Proteomics: the key to understanding AML hematopoietic aging Free

In this issue of Blood,Cheng et al¹ integrate genomic, transcriptomic, proteomic, and phosphoproteomic measurements of 374 patients with newly diagnosed acute myeloid leukemia (AML) to describe a protein signature of hematopoietic aging that can be used to both understand why AML is so deadly in older patients and to stratify patients by disease severity. Despite the recent popularity of proteogenomic data in AML, this study highlights how a deep molecular profiling approach can uncover novel disease mechanisms and provide new insight into why AML is so deadly, particularly in older patients.² 

AML treatment is stymied by its genetic heterogeneity with multiple molecular pathways that can cause disease progression. Patients are generally stratified by genetic markers at diagnosis to determine a specific treatment course³; however, each additional molecular data type that is collected (RNA sequencing [RNA-seq],⁴ mass spectrometry–based proteomics^5,6 single-cell RNA-seq [scRNA-seq]⁷) provides additional layers of complexity that are not typically captured in genetic profiles. Molecular subtyping is now a standard approach to stratify for disease heterogeneity; patients are clustered according to molecular measurements, and the subgroups obtained provide more homogeneous subsets of patients to study and compare across a given disease cohort. Subtyping via unsupervised clustering has now become standard in high-throughput patient profiling, particularly in AML, in which proteomics subtyping has been repeated across different cohorts,^5,6 each with different findings, such as the role of mitochondrial signaling in disease progression⁵ and identification of novel potential drug combinations.⁶ 

This study is another example of how proteomic subtyping can provide novel insights into AML progression. In the study, Cheng et al clustered the 374 patients in their cohort based on the global proteomics measurements and identified 8 distinct molecular subtypes, more than previous proteomic studies, thereby enabling them to uncover additional patterns that might have been missed in other cohorts. In choosing to cluster by proteomics alone, the authors were able to explicitly focus on the proteomic diversity of their cohort and identify subtypes with differing prognosis that overlapped with the existing genetic and World Health Organization classifications. Within these subtypes, there were notable differences in the age distribution and specific enrichment in aging-related pathway activity. Specifically, subtypes with a higher overall age were enriched in various messenger RNA, protein, and phosphosite markers of hematopoietic aging, including megakaryocyte expression, platelet expression, and immune-related pathways, suggesting concerted aging-related patterns of protein expression.

To tease out the proteins directly associated with hematopoietic aging, the authors sought out proteins that are known markers of hematopoietic stem cells (HSCs) in the scRNA-seq⁷ and proteomics data⁸ that were significantly correlated with patient age and survival in the 374 patient cohort, which led to the identification of 19 proteins. By applying a weighted scoring technique, the authors used these proteins to create a hematopoietic aging score (HAS) that can predict hematopoietic aging (and therefore overall survival) based on protein expression alone. A higher HAS correlated with biologic phenotypes, such as myelodysplasia-related AML, NPM1 mutations, and clonal hematopoiesis–related gene mutations. In addition, the HAS had significant prognostic value across the patient cohort and enabled the identification of potential drug combinations that were particularly efficacious in high-HAS samples.

The increase in AML treatment options over the past 2 decades has not improved the overall prognosis for older patients, because tumors acquire resistance to treatment over time in these patients. Unlike most pharmacologically targeted pathways that are altered only in AML, hematopoietic aging is a natural process that predisposes patients to AML development. Excitingly, by looking at the underlying aging process that is causing the genetic changes that lead to AML, this work identified proteins in a signaling pathway that could be targeted pharmacologically to prevent disease progression. Because the HAS itself has prognostic value, it can be used to assess at-risk patient populations to determine their risk of getting AML and enable clinicians to study strategies to prevent progression to overt AML.

In summary, this work underscores the value of large proteogenomic studies to uncover novel biology through careful measurement of genomics, transcriptomics proteomics, and phosphoproteomics. Through data integration, disease subtyping, and pathway enrichment, the authors identified elements of the HSC aging process that are linked to poor prognosis in AML that evade traditional genomics approaches. The development of the HAS from these elements has great potential to improve pharmacologic opportunities and patient care.

Conflict-of-interest disclosure: S.J.C.G. declares no competing financial interests.