Structural basis of nucleic acid recognition and 6mA demethylation by human ALKBH1

Dear Editor,

DNA N⁶-methyladenine (6mA) modification is common in prokaryotes¹ and eukaryotes,² involving in gene regulation, transposon, stem cell differentiation, and human tumors. At present, it has been confirmed that 6mA is ubiquitous in the human genome, and [G/C]AGG[C/T] is the most prominent motif for 6mA modification.³ Human ALKBH1 (hALKBH1), one of the nine human homologs of the AlkB family, is an Fe(II) and α-ketoglutarate (α-KG)-dependent dioxygenase and highly conserved in mammals. AlkB family proteins can repair damaged DNA/RNA or other lesions. hALKBH1 exhibits demethylation activity toward 6mA, and its abnormal expression has been found in many human cancers and developmental defects, such as tissue malformation and gender imbalance.^3,4,5 DNA methyltransferase N6AMT1 and demethylase hALKBH1 mediate the methylation and demethylation of DNA 6mA in the human genome, respectively. The abnormal distribution of 6mA has been found in many cancers.^3,4 Interestingly, hALKBH1 is also reported to have demethylation activity toward other four kinds of substrates, such as histone H2A,⁶ m³C on DNA and RNA,⁷ m⁵C or m¹A on tRNA.^8,9,10 Therefore, the molecular function of hALKBH1 is still controversial and the functional mechanism is unclear. In addition, the sequence identity between hALKBH1 and other solved structures is less than 19%, which hinders the understanding of action mechanism and the potential drug applications.

Here we determined two crystal structures of binary and ternary complex of hALKBH1_19–369, hALKBH1_19–369-Mn²⁺ and hALKBH1_19–369-Mn²⁺-α-KG at resolutions of 1.97 and 2.8 Å, respectively (Supplementary information, Fig. S1 and Table S1). hALKBH1_19–369 (referred to as hALKBH1 hereafter) contained the enzymatically active center and retained full demethylation activity toward 6mA ssDNA, but not toward 6mA on dsDNA or m¹A on ssDNA, the same as hALKBH1 WT (Fig. 1a, b). The overall structures of hALKBH1-Mn²⁺-α-KG and hALKBH1-Mn²⁺ are similar, with root mean square deviation of 1.4 Å. The hALKBH1 contains a unique N-terminal Flip0, the nucleotide recognition lid (NRL, containing Flip1 and Flip2), and the central catalytic core. A highly conservative double-stranded β-helix (DSBH) fold in the catalytic core is a characteristic of the α-KG-dependent dioxygenase superfamily (Fig. 1c; Supplementary information, Fig. S2). Four antiparallel β-strands β7, β9, β12, and β14 form the major sheet, whereas β8, β10, β11 and β13 form the minor sheet. hALKBH1 has a conservative HxD…H metal ion-binding sequence and an R…R α-KG-binding sequence. Apart from Mn²⁺, α-KG is further stabilized by three hydrogen bonds formed by the side chains of Asn220, Asn340 and Tyr222 as well as three salt bridges formed by Arg338 and Arg344 (Fig. 1d, e). Isothermal titration calorimetry analysis of the hALKBH1-α-KG interaction revealed a dissociation constant (K_d) of 4.5 μM and a 1:1 stoichiometry (Supplementary information, Fig. S3b). The total surface area of hALKBH1 decreased obviously from 14,445 to 13,893 Ų during the addition of α-KG, indicating that hALKBH1 whole structure shrank with the association of α-KG, which caused many conformational changes including those of active sites, β5, β8, Flip1 and Flip2 (Supplementary information, Movie S1, Fig.S3c, d). Moreover, of the eight catalytic residues, Arg344 and Tyr222 showed significant conformational changes, interacting with α-KG through moving 2.5 Å and 2.3 Å toward the active center, respectively. Besides, Tyr184 and Glu236 of Flip2 exhibited the most obvious conformational changes and moved to Arg344, forming a stabilization triangle by hydrogen bonds (Fig. 1d). The stability triangle of Tyr184-Arg344-Glu236 near the α-KG-binding site is essential for 6mA ssDNA demethylation activity. Any single mutation of the triangle abolished the demethylation activity (Fig. 1e, f), but did not reduce the DNA-binding affinity (Supplementary information, S3e). Interestingly, these large conformational changes induced by α-KG were not observed in other AlkB members,^11,12,13,14 revealing a novel function of the essential triangle as a scaffold for the catalytic activity of hALKBH1.

Fig. 1

Structural and biochemical studies of hALKBH1. a Primary structure of hALKBH1. b Detection of hALKBH1 demethylation activity by electrophoresis. The demethylated hALKBH1 product at 1 h was enzymatically digested using nuclease DpnII. The results showed that hALKBH1 exhibited demethylation activity toward ssDNA containing 6mA, but not 6mA dsDNA, m¹A ssDNA or m¹A dsDNA. c A cartoon representation of hALKBH1-Mn²⁺-α-KG structure. Mn²⁺, purple ball. The residues without visible electron density, dashed lines. Flip0, purple. Flip1, blue. Flip2, cyan. d Structural comparison of the active centers of hALKBH1-Mn²⁺-α-KG (purple) and hALKBH1-Mn²⁺ (yellow). Interactions between Arg-344, Tyr-184, and Glu-236 are shown in dashed lines. e The interaction network around Mn²⁺ and α-KG. The 2mF_o-DF_c electron density map (light blue), 1.5 σ. f Demethylation activity of the wide-type hALKBH1 and its mutants toward 6mA ssDNA at 1 h. g Structural comparison of hALKBH1 (purple) with other AlkB proteins. AlkB (cyan; PDB 2FDH), hALKBH2 (green; 3BTZ), hALKBH3 (yellow; 2IUW), hALKBH5 (tv-blue; 4NRM), and hFTO (gray; 4QKN). hALKBH1-Flip1, Flip1 of other AlkB proteins and Flip2 are highlighted in purple, blue and yellow boxes, respectively. h, i Mutations of the key residues in the lid region greatly impair hALKBH1 demethylation activity toward 6mA ssDNA at 1 h. j Structural comparison of hALKBH1 (purple) with AlkB, hALKBH2, hALKBH3, hALKBH5. The hALKBH1-Flip0 and hALKBH5-Flip3 are highlighted within the purple frame. k Structural alignment of hALKBH1 with hALKBH2-dsDNA complex (dsDNA in orange; 3BTZ). The unmethylated strand of dsDNA would sterically clash with Flip0 of hALKBH1. m¹A and α-KG (purple) are shown in sticks. l Mutations in the Flip0 region greatly impair hALKBH1 demethylation activity at 1 h. m Schematic diagram of bubble- and bulge-structured DNA. n, o Detection of hALKBH1 demethylation activity at 30 min by electrophoretic and statistical analysis. hALKBH1 had the strongest demethylation activity toward Bugle6. Error bars, SD of three replicates. p–r HPLC detection of 6mA demethylation of dsDNA, ssDNA or Bulge6 DNA respectively by hALKBH1. The relative content of 6mA is the same when using T, C or G as internal standard, respectively. s, t Proposed model of hALKBH1 binding to Bubble6 (s) or Bulge6 DNA (t). DNA substrates, orange. Surface colored as a gradient ranging from red (negative) to blue (positive). Surface potential (±72 kBT/e) of hALKBH1. The 6mA nucleobase inserted into the catalytic pocket is indicated

Compared with other members of the AlkB family, the NRL of hALKBH1 has several unique structural features (Supplementary information, S4). The hALKBH1 Flip1 region is unique and long, leaving a larger binding space over the active site pocket (Fig. 1g). K158A/R159A/R160A/R162A and K167A/R169A mutants completely lost the 6mA demethylation activity, and the activity of K158A/R159A/R160A mutant was significantly compromised (Fig.1h). Moreover, the Flip2 contains a pair of antiparallel β-sheets and a long loop with high B factors (Supplementary information, Fig. S5). Structure-based sequence analyses and mutagenesis experiments confirmed that key residues in the NRL region, such as Arg169, Trp170, Tyr177, and Trp179, are potential determinants of hALKBH1 for 6mA recognition and demethylation (Fig. 1i; Supplementary information, Fig. S6). The distinctive composition and conformation of Flip1 and Flip2 are likely to confer substrate selectivity on hALKBH1.

Intriguingly, a significant structural feature of hALKBH1 at the N-terminus is the Flip0 (residues 19–32). It is highly conserved in ALKBH1 among various mammalian species, but not in OsALKBH1¹⁵ or other AlkB family members (Fig. 1j; Supplementary information, Fig. S7). When superimposed with the hALKBH2-dsDNA complex structure, the Flip0 region of hALKBH1 is well accommodated by the modified strand, and the modified nucleotide enters the catalytic pocket. However, Flip0 seriously collides with the unmethylated strand (Fig. 1k). We hypothesized that Flip0 impedes the access of paired dsDNA to the active site and this may be the structural basis for the selectivity of hALKBH1 toward single-stranded substrates. EMSA data revealed that dsDNA and ssDNA can bind to hALKBH1, and EMSA and chromatographic analyses showed a direct binding between Flip0 and ssDNA/dsDNA (Supplementary information, Figs. S8 and S9). However, hALKBH1 has no demethylation activity on 6mA of dsDNA. Based on the above results, we believe that Flip0 is disadvantageous for 6mA dsDNA to enter the hALKBH1 catalytic pocket, and that dsDNA and hALKBH1 are a non-productive combination in the presence of Flip0, rather than a productive way. Flip0 binds tightly adjacent to the minor sheet of the DSBH fold, and most of the interactions are hydrophobic with all the involved hydrophobic residues conserved in ALKBH1 proteins of different species except for OsALKBH1 (Supplementary information, Fig. S7a). R24A/K25A, R28A/R31A, R24A/K25A/R28A/R31A mutants and Flip0-deleted mutant hALKBH1_36–369 completely lost demethylation activity toward 6mA ssDNA (Fig. 1l). Therefore, the unique Flip0 is essential for hALKBH1 activity and discriminating single-stranded from paired double-stranded substrates. This structural characteristic and function of hALKBH1 Flip0 are greatly different from those of Flip3 in hALKBH5 (Fig. 1j),¹¹ whose structure can be disrupted by reduction of the disulfide bond. So hALKBH1 uses a novel substrate recognition mechanism that is distinct from those of known AlkB demethylases.

Next, we examined the surface charge distribution of hALKBH1. The region of positive charge extends along with Flip1, Flip0, and Flip2 near the active center of the DSBH domain, and forms a large substrate-binding groove. Combining the substrate-binding ability and catalytic activity of hALKBH1 mutants with mutations in Flip0, Flip1 and Flip2 regions, a possible binding model of hALKBH1 with ssDNA is proposed: one end of single-stranded nucleic acid interacts with Flip1, and the other end extends along the direction of the groove formed by Flip2 and Flip0 (Supplementary information, Fig. S10d).

Compared with other AlkB family proteins (Supplementary information, Fig. S10b), the distribution of positive charges of hALKBH1 is different and the area is very large. The positive surface of hALKBH1 significantly exceeds that of ssDNA binding in the hALKBH1-ssDNA model. In addition, the demethylation activity of hALKBH1 on ssDNA is weak, probably because ssDNA is not the most suitable substrate. Based on structure, we introduced various substrates, such as hemi-methylated DNA bubble and bulge with different numbers of mismatched base pairs in the middle of double-stranded DNA (Fig. 1m; Supplementary information, Fig. S11). Because the HPLC method was not sensitive enough, we developed a high-throughput methylation-sensitive restriction digest assay to detect the demethylation activity of hALKBH1 toward DNA bubble/bulge. Among different DNA bubbles, hALKBH1 displayed the highest demethylation activity when the number of mismatched base pairs was 5–7. The demethylation activity of hALKBH1 on Bubble6 DNA was twice that toward ssDNA (Supplementary information, Fig. S12). Intriguingly, hALKBH1 had strongest activity on Bulge6 DNA (Fig. 1m–o). Similar results were obtained by using the method of HPLC (Fig. 1p–r; Supplementary information, Fig. S13). Furthermore, we tested the demethylation activity toward DNA bulge with 6mA at different mismatch position, and found that the activity was the highest at the third mismatched base pair. Notably, the demethylation activity of hALKBH1 to Bulge6 DNA was 3–4 folds relative to ssDNA (Fig. 1m–o; Supplementary information, Figs. S13 and S14), and the variation of hALKBH1 demethylation activity toward DNA bulge/bubble, ssDNA or dsDNA was not due to different binding affinities between substrates and hALKBH1 (Supplementary information, Fig. S15). Therefore, structure-based substrate screening reveals that DNA bubble/bulge is more suitable for hALKBH1. The 6mA demethylation activity of hALKBH5 on dsDNA, ssDNA, DNA bubble and bulge was also assayed. hALKBH5 is a member of the AlkB family, and is also a Fe(II)/2-OG-dependent dioxygenase. Results showed that when the substrate was Bubble6 or Bulge6, the 6mA demethylation activity of hALKBH5 was 41% or 52.5% of that toward ssDNA, respectively (Supplementary information, Fig. S16). Therefore, DNA bubble and bulge are specific substrates for hALKBH1.

Based on the binding and activity assays (Supplementary information, Fig. S17), we presented hypothetical structural models of hALKBH1-Bubble6 DNA and hALKBH1-Bulge6 DNA (Fig. 1s, t). The length of the substrate-binding groove is ~40 Å, suitable for binding 5–7-nt bubble-structured DNA, consistent with the experimental results. The opening of the bubble structure promotes the binding and makes the 6mA insert deeper into the active center (Fig. 1s). In the hALKBH1-Bulge6 DNA binding model, the double-stranded DNA extends along the groove formed by Flip0, Flip1, and Flip2. The Bulge6 DNA binds to the Flip1 and Flip2 near the active pocket (Fig. 1t). The conformation allows a DNA circle with diameter less than 12 Å, consistent with the experimental result that hALKBH1 displayed the highest activity toward 6-nt bulge-structured DNA. In this binding model, 6mA at the third mismatched base pairs is more suitable to penetrate into the active center. Compared with ssDNA, partly unpaired dsDNA is more biologically relevant in genomic DNA, such as in mismatch repair. Our structure-based substrate screening reveals that DNA bulge and bubble, rather than ssDNA, are preferential substrates of hALKBH1 and have more physiological significance. Further investigations are required to elucidate the detailed molecular mechanism.

In summary, our research revealed several unique structural features of hALKBH1 and found its novel native substrates. Our findings can further be used to study the regulatory mechanism of 6mA modification in different basic biological processes and in the field of DNA epigenetics, to guide future drug research.

The atomic coordinates and structure factors of the hALKBH1-Mn²⁺-α-KG and hALKBH1-Mn²⁺ complex have been deposited in the Protein Data Bank under the accession code 6IE2 and 6IE3.