To the Editor:

Acute myeloid leukemia (AML), an aggressive hematological malignancy, is frequently associated with mutations in genes encoding enzymes that regulate epigenetic patterns. Additionally, AML is associated with genomic rearrangements, including translocations [1]. Structural rearrangements can activate oncogenes by repositioning cis-regulatory elements into the vicinity of genes, normally not expressed in a certain cell lineage, a mechanism termed enhancer hijacking [2]. For example, in the AML-M4 cell line GDM-1, the oncogenic transcription factor Motor Neuron and Pancreas Homeobox 1 (MNX1) [3, 4] is aberrantly activated through enhancer hijacking due to a translocation t(6;7) which juxtaposes MNX1 on chromosome 7 (chr7) with enhancers from the MYB-AHI1 region on chr6 [3,4,5]. Similarly, MNX1 activation mainly by enhancer hijacking involving various hematopoietic enhancers is a rare but recurrent event and occurs in 1.4% of primary AML [6, 7]. In this study, we hypothesized that therapeutic downregulation of MNX1 could be achieved by an epigenetic compound disrupting the promoter-enhancer interaction. We first demonstrated with two MNX1-targeting shRNAs that MNX1 downregulation (Supplementary Fig. 1A) reduces GDM-1 viability (Supplementary Fig. 1B). This confirms a previous finding and suggests that MNX1 activation is required for leukemic growth [3]. To identify compounds that can reduce MNX1 expression, we screened a focused compound library of 174 epigenetic tool compounds including approved drugs in GDM-1 cells (Supplementary Table 1a,b). Twenty-one compounds reduced viability by approximately 70% (Fig. 1A). Based on their diverse modes of action, effectiveness, and commercial availability, we selected ten compounds mediating reduced viability and determined the IC80 concentrations. Of these compounds, only 5-aza-2’-deoxycitidine (decitabine, DAC), a DNA methyltransferase (DNMT) inhibitor in clinical use in myelodysplastic syndrome and other hematological malignancies including AML [1], significantly suppressed MNX1 (Fig. 1B). The reduction of viability by DAC treatment was more pronounced in GDM-1 as compared to other AML cell lines (HL-60, MOLM-13, OCI-AML-3) which do not express MNX1 (Supplementary Fig. 1C), indicating a lower sensitivity of these lines to DAC and suggesting a positive effect of MNX1 on AML cell viability in GDM-1. In addition to reduced cell viability, DAC treatment resulted in the activation of cancer testis antigens including MAGEA3, MAGEB2, MAGEB1, FMR1NB, PAGE2B and MAEL, potentially through DNA hypomethylation. On the other hand, MNX1 transcription (Fig. 1C), and protein levels (Fig. 1D, E) were significantly reduced upon DAC treatment. This suggested the presence of a negative regulator of MNX1 that becomes activated upon treatment. Gene ontology terms for upregulated genes were enriched for cell cycle, DNA replication & DNA repair-related processes, or tissue-specific/developmental processes, respectively (Supplementary Table 5A, B).

Fig. 1: Suppression of MNX1 expression mediated by 5-aza-2’-deoxycitidine (DAC).
figure 1

A Cell viability after treatment with the 50 most effective epigenetic compounds (refer to Supplementary Table 1). Green bars refer to compounds selected for further analysis, number 17 refers to DAC. B MNX1 mRNA expression was measured by qRT-PCR after treatment with ten selected compounds. The DMSO treatment was used for normalization. Quantification was performed using 2-Δ(Ct) method with GAPDH Ct-values, and the relative expression values from the compound treated and DMSO treated samples were divided by the average relative DMSO expression. These calculations were repeated for each compound. Expression after DAC treatment was reduced to 0.55-fold (Two-tailed T test. *p < 0.05) C Volcano plot showing transcriptional changes after DAC treatment. Red dots indicate significantly up- or downregulated transcripts. Genes with a false discovery rate (FDR)-adjusted p value (based on DESeq2) of less than 0.05 and an absolute (log2(fold change)) of greater than 1 were classified as significant. MNX1 and cancer testis antigens are highlighted. D Representative Western blot of MNX1 protein levels from GDM-1 cells extracted 120 h after DAC treatment. E Quantification of Western Blot in D. The p values come from a one-tailed t test.

Full size image

Since miRNAs are known to suppress MNX1 [8, 9], we performed miRNA-seq on DAC-treated GDM-1 cells and DMSO controls and used three prediction tools to search for miRNAs that become upregulated and could target the MNX1 3ʹ untranslated region (3ʹ UTR). Six miRNAs, predicted to bind to MNX1 3’ UTR by at least two of three prediction tools, were upregulated upon DAC treatment (Fig. 2A). Upregulation of three miRNAs, miR-381-3p, miR-410-3p, and miR-200a-3p, which were not expressed in DMSO-treated cells, was confirmed by qRT-PCR (Supplementary Fig. 2A, Supplementary Table 2). When we expressed miRNA mimics corresponding to these three miRNAs, we observed a significant downregulation of MNX1 only with the miR-200a-3p mimic (Fig. 2B), reaching approximately the same expression level as seen after DAC treatment. The strong reduction in MNX1 protein levels upon DAC treatment may indicate the existence of additional DAC-induced, yet unknown, miRNAs that may control MNX1 protein translation. As an indirect proof of the interaction between miR-200a-3p and the targeted MNX1 3’UTR and, hence, downregulation of MNX1 by the miRNA, we quantitated fluorescence signals in HEK293T cells transfected with the miR-200a-3p mimic and luciferase constructs bearing either a wildtype or a mutated binding site of the MNX1 3’UTR (Supplementary Fig. 2B). We observed a significant signal reduction in cells with the wildtype but not in cells with the mutated construct (Fig. 2C). We inferred that the miR-200a-3p mimic binds to the MNX1 3’UTR and leads to the degradation of the luciferase transcript, suggesting that MNX1 is downregulated by miR-200a-3p in GDM-1. Since the miR-200 family is activated by DNA demethylation of its transcription start site (TSS) in breast and bladder cancer [10, 11], we determined methylation levels in the CpG-rich TSS region of miR-200a-3p upon DAC treatment. We observed significant hypomethylation only on the 4th amplicon (Supplementary Fig. 2C), suggesting that DAC-induced promoter demethylation led to upregulation of miR-200a-3p. We also observed that overexpression of other miRNAs (miR-381-3p and miR-410-3p), which are found to be significantly upregulated upon DAC treatment, does not lead to reduction of MNX1 protein levels (Supplementary Fig. 2D).

Fig. 2: Activated miR-200a-3p suppresses MNX1 in GDM-1 and MNX1 expressing PDX samples.
figure 2

A Volcano plot of miRNA-seq expression data from GDM-1 cells treated with DAC. Genes with a false discovery rate (FDR)-adjusted p-value (based on DESeq2) of less than 0.05 and an absolute (log2(fold change)) of greater than 1 were classified as significant. B Representative Western blot (left) and quantification (right) showing MNX1 protein expression in GDM-1 cells upon treatment with miRNA-200a-3p mimic (miRNA-200a-3p) or scrambled control (miR-NC). Each of the three biological replicates was normalized to its specific control experiment. C Luciferase signal intensity in HEK293 cells with mimics for miR-200a-3p and negative control (used for normalization for each of the replicated separately), NC. D Quantification of MNX1 transcription in PDX491 and PDX661 after DAC treatment through qRT-PCR. The mean of the DMSO replicates was used for normalizing DAC-treated PDX samples. E Representative Western blot (left) and quantification (right) showing reduced MNX1 protein levels in DAC-treated PDX491 and PDX661. F Log2 of miR-200a-3p expression levels in DAC-treated PDX491 and PDX661. B–F The p values were obtained from one-tailed t-tests.

Full size image

Patient-derived xenografts (PDX) in mice enable to study cancer development in vivo. Previously, we demonstrated in an AML PDX mouse model with del(7q) and MNX1 expression that shRNA-mediated knockdown of MNX1 reduces the tumor load of AML PDX cells in vivo [6]. Here, we investigated the effect of DAC treatment on MNX1 regulation in two AML-derived PDX lines ex vivo, PDX491 and PDX661, which show aberrant MNX1 expression due to enhancer hijacking. As in the AML cell line GDM-1, we observed a significant reduction of MNX1 expression in both PDX lines (Fig. 2D, E). Moreover, DAC also induced upregulation of miR-200a-3p, supporting its role in MNX1 downregulation (Fig. 2F).

In summary, we demonstrated that treatment of GDM-1 with DAC leads to miR-200a-3p activation, which, in turn, suppresses MNX1 and reduces cell viability. Our results may have important implications for the clinical management of AML cases with ectopic MNX1 expression including infants with t(7;12) [12] and in elderly patients with del(7q) [6]. To enforce DAC treatment and advance in MNX1 overexpressing AML subtypes, combination therapies with other epigenetic compounds such as those that might disrupt the enhancer-promoter interaction activating MNX1 are warranted. Possible candidates are bromodomain and extraterminal (BET) protein inhibitors, where multiple clinically tested chemotypes have been previously developed [13] including JQ1, which was reported to reduce oncogenic expression by interference with enhancer-promoter interaction [14, 15].

A limitation of our study is the focus on miR-200a-3p as the mechanism mediating MNX1 downregulation upon DAC treatment. However, there may be other mechanisms at work as well. Future studies are needed to validate our findings in animal models.