A molecular‐level perspective on the frequency, distribution, and consequences of messenger RNA modifications

Chemical modifications have the potential to individually influence mRNA structure and dynamics, splicing and maturation, RNA–protein interactions, translation, and stability. The interconnected nature of the mRNA life cycle can intensify the effect of a modification through the modulation of downstream processes. For example, several mRNA modifications, m⁶A, m¹A, m¹G, Ψ, and f⁵C, have been shown to change the stability of RNA structures and would be predicted to redistribute the ensemble of mRNA secondary structures present in a cell (Charette & Gray, ; D. R. Davis, ; B. Liu et al., ; Roost et al., ; Spitale et al., ; R. Wang et al., ; Zhou et al., ). This alteration can modulate the ability to form RNA–protein interactions, which can in turn impact mRNA maturation, translation, and decay through pathways dependent on these interactions. Additionally, mRNA translation rates and mRNA decay rates are coupled, with poorly translated mRNAs being targeted more robustly for decay (Presnyak et al., ; Radhakrishnan et al., ). Thus, if an mRNA modification strongly impacts one step in an mRNA's life, this perturbation is likely to be observed in the outcome of related processes (e.g., modification induced perturbations in mRNA structure could slow translation, which in turn reduces the mRNA's half‐life)

Impact of messenger RNA (mRNA) modifications on mRNA–protein interactions and protein synthesis. (a) mRNA‐binding protein affinities are modestly altered by nucleoside modifications. The ratio of binding affinities (K_D) for proteins binding to m⁶A (gray bars) and Ψ (black bars) unmodified and modified mRNA transcripts binding to a Pumilio proteins (PUM2), and the m⁶A binding proteins proline‐rich coiled‐coil 2A (PRRC2A), YTHDF3, YTHDF2, YTHDC1, and insulin‐like factor 2 mRNA binding proteins 1, 2, and 3 (IGF2BP1, IGF2BP2, IGF2BP3) (H. Huang et al., ; Vaidyanathan et al., ; X. Wang et al., ; R. Wu et al., ). The affinity of PUM2 was measured for model mRNAs containing 1 (PUM2_1) or 3 (PUM2_3) modifications. (b) Reporter proteins were expressed from mRNAs containing a single nucleotide modification in commercially available fully reconstituted bacterial translation assays. The plot displays the amount of protein produced from the modified mRNA relative to the amount of protein produced from an unmodified transcript as a function of codon. The values in this table were from studies by Hoernes, Clementi et al. () and You et al. (). The magnitude of each modification's effect is depends not only on the identity of the modification, but also on the codon in which it is located, the position of the modification within that codon

LC–MS/MS measurements of mRNA modification abundance. All values displayed are the average values for mammals unless indicated otherwise (* yeast, ** plant) (Table ). The error bars reflect the range of values measured. Modifications without error bars have only one reported value in the literature

Methods to quantify messenger RNA (mRNA) m⁶A modification stoichiometry. (a) Site‐specific cleavage and radioactive‐labeling followed by ligation‐assisted extraction and thin‐layer chromatography (SCARLET) is the most direct method developed to quantify m⁶A frequency (N. Liu et al., ). During this process, a chimera is annealed to a specific mRNA where the DNA sequence is immediately upstream of the putative modification site. RNase H is used to cleave the mRNA to release an oligonucleotide containing the putative modification at the 5′ end. The 5′end of the oligonucleotide is ³²P‐labeled using T4 polynucleotide kinase and is splint ligated to a 116mer DNA oligomer. RNase T1/A is used to digest the resulting chimera to contain a single A or m⁶A at the 3′end of the 116mer DNA oligomer. The resulting oligonucleotide is gel purified, digested to nucleosides using nuclease P1, and analyzed using TLC. The stoichiometry is measure based on the relative intensity of the m⁶A and A phosphorescence. (b) m⁶A‐level and isoform‐characterization sequencing (m⁶A‐LAIC‐seq; Molinie et al., ) utilizes m⁶A modified external RNA controls consortium (ERCC) control RNAs to normalize the measured m⁶A stoichiometries to increase the accuracy of a standard m⁶A‐RIP assay. ERCC controls are added before and after m⁶A‐RIP to normalize the efficiency of the immunoprecipitation and detection by next generation sequencing, respectively. The relative counts of m⁶A positive and negative reads of the same transcript determine the stoichiometry. The occupancy levels measured by m⁶A‐LAIC‐seq correlate well with modification levels of mRNA standards (R = 0.995) despite the lack of single nucleotide resolution. (c) MAZTER‐seq also utilizes RNA‐seq to characterize m⁶A occupancy transcriptome wide (Garcia‐Campos et al., ). However, MAZTER‐seq does not rely on immunoprecipitation to isolate modified mRNAs, and instead identifies sites using the bacterial nuclease MazF to cleave immediately upstream of ACA sequences but not m⁶ACA sequences. Purified mRNA is digested using MazF, and an adapter is ligated to the 3′end of the digested products. The resulting oligonucleotides are reverse transcribed, 3′ adapter ligated, amplified, and end‐pair sequenced. Following read alignment and data processing, the stoichiometry is determined by calculating the cleavage efficiency of MazF at a specific ACA motif. While MAZTER‐seq has the advantage of quantifying m⁶A occupancy at single nucleotide resolution, there are some limitations to this approach. Namely, the lack of quantification at the ~50% of m⁶A found outside ACA motifs (Pandey & Pillai, ), MazF only cuts at ACA sites with 70% efficiency and can cleave at other sequences resembling ACA, and the values measured by MAZTER‐seq only modestly correlate with SCARLET measurements at similar sites (R values = 0.6–0.7; Garcia‐Campos et al., )

The implementation of quantitative approaches will allow us to critically assess some of the key questions in the epitranscriptome field and establish a molecular understanding of individual messenger RNA (mRNA) modifications. Here, we present several knowledge gaps that we think can be best filled using quantitative approaches: (a) Several mRNA modifications, m⁶A, m¹A, m¹G, Ψ, and f⁵C, affect the stability of RNA secondary structure, but limited knowledge is known about the effect of other chemical modifications (Charette & Gray, ; D. R. Davis, ; B. Liu et al., ; Roost et al., ; Spitale et al., ; R. Wang et al., ; Zhou et al., ). High‐resolution structural biology and secondary structure‐probing techniques are needed to uncover modification mediated structural changes. (b) Current transcriptome wide sequencing approaches have uncovered thousands of modification sites, but little is known about how modification insertion sites are selected. Modifications could be randomly incorporated on available sites, incorporated on specific locations of target transcripts, or there could be cross talk between sites on a single transcript (cooperative incorporation). Kinetic and thermodynamic investigations of modifying enzyme selectivity and broad analyses of the contributions of structure to selectivity (as in Carlile et al., ), coupled with measurements of the stoichiometry of multiple modifications on individual transcripts can help to distinguish between these models. (c) Targeted approaches will be required to discern which mRNA modification sites are biologically relevant. Measurements of modification stoichiometry, and assessment of how the stoichiometry at individual sites varies as a function of cell cycle, environment and disease is one example of experiments that could be done to identify significant sites of modification. (d) Occupancy of individual sites might be temporally controlled, and therefore the stoichiometry of individual sites need to be quantitatively assessed as a function of time. Without this information it is likely that biologically relevant sites may be overlooked. (e) It is difficult to deconvolute the impact of mRNA modifications on mRNA–protein, splicing, mRNA stability, and mRNA translation on protein output in cells (see Figure ). Reconstituted systems are ideally suited to overcome this challenge by allowing researchers to dissect how each individual interaction is influenced by mRNA modifications. These sorts of studies can help to establish which biological processes are likely more impacted by particular modifications, and have the potential to suggest likely consequences of mRNA modifications

Reported messenger RNA (mRNA) modifications. Unmodified nucleosides are shown in boxes, while the modified nucleosides are unboxed. The full names of the nucleosides are: A, adenosine; m⁶A, N⁶‐methyladenosine; m¹A, 1‐methyladenosine; Am, 2′‐O‐methyladenosine; I, inosine; m⁶Am, N⁶, 2′‐O‐dimethyladenosine; C, cytidine; Cm, 2′‐O‐methylcytidine; m⁵C, 5‐methylcytidine; m³C, 3‐methylcytidine; f⁵C, 5‐formylcytidine; ac⁴C, N⁴‐acetylytidine; hm⁵C, 5‐hydroxymethylcytidine; G, guanosine; m⁷G, 7‐methylguanosine; Gm, 2′‐O‐methylguanosine; m¹G, 1‐methylguanosine; U, uridine; Ψ, pseudouridine; Um, 2′‐O‐methyluridine