The 3D Solution Structure of the C-terminal Region ofKu86 (Ku86CTR)

Richard Harris1,2, Diego Esposito2, Andrew Sankar3

Joseph D. Maman4, John A. Hinks4, Laurence H. Pearl1,4 andPaul C. Driscoll1,2,3*

1Bloomsbury Centre forStructural Biology, UniversityCollege London, Gower StreetLondon WC1E 6BT, UK

2Department of Biochemistryand Molecular BiologyUniversity College LondonGower Street, London WC1E6BT, UK

3Ludwig Institute for CancerResearch, 91 Riding HouseStreet, London W1W 7BS, UK

4Section of Structural Biologyand CR-UK DNA RepairEnzyme Group, Institute ofCancer Research, 237 FulhamRoad, London SW3 6JB, UK

In eukaryotes the non-homologous end-joining repair of double strandbreaks in DNA is executed by a series of proteins that bring about thesynapsis, preparation and ligation of the broken DNA ends. The mechan-ism of this process appears to be initiated by the obligate heterodimer(Ku70/Ku86) protein complex Ku that has affinity for DNA ends. Kuthen recruits the DNA-dependent protein kinase catalytic subunit (DNA-PKcs). The three-dimensional structures of the major part of the Kuheterodimer, representing the DNA-binding core, both free and bound toDNA are known from X-ray crystallography. However, these structureslack a region of ca 190 residues from the C-terminal region (CTR) of theKu86 subunit (also known as Lupus Ku autoantigen p86, Ku80, orXRCC5) that includes the extreme C-terminal tail that is reported to besufficient for DNA-PKcs-binding. We have examined the structuralcharacteristics of the Ku86CTR protein expressed in bacteria. By deletionmutagenesis and heteronuclear NMR spectroscopy we localised a globu-lar domain consisting of residues 592–709. Constructs comprisingadditional residues either to the N-terminal side (residues 543–709), orthe C-terminal side (residues 592–732), which includes the putativeDNA-PKcs-binding motif, yielded NMR spectra consistent with theseextra regions lacking ordered structure. The three-dimensional solutionstructure of the core globular domain of the C-terminal region of Ku86(Ku86CTR592 – 709) has been determined using heteronuclear NMR spec-troscopy and dynamical simulated annealing using structural restraintsfrom nuclear Overhauser effect spectroscopy, and scalar and residualdipolar couplings. The polypeptide fold comprises six regions of a-helicalsecondary structure that has an overall superhelical topology remotelyhomologous to the MIF4G homology domain of the human nuclear capbinding protein 80 kDa subunit and the VHS domain of the Drosophilaprotein Hrs, though strict analysis of the structures suggests that thesedomains are not functionally related. Two prominent hydrophobicpockets in the gap between helices a2 and a4 suggest a potential ligand-binding characteristic for this globular domain.

q 2003 Elsevier Ltd. All rights reserved.

Keywords: Ku; DNA repair; protein structure; NMR spectroscopy; DNA-PK*Corresponding author

Introduction

It is well established that double strand breaks(DSB) in chromosomal DNA are generated by theaction of reactive oxygen species and ionisingradiation, as well as in normal nuclear processessuch as replication, recombination and generearrangements in eukaryotes (e.g. mating typeswitching; V(D)J recombination). It is crucial for

0022-2836/$ - see front matter q 2003 Elsevier Ltd. All rights reserved.

E-mail address of the corresponding author:p.driscoll@ucl.ac.uk

Abbreviations used: NMR, nuclear magneticresonance; CTR, C-terminal region; DNA-PK, DNA-dependent protein kinase; DNA-PKcs, DNA-PK catalyticsubunit; DSB, double strand break; NHEJ, non-homologous end-joining; NOE, nuclear Overhausereffect; HSQC, heteronuclear single quantum coherence;NOESY, NOE spectroscopy; TOCSY, total correlationspectroscopy; RMSD, root mean square difference.

doi:10.1016/j.jmb.2003.10.047 J. Mol. Biol. (2004) 335, 573–582

cells to recognise and repair this type of DNAlesion because un-repaired or improperly repairedDSB can lead to cell cycle arrest and cell death, orchromosome translocations that result in increasedrates of mutation and ultimately carcinogenesis.Eukaryotic cells repair DSB by either the homo-logous recombination pathway, or the non-homo-logous end-joining (NHEJ) pathway.1 Bothpathways are present in all eukaryotes. In yeast,repair of DSB is achieved primarily by homologousrecombination, whereas NHEJ repair is the pre-dominant mechanism in vertebrates.2 Recently ishas been shown that primitive NHEJ occurs incertain prokaryotic species too,3 suggesting thatthe repair of DSB by this mechanism is of ancientorigin on the evolutionary time-scale.

The eukaryotic NHEJ process is critically depen-dent upon the XRCC4/DNA ligase IV protein com-plex for the covalent rejoining of the broken DNAends, as demonstrated by the targeted disruptionof the XRCC4 or ligase IV genes in mice that leadsto embryonic lethality.4 – 6 Another obligate com-ponent of this system is a multi-protein complex,DNA-dependent protein kinase (DNA-PK) thatconsists of the DNA end-binding protein dimer,Ku, and a ,470 kDa catalytic subunit, DNA-PKcs,a distant homologue of phosphoinositide3-kinases. Other proteins that are thought to beinvolved in DNA processing at the site of the DSBare the Mre11/RAD50/NBS1 nucleases1 and arecently discovered protein called Artemis,7 whichhas DNA hairpin opening and single strand-specific exonuclease activities in purified form,and hairpin and single strand overhang endo-nucleolytic activity in the presence of DNA-PKcs.8

Unlike XRCC4 or ligase IV null mutants, animalslacking Ku or DNA-PKcs are viable but are unableto correctly rejoin V(D)J recombination inter-mediates and display severe combined immuno-deficiency (SCID) as well as deficiencies in DSBrejoining and hyper-sensitivity to ionisingradiation.2,9,10 The Ku protein, a heterodimer oftwo proteins termed Ku70 and Ku86 (Ku86 isoften alternatively labelled Ku80), binds tightly toDNA ends regardless of the nucleotide sequenceor secondary structure (i.e. blunt, recessed or hair-pinned) and is thought to function in the primaryrecognition of DSB. In mammalian cells, the hetero-dimeric Ku binds to the broken DNA ends andrecruits the catalytic subunit of DNA-PK, whichrequires the association of Ku and the presence ofDNA for full activation of its protein kinaseactivity.11 Kinase activity is apparently only maxi-mised following DNA-PK-mediated synapsis ofDNA ends,12 although it is also reported thatDNA-dependent autophosphorylation of DNA-PKcs and Ku components leads to the downregulation of DNA-binding by DNA-PK.13 In turn,Ku appears to be important in recruiting theXRCC4/DNA ligase IV protein complex, whichultimately rejoins the DNA ends, to DSB.14 Ku isalso implicated in the maintenance of telomeres:loss of Ku leads to telomeric shortening and to an

enhanced rate of chromosome end-fusion.15,16

Therefore another potential function of Ku inNHEJ is the protection of the broken DNA endsfrom excessive nucleolytic attack.

Recently the high-resolution three-dimensionalstructure of a truncated form of the Ku70/Ku86heterodimer, free and bound to duplex DNA, hasbeen solved by crystallisation and X-raydiffraction.17 However, the C-terminal domain ofKu86 (residues 545–732) is absent from the crystalstructure because part of this domain (residues579–732) has been truncated in the expression con-struct, presumably to allow crystallisation, and therest of this region (residues 546–578) is disorderedin the crystal. The structure clearly shows whyDNA end binding depends upon heterodimerisa-tion of the two chains: the DNA duplex is collaredby a pseudo-symmetric arrangement of the twoN-terminal portions of the Ku70 and Ku86 chains.Although low in overall sequence identity(,15%), there is a high structural similarity,reflected in the pseudosymmetry between Ku70and Ku86 except in the C-terminal regions that aredispensable for DNA binding. The small 51 residueC-terminal globular domain of Ku70 in the DNA-free crystal form packs against the main body ofthe DNA-collaring domains, to which it is linkedby a peptide chain that is not interpretable in theelectron density. The 3D solution structure of theKu70 C-terminal domain has been separatelyobtained in NMR studies.18 These characteristicssuggest that this domain is flexibly appended tothe core Ku70/Ku86 heterodimer. The Ku70C-terminal domain is homologous to the DNA-binding SAP domain,19 was shown to bind weaklyto DNA20 and is potentially involved in pausingKu at specific DNA sequences.17

Genetic and biochemical experiments suggestthat the C-terminal region of Ku86 is necessary forV(D)J recombination and for restoring resistanceto ionising radiation.21 DNA-PKcs is reported tohave DNA-binding and kinase activity in theabsence of Ku. However, it has been shown thatKu can enhance the binding of DNA-PKcs toDNA and stimulate DNA-PKcs kinase activity upto tenfold.22 Deletion of the C-terminal 178 residuesof Ku86 does not disrupt heterodimerisation of theKu subunits, or their DNA end-binding ability,but decreases Ku’s ability to stimulate DNA-PKkinase activity and induces a SCID-like phenotypedespite the presence of DNA-PKcs.21 It has beenproposed that the Ku heterodimer is responsiblefor recruiting DNA-PKcs to DSB, and it may bethat the C-terminal domain of Ku86 performs thisfunction.23 Although some micro-organismspossess Ku70/Ku86 orthologues,24,25 yeast andbacteria (and one archaeal species) do not appearto contain a DNA-PKcs homologue, and DSBrepair in these species must proceed by a differentmechanism. Interestingly, there is no similarC-terminal domain in the yeast or bacterial homo-logues of the mammalian Ku86 subunit.

Clearly the C-terminal region of Ku86 performs

574 3D Solution Structure of Ku86CTR

an important function within the process of DSBrepair, though the precise nature of its role has yetto be established. We report here on the 3Dsolution structure of this domain, which appearsto adopt a unique polypeptide fold, and provideevidence that a central globular portion of 118residues is flanked by unstructured N and C tails.We discuss the implication of these results for theputative function of this domain.

Results

Optimisation of Ku86 CTR construct

The X-ray crystal structure of the Ku70/Ku86heterodimer core obtained by Goldberg andco-workers17 shows that the Ku86 is well orderedup to residue 545, and the limit of the weaksequence and stronger structural similarity withthe Ku70 subunit extends as far as residue 540.This analysis defines the C-terminal domain ofKu86 as starting after residue 542. We initiallyattempted to solve the structure of the intact Ku86C-terminal domain (residues 543–732) by X-raycrystallography and NMR but we could not obtaincrystals or tractable NMR spectra. Controlledtryptic digests of Ku86CTR543 – 732 generated a com-plex pattern of peptides (results not shown) thatindicated both N and C-terminal deletions of theintact domain. We therefore generated a series ofdeletion polypeptides of Ku86CTR543 – 732. The firstindication that suggested a tractable globular formfor the Ku86 C-terminal region was in the NMRspectra of a construct comprising residues543–709 (i.e. lacking the final 23 residues). Wehave previously shown by heteronuclear NMRspectroscopy that the residues towards the Nterminus of Ku86CTR543 – 709 exhibit characteristicsof high internal mobility: {1H}15N heteronuclearnuclear Overhauser effect (NOE) values ,0.5,amide proton chemical shifts close to random coilvalues, and high NH/solvent exchange rates(amide protons not observed in [15N,1H]-HSQCspectra at pH 8).26 These data indicate a dynami-cally disordered region without well-definedregular secondary structure for residues 543–590(Figure 1(a)). A second construct (Ku86 residues592–732) was made to ascertain whether the final23 residues were similarly disordered or formadditional secondary structure elements. The[15N,1H]-HSQC spectra for this construct combinedwith analysis of the heteronuclear NOE values and1H chemical shifts of cross-peaks not appearing inthe Ku86CTR543 – 709 spectrum show that theseC-terminal residues are also highly flexible(Figure 1(c)). Following these observations, a finalexpression construct was made encoding Ku86residues 592–709. The NMR spectra of this proteinwere essentially devoid of cross-peaks with charac-teristics of large amplitude internal motion, andprovided the basis for further NMR studies and3D solution structure determination (Figure 1(b)).

Experiments were performed to obtain residue-by-residue longitudinal (R1) and transverse (R2)

15Nrelaxation rates, and quantitative {1H}15N hetero-nuclear NOEs.27 These data were analysed usingstandard methods28,29 within the so-called Lipari–Szabo model-free formalism30,31 yielding anestimate of the molecular rotational correlationtime, tm , 8.9 ns at 298 K. This value is broadly inline with expectations for a non-self-associating120 residue protein of molecular mass 14 kDa.Examination of the hydrodynamic properties ofKu86CTR constructs by dynamic light scatteringand sedimentation ultracentrifugation (data notshown) are also consistent with a monomericstatus.

Resonance assignments and structuredetermination of Ku86CTR592–709

A double [15N,13C]-isotope-labelled Ku86CTR592–709

sample was used in a standard set of triple reson-ance experiments to obtain sequence-specificresonance assignments using a similar strategy tothat reported.26 All amide 1H– 15N correlations,with the exception of His673, Glu689 and Leu705,were identified. Essentially complete side-chainresonance assignments were determined byanalysis of 3D HC(CO)NH, (H)C(CO)NH, CBCA-(CO)NH, HBHA(CO)NH and HC(C)H-TOCSYexperiments. Aromatic ring proton resonanceassignments were obtained using a combination of3D [1H,13C]–TOCSY–HSQC, [1H,13C]–NOESY–HSQC and a 3D version of the 2D (HB)CBHDexperiment.32 The latter experiment was recordedas a 3D spectrum where the extra dispersionafforded by the phenylalanine Hb chemical shiftswas essential for the unambiguous resonanceassignment of the Hd protons (Ku86CTR592 – 709

contains nine Phe residues).The Ku86CTR592 – 709 solution structure calcu-

lations were performed using well-establishedrestrained dynamical simulated annealingprotocols33 modified to allow floating stereo-chemistry, and active swapping, of prochiralcentres.34 Initially a set of conformers was calcu-lated using NOE, hydrogen bond, and dihedralangle restraints alone. The input data consisted ofa total of 1959 NOE-derived inter-proton distancerestraints, 96 hydrogen bond restraints, and 174dihedral angle restraints. Once the overall fold ofthe protein had been established, a further set ofstructures was calculated with 96 one-bond back-bone 1H–15N residual dipolar couplingrestraints included in the final stage of refinement.The inclusion of residual dipolar couplingrestraints does not alter the overall topology ofthe polypeptide chain but improves theprecision of the refined conformer bundle withoutsignificant penalty to any of the restraint terms(Table 1).

Table 1 shows the structural statistics for thefinal lowest energy bundle of 20 conformers, eachof which displays low restraint violations and

3D Solution Structure of Ku86CTR 575

good stereochemical and non-bonded interactionproperties. A best-fit superposition of the backbonefolds of the 20 lowest energy conformers is shownin Figure 2(a). The lowest energy structure isshown in ribbon representation in Figure 2(b).Ku86CTR592 – 709 consists of six a-helices connectedby loops. The six helices are defined by the follow-ing residue numbers: 594–602 (a1); 611–625 (a2);629–649 (a3); 652–668 (a4); 672–681 (a5); and698–704 (a6). A region of well-defined extendedstructure (residues 682–686) connects a5 with an11-residue loop (residues 687–697) that leads intoa6. The significant lower cross-peak intensity inthe [15N,1H]-HSQC spectra, and the relative paucityof medium and long-range NOEs for the residuesof this loop suggest a dynamically disorderedconformation giving rise to chemical exchangeline-broadening on the fast-to-intermediate time-scale. Omitting this loop and a small number ofresidues at both the N and C terminus, therefined conformer bundle provides a well-definedmodel for the Ku86CTR592 – 709 domain with rootmean square difference (RMSD) values of0.57(^0.08) A for the backbone and 1.08(^0.13) Afor all heavy-atoms (residues 593–683 and 694–704).

The overall fold of Ku80CTR592 – 709 can bedescribed as a five-helix bundle with right-handedsuperhelical topology. In the structure, helices a1and a2, and a1 and a4 are arranged in an antipar-allel fashion. The overall topology is characterisedby hydrophobic side-chains that populate interheli-cal interaction surfaces (see Figure 2(c)). In particu-lar, a hydrophobic cluster comprising Phe598 (a1),Phe609 (a2), and Phe642 (a3) fixes the packing ofthe first three helices. The position of helix a5 isdetermined through a hydrophobic cluster com-prising residues Leu671, Phe674, Val678 which

make key interactions with Leu624 (a2), Phe632(a3), and Leu662 (a4); Leu624 (a2), Phe632 (a3),Ile636 (a3), Ile639 (a3), Leu662 (a4); and Ile636(a3), respectively.

The extended structure connecting a5 with theloop connected to a6 is stabilised by hydrophobicinteractions through residues Ile683, Leu685,Ile686 with Ile636 (a3), Ile639 (a3), Val678 (a5);Leu659 (a4), Ala701 (a6), Phe704 (a6), Leu705;and Ile647 (a3), respectively. This packing helps toexplain why this region, although not recognisedas a regular secondary structure element, is rela-tively rigid and well defined by the NMR restraintdata.

Figure 3 shows a space-filling representation of theelectrostatic surface potential of Ku86CTR592–709.Whilst surface patches of charged residues exhi-bit a balanced distribution of positive and nega-tive potential, there is a distinct contiguouspatch of negatively charged residues that linethe groove formed by the helices a4, a5, andthe C terminus. Acidic amino acid residuesthat form this surface include Glu664, Glu667,Glu676 and Asp709. This patch is linkedthrough Glu699 (a6) to another pair of side-chain carboxylate groups (Glu700, a6 andGlu652, a4) to form an extended band of nega-tive electrostatic potential that wraps aroundone half of the C-terminal “end” of the domain(Figure 3(c)). It is particularly intriguing thatthere are two significant invaginations betweenhelices a2 and a4 (Figure 3(b)) that could bedocking sites for a putative protein or peptideligand. These holes are lined by hydrophobicside-chains (Leu624, Ile639, Leu662, Phe674 andIle617, Phe642, Phe655, respectively), and arelarge enough to accommodate at least a ligandmethylene or methyl group.

Figure 1. 2D [1H,15N]-HSQC spectra of the three different protein constructs analysed in this study: Ku86CTRresidues 543–709 (a), residues 592–709 (b), and 592–732 (c). Cross-peaks highlighted in red indicate residues whichgive a negative cross-peak in the {1H}15N heteronuclear NOE experiment, indicating a high degree of dynamic disorderfor these residues.

576 3D Solution Structure of Ku86CTR

Discussion

The overall topology of the Ku86CTR domain isfaintly reminiscent of other protein structures thatcontain superhelical repeats of a-helix secondarystructure elements. This similarity is illustrated inFigure 4, which shows a side-by-side comparison ofthe overall structures of Ku86CTR592-709 with theMIF4G-homology six-helix motif (repeat 1) from thehuman nuclear cap binding protein 80 kDa subunit(Protein Data Bank (PDB) code 1H6K)35 and theVps27p/Hrs/STAM (VHS) domain (PDB code1DVP) from the Drosophila melanogaster proteinHrs.36 These structures have similar overall architec-tures in the sense of having broadly equivalent poly-peptide chain topologies, but the details of the helixpacking arrangements are quantitatively dissimilar.We tested the structural homologies using the pro-gram DALI.37 The closest geometrical match to thehelical combination a2, a3, a4 of Ku86CTR592–709

occurs in the first of the three six-helix motifs fromthe 80 kDa human nuclear cap binding protein, inwhich a superposition over 100 residues yields aRSMD of 2.8 A, with DALI Z score of 5.6. However,the overall Ku86CTR592–709 fold that includes a5 posi-tioned roughly perpendicular to the other helices ismore similar to that of the Hrs VHS domain (3.7 ARMSD over 106 residues; DALI Z score 5.4). Interest-ingly, the VHS domain comprises a total of eighthelices spanning 143 residues. The size of the VHSdomain is the same as Ku86CTR including the extra23 residues that we have shown (on the basis ofqualitative NMR characteristics) to be disordered insolution. Whilst after a5 in Ku86CTR592–709 the struc-ture does not contain extended elements of regularsecondary structure, it is interesting to speculatethat this region including the final 23 residues thatare apparently unstructured in solution but whichare known to be important in binding to the targetprotein, could become ordered upon binding toDNA-PKcs, perhaps to adopt an overall structurethat contains the “missing” C-terminal helicalelements of the homologous VHS domain structure.

The emerging picture of the C-terminal region ofKu86 is a structured region of 118 residues flankedby an unstructured and flexible N-terminal linkerof 48 residues and C-terminal tail of 23 residues.The work of Gell & Jackson24 that investigated

Table 1. Summary of structure statistics forKu86CTR592 – 709

kSAl SAlowest

Experimental restraintsa

All (A) (1959) 0.021 ^ 0.002 0.019Unambiguous (A) (1804)Intraresidue (992) 0.012 ^ 0.001 0.012Sequential (373) 0.021 ^ 0.004 0.021Short (194) 0.031 ^ 0.005 0.033Long (245) 0.021 ^ 0.005 0.021

Ambiguous (A) (59) 0.031 ^ 0.003 0.013Hydrogen bond restraints (A)

(96)0.036 ^ 0.002 0.033

Dihedral angle restraints (8)(174)

0.51 ^ 0.10 0.51

Residual dipolar couplings(Hz) (96)

0.30 ^ 0.04 0.27

Number of residual restraints violationsb

NOE violations .0.4 A 0.4 ^ 0.5 0NOE violations .0.3 A 1.6 ^ 0.9 2Angle violations .38 0.0 0

Deviations from idealised covalent geometryc

Bonds (A) (1967) 0.0015 ^ 0.0001 0.0014Angles (8) (3565) 0.33 ^ 0.01 0.31Improper dihedrals (8) (1016) 0.33 ^ 0.02 0.27

Structural statistics for the ensembled

PROCHECK parametersMost favoured region (%) 80.3 ^ 1.5 79.5Additionally allowed (%) 14.2 ^ 1.8 14.3Generously allowed (%) 3.9 ^ 1.3 6.2Disallowed (%) 1.6 ^ 0.9 0Number of bad contacts 3.6 ^ 1.7 2

CHARMM Lennard-Jones energye

ELJ (kcal mol21) -995 ^ 10 -1005R-factor for residual dipolar

coupling restraintsf (%)0.7 ^ 0.1 0.7

RMSD from the average structureg

Backbone (N, Ca, C) (A) 0.57 ^ 0.08Heavy-atoms (A) 1.08 ^ 0.13

kSAl, represents the set of 20 selected conformers obtained byrestrained dynamical simulated annealing in CNS. SAlowest refersto the lowest energy structure of the set.

a Sum averaging of NOE distance restraints was used forgroups with degenerate proton chemical shifts. The interprotonunambiguous distance restraint list comprised 992 intraresidue,373 sequential ðli 2 jl ¼ 1lÞ; 194 short range ð1 , li 2 jl , 5Þ;and 245 long-range ðli 2 jl . 5Þ: Hydrogen bond restraints wereapplied as pairs of distance restraints: HN· · ·O, 1.2–2.2 A;N· · ·O, 1.2–3.2 A. The final values for the respective forceconstants were: NOE, 30 kcal mol21 A22; H-bonds,50 kcal mol21 A22; dihedral angles, 200 kcal mol21 rad22;residual dipolar couplings, 0.5 kcal mol21 Hz22.

b No experimental distance restraint was violated by morethan 0.5 A and no torsion angle restraint by more than 38.

c The final values for the respective force constants were:bond lengths, 1000 kcal mol21 A22; angles and impropertorsions, 500 kcal mol21 rad22; the improper torsion anglerestraints serve to maintain planarity and chirality.

d The program PROCHECK58 was used to assess the stereo-chemical parameters of the family of conformers. The figuresindicate the percentage of residues with backbone f and cangles in separate regions of the Ramachandran plot, definedin the program. The number of bad contacts per 100 residues isexpected to be in the range 0–30 for protein crystal structuresof better than 3.0 A resolution.

e The CHARMM59 Lennard–Jones van der Waals energyterm, which was not included in the force field of the simulatedannealing of restrained minimisation, was used to assess theatomic packing of the structure. A negative value is consistentwith the absence of any significant atomic clash.

f The dipolar coupling R-factor is defined as the ratio of theRMSD between observed and calculated values to the expected

RMSD if the vectors were randomly orientated.60 The former isgiven by ½

Pi ðD

obsi 2 Dcalc

i Þ2=N�1=2; where Dobs and Dcalc are theexperimentally measured and the predicted dipolar couplingswhich extends over N residues for which dipolar couplings areknown. The latter is given by {2D2

a½4 þ 3R2�=5}1=2; where Da isthe magnitude of the axial component of the alignment tensorand R the rhombicity.

g The precision of the atomic coordinates is defined as theaverage pairwise RMSD between each of the 20 conformersand a mean coordinate structure SA generated by iterativebest-fit of the backbone atoms (N, Ca, and C) over residues593–685 and 697–705 (comprising the core secondary structureelements and omitting the flexible N and C termini, and thedisordered loop between helices a5 and a6, residues 686–696)followed by coordinate averaging.

3D Solution Structure of Ku86CTR 577

Figure 3. Solvent accessible surface representation of Ku86CTR592 –709 coloured according to the electrostatic potential(positive potential blue, negative potential red). (a) The protein in the same orientation of the molecule shown inFigure 2, and (b) the molecule rotated 1808 to show the back face. (c) An orientation between those in (a) and (b) thathighlights the negatively charged region discussed in the text, with charged residues identified by the one-letter sym-bols and the sequence numbers. The positions of a-helices (green) and the location of the hydrophobic pocketsbetween a2 and a4 (yellow arrows) are indicated. These images were created in the program GRASP.55

Figure 4. Comparison of theKu86CTR592 – 709 domain (a) withstructurally homologous domainsidentified by the program DALI:the N-terminal MIF4G-homologysix-helix domain (repeat 1) fromhuman 80 kDa nuclear cap bindingprotein (PDB code 1H6K) (b); andthe VHS domain form D. melanoga-ster Hrs (PDB code 1DVP) (c). Thestructures were superimposedusing the program TOP56 and areshown in an orientation consistentwith the best-fit superposition. Thecartoon representation of theKu86CTR592 – 709 and VHS domainshighlights the similar topologicalarrangement of helices 1–5 of thesetwo folds.

Figure 2. (a) Representation of the bundle of the 20 lowest energy conformers of Ku86CTR592 – 709 after best-fit super-position of the backbone atoms. (b) A ribbon drawing of the lowest energy conformation of Ku86CTR592 –709 illustratingthe a-helical secondary structure elements. This Figure was prepared with the program MOLSCRIPT.54 (c) Represen-tation of the key hydrophobic side-chains that populate the interhelical interaction surfaces (see main text fordiscussion): helix a1 (red); a2 (orange); a3 (green); a4 (cyan); extended structure (magenta); a6 (violet).

578 3D Solution Structure of Ku86CTR

deletion mutants of Ku86 demonstrated that the last12 residues of Ku86CTR (whose sequence in oneletter amino acid code is ,EGGDVD DLLDMI-COOH) are necessary and sufficient to bind toDNA-PKcs in a glutathione S-transferase fusion pro-tein “pull-down” assay. Arguably, it is somewhatunexpected to find that the region apparently bear-ing the responsibility for binding DNA-PKcs is in arandom coil conformation in solution. However, it ispossible that this mainly hydrophobic and overallnegatively charged peptide, may bind a putativehydrophobic/basic pocket in the 470 kDa DNA-PKcs, possibly accompanied by coil-to-orderinduced-fit transition. A clear precedent for such amechanism is provided by the binding-inducedhelix formation that occurs when the acidic LXXLL-containing motif of the C-terminal domain of thetranscription factor IIF (TFIIF)-associated phospha-tase FCP1 interacts with the large subunit of TFIIF,RAP74.38 This peptide, whose sequence shows over-all similarity to the extreme C terminus of Ku86, isunstructured in the absence of its binding partnersbut folds in to an amphipathic a-helix in the complexwith its target. In this context it is noteworthy thatsecondary structure prediction algorithms suggestthat the extreme C-terminal tail of Ku86 (residues723–730) has a high probability to fold as an a-helix. Therefore, although the C-terminal 23 residuesof Ku86 are unstructured and flexible in solution, we

cannot exclude the possibility that this region isordered in solution in vivo either when bound toDNA ends or DNA-PKcs. It is also interesting tonote that the Arabidopsis thaliana plant homologue ofthe Ku86 protein contains the residues that are hom-ologous to the globular domain of mammalian iso-forms, but lack these last residues that we find aredisordered (Figure 5). However, DNA-PKcs, whichin mammalian cells is recruited by the Ku86CTR toDSB, is apparently absent in A. thaliana,39 supportingthe hypothesis that this disordered peptide is respon-sible for recruiting DNA-PKcs in mammalian cells.

It has been shown that the Ku heterodimer canbe phosphorylated in vitro by DNA-PKcs in aDNA-dependent manner.40 Ku70 is phosphoryl-ated on Ser6, Ku86 on Ser577, Ser580 and Thr715.All the three Ku86 phosphorylation sites lie justoutside the globular region in the C-terminaldomain. These modifications may play an import-ant role in the function of the Ku86CTR, as demon-strated by the work of Merkle et al.13 who showedthat holo-DNA-PK (i.e. DNA-PKcs and Ku)interaction with DNA is dramatically reduced byauto-phosphorylation. It is possible that phos-phorylation of the Ku86 C-terminal region maycause structural alterations that preclude the pro-ductive interaction of the flexible C-terminal tailwith DNA-PKcs, thereby severely reducing theaffinity of the kinase domain for DNA.

Figure 5. Sequence alignment of known Ku86CTR homologous proteins, together with a cartoon representation ofthe secondary structure elements of the Ku86CTR592 – 709 globular domain. The alignment includes the disorderedC-terminal tail region that is highly conserved at the extreme C terminus but is missing from the plant species(A. thaliana) homologue. The top line of the alignment summarises the relative surface exposure57 of that residue’sside-chain in the human Ku86CTR592 – 709 solution structure on a scale of 0 (,10% exposed) to 9 (.90% exposed).

3D Solution Structure of Ku86CTR 579

As yet there is no specific Ku function associatedwith residues embedded within the globularregion of the Ku86CTR. Since the structure of theglobular region is apparently unique and only dis-tantly related to other protein folds with diversebiological roles we cannot assign a putative func-tion by structural homology. It may be that thisdomain plays a role in other functions of Ku, suchas in recruiting or providing optimal subunit pack-ing for other members of the NHEJ pathway to theDSB site (e.g. DNA ligase IV/XRCC4). The analysisof the structure of the Ku86CTR provided hereinmay provide useful information for furtherexploration of the structural biology of NHEJ,which apart from the Ku heterodimer core haslargely proved intractable to date.

Materials and Methods

Sample preparation

The expression and purification of 13C, 15N-labelledKu86CTR592 – 709 in Escherichia coli was performed asdescribed for Ku86CTR543 – 709.

26 The NMR sample wasprepared in 330 ml 90% H2O/10% 2H2O solution contain-ing 20 mM potassium phosphate buffer at pH 7.0,100 mM NaCl, and 1 mM NaN3, resulting in approxi-mately 1.0 mM protein concentration.

NMR spectroscopy

NMR spectra were acquired at 298 K (except whereindicated) on Varian UnityPLUS and UnityINOVAspectrometers (operating at nominal 1H frequencies of600 MHz and 800 MHz, respectively) each equippedwith a triple resonance probe including a Z-axis pulsefield gradient coil. Sequence-specific backbone resonanceassignments were obtained using through-bond tripleresonance NMR spectroscopy as described forKu86CTR543 – 709.

26 Side-chain resonance assignmentswere obtained from 3D 15N-separated TOCSY-HSQC(80 ms mixing time), 3D H(CCO)NH-TOCSY (12 ms),3D (H)C(CO)NH-TOCSY (12 ms), and 3D [1H,13C]-HCCH-TOCSY (20 ms) spectra. Aromatic ring protonresonance assignments were obtained from a 3D versionof the 2D (HB)CBHD32 and [1H,13C]-TOWNY-HSQCspectra. Qualitative side-chain x1 angle restraints wereobtained from inspection of 3D HNHB, 3D [1H,15N]-NOESY-HSQC (40 ms mixing time), and 3D [1H,13C]-NOESY-HSQC (100 ms mixing time) spectra.41 Distancerestraints were derived from 3D 15N and 13C-editedNOESY-HSQC spectra with a mixing time of 100 ms.{1H}15N heteronuclear NOE data were recorded with3.0 seconds 1H saturation in the latter part of a3.5 seconds preparation period delay, which was alsoused without radio frequency pulses for the reference2D spectrum without NOE. One-bond 1H–15N residualdipolar couplings were obtained from [1H,15N]-IPAP-HSQC spectra42 acquired at 293 K in the presence andabsence of 5% n-octyl-penta(ethylene glycol):octanol,0.96:1.43 All NMR spectra were processed usingNMRpipe/NMRDraw44 and analysed using ANSIG foropenGL v1.0.3.45 1H, 13C, and 15N chemical shifts werereferenced indirectly to sodium 2,2-dimethyl-2-silane-pentane-5-sulphonate (DSS), using absolute frequencyratios for the 1H signals.46

Structure calculations

Interproton distance restraints were derived from theANSIG cross-peaks file of 3D [1H,15N]-NOESY-HSQCand [1H,13C]-NOESY-HSQC spectra. A proportion of theresonances were successfully assigned in a manualfashion without ambiguity. The remaining cross-peaksappearing at positions in the spectrum with overlappingresonances were labelled with ambiguous assignmentsby reference to the chemical shift list obtained withthrough-bond correlation spectra, using the “Connect”module from the program AZARA.47 The cross-peakswere grouped into four categories according to theirrelative peak intensities: strong, medium, weak, andvery weak, and were designated with the correspondinginterproton distance restraint limit of 1.8–2.5 A, 1.8-3.0 A, 1.8–3.5 A, 1.8–4.0 A, respectively. 0.5 A per methylgroup was added to the upper bound of the distancerestraint for NOE cross-peaks that involved methylgroups.

The structure calculations were carried out within theCNS program33 using the PARALLHDGv5.1 parameterset, with non-bonded energy function of PROLSQ,48

with modifications to allow floating stereochemistry,and to include active swapping, of the atom labels atprochiral centres.34 The ambiguous distance restraintswere then filtered iteratively, based upon the coordinatesof an ensemble of intermediate conformers, and redun-dant (duplicate) restraints were discarded. A total of1959 NOE-derived inter-proton distance restraints wereincluded in the final iterations of the structure calcu-lations (see Table 1).

Backbone torsion angle restraints for f and c werederived from analysis of 1Ha, 13Ca, 13Cb, 13C0, and15NH chemical shift databases as implemented in theprogram TALOS.49 TALOS dihedral restraints wereapplied to all residues for which a statistically signifi-cant fit was obtained and were assigned an errorrange of ^208. Hydrogen bond restraints for amideprotons were derived from an assessment of theregular secondary structure elements. This analysisincluded the overall and local patterns of NOEs, theanalysis of backbone atom chemical shifts using theprogram CSI,50 and the pattern of amide proton sol-vent exchange rates. A total of 163 dihedral angleand 96 hydrogen bond interatomic distance (48 H-bonds; two distance restraints per H-bond) restraintswere used. A total of 11 side-chain x1 anglerestraints were obtained from a qualitative inspectionof the intensities in 3D HNHB, [1H,15N]-NOESY-HSQC and [1H,13C]-NOESY-HSQC spectra and wereassigned an error range of ^308.

Initial estimates of the value of the axial and rhombiccomponents of the molecular alignment tensor wereobtained by examination of the distribution of NHresidual dipolar couplings,51 17.5 Hz and 0.476 Hz,respectively. These values were used as a starting pointfor a procedure that simultaneously refined the proteinstructure and ascertained the values of the alignmenttensor components using a grid search approach. Thefinal values of the axial component and rhombicity ofthe molecular alignment tensor used in the refinementstage were 28.7 Hz and 0.587 Hz, respectively. Theprograms SSIA52 and MODULE-153 were used tocompare the predicted and measured residual dipolarcouplings for models of the Ku86CTR molecular structureand for the calculation of the structure quality factors(see Table 1).

580 3D Solution Structure of Ku86CTR

Data Bank accession numbers

The atomic coordinates of the final 20 simulatedannealing Ku86CTR592 – 709 conformers and the list ofexperimental restraints have been deposited at theRCSB Protein Data Bank (accession code 1Q2Z). Chemi-cal shifts for resonance assignments for Ku86CTR592 – 709

have been deposited at the BioMagResBank (accessionnumber 5912).

Acknowledgements

This is a contribution form the Bloomsbury Centrefor Structural Biology supported by the BBSRC. Thiswork was additionally supported by the LICR(P.C.D. and A.S.) and by CR-UK (J.D.M., J.A.H. andL.H.P.). We are grateful to Dr G. Kelly for providingassistance with the 800 MHz spectrometer at theMRC National NMR Centre at the National Institutefor Medical Research, Mill Hill, London.

References

1. Khanna, K. K. & Jackson, S. P. (2001). DNA double-strand breaks: signaling, repair and the cancerconnection. Nature Genet. 27, 247–254.

2. Featherstone, C. & Jackson, S. P. (1999). Ku a DNArepair protein with multiple cellular functions.Mutat. Res. 434, 3–15.

3. Hiom, K. (2003). DNA repair: bacteria join in. Curr.Biol. 13, R28–R30.

4. Barnes, D. E., Stamp, G., Roswell, I., Denzel, A. &Lindhal, T. (1998). Targeted disruption of the geneencoding DNA ligase IV leads to lethality inembryonic mice. Curr. Biol. 8, 1395–1398.

5. Frank, K. M., Sekiguchi, J. M., Seidl, K. J., Swat, W.,Rathbun, G. A., Cheng, H. L. & Alt, F. W. (1998).Late embryonic lethality and impaired V(D)J recom-bination in mice lacking DNA ligase IV. Nature, 396,173–177.

6. Gao, Y., Sun, Y., Frank, K. M., Dikkes, P., Fujiwara, Y.,Seidl, K. J. et al. (1998). A critical role for DNA end-joining proteins in both lymphogenesis and neuro-genesis. Cell, 95, 891–902.

7. Moshous, D., Callebaut, I., de Chasseval, R., Corneo,B., Cavazzana-Calvo, M., Le Deist, F. et al. (2001). Arte-mis, a novel DNA double-strand break repair/V(D)Jrecombination protein, is mutated in human severecombined immune deficiency. Cell, 105, 177–186.

8. Ma, Y., Pannicke, U., Schwarz, K. & Lieber, M. R.(2002). Hairpin opening and overhang processing byan Artemis/DNA-dependent protein kinase complexin non-homologous end joining and V(D)J recombi-nation. Cell, 108, 781–794.

9. Smith, G. C. & Jackson, S. P. (1999). The DNA-depen-dent protein kinase. Genes Dev. 13, 916–934.

10. Critchlow, S. E. & Jackson, S. P. (1998). DNA end-join-ing: from yeast to man. Trends Biochem. Sci. 23, 394–398.

11. Gottlieb, T. M. & Jackson, S. P. (1993). The DNA-depen-dent protein kinase: requirement for DNA ends andassociation with Ku antigen. Cell, 72, 131–142.

12. DeFazio, L. G., Stansel, R. M., Griffith, J. D. & Chu, G.(2002). Synapsis of DNA ends by DNA-dependentprotein kinase. EMBO J. 21, 3192–3200.

13. Merkle, D., Douglas, P., Moorhead, G. B. G.,Leonenko, Z., Yu, Y., Cramb, D. et al. (2002). TheDNA-dependent protein kinase interacts with DNAto form a protein–DNA complex that is disruptedby phosphorylation. Biochemistry, 41, 12706–12714.

14. McElhinny, S. A. N., Snowden, C. M., McCarville, J.& Ramsden, D. A. (2000). Ku recruits the XRCC4-ligase IV complex to DNA ends. Mol. Cell Biol. 20,2996–3003.

15. Polotnianka, R. M., Li, J. & Lustig, A. J. (1998). Theyeast Ku heterodimer is essential for protection ofthe telomere against nucleolytic and recombinationalactivities. Curr. Biol. 8, 831–834.

16. Fagagna, F. A., Hande, M. P., Tong, W. M., Roth, D.,Lansdorp, P. M., Wang, Z. Q. & Jackson, S. P. (2001).Effects of DNA non-homologous end-joining factorson telomere length and chromosomal stability inmammalian cells. Curr. Biol. 11, 1192–1196.

17. Walker, J. R., Corpina, R. A. & Goldberg, J. (2001).Structure of the Ku heterodimer bound to DNA andits implications for double-strand brake repair.Nature, 412, 607–614.

18. Zhang, Z., Zhu, L., Lin, D., Chen, F., Chen, D. J. &Chen, Y. (2001). The three-dimensional structure ofthe C-terminal DNA-binding domain of Ku70. J. Biol.Chem. 276, 38231–38236.

19. Aravind, L. & Koonin, E. V. (2000). SAP—a putativeDNA-binding motif involved in chromosomal organ-ization. Trends Biochem. Sci. 25, 112–114.

20. Chou, C. H., Wang, J., Knuth, M. W. & Reeves, W. H.(1992). Role of a major epitope in forming the DNAbinding site of the p70 (Ku) antigen. J Expt. Med.175, 1677–1684.

21. Singleton, B. K., Torres-Arzayus, M. I., Rottinghaus,S. T., Taccioli, G. E. & Jeggo, P. A. (1999). The C termi-nus of Ku80 activates the DNA-dependent proteinkinase catalytic subunit. Mol. Cell BioI. 19, 3267–3277.

22. Dynan, W. S. & Yoo, S. (1998). Interaction of Kuprotein and DNA-dependent protein kinase catalyticsubunit with nuclei acids. Nucl. Acids Res. 26,1551–1559.

23. Gell, D. & Jackson, S. P. (1999). Mapping of protein–protein interactions within the DNA-dependentprotein kinase complex. Nucl. Acids Res. 17, 3494–3502.

24. Doherty, A. J., Jackson, S. P. & Weller, G. R. (2001).Identification of bacterial homologues of the KuDNA repair proteins. FEBS Letters, 500, 186–188.

25. Weller, G. R., Kysela, B., Roy, R., Tonkin, L. M.,Scanlan, E., Della, M. et al. (2002). Identification of aDNA non-homologous end-joining complex inbacteria. Science, 297, 1686–1689.

26. Harris, R., Maman, J. D., Hinks, J. A., Sankar, A., Pearl,L. H. & Driscoll, P. C. (2002). Backbone 1H, 13C and 15Nresonance assignments for the C-terminal region ofKu86 (Ku86CTR). J. Biomol. NMR, 22, 373–374.

27. Kay, L. E., Torchia, D. A. & Bax, A. (1989). Backbonedynamics of proteins studied by 15N inverse detectedheteronuclear NMR-spectroscopy—application tostaphylococcal nuclease. Biochemistry, 28, 8972–8979.

28. Kristensen, S. M., Siegal, G., Sankar, A. & Driscoll,P. C. (2000). Backbone dynamics of the C-terminalSH2 domain of the p85a subunit of phosphoinositide3-kinase. Effect of phosphotyrosine peptide bindingand characterization of slow conformationalexchange processes. J. Mol. Biol. 299, 771–788.

29. Mandel, A. M., Akke, M. & Palmer, A. G. (1995).Backbone dynamics of Escherichia coli ribonucleaseH1—correlations with structure and function in anactive enzyme. J. Mol. Biol. 246, 144–163.

3D Solution Structure of Ku86CTR 581

30. Lipari, G. & Szabo, A. (1982). Model-free approach tothe interpretation of nuclear magnetic-resonancerelaxation in macromolecules. 1. Theory and rangeof validity. J. Am. Chem. Soc. 104, 4546–4559.

31. Lipari, G. & Szabo, A. (1982). Model-free approach tothe interpretation of nuclear magnetic-resonancerelaxation in macromolecules. 2. Analysis of experi-mental results. J. Am. Chem. Soc. 104, 4559–4570.

32. Yamazaki, T., Julie, D., Forman-Kay, J. D. & Kay, L. E.(1993). Two-dimensional NMR experiments for cor-relating 13Cb and 1Hd/1 chemical shifts of aromaticresidues in 13C-labeled proteins via scalar couplings.J. Am. Chem. Soc. 115, 11054–11055.

33. Brunger, A. T., Adams, P. D., Clore, G. M., DeLano,W. L., Gros, P., Grosse-Kunstleve, R. W. et al. (1998).Crystallography MR system (CNS): a new softwaresystem for macromolecular structure determination.Acta Crystallog. sect. D, 54, 905–921.

34. Wakefield, R. I., Smith, B. O., Nan, X., Free, A.,Soteriou, A., Uhrin, D. et al. (1999). The solutionstructure of the domain from MeCP2 that binds tomethylated DNA. J. Mol. Biol. 291, 1055–1065.

35. Mazza, C., Ohno, M., Segref, A., Mattaj, I. W. & Cusack,S. (2001). Crystal structure of the human nuclear capbinding protein complex. Mol. Cell, 8, 383–396.

36. Mao, Y., Nickitenko, A., Duan, X., Lloyd, T. E., Wu,M. N., Bellen, H. & Quiocho, F. A. (2000). Crystalstructure of the VHS and FYVE tandem domains ofHrs, a protein involved in membrane trafficking andsignal transduction. Cell, 100, 447–456.

37. Holm, L. & Sander, C. (1993). Protein structure com-parison by alignment of distance matrices. J. Mol.Biol. 233, 123–138.

38. Nguyen, B. D., Abbott, K. L., Potempa, K., Kobor,M. S., Archambault, J., Greenblatt, J. et al. (2003).NMR structure of a complex containing the TFIIFsubunit RAP74 and the polymerase II carboxyl-terminal domain phosphatase FCP1. Proc. Natl Acad.Sci. USA, 100, 5688–5693.

39. Hays, J. B. (2002). Arabidopsis thaliana, a versatilemodel system for study of eukaryotic genome-main-tenance functions. DNA Repair, 1, 579–600.

40. Chan, D. W., Ye, R., Veillette, C. J. & Lees-Miller, S. P.(1999). DNA-dependent protein kinase phosphryl-ation sites in Ku 70/80 heterodimer. Biochemistry, 38,1819–1828.

41. Powers, R., Garrett, D. S., March, C. J., Frieden, E. A.,Gronenborn, A. M. & Clore, G. M. (1993). The high-resolution, three-dimensional solution structure ofhuman Interleukin-4 determined by multidimen-sional heteronuclear magnetic resonance spec-troscopy. Biochemistry, 32, 6744–6762.

42. Ottiger, M., Delaglio, F. & Bax, A. (1998). Measurementof J and dipolar couplings from simplified two-dimen-sional NMR spectra. J Magn. Reson. 131, 373–378.

43. Ruckert, M. & Otting, G. (2000). Alignment of bio-logical macromolecules in novel non-ionic liquidcrystalline media for NMR experiments. J. Am.Chem. Soc. 122, 7793–7797.

44. Delaglio, F., Grzesiek, S., Vuister, G. W., Zhu, G.,Pfiefer, J. & Bax, A. (1995). NMR-Pipe—a multi-

dimensional spectral processing system based onUNIX pipes. J. Biomol. NMR, 6, 277–293.

45. Kraulis, P. J. (1989). ANSIG: A program for theassignment of protein 1H 2D NMR spectra by inter-active graphics. J. Magn. Reson. 24, 627–633.

46. Wishart, D. S., Bigam, C. G., Yao, J., Abildgaard, F.,Dyson, H. J., Oldfield, E. et al. (1995). 1H, 13C and15N chemical shift referencing in biomolecular NMR.J. Biomol. NMR, 6, 135–140.

47. Boucher, W. (1996). AZARA v2.0, Department ofBiochemistry, University of Cambridge, UK.

48. Linge, J. P. & Nilges, M. (1999). Influence of non-bonded parameters on the quality of NMRstructures: a new force field for NMR structurecalculation. J. Biomol. NMR, 13, 51–59.

49. Cornilescu, G., Delaglio, F. & Bax, A. (1999). Proteinbackbone angle restraints from searching a databasefor chemical shift and sequence homology. J. Biomol.NMR, 13, 289–302.

50. Wishart, D. S. & Sykes, B. D. (1994). The 13Cchemical-shift index: a simple method for the identi-fication of protein secondary structure using 13Cchemical-shift data. J. Biomol. NMR, 4, 171–180.

51. Clore, G. M., Gronenborn, A. M. & Bax, A. (1998). Arobust method for determining the magnitude of thefully asymmetric alignment tensor of orientedmacromolecules in the absence of structural infor-mation. J. Magn. Reson. 133, 216–221.

52. Zweckstetter, M. & Bax, A. (2000). Prediction ofsterically induced alignment in a dilute liquid crys-talline phase: aid to protein structure determinationby NMR. J. Am. Chem. Soc. 122, 3791–3792.

53. Dosset, P., Hus, J. C., Marion, D. & Blackledge, M.(2001). A novel interactive tool for rigid-body model-ing of multi-domain macromolecules using residualdipolar couplings. J. Biomol. NMR, 20, 223–231.

54. Laskowski, R. A., MacArthur, M. W., Moss, D. S. &Thornton, J. M. (1993). PROCHECK: a program tocheck the stereochemical quality of protein struc-tures. J. Appl. Crystallog. 26, 283–291.

55. Brooks, B. R., Bruccoleri, R. E., Olafson, B. D., States,D. J., Swaminathanm, S. & Karplus, M. (1983).CHARMM—a program for macromolecular energy,minimization, and dynamics calculations. J Comput.Chem. 4, 187–217.

56. Clore, G. M. & Garrett, D. S. (1999). R-factor, free R,and complete cross-validation for dipolar couplingrefinement of NMR structures. J. Am. Chem. Soc. 121,9008–9012.

57. Kraulis, P. J. (1991). MOLSCRIPT—a program toproduce both detailed and schematic plots of proteinstructures. J. Appl. Crystallog. 24, 946–950.

58. Nicholls, A., Sharp, K. A. & Honing, B. (1991).Protein folding and association: insight from theinterfacial and thermodynamics of hydrocarbon.Proteins: Struct. Funct. Genet. 11, 281–296.

59. Lu, G. G. (2000). TOP: a new method for proteinstructure comparisons and similarity searches.J. Appl. Crystallog. 33, 176–183.

60. Hubbard, S. J., Thornton, J. M. (1993). “NACCESS”,Computer program, Department of Biochemistry andMolecular Biology, University College London, UK.

Edited by J. Thornton

(Received 29 July 2003; received in revised form 25 September 2003; accepted 1 October 2003)

582 3D Solution Structure of Ku86CTR