Genome-Wide Analysis of DNA Copy Number in Multiple Myeloma Using High-Density SNP Arrays Reveals Clustering Patterns with Distinct Transcriptional Profiles.

Multiple myeloma (MM) is characterized by a high genomic instability that involves both ploidy and structural rearrangements. Nearly half of MM tumors are non-hyperdiploid and are frequently associated with 13q deletion and chromosomal translocations involving the immunoglobulin heavy chain (IGH) locus on chromosome 14q32. The remaining tumors are hyperdiploid, showing low prevalence of both IGH translocations and chromosome 13 deletions. Our study was aimed at providing insights into the genomic heterogeneity associated with plasma cell neoplasms by defining the genome-wide pattern of genetic lesions in a representative and stratified panel of MM patients. To this end, genome-wide profiling data of 45 plasma cell dyscrasia patients (41 MM and 4 plasma cell leukemia) were generated on GeneChip® Human Mapping 50K Xba SNP arrays, and the local DNA copy number variations were calculated using the DNAcopy Bioconductor package. The patients were clustered using the non-negative matrix factorization (NMF) algorithm to identify, within the natural grouping of profiles, the strongest clusters on the basis of their genomic characteristics. We identified three consistent clusters, characterized by

• recurrent gains of odd-chromosomes, suggestive of the hyperdiploid status (Group A),

• high frequency of chromosome 13 deletion and 1q gains (Group B), or

• high frequency of chromosomes 13, 14, 16 and 22 deletions and losses of 1p and 4p regions, together with some cases showing 1q gains (Group C).

To determine whether peculiar transcription fingerprints characterized these groups, gene expression profiles of 40 out of 45 corresponding samples generated on GeneChip® HG-U133A arrays were analyzed using the Prediction Analysis of Microarray (PAM) software. The multi-class analysis identified 229 transcripts (corresponding to 195 genes), which specifically marked the three groups. In particular, Group A was characterized by the overexpression of genes involved in the translational machinery or thought to be involved in MM pathogenesis such as the HGF, the tumor necrosis factor ligand TRAIL, DKK1, and c-KIT. Upregulation of the CKS1B gene was present in Group B and C, most likely reflecting the high frequencies of 1q gains in tumors within group B and C and its consequent deregulation. Group C was marked by the specific downregulation of genes mainly mapped to 1p arm: AMPD1, CSDE1, AKR1A1 and the PRKACB kinase, suggesting a relationship with the recurrent 1p loss within the group. Our data further supported the notion that structural abnormalities in multiple myeloma are associated with gene expression imbalances, and provide novel analytical approaches for the identification of genetic lesions and molecular patterns of the disease.