MainATR, a member of the phosphoinositide 3 kinase-related protein kinase (PIKK) family, controls the replication checkpoint and is essential for preserving genomic integrity1,2 and cell viability3,4,5. Hypomorphic mutations in ATR cause Seckel syndrome, a developmental disorder associated with sensitivity to DNA-damaging agents6,7. ATR inhibitors have emerged as cancer therapeutics by exploiting oncogene-induced replication stress and genomic instability in tumor cells8,9.The replication checkpoint is triggered by replication-fork stalling due to DNA damage, transcriptional machinery, or other impediments to replication1,2. This results in persistent single-stranded DNA (ssDNA) coated by the heterotrimeric replication protein A (RPA)10, which recruits ATR–ATRIP, at least in part, through an ATRIP–RPA interaction5,11,12. The prototypical ATR activator TOPBP1 is independently recruited to the stalled fork by the Rad9–Hus1–Rad1 (9-1-1) complex, a checkpoint clamp loaded onto double-stranded (dsDNA) adjacent to a 5′-ended dsDNA–ssDNA junction13,14,15. TOPBP1-activated ATR phosphorylates the diffusible kinase CHK1 during replication stress, leading to cell-cycle arrest, suppression of origin firing, and replication-fork stabilization and repair1. By contrast, ETAA1, which is recruited directly by RPA–ssDNA16,17,18,19, activates ATR in unperturbed S phase without causing cell-cycle arrest. This branch of ATR signaling is thought to provide the basal ATR activity needed for suppressing premature mitosis and associated chromosomal instability20,21. ETAA1 also promotes the phosphorylation by ATR of mitotic proteins required for a functional spindle-assembly checkpoint22.The ATR-activating functions of TOPBP1 and ETAA1 reside in ~200-residue segments termed ATR-activating domains (AADs)16,18,19,23,24, which can induce phosphorylation of ATR substrates when overexpressed in cells11,16,18,23,25. The two AADs share little sequence homology except for two conserved aromatic residues that have been shown by mutagenesis to be required for activating ATR–ATRIP16,18,19,23,24. The six human PIKKs share sequence homology across their carboxy-terminal FAT and kinase domains (hereafter KD)26,27. Their larger amino-terminal portions are divergent, consist of at least two α–α solenoid domains formed by HEAT repeats and are often termed N-heat and M-heat28,29. Cryogenic electron microscopy (cryo-EM) structures of human ATR–ATRIP and the yeast ortholog Mec1–Ddc2 (refs. 30,31,32) have shown that they adopt structures analogous to those of other PIKKs33,34, but the mechanisms underlying ATR activation by TOPBP1 and ETAA1 remain unknown. To address these questions, we assembled ATR–ATRIP bound to the AADs of TOPBP1 and ETAA1 to analyze the structures of the complexes using cryo-EM.ResultsCryo-EM analysis of ATR–ATRIP–TOPBP1AADWe first compared the activity of the TOPBP1 AAD (hereafter TOPBP1AAD) with that of full-length TOPBP1, as the latter forms aggregates on cryo-EM grids that preclude cryo-EM analysis. Steady-state kinetic analysis showed that TOPBP1AAD increases the protein kinase catalytic constant (kcat) by ~100-fold, comparable to that of full-length TOPBP1 (Fig. 1a and Extended Data Fig. 1a). This indicates the two proteins induce a substantially similar activated state in ATR–ATRIP, although the half-maximal effective concentration (EC50) of TOPBP1AAD was fivefold higher than that of TOPBP1, likely owing to the lack of a dimerization domain located outside the AAD35.Fig. 1: Cryo-EM structure of apo-ATR–ATRIP.Full size imagea, Comparison of activation of ATR–ATRIP (19 nM) phosphorylating PHAS-I (100 μM) by full-length TOPBP1 and TOPBP1AAD. The graph shows reaction velocity normalized by enzyme concentration (v/[E]) at the indicated concentrations of TOPBP1 and TOPBP1AAD from three independent experiments, and the dose–response curves and EC50 values calculated from non-linear regression fitting of the data with the indicated Hill coefficients (nH). b, Composite map of four focused cryo-EM reconstructions. c, Cartoon representation (containing black sticks representing twofold rotation axes for ATRIP–N-heat and FAT–KD, and AMPPNP–Mg2+ in CPK representation) of apo-ATR–ATRIP from the ATR–ATRIP–TOPBP1AAD data (cryo-EM data processing workflow shown in Extended Data Fig. 2a), colored according to the linear schematics (gray for disordered regions). ATR starts with an N-terminal solenoid (N-heat, 1–1,071), followed by a shorter solenoid (M-heat, 1,072–1,376). It ends with the PIKK-specific FAT domain (1,381–2,206) and the kinase domain (KD, 2,207–2,644). The FAT domain contains an N-terminal helical repeat subdomain (1,381–1,562), in addition to the three canonical tetratricopeptide repeat domains (TRD1, TRD2 and TRD3) and a single HEAT repeat domain (HRD) that surround the kinase domain41. The kinase domain consists of the N and C lobes, with the catalytic cleft and AMPPNP binding site in between the two lobes. The ATRIP N-terminal RPA-binding checkpoint recruitment domain (CRD), the coiled-coil domain (CCD) and the linker connecting them to the ATR-interacting domain are disordered.Source dataWe next collected cryo-EM data on human ATR–ATRIP (1.8 and 4 μM) mixed with TOPBP1AAD (160 μM), as well as on apo-ATR–ATRIP (0.8 μM). In the apo structure, the ATR–ATRIP heterodimer forms a ring-like homodimer through two interfaces (Fig. 1b, Extended Data Fig. 1b–d and Table 1). The ATRIP α–α solenoid (residues 304–780) forms one dimer nestled into the space between the N-heat domains of two ATR protomers; at the opposing end, ATR dimerizes through a FAT–FAT interface. The N-terminal 300 ATRIP residues, which contain the RPA-binding motif and a coiled coil12,36, are disordered. In keeping with the previous low-resolution structure30, human ATR–ATRIP lacks global twofold symmetry. This is caused by the M-heat domain having distinct conformations in the two ATR protomers, with a relative rotation of ~45° about an internal hinge. The hinge splits M-heat into two rigid bodies, M-heat1 and M-heat2, which are linked to N-heat and FAT, respectively. The hinge rotation causes the twofold rotation axis of the ATRIP–N-heat portion of the complex to tilt relative to that of the FAT–KD portion by ~23° (Fig. 1b and Extended Data Fig. 1d).Table 1 Cryo-EM data collection, model refinement and validation statistics for the ATR–ATRIP–TOPBP1AAD data setsFull size tableThe ATR–ATRIP–TOPBP1AAD data exhibited substantially more heterogeneity in three-dimensional (3D) classification (Extended Data Fig. 2a). After 3D refinement of individual classes, one class with 48% of the particles was indistinguishable from the consensus reconstruction of apo-ATR–ATRIP (Extended Data Fig. 2b). The remaining eight classes featured TOPBP1AAD on either one ATR protomer or both. They were distinguished from apo-ATR–ATRIP by their substantially smaller tilt angles, including a class with zero tilt (Extended Data Fig. 2b). Their reduced tilts correlated with M-heat hinge rotations distinct from those of apo-ATR–ATRIP (Extended Data Fig. 2c). The N-heat–ATRIP density was indistinguishable from apo-ATR–ATRIP in all of the TOPBP1AAD-containing classes, as was the FAT-KD density with the notable exception of the zero-tilt class, which exhibited large conformational changes in both the FAT and KD domains. Specifically, the N lobe and C lobe of its KD domain, which form the catalytic cleft, were rotated in an arrangement very similar to that of the active mTOR and DNA-PKcs28,29, for which both inactive and active structures have been characterized (Extended Data Fig. 2d). The zero-tilt class exhibited global twofold symmetry, as determined by a comparison of the densities and structures of the two protomers from cryo-EM data refined without symmetry (Extended Data Fig. 3). The 239,725 particles in the zero-tilt class were thus converted to monomers through C2 symmetry expansion and partial signal subtraction, yielding a 3.2-Å consensus reconstruction. The ATR–ATRIP–TOPBP1AAD model was then refined using a composite map of three focused 3D refinements, each at 3.0-Å resolution (Fig. 2 and Extended Data Fig. 2a).Fig. 2: Cryo-EM structure of ATR–ATRIP–TOPBP1AAD.Full size imagea, Composite map of three focused cryo-EM reconstructions, each at 3.0 Å, of ATR–ATRIP–TOPBP1AAD. b, Cartoon representation of the structure in two orthogonal views colored as in Fig. 1. The twofold rotation axes of the ATRIP–N-heat and FAT-KD portions are represented by black sticks. AMPPNP–Mg2+ is in CPK representation. c, Cartoon representations of one copy of ATR–ATRIP–TOPBP1AAD (colored as in b), and open (gray) and closed (blue) apo-ATR–ATRIP superimposed on their M-heat2, showing the global conformational changes upon TOPBP1 binding. The M-heat hinge is indicated by an arrow.TOPBP1AAD binding to ATR–ATRIPTOPBP1AAD binds to ATR–ATRIP as three discontinuous segments (Fig. 2). The first segment forms the tα1 helix, which binds to the first HEAT repeat of N-heat, packing into an ATR groove formed by conserved residues from the nα1–nα3 helices (Fig. 3a,b and Extended Data Fig. 4). This mostly hydrophobic interface includes the conserved Phe1071 residue, which has been reported to be essential for ATR binding24. The second segment forms the tα2 helix, which packs into a cleft between N-heat and ATRIP, forming an interface of mixed polar and hydrophobic character (Fig. 3c and Extended Data Figs. 4 and 5). The third segment, which forms a β-strand followed by an α-helix (tβ1α3), binds to the nexus of the N-heat, M-heat1 and FAT domains (Fig. 3d,e). The tβ1 strand forms a mixed β-sheet with fβ1 at the N-terminal portion of FAT, stabilized by hydrophobic contacts and additional hydrogen bonds. The tα3 helix binds to the interface between N-heat and M-heat1, with two-thirds of the helix interacting with N-heat and the rest with M-heat1. The tight turn between tβ1 and tα3 contains the highly conserved Trp1145 residue, whose p.W1145R mutation abolishes ATR activation and causes embryonic lethality in mice23,37. Its side chain packs with both M-heat1 and FAT in a buried environment, and it likely has a key role locking these ATR domains together. The binding sites for tα1 and tα2 are preformed in apo-ATR–ATRIP complex, whereas that of tβ1α3 is not.Fig. 3: TOPBP1AAD binding sites and interactions.Full size imagea, One copy of the ATR–ATRIP–TOPBP1AAD complex, highlighting the connectivity of the TOPBP1AAD structural elements with red dashed lines. N- and C-termini are labeled. b–e, Close-up views of the tα1 (b), tα2 (c), tβ1 (d) and tα3 (e) binding sites, showing side chains involved in intermolecular contacts (yellow dotted lines, hydrogen bonds).TOPBP1 allosterically activates ATR–ATRIPThe activating conformational changes are triggered by tβ1α3 binding to N-heat, M-heat1 and FAT. This brings the three ATR domains together in a relative arrangement distinct from both protomers of apo-ATR, in which one protomer has the N-terminal portion of the FAT domain 10 Å away from M-heat1 (hereafter the ‘open’ protomer), whereas the other protomer has FAT packing at the N-heat–M-heat interface in a manner that partially obstructs tβ1α3 binding (‘closed’ protomer; Fig. 4a). The 3D classification indicates that the active state is in equilibrium with TOPBP1AAD-bound inactive states that have intermediate tβ1α3–ATR interface conformations with fewer intermolecular contacts and a smaller buried surface area than in the active state (Extended Data Fig. 6a,b).Fig. 4: TOPBP1AAD allosterically activates ATR–ATRIP.Full size imagea, FAT domain rotations in the open (left) and closed (right) apo-ATR protomers that reconstitute their tβ1α3 binding sites in the active state (center). ATR residues that contact tβ1α3 in the active state are shown as CPK representations in gray. Rotations of the FAT domains, shown with their axes (red sticks) and degrees, were calculated by aligning the M-heat1 portions of the respective ATR structures. b, Cartoon representation of the activating conformational changes, colored as in a. Approximate locations of the hinges are marked by their rotation axes (black lines or filled circles) and degrees. c, FAT-KD portions of active and inactive (gray) ATR superimposed on the KD C lobe. Hinge rotations and their axes (red sticks) are marked. d, Superposition of the kinase domains of the active and inactive (gray) conformations, aligned on their C lobes, shows the movement of the N lobe and its bound AMNPNP. Side chains of key catalytic cleft residues, including Lys2327, which orients the γ phosphate of AMPPNP, and the Mg2+ ligands Asn2480 and Asp2494, are shown. e, Activation of ATR–ATRIP (19 nM) phosphorylating PHAS-I (100 μM) by wild-type or mutant TOPBP1AAD from three independent experiments. Dose–response curves (bottom) could not be fit for mutants that abolished activation or shifted the activation EC50 beyond the highest TOPBP1AAD concentration tested. f, ATR–ATRIP pull-down assay using GST-tagged wild-type (WT) or mutant TOPBP1AAD, and the bar graph from three independent experiments indicating the percentage of ATR recovered relative to the wild type. g, Activation of ATR–ATRIP (19 nM) phosphorylating PHAS-I (100 μM) by wild-type or mutant TOPBP1AAD (128 μM) containing the indicated tα3 mutations, or mutations of ATR-phosphorylated residues. The bar graph from three independent experiments indicates the percentage of phosphorylated PHAS-I relative to the wild type.Source dataThe first TOPBP1AAD binds to the open protomer, where the engagement of tβ1α3 causes the FAT domain to rotate by ~27°; this movement brings the FAT domain in juxtaposition with M-heat1 and N-heat, resulting in an intermediate tβ1α3–ATR interface (Extended Data Fig. 6a). The FAT motion is transmitted through the FAT–FAT dimerization interface to the closed protomer (Fig. 4b), relieving the obstruction of its tβ1α3 site. The binding of the second TOPBP1AAD also creates an intermediate tβ1α3–ATR interface by inducing FAT movement at the closed protomer (Extended Data Fig. 6b). Because the mid-portion of FAT anchors M-heat2, its movements relative to M-heat1 are facilitated by hinge rotations of M-heat, approaching the active conformation.The conversion of the tβ1α3–ATR interfaces from the intermediate to the more extensive active state requires additional movement of the FAT domain relative to N-heat and M-heat1 (Extended Data Fig. 6a,b). This triggers a conformational change in the FAT-KD segment, presumably because the dimeric complex, which traces a closed-ring path through invariant ATRIP–ATRIP and FAT–FAT interfaces, cannot accommodate additional rigid-body motions of the FAT domain.The FAT conformational change occurs cooperatively across hinge regions, demarcating three FAT segments. The N-terminal FAT segment is connected to M-heat1 and to the KD C lobe, the middle segment to M-heat2 and to the second protomer, and the third segment to the KD N lobe (Fig. 4b,c and Extended Data Fig. 6c). Crucially, the FAT conformational change culminates in the relative rotation of the kinase N and C lobes that are anchored on different FAT segments. This realigns active-site residues on the N and C lobes to a catalytically competent configuration that is superimposable on that of the active mTOR structure (Fig. 4d and Extended Data Fig. 6d).The absence of 3D classes in which only one protomer is active indicates that the FAT-KD conformational change occurs cooperatively across the two protomers. This is likely because the two protomers are linked at both the N- and C-terminal portions. Thus, the motion of N-heat of one protomer is transmitted to the other through ATRIP–ATRIP dimerization, while the FAT conformational change is transmitted through FAT–FAT dimerization.Mutational analysis of activation and binding by TOPBP1AADThe structure suggests the model that tα1–tα2 binding to ATR serves primarily to recruit tβ1α3, which then uses its free energy change of binding to pay for the activating conformational change in ATR. To test this model, we first assessed the reported p.F1071A mutation24 on the tα1 helix. It reduced the phosphorylation of the PHAS-I substrate to 43% of the wild-type level when using 128 μM TOPBP1AAD, a saturating concentration for the wild-type protein (Fig. 4e). p.F1071A also substantially reduced TOPBP1AAD binding to 19% of the wild-type level, as determined using a pull-down assay with GST-tagged proteins (Fig. 4f). Mutation of tα2 residues had similar effects, with p.L1123A and p.R1127A reducing kinase activity and binding to 31% and 39%, respectively, of the wild-type levels (Fig. 4e,f).We next tested the p.W1145R mutation23,37 at tβ1α3 (Fig. 4e,f). In keeping with previous reports, this mutation abolished ATR–ATRIP activation to undetectable levels. By contrast, binding to ATR–ATRIP was minimally affected, with 85% of the wild-type TOPBP1AAD levels. This is consistent with Trp1145 having a key role at the M-heat1–FAT interface of the active state but not the intermediate TOPBP1-bound states (Extended Data Fig. 6a,b). We additionally mutated the tβ1 residue Ile1143, which binds to the FAT domain in both the intermediate and active states. The double mutant with p.I1143A and p.W1145A was inactive, as expected. It also caused a further reduction in ATR–ATRIP binding to 55% of the wild-type level (Fig. 4e,f). Nevertheless, the binding was still higher than the mutants of tα1 or tα2, consistent with the model that the energy of binding of tβ1α3 to the active ATR state is what drives the conformational change. As the tα3 helix cooperates with tβ1 in bringing together the N-heat, M-heat and FAT domains of ATR, we tested mutations at tα3 residues that hydrogen bond to N-heat and M-heat residues (Fig. 3e) and found that the p.D1147A, p.E1153A and p.R1154A mutations substantially reduced ATR activation (Fig. 4g). We also mutated Thr1062 and Ser1138, which were reported to be phosphorylated by ATR and the related kinase ATM38,39. In the structure, they are in disordered segments two residues N-terminal to the tα1 and tβ1, respectively, and their mutation to alanine had no effect on ATR activation (Fig. 4g).Cryo-EM structure of ATR–ATRIP bound to ETAA1AADWe next sought to investigate ATR–ATRIP activation by an ETAA1 fragment encompassing the reported AAD16,18,19 (ETAA1AAD), because we could not test full-length ETAA1 owing to limited solubility. Steady-state kinetic analysis showed that ETAA1AAD activates ATR–ATRIP at a much lower level than does TOPBP1AAD, with the dose–response curve plateauing at approximately fivefold activation, compared with ~100-fold for TOPBP1AAD (Fig. 5a). The low activity is not a result of lower affinity, as the 7.1 μM EC50 of activation of ETAA1AAD is comparable to the 10.5 μM EC50 of TOPBP1AAD, and GST-ETAA1AAD recovers 25 % more ATR–ATRIP compared to GST-TOPBP1AAD in pull-down assays (Extended Data Fig. 7a). However, there is substantial evidence that ETAA1 requires phosphorylation for robust activation of ATR, with the individual S95D and S111D phosphomimetic mutations potentiating ATR activation by ectopically expressed ETAA1AAD, and the double mutant acting synergistically20. We thus tested activation by the ETAA1AAD-S95D S111D double mutant and found its dose–response curve plateaued at ~70% of the TOPBP1AAD level, substantially higher than the wild-type ETAA1AAD (Fig. 5a). To further investigate this observation of robust binding and differing levels of activation, we collected cryo-EM data of ATR–ATRIP in the presence of either the wild-type ETAA1AAD or the ETAA1AAD-S95D S111D double mutant, both at ~140 μM (Extended Data Fig. 7b,c and Table 2).Fig. 5: Cryo-EM structure of ATR–ATRIP–ETAA1AAD.Full size imagea, Activation of ATR–ATRIP (19 nM) phosphorylating PHAS-I (100 μM) by wild-type or mutant (p.S95D and p.S111D) ETAA1AAD. A reaction with a saturating concentration of TOPBP1AAD (128 μM) is included for comparison. The graph shows reaction velocity (normalized by enzyme concentration) versus ETAA1AAD concentration from three independent experiments with the indicated markers, and the non-linear regression fit of the ETAA1AAD data. The inset shows the same plot for the wild-type ETAA1AAD data alone, with a truncated y axis to make the low-level activation easier to see. b, Composite map of three focused cryo-EM reconstructions of ATR–ATRIP–ETAA1AAD. c, Cartoon representation of the structure colored as in Fig. 1. Disordered regions of ETAA1AAD in the linear schematic are colored gray. The twofold rotation axes of the ATRIP–N-heat and FAT-KD portions are represented by black sticks. AMPPNP–Mg2+ is in CPK representation. d, One copy of the ATR–ATRIP–ETAA1AAD complex highlighting the connectivity of the ETAA1AAD structural elements with magenta dashed lines. N and C termini are labeled. The view is rotated about the vertical axis relative to b. e–i, Close-up views of the eβ1 (e), eα1 (f), eα2 (g), eα3 (h) and eβ1(mut) (i) binding sites showing the side chains involved in intermolecular contacts (yellow dotted lines, hydrogen bonds). j, Activation of ATR–ATRIP (19 nM) phosphorylating PHAS-I (100 μM) by full-length TOPBP1 (5 μM) in the presence of the indicated ETAA1AAD concentrations, and the corresponding IC50 curve and value calculated from incorporated 32P as fractions of the zero ETAA1AAD reaction from three independent experiments.Source dataTable 2 Cryo-EM data collection, model refinement and validation statistics for the ATR–ATRIP–ETAA1AAD and ATR–ATRIP–ETAA1AAD(mut) data setsFull size tableFor the wild-type ETAA1AAD, initial 3D classification of ~1.5 million particles showed that, in contrast to the TOPBP1AAD data, ETAA1AAD was bound to ATR–ATRIP in all eight classes (Extended Data Fig. 7d). A single symmetric class had four ETAA1AAD segments bound to each of the two protomers. The remaining classes had the same ETAA1AAD density on the closed protomer, but lacked density for two ETAA1AAD segments on the open protomer (Extended Data Fig. 7d). However, none of the classes had detectable conformational changes in the FAT-KD segment. This is overall consistent with robust binding but little activation. We presume that the low fraction of particles that might have undergone the activating FAT-KD conformational change is likely below the detection limit. The particles of the symmetric class were refined in point group C2 to yield a reconstruction at 3.3-Å resolution, and a composite map of subsequent focused refinements was used to refine the ATR–ATRIP–ETAA1AAD model (Fig. 5b and Extended Data Fig. 7b).The structure reveals that ETAA1AAD binding to ATR is substantially different from TOPBP1AAD binding to ATR. ETAA1AAD has four discontinuous segments that form a β-strand (eβ1) and three α-helices (eα1, eα2, eα3; Fig. 5c,d and Extended Data Fig. 8a,b). Although the eβ1 and eα3 elements have ATR-binding sites analogous to TOPBP1 tβ1 and tα1, the eα1 and eα2 helices bind to distinct sites, with the TOPBP1 tα2 and tα3 sites remaining unoccupied in the ETAA1 complex. In addition, the chain direction of ETAA1 is reversed, with eβ1 and eα3 located at the ETAA1AAD N and C termini, respectively.The eβ1 strand forms a mixed β-sheet with the N-terminal portion of FAT, like TOPBP1 tβ1, with Ile105 and Trp107 (refs. 16,18,19), analogous to the TOPBP1 Ile1143 and Trp1145 residues, packing with hydrophobic FAT side chains (Fig. 5e). However, unlike TOPBP1, the ETAA1 polypeptide chain becomes disordered after the end of eβ1. As eβ1 is not structurally coupled to an α-helix that can bind to N-heat and M-heat1, its binding does not cause the conformational change of the TOPBP1 tβ1α3 that causes FAT to move towards N-heat and M-heat1. Trp107 interacts with only the FAT domain, with its putative binding site on M-heat1 remaining ~17 Å away. After eβ1, a 19-residue unstructured segment leads to eα1, which uses 6 hydrophobic side chains to bind to the N-heat–M-heat interface, opposite from where the TOPBP1 tα3 binds (Fig. 5f and Extended Data Fig. 8c). The eα2 helix binds to a groove near the C terminus of N-heat, at a discontinuity in the regular arrangement of HEAT repeats (Fig. 5g). The eα3 helix binds to the same site as TOPBP1 tα1, but in the opposite N–C orientation. Nevertheless, eα3 contains a conserved phenylalanine, Phe198, that inserts into the same pocket as the TOPBP1 Phe1071 (Fig. 5h).We presume that the standalone nature of eβ1 fails to coalesce the N-heat, M-heat1 and FAT domains in a way that adequately stresses the FAT domain. Yet, some conformational stress must exist, as the M-heat hinges of both protomers assume a closed-like conformation (Extended Data Fig. 8d), a state that is not observed with the apo complex. The mechanism underlying the induction of this state seems to involve the binding of the ETAA1-specific eα1 and eα2 helices to hinge regions, inducing conformations distinct from apo- or TOPBP1AAD-bound ATR (Extended Data Fig. 7d).This altered conformational landscape might give rise to the transient coupling of FAT-bound eβ1 and M-heat1 that can trigger a reversible FAT-KD conformational change.For ETAA1AAD-S95D S111D, the data sets yielded ~1.3 million particles (Extended Data Fig. 7c and Table 2). Three-dimensional classification exhibited a conformational landscape and overall AAD density similar to the wild-type ETAA1AAD. One class with the smallest tilt exhibited density in both protomers at all four of the sites observed in the wild-type ETAA1AAD complex, whereas the same density was observed in the remaining classes on only one protomer, with the other protomer lacking density for eβ1 and eα2. As with the wild-type ETAA1AAD, we could not detect conformational changes in the FAT-KD segment in any of these classes, even after classification with signal subtraction outside FAT-KD. This suggests that the number of particles with an active FAT-KD segment is below the detection limit of our calculations.In the wild-type ETAA1AAD complex, Ser95 and Ser111 are in disordered segments N- and C-terminal of eβ1, respectively. In ETAA1AAD-S95D S111D, Asp95 remains disordered, with the density at the N terminus of eβ1 starting at residue 102 as is the case in the wild-type complex. At the C terminus of eβ1, by contrast, there is clear density extending for five additional residues (residues 113–117) (Extended Data Fig. 8a). To better evaluate the extended density, we performed focused refinement at the protomer with full eβ1 occupancy using the entire data set. The 2.9-Å reconstruction shows that the carboxylate group of Asp111 forms intramolecular hydrogen bonds with the backbone amide groups of Leu113 and Thr114, stabilizing an α-helical conformation that starts with Pro112 and extends to Leu117 (Fig. 5i). The helical extension, together with a conformational change in eβ1 residues 109–112, results in contacts with the ATR N-heat domain, which shifts closer to eβ1 compared with the wild-type complex devoid of N-heat contacts (Fig. 5i). However, these N-heat contacts are fewer and distinct from those made by the TOPBP1 tα3. In addition, there are no contacts to M-heat, which remains more than ~15 Å away relative to its position in the TOPBP1-bound active complex, the N-heat has not shifted position relative to FAT, and the FAT-KD conformation remains in the apo state.Because Asp111 is an imperfect mimetic of phospho-Ser111, it is conceivable that the latter will bring about additional conformational changes that would trigger the activating changes in FAT-KD more effectively. In this respect, we note that the Asp111 side chain is within hydrogen bonding distance, although with poor geometry, of Arg1461 on the ATR FAT domain. It is not clear how phospho-Ser111 would alter this interface, although it is unlikely to extend the short α-helix, as the last ordered AAD residue, Leu117, is followed by the Gly-Lys-Gly sequence.The robust binding of the wild-type ETAA1AAD to ATR that is apparent in the biochemical and cryo-EM data, coupled with the overlap with TOPBP1 in two of the four binding sites, indicates that ETAA1AAD should compete with TOPBP1 for binding to ATR–ATRIP, raising the possibility that unphosphorylated ETAA1 could act as an antagonist of TOPBP1 for activating ATR. To test this hypothesis, we added increasing amounts of ETAA1AAD to a phosphorylation reaction containing either 5 μM full-length TOPBP1 or 24 μM TOPBP1AAD. ETAA1AAD reduced phosphorylation by ~50% at two-thirds and one-third molar ratio to TOPBP1 and TOPBP1AAD, respectively (Fig. 5j and Extended Data Fig. 8e,f). Phosphorylation plateaued at a level of approximately 10% of the reactions lacking ETAA1AAD.DiscussionThe structure of the ATR–ATRIP–TOPBP1AAD complex shows that TOPBP1 binds to ATR in three discontinuous segments. The tα1 and tα2 segments bind to preformed sites on ATR and function primarily in recruiting TOPBP1 to ATR. The binding free energy of the tβ1α3 segment, which binds to and brings together the N-heat, M-heat and FAT domains, is used to induce conformational changes in the FAT domain. As the FAT is clamped onto the N and C lobes of the KD, the FAT conformational change is coupled to a relative rotation of the two kinase lobes, bringing their active-site residues into proper alignment for phosphotransfer. The activating FAT-KD conformational change is conserved in mTOR29 and DNA-PKcs28. Although the details differ in the three PIKKs, they share a common trigger point in the middle portion of the FAT domain (Extended Data Fig. 6c). This is where the M-heat domain of ATR and the N-heat domains of mTOR and DNA-PKcs allosterically pull on the FAT domain as a result of distal events initiated by their respective activator proteins.The FAT has been suggested to be an auto-inhibitory clamp that keeps the N and C lobes misaligned, given that activating mTOR mutations map to the interfaces between the FAT and KD domains or between the N and C lobes29. In this respect, the partially activated yeast Mec1-F2244L mutant31 could be analogous to the activating mTOR mutations that map to the N–C lobe interface. In ATR, the corresponding residue (Phe2495) is on the activation loop of the C lobe and packs with the N lobe inside the catalytic cleft. The p.F2244L mutation in Mec1 results in altered N–C lobe packing and an activation-loop conformation distinct from the active conformations of ATR and other PIKKs (Extended Data Fig. 8g).An earlier study utilizing yeast two-hybrid screens40 implicated the ATRIP dimer interface and kinase domain in TOPBP1AAD binding. However, our reconstructions do not exhibit any additional density that could be attributable to TOPBP1AAD in these regions. The suggested binding site on the kinase domain, termed the PIKK regulatory domain (PRD), includes the kα9 and kα10 helices connected by a segment that is disordered in both apo- and TOPBP1AAD-bound ATR–ATRIP; the conserved Lys2589 residue, whose mutation has been reported to disrupt TOPBP1AAD-mediated activation, is solvent-exposed on the kα10 helix and is not involved in any interactions. The analogous segment in mTOR comprises a helix that is also unchanged in both the active and inactive states29. In DNA-PKcs, by contrast, the segment between kα9 and kα10 contains two helices that pack against and occlude the substrate-binding site but are displaced following the activating conformational changes of the FAT-KD28. In this respect, the ATR PRD segment is functionally analogous to that of mTOR, without an apparent role in TOPBP1-mediated activation.Our cryo-EM analysis shows that the ATR–ATRIP–TOPBP1AAD complex exists in an equilibrium between the inactive and active states. The inactive TOPBP1-bound state is characterized by an intermediate TOPBP1–ATR interface whose conversion to a more extensive interface triggers a FAT conformational change that allows the kinase domain to assume the active configuration. With the unphosphorylated ETAA1AAD complex, our biochemical and structural data indicate that the equilibrium is substantially shifted to the inactive state owing to the distinct arrangements of the ETAA1 structural motifs that bind to ATR. The p.S95D and p.S111D phosphomimetic substitutions significantly increase the ATR activation potential of ETAA1AAD, as previously reported20, and the structure suggests that this is primarily due to the p.S111D substitution. Although we did not capture the active conformation of ATR bound to ETAA1AAD-S95D S111D, the structure suggests a disorder-to-order transition at the end of eβ1 as a contributory mechanism for phospho-Ser111 potentiating ATR activation. The disparate activation and competitive ATR binding of TOPBP1 and ETAA1 suggest a potential mechanism of tuning ATR activity for different signaling contexts and outcomes.MethodsProtein expression and purificationThe human ATR–ATRIP complex was expressed in a stably transfected HEK293F cell line generated from a pcDNA3.1 vector encoding ATR and ATRIP, with each containing their own CMV promoter, N-terminal FLAG tag and poly-A sequences. The FLAG-tagged complex was purified by affinity chromatography with anti-FLAG M2 agarose beads (Sigma) and anion exchange chromatography on a Mono Q 5/50 GL column (Cytiva). For cryo-EM sample preparation, the complex was concentrated to ~2–5 mg ml−1 in 20 mM Bis-Tris propane-HCl pH 8.0, ~250 mM NaCl, 5% glycerol and 0.5 mM tris-(2-carboxyethyl) phosphine (TCEP). For kinase assays, the complex was additionally purified by size exclusion chromatography on a Superose 6 Increase 10/300 GL column (Cytiva), concentrated to ~0.4 mg ml−1 and stored in 20 mM Tris-HCl pH 8.0, 250 mM NaCl, 10% glycerol and 10 mM dithiothreitol (DTT). Wild-type and mutant TOPBP1AAD (residues 978–1192) and PHAS-I were produced in Escherichia coli strain BL21(DE3) as N-terminally GST-tagged proteins. They were purified through affinity chromatography with glutathione Sepharose 4B resin (Cytiva), removal of the GST tag by TEV protease and anion exchange chromatography on a Mono Q 5/50 GL column (Cytiva). The GST-tagged TOPBP1AAD proteins used in pull-down assays contained a C-terminal StrepII tag (WSHPQFEK) and were additionally purified with Strep-Tactin XT resin (IBA Lifesciences) to improve purity. PHAS-I was concentrated to 14 mg ml−1 in 20 mM Tris-HCl pH 8.0, 95 mM NaCl, 5% glycerol and 0.5 mM TCEP. Following dialysis into a buffer containing 20 mM Tris-HCl pH 8.0, 100 mM NaCl, 5% glycerol and 0.5 mM TCEP, GST-tagged TOPBP1AAD proteins were concentrated to 5–7 mg ml−1 and their untagged versions to 20–22 mg ml−1. Full-length TOPBP1 was overexpressed in E. coli strain Rosetta(DE3) (Novagen) and purified similarly to TOPBP1AAD, except for the addition of a size exclusion chromatography step using a Superdex 200 Increase 10/300 GL column (Cytiva). The protein was concentrated to ~14 mg ml−1 and stored in 20 mM Tris-HCl pH 8.0, 180 mM NaCl, 5% glycerol and 0.5 mM TCEP. Wild-type and mutant ETAA1AAD (residues 50–250, GST-tagged and untagged) containing a C-terminal StrepII tag were expressed and purified similarly to TOPBP1AAD. Following dialysis into the same buffer as TOPBP1AAD, the GST-tagged protein was concentrated to ~19 mg ml−1 and the untagged protein to 10–12 mg ml−1. All purified proteins were aliquoted, flash-frozen in liquid nitrogen and stored at −80 °C.In vitro kinase assaysKinase assays were carried out in 25 mM HEPES pH 7.4, 100 mM NaCl, 5 mM MgCl2, 5% glycerol and 2 mM DTT, containing 19 nM ATR–ATRIP, 100 μM PHAS-I substrate and 1 mg ml−1 native bovine serum albumin (Sigma). Reactions (8.5 μl volume total) containing varying concentrations of activator were assembled and incubated on ice for 10 min. They were initiated by the addition of 1.5 μl of a mixture of ATP (final 0.5 mM) and [γ-32P]ATP (final 4 μCi) (6,000 Ci mmol−1, Perkin Elmer and Revvity). All reactions were performed at 30 °C for 20–30 min and stopped by the addition of 20 μl of 1.5× NuPAGE LDS sample buffer supplemented with TCEP (final 25 mM) and EDTA (final 30 mM). One-third of each reaction (10 μl) was resolved on 4–12% NuPAGE Bis-Tris gels, which were dried on DE81 paper (Whatman) and exposed to a phosphor imaging plate (Fuji) for image acquisition on a Typhoon FLA 7000 (GE Healthcare) or Amersham Typhoon imager (Cytiva). Bands were quantified using ImageJ (NIH) software42. Curve fitting for the calculation of steady-state kinetic constants and EC50 values was performed using the GraphPad Prism software. Given that the cryo-EM analysis indicates that fully TOPBP1AAD-bound ATR is in an equilibrium of active and inactive conformations, the intrinsic kcat value of active ATR is likely underestimated.GST pull-down assaysFor binding assays, a master mix was prepared containing binding buffer (25 mM HEPES pH 7.4, 150 mM NaCl, 5% glycerol, 0.5 mM TCEP, 5 mM MgCl2, 0.5 mM AMPPNP, 0.01% NP-40) and 60 nM ATR–ATRIP. To tubes containing 5 μl magnetic glutathione beads (Pierce), 50 μl of master mix was aliqouted and 2 μM GST, GST-tagged ETAA1AAD or TOPBP1AAD was added. The mixtures were incubated for 30 min on ice and mixed by gentle flicking of the microcentrifuge tubes every few minutes. The beads were washed three times with 200 μl wash buffer (binding buffer containing 0.1 mM AMPPNP instead) with gentle vortexing, and the samples were eluted with 30 μl wash buffer supplemented with 50 mM glutathione. The eluted proteins were resolved on 4–12% NuPAGE Bis-Tris gels and visualized with Coomassie blue staining, and bands corresponding to ATR were quantified using ImageJ software.Cryo-EM sample preparation and data collectionThe apo-ATR–ATRIP cryo-EM sample was diluted to 0.3 mg ml−1 (~0.8 μM) with a buffer containing 20 mM Tris-HCl pH 8.0, 250 mM NaCl and 0.5 mM TCEP, and supplemented with 1 mM AMPPNP and 2 mM MgCl2. The cryo-EM samples containing TOPBP1AAD were prepared by mixing ATR–ATRIP with an excess of TOPBP1AAD and AMPPNP, followed by dilution with a buffer containing 20 mM Tris-HCl pH 8.0, 5 mM MgCl2 and 0.5 mM TCEP to reduce NaCl concentration to ~80 mM. The final sample contained 0.7 mg ml−1 (~1.8 μM) or 1.6 mg ml−1 (~4 μM) ATR–ATRIP, 160 μM TOPBP1AAD and 1 mM AMPPNP. The cryo-EM samples containing ETAA1AAD (142 μM final concentration) or the ETAA1AAD mutant (137 μM final concentration) were prepared similarly, except the salt composition of the buffer was adjusted to either 100 mM NaCl or a mixture of 50 mM NaCl and 125 mM potassium acetate, which were screened to mitigate protein aggregation on cryo-EM grids. Additionally, to improve particle distribution of the wild-type and mutant ETAA1AAD samples, the mixtures were supplemented with 0.05% CHAPSO and centrifuged at high speed immediately before sample vitrification. Cryo-EM grids were prepared with an FEI Vitrobot Mark IV by applying 4 μl sample to glow-discharged UltrAuFoil 300 mesh R1.2/1.3 grids (Quantifoil). Grids were blotted for 2 s at 22 °C and 100% humidity, and plunged into liquid ethane cooled by liquid nitrogen. Two ATR–ATRIP-TOPBP1AAD data sets (containing 8,252 and 8,202 videos), three ATR–ATRIP-ETAA1AAD data sets (containing 8,575, 9,220 and 12,825 videos) and two ATR–ATRIP-ETAA1AAD-S95D S111D mutant (hereafter ETAA1AAD(mut)) data sets (9,754 and 15,753 videos) were collected using an FEI Titan Krios G2 300 kV microscope equipped with a Gatan K3 camera (MSKCC Cryo-EM facility). Videos were acquired using SerialEM43 with a 0.532 Å super-resolution pixel size and a 0.3–1.8 μm defocus range. Each 3-s exposure, dose-fractionated into 40 frames, was collected with a 20 e− pixel−1 s−1 dose rate and contained a total dose of 53 e− Å−2.Cryo-EM data processing and structure determinationATR–ATRIP–TOPBP1AAD data setsThe 16,454 super-resolution videos from the ATR–ATRIP–TOPBP1AAD data sets were gain-corrected, Fourier-cropped by a factor of two (to a pixel size of 1.064 Å) and aligned using UCSF MotionCor2 (ref. 44). Contrast transfer function (CTF) parameters were estimated using CTFFIND4 (ref. 45). All classifications, 3D refinements, local resolution estimations and other processing steps were carried out in RELION 3.0 (ref. 46). The two data sets were initially processed separately. Particles were autopicked using 2D templates generated from an initial analysis of the data that used Laplacian-of-Gaussian autopicking. Two rounds of 2D classification were performed to remove ice contaminants and false positives, followed by two rounds of Bayesian polishing and CTF refinement. Initial 3D classification yielded apo-ATR–ATRIP particles as well as a series of heterogeneous classes with global conformations distinct from the apo state; this classification also produced classes lacking density in parts of the ATRIP–N-heat portion, and particles from these classes were excluded from further analysis. The apo-ATR–ATRIP particles from the two data sets (1,697,213 particles total) were refined in C1 to yield a consensus reconstruction extending to 2.5-Å resolution as determined from the gold-standard Fourier shell correlation (FSC) procedure, and four partially overlapping soft masks were used to generate focused refinements as shown in Extended Data Fig. 1c. Particles of classes that differed from the apo conformation were combined (1,842,700 particles total) for an additional round of 3D classification. These classes belong to a continuum of conformational states, varying in the orientation of the twofold symmetry axis of N-heat–ATRIP relative to that of the FAT–KD. The tilt roughly correlated with the presence of additional density (attributable to the TOPBP1AAD) on one or both protomers, with the most tilt seen in singly occupied classes and the least tilt seen in doubly bound classes. The 3D classes with low-tilts were further analyzed by 3D sub-classification with partial signal subtraction outside a FAT-KD mask. The only 3D class that contained detectable particles with the activating FAT-KD conformational change was the zero-tilt, C2 symmetric class. The 239,725 particles belonging to the zero-tilt 3D class were refined in point group C2 to 3.2 Å and converted to monomers by symmetry expansion and signal subtraction. The resulting 479,450 monomer particles were then 3D-refined to yield a consensus reconstruction extending to 3.2 Å. Three partially overlapping soft masks were used to generate focused refinements, each at 3.0-Å resolution, as shown in Extended Data Fig. 2a. The individual focused maps were aligned to their respective consensus maps by obtaining the rotation-translation matrices in UCSF Chimera47 and applying them in CCP4 to generate a composite map using the composite sfcalc option in REFMAC5 (ref. 48). The composite maps were then used for model building, real-space refinement and map-to-model validations. On the basis of their sequence and structure homology, the FAT and kinase domains from the crystal structure of N-terminally truncated mTOR41 were docked as rigid bodies into the apo-ATR–ATRIP map in UCSF Chimera, and the model was manually adjusted to fit the density in Coot49. The ATR N-heat and M-heat domains and ATRIP were built de novo, as the previous 4.7-Å-resolution apo-ATR–ATRIP reconstruction30 lacked sufficient local resolution for side chain assignment, and the corresponding regions in the yeast Mec1–Ddc2 structure share little sequence homology with the human complex31. The TOPBP1AAD-bound intermediate and active-state models were built by manually docking in domains of apo-ATR–ATRIP to their respective composite maps as rigid bodies in UCSF Chimera. The models underwent iterative rounds of manual building in Coot, real-space refinement and validation with MolProbity50 as implemented in PHENIX51. ATR and ATRIP each contain a structural Zn2+ ion coordinated by Cys716, Cys753, His788 and His792 of ATR, and Cys624, Cys630, Cys680 and Cys682 of ATRIP.ATR–ATRIP–ETAA1AAD ATR–ATRIP–ETAA1AAD(mut) data setsThe three ATR–ATRIP–ETAA1AAD data sets were initially processed in CryoSPARC v4.3 (ref. 52). Videos were gain-corrected and Fourier-cropped by a factor of two and aligned by patch motion correction, and CTF parameters were estimated by patch CTF estimation. Particles were picked by blob-based and template-based autopicking, and clean particles from 2D classification were used to train a model for additional picking by the Topaz neural network program53 as implemented in CryoSPARC. One round of 2D classification was used to remove only hole edges and ice contaminants, and remaining particles were further processed by heterogeneous refinement using the following references: apo-ATR–ATRIP; a map containing only the FAT–KD portion used to capture particles lacking ATRIP–N-heat density; three noise classes generated from a small number of particles after the first few iterations of ab initio reconstruction used to remove false positives and low-quality particles. Particles classified to the FAT–KD and noise classes were discarded, and heterogeneous refinement was iterated until >90 % of remaining particles classified into the apo-ATR–ATRIP class. The particle metadata were then converted to a RELION STAR file using the UCSF pyem csparc2star.py script54. Particles were extracted in RELION 4.0 or 5.0beta following motion correction by the program’s own CPU implementation of UCSF MotionCor2 and CTF estimation by CTFFIND4. Particles were assigned to different optical groups on the basis of their grid of origin and image-shift positions (indicated by the movie file names from SerialEM), underwent two rounds of Bayesian polishing and CTF refinement, and subsequently classified by 3D classification, as in the ATR–ATRIP–TOPBP1AAD data sets. The 174,280 particles belonging to the zero-tilt 3D class were refined in C2 to 3.3 Å. Three partially overlapping soft masks were used to generate focused refinements, which were subsequently aligned to the consensus C2 map to generate a composite map using the combine_focused_maps tool in PHENIX. The model was built by manually docking the domains of apo-ATR–ATRIP to the composite map as rigid bodies in UCSF Chimera, and it then underwent iterative rounds of manual building in Coot, real-space refinement and validation with MolProbity as implemented in PHENIX. The two ATR–ATRIP–ETAA1AAD(mut) data sets were processed similarly to the wild-type ETAA1AAD data sets. Alignment-free, masked 3D classification of the FAT–KD segment following partial signal subtraction on ATR–ATRIP–ETAA1AAD wild-type and mutant particles, which could identify active-state ATR–ATRIP–TOPBP1AAD particles, failed to identify particles with a FAT–KD dissimilar to that of apo-ATR–ATRIP.Reporting summaryFurther information on research design is available in the Nature Portfolio Reporting Summary linked to this article.Data availabilityCryo-EM maps and models generated in this study have been deposited to the Electron Microscopy Data Bank (EMDB) and Protein Data Bank (PDB) with the following accession codes: EMD-74045 and 9ZCY for apo-ATR–ATRIP; EMD-74048 and 9ZD1 for ATR–ATRIP–TOPBP1AAD; EMD-74050 and 9ZD3 for ATR–ATRIP–ETAA1AAD; and EMD-74051 and 9ZD4 for the symmetric class of ATR–ATRIP–ETAA1AAD; EMD-76990 and 13DE for ATR–ATRIP–TOPBP1AAD(mut). Source data are provided with this paper.ReferencesSaldivar, J. C., Cortez, D. & Cimprich, K. A. The essential kinase ATR: ensuring faithful duplication of a challenging genome. Nat. Rev. Mol. Cell Biol. 18, 622–636 (2017).Article CAS PubMed PubMed Central Google Scholar Zeman, M. K. & Cimprich, K. A. Causes and consequences of replication stress. Nat. Cell Biol. 16, 2–9 (2014).Article CAS PubMed PubMed Central Google Scholar Brown, E. J. & Baltimore, D. ATR disruption leads to chromosomal fragmentation and early embryonic lethality. Genes Dev. 14, 397–402 (2000).Article CAS PubMed PubMed Central Google Scholar Cortez, D., Guntuku, S., Qin, J. & Elledge, S. J. ATR and ATRIP: partners in checkpoint signaling. Science 294, 1713–1716 (2001).Article CAS PubMed Google Scholar Zou, L. & Elledge, S. J. Sensing DNA damage through ATRIP recognition of RPA-ssDNA complexes. Science 300, 1542–1548 (2003).Article CAS PubMed Google Scholar Alderton, G. K. et al. Seckel syndrome exhibits cellular features demonstrating defects in the ATR-signalling pathway. Hum. Mol. Genet. 13, 3127–3138 (2004).Article CAS PubMed Google Scholar O’Driscoll, M., Ruiz-Perez, V. L., Woods, C. G., Jeggo, P. A. & Goodship, J. A. A splicing mutation affecting expression of ataxia-telangiectasia and Rad3-related protein (ATR) results in Seckel syndrome. Nat. Genet. 33, 497–501 (2003).Article PubMed Google Scholar Hall, A. B. et al. Potentiation of tumor responses to DNA damaging therapy by the selective ATR inhibitor VX-970. Oncotarget 5, 5674–5685 (2014).Article PubMed PubMed Central Google Scholar Qiu, Z., Oleinick, N. L. & Zhang, J. ATR/CHK1 inhibitors and cancer therapy. Radiother. Oncol. 126, 450–464 (2018).Article CAS PubMed Google Scholar Byun, T. S., Pacek, M., Yee, M. C., Walter, J. C. & Cimprich, K. A. Functional uncoupling of MCM helicase and DNA polymerase activities activates the ATR-dependent checkpoint. Genes Dev. 19, 1040–1052 (2005).Article CAS PubMed PubMed Central Google Scholar Ball, H. L. et al. Function of a conserved checkpoint recruitment domain in ATRIP proteins. Mol. Cell. Biol. 27, 3367–3377 (2007).Article CAS PubMed PubMed Central Google Scholar Ball, H. L., Myers, J. S. & Cortez, D. ATRIP binding to replication protein A-single-stranded DNA promotes ATR-ATRIP localization but is dispensable for Chk1 phosphorylation. Mol. Biol. Cell 16, 2372–2381 (2005).Article CAS PubMed PubMed Central Google Scholar Delacroix, S., Wagner, J. M., Kobayashi, M., Yamamoto, K. & Karnitz, L. M. The Rad9–Hus1–Rad1 (9-1-1) clamp activates checkpoint signaling via TopBP1. Genes Dev. 21, 1472–1477 (2007).Article CAS PubMed PubMed Central Google Scholar Lee, J., Kumagai, A. & Dunphy, W. G. The Rad9–Hus1–Rad1 checkpoint clamp regulates interaction of TopBP1 with ATR. J. Biol. Chem. 282, 28036–28044 (2007).Article CAS PubMed Google Scholar Majka, J., Binz, S. K., Wold, M. S. & Burgers, P. M. Replication protein A directs loading of the DNA damage checkpoint clamp to 5′-DNA junctions. J. Biol. Chem. 281, 27855–27861 (2006).Article CAS PubMed Google Scholar Bass, T. E. et al. ETAA1 acts at stalled replication forks to maintain genome integrity. Nat. Cell Biol. 18, 1185–1195 (2016).Article CAS PubMed PubMed Central Google Scholar Feng, S. et al. Ewing tumor-associated antigen 1 interacts with replication protein A to promote restart of stalled replication forks. J. Biol. Chem. 291, 21956–21962 (2016).Article CAS PubMed PubMed Central Google Scholar Haahr, P. et al. Activation of the ATR kinase by the RPA-binding protein ETAA1. Nat. Cell Biol. 18, 1196–1207 (2016).Article CAS PubMed Google Scholar Lee, Y. C., Zhou, Q., Chen, J. & Yuan, J. RPA-binding protein ETAA1 is an ATR activator involved in DNA replication stress response. Curr. Biol. 26, 3257–3268 (2016).Article CAS PubMed PubMed Central Google Scholar Achuthankutty, D. et al. Regulation of ETAA1-mediated ATR activation couples DNA replication fidelity and genome stability. J. Cell Biol. 218, 3943–3953 (2019).Article CAS PubMed PubMed Central Google Scholar Saldivar, J. C. et al. An intrinsic S/G2 checkpoint enforced by ATR. Science 361, 806–810 (2018).Article CAS PubMed PubMed Central Google Scholar Bass, T. E. & Cortez, D. Quantitative phosphoproteomics reveals mitotic function of the ATR activator ETAA1. J. Cell Biol. 218, 1235–1249 (2019).Article CAS PubMed PubMed Central Google Scholar Kumagai, A., Lee, J., Yoo, H. Y. & Dunphy, W. G. TopBP1 activates the ATR–ATRIP complex. Cell 124, 943–955 (2006).Article CAS PubMed Google Scholar Thada, V. & Cortez, D. Common motifs in ETAA1 and TOPBP1 required for ATR kinase activation. J. Biol. Chem. 294, 8395–8402 (2019).Article CAS PubMed PubMed Central Google Scholar Toledo, L. I., Murga, M., Gutierrez-Martinez, P., Soria, R. & Fernandez-Capetillo, O. ATR signaling candrive cells into senescence in the absence of DNA breaks. Genes Dev. 22, 297–302 (2008).Article CAS PubMed PubMed Central Google Scholar Abraham, R. T. PI 3-kinase related kinases: ‘big’ players in stress-induced signaling pathways. DNA Repair 3, 883–887 (2004).Article CAS PubMed Google Scholar Keith, C. T. & Schreiber, S. L. PIK-related kinases: DNA repair, recombination, and cell cycle checkpoints. Science 270, 50–51 (1995).Article CAS PubMed Google Scholar Chen, X. et al. Structure of an activated DNA-PK and its implications for NHEJ. Mol. Cell 81, 801–810 (2021).Article CAS PubMed Google Scholar Yang, H. et al. Mechanisms of mTORC1 activation by RHEB and inhibition by PRAS40. Nature 552, 368–373 (2017).Article CAS PubMed PubMed Central Google Scholar Rao, Q. et al. Cryo-EM structure of human ATR-ATRIP complex. Cell Res. 28, 143–156 (2018).Article CAS PubMed Google Scholar Tannous, E. A., Yates, L. A., Zhang, X. & Burgers, P. M. Mechanism of auto-inhibition and activation of Mec1(ATR) checkpoint kinase. Nat. Struct. Mol. Biol. 28, 50–61 (2021).Article CAS PubMed Google Scholar Wang, G. et al. Molecular architecture and inhibition mechanism of human ATR-ATRIP. Sci. Bull. 70, 2137–2146 (2025).Article CAS Google Scholar Baretic, D. et al. Structural insights into the critical DNA damage sensors DNA-PKcs, ATM and ATR. Prog. Biophys. Mol. Biol. 147, 4–16 (2019).Article CAS PubMed Google Scholar Imseng, S., Aylett, C. H. & Maier, T. Architecture and activation of phosphatidylinositol 3-kinase related kinases. Curr. Opin. Struct. Biol. 49, 177–189 (2018).Article CAS PubMed Google Scholar Kim, A. et al. Biochemical analysis of TOPBP1 oligomerization. DNA Repair 96, 102973 (2020).Article CAS PubMed PubMed Central Google Scholar Ball, H. L. & Cortez, D. ATRIP oligomerization is required for ATR-dependent checkpoint signaling. J. Biol. Chem. 280, 31390–31396 (2005).Article CAS PubMed PubMed Central Google Scholar Zhou, Z. W. et al. An essential function for the ATR-activation-domain (AAD) of TopBP1 in mouse development and cellular senescence. PLoS Genet. 9, e1003702 (2013).Article CAS PubMed PubMed Central Google Scholar Liu, S. et al. ATR autophosphorylation as a molecular switch for checkpoint activation. Mol. Cell 43, 192–202 (2011).Article CAS PubMed PubMed Central Google Scholar Yoo, H. Y., Kumagai, A., Shevchenko, A., Shevchenko, A. & Dunphy, W. G. Ataxia-telangiectasia mutated (ATM)-dependent activation of ATR occurs through phosphorylation of TopBP1 by ATM. J. Biol. Chem. 282, 17501–17506 (2007).Article CAS PubMed Google Scholar Mordes, D. A., Glick, G. G., Zhao, R. & Cortez, D. TopBP1 activates ATR through ATRIP and a PIKK regulatory domain. Genes Dev. 22, 1478–1489 (2008).Article CAS PubMed PubMed Central Google Scholar Yang, H. et al. mTOR kinase structure, mechanism and regulation. Nature 497, 217–223 (2013).Article CAS PubMed PubMed Central Google Scholar Schneider, C. A., Rasband, W. S. & Eliceiri, K. W. NIH Image to ImageJ: 25 years of image analysis. Nat. Methods 9, 671–675 (2012).Article CAS PubMed PubMed Central Google Scholar Mastronarde, D. N. Automated electron microscope tomography using robust prediction of specimen movements. J. Struct. Biol. 152, 36–51 (2005).Article PubMed Google Scholar Zheng, S. Q. et al. MotionCor2: anisotropic correction of beam-induced motion for improved cryo-electron microscopy. Nat. Methods 14, 331–332 (2017).Article CAS PubMed PubMed Central Google Scholar Rohou, A. & Grigorieff, N. CTFFIND4: fast and accurate defocus estimation from electron micrographs. J. Struct. Biol. 192, 216–221 (2015).Article PubMed PubMed Central Google Scholar Zivanov, J. et al. New tools for automated high-resolution cryo-EM structure determination in RELION-3. eLife 7, e42166 (2018).Article PubMed PubMed Central Google Scholar Pettersen, E. F. et al. UCSF Chimera—a visualization system for exploratory research and analysis. J. Comput. Chem. 25, 1605–1612 (2004).Article CAS PubMed Google Scholar Brown, A. et al. Tools for macromolecular model building and refinement into electron cryo-microscopy reconstructions. Acta Crystallogr. D Biol. Crystallogr. 71, 136–153 (2015).Article CAS PubMed PubMed Central Google Scholar Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot. Acta Crystallogr. D Biol. Crystallogr. 66, 486–501 (2010).Article CAS PubMed PubMed Central Google Scholar Davis, I. W. et al. MolProbity: all-atom contacts and structure validation for proteins and nucleic acids. Nucleic Acids Res. 35, W375–W383 (2007).Article PubMed PubMed Central Google Scholar Liebschner, D. et al. Macromolecular structure determination using X-rays, neutrons and electrons: recent developments in Phenix. Acta Crystallogr. D Struct. Biol. 75, 861–877 (2019).Article CAS PubMed PubMed Central Google Scholar Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination. Nat. Methods 14, 290–296 (2017).Article CAS PubMed Google Scholar Bepler, T. et al. Positive-unlabeled convolutional neural networks for particle picking in cryo-electron micrographs. Nat. Methods 16, 1153–1160 (2019).Article CAS PubMed PubMed Central Google Scholar Asarnow, D., Palovcak, E. & Cheng, Y. asarnow/pyem: UCSF pyem v0.5. Zenodo https://doi.org/10.5281/zenodo.3576630. (2019).Abraham, R. T. Cell cycle checkpoint signaling through the ATM and ATR kinases. Genes Dev. 15, 2177–2196 (2001).Article CAS PubMed Google Scholar Download referencesAcknowledgementsWe thank the staff of the Memorial Sloan Kettering Cancer Center cryo-EM facility for help with data collection.FundingB.L. is supported by a Canadian Institutes of Health Research training award (DFS-164271). This work is supported by NIH grants (5R01GM149915 and P30CA008748). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the paper.Author informationAuthors and AffiliationsStructural Biology Program, Memorial Sloan Kettering Cancer Center, New York, NY, USABuren Li, Ayat Yaseen & Nikola P. PavletichLouis V. Gerstner, Jr. Graduate School of Biomedical Sciences, Memorial Sloan Kettering Cancer Center, New York, NY, USABuren LiAuthorsBuren LiView author publicationsSearch author on:PubMed Google ScholarAyat YaseenView author publicationsSearch author on:PubMed Google ScholarNikola P. PavletichView author publicationsSearch author on:PubMed Google ScholarContributionsB.L. carried out biochemical experiments and collected and analyzed the cryo-EM data. A.Y. carried out cell culture and protein purification. B.L. and N.P.P. analyzed the data and wrote the paper.Corresponding authorCorrespondence to Nikola P. Pavletich.Ethics declarationsCompeting interestsThe authors declare no competing interests.Peer reviewPeer review informationNature Structural & Molecular Biology thanks Wei Yang, Lee Zou and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: Dimitris Typas, in collaboration with the Nature Structural & Molecular Biology team.Additional informationPublisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Extended dataExtended Data Fig. 1 Steady-state kinetic analysis, cryo-EM data processing and structure of ATR–ATRIP.a Steady-state kinetic analysis of the phosphorylation of PHAS-I, a commonly employed ATR substrate55, by ATR–ATRIP (19 nM) alone and in the presence of 32 μM TOPBP1AAD. The graph shows reaction velocity (normalized by enzyme concentration) versus PHAS-I concentration from three independent experiments with indicated markers, and the dose-response curves and Km and kcat values calculated from nonlinear regression fitting of the data. b Workflow of cryo-EM data processing of the apo-ATR–ATRIP sample. The consensus reconstruction (masked and post-processed) is colored by local resolution according to the color key. The graph shows the gold-standard Fourier shell correlation (FSC) curve calculated from two independently refined half-maps, and the dashed line indicates the FSC cutoff at 0.143 marking the resolution of the consensus reconstruction. c Consensus and focused reconstructions (masked and post-processed) colored by local resolution according to the color key below each reconstruction. The graph shows the gold-standard FSC curves calculated from two independently refined half-maps, as well as the FSC curve (colored red) between the model and the composite map. The dashed lines indicate the FSC cutoffs at 0.143 and 0.5 marking the resolutions of the consensus and focused-refined reconstructions and of the correlation of the model with the map, respectively. d 3D classification of apo-ATR–ATRIP particles from the ATR–ATRIP–TOPBP1AAD data showing two orthogonal views (top and bottom rows) of the reconstruction from each class. The twofold rotation axes of the ATRIP–N-heat and the FAT–KD portions are shown as black lines, with the angle between the two indicated below each map. Also indicated are the percentage of particles in each class. The ~45° relative rotation of the M-heat hinge in the two protomers of the consensus reconstruction is represented by a range of 40° to 54° in the 3D classes. There are two other hinges that exhibit minor rotations of up to −3° to 3° relative to the consensus structure. One is at a discontinuity in the regular arrangement of HEAT repeats near the C-terminus of N-heat (between helices nα34 and nα35, Extended Data Fig. 4), and the other is at the interface between N-heat and M-heat1.Source dataExtended Data Fig. 2 Cryo-EM data processing and conformational flexibility of ATR–ATRIP–TOPBP1AAD.a Representative cryo-EM micrograph and data processing workflow of the ATR–ATRIP–TOPBP1AAD sample. Consensus and focused reconstructions (masked and post-processed) are shown for the active-state particles (rightmost 3D class in Extended Data Fig. 2b) converted to monomers by C2 symmetry expansion and partial signal subtraction that are colored by local resolution according to the color key below each reconstruction. The graph shows the corresponding gold-standard FSC curves calculated from two independently refined half-maps, as well as the FSC curve (colored red) between the refined model and the composite map, with the FSC cutoffs at 0.143 and 0.5 indicated by dashed lines. Also shown are orthogonal views of the consensus reconstruction in C2 symmetry (prior to symmetry expansion and partial signal subtraction) overlaid with the 3D histogram of the Euler angles of the particles used in the refinement. b 3D classification of the ATR–ATRIP–TOPBP1AAD data showing each class in two orthogonal views. The TOPBP1AAD density is colored red. The tilt axes, angles and percent of the particles are indicated as in Extended Data Fig. 1d. The resolution of each class after 3D refinement is indicated at the bottom row. c The M-heat hinge rotations associated with the conversion of the open and closed protomers to the active state occur along axes distinct from those relating the open protomer to the closed one. The M-heat structures of the two apo protomers and of the active protomer were superimposed by aligning their M-heat1 subdomains. M-heat1 domains are colored gray, except for the C-terminal portion of mα9 that is colored as indicated for the three M-heat2 domains. The axes of rotations are shown as red sticks, with the transition and rotation angle indicated. The C-termini (residue 1371) are labeled to highlight the extent of the active conformation’s M-heat2 movement. d As with the activation of mTOR29 (left) and DNA-PKcs28 (right), the ATR KD N lobe is rotated relative to the C lobe in the symmetric class. The kinase domains of mTOR (left) and DNA-PKcs (right) in both their inactive and active states were superimposed on the apo and symmetric 3D class of ATR by aligning the KD C lobe segment containing the catalytic and activation loops (residues 2474-2498 of ATR). For the color key shown in the figure, the inactive states are indicated by the prefix ‘i’ and the active states by ‘a’.Source dataExtended Data Fig. 3 Comparison of the closed and open protomer kinase domains of active ATR–ATRIP.Models of the kinase domains of the closed and open ATR protomers refined against the C1 FAT-KD focused reconstruction from particles of the zero-tilt 3D class (Extended Data Fig. 2b) overlaid with their respective density. Also shown are the model and density from the reconstruction following C2 symmetry expansion and partial signal subtraction on the same particles, and the three models superimposed on their C lobes showing that the kinase domains of the closed and open protomers adopt essentially the same conformation in the active state. The KD backbone is shown as coils, AMPPNP in stick representation, and Mg2+ as spheres. Backbone and sidechain atoms are not shown for clarity.Extended Data Fig. 4 Conservation, secondary structure and interactions of ATR.Conservation is shown in the blue graph above the amino acid sequence, α-helices are shown as rectangles, β-strands as arrows, coils/loops as solid lines, and unmodeled, disordered regions as dashed lines. ATR–ATR interactions are indicated by asterisks, ATR–ATRIP interactions by triangles, ATR–TOPBP1 interactions by hollow circles, and ATR–ETAA1 interactions by gray circles.Extended Data Fig. 5 Conservation, secondary structure and interactions of ATRIP and TOPBP1AAD.a Conservation, secondary structure, and interactions of ATRIP. Conservation and secondary structure are marked as in Extended Data Fig. 4. ATR–ATRIP interactions are indicated by triangles, ATRIP–ATRIP interactions by diamonds, and ATRIP–TOPBP1 interactions by circles. b Conservation, secondary structure, and interactions of TOPBP1AAD. Conservation and secondary structure are marked as in Extended Data Fig. 4. ATR–TOPBP1 interactions are indicated by hollow circles, and ATRIP–TOPBP1 interactions by filled circles. c Cryo-EM density of the TOPBP1 tα1, tα2 and tβ1α3 segments bound to ATR–ATRIP. d Cryo-EM density of AMPPNP, Mg2+, and key N and C lobe residues (as shown in Fig. 4d) in the active states of ATR, mTOR and DNA-PKcs. Note that the DNA-PKcs structure was determined in the absence of ATP, its analogues, or Mg2+. e Surface representation of one copy of ATR–ATRIP–TOPBP1AAD and the close-up views of the tα1, tα2 and tβ1α3 binding sites colored by sequence conservation in 14 orthologues of ATR and ATRIP. Red is conserved in 14 orthologues, orange in 13, yellow-orange in 12, and yellow in 10 and 11. ATR domains and ATRIP are delineated by dashed lines.Extended Data Fig. 6 TOPBP1 allosterically activates ATR–ATRIP.a Intermediate tβ1α3–ATR interface of the open protomer, colored gray for ATR and purple for TOPBP1, superimposed on the active-state interface by aligning their respective M-heat1 domains. The side chains of TOPBP1 Trp1145 and FAT residues that interact with it are shown in stick representation. The intermediate 3D class shown is the second of the three single-TOPBP1AAD classes (Extended Data Fig. 2b). They are related through relative rotations of the M-heat hinge, of 13° to 16°, partway to the rotation of the active conformation (Extended Data Fig. 2c), and additional rotations at the global hinges common to apo-ATR–ATRIP. The surface area buried at the tβ1α3–ATR interface increases by 23 % in the active state compared to this intermediate state (from 884 Å2 to 1087 Å2). This is in part due to Trp1145 packing with M-heat1, and also tα3 packing more extensively with N-heat in the active state. b Intermediate tβ1α3–ATR interface of the closed protomer superimposed on that of the active state as in Extended Data Fig. 6a. The remodeling of this interface to the active state increases its surface area buried by 16 %. The intermediate 3D class shown is the first of the four double-TOPBP1AAD inactive classes (Extended Data Fig. 2b). The three 3D classes between this intermediate and the active state exhibit fluctuations in the conformations of the tβ1α3–ATR interfaces on both protomers, accompanied by rigid body FAT movements and rotations at their M-heat hinges. c The ATR, mTOR29 and DNA-PKcs28 PIKKs share a common trigger point in the middle portion of the FAT domain, at the interface of the TRD2 and TRD3 subdomains41 (shown superimposed in gray for the indicated residues of the three proteins). This is where the M-heat2 domain of ATR (orange) and the N-heat domains of mTOR (blue) and DNA-PKcs (green) pull on the FAT domain as a result of distal events initiated by their respective activators. d Comparison of the positions of critical catalytic residues in the active ATR state relative to those of mTOR in which their roles were previously described41. For clarity, only the AMPPNP of the ATR structure is shown. In keeping with the structures of the inactive and active states of mTOR and DNA-PKcs, the activation-loop conformation of ATR is structurally invariant between the inactive and active states.Extended Data Fig. 7 Cryo-EM data processing of ATR–ATRIP–ETAA1AAD.a ATR–ATRIP (60 nM) pull-down assay using GST-tagged TOPBP1AAD and ETAA1AAD (2 μM), with the bar graph from two independent experiments indicating the percentage of ATR recovered relative to that of the GST-tagged TOPBP1AAD pull-down. b Representative cryo-EM micrograph and data processing workflow of the ATR–ATRIP–ETAA1AAD sample. Consensus and focused reconstructions (masked and post-processed) are shown for the particles of the C2 symmetric 3D class from the ATR–ATRIP–ETAA1AAD data sets (rightmost 3D class in Extended Data Fig. 7d) colored by local resolution according to the color key below each reconstruction. The graph shows the corresponding gold-standard FSC curves calculated from two independently refined half-maps, as well as the FSC curve (colored red) between the refined model and the composite map, with the FSC cutoffs at 0.143 and 0.5 indicated by dashed lines. Also shown are orthogonal views of the consensus reconstruction in C2 symmetry overlaid with the 3D histogram of the Euler angles of the particles used in the refinement. c Representative cryo-EM micrograph and cryo-EM data processing workflow of the ATR–ATRIP–ETAA1AAD(mut) sample. d 3D classification of the ATR–ATRIP–ETAA1AAD data showing each class in two orthogonal views. The ETAA1AAD density is colored magenta. The tilt axes, angles and percent of the particles are indicated as in Extended Data Fig. 1d. The large-tilt classes (first seven classes from the left) lack density for the eβ1 and eα2 segments of ETAA1 on the open protomer of ATR (right protomer on the top panel). The classes exhibit conformational flexibility at the three hinges that are common to apo-ATR–ATRIP (Extended Data Fig. 1d legend). However, two of these hinges, one between N-heat helices nα34 and nα35, and the other at the interface between N-heat and M-heat1 display distinct conformational landscapes associated with the binding of ETAA1 eα2 and eα1, respectively, to these sites. The distinct rotations are most pronounced for the symmetric class.Source dataExtended Data Fig. 8 ETAA1AAD conservation, secondary structure, interactions and competitive inhibition of TOPBP1.a Cryo-EM density of the ETAA1 eβ1, eα1, eα2 and eα3 and eβ1(mut) segments bound to ATR. Also shown is the density (right) for AMPPNP, Mg2+, and key N and C lobe residues from the consensus ATR–ATRIP–ETAA1AAD reconstruction. b Conservation, secondary structure and interactions of ETAA1AAD. Conservation is shown in the blue graph above the sequence, and α helices are shown as rectangles, β strands as arrows, coils/loops as solid lines, and unmodeled, disordered regions as dashed lines. ATR–ETAA1 interactions are indicated by gray circles. c Surface representation of one copy of ATR–ATRIP–ETAA1AAD and the close-up views of the, eβ1, eα1, eα2 and eα3 binding sites colored by sequence conservation in 14 orthologues of ATR and ATRIP. Red is conserved in 14 orthologues, orange in 13, yellow-orange in 12, and yellow in 10 and 11. ATR domains and ATRIP are delineated by dashed lines. d Comparison of the M-heat hinges between the inactive open and closed protomers, the active TOPBP1AAD-bound conformation, and the symmetric ETAA1AAD-bound conformation. For clarity, only the first helix (mα10) of each M-heat2 is shown. The structures were superimposed on their M-heat1 subdomains (gray, except the C-terminal portions of mα9, which are colored as the mα9-mα10 linkers and mα10 helices). e The IC50 value of ETAA1AAD inhibiting ATR-ATRIP activation by TOPBP1, shown in Fig. 5j of the main text, does not change significantly if phosphorylation of TOPBP1 and ETAA1AAD are included in the calculation. Plots are shown for the PHAS-I substrate alone (blue, as in Fig. 5j), and also in combination with the phosphorylated TOPBP1 and ETAA1AAD (red, labeled Total). f Autoradiogram showing activation of ATR–ATRIP (19 nM) phosphorylating PHAS-I (100 μM) by TOPBP1AAD (24 μM) in the presence of indicated ETAA1AAD concentrations. IC50 curves and values were calculated from nonlinear regression fitting of incorporated 32P from three independent experiments, which were plotted as fractions of the zero ETAA1AAD reaction. Plots are shown for the PHAS-I substrate alone (blue) and also in combination with phosphorylated TOPBP1AAD and ETAA1AAD (red, labeled Total). g Comparison of the kinase domains of ATR–ATRIP–TOPBP1AAD and the partially active Mec1 mutant superimposed on their C lobes. The F2244L mutation in Mec1 (indicated in parentheses, corresponding to ATR Phe2495), located at one end of the activation loop, results in an activation-loop conformation distinct from that of the ATR active state31.Source dataSupplementary informationReporting Summary (download PDF )Source dataSource Data Figs. 1, 4 and 5 and Extended Data Figs. 1, 2, 7 and 8 (download XLSX )Statistical source data.Source Data Figs. 1, 4 and 5 and Extended Data Figs. 7 and 8 (download PDF )Unprocessed gels.Rights and permissionsOpen Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.Reprints and permissionsAbout this article