The genetic hallmark of retinoblastoma is mutation or deletion of the RB1 gene, whereas other genetic alterations that are also required are largely unknown. To screen for genomic imbalances on a genomewide level, we studied a series of 17 primary retinoblastomas by matrix-based comparative genomic hybridization (matrix-CGH). The matrix-CGH chip contained 6,000 immobilized genomic DNA fragments covering the human genome, with an average resolution of about 500 kb. The most frequent imbalances detected were gains on chromosome arms 1q (12 of 17), 6p (10 of 17), 2p (5 of 17), and 19q (4 of 17) and loss on 16q (7 of 17). Candidate regions could be narrowed to small intervals by the identified minimally overlapping regions on 1q22, 1q32.1q32.2, 2p24.1, and 6p21.33-p21.31. Furthermore, two as-yet-unknown high-level amplifications were detected, each in a single patient, on chromosome bands 1p34.2 and 1p33. Thus, this study identified new chromosomal regions and therefore potential candidate genes that may play a role in retinoblastoma.

The paediatric eye tumour retinoblastoma is initiated by inactivation of RB1, a tumour suppressor on chromosome 13q. In addition to RB1 loss, many retinoblastomas show other genetic alterations including gains on chromosomes 6p21-pter and 1q31-q32. Recently, the minimal region of gains on chromosome 6 was narrowed to band p22. We examined genomic gains and expression changes in primary retinoblastomas to identify potential target genes in 6p22. Quantitative multiplex PCR detected copy numbers X3 in 25 (33%) tumours and no gains in 31 of 76 (40%) tumours. The remaining 20 (26%) samples showed gains only at some loci, most often including E2F3 and DEK in 6p22.3. Analysis of RNA from 21 primary retinoblastomas showed that expression levels of these and some other genes in 6p22 correspond to DNA gains. However, KIF 13A, a reported candidate oncogene on 6p, was expressed at low levels or absent. Clinical manifestation of tumours with gains at all 6p22 loci was distinct in that distribution of age at diagnosis was markedly shifted to older age compared to tumours with no or partial gains. In summary, our results suggest that DEK and E2F3 are potential targets of 6p gains in retinoblastoma.

Many retinoblastomas (Rbs) show genomic alterations in addition to mutational loss of both normal RB1 alleles. The most frequent of these changes are gains on chromosomes 1q and 6p and losses on 16q. To identify the genes targeted by gains on chromosome 1q, we used quantitative-multiplex PCR to determine DNA copy number changes in 76 primary tumors and 6 Rb cell lines. In addition, in 21 of these tumors, gene expression was analyzed by cDNA microarray hybridization. Increased copy numbers of loci on chromosome 1q were present in 34 (45%) primary tumors and in all 6 cell lines. Two regions of gain emerged, one in 1q32 and another in 1q21. Tumors with 1q gains showed higher RNA expression of several genes in these 2 regions. The clinical manifestation of tumors with and without gains was similar with regard to many aspects, including size, necrosis and calcification. However, the distribution of age at diagnosis was remarkably distinct, with earlier diagnosis in tumors without gains. This suggests that these tumors either are initiated earlier or grow faster than tumors with gains. This association with clinical manifestation indicates that gains on 1q are significant for the biology of Rb. The genes on 1q with copy number gains and overexpression are candidates that need to be tested for their individual contribution to the progression of Rb. ' 2005 Wiley-Liss, Inc.

In addition to RB1 gene mutations, retinoblastomas frequently show gains of 1q and 6p and losses of 16q. To identify suppressor genes on 16q, we analyzed 22 short tandem repeat loci in 58 patients with known RB1 mutations. A subset of tumors was also investigated by conventional and matrix comparative genomic hybridization. In 40 of 58 (69%) tumors, we found no loss of heterozygosity (LOH) at any 16q marker. LOH was detected in 18 of 58 (31%) tumors, including five with allelic imbalance at some markers. In one tumor LOH was only observed at 16q24. As the parental origin of allele loss was unbiased, an imprinted locus is unlikely to be involved. Analysis of gene expression by microarray hybridization and quantitative RT real-time PCR did not identify a candidate suppressor in 16q24. Cadherin 13 (CDH13), CBFA2T3, and WFDC1, which are candidate suppressors in other tumor entities with 16q24 loss, did not show loss of expression. In addition, mutation and methylation analysis showed no somatic alteration of CDH13. Results in all tumors with chromosome 16 alterations define a single minimal deleted region of 5.7 Mb in the telomeric part of 16q24 with the centromeric boundary defined by retention of heterozygosity for a single nucleotide variant in exon 10 of CDH13 (Mb 82.7). Interestingly, clinical presentation of tumors with and without 16q alterations was distinct. Specifically, almost all retinoblastomas with 16q24 loss showed diffuse intraocular seeding. This suggests that genetic alterations in the minimal deleted region are associated with impaired cell-to-cell adhesion.