Gene expression patterns in renal cell carcinoma assessed by complementary DNA microarray

Abstract

Renal cell carcinoma comprises several histological types with different clinical behavior. Accurate pathological characterization is important in the clinical management of these tumors. We describe gene expression profiles in 41 renal tumors determined by using DNA microarrays containing 22,648 unique cDNAs representing 17,083 different UniGene Clusters, including 7230 characterized human genes. Differences in the patterns of gene expression among the different tumor types were readily apparent; hierarchical cluster analysis of the tumor samples segregated histologically distinct tumor types solely based on their gene expression patterns. Conventional renal cell carcinomas with clear cells showed a highly distinctive pattern of gene expression. Papillary carcinomas formed a tightly clustered group, as did tumors arising from the distal nephron and the normal kidney samples. Surprisingly, conventional renal cell carcinomas with granular cytoplasm were heterogeneous, and did not resemble any of the conventional carcinomas with clear cytoplasm in their pattern of gene expression. Characterization of renal cell carcinomas based on gene expression patterns provides a revised classification of these tumors and has the potential to supply significant biological and clinical insights.

Figure 1.

Hierarchical clustering analysis for all genes across all arrays. A representation of the variation in expression of 1550 genes in the 38 tumors and three normal samples is shown at the left. The right panel depicts the detailed sample dendrogram in which three primary tumors and their associated renal vein/vena cava thrombi are designated on the dendrogram by branches with filled circles at the ends. Each row represents a single gene. Each column represents a single sample. The hierarchical clustering algorithm organizes the experimental samples only on the basis of overall similarity in their gene expression. Green squares indicate transcript levels below the mean; black squares, transcript levels equal to the mean; red squares, transcript levels greater than the mean; gray squares, technically inadequate or missing data. Gene filtering criteria were for fourfold variation from the normalized mean and at least 80% well measured spots. Colored bars adjacent to the clustering table indicate the position of the enlarged images. The color of the bar indicates the type of sample in which the cluster of genes is expressed. Green, papillary gene cluster; blue, normal gene cluster; orange, chromophobe-oncocytoma gene cluster; red, conventional clear cell carcinoma gene cluster. Selected gene names are shown.

Figure 2.

Immunohistochemical analysis of CK7, CD10, and vimentin and correlated mRNA expression data. mRNA expression data for the keratin 7, CD10, and vimentin genes, displayed using the same color key used in Figure 1 ▶ . Immunohistochemical stains for the proteins encoded by these genes are shown for selected tumors below the corresponding mRNA expression results.