Proteomic Profiling Reveals Protein Expression Patterns Correlated with Outcome in Acute Promyelocytic Leukemia (APL)

Abstract

Background: Acute promyelocytic leukemia (APL) is the most favorable prognostic form of acute myeloid leukemia (AML). However, 5-20% of the patients relapse, which indicates that the clinical need to identify those who are likely to relapse remains. In this study we aimed to define protein expression patterns of APL with the goal to find certain protein patterns that would be prognostic for outcome.Methods: Custom reverse phase protein arrays (RPPA) were performed on 20 newly diagnosed APL patients, 205 newly diagnosed AML patients, and 10 normal CD34+ bone marrow samples, probed with 230 strictly validated antibodies. First, proteins were divided into 31 functionally related groups to analyze them in the context of other proteins. Progeny clustering analysis was applied to each of the functional groups to identify an optimal number of protein clusters; cases with similar expression of proteins. Then, block clustering was done on a binary matrix that indicated the protein cluster membership for all protein groups, for selected APL patients only, to assess whether we could identify subsets of APL cases (signatures) that expressed similar patterns of correlated protein clusters from various protein groups. Survival curves were generated for the identified signatures and individual proteins were compared between signatures using the student t-test with the Bonferroni correction (P<0.05).Results: Protein expression levels were measured for APL and AML patients relative to the normal CD34+ samples. Progeny clustering analysis revealed protein clusters for each functionally related protein group. All protein clusters that included APL patients were also found in AML patients. Only for the functional protein groups “Apoptosis regulating”, “HIPPO” and “TP53”, we observed protein patterns specific to APL. Block-clustering on APL patients showed strong correlations between various protein clusters and identified two subsets of patients, which we called protein signatures. Interestingly, survival curves showed that all of the relapse patients (N=4/20, 80.0%) were present in signature 2 (N=4/13, 30.8% vs. signature 1 N=0/7, 0%). Due to the small number of patients in our cohort, no level of significance was reached (P=0.084). Moreover, the number of promyelocytes in the bone marrow was lower in signature 1 (median 47 vs. 71 overall, P=0.070) and fibrinogen levels were higher in signature 1 (median 235 vs. 159, P=0.052) compared to signature 2. Direct comparison of individual proteins between signature 1 and 2 found 38 differentially expressed proteins. Notably, most protein had opposite protein levels relative to the normal cells in both signatures. For instance, ERG, STMN1, SSBP2, PRKAA1_2.pT172 and hnRNPK were higher in signature 2 and lower in signature 1, whereas CTSG, ITGB3 and ITGA2 were higher in signature 1.Conclusion: We demonstrated that although APL is genetically and therapeutically unique form of AML, most expression patterns were comparable between APL and AML. Most interestingly, we found that a subset of APL patients could be distinguished as high-risk solely based on their protein expression patterns. By knowing the signature membership at an early stage could, we could then decide whether additional therapy is required or not. Recognition of proteins that are higher or lower expressed in both signatures could be used to identify novel therapeutic targets. However, validation of our findings in a larger cohort is required. DisclosuresNo relevant conflicts of interest to declare.