Clear cell (or conventional) renal cell carcinoma (RCC) accounts for about 85% of all primary kidney malignancies and, although familial forms of RCC exist, the disease is more often sporadic[1]. This pathology is associated with genetic alterations affecting particular chromosomes [2, 3]. The most frequent findings are deletions and unbalanced translocations involving chromosome 3p, resulting in the loss of specific regions, e.g. 3p25-p26 including the von Hippel-Lindau (VHL) gene locus [4]. Duplications of chromosomes 5q and 7 and deletions on chromosomes 6q, 8p, 9p and 14q are other typical chromosomal abnormalities. Another recurrent RCC genetic feature is a particular pattern of loss of heterozygosity (LOH), i.e. the change from a heterozygous genotype in a normal sample to a homozygous one in a tumor, with a high frequency of allelic imbalances on chromosome 3p in conjunction with 6q, 8p, 9p and q, and 14q [5]. These DNA alterations define a specific pattern of genetic instability, which represent a tumor-specific molecular fingerprint potentially useful for diagnostic applications [3]. In addition, some of these DNA alterations have been associated with tumor progression and metastatic potential, and so may be useful prognostic indicators [2]. Furthermore, many studies have profiled transcriptional patterns in RCC samples, but a univocally accepted gene expression signature is lacking [6].

Until now, chromosomal imbalances (i.e. amplifications and deletions) in neoplastic diseases have been studied with a variety of methods, such as cytogenetic techniques, comparative genomic hybridization (CGH) and the newest array-CGH, and fluorescence in situ hybridization (FISH), but each suffers limitations in resolution or high-throughput capacity [7, 8]. The recent development of single nucleotide polymorphism (SNP) array technology has greatly improved the detection of DNA copy number (CN) changes, and now allows the simultaneous genotyping of more than 100,000 polymorphic loci distributed across all human chromosomes and thus the high-resolution scanning of the entire genome [9]. Moreover, by allelotyping the DNA sequence, this technology permits the detection of LOH events, which has principally been performed by investigating microsatellite markers at specific genomic regions. The assessment of areas rich in LOH events is a useful approach for identifying regions potentially harboring novel tumor suppressor genes (TSGs) [10]. Thus, SNP mapping technology permits the simultaneous analysis, at whole genome level and on the same platform, of chromosomal and allelic imbalances. Moreover, it permits the distinction between LOH associated with CN changes (such as hemizygous deletions) and that associated with a CN neutral status, due to different mechanisms including mitotic recombination, gene conversion and mitotic non-disjunction leading to uniparental disomy (UPD). Simultaneous assessment of LOH and CN changes at high resolution (more than 50,000 SNPs) has been recently performed in prostate cancers [11] and gliomas [12], but not yet in RCC.

There is an increasing tendency to combine genomic analysis with transcriptomic profiles, in order to study the relation between CN changes and gene expression levels. The integration of CGH profiles and transcriptional data has demonstrated that CN alterations have a clear impact on expression levels, in a variety of tumors [13–15]. Currently, this dual strategy is considered the most effective approach for interpreting genome-wide data about DNA and RNA anomalies in cancer, in order to identify chromosomal regions and genes involved in tumor initiation and progression. The proof of principle comes from the study by Garraway et al. [16], in which the combined analysis of genome-wide SNP-based CN data and expression profiles led to the identification of a candidate lineage-specific oncogene associated with a novel amplification in malignant melanoma. Concerning RCC, Furge et al. [17] developed a computational approach to predict cytogenetic abnormalities from gene expression data and confirmed alterations on chromosomes 3 and 5 as the typical DNA signature of this pathology. However, an integrated analysis of DNA and RNA from RCC samples using genome-wide approaches has not yet been reported.

In this study, we applied the 100K SNP mapping array technology to assemble a detailed genome-wide map of copy number alterations (CNAs) and LOH events in a collection of 27 RCC tissue samples. Moreover, we combined genomic data with the transcriptomic profile to identify chromosomal regions and individual genes modulated by the underlying DNA alterations and thus potentially useful in clinical applications as novel RCC constitutive markers.

In this study, we used the 100K SNP mapping array technology to investigate DNA copy number alterations and LOH events in a collection of 27 RCC tissue samples. Simultaneously, transcriptomic analysis was performed to identify differentially expressed genes potentially involved in tumor etiology. To narrow down the number of candidate genes to those whose altered expression level was attributable to underlying DNA alterations, genomic and transcriptomic data were combined at genome level. This permitted us to identify specific chromosomal regions with concordant alterations in DNA and RNA profiles. These regions are of particular interest since they contain genes which, being transcriptionally modulated due to genomic alterations, could be novel RCC constitutive markers actively involved in tumor initiation and progression and useful in clinical applications.

To our knowledge, this is the first study assembling a genome-wide map of CNAs and LOH events characterizing RCC pathology. We documented CN loss regions on chromosomes 3p and 6q, thus confirming the genetic signature typical of RCC while greatly improving the resolution compared to previous studies employing cytogenetic and CGH techniques [2, 3, 19]. We also observed novel CN loss regions on chr. 1p, 9q and 22q. We confirmed the presence of genomic regions affected by LOH on chromosomes 3p, 6q and 9q, as previously identified by Thrash-Bingham et al. in a microsatellite analysis on all chromosomes [5]. In addition, our data revealed novel LOH sites on chromosomes 1p, 2q, 3q and 13q. Taking into account the DNA copy number status, we distinguished LOH events associated with CN loss (e.g. on chromosomes 1p and 9q) highlighting the presence of hemizygous deletions, and LOH events with CN neutral status (e.g. on chromosomes 2q and 13q) indicating the occurrence of concomitant genetic mechanisms (including mitotic recombination, gene conversion or UPD events) to maintain the normal diploidy status. It is widely accepted that regions of LOH accompanied by deletion (representing the second hit of Knudson's hypothesis) are of particular interest since they may contain TSGs actively involved in tumor etiology and thus completely inactivated [10]. Interestingly, we identified a novel LOH and CN loss site on chromosome 1p36.22-p36.21, a region frequently deleted also in glioma and neuroblastoma [20] and harboring potential TSGs, such as IIp45 (invasion inhibitory protein 45), found down-regulated in glioblastomas [21], and DHRS3, a dehydrogenase/reductase enzyme frequently inactivated by deletion in human neuroblastoma cell lines [22] and by promoter methylation in melanoma cell lines [23].

It is known that RCC is characterized by a particular genetic signature on chr. 3p [24]. We observed deletions affecting chr. 3p26.3-p26.2 and 3p25.3-p25.1 in 96% samples, thus confirming the crucial importance of inactivation of these regions in RCC oncogenesis [2]. Moreover, in 30% of samples we observed LOH with CN neutral status on chr. 3p14.3, a region already reported to be frequently affected by LOH in RCC pathology and harboring the FHIT gene, thought to be involved in tumor suppression in many neoplasms [25]. Differently, the most frequent LOH area, mapped on chr. 3p26.2-p25.3 and observed in 41% samples, showed a more complex DNA arrangement, presenting both CN loss and neutral status in different tumor samples. This region harbors in 3p25.3 the VHL gene, the main TSG implicated in the early steps of RCC oncogenesis[4]. As previously reported, in 50–70% of RCC sporadic forms, VHL gene is completely inactivated through a combination of various mechanisms such as mutations, promoter methylation, LOH and deletions [4]. VHL protein is a key regulator of the hypoxic response and controls the transcription of several genes. Concordantly with a VHL loss of function, we found increased expression of three known hypoxia-inducible genes: the carbonic anhydrase CA9, a cell growth regulator; the chemokine receptor CXCR4, which controls metastatic spread; and the vascular endothelial growth factor (VEGF), involved in cell adhesion and angiogenesis. These genes have already been reported to be up-regulated in RCC and other human cancers [26].

Although our analysis is limited by its relatively small sample collection compared to that of Zhao et al. [27], this is the first study to perform a genome-wide CNAs and LOH mapping using RCC and matched normal DNA from the same patients, together with gene profiling analysis. We documented a strong association between DNA alterations and gene expression levels and observed that CN gain events have a stronger influence on transcriptional modulation than CN loss events, in agreement with previous studies on other tumor types [28]. That gene amplification greatly enhances expression levels is supported by ample evidence from mammalian cell lines and tumors [29]. On the contrary, there are many aneuploidy-independent mechanisms leading to transcriptional down-modulation (e.g. mutation, promoter hypermethylation) [30]. Thus, we suggest that the recently developed approaches to infer "functional aneuploidy" of tumor samples by their transcriptomic profile [17, 31] should take into account these differences.

Focusing on regions with concomitant CN gain events and up-regulated expression levels, we observed that over-expressed genes often mapped close to one another to form blocks of concordantly regulated genes. These regions may contain novel RCC constitutive markers, i.e. genes with up-regulated expression levels due to DNA amplification events particular of RCC histotype and necessary for its malignant transformation or progression. Up-regulated genes whose protein products are released to the extracellular space are ideal tumor markers for clinical applications, since they are detectable in patients' biological fluids without invasive procedures. Interestingly, within the amplified region on chromosome 5q11-q12, we found strongly increased expression levels of two adjacent genes encoding secreted proteins. ESM1 (endothelial cell-specific molecule 1, alias endocan), crucially involved in cell growth regulation, vascular remodeling and angiogenesis, is dramatically over-expressed in endothelial cells of kidney cancers and other tumors, in association with increased tumor aggressiveness and vascularity [32, 33]. Moreover, elevated protein levels have been observed in serum of late-stage lung cancer patients, correlating inversely with both survival and time to tumor progression [34, 35]. Next, GZMA gene, encoding a secreted granzyme involved in cytotoxic T cell-mediated immune response, is over-expressed in RCC tissues with high contents of natural killer cells and tumor-infiltrating lymphocytes [36]. Furthermore, in the amplified region on chromosome 11q22.3-q23, we found four adjacent over-expressed genes, including PDGFD, which encodes a secreted activator of the inflammatory response playing an important role in angiogenesis and is over-expressed in prostate carcinoma [37] and many tumor, including RCC, cell lines [38], and CASP1, which is up-regulated in pancreatic adenocarcinomas [39] and presents increased plasma levels in metastatic melanoma patients [40]. These genes and their encoded proteins will be further studied to determine if they have roles in RCC pathology and if they could be novel RCC biomarkers detectable in clinical biological fluids.

By combining genomic and transcriptomic profiles from a collection of RCC samples, we identified at high resolution specific genomic regions concomitantly affected by DNA alterations and modulated expression levels. This study lays the groundwork for a more detailed search for novel RCC biomarkers associated with these aberrant regions, in particular in areas with increased DNA copy number. The integration of data from both DNA and RNA profiles should greatly improve the specificity of analysis and increase the possibility of identifying the genetic causes of tumor pathologies. Moreover, data from the follow-up evaluations of these RCC patients should enable us to validate this novel genomic RCC fingerprint.