Mining for METTL3 inhibitors to suppress cancer

Abstract

The RNA methyltransferase METTL3 catalyzes N⁶-methyladenosine (m⁶A) modification of messenger RNAs (mRNAs). It is overexpressed in many types of cancer, including acute myelogenous leukemia (AML), and promotes cancer cell growth and tumorigenicity. Now, a selective small molecule inhibitor of METTL3 shows significant anti-leukemic effects in preclinical AML models, highlighting the promise of pharmacological METTL3 inhibition as a new cancer therapy.

While the importance of DNA methylation and histone modification in gene regulation is well established and chromatin-associated proteins are the subject of current drug-development efforts to treat cancer and other diseases¹, RNA modifications are less well understood and appreciated. m⁶A is the most prevalent mRNA modification. It occurs at a particular sequence motif and is enriched at sites close to the translation stop codon of a large subset of the transcriptome^2,3. A protein complex comprising METTL3 (methyltransferase-like 3) together with its essential cofactor METTL14 and accessory proteins is responsible for the deposition of almost all m⁶A on mRNA. The m⁶A ‘epitranscriptome’ plays important roles in the posttranscriptional control of gene expression including mRNA splicing, nuclear export, stability, and translation⁴.

METTL3 belongs to the large family of class I methyltransferases (MTases) that contain a Rossmann fold that binds the S-Adenosyl Methionine (SAM) methyl donor. Over recent years it has become increasingly clear that METTL3 functions as an oncogene in a variety of different cancer types. Studies in human cancer cell lines and mouse models have established METTL3 as a possible new anti-cancer^5–8. Reporting now in Nature, Yankova, Blackaby et al., identified a selective inhibitor of METTL3 catalytic activity and demonstrate its efficacy in vitro and in vivo⁹.

The authors employed a RapidFire™ mass spectrometry-based assay to monitor METTL3 activity by measuring levels of S-Adenosyl homocysteine (SAH) generated by the MTase reaction A high throughput screen (HTS) of 250,000 compounds identified only two non-SAM-related hits. Chemical optimization dramatically improved the potency of one of the original hits, leading to the development of the small molecule STM2475 with an IC₅₀ of 16.9 nM. Surface plasmon resonance (SPR) and a crystal structure of STM2475-bound METTL3 show that STM2457 adopts a co-factor competitive mode and binds within the SAM-binding pocket of METTL3 with high affinity (Kd = 1.4 nM) (Figure 1a). STM2457 exhibited high selectivity for METTL3 compared to a broad MTase panel and potently inhibited METTL3/14 catalytic activity. Using a structure-based drug discovery approach, an independent group recently also identified a selective small molecule inhibitor (UZH1a) of METTL3 activity. Like STM2457, UZH1a occupies the SAM binding site (Figure 1a) of METTL3¹⁰. The strong selectivity of both of these inhibitors can be explained at least in part by the conformational reorganization of lysine 513 (K513) upon binding of the small molecule. This feature is common to STM2457 and UZH1a, but does not occur in METTL3 bound to SAM or the non-specific MTase inhibitor Sinefungin (Figure 1a). As expected, considering the SAM-competitive mechanism of action of these METTL3 inhibitors and the relatively high cellular concentration of SAM, a much higher concentration of STM2457 (IC₅₀ = 2.2 uM), or UZH1a (IC₅₀ = 4.6 uM) was required to inhibit METTL3 in cells as compared to biochemical assays and to reduce the overall m⁶A mRNA modification level.

Figure 1. METTL3 inhibition is a promising therapy against acute myelogenous leukemia (AML).

(a) The position of small molecule or ligands within the SAM-binding pocket of human METTL3. Surface representation of the crystal structure of METTL3 bound to STM2457 (PDB 7O2I), UZH1a (PDB 7ACD), SAM (PDB 5K7U), and Sinefungin (PDB 6Y4G) are displayed. Residue K513 of METTL3 is highlighted. (b) Inhibition of METTL3 methyltransferase activity affects m⁶A modification and translation levels of METTL3-modified mRNAs. (c) STM2457 treatment reduces proliferation and colony formation of AML cells, and impairs AML cell engraftment and expansion in human PDX models by inhibiting the self-renewal of leukemic stem cells.

Consistent with previous METTL3-depletion studies, inhibition of the catalytic activity of METTL3 with either STM2457 or UZH1a blocked proliferation of a panel of human AML cell lines including MOLM-13 (Figure 1b). Moreover, while METTL3 inhibition by STM2457 strongly reduced the clonogenic capacity and induced apoptosis of mouse primary AML cells, it had no effect on normal mouse or human hematopoietic stem and progenitor cells (HSPCs) in these in vitro assays.

These dramatic results raise the question of the underlying mechanism responsible for these phenotypes. While the biological importance and pathological relevance of METTL3 is now well-established, the molecular function of METTL3 and m⁶A mRNA modification remains unresolved and somewhat contentious, because effects on many different aspects of mRNA metabolism have been reported to-date. To address this question in the context of AML, Yankova and, Blackaby et al. initially performed RNA-seq on control and STM2457-treated MOLM-13 cells. Gene Ontology analysis showed that differentially expressed transcripts were enriched for genes related to myeloid differentiation and leukemia. Given the analyzed cell-type and the phenotypic changes caused by METTL3 inhibition, this result is not unexpected and at least some of the expression changes are likely due to indirect effects on the cellular state. The authors then performed m⁶A methylated RNA immuno-precipitation and sequencing (meRIP-Seq) to identify m⁶A-modified mRNAs. Identification of mRNAs with reduced m⁶A levels upon METTL3 inhibition is challenging, because meRIP-Seq is not very quantitative. Nevertheless, the authors perform quantitative RT-PCR to support the observation that m⁶A modification of several mRNAs is strongly decreased upon STM2457 treatment. These mRNAs include SP1 and BRD4, two mRNAs that are known to be regulated by METTL3 and relevant to AML.

m⁶A modification has been linked to various aspects of posttranscriptional mRNA control. While much focus has been on the role of m⁶A in targeting mRNAs for decay^11,12, we previously found no major changes in steady state mRNA levels or mRNA stability upon METTL3 depletion. Instead, we observed a pronounced decrease in translation of multiple m⁶A-modified mRNAs^5,6. Similarly, Yankova, Blackaby et al. found a substantial decrease in translation upon METTL3 inhibition . Importantly, they observe that m⁶A-modified target mRNAs, including SP1 and BRD4, shifted from the heavy polysome fraction to sub-polysome fractions upon METTL3 inhibition. No change in mRNA levels were detected. Overall, these findings support a major role for METTL3 and m⁶A in promoting translation of these (and many other) mRNAs.

Conversely, YTHDF2, one of several m⁶A reader proteins linked to mRNA decay, reportedly plays an important role in AML¹³. Even more confounding are results reporting that the putative m⁶A demethylases ALKBH5, and FTO promote AML and that a small molecule inhibitor of FTO effectively suppresses AML progression in mouse models^14–17. The apparent discrepancy in the observation that blocking either the m⁶A-writer METTL3 or the m⁶A-erasers ALKBH5 and FTO suppress AML might be explained at least in part by recent reports that the major substrate of FTO is m⁶Am, a modification that occurs specifically at the 5’ CAP structure of mRNA. Moreover, FTO might primarily regulate small nuclear RNA (snRNA) rather than mRNAs^18,19. More work is needed to fully reconcile these findings.

A highlight of the work by Yankova, Blackaby et al. is the striking anti-leukemic effect of STM2457 in vivo. The authors use both AML patient-derived xenografts (PDX) as well as a primary murine MLL-AF9/Flt3^Itd/+ model to test the efficacy of pharmacological inhibition of METTL3. Strikingly, they found that daily injection of STM2457 suppressed AML engraftment and expansion of PDXs with different genetic cancer drivers, and extended the lifespan of recipient mice. Furthermore, these effects correlate with decreased global m⁶A levels in treated mice as measured by mass-spectrometry as well as reduced protein expression of several m⁶A-modified mRNAs. Since AML is a cancer with a well-defined cancer stem cell contribution, the authors also examined the effect of STM2457 treatment on the CD93⁺ cell population and found strongly diminished numbers of these leukemia stem cells upon METTL3 inhibition (Figure 1c). Prolonged survival and comparable molecular effects were also found in the mouse AML model upon treatment with STM2457. The results from secondary cell transplantations treated with STM2457 (or control) provide additional functional support that METTL3 inhibition depletes the leukemic stem cell component and results in impaired AML propagation and extended survival of mice receiving STM2457-treated cells. Gratifyingly, no overt toxicity due to systemic METTL3 inhibition was reported in the STM2457-treated mice.

Overall, these findings represent an important step towards the development of METTL3 inhibitors and highlight the potential of RNA MTases and the epitranscriptome as a largely unexplored territory for cancer drug discovery. Nevertheless, questions remain: why are cancer cells more sensitive to METTL3 inhibition than their non-transformed counterparts? What downstream gene(s) are responsible for the phenotypic effects of METTL3 inhibition? How does METTL3 influence mRNA translation and what is the relevance of altered mRNA stability to its oncogenic role? How is METTL3 expression or its activity dysregulated in AML? Are there particular AML patient subsets or cancer genotypes that are expected to be more responsive to METTL3 inhibition? Which biomarkers might help identify cancer patients for METTL3 inhibitor therapy and monitor responsiveness? Despite these open questions, this landmark study provides a proof-of-concept that pharmacological inhibition of METTL3 and manipulation the m⁶A epitranscriptome can suppress cancer and substantially extend survival in mouse AML models.. Given this precedent, and considering that there are well over 100 different RNA modifications present on various classes of RNAs including mRNA, rRNA, tRNA, snRNA, and other non-coding RNAs (ncRNA), it is reasonable to speculate that other RNA-modifying enzymes may represent similarly promising cancer drug targets²⁰.