Powered By Bing

Supplementary Information

Gene Expression Profiling of Pediatric Acute Myelogenous Leukemia

Section III: Genetic Subtype Discriminating Genes

Figure S2. Heirarchical clustering with class discriminating genes

Figure 2

Figure S2. Characteristics of AML cases clustered in Figure 2 using the top 50 class discriminating genes. The dendogram generated in Figure 2 is reproduced in this figure, with the sample number, genetic lesion, karyotype, and FAB morphology indicated next to each case.

Close examination of the dendogram shown in figure S2 above suggests the existence of several distinct subgroups within the cases with MLL chimeric fusion genes. Clustered within one of the major branches of the dendogram are 80% of the MLL cases along with two non-MLL cases (Figure 2 and Figure 2S). The cases within this cluster are characterized by FAB-M5a morphology (90%) and include all t(9;11) and t(10;11) cases along with several other rarer MLL translocations. The two non-MLL cases each have M5a morphology. The remaining 4 MLL cases cluster together on a separate branch along with one non-MLL case. Within this cluster are both t(11;19) cases along with a t(6;11) and t(1;11). The single non-MLL case within this cluster had M5a morphology. Interestingly, the 4 non-MLL cases that separate the two major MLL clusters include two cases with a t(16;21)(p11.2;q22). The molecular targets of this translocation encode FUS-ERG 4,5. Collectively, these data suggest that the variation in gene expression signatures among cases with MLL chimeric fusion genes is likely to result not only from differences in the nature of the MLL translocation, but also from differences in the extent of differentiation and the range of secondary mutations required for induction of a full leukemic phenotype. Although two cases lacking MLL chimeric fusion genes tightly clustered with the MLL cases, their expression signatures differed enough from true MLL chimeric fusion gene cases so that they were appropriately classified using supervised learning algorithms.

Similar to MLL cases, two subgroups of CBFβ-MYH11 expressing leukemias were suggested by this analysis, along with a group of five “other” cases that lacked evidence of CBFβ-MYH11 but nevertheless clustered with these cases (Figure 2 and Figure 2S). Within this extended cluster the majority (90%) of the cases had FAB-M4/M4E/M5 morphology, including the five non-CBFβ-MYH11 cases. One group of the CBFβ-MYH11 cases (6/14), expressed a set of genes that were also overexpressed in the five “other” cases, and in a subset of cases with MLL chimeric fusion genes (Figure 2), whereas the other subgroup of CBFβ-MYH11 cases (8/14 cases), expressed these genes at a lower level. Moreover, the later cases also expressed a small set of genes that were overexpressed in AML1-ETO leukemias (Figure 2). A careful analysis of the morphology and cytogenetics of the CBFβ-MYH11 expressing cases failed to reveal any significant differences between these two potential subgroups. Because of the marked variability in genomic breakpoint location in both the CBFβ and MYH11 genes, we did not try to correlate these features with the observed pattern of clustering in this relatively small cohort of cases. Thus, the underlying reason for the observed heterogeneity in CBFβ-MYH11 expressing cases remains unknown.