Single-Cell Proteomics Identifies Leukemia Landscape Associated with Clinical Outcomes in R/R AML Treated with MDM2i (Milademetan) and FLT3i (Quizartinib): Putative Role of CD68 and Diversity Index

Distinct mutations could differentially regulate cellular programs and alter the proteomic landscape in AML. Sequential acquisition of various mutations not only leads to clonal diversification but also alters the leukemia proteomic landscape through activation of mutation-specific gene programs. Characterization of AML proteomic profiles and diversity could be utilized as a measure of genetic imprint on leukemia proteome to inform clinical decision-making. We reasoned that diverse leukemia-specific proteomic profile could be indicative of the presence of multiple mutations activating numerous pathways, thus leading to a heterogenous clonal composition and more likely to therapy resistance. We aimed to test this hypothesis by assessing the proteomic profiles of FLT3-ITD AML patients treated with MDM2i (Milademetan) plus FLT3i (Quizartinib) (NCT03552029) and interrogate the association between proteomic landscape and therapy response.

We assessed single-cell proteomic profiles of 35 sequentially collected samples for six selected patients treated with MDM2i+FLT3i using CyTOF, enabling us to assess expression of 51-parameters across leukemia compartments and identify leukemic clones with distinct proteomic profiles. Three patients achieved CRi while three patients did not respond. This allowed us to start interrogating proteomic signatures for their ability to predict response to therapy. We performed single-cell analysis and interrogated the phenotypic profiles of leukemia compartments to assess leukemia hierarchies, defined by spatial organization of leukemic subpopulations, and whether mutations in AML were associated with unique phenotypes. Notably, we found that NPM1-mutant (Mt) leukemia cells lacked CD34 expression, expressed high levels of CD99 and had patchy c-kit expression. Despite lacking a canonical marker, CD34, high-dimensional analysis positioned NPM1-Mt leukemia cells spatially in close proximity to CD34+ leukemia cells (NPM1 WT), indicating that NPM1 WT and Mt leukemia cells are closely related. CD34+ expressing cells most likely serve as the founding clone and acquisition of NPM1 mutation led to emergence of CD34- leukemia clones. As expected, all three patients who achieved CRi were NPM1-Mt and NPM1-Mt leukemia cells in CRi patients expressed CD68. Importantly, we also observed that CD68+ leukemia cells were eradicated in a NR patient where only a fraction of leukemia cells expressed CD68. This suggests that NPM1 mutations could activate unique cellular programs and induce distinct differentiation states (CD68), which could sensitize leukemia cells to MDM2+FLT3 inhibition. Altogether, NPM1 mutation status and CD68 expression level were associated with therapy response. Next, we mapped the response kinetics and quantified survived leukemia cells across multiple timepoints. Strikingly, MDM2i+FLT3i almost completely eliminated circulating blasts in responders (R) by day 8 while leukemia blasts persisted in NR (median blast %: 0.11 in R vs 19.8 in NR). This indicates that assessment of therapy response as early as day 8 could provide insights into the overall response and identify patients who will fail to achieve CR. Importantly, patients with reduced leukemia blasts at day 8 were also leukemia-free in BM at the end of cycle 1.

Lastly, we sought to investigate the association between proteomic landscape diversity and therapy response, and quantified the number of leukemia subpopulations by unsupervised clustering. The median number of subpopulations detected in R vs NR at baseline were 3 and 9, respectively. We also utilized the inverse Simpson index to quantify the proteomic diversity of leukemia compartments and to further investigate the association between proteomic diversity and therapy outcome in an unbiased manner. The median diversity indices in R vs NR were 64 vs 212, revealing that patients with CR had restricted pre-treatment proteomic diversity. These findings suggest that a pre-treatment diverse phenotypic landscape could portend poor therapeutic outcome.

Altogether, single-cell proteomic analysis identified correlates associated with overall clinical response in AML patients treated with MDM2i+FLT3i. Further validation is needed in a larger cohort of patients. Such approaches could be utilized in clinical-trial settings to predict therapy response with targeted agents and inform clinical decision-making.