Ronglai Shen

EZH2 is a marker of aggressive breast cancer and promotes neoplastic transformation of breast epithelial cells

The Polycomb Group Protein EZH2 is a transcriptional repressor involved in controlling cellular memory and has been linked to aggressive prostate cancer. Here we investigate the functional role of EZH2 in cancer cell invasion and breast cancer progression. EZH2 transcript and protein were consistently elevated in invasive breast carcinoma compared with normal breast epithelia. Tissue microarray analysis, which included 917 samples from 280 patients, demonstrated that EZH2 protein levels were strongly associated with breast cancer aggressiveness. Overexpression of EZH2 in immortalized human mammary epithelial cell lines promotes anchorage-independent growth and cell invasion. EZH2-mediated cell invasion required an intact SET domain and histone deacetylase activity. This study provides compelling evidence for a functional link between dysregulated cellular memory, transcriptional repression, and neoplastic transformation.

Androgen-Independent Prostate Cancer Is a Heterogeneous Group of Diseases Lessons from a Rapid Autopsy Program

Understanding the biology of prostate cancer metastasis has been limited by the lack of tissue for study. We studied the clinical data, distribution of prostate cancer involvement, morphology, immunophenotypes, and gene expression from 30 rapid autopsies of men who died of hormone-refractory prostate cancer. A tissue microarray was constructed and quantitatively evaluated for expression of prostate-specific antigen, androgen receptor, chromogranin, synaptophysin, MIB-1, and α-methylacylCoA-racemase markers. Hierarchical clustering of 16 rapid autopsy tumor samples was performed to evaluate the cDNA expression pattern associated with the morphology. Comparisons were made between patients as well as within the same patient. Metastatic hormone-refractory prostate cancer has a heterogeneous morphology, immunophenotype, and genotype, demonstrating that “metastatic disease” is a group of diseases even within the same patient. An appreciation of this heterogeneity is critical to evaluating diagnostic and prognostic biomarkers as well as to designing therapeutic targets for advanced disease.

Comprehensive assessment of TMPRSS2 and ETS family gene aberrations in clinically localized prostate cancer

Novel recurrent gene fusions between the androgen-regulated gene TMPRSS2 and the ETS family members ERG, ETV1, or ETV4 have been recently identified as a common molecular event in prostate cancer development. We comprehensively analyzed the frequency and risk of disease progression for the TMPRSS2 and ETS family genes rearrangements in a cohort of 96 American men surgically treated for clinically localized prostate cancer. Using three break apart (TMPRSS2, ERG, ETV4) and one fusion (TMPRSS:ETV1) fluorescence in situ hybridization (FISH) assays, we identified rearrangements in TMPRSS2, ERG, ETV1, and ETV4 in 65, 55, 2, and 2% of cases, respectively. Overall, 54 and 2% of cases demonstrated TMPRSS2:ERG and TMPRSS2:ETV1 fusions, respectively. As intronic loss of genomic DNA between TMPRSS2 and ERG has been identified as a mechanism of TMPRSS2:ERG fusion, our assays allowed us to detect deletion of the 3′ end of TMPRSS2 and the 5′ end of ERG in 41 and 39% of cases rearranged for respective genes. Prostate cancers demonstrating TMPRSS2 gene rearrangement were associated with high pathologic stage (P=0.04). Our results confirm that recurrent chromosomal aberrations in TMPRSS2 and/or ETS family members are found in about 70% of prostate cancers. Importantly, we define a novel approach to study these gene fusions and identified cases where TMPRSS2 was rearranged without rearrangement of ERG, ETV1 or ETV4 and cases with ETS family gene rearrangement without TMPRSS2 rearrangement, suggesting that novel 5′ and 3′ partners may be involved in gene fusions in prostate cancer.

Identification of GATA3 as a Breast Cancer Prognostic Marker by Global Gene Expression Meta-analysis

GATA binding protein 3 (GATA3) is a transcriptional activator highly expressed by the luminal epithelial cells in the breast. Here we did a meta-analysis of the available breast cancer cDNA data sets on a cohort of 305 patients and found that GATA3 was one of the top genes with low expression in invasive carcinomas with poor clinical outcome. To validate its prognostic utility, we did a tissue microarray analysis on a cohort of 139 consecutive invasive carcinomas (n = 417 tissue samples) immunostained with a monoclonal antibody against GATA3. Low GATA3 expression was associated with higher histologic grade (P < 0.001), positive nodes (P = 0.002), larger tumor size (P = 0.03), negative estrogen receptor and progesterone receptor (P < 0.001 for both), and HER2-neu overexpression (P = 0.03). Patients whose tumors expressed low GATA3 had significantly shorter overall and disease-free survival when compared with those whose tumors had high GATA3 levels. The hazard ratio of metastasis or recurrence according to the GATA3 status was 0.31 (95% confidence interval, 0.13-0.74; P = 0.009). Cox multivariate analysis showed that GATA3 had independent prognostic significance above and beyond conventional variables. Our data suggest that immunohistochemical analysis of GATA3 may be the basis for a new clinically applicable test to predict tumor recurrence early in the progression of breast cancer.

Heterogeneity of TMPRSS2 Gene Rearrangements in Multifocal Prostate Adenocarcinoma: Molecular Evidence for an Independent Group of Diseases

Recurrent gene fusions between the androgen-regulated gene TMPRSS2 and the ETS family transcription factors ERG, ETV1, and ETV4 have been identified in the majority of prostate adenocarcinomas (PCA). PCA is often multifocal with histologic heterogeneity of different tumor foci. As TMPRSS2 is a common 5 partner of ETS gene fusions, we monitored TMPRSS2 rearrangement by fluorescence in situ hybridization (FISH) to study the origin and molecular basis of multifocal PCA heterogeneity. TMPRSS2 rearrangement was evaluated by FISH on a tissue microarray representing 93 multifocal PCAs from 43 radical prostatectomy resections. Overall, 70% (30 of 43) of the cases showed TMPRSS2 rearrangement, including 63% through deletion (loss of the 3 TMPRSS2 signal), 27% through translocation (split of 5 and 3 TMPRSS2 signals), and 10% through both mechanisms in different tumor foci. Of the 30 TMPRSS2 rearranged cases, 30% showed concordance in all tumor foci, whereas 70% were discordant in at least one focus. In TMPRSS2 rearranged cases, the largest (index) tumor was rearranged 83% of the time. Pathologic stage, size, or Gleason grade of the multifocal PCA did not correlate with overall TMPRSS2 rearrangement. Our results suggest that multifocal PCA is a heterogeneous group of diseases arising from multiple, independent clonal expansions. Understanding this molecular heterogeneity is critical to the future development and utility of diagnostic and prognostic PCA biomarkers.

Changes in differential gene expression because of warm ischemia time of radical prostatectomy specimens.

The expression of thousands of genes can be monitored simultaneously using cDNA microarray technology. This technology is being used to understand the complexity of human disease. One significant technical concern regards potential alterations in gene expression because of the effect of tissue ischemia. This study evaluates the increase in the differential gene expression because of tissue processing time. To evaluate differential gene expression because of ischemia time, prostate samples were divided into five time points (0, 0.5, 1, 3, and 5 hours). Each time point consisted of a homogeneous mixture of 12 to 15 prostate tissue cubes (5 mm(3)). These tissues were maintained at room temperature until at the assigned time point the tissue was placed in OCT, flash frozen in liquid nitrogen, and stored at -80 degrees C until RNA extraction. RNA from each time point was hybridized against an aliquot of 0 time point RNA from the same prostate. Four prostate glands were used in parallel studies. M-A plots were graphed to compare variability between time point sample hybridizations. Statistical Analysis of Microarray software was used to identify genes overexpressed at the 1-hour time point versus the 0-hour time with statistically significance. Microarray analysis revealed only a small percentage of genes (<0.6%) from more than 9000 to demonstrate overexpression at the 1-hour time point. Among the 41 statistically significant named overexpressed genes at the 1-hour time point were early growth response 1 (EGR1), jun B proto-oncogene (jun B), jun D proto-oncogene (jun D), and activating transcription factor 3 (ATF3). Genes previously associated with prostate cancer did not have significantly altered expression with ischemia time. Increased EGR1 protein expression was confirmed by Western blot analysis. Microarray technology has opened the possibility of evaluating the expression of a multitude of genes simultaneously, however, the interpretation of this complex data needs to be assessed circumspectly using refined statistical methods. Because RNA expression represents the tissue response to insults such as ischemia, and is also sensitive to degradation, investigators need be mindful of confounding artifacts secondary to tissue processing. All attempts should be made to process tissue rapidly to ensure that the microarray gene profile accurately represents the state of the cells and confirmatory studies should be performed using alternative methods (eg, Northern blot analysis, Western blot, immunohistochemistry).

Prognostic meta-signature of breast cancer developed by two-stage mixture modeling of microarray data

BackgroundAn increasing number of studies have profiled tumor specimens using distinct microarray platforms and analysis techniques. With the accumulating amount of microarray data, one of the most intriguing yet challenging tasks is to develop robust statistical models to integrate the findings.ResultsBy applying a two-stage Bayesian mixture modeling strategy, we were able to assimilate and analyze four independent microarray studies to derive an inter-study validated meta-signature associated with breast cancer prognosis. Combining multiple studies (n = 305 samples) on a common probability scale, we developed a 90-gene meta-signature, which strongly associated with survival in breast cancer patients. Given the set of independent studies using different microarray platforms which included spotted cDNAs, Affymetrix GeneChip, and inkjet oligonucleotides, the individually identified classifiers yielded gene sets predictive of survival in each study cohort. The study-specific gene signatures, however, had minimal overlap with each other, and performed poorly in pairwise cross-validation. The meta-signature, on the other hand, accommodated such heterogeneity and achieved comparable or better prognostic performance when compared with the individual signatures. Further by comparing to a global standardization method, the mixture model based data transformation demonstrated superior properties for data integration and provided solid basis for building classifiers at the second stage. Functional annotation revealed that genes involved in cell cycle and signal transduction activities were over-represented in the meta-signature.ConclusionThe mixture modeling approach unifies disparate gene expression data on a common probability scale allowing for robust, inter-study validated prognostic signatures to be obtained. With the emerging utility of microarrays for cancer prognosis, it will be important to establish paradigms to meta-analyze disparate gene expression data for prognostic signatures of potential clinical use.

Basal Cell Cocktail (34βE12 + p63) Improves the Detection of Prostate Basal Cells

Antibodies against high molecular weight cytokeratin (34&bgr;E12) and p63 are frequently used basal cell markers to aid in the diagnosis of prostate cancer (Pca). Absence of a basal cell marker in an atypical lesion histologically suspicious for cancer supports a diagnosis of Pca. However, absence of basal cells demonstrable by basal cell immunohistochemistry (IHC) is not always conclusive for PCa. Some benign prostatic lesions may have inconspicuous or even lack basal cell lining focally. Technical factors such as tissue fixation and antigen retrieval techniques may also make the detection of basal cells difficult. Improving the sensitivity of current basal cell markers is critical if these tests are being used to help make diagnostic decisions in conjunction with standard histology. In this study, we test the hypothesis that that inclusion of both 34&bgr;E12 and p63 in the same IHC reaction (basal cell cocktail) is advantageous over either marker used alone. One thousand three hundred fifty glands from 9 trans-urethral resectioned of prostate specimens with benign prostatic hypertrophy were used to study the immunostaining intensity and pattern for 34&bgr;E12, p63, and the basal cell cocktail. Basal cell marker expression was scored as strong, moderate, weak, or negative. Basal cell staining was considered complete if 75% of the glands circumference was positive for the basal cell marker and partial if <25% of the circumference was stained. The mean staining intensity and variance were calculated for 34&bgr;E12, p63, and the basal cell cocktail. A paired t test was used to evaluate whether the overall basal cell staining was significantly different between 34&bgr;E12, p63, and the basal cell cocktail. F-test was used to assess the variances for 34&bgr;E12, p63, and the basal cell cocktail. A high-density tissue microarray (TMA) comprising prostate tissue from 103 tumors from men with clinically localized Pca and a separate TMA comprising metastatic hormone-refractory Pca samples from 23 rapid autopsy cases were used to study the aberrant expression of 34&bgr;E12 and p63 in clinically localized and poorly differentiated Pca. The prostate glands in transition zone have variable basal cell staining intensity and pattern with 34&bgr;E12, p63, or the cocktail. Histologically, benign glands lack basal cell lining in 2%, 6%, and 2% of glands with cocktail, 34&bgr;E12, and p63 staining, respectively. The staining variance for the cocktail is significantly smaller than that for 34&bgr;E12 (0.0100 vs 0.1559, p = 0.0008). It is also smaller than that for p63, although a statistical significance has not been reached (0.0100 vs 0.0345, p = 0.099). The basal cell cocktail stains the basal cell layers more intensely than either 34&bgr;E12 or p63 alone, with complete and partial strong basal cell staining in 93% and 1% of benign glands, compared with 55% and 4% with 34&bgr;E12 and 81% and 1% with p63. Complete and partial weak staining is seen in 0% and 0% of benign glands with basal cell cocktail, compared with 8% and 7% with 34&bgr;E12 and 4% and 1% with p63 (p = 0.007 and 0.014 for cocktail vs 34&bgr;E12 and cocktail vs p63, respectively). A total of 2.8% clinically localized Pca had positive 34&bgr;E12 staining and 0.3% had positive p63 staining. Five (22%) of the metastatic Pca is positive for 34&bgr;E12. However, none had p63 expression. The basal cell cocktail had a staining pattern identical to that of 34&bgr;E12. IHC of the prostatic glands from the transition zone is subjected to staining variability that results in frequent variable and occasional negative basal cell staining in histologically benign glands; 34&bgr;E12 is most susceptible, and basal cell cocktail is least susceptible to such variability. Basal cell cocktail not only increases the sensitivity of the basal cell detection, but also reduces the staining variability and therefore renders the basal cell immunostaining more consistent. We recommend this basal cell cocktail for routine Pca diagnostic work-up.

Eigengene-based linear discriminant model for tumor classification using gene expression microarray data

MOTIVATIONnThe nearest shrunken centroids classifier has become a popular algorithm in tumor classification problems using gene expression microarray data. Feature selection is an embedded part of the method to select top-ranking genes based on a univariate distance statistic calculated for each gene individually. The univariate statistics summarize gene expression profiles outside of the gene co-regulation network context, leading to redundant information being included in the selection procedure.nnnRESULTSnWe propose an Eigengene-based Linear Discriminant Analysis (ELDA) to address gene selection in a multivariate framework. The algorithm uses a modified rotated Spectral Decomposition (SpD) technique to select hub genes that associate with the most important eigenvectors. Using three benchmark cancer microarray datasets, we show that ELDA selects the most characteristic genes, leading to substantially smaller classifiers than the univariate feature selection based analogues. The resulting de-correlated expression profiles make the gene-wise independence assumption more realistic and applicable for the shrunken centroids classifier and other diagonal linear discriminant type of models. Our algorithm further incorporates a misclassification cost matrix, allowing differential penalization of one type of error over another. In the breast cancer data, we show false negative prognosis can be controlled via a cost-adjusted discriminant function.nnnAVAILABILITYnR code for the ELDA algorithm is available from author upon request.

A Latent Variable Approach for Meta-Analysis of Gene Expression Data from Multiple Microarray Experiments

BackgroundWith the explosion in data generated using microarray technology by different investigators working on similar experiments, it is of interest to combine results across multiple studies.ResultsIn this article, we describe a general probabilistic framework for combining high-throughput genomic data from several related microarray experiments using mixture models. A key feature of the model is the use of latent variables that represent quantities that can be combined across diverse platforms. We consider two methods for estimation of an index termed the probability of expression (POE). The first, reported in previous work by the authors, involves Markov Chain Monte Carlo (MCMC) techniques. The second method is a faster algorithm based on the expectation-maximization (EM) algorithm. The methods are illustrated with application to a meta-analysis of datasets for metastatic cancer.ConclusionThe statistical methods described in the paper are available as an R package, metaArray 1.8.1, which is at Bioconductor, whose URL is http://www.bioconductor.org/.