Objectives: Renal epithelial neoplasms are classified histologically into subtypes with features of different normal nephron components. Accurate classification can be difficult by routine histopathology, but is important because subtypes are associated with distinct clinical findings and prognoses. We have developed a custom cDNA microarray system for renal tumor classification based on differential gene expression. We used the arrays to group tumors into clusters that correlate with histopathologic subtypes. Based on distinct expression patterns, we identified specific candidate markers for renal tumor diagnosis and validated them by quantitative RT-PCR. Methods: An initial cohort of 14 random frozen kidney tumors [7 conventional renal cell carcinomas (RCC), 3 papillary RCC, 2 chromophobe RCC, and 2 angiomyolipomas (AML)] was analyzed with cDNA microarrays featuring ∼1000 gene probes shown previously to be expressed differentially in renal neoplasia. Unsupervised clustering algorithms were used to correlate differential gene expression with histologic diagnosis. An independent cohort of formalin-fixed renal tumors was then analyzed by quantitative RT-PCR to validate differential expression of the following genes: megalin and cadherin 6 (proximal nephron antigens, and candidate markers for conventional renal cell carcinoma (RCC)); alpha-methylacyl CoA racemase (AMACR; candidate marker for papillary RCC); beta defensin 1 and chloride channel Kb (distal nephron antigens, and candidate markers for chromophobe RCC and oncocytoma). Results: Histopathologic subtypes of renal tumor were characterized by distinct gene expression profiles determined with cDNA microarrays. Consistent with micorarray findings, conventional RCC overexpressed megalin and cadherin 6; chromophobe RCC and oncocytoma overexpressed beta defensin-1 and chloride channel Kb; and papillary RCC overexpressed AMACR by quantitative RT-PCR. Papillary carcinomas also expressed markers of proximal and distal nephron. These differential expression patterns are consistent with current models of renal tumor histogenesis, which relate conventional RCC to proximal nephron epithelium and chromophobe RCC/oncocytoma to distal nephron intercalated cells. Conclusions: Renal epithelial tumors are characterized by unique gene expression profiles determined by cDNA microarray and quantitative RT-PCR. Expression profiles are consistent with biologic features of renal tumor subtypes. Molecular classification may yield new insights into renal tumor pathobiology and be useful clinically to enhance accuracy of pathologic renal tumor classification.

[Proc Amer Assoc Cancer Res, Volume 45, 2004]