Complex 3q26/EVI1 Rearrangements Genocopy Inv(3)/t(3;3) Acute Myeloid Leukemias By Enhancer Hijacking, EVI1 Overexpression, Absent MDS1-EVI1 and Low GATA2 Expression

Introduction

Acute myeloid leukemia (AML) with inv(3)(q21q26) or t(3;3)(q21;q26) overexpress EVI1 and have a very poor prognosis. EVI1 is part of the MECOM (MDS1-EVI1-Combination) locus from which MDS1-EVI1 and EVI1 can be transcribed from two different promoters. Although EVI1 is expressed at high levels, MDS1-EVI1 is absent or expressed at very low levels in inv(3)/t(3;3)-AMLs. Aberrant EVI1 expression in these leukemias is driven by the long-distant GATA2 enhancer, translocated from 3q21 to EVI1 at 3q26 (Gröschel et al, 2014). As a result of this GATA2 enhancer hijack by EVI1, GATA2 is switched off on the rearranged allele, resulting in mono-allelic and low GATA2 expression. We hypothesize that leukemic transformation of inv(3)/t(3;3)-AMLs is driven by EVI1 overexpression and by low GATA2 and that these leukemias are marked by the absence of MDS1-EVI1 expression. We previously reported about a group of AML patients that presented with complex rearrangements of 3q26 (refer to as variant-3q26-AML) with frequent MECOM involvement and very poor survival (Lugthart et al, 2010). Here we address the questions if these variant-3q26-AMLs 1) overexpress EVI1 by enhancer hijacking, 2) are marked by absent MDS1-EVI1 and 3) express low levels of GATA2. Accordingly, the variant-3q26-AMLs should be classified as inv(3)/t(3;3)-AMLs.

Results

We identified 37 variant-3q26-AMLs with MECOM rearrangement as determined by Fluorescent in-situ hybridization (FISH). RNA-seq of these AMLs revealed EVI1 overexpression but also demonstrated the absence of MDS1-EVI1 in 90% of patient samples. Applying 3q-capture DNA-seq, we found that in 2 cases the patient cells harboured a "hidden" inv(3)(q21q26) with involvement of the GATA2 enhancer. In 7 cases recurrent 3q26/EVI1 translocations were identified, e.g. t(2;3)(p21;q26), t(3;8)(q26;q24), t(3;7)(q26;q11), involving the THADA, MYC or CDK6 loci respectively as previously described. Interestingly, we identified new translocations to the EVI1 locus in 13 AMLs, including a t(3;6)(q26;q21) and a t(3;4)(q26;p15), involving the CD164, and PROM1 loci respectively. In these samples we find clearly skewed expression of these genes to one allele, suggestively caused by the rearrangement and enhancer hijacking. CD164 plays a key role in adhesion, proliferation and migration of CD34+ hematopoietic progenitor cells (Watt et al, 2000). PROM1 (CD133) is expressed in human hematopoietic stem and progenitor cells and is thought to be involved in maintaining stem cell properties by suppressing differentiation (Bauer et al, 2008). We argue that EVI1 overexpression in these variant-3q26-AMLs is driven by hijacking enhancers of genes that are normally active in myeloid progenitors.

In most of the patients the translocation breakpoints are in between the promoters of MDS1 and EVI1, explaining absence of MDS1-EVI1 expression. In addition, analysis of SNP-array data of these patients (N=33) showed Copy Number Loss (CNL) of the MDS1 exon(s) and not the EVI1 exons in at least 5 cases. Together these data suggest the importance of MDS1-EVI1 loss in 3q26-AMLs.

Furthermore we wondered whether low GATA2 expression is an important event in variant-3q26-AMLs. Similar to inv(3)/t(3;3)-AMLs (Gröschel et al, 2014), RNA-seq revealed that the GATA2 expression was on average a two-fold lower in the variant-3q26-AMLs (N=37), compared to non-3q26 rearranged AMLs (N=114). Surprisingly, SNP-array analysis in 26 variant-3q26-AMLs revealed CNLs of GATA2 and/or its enhancer in 7 patients. Detailed SNP analysis in GATA2 exons by combined 3q-capture DNA-seq and RNA-seq uncovered another 7 cases with mono-allelic GATA2 expression or skewing to expression of one allele (allele_freq<0.4, P<0.05). Hence, in 53% of these patient samples one of the GATA2 alleles appeared to be affected. These results are unexpected, as the GATA2 locus was not involved in any of these rearrangements.

Conclusion

Given their complex karyotype, variant-3q26-AMLs are often not recognised as 3q26/EVI1 AMLs. Although the exact mechanism remains elusive, the overall effect seems to be alike. EVI1 overexpression, potentially driven by enhancer hijacking of genes that are active in early myeloid progenitors, combined with absent MDS1-EVI1 and mono-allelic/low GATA2 expression results in AML with very poor survival. Given these data we believe variant-3q26-AMLs genocopy inv(3)/t(3;3)-AMLs and should be classified as such.