Abstract
Objectives
To identify gene signatures in transitional cell carcinoma that can differentiate high-grade T1 nonprogressive (T1NP) bladder cancer (BCa) from those T1 progressive (T1P) tumors that progress to muscularis propria–invasive T2 tumors.
Materials and methods
We performed a high-throughput RNA sequencing (RNA-Seq) on formalin-fixed and paraffin-embedded BCa specimens with clinical pathologic characteristics best representing the general clinical development of the disease. For the T1NP group, only patients with long-term follow-up (6–17 y) and periodic examinations (average of 4 resections and 9 cytology tests) were selected. For the T1P group, only patients in whom a complete resection was performed after a minimum of 8 months after the initial T1 diagnosis were selected, therefore eliminating the possibility of underdiagnosis. Only samples in which muscularis propria was present and uninvolved were included, further assuring a correct diagnosis. The RNA-Seq reads were mapped to the human genome build NCBI 36 (hg18) using TopHat with no mismatch. After alignment to the transcriptome and expression quantification, a linear statistical model was built using Limma between T1NP and T1P samples to identify differentially expressed genes.
Results
Overall, 5,561 genes were mapped to all samples and used for RNA-Seq analysis to identify a gene signature that was significantly and differentially expressed between patients with T1NP BCa and patients with T1P BCa. Signature-based stratification indicated the gene signature correlated notably with the time of T1 development to T2 tumor, suggesting that the molecular signature might be used as an independent predictor for the pace of high-grade T1 BCa progression.
Conclusions
This is the first demonstration that RNA-Seq can be applied as a powerful tool to study BCa using formalin-fixed and paraffin-embedded specimens. We identified a gene signature that can distinguish patients diagnosed with high-grade T1 BCas that remain as non–muscle invasive tumors from those patients with cancers progressing to muscle-invasive tumors. Our findings will make future large-scale clinical cohort studies and clinical trial–based studies possible and help the development of prognostic tools for accurate prediction of T1 BCa progression that may considerably influence the clinical decision–making process, treatment regimen, and patient survival.
Keywords: T1 bladder cancer, RNA-Seq, Tumor progression, Molecular signature
1. Introduction
Management of high-grade T1 (lamina propria invasive) bladder cancer (BCa) poses a challenging dilemma to clinicians and patients. Most cases of patients with BCa present initially as non–muscle invasive transitional cell carcinoma (TCC). There are a significant number of high-grade T1 lesions that have the potential to progress to muscle-invasive BCa with increased risk for developing metastatic disease [1,2]. For high-grade T1 TCC, approximately 80% of treated tumors recur and 5% to 15% of recurrent tumors progress to invasive disease. Once BCa becomes metastatic, the cancer-related survival is approximately 5.4% at 5 years [3]. Currently, no definitive risk criteria exist to distinguish patients with T1 nonprogressive (T1NP) BCa who may suffer multiple recurrences of the disease without developing muscle-invasive tumors from those patients with the T1 progressive (T1P) BCa whose cancer first presented as T1 disease but eventually developed into muscle-invasive and metastatic disease. Therefore, establishing new prognostic criteria to distinguish high-grade T1NP from T1P would meet a great clinical need.
Previously, BCa prognostic factors have been reported and specific markers and molecular pathways have been linked with BCa tumor stage and progression [4–6]. Thus far, no standard guidelines have been adopted into clinical practice. Most of these studies used samples pooled from distinct groups of noninvasive BCa (Ta and T1) tumors and compared them with muscularis propria–invasive BCa (T2, T3, and T4) tumors. Specimens with long-term clinical T1 BCa follow-up were lacking. This makes distinguishing patients with T1NP BCa from patients with T1P particularly difficult, if not impossible from these studies.
Gene expression profiling has been used to develop prognostic signatures in a wide range of diseases [7,8], but its application has been limited, to an extent, by the fact that gene expression technologies work best with fresh frozen tissue [9–11]. By far, the vast majority of human disease tissue samples and those with the best outcome data are archived formalin-fixed paraffin-embedded (FFPE) surgical specimens. Important clinical information and disease outcome are often collected years after the initial specimen collection. However, FFPE tissue processing is known to cause fragmentation and chemical modification of RNA, presenting challenges for gene expression profiling. The analysis of BCa specimens is potentially more difficult as they are typically obtained by transurethral resection (TUR) and are associated with cautery artifacts that may further degrade nucleic acids.
High-throughput RNA sequencing (RNA-Seq) is a recently developed gene expression profiling method that has several advantages over other expression profiling technologies, including higher sensitivities, low background noise, and the ability to detect splicing variants and somatic mutations, resulting in precise measurements of levels of transcripts and their isoforms [12,13]. Using RNA-Seq, expression profiling genes involved in various cancers were identified and showed potential for biomarker-driven clinical trials and targeted cancer therapy [14–16]. To this date, the RNA-Seq technology using FFPE samples on BCa analysis has not been published.
In this pilot study, our goal was to test whether RNA-Seq analysis with archival FFPE BCa specimens could be used to identify a genomic signature capable of differentiating high-grade BCa of T1NP from those T1 diseases that eventually progress to muscle-invasive tumors. We used FFPE samples of patients with BCa with long-term follow-up to determine the natural history of the disease. RNA-Seq analysis performed on these FFPE samples identified a gene expression signature associated with disease progression. Our results demonstrated the applicability of using RNA-Seq to study bladder tumors obtained by TUR and stored long term as FFPE specimens. The gene signature identified by this study may lead to a promising diagnostic tool for high-risk patients with tumors that rapidly progress to muscle-invasive BCa.
2. Materials and methods
2.1. Patients and tumor specimens
Our research was conducted with the approval by the hospital Institutional Review Board at Massachusetts General Hospital. All the samples were obtained from patients treated at Massachusetts General Hospital between 1994 and 2011. BCa tumor FFPE specimens from 3 patients with T1NP and 4 patients with T1P were analyzed for each experiment condition. An expert urologic pathologist (C.L.W) carefully evaluated specimens of papillary BCa, and the tumor tissues were identified, marked, and highlighted. Two cores with only papillary bladder tumors and without any stromal or muscular tissues were collected from highlighted areas using a biopsy punch with a plunger (1.5 mm in diameter) in a RNA-free environment. Cancer grade was assigned using the 1973 World Health Organization pathology (WHO) criteria for BCa, and cancer stage was assigned according to the seventh edition of American Joint Committee on Cancer TNM system [17,18].
2.2. RNA extraction, rRNA removal, and sequencing library construction
Total RNA was extracted using hot phenol with additional purification using the RNeasy mini kit (Qiagen, Germantown, MD) following the manufacturer's instructions. RNA integrity was assessed using an Agilent bioanalyzer (Agilent, Santa Clara, CA), and the RNA integrity number was calculated for each sample and the average RNA integrity number for all samples was 2.6, ranging from 2.3 to 3.3. A cDNA library was constructed for each sample using Illumina's mRNA-Seq Sample Prep Kit (Illumina San Diego, CA). Briefly, for each sample, 100 ng of total RNA was used to generate a sequencing library and the RNA was directly subjected to fragmentation without the mRNA purification step. The resulting sample libraries were subjected to double-specific nuclease (DSN) treatment using the Trimmer-Direct cDNA Normalization kit (Evrogen, Moscow, Russia).
2.3. RNA-Seq read mapping and annotation
The cDNA library of each sample was loaded to a single lane of Illumina flow cell and the libraries were sequenced on Illumina Genome Analyzer II. Image deconvolution and calculation of quality value were performed using the Boat module (Firecrest v1.1.4.0 and Bustard v.1.4.0 programs) of Illumina pipeline V1.4. Sequence base calls were assigned using Illumina CASAVA software. The sequence reads were 36 bases long and each lane produced on an average 30 million of 36-mer raw sequence reads. Reads were mapped to the human genome build NCBI 36 (hg18) using TopHat [19] with no mismatch. The mapped reads were assembled and annotated using Cufflinks software tools [20].
2.4. Transcript quantification and gene expression consolidation
Transcript abundances were quantified in fragments per kilobase of exon per million fragments mapped by Cufflinks by taking into account both the gene length and the mapped reads for each sample and normalized accordingly [21]. Owing to the fragmented nature of mRNA in FFPE samples, we focused on the abundance measurement at the gene level. When multiple transcript abundance measurements were reported for a gene, the maximum value was chosen to represent the expression level of that gene.
2.5. Differentially expressed genes identification in RNA-Seq data
After alignment to the transcriptome and expression quantification, a linear statistical model was built using limma [21,22] between 3 T1NP samples and 4 T1P samples to identify differentially expressed genes with T1NP samples as the reference. The analyses were accomplished using R and Bioconductor packages [23]. To validate our gene signature, Illumina cDNA-mediated Annealing, Selection, extension, and Ligation Cancer Panel (DASL) Assay was performed independently with the same specimens and genes were identified to show significantly differential expressions between patients with T1NP and patients with T1P using both the RNA-Seq and the DASL platforms.
2.6. Functional enrichment and network analysis
Network enrichment for the significantly differentially expressed genes was analyzed using Ingenuity Pathway Analysis (IPA) software. The network interaction of the focused genes in the network is based on their connectivity in Ingenuity Knowledge Base.
3. Results
3.1. Patient cohort characteristics
To ensure the high specificity of the gene signature obtained, we selected tissue samples from patients with the clinicopathologic characteristics that best represented the general clinical development of nonprogressive from progressive disease. We selected 7 patients in this study based on several considerations. First, all patients had well-established lamina propria–invasive TCC confirmed by an expert urologic pathologist. Second, every patient's disease history and his/her pathology report was carefully evaluated to ascertain BCa progression. Third, only patients with extended follow-up and sufficient tumor cells in the paraffin blocks were included in this study. All specimens had muscularis propria present that was uninvolved. Patients had a short-interval follow-up biopsy after initial diagnosis, which demonstrated no tumor. As shown in Table 1, for the T1NP group, these 3 patients were confirmed as non-progressive disease by subsequent biopsies and with significantly extended follow-up times, ranging from 6 to 17 years (average 9 y). For patients with T1P, we excluded any cases with short interval between T1 and T2 diagnosis, eliminating the possibility of T1 BCa being understaged. The 4 patients with T1P had an average time to development of muscularis propria–invasive cancer of 2.43 years, ranging from 0.8 to 4.5 years and with an average follow-up time of 4.75 years, ranging from 1 to 8 years. All these data coincided with the general clinical observation of T1 BCa progression.
Table 1.
Clinical pathologic characteristics of study cohort
| T1 nonprogressive (T1NP) | T1 progressive (T1P) | |
|---|---|---|
| Samples | n = 3 | n = 4 |
| Median age, y (range) | 69 (56–79) | 73 (62–85) |
| Male:female | 2:1 | 3:1 |
| Clinical tumor stage | T1 | T1 |
| T1 | 3 | 4 |
| Grade 1–2 | 0 | 0 |
| Grade 3 | 3 | 4 |
| Average follow-up years (range) | 9 (6–17) | 4.75 (1–8) |
| Average progression years (range) | 0 | 2.43 (0.8–4.5) |
3.2. RNA-Seq analysis of T1NP and T1P tumors
We performed RNA-Seq using the Illumina GAII platform. The study design and the workflow for the RNA-Seq are illustrated in Fig. 1 and are described in detail in Materials and Methods. After transcript quantification, 11,092 genes were detected in at least 1 of the 7 samples, with 6,143 genes with multiple transcripts and 4,929 genes with one transcript. An unbiased analysis of the expression data revealed that 5,561 genes were found to be expressed in all samples and it is this final set that was used for further analysis. The characteristics of the RNA-Seq data are summarized in Table 2. On average, approximately 21 million sequencing reads were generated covering 47 million exon bases. Not all the genes could be mapped to the transcribed database, likely owing to genetic variations and repetitive elements and tandem repeats [24]. The average number of genes mapped was 8,149, which is substantially smaller than what has been reported when using RNA derived from fresh or frozen samples [24].
Fig. 1.
Study design and RNA-Seq analysis workflow.
Table 2.
RNA-Seq data summary. The read coverage of the 3 T1NP and 4 T1P, patients’ samples were shown.
| Sample ID | Block age (y) | BCa group | Number of reads | Number of detected genes | Number of exon bases covered |
|---|---|---|---|---|---|
| BLK004 | 11 | NP | 21,879,070 | 8,790 | 42,922,505 |
| BLK007 | 16 | NP | 26,555,690 | 9,092 | 47,994,767 |
| BLK016 | 10 | NP | 33,927,499 | 7,567 | 43,266,436 |
| BLK020 | 5 | P | 10,888,159 | 7,358 | 50,383,266 |
| BLK021 | 15 | P | 25,313,779 | 8,935 | 48,868,444 |
| BLK025 | 6 | P | 18,278,586 | 8,789 | 35,467,184 |
| BLK034 | 6 | P | 12,497,222 | 6,513 | 61,692,969 |
3.3. Identification of differentially expressed genes in T1NP compared with T1P samples
We used Limma [22] to construct a linear model and identified 181 significantly differentially expressed genes (Table 3; P < 0.05) between the nonprogressive and progressive patients whose original diagnosis was high-grade T1 BCa. Among them, we found 101 up-regulated and 80 down-regulated genes in T1P relative to T1NP. The results were validated with the DASL Cancer Panel analysis using the same samples. We then performed average-linkage hierarchical clustering using a Pearson correlation coefficient distance metric using the gene signature identified (Fig. 2A). The green bar indicates 3 T1NP samples and the red bar for 4 T1P samples. Rows are median-centered log2 (fragments per kilobase of exon per million fragments mapped) for each gene with relative high levels of expression in red and low levels in blue. On examining the hierarchical clustering dendrogram, one can see a correlation between gene expression levels and the time for disease progression for the 4 T1P samples. As shown in Fig. 2B, it took 0.8, 1.4, 3, and 4.5 years for patient BLK25, BLK20, BLK34, and BLK21, respectively, for the disease to progress from T1 to muscle-invasive BCa. This suggests that the biomarker we identified may also help to predict the pace of progression from high-grade T1 BCa to muscle-invasive BCa, although a much larger sample size would be necessary to validate this observation. However, if validated, this result could have a direct effect on clinical decision–making. To further understand the biology underlying the patterns of differential expression, we used IPA software, to search for overrepresentation of biological pathways and annotated gene functional classes among the genes found to be significantly differentially expression between patients with T1P and T1NP and the top enriched network associated with T1 BCa progression was shown in Fig. 3.
Table 3.
Annotation of significantly differentially expressed genes between patients with T1P BCa and patients with T1NP BCa
| Gene symbol | log FC(P/NP) | t | P value |
|---|---|---|---|
| IGFBP5 | 5.28378713 | 6.394460723 | 0.000213397 |
| LSP1 | 4.171875655 | 5.476182971 | 0.000596924 |
| STIM1 | 3.420212531 | 4.654638098 | 0.001649362 |
| APOL4 | –3.221334132 | –4.639699078 | 0.001681636 |
| CCPG1 | 2.172448093 | 4.391558528 | 0.002331371 |
| ANTXR2 | 2.254543518 | 4.232024004 | 0.002890112 |
| C10orf76 | –3.818675576 | –4.214632711 | 0.002959271 |
| ABCA5 | –2.381645938 | –4.029101417 | 0.003819003 |
| TBC1D4 | 2.729309403 | 3.930572519 | 0.004381905 |
| OPTN | 2.1810596 | 3.889946098 | 0.004639408 |
| CYP4Z2P | –3.076981378 | –3.849316111 | 0.004913234 |
| VEGF | –2.408409244 | –3.748738239 | 0.00566838 |
| MDK | –2.212910642 | –3.745344698 | 0.005695932 |
| CYP4B1 | –3.622571589 | –3.717678918 | 0.005925963 |
| AGPS | 2.694819961 | 3.687937586 | 0.006184361 |
| NLRP1 | 3.727425324 | 3.64172336 | 0.006610026 |
| NFIA | 1.838046144 | 3.604198172 | 0.006978631 |
| CDC42BPG | –2.284311666 | –3.549514259 | 0.007555489 |
| SH3D19 | 1.911486029 | 3.548235041 | 0.007569576 |
| KALRN | –3.728133666 | –3.544078123 | 0.007615546 |
| ITGA2 | 2.800984034 | 3.511152117 | 0.007990317 |
| PLA2G2F | –4.156008011 | –3.508019408 | 0.008026982 |
| ALDH16A1 | –2.862696914 | –3.505490848 | 0.008056706 |
| LTBP1 | 2.656543111 | 3.478779402 | 0.008377941 |
| DHRS7 | 2.279694661 | 3.462326805 | 0.008582544 |
| 5S_rRNA | 5.028224923 | 3.436200666 | 0.008918409 |
| CASP8 | 2.615453795 | 3.415985067 | 0.009187841 |
| PDLIM5 | 1.708273735 | 3.39620038 | 0.009459889 |
| TRAF4 | –2.350257868 | –3.381925006 | 0.00966147 |
| RAB31 | 2.357119785 | 3.370061392 | 0.009832451 |
| ANXA7 | 2.192966094 | 3.324721288 | 0.010515958 |
| CAMK2N1 | –2.520187721 | –3.293483171 | 0.011015952 |
| PICK1 | –2.627547988 | –3.274092328 | 0.011338883 |
| AUH | 1.566006828 | 3.258177995 | 0.01161137 |
| ACVR1 | 2.242422498 | 3.245297375 | 0.011836961 |
| HLA-K | 2.547641918 | 3.2147809 | 0.012390042 |
| RNF115 | –1.854308544 | –3.17808107 | 0.013091395 |
| RNF145 | 1.921960747 | 3.165520893 | 0.013340906 |
| AMD1 | 1.872288678 | 3.157376316 | 0.013505363 |
| F3 | 3.925845101 | 3.152687846 | 0.013600998 |
| SLC19A2 | –2.408176319 | –3.14173746 | 0.013827139 |
| MAN2A1 | 1.853521362 | 3.141501987 | 0.013832045 |
| SNX14 | 2.178844054 | 3.10806374 | 0.014547586 |
| LY75 | 1.982609555 | 3.089435701 | 0.014962925 |
| TCIRG1 | 1.519819127 | 3.078512561 | 0.015212227 |
| TOR1B | 1.645119041 | 3.057636916 | 0.01570084 |
| PTGR1 | –2.445310476 | –3.044717278 | 0.016011429 |
| CRISPLD2 | 2.737330915 | 3.035132503 | 0.016245991 |
| BHLHE41 | –2.305189141 | –3.028038491 | 0.016421907 |
| SIPA1L2 | 1.830611683 | 3.025830923 | 0.016477054 |
| CEP164 | 3.297527769 | 3.008038912 | 0.016928636 |
| PAPSS1 | 1.73903819 | 2.975601371 | 0.017785457 |
| GNPDA1 | 1.762492611 | 2.933743377 | 0.018958336 |
| LONP1 | –1.56359523 | –2.933095568 | 0.018977107 |
| KIAA0182 | –1.622904521 | –2.913291303 | 0.019560385 |
| LIG1 | –2.313584528 | –2.906708453 | 0.019758362 |
| CD44 | 4.002931343 | 2.898246336 | 0.020015917 |
| PLEKHG6 | –1.982202313 | –2.880303583 | 0.020573615 |
| FAM46A | 2.044259415 | 2.880296689 | 0.020573832 |
| PPFIBP1 | 1.789928588 | 2.879589157 | 0.020596151 |
| RP11-345P4.4 | –1.757416661 | –2.875835802 | 0.020714972 |
| POLE | –2.623991911 | –2.851535329 | 0.021501641 |
| PON2 | 2.032836251 | 2.826353505 | 0.022349529 |
| HMGN3 | 1.72280555 | 2.825036704 | 0.022394807 |
| COL7A1 | 3.970018254 | 2.813795234 | 0.022785223 |
| C19orf33 | 1.440515121 | 2.792784855 | 0.023533875 |
| POFUT2 | –1.487815721 | –2.787720939 | 0.023718083 |
| AKAP1 | –1.803718104 | –2.786099847 | 0.023777367 |
| SLC35D1 | 1.466010414 | 2.777427063 | 0.02409714 |
| ASCC2 | –2.033023461 | –2.773047563 | 0.024260298 |
| CCND1 | 3.259008872 | 2.772126521 | 0.024294756 |
| BTN3A3 | 1.72513362 | 2.762907815 | 0.024642439 |
| TAF10 | 1.657211243 | 2.759334651 | 0.024778575 |
| SIPA1L3 | –1.669399938 | –2.745252431 | 0.025322692 |
| ANXA1 | 4.206731688 | 2.745215675 | 0.025324128 |
| TINF2 | 1.890822482 | 2.743158705 | 0.025404634 |
| CSPG2 | 3.90783955 | 2.73594918 | 0.025688889 |
| KIAA0415 | –2.105622992 | –2.730688823 | 0.025898357 |
| ITGB7 | –1.908655609 | –2.725256153 | 0.02611653 |
| NUMB | 1.394564947 | 2.722490684 | 0.026228315 |
| GPC1 | 2.378547736 | 2.703836999 | 0.026995286 |
| C20orf194 | 2.388977829 | 2.678685101 | 0.02806606 |
| GGA1 | –1.922870984 | –2.672079037 | 0.028354455 |
| PAK1 | 1.547808712 | 2.663626083 | 0.028727912 |
| SLC12A6 | 1.791246724 | 2.660450487 | 0.028869511 |
| DDB2 | –1.762351611 | –2.659926863 | 0.028892928 |
| SAMD9L | 2.379492442 | 2.657272638 | 0.029011927 |
| UNC13B | –1.938490147 | –2.654444283 | 0.029139285 |
| RNF141 | 1.824007956 | 2.64865981 | 0.02940154 |
| PSEN1 | 1.614933918 | 2.644278862 | 0.029601769 |
| MAPKAPK3 | 2.450534752 | 2.640518671 | 0.029774739 |
| TMEM201 | –1.59423922 | –2.619510743 | 0.030760314 |
| MORC2 | –1.628752784 | –2.618769431 | 0.030795695 |
| NEDD9 | –1.961975214 | –2.618627696 | 0.030802465 |
| GATA2 | –2.511628364 | –2.618350233 | 0.030815721 |
| FBXL5 | 1.345326223 | 2.607768178 | 0.031325674 |
| FARP2 | 1.725692393 | 2.60722749 | 0.03135196 |
| AARS2 | –1.977567987 | –2.60127093 | 0.03164304 |
| FAM73B | –2.597450137 | –2.59632296 | 0.031886929 |
| AKR1C3 | –1.563447818 | –2.595660351 | 0.031919735 |
| DOT1L | –1.418412546 | –2.590124605 | 0.03219516 |
| DHX8 | –1.793769773 | –2.585869975 | 0.03240849 |
| C9orf84 | 2.742453953 | 2.581254025 | 0.032641569 |
| DSCR1 | 1.822394025 | 2.570407591 | 0.033195996 |
| RECQL5 | –1.830666864 | –2.563814258 | 0.033537701 |
| REEP3 | 1.698774983 | 2.562182046 | 0.033622843 |
| MAP3K7IP2 | 1.596373857 | 2.559860853 | 0.033744304 |
| SS18L1 | –1.39546114 | –2.558763081 | 0.033801903 |
| SKAP2 | 1.821616802 | 2.549410379 | 0.034296703 |
| FKBP3 | 1.642847336 | 2.546576449 | 0.03444808 |
| SLC25A24 | 1.433158668 | 2.544613401 | 0.034553337 |
| PDE4DIP | –2.144480334 | –2.541263209 | 0.034733726 |
| KLHL22 | –1.580374914 | –2.539279442 | 0.034840992 |
| DST | 2.17427338 | 2.536334286 | 0.035000864 |
| AHCTF1 | 1.574229907 | 2.532632073 | 0.035202889 |
| CEACAM1 | –1.726409623 | –2.531454922 | 0.035267373 |
| ENGASE | –2.276362919 | –2.525395678 | 0.035601197 |
| DCBLD1 | 1.875591159 | 2.521535273 | 0.035815552 |
| RPL36AL | 1.54280083 | 2.511514843 | 0.036378087 |
| PLSCR3 | 1.420903374 | 2.51142349 | 0.036383256 |
| PPP2R5C | 1.307050755 | 2.510440903 | 0.036438906 |
| GRB7 | –1.720623113 | –2.508716287 | 0.03653679 |
| UGCGL2 | –2.647291621 | –2.50680118 | 0.036645799 |
| CARD14 | –1.409730182 | –2.502937857 | 0.036866705 |
| USP36 | –1.252316221 | –2.491910656 | 0.037504699 |
| C22orf9 | 1.721359342 | 2.486442274 | 0.037825219 |
| SH3BGRL | 2.009799067 | 2.48497519 | 0.03791168 |
| TNFAIP2 | –1.661063451 | –2.484647383 | 0.037931027 |
| MAP3K7 | 1.282293995 | 2.479406568 | 0.038241689 |
| OTUD7B | –1.896353997 | –2.476276125 | 0.038428481 |
| PDSS2 | –2.338248486 | –2.476129154 | 0.038437273 |
| MGST2 | 1.462079037 | 2.472563455 | 0.038651211 |
| NUPL2 | 1.628957352 | 2.467877477 | 0.038934198 |
| DPYSL3 | 1.91576424 | 2.464112227 | 0.0391631 |
| ZBTB4 | 1.21710635 | 2.463520042 | 0.039199225 |
| ACOT11 | –2.070513114 | –2.462763559 | 0.039245421 |
| HEG1 | 2.584471663 | 2.460518133 | 0.039382866 |
| UGT1A8 | 2.849211631 | 2.458255547 | 0.039521854 |
| PAN2 | –2.293607666 | –2.455013779 | 0.039721857 |
| RAD51L3 | –1.29829818 | –2.45280216 | 0.039858891 |
| CYLD | 1.374783471 | 2.448055195 | 0.040154632 |
| CRAT | –2.755364275 | –2.448014399 | 0.040157183 |
| ACACA | –1.275498781 | –2.445917734 | 0.040288521 |
| HDAC5 | –1.451246253 | –2.441489292 | 0.040567351 |
| CHAF1A | –1.874461565 | –2.427956095 | 0.041431578 |
| PNMA1 | 1.646612223 | 2.427569109 | 0.041456562 |
| UBE2D3 | 1.170890055 | 2.426386474 | 0.041533008 |
| ZNF704 | –2.638607084 | –2.424684105 | 0.041643299 |
| FAM117B | 1.969329664 | 2.421711392 | 0.041836601 |
| ACOX3 | –1.370497072 | –2.421331019 | 0.041861401 |
| SETD7 | 1.668513412 | 2.415543953 | 0.042240531 |
| C4orf34 | 1.345863238 | 2.411436977 | 0.042511691 |
| IL6ST | 1.491818909 | 2.41008791 | 0.042601144 |
| PTEN | 1.251488632 | 2.408825886 | 0.042684997 |
| FAM179B | 1.536096482 | 2.405452694 | 0.04290994 |
| CBLC | –1.352667824 | –2.404041938 | 0.04300437 |
| HOOK2 | –1.902609985 | –2.403003647 | 0.043074003 |
| CGN | –2.341892273 | –2.401814619 | 0.043153884 |
| EZH2 | –1.845958401 | –2.390117617 | 0.043947702 |
| XPNPEP3 | –1.735687018 | –2.388371373 | 0.044067465 |
| LSG1 | –1.353533592 | –2.382805061 | 0.044451413 |
| CCDC109A | 1.486079427 | 2.382529921 | 0.044470479 |
| GPT2 | –2.301677625 | –2.378141626 | 0.044775671 |
| CACNA1D | –3.552311409 | –2.371912571 | 0.045212499 |
| AFMID | –1.738027331 | –2.364302478 | 0.045751988 |
| ZNF330 | 1.623902688 | 2.359784763 | 0.046075305 |
| DGKZ | –1.828659304 | –2.357378201 | 0.046248469 |
| CSAD | –2.268177751 | –2.354431508 | 0.046461388 |
| ESD | 1.479256457 | 2.351836341 | 0.04664972 |
| DOPEY2 | –1.502561646 | –2.34891318 | 0.046862771 |
| GPRC5A | –1.80242518 | –2.342122595 | 0.047361468 |
| C6orf134 | –1.73080824 | –2.340904003 | 0.047451521 |
| MARCH5 | 1.421978347 | 2.337675959 | 0.047690903 |
| CD164 | 1.263919165 | 2.337024623 | 0.04773935 |
| LFNG | 1.815659694 | 2.333772259 | 0.047982005 |
| BIRC2 | 2.272732375 | 2.321656288 | 0.048896876 |
| AEBP1 | 3.476393655 | 2.319486911 | 0.049062517 |
| CPSF3 | 1.433022399 | 2.319243225 | 0.049081158 |
| COPB2 | –1.174259775 | –2.318070989 | 0.049170931 |
| RPS23 | 1.364880755 | 2.31757051 | 0.049209308 |
| NARF | –2.194601067 | –2.307580216 | 0.04998168 |
Annotation of significantly differentially expressed genes between the nonprogressive and progressive groups of patients with T1 BCa.
The 181 most significantly differentially expressed genes (P < 0.05) that can distinguish high-grade T1NP tumors with nonprogressive recurrence from those T1P muscle-invasion tumors are listed. The genes were sorted based on the P value and their expression levels were indicated by the log FC (P/NP) values, genes that overexpressed were with positive log FC (P/NP) values and genes that underexpressed were with negative log FC (P/NP) values.
Fig. 2.
Heatmap and sample clustering analysis of the significantly differentially expressed genes between 3 T1NP and 4 T1P samples. Hierarchical cluster analysis of RNA-Seq data profiling bladder tissue obtained from 3 patietns with T1NP of know BCa with nonprogressive status and 4 patients with T1P BCa with known muscle-invasive progressive disease, first diagnosed as nonprogressive (T1) and then progressed to T ≥ 2 progressive BC. There were 181 genes significantly differentially expressed between 3 T1NP and 4 T1P samples with P < 0.05. (A) Each column represents a sample and each row a gene. Red represents a higher level of gene expression and blue represents a lower level gene expression, relative to the median across all samples for each gene. The green bar above the heatmap represents T1NP samples and red bar T1P samples. (B) the hierarchical clustering dendrogram from data obtained from the 4 T1P samples shown in (A) was plotted regarding the time for T1 progression to muscle-invasive tumor. For patients with BLK21, BLK20, BLK24, and BLK34, it took 4.5, 1.4, 0.8, and 3 years for their disease to progress from T1 to T2 (muscle invasive) BCa. y-Axis indicates the time (years) for the progression from T1 to muscle-invasive BCa.
Fig. 3.
The top enriched network associated with BCa progression. Map was created by Ingenuity Pathway Analysis (IPA) software. Red icons depict genes up-regulated and green icons depict gene down-regulated. Lines report interactions between proteins.
4. Discussion
Genomics has created an unprecedented opportunity to survey expression patterns across the genome and to use the resulting data to develop diagnostic and prognostic bio-markers. However, doing this requires the availability of well-annotated clinical samples with extensive clinical data so that patterns of gene expression can be linked to outcome or other relevant end points.
Although there are many archival pathologic samples reserved as FFPE tissues, these have proven difficult to analyze using most of the available genomic technologies. This is largely owing to the fact that the process of creating FFPE samples is known to introduce chemical modification and cross-linking between DNA, RNA, and proteins in these samples. In BCa, the situation has proven particularly difficult as specimens are typically obtained through TUR and the cautery effect associated with the procedure may further degrade nucleic acids. To overcome these limitations, we used a variation on RNA-Seq technology that relies on short nucleic acid fragments coupled with DSN normalization. DSN removes ribosomal RNA and other abundant double-stranded DNA and DNA-RNA hybrid complexes, allowing creation of RNA-Seq libraries from most highly degraded samples [25,26]. Using this approach, we identified a gene signature that significantly differentially expressed between the nonprogressive and progressive T1 BCa (Table 3). Our data demonstrated that FFPE specimens obtained by TUR could be used in RNASeq, genome-wide expression analysis.
In our study, we went to great lengths to select samples that had an appropriate diagnosis and were of high quality. It is known that up to 25% of T1 G3 tumors are incorrectly staged on initial surgical resection (TUR of bladder [TURB]). Studies without systematically repeated TURB could have been hampered by the lack of patients with correctly documented T1 who have tumors that subsequently progress at a later date. It is therefore critical that a re-resection be performed, particularly if muscularis propria is not present in the specimen. In our study, a complete resection was performed for each and all patients with a minimum of at least 8 months after the initial T1 diagnosis. Only samples in which muscularis propria was present and uninvolved were included, further assuring a correct diagnosis. For each patient included in our study, there were longitudinal data, including TURB and cytology tests (between 14 and 24, ranging from 2–40), which followed the original diagnosis (data not shown). For the non-progressive group, one of the most critical issues is the length of the follow-up. In our study, patients with T1NP were followed up to 17 years with periodic examinations (average being more than 4 resections and 9 cytology tests), providing strong evidence to support our assignment of nonprogressive status. For the progressive group, the important issue was that a second resection should be performed following an initial T1 diagnosis without evidence of T2 disease. In our study, a sample was required to have a second resection to be accepted as T1 progression, therefore eliminating the possibility of underdiagnosis. Only with careful exclusion of patients not fitting these criteria can one be certain of a true T1 diagnosis and not a misdiagnosed T2 BCa.
Previous studies of superficial papillary BCa have associated specific genomic alterations with the disease, including mutations in and dysregulation of FGFR3, PI3K, KRAS, HRAS, TP53, P16, TSC1, and PTEN and loss of chromosome 9, 9p, or 9q as well as loss of RB1 [9,10,27–29]. Recently, mutations in genes responsible for chromatin remodeling were identified [30], and high-throughput assays including RNA-Seq performed on fresh tissue samples have shown promising results, demonstrating molecular assessment can be used for the development of mutation and pathway-based trials of targeted therapy for cancer patients [16]. Ultimately, a list of genes provides little insight into the biology underlying the patterns of differential expression. Consequently, we used IPA software to search for overrepresentation of biological pathways and annotated gene functional classes among the genes found to be significant. Among the highest ranking were cancer-related pathways associated with cell death, cellular growth and proliferation, and cell cycle (Fig. 3). Future studies are needed to understand the involvement of these molecules in BCa progression. In this study, we selected the portion of BCa tissue that was primarily epithelial cells and limited any stromal contaminant. Even so, it is conceivable that a small amount of stromal contamination could have occurred, but in view of the small volume, it is unlikely that it influenced the results.
In this exploratory study, we were able to identify a gene signature that was able to distinguish patients with T1NP from patients with T1P. Although encouraging, this study has some limitations including a small number of samples and a potential undersampling of tumor suppressor genes that may have been lost in the T1P patient group. Regardless, the fact that we are able to find an apparently robust signature that seems to track with progression is encouraging and suggests a more thorough prospective study is in order. We hope our study would have an effect on T1 BCa clinical management, other bladder diseases, and different types of cancers, and ultimately, play a role in facilitating the use of biomarker and gene signatures in the clinical settings.
5. Conclusions
This study identifies a gene signature that can distinguish patients diagnosed with high-grade T1 BCa that remain as non–muscle invasive tumors from those patients with cancers progressing to muscle-invasive tumors. This is the first demonstration that RNA-Seq can be applied as a tool to study BCa using FFPE specimens. Our findings could influence clinical T1 BCa treatment regiment and thereby effect patient survival.
Acknowledgments
We thank Renee Rubio and Fieda Abderazzaq, Department of Biostatistics and Computational Biology and Center for Cancer Computational Biology, Dana-Farber Cancer Institute for excellent technical support.
References
- 1.Cookson MS, Hen HW, Zhang ZF, et al. The natural history of high risk superficial bladder cancer: 15 year outcome. J Urol. 1997;158:62–7. doi: 10.1097/00005392-199707000-00017. [DOI] [PubMed] [Google Scholar]
- 2.William SG, Stein JP. Molecular pathways in bladder cancer. Urol Res. 2004;32:373–85. doi: 10.1007/s00240-003-0345-y. [DOI] [PubMed] [Google Scholar]
- 3. 〈 http://seer.cancer.gov/statfacts/html/urinb.html#survival 〉 .
- 4.Sanchez-Carbayo M, Socci ND, Lozano J, Saint F, Cordon-Cardo C. Defining molecular profiles of poor outcome in patients with invasive bladder cancer using oligonucleotide microarrays. J Clin Oncol. 2006;24:778–89. doi: 10.1200/JCO.2005.03.2375. [DOI] [PubMed] [Google Scholar]
- 5.Schiffer E, Vlahou A, Petrolekas A, Stravodimos K, Tauber R, Geschwend JE, et al. Prediction of muscle-invasive bladder cancer using urinary proteomics. Clin Cancer Res. 2009;15:4935–43. doi: 10.1158/1078-0432.CCR-09-0226. [DOI] [PubMed] [Google Scholar]
- 6.Wiklund ED, Bramsen JB, Hulf T, Dyrskjøt L, Ramanathan R, Hansen TB, et al. Coordinated epigenetic repression of the miR-200 family and miR-205 in invasive bladder cancer. Int J Cancer. 2011;128:1327–34. doi: 10.1002/ijc.25461. [DOI] [PubMed] [Google Scholar]
- 7.Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, Mesirov JP, et al. Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science. 1999;286:531–7. doi: 10.1126/science.286.5439.531. [DOI] [PubMed] [Google Scholar]
- 8.Stuart RO, Wachsman W, Berry CC, et al. In silico dissection of cell-type-associated patterns of gene expression in prostate cancer. Proc Natl Acad Sci USA. 2004;101:615–20. doi: 10.1073/pnas.2536479100. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Blaveri E, Brewer JL, Roydasgupta R, et al. Bladder cancer stage and outcome by array-based comparative genomic hybridization. Clin Cancer Res. 2005;11:7012–22. doi: 10.1158/1078-0432.CCR-05-0177. [DOI] [PubMed] [Google Scholar]
- 10.Blaveri E, Simko JP, Korkola JE, et al. Bladder cancer outcome and subtype classification by gene expression. Clin Cancer Res. 2005;11:4044–55. doi: 10.1158/1078-0432.CCR-04-2409. [DOI] [PubMed] [Google Scholar]
- 11.Dyrskjot L, Kruhoffer M, Thykjaer T, et al. Gene expression in the urinary bladder: a common carcinoma in situ gene expression signature exists disregarding histopathological classification. Cancer Res. 2004;64:4040–8. doi: 10.1158/0008-5472.CAN-03-3620. [DOI] [PubMed] [Google Scholar]
- 12.Mardis ER. The impact of next-generation sequencing technology on genetics. Trends Genet. 2008;24:133–41. doi: 10.1016/j.tig.2007.12.007. [DOI] [PubMed] [Google Scholar]
- 13.Wang Z, Gerstein M, Snyder M. RNA-Seq: a revolutionary tool for transcriptomics. Nat Rev Genet. 2009;10:57–63. doi: 10.1038/nrg2484. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Huang Q, Lin B, Liu H, Ma X, Mo F, Yu W, et al. RNA-Seq analyses generate comprehensive transcriptomic landscape and reveal complex transcript patterns in hepatocellular carcinoma. PLoS One. 2011;6:e26168. doi: 10.1371/journal.pone.0026168. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Berger MF, Levin JZ, Vijayendran K, Sivachenko A, Adiconis X, et al. Integrative analysis of the melanoma transcriptome. Genome Res. 2010;20:413–27. doi: 10.1101/gr.103697.109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Roychowdhury S, Iyer MK, Robinson DR, Lonigro RJ, Wu YM, Cao X, et al. Personalized oncology through integrative high-throughput sequencing: a pilot study. Sci Transl Med. 2011;3:1–10. doi: 10.1126/scitranslmed.3003161. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Mostofi FK, Sobin LH, Torloni H. Histological Typing of Urinary Bladder Tumours. World Organization; Geneva: International histology classification of tumours, No. 10. p. 1973. [Google Scholar]
- 18.Edge SB, Byrd DR, Compton CC, Fritz AG, Greene FL, Trotti A, editors. American Joint Committee on Cancer (AJCC) cancer staging manual. 7th ed. Springer-Verlag; New York, NY: p. 2010. [Google Scholar]
- 19.Trapnell C, Pachter L, Salzberg SL. TopHat: discovering splice junctions with RNA-Seq. Bioinformatics. 2009;25:1105–11. doi: 10.1093/bioinformatics/btp120. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Trapnell C, Williams BA, Pertea G, Mortazavi AM, Kwan G, van Baren MJ, et al. Transcript assembly and quantification by RNA-Seq reveals unannotated transcripts and isoform switching during cell differentiation. Nat Biotechnol. 2010;28:511–5. doi: 10.1038/nbt.1621. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Mortazavi A, Williams BA, McCue K, Schaeffer L, Wold B. Mapping and quantifying mammalian transcriptomes by RNA-Seq. Nat Methods. 2008;5:621–8. doi: 10.1038/nmeth.1226. [DOI] [PubMed] [Google Scholar]
- 22.Smyth GK. In: Limma: linear models for microarray data. Bioinformatics and Computational Biology Solutions using R and Bioconductor. Gentleman R, Carey VJ, Huber W, Irizarry RA, Dudoit S, editors. Springer; New York: 2005. pp. 397–420. [Google Scholar]
- 23. 〈Available at: http://www.bioconductor.org/〉.
- 24.Li B, Ruotti V, Stewart RM, Thomson JA, Dewey CN. RNA-Seq gene expression estimation with read mapping uncertainty. Bioinformatics. 2010;26:493–500. doi: 10.1093/bioinformatics/btp692. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Toung JM, Morley M, Li M, Cheung VG. RNA-sequence analysis of human B-cells. Genome Res. 2011;21(6):991–8. doi: 10.1101/gr.116335.110. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Rupp GM, Locker J. Purification and analysis of RNA from paraffin-embedded tissues. Biotechniques. 1988;6:56–60. [PubMed] [Google Scholar]
- 27.Shujun L, Khrebtukova I, Perou C, Schroth GP. Complete RNA-seq analysis of cancer transcriptomes from FFPE samples. Genome Biol. 2010;11(Suppl. 1):P35. [Google Scholar]
- 28.Lindgren D, Frigyesi A, Gudjonsson S, Sjödahl G, Hallden C, Chebil G, et al. Combined gene expression and genomic profiling define two intrinsic molecular subtypes of urothelial carcinoma and gene signatures for molecular grading and outcome. Cancer Res. 2010;70:3463–72. doi: 10.1158/0008-5472.CAN-09-4213. [DOI] [PubMed] [Google Scholar]
- 29.McConkey DJ, Lee S, Choi W, Tran M, Majewski T, Lee S, et al. Molecular genetics of bladder cancer: emerging mechanisms of tumor initiation and progression. Urol Oncol. 2010;28:429–40. doi: 10.1016/j.urolonc.2010.04.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Gui Y, Guo G, Huang Y, Hu X, Tang A, Gao S, et al. Frequent mutations of chromatin remodeling genes in transitional cell carcinoma of the bladder. Nat Genet. 2011;43:875–8. doi: 10.1038/ng.907. [DOI] [PMC free article] [PubMed] [Google Scholar]