Identification of YTH Domain-Containing Proteins as the Readers for N1-Methyladenosine in RNA

Abstract

N1-methyladenosine (m¹A) is an important post-transcriptionalmodification in RNA; however, the exact biological role of m¹A remains to bedetermined. By employing a quantitative proteomics method, we identified multipleputative protein readers of m¹A in RNA, including several YTH domain familyproteins. We showed that YTHDF1–3 and YTHDC1, but not YTHDC2, could binddirectly to m¹A in RNA. We also found that Trp⁴³² in YTHDF2, a conservedresidue in the hydrophobic pocket of the YTH domain that is necessary for itsbinding to N⁶-methyladenosine (m⁶A), is required for its recognition of m¹A. An Ananalysis of previously published data revealed transcriptome-wide colocalization of YTH domain-containing proteins and m¹A sites in HeLa cells, suggesting that YTH domain-containing proteins can bind to m¹A in cells. Together, our results uncovered YTH domain-containing proteins as readers for m¹A in RNA and provided new insight into the functions of m¹A in RNA biology.

Graphical abstract

RNA is known to contain more than 100 distinct types of post-transcriptional modifications, which can modulate its splicing, localization, stability, and translation.¹ It was found that methylation at the N⁶ position of adenosine (m⁶A) is involved in the epigenetic control of gene regulation.²⁻⁶ Recent transcriptome-wide mapping also revealed the presence of m¹A in human mRNA,⁷⁻⁹ though there is some controversy about the accuracy and specificity for some of the m¹A mapping methods.¹⁰ These studies also suggested the potential role of m¹A in modulating mRNA splicing and translation.⁷⁻⁹ TRMT6, TRMT61A, and TRMT10C are three known m¹A methyltransferases in human cells (writers),⁹ and m¹A can be demethylated to adenosine by ALKBH1 and ALKBH3 (erasers).^8,11 However, it remains unknown which cellular proteins are involved in binding to m¹A in RNA (readers).

YTH domain-containing proteins can bind directly to m⁶A in RNA. YTHDF1 promotes protein synthesis by interacting with the translation machinery;¹² YTHDF2 accelerates the degradation of m⁶A-carrying mRNA and affects the translation of heat shock proteins;^13,14 YTHDF3 facilitates the translation and decay of m⁶A-bearing RNA in synergy with YTHDF1 and YTHDF2;^15,16 YTHDC1 regulates mRNA splicing by modulating pre-mRNA splicing factors and mediates nuclear export of m⁶A-harboring RNA;^17,18 YTHDC2 was found to be essential for meiosis in mammalian germline.¹⁹⁻²¹ Together, the YTH domain-containing proteins modulate various biological processes through binding to m⁶A-modified RNA.

In this study, we employed a mass spectrometry-based method to identify the m¹A-binding proteins in RNA. We revealed that several YTH domain-containing proteins directly interact with m¹A in RNA. In addition, functional analysis suggested important roles of YTH domain family proteins in the regulation of m¹A-carrying RNA. Our results provided a foundation for further understanding of the biological functions of this important RNA modification.

RESULTS AND DISCUSSION

To understand better the biological functions of m¹A in RNA, we employed the stable isotope labeling by amino acid in cell culture (SILAC)²²-based quantitative proteomics method to screen systematically for cellular proteins that bind to m¹A-bearing RNA (Figure 1a). To this end, we used a 5′-biotin-labeled m¹A-carrying RNA sequence derived from human SOX18 gene, which was previously shown to harbor an m¹A site near the start codon in both HEK293T and HeLa cells,^7,8 as the probe bait and the corresponding unmethylated sequence as the control bait. We cultured the cells separately in heavy and light culture media and extracted proteins from these cells. An equal amount of proteins from the heavy- and light-labeled cells were incubated with biotin-conjugated m¹A-bearing RNA and the corresponding unmethylated RNA, respectively, which was designated as the forward SILAC experiments. The opposite incubations in reverse SILAC experiments were conducted to remove experimental bias. After the incubation, the RNA-conjugated beads were extensively washed to remove nonspecific proteins, and the bound proteins were isolated from the beads, digested with trypsin, and analyzed using LC−MS/MS (Figure 1a).

Figure 1.

Identification of m¹A-binding proteins. (a) Schematic overview of the SILAC-based quantitative proteomics method for discovering m¹A-binding proteins. Shown is the workflow for a forward SILAC labeling experiment. (b,c) Scatterplot of the proteins identified from pull-down assays using m¹A- (b) or m⁶A- (c) carrying RNA over the corresponding unmodified RNA and the whole-cell protein lysate of HEK293T cells. The proteins present in top right quadrant are m¹A- (b) or m⁶A (c)-binding proteins, whereas those in the bottom left quadrant are proteins which bind more weakly to the m¹A- or m⁶A-bearing RNA than the control RNA probe. The proteins that fall in the bottom right quadrant represent those proteins displaying large light/heavy ratios in both forward and reverse SILAC labeling experiments, which are likely due to incomplete SILAC labeling (for those proteins with slow turnover rate). The data were based on results from two forward and one reverse SILAC experiments.

The LC−MS/MS results revealed multiple proteins exhibiting preferential binding toward the m¹A-bearing RNA over the corresponding rA-containing RNA (Figure 1b, Figure S1). In HEK293T cells, these proteins include YTH domain family members YTHDF3 and YTHDF2 and heterogeneous nuclear ribonucleoprotein hnRNPD, with the SILAC protein ratios (m¹A/rA) being 6.9 ± 1.7, 4.1 ± 2.7, and 2.3 ± 0.3, respectively (Figure 1b, Table S2). In HeLa cells, the proteins which bind more strongly to m¹A-bearing RNA probe over rA-bearing RNA probe encompass TAR DNA-binding protein (TARDBP), DEAD-box helicase DDX56, YTHDF proteins, and others (Figure S1, Table S2). Among them, YTHDF1, YTHDF2, and YTHDF3 displayed SILAC ratios (m¹A/rA) of 2.3 ± 0.7, 1.9 ± 0.3, and 2.1 ± 0.4, respectively (Table S2). It is worth noting that our LC−MS/MS results also identified some proteins that bind more strongly to rA-containing RNA probe over m¹A-bearing RNA probe, such as interleukin enhancer binding factors ILF2 and ILF3, which form a complex and negatively regulates microRNA processing,²³ KU70 and KU80, which forms a heterodimer and functions in DNA double strand break repair via the nonhomologous end-joining pathway,²⁴ etc. (Figure 1b and Table S2).

We also performed the SILAC experiments using the corresponding m⁶A and rA probes in HEK293T cells. As expected, the SILAC-based interaction screening resulted in the identification of YTHDF3, YTHDF2, and YTHDF1 as proteins displaying preferential binding toward the m⁶A-containing RNA over the rA-containing RNA substrate, with the SILAC protein ratios (m⁶A/rA) being 14.6 ± 7.9, 9.7 ± 3.6, and 9.0 ±2.7, respectively (Figure 1c, Table S3). This result is consistent with the previous finding,¹³ validating that our SILAC-based method is capable of identifying specific RNA-binding proteins. Aside from YTH domain family proteins, our SILAC-based interaction screening also identified several other m⁶A-binding proteins, such as PCBP1 and CNBP, which exhibited the SILAC ratios (m⁶A/rA) of 2.1 ± 0.8 and 1.9 ± 0.3, respectively (Figure 1c, Table S3). Moreover, our results showed that some proteins exhibit stronger binding affinities to the rA-containing RNA probe over the corresponding m⁶A-bearing RNA probe, such as heterogeneous nuclear ribonucleoprotein hnRNPA0, and the aforementioned ILF2, ILF3, KU70, and KU80, etc.

Representative LC−MS results for a tryptic peptide derived from YTHDF2, SINNYNPK are shown in Figure 2, which clearly revealed the stronger binding of YTHDF2 to the m¹A-containing RNA than the corresponding unmethylated probe in both forward and reverse SILAC experiments (MS/MS shown in Figure S2). The selective binding of YTHDF3 toward m¹A-containing RNA also found its support from a representative tryptic peptide derived from YTHDF3, i.e., IGGDLTAAVTK (Figure 2, MS/MS shown in Figure S3).

Figure 2.

Representative ESI-MS for the [M + 2H]²⁺ ions of a tryptic peptide of YTHDF2, SINNYNPK (a, b) and a tryptic peptide of YTHDF3, IGGDLTAAVTK (c, d) revealing the preferential binding of YTHDF2 and YTHDF3 toward the m¹A-containing RNA probe in both forward (a, c) and reverse (b, d) SILAC experiments.

SILAC-based quantitative proteomic approach may also result in the discovery of those proteins that interact indirectly with m¹A-bearing bait through protein−protein interaction(s). Thus, we examined whether YTH domain-containing proteins can interact directly with m¹A in RNA. Our results from electrophoretic mobility shift assay (EMSA) showed that YTHDF1–3 and the YTH domain of YTHDC1 (YTHDC1^YTH), but not the YTH domain of YTHDC2 (YTHD C2^YTH), bind more strongly to m¹A-carrying RNA substrate than its unmethylated counterpart (Figure 3a–d, Figure S4). We also compared the binding difference of YTH domain-containing proteins toward m¹A and m⁶A-bearing RNA baits. The EMSA results showed that YTHDF1–3 and YTHDC1^YTHbind more strongly to m⁶A-carrying RNA than m¹A-carrying RNA, with the K_d values being 1.3 ± 0.1/16.5 ± 1.5, 1.3 ± 0.1/5.8 ± 1.7, 1.9 ± 0.1/7.0 ± 1.1, and 0.7 ± 0.1/23.3 ± 2.1 μM (m⁶A/m¹A), respectively (Figure 3a–d). Together, our results support that YTHDF1–3 and YTHDC1^YTH can bind directly to m¹A-containing RNA substrate, albeit at affinities that are lower than those toward the m⁶A-containing RNA probe, which are in keeping with the findings made from SILAC-based interaction screening.

Figure 3.

YTH domain-containing proteins bind directly to m¹A-carrying RNA. (a−d) Electrophoretic mobility shift assay (EMSA) for measuring the binding affinity of YTHDF1 (a), YTHDF2 and W432A mutant proteins (b), YTHDF3 (c), and YTH domain of YTHDC1 (d) with m⁶A, A, and m¹A-carrying RNA probes. The K_d values (in μM) are listed in (a)−(d), and error bar represents the SEM (n = 3). (e) Overlap between the m¹A sites (200 bp window centered on the m¹A sites) and YTHDF1–3, and YTHDC1 PAR-CLIP binding sites in HeLa cells.

The X-ray crystal structure of YTHDF2 revealed three aromatic amino acid residues in the hydrophobic pocket of the YTH domain of YTHDF2 that are crucial for its binding toward m⁶A.^13,25 To explore whether this hydrophobic pocket also assumes an important role in binding toward m¹A, we conducted EMSA assay with a mutant form of YTHDF2 protein where the conserved Trp⁴³² was mutated to an alanine (W432A). The result indeed showed that the mutation led to a pronounced attenuation in binding affinity toward the m¹A-containing probe (Figure 3b), suggesting that m¹A may occupy the same hydrophobic pocket of YTHDF2 that is required for m⁶A recognition.

We further interrogated the publicly available transcriptome-wide m¹A mapping data⁷ and photoactivatable ribonucleoside cross-linking and immunoprecipitation (PAR-CLIP) data^12,16,18,26 of YTH domain-containing proteins in HeLa cells to assess whether YTH domain-containing proteins bind to m¹A-bearing RNA in cells. In this aspect, we analyzed the overlap between the m¹A sites (200 bp window centered on the m¹A sites) and YTHDF1–3, and YTHDC1 PAR-CLIP binding sites in HeLa cells. The result showed that 24.4%, 23.1%, 4.5%, and 12.5% of the m¹A sites overlapped with the YTHDF1, YTHDF2, YTHDF3, and YTHDC1 PAR-CLIP peaks, respectively (Figure 3e, Table S4). Therefore, these results suggest that YTH domain-containing proteins can recognize m¹A in transcripts in cells. Analysis of the ribosome profiling data and RNA lifetime data^12,16,26 in HeLa cells revealed that, upon YTHDF2 knockdown, the YTHDF2-bound m¹A-containing mRNA targets experienced a significant increase in lifetime compared with other targets (p = 1.0 × 10⁻²⁸, Mann− Whitney U-test, Figure S5, Table S5), whereas the YTHDF1-binding m¹A-containing targets exhibited a significant decrease in translation efficiency compared with other targets upon genetic depletion of YTHDF1 (p = 8.9 × 10⁻¹², Mann− Whitney U-test, Figure S6, Table S6), suggesting that YTHDF1 may promote translation whereas YTHDF2 mainly accelerates the degradation of m¹A-carrying transcripts.

CONCLUSIONS

In summary, we identified by using an unbiased quantitative proteomic method, for the first time, YTH domain-containing proteins as the readers of m¹A in RNA, which expands the repertoire of RNA modifications that can be regulated by this protein family. We demonstrated that YTH domain-containing proteins directly interact with m¹A-bearing RNA. Furthermore, we identified multiple other putative m¹A-binding proteins by using SILAC-based quantitative proteomic screening. Further characterizations of these proteins in m¹A recognition and RNA metabolism will improve our understanding about the biological functions of m¹A in RNA.