Proteomic signatures of acute myeloid leukemia (AML) distinguishes different outcome groups across cytogenetics and identified potential therapy targets
6523 Background: Cytogenetics (CG) guide AML treatment but reliable markers predicting response and relapse within CG groups are missing. We therefore determined whether functional proteomic signatures can classify AML into groups with different outcomes and risk of relapse. Method: Using Reverse Phase Protein Array, total and phospho-site specific expression of 37 proteins in 73 primary AML was measured. Outcomes in the set were comprised equally of primary refractory (PR), relapsed (Rel) and continuous complete remission (CCR) patients. Cell lysates were spotted on nitrocellulose coated slides, probed with validated antibodies, expression intensities were quantified, data was standardized and analyzed for correlations using different clustering approaches. Results: Unsupervised hierarchical clustering based on Pearsons’ correlation distance yielded 4 large clusters. Subsequent perturbation bootstrap re-sampling arranged samples into four classes that correlated with initial response to therapy and risk of relapse (see Table ). Protein profiles in each of he 4 classes differed. Cytogenetic marker distribution were similar across the 4 clusters. Class 1 and 4 demonstrated a similar predictive value of patient outcome as cytogenetics. In classes at highest risk of relapse (2, 3) different proteins were predictive of response. In class 2, the most discriminatory proteins predicting CCR were elevated AMPK, p27, 4-EBP1, BclXL. In class 3, relapsed patients had elevated PTEN, phospho-Stat3, total Stat3, and phospho-PKCα compared to CCR patients. Conclusion: Pretreatment protein expression signatures divide AML into classes that predict for initial achievement of CR and subsequent relapse independent of CG. Poteomic profiling may suggest potential therapy targets as opposed to CG or transcriptional profiling. These preliminary results need to be confirmed in formal training and test sets prior to changing patient management. [Table: see text] No significant financial relationships to disclose.