Human DNA-dependent protein kinase activation mechanism

DNA-dependent protein kinase (DNA-PK), a multicomponent complex including the DNA-PK catalytic subunit and Ku70/80 heterodimer together with DNA, is central to human DNA damage response and repair. Using a DNA-PK-selective inhibitor (M3814), we identified from one dataset two cryo-EM structures of the human DNA-PK complex in different states, the intermediate state and the active state. Here we show that activation of the kinase is regulated through conformational changes caused by the binding ligand and the string region (residues 802–846) of the DNA-PK catalytic subunit, particularly the helix-hairpin-helix motif (residues 816–836) that interacts with DNA. These observations demonstrate the regulatory role of the ligand and explain why DNA-PK is DNA dependent. Cooperation and coordination among binding partners, disordered flexible regions and mechanically flexible HEAT repeats modulate the activation of the kinase. Together with previous findings, these results provide a better molecular understanding of DNA-PK catalysis.

The DNA-PK of vertebrate cells is central to the DNA damage response and the nonhomologous end-joining (NHEJ) pathway for the repair of DNA double-strand breaks, the most dangerous form of DNA damage, which may lead to genomic instability, carcinogenesis and cell death 1,2 . DNA-PK was first identified when addition of double-stranded DNA (dsDNA) to cell extracts was observed to stimulate the phosphorylation of certain proteins. The catalytic activity was found to be related to a very large polypeptide, now known as the DNA-PK catalytic subunit (DNA-PKcs), with 4128 amino acid residues and a molecular weight of ~470 kDa [3][4][5] . Later, it was discovered that the effective recruitment and activation of DNA-PKcs by DNA requires the heterodimer Ku70/80, which is formed by two subunits of molecular weights approximately 70 kDa and 80 kDa 6,7 . The kinase domain of DNA-PKcs was found to be highly similar to that of phosphoinositide 3-kinase, belonging to the family of phosphoinositide 3-kinase-related kinases (PIKKs), which includes ATM, ATR, mTOR, SMG1 and TRRAP 8 . PIKK family members have the greatest similarity in the kinase domain, which is surrounded by a conserved α-helical FAT (FRAP, ATM and TRAPP) domain together with the PIKK regulatory domain (PRD) and the FATC domain. All PIKKs also have an α-solenoid amino-terminal region, which consists mainly of HEAT repeats that have important regulatory roles.
Our previous work suggested that inhibitors have a similar mode of structural regulation to that of ATP binding DNA-PKcs 18 . Using the latest-generation DNA-PK selective inhibitor M3814 to mimic ATP binding, while at the same time slowing down and inhibiting kinase catalysis, we succeeded in obtaining two different conformations of the DNA-PK-ligand complex from one cryo-EM dataset: the intermediate state, where the inhibitor binds to the ATP pocket before the closure of the kinase domain, and the active state, where the kinase domain is closed up and the ligand is brought next to the substrate-binding site. Together with the results of previous work on DNA-PKcs, DNA-PK and DNA-PKcs-ligand complexes, these findings provide structural insight Article https://doi.org/10.1038/s41594-022-00881-w flexible, including the vWA domain, the six-helical bundle (595-706) and the 12 C-terminal residues (721-732). However, in DNA-PKcs, the previously flexible N-terminal region and HhH motif (816-836) are now stabilized between the FAT region and the DNA terminus (Fig. 3d).
On one side of the HhH motif, there are electrostatic interactions between positively charged residues, including Arg820, Lys824, Lys832 and His833, and the negatively charged DNA phosphate backbone. On the other side of the HhH motif, hydrophobic interactions define the binding with the DNA-PKcs N-terminal region. Phe826 of the HhH motif docks into the groove between the two helices that pack onto the into the progression from ligand binding through the full catalytic pathway 13,14,18 .

Structure overview
DNA-PKcs natively purified from HeLa cell nuclear extract was incubated on ice for 1 h with purified recombinant Ku70/80, DNA and an excess of DNA-PK selective inhibitor M3814. The dataset was collected on a Gatan K3 detector and processed (Extended Data Fig. 1 and Table 1). From the dataset, we identified an intermediate state of DNA-PK, allowing definition at 3.2 Å resolution following density modification with ResolveCryoEM, before closure of the kinase domain and catalysis (Fig. 1a). Moreover, we managed to stabilize and lock the active DNA-PK in the transient conformation adopted for catalysis with a closed kinase domain, giving a cryo-EM model at a resolution of 2.8 Å following density modification with ResolveCryoEM (Fig. 1b), in which the inhibitor occupies a position next to the substrate-binding site.

Cryo-EM structure of DNA-PK in the intermediate state
The inhibitor docks in the center of the ATP-binding groove in the intermediate-state DNA-PK (Fig. 2a,b). The overall assembly of the kinase complex is similar to those previously defined [12][13][14]17 . The DNA-Ku70/80 core complex interacts with the N-terminal unit and circular cradle of DNA-PKcs, and the DNA passes under the N-terminal unit, while the carboxy-terminal region of Ku80 docks on the circular cradle. However, there are many differences in the conformations and structural details.
DNA-PK in the intermediate state has several structured regions that are flexible. In Ku80, these include the vWA domain (6-242), the six-helical bundle (595-706) and the 12 residues (721-732) of the C-terminal region (543-732) 19,20 . Within the DNA-PKcs component of DNA-PK, there are several structured regions that are flexible, including the N-terminal region, close to where the DNA end enters, and a part of the kinase region within the head unit, especially the four-helical FKBP12-rapamycin-binding (FRB) domain (3582-3675) 21 . Another interesting flexible region in DNA-PKcs is the helix-hairpin-helix (HhH) motif between the N terminus of DNA-PKcs and the DNA end ( Fig. 2c), which is likely to come from a flexible region (802-846) hanging down in the DNA-PKcs central cavity formed by the circular cradle and N-terminal unit. It is clear that in the higher-resolution active-state DNA-PK structure, the HhH motif (816-836) is essential not only for allosteric activation and conformational transition but also for stabilizing the interaction between the DNA-PKcs N-terminal region and the head unit.
Compared with the previously published structures of the inactive-state apo-DNA-PK, the major conformational differences lie in the head unit of DNA-PKcs (Extended Data Fig. 2a,b) 13,14 . In the intermediate-state complex, the head unit moves towards the DNA end and the N-terminal unit of DNA-PKcs (Extended Data Fig. 2a,b). When the ligand binds, the p-loop moves close to it (Extended Data Fig. 2c). This leads to concerted movements involving conformational changes of the flanking β-sheets of the p-loop and movement in the hydrophobic core of the head unit, resulting in overall conformational changes. More importantly, the outward movement of the head unit releases the PRD (4009-4039) from its position in the apo kinase ( Fig. 2b and Extended Data Fig. 2c). In the intermediate complex, there is only weak signal near the PRD position, which indicates that the PRD is destabilized. When the blockage of PRD is removed, the kinase active site becomes available for substrate peptide binding and subsequent catalytic activity.

Cryo-EM structure of DNA-PK in the active state
In the active state of DNA-PK, the kinase core is closed and the ligand M3814 is moved close to the substrate-binding site (Fig. 3a-c). Similar to the intermediate-state kinase complex, some structured regions are flexible in the active DNA-PK structure. Most parts of Ku80 remain HhH motif, interacting with the side chains of Phe88 and Ile91 and the main chain atoms of Arg36, Gln40, Glu84 and Lys87. The side chains of Leu818, Val830, Leu831 and Leu834 also interact with the hydrophobic surface composed of the side chains of Val43, Leu44 and Ile91 and the main chain atoms of Glu94 and Lys95. In addition to the hydrophobic interaction, there is an electrostatic interaction between Lys835 and Glu94 and hydrogen bonds among Gln823, Arg36 and Gln40 (Fig. 3d). The position of the N-terminal region is also defined by interactions with the FAT region of the head unit. Tyr3101 lies in the hydrophobic pocket formed by Ala49, Val50 and Leu53. There are water-mediated side chain interactions among the polar and charged residues of the interface, including Asp3066, His3070, Lys3075, Tyr3090, Arg3098 and Tyr3102 from the FAT region and Arg9, Ser60 and Asp62 from the N-terminal unit (Fig. 3f). Furthermore, in the active form of DNA-PK, a DNA-end-binding helix (2735-2768) from the flexible region of DNA-PKcs lies near the ABCDE cluster in the middle of the DNA-PKcs central cavity, restricting inward movement of the DNA (Fig. 3e). The stabilization of the DNA-end-binding helix is likely to be caused by the fixation of the activation HhH motif together with its flanking peptide string and the DNA end.
Compared with the previously proposed model of an activated DNA-PK, the structure of the active holoenzyme, described here, is more compact, showing different conformational changes (Extended Data Fig. 3a) 13 . The DNA end lies in the center of the holoenzyme and to different extents all regions of the holoenzyme move towards it (Extended Data Fig. 3a). Moreover, the lengths of the DNA strands coming into the DNA-PKcs central cavity are different (Extended Data Fig. 3b). A further difference is that the helix from the region around the PQR cluster continues to dock in the cleft between the N-terminal unit and the circular cradle in our active form, similar to the intermediate DNA-PK (Extended Data Fig. 3c). In addition, the PRD, instead of having a 115° outward rotation, is destabilized and removed from the substrate-binding site (Extended Data Fig. 3d), although different, destabilization and rotation have the same effect in making the substrate-binding site available for substrates to bind.

Intermediate-to-active-state transition of DNA-PK
The holoenzyme undergoes a series of conformational changes during the transition from the intermediate to the active state (Fig. 4a). Most changes lie in DNA-PKcs, especially in the N-terminal Article https://doi.org/10.1038/s41594-022-00881-w and head units and the part of the circular cradle unit that is close to them (893-1469) (Fig. 4a-d).
According to our structures, the main trigger for kinase activation and transition is the fixation of the HhH motif (816-836) (Fig. 4b). Stabilization of the activation HhH motif between the DNA terminus and the N-terminal region induces a conformational change of the latter. Furthermore, it restricts the conformation of the previously disordered flexible peptide (837-846) linking to the following α-helix (847-862) from the top of the DNA-PKcs central cavity (Fig. 4b). The peptide string pulls the connected helix closer to the DNA end and leads to further serial downward movements of the nearby FAT region, which comes in contact with the N-terminal region. The conformational changes are also passed on to the whole molecule, especially to the kinase region ( Fig. 4a-d). The peptide string acts as a 'pull switch' for activating the kinase, and the inhibitor in the active state is moved more than 10 Å away from its position in the intermediate state (Fig. 4d).
To understand how the kinase region is closed during activation, the kinase regions of DNA-PK in the two states were aligned by superimposing the C-lobes of the two kinase structures (Fig. 4e). This demonstrates a concerted movement of the N-lobe and neighboring FAT region (3490-3565), pushing the p-loop towards the C-lobe. The inhibitor molecule moves about 3.5 Å from its previous position relative to the C-lobe (Fig. 4e) towards the catalytic loop that contains the catalytic aspartate (Asp 3922) and the neighboring positively charged His 3924, which helps to localize the γ-phosphate of ATP together with Ser 3731 from the p-loop (Fig. 4e).
Previously, it was proposed that YRPD (2775-2795) and its interacting region (2569-2585) are important in regulation of the ABCDE cluster and autophosphorylation 22,23 . During the transition from the intermediate state to the active state, YRPD domain and its interacting region move closer to the catalytic site, indicating the cis-autophosphorylation of the ABCDE domain (Fig. 4e).

Structural mechanism of DNA-PK activation
The structural studies of DNA-PKcs and DNA-PK, together with the intermediate and active DNA-PK, now provide a better understanding of the concerted movements of different domains during allosteric activation.
Here, we focus on comparison of the kinase domains of different DNA-PKcs-related models, aligned by the C-lobe. Comparison of apo-DNA-PKcs and apo-DNA-PK indicates that the assembly of DNA-PK holoenzyme itself leads to an opening of the N-lobe, involving outward moving of the FRB domain and partial relaxation of the PRD (Fig. 5a) 13,14 . When a ligand binds to the ATP-binding site in DNA-PKcs, the PRD tilts but remains docked on the substrate-binding site 18 . With binding of the inhibitor to the ATP pocket in DNA-PK, the FRB domain further opens Article https://doi.org/10.1038/s41594-022-00881-w compared with that in apo-DNA-PKcs and apo-DNA-PK, leading to the release of the PRD (Fig. 5a).
Moreover, we can now better understand the progression from DNA-PKcs to DNA-PK in the catalytic cycle. The assembly of DNA-PK from ligand-bound DNA-PKcs or ligand binding to apo-DNA-PK affect the conformation of the head unit, open the FRB domain and release PRD from the substrate-binding site. Stabilization of the activation HhH motif between the DNA end and the N-terminal bridge induces global conformational changes within the holoenzyme complex, including N-lobe movement and closure of the kinase domain to move the ligand towards the substrate-binding site for catalysis (Fig. 5b). Of note, considering the high concentration of ATP in the cell, the in vivo structural progression of DNA-PK activation should start from ligand-bound DNA-PKcs and proceed to intermediate and active DNA-PK. In general, DNA, Ku70/80 and ligand binding, together with the activation HhH motif far from the DNA-PKcs head region, allosterically activate the kinase. Similar concerted movements among HEAT repeats and the kinase domain have previously been visualized in other PIKK members,  Article https://doi.org/10.1038/s41594-022-00881-w and it is tempting to hypothesize that the cooperation among binding partners, the disordered flexible region and the structurally plastic HEAT repeats regulate functions of PIKK family 24 .
Among the phosphor ylation targets of DNA-PK, self-phosphorylation is probably the most studied [25][26][27][28] . According to our active kinase model, the ABCDE cluster (2609-2647) is on a flexible region of polypeptide, which could have access to the kinase catalytic site and therefore could be phosphorylated in cis. The movement of the YRPD domain and its interacting region also suggests autophosphorylation of the ABCDE cluster in a cis manner. However, the PQR cluster (2023-2056) of DNA-PKcs can only be autophosphorylated in trans owing to its distance from the kinase catalytic region. Studies of NHEJ complexes containing DNA-PK have revealed the assembly of different dimers that probably mediate autophosphorylation in trans [14][15][16] . In these models, the Ku80 C-terminal region and other NHEJ components including XLF, XRCC4 and DNA ligase IV appear to assemble as a scaffold. A recent publication has revealed an interesting new conformation of DNA-PK after self-phosphorylation on the ABCDE cluster that is in line with the previous SAX model on the displacement of the N-terminal HEAT domain, as the N-terminal HEAT domain opens greatly 17,29 . Of note, in the autophosphorylated DNA-PK in complex with Artemis, the activation HhH motif cannot be stabilized by interactions between the DNA end and N-terminal arm owing to the Artemis blockage and remains in a destabilization state 17 . Therefore, the kinase domain is not closed, and DNA-PK is in an inactive state halting further phosphorylation before end processing and removal of the Artemis catalytic domain. However, how the phosphorylation on the flexible ABCDE domain affects the overall conformation of the HEAT repeats, especially the N-terminal region, remains unclear and requires further investigation.
Nevertheless, with high-resolution DNA-PK structures in different states, our findings reveal the structural mechanism of allosteric Article https://doi.org/10.1038/s41594-022-00881-w activation of the atypical kinase DNA-PK, the key step of its catalytic cycle, and offer improved models for future structure-guided drug discovery of new cancer therapeutics targeting the DNA-PK holoenzyme.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41594-022-00881-w.

Purification of Ku70/80 and DNA-PKcs
Purifications of Ku70/80 and DNA-PKcs were performed as described previously 30 .

DNA annealing
DNA oligonucleotides (43 nucleotides) designed for blunt-end dsDNA were purchased from Sigma Aldrich. The complementary oligonucleotides were dissolved in ultrapure water to a final concentration of 100 μM, mixed at a molar ratio of 1:1 and annealed in a thermocycler. The mixed volume was first heated to 98 °C for 5 min and then cooled to 25 °C for 2 h to form a 43-base-pair (bp) dsDNA. The DNA sequences used for annealing can be found below.

Sample preparation and cryo-EM data collection
Purified DNA-PKcs, Ku70/80 and 43-bp DNA were mixed at a molar ratio of 1  (Table 1).

Image processing and model refinement
Data processing was done using Warp and cryoSPARC [31][32][33] . CTF correction, motion correction and particle auto-picking were carried out using Warp. All particles were subjected to ab initio reconstruction in cryoSPARC to generate four initial three-dimensional models. Classes of DNA-PKcs only and contamination were discarded and classes of DNA-PK were selected for further optimization through iterative rounds of heterogeneous refinement to filter out unwanted particles. The two models of DNA-PK in different states were then further refined using homogeneous refinement, global CTF refinement and nonuniform refinement in cryoSPARC. The reported resolutions were determined based on the 'gold standard' of the 0.143 Fourier shell correlation (FSC) criterion 34 . Maps were further improved using ResolveCryoEM in PHENIX 35 . Regarding the modeling of the complexes, DNA bases were manually built in COOT, and previously published cryo-EM structures of DNA-PK (PDB: 6ZHA, 7K1N, 7K0Y) were used as an initial reference for the protein modeling 13,14,36 . The templates were first rigid-body-fitted into the maps in CHIMERA and CHIMERAX, followed by real-space refinement in PHENIX [37][38][39] . The ligand was then modeled into the corresponding density in the cryo-EM maps in COOT, followed by multiple refinement rounds in PHENIX and model building in COOT 36,38 .

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
Cryo-EM maps have been deposited in the Electron Microscopy Data Bank under accession numbers EMD-14546 (DNA-PK in the intermediate state) and EMD-14545 (DNA-PK in the active state). Model coordinates have been deposited in the Protein Data Bank under accession numbers 7Z88 (DNA-PK in the intermediate state) and 7Z87 (DNA-PK in the active state). Other structures used for modeling and comparison in this study were retrieved from the Protein Data Bank with accession codes 6ZFP for DNA-PKcs; 7OTY for the DNA-PKcs-inhibitor complex; and 6ZHA, 7K1N and 7K0Y for DNA-PK.