Improving prognostication for patients with myelodysplastic syndromes

Abstract

Background and aims: MDS constitute a heterogenous group of myeloid malignancies mainly characterized by dysfunctional hematopoiesis. Although cytopenia, dysplastic features and evidence of clonality are essential criteria for the diagnosis of all MDS, the several subtypes of the disease have a highly variable prognosis. The increasing quality and accessibility of DNA sequencing techniques have enabled huge advances for molecular characterization of the disease, and the prognostic impact of specific molecular markers in MDS is now well established. Several prognostic scoring systems have been developed during the last two decades but none of these tools accounted for the effect of molecular markers on outcome. MDS with RS is easily recognizable by the intra-cellular presence of iron-loaded mitochondria and this subtype reflects the heterogeneity of MDS. Hence, while RS are classically associated with SF3B1 mutations and an indolent disease course, RS are sometimes found in aggressive subtypes of MDS or AML. Patients and diseases change over time, and evolution patterns themselves can tell us something about disease biology and outcome. Clinicians account for these variations in practice, but current prognostic models do not. This may partly explain remaining discrepancies between observed and predicted prognosis. Hence, in this thesis we aimed to i) develop a novel prognostic score including molecular markers to refine prognosis prediction at diagnosis, ii) study the prognostic impact of combined gene mutation and gene expression in MDS with RS and iii) assess whether changes in erythrocyte (E) transfusion patterns during the early disease course can refine outcome prediction.

Methods: Study I – an international cohort of 2957 patients with MDS, MDS/myeloproliferative neoplasms (MPN) were retrospectively collected. DNA sequencing with a panel of 152 genes known to be involved in myeloid malignancies was performed on all samples. Clinical data, cytogenetic and molecular features were retrieved and their association with outcomes was studied. A Cox multivariable model was used to estimate relative weights of selected explanatory variables. The score was validated on an independent cohort of 754 Japanese patients with MDS. Study II – A total of 129 patients with MDS and RS (MDSRS+) was assembled. All samples underwent DNA sequencing according to study I and thereafter RNA sequencing of CD34 sorted bone marrow mononuclear cells. Supervised/unsupervised clustering analysis and digital sorting were performed. A Cox multivariable model was used to assess association between clinical and genomic/transcriptomic patterns and outcome. Study III – a cohort of 677 Swedish patients was gathered from study I. We collected complete information on administered E-transfusions through the nationwide SCANDAT3-S database. Cox regression analyses were used to assess associations between clinical, molecular and transfusion data, and outcome. A Markov multistate model was used to assess association between changes in transfusion patterns and outcome.

Results: Study I – TP53multi, MLL-PTD and FLT3 mutations were shown to be predictive of dismal outcome. In contrast, SF3B1 mutations were associated with favorable prognosis, however this effect was significantly influenced by the co-mutation patterns. A total of 22 variables (clinical, cytogenetic, and molecular markers) were used to build the IPSS-M score, each of them carrying a specific mathematic weight according to their individual impacts on endpoints. The calculation of the IPSS-M resulted in a unique score for individual patients and assigned each case to one of the 6 IPSS-M risk categories. When compared to the IPSS-R, the IPSS-M score clearly improved outcome prediction and led to the restratification of 46% of patients. The IPSS-M is validated both in MDS/MPN with WBC count below 13x109/L and in therapy related MDS (t-MDS). Study II – Most (~90%) MDSRS+ cases were found to have a mutation in SF3B1, SRSF2 or TP53multi. Overall, TP53multi and splice factors mutations were mutually exclusive, and SF3B1 and SRSF2 mutations cooccurred in only 3% of the patients. The three genetic subgroups were shown to have very different outcomes. Supervised transcriptome analysis confirmed the distinction between SF3B1-, SRSF2- and TP53multi-mutated MDS with RS. Unsupervised clustering analysis found three transcriptomic groups, each with distinct erythroid/megakaryocytic progenitor fraction, which predicted OS independently of IPSS-M. Study III – Whereas TP53multi, poor cytogenetic and higher bone marrow blasts predicted shorter time to first E-transfusion event, higher hemoglobin level and SF3B1alpha only were associated with longer time to first E-transfusion event. Next, E-transfusion state at 8 months after diagnosis was shown to be a strong predictor of OS independently of IPSS-M. Our model based on E-transfusion state at 8 months and IPSS-M (model 2) improved significantly prognostic prediction compared to IPSS-M only (model 1). Finally, a multistate model revealed that individual transfusion trajectories during the early disease course impacted both future transfusion requirement and OS.

Conclusion: This thesis provides evidence that integration of genomic data to clinical characteristics improves greatly prognostication in MDS and we suggest that the novel IPSS-M prognostic score is implemented in clinical practice to provide further guidance in therapeutic decision-making. Our work also indicates that the heterogeneity of outcome in MDS cannot be explain by genetic profiling only and that studies of gene expression and integration of dynamic parameters among other techniques will contribute to a better understanding of the clinical course. In general, this thesis advocates for the need of a holistic approach of the disease to deepen our understanding of underlying mechanisms and ultimately to improve the care of patients with MDS. Enormous efforts are currently put in the field of precision medicine in cancer. Future integrative multiomics studies will hopefully improve individualized care to increase survival and quality of life of patients with MDS.