Structural basis of pyrimidine-pyrimidone (6–4) photoproduct recognition by UV-DDB in the nucleosome

UV-DDB, an initiation factor for the nucleotide excision repair pathway, recognizes 6–4PP lesions through a base flipping mechanism. As genomic DNA is almost entirely accommodated within nucleosomes, the flipping of the 6–4PP bases is supposed to be extremely difficult if the lesion occurs in a nucleosome, especially on the strand directly contacting the histone surface. Here we report that UV-DDB binds efficiently to nucleosomal 6–4PPs that are rotationally positioned on the solvent accessible or occluded surface. We determined the crystal structures of nucleosomes containing 6–4PPs in these rotational positions, and found that the 6–4PP DNA regions were flexibly disordered, especially in the strand exposed to the solvent. This characteristic of 6–4PP may facilitate UV-DDB binding to the damaged nucleosome. We present the first atomic-resolution pictures of the detrimental DNA cross-links of neighboring pyrimidine bases within the nucleosome, and provide the mechanistic framework for lesion recognition by UV-DDB in chromatin.


UV-DDB efficiently binds to the 6-4PPs in nucleosomes.
To study the UV-DDB binding to the nucleosomal 6-4PPs, we reconstituted the nucleosome core particles with a palindromic DNA sequence, in which two 6-4PPs were introduced at symmetric sites ( Supplementary Fig. S1). These symmetric positions of the 6-4PPs within a nucleosome facilitated the determination of the crystal structures of nucleosomal 6-4PPs (see below), although they may not occur in such close proximity in vivo. To detect the nucleosome-UV-DDB complex, nucleosomes were reconstituted with histones H2A, H3.1, H4, and the H2B T122C mutant, in which the Thr122 residue of H2B was replaced by Cys for fluorescent labeling by Alexa488. We then performed the nucleosome binding assay with the purified UV-DDB and purified Alexa488-labeled nucleosomes ( Supplementary Figs S2 and S3). To do so, the 6-4PP(outside) and 6-4PP(inside) nucleosomes, in which the 6-4PPs were rotationally positioned in a DNA strand exposed toward the solvent and contacting the histone surface, respectively, were prepared (Fig. 1a). The 6-4PP positions within the nucleosomes were confirmed by X-ray crystallography (see below).
The nucleosome binding assay was conducted in the presence of a 19-fold molar excess amount of unlabeled nucleosomes without 6-4PP lesions. Under these conditions, UV-DDB binding to the undamaged nucleosome was only weakly detected, while UV-DDB bound tightly to the 6-4PP(outside) nucleosome (Fig. 1b,c). We detected two bands with the 6-4PP(outside) nucleosome in the presence of UV-DDB (Fig. 1b). The lower and upper bands represent the nucleosomes complexed with one UV-DDB and two UV-DDBs, respectively. Unexpectedly, UV-DDB also bound to the 6-4PP(inside) nucleosome with comparable efficiency to the 6-4PP(outside) nucleosome (Fig. 1b,c).
A third species, just above the major UV-DDB-nucleosome complex band (the lower band), was specifically observed in the UV-DDB binding experiments with the 6-4PP(inside) nucleosome (Fig. 1b, lanes [8][9][10]. This suggests that UV-DDB has two binding modes to the 6-4PP(inside) nucleosome, although the underlying mechanisms are not currently understood.
Crystal structures of the 6-4PP nucleosomes. We then determined the crystal structures of the nucleosomes containing the 6-4PP(outside) or 6-4PP(inside) lesions (Fig. 2, Supplementary Table S1). For the nucleosome crystallization, we used 6-4PP(outside) and 6-4PP(inside) DNAs containing sequences identical to those used in the nucleosome binding assays ( Supplementary Fig. S1). Our design ensured that, in the crystals, one of the two 6-4PP DNA regions was exposed to the solvent without a crystal packing contact. As expected, in both the 6-4PP(outside) and 6-4PP(inside) nucleosome structures, one 6-4PP site is completely exposed to the solvent, and no direct contact is observed with other nucleosome molecules in the crystal lattice ( Supplementary Fig. S4).
Two types of solution structures, one with a sharp kink 14,15 and the other without a large distortion 16,17 , have been reported for 6-4PP-containing DNA. In the present nucleosome structures, the DNA regions around the 6-4PP sites are not kinked within the nucleosomal DNA (Fig. 2). Actually, the 6-4PP DNA model with the same kinking angle as that in a previously published 6-4PP DNA structure 14 did not fit well with the nucleosomal 6-4PP DNA structure ( Supplementary Fig. S5). However, in both the 6-4PP(outside) and 6-4PP(inside) nucleosomes, the electron densities around the 6-4PPs are quite ambiguous in both strands, indicating that the DNA region containing the 6-4PP is flexible in the nucleosomes (Fig. 2a,b). Especially, in the 6-4PP(outside) nucleosome, the affected thymine dimer nucleotides are entirely disordered (Fig. 2a). The 6-4PP thus may become flipped-out more easily, if it exists on the strand exposed to the solvent. Scrima et al. reported that UV-DDB flips 6-4PPs out of the DNA double helix, and directly interacts with both strands of the damaged site 10 . The superimposition of the present structure with the UV-DDB (DDB2)-6-4PP DNA complex illustrates that DDB2 is capable of recognizing 6-4PP within the nucleosome, if it is located on the strand exposed to the solvent ( Supplementary  Fig. S6).
In the 6-4PP(inside) nucleosome structure, the electron densities for the 6-4PP bases are interpretable, and the 6-4PP bases are not flipped out of the DNA double helix (Fig. 2b). However, we found that the electron densities for the backbone atoms of the strand complementary to the 6-4PP bases are extremely weak, as compared to the electron densities outside the lesion (Fig. 2b). These findings suggest that, in the nucleosome, the DNA backbone at the 6-4PP sites is disordered in the strand exposed to the solvent, regardless of the existence of 6-4PP bases in the flexible strand. UV-DDB preferentially binds to damaged nucleosomes containing a flexible strand. To test whether the UV-DDB binding occurs on the region containing a flexible DNA strand, we reconstituted the AP(outside) and AP(inside) nucleosomes. These represented model nucleosomes containing a flexible strand, and contained two missing consecutive thymine bases in a strand exposed to the solvent or contacting the histone surface, respectively (Fig. 3a, Supplementary Fig. S2c). Interestingly, we found that UV-DDB robustly bound to the AP(outside) nucleosome, although the bases were missing on the strand exposed to the solvent (Fig. 3b,c). The UV-DDB binding to the AP(inside) nucleosome was slightly less efficient, as compared to the binding to the AP(outside) nucleosome (Fig. 3b,c). However, we confirmed that UV-DDB bound to the naked AP(outside) and AP(inside) DNAs with exactly equal efficiency (Fig. 3d).
We next tested the possibility that UV-DDB may bind to the AP regions partially unwrapped from the histone surface. To do so, we reconstituted the nucleosomes containing the AP lesion at a single site, located at either the original position (single AP(outside) nucleosome) or 21 base-pairs away from the original position, toward the nucleosomal dyad (single AP(outside+ 21) nucleosome) (Fig. 4a, Supplementary Figs S7 and S8). The upper band observed in the nucleosome containing two symmetric AP sites was absent in the analysis of the single AP(outside) nucleosome, indicating that the upper band corresponds to the nucleosome complexed with two UV-DDB molecules (Fig. 4b). To our surprise, we found that UV-DDB binding to the single AP(outside+ 21) nucleosome is robustly enhanced, as compared to that to the single AP(outside) nucleosome ( Fig. 4b-d). The spontaneous unwrapping around the nucleosomal AP(outside+ 21) DNA region is reportedly extremely slow 18 , indicating that the UV-DDB binding occurs without unwrapping of the DNA from the histone core.
These results support the idea that the DNA flexibility at the damaged site of the nucleosome may provide the initial recognition parameter for UV-DDB. It should be noted that UV-DDB binding to the AP(outside) nucleosome was clearly weaker than that to the 6-4PP(outside) or 6-4PP(inside) nucleosome (Figs 1 and 3). These facts suggest that the existence of the flipped bases may stabilize the UV-DDB binding to damaged nucleosomes.

Discussion
In the eukaryotic NER pathway, two damage surveillance protein complexes, UV-DDB and XPC, have been identified 2,19-23 . XPC recognizes a wide range of DNA lesions, including 6-4PP 24,25 . Nucleosome formation reportedly inhibits 6-4PP binding by XPC 26 as well as excision of the lesion in vitro 27,28 , suggesting that the presence of a histone octamer may impede direct damage recognition by XPC and the following recruitment of other NER proteins. In contrast to XPC, we found that UV-DDB efficiently binds to nucleosomal 6-4PPs, even when located on a strand directly interacting with histones, probably by recognizing the DNA backbone flexibility at the damaged site.
While it is possible that the UV-DDB binding may occur at a damaged site that is partially unwrapped from the histone surface 18,29 , we consider this unlikely, because UV-DDB binding was enhanced when the damaged site was moved 21 base-pairs (+ 21) toward the nucleosomal dyad. This + 21 position is stably wrapped around the histone octamer, and rarely unwraps spontaneously 18 . It is intriguing that UV-DDB exhibited better binding to the + 21 position, as compared to the original position. According to the previous high-resolution crystal structure, the N-terminal H2A tail directly binds to the minor groove at the original position, but no histone tail interacts with the minor groove at the + 21 position 30 ( Supplementary Fig. S9). Thus, this DNA-histone tail interaction may affect the UV-DDB binding to the nucleosomal damaged site.
UV-DDB binding to the flexible DNA region functions during the initial search for photolesions within chromatin. UV-DDB reportedly forms a stable complex with the flipped photodimer, which is accommodated within the specific binding pocket of DDB2 10,11 . After the initial coarse search process with the flexible DNA backbone, UV-DDB would adopt a more stable state by interacting with the flipped damaged bases, if appropriate damage is present. The nucleosome could be an obstacle to stable complex formation, and thus probably needs to be remodeled or removed.
UV-DDB associates in vivo with the CUL4 ubiquitin ligase 31 and the CBP/p300 histone acetyltransferases 32 , suggesting that, after UV-DDB binds to the damaged site, the histone modifications catalyzed by these factors may promote the reorganization of the nucleosome structure. In response to UV irradiation, the CUL4 ubiquitin ligase reportedly ubiquitylates histones H3 and H4, which seemed to promote the dissociation of the modified histone octamer from the DNA 33 . The modification and removal of histones may then contribute to the conversion of the bound UV-DDB from the initial binding state to the stable binding state, followed by XPC recruitment.
In the XP-E patient cells lacking UV-DDB, the photolesions must be recognized in the absence of UV-DDB. Therefore, sliding and/or dissociation of histone octamers may be required for the efficient recognition of 6-4PPs by XPC, which could be facilitated by the intrinsic thermodynamic instability of the damaged nucleosomes 34 , with the aid of some chromatin remodeling factors. It is extremely intriguing to study the mechanism by which UV-DDB and XPC overcome the nucleosome barrier, after the initial UV-DDB binding to the damaged nucleosome, during the NER process.

Methods
Overexpression and purification of human histones. Human histones were expressed and purified as described previously 35,36 . The DNA fragment encoding the histone H2B T122C mutant, in which the Thr122 residue was replaced by Cys, was constructed by site-directed mutagenesis, and the H2B T122C mutant was prepared by the method described previously 35,36 . Reconstitution and purification of the H2A-H2B T122C complex, the H3.1-H4 complex, and the histone octamer were performed as described previously [35][36][37] . Freeze-dried histones were mixed at an equal molar ratio in 20 mM Tris-HCl (pH 7.5) buffer, containing 7 M guanidine hydrochloride and 20 mM 2-mercaptoethanol. Samples were dialyzed against 10 mM Tris-HCl (pH 7.5) buffer, containing 2 M NaCl, 1 mM EDTA, and 2 mM 2-mercaptoethanol. The resulting histone complexes were purified by Superdex 200 gel filtration chromatography. Preparation of DNAs. The oligonucleotides containing the 6-4PP (6-4PP ssDNA) were synthesized with the 6-4PP building block, using an Applied Biosystems 3400 DNA synthesizer 38 . Benzimidazolium triflate was used as an activator on Universal Support II PS (Glen Research), as described previously 39 . For the preparation of the 5′ -phosphorylated oligonucleotide, Chemical Phosphorylation Reagent (Glen Research) was used on the synthesizer, and at the deprotection step, the treatment with ammonia water at room temperature was prolonged to 6 h to remove the protecting group for the 5′ -phosphate. After deprotection, the products were purified by HPLC using a Waters μ Bondasphere C18 15 μ m 300A column (7.8 × 300 mm) at 60 °C, with a linear gradient of acetonitrile in 0.1 M triethylammonium acetate (pH 7.0). The eluate was concentrated in vacuo, and after desalting on an illustra NAP-10 column (GE Healthcare), the counter cation was exchanged to Na + using AG 50W-X2 resin (Bio-Rad).

Preparation of nucleosomes.
The diffraction data of the 6-4PP(inside) and 6-4PP(outside) nucleosomes were integrated and scaled with the HKL2000 program 40 . The data were processed with the CCP4 program suite 41 . The structures were solved by the molecular replacement method, using the Phaser program 42 with the human nucleosome structure (PDB ID: 3AFA) as the search model. The structures of the 6-4PP(inside) and 6-4PP(outside) nucleosomes were initially calculated at 4.0 Å and 3.5 Å resolutions, respectively. Rigid body refinement of the obtained solution was performed using the Phenix program 43 . Further structural refinement consisted of iterative rounds of energy minimization and B factor refinement using the Phenix program 43 , and model building using the COOT program 44 . The Ramachandran plot of the final 6-4PP(inside) nucleosome structure showed 100% of the residues in the most favorable and additional Scientific RepoRts | 5:16330 | DOI: 10.1038/srep16330 allowed regions, and no residues in the disallowed region. Similarly, the Ramachandran plot of the final 6-4PP(outside) nucleosome structure showed 99.4% of the residues in the most favorable and additional allowed regions, and no residues in the disallowed region. Summaries of the data collection and refinement statistics are provided in Supplementary Table S1. All structure figures were created using the PyMOL program (http://pymol.org). The atomic coordinates of the 6-4PP(inside) and 6-4PP(outside) nucleosomes have been deposited in the RCSB, with the ID codes 4YM5 and 4YM6, respectively.