Construction of a Myeloid Gene Signature Revealing Novel Therapeutic Targets for Myeloid Malignancies

FH, SL and BG equal contributors

Myeloproliferative neoplasms (MPN) and acute myeloid leukemia (AML) exhibit shared hallmarks of hyperinflammation and aberrant clonal expansion. Chronic MPNs exhibit a propensity for transformation to secondary AML (sAML), which imparts a dismal prognosis with limited treatment options. However, the pathogenesis of leukemia transition for MPN and the bidirectional relationship between MPN and AML remain incompletely understood. In this study, we investigated commonly-featured genes and pathways of these two diseases to explore novel therapeutic strategies.

To dissect the causal association between MPN and AML, we first performed two-sample bidirectional mendelian randomization (MR) analysis which revealed that MPN predicts the risk of AML across multiple statistical methods with consistent directionality and significance. Heterogeneity, pleiotropy, or outlier effects were ruled out by sensitivity analyses with no reverse causality from AML to MPN. To identify a potential myeloid signature, we estimated causal effects of 2,940 plasma proteins on MPN via a two-sample MR framework and identified 55 candidates significantly associated with increased risk of MPN (P < 0.05, OR > 1).

Utilizing data from the BEAT-AML cohort, we further screened the prognostic value of these candidates using Cox and log-rank test survival analyses and identified a 4-gene myeloid signature comprising MPO, CDCP1, CRISP3 and DXCR, with each conferring prognostic significance individually. Our scRNA-seq results confirmed enrichment of myeloid signature genes in myeloid progenitor cells, with further elevation in MPN. Of note, expression of myeloid signature genes was induced by Jak2^V617F and MPL^W515L in Ba/F3 cells. We then investigated the combined prognostic significance of these four genes by constructing a risk score via LASSO regression modeling and divided patients into high- and low-risk groups accordingly. Cox analysis validated the risk score as an independent prognostic factor (HR: 3.01 (1.2 - 7.6), P = 0.019) and Kaplan-Meier survival analysis demonstrated that low-risk AML patients had a significantly better survival than high-risk AML patients with median survival of 21.9 months vs. 11.9 months (P < 0.0001).

The prediction accuracy of the myeloid signature was successfully validated in the TCGA-LAML cohort. The risk score significantly correlated with several key clinical parameters, including age, platelet count, ELN2017 staging, monocyte percentage and bone marrow cellularity. To interrogate the underlying mechanism of the myeloid signature, we performed gene set enrichment analysis (GSEA) and observed enrichment of several inflammatory pathways in high-risk AML patients. A similar enrichment of inflammatory pathways in MPN samples compared to healthy donors observed in our previous scRNA-seq results reinforced hyperinflammation as a shared etiology of myeloid malignancies.

To explore potential therapeutic strategies, we performed in silico drug sensitivity screening which predicted specific vulnerability of high-risk AML to compounds targeting PI3K/AKT/mTOR pathway signaling. Activation of PI3K/AKT/mTOR signaling in both MPN and AML samples observed across multiple single cell and bulk RNA-seq datasets further prompted us to evaluate targeted therapies against this pathway in myeloid malignancies. MPN and AML cells exhibited sensitivity to mTOR inhibitors as indicated by suppression of cellular proliferation, metabolism, colony formation, and inflammatory cytokine secretion in conjunction with induction of apoptosis. We further evaluated the therapeutic effects of mTOR inhibition in vivo via administration of omipalisib, a potent dual PI3K/mTOR inhibitor, to JAK2^V617F knock-in mice. Omipalisib treatment significantly ameliorated features of myeloid malignancies including splenomegaly and leukocytosis which were both significantly reduced following 4 weeks of treatment, without affecting body weight.

In summary, our MR analyses reveal that MPN predicts the risk of AML and enabled the construction of a novel myeloid signature which risk stratified AML patients across two different cohorts. We further demonstrate shared activation of PI3K/AKT/mTOR pathway across myeloid malignancies, with in vitro and in vivo data providing a rationale for therapeutic targeting of PI3K/AKT/mTOR pathway in these diseases.

Oh:CTI BioPharma: Consultancy; Blueprint Medicines: Consultancy; Geron: Consultancy; PharmaEssentia: Consultancy; Constellation: Consultancy; Celgene/Bristol Myers Squibb: Consultancy; Disc Medicine: Consultancy; Kartos Therapeutics: Consultancy; Novartis: Consultancy; AbbVie: Consultancy; Sierra Oncology: Consultancy; Incyte: Consultancy.