Structure of a Novel DNA-binding Domain of Helicase-like Transcription Factor ( HLTF ) and Its Functional Implication in DNA Damage Tolerance *

Asami Hishiki (菱木 麻美), Kodai Hara (原 幸大), Yuzu Ikegaya (池ヶ谷 柚子), Hideshi Yokoyama (横山 英志), Toshiyuki Shimizu (清水 敏之) , Mamoru Sato (佐藤 衛), and Hiroshi Hashimoto (橋本 博)** From the School of Pharmaceutical Sciences, University of Shizuoka, 52-1 Yada, Suruga-ku, Shizuoka, Shizuoka 422-8002, Japan, the Graduate School of Medical Life Sciences, Yokohama City University, 1-7-29 Suehiro-cho, Tsurumi-ku, Yokohama, Kanagawa 230-0045, Japan, the Sanford-Burnham Medical Research Institute, La Jolla, California 92037, the Graduate School of Pharmaceutical Sciences, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan, and the **Institute for Protein Research, Osaka University, 3-2 Yamadaoka, Suita, Osaka 565-0871, Japan

DNA damage occurs both naturally and artificially. Damaged template DNA blocks accurate and processive DNA synthesis by replicative DNA polymerase, DNA polymerase ␦ or DNA polymerase ⑀. Replication arrest leads to collapse of the replication fork and results in genomic instability causing various human diseases, including cancer. To avoid replication arrest and to restart DNA synthesis at the damaged site, cells initiate DNA damage tolerance, also known as post-replication repair (1). DNA damage tolerance can be divided into two pathways that synthesize DNA, translesion and template-switching DNA syntheses. Translesion DNA synthesis (TLS) 2 is transient DNA synthesis using a damaged template by error-prone DNA polymerases specialized for DNA damage (TLS polymerases). Template-switching (TS) DNA synthesis, also known as the damage avoidance pathway, in which one newly synthesized strand is utilized as an undamaged template for replication by replicative polymerases, is an error-free process.
DNA damage tolerance is initiated by the ubiquitination of the Lys-164 of PCNA, a scaffold protein that stimulates DNA synthesis by DNA polymerases. Mono-ubiquitination of PCNA at Lys-164 by RAD6-RAD18, an E2-E3 complex, activates the TLS pathway. Poly-ubiquitination of PCNA at Lys-164 promotes the TS pathway. The ubiquitin chain is formed by Lys-63 linkage of ubiquitin. The Lys-63-linked ubiquitin chains are involved in various cellular signaling processes including DNA repair. The Lys-63linked poly-ubiquitination of PCNA is performed by MMS2-UBC13 (E2) and RAD5 (E3). RAD5 (REV2) was initially identified in yeast as a repair protein with DNA-dependent ATPase activity (2,3). Yeast RAD5 has been shown to be an E3 ubiquitin ligase responsible for Lys-63-linked poly-ubiquitination of PCNA (4).
Mammalian cells contain two RAD5 homologs, HLTF (helicaselike transcription factor) and SHPRH (SNF2 histone-linker PHD finger RING finger helicase), both of which are involved in Lys-63linked poly-ubiquitination of PCNA at Lys-164 and in the regulation of DNA damage tolerance (5)(6)(7)(8)(9). Loss of either HLTF or SHPRH increases the frequency of chromosome abnormalities under DNA damage, indicating that both proteins are important for the maintenance of genomic stability under such conditions (6,7). Recently, it has been shown in vitro that HLTF forms a thiollinked ubiquitin chain on UBC13 and then transfers the chain to RAD6, after which the chain is transferred to un-modified Lys-164 of PCNA by RAD18 (10).
HLTF and SHPRH possess a SWI/SNF helicase domain that is divided into N-and C-terminal parts by an insertion of RING domain involved in the poly-ubiquitination of PCNA (see Fig.  1A). For several years, helicase activities of HLTF have been characterized in vitro. Human HLTF shows double-stranded DNA translocase activity with 3Ј-5Ј polarity, thereby facilitating regression of the replication fork (11). Furthermore, a recent study has shown that HLTF facilitates the strand invasion and formation of a D-loop structure (12). Although these biochemical studies have revealed the functional significance of HLTF in the TS pathway, the structural basis of this HLTF function remains unclear. Interestingly, sequence analysis has suggested that HLTF contains a novel domain referred to as the HIRAN (HIP116 and RAD5 N-terminal) domain that is hypothesized to be involved in DNA binding (see Fig. 1A) (13). The HIRAN domain has been identified in various proteins involved in DNA metabolism; however, the structural basis of, and functional evidence for, the HIRAN domain in DNA binding has remained unclear until now.
Here we show the crystal structures of the HIRAN domain of human HLTF and its DNA complex. This study reveals the function of the novel DNA-binding domain as a sensor to the 3Ј-end of the primer strand and provides a significant clue to clarifying the molecular mechanism behind TS in DNA damage tolerance pathways.
Crystallographic study-An ssDNA mt04-13 (5Ј-ACCGC-CGGGTGCC-3Ј) for crystallization of the DNA complex was commercially synthesized (Invitrogen) (see Table 2). Prior to crystallization, equal volumes of HLTF 58 -174 (0.7 mM) and ssDNA (0.8 mM) were mixed on ice. Suitable crystals for xray diffraction study were obtained by hanging drop vapor diffusion using a reservoir solution composed of 0.1 M Bis-Tris propane, pH 6.5, 0.2 M sodium acetate, and 20% PEG3350. For phase determination, selenomethionyl protein was overexpressed in methionine auxotrophic E. coli strain B834(DE3)pLysS with a minimum medium including 50 g/ml ampicillin, 34 g/ml chloramphenicol, and 25 g/ml seleno-Lmethionine. Selenomethionyl protein was purified and crystallized by a procedure similar to that of non-labeled protein.
Crystals were transferred to a cryo-buffer composed of a reservoir solution and ethylene glycol in a ratio of 4:1. Then, the crystals were frozen in N 2 gas stream at 100 K. X-ray diffraction data were collected at the Photon Factory beamline NE-3A (Tsukuba, Japan). Diffraction data were processed using the program HKL2000 (15). Phase calculation was performed using the program SOLVE (16), and the initial phase was improved using the program RESOLVE (17). Model building and fitting were performed using the program COOT (18). The structure was refined using the programs CNS (19) and REFMAC (20).
Crystals of His-HLTF 58 -174 in DNA-free form were obtained by sitting drop vapor diffusion method using a reservoir solution containing 0.2 M lithium acetate and 20% PEG3350. Prior to the diffraction study, crystals were transferred in buffer composed of 0.2 M lithium acetate, 20% PEG3350, and 20% ethylene glycol for cryo-protection, and then the crystals were frozen in liquid N 2 . X-ray diffraction data were collected at the SPring-8 beamline BL-26B2 (Sayo, Japan). Diffraction data were processed using the programs XDS (21) and SCALA (22). Crystal structure of the DNA-free form of His-HLTF 58 -174 was determined by molecular replacement methods, followed by structure refinement using the programs PHENIX (23) and REFMAC (20). Data collection, phasing, and refinement statis-tics are given in Table 1. Structural superimposition was performed using the program LSQKAB (24). The evolutionary conservation scores of amino acid positions in the HIRAN domains are calculated by the program CONSURF (25).
DNA Binding Assay-To perform an EMSA, an alanine mutation was introduced into HLTF 58 -174 by the QuikChange protocol, a PCR-based site-directed mutagenesis using mismatched primers (Stratagene). The GST-fused HLTF 58 -174 mutant was overexpressed and purified with glutathione-Sepharose 4B resin. The GST tag of the fusion protein was cleaved by HRV3C protease, followed by gel filtration using HiLoad Superdex 75. Purified mutant proteins were concentrated, frozen with liquid N 2 , and stored at Ϫ80°C until use.

Results and Discussion
Overall structure of the DNA Complex of Human HLTF HIRAN Domain-The crystal structures of HLTF 58 -174 in complex with DNA and its DNA-free form were determined at 1.38 and 2.40 Å resolution, respectively (Fig. 1, B and D). The results showed that the HIRAN domain adopts a ␤-barrel structure composed of six ␤-strands flanked with two ␣-helices (Fig.  1, B and C), resulting in an OB-fold structure that contains oligonucleotide-binding proteins such as RPA or telomerebinding protein (26). Typical OB-fold proteins possess a concave surface to accommodate a ligand. In general, ssDNA-binding proteins such as RPA or telomere-binding protein are composed of multiple OB-fold domains to extensively accommodate ssDNA. For instance, RPA70 (residues 181-422) adopts a tandem OB-fold structure, and an ssDNA traverses the successive concave areas of the tandem OB-folds (27). HLTF 58 -174 also has a concave surface that interacts with the two bases of ssDNA (Fig. 2B), and the concave surface is evolutionally conserved (Fig. 2C). In contrast to RPA and other ssDNA-binding proteins, the 3Ј-end of DNA strand is specifically bound to a pocket composed of ␤3, ␤4, and ␣1 of HLTF 58 -174 (Figs. 1B  and 2, B and D). Interestingly, ssDNA used in crystallization forms a duplex DNA due to the 2-fold crystallographic symmetry, despite some pairing of mismatched bases ( Fig. 2A). Gua4 and Thy10 formed typical TG-mismatched structure. In contrast, electron densities of Gua7 and Gua7 were ambiguous, and thus these could form a Hoogsteen base pair with dual conformations. Details of the interaction of HLTF 58 -174 with DNA are described and discussed in later sections.
An NMR structure of the HIRAN domain of human HLTF (residues 50 -171) had previously been deposited in the Protein Data Bank (PDB ID: 2L1I). That structure was distinctly different from our structure in complex with DNA, as it was composed of one ␤-sheet flanked with four ␣-helices. Root mean square deviation values for structural superimposition between the crystal structure and the NMR structures (20 ensembles) are in the range of 7.248 -8.509 Å for 114 comparable C␣ atoms in residues 58 -171 (Fig. 1D). To investigate whether the DNA binding imposes structural change on the HIRAN domain, we also determined the crystal structure of a DNA-free form of His-HLTF 58 -174 by molecular replacement using the structure of the DNA-bound form. Nine molecules of the DNA-free form were contained in the asymmetric unit. Our DNA-free structures were virtually identical to the DNA-bound structure with root mean square deviation values in the range of 0.469 -0.646 Å for superimposable 116 C␣ atoms in residues 58 -173 (Fig.  1D), indicating that the HIRAN domain does not undergo a large structural change upon DNA binding. Actually, residues responsible for DNA binding in DNA-free and DNA-bound forms are located on similar positions (Fig. 1D). It is possible that the structural differences between the structure previously deposited in the Protein Data Bank and our own, if these differences do exist, may be due to the structural plasticity of the HIRAN domain.
Interaction between HIRAN Domain and DNA-As described above, the crystal structure showed that HLTF 58 -174 interacts with dsDNA ( Fig. 2A). Thus, to investigate that the HIRAN domain interacts with duplex DNA, we performed EMSA using blunt-ended and overhanging dsDNAs (Fig. 3A). The result clearly shows that the HIRAN domain actually interacts not only with the ssDNA (Fig. 3A, lanes 3-5) but also with dsDNAs (Fig. 3A, lanes 7-9, 11-13, and 15-17). Interestingly, the HIRAN domain appears to prefer 3Ј-overhanging dsDNA rather than blunt-ended dsDNA (Fig. 3A, lanes 7-9 and 15-17), suggesting that the HIRAN domain interacts with DNA including the exposed single-stranded structure. In fact, the crystal structure reveals that the single-stranded part of DNA is bound to HLTF 58 -174 along with the positive concave (Fig. 2B), and the 3Ј-end of the DNA is specifically bound to the conserved pocket of the HIRAN domain (Fig. 2, B-D). Asn-91 interacts with the Cyt13 of the 3Ј terminus by hydrogen bond. His-110 stacks with the ribose ring of Cyt13. Furthermore, Asp-94 interacts with the 3Ј-OH of Cyt13 at the terminus of DNA. Tyr-73 forms hydrogen bonds with the Cyt12 adjacent to the 3Ј-terminal base and with the His-110 that recognizes the ribose ring at the 3Ј terminus. Two aromatic residues, Tyr-72 and Tyr-93, stack with these two bases, Cyt12 and Cyt13, respectively. Those stacking interactions pinch the two bases of the single-strand moiety (Fig. 2, B and C). In addition, Phe-142 stacks with the Ade1 of the symmetry-related DNA strand that forms dsDNA in the crystal. More interestingly, the HIRAN domain has affinity for the 5Ј-overhanging dsDNA at a level similar to 3Ј-overhanging dsDNA (Fig. 3A, lanes 11-13 and 15-17). This implies that the HIRAN domain may facilitate unwinding of 3Ј-overhanging dsDNA, a simple model of the primer-template DNA, thereby interacts with the single-stranded 3Ј-end. That issue is discussed in the later section.  3-5, 7-9, 11-13, 15-17, 19 -21, and 23-25 are 5Ј-modified ssDNA, 3Ј-modified ssDNA, 3Ј-modified-5Ј-overhanging dsDNA, 3Ј-modified-3Ј-overhanging dsDNA, 5Ј-modified-5Ј-overhanging dsDNA, and 5Ј-modified-3Ј-overhanging dsDNA incubated with HLTF 58 -174 , respectively. Schematic drawings of DNA structures are shown above the blot. Asterisks indicate TAMRA modification at 5Ј-OH or 3Ј-OH.
To clarify the significant residues that interacted with DNA, we performed an EMSA (Fig. 3B). Mutations in residues involved in binding to the terminal nucleotide (N91A and H110A) drastically decreased DNA binding (Fig. 3B, lanes  11-13 and 20 -22), and mutation in Asp-94 interacting with 3Ј-OH caused little defect in DNA binding (Fig. 3B, lanes  17-19) when compared with wild type (Fig. 3B, lanes 2-4, WT). Y73A mutation also remarkably impaired DNA binding (Fig.  3B, lanes 8 -10). As described above, Tyr-73 interacts not only with Cyt12 but also with the His-110 that stacks with the terminal ribose moiety. Thus, the Y73A mutation may have affected the interaction of His-110 with the ribose, thereby indicating a significant reduction in DNA binding. As expected from the crystal structure, the mutation in Tyr-72 involved in pinching Cyt12-Cyt13 of the ssDNA strongly reduced DNA binding (Fig. 3B, lanes 5-7). In contrast, the mutation in Tyr-93 that was also involved in the base pinching did not have a large effect on DNA binding (Fig. 3B, lanes 14 -16), indicating that role of Tyr-72 is more significant than that of Tyr-93 in the base pinching. The Y72A/Y93A double mutant protein lost DNA binding activity (Fig. 3B, lanes 26 -28). Mutation in Phe-142 stacking with the symmetry-related Ade1 in the crystal structure slightly decreased the DNA binding of HLTF 58 -174 (Fig.  3B, lanes 23-25). This EMSA has revealed that Tyr-72, Tyr-73, Asn-91, and His-110 are crucial for DNA binding.
Recruitment of HLTF to the Stalled Replication Fork and the Mechanism of the TS Pathway-As described above, DNA bound to HLTF 58 -174 forms DNA duplex by crystallographic 2-fold symmetry ( Fig. 2A). Furthermore, EMSA reveals that HLTF 58 -174 actually interacts with dsDNA (Fig. 3A). Thus, it is possible that the duplex mimics a primer-template DNA and that the crystal structure presents a snapshot of how HLTF interacts with the stalled replication fork. Moreover, specific recognition of the 3Ј-end and the base-pinching activity of the HIRAN domain imply a significant role for the domain in the process of fork regression. The crystal structure reveals specific interaction of the HIRAN domain with the 3Ј-end, indicating that the domain acts as a sensor of the 3Ј-end. Furthermore, the structure suggests that the HIRAN domain may be capable of unwinding the primer-template DNA by the base pinching with 3Ј-5Ј polarity. Consistent with this hypothesis, HLTF actually exhibits 3Ј-5Ј translocase activity (11). Structural study of UvrD DNA helicase with 3Ј-5Ј polarity has revealed that an aromatic residue, Tyr-621, acts as the separation pin that buttresses the end of DNA duplex (28). The Tyr-72 and Phe-142 of the HIRAN domain are located in the end of DNA duplex (Fig. 2D). Those aromatic residues may play a role similar to the separation pin of UvrD to drive unwinding of the primer-template DNA. In fact, Tyr-72 is significantly involved in interaction with the primer strand (Fig. 3B, lanes 5-7). Therefore, Tyr-72 could be crucial for both separation and DNA binding.
The process of DNA damage tolerance involves either the TLS or the TS pathway, but how one pathway is chosen over the other remains a mystery. TLS appears to be a rapid process because it is performed by several TLS polymerases and does not require fork regression (1). In contrast, TS is a much more complicated process including regression of the stalled replication fork, DNA synthesis, and reversal of the regressed fork (1). It is plausible that the TS pathway is actuated when TLS fails due to DNA damage that is too severe for TLS polymerases to use it as a template (Fig. 4). In such a situation, the HIRAN domain of HLTF may sense the exposed 3Ј-end of the primer stand in the stalled replication fork, thereby recruiting MMS2-UBC13-HLTF to the stalled replication fork. If MMS2-UBC13-HLTF and RAD6-RAD18 had already been recruited to the damaged site to initiate the TLS pathway by mono-ubiquitination of PCNA, they could also perform poly-ubiquitination of PCNA. Additional RAD6-RAD18 might accompany MMS2-UBC13-HLTF, as it has been shown that RAD18 interacts with RAD5 (4). Recently, several groups have reported that the Lys-63-linked poly-ubiquitin chain of PCNA at Lys-164 is recognized by ZRANB3 (zinc finger, RAN-binding domain containing 3), alternatively termed AH2 (annealing helicase 2), that plays an important role in restarting the stalled replication fork (29 -31). ZRANB3 contains not only PCNA-binding motifs but also the NZF (NPL4 zinc finger) domain that is a specialized ubiquitin-binding domain for the Lys-63-linked poly-ubiquitin chain. ZRANB3 is also a SWI/SNF DNA helicase that facilitates fork regression but disrupts the D-loop structure that is a product of strand invasion, suggesting that ZRANB3 suppresses unfavorable sister-chromatid exchange, also known as recombination-mediated TS (29). Interestingly, HLTF is capable of strand invasion and D-loop formation in a RAD51-independent manner (12). Considering this, these two SWI/SNF DNA helicases, HLTF and ZRANB3, might be involved in the early and late stages of the TS pathway, respectively. More complicatedly, mammalian cells have homologs that are functionally related to these DNA helicases. As described above, SHPRH is another RAD5 homolog, whereas interestingly, SHPRH lacks the HIRAN domain, suggesting that HLTF and SHPRH are not functionally redundant. In fact, HLTF and SHPRH have been shown to suppress mutagenesis in a damage-specific manner (32). Furthermore, mammalian cells have a ZRANB3 homolog, SMARCAL1 (SWI/SNF-related, matrix-associated, actin-dependent regulator of chromatin, subfamily A-like 1), that is responsible for Schimke immuno-osseous dysplasia (33). SMARCAL1 catalyzes fork regression and Holliday junction migration (34). It is a puzzle how cells differently utilize these SWI/SNF DNA helicases to maintain genome stability during DNA replication. Structural studies of these DNA helicases in complex with DNA are urgently required to unravel the mystery of the functional diversity of the helicases and the molecular mechanism behind TS. We are continuing to work on these problems.
The HIRAN domain has been identified in proteins involved in DNA transaction events, including DNA repair and chromatin remodeling (13). Here we have shown the first structure of the HIRAN domain in complex with DNA. This is the first evidence showing that the HIRAN domain is a DNA-binding domain. The structure reveals the function of the novel DNAbinding domain as a sensor of the 3Ј-end and implies a significant role in recruiting HLTF to the damaged site, thereby facilitating poly-ubiquitination of PCNA and fork regression. Although the detailed molecular mechanism behind the TS pathway is still an open question and structural study of fulllength HLTF is required, our results are the first to shed light on the mechanisms underlying the DNA transaction events in which the HIRAN domains are involved.