Advances in targeting RNA modifications for anti-cancer therapy

Abstract

Numerous strategies are employed by cancer cells to control gene expression and facilitate tumorigenesis. In the study of epitranscriptomics, a diverse set of modifications to RNA represent a new player of gene regulation in disease and in development. N⁶-methyladenosine (m⁶A) is the most common modification on mammalian messenger RNA and tends to be aberrantly placed in cancer. Recognized by a series of reader proteins that dictate the fate of the RNA, m⁶A-modified RNA could promote tumorigenesis by driving pro-tumor gene expression signatures and altering the immunologic response to tumors. Preclinical evidence suggests m⁶A writer, reader, and eraser proteins are attractive therapeutic targets. First-in-human studies are currently testing small molecule inhibition against the METTL3/METTL14 methyltransferase complex. Additional modifications to RNA are adopted by cancers to drive tumor development and are under investigation.

RNA modifications regulate gene expression

Gene expression relies on dynamic amplification and suppression of genetic signals into functional biomolecular units of RNA and protein, a process that is tightly regulated in time and space [1]. The word epitranscriptomics refers to a group of post-synthesis modifications to canonical A, C, G, and U bases on RNA [2–4]. Both coding and noncoding RNA can undergo more than 150 distinct alterations, which have an impact on a wide range of biological processes (Figure 1) [2]. tRNA and rRNA contain the highest number of modifications, additions that modulate structure and function of these RNAs and eventually alter translation of mRNA into protein. Modifications are placed by several writers, or enzymes that catalyze the deposition of RNA modifications.

Figure 1. RNA modifications are distributed across mRNA, noncoding RNA, rRNA, and tRNA.

m⁶A is the most abundant modification on mRNA. Marks can be distributed across the transcript body and at locations important for RNA processing, such as at splice sites and the 5’ cap of mRNA. tRNA contain a highly diverse set of modifications, few select examples shown in this figure. These modifications can alter the protein coding ability of transcripts. Created with BioRender.com.

The most common internal modification to mammalian mRNA is N⁶-methyladenosine, often studied in human disease due to the ability of m⁶A to affect the fate of protein-coding mRNA [5]. m⁶A is placed by a methyltransferase complex composed primarily of methyltransferase-like 3 (METTL3) and methyltransferase-like 14 (METTL14) [6–9]. m⁶A is detected by an expanding list of reader proteins that recognize marked RNA and enable a diverse set of cellular functions that mediate tumorigenesis [2, 10]. The discovery of eraser proteins fat mass and obesity associated protein (FTO) and alkB homolog 5 (ALKBH5) demonstrated the reversible nature of m⁶A [11, 12]. Pseudouridine is an abundant modification found on most types of RNA, including eukaryotic mRNA [13–17]. Commonly referred to as the “fifth nucleotide”, pseudouridine is the isomerization product of uridine [18, 19]. N¹-methyladenosine (m¹A) has been primarily characterized as a component of tRNA, deposited by the tRNA methyltransferase (TRM) family of proteins [20]. 7-methylguanosine (m⁷G) is found on multiple types of RNAs. Base editing of RNA is marked by adenosine residues switching to inosine, carried out by adenosine deaminase acting on double-stranded RNA (ADAR) family of proteins [21, 22].

Many cancers utilize RNA modifications as a mechanism to enhance or repress expression of genes in favor of an oncogenic state [20]. Efforts to characterize the function of RNA modifications have been aided by sequencing innovations that have hastened insight into their function in development and in disease [23, 24]. Targeting RNA modifications is a budding strategy to target various cancers, and determination of effective settings in which to modulate their function is underway, along with efforts to develop specific and effective compounds that target the enzymes that deposit such modifications.

As the role for epitranscriptomics in cancer has been unveiled, an outstanding question in the field is addressing the ability to therapeutically target the proteins that install these marks. Essentiality and tissue specificity of such broadly functioning proteins are concerns, though many groups have compared modification levels and methyltransferase levels in cancerous tissue to matched normal control tissue with notable differences. Genetic knockdown and knockout experiments are helpful in the study of methyltransferase proteins but can hinder the identification of catalytic effects due to the simultaneous obstruction of catalytic independent roles in cells. Despite these challenges, small molecules have been synthesized and have exhibited promising effects in early studies, with first-in-human studies currently ongoing. In this review, we look to critically evaluate the targeting potential of catalytic activity of proteins that deposit RNA modifications.

The role of METTL3 and METTL14

The METTL3/METTL14 complex as writers of m⁶A

Co-transcriptional placement of m⁶A on mRNA is guided by a methyltransferase complex composed of METTL3, METTL14, and Wilms-tumor associated protein (WTAP) together with accessory proteins [2, 6, 8]. m⁶A is found at a consensus DRACH sequence motif, where D = A, G, or U; R = A or G; and H = A, C, or U. Exon architecture guides the selection of transcripts for m⁶A installation, with areas encompassing exon-junction complexes (EJC) insulated from m⁶A placement [25–27]. m⁶A is recognized by nuclear and cytoplasmic RNA-binding reader proteins that dictate the fate of marked RNA and is removed by eraser proteins FTO and ALKBH5 [2, 11, 12].

From early studies, it was apparent that m⁶A influences ubiquitous aspects of tumorigenesis, such as cell proliferation, differentiation from a stem-like state, and invasion [2, 20]. m⁶A is placed co-transcriptionally and serves as a central mediator to gene expression, and a bridge to translation. To note, the activity of METTL3 can be disease-specific, as there are differences in cellular response to depletion of METTL3; in some instances, METTL3 promotes tumorigenesis, while few others use METTL3-mediated m⁶A deposition to drive tumor suppression [28, 29]. Mutations of methyltransferase components can also occur, such as a hotspot mutation of METTL14 that appears to promote progression of endometrial cancer [30]. These contradictions together with findings of concomitant catalytic dependent and catalytic independent functions of METTL3 have increased the complexity of characterizing the translational impact for inhibitors against the methyltransferase complex.

Many initial studies in cancer revealed that METTL3 functions in hematologic malignancies [20, 31]. Depletion of METTL3 generated a prominent influence on differentiation of leukemic cells and reduced cell proliferation, an effect that was only rescued with wild-type METTL3, suggesting the necessity for intact catalytic function [32]. In recent years, the phenotypic effect of METTL3 has extended to solid tumors [33–39]. Akin to the influence of METTL3 on prevention of acute myelogenous leukemia (AML) cell differentiation, METTL3 also maintains a glioma stem-like state in glioblastoma multiforme (GBM) [40]. Moreover, METTL3 depletion in GBM cells sensitized cells to gamma irradiation and reduced the cellular ability to undergo DNA repair [40]. In a pediatric-derived osteosarcoma cell line, lack of METTL3 sensitized tumors to DNA damage-inducing agent cisplatin, in addition to irradiation therapy [41].

The METTL3 protein can also be modified to alter its activity. METTL3 can be phosphorylated by DNA repair protein ataxia-telangiectasia mutated (ATM) at position S43, thus integrating into DNA damage response and repair [41]. Phosphorylated METTL3 localized to foci of DNA double strand breaks, placing m⁶A onto DNA damage-associated RNAs for protection from degradation, eventually forming DNA:RNA hybrid structures that recruit repair machinery. Interestingly, METTL3 catalytic and phosphorylation sites were both needed to observe this phenotypic effect [41]. METTL3 was also found to have two sites on its zinc finger domain that are susceptible to lactylation, endowing an immunosuppressive effect by promoting an environment with immune escape-promoting tumor-infiltrating myeloid cells (TIMs) [42].

The overlap of METTL3 and METTL14 depletion on cancer cell phenotype is intriguing, especially given the thought that METTL3 plays the predominant catalytic role, while METTL14 is thought to play a rather supportive role. In a recent study assessing the role of m⁶A in the response to checkpoint blockade, both METTL3 and METTL14 were found to influence response to anti-PD-1 [43]. m⁶A appears to influence immunoregulation by both cell intrinsic and extrinsic immune responses. While in some cancers immune checkpoint inhibition using anti-PD-1 has been fruitful, in others response has been limited [44]. METTL3 and METTL14 perturbation can enhance the response to anti-PD1 treatment in certain types of colorectal cancer and in melanoma, with m⁶A bound by YTH domain-containing family 2 (YTHDF2) [43], and in breast cancer recognized by IGF2BP3 [45]. Experiments assessing host-derived immune response show that depletion of METTL3 can contribute to an immunosuppressive tumor microenvironment by redirecting macrophage response, leading to a pro-tumorigenic effect [46]. Others have shown that loss of m⁶A after METTL3 depletion diminished immunologic response to tumors, promoting tumor growth [47]. It is clear that both m⁶A writer- and reader-mediated manipulation impact the immunologic response to tumors. The therapeutic effect of using either in anti-cancer therapy may be context specific, as targets of METTL3 show different response in different cell lines. Alternative isoforms of METTL3 have been identified after knockout in cell line experiments, an effect that if present in vivo could account for variation in effects depending on the system used [48].

METTL3 and METTL14 function are tightly intertwined with transcription. In mouse embryonic stem (mES) cells, knockout of METTL3 led to globally increased chromatin accessibility and increased nascent transcript synthesis [49]. Through interaction of nuclear exosome targeting (NEXT) complex components with nuclear m⁶A reader protein YTHDC1, m⁶A-marked chromatin-associated retrotransposon RNA was targeted for degradation at baseline that ultimately affected development. In absence of METTL3, m⁶A was depleted and resulted in increased differentiation capacity of mES cells [49].

Methyltransferase complex components can also directly interact with DNA and histone proteins. Noting an inverse relationship between repressive DNA mark 5-methylcytosine (5-MC) and m⁶A levels, METTL3-mediated m⁶A methylated RNA was found to be an important factor in tet methylcytosine dioxygenase 1 (TET1) binding to target chromatin, demonstrating crosstalk between RNA modifications and DNA modifications [50]. The close relationship is further demonstrated by H3K36me3, which recruits METTL14 to induce deposition of m⁶A at specific loci during transcription [51]. m⁶A RNA-binding reader proteins can also directly influence chromatin dynamics. In roles that may have implications on mammalian development, the nuclear m⁶A reader protein YTHDC1 mediated mES cell state by affecting chromatin through mediation of retrotransposons or retroviral elements [52, 53]. Acting, again, in coordination with histone modifications, YTHDC1 recruited histone demethylase KDM3B to remove H3K9me2 from histone proteins [54]. In other contexts, YTHDC1 together with hnRNP protected newly transcribed RNA from termination using the integrator complex [55].

METTL16 deposits m⁶A on a subset of mRNA. In a CRISPR-Cas9 screen, METTL16 was essential for survival of many cancers [56]. Similar to METTL3, METTL16 also functions in the cytoplasm to enhance translation initiation by recognizing m⁶A-marked mRNA of a larger proportion of nearly 4,000 mRNAs, implying cooperative functions of methyltransferase members in enhancing expression of transcripts [56].

Catalytic-independent roles for METTL3/METTL14

An intriguing observation that METTL3 functioned in an oncogenic role, but that the effect was not dependent on m⁶A was identified by Lin et. al in 2016 [57]. In lung cancer, a reporter assay showed that METTL3 directly promoted translation of mRNA into protein by recruiting the translation initiation machinery. Expression of catalytic inactive METTL3 rescued protein translation deficiencies, suggesting that cytoplasmic METTL3 can function independent of m⁶A deposition. The mechanism was further characterized to show that METTL3 enhanced protein translation by promoting an mRNA looping-based structure with translation machinery [58]. Since then, this theme has been recapitulated in various settings and has shown that the ability of METTL3/METTL14 to induce oncogenicity can be both dependent on the ability of the methyltransferase complex to catalyze the deposition of m⁶A, and a function separate from catalytic deposition [59].

Using a system to induce cellular senescence, METTL3 and METTL14 were both able to drive a phenotype called senescence-associated secretory phenotype (SASP) [60]. SASP is characterized by various cytokines and chemokines that endow cells with enhanced ability to mediate both recognition and evasion of detection by the host immune response. METTL3 scattered to transcription start sites and METTL14 to gene enhancers, ultimately mediating an immunoregulatory state. Interestingly, gene expression changes from METTL14 and METTL3 knockdown cells could be rescued with both WT and catalytic mutant forms of each protein, suggesting that SASP was not dependent on catalytic m⁶A deposition [60].

METTL14 can function independently of the catalytic complex responsible for m⁶a deposition by mediating transcription [61]. In mES cells, where METTL3 depletion led to a global increase in transcription, METTL14 depletion resulted in a global decreased production of nascent RNA. Mechanistically, the histone mark H3K27me3 is recognized by METTL14 and ultimately recruits demethylase KDM6B to globally repress transcription [61]. Insights gained from studying differentiation capacity in mES cells in response to the m⁶A methyltransferase complex represents a mechanism that may be replicated in other disease processes such as cancer [62].

m⁶A reader proteins

Translating the message of m⁶A-marked RNA is undertaken by a set of reader proteins that serve to recognize and recruit processing factors that dictate the fate of modified RNA. The primary family members that have been characterized include the YTH domain-containing family (YTHDF1-3, YTHDC1-2), insulin-like growth factor 1 mRNA-binding protein (IGFBP2) family, proline-rich coiled-coil containing protein 2A (PRRC2A), and heterogenous nuclear ribonucleoproteins A2/B1 (HNRNPA2B1) [20]. YTHDF1 and YTHDF2 bind m⁶A to modulate RNA translation and RNA degradation, respectively, while YTHDF3 can enhance the function of both YTHDF1 and YTHDF3 [2, 63]. In an animal model of melanoma and colon cancer, YTHDF1 depletion enhanced recruitment of CD8+ cytotoxic T cells and natural killer cells into the tumor [64]. Loss of YTHDF1 decreased the translation of cathepsins, which allowed for decreased lysosomal-mediated degradation of tumor specific antigens. The effect of this was increased capacity for dendritic cells to promote cross-presentation to CD8+ cytotoxic T cells. In response, tumor growth was slowed, and response to checkpoint blockade therapy was markedly increased [64].

The host immune response to tumors is also mediated by reader protein YTHDF2. In a mouse model of melanoma and colorectal cancer, the antitumor CD8+ cytotoxic T cell response was blunted in presence of YTHDF2 [65]. By decreasing the stability of STAT1 mRNA in macrophages, YTHDF2 dampened cross presentation of tumor antigens to T cells, and subsequent downregulation of YTHDF2 led to reduced tumor growth and metastasis [65]. Enhanced response to checkpoint blockade in colorectal cancer (CRC) in the setting of METTL3 or METTL14 depletion was also dependent on YTHDF2 [43]. YTHDF2 is highly expressed in many different cancers, and tumors such as AML and glioblastoma exhibit a selective dependency on the protein [65–67]. In glioblastoma, YTHDF2 mediated the stability of LXRα and HIVEP2 mRNA, and depletion of YTHDF2 led to LXRα-mediated tumor growth inhibition by reducing intracellular cholesterol [68]. In AML, loss of YTHDF2 led to decreased proliferation in leukemic cells and sensitized them to tumor necrosis factor (TNF)-induced apoptosis [66]. Interestingly, sensitivity to YTHDF2 perturbation is dependent on the differentiation state of the cell. While YTHDF2 loss affected leukemic cells and led to expansion of HSCs, there was no impact on normal hematopoiesis [66].

The impact on HSC expansion has vast clinical applicability. In patients who undergo hematopoietic stem cell transplant (HSCT), the source for HSCs that produce the most optimal clinical outcome is often from cord blood donors. However, use of cord blood is limited by the low cell numbers derived that are needed for successful transplant. In mouse HSCs, YTHDF2 knockout led to expansion of phenotypical HSCs [69]. Knockdown of YTHDF2 in human cord blood samples led to increased expression of m⁶A-marked self-renewal-related transcription factors, supporting a role for YTHDF2-mediated mRNA degradation to maintain cord blood HSC self-renewal [69]. 10 weeks post-transplant, mice that were transplanted with YTHDF2 knockdown HSCs had a nine-fold increase in hematopoietic cell engraftment in the bone marrow, without any changes to the proportion of other cell lineages [69].

m⁶A reader proteins influence a variety of cellular processes linked to oncogenesis. m⁶A is detected in many DNA:RNA hybrid structures, otherwise known as R loops, which contribute to genomic instability [70]. YTHDF2 degrades m⁶A-marked RNA comprising R loop structures, which could have meaningful implications in cancer biology [70]. YTHDF1, YTHDF2, and YTHDF3 undergo liquid-liquid phase separation, capturing m⁶A-methylated RNA into organized structures such as P-bodies and stress granules for processing [71, 72]. As a binding partner of YTHDF1, fragile X mental retardation protein (FMRP) phosphorylation stimulated translational activity of YTHDF1, an effect that is recapitulated in fragile X syndrome [73].

m⁶A eraser proteins

The reversible nature of m⁶A is demonstrated by demethylases FTO and ALKBH5, which function to remove m⁶A from RNA in an iron(II)- and α-ketoglutarate-dependent manner [11, 74]. FTO can also remove methylation from N⁶,2’-O-dimethyladenosine (m⁶Am) and m¹A in tRNA, with the prominent number of sites comprising m⁶A. Consistent with the paradoxical effect of the METTL3/METTL14 complex as both tumor suppressive and oncogenic, FTO and ALKBH5 play diverse roles in driving cancer depending on the tumor and context in which they are studied. ALKBH5 and FTO both target a subset of genes in AML to influence cell growth and differentiation [75, 76]. In mES cells, FTO mediated the demethylation of m⁶A-containing long-interspersed element-1 (LINE1) RNA, which affected chromatin state and regulated gene expression during development [77]. These studies support the notion of RNA modifications as a fundamental component of human development and disease, and small molecules have been synthesized to target them (Table 1).

Table 1.

Small molecule inhibitors that target RNA modifications

Inhibitor Name	Cancer	Anticancer Potency	Ref
METTL3/METTL14
STM2457	AML	0.6 μM −10.3 μMa	[104]
STC-15	advanced tumors		NCT05584111
Multiple small molecules	AML, ovarian adenocarcinoma	80 nM–1.39 μM	[105–108, 132]
Multiple small molecules	AML	<1 μM, <10 μM	[109–111, 132]
UZH2	AML, prostate cancer	12 μM, 70 μM	[112]
UZH1a	AML, osteosarcoma	Not tested	[133]
CDIBA	AML	GI50 13–22 μM	[115]
Eltrombopag	AML	GI50 = 8.28 μM	[116]
Elvitegravir	esophageal squamous cell carcinoma	reduced metastasis	[134]
YTHDF
Ebselon	prostate cancer	26.83 μM	[119]
YTHDF1
Salvianolic Acid (SAC)	neuronal tissue	N/A	[73]
YTHDF2
CpG-siRNAYTHDF2	mouse melanoma, murine colon adenocarcinoma	reduced tumor growth and metastasis	[65]
FTO
CS1, CS2	AML	CS1:22 nM–753 nM CS2: 55.9 nM–1.127 μM	[122]
FB23–2	AML	1.9 μM −5.2 μM	[135]
MO-I-500	triple-negative inflammatory breast cancer	Inhibited survival and colony formation	[136, 137]
MA	HeLa	N/A	[138]
MA2	GSCs	Inhibited growth and self-renewal of GSCs, suppressed tumor progression and prolonged lifespan of GSC-grafted mice	[29]
FTO-04	GSCs, glioblastoma	Impaired self-renewal properties	[139]
Dac51	Melanoma	Inhibited tumor growth and prolonged survival of melanoma mouse models in combination with immunotherapy	[140]
R-2HG	Non-IDH mutant	decreased cell	[141]
	AML, Glioma	proliferation
ALKBH5
2-[(1-hydroxy-2-oxo-2-phenylethyl)sulfanyl]acetic acid, 4-{[(furan-2-yl)methyl]amino}-1,2-diazinane-3,6-dione	AML	1.38 μM −47.8 μM	[142]
PUS7
C17	Glioblastoma, GSCs	56.77 nM–177.4 nM	[94]

^adenotes IC₅₀, unless otherwise specified

AML, acute myeloid leukemia; GSCs, glioma stem-like cells; METTL3, methyltransferase-like 3; METTL14, methyltransferase-like 14; YTHDF, YTH domain-containing family protein; FTO, fat mass and obesity-associated protein; ALKBH5, alkB homologue 5; PUS, pseudouridine synthase; MA, meclofenamic acid; N/A, not available; IDH, isocitrate dehydrogenases

Additional writers of RNA modifications

METTL1/WDR4 complex

METTL1, together with WD repeat domain 4 (WDR4), catalyze the synthesis of N⁷-methylguanosine (m⁷G), which is the addition of a methyl group to ribo-guanosine at the N⁷ position [78]. m⁷G is found on tRNA, mRNA, miRNA, and rRNA, representing the most common modification to the cap of mRNA, and found internally in the additional types of RNA [79, 80]. m⁷G functions to influence gene expression through its role in RNA processing, metabolism, and function. METTL1 is found at higher levels in cancerous tissue compared to non-cancerous tissue [81]. The METTL1/WDR4 complex has potential opposing mechanisms on tumorigenicity depending on the RNA modified; when found on tRNA m⁷G is tumor promoting, while on miRNA is tumor suppressing in cases known to date [80].

Knockdown of METTL1 or WDR4 has led to decreased m⁷G modifications on tRNA in nasopharyngeal carcinoma, neuroblastoma, hepatocellular carcinoma, bladder cancer, and AML, among others [81–85]. METTL1 knockdown in these instances impaired aspects of codon recognition and tRNA functionality that ultimately impaired protein translation. Collectively, this perturbation has led to features of decreased tumorigenicity such as increased apoptosis and decreased cell cycle progression, chemoresistance, and tumor growth in mice. Using a sensitive strategy to map m⁷G on miRNA, heightened let-7 miRNA processing was shown to inhibit cell migration and decrease tumorigenicity [86].

NOL1/NOP2/SUN domain family of proteins

Methylation of cytosine is common to multiple types of RNAs, such as rRNA, tRNA, mRNA, and non-coding RNA [87]. m⁵C is catalyzed by a family of proteins of the NOL1/NOP2/SUN domain family of SAM-dependent methyltransferases NSUN1 to NSUN7, and DNA methyltransferase homolog DNMT2 [88]. NOP2/NSUN1, previously known as p120, has classically been found upregulated in proliferating cancer cells [89, 90]. Their role in regulating gene expression stems from the ability to influence ribosomal RNA processing, which in turn has widespread effects on gene expression into proteins. m⁵C has also been implicated in leukemia and in glioma [91, 92]. Similar to the diverse effects of the m⁶A methyltransferase complex, NOP2/NSUN1 were shown to have both catalytic and non-catalytic functions. Using cells derived from a patient with colon cancer, NOP2/NSUN1 recruited U3 and U8 snoRNAs to pre-90s ribosomes and influenced the assembly into complete snoRNP complex, an effect that was rescued by both catalytic inactive and wild type NOP2/NSUN1 [89].

Pseudouridine synthases

Often activated by stressors, the isomerization product of uridine, pseudouridine is placed by a family of pseudouridine synthase enzymes PUS1-PUS7, PUS7L, RPUSD1 to RPUSD4, and DKC1 [20]. While in the context of embryonic stem cell development perturbations of epitranscriptomic writers affect broad swaths of gene expression [93], it appears as though this effect can also be used to direct specific cohorts of biological pathways in cancer. In glioblastoma, PUS7 is more highly expressed in cancer tissue compared to control tissue, and its high expression correlates with poor prognosis. Knockout of PUS7, instead of inducing global alterations in protein translation, led to codon-specific translation alterations that were vital in regulating glioblastoma stem cells (GSCs) [94]. Interestingly, a catalytic-independent role for PUS7 has also been discovered in CRC [95]. In probing for enhanced mechanistic understanding of the function of pseudouridine, sequencing of the marks revealed enrichment at RNA regulatory regions of nascent pre-mRNA. Pseudouridine was found at sites of alternative splicing and RNA binding sites, further supporting the role for such modifications in regulating gene expression [13]. A new method using bisulfite-induced deletion sequencing (BID-seq) enabled stoichiometric mapping of pseudouridine on mouse tissue mRNA with low sample input [96]. Using this method, pseudouridine was found to increase stability of mRNA, and writer-specific pseudouridine deposition could be characterized.

RNA editing

The facilitation of RNA base editing from adenosine to inosine is carried out by the ADAR family of proteins ADAR1, ADAR2, and ADAR3, with ADAR1 being the most actively studied [97] and ADAR2 that may be tumor-inhibitory [98]. The ADAR family of proteins alter gene expression by modulating sequence and structure of both RNA and protein [21, 22]. When found within the coding sequence, inosine is interpreted as guanosine, leading to codon changes that subsequently alter the amino acid sequence and affect protein production. During RNA processing, RNA editing leads to alternatively spliced RNA [21]. Structurally, inosine preferentially pairs with cytidine instead of adenosine to uridine [21]. Like other RNA modification writers, ADAR1 and ADAR2 have both tumor suppressive and tumor promoting functions, shown by knockdown studies [20]. A large portion of RNA editing occurs in the non-coding portions of RNA. A mechanism prominent in human disease, ADAR1 is used to edit endogenous double-stranded RNA (dsRNA) to avoid activation of the innate immune system that would typically recognize dsRNA as non-self and mark for destruction [97, 99]. Cancer cells adopt this process to avoid perception as non-self.

A genome-wide screen of dependencies in multiple lung cancer cells placed ADAR1 as a common node of susceptibility in cells with high expression of interferon response genes [100]. In cancer, depletion of ADAR1 has shown varying effects on tumor growth. In a solid tumor model, where ADAR1 depletion has led to negligible effects on tumor volume, tumors demonstrated a markedly enhanced response to PD-1 antibody treatment [101]. Loss of ADAR1 also was able to mitigate common methods of resistance to immune checkpoint blockade using PD-1 antibody treatment [101]. In a model of triple-negative breast cancer (TNBC), loss of ADAR1 led to decreased cell proliferation, however this was not completely dependent on the RNA editing effects of ADAR1 or on the interferon response [102]. Instead, a concomitant effect on interferon response and on protein translation was seen. Similarly, in lung cancer, catalytic mutant ADAR1 partially rescued the effect of ADAR1 depletion, suggesting both catalytic and noncatalytic functions of ADAR1 [100]. A relationship between m⁶A and ADAR1 is present, whereby deaminase independent ADAR1 is a target of METTL3 in glioblastoma, and perturbation impairs glioblastoma growth [103].

Optimizing anticancer therapy by targeting RNA modifications

Pharmacologic approaches to targeting m⁶A

Small molecules that exhibit specificity against RNA modification machinery are key tool compounds to dissect roles of these modifications in different cancer types. A prominent development in the targeting of RNA modifications was the development of a catalytic inhibitor against METTL3 [104] (Table 1). Catalytic inhibitors function to interfere with the methyltransferase dependent function of the METTL3/METTL14 complex (Figure 2). The first bioavailable synthesized small molecule STM2457 was tested in vitro against a panel of structurally similar methyltransferases and exhibited specificity primarily to the METTL3/METTL14 complex. What was distinctly promising about this small molecule was the ability to target bulk AML as well as leukemia stem cells in vitro and in vivo without notable differences in animal weight, impact on bone marrow-derived hematopoietic stem cells, or early progenitor cells. Mice treated with the small molecule at a dose of 50mg/kg showed decreased disease burden while leaving hematopoietic stem cell precursors (HSCPs) untouched and with no appreciable toxicity to animals [104]. Since then, an orally bioavailable small molecule, STC-15, was synthesized in the United Kingdom that is now in active multi-center phase I clinical trials to assess for safety in advanced tumors (NCT05584111). Other patents filed by companies Accent Therapeutics and Storm Therapeutics suggest an influx of newly synthesized small molecules with potential for enhanced specificity [105–111]. Structure-based screening also identified small molecule UZH2, which was tested in a leukemia and prostate cancer cell line and decreased m⁶A levels on mRNA without affecting other related modifications [112]. Virtual screening has detected small molecules amentoflavone and methyl heperidin that could be considered as tool compounds, though these molecules have not been validated [113]. Adenosine analogues have been described, albeit with lower sensitivity, as well as allosteric inhibitor 43n, which showed an antiproliferative effect in AML cell lines [114, 115]. Eltrombopag, a medication typically used to treat cytopenia, demonstrated properties of noncompetitive inhibition of METTL3 [116]. Additional proposed METTL3 inhibitors were less specific, such as the general nucleoside analogue sinefungin and s-adenosylhomocysteine, with additional molecules reviewed by Fiorentino et al. [117, 118].

Figure 2. Targeting the catalytic activity of the METTL3/METTL14 complex.

Small molecule inhibition of the METTL3/METTL14 m⁶A methyltransferase complex interferes with co-transcriptional deposition of m⁶A on target transcripts. Catalytic inhibition does not directly mediate the scaffold effect of METTL3. In the cytoplasm, a catalytic-independent function of METTL3 is to recruit proteins such as translation initiation factors to enhance translation of target transcripts. Created with BioRender.com.

Many m⁶A-mediated tumorigenic effects are mediated by downstream reader proteins. Identification of additional facets to the m⁶A-reader axis could help define therapeutic specificity. The fact that m⁶A-marked RNA can be read by multiple different reader proteins suggests that targeting specific pathways has potential to increase disease selectivity. Early studies shed light onto this possibility. Screening for molecules that fit inside the hydrophobic pocket of the YTH family of proteins identified ebselon, a small molecule that was able to interact with each YTHDF paralog [119]. Identification of the small molecule inhibitor salvianolic acid C (SAC) was able to ablate condensates containing YTHDF1 and inhibit hyperactive YTHDF1 in neurons, demonstrating the potential for small molecule therapy against YTHDF1 [73]. Despite YTHDF1 being structurally similar to YTHDF2, SAC showed multi-fold increase in specificity toward YTHDF1. While small molecules have been identified as a YTHDF2-based therapeutic dependency, therapies that directly and specifically target YTHDF2 have not been identified to date [67, 120]. In an alternative strategy, YTHDF2 siRNA was conjugated to a toll-like receptor 9 (TLR9) agonist, that when internalized by antigen presenting cells slowed tumor growth and decreased metastasis [65].

A role for reader protein YTHDC1 has been described in AML, where m⁶A facilitated YTHDC1-containing condensates, restraining AML cells in an undifferentiated state [121]. These findings, together with those from YTHDF1 suggest disruption of phase separation may be a tenable mechanism to disrupt m⁶A-mediated oncogenesis. Multiple inhibitors have been identified against FTO and tested in various cancers, most recently CS1 and CS2, where at low nanomolar levels showed specificity against FTO in AML [122].

Targeting additional RNA modification proteins

In an effort to identify compounds with enzymatic activity against PUS7, a virtual screening platform of over 270,000 chemicals was used [94]. Prospective small molecules from the screen were then tested using an in vitro assay to assess target specificity, identifying compound NSC107512, or C17, as a top candidate. Addition of C17 decreased growth of GSC-derived tumors and extended lifespan of mice [94]. Though additional inhibitors of pseudouridine synthase enzymes have been proposed, their intracellular targets are rather widespread [123]. Analogues of adenosine substrates have been proposed as ADAR inhibitors [124]. However, characterization of these small molecules have questioned the specificity of both [125]. Three FDA-approved compounds were also predicted to target the interferon-responsive catalytic structure Zα domain of ADAR1, though untested in a biological system [126].

Implementation of targeted RNA modification therapies into the clinic

The failure of phase I clinical trials in cancer reaches approximately 90 percent [127]. Furthermore, limited scientific reproducibility hinders reliable drug development. It is imperative for those involved in preclinical target identification and drug design to consider the end goal of treatment response throughout initial mechanistic studies. Incorporating appropriate positive and negative controls, performing rescue experiments, and validating on-target effect are critical to improve preclinical studies [128]. In the case of targeting RNA methyltransferase proteins, catalytic inactive mutants have helped to dissect important functions of individual proteins and predict response to catalytic inhibition. Knockout of METTL3 in mES cells is embryonic lethal, consistent with the finding that loss of METTL3 can lead cells to lose self-renewal capabilities [9]. It is possible that the effect is a combination of the catalytic activity of METTL3 and its scaffold effect; catalytic inhibition could enhance specificity so that cancerous tissue is affected, leaving minimal toxicity to non-cancerous tissue. Indeed, a common theme among expression of RNA modification-inducing proteins is that the cancer cell often possesses higher levels of any given RNA modification enzyme compared to non-cancerous tissue. This has been demonstrated in the case of the m⁶A methyltransferase complex, enzymes that induce pseudouridylation, and RNA editing proteins, enabling a potential therapeutic window. If elimination of the whole protein is desired, protein degradation technology is an alternative approach [129]. This technique has the potential to disrupt both methyltransferase-dependent and methyltransferaseindependent functions of methyltransferase proteins.

The therapeutic course of a patient’s disease presents multiple opportunities for use of RNA modification mediators to enhance response to therapy (Figure 3). METTL3 inhibition can synergize with common chemotherapy agents, radiation therapy, and in the case of reader proteins aid in the process of hematopoietic stem cell transplant. Leveraging RNA modifications to mitigate common barriers to anti-cancer therapy should be considered in future work. In breast cancer, YTHDF1 promoted resistance to standard chemotherapy agents, yielding an opportunity to deter resistance in the setting of YTHDF1 depletion [130]. RNA sequencing analysis of patients with AML found METTL3 to be a mediator of chemoresistance in patients with relapsed disease, an effect that was dependent on the catalytic activity of METTL3 [131]. Treatment with catalytic METTL3 inhibitor STM2457 reduced proliferation and increased apoptosis of chemoresistant AML cells. The orally available METTL3 inhibitor currently in phase I clinical trials, STC-15, also enhanced response to treatment with BCL2 inhibitor Venetoclax, an agent used in AML therapy protocols (Storm Therapeutics). Opportunities to reduce dosages of standard chemotherapy that carry lifetime toxicities could be opened in presence of strategies that reduce their need.

Figure 3. Using RNA modification enzymes to amplify the anti-tumor response of standard therapies.

Modifications mediated by m⁶A writers METTL3 and METTL14, pseudouridine synthase PUS7, m⁶A demethylases FTO and ALKBH5, reader proteins YTHDF1 or YTHDF2, and enzymes responsible for adenosine to inosine (A to I) RNA editing can synergize with existing modalities of treatment. Enhanced anti-tumor activity in the preclinical setting has been observed when combined with strategies such as standard chemotherapy and radiotherapy-induced DNA damage. Additional anti-tumor activity has been observed in combining depletion of select proteins with immune checkpoint inhibitors and by stimulating the host anti-tumor response. RNA modifications can also mediate differentiation of some cancer cells from a stem cell-like state. Created with BioRender.com.

Concluding Remarks

The arc of progression from a gene unit to functioning RNA and protein biomolecules is traditionally characterized as the output of transcription and translation. We discover that this process is more interconnected than a silo-like separation of systems. An extraordinarily functionally diverse collection of RNA alterations serve as a bridge to direct gene expression, which has a significant impact on tumor formation. When considering the efficacy of cancer treatment approaches, the targeting of a single identified driver has frequently failed for a variety of reasons, including the development of resistance and intra- or intertumoral heterogeneity. Hence, it is advantageous to consider nodes that can influence a shared network of targets within a tumor. Such is the possibility of RNA modification targeting. Furthermore, it appears that these methyltransferase-depositing proteins may be more active or expressed at higher levels in cancer cells compared to matched non-cancerous tissue.

Several studies have attributed the anti-tumor activity of methyltransferase depletion to catalytic function, demonstrating that pharmacologic approaches to targeting these proteins is a promising technique. Additional medicinal chemistry and characterization efforts are needed to develop new relevant small molecule inhibitors as well as enhance the efficacy of the existing ones, assess safety profiles, and identify responsive cancers (see Outstanding questions). In the future, it will be essential to consider clinical and genetic biomarkers to identify optimal patient selection and evaluate treatment response. Overall, targeting the catalytic activity of RNA modification proteins is a promising anti-cancer treatment strategy.

Outstanding Questions.

• Initial studies used genetic methods of depletion to characterize the role of RNA modifications in cancer. Will the phenotypic findings from these studies be recapitulated with pharmacologic based activity inhibition?

• Writers, readers, and erasers of RNA modifications are often more highly expressed in cancers compared to matched tissue controls. However, expression is not always limited to cancer tissue. Is there an appropriate therapeutic window to target these RNA modification proteins without inducing toxicity to patients?

• The METTL3/METTL14 complex has shown both pro- and anti-tumorigenic functions in some cancers, such as glioblastoma. What accounts for this discrepancy?

• Though most m⁶A deposition is catalyzed by the methyltransferase complex composed of METTL3 and METTL14, other proteins such as METTL16 have been shown to install m⁶A in cancers. Is there a redundant role for methyltransferase proteins that cancer cells could exploit to induce resistance to pharmacologic inhibition?

Highlights.

• Epitranscriptomics is the study of a diverse set of chemical modifications to RNA that are installed, detected, and removed by a series of proteins.

• RNA modifications mediate the expression of genes by affecting a diverse set of cellular processes.

• Cancer cells commonly express high levels of certain writers, readers, and erasers of RNA modifications compared to matched non-cancerous tissue, and high expression is often associated with poor prognosis.

• Preclinical evidence provides proof-of-concept that therapeutic targeting of RNA modifications is a promising anti-cancer strategy and has potential to synergize with existing treatment modalities.