Serial Gene Expression Studies of Malignant Plasma Cells before Treatment and after Relapse: Identifying Genes Linked to Drug Resistance in Myeloma.

Virtually all patients with multiple myeloma (MM) eventually develop drug resistance. It is unclear whether drug resistance derives from the emergence of a subclone with de novo, genetically programmed resistance, or whether it develops through accumulation of genetic lesions or epigenetic mechanisms facilitated by microenvironment interactions. We investigated changes in gene expression profiles (GEP) that accompany drug resistance by comparing microarray signatures of purified plasma cells from newly diagnosed MM patients before therapy and after relapse. Paired samples (n=20) were obtained from treated with high-dose chemotherapy and tandem PBSCT. RNA was isolated from plasma cells labeled and hybridized to U133Plus2.0 high-density oligonucleotide microarrays capable of investigating ~33,000 genes. Baseline and relapse samples were grouped together and the GEP of the two groups compared. Only genes with Chi-square >3.84 (P <0.05) or present in >50% of samples were retained. 15,000 genes were analyzed with SAM in R (two-class paired case). The cut-off point was the smallest such that the estimate of the false positives was zero. A total of 234 genes were significant by SAM, with a median FDR of 1% and a 90^th percentile FDR of 3%. Of the 234 genes, 199 were down-regulated and 35 were up-regulated at relapse. Thus, with 90% confidence, the FDR among genes found significant was no higher than 3%. Chromosomal translocations involving the immunoglobulin locus are likely to be initiating events in nearly half of all MM cases and we have previously shown that GEP is a robust method for identifying all the common translocations in MM. Because translocations result in the spiked expression of target oncogenes and represent useful tumor-specific landmarks, we also investigated the distribution of the common spikes in both baseline and relapse samples. For the most part, translocations present at baseline were also present at relapse. As expected from our previous work, whereas MMSET spikes were consistently found associated with t(4;14)(p16;q32), FGFR3 spikes could be absent at baseline and lost during progression. A CCND3 spike, not present in the baseline sample, appeared in the relapse sample. This is consistent with this translocation being a secondary event associated with tumor progression. MAF expression was lost in one relapse. This was unexpected and may be due to mislabeling. No MAFB spikes were found in any of the baseline or relapse samples. Interestingly, at the time of this analysis, there were no CCND1 spikes in the relapse cases, consistent with this translocation being a good prognostic marker. Thus, significant and recurrent changes in GEP accompany the development of drug resistance in MM. Surprisingly, the majority of altered genes were down-regulated in relapse. The significance of this finding is not clear, but may reflect loss of chromosome material and/or epigenetic silencing of genes through increased DNA methylation. Validation of our findings may provide important insight into the mechanisms of MM drug resistance.