N⁶-methyl-Adenosine modification in messenger and long non-coding RNA

Abstract

N⁶-methyl-Adenosine (m⁶A) is the most abundant modification in mammalian mRNA and long noncoding RNA. First discovered in 1970s, m⁶A modification has been proposed to function in mRNA splicing, export, stability and immune tolerance. Interest and excitement in m⁶A modification has recently been revived based on the discovery of a mammalian enzyme that removes m⁶A and the application of deep sequencing to localize modification sites. The m⁶A demethylase FTO controls cellular energy homeostasis and is the first enzyme discovered that reverses an RNA modification. m⁶A sequencing demonstrates cell type and cell state-dependent m⁶A patterns, indicating that m⁶A modifications are highly regulated. This review describes the current knowledge of mammalian m⁶A modifications and future perspectives on how to push the field forward.

Modifications in messenger and long non-coding RNA

Over one hundred types of RNA modifications have been identified in all three kingdoms of life (http://rna-mdb.cas.albany.edu/RNAmods/). The most chemically diverse modifications are present in ribosomal RNA (rRNA) and transfer RNA (tRNA); rRNA and tRNA modifications are commonly proposed to find-tune their structure and function ¹. Most modification studies have focused on rRNA and tRNA due to their high cellular abundance, large chemical diversity and functional importance in translation.

Modifications are also present in messenger RNA (mRNA) and long non-coding RNA (lncRNA) in eukaryotes (Figure 1). Additional chemical changes of specific RNA residues include adenosine to inosine (A-to-I) and cytosine to uridine (C-to-U), commonly referred to as “RNA editing” (Figure 1). The 5′ cap of mRNA, some lncRNA, and spliceosomal RNA (snRNA) is the best known example of such modifications. The 5′cap of mRNA is created by both 7-methyl-G addition to the 5′ triphosphate as well as 2′OMe modification of the first and second mRNA residues. The 5′ cap in mRNA recruits the cap-binding protein eIF4E, other initiation factors and the 40S ribosomal subunit to initiate mRNA scanning and translation at the first codon of the coding sequence. The 5′ cap is also required for mRNA stability; its removal is often a necessary step for complete mRNA degradation ^{2, 3}. Up to three types of mRNA/lncRNA modifications are known to be present away from the 5′ cap. Two types, N⁶-methyl-adenosine (m⁶A) and 5-methyl-cytosine (m⁵C), have been shown to occur in significant amounts ^4–7. One other type, pseudouridine (Ψ) has not been shown conclusively to be present in mRNA/lncRNA, although Ψ-modification of codons can have a profound effect on decoding ^{8, 9}. A-to-I and C-to-U editing changes the base pairing property of the edited bases, therefore resulting in recoding of the edited codons. Biological functions of the m⁶A and m⁵C modifications in mRNA/lncRNA, however, are much less well understood. This review discusses the current knowledge and future perspectives of mRNA/lncRNA modification, focusing on m⁶A modification in mammals.

Fig. 1. mRNA and lncRNA modifications.

The functional groups introduced through modifications are shown in red. (A) The three abundant modifications are m⁶A, 5′ cap and m⁵C. The 5′ cap contains both modifications at the 5′ triphosphate and 2′O-methylation of the first and sometimes the second nucleotide. The estimated number of modifications per type, per RNA molecule, is shown in parentheses ^{4, 6, 10}. Some known human enzymes that catalyze the forward and reverse reactions are also shown ^{7, 10, 11, 40}. MTA70: METTL3 methyltransferase like 3; FTO: fat mass and obesity associated; AlkBH5: alkB alkylation repair homolog 5; NSUN2: NOP2/Sun RNA methyltransferase family, member 2. (B) Pseudouridine modification (Ψ) may also be present, although the extent of Ψ is currently unknown. (C) RNA editing commonly refers to deamination of A to inosine and C to uridine at specific sites in mRNA.

N⁶-methyl-A, discovered in the 1970s ⁵, is the most abundant modification in eukaryotic mRNA/lncRNA. N⁶-methyl-A modifications in polyadenylated RNA have been found in mammals, plants and yeast ¹⁰. In mammals, m⁶A occurs on average in 3–5 sites per mRNA molecule, and up to 15 sites in viral RNA. Since a mammalian cell contains over 10,000 mRNA species, tens of thousands of m⁶A sites may exist in the mRNA population in each cell. Two recent developments have revived the interest and excitement in understanding the biological function of m⁶A modification. The first is the discovery in 2011 that m⁶A modification can be reversed by a mammalian enzyme involved in diabetes and obesity ¹¹. The second is the transcriptome-wide mapping of tens of thousands of m⁶A-containing segments in mRNA and long non-coding RNA in 2012 ^{12, 13}. These developments have elevated m⁶A modification to a new level in understanding RNA modification in biology.

m⁶A studies before the advent of deep sequencing

The classical phase of m⁶A modification studies in mRNA and viral RNA has been comprehensively described in a 2005 review ¹⁰. In contrast to abundant rRNA and tRNA, the low abundance of mRNA presents a huge challenge in the identification of m⁶A sites in mRNA. By 2005, specific sites for m⁶A modification were determined in only two individual mRNAs. The m⁶A site in bovine prolactin mRNA is present in the 3′ UTR and near a consensus 3′ polyadenylation sequence. Interestingly, the extent of modification appeared to be only ~20%, suggesting that m⁶A modifications in mRNA may be incomplete at other sites as well. Over ten m⁶A sites have been found in the Rous sarcoma virus (RSV) RNA. Again, modification fractions of the RSV sites are incomplete, ranging between 20–90%. Through mutational analysis of m⁶A sites in vivo and in vitro, the consensus sequence of m⁶A modification has been defined as RRm⁶ACH, where R is A/G and H is A/C/U ^14–17. This consensus sequence can occur once every 85 nucleotides, so that an average mRNA can have more than 30 consensus sites. Since only ~3–5 m⁶A sites were estimated to be present in an mRNA, the majority of the consensus sequences are either not modified, or many of them are modified but at substoichiometic amounts.

A major advance in m⁶A study was the identification of a human m⁶A methyl-transferase subunit of a nuclear-localized complex ^{18, 19}. This methyltransferase, MTA70 (or METTL3), is a homolog of yeast IME4, which is known to introduce m⁶A modification in yeast mRNA during sporulation ^{20, 21}. MTA70 is a subunit of a nearly one mega-Dalton complex. RNA interference of this methyltransferase in HeLa cells resulted in cell death through apoptosis, indicating that m⁶A modifications perform crucial regulatory functions. The MTA-homologs in plant and in Drosophila are essential for viability and are particularly important in certain developmental pathways ^{22, 23}.

Proposed functions that m⁶A modification affects include mRNA splicing, nuclear export, stability and translational efficiency ¹⁰. However, it has been extremely difficult to test m⁶A function for an individual RNA. Unexpectedly, mutating the modified adenosine or the 3′ cytidine residue in the consensus sequence often resulted in the appearance of new m⁶A modification at a nearby consensus site in the same RNA ¹⁰. Further, because many sites are potentially modified in each mRNA, one may not expect a very large effect upon mutating a single modification site.

An exciting development in the studies of m⁶A modification function was the discovery that it can prevent recognition of the modified RNA by cellular innate immunity ²⁴. The innate immune system recognizes self versus foreign molecules and triggers a cascade of cellular activities upon activation. Known components in the innate immune system that directly recognize RNA include toll-like receptor 3 (TLR3), which recognizes double-stranded RNA, and TLR7, which recognizes single-stranded RNA ²⁵. The innate immune system is activated upon transfection of unmodified RNA oligonucleotides, but remains dormant upon transfection of the same RNA containing m⁶A modification. Interestingly, transfection of modified RNA containing m⁵C or Ψ also markedly reduced innate immune activation, although m⁶A modification was most effective in preventing activation. By contrast, compared to unmodified mRNA transcripts, complete substitution of A to m⁶A was highly detrimental to translation, whereas substitution of C/U to m⁵C/Ψ increased the level of translation ²⁶. However, substituting ~5% of A to m⁶A did not affect translation levels; this low level of A-to- m⁶A substitution more closely resembles the m⁶A levels observed in natural mRNAs ²⁶. These results suggest that one function of m⁶A (and potentially m⁵C or Ψ) modification in mRNA is to tag cellular mRNA to be “self” when presented to the innate immune system ^{24, 26}. This insight led to the remarkable success in efficiently generating induced pluripotent stem cells (iPS) simply by transfecting m⁵C/Ψmodified mRNAs of the four key factors in iPS formation ²⁷.

The discovery of two m⁶A demethylases

A major breakthrough was the discovery in 2011 of a mammalian enzyme that specifically removes m⁶A modification in polyadenylated RNA, which are primarily composed of mRNA and lncRNA ¹¹. The fat mass and obesity associated protein (FTO) belongs to a human family of AlkB-homologues that can remove N-methylated RNA or DNA bases. Escherichia coli AlkB is a repair enzyme capable of demethylating N¹-modified A or N³-modified C both in DNA and RNA ^{28, 29}. Many AlkB-homologues or closely related proteins have been found in mammalian genomes including AlkBH1–8, FTO and Tet1–3. AlkBH8 was found to be a tRNA modification enzyme containing two catalytic domains: one catalyzes methylation of several wobble uridine modifications ^{30, 31} and the other catalyzes hydroxylation of the hypermodified U34 wobble base of tRNA^Gly(UCC) ^{32, 33}. The Tet enzymes were found to catalyze the oxidation of m⁵C in DNA to 5-hydroxymethyl-C, which is a new form of DNA modification and may represent the first step in the reversal of m⁵C modification in chromosomal DNA ^{34, 35}. The FTO enzyme is one of the key factors in regulating mammalian energy homeostasis; the FTO gene was identified in genome-wide association studies to be strongly linked to diabetes and obesity ³⁶, and loss of FTO function in mice leads to a significant reduction in body mass ³⁷. On a mechanistic level, FTO was found to selectively remove m⁶A in polyadenylated RNAs both in vitro and in vivo. In vivo, siRNA knockdown or overexpression of FTO increased or decreased total m⁶A content by ~15–20%, respectively. The FTO protein partially co-localizes with markers of nuclear speckles, which are sites of mRNA splicing. This localization of FTO activity is consistent with previous suggestions that m⁶A is involved in mRNA splicing. These results indicate that m⁶A modifications in mRNA/lncRNA play a significant role in cellular energy homeostasis and metabolism through their influences on the spliced mRNA/lncRNA populations.

FTO is the first enzyme discovered that can reverse an endogenous RNA modification. Base methylation is a dominant form of RNA modification in all cellular RNA types. Aside from m⁶A, which occurs in polyadenylated RNA as well as in rRNA/tRNA and snRNA, other known N-methylations in rRNA/tRNA include N¹-methyl-A (m¹A), N³-methyl-C (m³C), N⁴-methyl-C (m ⁴C), N¹-methyl-G (m¹G), N²-methyl-G (m²G), N³-methyl-U (m³U), and N³-methyl-pseudoU (m³Ψ), all of which can potentially be subjected to reversal by a similar or other mechanisms. The discovery of m⁶A as a cellular substrate of FTO indicates that other cellular enzymes may exist to catalyze the reversal of other base methylations ^{38, 39}.

A second human m⁶A-demethylase has just been described ⁴⁰. Like FTO, AlkBH5 is another member of the AlkB-homologous gene family of dioxygenases. Perturbation of AlkBH5 significantly affects nuclear mRNA export and RNA metabolism. Furthermore, AlkBH5 appears to be involved in the assembly of mRNA processing factors in nuclear speckles. The physiological effect of AlkBH5 deficient mice includes impaired male fertility. The discovery of a second m⁶A-demethylase strongly indicates that reversible m⁶A modification likely has broad functions in mammalian biology.

Both mRNA m⁶A methyltransferase (MTA70) and mRNA m⁶A demethylase (FTO and AlkBH5) are catalytic subunits of much larger complexes in the cell that presumably confer specificity of m⁶A addition and removal. These accessory factors need to be identified using modern proteomic techniques to further advance studies of m⁶A function.

Transcriptome-wide profiling of m⁶A in mRNA and lncRNA

A new wave of breakthroughs in 2012 took advantage of recent advances in high throughput sequencing ^{12, 13}. Reverse transcriptase does not discriminate m⁶A from A in bulk cDNA synthesis, therefore, m⁶A cannot be currently detected by standard high throughput sequencing methods. Using m⁶A-specific antibodies originally generated for m⁶A modification studies of spliceosomal RNA, two independent studies developed similar methods that enabled the identification of ten thousands of m⁶A-containing mRNA/lncRNA segments in different tissues and stress conditions (Figure 2). In these studies, polyadenylated RNAs are first fragmented to ~50–100 nucleotides then pulled down with an m⁶A-antibody, followed by sequencing of the immunoprecipitated RNA fragments. Analysis of the high throughput sequences reveals ~200-residue-wide peaks; many, although not all contain the known m⁶A consensus sequences. These m⁶A profiling studies showed that m⁶A modifications are enriched around the stop codon of the coding regions of mRNA, although the function of this enrichment is currently unknown. Further, m⁶A modifications have distinct distributions among tissue types and stress conditions, thus indicating that each cell type and cell state likely have unique m⁶A modification patterns. Another very exciting finding from these studies was the discovery of m⁶A modification in many long non-coding RNAs, indicating that m⁶A modifications are not confined to mRNA. The transcriptome-wide m⁶A profiles are similar to the profiles for m⁵C modifications ⁷. Both modifications are non-randomly distributed in mRNA with enrichment in and near the 3′UTR, and both are also found in non-coding RNAs.

Fig. 2. m⁶A sequencing method and results ^{12, 13}.

Starting from total polyadenylated RNA, step 1 is to chemically fragment RNA to ~50–100 nucleotides. This sample is split in two; one is used as input control, and the other immunoprecipitated with anti-m⁶A antibody to isolate RNA fragments containing m⁶A modification. Both samples are deep sequenced (step 2), and the sequencing results compared to identify m⁶A-containing RNA segments (step 3). The m⁶A modification is markedly enriched around the stop codon of mRNA, although the reason for this enrichment is unclear.

The demonstration that m⁶A modification in mRNA/lncRNA has distinct patterns in response to developmental or environmental cues is akin to m⁵C modification in chromosomal DNA where distinct patterns are present depending on cell type and cell state in order to establish epigenetic maintenance and regulation. It is likely that m⁶A modification patterns also follow a well defined but distinct pattern depending on cell type and cell state. The identification of m⁶A patterns in mRNA/lncRNA that correlate to cell type and states could therefore establish the concept of “RNA epigenetics” as another layer of epigenetic regulation ^{38, 39}.

A major challenge will be the development of a high throughput technique that identifies m⁶A modification in mRNA/lncRNA at single nucleotide resolution. The current technique of using an m⁶A antibody generates m⁶A candidate sites within ~200 nucleotide segments, but there is no simple method to actually validate the presence of m⁶A modification there. Furthermore, the known m⁶A consensus sequence cannot explain many sequencing peaks, suggesting that more than one m⁶A modification enzyme is present in cells, and each enzyme may have distinct requirements for m⁶A modification. The development of an m⁶A sequencing technique with single nucleotide resolution may involve chemical reactions that are specific for m⁶A, but not for unmodified A. Chemically reacted m⁶A-containing RNA fragments can then be enriched by affinity pull-down followed by sequencing. Another potential method involves single molecule sequencing which has been shown to be capable of detecting m⁶A modification in DNA ⁴¹. Another highly desirable feature of new m⁶A sequencing technologies would be the determination of the m⁶A modification fraction at each site. All m⁶A modifications in cellular mRNA or viral RNA that are mapped to single nucleotide resolution have 20–90% methylation for each site. Why cells maintain fractional m⁶A modification at these sites is unclear; it is difficult to derive functional hypotheses given the paucity of such experimental data. Fractional modification can be anywhere from 0–100%, so multiple regulatory states may be available depending on the extent of modification at each site.

Several proteins that seem to prefer binding to m⁶A modified RNA have also been identified from cell lysates using a known viral m⁶A-containing RNA oligonucleotide as in vitro bait ¹². Some of these m⁶A binding proteins have been shown previously to bind RNA, although a few had no previously known association with RNA. At this time, it remains to be determined whether these proteins actually bind m⁶A-containing RNA in vivo and whether m⁶A binding of these proteins is functional. In any case, identification of such proteins strongly suggests that an important function of m⁶A modification is to recruit specific proteins to modified mRNA or lncRNA. Thus, m⁶A modification has potential to introduce a new layer of biological regulation through m⁶A binding proteins and beyond.

Concluding remarks

N⁶-methyl-A modification can exert its function in at least two ways: alteration of RNA structure or recruitment of specific m⁶A binding proteins (Figure 3). The addition of a single methyl-group to the N⁶-position of adenosine does not perturb Watson-Crick base pairing. However, m⁶A modification is known to weaken RNA secondary structure ⁴², and can affect RNA tertiary structure, in particular tertiary interactions involving base triples or Hoogsteen pairs that rely on the N⁶ proton as a Hydrogen bond donor. N⁶-methyl-A modification may also prevent RNA folding into alternate structures, as demonstrated for certain base methylations in tRNA ⁴³. That N⁶-methylation might create new binding sites for specific proteins is an attractive and readily identifiable function for specific m⁶A modification sites. Recruitment of m⁶A binding proteins to specific sites in the RNA would be a way to concurrently recruit additional complexes that bind m⁶A, such as the spliceosome, the mRNA export machinery, or chromatin remodeling enzymes.

Fig. 3. Functional effects of m⁶A modification.

(A) m⁶A modification could alter RNA structure either through weakening of base pairing or through the loss of one of the two hydrogen bond donors at the N⁶-position of A. This H-bond donor could be involved in RNA tertiary structure. (B) m⁶A modification could generate a new binding site for proteins that recognize m⁶A modified RNA. The m⁶A binding proteins can recruit additional complexes involved in cellular processes such as mRNA splicing, export, stability and immune tolerance.

In my opinion, the current biggest challenge for the field is to develop methods that can perturb m⁶A modification at specific sites in order to directly assess m⁶A function in specific genes. RNA interference or over-expression of an mRNA may simply decrease or increase both modified and unmodified RNA alike. In a few cases (summarized in ¹⁰), mutation of a known m⁶A site in an mRNA resulted in additional modification at a nearby consensus site, so that one cannot simply assume that mutation of a known site would not lead to the modification of cryptic sites nearby that may perform the same function. Further, functional understanding of a specific site should also take into account that all currently known m⁶A sites in mRNA and viral RNA are incompletely modified, so that one may need to explain why cells simultaneously maintain two RNA species that differ only at the site of m⁶A modification.

In summary, since the discovery of m⁶A modification in mRNA in 1970s, the field has come a long way to establish this modification as a wide spread biological process in many branches of eukaryotes. The m⁶A modifications have been proposed to play significant roles in cellular RNA functions including mRNA splicing, nuclear export, stability, and immune tolerance. The presence of appreciable amounts of m⁶A in lncRNAs suggests that m⁶A can also be important in cellular processes other than protein synthesis. Despite these advances, many technical, conceptual and biological challenges remain, and the functional significance of m⁶A modification in specific RNAs remains to be determined. The field is wide open for exploration.

Abbreviations

m⁶A

N⁶-methyl-adenosine

m⁵C

5-methyl-cytosine

Ψ

pseudouridine