The m⁶A RNA modification in AML

Abstract

Purpose of review

In recent years, the N⁶-methyladenosine (m⁶A) modification of RNA has been shown to play an important role in the development of acute myeloid leukemia (AML) and the maintenance of leukemic stem cells (LSCs). In this review we summarise the recent findings in the field of epitranscriptomics related to m⁶A and its relevance in AML.

Recent findings

Recent studies have focused on the role of m⁶A regulators in the development of AML and their potential as translational targets. The writer METTL3 and its binding partner METTL14, as well as the reader YTHDF2, were shown to be vital for LSC survival, and their loss has detrimental effects on AML cells. Similar observations were made with the demethylases FTO and ALKBH5. Of importance, loss of any of these genes has little to no effect on normal haemopoietic stem cells, suggesting therapeutic potential.

Summary

The field of epitranscriptomics is still in its infancy and the importance of m⁶A and other RNA-modifications in AML will only come into sharper focus. The development of therapeutics targeting RNA-modifying enzymes may open up new avenues for treatment of such malignancies.

1. Introduction

RNA is amenable to nucleotide modifications that influence aspects of its metabolism such as stability, exportability, localisation and translation[1]. Over 170 chemically distinct epimarks have been identified on RNA species, representing a dynamic mechanism of gene regulation at the epitranscriptomic level[2]. The selective methylation of adenosine residues at position N(6) (referred to as N⁶-methyladenosine or m⁶A) constitutes the most abundant and conserved internal modification in eukaryotic mRNAs. m⁶A deposition on mRNA is catalysed either post- or co-transcriptionally by methyltransferases and associated complexes known as “writers”, is recognised by RNA-binding proteins termed “readers”, and is removed by demethylating enzymes or “erasers”. Following the advancement of transcriptome-wide m⁶A-mapping technologies, its presence has been identified on approximately a quarter of all mRNAs, with the majority decorated with only a single m⁶A despite numerous sites fulfilling the conserved motif DRACH (D=G/A/U, R=G/A, H=A/U/C), whereas a few may contain 20 or more m⁶A sites with a transcriptome-wide pronounced density near stop codons or the 3’ untranslated region (UTR)[3–5].

The main m⁶A installer is the enzyme Methyltransferase Like 3 (METTL3), which deposits m⁶A on mRNA via its S-adenosylmethionine (SAM)-binding domain and primarily in association with Methyltransferase Like 14 (METTL14), forming the heterodimeric m⁶A-METTL Complex (MAC)[6, 7]. This enzymatic core interacts with a more extensive multicomponent regulatory complex, termed the m⁶A-METTL-associated complex (MACOM), comprising several proteins including Wilms Tumor 1 Associated Protein (WTAP), Vir Like m⁶A Methyltransferase Associated (VIRMA or KIAA1429), Cbl Proto-Oncogene Like 1 (CBLL1 or Hakai), RNA Binding Motif Protein 15 (RBM15), and Zinc Finger CCCH-Type Containing 13 (ZC3H13)[8]. The fate of m⁶A modified transcripts is mediated by readers, including the YT512-B Homology (YTH) domain-containing protein family (YTHDC1/2, YTHDF1/2/3,) and Insulin Like Growth Factor 2 mRNA Binding Protein family (IGF2BP1/2/3)[9, 10], the Fragile X Mental Retardation Protein (FMRP), the Eukaryotic Translation Initiation Factor 3 subunit h (eIF3h) and members of the Heterogeneous Nuclear Ribonuclease (HNRNP) family (HNRNPA2B1, HNRNPC, HNRNPG)¹¹. The reversibility of m⁶A modifications is attributed to two demethylases proposed as erasers of m⁶A, the Fat Mass and Obesity Associated protein (FTO) and AlkB Homologue 5 RNA Demethylase (ALKBH5).

In acute myeloid leukemia (AML), genetic aberrations and molecular imbalances accrued in myeloid progenitors or blasts lead to their uncontrolled clonal expansion and accumulation in the bone marrow. The cancer is maintained by leukemic stem cells (LSCs), a portion of which also serves as the persisting reservoir in resistance to current treatment or disease relapse frequently observed in patients. The role of m⁶A in driving or maintaining the leukemic state has recently started to be unravelled[11]. Importantly, most of the MACOM subunits represent genetic vulnerabilities of AML cells as indicated by CRISPR loss-of-function screens[12, 13]. Multiple studies have shown that METTL3 depletion in mammalian cells leads to an acute deficit of m⁶A on polyadenylated RNA in line with the enzyme being the primary mRNA m⁶A installer, and the paramount importance of specific and balanced dosage of this modification on RNA in normal development and homeostasis has been confirmed for several physiological processes including haematopoiesis[14–16].

Here we employ an outline of m⁶A writers, readers and erasers to summarize the role this modification has only recently been shown to play in the pathogenesis and maintenance of AML.

2. m⁶A Writers

a. METTL3

In experiments involving human cord blood-derived CD34+ hematopoietic stem/progenitor cells (CB-HSPCs), m⁶A loss by METTL3 knock-down (KD) was shown to promote terminal myeloid differentiation in vitro while decreasing cell proliferation. In contrast, overexpression of wild-type METTL3 inhibits myeloid differentiation and promotes cell growth, an effect which is not phenocopied by a catalytically inactive form of the enzyme[15]. Importantly, METTL3 mRNA and protein levels are significantly more abundant in human patient-derived AML cells than in primary CB-HSPCs. METTL3 depletion in AML cell lines induces differentiation and apoptosis, with in vivo experiments showing delayed AML progression in recipient mice[15]. These effects are traced to increased m⁶A which alters the translation of oncogenic c-MYC, BCL2 and PTEN mRNAs in AML models[15]. Cell differentiation upon METTL3 depletion appears to be orchestrated by the phosphatidylinositol 3-kinase (PI3K)/AKT pathway[15]. In an independent study, we demonstrated that, in AML, METTL3 installs m⁶A co-transcriptionally in a METTL14-independent fashion. The gene essentiality of METTL3 was validated by demonstrating that its disruption by either knock-out (KO) or KD reverses the myeloid differentiation block in AML cells[16]. Ectopic expression of METTL3 rescued the leukemic self-renewing phenotype, while a catalytically inactive mutant failed to recapitulate this result[16]. Engraftment of human leukemic cells into immunocompromised mice was also markedly impaired following targeting of METTL3, leading to significantly longer survival[16].

We identified METTL3 and METTL14 associating with chromatin fractions in the AML cell line MOLM13, and we observed them binding to transcriptional start sites (TSSs) of different coding genes, with the CCAAT-box specifically enriched under METTL3 peaks[16]. We further identified CCAAT-binding factor CEBPZ as the recruiter of METTL3 to these TSSs[16]. Contrary to other m⁶A-containing mRNAs which generally exhibited enhanced transcription upon METTL3 depletion, the transcripts from METTL3-bound genes were translated less efficiently. Mapping of ribosomal pausing sites on the pertinent mRNAs indicated stalling at GAN codons in METTL3-depleted cells[16]. One oncogenic driver requiring METTL3-mediated m⁶A for enhanced translation is SP1 which itself regulates c-MYC expression and which is also shown to be a dependency of AML[16]. Although METTL3 depletion has been reported to compromise normal haematopoiesis[15], the enzyme’s essentiality in AML maintenance makes it a promising potential therapeutic target especially because inhibition by small molecule is not expected to achieve global reduction of METTL3 levels similar to gene deletion; rather, the response to a small molecule can be both dosage-dependent and reversible, with minimum or no effects on normal haematopoietic cells.

b. METTL14

While previously believed to possess weak methyltransferase activity, METTL14 is now thought to mainly act as an RNA-binding scaffold and allosteric activator to METTL3[17]. METTL14 is highly expressed in CB-HSPCs and is downregulated upon myeloid differentiation[17]. Importantly, when compared to healthy donor cells, METTL14 levels are higher in primary AML samples and in AML cell lines, while also being higher in AML than in non-haematological cancers[17]. These results are indicative of METTL14’s role in maintaining the self-renewing capacity of LSCs. METTL14 is indeed essential for the preservation of HSPCs and AML’s proliferative capacity as its silencing triggers myeloid differentiation, with its expression seemingly maintained by MLL-rearranged chimeras and diminished by differentiation-inducing agents (e.g. all-trans retinoic acid, ATRA)[17].

More specifically, Mettl14 KO/KD disrupts MLL/AF9-induced cell immortalization in vitro, whereas ectopic expression of wild-type, but not mutant, protein increases transcriptomewide m⁶A thereby promoting MLL/AF9-mediated immortalization[17]. The LSC frequency is also shown to be lower in Mettl14 KO murine cells vs controls, leading to delayed leukemia onset and prolonged survival in vivo[17]. Mechanistically, SPI1 binds the METTL14 TSS and is negatively correlated with METTL14 expression; its silencing leads to an increase in METTL14 mRNA and protein levels in both HSPCs and human AML cells whereas its overexpression yields the reverse[17]. METTL14 expression positively modulates expression of oncogenic MYB and MYC by mediating m⁶A deposition; the latter is lost following METTL14 disruption which also produces a concomitant downregulation of the oncogenes[17]. Conversely, overexpression of METTL14 increases expression of both MYB and MYC, implicating this axis in leukemogenesis[17].

c. WTAP

WTAP acts as a binding partner for MAC and MACOM protein elements and helps direct the complex at nuclear speckles[18]. Similarly to METTL3 and METTL14, WTAP is overexpressed in AML samples and AML cell lines compared to normal mononuclear cells. WTAP is strikingly upregulated in both METTL3 KD and overexpression scenarios, indicating a precise dosagedependent regulatory mechanism of WTAP homeostasis[19]. Overexpression of METTL3 does not increase WTAP mRNA levels, however, it mediates WTAP downregulation whether catalytically active or not[19]. METTL3 KD on the other hand leads to increased levels of both WTAP mRNA and protein[19]. In the absence of METTL3, WTAP upregulation alone does not suffice in driving proliferative growth in AML cells, further narrowing its functional role to its involvement in the m⁶A methylation complex[19].

3. m⁶A Readers

a. YTH domain-containing proteins

The effects of m⁶A on RNA molecules are dependent on reader proteins the modified transcripts interact with. YTHDC1 is the only nuclear member of the protein family and localises at nuclear speckles. In the cytoplasm, YTHDF1 and YTHDF3 enhance mRNA translation, while m⁶A recognition by YTHDF2 leads to mRNA degradation[20–22]. Interestingly, YTHDC2 can both enhance mRNA translation and mediate mRNA degradation. In the context of AML, YTHDF2 is upregulated in patient-derived AML samples compared to healthy controls[23*]. The protein is essential for LSC self-renewal and maintenance of AML and whereas its depletion has no detrimental effects on HSPCs and instead causes their expansion, it compromises the ability of LSCs to expand and thus propagate AML[23*, 24]. Furthermore, YTHDF2 KO cells show upregulation of the m⁶A-modified TNFRSF1B gene (encoding TNF Receptor Superfamily Member 1B); when conditional YTHDF2 KO pre-leukemic cells were treated with TNF, they showed significant increase in apoptosis compared to controls, suggesting YTHDF2 as a potential therapeutic target[23*]. YTHDF1 and YTHDF2 have a large number of overlapping target mRNAs, suggesting a complex regulatory mechanism controlling their two distinct functions mentioned above[25]. A recent study showed that members of the YTHDF family preferentially bind m⁶A modified mRNAs, while YTHDC1 has a higher affinity for lncRNAs, suggesting a potentially distinct role in their regulation[26].

b. IGF2BP1/2/3

In contrast to YTHDF2, the members of the IGF2BP family increase the stability of modified mRNAs. However, there are conflicting reports about whether IGF2BP1/2/3 bind – and hence are direct readers of – m⁶A. Some studies suggest that by directly binding mRNA at the GG(m⁶A)C consensus sequence via their K Homology (KH) domain, IGF2BPs promote oncogenesis[9]. In contrast, other work using pull-down experiments has shown that IGF2BPs interact directly with YTHDF proteins, thus binding to the modification indirectly, via an m⁶A-binding intermediary[27]. Alternatively, m⁶A binding by IGF2BPs has also been suggested to be a result of conformation changes in mRNA caused by the modification[28].

Similarly to IGF2BPs, FMRP was also thought to interact with m⁶A[29], however, it has recently been shown to directly bind to YTHDF2[30]; thus, its role and status as an m⁶A reader remain elusive.

4. m⁶A Erasers

Conflicting reports exist about FTO’s involvement in the demethylation of m⁶A. The initial studies demonstrated that FTO was highly expressed in a subset of AMLs (e.g. those with chromosomal rearrangements or NPM1 mutations)[31]. The study then showed that through its eraser activity, FTO can regulate the expression and decrease the stability of ABS2 and RARA mRNA[31]. RARA is involved in leukemogenesis and ABS2 inhibits the growth of AML and promotes differentiation; as a result, increased levels of FTO lead to a decrease in differentiation and an increase in pro-survival signalling in AML cells. Other studies have suggested that FTO positively regulates the expression of AML oncogenes such as MYB and CEBPA[32]. FTO KD led to human AML being sensitised to treatment with ATRA, which induced LSC differentiation[31]. The potential of drug inhibition of FTO in AML has also been investigated using two different compounds. Su et al[32] reported that when treating AML cells with R-2-Hydroxyglutarate (R-2HG) the results were similar to FTO KD, and sensitized the cells to commonly used chemotherapy agents both in vitro and in vivo; additionally, the treatment with R-2HG negatively affected the ability of AML cells to proliferate and increased their differentiation. Another small molecule FTO-inhibitor, FB23-2, was later developed and exhibited results similar to genetic silencing of FTO and treatment with R-2HG – decrease of proliferation and increase of LSC differentiation, thereby stalling AML maintenance[33**]. However, a recent study has shown that FTO has a significantly higher catalytic activity when demethylating m⁶A_m, compared to m⁶A, with the rate of demethylation of the former by FTO being 100-fold higher[34]. Moreover, while the data suggesting FTO is only acting on m⁶A_m is compelling, an increase of ~20% in m⁶A is observed when FTO is depleted[31]. Despite the strong data supporting the role of FTO in AML and its potential as a therapeutic target, further studies are needed to clarify its molecular function and specificity as an m⁶A eraser.

The second m⁶A eraser, ALKBH5, is localized in the nucleus, suggesting that its demethylase activity is focused on newly formed mRNA prior to its export to the cytoplasm[35]. Unlike FTO, ALKBH5 has no activity towards m⁶A_m[34]. ALKBH5 is overexpressed in LSCs compared to normal blasts and is correlated with poor prognosis in patients. Wang and colleagues[36] observed that the ALKBH5 chromatin locus is more accessible in AML compared to healthy controls. They further investigated proteins which bind the ALKBH5 promoter region and observed enrichment of histone modifiers, including the H3K9me3 demethylase Lysine Demethylase 4C (KDM4C). KDM4C can increase chromatin accessibility and transcription factor binding to the promoter site, thus increasing the expression of ALKBH5[36]. In vitro KDM4C depletion phenocopied the depletion of ALKBH5[36].

Another study investigating the specific role of ALKBH5 in AML observed that ALKBH5 is often overexpressed in patients and this correlates with poor prognosis[37*]. Depletion of ALKBH5 in both AML cell lines and primary Lin^-CD34⁺ cells was shown to decrease the survival of the cancer and increase the apoptosis of cancer cells[37*]. Overexpression of ALKBH5 showed the reverse effect by promoting the growth of cancer cells; this phenotype was not recapitulated with catalytically inactive ALKBH5[37*]. In vivo studies using Alkbh5 KO mice showed that the lack of ALKBH5 leads to delayed onset of AML and decrease in the AML load[36, 37*]. Secondary transplantation of Alkbh5 KO cells did not lead to AML, suggesting Alkbh5 plays an important role for the maintenance of AML by controlling the self-renewal of leukemic stem cells[37*]. Further support for the role of ALKBH5 in LSC self-renewal is that, in its absence, human and mouse leukemic cells both exhibit reduced propagation after xenograft transplantation as well as diminished colony-forming potential. These studies also demonstrate that ALKBH5 is not essential for normal haematopoiesis and thus highlight its potential as a therapeutic target.

Two different explanations have been put forth to account for the action of ALKBH5. One suggested mechanism is via TACC3, an oncogene and upstream regulator of MYC[37*] and an alternative mechanism of action is through the ALKBH5-m⁶A-AXL kinase signalling pathway[36]. Both studies suggest that ALKBH5 plays a role in the control of AML cell growth by impacting mRNA stability[36, 37*].

There is, however, some conflicting evidence relating to the role of ALKBH5 in AML progression. Based on genome-wide CRISPR screens, ALKBH5 was proposed as either a tumour suppressor gene or as non-essential for AML[15, 16]. A possible explanation for the discrepancies could be the difference in AML models used.

5. Conclusions

The complex network of m⁶A writers, readers and erasers plays a clear and crucial role in the development and maintenance of AML. Since targeting most of the relevant enzymes seems to specifically inhibit the survival of LSCs, it makes them potential targets for AML treatment as demonstrated recently with the FTO inhibitors. A number of groups and biotechnology companies are now racing to develop small molecule inhibitors against these. The first in the RNA-methyltransferase inhibitor class is a METTL3-specific bioavailable small molecule whose strong potential as a treatment for AML has been exhibited through work from our own group[38**]. The success of such an inhibitor molecule would open a new field in anti-cancer therapy. Of note, all current studies focusing on m⁶A and other RNA-modifications are heavily relying on antibody-specificity and there is need for the development of technologies complimenting or substituting antibody-based methods. Some examples here include pipelines for detection of RNA methylation on native RNA using Oxford Nanopore sequencing[39–41] or our recently published meCLICK-seq[42*].

Key points.

• m⁶A is important for myeloid leukemogenesis

• changes in m⁶A levels affect specifically LCSs but not normal HSCs

• m⁶A regulators are promising therapeutic targets for AML