EVI1 expression in childhood acute lymphoblastic leukaemia is not restricted to MLL and BCR/ABL rearrangements and is influenced by age

EVI1 is a transcriptional regulator with an important function in haematopoiesis and self-renewal. Aberrant overexpression of EVI1 has been firmly established as one of the most adverse prognostic markers in acute myeloid leukaemia (AML), implying that EVI1 is one of the most aggressive oncogenes in AML. Importantly, a recent report in Leukemia from Konantz et al. suggests that EVI1 might also have a role in paediatric acute lymphoblastic leukaemia (ALL), where high expression confers apoptosis resistance, and possibly also an adverse prognosis. Rearrangements of the 3q26 region, which encompasses the MECOM (MDS–EVI1 complex) gene that encodes EVI1 transcripts and that are commonly associated with EVI1 overexpression in adult AML, rarely occur in childhood ALL or AML. However, when EVI1 is expressed in childhood AML it seems to be predominate in MLL-rearranged cases, which may then confer an adverse prognosis, as illustrated by the correlation with the t(6;11) subtype, and with complex karyotype cases. However, in general high EVI1 expression is not seen in AML with good prognosis cytogenetics such as core-binding factor-rearranged AML with t(8;21) or inv(16) (Figure 1a). In addition, in chronic myeloid leukaemia, the BCR–ABL fusion tyrosine kinase sustains EVI1 expression. To complement the work reported by Konantz et al., we analysed gene expression data generated from nucleated cells obtained from diagnostic bone marrow aspirates of 70 de novo ALL (31 female subjects, median age at diagnosis: 4.4 years, range: 1.1–14.6 years), using the Affymetrix U133þ 2 platform from our previously published and verified data set. We analysed nine T-cell ALL, three BCR–ABL-rearranged Philadelphia-positive, one MLL-rearranged and 57 ALL with other cytogenetics, of which most were hyperdiploid. We focussed on EVI1 transcripts targeted by probes 226420_at and 221884_at, which detect exonic sequences of EVI1 transcripts. We also included two large and sufficiently annotated gene expression data sets in our analysis, together comprising 455 de novo ALL samples (GEO Data sets GES28497 and GES13425). To determine whether high EVI1 expression in ALL is also associated with specific chromosomal changes involving BCR/ABL or MLL rearrangements, or T-cell phenotype, we divided de novo childhood ALL into four subgroups. We analysed T-cell ALL separately from BCR–ABL fusion (Philadelphia)-positive, MLL gene rearrangement-positive and ALL with other cytogenetic abnormalities (the largest group). In addition, and separately, we also analysed paired data sets of de novo and subsequently relapsed childhood ALL (GSE28460). EVI1 expression is variable in de novo paediatric ALL. However, the range of EVI1 expression levels is much smaller as compared with childhood AML (Figure 1a). When applying the criteria used for paediatric AML with samples considered EVI1-positive (þ ) when EVI1 expression is higher than log0.5 normalised fold change, the proportion of EVI1þ ALLs is larger than AML and includes cases of T-cell ALL. There was no clear association of higher EVI1 expression levels with BCR–ABL or MLL rearrangements in the analysis of our own samples and results of published data sets (Figure 1a). The variability of EVI1 expression in the MLLrearranged ALL group might be partly dependent on specific MLL gene rearrangements, as AMLs with MLL/AF6 and MLL/AF9 fusions are associated in particular with high EVI1 expression in adults and children. We found no obvious possible cause for variability of EVI1 expression in ALL of other cytogenetic subgroups. Given the rarity of 3q rearrangements in ALL, high EVI1 expression is likely to be a secondary event. Cytogenetic features of high EVI1expressing ALL, which include hyperdiploid and normal karyotype disease, are listed in Supplementary Table 1. Konatz et al. observed that EVI1 modulated the expression of apoptosis-related genes in paediatric ALL. In accordance with their data, we identified BCLX (fold change 1.8, P1⁄4 0.007) and PUMA (FC1⁄4 1.9, P1⁄4 0.005) in T-cell ALL when determining EVI1-coregulated genes by selecting for significantly changed (Po0.01, analysis of variance) low and high expressed genes in paediatric Tand B-cell ALL with EVI1 expression 4log1.5. To determine the overlap in global expression patterns associated with high EVI1 transcripts, we chose a 4log1.5 cutoff to analyse a meaningful sample size for comparison. In general, the overlap was greater between T-cell ALL and B-cell ALL than either of these groups with AML (Supplementary Information, Figure 1). Only 12 genes were significantly co-regulated in all the subgroups of childhood leukaemia (Po0.01, analysis of variance) (Supplementary Information, Figure 1), with seven regulated in different directions between the subgroups. This included SMARCA5, of which the encoded protein recently has been shown to directly interact with the EVI1 protein in SKOV ovarian cancer cells and K562 leukaemia cells. Importantly, we noted that several of the high EVI1-expressing ALLs were from patients in late childhood or adolescence. To further explore a possible association of EVI1 expression with age in de novo childhood ALL, we carried out rank regression analysis by age on our B-cell ALL cases, which is the largest group of our ageand sex-annotated data set (n1⁄4 51), excluding those with MLL or BCR–ABL rearrangements. We found a highly significant increase in EVI1 expression with age at diagnosis of ALL (Figure 1bi, upper panel) during childhood years (age 1–10, r1⁄4 0.29, P1⁄4 0.05) but in particular obvious with the onset of adolescence (age 1–14, r1⁄4 0.53, Po0.003). To further explore the relationship of patient age on gene expression patterns in childhood ALL we applied this approach to the entire data set. We identified 415 probe sets corresponding to 341 genes with significant ageassociated changes in expression levels (P1⁄4 0.01) (Figure 1bii, lower panel). Of these, 130 genes have the highest expression in adolescence (cluster 3). When we associated the upstream regulators of these 130 genes (Ingenuity pathway analysis), we found that 21 are regulated directly by transforming growth factor-b (P1⁄4 1.8 10 ). This resembles our findings with respect to the impact of age on gene expression patterns in general and implies that transforming growth factor-b might in particular Citation: Blood Cancer Journal (2014) 4, e179; doi:10.1038/bcj.2013.76 & 2014 Macmillan Publishers Limited All rights reserved 2044-5385/14