Proteomic Profiling of AML Bone Marrow-Derived Mesenchymal Stem Cells (MSC) Reveals Marked Differences From Normal MSC.

Abstract 2355

The interaction of bone marrow (BM) mesenchymal stem cells (MSC) and acute myeloid leukemia (AML) cells creates a microenvironment (McrEnv) that supports and regulates the survival and proliferation of leukemic cells. These same BM McrEnv interactions can also create a sanctuary that protects subpopulations of AML blasts from chemotherapy. The mechanisms by which the BM-MSC McrEnv effects these changes remain unclear and the heterogeneity of these effects across different subtypes of AML (FAB, WHO or cytogenetics) is unknown. Furthermore, the functional differences between normal BM-MSC and AML BM-MSC biology are undefined. To address this we set out to perform proteomic profiling comparing normal and AML BM derived-MSC and to ascertain how AML BM-MSC protein expression patterns correlated with AML blast protein expression. We cultured BM samples for 4 to 8 weeks in MEM-alpha media with 20% fetal calf serum to isolate AML-MSC (n=106), and NL-MSC (n=70). Cells defined as MSC were positive for CD90 and CD105 and negative for CD45 by flow cytometry. Whole cell protein lysates were prepared. Matched AML protein samples prepared from mononuclear cell fractions from the same BM collection were available for most cases (n=96). We generated a custom Reverse Phase Protein Array (RPPA) from these samples and probed the array with 151 validated antibodies. Statistical analysis was performed by two-way ANOVA using Tukey's test to identify differential expression of individual proteins between the samples and Ingenuity pathway analysis was performed to elucidate differential pathway utilization.

Comparison of AML-BM-MSC to NL-BM-MSC demonstrated similar levels of expression for 66 proteins (43.7%) but 85, 67 and 28 were different at the p-value of < 0.05, < 0.01 and <0.0001 with a false discovery rate (FDR) of 0.05, 0.01, 0.0001 respectively. By function, the 28 proteins with greatest differential expression (Italic = lower in AML-MSC) included cell cycle regulators (P21, Cyclin D1, CDK4), adhesion/integrins (CD49b, CD31, and galectin 3, signaling pathway members and their targets (Smad1, Smad4, Stat1, Stat5, pPDK1, GSK3), apoptosis regulators (Bak, BCLX_L, Smac,) growth and proliferation regulators (pCREB, EIF2α, FOXO1α, Sirt1 Strathmin) and pIRS, cleaved Notch1, HSP90, TP53 CK2 and PP2A. Using mixed linear effectors protein expression levels in AML-MSC did not show correlation with patient's age (< 50 >), gender, blast count in BM or peripheral blood, or the percent of CD34 positivity. Protein expression in AML-BM-MSC from cases with favorable cytogenetics had significantly lower levels of GAB2, P27 P70S6K, SMAC and 14.3.3e and cases with unfavorable cytogenetics had significantly lower levels of antiapoptosis proteins Bax, Bad and BCL-X_L and higher levels of Smac as well as lower levels of phosphor-FOXO3a and pELK. Levels of ARC were higher in cases with intermediate cytogenetics. Ingenuity pathway analysis also demonstrated differential utilization of several families of proteins regulating signal transduction, apoptosis and transcription and connected to surface growth factor receptors and adhesion molecules. As anticipated for cells of different origin, the expression patterns were completely different between AML BM-MSC and AML blasts for 131 of 151 proteins (86.1%) (Tukey's p-value <0.0001 and FDR 0.0001). These results suggest that protein expression in AML MSC is markedly different from that of NL-MSC. Differential expression was observed in multiple functional groups suggesting that AML-MSC are functionally distinct from NL-MSC. Since MSC influence adjacent and nearby AML blasts it is likely that these variances impact AML blast biology. Additional analysis is underway to determine if recurrent patterns of protein expression exist in AML-BM-MSC, how these differ from protein expression patterns in NL-MSC, and whether AML-BM-MSC protein expression patterns correlate with AML-blast protein expression patterns. Correlation between MSC patterns and AML-blast patterns would provide therapeutically targetable sites in MSC that could be exploited to influence AML blast biology.