The RNA Epitranscriptome of DNA Viruses

RNA modifications have generated much interest in the virology field, as recent works have shown that many viruses harbor these marks and modify cellular marks. The most abundant mRNA modification in eukaryotic cells, N⁶-methyladenosine (m⁶A), has been examined extensively at the genome-wide scale in both cellular and viral contexts.

ABSTRACT

RNA modifications have generated much interest in the virology field, as recent works have shown that many viruses harbor these marks and modify cellular marks. The most abundant mRNA modification in eukaryotic cells, N⁶-methyladenosine (m⁶A), has been examined extensively at the genome-wide scale in both cellular and viral contexts. This Gem discusses the role of m⁶A in gene regulation and describes recent advancements in Kaposi's sarcoma-associated herpesvirus (KSHV) and simian virus 40 (SV40) research. We provide insights into future research related to m⁶A in DNA viruses.

INTRODUCTION

The flow of information orchestrated in myriad molecular events affords the cell with opportunities to maintain a steady state in response to diverse intracellular and extracellular signals. As an intermediate molecule between DNA and protein, mRNA offers a critical junction for the cell to tightly regulate the flow of information. With the half-life of mRNA being as short as 1 h, the cell needs to closely sense and regulate levels of mRNA to either replenish the existing pool of mRNA through transcription or remove mRNA through degradation (1). Additionally, the flow of information can be regulated by mechanisms such as alternative splicing and alternative polyadenylation, giving rise to different mRNA and protein isoforms from a single gene. Furthermore, regulation of mRNA is spatial, through controlled export from the nucleus into the cytoplasm. An apparent question is how the cell regulates these complex processes. Recent advances reveal that part of the answer might lie in mRNA modifications.

A plethora of mRNA modifications, including N⁶-methyladenosine (m⁶A), N¹-methyladenosineosine (m¹A), pseudouridine, and 5-methylcytosine (m⁵C), were unearthed in the 1970s (2–4). m⁶A is the most abundant modification on mRNA and has been shown to influence all the mRNA processes discussed above (5). In fact, early m⁶A studies on viruses, such as influenza virus (6), Rous sarcoma virus (7), and simian virus 40 (SV40) (8), excited the field of RNA modifications. However, owing to technological limitations, elucidating the topology and function of m⁶A remained elusive until the advent of m⁶A sequencing (m⁶A-seq) (9, 10). This technique enabled the profiling of the global m⁶A epitranscriptome and led to the identification of key characteristics of m⁶A methylation. Methylation has been found on about one in four transcripts, within RRACH motifs, and on the entire length of mRNA, with enrichment in the 3′ untranslated region (3′UTR) and the stop codon (9–11). Analogous genes in both humans and mice are methylated with 49% conservation (9), suggesting the occurrence of rapid evolutionary changes over time. Since methylation occurs cotranscriptionally, newly synthesized RNA often contains m⁶A methylation in intronic regions near splice junctions (12). With the discovery of two m⁶A demethylases, encoded by FTO and ALKBH5 (13, 14), a dynamic model of m⁶A has been proposed, which elevates the complexity of epitranscriptomics. This Gem provides an overview of the field of m⁶A epitranscriptomics and the cellular m⁶A machinery as well as their roles in regulating gene expression in the cell. We then focus on recent developments in DNA virus epitranscriptomics, with an emphasis on Kaposi's sarcoma-associated herpesvirus (KSHV) and SV40.

CELLULAR m⁶A MACHINERY AND ITS ROLE IN GENE REGULATION

Methyltransferases.

The 70-kDa METTL3 protein was the first m⁶A methyltransferase to be identified and was known to be a subunit of a larger, ∼1-MDa complex that “writes” m⁶A onto mRNA (15, 16). It contains a binding pocket for the methyl group donor S-adenosylmethionine (SAM), adjacent to a methyltransferase domain. METTL3 exists in a heterodimer with METTL14, and depletion of either subunit significantly impairs methyltransferase activity in vitro (16, 17). This observation agrees with structural data showing that METTL14 has a catalytically inactive methyltransferase domain that enhances RNA binding of the METTL3-METTL14 dimer (18–21). Photoactivatable ribonucleoside-enhanced crosslinking and immunoprecipitation (PAR-CLIP) and in vitro methylation studies show that these proteins bind to and methylate mRNA at RRACH motifs (17). Methylation of RNA occurs as it is being transcribed, with intronic regions containing METTL3-dependent m⁶A sites that promote splicing (12, 22). Under steady-state conditions, the methyltransferase complex is localized in the nucleus. However, METTL3 has been observed in the cytoplasm of lung cancer cells, mouse embryonic stem cells (mESC), and hepatitis C virus (HCV)-infected cells (23–25). In lung cancer cells, METTL3 is associated with eIF3 at the 5′UTR of methylated transcripts and promotes translation (24). METTL3-METTL14-dependent methylation is critical for HIV replication (26, 27). A single m⁶A site at the Rev responsive element (RRE) on the HIV genome has been proposed to enhance binding of Rev to the genome and to promote nuclear export of HIV RNA (27).

With the RRACH motif occurring every 86th nucleotide on the transcript, how the METTL3-METTL14 complex preferentially methylates sites on mRNA is not well understood. However, new evidence indicates that multiple subunits of the methyltransferase complex confer this specificity. Other identified subunits of the methyltransferase complex are WTAP (28, 29), RBM15/15B (30, 31), KIAA1429/VIRMA (28, 31, 32), HAKAI (31, 33), and ZC3H13 (25, 34). WTAP has no methyltransferase activity, yet knockdown of WTAP dramatically decreases the level of m⁶A in the cell. WTAP acts as a scaffold protein that recruits METTL3-METTL14 to nuclear speckles (28, 29). RBM15/15B is associated with the methyltransferase complex and required for methylation of the long noncoding RNA (lncRNA) Xist in humans. RBM15/15B regulates neuronal function and sex determination in Drosophila (30, 34). KIAA1429/VIRMA is important for methylation at the 3′UTR and is associated with the polyadenylation factors CPSF5 and CPSF6, hinting at a possible role of m⁶A in the selection of polyadenylation sites (32).

Knockdown of ZC3H13 in mouse embryonic stem cells and a Drosophila cell line resulted in reduced levels of m⁶A in the 3′UTR and increased intron retention (25, 34). Knuckles et al. demonstrated that ZC3H13 is required for the interaction between RBM15 and the rest of the methyltransferase complex, through WTAP, and that ZC3H13 regulates Drosophila sex determination by affecting alternative splicing of Sxl (34). Wen et al. showed that depletion of ZC3H13 causes other subunits of the methyltransferase complex, such as METTL3, METTL14, VIRMA, HAKAI, and WTAP, to relocalize to the cytoplasm and that ZC3H13 is important in regulating mESC self-renewal (25).

An additional methyltransferase, METTL16, methylates the 3′UTR of mRNA and U6 snRNA and regulates SAM levels in the cell by affecting splicing and intron retention of SAM synthetase (35, 36). However, METTL16 is unlikely to be associated with the METTL3-METTL14 complex, as it mediates methylation at UACm⁶AGAGAA motifs and depends on a hairpin structure for its activity (35, 36). Our knowledge of m⁶A biology prior to this study was limited to RRACH motifs, which make up 20 to 60% of all motifs found within m⁶A peaks (10). Hence, further elucidation of the topology and roles of these non-RRACH m⁶A sites in the cell is warranted.

m⁶A demethylases.

There are currently two known demethylases in mammals that can reverse the m⁶A mark (FTO and ALKBH5), and they belong to the family of Fe²⁺/alpha-ketoglutarate-dependent dioxygenases. FTO was recently shown to demethylate N⁶,2′-O-dimethyladenosine (m⁶Am), suggesting that FTO may remove both m⁶A and m⁶Am (37, 38). Both FTO and ALKBH5 are localized to nuclear speckles (13, 14), though FTO can shuttle between the cytoplasm and the nucleus in rat neurons under starvation conditions (39) and is mediated by XPO2 (40). FTO regulates splicing by demethylating intronic m⁶A sites (41). The demethylation activity of ALKBH5 is critical for nuclear export of mRNA and for splicing in spermatocytes (14, 42). Studies on the dynamic nature of m⁶A have been limited to nuclear functions, such as alternative splicing and nuclear export, which agrees with the predominantly nuclear localization of FTO and ALKBH5 in most cell models. It remains to be seen how the cell removes m⁶A/m⁶Am once methylated mRNA has left the nucleus.

m⁶A-binding proteins.

YTHDF1, YTHDF2, YTHDF3, YTHDC1, and YTHDC2 are part of the YTH domain family of m⁶A-binding proteins (43). YTHDF1, YTHDF2, and YTHDF3 are most conserved and are predominantly cytoplasmic (43). YTHDF1 recruits translation initiation factors to mRNA to facilitate translation (44), YTHDF2 promotes deadenylation of methylated transcripts by the CCR4-NOT complex (45, 46), and YTHDF3 promotes either translation or degradation of transcripts by interacting with YTHDF1 or YTHDF2, respectively (47, 48). Despite differing functions, all three proteins bind to ∼40% of the total YTHDF target mRNAs (48). Interestingly, all three proteins have similar affinities for m⁶A on RNA (1 to 2 μM), and hence competition for m⁶A sites among the proteins may depend on other RNA-binding factors or the secondary structure of the RNA (49–52). How YTHDF proteins choose to bind preferentially to certain transcripts is unclear and warrants a systematic study. YTHDC1 is localized to the nucleus and plays a role in splicing, alternative polyadenylation, and nuclear export (53–55). The splice factor SRSF3 together with YTHDC1 promotes exon inclusion, whereas SRSF10 competes with SRSF3/YTHDC1 to promote exon skipping (55). YTHDC1-facilitated nuclear export occurs through two subunits: SRSF3 and NXF1 (53). YTHDC2 is a nucleocytoplasmic protein that favors translation and degradation of mRNA by associating with the small ribosomal subunit and XRN1, respectively (56–58). How YTHDC2 is regulated to have such contradictory functions is unknown. On the other hand, YTHDF1 and YTHDF2 promote translation and degradation by binding to different complexes (58). It is likely that these different proteins regulate translation and degradation at different phases of the mRNA life cycle. For example, YTHDF1 might recruit translation initiation factors to mRNA preceding the binding of YTHDC2 to mRNA in the ribosome.

Additional m⁶A-binding proteins have been discovered. eIF3 promotes cap-independent translation by binding to 5′UTR m⁶A sites (59), while IGF2BP1/2/3 proteins promote translation and stability of transcripts by binding to m⁶A sites (60). FMR1 also binds to m⁶A sites but has an undefined function (61, 62).

m⁶A IN INFECTION OF DNA VIRUSES

m⁶A enhances SV40 replication.

By applying photo-cross-linking-assisted m⁶A sequencing (PA-m⁶A-seq), Tsai and colleagues found about 11 and 2 potential m⁶A sites in the late and early regions of SV40, respectively (63). They introduced silent mutations at some of these m⁶A sites, which caused decreases in viral gene expression and virion production. Both METTL3 and YTHDF2, as well as YTHDF3 (to a lesser extent), were shown to promote lytic replication. Inhibition of these enzymes by 3-deazaadenosine (3DAA), as shown by Tsai et al., may have a therapeutic effect for treating SV40 infection and polyomas. The significance of this finding can be extended to human polyomaviruses, including Merkel cell polyomavirus, BK virus, and JC virus (64). Understanding the effect of SV40 infection on the host cell epitranscriptome may shed light on new strategies for treating polyomas.

Role of m⁶A in KSHV life cycle.

Kaposi's sarcoma-associated herpesvirus (KSHV) is an oncogenic virus etiologically associated with Kaposi's sarcoma (KS), primary effusion lymphoma (PEL), multicentric Castleman's disease (MCD), and KSHV-induced cytokine syndrome (KICS) (65–68). KSHV infects diverse types of cells and establishes latency, but KSHV infection of mesenchymal stem cells transforms them into tumor cells with KS-like features (65, 69–74). During latency, only a few viral genes are expressed, chiefly LANA, vFLIP, and vCyclin, as well as a dozen microRNAs (miRNAs), all of which are essential for tumorigenesis (71, 75–81). In tumors, a small proportion of cells undergo constant lytic replication, with most viral genes being expressed, culminating in the release of virions. These spontaneously lytic cells contribute to a protumor microenvironment by secreting a viral homolog of human interleukin-6 (IL-6), vIL6, as well as inducing proinflammatory cytokines, such as oncostatin M, vascular endothelial growth factor (VEGF), and IL-6 (82–85). Like other herpesviruses, KSHV encodes lytic transcripts expressed in a temporal manner, characterized as immediate early, early, and late transcripts (86–88). Our knowledge of how the expression of these transcripts is regulated is limited, and we sought to determine if m⁶A is involved in KSHV lytic replication. We investigated the viral and cellular m⁶A and m⁶Am epitranscriptomes during the latent and lytic phases of the KSHV life cycle by using m⁶A-seq and by modulating the cellular m⁶A machinery.

The KSHV latent epitranscriptome is conserved among different cell types during latency.

We examined the following five latently infected KSHV cell lines: BCBL1-R cells (PEL cells), KiSLK cells (renal cell carcinoma cells), KTIME cells (telomerase-immortalized microvascular endothelial cells), KMM cells (rat metanephric mesenchymal precursor cells), and KMSC cells (adipose tissue-derived mesenchymal stem cells). The topology of the latent KSHV epitranscriptome was remarkably conserved across all five types of cells. The latent locus (ORF71/72/73) was the most consistently methylated locus in these latently infected cells, indicating a potentially important role of m⁶A in viral latency (89).

The cellular epitranscriptome is remodeled during latent infection.

Utilizing four pairs of latently infected and uninfected cells (KiSLK versus iSLK, KMM versus MM, KTIME versus TIME, and KMSC versus MSC), we investigated the effects of KSHV latency on the cellular epitranscriptome by m⁶A-seq. Overall, the number of methylated transcripts and m⁶A motifs did not change significantly after KSHV infection. However, we observed striking 5′UTR hypomethylation and 3′UTR hypermethylation of m⁶A following viral infection for three of the four pairs of cells (KiSLK versus iSLK, KMM versus MM, and KMSC versus MSC), all of which were transformed cells, while the only pair of cells (KTIME versus TIME) that did not have 5′UTR hypomethylation was untransformed (Fig. 1A). Since KMM and KMSC cells are primary cells transformed by KSHV, we chose these two cell types for further analysis of cellular pathways affected by m⁶A. Interestingly, the most enriched pathways are those that regulate KSHV latency and cellular transformation, such as cytoskeleton and extracellular matrix signaling, cell-cell junction signaling, Rho signaling, ILK signaling, Ephrin signaling, endocytosis, and protein ubiquitination pathways. These results indicate that the m⁶A epitranscriptome is dynamic and that KSHV might alter the distribution of m⁶A on cellular transcripts to promote viral latency and tumorigenesis. Further studies are needed to identify viral factors involved in altering the distribution of cellular m⁶A and the effect of 5′UTR hypomethylation/3′UTR hypermethylation on these transcripts.

FIG 1.

(A) KSHV reprograms the cellular epitranscriptome during latency, causing 5′UTR hypomethylation and 3′UTR hypermethylation (89). (B) Regulation of SV40 and KSHV replication by m⁶A. The influence of m⁶A on key events during SV40 and KSHV viral replication is highlighted. In the nucleus, the methyltransferases METTL3 and METTL14 favor the accumulation of SV40 and KSHV transcripts (90, 92). YTHDC1 enhances splicing of KSHV RTA transcripts (92). In the cytoplasm, YTHDF2 mediates degradation of KSHV transcripts, leading to repression of viral replication (89). YTHDF2 and YTHDF3 promote viral replication in KSHV (90) and SV40 (63), through a noncanonical role of these proteins.

Role of m⁶A during KSHV lytic replication.

When we profiled the viral m⁶A epitranscriptomes of renal carcinoma cells (KiSLK) and PEL cells (BCBL1-R), we found that almost every viral transcript (94 viral genes) contained one or more m⁶A sites. Interestingly, the KSHV epitranscriptomes from both the KiSLK and PEL cell lines were at least 70% conserved, indicating an important role for m⁶A in viral lytic replication, regardless of cell type. We investigated the function of m⁶A during viral lytic replication by knocking down YTHDF2 and found elevated levels of viral transcripts, proteins, and release of virions. Furthermore, we observed an increase in the half-life of viral transcripts after YTHDF2 knockdown, which is consistent with the reported role of YTHDF2 in promoting mRNA degradation. Specifically, YTHDF2 binds to and degrades viral transcripts and negatively impacts KSHV replication (Fig. 1B). Using RNA-binding protein immunoprecipitation and reverse transcriptase quantitative PCR (RIP-RT-qPCR), we also showed that YTHDF2 selectively binds to certain viral transcripts at 48 h post-lytic induction. Hence, YTHDF2 might be a novel restriction factor for KSHV lytic replication. A conflicting role for YTHDF2 was observed in a different study, as knockdown resulted in decreased levels of viral transcripts, proteins, and release of virions (Fig. 1B) (90). The reason for this disparity is unknown, even though both studies used similar cell types, methods of gene silencing, and experimental time points for viral induction. However, the proviral role of YTHDF2 observed in that study contradicts the observation that YTHDF2 mediates viral mRNA degradation. Although it is unusual, this noncanonical role was also observed in influenza virus and SV40 replication (63, 91). The mechanism behind this noncanonical role of YTHDF2 remains unknown. Since KSHV has a complex temporal lytic replication cycle, YTHDF2 and other reader proteins may prefer distinct sets of viral transcripts at different time points. Both studies probed YTHDF2 knockdown at a single time point in KSHV lytic reactivation. This problem is further exacerbated by the nonsynchronous nature of KSHV lytic replication. Until we can develop technologies such as rapid conditional knockdown of m⁶A-binding proteins or specific inhibitors, the effects of m⁶A on different classes of KSHV transcripts at different time points might remain elusive.

Another study, by Ye et al., showed that inhibition of methylation by use of 3DAA impaired lytic replication, whereas silencing of FTO slightly increased lytic replication (92). They showed that YTHDC1 promotes splicing of KSHV RTA, which is essential and sufficient for inducing lytic replication. RTA itself was shown to increase levels of m⁶A in the cell, which agrees with the abundance of methylated viral transcripts during lytic replication (Fig. 1B) (89, 90, 92). However, it is worth mentioning that 3DAA has unspecific activities, while FTO also acts on m⁶Am, both of which may confound the observed effects.

The cellular epitranscriptome is modified during lytic replication.

Lytic replication affected the cellular m⁶A epitranscriptome in KiSLK and BCBL1-R cells in distinct ways. 5′UTR hypermethylation and 3′UTR hypomethylation were observed in KiSLK cells, whereas the opposites were seen in BCBL1-R cells. We noticed a higher level of KSHV host shutoff and endonuclease (SOX) activity, which degrades host mRNA, in BCBL1-R cells than in KiSLK cells, as reflected by the different percentages of mapped cellular reads. Pathway analysis of hypermethylated and hypomethylated genes from both cell lines identified those associated with KSHV replication, such as protein kinase A signaling, ILK signaling, extracellular signal-regulated kinase (ERK)/mitogen-activated protein kinase (MAPK) signaling, phosphatidylinositol 3-kinase (PI3K)/AKT signaling, adipogenesis, and integrin signaling (93).

CONCLUDING REMARKS

Modern studies on m⁶A of DNA viruses provide evidence that m⁶A can either promote or attenuate viral (SV40 and KSHV) replication. Given that these viruses replicate in the nucleus, the cellular m⁶A machinery is expected to regulate viral transcripts in a fashion similar to that for cellular transcripts. Smaller viruses, such as SV40, have fewer m⁶A sites, which are likely critical for viral replication. Attenuating m⁶A on the transcripts of these viruses by inhibiting the methyltransferase complex or activating the demethylases FTO and ALKBH5 might decrease viral replication. In contrast, as a large virus, KSHV might utilize m⁶A in a transcript-specific and temporal manner, making it difficult to broadly place m⁶A as a proviral or antiviral factor. The multitude of m⁶A-binding proteins and their contrasting functions, such as in the case of YTHDF1 and YTHDF2, complicate this matter.

Our knowledge of the epitranscriptomics of DNA viruses is in its infancy, and only a few observations on the role of m⁶A in KSHV and SV40 replication have been made so far. Other posttranscriptional mRNA modifications, such as m¹A, pseudouridine, and m⁵C, have been studied less in viral contexts and thus require further investigation (94–96). For KSHV, it is difficult to pinpoint if m⁶A has a clear proviral or antiviral role. Hence, the use of inhibitors targeting the cellular m⁶A machinery for treatment of viral replication might be challenging. In contrast, for small DNA viruses, such as SV40 and its closely related relatives, such as JC virus, BK virus, and Merkel cell polyomavirus, 3DAA or other methyltransferase inhibitors may be examined for antiviral activity. For KSHV-associated tumors, identification of viral latent proteins or miRNAs that mediate the reprogramming of the cellular epitranscriptome is of great interest. Defining the mechanism by which m⁶A mediates the dysregulated cellular pathways involved in tumorigenesis might lead to the identification of potential novel therapeutic targets.