1. ¹Laboratory of Cancer Genetics, Van Andel Research Institute, 333 Bostwick NE, Grand Rapids, MI 49503, USA
2. ²Department of Urology, School of Medicine, The University of Tokushima, Tokushima 770-8503, Japan
3. ³Department of Pathology, University of Chicago, 5841 South Maryland Avenue, Chicago, IL 60637, USA
4. ⁴Department of Surgery/Urology, University of Chicago, 5841 South Maryland Avenue, Chicago, IL 60637, USA
5. ⁵Department of Pathology, Spectrum Health Hospital, Michigan St, NE, Grand Rapids, MI 49503
6. ⁶Division of Urology, Spectrum Health Hospital, Michigan St NE, Grand Rapids, MI 49503, USA
7. ⁷Department of Urology, Princess Alexandra Hospital, Queensland, 4102 Australia
8. ⁸Department of Medicine, University of Chicago, 5841 South Maryland Avenue, Chicago IL 60637, USA
9. ⁹Bioinformatics Program, Van Andel Research Institute, 333 Bostwick NE, Grand Rapids, MI 49503, USA
10. ¹⁰Department of Urology, School of Medicine, The University of Tokushima, Tokushima 770-8503, Japan
11. ¹¹Department of Urology, School of Medicine, Iwate medical University, Morioka 020-8505, Japan

Correspondence: BT Teh, E-mail: bin.teh@vai.org

Received 19 February 2003; Revised 19 June 2003; Accepted 24 June 2003.

Abstract

We analysed the expression profiles of 70 kidney tumors of different histological subtypes to determine if these subgroups can be distinguished by their gene expression profiles, and to gain insights into the molecular mechanisms underlying each subtype. In all, 39 clear cell renal cell carcinomas (RCC), seven primary and one metastatic papillary RCC, six granular RCC from old classification, five chromophobe RCC, five sarcomatoid RCC, two oncocytomas, three transitional cell carcinomas (TCC) of the renal pelvis and five Wilms' tumors were compared with noncancerous kidney tissues using microarrays containing 19 968 cDNAs. Based on global gene clustering of 3560 selected cDNAs, we found distinct molecular signatures in clear cell, papillary, chromophobe RCC/oncocytoma, TCC and Wilms' subtypes. The close clustering in each of these subtypes points to different tumorigenic pathways as reflected by their histological characteristics. In the clear cell RCC clustering, two subgroups emerged that correlated with clinical outcomes, confirming the potential use of gene expression signatures as a predictor of survival. In the so-called granular cell RCC (terminology for a subtype that is no longer preferred), none of the six cases clusters together, supporting the current view that they do not represent a single entity. Blinded histological re-evaluation of four cases of 'granular RCC' led to their reassignment to other existing histological subtypes, each compatible with our molecular classification. Finally, we found gene sets specific to each subtype. In order to establish the use of some of these genes as novel subtype markers, we selected four genes and performed immunohistochemical analysis on 40 cases of primary kidney tumors. The results were consistent with the gene expression microarray data: glutathione S-transferase was highly expressed in clear cell RCC, methylacyl racemase in papillary RCC, carbonic anhydrase II in chromophobe RCC and K19 in TCC. In conclusion, we demonstrated that molecular profiles of kidney cancers closely correlated with their histological subtypes. We have also identified in these subtypes differentially expressed genes that could have important diagnostic and therapeutic implications.