3.9 Å structure of the yeast Mec1-Ddc2 complex, a homolog of human ATR-ATRIP

See allHide authors and affiliations

Science  01 Dec 2017:
Vol. 358, Issue 6367, pp. 1206-1209
DOI: 10.1126/science.aan8414

Holding a master regulator in check

A family of eukaryotic protein kinases, the phosphatidylinositol 3-kinase-related kinases (PIKKs), has key functions in DNA repair and nutrient sensing. In humans, ATR kinase locates DNA damage through its partner, ATRIP. Once activated, ATR initiates a cell-cycle cascade that culminates in cell-cycle arrest. Wang et al. determined the high-resolution structure of Mec1-Ddc2 (the yeast homolog of ATR-ATRIP) by electron microscopy. The structure shows the detailed architecture of the multidomain complex that overall forms a dimer of heterodimers. The detailed analysis of the structure reveals how an allosteric mechanism may activate the kinase.

Science, this issue p. 1206


The ataxia telangiectasia–mutated and Rad3-related (ATR) kinase is a master regulator of DNA damage response and replication stress in humans, but the mechanism of its activation remains unclear. ATR acts together with its partner ATRIP. Using cryo–electron microscopy, we determined the structure of intact Mec1-Ddc2 (the yeast homolog of ATR-ATRIP), which is poised for catalysis, at a resolution of 3.9 angstroms. Mec1-Ddc2 forms a dimer of heterodimers through the PRD and FAT domains of Mec1 and the coiled-coil domain of Ddc2. The PRD and Bridge domains in Mec1 constitute critical regulatory sites. The activation loop of Mec1 is inhibited by the PRD, revealing an allosteric mechanism of kinase activation. Our study clarifies the architecture of ATR-ATRIP and provides a structural framework for the understanding of ATR regulation.

Ataxia telangiectasia–mutated (ATM) and ATM-Rad3–related (ATR) are master regulators of the DNA damage response and are highly conserved among eukaryotes. The ATR kinase is essential for the maintenance of genomic integrity, which is activated by DNA double-strand breaks (DSBs) as well as various types of DNA replication problems (1). ATR forms a stable complex with ATRIP (ATR-interacting protein), which regulates the localization of ATR and is essential for ATR signaling (13). Mutations in ATR are associated with Seckel syndrome, a clinically distinct disorder characterized by proportionate growth retardation and severe microcephaly (4). Given the central role of ATR in genome integrity and human disease, it is essential to understand the mechanism of its regulation. Molecular and structural insights into ATR are critical to facilitate the design of therapeutic agents (5). A negative-stain structure of Saccharomyces cerevisiae Mec1-Ddc2 (homolog of human ATR-ATRIP) was recently reported (6). However, the resolution obtained by electron microscopy in that study was low (22.5 Å). Here, we report the cryo–electron microscopy (cryo-EM) structure of the Mec1-Ddc2 complex at 3.9 Å resolution (table S1).

The endogenously purified Mec1-Ddc2 complex has a 1:1 stoichiometry, which is consistent with previous biochemical and functional studies (fig. S1) (2, 7). The preparation displays basal kinase activity, which could be stimulated by incubation with the endogenous Mec1 activator Dpb11, the homolog of human TopBP1 (fig. S2). Each Mec1-Ddc2 complex contains two copies of Mec1 and two copies of Ddc2, such that the complex has a butterfly-like dimeric architecture with a two-fold rotational (C2) symmetry (Fig. 1, A and B, and figs. S3 and S4). The dimeric architecture is similar to that of the Tel1 (homolog of human ATM) homodimer (8) but is distinct from that of the mTOR complex (9, 10) and Tor-Lst8 (11) (figs. S11 to S15). Human ATR can be autophosphorylated at Thr1989 (12, 13), which is neither close to the Mec1 kinase domain nor in proximity to the dimer interface (Fig. 1C). Thus, Thr1989 seems unable to access either active site in the Mec1 dimer without substantial conformational changes.

Fig. 1 Architecture of Mec1-Ddc2.

(A) Front view of the density map of the Mec1-Ddc2 homodimer at 3.9 Å. One monomer is color-coded by domain assignment: FAT-KD-PRD-FATC in blue, N-terminal α-solenoid in orange, and Ddc2 in light green. The other monomer is shown as a solid gray surface. (B) Bottom view of the density map. (C) Three views of the Mec1-Ddc2 monomer. The cryo-EM density is shown as a translucent surface and fitted with the ribbon diagram model. The N-terminal α-solenoids of Mec1 are in orange, FAT in cornflower blue, kinase N-lobe in yellow, C-lobe in hot pink, PRD in red, and FATC in blue. Also shown are the Bridge of Mec1 in dark orange and Ddc2 in light green. Each successive view is rotated as indicated. The red stars highlight the Thr1989 autophosphorylation site in human ATR. (D) Schematic representation highlighting the functional domains of Mec1 and Ddc2. The N-terminal checkpoint protein recruitment domain (CRD) and Ddc2’s interface with Dpb11 are denoted. Three units of Mec1 are labeled: the N-terminal α-solenoid and its Bridge region; the FAT region (TRD1, 2, 3 and HRD); and the KD, PRD, and FATC. The flexible regions controlling the Bridge domain and loops critical for Mec1 activity are shown above the schematic.

Mec1 contains a canonical two-lobe kinase domain (KD) spanning about 400 C-terminal residues, with three characteristic insertions: FATC (∼30 residues), LBE (∼40 residues), and PRD (∼20 residues). The elements crucial for catalysis are ordered in the structure, including the activation loop, catalytic loop, P-loop, and FATC (Fig. 1, C and D). Immediately preceding the KD is an array of helical repeats constituting the FAT domain (fig. S5), which extends toward the N-terminal α-solenoid. The Mec1 N terminus is highly flexible, with more than 200 amino acids (residues 1 to 235) invisible in the structure. Consistently, no cross-linking signals were detected within this region (14). Similarly, the N terminus of the mTOR is also invisible in the 4.4 Å cryo-EM reconstruction (10).

The Ddc2 density is well defined; most of the side chains are discernible (fig. S9). Ddc2 has a sinuous superhelical structure containing 26 helices (13 pairs) of HEAT repeats. Two long inter-HEAT repeat loops mediate multiple interactions with Mec1 (fig. S10). One loop (residues 338 to 358) abuts the FAT domain (Tyr1573 and Asn1596) and the upper spiral (Lys1110); the other one (residues 463 to 535) runs over peripheral parts of the central hollow region and intertwines with a long inter-HEAT repeat loop of Mec1 (residues 473 to 516). Interestingly, the reported DNA binding region (Lys177-Lys178-Arg179-Lys180) in Ddc2 (15) is in close proximity to the interface with the inter-HEAT repeat of Mec1 (fig. S8).

There are three major dimer interfaces in the Mec1-Ddc2 complex (Fig. 2A). The Mec1 PRD [phosphatidylinositol 3-kinase–related kinase (PIKK) regulatory domain] constitutes an important Mec1-Mec1 dimer interface that critically regulates the kinase activity of PIKKs (16). The PRD loop interacts with both the kα9b helix and the TRD3 region of another monomer by putative hydrogen bonds and electrostatic interactions (Fig. 2B). The TRD2-TRD2 dimer is a conserved interface shared by Tel1 (8) and Mec1 (Fig. 2C). Several polar or charged residues at the TRD2 and TRD3 regions are responsible for making the contacts. The Ddc2-Ddc2 dimer is formed by the coiled-coil (CC) domain, which plays a critical role in Mec1 activation (16, 17). The helix bundles and upper loops are stabilized by the Leu-mediated hydrophobic interactions (Fig. 2D). The bottom loops in the Ddc2 dimer interface are stabilized by the intermolecular Asn196-Arg197 salt bridges (Fig. 2D). In addition, Ddc2 also contributes to the Ddc2-Mec1 dimer interface and interacts with the α-solenoid of another Mec1 by putative Arg260-Asp317, Lys263-Asp322, and Lys263-Gln323 salt bridges (Fig. 2E).

Fig. 2 Three prominent dimer interfaces of the Mec1-Ddc2 complex.

(A) Cryo-EM structure of the Mec1-Ddc2 homodimer highlighting the three dimer interfaces. Two monomers are respectively displayed in surface and mesh. The interfaces of Mec1-Ddc2 are color-coded: PRD in red, TRD2 domain in cyan, TRD3 domain in purple, α-solenoids in orange, and Ddc2 in light green. (B) Enlarged view of the PRD-PRD and PRD-TRD3 interface. The potential hydrogen bonds are denoted by black dashed lines. (C) Enlarged view of the TRD2-TRD2 and TRD2-TRD3 interface. The intermolecular hydrogen bonds between the TRD2 and TRD3 domains are denoted by black dashed lines. (D) Enlarged view of the Ddc2-Ddc2 interface. The side chain of Leu residues mediating putative hydrophobic interactions is shown. The potential electrostatic interactions at the bottom loop are denoted by black dashed lines. (E) Enlarged view of the Mec1 α-solenoid–Ddc2 interface. The potential hydrogen bonds are denoted by black dashed lines. Amino acid abbreviations appearing here or in other figures: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; Y, Tyr.

The Mec1 kinase domain has the signatures of an active conformation, indicating that it is an intrinsically active enzyme. The 23-residue activation loop (2241HVDFDCLFEKGKRLPVPEIVPFR2263) adopts a fully extended conformation that stabilizes the active site. The conserved residues of both the Asp-Arg-His (DRH) and Asp-Phe-Gly (DFG) motifs of Mec1 are positioned toward the active site (Fig. 3A). The 23-residue PRD (2315SIQKALKVLRNKIRGIDPQDGLV2337) is targeted by Dpb11 as a direct activator of Mec1 (16, 18). The kα9b helix of Mec1 packs against the activation loop and caps the catalytic site. Furthermore, the Met2312 residue in the hinge between kα9 and kα9b directly contacts with the exposed Phe2244 of the DFG motif on the activation loop by hydrophobic interaction (Fig. 3B). Thus, the Mec1 PRD could clamp the activation loop through the Phe2244-Met2312 interaction and block substrate entry into the active site. In addition, five basic residues (Lys2318, Lys2321, Lys2326, Arg2324, and Arg2328) condense in the short kα9b helix of the PRD of Mec1. In human ATR, the Lys2589 → Glu (K2589E) mutation in the corresponding basic patch specifically affects TopBP1 activation (16), highlighting a critical role of the PRD in both ATR and Mec1 activation (Fig. 3B).

Fig. 3 The active site of Mec1.

(A) Detailed view of the active site. The color scheme of the N-lobe, C-lobe, and PRD is the same as in Fig. 1. The activation loop (cyan), the catalytic loop (orange), the P-loop (green), and the kα1 helix (purple) are labeled. The critical residues in the catalytic and activation loops and the contact between Phe2244 and Met2312 are highlighted. (B) Two other views of the active site, highlighting the supersecondary structures that enclose the active site. The activation and catalytic loops are besieged by the FATC and PRD. The high-abundance basic residues in the PRD are labeled. A blue star denotes the position of Lys2589 of human ATR. (C) Enlarged view of the Bridge domain, with one helix of the HEAT 32R colored in red. (D) The interfaces of the α-solenoid with the FAT and kinase domains. Cryo-EM densities of the two extended loops (linker and railing) of the Bridge are highlighted. The Ser1333 residue in human ATR (corresponding to Thr1092 in Mec1) creating a hyperactive kinase is shown in red. (E) The active site is marked by an adenosine diphosphate (ADP; red) that was modeled using the mTOR-ADP structure (PDB ID 4JSV). Two helices of the HEAT 32R extend toward the active site.

The structure clearly resolves the side-chain densities in the majority of the N-terminal α-solenoid of Mec1 (figs. S6 and S7). The N-terminal Spiral region (residues 236 to 1121) is followed by a linker (residues 1122 to 1148) that runs along the surface of the FAT and kinase domains, thereby connecting the Spiral region to a region that we refer to as the “Bridge” (residues 1149 to 1409). Strikingly, two extended helices of the N-terminal solenoid (HEAT 32R) that are in close proximity to the active site may generate steric hindrance for substrate entry to the catalytic cavity of Mec1 (Fig. 3C). A point mutation of Ser1333 in ATR creates a hyperactive kinase (19), which is in proximity to the linker region stabilizing the Bridge region (Fig. 3, D and E). The critical regulatory sites separating the S-phase and G2-phase functions of both human ATR (Thr1566, Thr1578, Thr1589) and yeast Mec1 (Phe1179) are coincidentally located in the Bridge domain (20, 21). Interestingly, the structural features of Bridge are shared by ATM (8), DNA-PKcs (22), and mTOR (9) (figs. S13 and S14). In both ATM and DNA-PKcs, the Bridge domain contains critical regulatory autophosphorylation sites (20). The mTOR Bridge domain is responsible for Raptor binding (9). These observations all indicate that the conserved Bridge domain critically regulates the kinase activity of PIKKs and constitutes an important regulation site.

The kinase activation loop generally undergoes substantial conformational changes during catalysis that are intimately tied to activation (23). The structure discloses that the PRD, LBE domain, and FATC domain of Mec1 inhibit the substrate binding and enzyme activation by enclosing the catalytic and activation loops (Fig. 3, A and B). The relief of such inhibition critically relies on several specific activators, such as Dpb11. The ATR/Mec1 activation domain (AAD) of these activators is generally unstructured and contains two critical aromatic residues, which may initiate AAD binding to Mec1 (or, in humans, to ATR) (24, 25).

Although the known AADs share little sequence homology, the region around the conserved aromatic residues generally contains several consecutive pairs of acid residues (Fig. 4A). The Mec1 activation mechanism may be mediated by both the conserved aromatic residues and acidic patch in the specific AADs (Fig. 4B). In agreement with this deduction, the sequence surrounding a conserved aromatic residue in AADs is essential for its ability to activate ATR (26). ATR activation by AAD is salt-sensitive and the activation effect is lost at higher salt concentration (27). The effects of the ATR K2589E mutation [i.e., decreasing the TopBP1 association and abolishing ATR activation (16)] probably result from altering the PRD’s basic patch for AAD binding.

Fig. 4 Model of Mec1 activation by Dpb11.

(A) Sequence characteristics of the Mec1 and ATR activators in different species. The unstructured AAD domain harbors high-abundance hydrophobic residues (green) and contains essential aromatic amino acids (red) as well as consecutive acidic patches (yellow). (B and C) Cartoon of Mec1 activation showing how the opening up of the active site leads to full kinase activity. (B) Mec1 active site before AAD binding; (C) tethering of AAD onto the PRD, which could trigger conformational changes destabilizing the hydrophobic interaction between Phe2244 of the activation loop and Met2312 in the PRD. The resulting movement of the activation and catalytic loops may open up the active site, culminating in full kinase activity.

When AAD binds to the PRD, the conserved aromatic residues could first target the exposed Phe2244-Met2312 interface and destabilize the hydrophobic interaction, which clamps the activation loop. Then, the acid patch surrounding the aromatic residues could attract the basic patch on the kα9b, increasing the overall binding affinity and the stability of the AAD-Mec1 complex (Fig. 4C). These multiple interactions could further trigger substantial conformational changes in the PRD and in the activation and catalytic loops. The synergetic movements may expose the substrate binding site and culminate in full kinase activity.

Supplementary Materials

Materials and Methods

Figs. S1 to S18

Table S1

References (2842)

References and Notes

  1. Acknowledgments: The 3D cryo-EM density map reported in this paper has been deposited in the EM Databank under accession number EMD-6708 and the corresponding model in the Protein Data Bank as PDB ID 5X6O. EM data were collected at the Center for Bio-imaging, Institute of Biophysics, Chinese Academy of Sciences. We thank G. Ji and X. Huang for technical help and support with electron microscopy, X. Shen (University of Texas M. D. Anderson Cancer Center) for providing the N-terminal Mec1 tagged yeast strain MEC1-FLAG (MATa MEC1-FLAG his3 Δ1 leu2 Δ0 met15 Δ0 ura3 Δ0), and S. Ma and H. Wu (Shanghai Chenglan Technology Co.) for help with GPU computation. Supported by National Basic Research Program grants 2014CB910700 and 2013CB910200 and National Natural Science Foundation of China grants 31222017 and 31770808. Author contributions: X.W. ran the kinase assay, froze the grids, performed the 3D reconstructions, analyzed the cryo-EM reconstruction of the Mec1-Ddc2 complex, and built the models with W.W. and T.R.; X.Z. and T.W. purified the Mec1-Ddc2; J.X. prepared the Dpb11; Z.Z. and X.W. collected the cryo-EM data; and G.C. designed experiments, analyzed data, and wrote the manuscript with X.W.
View Abstract


Navigate This Article