Crystal Structure of Human REV7 in Complex with a Human REV3 Fragment and Structural Implication of the Interaction between DNA Polymerase ζ and REV1*

DNA polymerase ζ (Polζ) is an error-prone DNA polymerase involved in translesion DNA synthesis. Polζ consists of two subunits: the catalytic REV3, which belongs to B family DNA polymerase, and the noncatalytic REV7. REV7 also interacts with REV1 polymerase, which is an error-prone Y family DNA polymerase and is also involved in translesion DNA synthesis. Cells deficient in one of the three REV proteins and those deficient in all three proteins show similar phenotype, indicating the functional collaboration of the three REV proteins. REV7 interacts with both REV3 and REV1 polymerases, but the structure of REV7 or REV3, as well as the structural and functional basis of the REV1-REV7 and REV3-REV7 interactions, remains unknown. Here we show the first crystal structure of human REV7 in complex with a fragment of human REV3 polymerase (residues 1847–1898) and reveal the mechanism underlying REV7-REV3 interaction. The structure indicates that the interaction between REV7 and REV3 creates a structural interface for REV1 binding. Furthermore, we show that the REV7-mediated interactions are responsible for DNA damage tolerance. Our results highlight the function of REV7 as an adapter protein to recruit Polζ to a lesion site. REV7 is alternatively called MAD2B or MAD2L2 and also involved in various cellular functions such as signal transduction and cell cycle regulation. Our results will provide a general structural basis for understanding the REV7 interaction.

Large numbers of DNA lesions occur daily in every cell, and the majority of the DNA lesions stall replicative DNA polymerases. This results in the arrest of DNA replication, which causes lethal effects including genome instability and cell death.
Translesion DNA synthesis (TLS) 2 releases this replication blockage by replacing the stalled replicative polymerase with a DNA polymerase specialized for TLS (TLS polymerase). It is generally considered that TLS includes two steps performed by at least two types of TLS polymerases, namely inserter and extender polymerases (reviewed in Refs. 1 and 2). In the first step, the stalled replicative polymerase is switched to an inserter polymerase such as Pol, Pol, Pol, or REV1, which are classified as Y family DNA polymerases (3) and have different lesion specificity (reviewed in Refs. 4 -8), and an inserter polymerase incorporates nucleotides opposite the DNA lesion instead of the stalled replicative polymerase. In the second step, an inserter polymerase is switched to the extender polymerase DNA polymerase (Pol), and then Pol extends a few additional nucleotides before a replicative polymerase restarts DNA replication.
Pol consists of the catalytic REV3 and the noncatalytic REV7 subunits. REV3 is classified as a B family DNA polymerase on the basis of the primary sequence. The catalytic activity of yeast REV3 is stimulated by yeast REV7 (9). Biochemical analysis has been done only for yeast REV3 but not mammalian REV3, because the molecular mass of human REV3 is larger (ϳ350 kDa) than that of yeast REV3 (ϳ150 kDa). Disruption of the mouse REV3 gene causes embryonic lethality accompanied by massive apoptosis (10 -12), suggesting that the function of mammalian REV3 is essential for embryogenesis. Although REV7 is a smaller protein with a molecular mass of 24 -28 kDa compared with REV3, the function of REV7 is less understood. REV7 is a member of the HORMA (Hop1, Rev7, and Mad2) family of proteins (13). REV7, which is alternatively called MAD2B or MAD2L2, appears to be involved in multiple cellular functions including not only TLS but also cell cycle regulation (14), bacterial infection (15), and signal transduction (16,17).
In this study, we investigated the function of REV7 in TLS from structural analysis. Previous studies have reported that human REV7 interacts with the central region (residues 1847-1892) of human REV3 by yeast two-hybrid and in vitro interaction assays (18). Interestingly, human REV7 also interacts with the C-terminal region (residues 1130 -1251) of human REV1 polymerase as shown by yeast two-hybrid, in vitro interaction and co-immunoprecipitation assays (19 -21). Furthermore, human REV7 and human REV1 were co-expressed by Escherichia coli, and the REV7-REV1 complex was purified, whereas REV7 does not affect the polymerase activity of REV1 (22). The three yeast rev mutants and the triple mutant show very similar sensitivity to various genotoxic treatments (23)(24)(25). Furthermore, chicken DT40 cells deficient in one of the three REV proteins and those deficient in all three proteins show hypersensitivity to various genotoxic treatment including cisplatin (cis-diaminedichloroplatinum (II)), indicating the functional collaboration of the three REV proteins (26). However, these previous analyses failed to determine the mechanism underlying the protein-protein interactions on the atomic revel, because they tried to analyze without data of the three-dimensional structures. In addition, it remains unclear whether mammalian REV1, REV3, and REV7 can form the Pol-REV1 ternary complex. It has been considered that switching of DNA polymerase occurred at least twice in TLS: the switching from a stalled replicative polymerase to an inserter polymerase and from an inserter polymerase to the extender polymerase. Recently, the structural implications of the first polymerase switching have been reported (27). However, the mechanism underlying the recruitment of the extender polymerase to the lesion site and the second polymerase switching as well as the physical and functional interactions of REV1, REV3, and REV7 remains unclear. Here we report the first crystal structure of human REV7 in complex with a fragment of human REV3 (residues 1847-1898). The structure reveals the mechanism underlying Pol formation and shows that the REV7-REV3 interaction unexpectedly provides a structural interface for REV1 binding. Furthermore, we show that these REV7-mediated interactions with REV1 and REV3 are responsible for DNA damage tolerance. Lastly, we propose a model of the structural interplay of REV1, REV3, and REV7 in TLS. Our results will provide a general structural basis for understanding the REV7 interaction in various cellular functions.

EXPERIMENTAL PROCEDURES
Crystallographic Analysis of Human REV7 in Complex with REV3 Fragment-In the present crystallographic study, REV7 with an R124A mutation, REV7(R124A), was used instead of wild type REV7, REV7(WT) (28). It has been shown that a human REV3 fragment (residues 1847-1892) interacts with human REV7 (18). Thus, based on the result of secondary structure prediction, we constructed the REV3 fragment carrying residues 1847-1898, REV3(1847-1898), for this crystallographic study (28). The REV7(WT)-REV3(1847-1898) complex was polydisperse and did not crystallize, whereas REV7(R124A)-REV3(1847-1898) complex was a monodisperse (28). In addition, REV7(R124A) efficiently binds REV3(1847-1898). Preparation and crystallization of the human REV7(R124A)-REV3(1847-1898) complex have been described before (28). In brief, recombinant human REV7(R124A) with an N-terminal hexameric His tag in complex with human REV3(1847-1898) was expressed in E. coli BL21(DE3) harboring the REV7(R124A)-REV3(1847-1898) co-expression vector. The protein was purified by nickel-Sepharose resin (GE Healthcare), HiTrap Q HP (GE Healthcare), and HiLoad Superdex200 (GE Healthcare). Monoclinic and tetragonal crystals of the REV7(R124A)-REV3(1847-1898) complex were obtained in different conditions. Heavy atom derivatives of monoclinic crystals were prepared by the soaking method using a solution of 10 mM ethylmercurithiosalicylate, 100 mM Tris-HCl, pH 7.5, 800 mM sodium formate, and 25% (w/v) polyethylene glycol 2000 monomethyl ether for 20 h. X-ray diffraction data for native crystals were collected by using a Quantum 315 CCD detector (Area Detector Systems Corp.) on Beamline BL-5A at Photon Factory. X-ray diffraction data for derivative crystals were collected by using an FR-D in-house x-ray generator with an R-AXIS IV ϩϩ imaging plate detector (Rigaku). All of the diffraction data were processed with the program HKL2000 (29). The structure of the REV7(R124A)-REV3(1847-1898) complex was solved by the single isomorphous replacement method using the programs SOLVE and RESOLVE (30,31). Model building was performed with the programs O (32) and COOT (33). Structure refinement was performed at 1.9 Å resolution with the programs CNS (34) and REFMAC (35). P2 1 crystal contains two REV7(R124A)-REV3(1847-1898) complexes in the asymmetric unit. The structure in the P4 1 2 1 2 crystal was solved at 2.6 Å resolution by the molecular replacement method with the program MOLREP (36) using one of P2 1 structures. The structure was refined with a procedure similar to that of the monoclinic case. The data collection and refinement statistics are given in Table 1. The coordinates and structure factors have been deposited in the Protein Data Bank Japan.
Rapid Survival Assays Using Chicken DT40 Cells-For retrovirus infection, a pMSCV-IRES-GFP recombinant plasmid was constructed by ligating the 5.2-kb BamHI-NotI fragment from pMSCVhyg (Clontech) with the 1.2-kb BamHI-Not1 fragment from pIRES2-EGFP (Clontech). cDNA of chicken REV7 (GdREV7) was inserted between the BglII and EcoRI sites of pMSCV-IRES-GFP. Virus was prepared by using 293T cells and 1 l of Gene juice (Novagen), 1 g of pMSCV-GdREV7-IRES-GFP, and 1 g of pCL-Ampho. After the 293T cells were cultured with the above reagents and plasmids at 37°C for 2 days, the cells were centrifuged, and the supernatant was stored at Ϫ80°C. Retrovirus infection was done by centrifugation (3000 rpm, 30 min, 32°C) of the DT40 REV7 Ϫ/Ϫ cells (26) and the retroviral solution. A day after infection, expression of GFP was confirmed by flow cytometry. The efficiency of infection was more than 50%, as assayed by GFP expression. The cells were subcloned into 96-well plates, and clones displaying high levels of GFP were determined by a fluorescence-activated cell sorter. To test for differential sensitivity to cisplatin, we performed rapid survival assays using chicken DT40 cells. 3 The cells (1 ϫ 10 4 ) were exposed to various concentrations of cisplatin and incubated at 39.5°C for 48 h. We analyzed each cell type with at least three clones, and at least three independent experiments were carried out to obtain individual data. The cell number was counted by the Cell Titer-Glo luminescent cell viability assay (Promega) according to the manufacturer's instructions. We calculated the extent of cytotoxicity.
Human REV7 shares 22% amino acid identity with human Mad2, another member of the HORMA family (13). Mad2 functions in spindle assembly checkpoint by binding directly to Mad1 (40 -42). Mad2 undergoes a striking conformational change from the open (O-Mad2) to the closed (C-Mad2) form, in which the C-terminal region known as the "seatbelt" following ␤6 moves toward the edge of ␤5 to wrap around the ligand, and ␤1 is relocated. Concomitantly, ␤7 and ␤8 are rearranged to form ␤8Ј and ␤8Љ (43, 44) (Fig. 1A). Structural alignment between REV7 bound to the REV3 fragment and Mad2 bound to Mad1 shows an root mean square deviation value of 2.0 Å for 183 superimposable C␣ atoms (Fig. 1D), indicating that the overall structure of REV7 in the REV7(R124A)-REV3(1847-1898) complex may be very similar to the structure of the closed Mad2 form. Conceivably, a large conformational change may occur in the seatbelt (residues 153-211) of REV7 upon interaction with the REV3 fragment.
Structural Details of the Interaction between REV7 and REV3 in Pol Formation-Our novel structural analysis of the human REV7(R124A)-REV3(1847-1898) complex allowed for the pre-3 T. Kogame and S. Takeda, unpublished data.  (44). However, a similar interaction observed in the Mad2-Mad1 complex is not observed in the REV7(R124A)-REV3(1847-1898) complex, implying that the mechanism underlying the REV7-REV3 interaction may be distinct from that of Mad2-Mad1 interaction. To elucidate the crucial residues responsible for the REV7-REV3 interaction, we performed in vitro binding assays using alanine mutants of REV7. Of these mutants, Y63A or W171A mutation significantly reduced affinity for REV3(1847-1898), indicating that Tyr-63 REV7 and Trp-171 REV7 are crucial for the physical interaction with REV3(1847-1898) (Fig. 2B, lanes 5 and 8). In contrast, Lys-159 Mad2 , Leu-161 Mad2 , Tyr-64 Mad2 , and Trp-167 Mad2 were crucial for the physical interaction with Mad1 in the Mad2-Mad1 interaction (44). Thus, the mechanism underlying  the REV7-REV3 interaction is distinctly different from that underlying the Mad2-Mad1 interaction. In addition, the ␣-helix of REV3(1847-1898) might be required for interaction with REV7, because a REV3 fragment (residues 1847-1886) lacking the ␣-helix did not form a stable complex with REV7 (data not shown). This suggests that the van der Waal's interactions by the ␣-helix of REV3(1847-1898) also contribute to the formation of the REV7-REV3 complex, Pol.
Furthermore, to investigate the REV7-REV3 interaction in vivo, we carried out co-immunoprecipitation assays using HEK293 cells. Consistent with the in vitro results, REV7(Y63A/ W171A) showed no binding affinity for REV3(1776 -2044), although REV7(Y63A) and REV7(W171A) retained affinity (Fig. 2C, lanes 3-5). It is also noteworthy that the expression level of FLAG-REV3(1776 -2044) in lane 5 of Fig.  2C is considerably lower compared with other lanes, even though the same amount of plasmid DNA was used for transfection in each cells, suggesting that REV7-unbound REV3(1776 -2044) may be unstable in vivo. Interestingly, REV7(R124A) had markedly higher affinity for REV3(1776 -2044), as compared with REV7(WT) (Fig. 2C, lane 2). This observation suggests that the R124A substitution stabilized the closed conformation of REV7. In fact, the analogous substitution in Mad2, Mad2(R133A), pushes the conformation toward the closed form and enables structure determination of the ligand-free C-Mad2 (45), although why this substitution stabilizes the closed form is less understood. The CD spectrum of the REV7(R124A)-REV3(1847-1898) complex is similar to that of the REV7(WT)-REV3(1847-1898) complex (28), whereas the CD spectrum of REV7(R124A) is distinct from that of REV7(WT) (supplemental Fig. S2). This observation indicates that REV7 also supposedly undergoes a significant structural change, in which the seatbelt region is expected to move upon the ligand binding as observed in Mad2.
To identify amino acid residues of REV7 responsible for interaction with REV1, we performed comprehensive alanine substitutions in the solvent-exposed residues of REV7 (Fig. 3A). Mad2 has two independent binding sites for different proteins; one is the seatbelt region, where Mad1 interacts with Mad2, and the other is the ␣C helix, which is the binding site of O-Mad2 for checkpoint activation (47) or p31 comet for checkpoint inhibition (48). However, mutations in the ␣C helix in REV7 did not affect binding to REV1(1130 -1251) (data not shown). Our results show that L186A or Q200A mutation significantly depletes REV1 binding, whereas those mutations have no effect on REV3 binding (Fig. 3B, lanes 3 and 4). The fact that Leu-186 and Gln-200 are exposed to solvent (Fig. 3C) indicates that these residues are directly involved in REV1 binding. Leu-186 and Gln-200 are present in ␤8Ј and ␤8Љ, respectively (Fig. 3C). Therefore, the REV1-binding site is unprecedented and represents a novel interface for protein-protein interactions of HORMA family proteins. Consistent with this finding, Leu-186 and Gln-200 are not conserved in Mad2 (Fig. 1A), whereas they are conserved in yeast REV7. Leu-186 and Gln-200 of REV7 are positioned close to each other and are thus likely to provide a platform for REV1 binding on the anti-parallel ␤-sheet composed of ␤8Ј and ␤8Љ. This observation implies that formation of the C-terminal ␤-sheet is significant for REV1 binding. To verify this idea, we performed further surveys by alanine substitution for residues that are relatively buried and seemingly involved in stabilizing the structure of the ␤-sheet. Consequently, we found that the Y202A mutation impaired the REV1 interaction (Fig. 3B, lane 5). Tyr-202 is located in ␤8Љ and directly interacts with both Leu-186 and Gln-200 (Fig. 3C).
These results indicate that Tyr-202 stabilizes the REV1-binding platform.
To examine whether Leu-186 REV7 is important for interaction with REV1 in vivo, we performed binding assays using human cells and showed that the L186A mutation greatly reduced the interaction with REV1(826 -1251), although it did not reduce the interaction with REV3(1776 -2044) (Fig. 2C,  lane 6, and 3D, lane 4). On the other hand, the W171A mutation unexpectedly decreased the interaction with not only REV3(1776 -2044) but also REV1(826 -1251) (Fig. 3D, lane 3). Most interestingly, the R124A mutation, which stabilized the closed conformation of REV7, increased REV1 binding; furthermore, the double mutation R124A/W171A in REV7 brought back REV1 binding to the level of REV7(WT) (Fig. 3D,  lanes 2, 3, and 5). Therefore, we conclude that REV1 interacts with the closed form of REV7 and that the REV7-REV3 interaction precedes the REV7-REV1 interaction during formation of the Pol-REV1 complex.
Functional Role of REV7-mediated Interactions in DNA Damage Tolerance-To examine the contribution of REV7mediated interactions to DNA damage tolerance, we performed rapid survival assays using chicken DT40 cells. Chicken REV7 shares a high degree of amino acid identity with human REV7 (96%). To functionally analyze the REV7-mediated interactions, we expressed chicken REV7(WT) and REV7(Y63A/ W171A) in DT40 REV7 Ϫ/Ϫ cells and measured the sensitivity to cisplatin in the resulting reconstituted clones (Fig. 4). We chose cisplatin, because REV1 Ϫ/Ϫ , REV3 Ϫ/Ϫ , or REV7 Ϫ/Ϫ DT40 cells show strong sensitivity to cisplatin (26,49). The reconstitution of REV7 Ϫ/Ϫ cells with the REV7 transgene completely normalized cellular sensitivity to cisplatin (white square). In contrast, expression of REV7(Y63A/W171A), which lack both REV3 and REV1 interactions, had no impact on cellular sensitivity of REV7 Ϫ/Ϫ cells to cisplatin (black triangle). Our results clearly indicate that REV7-mediated interactions are essential for resistance to the DNA damage caused by cisplatin, implying that formation of the Pol-REV1 complex is responsible for resistance and that REV7 functions as an adapter protein between REV3 and REV1 polymerases.
Recruitment of Pol and Polymerase Switching-Taking our results altogether, we can propose a model of the interactions involving REV1, REV3, and REV7 (Fig. 5A). In the absence of REV3, REV7 can adopt an open form. In the presence of REV3, REV7 undergoes structural rearrangement of the seatbelt by REV3 binding, resulting in the formation of Pol, where Tyr-63 REV7 and Trp-171 REV7 are crucial for the interaction (Fig. 2). The conformational change in the seatbelt of REV7 provides an interface for REV1 interaction (Fig. 3C) and therefore enables formation of the Pol-REV1 complex. REV1, which is the inserter polymerase for an abasic lesion (50,51), is also supposed to function as a scaffold protein for polymerase switching at a lesion site, because the C-terminal region (residues 1130 -1251) of REV1 interacts with the other inserter polymerases, namely Pol, Pol, and Pol (21). In addition, it has been shown that mouse Pol and Rev7 compete directly for binding to Rev1 (20). Based on our results, we can propose a model of the recruitment of Pol and the second polymerase switching from an inserter polymerase to the extender polymerase, as well (Fig. 5B). An inserter polymerase suitable for a DNA lesion could perform bypass synthesis after the first polymerase switching from a replicative polymerase to an inserter polymerase through interactions with ubiquitinated PCNA and/or the C-terminal region of REV1. Then Pol (REV7-REV3 complex) could be recruited to the lesion site by the interaction between REV7 and the C-terminal region of REV1, where the C-terminal ␤-sheet of REV7 is crucial. The REV7-REV1 interaction would release an inserter polymerase from the C-terminal region of REV1, and Pol subsequently could perform extension from the nucleotide inserted by an inserter polymerase (Fig. 5B). Depending on DNA lesions, it is considered that Pol performs both nucleotide insertion and extension. In this case, Pol is recruited to the lesion site through the REV7-REV1 interaction in a similar way. In contrast to the interaction of human REV7 and REV1, it has been reported that yeast Rev7 interacts with various regions of yeast Rev1 (the N-terminal BRCT domain; the polymerase-associated domain, which is alternatively called the little finger domain; and the C-terminal domain) (52-54). Furthermore, yeast Pol, an inserter polymerase, interacts with the polymerase-associated domain of yeast Rev1 (55). Therefore, the Rev1 interactions in yeast are complicated, and the interactions might diverge between yeast and human.
In this work, we have performed structural and functional analysis of REV7-mediated interactions, and clarified the structural basis of the REV7-REV3 interaction and obtained structural implication of Pol-REV1 interactions. We propose that REV7 functions as the adapter protein between REV3 and REV1 polymerases, thereby mediating the second polymerase switching. Human REV7 interacts with various proteins including the Shigella effector IpaB in bacterial infection and ELK1 and TCF4 in signal transduction (15)(16)(17). Those proteins have sequences similar to the REV7-binding region of human REV3, indicating that the mechanisms underlying the interactions of those proteins with REV7 are similar to that of the REV7-REV3 interactions described here. Thus, our finding will provide a general structural basis for understanding the interactions mediated by REV7 in various cellular functions.