Impressions of the myeloma landscape

Multiple myeloma (MM) is a heterogeneous disease. In this issue of Blood, Broyl and colleagues¹  present a comprehensive analysis of gene expression profiles that leaves an overall impression consistent with previous molecular classifications^2,3  and highlights 3 additional molecular subtypes of the disease.

A useful molecular classification forms the basis for identifying patients with shared biology, prognosis, and response to treatment. Although the use of a molecular classification is firmly established in leukemia and lymphoma, it has been only slowly adopted for myeloma. In part this is because, unlike in the former 2 diseases, the molecular subtypes of myeloma are not associated with very distinct morphologic or histologic appearances. They do, however, have distinct prognoses and response to treatment, which is leading to the more widespread use of molecular markers in the management of myeloma.⁴ 

Myeloma can be thought of has having primary and secondary genetic events. The primary events include recurrent, nonoverlapping, immunoglobulin gene chromosome translocations involving 4p16 (FGFR3 and MMSET), 6p21 (CCND3), 11q13 (CCND1), 16q23 (MAF), and 20q11 (MAFB). The patients lacking 1 of these translocations are primarily characterized by hyperdiploidy, with trisomies of chromosomes 3, 5, 7, 9, 11, 15, 19, and 21. A unifying feature appears to be dysregulation of a D-type cyclin, and, to a large extent, the primary genetic events can be considered disease-defining events that will not change during the course of a given patient's disease. Secondary genetic events include mutations that activate RAS and the NFKB pathway, inactivating mutations of p53 and KDM6A (UTX), rearrangements that dysregulate MYC, and closely associated 1p deletion/1q amplification. Although not disease-defining, the secondary genetic events may have important implications for prognosis and response to treatment, and be subject to modulation by effective therapy.

Some of these genetic events are associated with very distinctive gene expression profiles, most notably the MAF translocations, which are characterized by the unique and high level of coexpression of a number genes, many of which are thought to be direct targets of the MAF transcription factors. Similarly, activation of the NFKB pathway results in the coexpression of a set of characteristic NFKB target genes. Hyperdiploid MM is characterized primarily by the low level overexpression of a very large set of genes located on the chromosomes involved in the trisomies. Finally, some genetic events are not correlated with any patterns of gene expression, most notably RAS mutations.

Conversely, there are some distinctive patterns of gene expression that are not associated with known underlying genetic events. These include the coexpression of myeloid lineage genes (often interpreted to represent contamination of the CD138-selected plasma cells with myeloid cells), the coexpression of genes associated with proliferation, the coexpression of a number of cancer testis antigens (CTA), and, as noted by Broyl and colleagues,¹  the coexpression of a set of genes that include PTP4A3 (PRL3), PTPRZ1, and SOCS3.

To take an unbiased approach to the molecular characterization of MM, the authors have analyzed the gene expression profiles of a new cohort of untreated MM patients, and using somewhat different analytical methods, repeated the basic analysis first reported by Zhan et al from the University of Arkansas for Medical Sciences (UAMS).³  Most importantly they reproduced the unbiased identification of the 8 main subgroups (CD1, CD2, MF, MS, PR, HY, LB, MY) and identified 3 additional subgroups (NFKB, CTA, and PRL-3). Although the impression for the entire cohort was the same, at the individual patient level there were some important differences. In contrast to the previous report, only a third of the CD1 had a t(11;14), and there was a 38% discordance with the UAMS CD1. In addition, there was a discordance of 40% for the PR subgroup, 30% for the NFKB, and 18% for both the CD-2 and HY. There are some technical aspects that confound the analysis, most notably the divergent handling of the MY samples, the lack of an NFKB subgroup in the original UAMS classification, and the apparent use of the original translocation and Cyclin D (TC) classification based on the Affymetrix Hu95A GeneChip, as opposed to the one updated for the HU133Plus2 GeneChip actually used.⁵ 

Should one be concerned at this high level of discordance? I do not think so, as it is not that important, and in fact quite predictable. The first part of the analysis identifies the 5% most variable genes, and uses these to perform hierarchical clustering. As one varies the number and identity of this list of genes (eg, 1% most variable, 10% most variable, etc), patients move between clusters, new clusters are identified, and old clusters split. Three different patients may express a variable number of genes characteristic of hyperdiploidy, proliferation, and NFKB (see figure). As one varies the number of genes from each subgroup within the list used for clustering, a given patient may be classified differently. In part it is a problem of taking multidimensional data and reducing it to a single dimension. An alternative would be to keep all of the dimensions, starting with the first dimension—the primary genetic event (captured to a great extent by the TC classification), and adding additional dimensions to represent the dominant secondary transcriptional signatures (PR, LB, NFKB, CTA, PRL-3).

So which molecular classification (TC, UAMS, H65) should be used going forward? In my opinion, whichever one works best for the question being addressed, given the inherent limitations of all such classification. In the future we will have molecularly targeted therapies (eg, NFKB inhibitors, STAT3 inhibitors, FGFR3 inhibitors, CTA immunotherapy, antiproliferatives) and the molecular classification used will follow logically from the therapy being considered.