Molecular basis for assembly of the shieldin complex and its implications for NHEJ

Shieldin, including SHLD1, SHLD2, SHLD3 and REV7, functions as a bridge linking 53BP1-RIF1 and single-strand DNA to suppress the DNA termini nucleolytic resection during non-homologous end joining (NHEJ). However, the mechanism of shieldin assembly remains unclear. Here we present the crystal structure of the SHLD3-REV7-SHLD2 ternary complex and reveal an unexpected C (closed)-REV7-O (open)-REV7 conformational dimer mediated by SHLD3. We show that SHLD2 interacts with O-REV7 and the N-terminus of SHLD3 by forming β sheet sandwich. Disruption of the REV7 conformational dimer abolishes the assembly of shieldin and impairs NHEJ efficiency. The conserved FXPWFP motif of SHLD3 binds to C-REV7 and blocks its binding to REV1, which excludes shieldin from the REV1/Pol ζ translesion synthesis (TLS) complex. Our study reveals the molecular architecture of shieldin assembly, elucidates the structural basis of the REV7 conformational dimer, and provides mechanistic insight into orchestration between TLS and NHEJ. Shieldin, including SHLD1, SHLD2, SHLD3 and REV7, functions to suppress the DNA termini nucleolytic resection during non-homologous end joining (NHEJ). Here the authors present the crystal structure of the SHLD3-REV7-SHLD2 ternary complex revealing insights into the mechanism of the complex.

In the four-subunit protein complex, REV7 was the first component to be reported as a counteractor of DNA DSB resection and an important regulator in choice of DSB repair pathway in 2015 10,11 . REV7, also known as MAD2B and MAD2L2, is a conserved protein of the HORMA domain (named after the Hop1, REV7 and Mad2 proteins) family and undergoes an open (O)-to-closed (C) transition when partner proteins bind to the safety-belt 12 . REV7 was first found to be a subunit of Pol ζ, a translesion synthesis (TLS) polymerase that enables replication of damaged DNA [12][13][14] . Pol ζ is composed of REV3, REV7, PolD2 and PolD3, in which REV3 is the catalytic subunit and REV7 acts as a multitasking scaffolding protein 13,15,16 . Human REV3 (hREV3, hereafter called REV3) has over 3,000 residues and is twice as large as its homolog in yeast. Except for the relatively conserved N-terminal 250 amino acids and C-terminal 800 amino acids, which are homologous to Pol α, δ, and ε, but without 3′-5′ exonuclease activity, REV3 has a positively charged domain (PCD, amino acids 1042-1251) for DNA binding and two REV7-binding motifs (RBMs). The two RBMs in REV3 (amino acids 1847-2021, hereafter called REV3 (1847-2021)) are characterized as PXXXpP motif and both RBM 1 (amino acids  and RBM 2 (amino acids 1977-2021) bind to REV7 underneath the safety belt loop as other reported REV7 binding proteins, such as human RAN, ELK-1, chromosome alignment-maintaining phosphoprotein (CAMP) and shigella IpaB [17][18][19][20][21][22] . To date, all reported crystal structures of REV7 complexed with its partners adopt a similar closed conformation as a monomer with RBMs bound underneath the safety-belt loop 12,17,[21][22][23][24][25] .
Since SHLD3 has two REV3-like RBM motifs, it is supposed that SHLD3 interacts with the C-terminal safety-belt of REV7 in a similar manner 6 . However, whether these two RBMs interact with REV7 in the same way as REV3 is uncertain. On the other hand, although it is known that the N terminus of SHLD2 (amino acids 1-60, hereafter called SHLD2(1-60)) is sufficient for its interaction with upstream molecules SHLD3 and REV7 1,4 , neither SHLD3 nor REV7 interacts with SHLD2 solely 4,6 , the details of their interactions needs to be further explored to understand how shieldin is assembled.
In this study, we solved the crystal structure of the SHLD3-REV7-SHLD2 ternary complex. We demonstrate that SHLD3 binds to REV7 in a completely different way from that of other REV7 binding proteins. Two copies of REV7 bind to SHLD3, and REV7 adopts two conformations with different topologies, closed and open states. O-REV7 is essential for the interaction between SHLD3-REV7 sub-complex and SHLD2 by forming β sheet sandwich occupying the position of the safety belt. Further evidence shows the conformational dimer precludes the binding of C-REV7 to REV1 C-terminal domain (CTD) and may act as a platform to interact with other REV7 binding proteins, such as REV3. Taken together, our work illustrates how REV7 interacts with other components of shieldin through its conformational change, and reveals NHEJ and TLS are mutually exclusive events coordinated by REV7.

Results
Overall structure of the SHLD3-REV7-SHLD2 complex. Shieldin complex is composed of REV7 and three newly characterized proteins SHLD1, SHLD2 and SHLD3 1,4,6 . REV7 contains a HORMA domain that usually acts as an adaptor to recruit other proteins. SHLD2 contains an N-terminal REV7 interacting motif (RIM) and a C-terminal OB fold domain that resembles RPA70, which is connected by a predicated disordered linker 1,4 . SHLD3 contains two putative REV7-binding motifs (RBM) in N terminus and an EIF4E-like motif in C terminus 6 (Fig. 1a).
To elucidate how shieldin complex is assembled, we constructed full length REV7 and the N-terminal REV7-binding domains of both SHLD3 and SHLD2 (designated as SHLD3  and SHLD2(1-52)), co-purified REV7-SHLD3(1-64)-SHLD2 , and solved the crystal structure of the complex to 3.5 Å resolution ( Fig. 1a and Table 1). The high-quality electron density map made unambiguous building of the structure model possible ( Supplementary Fig. 1). Unexpectedly, the crystal structure shows that SHLD3-SHLD2 binds two REV7 molecules, one in its closed (C) state and the other in an open (O) conformation, which are characterized by the conformation of the safety belt, thereafter we name it as C-REV7-O-REV7 conformational dimer (Fig. 1b). Compared with C-REV7, the αC, β1, β2 and β7 of O-REV7 show obvious rearrangement despite the safety belt (Fig. 1c, d). As a consensus REV7-binding motif PXXXpP, RBM 2 of SHLD3 interacts with C-REV7 in a canonical manner as RBMs of REV3 ( Supplementary Fig. 2). However, the recognition mechanism between SHLD3 and O-REV7 is completely different (Fig. 1b). Indeed, RBM 1 ( 11 PCESDP 16 ) of SHLD3 is not a consensus REV7binding motif PXXXpP, and this is in accordance with the observation that only Pro11 is conserved across species as Pro53 and Pro58 while Pro16 is not (Fig. 1e). Instead, SHLD2 helps SHLD3 interact with O-REV7 and acts as a bolt to lock SHLD3 and O-REV7 tightly, which we will discuss in detail later (Fig. 1b).
The interface between C-REV7 and O-REV7. In our crystal structure, REV7 forms a conformational dimer which resembles C-Mad2-O-Mad2 but is different 26 . Compared with the known C-Mad2-O-Mad2 conformational dimer, the obvious difference is that SHLD3 acts as a bridge to link C-REV7 and O-REV7 ( Supplementary Fig. 3a). Similar to C-Mad2-O-Mad2, the REV7 conformational dimer uses the conventional HORMA interface centered around αC helix to dimerize and the dimer interface is mainly connected by hydrogen bonds and electrostatic interactions with a buried area of 1670 Å 2 (Fig. 2a, b). Several residues make key interactions including Glu35, Lys44, Arg124, Lys129, Asp134 and Lys190 (Fig. 2a). Discriminatively, Arg124 and Asp134 of both REV7 molecules are located at the interface and contributes to the interaction, while Glu35, Lys44 of C-REV7 and Lys129, Lys190 of O-REV7 diverge from the interface and have no effect on the interaction (Fig. 2a,  (1-82) hardly binds to REV7 K129A ( Fig. 2d-g). This is because Lys129 and Lys190 of C-REV7 locate at the asymmetric interface of C-REV7-O-REV7, and contribute to the interaction, while Glu35, Lys44 of C-REV7 diverge from the interface and have no effect on the interaction. Collectively, REV7 utilizes the conventional HORMA interface to form an asymmetric conformational dimer.
SHLD3 enhances the interaction between C-REV7 and O-REV7. SHLD3 acts as a bridge to link C-REV7 and O-REV7, we wondered whether SHLD3 further strengthens the REV7 conformational dimer. Although RBM 2 of SHLD3 interacts with C-REV7 in a canonical manner, the structure also shows that 38 FXPWFP 43 makes extensive interactions with C-REV7 despite RBM 2 , and these residues are all highly conserved across species, which indicates FXPWFP may be an unrevealed C-REV7 binding motif and exerts important functions ( Fig. 3a and Fig. 1e). The conserved FXPWFP motif is composed of hydrophobic residues and binds a relatively hydrophobic surface of C-REV7 (Fig. 3a). Despite the hydrophobic interactions, many hydrogen bonds also contribute to the interaction. In brief, REV7 residues Lys82, Glu101, Gln200 and Tyr202 form hydrogen bonds with backbone of the FXPWFP motif and Gln200 of REV7 also forms a hydrogen bond with the side chain of Trp41 SHLD3 (Fig. 3b). Moreover, with Phe38 SHLD3 in the center, residues Phe38, Pro40 of SHLD3, residues Leu186, Pro188, Tyr202 of C-REV7 and residues Val132, Ala135, Val136 of O-REV7 make extensive hydrophobic interactions (Fig. 3b).
C-REV7-O-REV7 is essential for the recruitment of SHLD2. To determine whether SHLD3 mediated REV7 conformational dimer is essential for binding to SHLD2, we performed gel filtration assay. Because SHLD2(1-60) cannot be expressed solely in a soluble state, we fused a MBP tag to its N terminus, which is designated as MBP-SHLD2(1-60). As shown in Fig. 5a and interface impairs the assembly of the SHLD3-REV7-SHLD2 complex. REV7 K190A -SHLD3(1-82) fails to interact with O-REV7 as we previously showed (Fig. 2g), which is essential for the assembly of shieldin, we determined whether it interferes with NHEJ. As expected, overexpression of REV7 K190A but not REV7 K44A reduced NHEJ efficiency in a dominant-negative manner (Fig. 5e). This is because REV7 K44A -SHLD3 still recruits endogenous REV7 WT to assemble the shieldin complex while REV7 K190A -SHLD3 precludes REV7 WT binding capability, which leads to the deficiency of the shieldin complex assembly. Therefore, these results demonstrate that SHLD3 mediated REV7 conformational dimerization is essential for the recruitment of SHLD2 and efficient NHEJ. Leu74, Val148 and Val150 of O-REV7 (Fig. 6a, right panel). On the reverse side, extensive hydrophobic interactions are made by lots of hydrophobic residues, including Val7, Ile9, Trp11, Ile39, Leu41, Tyr43, Leu48, Leu50 of SHLD2, Val5, Leu7, Tyr9, Pro16, Leu19, Pro20, Ile27 of SHLD3 and Leu149 of O-REV7 (Fig. 6b).
The Cβ-Cδ of K37 SHLD2 is also involved in this interaction. Totally, the interface between SHLD2 and O-REV7 buries 1678 Å 2 area and an area of 2471 Å 2 is buried between SHLD2 and SHLD3 interface. Interestingly, the structural alignment also shows β1 of SHLD2 occupies the positon of RBM 2 of SHLD3 in C-REV7-SHLD3 and the N-terminal region of SHLD3 displaces the safety belt of C-REV7 (Fig. 6c). This unique binding mode and these extensive contacts make the complex very stable. Furthermore, the amino acid residues between the parallel β sheet formed by β1 of SHLD2 and β1 of SHLD3 are similar ( Supplementary Fig. 6a). Therefore, it is possible that without SHLD2, SHLD3 interacts with O-REV7 in a similar manner. To verify this, we performed molecular dynamics simulations by deletion of SHLD2. Notably, after simulation, SHLD3-C-REV7-O-REV7 forms a stable complex and β1 of SHLD3 binds to O-REV7 in a similar manner as β1 of SHLD2 ( Fig. 6d and Supplementary Fig. 6b). Hence the N terminus (amino acids 1-27) of SHLD3 interacts with O-REV7 when it forms a conformational dimer with C-REV7. However, the binding affinity between SHLD3(1-27) and O-REV7 is extremely weak because deletion of SHLD3(1-27) has little impact on the interaction between REV7 K44A -SHLD3 and REV7 K129A as shown in Fig. 2e and Fig. 3e (0.11 ± 0.01 μM versus 0.20 ± 0.01 μM). Therefore, SHLD2 displaces β1 of SHLD3 to further lock the complex in a more stable state. These results show that SHLD2 acts as a bolt to lock SHLD3 and O-REV7 tightly.
O-REV7-SHLD2 interaction impairment leads to NHEJ deficiency. We further investigated how the interaction between O-REV7 and SHLD2 impact NHEJ efficiency. Tyr63 and Trp171 are two evolutionarily conserved residues that contribute equally to interaction between safety-belt and RBMs ( Supplementary  Fig. 7a). Our structure shows O-REV7 Tyr63 but not Trp171 locates at the interface between SHLD2 and O-REV7 (Fig. 7a). We suppose that REV7 Y63A may influence the interaction between the SHLD3-REV7 conformational dimer and SHLD2. We purified the recombinant expressed REV7 Y63A -SHLD3(1-82) and REV7 W171A -SHLD3(1-82), which form homogeneous conformational dimer and both form stable complexes with a b c d e Fig. 3 The highly conserved FXPWFP motif of SHLD3 binds to C-REV7 and enhances the binding affinity between C-REV7 and O-REV7. a The interface between SHLD3 and C-REV7. C-REV7 and O-REV7 are shown in electrostatic surface representation (positive potential, blue; negative potential, red), SHLD3 in ribbon view, and the FXPWFP motif in sticks. The N terminus of SHLD3 is also indicated. Residues 1-37 of SHLD3 are shown in ribbon. b Close-up view of the interface between the FXPWFP motif (magenta) and C-REV7 (green), O-REV7 (cyan). Hydrogen bonds are indicated by dashed lines and the residues that are involved in the hydrophobic interactions are shown in sticks. c-e ITC measurements of interaction between distinct REV7 K44A -SHLD3 (1-82) truncations or mutants and REV7 K129A . The calculated N and K D are indicated as described in Fig. 2d. Source data are provided as a Source Data file.
Afterwards, we also detected the interaction between SHLD3 and REV3(1847-2021) in vivo and the interaction is enhanced after doxorubicin treatment that causes DSBs (Fig. 8e, lane 4 versus lane 2). Meanwhile, to determine whether REV3 accumulates at DSB sites in living cells, we induced DNA damage tracts (laser lines) by employing laser micro-irradiation. REV3 TR1 is an approximately minimal truncation of REV3 that remains its normal function according to previous report 28 (Fig. 8f). Both full length and the functional truncated GFPtagged REV3 (REV3 TR1 ) accumulated in laser-lines, which were discerned by their co-localization with mCherry-tagged SHLD3 (Fig. 8g, h). These results suggest that REV3 accumulates at DNA damage sites and SHLD3 mediated REV7 conformational dimer interact with REV3 both in vitro and in vivo and that SHLD3 mediated REV7 conformational dimer may act as a platform to coordinate various proteins to accomplish NHEJ.

Discussion
Shieldin is an important downstream effector of 53BP1-RIF1 to regulate the repair of DNA DSBs. In this study, we successfully solved the crystal structure of the SHLD3-REV7-SHLD2 complex, which reveals a striking SHLD3 mediated conformational dimer of C-REV7-O-REV7 and established the molecular architecture of the shieldin complex ( Supplementary Fig. 8d).
The previous report shows that the interaction between SHLD3 and REV7 is abolished by REV7 Y63A but unaffected by REV7 W171A in yeast two-hybrid (Y2H) assay 6 . However, our structural and biochemical studies demonstrate that Tyr63 and Trp171 of REV7 play equal important roles in the REV7-SHLD3 interaction. This controversial situation may be a result of the technical limitation of the Y2H assay. Our structural analysis show that SHLD3 makes much more extensive contacts with REV7 than the RBM 1 of REV3 does with REV7. Thus, neither REV7 Y63A nor REV7 W171A mutant abolishes the interaction between REV7 and SHLD3. On the other hand, O-REV7 Tyr63 but not Trp171 locates at the interface between SHLD2 and O-REV7, and REV7 Y63A diminishes its interaction with SHLD2 while REV7 W171A has no effect. This is consistent with the phenomenon previously reported that cells expressing REV7 Y63A exhibit lower CSR efficiency than those expressing REV7 W171A mutant 6 . Furthermore, it is reported that REV7 K129A fails to interact with SHLD1-SHLD2 in vivo and shows CSR deficiency 6 . Our studies show that Lys129 of REV7 is essential for the integration of the REV7 conformational dimer and this conformational dimer is essential for the interaction with SHLD2 can perfectly explain the CSR deficiency in cells expressing REV7 K129A mutant. Moreover, we found that SHLD3(1-27) is also essential for priming SHLD2 binding to O-REV7.
Our results show that SHLD3 mediated REV7 conformational dimer not only recruits SHLD2, but also interacts with other proteins like REV3 that can bind to O-REV7. This suggests that the SHLD3-C-REV7-O-REV7 trimer acts as a platform to recruit various proteins other than only SHLD2 for exerting different cellular functions, which is in accordance with the observation that SHLD3 and REV7 have similar abundance while SHLD1 and SHLD2 have a much lower abundance in the affinity-purification mass spectrometry data 1 . Since both Mad2 and REV7 form a conformational dimer to recruit downstream effectors, this may be a universal mechanism utilized by HORMA proteins 26 . It is known that Mad2 is regulated by p31 comet and ATPase TRIP13 through structural remodeling 29,30 . However, how shieldin is regulated is unknown. Whether shieldin is regulated in a similar manner needs to be tested. Since disruption of the conformational dimer disables the assembly of the SHLD3-REV7-SHLD2 complex and TRIP13 is found to interact with REV7 4 , it is possible that TRIP13 regulates shieldin through structural remodeling of REV7 to disrupt the conformational dimer. Meanwhile, since mutants that abolish the conformational dimer (such as R124A, K129A) show stronger interaction with SHLD3, these mutants could function in dominant-negative manner.
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15879-5 ARTICLE by conditional knockout of Rev3 10,33 . CH12 cells with Rev3 depletion exhibits normal RPA recruitment but CSR deficiency 10 , indicating that Rev3 ablation does not affect the protection of ssDNA, and suggesting that REV3 is essential for the mediated step(s) between ssDNA protection and the final step of NHEJ. Given the fact that Pol ζ has been designated as an extender polymerase, and our results show its catalytic subunit REV3 can be recruited to DSB sites, we propose that Pol ζ may orchestrate with Pol α to complement ssDNA to double-strand DNA, especially when the ssDNA is longer than 20 nt, which is beyond the processive capacity of Pol α, to trigger NHEJ.
Pol ζ/REV1 plays an essential role in the TLS that usually leads to resistance of cancer cells to chemotherapy 34  Recently, a small molecule, JH-RE-06, was discovered to block the REV1-REV7 interaction by inducing REV1 dimerization 27 . JH-RE-06 shows promising effects to improve chemotherapy both in vitro and in vivo, indicating that inhibition of the REV1 CTD-REV7 interaction has therapeutic potential. Our structure shows the highly conserved FXPWFP motif blocks the REV7-REV1 CTD interaction by masking the REV1 CTD-binding surface of REV7, here we propose this unrevealed REV7 surface will help develop inhibitors to improve chemotherapy, and high-affinity inhibitors targeting this unrevealed REV7 surface is worth further development.
Taken together, our results reveal the unexpected architecture of SHLD3-C-REV7-O-REV7-SHLD2 tetramer, which provides new biological insight into how SHLD3-REV7 coordinates with different downstream mediators and how to develop new small molecules for improving chemotherapy.
Cloning. cDNAs encoding human REV7 and SHLD2 were kind gifts from Han lab. cDNA of REV7 full length was cloned into ORF1 of pETduet by EcoRI and NotI. For co-expression with REV7 and SHLD3, the coding sequences of truncations of SHLD2 were cloned into pET28a(+) by NcoI and XhoI without tag. For biochemical studies, the coding sequences of SHLD2(1-60), REV3(1847-1906) and REV3(1847-2021) were cloned into a modified pET28a(+) with His 6 -MBP tag. For co-expression with REV7, the coding sequence of SHLD3 was synthesized by Synbio Technologies and truncations of it were cloned into ORF2 of REV7 inserted pETduet by NdeI and XhoI without His-tag. For in vivo studies, the coding sequence of SHLD3 was cloned into pmCherryC1 with an N-terminal mCherry tag or pCDH-puro with an N-terminal S-tag-HA tag. For biochemical studies, the coding sequences of SHLD3(1-82) and cREV1 CTD were cloned into pET28a(+) with a C-terminal His 6 -tag. The coding sequence of the truncated REV3 was amplified from the cDNA of Jurkat cells and truncations of it was cloned into pET28a(+) with a C-terminal His 6 -tag or pEGFPC1 with an N-terminal GFP tag. The coding sequence of full length REV3 was amplified from JT113-pETDuet1-(R)-hREV3L, which was a gift from Richard Wood (Addgene plasmid # 64872) 35 . Mutants of REV7 and SHLD3 were generated by site-directed mutagenesis. All plasmids were verified by sequencing (Ruibiotech). The supernatant was applied to a Ni-IDA beads (Smart-Lifesciences) and washed with buffer containing 20 mM Tris, pH 8.0, 500 mM NaCl, 1% glycerol and 0.5 mM TCEP with appropriate concentrations of imidazole. After that, proteins were eluted with elution buffer containing 20 mM Tris, pH 8.0, 500 mM NaCl, 1% glycerol, 0.5 mM TCEP and 300 mM imidazole. Proteins were concentrated to 500 μl and loaded onto a Superdex200 Increase 10/300 (GE Healthcare) equilibrated with gel filtration buffer containing 20 mM Tris, pH 8.0, 150 mM NaCl, 1% glycerol and 1 mM TCEP. Peaks containing proteins were collected and concentrated to a small volume. Protein concentrations were determined with microspectrophotometry using the theoretical molar extinction coefficients at 280 nm, and protein purity was evaluated with Coomassie blue staining of SDS-PAGE gels. Samples were flash frozen with liquid nitrogen in aliquots of 40 μl and stored at −80°C until use. All proteins used in this study were purified using the same methods and buffer as described above.
Crystallization and data collection. The ternary complex crystals were obtained by sitting drop vapor diffusion in 0.02 M magnesium chloride hexahydrate, 0.1 M HEPES 7.5, 22% w/v Poly (acrylic acid sodium salt) 5100 at 20°C with a concentration of 5 mg/ml. For data collection, the crystals were rapidly dipped in reservoir solution with 25% ethylene glycol and were flash frozen with liquid nitrogen. X-ray diffraction data were collected at beamline BL18U1 at Shanghai Synchrotron Radiation Facility (SSRF). The diffraction data were processed using HKL2000 and the CCP4 program suite 36 .
Structure determination. The structure was determined by molecular replacement with Phaser 37 using the known REV7 structure (PDB ID 3VU7) 23 as the starting model. Multiple rounds of manual building and refinement were then performed using COOT 38 and PHENIX 39 . Densities of SHLD3 and SHLD2 become more and more clear during the refinement and the model of them were manually built. Diffraction data, refinement statistics, and quality of the structure are summarized in Table 1. The areas of the interfaces were calculated using the PISA server 40 .
Size-exclusion chromatography multi-angle light scattering. SEC-MALS analysis was performed on a high-pressure injection system (Wyatt Technology) and chromatography system equipped with a DAWN HELEOS-II MALS detector and an Optilab T-rEX differential refractive index detector. An aliquot of 100 μl protein at 5 mg/ml was loaded onto a WTC-015S5 column (7.8 × 300 mm, 5 μm, Wyatt Technology) and eluted in buffer (20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% glycerol, 1 mM TCEP, 0.01% NaN 3 ) at a flow rate of 0.4 ml/min. The outputs were analyzed by the ASTRA VI software (Wyatt Technology). The molecular mass was determined using the Astra 6 software program (Wyatt Technology) from the Raleigh ratio calculated by measuring the static light scattering and corresponding protein concentration of a selected peak.
Isothermal titration calorimetry (ITC) measurements. The interactions among the complexes were thermodynamically characterized using isothermal titration calorimetry on an ITC200 instrument (Malvern Instruments). All measurements were done in ITC buffer containing 20 mM Tris pH 8.0, 150 mM NaCl, 1% glycerol and 1 mM TCEP at 25°C. Each titration consisted of 20 successive injections (the first at 0.4 μl and the remaining 19 at 2 μl). The heating power per injection was recorded and plotted as a function of time. The background was deduced either by the last several saturated titrations or by ligand-to-buffer titration (For those Fig. 8 REV3 interacts with SHLD3 mediated C-REV7-O-REV7 and locates at DSB sites. a ITC measurement of interaction between REV7 E35A -SHLD3 (1-82)-REV7 K129A andREV3(1847-1906). REV7 K129A was first saturated by excessive REV7 E35A -SHLD3(1-82) as shown in Fig. 2d. The calculated N and K D are indicated as described in Fig. 2d. b Gel filtration profiles show the interaction between MBP-REV3(1847-1906) and REV7 E35A -SHLD3(1-82)-MBP-REV7 K129A (short as 1906, E35A-1-82 and K129A respectively in the figure). MBP-REV7 K129A was first saturated by excessive REV7 E35A -SHLD3(1-82) as shown in red line. Co-elutions were analyzed by SDS-PAGE and stained by Coomassie brilliant blue. c ITC measurement of interaction between REV7 E35A -SHLD3(1-82) and REV7 K129A -REV3(1847-1906). REV7 K129A was first saturated by excessive REV3 (1847-1906). The calculated N and K D are indicated as described in Fig. 2d. d Gel filtration profiles show the interaction between MBP-REV3(1847-2021) and REV7 WT -SHLD3(1-82) (short as MBP-1847-2021 and WT-1-82 in the figure). Co-elutions were analyzed by SDS-PAGE and stained by Coomassie brilliant blue. e S-tag-HA-SHLD3 was co-expressed with GFP-tagged REV3(1847-2021) (short as GFP-REV3-2021 in the figure) in HEK293T cells. Indicated cells were treated with DMSO or 1 μM doxorubicin for 24 h. S-tag-HA-SHLD3 and its associated proteins were purified by S-protein beads. HA and GFP antibodies were used to detect S-tag-HA-SHLD3 and GFP-tagged REV3(1847-2021), respectively. The input panel shows the transfection efficiency and the S-tag pulldown panel shows the interaction between SHLD3 and REV3(1847-2021). Source data are provided as a Source Data file. f Schematic representation showing conserved domains of human REV3 and the truncation used in laser micro-irradiation assay. The N-terminal domain (NTD) and polymerase domain are shaded in yellow green, REV7 binding motif (RBM) in yellow, and the positively charged domain (PCD) in blue. Domain boundaries are indicated by residue numbers. In the REV3 deletion construct REV3 TR1 , predicted unstructured portions with boundaries marked by residue numbers were deleted. FL: full length. g, h mCherry-SHLD3 and GFP-REV3 localization to DNA damage were monitored after laser micro-irradiation of 293FT and HeLa cells. Cells were imaged 5 min after damage induction except REV3 FL (10 min) due to its large size. n = 2 biologically independent experiments in b, d, e and h, except g was assessed once.
titrations that could not be saturated). The binding isotherms were fitted to a one set of sites model using the MicroCal software. The stoichiometry of binding (N) and the equilibrium-association constant (K A ) were obtained directly. The equilibrium-dissociation constant (K D ) were derived from K A .
Size-exclusion chromatography. Size-exclusion chromatography runs were performed on a Superdex 200 Increase 10/300 GL column (GE healthcare) using buffer (20 mM Tris/HCl pH 8.0, 150 mM NaCl, 1% glycerol and 1 mM TCEP). 400 μg of REV7 WT -SHLD3(1-82), REV7 R124A -SHLD3(1-82) or other mutant heterodimers were mixed with 400 μg of MBP-SHLD2(1-60) and incubated for 10 min on ice prior to the SEC run. For the reconstitution of the conformational dimer, 200 μg of REV7-SHLD3(1-82) mutants were first mixed with 200 μg REV7 K129A to pre-form the complex, and then 400 μg of MBP-SHLD2(1-60) was added prior to the SEC run. The other runs were done by mixing 400 μg of two proteins prior to the SEC run. Samples were all loaded on the same column with a volume of 100 μl. The fractions (at a volume of 0.5 ml) obtained from the SEC runs were analyzed by 13% SDS-PAGE and stained by Coomassie brilliant blue.
Anion exchange chromatography. Prior to the anion exchange chromatography run, the concentration of NaCl was diluted to 80 mM. Buffer A was 20 mM Tris pH 8.0, 20 mM NaCl, 2 mM DTT, buffer B was 20 mM Tris pH 8.0, 500 mM NaCl, 2 mM DTT. The sample was eluted with the following gradient: 0-10% 2 ml, 10-50% 30 ml, 50-100% 10 ml. The fractions (at a volume of 1.0 ml) were analyzed by 13% SDS-PAGE and stained by Coomassie brilliant blue.
Molecular dynamics simulation. The crystal structure of the SHLD3-REV7-SHLD2 complex described in this manuscript was used for molecular dynamics simulations by removing SHLD2. Hydrogen atoms were added by SWISS PDB VIEWER. The system was solvated in a cuboid water box with 8.6 × 7.5 × 10.6 Å 3 buffer and neutralizing counter ions were added. A concentration of 0.15 M NaCl salt bath was introduced to mimic experimental assay conditions. We used the OPLS-AA/L all-atom force field (2001 amino acid dihedrals) parameter set for the protein, and TIP3P model for water. Simulations were performed with GRO-MACS 41 . Prepared systems were first minimized using 5000 steps of a steepest descent algorithm, then equilibrated as follows: the system was heated to 310 K by Nose-Hoover in the isothermal-isobaric (NpT) ensemble over 25 ps. Production runs were then made for 120 ns duration in the NpT ensemble. The short-range electrostatic and Lennard-Jones interactions were calculated within a cut-off of 12 Å. Particle Mesh Ewald was used for long-range electrostatics. Trajectory analysis was conducted with GROMACS. Before processing, the trajectories were aligned to calculate the RMSD between backbone atoms of the initial equilibrated structure and all subsequent frames.
NHEJ assay. Linearized DNA containing the EF1α promoter, the open reading frame of puromycin resistance and the WPRE element was expanded by PCR, using pCDH-CMV-MCS-EF1-Puro (System Biosciences) as template. Indicated cells were transfected with this linearized DNA and the pEGFP-C1 plasmid. Sixty hours later, the cells were collected and flow cytometry analysis was used to detect the efficiency of EGFP expression thus to determine the transfection efficiency. After incubation with medium containing puromycin for 14 days, the cells were fixed by 70% ethanol and stained with 0.1% Coomassie Brilliant Blue for 30 min at room temperature. The stained dishes were washed with water, and the colonies were counted by ImageJ (version 1.52a).
Laser micro-irradiation and imaging of live cells. Thirty-six hours after transfected with indicated plasmids, cells were plated on glass-bottomed dishes (NEST #801002) and sensitized with 5 μM Hoechst 33342 prior to exposure to a 405 nm localized laser beam (100% laser power, 24 s) on an inverted Nikon A1R microscope. Following micro-irradiation, cells were subject to live cell imaging.
Quantification and statistical analysis. Data were tested for statistical significance with GraphPad Prism 5 (Version 5.01). The tests performed, the number of biologically independent replicates (n) for each experiment are indicated in the Figure legends.
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Coordinates and structure factors for the crystal structure of human SHLD3-C-REV7-O-REV7-SHLD2 complex have been deposited into the Protein Data Bank with the accession code 6KTO. Structural details about REV7-REV3-REV1 (PDB ID: 3VU7) and C-Mad2-O-Mad2 (PDB ID: 2V64) are accessible in the Protein Data Bank (PDB). Source data for Figs. 2d-g, 3c-e, 4b-g, 5a-e, 7b-f, 8a-e and Supplementary Figs. 4a, b, 5a, b, 7d-f, 8a-c are provided as a Source Data file. All other data that support the study are available from the corresponding authors upon reasonable request.