Defects in apoptotic mechanisms play an important role in chemotherapy resistance in AML, resulting in either refractoriness or high levels of Minimal Residual Disease (MRD) ultimately leading to relapse. For that reason the search for prognostic markers has focused on genes involved in apoptosis. At present the general concept encompasses a balance between pro- and anti apoptotic family members, but few studies have addressed large sets of genes. We tried to identify gene-profiles with maximal predictive power for patient survival by simultaneous quantification of 35 apoptosis related target sequences. For this purpose we have used the Reverse Transcriptase-Multiplex Ligation dependent Probe Amplification (RT-MLPA, Hess,
Leukemia 2004; 18:1981
) method in viable (7AAD− /AnnexinV-) blasts of 120 newly diagnosed AML. Eight out of 35 genes predicted for poor survival and these surprisingly contained both anti-apoptotic (n=4) and pro-apoptotic (n=4) genes. These were used to compose 3-gene-signatures. High gene expression in these signatures predicted for poor survival in 40 out of the 56 possible combinations, with no preference for the anti-apoptosis genes. This is quite unlike the current concepts and evoked the idea that a more general mechanism is responsible for defective apoptosis regulation. To identify such mechanism we extracted the latent variable structure from our dataset using Principal Component Regression analysis. Several components, each containing the contribution of all 35 genes, were identified; of these, the fourth component was predictive for achievement of complete remission (CR) (p=0.009, B(exp)=2.47). Of all genes, the anti-apoptotic gene XIAP had the highest impact on this component. Its negative impact on CR rate in AML is in agreement with other reports. Most importantly, 33/35 genes had a positive impact on the main component. This component turned out to be highly associated with the median expression of all 35 genes present in each sample (p=0.0005, R=0.664). In turn, high median gene expression was predictive for poor OS (N=120, p=0.01). In agreement with our hypothesis this points to activation/deregulation of the complete apoptosis pathways instead of deficiencies in individual gene expression. To confirm this observation we have cross-validated our MLPA findings using a published micro-array dataset (N Engl J Med 2004; 350:16
). Fifteen patients were present in both datasets and showed positive correlation in 21/24 parallel gene sequences. With the exclusion of these 15 patients, in the array data too, both increased pro and anti apoptotic gene expression correlated with poor prognosis. For improvement of predictive strength of gene expression on patient survival, we have maximized the number of genes under analysis (N=33). To increase the distinctive power of each gene, we dichotomized our data based on arbitrary thresholds (10%, 50%, 90%) in the ranges of individual gene expression. This pathway approach showed in both methods that expression of multiple genes above the 90% threshold is associated with short survival. For MLPA this was also found at lower thresholds of 50% or 10%. In conclusion, with the use of the novel and very sensitive RT-MLPA technique we have identified general deregulation of the apoptosis cascade as a cause of treatment failure in AML.
