Genomic Landscape Predictive of Minimal Residual Disease (MRD) in Multiple Myeloma (MM)

Progress in the treatment of multiple myeloma (MM) has increased extent and frequency of response, as well as prolonged progression-free (PFS) and overall survival (OS). Today complete remission (CR) rates up to 70% are achieved with new drug combinations. This has lead to development of sensitive next generation sequencing (NGS) -based methods to predict deeper responses that may more accurately predict survival outcomes in MM. Our large recent study has confirmed the clinical impact of achieving MRD- status in MM. Here we are evaluating the genomic alterations that may predict attainment of MRD negative status in MM.

MRD status was evaluted in 279 patients from IFM/DFCI 2009 trial. We obtained gene expression by RNA-seq, and copy number profile by cytoScan HD array to evaluate genomic differences between MRD negative and MRD positive groups. We generated copy number data for 175 / 279 patients (72 MRD- and 103 MRD+) with Affymetrix Cytoscan HD array and compared genome wide copy number alterations. We observed statistically significant copy number alterations in chromosome 1p, 2, 4q, 11q, 13, 14 and 20 between MRD- and MRD+ patients. However, the extent of alterations in these regions is limited. The largest difference was on chromosome 11q arm where MRD- patients had 2.2 copies on average and MRD+ had 2.4 (p value < 0.001).

Similarly, we generated gene expression profiles with RNAseq for 69 MRD- patients and 92 MRD+ patients to study gene expression alterations that may predict attainment of MRD negative status and to examine possible biological pathways. Although first two component of principle component analysis (PCA) showed that two groups have similar expression profile, we were able to identify 586 differentially expressed genes; 333 of those were up and 253 were down regulated in MRD+ compared to MRD- groups. We found that seven oncogenes (CCND1, CD79B, IDH1, PATZ1, PAX5, POU2AF1, RUNX1) were significantly high in MRD+ and two (CCND2 and MYCN) were high in MRD-. Additional genes that were high in MRD+ samples were enriched in genes regulated by NF-kB in response to TNF, P53 pathway, KRAS signaling and genes down-regulated in response to ultraviolet (UV) radiation. Genes that were high in MRD- compared to MRD+ were also enriched in genes up-regulated by STAT5 in response to IL2 stimulation, p53 pathways and networks, and genes up-regulated in response to ultraviolet (UV) radiation pathways.

Finally, we have created a signature to predict MRD+ and MRD- in MM samples from differentially expressed genes. We used 40 genes that has at least 2 fold change difference between MRD+ and MRD- groups as a predictor and we randomly separated 161 RNAseq samples into train (n=99) and test group (n=62). We developed our classifiers with diagonal discriminant analysis and we achieved 0.79 classifier performance on test dataset. Then we tested our signature against 1000 random signature and it was significantly different than random signatures (Figure).

In conclusion, we here report a first genomic landscape predictive of minimal residual disease (MRD) in Multiple Myeloma (MM). This analysis will help understand genomic and molecular correlates of achieving minimal residula disease and confirms feasibility of using RNAseq data from diagnosis sample to predict MRD status. The ongoing  integration of other genomic correlates such as copy number status as well as alternate splicing may allow further improvement in  the performance of prediction.