Large-Scale Transcriptome Analysis Supports Genomic Taxonomy and Predicts Response to Hypomethylating Agents in Myelodysplastic Neoplasms

Genomics characterization of myelodysplastic neoplasms (MDS) provided a substantial contribution to elucidate myeloid neoplasms pathobiology and opened the way for disease classification based on genetic features. Converging evidence from several correlation studies uncovered specific disease subsets with homogenous phenotype driven by somatic mutation in a restricted number of genes. A recent study focusing on co-mutation patterns and clonal hierarchy provided additional evidence supporting genetic classification of MDS (i.e. genetic taxonomy, PMID: 38958467), which overcomes limitation of current classifications (still relying on morphologic criteria) and displays a high translational therapeutic potential. In spite of that, systematic characterization of each genetic subgroup pathobiology is lacking, preventing genetic taxonomy from full application in clinical practice. To overcome this limitation, we studied the combined genetic and transcriptomic profiles of two independent MDS prospective cohorts, aiming at improving our understanding of the MDS biology, identifying new molecular targets suitable for innovative personalized treatment and improving treatment strategy development.

Collectively, we studied 708 MDS patients from two referral centers, divided in training (n=399) and validation (n=309) cohorts. All cases underwent targeted DNA-sequencing on peripheral blood or bone marrow mononuclear cells (BMMNC), together with bulk CD34⁺ BMMNC whole transcriptome RNA-sequencing. Normal CD34⁺ BMMNC from 16 healthy age-matched donors were also included in the study as controls. Differential gene expression and regression models were analyzed separately in the training and validation cohorts, with P value of < 0.01 considered significant for downstream analysis.

Differential gene expression analysis of genetically-defined MDS subgroups compared to healthy controls identified two major gene groups. The first one consisted of genes differentially expressed in hemopoietic stem and precursors cells (HSPC), confirming previous reports suggesting that HSPC composition was the major determinant of CD34⁺ BMMNC transcriptome (PMID: 37498312, 35618837). The second group consisted of genes specific for each genetic taxonomy group. Gene-set enrichment analysis revealed several enriched oncological signatures and various metabolic processes. For instance, MDS with biallelic TET2 mutation displayed an enrichment of interleukins and rhodopsin-like receptor genes, whereas MDS with SF3B1 mutation showed enrichment of adhesion molecules (ITGA3, VCAM), MET signaling and heme degradation. MDS with del5q displayed deregulation of iron homeostasis (TRF2, FDXR) and epigenetic regulators (L3MBTL4). Downregulation of FOS transcription factor was specific for MDS with DDX41 mutation. Overall, transcriptome supervised analysis supported the validity of genetic taxonomy in MDS identifying transcriptomic signatures restricted for specific genotypes. Dysregulation of cancer and inflammatory pathways was often associated with enrichment in genes involved in cell adhesion and interaction with extracellular matrix, providing further evidence to the relevance of bone marrow microenvironment for innovative treatment development.

Considering that no score is currently available in clinical practice to accurately predict treatment response to hypomethylating agents (HMA) in MDS, we applied least absolute shrinkage and selection operator regression to somatic mutation, cytogenetic abnormalities and normalized gene expression levels from ~18000 genes to derive a molecular signature predicting HMA response. Results from the exploratory cohort, validated in the validation cohort, led to a 10 gene expression signature predicting HMA response with 85% accuracy (PPV 82%, NPV 87%, kappa=0.69, P<1e-6). Genes expression positively correlating with HMA refractoriness were BTK, ASGR1, BFSP1, SLC25A21, CTSE and TGFBI, whereas KCNQ4, FGL2, PUDP and XK expression was positively correlated with HMA response.

In conclusion, the present study identified distinct gene expression profiles supporting MDS genetic taxonomy and correlating with HMA response. Further studies are warranted to functionally characterize the pathobiology of each genetically-defined MDS subgroup, which will foster new targeted and personalized therapeutic strategies in MDS.