Proteomic Profiling In CD34+CD38- Stem Cells and CD34+ Cells From 185 Myelodysplasia Patients Using Reverse Phase Proteins Arrays (RPPA) Reveals Recurrent Proteins Expression Signatures with Prognostic Implications

Studying the cellular biology of myelodysplasia (MDS) has been greatly hindered by limitations in the amount of material that can be obtained from patients and by fact that the neoplastic MDS cells are a small fraction of the total bone marrow population. Sample requirements for Reverse Phase Protein Array (RPPA) are small enough to enable array like analyses that were previously impossible. We therefore constructed an RPPA to perform proteomic profiling in MDS that analyzed both the CD34+ and the stem cell enriched CD34+CD38- fractions.

Fresh (F) or cryopreserved (C) primary bone marrow (BM, n=153 F and 51C) or blood (PB n=84 F and 7 C) samples from 185 MDS patients (123 diagnosis, 62 relapse) underwent sequential CD34 selection followed by CD38 depletion using magnetic antibody sorting to produce CD34+ and CD34+CD38- (SC) fractions. All samples yielded sufficient CD34+ cells, with a median yield of 4×10⁵ cells and recovery of 3.2%. CD34+CD38- recovery correlated with the starting number of cells and was successful 50% of the time when starting with >1×10⁷ cells, with a median yield of 3×10⁵ cells and recovery of ∼ 2% of the starting number of cells. There were 91 Paired CD34+/SC samples from 82 cases and 27 paired BM/PB and 16 paired DX/Rel samples. Slides were probed with 128 antibodies (ABs) against apoptosis, cell cycle, signaling (STP), regulating proteins, integrins, phosphatases etc including 90 vs. total protein, 33 vs. phos-specific sites and 5 vs. caspase or PARP cleavage sites. Spot intensities were quantified using MicroVigene software. Data was analysed using R, with loading control and topographical background normalization being utilized.

A mixed linear model was used to compare protein expression in CD34+ vs. SC. Of the 128 proteins 82 were significantly different with a FDR of 5%. The majority of SC clustered separately from the CD34+ cells but there were a few SC with CD34+ like expression and about 25% of CD34+ cases were found in the SC group. We therefore looked at protein expression patterns in CD34+ and SC separately. Protein expression patterns distinguished FAB and IPSS groups for both CD34+ and SC, and with 8 and 11 proteins being significantly different among CD34 + cells and 15 and 8 being different in the SC. Proteins were clustered based on Pearson and complete linkage. In CD34+ cells, the Gap statistic suggested that there were 11 constellations of protein with correlated expression. Principle component analysis identified 6 distinct Protein Expression Signatures (ProExpSig) based on the expression of the 11 constellations. In the SC analysis there were 8 protein constellations and principle component analysis identified 4 distinct ProExpSig. Perturbation boostrap clustering of the protein expression constellations and the ProExpSig showed that they were highly reproducible for both the CD34+ and SC analyses. Signature assignment was more strongly associated with FAB in the SC than CD34+ analysis. In contrast IPSS score was more strongly associated with signature assignment. Concordance of ProExpSig between the CD34+ and SC groups was minimal reflecting the differences in protein expression. Among the patients that were treated is was obvious that there were favorable and unfavorable signatures. For the CD34+ analysis ProExpSig 1,2 and 3 were unfavorable and had a median survival of 71 weeks compared to favorable groups 4, 5, 6 which had a median survival of 127 weeks (p=0.009). A similar finding was present in the SC analysis where ProExpSig 1 and 2 were unfavorable with a median survival of 80 weeks compared to favorable ProExpSig 3 ad 4, which had a median survival of 128 weeks (p=0.08). When combined, the information from both the CD34+ and SC analyses improved the prognostic discrimination (P=0.007) with the four groups having median survival of 50, 53, 97 and 248 weeks.

This is the first study to perform proteomic profiling in MDS or in MDS stem cells. Protein expression in SC was markedly different from CD34+ cells stressing the importance of studying this rare population to truly understand the biology of MDS. Similar to our findings in AML and ALL, MDS is characterized by several distinct ProExpSig which form favorable and unfavorable prognostic groups. The differences in the patterns of protein expression can be exploited therapeutically to target specific agents to those cases where that particular pathway or protein is active.

Kantarjian: AstraZeneca: Research Funding.