Epithelial renal cell carcinoma (RCC) is the most common malignancy of the adult kidney. RCC is a clinicopathologically heterogeneous disease that is traditionally classified by morphology into clear cell, papillary, chromophobe, and collecting duct carcinoma. Chromophobe renal cell carcinoma (chRCC) and renal oncocytoma are two distinct but related entities, with strong morphologic and genetic similarities 1 . Distinguishing between the two tumors may present a significant diagnostic challenge on routine hematoxylin-eosin stained sections, especially in cases with features resembling both chRCC and oncocytoma, oncocytoma with associated invasion and even metastasis 2 , and the eosinophilic variant of chRCC.

ChRCCs account for about 4-8% of all renal tumors, with a more favorable prognosis relative to clear cell renal cell carcinoma, which comprises the majority of all RCCs 3 . On the other hand, oncocytoma is the most common benign renal tumor, comprising 5-8% of resected renal masses. The overlapping characteristics of these entities may be explained by a possible common origin from the intercalated cells of the distal tubule 4 . Patients with Birt-Hogg-Dubé syndrome, a familial multi-tumor syndrome linked to mutation of the BHD gene, exhibit bilateral oncocytomas, chRCC and hybrid tumors 5 6 .

In our previous gene expression profiling studies of a limited number of chRCC and oncocytoma 7 , we demonstrated that both tumors showed strong similarities in expression patterns suggesting a common underlying biology 8 and this was supported by subsequent expression profiling studies by other groups 9 . We hypothesized that more effective discrimination might be achieved with a larger sample number with additional analyses, and that the differences might shed light on the underlying genetic drivers of tumorigenesis, diagnosis and clinical management. We set out to perform a comprehensive characterization of both entities by integrating gene expression and high resolution single-nucleotide polymorphism (SNP) profiling, proceeding to identify a useful and valid molecular predictor, as well as identifying novel immunohistochemical markers for each entity.

ChRCC and oncocytoma are morphologic and genetically related entities, and distinction between these two tumors is important because of their different biological behaviors. However, these entities can be difficult to distinguish morphologically. We report the derivation of a novel and useful gene predictor validated both on an internal and an independent external data-set, implying its generalizability. Our results suggest that it is possible to classify accurately histopathologically challenging tumors. The degree of accuracy achieved at 93% is reasonable for a genetic classifier. However, integration into clinical practice requires a comprehensive evaluation of these classifiers within a clinical setting, comparing clinical outcomes in routine pathologic evaluation relative to that derived from novel classifiers. This may be most practically if not most ideally done in a retrospective fashion on paraffin-embedded tissue in a large multi-institutional collaboration, which we are currently pursuing. This issue may become progressively more important with the increase in incidentally detected small tumors on radiologic surveillance, where the dilemma between observation or intervention is commonly posed.

Integrating RNA and DNA genomic data allows us to verify genomic alterations in tumor samples and distinguish the genomic signatures of different tumor subtypes. Frequent losses of chromosome 1, 2, 6, 10, 13, 17, and 21 and gains in chromosome 4, 7, 11, 12, 14q and 18q were observed in chRCC, consistent with previously reported data 18 19 . For renal oncocytoma, we show a high prevalence of chromosome 1p loss. Both chromophobe RCC and oncocytoma share this chromosomal alteration, consistent with a speculation that this may represent an early event in neoplastic transformation of a common progenitor cell.

Chromosome 1p loss represents a common cytogenetic alteration in both chRCC and renal oncocytoma identified by high-throughput SNP studies. This may suggest that this is an early event in the histogenesis of both tumors, before additional cellular events lead to malignancy in lesions that progress to chRCC, similar to chromosome 3p loss in clear cell renal cell carcinoma, which is thought to be an early event in carcinogenesis. Loss of chromosome 1p has been identified recently in renal oncocytoma 20 , but this has not been previously shown to be a common cytogenetic alteration common to both entities, which is the key insight. Our delineation of the nature of chromosome 1p loss in renal oncocytoma provides the opportunity to identify novel tumor suppressor genes in future studies, and in establishing a possible carcinogenesis progression sequence.

There has been a recent advent of targeted therapies for a wide variety of cancers. Given the relative rarity of chRCC, there is no current standard of care and it is unlikely that any specific clinical trial is feasible or will be initiated. Here, we report two clinically relevant pathways--the c-erbB2/HER2 pathway and the mTOR signaling pathway--are dysregulated in chRCC on exploratory pathway analysis of mRNA expression, but our evaluation of extracellular HER2 and phospho-AKT immunohistochemical expression has not provided direct support for this mRNA finding. On a clinical trial level, in a subgroup analysis of a Phase III trial of temsirolimus, an mTOR inhibitor, in poor-prognosis RCC of all subtypes, patients of non-clear cell histology benefited as much as patients with clear cell histology, if not more 21 . Our findings do not permit a single definitive conclusion about the nature of pathway activation in these two entities. Currently, mTOR inhibitors remain a clinical standard of care for poor-risk metastatic non-clear cell renal cell carcinoma. HER2 expression has been evaluated in chromophobe RCC and oncocytoma, with distinct patterns of peptide expression varying according to epitope 22 . Interestingly, this study showed that strong intracellular HER2 expression (as defined by a 3+ expression) was strongly expressed in chromophobe RCC (9/19) but not in oncocytoma (1/11), whereas neither chromophobe RCC nor oncocytoma showed strong extracellular HER2 expression. Further evaluation of this is warranted, in conjunction with relevant fluorescent in-situ hybridization studies.

It has been previously reported that oxidative phosphorylation and energy pathway genes are overexpressed in chRCC and renal oncocytoma relative to the other subtypes of RCC 9 . We are able to clarify this issue, demonstrating that even between these two entities, there are major differences in quantitative expression of the same pathways discriminating the two entities. Consistent with these results, it has been recently reported that oncocytomas exhibit mitochondrial DNA mutations with clonal expansion and complex I deficiencies 23 . Oncocytoma contains a large number of mitochondria, and the overexpression of these genes involved in cellular metabolism may reflect the relative quantitative excess of the mitochondria. A similar profound modification in energy metabolism genes has been observed in thyroid oncocytomas, with high activity of the aerobic respiratory pathway 24 . It may be speculated that potential inhibition of autophagy in the chromophobe RCC may correspond to this difference as well. Rohan et al have previously reported in a smaller data-set that gene expression profiling is able to discriminate oncocytomas and chRCC 11 , and has reported that vesicular transport and cell junction proteins are relatively upregulated in chRCC.

In the process of validating our high-throughput expression studies, we report three novel markers discriminating between chRCC and oncocytoma: parafibromin, aquaporin 6, and synaptogyrin 3. Parafibromin, the protein product of the HRPT2 tumor suppressor gene, has been reported to be downregulated in a variety of tumors 17 25 , and a role has been assigned to it in the Wnt signaling pathway 26 . While the mechanism of parafibromin downregulation in parathyroid carcinoma appears to be mediated through gene mutation, this does not seem to be the mechanism in chRCC, as we have not identified any HRPT2 mutations after analyzing DNA samples from 5 chRCC tumors (data not shown). Similarly, other investigators have reported allelic imbalances in the HRPT2 gene in oncocytoma and chromophobe RCC, but no mutations 27 . Aquaporin 6 is an intracellular vesicle water channel protein reported to be expressed in the intercalated cells of the collecting duct 28 , which is hypothesized to be the originating cell for oncocytoma and chRCC 4 . Little is known about synaptogyrin-3, a tyrosine-phosphorylated protein that is expressed in synaptic vesicles 29 . The reasons underlying the reduced expression of aquaporin 6 and increased expression of synaptogyrin-3 in chRCC, relative to oncocytoma are uncertain.

In summary, we have comprehensively characterized the molecular profiles of chRCC and oncocytoma using high throughput expression and SNP profiling. We have consequently derived discriminating expression signatures, pathways, cytogenetic profiles and protein markers that are of biologic, clinical and therapeutic interest.

chRCC: Chromophobe renal cell carcinoma; SNP: single-nucleotide polymorphism; RCC: renal cell carcinoma; RMA: robust multichip average; SAM: significance analysis of microarrays; PAM: prediction analysis of microarrays; CGMA: comparative genomic microarray analysis; GO: gene ontology; CN: copy number; CNA: copy number alteration; mTOR: mammalian target of rapamycin; EGF: epidermal growth factor.

The authors declare that they have no competing interests.

MHT, HLT, JD, DM, SKK and JS performed the molecular genetic studies. MHT, HLT, KAF, EK, and DM analyzed the data. CFW, XJY, and PHT provided pathologic expertise. SG, SF, DA-O, TF, NT, MZ, GB, SD, VM, AV, and SR contributed and evaluated the samples. BTT conceived of the study and participated in its design and coordination. MHT and BTT wrote the manuscript. All authors reviewed and approved the final manuscript.