Retinoblastoma-binding Protein 1 Has an Interdigitated Double Tudor Domain with DNA Binding Activity*

Background: Retinoblastoma-binding protein 1 (RBBP1), a tumor suppressor, has a Tudor domain with unknown function. Results: The Tudor domain adopts an interdigitated double Tudor structure with DNA binding activity. Conclusion: The Tudor domain of the RBBP1 family is a DNA binding module. Significance: Our research provides further structural insights into RBBP1 function and the diversity of Tudor domains. Retinoblastoma-binding protein 1 (RBBP1) is a tumor and leukemia suppressor that binds both methylated histone tails and DNA. Our previous studies indicated that RBBP1 possesses a Tudor domain, which cannot bind histone marks. In order to clarify the function of the Tudor domain, the solution structure of the RBBP1 Tudor domain was determined by NMR and is presented here. Although the proteins are unrelated, the RBBP1 Tudor domain forms an interdigitated double Tudor structure similar to the Tudor domain of JMJD2A, which is an epigenetic mark reader. This indicates the functional diversity of Tudor domains. The RBBP1 Tudor domain structure has a significant area of positively charged surface, which reveals a capability of the RBBP1 Tudor domain to bind nucleic acids. NMR titration and isothermal titration calorimetry experiments indicate that the RBBP1 Tudor domain binds both double- and single-stranded DNA with an affinity of 10–100 μm; no apparent DNA sequence specificity was detected. The DNA binding mode and key interaction residues were analyzed in detail based on a model structure of the Tudor domain-dsDNA complex, built by HADDOCK docking using the NMR data. Electrostatic interactions mediate the binding of the Tudor domain with DNA, which is consistent with NMR experiments performed at high salt concentration. The DNA-binding residues are conserved in Tudor domains of the RBBP1 protein family, resulting in conservation of the DNA-binding function in the RBBP1 Tudor domains. Our results provide further insights into the structure and function of RBBP1.

Retinoblastoma-binding protein 1 (RBBP1) 3 and its homolog RBBP1-like protein 1 (RBBP1L1) are leukemia and tumor suppressors that specifically interact with retinoblastoma protein (RB) and the mSin3A complex to suppress gene repression (1)(2)(3)(4)(5)(6)(7)(8). Peptide sequences, including KASIFLK in RBBP1 (9) and SSKKQKRSHK and IKPSLGSKK in RBBP1L1 (6,10), have been identified as breast carcinoma antigen epitopes. RBBP1 and RBBP1L1 regulate epigenetic marks, such as methylation of lysines in histones H3 and H4, which are observed in leukemia and Prader-Willi/Angelman syndromes (2,11). The epigenetic regulation function of RBBP1 was proposed to be mediated by its three Royal Family domains (2,11,12), including the chromobarrel domain, Tudor domain, and Pro-Trp-Trp-Pro (PWWP) domain, but our previous study indicated that of these three domains, only the chromobarrel domain can bind with histone tails and is responsible for the epigenetic regulation function (13). RBBP1 and RBBP1L1 are also known as ARID4A and ARID4B, respectively, because they each contain an ATrich interaction domain (ARID) and bind to DNA nonspecifically, presumably through this domain (14). The RBBP1 chromobarrel domain can also bind to DNA weakly, and this binding can increase the affinity of the chromobarrel domain for methylated histone tails (13).
Bioinformatic analysis indicates that RBBP1 contains a Tudor domain at its N terminus (13). The Tudor domain is a conserved protein structural motif, originally identified as a region of 60 amino acids in the Drosophila Tudor protein (15,16). Tudor domains have been found to regulate RNA metabolism, histone modification, DNA damage response, and a number of other processes (17)(18)(19). They are usually classed into two subsets. One subset of Tudor domain-containing proteins binds methylated lysines, whereas the other subset has methylarginine-binding capacity (17,18). Besides typical single Tudor domains, Tudor domains exist in many proteins as double Tudor domains or tandem Tudor domains. The mammalian protein p53-binding protein 1 (53BP1) contains a double Tudor domain and recognizes methylated histones mainly through its first Tudor domain, and this domain also binds DNA (20). Its homologues Crb2 and Rad9 also contain double Tudor domains with DNA binding ability (21,22). These three proteins have been proposed to play similar roles in DNA damage signaling and repair (21). DNA/RNA binding affinity is critical for the role of the Esa1 knotted Tudor domain (23). Some proteins, such as TDRD and TUDOR, have tandem Tudor domains, which cooperatively recognize combinational histone marks (24). An interdigitated double Tudor domain was found in the three Jmjc domain-containing histone demethylases (JMJD2A/2B/2C), in which two ␤ strands at the N terminus and two ␤ strands at the C terminus form a hybrid Tudor domain (HTD-1), and the middle four ␤ strands form another Tudor domain (HTD-2) (25). This interdigitated double Tudor domain specifically binds with trimethylated histone H3 Lys 4 by a conserved aromatic box in HTD-2.
Our previous studies indicated that the RBBP1 Tudor domain has low sequence homology to other known Tudor domains and cannot recognize epigenetic marks (13). In order to further understand the function of the RBBP1 Tudor domain, we have now determined its solution structure, which shows an interdigitated double Tudor fold similar to the epigenetic mark reader JMJD2A. Structural analysis reveals that the RBBP1 Tudor domain does not contain an aromatic cage to recognize an epigenetic mark. However, a positive electrostatic surface on the Tudor domain suggests that the RBBP1 Tudor domain has the capability of binding to nucleic acids. NMR and isothermal titration calorimetry (ITC) experiments demonstrated that the RBBP1 Tudor domain can bind to both dsDNA and ssDNA with an apparent K D of 10 -100 M; no sequence specificity was detected. The DNA-binding mode of the RBBP1 Tudor domain was analyzed by docking and modeling and was compared with other Tudor domains. Our data indicate diversity in the structures and functions of Tudor domains. The detailed analysis suggests that the Tudor, chromobarrel, and ARID domains of RBBP1 may act cooperatively to play roles in chromatin recognition.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The RBBP1 Tudor domain was expressed and purified as reported previously (13). 15 N-13 C-Labeled RBBP1 Tudor domain was prepared using the same procedures, except that cells were grown in M9 minimal medium containing 15  H-13 C NOESY-HSQC spectra with mixing times of 120 ms were collected to generate distance restraints. All data were processed with NMRPipe (27) and analyzed with NMRViewJ (28). Proton chemical shifts were referenced to the internal sodium 2,2-dimethylsilapentane-5-sulfonate, and 15 N and 13 C chemical shifts were referenced indirectly (29).
Structure Calculations-The RBBP1 Tudor domain structures were initially calculated with the program CYANA (31) and then refined using CNS (32) with manual assignments as well as semiautomated NOE assignments by SANE (33). Backbone dihedral angle restraints obtained using CSI (34) and TALOSϩ (35) as well as hydrogen bond restraints according to the regular secondary structure patterns were also incorporated into the structural calculation. From 100 CNS-calculated structures, the 50 lowest energy conformers of the Tudor domain were selected for further water refinement using CNS and RECOORDScript (36). The resulting 20 energy-minimized conformers were used to represent the solution structure of the Tudor domain. The quality of the determined structures (Table  1) was analyzed using PROCHECK-NMR (37), MolMol (38), and WHAT_CHECK (39). Structural figures were created with MolMol (38) and PyMOL (40).
DNA Titration-DNAs used in the titration experiments were AT-rich double-stranded DNA (dsAT), AT-rich singlestranded DNA (ssAT), GC-rich double-stranded DNA (dsGC), GC-rich single-stranded DNA (ssGC), double-stranded DNA1, single-stranded DNA1N, double-stranded DNA2, and singlestranded DNA2C (41) ( Table 2). Double-stranded DNA was made by annealing equimolar amounts of the two synthesized single-stranded DNA (1:1 molar ratio), which were dissolved in a buffer containing 50 mM Tris-HCl (pH 7.6) and 50 mM NaCl, heated to 94°C for 3 min, and then allowed to cool slowly to room temperature. The ssDNAs and dsDNAs were further purified by gel filtration and then concentrated. The stock solution of dsDNA and ssDNA contained 5 mM DNA in the same buffer as that used for the RBBP1 Tudor domain sample.
Interaction of the RBBP1 Tudor domain with DNA was monitored by recording a series of two-dimensional 1 H-15 N HSQC spectra of proteins at each DNA titration point. The observed chemical shift perturbations (CSPs) of the protein resonances were calculated using the equation, where ␦ HN and ␦ N are the changes of 1 H N and 15 N chemical shifts, respectively. The equilibrium dissociation constants (K D ) of protein with DNA were estimated by fitting the CSPs to the equation, where CSP max is the CSP value at the theoretical saturated condition obtained from the titration curve fitting; r is the molar ratio of DNA to protein; C pro is the concentration of initial protein solution; C lig is the stock concentration of DNA; and n is the number of equivalent and independent binding sites on the DNA. Fitting of the titration curves suggested that the 12-bp DNAs can accommodate simultaneously at least two RBBP1 Tudor domain molecules. When n was floating, a value of 3-5 was obtained for 12-bp DNAs and 5-9 for 18-bp DNAs. The physical meaning of the obtained value of n is complicated because it could also account for any uncertainty in DNA and protein concentrations that were fixed in fitting. To be consistent, n was fixed as 3 in the fitting of 12-bp DNA titration curves and as 5 in the fitting of 18-bp DNA titration curves. HADDOCK Docking-Modeling of the RBBP1 Tudor domain-DNA complex was performed on the HADDOCK Web server (42). The starting structural coordinate files for HADDOCK were generated from the 20 solution structures of the RBBP1 Tudor domain and a 12-bp B form DNA duplex built using 3D-dart (43). For HADDOCK calculations, active residues for the RBBP1 Tudor domain were defined as those having weighted CSPs larger than 0.5 ppm, which are residues at sequence positions 8 -115 of the RBBP1 Tudor domain. Residues 4 -7 and 116 -121 of the RBBP1 Tudor domain, which contain several negatively charged residues, are flexible in the structure and showed smaller CSP values during NMR titration. They were deleted in the docking because they are not important for DNA binding and their flexibility could lead to steric hindrance during the docking process. Passive residues were automatically defined around the active residues by HADDOCK. The active residues were then optimized according to the initial docking result, and the final active residues include Tyr 9 -Leu 10 , Trp 88 , Arg 99 -Thr 104 , Lys 109 -Phe 114 , Gln 52 , and Ala 18 -Arg 21 . All of the bases of the dsDNA2 5Ј sequence were considered active in the docking. A total of 1000 initial complex structures were generated for rigid body docking, and the 200 lowest energy structures were further refined in explicit water after semiflexible simulated annealing. A cluster analysis was performed on the finally docked structures corresponding to the 200 best solutions with lowest intermolecular energies based on a 7.5-Å root mean square deviation (RMSD) cut-off criterion. The clusters were ranked based on the averaged HADDOCK score of their top 10 structures.

RESULTS
The RBBP1 Tudor Domain Is an Interdigitated Double Tudor Domain-Secondary structure and disorder predictions suggest that residues 1-152 of RBBP1 might form a structural domain, and residues 58 -103 show homology to Tudor domains (13). In order to study the structure of this RBBP1 region, a number of different length protein constructs were obtained. A construct containing residues 4 -121 was used for structure determination because it is most stable and provided the best NMR spectra. More than 90% backbone and side chain resonances of residues 4 -121 of RBBP1 (we will refer to residues 4 -121 of RBBP1 as the RBBP1 Tudor domain hereafter) were assigned, and the structure of the RBBP1 Tudor domain was determined at pH 7.8 ( Fig. 1, A-C). Interestingly, the structure of the RBBP1 Tudor domain is similar to the Jmjc domaincontaining histone demethylase (JMJD2A) (25), exhibiting an interdigitated double Tudor domain (Fig. 1C). Such interdigitated double Tudor domains are well conserved among the whole RBBP1 protein family throughout the animal kingdom ( Fig. 2A). The overall structure comprises eight ␤-strands (␤1, 13-20; ␤2, 23-33; ␤3, 36 -42; ␤4, 47-52; ␤1Ј, 63-70; ␤2Ј, 73-85; ␤3Ј, 88 -93; ␤4Ј, 96 -100), which form two hybrid Tudor domains: HTD-1 and HTD-2 according to the nomenclature in JMJD2A (25). Each HTD adopts a ␤-barrel fold as for typical Tudor domains. The RMSD between the two HTDs is 0.91 Å for the backbone of secondary structure regions (Fig.   1D). However, the sequence topologies of the two HTDs of the RBBP1 Tudor domain are different. The HTD-1 (residues 8 -34 and 87-114) is a swapped Tudor domain consisting of ␤1 and ␤2 strands from the N-terminal region of the protein and ␤3Ј and ␤4Ј strands from the C-terminal region. The HTD-2 (residues 35-86) is a canonical Tudor domain consisting of ␤3, ␤4, ␤1Ј, and ␤2Ј ␤-strands from the N terminus to the C terminus.
The structure-based sequence alignment indicates 31% identity for the residue 58 -103 fragment of RBBP1 to JMJD2A, whereas it shows only 23% identity for the residue 4 -121 fragment (Fig. 2B). However, the overall structure of the RBBP1 Tudor domain is highly similar to the JMJD2A Tudor domain (Protein Data Bank code 2GF7) (Fig. 4A). The overall RMSD is about 5.2 Å when the two structures are aligned. The structural alignment between each HTD of the two proteins (Fig. 4B) generated an RMSD value of 1.0 Å for HTD-1 and 1.3 Å for HTD-2. Basically, the two HTDs are in a similar orientation to each other in the RBBP1 Tudor domain as in the JMJD2A Tudor domain. On the other hand, the structural comparison reveals some conformational differences between the two proteins. In the crystal structure of the JMJD2A Tudor domain, the backbone of the linker residues between the two HTDs adopts a ␤-strand conformation, making ␤2-␤3 and ␤2Ј-␤3Ј the two long ␤-strands. Unlike the JMJD2A Tudor domain, the linker residues between the two HTDs of the RBBP1 Tudor domain are in a loop conformation as identified by TALOSϩ (Table 3). Moreover, the RBBP1 Tudor domain lacks the conserved aromatic cage as observed in the HTD-2 of JMJD2A. The aromatic cage residues, Phe 932 , Trp 967 , and Tyr 973 , in the HTD-2 of JMJD2A are replaced by Leu 42 , Thr 70 , and Gln 76 in the HTD-2 of the RBBP1 Tudor domain (Fig. 4, B and C). Such a significant difference must lead to totally different functions between the two interdigitated double Tudor domains. This may explain why JMJD2A specifically binds trimethylated histone H3 Lys 4 , whereas the RBBP1 Tudor domain does not bind any methylated ligand, as reported previously (13). In addition, the loop L 41Ј of the RBBP1 Tudor domain is 5 residues shorter than that of JMJD2A.
The RBBP1 Tudor Domain Can Bind ssDNA and dsDNA-In exploring the function of the RBBP1 Tudor domain, we noticed that the RBBP1 Tudor domain has many more basic residues than the JMJD2A Tudor domain. The JMJD2A Tudor domain is an acidic protein, with pI values of 6.8 for HTD-1 and 3.9 for HTD-2, whereas the RBBP1 HTD-1 (residues 8 -34 and 87-114) and HTD-2 (residues 35-86) are both positively charged, with pI values of 9.4 and 8.4, respectively, and the RBBP1 Tudor domain has a significant area of positively charged surface (Fig. 4C). It has been reported that other Tudor domains, such as 53BP1 TT and the Esa1 knotted Tudor domain, bind other DNA sequences besides methylated lysines (20,23). The 53BP1 TT Tudor domain 1, the key domain for DNA binding, has a pI of 9.8, whereas the pI of the Tudor domain 2 is 4.4, and the overall pI of 53BP1 TT is 6.8. This led us to suspect that the RBBP1 Tudor domain may be involved in interaction with nucleic acids.
We checked the possibility of interactions between the RBBP1 Tudor domain and various sequences of dsDNA and ssDNA by monitoring the two-dimensional 1 H- 15 Table 2). Comparison of these K D values (Table 2) and chemical shift changes at a protein/DNA ratio of 1:1 (Fig. 6, A  and B) obtained for the different DNA sequences suggests that double-stranded DNA (e.g. dsAT and dsGC) has similar binding affinity for the RBBP1 Tudor domain as the equivalent single-stranded sequences (e.g. ssAT and ssGC), and AT-rich sequences showed similar binding affinity as GC-rich sequences. The NMR titration result was confirmed by ITC measurements, which gave binding affinities of a similar order of magnitude, namely ϳ150 and ϳ70 M for ssAT and dsAT, respectively (Fig. 7, A and B). This indicates that the RBBP1 Tudor domain is able to bind both single-stranded DNA and double-stranded DNA and both AT-and GC-rich sequences. Unlike the simple AT-rich or GC-rich sequences, binding of dsDNA2 and dsDNA1 to the RBBP1 Tudor domain consistently showed stronger binding than the equivalent singlestranded sequences ( Table 2 and Fig. 6, C and D). This suggests that the RBBP1 Tudor domain may have some weak binding specificity toward a double-stranded GTCAAAGGT sequence because both dsDNA1 and dsDNA2 contain this sequence. However, the two-dimensional HSQC spectrum of the RBBP1 Tudor domain in the presence of 350 mM NaCl showed no chemical shift perturbations compared with that of free RBBP1 Tudor domain (data not shown), implying that high salt concentration can abolish the DNA binding and thus the binding of DNA with RBBP1 Tudor domain is mainly mediated through electrostatic interactions.
DNA Binding Site of the RBBP1 Tudor Domain-The similar chemical shift perturbation patterns caused by binding of different DNA sequences to the RBBP1 Tudor domain (Fig. 5C) indicate that the different DNA sequences interact with similar   (13), little change in signal intensity was observed for most NH signals of the RBBP1 Tudor domain when binding with DNA. In fact, the NH signals of residues Gly 22 , Cys 107 , and Gly 110 are enhanced during DNA titration, suggesting that DNA binding stabilizes the C-terminal residues and may further stabilize the loop L 12 (residues Arg 21 and Gly 22 ) between the ␤1 and ␤2 strands. Mapping the CSP results to the RBBP1 Tudor domain structure reveals that the DNA binds to both the N and C termini and to the ␤-barrel of RBBP1 HTD-1 (Fig. 5D). Nevertheless, the HTD-2 may also be involved in DNA binding because a few residues in HTD-2, such as Gln 52 , Lys 82 , and Thr 84 , also show moderate chemical shift changes as a result of DNA interactions. Fig. 5E shows that more than half of the residues in the binding site form a positive surface that contains all 10 positively charged residues (Lys 19 , Arg 21 , Lys 37 , Lys 39 , Lys 82 , Arg 99 , Arg 102 , Arg 103 , Lys 109 , and Arg 112 ). Sequence alignment reveals that most of these positively charged residues are quite conserved among the RBBP1 family ( Fig. 2A), suggesting a conserved DNA-binding function of Tudor domains in the RBBP1 family. Therefore, the electrostatic interaction between these residues and DNA is likely to be important for DNA binding, which is consistent with the fact that high salt concentrations can dramatically reduce the DNA binding ability of the RBBP1 Tudor domain.

Structural Basis of the Unique Double Interdigitated Tudor
Domain in RBBP1 and JMJD2A Family Proteins-Unlike the tandem Tudor domains that exist in many proteins, a double interdigitated Tudor domain is only found in RBBP1 and JMJD2 family proteins. Sequence analysis suggests that the RBBP1 family is represented throughout the animal kingdom, and its Tudor domain is highly conserved (Fig. 2). This raises the interesting question of why the RBBP1 and JMJD Tudor domains can form an interdigitated Tudor domain, but not a tandem double Tudor domain, and what are the structural features that define formation of this unusual structure. A possible explanation is as follows. In the structure of the double interdigitated Tudor domains, the secondary structural elements are organized in the sequence ␤1-␤2-␤3-␤4-␤1Ј-␤2Ј-␤3Ј-␤4Ј from the N to C terminus. However, according to the structurebased sequence alignments of HTD-1/HTD-2 and typical tandem Tudor domains (Fig. 8, A-C), the secondary structural elements are arranged as ␤3-␤4-␤1Ј-␤2Ј in HTD-2 and ␤3Ј-␤4Ј-␤1-␤2 in HTD-1. In other words, ␤1-␤2 and ␤1Ј-␤2Ј correspond to the C-terminal parts of a typical Tudor domain (TD C ), whereas ␤3-␤4 and ␤3Ј-␤4Ј correspond to the N-terminal parts of a typical Tudor domain (TD N ). This means that the HTD-2 of the RBBP1 Tudor domain should be a typical Tudor domain because ␤3-␤4 are at the N terminus of ␤1Ј-␤2Ј, whereas the HTD-1 is an N-C swapped Tudor domain because the N-terminal ␤1-␤2 in the RBBP1 Tudor domain sequence correspond to the C-terminal parts of a typical Tudor domain (TD C ) (Fig. 8B). Consequently, the sequence order in a double interdigitated Tudor domain is TD C -TD N -TD C -TD N . Because a tandem Tudor domain requires the order TD N -TD C -TD N -TD C , it is not possible for a protein having the sequence organization TD C -TD N -TD C -TD N to form a typical tandem Tudor domain.
On the other hand, the potential to form a tandem N-C swapped Tudor domain cannot be excluded in such an explanation. In other words, ␤1-␤2-␤3-␤4 might first form an N-C swapped Tudor domain, and ␤1Ј-␤2Ј-␤3Ј-␤4Ј might form the second N-C swapped Tudor domain. Therefore, there must be some factors that influence the RBBP1 Tudor domain to form a double interdigitated Tudor domain instead of a tandem N-C swapped Tudor domain. Generally, the driving forces for protein folding include hydrophobic interactions, hydrogen bonds, electrostatic interactions, and Van der Waals interactions. From the viewpoint of hydrophobic interactions, two hydrophobic cores are formed in the two HTDs of the RBBP1 Tudor domain. The hydrophobic core of HTD-1 involves residues Tyr 89 , Val 91 , Phe 93 , Leu 101 , Leu 106 , and Leu 108 in ␤3Ј and ␤4Ј and residues Val 12 , Val 16 80 , and Leu 83 in ␤1Ј and ␤2Ј form a hydrophobic core in HTD-2. However, structure alignment indicates that the relative positions of these hydrophobic residues in the two domains are well conserved (Fig. 8, A and D). Apparently, hydrophobic interactions are not the driving force for the interdigitated fold of the RBBP1 Tudor domain.
Good E a Classes are as follows: Good, the prediction is faithful; Warn, the prediction is not faithful; Dyn, dihedral angle of residue is dynamic according to TALOSϩ prediction; None, no chemical shifts available for prediction. Amino acids in boldface type and underlined refer to linker residues. b ss-Classes are as follows: L, loop conformation; E, ␤-stranded conformation.

JOURNAL OF BIOLOGICAL CHEMISTRY 4889
A detailed structural analysis reveals that salt bridges are formed in HTD-1 by Lys 30 -Glu 98 between ␤2 and ␤4Ј, and in HTD-2 by Lys 43 -Glu 77 between ␤3 and ␤2Ј. However, residues Lys 30 and Glu 98 in HTD-1 correspond to Ser 81 and Thr 48 in HTD-2, and Lys 43 and Glu 77 in HTD-2 correspond to Asp 94 and Glu 26 in HTD-1 (Fig. 8E). Thus, the cross-strand ion pair interactions in the two HTDs prevent the RBBP1 Tudor domain from forming two N-C swapped Tudor domains. Similarly, there are two cross-strand salt bridges in the JMJD2A Tudor domain, namely Lys 910 -Asp 992 in HTD-1 and Asp 933 -Lys 976 in HTD-2. These four residues, in turn, correspond to residues Thr 968 and Asp 934 in HTD-2 and Glu 991 and Glu 918 in HTD-1. Construction of a tandem N-C swapped Tudor domain in JMJD2A is also prevented by these two salt bridges. This implies that these electrostatic interactions may be the key factor for the interdigitated folding of the RBBP1 and the JMJD2A Tudor domains.

Structural Model of the RBBP1 Tudor Domain-DNA
Complex-A structural model of the interaction of the RBBP1 Tudor domain with dsDNA2 ( Table 2) was generated using the modeling program HADDOCK (42). The chemical shift perturbations obtained from the NMR titration experiments were used to define residues involved in the interaction. The resulting model of the complex can be seen in Fig. 9. Analysis of the final 200 HADDOCK models for the RBBP1 Tudor domain and dsDNA2 complex resulted in two top clusters containing 23 and 11 structures of the complex, respectively, with cluster 1 showing a Z-score of Ϫ1.9 (HADDOCK score Ϫ124.0) and cluster 2 showing a Z-score of Ϫ1.8 (HADDOCK score Ϫ122.7). The HADDOCK score is an arbitrary score that is intended to group solutions into clusters, whereas the Z-score is a standard score reflecting the statistical significance of the HADDOCK scores. A lower Z-score and a larger gap with the next cluster indicates a better cluster in docking. The Z-scores   In both clusters, the DNA duplex mainly binds to the RBBP1 HTD-1 at similar sites containing ␤1, ␤4Ј, loop L 12 , the C-terminal part of HTD-1, and the linker between HTD-1 and HTD-2 (Fig. 9). A few residues of HTD-2 are also involved in the interaction with DNA. A positively charged surface is seen on one side of the RBBP1 Tudor domain surrounding the DNA binding sites (Fig. 9, B and C), which is likely to contribute to the direct contact with DNA. However, residues Asp 94 -Glu 98 form a negatively charged surface on the opposite side of the RBBP1 Tudor domain. Presumably, the electrostatic potential mediates the DNA binding, driving the DNA from the negatively charged surface to the positively charged surface. Consequently, electrostatic interactions make a dominant contribution to the binding of DNA to the RBBP1 Tudor domain. Inspection of the model structures of the complex reveals that residues Lys 19 -Arg 21 , Lys 37 , Arg 102 -Ser 105 , and Arg 112 -His 113 of the RBBP1 Tudor domain make direct contact with the DNA duplex. Of these, the C-terminal part (Lys 19 ) of ␤1 and the N terminus (Arg 21 ) of L 12 extend into the major groove of the DNA duplex, whereas the side chain of Arg 102 of the C terminus extends into the minor groove. In addition, residue Lys 37 from ␤3 also forms electrostatic interactions with DNA. Indeed, the positively charged residues Lys 19 -Arg 21 , Lys 37 , Arg 102 , Arg 103 , and His 113 are basically conserved throughout the animal kingdom, as indicated by the sequence alignment ( Fig. 2A).
Comparing Nucleic Acid Binding Properties of the RBBP1 Tudor Domain with Other Tudor Domains-In general, Tudor domains have been identified as an important structural module for recognizing methylated histones (17,18). A few Tudor domains, some of which lack the ability to bind methylated histones, have been shown to be involved in DNA/RNA binding, such as the 53BP1 tandem Tudor domain (53BP1 TT ) (20), Esa1 knotted Tudor domain (23), and Rad9 tandem Tudor domain (Rad9 TT ) (21). The RBBP1 Tudor domain can bind DNA but lacks methylated ligand binding ability, so it is interesting to compare its nucleic acid binding properties with other Tudor domains.
A common feature for these Tudor domains is that they all have relatively weak binding affinity for DNA. Esa1  Residues that can be aligned in the structure alignment are capitalized, and residues that cannot be aligned are in lowercase type. Residues within the hydrophobic core are in boldface type and colored red. The N-and C-terminal residues, Ala 4 and Asp 121 , are numbered. C, structure alignment of RBBP1 HTD-1 (dark gray) and the 53BP1 Tudor domain 1 (green). D and E, structure alignment of the hydrophobic core (D) and salt bridges (E) of RBBP1 HTD-1 (magenta) and HTD-2 (yellow).
weak binding affinities can provide more precise regulation (44).
DNA binding mechanisms of these Tudor domains are diverse. The DNA binding sites in the 53BP1 TT and Esa1 knotted Tudor domains involve L 34 and ␤4, corresponding to L 3Ј4Ј and ␤4Ј in the HTD-1 of the RBBP1 Tudor domain, which is at a different site (rotation by 120°) of the protein compared with the DNA binding region in the RBBP1 Tudor domain (Fig. 10). Loop L 12 of 53BP1 TT is also involved in DNA binding, and the DNA binding interface is a negatively charged surface (Fig.  10A), whereas the DNA binding surface of Esa1 knotted Tudor domain is composed of negative and positive patches (Fig. 10B).
The nucleic acid binding of 53BP1 TT and Esa1 knotted Tudor domain depends on multiple types of interactions, including hydrophobic, cation-, ring stacking, and hydrogen bond interactions (20,23). Rad9 TT has a similar tandem Tudor domain structure as 53BP1 TT , but electrostatic interactions around the loop L 3Ј4Ј dominate the binding of DNA to Rad9 TT (21). As shown in Fig. 10C, the DNA binding of RBBP1 Tudor domain is mainly dominated by electrostatic interactions around ␤1, ␤4Ј, and loop L 12 . The relationship of sequence, structure, and function of the Tudor domains of the RBBP1 protein family indicates greater diversity in the structures and functions of Tudor domains than had previously been appreciated.
DNA-binding Function of the RBBP1 Tudor Domain May Act Together with the ARID and Chromobarrel Domains in Biolog- FIGURE 9. Structural model of the RBBP1 Tudor domain in the complex with dsDNA2. The model was calculated using the HADDOCK Web server (42). A, the best four structures of the top 1 cluster. B, contacts at the interface between the RBBP1 Tudor domain and the DNA duplex in the best structure of cluster 1. The RBBP1 Tudor domain is shown in a ribbon representation, and the residues potentially involved in the binding are shown as sticks. The dsDNA is shown in yellow. C, the electrostatic surface of the RBBP1 Tudor domain in the model of the complex. The charged residues involved in binding are labeled.

RBBP1 Has a DNA-binding Interdigitated Double Tudor Domain
FEBRUARY 21, 2014 • VOLUME 289 • NUMBER 8 ical Processes-Both RBBP1 and RBBP1L1 contain five conserved domains: a Tudor domain, an ARID, a chromobarrel domain, a PWWP domain, and a C-terminal R2 domain. It has been demonstrated that the RBBP1 chromobarrel domain recognizes methylated lysines of histone peptides with high affinity through its aromatic cage, and DNA also binds to the regions near the aromatic cage. The binding affinity of the chromobarrel domain for methylated ligands is enhanced by the binding of DNA (13). Moreover, the RBBP1 ARID can bind DNA nonspecifically with low affinity (14), which may also affect the binding affinity of the chromobarrel domain with methylated ligands. The finding that the RBBP1 Tudor domain recognizes DNA rather than methylated histones sheds light on the combinational functional binding mode of the different RBBP1 domains. This raises the possibility that the DNA-binding function of the Tudor domain may also facilitate the binding of RBBP1 to methylated histones. Several studies have demonstrated that RBBP1 interacts with RBBP1L1, and the two proteins usually function in the same protein complex, such as mSin3A (1,4,45) and an epigenetic regulation complex together with RB protein (11). Such a cooperational pattern may favor precise regulation of biological processes mediated by RBBP1 and RBBP1L1 (44). Thus, the Tudor, ARID, and chromobarrel domains of RBBP1 and RBBP1L1 may cooperate within the related protein complexes to recognize specific sites in chromatin and provide sufficient binding affinity for function.
Biological Implications-Our finding that an interdigitated double Tudor domain can bind DNA indicates a higher degree of functional diversity among Tudor domains than previously anticipated. Besides important information about the relationship between structure and function for the Tudor domain, our findings also have biological implications regarding the various functions of RBBP1. RBBP1 and RBBP1L1 are localized in the nucleus (5,46,47), and epigenetic roles in chromatin or direct DNA-binding as a transcription factor have been proposed or demonstrated as functions of RBBP1 in the cell cycle, osteoblast differentiation, male fertility, cancer, and other diseases (2, 5, 11, 48 -50). For example, a recent study indicated that RBBP1L1 is a breast cancer progression modifier gene, and it was proposed that the suppression role was accomplished by recruiting the mSin3A complex for epigenetic modification on specific gene sites through the DNA-binding activity of RBBP1L1 because the mSin3A complex itself does not possess a DNA binding activity (49). A more recent paper indicated that RBBP1 and RBBP1L1 control male fertility as transcriptional coactivators for the androgen receptor and RB (48). However, the molecular and structural mechanisms are not yet clear because of the lack of detailed research on the DNA-binding activity of RBBP1, beyond a preliminary study, which indicated that the RBBP1 ARID can weakly and nonspecifically bind to DNA (14). Our finding that the RBBP1 Tudor domain has DNA-binding activity provides new structural and functional insight into the binding activity of RBBP1. This and our previous study (13) indicate that, in addition to the ARID, both the chromobarrel domain and the Tudor domain participate in DNA binding, and thus the DNA-binding affinity of full-length RBBP1 may be much stronger than the affinity of any one of the individual domains, which could then provide sufficient affinity to bind a specific gene site directly or to recruit the mSin3A complex to the site to regulate gene expression. Considering the important functions of RBBP1 in various biological processes, the DNA site bound by RBBP1 in vivo must be identified in order to elucidate the underlying molecular mechanism. Our structural and DNA-binding studies of the RBBP1 Tudor domain presented here will provide a basis for further studies to reveal the site recognition mechanism.