Sub-Classification of Hyperdiploid Myeloma Using Global Gene Expression Profiling and SNP-Based Mapping Arrays.

The translocation/cyclin classification system in myeloma does not neatly define subgroups of hyperdiploidy (HRD) and we sought a more definitive sub-classification. Using 131 pre-treatment samples (49 HRD with no split IgH locus by FISH) we defined subgroups using both supervised and unsupervised hierarchical clustering of gene expression profiles. RNA was purified from CD138+ cells, amplified using a 2-cycle IVT and hybridised onto U133 Plus 2 GeneChips. On 30 of the 49 HRD samples we also performed 500K SNP mapping arrays to define the true extent of the genomic change in HRD. The most common trisomic chromosomes were 15 (97%), 9 (86%), 19 (80%), 5 (77%), 11 (74%), 3 (64%), 21 (54%) and 7 (54%). There was no association between HRD and any of the major genetic abnormalities (1p, 1q, 6q, 8p, 13, 16q and 17p) compared to the non-HRD (NHRD) group. Many interstitial deletions were seen in all HRD samples, on both odd and even numbered chromosomes. However, using gene mapping alone it was not possible to globally sub-classify HRD myeloma. We compared NHRD and HRD sample gene expression profiles, removing differences between t(4;14) and t(11;14) cases in the NHRD group. This analysis showed that HRD samples segregate into 2 groups; one with a pattern distinct to NHRD samples and another containing genes that are up-regulated in both HRD and NHRD samples. In this analysis 176 genes were up-regulated in the HRD samples and were predominantly located on the trisomic chromosomes, especially 19, 11, 9 and 5. These genes showed a predominant upregulation of HGF and TRAIL, and down-regulation of TRAIL-R2 compared to NHRD samples. Unsupervised hierarchical clustering split the HRD samples into 5 distinct groups suggesting that there are distinct pathological entities. Group 1 overexpressed 90 genes including BCL2, CCNL1 (cyclin L1) and CDK6, consistent with a proliferation signature. Group 2 overexpressed interferon inducible genes including IFI6, IFI27, IFIT1 as well as TRAIL. Group 3 upregulated genes included IL8, MMP9 and TIMP2. Group 4 upregulated transcripts include neurexophilin 3. Group 5 was less well defined but contained transcripts for CCND2, WNT5A and CXCR4. To define clinically relevant subgroups the HRD samples were clustered comparing response or no response to induction chemotherapy. Analysis showed that Group 1 cases cluster together and were either non or minimal responders. This is consistent with the Group 1 cases over-expressing cell-cycle and proliferation related genes. Group 5 clustered together and were either complete or partial responders, and had a low expression of the genes over expressed by Group 1. The non-responder group overexpressed 58 genes and include MMSET-like 1 (in a region on 8p paralogous to 4p containing FGFR1), DVL3 (dishevelled homolog 3) and CCNL1. 23 genes were over expressed in the complete response group including caspase 1 and manic fringe homolog. The unsupervised HRD cluster and the supervised response cluster shared 10 genes, including CCNL1 and ASS. We have used both genetic and expression data to further define the HRD sub-group in terms of gene expression signatures and response to therapy and have identified 5 groups, of which Group 1 has a proliferation signature and poor response to induction therapy.