Oncomir-1 is an oncogenic cluster of microRNAs (miRNA) located on chromosome 13. Previous in vitro studies showed that it is transcriptionally regulated by the transcription factor E2F3. In this report, we combine expression profiling of both mRNA and miRNAs in Wilms' tumor (WT) samples to provide the first evidence that the E2F3-Oncomir-1 axis, previously identified in cell culture, is deregulated in primary human tumors. Analysis of RNA expression signatures showed that an E2F3 gene signature was activated in all WT samples analyzed, in contrast to other kidney tumors. This finding was validated by immunohistochemistry on the protein level. Expression of E2F3 was lowest in early-stage tumors and highest in metastatic tissue. Expression profiling of miRNAs in WT showed that expression of each measured member of the Oncomir-1 family was highest in WT relative to other kidney tumor subtypes. Quantitative PCR confirmed that these miRNAs were overexpressed in WT relative to normal kidney tissue. These results suggest that the E2F3-Oncomir-1 axis is activated in WT. Our study also shows the utility of integrated genomics combining gene signature analysis with miRNA expression profiling to identify protein-miRNA interactions that are perturbed in disease states. [Cancer Res 2008;68(11):4034–8]
- Pediatric cancers
- Gene expression profiling
- Cancer genome anatomy: comparative expression patterns
- Oncogenic transcription factors: leukemias/lymphomas/solid tumors
Wilms' tumor (WT) is associated with mutations and expression abnormalities in the transcription factor and putative tumor suppressor WT1. However, there has been a call for further investigation into other genes that participate in the development of WT ( 1).
Here, we combined expression profiling of both mRNA and microRNAs (miRNA) to identify expression patterns unique to WT. Microarray analysis is often plagued by poor signal-to-noise characteristics. We mitigated this challenge through the use of parametric gene set analysis (PGSEA). PGSEA increases the power of microarray expression analysis by analyzing sets of genes (curated from the literature) that have been empirically shown to change in expression in response to a defined stimulus. This approach has already been successfully used by our laboratory in the context of adult renal tumors ( 2, 3).
Our results indicated that the E2F3 signature is strongly activated in WT samples—a phenomenon not consistently observed in other renal tumors. E2F3 is an activator of transcription that is amplified or overexpressed in several tumors, including those of the bladder ( 4), prostate ( 5), and lung ( 6). Interestingly, E2F3 has been shown to drive the expression of Oncomir-1 in vitro ( 7). Oncomir-1 is an oncogenic cluster of miRNAs located on chromosome 13 that has been shown to play an important role in promoting tumor cell proliferation ( 8). Here, we show that both E2F3 and Oncomir-1 are overexpressed in vivo in WT. These results provide the first evidence that the E2F3-Oncomir-1 axis may be activated in a human tumor.
Materials and Methods
Tumor tissue samples. Gene expression was done on tissue specimens obtained from patients undergoing tumor resection at collaborating institutions. Additional tissues were acquired from the Cooperative Human Tissue Network. Immunohistochemistry validation of E2F3 protein expression was carried out using a WT tissue microarray assembled at the University of Chicago. For the tissue microarray, 57 WT (39 primary and 18 metastatic) were identified along with corresponding clinicopathologic data.
mRNA expression profiling. For each sample, mRNA was isolated by Trizol (Invitrogen) extraction. Gene expression profiles were then generated using the Affymetrix HG-U133 Plus 2.0 GeneChip from 27 WT samples, 10 clear cell renal cell carcinoma samples, 6 chromophobe renal cell carcinoma samples, 7 oncocytoma samples, 17 papillary renal cell carcinoma samples, and 12 samples of normal kidney tissue. Before data processing, probe set mappings were updated as described ( 9). Microarray data were uploaded to the Gene Expression Omnibus database, accession number GSE11024. Myc, E2F, SRC, RAS, β-catenin, synergistic HGF/VEGF, and von Hippel-Lindau (VHL) gene signatures were obtained from the literature as previously described ( 3). HGF and VEGF signatures were generated using data from the Gene Expression Omnibus (GDS406 and GDS495, respectively). In addition, we analyzed a hypoxia gene signature obtained from the literature ( 10) and an angiogenesis gene set generated by our group.
PGSEA was applied to identify up-regulated or down-regulated expression signatures in each sample ( 3, 11– 13). The resulting data set summarizing the relative up-regulation or down-regulation of each pathway for each sample was then processed using the Limma Bioconductor package ( 14) to identify pathways whose expression was differentially expressed compared with normal kidney tissue. P values for the Limma analysis were adjusted to control the false discovery rate at 5% ( 15).
Quantitative reverse transcription-PCR analysis of E2F3 isoforms. First-strand synthesis for each mRNA sample was done using oligo(dT)18 primer and the High Capacity cDNA Archive kit (Applied Biosystems). For each sample, 1 μg of total RNA was reverse transcribed into cDNA. For miRNA assays, 10 ng of total RNA were reverse transcribed using a sequence-specific primer provided with the miRNA TaqMan assay (Applied Biosystems). Quantitative PCR analysis was done using TaqMan assays. The commercially available TaqMan gene expression assay Hs01076037_m1, which spans the exon 1–2 boundary of E2F3a, was used to measure E2F3a levels. A custom TaqMan assay was designed to quantify E2F3b levels using forward primer TGCTTTCGGAAATGCCCTTACA, reverse primer GATGACCGCTTTCTCCTAGCT, and probe sequence CTTCGCTTTGCCTGCTGC. Expression levels were normalized to glyceraldehyde-3-phosphate dehydrogenase levels (using TaqMan gene expression assay Hs99999905_m1) by the ΔΔCt method.
Immunohistochemistry. Mouse anti-human E2F3 monoclonal antibody to E2F3 was purchased from Upstate UK and diluted 1:100 in antibody diluent. Sections of the Wilms' tissue microarray were obtained at 4-μm thickness, transferred to slides, deparaffinized, and rehydrated in the usual manner. Endogenous peroxidase was blocked with 3% H2O2. Microwave antigen retrieval was done by boiling the sections in 10 mmol/L EDTA buffer (pH 9.0) at low power for 12 min. After rinsing, sections were incubated with the primary antibody for 40 min at room temperature and washed with TBS (0.05 mol/L Tris, 0.12 mol/L sodium chloride, 0.05% Tween 20, pH 7.6). This step was followed by a 30-min incubation with goat anti-mouse IgG conjugated to a horseradish peroxidase–labeled polymer (EnVision+, DAKO). Slides were then developed for 5 min with 3,3′-diaminobenzidine chromogen and counterstained with hematoxylin. To assess proliferating cells, we used the Ki67 marker (clone Ki-S5, DAKO; 1:50 dilution) in a similar fashion. Negative controls were done by substituting the primary antibody with nonimmune mouse immunoglobulins. Nuclear expression of E2F3 and Ki67 was quantified by using the Automated Cellular Imaging System (Clarient). The percentage of cells with positive staining was determined for each tissue core by a trained pathologist (M.T.) as previously described ( 16).
MiRNA expression profiling. MiRNA was isolated with the mirVana extraction kit (Ambion). Expression levels relative to pooled normal kidney miRNA were measured by microarray analysis done by LC Sciences as described elsewhere. 8 We compared the miRNA expression of WT samples to miRNA expression data for clear cell renal cell carcinoma and papillary renal cell carcinoma described in a separate report. 8 Microarray data were uploaded to the Gene Expression Omnibus database, accession number GSE11016. Genes that were overexpressed in WT relative to other kidney tumor subtypes were identified by Kolmogorov-Smirnov rank-sum analysis ( 17).
Oncomir-1 expression levels were validated by quantitative reverse transcription-PCR (RT-PCR) using commercial TaqMan assays targeting human mir-17-5p, mir-18a, mir-19b, mir-20a, and mir-92. Expression levels of the miRNAs were normalized to the average expression of three endogenous controls, RNU6B, RNU49, and U47, using the corresponding TaqMan assays.
Results and Discussion
WT samples exhibit unique gene expression patterns as compared with other renal tumors. Unsupervised hierarchical clustering of genome-wide expression data revealed that WT specimens are more similar to each other in terms of gene expression than to other common types of kidney tumor or normal kidney tissue ( Fig. 1A ). PGSEA was then used to identify which biologically defined expression signatures are responsible for this unique gene expression pattern. Of the 11 malignancy-related pathways analyzed, two were identified whose expression deviated significantly from that of normal kidney tissue after P value correction: those related to E2F3 and Ras, the more significant of which was E2F3. As shown in Fig. 1B, PGSEA revealed that a set of genes previously identified as being up-regulated by E2F3 in vitro was also up-regulated in WT tumors relative to normal tissue. Consistent up-regulation of the E2F3 gene expression signature was unique to WT as compared with other renal tumors ( Fig. 1B).
As a control for our analytic approach, we also assessed the expression of genes that were down-regulated by VHL in vitro. As expected, the VHL derepression signature is evident in the clear cell renal cell carcinoma samples. As these are features unique to the clear cell renal cell carcinoma subtype, we expected that these changes would not be evident in the WT samples, and this was indeed confirmed.
The E2F3a isoform is preferentially up-regulated in WT tissue. E2F3 has two isoforms, E2F3a and E2F3b, which have different first exons and are believed to be driven by different promoters. The relative mRNA expression of these two isoforms was quantified by quantitative RT-PCR. E2F3a and E2F3b expression in the WT sample was normalized to the mean expression of these isoforms in clear cell renal cell carcinoma samples. E2F3 is known to be regulated in a cell cycle–dependent fashion. Therefore, it may be somewhat overexpressed in a nonspecific fashion simply due to the relatively elevated rate of proliferation exhibited by tumor tissue. To control for this nonspecific effect, we chose clear cell renal cell carcinoma samples as controls for this analysis as opposed to less proliferative normal tissue. The expression of E2F3a in WT samples was on average 8-fold higher than in clear cell samples, whereas the expression of E2F3b in WT samples was only 1.7-fold higher ( Fig. 1C), and this difference was significant by t test (P = 0.01).
These findings show that in contrast to the evidence in bladder tumors, E2F3a seems to be preferentially up-regulated in WT specimens ( 18).
E2F3 protein is overexpressed in WT and exhibits nuclear localization. Immunohistochemistry was done on a WT tissue array to confirm that the E2F3 activation apparent from gene expression data corresponded to an increase in E2F3 protein expression. Normal kidney tissue exhibited limited or no E2F3 positivity by immunohistochemistry, with the exception of medullary tissue, which exhibited moderate positivity in some cells. Where present in normal tissue, the E2F3 staining was confined to the cytoplasm ( Fig. 2A ). By contrast, WT tissue exhibits E2F3 expression localized to the nucleus.
E2F3 protein expression correlates with cell proliferation and is highest in metastatic tissue. No E2F3-positive nuclei were detected in the normal kidney samples, whereas up to 85% of nuclei were identified as E2F3 positive in tumor tissue, with the highest proportions of positive nuclei being found in high-grade tumors and metastatic lesions ( Fig. 2B). These samples were also stained for Ki67 as a marker of cellular proliferation. The proportion of E2F3-positive nuclei was correlated with the proportion of Ki67-positive nuclei in these samples (r = 0.66, P < 0.0001; Fig. 2C).
These results indicate that E2F3, previously known to be up-regulated in bladder ( 4), prostate ( 5), and lung tumors ( 6), is overexpressed on the mRNA and protein levels in WT. Its expression is highest in metastatic tissue, although it is unclear whether this is functionally related to the process of metastasis.
Oncomir-1, a regulatory target of E2F3, is also overexpressed in WT cells. Woods and colleagues have previously shown that E2F3 drives the expression of an oncogenic cluster of miRNAs, termed Oncomir-1 ( 3). Therefore, we hypothesized that the E2F3 overexpression observed in our WT samples may also be associated with Oncomir-1 overexpression. We examined the expression of members of the Oncomir-1 in miRNA profiles generated by the laboratory.
First, unsupervised clustering of our miRNA data was done to verify the quality of the data. All the WT samples clustered together as compared with other types of renal tumors, based on global miRNA expression ( Fig. 3A ). We then carried out a Kolmogorov-Smirnov rank-sum analysis to identify those miRNAs that were uniquely overexpressed in WT samples relative to other subtypes. Consistent with our hypothesis, of the five Oncomir-1 members assayed on the miRNA microarray, three were in the top 20 miRNAs uniquely overexpressed by WT ( Table 1 ). For each of the five miRNAs of the Oncomir-1 members assayed on the miRNA microarray, the highest expression was observed among the WT samples ( Fig. 3B) as compared with other renal tumor subtypes.
The gene expression data were validated by quantitative RT-PCR of five members of the Oncomir-1 family from WT samples. Consistent with the microarray results, each of these miRNAs was overexpressed between 2- and 25-fold in WT samples compared with normal tissue (P < 0.05 for each miRNA).
Oncomir-1 has previously been shown to promote proliferation in tumor cells ( 8), and Woods and colleagues ( 7) have shown in vitro that its expression is positively regulated by E2F3. Our results provide the first data that this E2F3-Oncomir-1 axis is activated in a human tumor.
Conclusions. In this study, we have integrated PGSEA analysis of mRNA expression data with miRNA expression profiling to identify novel features of WT. Whereas the body of literature describing abnormalities in miRNA expression in cancer is rapidly growing, much remains to be determined about both the upstream causes and the downstream consequences of the miRNA abnormalities. We are confident that integration of mRNA and miRNA expression profiling will be a useful tool in elucidating these relationships, such as the activation of the E2F3-Oncomir-1 axis we have observed in WT.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant support: The Van Andel Family, the Gerber Foundation, the Hauenstein Foundation, and in part by NIH grant HD046377-01A1 (E.J. Kort).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank the DeVos Children's Hospital and the Cooperative Human Tissue Network of the National Cancer Institute for providing samples, and Sabrina Noyes for administrative support.
↵8 D. Petillo, E.J. Kort, B. Teh. MicroRNA profiling of human kidney cancer subtypes, submitted for publication.
- Received February 18, 2008.
- Revision received April 4, 2008.
- Accepted April 7, 2008.
- ©2008 American Association for Cancer Research.