Unravelling molecular determinants of ruxolitinib treatment response in myelofibrosis using single cell multiomics

Abstract Myelofibrosis (MF) originates from the acquisition of somatic mutations in hematopoietic stem and progenitor cells (HSPCs). MF driver mutations affecting JAK2, MPL or CALR genes induce the constitutive activation of JAK/STAT signaling that sustain myeloproliferation. The JAK inhibitor Ruxolitinib (Ruxo) is effective in reducing splenomegaly and constitutional symptoms but can't eradicate MF neoplastic cells; indeed, loss of response is observed due to disease progression and symptoms worsening. The aim of this study is to unveil molecular determinants of Ruxo resistance in MF patients. To investigate clonal dynamics associated with Ruxo response, we performed single-cell (SC) genomics coupled with SC proteomics using the Tapestri platform (Mission Bio) in 12 MF patients before and during Ruxo treatment. Patients were classified as Ruxo responders and non-responders based on symptoms and splenic response. Circulating CD34⁺ HSPCs were isolated from cryopreserved peripheral blood mononuclear cells (PBMCs), mixed at a 1:1 ratio with PBMCs, and stained using the TotalSeq-D Human Heme Oncology antibody panel (BioLegend), which includes 42 oligonucleotide-conjugated antibodies targeting HSPC and lineage-specific markers. For SC genomic analysis we used a custom 299-amplicon panel targeting 29 genes frequently mutated in myeloid neoplasms. Two patients, one responder and one non-responder, were subjected to SC transcriptomics integrated with SC proteomics using the 10X Genomics platform. Cells were stained with the TotalSeq-B Human Universal Cocktail (BioLegend). According to clinical data, circulating CD34⁺ cell level was increased after treatment in non-responders, whereas it decreased in responders. Furthermore, a partial reduction in driver mutation variant allele frequency (VAF) was observed only in granulocytes from responder patients. All patients harbored a driver mutation, with eight patients also carrying at least one additional mutation. The most frequently mutated genes were epigenetic regulators, particularly TET2 and ASXL1. SC multiomic analysis allowed the reconstruction of the mutation acquisition order and clonal architecture in immunophenotypically defined hematopoietic cell populations. CD34+ HSPCs were reduced in responder patients after Ruxo treatment while CD14+ monocytes were increased in both responders and non-responders. As for non-responders, the acquisition of MF driver mutation was preceded by hits in epigenetic modifier genes, TET2 or ASXL1. In these patients, the most mutated malignant clone which dominated hematopoiesis before treatment further expanded or at least remained stable during time in both CD34+ and CD34- cells. On the contrary, in responder patients, driver mutation was the first molecular hit and non-mutated cells expanded over time, albeit differences in the clonal dynamics of CD34+ and CD34- cells were observed. According to SC genomic information, we classified cells based on the acquisition order of MF driver mutations, epigenetic variants and zygosity. Interestingly, as regards CD34- cell clusters, in non-responders co-mutated clones expanded at the expense of WT cells in monocytes and T and B lymphocytes. Conversely, WT cell frequency increased at the expense of MF driver homozygous cells in responders. In CD34+ cells, WT cells expanded while MF driver homozygous cell frequency was reduced in responders. Nevertheless, clones displaying both MF driver and epigenetic variants expanded despite Ruxo treatment, in both responders and non-responders patients, To deepen understand molecular events associated with Ruxo response we performed SC transcriptomic analysis coupled with SC proteomics in two patients. Thanks to immunophenotypic information we were able to identify the same cell populations identified in SC genomic study. Strikingly, differential gene expression analysis revealed the increased activation of JAK/STAT signaling pathway in non-responder monocytes despite Ruxo treatment. Collectively, our results demonstrate that the order of mutation acquisition impacts Ruxo response in myelofibrosis patients. Ruxo suppresses the expansion of MF driver homozygous clones but has only limited effects on co-mutated clones. As a result, co-mutated clones may benefit from Ruxo treatment by outcompeting other neoplastic cell populations and persist as a disease reservoir within the stem and progenitor cell compartments.