Genetic Determinants Of Disease Phenotype In Myelodysplastic Syndromes

Diagnosis of myelodysplastic syndromes (MDS) is greatly limited by the scarce reproducibility of morphological analysis of dysplasia and the poor specificity of dysplastic changes. In a group of myeloid disorders classified on the basis of morphological criteria, identifying specific associations between genotype and disease phenotypes is essential to recognize disease entities according to distinctive genetic profiles. These genotype-phenotype correlations are illustrated by the 5q- syndrome, and more recently by the association of SF3B1 mutations with ring sideroblasts, which makes SF3B1 the first gene to be strongly associated with a specific morphological feature in myeloid neoplasms. In this work, we performed a comprehensive mutation analysis in a large and well clinically characterized cohort of MDS patients, and sought for associations between genotype and disease phenotype.

Massively parallel pyrosequencing was used to screen coding exons of 111 genes known to be implicated in myeloid neoplasms. We studied 245 patients diagnosed with MDS according to WHO criteria at the Department of Hematology, Fondazione IRCCS Policlinico San Matteo, University of Pavia, Italy: 26 were classified as RA, 35 as RARS, 69 as RCMD, 27 as RCMD-RS, 10 MDS with isolated del(5q), 36 as RAEB-1, and 42 as RAEB-2.

The most frequently mutated gene categories were splicing factors (122/245, 49.7%), chromatin modifiers (91/245, 37.1%), DNA methylation (71/245, 29%), transcription factors (58/245, 23. 7%), signalling (44/245, 18%), and RAS pathway (27/245, 11%).

We first focused on nonsideroblastic MDS categories (RA, RCMD, RAEB-1, RAEB-2) with the aim to identify genetic determinants of multilineage dysplasia and excess of bone marrow blasts. We found that mutations of genes involved in DNA methylation (DNMT3A, TET2, IDH1, IDH2) were significantly associated with multilineage dysplasia or excess blasts (P=.007). In fact, of 48 patients carrying these mutant genes, only two had isolated erythroid dysplasia, whereas 26 of 48 showed multilineage dysplasia and 20 excess blasts. The positive predictive value of mutations in this gene category for multilineage dysplasia was 92.9%, and 95.8% for advanced disease, defined as either multilineage dysplasia or excess blasts.

We then studied patients with mutations in splicing factors other than SF3B1 (SRSF2, U2AF1, ZRSR2), and found that, of 57 patients carrying these mutaions, 3 showed unilineage dysplasia, whereas multilineage dysplasia or excess blasts were observed in 27 patients each (P=.016), resulting in a positive predictive value for multilineage dysplasia of 90% and of 94.7% for advanced disease.

Mutations in RAS pathway (KRAS, NRAS, CBL, NF1) were observed in 19 patients, 7 classified as RCMD and 12 as RAEB, whereas no cases with isolated erythroid dysplasia were observed (P=.033). The resulting positive predictive value of these mutations for multilineage dysplasia or excess blasts was of 100%.

Finally, we found a significant association of RUNX1 mutations with advanced disease (P=.024), resulting in a positive predictive value for multilineage dysplasia of 83% and of 100% for advanced disease. Overall, having at least one mutation in genes involved in DNA methylation, splicing factors other than SF3B1, RAS pathway or in RUNX1 had a positive predictive value for multilineage dysplasia or excess blasts of 96.5%.

We then focused on sideroblastic categories (RARS and RCMD-RS). A significantly different prevalence of SF3B1 mutations was observed in patients with isolated erythroid dysplasia versus multilineage dysplasia (91.4% in RARS vs 55.6% in RCMD-RS, P=.002). Conversely, a significantly higher prevalence of mutations in other splicing factors (SRSF2, U2AF1, ZRSR2) was found in patients with RCMD-RS compared with those with RARS (2.9% vs 22.2% respectively, P=.023).

In conclusion, in a cohort of MDS patients well characterized with respect to clinical features and mutation status, we identified significant associations between genotype and disease phenotype. In particular, we found that mutations in genes involved in DNA methylation, splicing factors (other than SF3B1), RAS pathway and in RUNX1 are highly predictive of multilineage dysplasia and excess of BM blasts. These observations, together with the close association of SF3B1 with disease phenotype with ring sideroblasts, may represent the basis for a molecular classification of MDS.