCA1 and MALAT1 (Fig 4A), suggesting that patients with low expression of these lncRNAs might also belong to the BASQ subtype.

A recent large-scale study on NMIBC by Hedegaard and colleagues reported that NMIBC can be subgrouped into at least three molecular subclasses [29]. This study compared tumour samples among each other and not differential expression compared to benign tissues, which are not usually available from these patients. Therefore, we could not simply use this data to confirm our results on differential expression between NMIBC, MIBC and benign tissues. However, some of our lncRNA biomarker candidates were also found to be differentially expressed among subclasses of NMIBC. GAS5, H19 and ncRAN displayed the highest expression levels in differentiated class 1 tumours, MALAT1 and TUG1 were rather highly expressed in basal-like class 3 tumours (Fig 4B, see S3 Table in Ref. 29). Taken together, these data suggest that lncRNA expression may not only vary between NMIBC vs MIBC, but even between their respective molecular subtypes.

Discussion

The major aim of this study was to validate the diagnostic and prognostic value of lncRNAs that were postulated in original publications and reviews as individual biomarkers for urothelial carcinoma [4,5,6,7,8,9,10,11,30]. With the exception of GAS5, all other candidates were reported to be significantly upregulated.

While urinary markers are evidently clinically useful, our study focused on tumour tissues based on the postulates that a robust and clinically applicable RNA biomarker for the identification of UC should be strongly and regularly overexpressed across large and independent cohorts of patients with UC and that its degree of overexpression should correlate with histopathological parameters indicating worse prognosis or independently predict prognosis. To this end, we analysed two distinct lncRNA expression datasets from different populations and obtained by different assays. Our own tissues sample set investigated by RT-qPCR came from an urban population in an industrial region of Germany (set 1), whereas the TCGA dataset (set 2) was assembled from US-American patients using next generation RNA-sequencing. Briefly, our analysis of public TCGA data for seven lncRNA candidates by means of the TANRIC database tool identified four of them to be significantly higher expressed in tumour tissues compared to control tissues, but for all but one the differences were slight, with medians differing by a factor less than 2-fold. In set 1, the direction of the expression changes between tumour and benign tissues concurred for all seven candidates with the TCGA data, but none of the candidate lncRNAs showed significant overexpression in tumours or reproducible strong prognostic value. Strikingly, for two candidates we obtained results from both datasets that were opposed to previous reports. MALAT1 was reported to be overexpressed in UC [6], whereas we observed reduced expression in both sets. Conversely, GAS5 was expected to be downregulated in UC [8], but was rather upregulated in both our cohorts.

There are many potential confounders that might account for the differences between the present study and previous reports, including technical issues like the cellularity of samples, quality of RNA and of the assays used as well as differences relating to patient populations. A closer analysis, as detailed below for the individual lncRNAs, suggests however that the most important factors relate to the biological heterogeneity of UC. Specifically, it is well known that non muscle-invasive tumours (NMIBC) differ substantially from muscle-invasive tumours (MIBC) with regard to tumour biology and molecular alterations. Thus, divergent results from different studies might originate from the varying abundance of tumour stages across the investigated cohorts. Most previous reports have investigated mixed cohorts of NMIBC and MIBC, whereas especially the TCGA cohort is heavily biased towards MIBC. However, expression differed significantly between NMIBC and MIBC only for UCA1 and MALAT1 and only in either set 1 or set 2. More prominently, our own results for MIBC along with a recent comprehensive transcriptome analysis of NMIBC by Hedegaard et al. [29] suggest a high degree of heterogeneity in the expression of lncRNAs between molecular subtypes of UC. Specifically, these authors reported that GAS5, H19 and ncRAN were more strongly expressed in class 1 NMIBC, representing a more differentiated subtype with a lower risk for progression, whereas MALAT1 and TUG1 were more strongly expressed in the newly defined basal-like class 3 NMIBC [29]. Notably, in accord with this idea, lncRNAs have been shown to often exhibit strong cell-type and tissue-specific patterns of expression [2]. Therefore, it could be rather difficult to identify a single lncRNA biomarker that robustly distinguishes all stages and subtypes of UC in the same fashion as certain lncRNAs like PCA3 are able to identify almost all prostate cancer cases [31]. Rather, lncRNAs like TUG1 might turn out to be most valuable as biomarkers as part of a panel to cover UC subtypes for diagnosis.

The best studied lncRNA candidate to date is arguably UCA1 which has been described as a potential urinary biomarker for UC by several groups. Notably, most of these studies have investigated urine sediments [32] rather than tissue samples, applying RT-qPCR [5] or semi-quantitative RT-PCR [33,34]. UCA1 was originally reported to be largely restricted to embryonic tissues and cancers and not to be expressed in normal bladder tissue [34], but it is clearly detectable by sensitive techniques in benign bladder tissues. In UC tumour tissues, we found a moderate increase of UCA1 expression, which in the TCGA set (set 2) resulted mainly from high expression levels in papillary tumours. Concurringly, a follow-up study on UCA1 as a urinary biomarker found UCA1 urine analysis to be particularly efficient for detection of pT1 tumours [33]. Conceivably, increased detection of UCA1 in urinary sediments might reflect only partly increased expression in cancer tissues, but instead increased shedding of urothelial cells into urine in tumour patients. Analogously, Cytokeratin 20 mRNA, a marker of differentiated urothelial cells expressed in many urothelial carcinomas, is increased in urinary sediments of UC patients [35].

As for UCA1, we observed overexpression of linc-UBC1 in many UC, but significant increases only in the TCGA dataset. This finding fits reasonably with a previous report, whereby linc-UBC1 was upregulated in 60% of 102 bladder cancer tissues by more than 1.5-fold [11]. Notably, linc-UBC1 expression was very low in many tissues. Concordant with the previous report, high linc-UBC1 expression was associated with worse survival in the TCGA cohort, but neither associated with survival nor lymph node status in the samples analysed by RT-qPCR. However, expression near the detection limit in many cancer tissues would limit assay sensitivity and thus suitability of linc-UBC1 as a biomarker.

Unexpectedly, in view of previous reports on MALAT1 overexpression in UC [2, 26], MALAT1 expression was rather diminished in UC cell lines and tumour tissues compared to normal tissues in our cohorts. This discrepancy may relate to significant differences in MALAT1 expression between NMIBC and MIBC. In our sample set 1 MALAT1 expression was diminished in invasive tumours compared to benign tissues, whereas NMIBC displayed rather upregulation. Low MALAT1 expression in MIBC, moreover, appears to be associated with the BASQ subtype, as revealed by hierarchical clustering analysis (Fig 4). The association of low MALAT1 expression with poor survival may therefore reflect its association with the more clinically aggressive basal-like subtype. This contention is partly compatible with other reports in the literature. In a study by Fan et al. [25] 62 of 95 investigated bladder cancers were NMIBC and MALAT1 was rather upregulated. In the large-scale transcriptome analysis of early-stage UC by Hedegaard et al [29] MALAT1 was likewise expressed most prominently in one NMIBC subgroup, basal-like class 3. Thus, opposite to the results for muscle-invasive UC where low MALAT1 expression was associated with basal-like characteristics, in NMIBC MALAT1 upregulation correlated with basal-like characteristics, but not necessarily with disease progression. Additional published studies on MALAT1 in UC are difficult to compare as only small sample numbers were investigated and information on histopathological characteristics is not comprehensive [26]. Two meta-analyses of 10 studies on MALAT1 in various cancer types indicated a relatively consistent association of increased MALAT1 expression with survival in various types of cancer, but some studies analysed only few cases [36, 37]. Clearly, further studies on MALAT1 should be conducted to ascertain the association of its expression with tumour subtypes. Moreover, given the assumption that lncRNAs commonly act in a context-dependent fashion, the function of MALAT1 should be investigated in different UC molecular subtypes.

H19 expression was not significantly altered between tumour and benign tissues of both sample sets and no association with tumour stage or other clinical parameters was found. Notably, H19 expression varied across a broad range. H19 is one of the first discovered long noncoding RNAs and its gene is imprinted. Altered expression could therefore be due to epigenetic as well as genetic changes of the imprinted gene cluster at 11p15.5. Previously, Luo et al. reported a strong increase in H19 expression in 24 bladder tumour specimens over adjacent normal tissue [8], but strangely, expression measured by RT-qPCR ranged across 15 orders of magnitude. More recently, Li et al. [38] reported more moderate increases in 40 tumours (mostly NMIBC) vs. 19 benign tissues and an excellent association with grading. Hedegaard et al observed significant expression changes in their large NMIBC cohort and found H19 the highest expressed lncRNA in well-differentiated class 1 tumours with a low risk for progression [29]. In an older report, H19 mRNA expression in bladder cancer was studied by in situ hybridization on paraffin sections [39]. Thus, again, varying tumour population and detection techniques might account for the differences among the findings.

For GAS5, we observed neither significant differential expression between tumour and benign tissues nor associations of expression with histopathological parameters or prognosis. However, expression was increased in many NMIBC tumours compared with benign tissues. Accordingly, Hedegaard et al [29] observed significantly increased GAS5 expression in class 1 NMIBC tumours. Previously, GAS5 had been reported to be downregulated based on a cohort of 12 NMIBC and 16 MIBC [8], but the differences between these subtypes were not detailed. Based on these results and experiments in one bladder cancer cell line, a tumour suppressive function was postulated [40], which would not fit well with the newer results in tissues.

For ncRAN, deregulation has been reported in neuroblastoma [41], colorectal cancer [42] and in bladder cancer, where a limited number of cancer tissues were investigated by semi-quantitative PCR [10]. Experimental overexpression of ncRAN enhanced cell proliferation, migration and invasion in the well-differentiated UC cell line RT-4. Accordingly, ncRAN tended to be overexpressed, albeit moderately, in the cancer tissues in our study, but the association of lower expression with worse prognosis in one sample set argues against an important function of ncRAN upregulation in tumour progression. In addition, Hedegaard et al reported increased ncRAN expression in differentiated and good prognosis class 1 tumours likewise arguing against a decisive function in tumour progression [29].

Upregulation of TUG1 and an association with higher tumour stages in UC was reported in previous studies. In the first study, TUG1 was elevated by 1.74-fold in the mean in 44 cancer tissues compared to adjacent normal tissues [9]. Another study detected TUG1 transcripts in 54 bladder cancer tissues by RT-qPCR; expression in normal tissues was not investigated [43]. A third study observed significant overexpression in 47 muscle-invasive tumours compared to paired benign tissues, an association with metastasis and shorter overall survival [44]. We found modest, but not consistently statistically significant increases of TUG1 in cancer tissues across both tissue sets. Strikingly, TUG1 was the only investigated lncRNA candidate that was significantly associated with survival in Kaplan-Meier analyses across both cohorts, but unexpectedly, low rather than high TUG1 expression was significantly correlated with poor overall survival, which remained significant in multivariate analysis. Interestingly, Zhang et al. likewise found diminished TUG1 expression in non-small cell lung carcinoma (NSCLC) to be associated with higher tumour stage and size as well as significant poorer overall survival in uni- and multivariate analyses [45]. In their study, TUG1 knockdown in NSCLC cell lines promoted cellular proliferation, whereas in T24 bladder cancer cells TUG1 overexpression increased cell invasion [9, 44].

A new finding in our study was the association of low TUG1 expression with a particular molecular subtype of muscle-invasive UC, the basal subtype with squamous differentiation patterns (BASQ). While the definition of molecular subtypes in UC is still somewhat controversial, a consensus has been reached on the existence of this particular subtype among MIBC, its association with poor prognosis and a minimal set of defining markers [27]. Among five markers for BASQ, low TUG1 expression significantly correlated with four, i.e. high expression of Cytokeratins 5 and 14 and low expression of FOXA1 and GATA3. The correlation of low TUG1 expression with poor prognosis might therefore reflect its relation to the BASQ molecular subtype with its known worse clinical behaviour. As discussed above, low expression of MALAT1 and UCA1 might similarly be related to the BASQ subtype.

Taken together, these observations hint at cancer subtype-specific expression and context-dependent functions of TUG1 which should be considered in future studies. Of note, Zhang et al. proposed TUG1 as a direct transcriptional target of p53 and reported that TUG1 was not induced in cell lines depleted of p53 or harbouring common p53 mutations [45]. P53 mutations are common in urothelial carcinomas, especially in the BASQ subtype, and might explain low TUG1 expression in some tumour tissues.

In conclusion, we observed that few of the previously reported changes in lncRNA expression or their association to histopathological parameters and patient prognosis could be robustly confirmed across two further large independent patient cohorts. In particular, overexpression of lncRNAs was often restricted to a subset of the cases or was moderate in extent. In addition to technical issues, varying proportions of tumour stages and molecular subtypes are likely to account for the differences between studies. Notably, both our sample set and the TCGA cohort consist predominantly of muscle-invasive tumours and our data are not representative for NMIBC. Moreover, the case of UCA1 demonstrates that tissue and urinary biomarkers might differ. Taken together, our study clearly demonstrates that the identification of reliable lncRNA biomarkers for urothelial carcinoma demands validation studies in independent patient cohorts with large sample numbers which take tumour stages and molecular subtypes into account. In addition, lncRNAs might deserve investigation as components of biomarker panels for UC subtypes.

Supporting information

S1 Fig. Quality controls for tissues of set 1.

(a) To further characterize the quality/ purity of tissue samples from set 1 expression of the proliferation marker Ki67 (gene: MKI67) was determined in all samples by RT-qPCR. RNA expression illustrated as boxplot representation (relative expression to geometric mean of reference genes SDHA and TBP). P-values for difference between benign (Normal) and tumour samples (Tumour) were calculated by Mann-Whitney U-test. (b) Two representative H&E stainings of tumour tissue sections used to evaluate histology, quality and cellularity of the sample are given.

https://doi.org/10.1371/journal.pone.0176287.s001

(TIF)

S2 Fig. lncRNA expression in UC cell lines.

Relative expression of the seven lncRNA candidates across 12 UC cell lines and a benign control cell line (TERT-NHUC) is illustrated in bar graphs. Cell lines were classified into non-basal like and basal-like according to Earl et al [20]. Two different primer assays detecting various numbers of transcript variants had been evaluated for MALAT1 and ncRAN across the cell lines.

https://doi.org/10.1371/journal.pone.0176287.s002

(TIF)

S3 Fig. lncRNA expression data of set 1 and 2 analysed separately for NMIBC and MIBC.

Boxplot representations of lncRNA expression in set 1 (left, RT-qPCR, relative expression to geometric mean of reference genes SDHA and TBP) and set 2 (right, RNA-Seq in the TCGA bladder cancer cohort, expression as log2 RPMK, data obtained from the TANRIC database). P-values for difference between control (N) samples, non-muscle invasive stages (pTa/pT1, NMIBC) and muscle-invasive tumours (T2-T4, MIBC) were calculated by Mann-Whitney U-test.

https://doi.org/10.1371/journal.pone.0176287.s003

(TIF)

S4 Fig. Kaplan Meier Analysis for overall survival of patients of set 1 divided in subgroups based on lncRNA expression.

Kaplan Meier analysis for quartiles of lncRNA expression in set 1 as indicated and overall survival (stratified by three cutpoints, time in months).

https://doi.org/10.1371/journal.pone.0176287.s004

(TIF)

S5 Fig. Impact of lncRNA expression levels on patient overall survival in tissue set 1 and 2.

Kaplan-Meier curves are shown for each lncRNA stratified by median expression. Kaplan-Meier curves for set 2 were obtained from the TANRIC-database. Only data for lncRNAs which did not show a significant difference for set 1 are shown (p-values for Cox regression analysis).

https://doi.org/10.1371/journal.pone.0176287.s005

(TIF)

S6 Fig. Analysis of RT minus controls.

RT minus controls without reverse transcriptase were included in the analysis of assays using primers that were not exon spanning to exclude that results were affected by contamination with genomic DNA. Raw quantified values are given (related to the internal standard curve for the respective gene). Without RT fluorescence signals remained below the detection limit (n.d.).

https://doi.org/10.1371/journal.pone.0176287.s006

(TIF)

S1 Table. Primer sequences and PCR conditions.

Primer sequences used in RT-qPCR analyses in 5’->3’ orientation and further information regarding PCR assay design, reaction conditions and the reference bladder or prostate cancer cell line used for standard curve (Ref. std. curve) are given.

https://doi.org/10.1371/journal.pone.0176287.s007

(PDF)

S2 Table. Information on RT-qPCR run performance.

For each lncRNA three RT-qPCR runs were conducted with normal and tumor samples distributed equally. Information on slope of the standard curve, resulting efficiency, R^2, melting temperature Tm, the Y-Intercept and Cq values of negative controls (Cq neg. “-”equivalent to undetectable) are given for each run. *According to the melting curve analysis the Cq result for the negative control did not result from a contamination by the specific amplicon.

https://doi.org/10.1371/journal.pone.0176287.s008

(PDF)

S3 Table. Information on RT-qPCR run performance (2nd).

For the second reverse transcription of tissue sample RNA RT-qPCR runs were conducted with normal and tumor samples distributed equally. Information on slope of the standard curve, resulting efficiency, R^2, melting temperature Tm, the Y-Intercept and Cq values of negative controls (Cq neg. “-”equivalent to undetectable) are given for each run. *According to the melting curve analysis the Cq result for the negative control did not result from a contamination by the specific amplicon.

https://doi.org/10.1371/journal.pone.0176287.s009

(PDF)

S4 Table. Univariate analyses of the impact of lncRNA expression on survival of patients with MIBC.

Hazard Ratios (HR) with a 95% Confidence Interval (CI) and p-values (P) were calculated by Cox regression analyses on overall and disease-specific survival for lncRNA expression levels for all patients with T2-T4 tumors in set 1. Patients were divided into a low- and a high-expression group for each lncRNA by median expression. The cut-off is based on the whole Ta-T4 cohort. Bold printed p-values were significant (≤0.05).

https://doi.org/10.1371/journal.pone.0176287.s010

(PDF)

S5 Table. Multivariate analyses of the impact of lncRNA expression on survival of patients with MIBC.

UCA1, TUG1, ncRAN and MALAT1 expression had a statistically significant impact on patient overall or disease-specific survival and were further analysed in the multivariate analyses including tumor stage, grade and lymph node metastasis as parameters. Hazard Ratios (HR) with a 95% Confidence Interval (CI) and p-values (p) are given for overall and disease-specific survival. Bold printed p-values were significant (≤0.05).

https://doi.org/10.1371/journal.pone.0176287.s011

(PDF)

S6 Table. Univariate analyses of the impact of lncRNA expression on patient survival (TCGA).

Hazard Ratios (HR) with a 95% Confidence Interval (CI) and p-values (P) were calculated by Cox regression analyses on overall and disease-specific survival for lncRNA expression levels in set 1. Patients were divided into a low- and a high-expression group for each lncRNA by median expression. Bold printed p-values were significant (≤0.05).

https://doi.org/10.1371/journal.pone.0176287.s012

(PDF)

Author Contributions

  1. Conceptualization: WAS KS MJH.
  2. Data curation: JD MJH TS.
  3. Formal analysis: JD MJH TS CN.
  4. Funding acquisition: WAS MJH KS.
  5. Investigation: JD MJH TS CN.
  6. Methodology: JD MJH TS WAS GN.
  7. Project administration: WAS MJH KS.
  8. Resources: TS CN GN KS.
  9. Software: JD MJH TS CN.
  10. Supervision: WAS KS MJH GN CN.
  11. Validation: JD MJH.
  12. Visualization: JD MJH TS.
  13. Writing – original draft: WAS MJH JD TS.
  14. Writing – review & editing: MJH JD WAS TS CN GN KS.

References

  1. 1. Buttigliero C, Tucci M, Vignani F, Scagliotti GV, Di Maio M. Molecular biomarkers to predict response to neoadjuvant chemotherapy for bladder cancer. Cancer Treat Rev. 2017; 54:1–9. pmid:28135623
  2. 2. Fatica A, Bozzoni I. Long non-coding RNAs: New players in cell differentiation and development. Nat Rev Genet. 2013; 15: 7–21. pmid:24296535
  3. 3. Haemmerle M, Gutschner T. Long non-coding RNAs in cancer and development: where do we go from here? Int J Mol Sci. 2015; 16:1395–1405. pmid:25580533
  4. 4. Martens-Uzunova ES, Böttcher R, Croce CM, Jenster G, Visakorpi T, Calin GA. Long noncoding RNA in prostate, bladder, and kidney cancer. Eur Urol. 2014; 65: 1140–1151. pmid:24373479
  5. 5. Srivastava AK, Singh PK, Rath SK, Dalela D, Goel MM, Bhatt MLB. Appraisal of diagnostic ability of UCA1 as a biomarker of carcinoma of the urinary bladder. Tumour Biol. 2014; 35: 11435–11442. pmid:25123267
  6. 6. Han Y, Liu Y, Nie L, Gui Y, Cai Z. Inducing cell proliferation inhibition, apoptosis, and motility reduction by silencing long noncoding ribonucleic acid metastasis-associated lung adenocarcinoma transcript 1 in urothelial carcinoma of the bladder. Urology. 2013; 81: 209.e1.
  7. 7. Luo M, Li Z, Wang W, Zeng Y, Liu Z, Qiu J. Upregulated H19 contributes to bladder cancer cell proliferation by regulating ID2 expression. FEBS J. 2013; 280: 1709–1716. pmid:23399020
  8. 8. Liu Z, Wang W, Jiang J, Bao E, Xu D, Zeng Y et al. Downregulation of GAS5 promotes bladder cancer cell proliferation, partly by regulating CDK6. PLoS ONE. 2013; 8: e73991. pmid:24069260
  9. 9. Han Y, Liu Y, Gui Y, Cai Z. Long intergenic non-coding RNA TUG1 is overexpressed in urothelial carcinoma of the bladder. J Surg Oncol. 2013; 107: 555–559. pmid:22961206
  10. 10. Zhu Y, Yu M, Li Z, Kong C, Bi J, Li J et al. ncRAN, a newly identified long noncoding RNA, enhances human bladder tumor growth, invasion, and survival. Urology. 2011; 77: 510.e1–5.
  11. 11. He W, Cai Q, Sun F, Zhong G, Wang P, Liu H et al. linc-UBC1 physically associates with polycomb repressive complex 2 (PRC2) and acts as a negative prognostic factor for lymph node metastasis and survival in bladder cancer. Biochim Biophys Acta. 2013; 1832: 1528–1537. pmid:23688781
  12. 12. Heubach J, Monsior J, Deenen R, Niegisch G, Szarvas T, Niedworok C et al. The long noncoding RNA HOTAIR has tissue and cell type-dependent effects on HOX gene expression and phenotype of urothelial cancer cells. Mol Cancer. 2015; 14: 361.
  13. 13. Yan T, Lu S, Huang Y, Que G, Chen J, Chen Y et al. Upregulation of the long noncoding RNA HOTAIR predicts recurrence in stage Ta/T1 bladder cancer. Tumor Biol. 2014; 35: 10249–10257.
  14. 14. Greife A, Knievel J, Ribarska T, Niegisch G, Schulz WA. Concomitant downregulation of the imprinted genes DLK1 and MEG3 at 14q32.2 by epigenetic mechanisms in urothelial carcinoma. Clin Epigenetics. 2014; 6: 29. pmid:25741387
  15. 15. Ying L, Huang Y, Chen H, Wang Y, Xia L, Chen Y et al. Downregulated MEG3 activates autophagy and increases cell proliferation in bladder cancer. Mol BioSyst. 2013; 9: 407. pmid:23295831
  16. 16. Zhu H, Li X, Song Y, Zhang P, Xiao Y, Xing Y. Long non-coding RNA ANRIL is up-regulated in bladder cancer and regulates bladder cancer cell proliferation and apoptosis through the intrinsic pathway. Biochem Biophys Res Commun. 2015; 467: 223–228. pmid:26449463
  17. 17. Martínez-Fernández M, Feber A, Dueñas M, Segovia C, Rubio C, Fernandez M et al. Analysis of the Polycomb-related lncRNAs HOTAIR and ANRIL in bladder cancer. Clin Epigenet. 2015; 7: 109.
  18. 18. Hoffmann M, Dehn J, Droop J, Niegisch G, Niedworok C, Szarvas T et al. Truncated isoforms of lncRNA ANRIL are overexpressed in bladder cancer, but do not contribute to repression of INK4 tumor suppressors. Non-coding RNA. 2015; 1: 266–284.
  19. 19. Li J, Han L, Roebuck P, Diao L, Liu L, Yuan Y et al. TANRIC: an interactive open platform to explore the function of lncRNAs in cancer. Cancer Res. 2015; 75: 3728–3737. pmid:26208906
  20. 20. Earl J, Rico D, Carrillo-de-Santa-Pau E, Rodriguez-Santiago B, Mendez-Pertuz M, Auer H et al. The UBC-40 Urothelial Bladder Cancer cell line index: a genomic resource for functional studies. BMC Genomics. 2015; 16: 403. pmid:25997541
  21. 21. Chapman EJ, Williams SV, Platt FM, Hurst CD, Chambers P, Roberts P et al. Integrated genomic and transcriptional analysis of the in vitro evolution of telomerase-immortalized urothelial cells (TERT-NHUC). Genes Chromosomes Cancer. 2009; 48: 694–710. pmid:19405089
  22. 22. Ohl F, Jung M, Radonic A, Sachs M, Loening SA, Jung K. Identification and validation of suitable endogenous reference genes for gene expression studies of human bladder cancer. J Urol. 2006; 175:1915–1920. pmid:16600798
  23. 23. Vandesompele J, Preter K de, Pattyn F, Poppe B, van Roy N, Paepe A de et al. Accurate normalization of real-time quantitative RT-PCR data by geometric averaging of multiple internal control genes. Genome Biol. 2002; 3: RESEARCH0034. pmid:12184808
  24. 24. Eissmann M, Gutschner T, Hammerle M, Gunther S, Caudron-Herger M, Gross M et al. Loss of the abundant nuclear non-coding RNA MALAT1 is compatible with life and development. RNA Biol. 2012; 9: 1076–1087. pmid:22858678
  25. 25. Fan Y, Shen B, Tan M, Mu X, Qin Y, Zhang F et al. TGF-β-induced upregulation of malat1 promotes bladder cancer metastasis by associating with suz12. Clin Cancer Res. 2014; 20: 1531–1541. pmid:24449823
  26. 26. Ying L, Chen Q, Wang Y, Zhou Z, Huang Y, Qiu F. Upregulated MALAT-1 contributes to bladder cancer cell migration by inducing epithelial-to-mesenchymal transition. Mol BioSyst. 2012; 8: 2289–2294. pmid:22722759
  27. 27. Lerner SP, McConkey DJ, Hoadley KA, Chan KS, Kim WY, Radvanyi F et al. Bladder cancer molecular taxonomy: summary from a consensus meeting. Bl Cancer. 2016; 2(1):37–47. pmid:27376123
  28. 28. Dadhania V, Zhang M, Zhang L, Bondaruk J, Majewski T, Siefker-Radtke A et al. Meta-Analysis of the luminal and basal subtypes of bladder cancer and the identification of signature immunohistochemical markers for clinical use. EBioMedicine. 2016; 12:105–117. pmid:27612592
  29. 29. Hedegaard J, Lamy P, Nordentoft I, Algaba F, Hoyer S, Ulhoi BP et al. Comprehensive transcriptional analysis of early-stage urothelial carcinoma. Cancer Cell. 2016; 30:1–16
  30. 30. Chen J, Miao Z, Xue B, Shan Y, Weng G, Shen B. Long non-coding RNAs in urologic malignancies: functional roles and clinical translation. J Cancer. 2016; 7:1842–1855. pmid:27698924
  31. 31. Ronnau CGH, Verhaegh GW, Luna-Velez MV, Schalken JA. Noncoding RNAs as novel biomarkers in prostate cancer. Biomed Res Int. 2014; 2014:591703. pmid:25243154
  32. 32. Eissa S, Matboli M, Essawy NOE, Kotb YM. Integrative functional genetic-epigenetic approach for selecting genes as urine biomarkers for bladder cancer diagnosis. Tumour Biol. 2015; 36: 9545–9552. pmid:26138586
  33. 33. Milowich D, Le Mercier M, Neve N de, Sandras F, Roumeguere T, Decaestecker C et al. Diagnostic value of the UCA1 test for bladder cancer detection: A clinical study. SpringerPlus. 2015; 4: 612.
  34. 34. Wang X, Zhang Z, Wang H, Cai J, Xu Q, Li M et al. Rapid identification of UCA1 as a very sensitive and specific unique marker for human bladder carcinoma. Clin Cancer Res. 2006; 12: 4851–4858. pmid:16914571
  35. 35. Mengual L, Burset M, Ribal MJ, Ars E, Marin-Aguilera M, Fernandez M et al. Gene expression signature in urine for diagnosing and assessing aggressiveness of bladder urothelial carcinoma. Clin Cancer Res. 2010; 16: 2624–2633. pmid:20406841
  36. 36. Yan J, Zhou X, Dang Y, Yin C, Zhang G. Prognostic role of the long non-coding RNA metastasis-associated lung adenocarcinoma transcript 1 in various cancers: A meta-analysis. Mol Clin Oncol. 2016; 4: 100–106. pmid:26870366
  37. 37. Wu Y, Lu W, Xu J, Shi Y, Zhang H, Xia D. Prognostic value of long non-coding RNA MALAT1 in cancer patients. Tumour Biol. 2015: 1–7.
  38. 38. Li S, Yu Z, Chen S, Li F, Lei C, Chen X et al. The YAP1 oncogene contributes to bladder cancer cell proliferation and migration by regulating the H19 long noncoding RNA. Urol Oncol. 2015; 33: 1078–1439.
  39. 39. Ariel I, Lustig O, Schneider T, Pizov G, Sappir M, De-Groot N et al. The imprinted H19 gene as a tumor marker in bladder carcinoma. Urology. 1995; 45: 335–338. pmid:7855987
  40. 40. Cao Q, Wang N, Qi J, Gu Z, Shen H. Long non-coding RNA-GAS5 acts as a tumor suppressor in bladder transitional cell carcinoma via regulation of chemokine (C-C motif) ligand 1 expression. Mol Med Report. 2015; 13: 27–34.
  41. 41. Yu M, Ohira M, Li Y, Niizuma H, Oo ML, Zhu Y et al. High expression of ncRAN, a novel non-coding RNA mapped to chromosome 17q25.1, is associated with poor prognosis in neuroblastoma. Int J Oncol. 2009; 34: 931–938. pmid:19287950
  42. 42. Qi P, Xu M, Ni S, Shen X, Wei P, Huang D et al. Down-regulation of ncRAN, a long non-coding RNA, contributes to colorectal cancer cell migration and invasion and predicts poor overall survival for colorectal cancer patients. Mol Carcinog. 2015; 54: 742–750. pmid:24519959
  43. 43. Tan J, Qiu K, Li M, Liang Y. Double-negative feedback loop between long non-coding RNA TUG1 and miR-145 promotes epithelial to mesenchymal transition and radioresistance in human bladder cancer cells. FEBS Letters. 2015; 589: 3175–3181. pmid:26318860
  44. 44. Iliev R, Kleinova R, Juracek J, Dolezel J, Ozanova Z, Fedorko M et al. Overexpression of long non-coding RNA TUG1 predicts poor prognosis and promotes cancer cell proliferation and migration in high-grade muscle-invasive bladder cancer. Tumour Biol. 2016 Jul 27;
  45. 45. Zhang E, Yin D, Sun M, Kong R, Liu X, You L et al. P53-regulated long non-coding RNA TUG1 affects cell proliferation in human non-small cell lung cancer, partly through epigenetically regulating HOXB7 expression. Cell Death Dis. 2014; 5: e1243. pmid:24853421