E科研设计

. information, and the backbones ligated to restore strand continuity. In addition to its role in repair of adventitious DSBs, the NHEJ system is responsible for repair of pro- grammed DSBs generated as intermediates in V(D)J recombination in lymphocytes during development of a bulky and largely globular ‘‘head’’ structure connected to a flat tubular ‘‘arm’’ segment terminating in two pro- jecting ‘‘claws’’ (Rivera-Calzada et al., 2005). From a combination of antibody labeling, bioinformatics, do- main docking, and comparison with a low-resolution structure of the related ATM (Llorca et al., 2003), the ma- jority of the DNA-PKcs structure has been assigned to*Correspondence: laurence.pearl@icr.ac.uk (L.H.P.); ollorca@cib. Molecular Cell 22, 511–519, May 19, 2006 ª2006 Elsevier Inc. DOI 10 Three-Dimensional Structu DNA-PKcs/Ku70/Ku80 Com on DNA and Its Implication Laura Spagnolo,2 Angel Rivera-Calzada,1 Laurence H. Pearl,2,* and Oscar Llorca1,* 1Centro de Investigaciones Biolo´gicas Consejo Superior de Investigaciones Cientı´ficas Ramiro de Maeztu, 9 Campus Complutense University 28040 Madrid Spain 2Section of Structural Biology and Cancer Research UK DNA Repair Enzyme Research Group Institute of Cancer Research Chester Beatty Laboratories 237 Fulham Road London SW3 6JB United Kingdom Summary DNA-PKcs is a large (w470 kDa) kinase that plays an essential role in the repair of DNA double-strand breaks (DSBs) by nonhomologous end joining (NHEJ). DNA-PKcs is recruited to DSBs by the Ku70/ Ku80 heterodimer, with which it forms the core of a multiprotein complex that promotes synapsis of the broken DNA ends. We have purified the human DNA- PKcs/Ku70/Ku80 holoenzyme assembled on a DNA molecule. Its three-dimensional (3D) structure at w25 A˚ resolution was determined by single-particle electron microscopy. Binding of Ku and DNA elicits conformational changes in the FATandFATCdomains of DNA-PKcs. Dimeric particles are observed in which two DNA-PKcs/Ku70/Ku80 holoenzymes interact through the N-terminal HEAT repeats. The proximity of the dimer contacts to the likely positions of the DNAendssuggests that these represent synaptic com- plexes thatmaintain brokenDNAends inproximity and provide a platform for access of the various enzymes required for end processing and ligation. Introduction DSBs are the most toxic form of DNA damage, with only one or two unrepaired breaks being sufficient to kill a di- viding cell. In S and G2 phases of the cell cycle, where sister chromatids are available, DSBs are efficiently repaired in an essentially error-free process of homolo- gous recombination dependent on the Rad51 pathway (West, 2003). In G0 and G1, however, DSBs are repaired by an NHEJ process in which the broken ends are re- sected and/or processed, with potential loss of genetic csic.es (O.L.) 1016/j.molcel.2006.04.013 re of the Human plex Assembled s for DNA DSB Repair immune diversity. Central to the NHEJ process is the for- mation of a synaptic complex that seems to be driven by the DNA-dependent protein kinase catalytic subunit (DNA-PKcs) and the Ku70 and Ku80 proteins, which maintain the broken ends in proximity and provide a plat- form for the recruitment of enzymes such as Artemis (Ma et al., 2002) and PNK (Koch et al., 2004), required to restore normal DNA structure at the broken ends, X family polymerases, which promote microhomology and cohesion between the broken ends (Ma et al., 2004), and DNA ligase IV-XRCC4 (Hsu et al., 2002), which closes the phosphodiester backbone on both strands. Primary recognition of free DNA ends is a function of the Ku protein, a highly abundant heterodimer of the ho- mologous Ku70 and Ku80 proteins. The crystal structure of the Ku70/Ku80 heterodimer (except the evolutionary distinct C-terminal domain of Ku80) shows an extended platform structure with a narrow bridge, creating a pre- formed ring that can sterically encircle DNA without es- tablishing sequence-specific contacts (Walker et al., 2001) (Figure 1A). DNA bound Ku directs the recruitment of the catalytic subunit DNA-PKcs via a small helical do- main at the C terminus of Ku80, whose structure has been determined in isolation (Harris et al., 2004; Zhang et al., 2004). DNA-PKcs itself is a large w470 kDa ser- ine/threonine protein kinase belonging to the phospha- tidylinositol-3-OH kinase (PI3K)-related kinase family (PIKKs), which includes other DNA damage-sensing en- zymes, such as ATM and ATR, as well as proteins in- volved in nutrient sensing (mTOR) and nonsense-medi- ated decay (hSMG1) (Abraham, 2004). DNA-PKcs binds DNA ends independently of Ku with concomitant stimulation of its kinase activity (Yaneva et al., 1997). Ac- tivated DNA-PKcs can phosphorylate itself and a variety of other proteins, including other NHEJ components. Al- though autophosphorylation appears to be essential to DNA-PK function in NHEJ (Ding et al., 2003; Block et al., 2004), apart from activating Artemis (Ma et al., 2005) and promoting dissociation of histone H1 from nucleosomes (Kysela et al., 2005), the physiological rel- evance and mechanistic function of other phosphoryla- tion reactions is far from clear (for example Douglas et al. [2005]). The size and complexity of this multicomponent repair system presents severe challenges for structural analy- sis. Several electron microscopy (EM) studies have pro- vided low-resolution views of DNA-PKcs (Chiu et al., 1998; Leuther et al., 1999; Boskovic et al., 2003), but only recently have we attained sufficient resolution to provide a reasonable model of its molecular architecture (Rivera-Calzada et al., 2005). The cryo-EM structure of DNA-PKcs shows the organization of the protein into conserved regions of its 4128 residue amino acid sequence (Figure 1B). The long and distinctive N-termi- nal region of DNA-PKcs, probably formed by an extended series of HEAT and related helical repeats (Brewerton et al., 2004), maps into the long curved tubu- lar-shaped domains within the arm region (Figure 1B, orange). The C terminus containing the conserved PI3K-related catalytic domain (residues 3649–4011 in DNA-PKcs), a weakly conserved w500 residue helical- repeat region immediately N-terminal of this (the FAT (Figure 1B, magenta, green, and blue). A recent solution structure for the last part of the FATC domain from the related mTOR is fully consistent with this assignment in the EM structure (Dames et al., 2005). Results and Discussion We have now purified to homogeneity a human DNA- PKcs/Ku70/Ku80 complex assembled on DNA (Fig- Figure 1. Purification and EM of DNA Bound DNA-PKcs/Ku70/Ku80 Complexes (A) Atomic structure of the Ku70/Ku80 dimer (PDB entry 1JEQ), lacking the C-terminal evolutionary distinct domains, before (top panels) and after (bottom panels) filtration to 25 A˚ resolution. Each subunit in the atomic representation has been colored differently. The scale bar represents 70 A˚. (B) 3D structure of DNA-PKcs taken from Rivera-Calzada et al. (2005) filtered at 30 A˚. Coloring depicts the assignment of domains in the sequence of DNA-PKcs (top row) into the 3D structure, according to Rivera-Calzada et al. (2005). The scale bar represents 70 A˚. (C) Purification of DNA-PKcs, Ku70, and Ku80 and isolation of their complex. Glycerol gradients performed with the purified proteins in the absence (i) and presence of DNA (ii). Only the ‘‘short’’ DNA is shown, but similar results were obtained with the ‘‘long’’ DNA (see Experimental Procedures and Figure S1). The purified complexes (iii) contain DNA-PKcs, Ku70, Ku80, and DNA. (1) DNA-PKcs alone, (2) complexes with short DNA, and (3) complexes with long DNA. (D) EM field of purified complexes. Asterisks (*) point to some representative views. Arrows point to putative dimeric aggregates. The scale bar represents 140 A˚. (E) A collection of selected pairs of projections and their class averages of the complex as provided by EMAN (i) from untilted micrographs and (ii) from tilted micrographs. (F) 2D projections of the Ku dimer crystal structure (i), DNA-PKcs cryo-EM structure (ii), and the DNA-PKcs/Ku70/Ku80 complex (iii). Molecular Cell 512 domain), and a narrow w100 residue C-terminal exten- sion at the end of the catalytic domain (FATC domain) (Bosotti et al., 2000) all locate in the head region ure 1C). Full-length DNA-PKcs, Ku70, and Ku80 were purified from HeLa cell nuclear extracts by using column chromatography and glycerol gradient centrifugation (see Experimental Procedures). Under these conditions, and in the absence of added DNA, the Ku70 and Ku80 proteins copurified as expected but did not copurify with DNA-PKcs (Figure 1C, i). After observations of neg- ative regulation of DNA binding by DNA-PKcs autophos- phorylation (Merkle et al., 2002), DNA-PKcs fractions were dephosphorylated with lambda phosphatase to improve the homogeneity of the preparation and restore the DNA-PKcs to a high-affinity DNA binding state more in accordance with an initial stage of the NHEJ reaction. To form a DNA complex, a DNA molecule containing a 35 base pair duplex segment with a 19-base Y structure at one end (short DNA) was incubated with the purified DNA-PKcs and Ku70/Ku80 proteins. A DNA molecule containing 50 bp plus the 19 base Y structure (long DNA) was also used and similar results obtained (Figure S1 available in the Supplemental Data with this article online). The DNA was designed on the basis of previous structural studies of DNA-PKcs (Boskovic et al., 2003) and Ku (Walker et al., 2001) to be long enough to bind both but with a blocked end to prevent migration along the DNA and no protruding duplex DNA that might permit loading of multiple Ku hetero- dimers. The rationale behind this design was the result of a compromise between two independent require- ments: assure simultaneous binding of both Ku70/ Ku80 and DNA-PKcs and reduce the conformational flexibility and structural heterogeneity of the DNA-pro- tein complex, also avoiding DNA protruding in several orientations. The analysis of macromolecules using sin- gle-particle EM has the requisite of averaging views of the same complex in the same conformation, and con- formational heterogeneity can distort and complicate these types of studies. We added either 21 or 36 addi- tional bp to the DNA length used by Walker et al. (2001) to allow for the accommodation of DNA-PKcs. Both DNA sizes were capable of assembling a DNA/ Ku70/Ku80/DNA-PKcs complex, and we decided to fo- cus on the short DNA, which should be more adequate for the EM studies. When this DNA-protein mixture was resubjected to glycerol gradient centrifugation, a heavy species sedi- menting distinctly from any of the individual compo- nents was isolated in good yield (Figure 1C, ii) and found to contain a stoichiometric complex of DNA-PKcs, Ku80, Ku70, and DNA as judged from silver-stained gels (Figure 1C, iii). This complex was loaded onto EM grids and observed under the electron microscope after negative staining with uranyl acetate (Figure 1D and Figure S2). Fields showed the characteristic views of DNA-PKcs particles seen in previous studies (Chiu et al., 1998; Leuther et al., 1999; Boskovic et al., 2003; Rivera-Calzada et al., 2005) but with some additional density clearly evident (Figure 1D, asterisk; and Figure S2, squares). In addition, some particles seemed to correspond to larger assemblies (Figure 1D, arrowed; and Figure S2, circles), which were analyzed separately (see below). Initially,w6000 particles of individual mole- cules were extracted and processed in 2D and 3D with the EMAN software package (Ludtke et al., 1999). This EM Structure of DNA-PK/Ku-DNA Complex 513 analysis revealed that the large majority of particles cor- responded to side views (Figure 1E, i), indicating a pre- ferred orientation in binding to the EM grid. To overcome this problem, the sample holder was tilted 40º and an- other w7000 images collected (Figure 1E, ii). Finally, 14,239 images were used and found to cover Euler an- gles adequately (data not shown). The 2D projections and averages obtained from the combined data dis- played the distinctive morphology of DNA-PKcs in side view, with additional lobules of density at the level of the arm region, very evocative of a projection of the Ku dimer (Figure 1F). A 3D reconstruction of DNA bound DNA-PKcs/Ku70/ Ku80 at 25 A˚ resolution was built by angular refinement (Figures 2B, i, displayed at a threshold showing w55% of the calculated protein mass in order to highlight struc- tural details; iii, displayed at a threshold to show 100% of the protein mass). Minimization of model bias was one of the main goals during 3D refinement and was achieved by the use of several starting volumes, including ran- domly generated Gaussian blobs, all of which pro- gressed toward similar solutions. The structure of the complex is fully compatible with the cryo-EM map of isolated DNA-PKcs (Figures 2A and 1B) so that domains can be mapped by visual comparison. Ku is found to sit on top of DNA-PKcs with contact points that expand from the back of the head region to the tubular N-terminal arm. The location of Ku was scrutinized more precisely by calculating the difference between the cryo-EM map of DNA-PKcs and the 3D structure of the complex after proper filtering, scaling, and align- ment (Figure 2B, ii). A substantial region of additional density is present in the complex compared to the iso- lated kinase, indicating the probable location of the bound Ku. The structure of Ku70/Ku80 (without DNA) was fitted into this map after removing the small C-ter- minal domains of both subunits, which are known to dis- locate when bound to DNA, and whose individual posi- tion could not be reliably determined at this resolution (Figure 2B, iv). Although correlation-based density fit- ting (Wriggers et al., 1999) is less accurate in negative stain maps compared to cryo-EM maps, all top-ranked solutions obtained by using Situs (Wriggers et al., 1999) placed Ku such that it fully occupied the additional lobular density, specifically within the main volume re- vealed by the difference map. The best solution was readjusted manually to better account not only for the new densities but also for their connectivity (Figure 2B, iv, DNA shown as yellow balls). The docked position of Ku within the complex suggests that Ku makes exten- sive interactions with several distinct regions of DNA- PKcs, including the HEAT repeats and their projecting claws, and contacts the head close to the expected po- sition of the kinase domain. Consistent with this, a region immediately N-terminal to the kinase domain (residues 3002–3850) has previously been implicated in interac- tions with Ku (Jin et al., 1997). The C-terminal domain (CTD) of Ku80 comprises a small helical region (Harris et al., 2004; Zhang et al., 2004), which has been shown to contain a module that drives the interaction with DNA-PKcs, in a mechanism that might be shared by other PIKK kinases (Falck et al., 2005). Our structure suggests that this CTD might be required for initial re- cruitment of DNA-PKcs, but more extensive contacts in- volving large parts of DNA-PKcs probably stabilize the interaction. Additionally, DNA-loaded DNA-PKcs/Ku70/ Ku80 complexes were incubated with ATP to permit au- tophosphorylation, and the sample was then studied by Figure 2. 3D Structure of DNA-PKcs/Ku70/Ku80 Using Negative Staining EM (A) Front and back views of the 3D structure of DNA-PKcs taken from Rivera-Calzada et al. (2005). Regions approximately corresponding to the catalytic PIKK domain and the N-terminal repeats are labeled. (B) Front and back views of the 3D structure of DNA bound DNA- PKcs/Ku70/Ku80 complexes at a threshold displaying w55% of the volume (i) and 100% (iii). Panel (ii) shows the difference map be- tween the structures of free DNA-PKcs and the complex shown at a similar threshold to that in (i). Red areas denote the presence of ad- ditional density in the Ku-containing complex. (iv) Fitting of core Ku dimer into the 3D reconstruction of the complex, using the DNA bound Ku structure (PDB entry 1JEY). Ku70 and Ku80 subunits in Ku are colored as red and blue ribbons, whereas DNA, as taken from the DNA bound atomic structure of Ku, is shown as yellow spheres. Thresholds showing w55% of the protein mass are used Molecular Cell 514 to visually highlight some of the structural features present in the map that can be masked when displaying a 100% of the density. EM. We derived two main kinds of averages: one type re- sembling those previously found in isolated DNA-PKcs (Rivera-Calzada et al., 2005) and another that had a smaller size compatible with a Ku70/Ku80 dimer (data not shown). Therefore, this result supported that the density assigned to Ku70/Ku80 in the complex was probably so because this density was missing under conditions promoting a disassembling of DNA-PKcs/ Ku70/Ku80. To further support the results obtained from negative stain, we also performed a 3D reconstruction of the complex from specimens vitrified in liquid ethane (cryo-EM) so to better preserve the native conformation of the molecule. Limitations due to the buffers used in purification, the amount of purified material, and its con- centration precluded the collection of very large num- bers of images, but w4000 images were obtained to generate aw30 A˚ map (Figure 3). This map was recon- structed independently, without using information from the negative stain reconstruction, using either random Gaussian blobs or a model built after reference-free im- age classification and common lines as implemented in Eman (Ludtke et al., 1999). The problem of preferred ori- entation encountered in negative staining was greatly reduced in the cryo-EM data, allowing a 3D reconstruc- tion without tilting the sample holder. Although the res- olution of the cryo-EM structure is lower than the struc- ture derived from negative stain, and the different domains in Ku are less well defined, the two structures are extremely similar and the location of Ku in the com- plex is unambiguous in both cases (Figure 3B). We per- formed a fitting of Ku70/Ku80 into this cryo-EM volume but constrained to the Ku density revealed after differ- ence mapping with the cryo-EM structure of isolated DNA-PKcs. Under these conditions, SITUS provided a solution fully compatible with that obtained for the negatively stained structure (Figure 3B, ii). The different regions in this volume could be readily assigned to domains in DNA-PKcs by comparison with its cryo-EM structure in isolation (Rivera-Calzada et al., 2005) using both the volume data and a partial pseudo-atomic model (data not shown). This compari- son reveals substantial conformational changes upon DNA and Ku binding (Figures 3D and 3E) within the ki- nase itself, which are reminiscent of those previously observed on DNA binding by DNA-PKcs (Boskovic et al., 2003), and can be related to changes in the posi- tion of individual segments of the protein by reference to the much higher resolution cryo-EM structure of iso- lated DNA-PKcs. In particular, there are changes at the level of the head region that correspond to movements of the projecting segment of the FAT domain and changes in the N-terminal repeat region that correspond to movements of the distal claw. Most striking is a change in the conformation of the extreme C-terminal FATC domain, which forms a slender tubular projection from the head. In the DNA-PKcs/Ku70/Ku80 complex, this feature is rotated by w90º from its position in the apo-DNA-PKcs to interact with the distal claw of the N-terminal HEAT-repeat segment and with the addi- tional density corresponding to the Ku heterodimer itself. The direct connection between the FATC domain and the C-terminal lobe of the kinase domain of DNA- PKcs suggests a key role for the FATC domain in regulating the kinase activity of DNA-PKcs in response to binding