RNA epitranscriptomics: Regulation of infection of RNA and DNA viruses by N⁶-methyladenosine (m⁶A)

Summary

N⁶-methyladenosine (m⁶A) was discovered 4 decades ago. However, the functions of m⁶A and the cellular machinery that regulates its changes have just been revealed in the last few years. m⁶A is an abundant internal mRNA modification on cellular RNA and is implicated in diverse cellular functions. Recent works have demonstrated the presence of m⁶A in the genomes of RNA viruses and transcripts of a DNA virus with either a proviral or antiviral role. Here, we first summarize what is known about the m⁶A “writers,” “erasers,” “readers,” and “antireaders” as well as the role of m⁶A in mRNA metabolism. We then review how the replications of numerous viruses are enhanced and restricted by m⁶A with emphasis on the oncogenic DNA virus, Kaposi sarcoma-associated herpesvirus (KSHV), whose m⁶A epitranscriptome was recently mapped. In the context of KSHV, m⁶A and the reader protein YTHDF2 acts as an antiviral mechanism during viral lytic replication. During viral latency, KSHV alters m⁶A on genes that are implicated in cellular transformation and viral latency. Lastly, we discuss future studies that are important to further delineate the functions of m⁶A in KSHV latent and lytic replication and KSHV-induced oncogenesis.

1 |. N⁶-METHYLADENOSINE STUDIES PRIOR TO NEXT GENERATION SEQUENCING

More than 100 posttranscriptional chemical modifications are present on RNA from all kingdoms of life. Most of these modifications are found on ribosomal RNA (rRNA) and transfer RNA (tRNA), which modulate their structures and functions, and hence translation, as they are accessory molecules in these processes.¹ Messenger RNA (mRNA), which is primarily an information-bearing molecule, is also posttranscriptionally modified albeit with fewer types of modifications compared with other RNA species.¹ Early studies on mRNA modifications revealed that N⁶-methyladenosine (m⁶A) was the most abundant internal modification on poly(A) RNA in hepatoma cells and mouse myeloma cells.^2-4 m⁶A was subsequently detected in both adenovirus and influenza A virus (IAV) with an average of 3 m⁶A modifications per viral mRNA in IAV, a level that is similar to that of cellular m⁶A.^5-7

Further studies in the 1970s to 1980s detected m⁶A in the RNA of human cancer cell lines, mouse white blood cells, bovine mRNA, mosquito cells, and in a variety of viruses that replicate in the nucleus such as herpes simplex virus type 1 (HSV-1), Rous sarcoma virus (RSV), simian virus 40 (SV40), B77 avian sarcoma virus, and feline leukemia virus.^3,4,8-28 These early studies showed that viral transcripts contain m⁶A levels similar to cellular RNA. Two studies mapped a cluster of 7 m⁶A bases on the src- and env-coding regions of RSV RNA at the single nucleotide level and revealed that each site was heterogeneously methylated, indicating different stoichiometries for each m⁶A site. These results suggested potential host-pathogen interactions that converge on viral RNA.^9,18 It was also revealed that in cells infected by adenovirus, viral nuclear pre-mRNA had a higher level of m⁶A than viral cytoplasmic mRNA (2.5 vs 1.5 m⁶A bases/transcript), hinting that either m⁶A located in intronic regions is lost during splicing or m⁶A methylation of mRNA is a dynamic process.^6,7 Since these early studies did not map m⁶A at the transcriptome-wide level, and no knowledge of the methyltransferases, demethylases, or m⁶A-binding proteins was available, the functions of m⁶A remained elusive for decades. Nevertheless, results of these studies have pointed to potential important roles of m⁶A in the life cycle of RNA and DNA viruses, which will be the focus of the current review.

2 |. NEXT GENERATION SEQUENCING UNRAVELS THE m⁶A EPITRANSCRIPTOME

Transcriptome-wide mapping of m⁶A was unavailable until N⁶-methyladenosine-sequencing (m⁶A-seq) was developed by 2 independent groups in 2012.^29,30 In this technique, total RNA or poly(A)-selected RNA was isolated from cells and fragmented to approximately 100 nucleotides. Then, an m⁶A-specific antibody was used to pull down the fragmented RNA followed by deep sequencing of the immunoprecipitated and input fractions. m⁶A peaks on the transcripts were determined by comparing immunoprecipitated and input reads. Both studies found that m⁶A on cellular mRNA was enriched in the 3′ untranslated region (UTR) and with RRm⁶ACH motifs. As an epitranscriptomic mark, m⁶A was found on orthologous genes in both human and mouse cell lines.²⁹ Analysis of methylated genes revealed enrichments of pathways related to RNA metabolism, transcriptional regulation, splicing, and developmental pathways.^29,30 One difference between these two studies was that m⁶A was found to be enriched at the 5′UTR near the transcription start site in the Dominissini et al study but not in the Meyer et al study.^29,30 This difference was attributed to different peak calling methods, but it could also be due to the limitation in m⁶A-seq, which is unable to distinguish m⁶A from another RNA modification, N⁶,2′-O-dimethyladenosine (m⁶Am), mostly found on the first few nucleotides of mRNA.^29,31 Due to the resolution of m⁶A-seq, which is limited to 100 to 200 nucleotides, clusters of m⁶A within 200 nucleotides cannot be sufficiently resolved.

Since then, other techniques have been invented to overcome the limitations of m⁶A-seq (Table 1). One of them called photo-crosslinking-assisted m⁶A sequencing (PA-m⁶A-seq), which is involved with UV cross-linking the anti-m⁶A antibody to poly(A) RNA from cells grown in media containing the photoactivatable ribonucleosides, 4-thiouridine (4SU), or 6-thioguanosine (6SG), enhances the resolution to approximately 23 nucleotides.³² A few months later, a technique called m⁶A individual nucleotide resolution cross-linking and immuno-precipitation (miCLIP) was published, which enabled transcriptomewide mapping of m⁶A or m⁶Am at a single nucleotide resolution.³³ This technique improved on previous techniques by UV cross-linking the anti-m⁶A antibody to RNA and then digesting all but a small part of the antibody in contact with the RNA using proteinase K. The remaining antibody fragment caused mutations during the preparation of a sequencing library, resulting in an antibody-dependent mutational signature or a truncation in sequencing reads close to m⁶A or m⁶Am sites. The truncations allow this technique to differentiate between m⁶A and m⁶Am. In contrast to m⁶A, m⁶Am is not found within the RRACH motif; instead, it is present around BCA motifs and mainly in the 5′UTR; thus, it is not able to generate the same mutational signature as m⁶A. Another technique, m⁶A level and isoform characterization sequencing (m⁶A-LAIC-seq), enables the quantification of the stoichiometry and identification of isoforms of methylated transcripts by using excess anti-m⁶A antibody to pull down full-length RNA.³⁴ Deep sequencing of the pull down and flow-through fractions enables quantification of methylated vs unmethylated transcripts at the transcriptome-wide level, without site-specific mapping of m⁶A. They found diversity in m⁶A stoichiometry among different cell types and that methylation influences the choice of alternative polyadenylation sites. Another group capitalized on m⁶A's slight interference on A-T/A-U base pairing and developed a tiling microarray technique to detect m⁶A.³⁵ Tiling RNA or DNA probes of 25 nucleotides in length complementary to the RNA sequences of interest was constructed and hybridized to the target RNA. This technique avoids the biases introduced by antibody-based techniques. However, it is less sensitive compared with the antibody-based techniques.

TABLE 1.

Comparison of current m⁶A transcriptome-wide profiling techniques

Technique	Description	Advantages	Disadvantages	Reference
m6A-seq	Immunoprecipitation of fragmented RNA with an anti-m6A antibody followed by deep sequencing.	High sensitivity, widely adopted.	Resolution at 100-200 nucleotides; Does not discriminate between m6A and m6Am; False positives due to nonspecific antibody interactions.	Dominissini et al29 and Meyer et al30
PA-m6A-seq	Cells grown in 4SU or 6SG followed by UV cross-linking of isolated RNA to an anti-m6A antibody. Mutations generated during library preparation offer improved resolution over m6A-seq.	Approximately 23 nucleotide resolution.	Incorporation of 4SU or 6SG requires live cells; Does not discriminate between m6A and m6Am; False positives due to nonspecific antibody interactions.	Chen et al32
miCLIP	UV cross-linking of an anti-m6A antibody to RNA during the immunoprecipitation step results in mutations or truncations during the preparation of sequencing library; Detecting precise location of m6A/m6Am.	Single nucleotide resolution; Can distinguish between m6A and m6Am.	Mutational signature is dependent on antibody type; False positives due to nonspecific antibody interactions.	Linder et al33
m6A-LAIC-seq	Full-length RNA is immunoprecipitated with excess anti-m6A antibody followed by sequencing of immunoprecipitated and flowthrough fractions.	Identifies stoichiometry of m6A and methylated vs unmethylated isoforms.	Since full-length transcripts are used, site-specific detection of m6A is not possible.	Molinie et al34
Microarray	A tiling array of RNA/DNA probes of 25 nucleotides complementary to RNA of interest is generated. m6A disrupts A-T or A-U base pairing, resulting in weaker hybridization with the probe.	Free from nonspecific interactions of antibody-based methods.	Low sensitivity.	Li et al35

To date, m⁶A, which is cotranscriptionally added to nascent RNA,^36,37 has been shown to affect alternative splicing by promoting exon inclusion or skipping in a splice factor-dependent manner,^38,39 nuclear export,⁴⁰ both cap-dependent and cap-independent translations,^41,42 miRNA biogenesis and binding,^43,44 and P body-mediated RNA degradation.^45,46

Using the modern techniques outlined above, evidence of dynamic regulation of the m⁶A epitranscriptome has been shown during DNA damage,⁴⁷ heat shock,^42,48 response to interferon-γ, stem cell differentiation,^49-52 spermatogenesis and oogenesis,^40,53-56 yeast sporulation,⁵⁷ circadian rhythm,⁵⁸ and plant development.^59-61 In the remaining parts of the review, we discuss the role of cellular m⁶A machinery and regulation of viral replication of RNA and DNA viruses by cellular m⁶A machinery.

3 |. CELLULAR m⁶A MACHINERY

The cellular machinery driving the m⁶A dynamics can be divided into 4 main groups: methyltransferases or “writers,” demethylases or “erasers,” m⁶A-binding proteins or “readers,” and m⁶A-repelled proteins or “antireaders” (Figure 1). In humans, the m⁶A writer is a large, nearly 1 mega Dalton complex; however, only a few subunits have been identified to date—methyltransferase-like 3 (METTL3), methyltransferaselike 14 (METTL14), Wilms tumor 1–associating protein (WTAP), Vir-like m⁶A methyltransferase associated (KIAA1429), RNA-binding motif protein 15 (RBM15), and RNA-binding motif protein 15B (RBM15B).^62-67 Of these subunits, METTL3 and METTL14 contain domains of methyltransferase, but only METTL3 is catalytically active. METTL14 forms a heterodimer with METTL3 and plays a role in substrate recognition.^68,69 WTAP, KIAA1429, RBM15, and RBM15B function as regulatory subunits for this complex and are likely involved in the selective methylation of m⁶A sites.^65,67 Knockdown of WTAP decreased the amount of METTL3 and METTL14 found in nuclear speckles, suggesting a possible role for WTAP in stabilizing the methyltransferase complex.⁶⁶ Depletion of WTAP or KIAA1429 in the cell decreases the amount of m⁶A, and RBM15/RBM15B is required for methylation of the long noncoding RNA X inactive specific transcript (XIST).^65-67 Since the RRACH motif is prevalent throughout the transcriptome, it is still not well understood how specificity for an m⁶A site is achieved in different physiological conditions. Under steady-state conditions, the methyltransferase complex is localized to the nucleus; however, METTL3 is also present in the cytoplasm of cancer cells and is associated with eIF3 to enhance translation.⁷⁰ This demonstrates that the localization of METTL3 could be cell type or cell condition dependent and that it may have other functions in addition to its role as a methyltransferase.

FIGURE 1.

Cellular m⁶A machinery and their roles in mRNA metabolism. In steady-state cells, the methyltransferase complex (“writers”) and demethylases (“erasers”) are localized in the nucleus, which can affect splicing and nuclear export of mRNAs. The nuclear reader YTHDC1 has been implicated in both splicing and nuclear export. In the cytoplasm, YTHDF3 can recruit either YTHDF1 or YTHDF2 to promote translation or degradation of mRNA, respectively. YTHDC2 promotes both translation and degradation of mRNA

To date, 2 m⁶A “erasers” have been characterized. Fat mass and obesity associated (FTO) and AlkB homolog 5 (ALKBH5) are oxygen-, alpha-ketoglutarate-, and iron-dependent enzymes.^40,71 Before its function as an “eraser” was known, a single nucleotide polymorphism (SNP) in the first intron of FTO was shown to be strongly correlated with obesity.⁷² Since it does not affect the protein coding sequence of FTO, the mechanism behind FTO's effect on obesity remains unknown.⁷³ Recently, one SNP in the first FTO intron was shown to affect long-range DNA-DNA interactions in the promoter of Iroquois homeobox 3 (IRX3), resulting in the increase of IRX3 expression and an obesity phenotype.⁷⁴ FTO deficiency in mouse models and humans resulted in growth retardation, malformations, metabolic changes, and abnormal neuronal signaling.^75-77 After its role as a demethylase was elucidated, FTO was shown to affect global alternative splicing³⁸ and to play a role in adipogenesis by regulating the alternative splicing of RUNX1 translocation partner 1 (RUNX1T1).⁷⁸ Recently, it has been reported that FTO prefers demethylation of m⁶Am over m⁶A in vivo, leaving ALKBH5 as the sole m⁶A demethylase.⁷⁹ Knockdown of ALKBH5 in HeLa cells accelerated mRNA export from the nucleus, and male ALKBH5-deficient mice were infertile due to aberrant spermatogenesis.^40,80 In spermatocytes, ALKBH5 is essential for the correct splicing of the 3′UTR of transcripts.⁸⁰ ALKBH5 also promotes the maintenance of glioblastoma stem-like cells by upregulating forkhead box M1 (FOXM1) expression.⁸¹ Another study found that ALKBH5 expression was induced during hypoxia, leading to increased Nanog homeobox (NANOG) expression in breast cancer stem cells.⁸²

Among the “reader” proteins, the YTH N⁶-methyladenosine RNA-binding protein family YTHDF1, YTHDF2, YTHDF3, YTHDC1, and YTHDC2 have been most well studied. YTHDF1 recruits the translation preinitiation complex to methylated transcripts to enhance translation⁴¹; YTHDF2 promotes the degradation of methylated transcripts by shuttling them to the CCR4-NOT complex and DCP1/2 in P bodies^45,46; YTHDF3 facilitates translation or degradation of methylated transcripts by binding directly to YTHDF1 or YTHDF2^83,84; YTHDC1 regulates RNA splicing by competitively recruiting serine-and arginine-rich splicing factor 3 or 10 (SRSF3 or SRSF10) to mRNA and mediates nuclear export by interacting with nuclear RNA export factor 1 (NXF1)^39,85; and YTHDC2 affects both translation and degradation of RNA by interacting with the 40-80S subunit and XRN1, respectively.^55,86 Recently, 2 independent studies screened for additional “reader” proteins and characterized FMR1 as a sequence context-dependent m⁶A reader that promotes translation of methylated transcripts.^87,88 Interestingly, a class of “antireader” proteins have been discovered where they preferentially bind to GGACU motifs in the absence of m⁶A.^87,88 Two of these proteins, G3BP stress granule assembly factor 1 (G3BP1) and G3BP stress granule assembly factor 2 (G3BP2), are stress granule proteins that stabilize unmethylated transcripts.^87,88

To date, numerous studies have revealed a role of m⁶A in viral replication by modulating the levels of “writers,” “erasers,” and “readers” in cells. As viruses hijack cellular pathways to favor their replication, it is not surprising that these proteins either promote or inhibit viral replication, depending on the virus or infected cell type. Conversely, it is also possible that the host cells use m⁶A and its associated proteins as an antiviral mechanism to restrict viral replication.

4 |. REGULATION OF REPLICATION OF RNA VIRUSES BY m⁶A

As intracellular parasites, viruses depend on cellular machinery for replication. Since the “writers” and “erasers” are found in the nucleus of resting cells, it is assumed that only viruses that replicate in the nucleus, eg, HIV, IAV, adenovirus, and HSV-1 can use m⁶A in their life cycle. Indeed, m⁶A was not detected on the genomes of RNA viruses that replicate in the cytoplasm such as vesicular stomatitis virus, reovirus, and vaccinia virus in studies done prior to next generation sequencing.^89-91 However, a more recent work with hepatitis C virus (HCV) and liver cancer cell lines shows that the methylation machinery and the demethylase FTO are present in the cytoplasm.⁹²

4.1 |. HIV

Three independent groups have characterized the involvement of m⁶A in HIV replication using CD4+ T cells, 293T cells, and HeLa cells.^93-95 All 3 studies showed an enrichment of m⁶A at the 3′ UTR of HIV-genomic RNA, but two of the studies showed additional m⁶A sites throughout the viral genome.^94,95 This discrepancy could be due to the different mapping techniques used. Kennedy et al used PA-m⁶A-seq, whereas the other 2 studies used m⁶A-seq. Cell line or HIV strain could also cause variations even though all 3 studies included CD4+ T cells. Two of these studies demonstrated a proviral role of m⁶A as knockdown of METTL3/METTL14 decreased viral replication while knockdown of ALKBH5 had the opposite effect.^93,94 All 3 YTHDF proteins were shown to bind to HIV RNA and favor HIV replication.^93,94 However, the study by Tirumuru et al showed that all 3 YTHDF proteins antagonize HIV replication.⁹⁵ The reason for this discrepancy is unclear but could be attributed to the use of a genetically modified virus that contains a luciferase reporter in its genome. Lichinchi et al also characterized the function of 2 potential m⁶A sites within the Rev response element (RRE) of HIV. The presence of m⁶A on the RRE enhanced binding of Rev to viral RNA, facilitating export of viral RNA. However, m⁶A mapping done by the 2 other studies failed to identify these m⁶A sites on the RRE. It is possible that the plastic nature of the HIV genome may contribute to this discrepancy; however, the RRE is a relatively stable region of the HIV genome. If m⁶A is a positive regulator of HIV replication, it should be evolutionarily conserved in this polymorphic virus.

4.2 |. Flaviviruses

Two independent groups simultaneously published epitranscriptomic maps of m⁶A on flaviviral genomes and proposed m⁶A-related mechanisms that regulate the replication of Zika virus (ZIKV) and HCV.^92,96 The m⁶A profiles in the genomes of the positive single-stranded flaviviruses such as HCV, Dengue virus, yellow fever virus, ZIKV, and West Nile virus are conserved.⁹² Knockdown of METTL3 or METTL14 in host cells enhanced the viral titers of HCV and ZIKV. Knockdown of FTO lowered the viral titers of both viruses, whereas knockdown of ALKBH5 had no effect on HCV viral production but lowered ZIKV viral titers in the supernatants. Since flaviviruses replicate in the cytoplasm, both studies presented evidence that the “writers” and “erasers” can be found in the cytoplasm of the host cells. Despite the different cellular functions of the YTHDF1, YTHDF2, and YTHDF3 “readers,” all of them negatively impact HCV and ZIKV replication. In the context of HCV, Gokhale et al mapped the binding sites of YTHDF1, YTHDF2, and YTHDF3 on the viral genome and showed that these proteins compete with HCV core protein for binding to regions on the Env gene to suppress packaging of viral RNA into new virions. This suppressive effect was hypothesized by Gokhale et al to be advantageous for HCV infection as a slower replication rate reflects chronic infection in the liver. The study by Lichinchi et al revealed 5′UTR hypermethylation of host transcripts after ZIKV infection. In addition, host immune-related transcripts were dynamically modified during ZIKV infection, indicating that m⁶A is involved in promoting an antiviral response. It is also possible that the virus usurps m⁶A machinery to suppress the host antiviral response by recruiting “writers” or “erasers” to specific cellular transcripts.⁹⁶ In contrast to HIV, m⁶A negatively affects flavivirus replication by affecting viral packaging.

4.3 |. Influenza A virus

A work by Courtney et al showed that inhibition of methylation with 3-deazaadenosine (3DAA) and METTL3 knockout in A549 lung cancer cells decreased the replication of IAV by reducing both viral mRNA and protein levels.⁹⁷ Contrary to YTHDF2's role in promoting RNA degradation, overexpression of YTHDF2 during IAV infection enhanced levels of viral mRNA, protein, and the release of infectious virions. It is possible that YTHDF2-mediated mRNA degradation decreases host antiviral gene transcripts, thus enhancing viral replication. Overexpression of YTHDF1 and YTHDF3 had no effect on IAV replication and viral production even though all 3 “readers” were found to bind to viral RNA. Using the photoactivatable ribonucleo-side-enhanced cross-linking and immunoprecipitation (PAR-CLIP) and PA-m⁶A-seq binding data of YTHDF1, YTHDF2, and YTHDF3, Courtney et al found numerous m⁶A sites on the negative sense vRNA and positive sense mRNA of IAV. They generated 2 mutant viruses by making silent mutations of m⁶A sites on the positive and negative sense RNAs of the hemagglutinin segment, respectively. However, they could not mutate all the m⁶A sites as some could introduce nonsynonymous mutations of the hemagglutinin protein. These two mutants had decreased levels of hemagglutinin protein, viral replication in culture, and IAV pathogenicity in a mouse infection model. The mechanism behind the positive effect of METTL3 or YTHDF2 on IAV replication remains unknown. The authors also investigated the possibility that methylation of viral RNAs might prevent the activation of innate immune sensors such as RIG-I or MDA5 but saw no additional activation of interferon-β when cells were infected with their mutant virions that carried fewer m⁶A sites. It is possible that YTHDF2 binding to viral RNAs sequesters it away from innate RNA immune sensors but no loss of function data was shown.

The function of YTHDC1 and YTHDC2 in the life cycle of RNA viruses has not been investigated so far. Since splicing is critical for HIV replication, it is likely that YTHDC1 could be involved in regulating this process to promote HIV replication, which would agree with the proviral role of m⁶A in the context of HIV. For flaviviruses, it is unclear if the nuclear “reader” YTHDC1 is present in the cytoplasm, but it is possible since the “writers” and “erasers” are involved in their cytoplasmic replicative cycles.

4.4 |. m⁶A in the life cycle of Kaposi sarcoma-associated herpesvirus

The role of m⁶A in the life cycles of RNA viruses has predominantly been investigated at the genomic RNA level in positive-stranded RNA viruses except for IAV, where both negative-stranded genomic RNA and positive-stranded mRNA have been investigated. DNA viruses such as Kaposi sarcoma-associated herpesvirus (KSHV), which replicates in the nucleus using the host machinery, is likely to usurp m⁶A machinery to promote its replication.

KSHV is the etiologic agent of Kaposi sarcoma, primary effusion lymphoma (PEL), multicentric Castleman disease, and KSHV-induced inflammatory cytokine syndrome (KICS).^98-101 KSHV latently infects endothelial progenitor cells, B cells, and mesenchymal stem cells.^98,102-107 Results of recent studies show that KS might originate from mesenchymal stem cells.^103,105,107 During viral latency, a few viral genes and a cluster of miRNAs are expressed,^108-112 which are essential for KSHV-induced cellular transformation.^105,113,114 Like other herpesviruses, KSHV latently infected cells can be reactivated into a lytic life cycle, expressing viral lytic genes in a cascade manner in the order of immediate early, early, and late genes that culminates in virion production.^115-117 Hence, KSHV lytic replication is a complex process regulated by multiple cellular processes, possibly including m⁶A and its related machinery.

Although most tumor cells are latently infected by KSHV in KS tumors, a small subset of them also undergo spontaneous lytic replication. These cells secrete viral cytokines such as viral interleukin-6 (vIL6) and induce proinflammatory and proangiogenic cytokines such as IL6, basic fibroblast growth factor (bFGF), and oncostatin M.^118-121 This cytokine milieu promotes the growth of KS cells via an autocrine and paracrine mechanism.¹¹⁸ Another lytic viral protein, KSHV G protein-coupled receptor (vGPCR), promotes the expression of VEGF to stimulate angiogenesis and survival.^122,123 Hence, KSHV lytic replication in a small subset of tumor cells is a key contributor to local inflammation and angiogenesis, which are the features of KS tumors. Understanding the alterations of viral and cellular m⁶A modifications during KSHV latent and lytic infection could provide insights into the mechanism of KSHV-induced tumorigenesis.

Using m⁶A-seq, we have mapped the KSHV m⁶A epitranscriptome during both viral latency in a variety of cell types, including primary cell lines. We have found that during latent infection, m⁶A is well conserved in KSHV transcripts across 5 cell lines and is predominantly found in the latent locus carrying latency-associated nuclear antigen (LANA), viral Fas-associated death domain (FADD)-like interleukin-1 betaconverting enzyme (FLICE)-inhibitory protein (vFLIP), and viral cyclin (vCyclin). The KSHV lytic epitranscriptome had widespread m⁶A methylation of viral genes. The lytic epitranscriptome was also well conserved between 2 models of lytic replication, KiSLK and BCBL1-R cells. The “reader” YTHDF2 was shown to suppress viral lytic replication by promoting the degradation of viral transcripts (Figure 2). Taken together, m⁶A is a novel viral restriction factor in the context of KSHV reactivation. Conflicting evidence was proposed by another study, which demonstrated a proviral role for m⁶A.¹²⁴ It was found that knockdown of METTL3 and inhibition of the methyltransferase complex with 3DAA in BCBL1 cells decreased viral lytic replication, whereas knockdown of FTO modestly increased viral lytic replication. This study also presented evidence for a role of YTHDC1 in promoting splicing of KSHV replication transcription activator (RTA, ORF50), an immediate early gene required and sufficient to induce KSHV lytic replication. Using RNA-binding protein immunoprecipitation and quantitative reverse transcription PCR (RIP-RT-qPCR), they showed that four m⁶A peaks on the RTA transcript are bound by YTHDC1 and mutation of each of these peaks results in decreased RTA splicing. This observation was made by overexpressing the mutant RTA transcripts in 293T cells. In the context of whole virus and by inducing the expression of an exogenous RTA gene with doxycycline, we did not observe an effect on viral lytic replication in YTHDC1-deficient KiSLK cells.

FIGURE 2.

The role of m⁶A during KSHV-induced cellular transformation and KSHV lytic replication. A, In cells that are latently infected and transformed by KSHV, we observed 5′ UTR hypomethylation and 3′ UTR hypermethylation. B, During viral lytic replication, KSHV mRNAs contain high levels of m⁶A. The “reader” protein YTHDF2 binds to and promotes the degradation of viral mRNAs. In YTHDF2-deficient cells, we observed an increase in the half-life of viral transcripts and increased production of virions

During lytic replication, we observed 5′UTR hypermethylation and a slight 3′ hypomethylation in KiSLK cells, whereas in BCBL1-R cells, there was 5′UTR hypomethylation and 3′UTR hypermethylation. The differences between these 2 cell lines could be attributed to the stronger activity of KSHV host shutoff, the result of an exonuclease (SOX) in BCBL1-R cells. It is possible that SOX activity is influenced by m⁶A, but further studies are needed to confirm this hypothesis. In both cell models, hypermethylated and hypomethylated pathways are important for KSHV lytic replication, such as adipogenesis, protein kinase A signaling, ILK signaling, ERK/MAPK signaling, PI3K/AKT signaling, and integrin signaling. Since many of these pathways are known to mediate KSHV lytic replication, m⁶A might be an additional mechanism regulating KSHV lytic replication via modulating these pathways. We did not observe any significant effect of m⁶Am on viral lytic replication.

We characterized the effect of KSHV latent infection on the cellular epitranscriptome in 4 pairs of uninfected and latently infected cells. The number of methylated genes and the m⁶A motifs did not change dramatically after viral latent infection of all 4 types of cells; however, there was 5′UTR hypomethylation and 3′UTR hypermethylation in 3 of the 4 cell lines (Figure 2). The 5’UTR hypomethylated and 3’UTR hypermethylated pathways are involved in oncogenic/mitogenic signaling, epithelial-mesenchymal transition, cytoskeleton and extracellular signaling, endocytosis, loss of contact inhibition, remodeling of adherens junctions, and cellular adhesion/invasion, indicating that m⁶A might regulate KSHV latency and cellular transformation. These observations are consistent with results of several recent studies showing the involvement of m⁶A in numerous types of cancer.^{70,81,125,126} Hence, it can be speculated that KSHV might hijack the m⁶A machinery to promote viral latency and induce cellular transformation. Overall, our study has provided a valuable epitranscriptomic resource for studying KSHV life cycle and KSHV-induced cellular transformation and tumorigenesis.

5 |. CONCLUSIONS AND PERSPECTIVES

We have only begun to understand how m⁶A regulates the life cycles of viruses. In large and complex viruses such as KSHV, it is possible that m⁶A on different viral transcripts mediate different functions, such as degradation, splicing, or translation. Assessing this at the global level as our study has done cannot tease apart these transcript-specific differences. Therefore, ablating methylation at specific sites on important viral transcripts such as LANA and RTA, for example, might provide more specific mechanisms on how m⁶A might regulate KSHV life cycle in future studies. Generation of point mutations in the context of whole viral genome by using the bacterial artificial chromosome (BAC) technology could achieve this goal.^127,128 Using the clustered regularly interspaced short palindromic repeats (CRISPR-cas9) technology to mutate viral DNA is possible but difficult because of the high copy number of KSHV episomes per cell.

KSHV encodes an RNA-binding protein, mRNA transcript accumulation protein (MTA or ORF57), which has functions analogous to m⁶A such as splicing, transcript stability, and translation. MTA-binding sites on PAN RNA, vIL6, and ORF59 have stem-loop structures; the presence of m⁶A might destabilize stem-loop formation, affecting its accessibility to RNA-binding proteins.^129-131 Therefore, it is possible that m⁶A might act as a repellent to MTA binding. Although this would be in contrast to m⁶A on HIV RRE, which favors binding of HIV Rev protein to its RNA, it remains possible particularly since m⁶A has both “reader” and “antireader” proteins.⁹⁴

Targeting “writers” and “erasers” with small molecule inhibitors is a potential approach for antiviral therapies. Various studies have shown that 3DAA, an methyltransferase inhibitor, can reduce viral replication^124,132,133; however, 3DAA also inhibits histone and DNA methylation.⁵⁸ A more specific inhibitor of the “writers” is required to eliminate off target effects. Similarly, an inhibitor of the “erasers,” meclofenamic acid, could be used to inhibit HCV replication since m⁶A antagonizes its replication. Both KSHV- and HCV-induced tumors are dependent on a small subset of cells undergoing viral lytic replication. Hence, inhibition of the m⁶A machinery might have an antitumor effect.

Viruses have coevolved with the host cells. How viruses hijack the m⁶A machinery remains to be understood. For example, during HCV replication, the “writers” and “erasers” are relocalized to the cytoplasm. These changes can lead to alterations in the host m⁶A methylome during viral infection or latency as we have shown in the context of KSHV infection. Current literature lacks studies on the functions of m⁶A in host-pathogen interactions, which could be especially important for oncogenic viruses. The extent of m⁶A's role in cellular transformation during oncogenic virus infection/latency remains unknown.

Abbreviations used:

3DAA

3-deazaadenosine

4SU

4-thiouridine

6SG

6-thioguanosine

ALKBH5

AlkB homolog 5

BAC

bacterial artificial chromosome

BCBL1-R

primary effusion lymphoma cell line

bFGF

basic fibroblast growth factor

CRISPR-Cas9

clustered regularly interspaced short palindromic repeats

FOXM1

forkhead box M1

FTO

fat mass and obesity associated

G3BP1

G3BP stress granule assembly factor 1

G3BP2

G3BP stress granule assembly factor 2

HCV

hepatitis C virus

HIV

human immunodeficiency virus

HSV-1

herpes simplex virus type 1

IAV

influenza A virus

IRX3

Iroquois homeobox 3

KIAA1429

Vir-like m⁶A methyltransferase associated

KICS

KSHV-induced inflammatory cytokine syndrome

KiSLK

latently infected renal carcinoma cells

KSHV

Kaposi sarcoma-associated herpesvirus

LANA

Latency-associated nuclear antigen

m⁶A

N⁶-methyladenosine

m⁶A-LAIC-seq

m⁶A level and isoform characterization sequencing

m⁶Am

N⁶,2′-O-dimethyladenosine

METTL3

methyltransferase-like 3

METTL14

methyltransferase-like 14

miCLIP

individual nucleotide resolution cross-linking and immunoprecipitation

mRNA

messenger RNA

MSC

adipose tissue-derived mesenchymal stem cells

MTA

mRNA transcript accumulation

NANOG

Nanog homeobox

NXF1

nuclear RNA export factor 1

PA-m⁶A-seq

photo-cross-linking-assisted m⁶A sequencing

PAR-CLIP

photoactivatable ribonucleoside-enhanced cross-linking and immunoprecipitation

PEL

primary effusion lymphoma

RBM15

RNA-binding motif protein 15

RBM15B

RNA-binding motif protein 15B

RIP-RT-qPCR

RNA-binding protein immunoprecipitation and quantitative reverse transcription PCR

RRE

Rev response element

rRNA

ribosomal RNA

RSV

Rous sarcoma virus

RTA

KSHV replication transcription activator

RUNX1T1

RUNX1 translocation partner 1

SOX

KSHV host shut off an exonuclease

SNP

single nucleotide polymorphism

SRSF3

serine-and arginine-rich splicing factor 3

SRSF10

serine-and arginine-rich splicing factor 10

SV40

simian virus 40

tRNA

transfer RNA

UTR

untranslated region

vCyclin

viral cyclin

vFLIP

KSHV Fas-associated death domain (FADD)-like interleukin-1 beta-converting enzyme (FLICE) inhibitory protein

vGPCR

KSHV G protein-coupled receptor

vIL6

viral interleukin-6

WTAP

Wilms tumor 1-associating protein

XIST

X inactive specific transcript

YTHDC1

YTH domain containing 1

YTHDC2

YTH domain containing 2

YTHDF1,

YTH N⁶-methyladenosine RNA-binding protein 1

YTHDF2

YTH N⁶-methyladenosine RNA-binding protein 2

YTHDF3

YTH N⁶-methyladenosine RNA-binding protein 3

ZIKV

Zika virus