| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Imaging, Diagnosis, Prognosis |
Authors' Affiliations: Departments of 1 Laboratory Medicine and Pathology, 2 Health Sciences Research, and 3 Urology, Mayo Clinic, Rochester, Minnesota, and 4 Department of Health Sciences Research, Mayo Clinic, Jacksonville, Florida
Requests for reprints: George Vasmatzis or John C. Cheville, Department of Laboratory Medicine and Pathology, 200 First Street SW, Rochester, MN 55905. Phone: 507-266-4617 and 507-284-3867; E-mail: vasmatzis.george{at}mayo.edu.
 |
Abstract |
|---|
 Top Abstract Materials and MethodsResultsDiscussionReferences |
|---|
Experimental Design: Candidate genes were discovered using Human Genome U133 Plus 2 Arrays and validated on independent samples by quantitative reverse transcription-PCR (RT-PCR). Both the discovery and the validation cohorts included nonaggressive primary CCRCC, aggressive primary CCRCC, metastatic CCRCC, and nonneoplastic kidney adjacent to tumor.
Results: Aggressive primary and metastatic CCRCC displayed no significant differences in gene expression. In contrast, we identified significant differences in gene expression between nonaggressive and aggressive CCRCC (including metastatic CCRCC). Thirty-four of the 35 transcripts that displayed the most significant differential expression by microarray analysis also displayed significant differential expression in independent validation studies using quantitative RT-PCR (P < 0.001 for 31 candidates and P < 0.005 for the remaining three candidates). Hierarchical clustering of the quantitative RT-PCR data using our candidate markers accurately grouped 88% (23 of 26) of aggressive and metastatic CCRCC samples, 100% (14 of 14) of nonaggressive CCRCC samples, and 100% (15 of 15) of nonneoplastic samples into separate clusters. Finally, we evaluated the ability of protein expression levels of one of our candidate markers (survivin) to predict survival among a cohort of 183 CCRCC patients treated surgically at Mayo Clinic from 1990 to 1992. In multivariate analysis, expression of survivin (BIRC5) was inversely associated with cancer-specific survival (P = 0.017).
Conclusion: We used a combination of genomic profiling and validation by quantitative PCR to identify a panel of candidate biomarkers for determining CCRCC aggressiveness. Our data also indicate that the gene expression alterations that result in aggressive behavior and metastatic potential can be identified in the primary tumor.
The standard of care for patients presenting with localized CCRCC remains surgical excision. Unfortunately, whereas the majority of patients with localized CCRCC will be cured by surgery, 
30% will develop metastases and die following removal of a confined tumor. Indeed, a significant problem in the clinical management of patients presenting with localized CCRCC is the inability to determine tumor aggressiveness and accurately predict which patients are at greater risk of experiencing distant metastases following surgery. Recently, several groups of investigators have combined individual clinical and pathologic predictors of CCRCC aggressiveness into composite multivariate scoring systems that possess a high degree of prognostic ability (4–7). Whereas these scoring systems have improved our ability to predict CCRCC aggressiveness, they remain surrogate measures of the true underlying molecular alterations that are associated with CCRCC aggressiveness. This limitation underscores the need to identify molecular markers of CCRCC aggressiveness within the tumor that have the potential to not only improve outcome prediction but also inform on new targets for therapies to be used in combination with surgical excision. Unfortunately, to date there are no widely accepted molecular biomarkers for CCRCC aggressiveness.
In recent years, gene expression profiling has emerged as a means of rapidly identifying candidate genes and gene expression patterns that are associated with aggressive tumors (8, 9) including CCRCC (10, 11). To date however, comprehensive gene expression profiling studies of well-characterized CCRCCs with the express purpose of identifying markers of aggressiveness are lacking. Moreover, the studies that do exist often omit analyses of normal and metastatic tumor samples or rely on CCRCC cell lines rather than actual patient tumors (12). More importantly, findings from gene profiling studies of CCRCC aggressiveness are often not validated on an independent cohort of samples using more accurate means of evaluating gene expression (i.e., quantitative PCR).
The objective of this study was to first do microarray analysis on a group of aggressive and nonaggressive CCRCC tumors to identify candidate markers of aggressiveness and then validate these candidates using quantitative PCR on an independent set of tumors. By comparing gene expression profiles of nonaggressive CCRCC and aggressive CCRCC, a panel of biomarkers with potential prognostic significance was identified. The potential prognostic significance of the set of genes identified by our microarray study was validated using quantitative reverse transcription-PCR (RT-PCR) on an independent cohort of nonneoplastic kidney and well-characterized CCRCC samples. Finally, the prognostic significance of one of these biomarkers was assessed at the protein level using immunoperoxidase staining methods.
|
Materials and Methods |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
3 and consequently had a predicted 5-year cancer-specific survival in excess of 88% and also had a predicted 5-year metastasis-free survival in excess of 84%. In fact, all but one of the patients with nonaggressive CCRCC had expected survivals in excess of 95%, and none of these patients have developed metastases or died of tumor (Table 1).
|
Before all experiments, H&E-stained sections from frozen tissue blocks were reviewed by a urologic pathologist (J.C.C.) to insure appropriate tissue diagnosis as well as quality and quantity of the tumor samples. Frozen tissue sections were also reviewed for pathologic features predictive of tumor aggressive behavior (nuclear grade and necrosis). Tumor blocks were selected to insure that aggressive CCRCC samples were predominantly high grade (nuclear grades 3 and 4) and nonaggressive CCRCC were all low grade (nuclear grades 1 and 2). In tumor blocks, all nonneoplastic tissue was removed from the frozen block before any analyses.
Oligonucleotide microarray experiments. Microarray experiments were done by the Mayo Clinic Microarray Core Facility; 30 mm3 of each tissue were sectioned at 20 or 35 µm, collected in buffer RLT (Qiagen, Valencia, CA) supplemented with ß-mercaptoethanol and homogenized using a PT 1200C (Kinematica AG, Luzerne, Switzerland) rotor/stator homogenizer. Total RNA was isolated using the RNeasy kit (Qiagen) following manufacturer's specifications. Quality and quantity of RNA samples were analyzed by spectrophotometry and Agilent 2100 Bioanalyzer (Agilant Technologies, Palo Alto, CA). Synthesis of biotin-labeled cRNA, hybridization, washing, staining, and scanning were done following manufacture's protocols (Affymetrix Corp., Santa Clara, CA). Microarray experiments were carried out using the Human Genome U133 Plus 2 Arrays.
Microarray data analysis. Affymetrix microarray analysis software GCOS was used to process scanned chip images. The software generates a cell intensity file for each chip, which contains a single intensity value for each probe cell (.CEL file). dChip 1.3 (http://biosun1.harvard.edu/complab/dchip/) was used to calculate model-based expression index after data from all chips were normalized against an array with median overall intensity using invariant set method (15). The model-based expression index was calculated using perfect match/mismatch model with outlier detection and correction, and the calculated expression values were log2 transformed.
To identify differentially expressed genes in aggressive and nonaggressive cases, three algorithms were used. First, using the dChip program, probe sets with a difference of >2.0 on the log scale (>4.0-fold change) between the average expression levels of the aggressive and nonaggressive cases and a P < 0.001 were identified (129 genes, list 1). To estimate the number of false positives in this list, the cases were randomly assigned to two groups 1,000 times and the same criteria were applied to identify differentially expressed genes. The median false discovery rate (FDR) by this process was 0.8% (one gene) and a 90th percentile of 3.9% (five genes). Second, expression values of probe sets (11,715) determined by the dChip program to be most variable across the aggressive and nonaggressive cases were exported to GeneCluster 2.0 (Whitehead/MIT for Genome Research) to identify 120 probe sets with highest signal to noise ratios (list 2). The signal to noise ratio estimate, also called the discriminate score (10), is computed as SNR = (µ1 – µ2) / (
1 + 2), where µ and refer to the mean and SD, respectively. A high SNR typically suggests that the expression level of a gene displays a much larger variation between the two groups compared with the variation within each group. Finally, probe set expression levels (54,605) from dChip were imported to the prediction analysis of microarray (PAM) algorithm to identify 117 genes that best distinguish aggressive and nonaggressive cases (list 3). PAM uses the "shrunken centroid" approach to reduce the effects of "noisy" genes (16). The threshold for shrinking the centroids was set at 3.5, producing an overall misclassification error of 11% (3 of 28). Probe sets common to the three lists were identified. From this list, candidates with >30% absent calls in the group determined to overexpress the gene were discarded. Finally, the redundant probe sets representing a gene were removed. The final list (Table 2) included 35 probe sets. We used this list for supervised clustering in the dChip program using the centroid linkage method and Euclidean distance metric (Fig. 2).
|
|
Quantitative RT-PCR experiments were done on an ABI 7900 HT system (Applied Biosystems, Foster City, CA). For each primer set, the optimum primer concentration (typically 0.15 µmol/L final concentration) was determined and standard curves were generated using four to five dilutions of a pooled cDNA sample from the validation cohort. Typical standard curve included 4, 1, 0.25, 0.0625, 0.0156, and 0 ng (no template control) of total RNA equivalents of cDNA. To confirm that the amplification occurred on the target sequences, the amplicons were analyzed by gel electrophoresis and the dissociation curves were examined for the presence of a single sharp peak at the melting temperature of the amplicon. The expression level of each gene was normalized by karyopherin 
6 (KPNA6) as 
CT = CT-KPNA6 – CT-gene, where CT is the threshold cycle in the quantitative PCR experiment (see Results). To select the most significantly differentially expressed genes (Fig. 4), we used the z score from the Mann-Whitney test (http://faculty.vassar.edu/lowry/utest.html).
|
|
Five-micrometer sections were prepared from formalin-fixed paraffin-embedded tissue blocks containing the highest grade. Immunoperoxidase staining with anti-survivin antibody (DAKO, Carpinteria, CA; 1:100 dilution) was done using standard techniques. Slides were reviewed (J.C.C.), and the area of greatest staining was circled (diameter, 1-2 mm) for analysis. Each slide was then scanned using the Bacus Slide Scanner (Bacus Laboratories, Inc., Lombard, IL). The BLISS system (Bremson Company, Lenexa, KS) was used to digitally capture images at 480 x 752 pixel resolution and at 40x magnification. The entire slide is composed of multiple tiles to create a mosaic or composite picture. Computer assisted analysis was done by using the IHCScore software (Bacus Laboratories). This software allows for analysis measurements of (a) total area, (b) immunohistochemistry area, (c) total immunohistochemistry stain density, (d) average immunohistochemistry stain density, and (e) immunohistochemistry percentage area (referenced to the total area). For this study, the all filter was selected to capture the staining intensity of anti-survivin antibody. A series of images were viewed to subjectively create an immunohistochemistry and reference threshold. Each slide was segmented to capture the entire area of interest. Thresholds were adjusted for each slide and data were exported to Excel spreadsheet.
Cancer-specific survival was estimated using the Kaplan-Meier method. The duration of follow-up was calculated from the date of surgery to the date of death or last follow-up. Scatter plots of survivin expression versus the difference in observed survival and the survival expected form a Cox proportional hazards regression model (formally known as the Martingale residuals) were used to identify a potential cut point for survivin expression for illustration purposes. The associations of survivin expression with death from CCRCC were evaluated using Cox proportional hazards regression models univariately and after adjusting for pathologic features important in outcome prediction among patients with CCRCC including primary tumor size, TNM classification, nuclear grade, and the presence of tumor necrosis. Survival analyses were done using the SAS software package (SAS Institute, Cary, NC).
|
Results |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
|
Comparison of the expression profiles of the aggressive clear cell renal cell carcinoma primary tumors and the clear cell renal cell carcinoma metastatic cases. Expression profiles were examined to determine if they could discriminate between aggressive primary CCRCC and metastatic CCRCC. In the unsupervised clustering analysis, aggressive primary and metastatic CCRCC did not separate into unique clades (Fig. 1A). To insure that the clustering pattern was not influenced by the expression profiles of the nonneoplastic and nonaggressive CCRCC, expression profiles from the aggressive primary CCRCC and metastatic samples were analyzed separately. Here again, it was noted that the aggressive CCRCC cases were interspersed among the metastatic tumor samples (Fig. 1C). We compared the expression profiles of the two groups using the dChip and PAM algorithms. By the dChip algorithm, the number of differentially expressed genes between the two groups was comparable with the number of differentially expressed genes found by randomly assigning the metastatic samples and the aggressive primary CCRCC to two groups. The median FDR was 100% and the 90th percentile FDR was 300% to 400%, depending on the significance criteria. We used PAM to identify a group of genes that could be used for classification of aggressive primary and CCRCC metastasis cases (see Materials and Methods). The average misclassification error with any possible threshold for "shrinking centroids" was 40% to 60%. These analyses suggested that no set of genes could separate the metastatic and aggressive primary CCRCC into two distinct groups.
Comparison of expression profiles of nonaggressive clear cell renal cell carcinoma against aggressive and metastatic clear cell renal cell carcinoma. Because the aggressive primary and metastatic CCRCC samples showed similar expression profiles, they were grouped together and compared with nonaggressive CCRCC. This increased the statistical power for identification of significantly differentially expressed genes.
To identify probe sets that are most relevant to CCRCC behavior, we used the signal to noise selection criteria and the PAM algorithm in addition to the fold change and P value criteria provided by the dChip software (see Materials and Methods). In each case, comparisons were made for the gene expression values in the primary tumors from nonaggressive CCRCC versus aggressive primary tumors and metastatic CCRCC samples. The top 120 to 130 candidate prognostic biomarkers were selected using the three statistical algorithms. We identified 129 probe sets that displayed a fold change of at least 4.0 and P < 0.001 (median FDR = 0.8% and 90 percentile FDR = 3.9%) by dChip, 125 probe sets with highest signal to noise ratio by GeneCluster, and 120 probe sets by PAM after the centroids were "shrunken" by a factor of 3.75. Probe sets common in the three lists that also had a present (P) call by the dChip algorithm in at least 70% of the cases were selected. Finally, multiple probe sets representing the same gene were discarded so that the listing would represent unique individual genes. The final candidate list included 35 probe sets corresponding to 35 unique transcripts (Table 2). The majority of the candidate biomarkers identified by our analysis (29 of 35, 83%) displayed down regulation of expression in the aggressive CCRCC compared with the nonaggressive CCRCC.
With this set of 35 differentially expressed genes, hierarchical clustering of the 28 CCRCC tissues was repeated. This analysis produced two major subgroups. One subgroup contained 17 of 18 (94%) of the aggressive CCRCC and metastatic CCRCC samples. The other subgroup included all 10 (100%) tissues from the nonaggressive CCRCCs and the remaining case of aggressive CCRCC (Fig. 2).
Validation by quantitative reverse transcription-PCR. The results from the microarray experiments were validated by determining the expression levels of the 35 candidate biomarkers in an independent cohort of CCRCC and nonneoplastic samples using quantitative RT-PCR. Compared with the microarray technology, quantitative RT-PCR provides a much wider dynamic range (5-6 orders of magnitude) and thus a more accurate measurement of the relative gene expression values.
We first used the microarray data to identify genes that could be used for normalization of expression levels determined by quantitative PCR. Two genes, eukaryotic translation elongation factor 1 1 (EEF1A1) and karyopherin 6 (KPNA6), were selected from the five genes with the lowest SDs across all samples in our microarray expression data. In addition, two common housekeeping genes, ß-2-microglobin (B2M) and glyceraldehyde 3-phosphate dehydrogenase (GAPDH), were examined. The expression levels of all four genes were measured by quantitative RT-PCR (Fig. 3). As expected from the microarray data, expression of GAPDH was most variable across the samples followed by B2M. On the other hand, KPNA6 expression was least variable across all samples. More importantly, expression levels of GAPDH and B2M were lower in nonneoplastic kidney than in the RCC cases (P < 0.001 for both genes). In contrast, KPNA6 expression was not statistically different among the CCRCC and nonneoplastic tissues. Furthermore, expression levels of KPNA6 were comparable with those of most of the candidate biomarkers and on average 10- to 20-fold (approximately four cycles in a quantitative RT-PCR experiment) lower than the expression levels of GAPDH. Thus, KPNA6 was selected for normalization of the quantitative RT-PCR data.
|
2-fold difference in expression. For ECRG4, for example, we detected a difference of more than four cycles in the mean expression levels between nonaggressive CCRCC and aggressive CCRCC, indicating an 16-fold difference in expression levels between the two groups. Hierarchical clustering of the quantitative RT-PCR data confirmed that the 35 genes selected from the microarrays had prognostic significance for CCRCC. As Fig. 5 shows, two main subgroups were identified in the validation cohort. One subgroup included 23 of the 26 aggressive and metastatic CCRCC cases (88%) and none of the nonaggressive CCRCC cases. The other main subgroup included two further clusters, one containing all 15 (100%) of the nonneoplastic tissues and the other containing all 14 (100%) of the nonaggressive cases and the remaining three cases of aggressive primaries.
Survivin immunoperoxidase staining and analysis. At last follow-up, 113 of the 183 patients with adequate tissue specimens available for immunoperoxidase analysis had died, including 63 patients who died of CCRCC at a median 2.4 years following surgery (range, 0-13). Among the 70 patients who were still alive, the median duration of follow-up was 12.5 years (range, 1-15). The median survivin expression was 0.87 (mean, 2.37; range, 0.03-35.80). The staining was predominantly nuclear (Fig. 6A and B). A scatterplot of survivin expression versus the expected risk of death for each patient suggested that a cut point value of 2 would be most appropriate (17). There were 58 (31.7%) patients with survivin expression levels of 
2. Univariately, the risk ratio for a 1-unit increase in survivin expression was 1.14 (95% confidence interval, 1.10-1.19; P < 0.001). In our categorical analysis, patients with survivin expression levels of 2 were over four times more likely to die from CCRCC compared with patients with survivin expression of <2 (risk ratio, 4.27; 95% confidence interval, 2.59-7.05; P < 0.001). The association between survivin expression levels and cancer-specific survival is shown in Fig. 7. The estimated cancer-specific survival rates at 1, 5, and 10 years following surgery were 96.7%, 85.3%, and 75.4%, respectively, for patients with survivin expression levels of <2 compared with 71.2%, 47.6%, and 38.0%, respectively, for patients with survivin expression of 2. After multivariate adjustment for primary tumor size, TNM classification, nuclear grade and tumor necrosis, the risk ratio for a 1-unit increase in survivin expression was 1.10 (95% confidence interval, 1.04-1.16; P < 0.001), whereas the risk ratio for survivin expression of 2 was 2.07 (95% confidence interval, 1.14-3.76; P = 0.017).
|
|
|
Discussion |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
The pursuit to accurately predict outcome among patients with primary CCRCC as well as metastatic CCRCC has resulted in several recently published scoring systems that incorporate clinical and pathologic features to determine prognosis (6, 7, 13, 18). These prediction models have improved our ability to predict outcome in patients with CCRCC, but the incorporated clinical and pathologic features are surrogate measures of the molecular alterations that determine CCRCC aggressiveness. As such, they do not provide insight regarding the biology of CCRCC nor do they provide potential targets for adjuvant therapies. In this study, the use of CCRCC primary tumors as well as metastatic CCRCC samples allowed for the identification of an expression profile indicative of tumor aggressiveness. In addition, we analyzed one of these biomarkers, survivin, using a immunoperoxidase assay, and found that overexpression of the survivin protein was significantly associated with decreased cancer-specific survival in a multivariate model. Survivin is a member of the inhibitor of apoptosis protein family, and its expression both at the mRNA and protein level has been associated with more aggressive tumor behavior in carcinomas of the larynx, liver, prostate, lung, ovary, and stomach (19–22). This is the first study to show that survivin protein overexpression is predictive of outcome in patients with CCRCC. Interestingly, it has been shown that inhibition of survivin activity and down-regulation of nuclear factor-
B, both mediated by VHL protein, are critical to conferring sensitivity of RCC cells to tumor necrosis factor- cytotoxicity (23). This suggests that survivin has potential not only as a prognostic biomarker but also as a therapeutic target, although additional confirmatory studies are needed.
A number of other genes that showed increased expression in aggressive as compared with nonaggressive CCRCC are noteworthy. IL-8, a potent chemotactic cytokine for inflammatory cells, exhibited higher expression levels in the aggressive compared with the nonaggressive CCRCCs in our microarray experiments. IL-8 is implicated in the migration of lymphocytes into tumors through an -1 integrin-mediated pathway in the extracellular matrix, and studies show that neutralizing antisera specific to IL-8 inhibit tumor-infiltrating lymphocyte migration (24). Although our microarrray experiments identified significantly increased expression of IL-8 in aggressive primary and metastatic CCRCC, this differential expression was marginally significant (P < 0.055) by our quantitative RT-PCR experiments. Another gene, serum amyloid A, has been identified in the serum of CCRCC patients, and elevated serum levels are associated with aggressive CCRCC (25). Serum amyloid A1 and A2 are acute-phase reactants whose expression is regulated in part by IL-1 and IL-6 (26–28). Serum amyloid A can be induced in renal tubular epithelial cells, but previous to this study, serum amyloid A mRNA had not been associated with CCRCC aggressive behavior. CKS2, determined to be up-regulated in aggressive CCRCC, has been associated with cancer (up-regulated in metastatic colon cancer; ref. 29), but its function and significance in CCRCC will require further study. Finally, HSP150 was found to be up-regulated in aggressive CCRCC. This gene has not been associated with CCRCC or any cancer to the best of our knowledge. In contrast to a limited number of up-regulated genes in aggressive CCRCC, there were numerous genes that exhibited decreased mRNA levels relative to nonaggressive CCRCC. Several of these genes have been described previously, yet their functional role in CCRCC remains unknown. Esophageal cancer-related gene 4 (ECRG4) has been identified to be down-regulated in squamous cell carcinoma of the esophagus through hypermethylation of the CpG islands (30, 31). In this study, ECRG4 was identified as one of the most significantly down-regulated genes in aggressive tumors and we are investigating the influence of this gene in progression of CCRCC.
At the present time, there is no standard method for the analysis of microarray data. In this report, we used three algorithms that were relevant to our analysis and identified the best candidate biomarkers common to all three of the algorithms. The fact that all of the candidate biomarkers on the list were validated by the quantitative RT-PCR experiments suggests that our approach for the analysis of microarray data was justified. To identify genes for normalization of quantitative RT-PCR results, we searched our microarray data for transcripts that displayed minimum variation in normalized expression values across the samples. The two transcripts selected by this analysis, EEF1A1 and KPNA6, were confirmed by quantitative RT-PCR to have considerably less variation across the 55-sample validation cohort than the commonly used GAPDH and B2M. Furthermore, GAPDH and B2M had significantly higher expression levels in CCRCC samples than in nonneoplastic kidney. Increased expression of GAPDH mRNA in tumor samples is consistent with reports suggesting increased expression of GAPDH protein in renal cell carcinoma to meet the energy demands of the tumor cells following diminished oxidative phosphorylation in the mitochondria (32). Similarly, increased expression of B2M is consistent with reports indicating elevated levels of B2M protein in the serum of patients with RCC (33). Comparing the expression levels of EEF1A1 and KPNA6, we chose KPNA6 for normalization because the expression levels of KPNA6 across the validation samples were more comparable to the expression levels of the selected biomarkers.
Other investigators have examined gene expression in CCRCC using microarrays. Takahashi et al. examined gene expression profiling in 26 cases of CCRCC and reported distinctive gene expression profiles associated with tumor aggressiveness similar to our findings (10). In that study, 17 CCRCC cases were nonaggressive and nine were aggressive. In the gene expression analyses, tumors clustered into defined outcome groups based on 40 transcripts. However, the expression profile was not validated either on the original cohort or an independent set of tumors. Univariate analyses showed a significant correlation of expression profile with outcome as well as histologic tumor grade. However, the number of patients was limited for these types of analyses, and the selection criteria for inclusion into the study were not provided. It is difficult to compare our analyses to that of Takahashi et al. We employed the latest Affymetrix oligonucleotide arrays, which contain more genes than the cDNA arrays used by Takahashi et al. In addition, patient selection and criteria for determining differential gene expression differed between our study and that of Takahashi et al.
The selection of cases in our study warrants comment. CCRCC samples were selected based on a combination of outcome and pathologic features (Table 1). In cases of nonaggressive CCRCC with limited follow-up data, two prediction models were employed to insure that patients considered to have nonaggressive CCRCC had a predicted 5-year metastasis-free survival of at least 84% and 5-year cancer-specific survival of at least 88%. In addition, all frozen tissue blocks were reviewed to insure that nonaggressive tumors were all low grade (nuclear grades 1 and 2). In contrast, patients with CCRCC considered aggressive died of disease or developed metastases within 4 years of diagnosis. In addition, because of the grade heterogeneity found within many CCRCC, aggressive tumors used in the analysis were predominantly nuclear grades 3 and 4. It is possible that this selection process, which included pathologic features, improved our ability to identify significant differences in gene expression and confirms the importance of pathologic review of all tumors analyzed by microarrays and quantitative RT-PCR.
At least one of the transcripts in our list of differentially expressed genes, endomucin (EMCN), is believed to be associated with the nonepithelial renal components. Because bulk tissues were used in this study, we cannot determine if these genes are expressed in nontumor or tumor cells, although tissues were selected to be composed entirely of tumor with limited stromal and inflammatory components. Studies using laser capture microdissection, quantitative RT-PCR, and immunohistochemistry are under way to localize mRNA and protein expression patterns.
In conclusion, our experimental analyses identified a panel of potential biomarkers that are differentially expressed between the aggressive and the nonaggressive CCRCC. In addition, protein expression of one of these genes, survivin, was an independent predictor of patient outcome in a large cohort of patients with surgically treated CCRCC. Further studies are required to determine if expression of the other genes identified in this study provide prognostic information beyond that provided by routine pathologic examination and prognostic scoring systems and algorithms.
|
Footnotes |
|---|
Received 1/11/05; revised 4/ 5/05; accepted 4/27/05.
|
References |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
B-dependent antiapoptotic pathway. Cancer Res 2003;63:7076–80.1 integrin I-domain. Eur J Immunol 1998;28:2530–6.[CrossRef][Medline]
This article has been cited by other articles:
|
|
 |
|
  A. E. Krambeck, H. Dong, R. H. Thompson, S. M. Kuntz, C. M. Lohse, B. C. Leibovich, M. L. Blute, T. J. Sebo, J. C. Cheville, A. S. Parker, et al. Survivin and B7-H1 Are Collaborative Predictors of Survival and Represent Potential Therapeutic Targets for Patients with Renal Cell Carcinoma Clin. Cancer Res., March 15, 2007; 13(6): 1749 - 1756. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Jones and T. A. Libermann Genomics of Renal Cell Cancer: The Biology Behind and the Therapy Ahead Clin. Cancer Res., January 15, 2007; 13(2): 685s - 692s. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. S. Lam, A. J. Pantuck, A. S. Belldegrun, and R. A. Figlin Protein Expression Profiles in Renal Cell Carcinoma: Staging, Prognosis, and Patient Selection for Clinical Trials Clin. Cancer Res., January 15, 2007; 13(2): 703s - 708s. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. C. Hsu, J. Chang, T. Wang, E. Steingrimsson, M. K. Magnusson, and K. Bergsteinsdottir Statistically designing microarrays and microarray experiments to enhance sensitivity and specificity Brief Bioinform, January 1, 2007; 8(1): 22 - 31. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K.-i. Shioi, A. Komiya, K. Hattori, Y. Huang, F. Sano, T. Murakami, N. Nakaigawa, T. Kishida, Y. Kubota, Y. Nagashima, et al. Vascular Cell Adhesion Molecule 1 Predicts Cancer-Free Survival in Clear Cell Renal Carcinoma Patients Clin. Cancer Res., December 15, 2006; 12(24): 7339 - 7346. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. S. Lam, A. Breda, A. S. Belldegrun, and R. A. Figlin Evolving Principles of Surgical Management and Prognostic Factors for Outcome in Renal Cell Carcinoma J. Clin. Oncol., December 10, 2006; 24(35): 5565 - 5575. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Cancer Research | Clinical Cancer Research |
| Cancer Epidemiology Biomarkers & Prevention | Molecular Cancer Therapeutics |
| Molecular Cancer Research | Cancer Prevention Research |
| Cancer Prevention Journals Portal | Cancer Reviews Online |
| Annual Meeting Education Book | Cell Growth & Differentiation |
