5. PHARMACOLOGICAL STRATEGIES AIM TO TARGET THE EPITRANSCRIPTOME.
Taking into consideration the aforementioned dysregulation of the epitranscriptome in cancer, the design and development of small molecules that potentially revert defects in the epitranscriptome opens new and exciting opportunities for drug discovery in oncology (Table 2). Inevitabily, the successful introduction of epigenetic-based drugs into the clinic delineates a model for chemical biology and drug discovery research also on RNA-epitranscriptomics. Although the field of epitranscriptome-targeting is still in its beginnings, preclinical evidences of the benefits of RNA-modifying therapy are found in specific experimental models. Most important, several biotech companies are addressing the therapeutic potential of this promising field. Below, we will highlight the emerging results and challenges for the use of RNA- modifications as actionable targets for cancer drug discovery.
5.1. RNA methyltransferase inhibitors. There are several parallelisms between DNA and RNA methylation that support their exploitation within the framework of pharmacological inhibition. The most similar chemical modification is the addition of a methyl group at position 5 of cytidine resulting in 5-methylcytosine both at DNA or RNA molecule. It is however not unreasonable to assume that current DNA methyltransferases influence the dynamism of m5C RNA. As mentioned before, drugs inhibiting DNA methylation (e.g., 5-Azacytidine (AZA) or decitabine) have been FDA- approved and included in a clinical setting for the treatment of haematological tumours (Berdasco and Esteller, 2018). However, it is known that the vast majority (~90%) of AZA is incorporated into the RNA molecule and that DNA methylation status does not correlate with the clinical response to hypomethylating treatment. Whether the antiproliferative effect is mediated by RNA or DNA methylation is still under debate and further investigation is needed. Recently, a mechanism involving members of the RNA methyltransferases NSUN (NSUN1 and NSUN3) together with DNMT2 has been proposed as a mediator of AZA response in a AML and myelodysplasic syndrome model (Cheng et al., 2018). Specifically, authors proposed a mode of action that involves the formation of two chromatin complexes including distinct RNA modifiers to explain positive response or resistance to AZA treatments. In AZA sensitive cells, the reader hnRNPK directly recognized NSUN3, DNMT2 and CDK9/P-TEFb to recruit RNA-pol-II \soutand resulting in an active conformation of chromatin in sensitive AML cells. In AZA-resistant cells, the interaction of NSUN1 with the chromatin remodelling factor BRD4 (but not hnRNPK) and RNA pol-II results in an active chromatin structure that is resistant to AZA. However, these AZA-resistant cells are sensitive to the BRD4 inhibitor JQ1 and NSUN1 interference by siRNA (Cheng et al., 2018). Supporting evidences on the impact of NSUN2 in AZA response previously showed that NSUN2 and METTL1 abrogation (by genetic knockdown) results in increased hypomethylating drug sensitivity in HeLa cells (Okamoto et al., 2014). Since acquired resistance to chemotherapy is a major bottleneck in cancer treatments, unravelling the multiple molecular mechanisms that guides therapy response is a mandatory concern to improve precision medicine in cancer.
Histone lysine methyltransferases (KMT) also contain SAM-binding pocket and a substrate-binding domain that have been successfully targeted (Ganesan et al., 2019). The first attempt to inhibit KMT activity was based on the discovery that the natural product sinefungin reversibly competes with SAM for its binding site (Kaniskan et al., 2015a). Thereafter, several potent SAM-mimetics that are selective by taking advantage of differences in the cofactor binding pocket have been developed as KMT inhibitors. The DOT1L inhibitor, Pimenostat, was the first KMT inhibitor to enter clinical trials for leukaemia therapy, followed by EZH2 inhibitors (GSK2816126 and tazemetostat) approved for B-Cell lymphoma treatments (Berdasco and Esteller, 2018). Although the protein structure of specific RNA methyltransferases has specific and unique features, like DOT1L, m5C and m6A RNA methyltransferases belong to the Rossmann fold family of methyltransferases. This similarity portends that KMT inhibitors could be used as a starting point for the chemical design and drug development of RNA methyltransferases; however, no strong preclinical data has been given to support this observation.
Considering that m6A is the most universal RNA modification together with the well-defined aberrant m6A patterns associated with cancers, the research community has already drawn attention to the importance of strengthening the pharmacological manipulation of m6A methyltransferase activity. METTL3-METTL4 is upregulated in cancers, and it has been previously showed that genetic manipulation by CRISPR-Cas9 technology guided to downregulate METTL3 enzyme prevent cell proliferation and invasion in AML in vitro and in vivo models (Barbieri et al., 2017). Consequently, drug developers from biotech companies have started the race for early drug discovery targeting RNA methyltransferases, mainly METTL3. Three companies, STORM Therapeutics (Cambridge, UK), Accent Therapeutics (MA, USA) and Gotham Therapeutics (NYC, USA), have announced to have METTL3 inhibitors in preclinical phases ready for phase I clinical trials (Cully, 2019). To date, the most advanced results have been achieved by STORM Therapeutics. In October 2020, STORM \southas announced that its first-in-class drug candidate targeting METTL3, named STC-15, has been selected to enter human Phase I clinical trial as a therapy for refractory AML. Preclinical studies on a mouse model of AML, showed that oral administration of SCT-15 reduced both splenomegaly and the number of circulating monocytes. Similarly, tumour growth was reduced in patient-derived-xenografts (PDX)-AML models after treatment with the METTL3 inhibitor (Cully, 2019). The company is now studying the application on solid tumours. Accent Therapeutics has started its drug discovery program with an initial investment of $40M to optimize the selection of RNA-modifier inhibitors. They have announced that the company has already found around 20 targets, with METTL3/METTL14 inhibitors for AML treatment at the front of their research (Boriack-Sjodin et al., 2019; Cully, 2019). Similarly, Gotham Therapeutics launched in October 2018 with a $54 million program is the third company with a METTL3 inhibitor in preclinical development for AML therapy (Cully, 2019).
Out of these few examples, discovery of METTL3-METTL14 inhibitors are also explored in academia. Adenosine, one of the two moieties of SAM, is a SAM-competitive inhibitor of METTL3 activity. Recently, starting from a library of 4000 analogues and derivatives of the adenosine moiety of SAM and using high-throughput docking into METTL3 and protein X-ray crystallography, an adenosine derivative showed low μM potency and good ligand efficiency (Bedi et al., 2020). Interestingly, authors showed that the ribose of adenosine can be replaced by other ring systems, opening new opportunities for additional modifications (Bedi et al., 2020). Further development in preclinical models is still needed to explore the biological effect and mode of action of these adenosine derivatives.
5.2. RNA demethylases inhibitors. RNA demethylases also exhibit structural similarities with the protein lysine demethylases from the Jumonji C (JMJC) family. These are part of the 2-oxoglutarate and iron (II)-dependent dioxygenase family. This similarity is very interesting given that current inhibitors of JMJC proteins are available (Hauser et al., 2018) and that the mechanistic similarity between JMJC and RNA demethylases could facilitate the drug discovery for the inhibition of RNA modifications.
Interestingly, RNA demethylases have been targeted by specific small-molecule inhibitors. As m6A dysregulation impact in normal development and disease, its inhibition has been in the spotlight in the last years (Table 2 ). A pioneer study of small-molecule inhibitors of the human FTO demethylase was achieved by an chemical optimization of the natural product rhein (Chen et al., 2012). Rhein competitively binds to the FTO active site in vitro and globally increases cellular m6A on mRNA in the BE-2(C) cell line (Chen et al., 2012). Rhein also binds to ALKBH2 and ALKBH3 m1A and m3C demethylase, respectively; however, different binding sites are involved for ALKB or FTO inhibition (Li et al., 2016a).
A selective inhibition of m6A demethylase FTO rather than ALKBH5 was reported (Huang et al., 2015). The work was based on the identification of the differences in the displacement of m6A-containing ssDNA binding to FTO and ALKBH5. This screening provides meclofenamic acid (MA), previously identified as an anti-inflammatory, as the best match to specifically inhibit FTO over ALKBH5 (Huang et al., 2015). In vitro studies demonstrated that treatment of HeLa cancer cells with the ethyl ester form of MA (MA2) increased m6A mRNA levels (Huang et al., 2015). Furthermore, the antiproliferative effect of MA2 treatment has been tested in in vivo models of glioblastoma (Cui et al., 2017). MA2 increased mRNA m6A levels in glioblastoma-stem cells (GSC) leading to suppression of the GSC-initiated brain tumour development and prolonged the lifespan of GSC-engrafted mice (Cui et al., 2017). This result suggests that targeting m6A methylation could be a promising strategy for the treatment of glioblastoma. Using leukaemia in vitro and in vivo models, a role for the pharmacological inhibition of FTO to prevent resistance to tyrosine kinase inhibitor (TKI) therapy has been demonstrated (Yan et al., 2018). Mechanistically, exposure to rhein or MA increases m6A and mRNA stability of survival and proliferation genes (e.g. BCL-2 or MERTK) improving the sensitivity of the tumour to the TKI nilotinib and PKC412 (Yan et al., 2018).
The knowledge gained on the basis for MA selectivity for FTO over ALKBH5 has facilitated the design of additional FTO inhibitors. In a later step, a study based on a screening of many fluorescent molecules with structures similar to MA revealed that fluorescein (and some of its derivatives) selectively inhibited FTO demethylation as well as directly labelled FTO protein (Wang et al., 2015). Two fluorescein derivatives with improved cell permeability, FL6 and FL8, could efficiently inhibit FTO demethylation and modulate the level of m6A in the mRNA of living cells (Wang et al., 2015).
Research aimed at developing selective and cell-active small molecule inhibitors of AlkB subfamilies of demethylases have also explored the nucleotide-binding site instead of the 2OG-binding site (Toh et al., 2015). Compound 12 exhibits 30-fold to 130-fold selectivity for FTO over other AlkB subfamilies, and what is probably more interesting, the compound also discriminates against other human 2OG oxygenases, as PHD2 and JMJD2A protein demethylases. Treatment with an ethyl ester derivative of compound 12 increases m6A in HeLa cells (Toh et al., 2015).
Using structure-based rational design, which maintains the benzyl carboxylic acid to keep MA selectivity for FTO but extends the dichloride-substituted benzene to a deeper pocket that could be fully occupied by a bulky ligand, two FTO inhibitors were developed (FB23 and FB23-2) (Huang et al., 2019c). Based on the discovery that FB23 showed increased m6A levels due to FTO inhibition in in vitro AML cell lines, researchers have optimized the physicochemical property of FB23 and produced FB23-2 compound with a significantly improved antitumoural ability in in vitro but also in vivo conditions. FB23-2 treatments reduces the proliferation of a panel of AML cell lines, but most importantly, FB23-2 also inhibits primary leukaemia stem cells in PDX –AML mice models. Mechanistically, gain of m6A levels after FB23-2 treatment modulates mRNA transcripts associated with proliferation (e.g., MYC, CEBPA, RARA, and ASB2) (Huang et al., 2019c).
Additional FTO inhibitors were discovered using an elegant high-throughput screen using the fluorogenic methylated Broccoli substrate HTS assay (Svensen and Jaffrey, 2016). These Broccoli assays are based on the construction of a fluorescent RNA-dye complex that appear non-fluorescent when it contains m6A but becomes fluorescent after demethylation. The study identified several selective compounds for FTO inhibition which increase m6A levels at FTO target mRNAs (bone morphogenetic protein 6 (BMP6) and ubiquitin C (UBC) in HEK293C cells (Svensen and Jaffrey, 2016).
The possibilities extend beyond rational drug design as we learn more of the mode of action of the m6A-FTO axis. For example, the oncometabolite R–2-hydroxyglutarate (R-2HG) is produced at high levels in mutant isocitrate dehydrogenase 1/2 (IDH1/2) leukaemia cells (Su et al., 2018). However, R-2HG also has an antitumoral effect through the inhibition of FTO activity. FTO inhibition results in a gain of m6A levels and the stabilization of the mRNA transcripts MYC/CEBPA, leading to the suppression of relevant proliferation pathways (Su et al., 2018).
Undoubtedly, the primary focus of attention is now in FTO inhibition. However, additional RNA demethylases could be “druggable” targets. For example, the small compound 1-(5-methyl-1H-benzimidazol-2-yl)-4-benzyl-3- methylpyrazol-5-ol (HUHS015) was able to inhibit the prostate cancer antigen-1 (PCA-1/ALKBH3) axis in prostate cancer cell lines and murine xenograft models for prostate cancer (Nakao et al., 2014).
5.3. Targeting other RNA modifications. Epigenetic modifications of DNA molecules are interpreted by a set of reader proteins with essential functions. Most importantly, these readers for epigenetic marks are altered in human diseases, leading to the discovery of small compound inhibitors of their activity (Ganesan et al., 2019). Potent drug inhibitors have been identified for the H3K27me3 reader Polycomb protein EED from the Polycomb repressive complex 2 (PRC2) family (He et al., 2017). Like The YTF family of RNA methyl readers, Polycomb protein EED contains an aromatic cage crucial for the recognition of the methyl group. Whether the success of drug discovery associated with methyl-lysine readers could be translated to the methyl-RNA-reader field is still uncertain and unexplored. One barrier to the development of YTF- inhibitors is that YTF members from the same family of proteins exhibit high structural homology. It still needs to be determined whether the application of Pan-YTF inhibitors could have a tissue-specific effect and contribute to increased specificity (Cully, 2019). Efforts to inhibit the m7G reader eukaryotic translation initiation factor 4E (eIF4E) as an actionable target have been proposed (Soukarieh et al., 2016). Through its role in the regulation of mRNA translation of oncogenic pathways, eIF4E is implicated in cell transformation, tumorigenesis, and angiogenesis. Guanine-based inhibitors of eIF4E were evaluated in in vitro cell-based assays and provides a set of compounds with inhibitory activity at physiological doses (Soukarieh et al., 2016).
There are unique RNA modifications that could not be compared with DNA or histone modifications. The RNA modifications involved in pseudouridylation or A-to-I editing have no precedents in drug discovery. However, chemical biology and drug discovery in this area requires a better characterization of the modes of action and pathological implications in a context-specific manner. The 3’ terminal uridylyl transferase Zcchc11 (TUT4) is recruited to precursor let-7 RNA to selectively block let-7 miRNA biogenesis, a miRNA with tumour suppressor properties. Downregulation of Let7 miRNA has been described in cancer. It is therefore of great application to develop an inhibitor for the uridylation of precursor let-7 resulting in restored Let-7 expression in cancer (Lin and Gregory, 2015). Using an automated high-throughput screen of ∼15,000 chemicals, some small compounds have been selected as putative TUTase inhibitors (Lin and Gregory, 2015). The understanding of the TUT4-let-7 mediated inhibition is not addressed, so the consequences in preclinical models need to be determined.
In terms of pharmacological inhibition of the A-to-I edition, targeting ADAR family of proteins could be a promising strategy for cancer therapy. As mentioned before, ADAR1 is involved in multiple cancer-related pathways, and loss of function of ADAR1 in tumour cells profoundly sensitizes tumours to anti-PD1 immunotherapy (Ishizuka et al., 2019). Consequently, a strategy to repress ADAR1 expression is particularly challenging. Nowadays, there is not any public compound targeting ADAR1; but biotech companies such as Accent Therapeutics have declared to be working on this target for NSCLC therapy (Cully, 2019). In addition, by studying analogues of a naturally adenosine analogue, the 8-azaadenosine compound showed in vitro inhibition activity of ADAR2 (Véliz et al., 2003).