MicroRNA Expression Signatures of Bladder Cancer Revealed by Deep Sequencing

1Bladder cancer is one of the most prevalent malignancies in the world. About 357,000 bladder cancer cases were newly diagnosed and 145,000 cancer-related deaths were estimated in 2002. Urothelial carcinoma of the bladder, the most common histopathologic type of bladder cancer, has a variety of genetic and phenotypic characteristics. Many factors, such as chromosomal anomalies, genetic polymorphisms, genetic and epigenetic alterations, contribute to tumorigenesis and progression of urothelial carcinoma of the bladder.

2MicroRNAs (miRNAs) are endogenous, noncoding RNA molecules of about 22 nucleotides in length that regulate gene expression. They join the RNA-induced silencing complex to regulate their targeted messenger RNA (mRNA) by repressing mRNA translation and/or directing mRNA cleavage. miRNAs play important roles in normal development, cell growth, differentiation, and apoptosis in mammals.

3More than half miRNA genes are located in cancer-associated genomic regions or in fragile sites. Aberrantly expressed miRNAs have been shown to be associated with many types of cancers. Both losses and gains of miRNA function contribute to cancer development. miRNAs act as oncogenes or tumor suppressors. Most importantly, different cancer types, stages or differentiation states have unique miRNA expression profiles, suggesting that miRNAs can function as novel biomarkers for cancer diagnosis,.

4Several previous researches used miRNA microarrays with limited and varied probes to profile the miRNA expression in bladder cancer and their results did not always indicate consistent results –. To better understand the role of miRNAs in bladder cancer development and progression, comprehensive analysis of the expression and abundance of miRNAs in this cancer is required. With the merit of the high-throughput deep sequencing technology, genome-wide cancer miRNAs can be quantitatively and accurately determined. Here, we present the genome-wide miRNA profiles in nine pairs of snap-frozen bladder urothelial carcinoma and matched histologically normal urothelium by deep sequencing. We found that a collection of miRNAs were aberrantly expressed in bladder urothelial carcinoma compared to matched histologically normal urothelium, several of which were evaluated by Real-Time qPCR in a total of fifty-one bladder urothelial carcinoma patients.

5Known miRNA expression files between bladder urothelial carcinoma and matched histologically normal urothelium from each patient were compared to find out the differentially expressed miRNAs. The expression of miRNAs in paired samples were shown by calculating log2Ratio. The procedures are shown as below: (1) Normalize the expression of miRNAs in two samples (tumor versus normal) to get the expression of transcript per million (TPM). Normalized expression = Actual miRNA count/Total count of clean reads*1000000. (2) Calculate fold-change and p value from the normalized expression. Then Calculate log2Ratio. Fold-change = log2Ratio (tumor/normal). We determined 656 differentially expressed known human miRNAs and miRNA antisense sequences (miRNA*s) in miRBase14.0 in nine bladder urothelial carcinoma patients (Table S1).

6We identified a great number of miRNAs and miRNA*s that were signicantly upregulated or downregulated in these patients and could discriminate bladder urothelial carcinoma from matched normal urothelium. hsa-miR-96 (log2Ratio = 4.664328) was the most significantly upregulated miRNA and hsa-miR-490-5p (log2Ratio = −5.79794) was the most significantly downregulated one (Table 1). Selected differentially expressed miRNAs were validated by Real-Time qPCR. The Real-Time qPCR findings correlated well with the sequencing analysis. The comparison between Real-Time qPCR findings and deep sequencing results is shown in Figure 1. The counts of upregulated and downregulated miRNAs varied in different patients. Upregulated miRNAs were more common than downregulated ones (Figure 2). Additionally, we identified a remarkable divergence of expression levels between miRNA and paired miRNA*. The expression levels of miRNAs were usually higher than that of paired miRNA*s (Figure 3).

7Besides the known miRNAs and miRNA*s, 92 novel miRNA sequence candidates were detected in our study (Table S2). Most of them were expressed at very low levels and only in certain samples. Their expression patterns and possible roles need further investigation.

8We found that a collection of deregulated miRNA clusters were expressed. The hsa-miR-183, hsa-miR-200b∼429, hsa-miR-200c∼141 and hsa-miR-17∼92 clusters were significantly upregulated. The hsa-miR-143∼145 cluster was significantly downregulated.

9hsa-miR-182, hsa-miR-183, hsa-miR-200a, hsa-miR-143 and hsa-miR-195 were evaluated by Real-Time qPCR in a total of fifty-one bladder urothelial carcinoma patients. hsa-miR-182, hsa-miR-183 and hsa-miR-200a were overexpressed hsa-miR-143 and hsa-miR-195 were underexpressed in bladder urothelial carcinoma compared to matched histologically normal urothelium (p<0.001 for each miRNA) (Table S3).

10The development of high throughput deep sequencing technology provides the possibility of a near complete view of miRNA profiles. Deep sequencing technology has the potential to identify novel tissue specific miRNAs. It determines the absolute abundance of miRNAs and can discover novel miRNAs that have been missed by common cloning and sequencing methods. Deep sequencing technology is superior to microarrays which determine limited known miRNAs and usually do not contain the full list of known miRNA antisense sequences. Up to now deep sequencing technology has been the gold standard for the comprehensive analysis of miRNAs.

11Researches have revealed that miRNAs can be used as biomarkers for different types of cancers,. Some miRNAs as biomarkers are able to trace the tissue of origin of cancers of unknown primary origin. Specific miRNA signatures are superior to mRNA signatures in predicting the prognosis of lung cancer.

12In this work, we used deep sequencing technology to determine the comprehensive miRNA expression profiles in nine pairs of snap-frozen bladder urothelial carcinoma and matched histologically normal urothelium. Many miRNAs were significantly upregulated or downregulated in bladder urothelial carcinoma compared to matched histologically normal urothelium in these patients, suggesting that these aberrantly expressed miRNAs might play roles in these cancers. The most significantly upregulated miRNA hsa-miR-96 has been reported to be oncogenic, but the role of the most significantly downregulated miRNA hsa-miR-490-5p is not clear. A lot of miRNA*s were significantly deregulated in bladder urothelial carcinoma compared to matched histologically normal urothelium in this study. Some miRNA*s can join the RNA-induced silencing complex and have inhibitory function.

13Many deregulated miRNA clusters were expressed according to our deep sequencing analysis. We found that the hsa-miR-183 cluster was overexpressed in bladder cancer. This cluster consists of hsa-miR-96, hsa-miR-182 and hsa-miR-183 and is located on chromosome 7. These three miRNAs are upregulated in prostate carcinoma. The coexpression pattern of the miRNAs of this cluster in cancers suggests they might play roles together. The hsa-miR-200 family of miRNAs (hsa-miR-200a/b/c, hsa-miR-141 and hsa-miR-429) were overexpressed in bladder cancer. The hsa-miR-200b∼429 cluster is located on chromosome 1 and the hsa-miR-200c∼141 cluster is located on chromosome 12. The coexpression of these clusters suggests that they might be controlled by common factors and play roles together. hsa-miR-200b, hsa-miR-200a and hsa-miR-429 miRNAs are encoded by a single polycistronic transcript and negatively regulated by ZEB1 and SIP1. TGFß 1 can downregulate the hsa-miR-200 family leading to the upregulation of ZEB1 and ZEB2. The hsa-miR-200 family are also overexpressed in ovarian and cervical cancers –, suggesting this miRNA family are oncogenic in seveval cancers. The hsa-miR-17-92 cluster is located on chromosome 13 and acts as oncogenes. E2F1 and E2F3 can directly activate the transcription of these miRNAs. The hsa-miR-143∼145 cluster is located on chromosome 5 and downregulated in many cancers, including bladder cancers and their cell lines,. Our findings provided more evidence to support that the hsa-miR-143∼145 cluster is tumor-suppressive in bladder cancer.

14In this study, Real-Time qPCR was performed to evaluate the expression patterns of hsa-miR-182, hsa-miR-183, hsa-miR-200a, hsa-miR-143 and hsa-miR-195 in a total of fifty-one bladder urothelial carcinoma patients. hsa-miR-182, hsa-miR-183 and hsa-miR-200a were overexpressed hsa-miR-143 and hsa-miR-195 were underexpressed in bladder urothelial carcinoma compared to matched histologically normal urothelium. These findings supported our deep sequencing analysis.

15We compared our results to published data to seacrch for independent external validations. hsa-miR-182, hsa-miR-183 and hsa-miR-224 are upregulated and hsa-miR-1, hsa-miR-101, hsa-miR-143, hsa-miR-145, hsa-miR-127 and hsa-miR-29c are downregulated in bladder urothelial carcinoma compared to matched histologically normal urothelium. Our deep sequencing results were largely consistent with these findings. The upregulation of hsa-miR-182 and hsa-miR-183 and the downregulation of hsa-miR-143 were found in bladder urothelial carcinoma compared to matched histologically normal urothelium in our Real-Time qPCR evaluation. The profiling of miRNAs in 106 bladder cancers and 11 normal samples using microarrays has revealed that a set of miRNAs are upregulated or downregulated in bladder cancers. Of these deregulated miRNAs, hsa-miR-21, hsa-miR-20a, hsa-miR-184, hsa-miR-26a, hsa-miR-125b, hsa-miR-29a, hsa-miR-29c and so on share the similar expression patterns with our results. hsa-miR-133a, hsa-miR-133b and hsa-miR-195 are downregulated in bladder cancers. These miRNAs showed downregulation in our sequencing analysis and hsa-miR-195 downregulation was comfirmed by Real-Time qPCR in this study.

16We used matched adjacent histologically normal urothelium as control in our study. It has been reported that select samples of histologically normal urothelium from bladder cancer patients have genetic alterations. Some chromosomal aberrations shared by both a tumor and a histologically normal looking tissue adjacent to the tumor can cause similar changes of miRNA expression. To compensate for this weakness, we compared our results to several published reports using urothelium from normals as control,,. A large collection of findings were comparable between theirs and ours. Our findings were also consistent with those papers using matched adjacent histologically normal urothelium as control concerning a lot of miRNAs,. Overlapping findings between published data and our results are shown in Table 2.

17To the best of our knowledge, this is the first genome-wide profiling of miRNAs in human bladder cancer by deep sequencing. We found that a collection of deregulated miRNAs were aberrantly expressed in bladder urothelial carcinoma compared to matched histologically normal urothelium, suggesting that they might play roles as oncogenes or tumor suppressors in the development and/or progression of this cancer. Our data provide novel insights into cancer biology. More work will be needed to determine the potential roles of miRNAs as diagnostic biomarkers and candidate therapeutic targets for bladder cancer.

18Written informed consent was obtained from all patients and the study was approved by the Institutional Review Board of Peking University Shenzhen Hospital. Fifty-one patients with bladder urothelial carcinoma who received partial or radical cystectomy were included in the study. Of these patients, nine were used for initial deep sequencing analysis of miRNAs and forty-two were used for an extra evaluation. Bladder urothelial carcinoma was diagnosed histopathologically. Bladder urothelial carcinoma and matched histologically normal urothelium from each subject were snap-frozen in liquid nitrogen immmediately after resection. Detailed information of nine bladder urothelial carcinoma patients in deep sequencing set is summarized in Table S4.

19When the proportion of cancer cells in a tissue section was greater than 80%, the frozen block was subjected to RNA extraction. Total RNA was extracted from fifty-one pairs of snap-frozen bladder urothelial carcinoma and matched histologically normal urothelium using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's protocol. The RNA integrity was evaluated by Agilent 2100 BioAnalyzer (Agilent Technologies, Palo Alto, CA, USA).

20Eighteen small RNA libraries prepared from nine pairs of snap-frozen bladder urothelial carcinoma and matched histologically normal urothelium were constructed, amplified and sequenced. Total RNA was used for miRNA sequencing. After 5′adapter and 3′adapter were ligated to small RNAs, Reverse transcription was performed. Then PCR was performed and PCR products were purified. Lastly, miRNA libraries were constructed and sequenced by the Illumina Cluster Station and Genome Analyze (Illumina Inc, CA, USA) at Beijing Genomics Institute at Shenzhen according to the manufacturer's protocol.

21Low quality reads were removed and adapter sequences were accurately clipped with the aid of a dynamic programming algorithm before further analysis. After elimination of duplicate reads, the remaining reads of at least 18 nt were mapped to a human reference genome (hg19) using SOAP V2.0. To identify sequence tags originating from coding exons, repeats, rRNA, tRNA, snRNA, and snoRNA, UCSC RefGene, RepeatMasker, NCBI Refseq data and the ncRNA annotations compiled from the NCBI Genbank data (http://www.ncbi.nih.gov) were used. To identify novel miRNA genes, all hairpin-like RNA structures encompassing small RNA tags were identified using MIREAP (http://sourceforge.net/projects/mireap).

22Three overexpressed (hsa-miR-182, hsa-miR-183 and hsa-miR-200a) and two underexpressed miRNAs (hsa-miR-143 and hsa-miR-195) were evaluated in all of the patients included in this study. These miRNAs were selected because they were significantly deregulated in the initial deep sequecing analysis. snRNA U6 was used as the endogenous control. Real-Time qPCR was performed using the All-in-One™ miRNA qRT-PCR Detection Kit (GeneCopoiea Inc, Rockville, MD, USA). 10 µg of total RNA was converted to cDNA according to the manufacturer's protocol. PCR was performed in a total reaction volume of 20 µl, including 10 µl of 2xAll-in-One™ qPCR Mix, 2 µl of Universal Adaptor PCR Primer (2 µM), 2 µl of All-in-One™ qPCR Primer (2 µM), 2 µl of First-Strand cDNA (diluted in 1∶5), 50xROX Reference Dye 0.4 µl and 3.6 µl double-distilled water. The reactions were performed and analyzed using the ABI PRISM 7000 Fluorescent Quantitative PCR System (Applied Biosystems, Foster City, CA,USA). PCR reactions were performed for cancer and normal cDNA in triplicate for each set. The cycling parameters for PCR were as follows: (1) an initial denaturation step of 15 min at 95°C; (2) 40 cycles, with 1 cycle consisting of 15 s at 95°C, 20 s at 55°C, and 30 s at 70°C. The catalog numbers of All-in-One™ miRNA qPCR Primers are listed in Table S5. The median in each triplicate was used to calculate relative miRNA concentrations (ΔCt = CtmedianmiRNA−CtmediansnRNAU6). Expression fold changes were calculated using 2−ΔΔCt methods. The miRNA expression differences between cancer and control were analysed using Student's t test within SPSS (Version 16.0 SPSS Inc.). A value of p<0.05 was considered as statistically significant.