We have shown that DNA copy-number analysis using high-density SNP arrays provides a molecular karyotype for every patient and leads to the identification of novel chromosomal abnormalities that impact prognosis. Our SNP array data mirror well the skyline recurrence plot using microarray-based comparative genomic hybridization for chromosomal aberrations derived from 67 newly diagnosed MM.18
High-resolution karyotyping is particularly relevant in MM since almost all patients (98% in this study) harbored chromosomal lesions and required only a small amount of purified myeloma cells (approximately 100,000 cells). In myeloma with lack of proliferative clone makes conventional cytogenetics informative in only limited number of patients (approximately 30% of patients), while use of interphase FISH circumvented the analysis to only very select loci in the genome. Use of SNP array as described here makes the technique applicable in most of the patients with evaluation of myeloma cells irrespective of their cell cycle status as well as investigation of the whole genome. Standardization of the technique and requirement of relatively smaller number of myeloma cells now provide general applicability of this technique in future.
We present here the first prognostic model based on frequent genome-wide DNA copy-number lesions in newly diagnosed MM. Univariate analyses revealed genetic lesions located on chromosomes 1, 5, 8q, 9, 11, 12p, 14q, 15q, 16q, 19q, 20p, 22q, significantly associated with survival (P < .001). After testing the robustness of these selected DNA segments by permutation and resampling only MCGL located at 1q, 5q, and 12p were retained in a multivariate predictive model.
The main interest of this model resides in its ability to identify patients who greatly benefit from high-dose therapy (patients with amp(5q31.3) alone and low Sβ2
M; 5-year overall survival, 87%). Of note, the prognostic impact of amp(5q31.3) over-rides that of hyperdiploidy and 5q lesion is easier to assess than hyperdiploid status. In addition, our model identifies high-risk patients (25% of patients) with a survival equivalent to that of gene expression models.26,36
Comparison of both CNA- and expression-based models suggest that these molecular predictors should be combined to form the most accurate prognostic model for MM. The CNA-based model is more discriminating than FISH-based models that are by definition targeted on combination of specific abnormalities such as t(4;14), del(17p), or del(16q) in the context of high or low Sβ2
For example, high-risk patients defined either by our CNA-based model or by del(17p) and/or a t(4;14) (approximately 20% of patients) had median overall survival of 29 months and 37 months, respectively. del(12p) is the strongest prognostic parameter in our cohort, this lesion has been previously reported in both newly diagnosed1,25
andrelapsed MM patients.38
del(12p) frequency calculated from Smadja et al1
(11.6%) is comparable with our results (13%). The strong prognostic value of del(12p) emerged solely in our study probably for several reasons including: the size of the cohort, the resolution of the genetic method used, and the median follow-up period available in the cohort.
Our results from LOH and copy-number analyses of 16q region in the 192 patients are in agreement with a recent study of 55 newly diagnosed patients with using the same Affymetrix platform.7
We identified 16q LOH in 38.3% of patients. The entire or interstitial deletions represent 27.8% (53 of 191) of patients and UPD represent 10.5% (20 of 191). Univariate analysis identified two MCGL located at 16q12 and 16q23 significantly associated with adverse overall survival (P
< .001), but their stability was not strong enough (permutation and resampling P
= 0.3; false discovery rate > 30%) to enter the multivariate model. In addition, Cox analysis revealed that UDP16q has no prognostic impact in our cohort (P
In addition to the strong prognostic impact of specific DNA copy-number changes revealed by high resolution SNP array, combination with expression profiles provides novel features of disease pathogenesis. NF45 stabilizes double-stranded RNA-binding protein NF90 complexes and reduction of NF45 or NF90 induces cell growth retardation.32
NF90 binds to vascular endothelial growth factor 3′ untranslated mRNA and contributes to hypoxia-induced vascular endothelial growth factor expression and tumorigenesis in vivo in a breast carcinoma.39
Of particular interest, recent results demonstrated an essential role of ADAR1 in hematopoiesis, where it acts to suppress interferon signaling and to block premature apoptosis.40
The authors suggest that the ADAR1-NF90 interaction may be critical in this process. These recent data support the hypothesis that concerted overexpression of ADAR1 and NF45 could contribute to MM pathophysiology.
CD27 is a member of the tumor necrosis factor receptor superfamily and like many other members of this family activates c-jun kinase and both the classical and alternative nuclear factor-κB pathways.41
We can hypothesize that diminution of CD27 in high-risk patients reduces NF-kB pathway activity in these patients. This hypothesis is supported by recent studies showing that the proliferation and MMSET
-spike subgroups of poor prognosis relative to the other MM subgroups had low expression NF-kB signature.42,43
In the same way, relative underexpression of CD27
in CD-1 compared to CD-2 myeloma subgroups is associated with a lower NF-kB signature.42
These results suggest that therapeutic strategies targeting the NF-kB pathway may not be tailored for high-risk patients defined by molecular profiling.
In conclusion, high-resolution karyotyping using SNP or CGH arrays is a powerful investigative tool to detect novel lesions that impact prognosis in MM (this study7,27
), thereby improving accuracy of the established predictive models. Molecular karyotype will occupy rapidly a place of choice in the armory of biologic tools used for prognostication and patient adapted-risk therapeutic strategy.