MDS and AML Are Closely Related Diseases: Gene Expression Patterns Reveal Clear Similarities with Respect to Karyotypes and Are Less Related to the Bone Marrow Blast Percentages.

MDS and AML are discriminated by percentages of blasts in the bone marrow (BM) according to the FAB as well as to the WHO classification. However, thresholds are arbitrary and demonstrate only a limited reproducibility in interlaboratory testings. Thus, other parameters have been assessed to discriminate these entities with respect to diagnosis and prognosis. In particular, in the majority of cases common karyotype aberrations have been observed between MDS and AML which have a higher prognostic impact than blast percentages. We applied gene expression profiling (U133A+B, Affymetrix) in 70 MDS and 238 AML cases. In accordance with the WHO classification we excluded cases with balanced translocations (i.e. t(8;21), t(15;17), inv(16), or 11q23) which are classified as AML irrespective of BM blast percentage. First we aimed at identifying genes of which the expression correlated to blast count (Spearman correlation). Out of the top 50 genes this analysis revealed only the FLT3 gene which showed a higher expression in cases with high blast count, while 12 genes with a higher expression in cases with lower blast counts were identified (ANXA3, ARG1, CAMP, CD24, CEACAM1, CEACAM6, CEACAM8, CRISP3, KIAA0922, LCN2, MMP9, STOM). Most of the latter genes are expressed in mature granulocytes and are involved in differentiation and apoptosis. In a second step we performed class prediction using support vector machines (SVM) to separate MDS and AML according to blast percentages as defined in the WHO classification (<5%: RA and 5q- syndrome; 5–9%: RAEB-1; 10–19%: RAEB-2; >19% AML). Using 10-fold cross validation and support vector machines the overall prediction accuracy was only 80%. In detail, 230/238 AML cases were correctly assigned to the AML group while 8 cases were classified as MDS RAEB-2. However, none of the RA, 5q- syndrome and RAEB-1 cases were correctly assigned to their groups, respectively, but were either classified as AML or RAEB-2. Furthermore, only 16 of 38 RAEB-2 cases were correctly predicted, while the 20 remaining cases were assigned to the AML group. Thus, no clear gene expression patterns were identified which correlated with AML and MDS subtypes according to WHO classification. Taking the common genetic background observed in MDS and AML into account, both entities were categorized in a third step according to cytogenetics and classified based on their gene expression profiles. In order to assess the impact of the common genetic background, the largest cytogenetically defined subgroups were compared to each other, i.e. AML and MDS with normal karyotype and with complex aberrant karyotype. Intriguingly, while correct classification of AML or MDS was found in 91%, classification into the correct cytogenetic groups was achieved in 95%. Consequently, all cases were devided into the two groups, complex aberrant karyotype (n=60) and other or no aberrations (n=248) irrespective of AML or MDS. A classification into these groups also yielded an accuracy of 93%. Our data suggests that gene expression profiling reveales the biology of MDS or AML to highly correlate with cytogenetics and less with the percentages of BM blasts. These results strengthen the need for a revision of the current MDS and AML classification centering now genetic abnormalities, which may also be used for clinical decisions.