A Novel CD34/ETO2/Ifngr gene Regulatory Axis Is Implicated in Poor-Prognosis Cases of t(8;21) AML

Background:Roughly 15% of AML cases are derived from the t(8;21) translocation, a molecular event that results in the production of the AML1-ETO (AE) fusion transcription factor. This aberrant protein induces a broad dysregulation of the transcriptome and causes expansion of leukemia stem cells (LSCs), which predisposes clones to an increased risk of second-hit mutations. In isolation, without the presence of type I mutations, the t(8;21) mark is considered to be one of favorable prognosis. However, a number of LSC characteristics have been shown to influence patient survival, including expression levels of the cell surface glycoprotein, CD34, and of the truncated isoform of AE, AML1-ETO9a (AE9a). Despite these findings, a unified model of how these factors interact to further the progression of t(8;21) AML has not been determined. Here, we use a combination of in vitro assays and bioinformatic analyses of patient datasets with the hopes of identifying an underlying connection between independent, prognostic factors.

Results: We first performed gene knockdown studies with shRNA constructs (against AE) in a pKLO.1 backbone, which was transduced into the AML-model cell line Kasumi-1. Subsequent RT-PCR showed a large decrease in CD34 expression and, serving as a control, an increase in RASSF2, a cytoplasmic signaling protein normally repressed by AE. This led us to perform gene overexpression of cDNA encoding AE or AE9a into CD34-enriched cord blood HSCs. Interestingly, we were able to show that both AE and AE9a increased the expression of CD34, and AE9a was able to do so to a greater extent. This prompted us to analyze previous ChIP-seq data from Kasumi-1 cells, and we discovered high levels of AE binding to an upstream enhancer of CD34. While this site was free of any E-box binding motifs, we nonetheless observed binding peaks of the E-protein HEB, and further gene knock-down studies showed that HEB recruitment to this site depended on AE and, in turn, CD34 expression depended on HEB recruitment. Next, we used differential gene expression analyses to determine the CD34-coexpressed genes from a publically available patient microarray dataset (GSE14468, n=516). CD34-high patient samples showed an accompanying enrichment of JUP, KIT, CD133, HEB, E2A and ETO2, and a decrease in immune signaling molecules such as IFNGR1 and IFNGR2. ETO2 is the only ETO family member significantly expressed in HSCs and is an important corepressor of E-proteins HEB and E2A; it is thought to help repress the pro-differentiation and anti-proliferation effects of E-proteins. ETO2, being an ETO family member, is also able to oligomerize with other ETO family members, such as ETO, and thus may play a role in modifying the functions of the AE-transcriptional complex. Intriguingly, we have evidence that AE and AE9a repress ETO2, although AE9a does so to a lessened extent. We then performed Kaplan-Meier analyses on subgroups of non-CBF AML and t(8;21) AML cases. Unsurprisingly, we found that both groups experienced lower event-free survival when CD34 and ETO2 were high (p=.028 and p=.040), and the t(8;21) group when IFNGR was low (p=.014). Additionally, although CD34 expression was higher in t(8;21) patients compared to non-CBF AML (p=.0027), ETO2 expression was lower (p=.025). If both subgroups were further divided into CD34-high and CD34-low groups, ETO2 expression was, in both cases, higher in the CD34-high patients and lower in the CD34-low patients, suggesting a possible regulatory link between these two genes. Finally, we performed ETO2 gene knockdown experiments in CD34+ HSCs and observed an increase in IFNGR1 and IFNGR2 expression.

Conclusions: Our findings suggest a clinically relevant association between CD34, ETO2, and IFNGR in both t(8;21) and non-CBF AML. We believe this regulatory axis is especially relevant in relapse likelihood and may be hijacked by LSCs to preserve their "stemness" and resistance to therapy. Our findings may suggest that AE simultaneously activates CD34 and represses ETO2, providing explanation for why full-length AE is slow to induce leukemia in mouse models and a relatively good prognostic mark in humans. Additionally, our findings suggest a loss of ETO2repression as a potential mechanism by which the AE9a isoform, and perhaps other second-hit mutations, converts LSCs to a more dangerous phenotype.

Supported by NIH grants R01HL093195-01A1 to JZ and T32 training grant T32GM008306-26A1