The PWWP domain of mammalian DNA methyltransferase Dnmt3b defines a new family of DNA-binding folds

Abstract

The PWWP domain is a weakly conserved sequence motif found in > 60 eukaryotic proteins, including the mammalian DNA methyltransferases Dnmt3a and Dnmt3b. These proteins often contain other chromatin-association domains. A 135-residue PWWP domain from mouse Dnmt3b (amino acids 223--357) has been structurally characterized at 1.8 A resolution. The N-terminal half of this domain resembles a barrel-like five-stranded structure, whereas the C-terminal half contains a five-helix bundle. The two halves are packed against each other to form a single structural module that exhibits a prominent positive electrostatic potential. The PWWP domain alone binds DNA in vitro, probably through its basic surface. We also show that recombinant Dnmt3b2 protein (a splice variant of Dnmt3b) and two N-terminal deletion mutants (Delta218 and Delta369) have approximately equal methyl transfer activity on unmethylated and hemimethylated CpG-containing oligonucleotides. The Delta218 protein, which includes the PWWP domain, binds DNA more strongly than Delta369, which lacks the PWWP domain.

Fig. 1

Dnmt3 MTase family. a, Schematic representation of the Dnmt3 family. The amino acid sequences of Dnmt3a and Dnmt3b are highly conserved with respect to one another in the PWWP domain, CXXC domain and C-terminal MTase domain. In contrast, the N-terminal variable domain differs substantially between these two proteins. The alternative splice site in Dnmt3b1 that generates 3b2 is indicated by a triangle. Below the full-length Dnmt3b2 (WT) are the depicted protein fragments (Δ218, Δ369, PWWP and CXXC) and the SDS-PAGE gels (right) of the recombinant proteins used throughout this study. b, Sequence alignment of Dnmt3 PWWP domains. The prefix M indicates mouse; H, human; and Zf, zebrafish. The mouse Dnmt3b residue numbering is shown above the sequences, with the secondary structure elements (numbers for strands and letters for α-helices). The highlighted amino acids are either invariant (cyan) or similar (yellow). The SWWP motif is colored in magenta. In zebrafish, there are three partial sequences for Dnmt3 orthologs. Two ESTs (expressed sequence tags) contain only the N-terminal region, including the PWWP domain: AW305577 is more similar to Dnmt3a, whereas BG307263 is more similar to Dnmt3b. The third EST (AAD32631) has a C-terminal portion (CXXC and MTase domains) that shares similarity to the corresponding portion of mammalian Dnmt3 family. However, its N-terminal PWWP domain deviates significantly from the sequences, including eight positively charged (blue) and three negatively charged residues (red), resulting in a pI of 6.3.

Fig. 2

Structure of Dnmt3b2 PWWP domain. a, RIBBON diagram, with invariant amino acids in cyan, conserved in yellow and varied in gray. The SWWP motif is colored in magenta. In the crystal packing contacts, strand β6 from one molecule interacts with strand β4, which is the edge strand of the β-sheet (4 3 2 1 5), of a neighboring molecule through backbone–backbone hydrogen bonds in an antiparallel arrangement. b, Molecular surface and charge distribution. The view is oriented similarly to that in (a). The surface is colored according to charge: positively charged groups (Arg and Lys) are blue, negatively charged groups (Glu and Asp) are red and uncharged groups are white. c, Contours of the electrostatic potential at ±2 KT e⁻¹ calculated and displayed in two orientations as a mesh surface. The positive potential is shown in blue, and the negative potential in red. Panels (b,c) were done using GRASP.

Fig. 3

DNA binding and methyl transfer activity of Dnmt3b2. a, Nitrocellulose filter binding experiments using different protein fragments of Dnmt3b2 (300 nM) and labeled oligonucleotides (5 nM), whose sequences contained a single CpG, hemimethlyated CpG (hemi-CpG), fully methylated CpG (MpG) or non-CpG (see Methods). b, Nitrocellulose filter-binding experiments using different amounts of Δ218 (0, 5, 15, 25, 50, 100, 150, 200 and 300 nM), PWWP (0, 25, 50, 100, 150, 200 and 300 nM) and a labeled CpG-containing oligonucleotide (5 nM). K_d values were estimated from a direct nonlinear regression fit (with the saturation level as one of the variables). c, CPM counts. Methyl transfer activities of purified recombinant Dnmt3b2 proteins (expressed in E. coli) and full-length Dnmt1 (ref. 5) (expressed in Sf21 insert cells using recombinant baculoviruses).

Fig. 4

Comparison of PWWP domain with its structural homologs. a, Ribbon diagram and stereo view of the superimposition of PWWP (yellow) and Sp100b SAND domain (PDB 1H5P) (orange). For clarity, the C-terminal helix of SAND domain is not shown in the superimposition. The structural similarity between the two domains was detected manually. Below is the structure-based sequence alignment between the β-barrel structure of PWWP and SAND. The line between the two sequences indicates the residues used in superimposition. b, Ribbon diagram and stereo view of the superimposition of PWWP (yellow) and the SMN Tudor domain (PDB 1G5V) (orange). The r.m.s. deviation is 0.8 Å when 43 Cα atoms (out of total 56 residues in Tudor) are structurally aligned between the two domains (DALI Z-score of 6.3). The largest difference lies in the loop between strands β2 and β3. In comparison, the five-stranded β-barrel of PWWP is remotely similar to the SH3 structures (DALI Z-score of ~3.3), a small basic folding unit. The structure-based sequence alignment between the β-barrel structure of PWWP and Tudor is shown below, with the line between the two sequences indicating the residues used in superimposition.

Fig. 5

Views of defined elements in the Dnmt3b2 PWWP structure. a, Hydrophobic core of the β-barrel centered on residue Trp 270 (β3), packing against Gly 238 (β1), Ile 240 (after strand β1), Ala 248 (β2), Val 268 (β3) and Ile 279 (β4). A hydrogen bond is formed between Trp 270 and Ser 277 (β4). These residues are from strands β1 to β4; therefore, these interactions appear to anchor the β-subdomain. b, Hydrophobic core of the helical bundle. The interactions involve helices αD (Phe 299), αE (Tyr 310, Met 309 and Leu 313), αF (Met 339, Trp 342 and Ala 343), αG (Leu 355), β6 (Phe 324) and two of their connecting loops — the loop prior helix αD (Phe 294) and the loop between helices αF and αG (Phe 347). c, The interactions involving the SWWP motif in the interface between the β-barrel and helical bundle substructures. The Pro 247 at the end of the SWWP motif is completely buried within the interface, whereas the other three residues are exposed on the surface. Trp 245 and Trp 246 point in opposite directions but have a similar network of interactions — the indole rings are flanked by a ring structure (His 293 imidazole ring and Tyr 305 aromatic ring) and by the aliphatic portion of a Lys side chain (Lys 239 of strand β1 and Lys 241 of loop L-12). One difference between the Trp residues is that a water molecule is held in position by three hydrogen bonds with Trp 246, Ser 244 and Lys 241, but there is no equivalent water associated with Trp 245. d, Network of intramolecular polar interactions (centered on residue Arg 319). The two terminal amino groups of the positive charged Arg 319 side chain (helix αE) interact with main chain carbonyl oxygen atoms of Gly 233 (N-terminal loop) and Lys 356 (C-terminal loop). The plane of the Arg 319 guanidino group makes van der Waals contacts with the proline ring of Pro 357. The negatively charged Asp 234 side chain (next to Gly 233) interacts with the main chain amino group of Gly 231. e, Stereo view showing an omit electron density map contoured at 4σ, superimposed with refined coordinates of SWWP motif.

Fig. 6

Sequence alignment of PWWP-containing proteins. a, The residues ‘WP’ (magenta) are conserved among all, except for the BR140 family in which a Tyr in the place of Trp. Other conserved residues are highlighted with a yellow background (hydrophobic), blue (basic) and red (acidic). The accession numbers are colored based on the range of calculated pI value: blue (8–10) and magenta (6–7). The prefix M indicates mouse; H, human; At, A. thaliana; and Sp, Schizosaccharomyces pombe. In addition to the sequences shown, the other PWWP-containing proteins used in alignment are from human (AAD53063, AAD32631, CAC28351, AAF25871, CAB75688, BAA86439, BAA86600, NP_055392, CAB66669, AAG44637); mouse (AAD33084, AF064553, AAB88445, BAA90478, BAA22896, AAF65469, BAA22895); A. thaliana (AAF64539, AAF23297, CAB86049, BAB11589, AAF29390, AAD24842, CAB93724, CAB87752, AAF18621, BAA97320, CAB71104); zebrafish (BG307263, AW305577, AAC40177, AI588565), Drosophila (AAF56762, AAF59276, AAF57004, AAF55003, AAF57177, AAF58494), Xenopus (AW636253); Saccharomyces. cerevisiae (AAB64719, NP_013758); S. pombe (CAA22121, AB027814); bovine (CAA40348); Onchocerca volvulus (BE132518); Ceanorhabditis elegans (AAB52437); and Triticum aestivum (BE500209). b, The schematic representation of PWWP-containing proteins. Abbreviations are NLS (nuclear localization signal), HMG (high mobility group), PCNA (PCNA binding motif), SET (Suvar39, Enhancer of zeste, Trithorax), bromo (bromodomain) and chromo (chromodomain). WHSC1 is a 136-kDa protein expressed ubiquitously in early development and preferentially in rapidly growing embryonic tissues. BS69 is a 69-kDa nuclear protein that specifically suppresses adenovirus E1A-activated transcription. BR140 shows similarity to a group of transcriptional coactivators, including the TAF250 subunit of TFIID. MSH6 is mammalian MutS homolog essential for DNA mismatch repair, which contains a PWWP domain at its extreme N-terminus. However, msh6 orthologs in S. cerevisiae, C. elegans and A. thaliana do not contain this domain. The fully sequenced C. elegans contains only one PWWP protein, AAB52437; this protein contains an N-terminal PWWP domain and a C-terminal SIR2-like domain. Although the known eukaryotic members of the NAD-dependent protein deacetylase SIR2 family often contains the N- and/or C-terminal extensions outside of a conserved core region, none of these contains a PWWP domain. A CXXC zinc-binding motif is found in the structure of an Archaeoglobus fulgidus SIR2 homolog, which contains a CXXC-X_15–20-CXXC zinc-binding sequence motif.