Crystal structure of the YTH domain of YTHDF2 reveals mechanism for recognition of N6-methyladenosine

Dear Editor,

N6-methyladenosine (m⁶A) has been demonstrated to be ubiquitous in several types of eukaryotic RNAs, including messenger RNA (mRNA), transfer RNA (tRNA), ribosomal RNA (rRNA), long non-coding RNA (lncRNA), and small nuclear RNA (snRNA)¹. The recent discoveries of RNA m⁶A methyltransferase complex METTL3/METTL14/WTAP and demethylases FTO and ALKBH5 prove the reversibility of m6A modification^2,3,4,5,6. This modification plays important roles in various biological processes, including circadian rhythms⁷, RNA splicing⁸, yeast meiosis⁹, and embryonic stem cell self-renewal¹⁰. Two recent studies show that YTH domain family 2 (YTHDF2) and other YTHDF proteins preferentially bind to m⁶A-containing mRNA in vivo and in vitro and regulate localization and stability of the bound mRNA^8,11. YTHDF2 is also known to be involved in development of acute myeloid leukemia¹². YTHDC1 (splicing factor YT521-B), another YTH domain-containing protein, is known to play an important role in Emery-Dreifuss muscular dystrophy. While the function of YTHDF2 in the regulation of mRNA stability has been explored, the molecular mechanism for specific recognition of m⁶A by the YTH domain remains elusive.

YTHDF2 consists of a C-terminal YTH domain (designated as YTH^YTHDF2), which specifically binds to m⁶A-containing RNA (m⁶A-RNA) with a preference for those containing a consensus motif of G(m⁶A)C¹¹. The YTH domains are highly conserved in YTH domain-containing proteins, including YTHDF1-3, YTHDC1, YTHDC2 (CsA-associated helicase-like protein), Mmi1 (Schizosaccharomyces pombe), and MRB1 (Saccharomyces cerevisiae), suggesting an important function of the YTH domain across species (Figure 1A).

Figure 1

Crystal structure of YTHDF2 reveals mechanism for recognition of N6-methyladenosine. (A) Schematic representation of YTHDF2 (NP_057342.2) and structure-based sequence alignment of selected YTH domains. Identical and highly conserved residues are highlighted in dark green, with conserved residues in light green and less conserved ones in light blue. Secondary structural elements are indicated above the sequence. Residues that may be involved in recognition of m⁶A-RNA are indicated as blue squares. (B) FP assay of YTH^YTHDF2 binding to A-RNA or m⁶A-RNA. In the two-step binding mode, the binding affinity (Kd) of the first step is indicated. The Kd shown might be lower than the real value as binding saturation could not be achieved during the FP assays. (C) EMSA using A-RNA or m⁶A-RNA with increasing amount of YTH^YTHDF2 proteins as indicated. (D) Ribbon representation of the crystal structure of YTH^YTHDF2. A close-up view of the residues for hydrophobic pocket formation is shown on the right. Critical residues that may be involved in m⁶A recognition are indicated as stick representations. (E) Electrostatic potential surface of YTH^YTHDF2. Positive charge is colored in blue with neutral charge in white and negative charge in red. Critical residues for RNA interaction are indicated. (F-G) Superimposed FP plots of RNA-binding affinities for wild-type and mutant YTH^YTHDF2 toward m⁶A-RNA (F) or A-RNA (G). The experiments were performed as in B. (H) Relative enrichment of m⁶A-RNA in RNA products immunoprecipitated by wild-type and mutant YTHDF2 full-length proteins from HEK293T cells. The levels of RNA products were determined by LC-MS/MS. Error bars represent SD for triplicate experiments.

To verify the specific recognition of m⁶A-RNA by YTH^YTHDF2, we performed fluorescence polarization (FP) assays and electrophoretic mobility shift assays (EMSA) using purified YTH^YTHDF2, unmodified RNA (A-RNA) and m⁶A-RNA. Interestingly, it appears that YTH^YTHDF2 binds to A-RNA in a one-step binding mode (with the goodness-of-fit R² value of 0.9943), but to m⁶A-RNA in a two-step binding mode (with R² of 0.9987, while the R² value is 0.9843 in one-step binding mode.) (Figure 1B, Supplementary information, Figure S1A and Table S1A). Both FP and EMSA experiments show that YTH^YTHDF2 bound to m⁶A-RNA with much higher binding affinity than to A-RNA (Figure 1B and 1C). Note that 30 pmol of YTH^YTHDF2 led to an almost complete shift of m⁶A-RNA (lane 8), while 100 pmol of YTH^YTHDF2 only shifted less than half of the A-RNA (lane 5) (Figure 1C). A similar two-step binding mode was observed in other RNA-binding proteins¹³. It is likely that binding to RNA facilitates the further recognition of m⁶A.

To reveal the molecular mechanism for specific recognition of m⁶A-RNA by YTH^YTHDF2, we solved the crystal structure of YTH^YTHDF2 at 2.1 Å resolution (Figure 1D). The YTH domain forms a dimer in the crystal due to crystal packing (Supplementary information, Figure S1B and Table S1B). The overall structure shows a globular fold with a central core of four-stranded β-sheets surrounded by four α helices and flanking regions on two sides. A Dali search indicates that YTH^YTHDF2 is structurally similar to YTHDC1 (PDB: 2YUD) with a root-mean-squared deviation (rmsd) of 1.54 Å for 121 Cα atoms, suggesting a conserved mechanism for m⁶A-RNA recognition (Supplementary information, Figure S1C).

Electrostatic potential surface of YTH^YTHDF2 shows a patch that is enriched in basic residues, which may be involved in RNA recognition (Figure 1E). The basic patch is formed by residues R411 and K416 on strand β1, R441 on helix α2, and R527 on the loop connecting helices α3 and α4 (Figure 1A and 1D). Close to this basic patch, a hydrophobic pocket is formed by aromatic residues Y418, W432, W486 and W491 and is supported by the loop connecting strand β1 and helix α1, the loop connecting strands β3 and β4, and the loop connecting helices α1 and α2 (Figure 1A and 1D).

Based on above structural analyses, we further characterized several residues that are potentially important for m⁶A-RNA recognition. Wild-type and mutant YTH^YTHDF2 were purified and used for the FP assays (Figure 1F-1G and Supplementary information, Table S1A). Consistent with the structural observations, K416A and R527A mutations of YTH^YTHDF2 significantly decreased the binding affinity to both A-RNA (∼5 and ∼10 folds, respectively) and m⁶A-RNA (∼25 folds), suggesting that these two residues are involved in binding to the backbone of RNA, but may not recognize the methyl-group of m⁶A. Similar results were obtained for the K416A mutant of YTH^YTHDF2 in EMSA (Supplementary information, Figure S1D). In contrast, R411A and R441A mutations of YTH^YTHDF2 slightly decreased RNA-binding affinity toward A-RNA (∼2 folds) and m⁶A-RNA (∼3 folds). It is worthy to note that we calculated binding affinities of wild-type and two mutants (R411A and R441A) of YTH^YTHDF2 to m⁶A-RNA based on both the two-step and one-step binding modes (Supplementary information, Figure S1A and Table S1). The binding affinities of other mutants to m⁶A-RNA were calculated based on the one-step binding mode (Supplementary information, Table S1A).

In the FP assays, mutating two hydrophobic residues W432 and W486 to alanine markedly decreased the binding affinity of YTH^YTHDF2 to m⁶A-RNA, but barely changed the binding affinity to A-RNA, suggesting that W432 and W486 are important for specific recognition of m⁶A. To test the involvement of these two residues in m⁶A recognition in vivo, we measured the ratio of m⁶A/A levels in the RNA products immunoprecipitated by wild-type and mutant YTHDF2 full-length proteins from HEK293T cells. Indeed, W432A and W486A mutations decreased the m⁶A-RNA selectivity of YTHDF2. Furthermore, the W432A mutant of YTH^YTHDF2 was unable to effectively pull down previously identified YTHDF2 targets, SON and CREBBP, in HeLa cells. Taken together, these data suggest that residues W432 and W486 are essential for specific recognition of m⁶A by YTHDF2 (Figure 1H and Supplementary information, Figure S1E and S1F).

Circular dichroism (CD) measurements show that wild-type and mutant YTH^YTHDF2 have similar secondary structure composition, suggesting that the overall structure of YTH^YTHDF2 is not disrupted by mutations of these residues (Supplementary information, Figure S1G). Notably, residues W432 and W486 are highly conserved among the YTHDF family members from yeast to human (Figure 1A), further supporting their important role in mediating specific recognition of m⁶A-RNA. When our manuscript was under revision, two crystal structures of m⁶A-RNA in complex with YTH domains from YTHDC1¹⁴ and Zygosaccharomyces rouxii MRB1 (ZrMRB1)¹⁵ were reported. Structural comparison shows that residues W432 and W486 of YTH^YTHDF2 adopt a similar conformation to that of m⁶A-binding residues in the two complex structures (Supplementary information, Figure S1H), suggesting a conserved mechanism for m⁶A-RNA recognition by YTH domains.

Taken together, our studies indicate that the basic residues K416 and R527 on the surface of YTH^YTHDF2 are involved in binding to the RNA backbone, and residues W432 and W486 within the hydrophobic pocket contribute to the specific recognition of m⁶A. Our study also provides a platform for further investigations of additional YTH-m⁶A-RNA complex structures, which would reveal molecular mechanisms for specific recognition of m⁶A-RNA by YTHDF2 and other YTH family proteins.

The coordinate and structure factor for the YTH^YTHDF2 structure have been deposited into the Protein Data Bank under the accession code of 4WQN.