1 Structure and function of the nucleosome-binding PWWP domain 2 3 Su Qin1 and Jinrong Min1, 2 4 1 Structural Genomics Consortium, University of Toronto, 101 College Street, Toronto, 5 Ontario M5G 1L7, Canada. 6 2 Department of Physiology, University of Toronto, Toronto, Ontario M5S 1A8, 7 Canada. 8 Corresponding author: Min, J. ([email protected]). 9 10 Keywords: PWWP domain; nucleosome binding; histone binding; DNA binding; 11 crosstalk; epigenetic code - 1 - 1 Abstract 2 3 PWWP domain-containing proteins are often involved in chromatin-associated 4 biological processes, such as transcriptional regulation and DNA repair, and recent 5 studies have shown that the PWWP domain specifies chromatin localization. 6 Mutations in the PWWP domain have been linked to various human diseases, 7 emphasizing its importance. Structural studies reveal that PWWP domains possess a 8 conserved aromatic cage for histone methyl-lysine recognition, and synergistically 9 bind both histone and DNA, which contributes to their nucleosome binding ability and 10 chromatin localization. Furthermore, the PWWP domain often cooperates with other 11 histone and DNA “reader” or “modifier” domains to evoke crosstalk between various 12 epigenetic marks. Here, we discuss these recent advances in understanding the 13 structure and function of the PWWP domain. 14 15 - 2 - 1 Structural characteristics of the PWWP domain 2 The PWWP domain is named after a conserved Pro-Trp-Trp-Pro motif [1, 2]. 3 However, the name can be misleading as only the fourth residue Pro is absolutely 4 conserved. The PWWP domain was also named the HATH domain (homologous to 5 the amino terminus of HDGF (Hepatoma-derived growth factor)) [3] and the 6 RBB1NT domain (RBBP1 N-terminal domain) (PDB entry: 2YRV and Pfam entry: 7 PF08169). It is found ubiquitously in eukaryotes, ranging from unicellular organisms 8 to humans, and there are more than 20 PWWP domain-containing proteins in the 9 human genome, most of which are chromatin-associated (Table 1). 10 The PWWP domain belongs to the Royal superfamily, which also includes the 11 chromodomain, Tudor domain, and the Malignant Brain Tumor (MBT) domain [4]. 12 The Royal superfamily shares a common structural feature, an antiparallel 13 β-barrel-like fold formed by 4-5 β-strands, except the canonical chromodomain, 14 which harbors only three β-strands and requires the binding ligand to complete the 15 β-barrel fold by forming an extra β-strand [5]. The PWWP domain contains a 16 complete β-barrel of 5 antiparallel β-strands (β1-β5), in which a short 310 helix is 17 often inserted between β4 and β5, and a highly variable linker that may form 18 additional secondary structure elements is inserted between β2 and β3 (Fig. 1 and Fig. 19 2) [6]. A unique structural feature of the PWWP domain is the presence of a helix 20 bundle of 1-6 α-helixes following the β-barrel (Fig. 1 and Fig. 3) [6]. This helix 21 bundle region is very variable and diverse at the sequence level; therefore, only the 22 β-barrel subdomain of the PWWP domain could be reliably predicted in the protein - 3 - 1 domain databases, such as SMART (Simple Modular Architecture Research Tool) [7] 2 and the Human Protein Reference Database [8]. Nevertheless, a “V” shaped motif 3 consisting of two helixes is relatively conserved in the helix bundle subdomain (Fig. 1 4 and Fig. 3). The Pro-Trp-Trp-Pro motif is located in the beginning of the β2 strand 5 and it is packed against the helix bundle (Fig. 1 and Fig. 2), underscoring its critical 6 roles in protein folding and stability. In most cases, the PWWP domain can fold as an 7 independent functional unit; however, recent studies reveal that the PWWP domain of 8 ZMYND11 (also known as BS69) folds together with the preceding bromodomain 9 and zinc finger to form an integral functional module (Fig. 1H) [9, 10]. Interestingly, 10 the PWWP domain of HDGF can form a homodimer through a domain exchange such 11 that β1-β2 of one molecule is swapped with that of the other molecule (Fig.1J) [11]. 12 13 DNA binding ability of the PWWP domain 14 The first three-dimensional structure of a PWWP domain was determined for the 15 murine DNA methyltransferase Dnmt3b [12]. Structural analysis of this PWWP 16 domain revealed a prominent positively charged surface, suggesting a potential role in 17 DNA binding, which was confirmed in vitro [12]. Sequence analyses revealed that a 18 common feature for the PWWP domain is that it is rich in lysine and arginine residues 19 and has a theoretical isoelectric point of more than 9, suggesting a general role of the 20 PWWP domain in DNA binding, which was later confirmed for PWWP domains in 21 other proteins, such as HDGF[13, 14], MSH6[15], PSIP1 (also known as LEDGF and 22 p75)[16, 17], and ZMYND11[9, 10]. A DNA binding assay also revealed that the - 4 - 1 murine Dnmt3b-PWWP was unselective for different kinds of DNA, and did not show 2 a preference for non-CpG DNA, unmodified CpG, hemimethylated CpG, or fully 3 methylated CpG DNA [12]. Based on a selected and amplified binding assay, the 4 PWWP domain of HDGF also did not discriminate between AT and GC base pairs 5 [13]. Electro mobility shift assays also revealed that the MSH6-PWWP has similar 6 affinity toward double-stranded, double-stranded G/T mismatch, or double-stranded 7 nicked DNA, but weaker affinity toward single-stranded DNA [15]. Taken together, 8 the PWWP domain is able to bind DNA in a nonspecific manner. 9 To date, no structure of a PWWP-DNA complex is available in the protein 10 structure database (Protein Data Bank). However, using NMR chemical shift 11 perturbation experiments, several groups tried to map the DNA binding site on 12 different PWWP domains (HDGF[13], MSH6[15], and PSIP1[16, 17]). Notably, the 13 residues potentially involved in DNA binding are consistently localized on one side of 14 the protein, centering on the β1-β2 arch region and the Pro-Trp-Trp-Pro motif, which 15 also overlaps with the patch of highly positively charged surface (Fig. 4). The 16 HDGF-PWWP is also able to bind heparin, a linear polymer consisting of repeating 17 units of 1→4-linked uronic acid and glucosamine residues. Similar to DNA, heparin 18 is highly negatively charged, and it binds to a similar positively charged surface on 19 the PWWP domain of HDGF [18]. These structural clues suggest that PWWP 20 domains interact with DNA’s phosphate backbone through electrostatic interactions, 21 thus lacking sequence specifity. To specify chromatin localization, additional 22 mechanism may be required. - 5 - 1 2 Histone binding ability of the PWWP domain 3 The structural similarity of the PWWP domain to other Royal superfamily 4 members, which can recognize methylated lysine and arginine [19], prompted 5 scientists to propose the PWWP domain as a potential histone “reader” in 2005 [20]. 6 Later on, it was demonstrated that the zebrafish Brpf1-PWWP could bind histones 7 directly [21] and the fission yeast Pdp1-PWWP could recognize H4K20me 8 specifically [22, 23]. The crystal structures of a BRPF1-H3K36me3 complex fully 9 established the notion that the PWWP domain can recognize methylated histones [6, 10 24]. Since then, many other PWWP domains were reported to possess methyl-lysine 11 recognition activity; for example, DNMT3A-PWWP binds H3K36me3 [25], 12 PSIP1-PWWP binds H3K36me3 [16, 17, 26], MSH6-PWWP binds H3K36me3 [27], 13 HDGF2-PWWP binds H3K79me3 and H4K20me3 [6], and ZMYND11-PWWP binds 14 H3.3K36me3 [9]. 15 Structural analysis of the PWWP-histone complex structures identified a 16 conserved cage for methyl-lysine binding, formed by three aromatic residues (Fig. 5). 17 The third residue (W/Y) of the P-W-W-P motif and the residue (F/Y/W) immediately 18 preceding this motif are involved in forming this cage. The third aromatic cage 19 residue (F/Y/W) comes from the end of the β3 strand [6, 9, 24]. Sequence alignment 20 reveals that most PWWP domains have this conserved cage for potential 21 methylated-histone binding (Fig. 1 and Fig. 2) [6]. Nevertheless, several exceptions 22 exist. The PWWP domains of RBBP1, RBBP1L1, MBD5, and NSD1 (N-terminal) - 6 - 1 have an incomplete aromatic cage. Consistently, the RBBP1-PWWP does not show 2 any binding to methylated histone peptides [28]. In the case of BRPF1, the peptide 3 residues (G33GV35) that precede the trimethylated K36 occupy a shallow groove on 4 the protein surface which involves the long β-β-α insertion between β2 and β3. For 5 the HDGF2, this insertion is a short loop and disordered, which may be responsible 6 for its unspecific binding to both H3K79me3 and H4K20me3 (Fig. 5). Following we 7 will discuss the histone binding ability of PWWP domains in a nucleosomal context. 8 9 Nucleosome binding ability of the PWWP domain 10 The aforementioned PWWP-binding histone lysine sites H3K36, H3K79, and 11 H4K20 are all in close proximity to DNA in the nucleosomal context. That the PWWP 12 domain can bind both DNA and methylated histone suggests a synergistic binding 13 mechanism among these interactions. Similar phenomena were observed for other 14 Royal superfamily modules; for example, compared to the chromodomain alone, the 15 pre-formed complex of the MSL3[29] or RBBP1[28] chromodomain with DNA 16 displayed enhanced binding ability to H4K20me1 or H4K20me3 peptides, 17 respectively. In addition, the Tudor domain of PHF1 concomitantly interacts with both 18 the H3K36me3 and DNA of the H3K36me3-nucleosome core particle with increased 19 binding ability [30].