Structural basis for nick recognition by a minimal pluripotent DNA ligase

Structural basis for nick recognition by a minimalpluripotent DNA ligasePravin A Nair, Jayakrishnan Nandakumar, Paul Smith, Mark Odell, Christopher D Lima & Stewart Shuman

Chlorella virus DNA ligase, the smallest eukaryotic ligase known, has pluripotent biological activity and an intrinsicnick-sensing function, despite having none of the accessory domains found in cellular ligases. A 2.3-A crystal structure of theChlorella virus ligase-AMP intermediate bound to duplex DNA containing a 3¢-OH–5¢-PO4 nick reveals a new mode of DNAenvelopment, in which a short surface loop emanating from the OB domain forms a b-hairpin ‘latch’ that inserts into the DNAmajor groove flanking the nick. A network of interactions with the 3¢-OH and 5¢-PO4 termini in the active site illuminates theDNA adenylylation mechanism and the crucial roles of AMP in nick sensing and catalysis. Addition of a divalent cation triggerednick sealing in crystallo, establishing that the nick complex is a bona fide intermediate in the DNA repair pathway.

This year marks the fortieth anniversary of the discovery of DNAligases1–10. Ligases are ubiquitous enzymes essential for DNA replica-tion, recombination and repair. They seal breaks in the phosphodiesterbackbone by joining 3¢-OH and 5¢-PO4 termini. The ligase mechanismentails three nucleotidyl transfer reactions11. In the first step, ligasereacts with either ATP or NAD+ to form a covalent ligase-adenylateintermediate in which AMP is linked via a phosphoamide bond tolysine. In the second step, the AMP is transferred to the 5¢-PO4 at thenick to form DNA-adenylate (AppDNA). In the third step, ligasecatalyzes attack of the 3¢-OH of the nick on AppDNA to join the twopolynucleotides and release AMP.

All DNA ligases are built around a core catalytic unit composed of anucleotidyltransferase (NTase) domain and an OB-fold domain12–19.The NTase domain includes six conserved motifs (I, Ia, III, IIIa, IVand V) that form the active site. The fold of the DNA ligase NTasedomain and the active site motifs are conserved in RNA ligases andRNA-capping enzymes20–22, which, together with DNA ligases, com-prise a superfamily of polynucleotide 5¢ end–modifying enzymes thatact through a lysyl-N-NMP intermediate23. The specificity of DNAligases for ATP or NAD+ is imparted by clade-specific structuralelements that bind and orient the PPi and nicotinamide mononucleo-tide leaving groups, respectively22–26.

DNA ligase functions include Okazaki-fragment joining, nucleotideand base excision repair, homologous recombination and nonhomol-ogous end joining. In mammals, these tasks are divided among fourdifferent ATP-dependent ligases (I, IIIa, IIIb and IV; reviewed inref. 11). Mammalian ligases are large polypeptides (4900 aminoacid residues), consisting of the core catalytic unit (NTase-OB)embellished by additional isozyme-specific structural domains; theseinclude zinc fingers, BRCT domains and nuclear or mitochondriallocalization signals. It is thought that the domains flanking the ligase

core confer specificity for a particular subset of nucleic acid trans-actions by mediating interactions with DNA or with other replicationand repair proteins that direct the ligases to their sites of action11. Alleukaryotic cellular ligases have a large globular DNA-binding domain(DBD) fused to the proximal end of the NTase component of the ligasecore15. The DBD of human ligase I (LIG1) is required for its nick-sealing activity15. LIG1 has modest affinity for singly nicked duplexDNA and it weakly discriminates nicked DNA from an intact DNAduplex15. Deletion of the DBD abolishes LIG1 binding to either nickedor intact duplex DNA15.

Virus-encoded ATP-dependent ligases from bacteriophage T7 (359amino acid residues) and Chlorella virus PBCV-1 (298 residues) aremuch smaller than cellular ligases because they have no extra domainsappended to the NTase-OB core12,14. Chlorella virus DNA ligaserepresents the minimal functional unit of the ATP-dependent ligaseclade. Chlorella virus ligase has an intrinsic nick-sensing function: itbinds stably and with high affinity to duplex DNA containing a single3¢-OH–5¢-PO4 nick but not to intact duplex DNA (the ligationreaction product) or to gapped DNA27–29. Nick recognition dependson the nick 5¢-PO4 and covalent adenylylation of the ligase27,28. Therequirement of enzyme adenylylation for nick binding is shared byother viral ATP-dependent polynucleotide ligases30,31. Perhaps themost impressive feature of Chlorella virus ligase is that it is able tosustain mitotic growth, excision repair and nonhomologous endjoining in budding yeast when it is the only ligase in the cell32,33.Thus, Chlorella virus ligase represents a stripped-down pluripotenteukaryotic ligase that can recapitulate most of the biological functionsof the two yeast ligases, Cdc9 (a LIG1 homolog) and Lig4.

The salient question is how Chlorella virus ligase is able to senseDNA nicks without the guidance of the flanking domains thatdecorate cellular ligases and contribute to their DNA interfaces.

Received 20 March; accepted 5 June; published online 8 July 2007; doi:10.1038/nsmb1266

Molecular Biology and Structural Biology Programs, Sloan-Kettering Institute, New York, New York 10021, USA. Correspondence should be addressed to S.S.([email protected]).

77 0 VOLUME 14 NUMBER 8 AUGUST 2007 NATURE STRUCTURAL & MOLECULAR BIOLOGY

ART IC L E S©

2007

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

smb

Crystal structures of human LIG1 and Escherichia coli LigA bound tothe nicked DNA-adenylate intermediate underscore a common modeof DNA binding in which the large multidomain eukaryotic andbacterial ligases envelop the DNA duplex within a circumferentialprotein clamp15,19. Although the topologies and domain connectivitiesof the human LIG1 and E. coli LigA clamps are quite different, theyboth rely on ‘kissing’ contacts between a large flanking domain andone of the modules of the ligase core to close the clamp around theduplex at the adenylylated nick15,19. The available structures ofpolynucleotide ligases bound at AppDNA nicks capture the enzymesat an advanced stage of the reaction15,19,20,34. The missing piece in theligase puzzle is a structure of ligase engaged at a nick before AMP istransferred to the broken 5¢ end. Here we report a 2.3-A crystalstructure of the Chlorella virus ligase-AMP intermediate bound toduplex DNA containing a single 3¢-OH–5¢-PO4 nick.

RESULTSCrystallization of Chlorella virus ligase bound at a nickCrystallization trials were performed by incubating recombinantChlorella virus ligase (consisting of a mixture of ligase apoenzymeand ligase-AMP) with nicked DNAs in the absence of a divalent cationand in the presence of EDTA. The rationale was that deprivation of thedivalent cation would permit nick binding by ligase-AMP but precludecatalysis of either the DNA adenylylation or strand-closure step of theligation pathway. The DNA ligand that resulted in successful crystal-lization was a 21-base-pair (bp) duplex with a central 3¢-OH–5¢-PO4

nick in one strand. The crystals formed in space group P212121 andcontained one ligase protomer and one DNA in the asymmetric unit.The structure of the ligase–DNA complex was determined at2.3-A resolution. The refined model (Fig. 1a) included the entireDNA duplex and a continuous ligase polypeptide (residues 1–293). Noelectron density was observed for the C-terminal 5 residues. Thispeptide was also disordered in the crystal structure of the ligase-AMPintermediate14. The 5¢ terminus of the nick in the ligase–DNA cocrystalwas a monophosphate, and AMP was attached covalently to Lys27 inthe ligase active site (Supplementary Fig. 1a online), signifying that(i) the ligase-AMP intermediate, but not the apoenzyme, hadrecognized the nick; and (ii) step 2 chemistry had not occurred during

crystal growth or in the crystal. Thus, this structure captures apolynucleotide ligase bound at a 3¢-OH–5¢-PO4 nick.

Ligase engages the nick as a novel protein-DNA clampChlorella virus ligase encircles the DNA as a ‘C’-shaped protein clamp(Fig. 1a). The clamp entails extensive DNA contacts by the NTasedomain (colored cyan) and the OB domain (beige) and by a novel‘latch’ module (magenta) that emanates from the OB domain. TheNTase domain binds the broken and intact DNA strands in the majorgroove flanking the nick and also in the minor groove on the 3¢-OHside of the nick (Fig. 1a). The OB domain binds across the minorgroove on the face of the duplex behind the nick. The latch module,consisting of a b-hairpin loop, occupies the major groove and com-pletes the circumferential clamp via kissing contacts between the tip ofthe loop and the surface of the NTase domain (Fig. 1b, red arrow).

The topology and connectivity of the Chlorella virus ligase clamparound DNA differs radically from the clamp formed by human LIG1bound to an adenylylated nick15 (Fig. 1b). Whereas Chlorella virusligase and LIG1 share the NTase and OB domains, which adopt similarconformations and positions when docked on their respective DNAs,LIG1 has a large N-terminal DBD proximal to the NTase that occupiesa place in the clamp toroid roughly analogous to that of the muchsmaller latch domain of Chlorella virus ligase (Fig. 1b). The DBD ofLIG1 has an all–a-helical secondary structure that has nothing incommon with the latch. The peptide linker covalently connecting theDBD to the NTase domain of LIG1 (Fig. 1b, black arrow) is situated inroughly the same angular position on the DNA circumference as thenoncovalent kissing contacts between the Chlorella virus ligase latchand NTase domains. LIG1 closes its clamp via contacts between theN-terminal DBD and the C-terminal OB module (Fig. 1b, red arrow).LIG1 has no latch; the kissing contacts from the OB domain to theDBD come from the short loop connecting the first and secondb-strands of the OB domain15, which has been called motif Va35 andis conserved in cellular and poxvirus DNA ligases that have anN-terminal DBD. The interdomain kissing interactions in LIG1occur at about the same angular position on the DNA circumferenceas the covalent junctions between the OB domain and the strands ofthe latch in Chlorella virus ligase (Fig. 1b). Thus, the Chlorella virusand human ligases have distinct modes of DNA envelopment.

The Chlorella virus ligase clamp is also radically different intopology and connectivity from the clamp formed by DNA-boundE. coli LigA, the prototypal NAD+-dependent ligase19. LigA has nocounterpart of the latch; rather, it closes its clamp via kissing contactsbetween its NTase domain and a large helix-hairpin-helix (HhH)domain located near the C terminus of the LigA polypeptide19. TheC-terminal HhH domain is a signature feature of the NAD+-dependentligase clade that occupies roughly the same angular position in theclamp as the N-terminal DBD of LIG1. On the basis of the available

NTase

NTase

OB OB

Latch

Latch

DBD

Human LIG1Chlorellavirus ligase

3′-OHside 5′-PO4

side

Latch

a

b

Figure 1 Comparison of the Chlorella virus ligase and human LIG1 DNA

clamps. (a) Space-filling models of the Chlorella virus ligase enveloping

a singly nicked duplex DNA (depicted as a cartoon trace). Cyan, NTase

domain; beige, OB domain; magenta, latch module. (b) Ribbon diagrams

of the DNA-bound structures of Chlorella virus ligase (left) and LIG1 (right,

PDB 1X9N). The proteins were superimposed with respect to their NTase

and OB domains. Red arrows indicate interdomain contacts that close the

ligase clamps. In Chlorella virus ligase, the latch and NTase domains make

kissing contacts. In contrast, LIG1 closes its clamp via contacts between the

N-terminal DBD (in magenta) and the C-terminal OB domain. Black arrow

indicates the peptide linker between the DBD and NTase domains of LIG1.

The images were prepared with PyMOL51.

ART IC L E S

NATURE STRUCTURAL & MOLECULAR BIOLOGY VOLUME 14 NUMBER 8 AUGUST 2007 7 7 1

©20

07 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nsm

b

structural data, we surmise that ligases have evolved at least threedifferent means of encircling DNA.

Large domain rearrangements accompany nick recognitionPrior inferences about the domain motions that drive ligation weregleaned by comparing crystal structures of ATP-dependent DNA andRNA ligases and GTP-dependent messenger RNA capping enzymescaptured at different stages along the reaction pathway11,14,15,20,22. Thepresent Chlorella virus ligase structure now makes it possible tocompare the same ATP-dependent DNA ligase before and afterbinding to nicked DNA. In the free ligase-AMP intermediate14, theOB domain is deflected away from the NTase domain to fully exposethe surface above the AMP-binding pocket, to which a sulfate anion isbound; the sulfate has been proposed to occupy the same site as the5¢-PO4 of the DNA nick (Fig. 2a). The polypeptide segment fromposition 205 to 224 (which is destined to become the latch) isdisordered in the free ligase14. Structure probing of the free ligaseby limited proteolysis has shown that sites of accessibility to chymo-trypsin and trypsin are clustered within this region29, suggestingthat the disorder of this loop in the crystal structure of free ligasefaithfully reflects its status in solution. Because this segment isprotected from proteolysis when Chlorella virus ligase binds nickedDNA29, it has been proposed that the disordered surface loop makesdirect contact with the DNA substrate and thereby becomes ordered.However, it was initially unclear how this loop, which is distant fromand pointing away from the DNA-binding surface of the NTasedomain where step 2 chemistry occurs, might directly engage theDNA substrate.

Alignment of the NTase domains of free ligase-AMP and the ligase-AMP–nicked DNA complex reveals that DNA binding entails a nearly1801 rotation of the OB domain around a swivel at residue Gln189(Fig. 2a), so that the concave surface of the OB b-barrel fits into theDNA minor groove. This transition results in a 63-A movement of theOB domain (measured at the Ca atoms of Ser205) and places the nowstructured latch module in direct contact with the DNA backbone inthe major groove (Fig. 2b).

Structure and function of the latchThe latch module (residues 202–231) is a b-hairpin loop thatpenetrates the major groove and makes numerous contacts to thephosphodiester backbone. The latch contacts each phosphate of thetemplate DNA strand within a 5-nucleotide segment on the 5¢-PO4

side of the nick (Fig. 2b; see also Supplementary Fig. 2 online).Hydrogen bonds to the phosphate oxygens are donated by the mainchain amides of Phe204, Asn206 and Ser218 and by the side chains ofArg220, Tyr217 and Ser218. The latch also engages a 2-nucleotidesegment of the broken 3¢-OH strand through hydrogen bonds to thevicinal phosphates from Ser221 Og, Thr222 Og and the Lys224 mainchain amide, and through van der Waals contacts of His233 to thebridging nucleoside sugar (Fig. 2b). The Lys210-Thr211, Tyr217-Ser218, Lys219-Arg220, Arg220-Ser221, His223-Lys-224 and Lys224-Ser225 dipeptides that are accessible to trypsin and chymotrypsin inthe free ligase and shielded in the DNA-bound enzyme29 are the veryelements that comprise the DNA-binding site of the latch.

The clamp-closing interface between the tip of the latch loop andthe NTase domain is also evident in Figure 2b. It entails a network ofvan der Waals contacts from Phe215 and Tyr217 of the latch andPhe44 and Lys5 of the NTase domain, plus a hydrogen bond fromAsn214 to the Thr43 main chain carbonyl and a van der Waals contactof Phe215 to the Arg42 main chain carbonyl. The 41-SRTFKP-46surface loop of the NTase domain (motif Ia), with which the latchloop is engaged, is itself a key component of the DNA-binding site:Ser41, Thr43 and Lys45 donate hydrogen bonds to two vicinalphosphates of the broken 3¢-OH strand flanking the nick (Fig. 2b).

Even before structural delineation of the latch, an alanine scan ofseveral of the basic amino acid residues in the disordered segment ofthe free Chlorella virus ligase (Lys205, Lys210, Lys212, Lys219, Arg220,Lys224 and Lys227) showed that all of the single alanine mutantsretained nick-sealing activity, in the range of 45% to 80% of the wild-type level33. The failure to discern a single critical side chain of thelatch now makes sense in light of the multiple latch-DNA contacts andthe fact that many of them involve the protein main chain. To examinethe effects of deleting the latch, we replaced the 27-residue segment

a b

Ligase-adenylate

+Nicked DNA

OB

OB

Latch

NTase

SO4

225

205

Figure 2 Protein domain movements and nucleation of the latch module accompany nick recognition and clamp closure. (a) The structures of the free ligase-

AMP intermediate (left) and ligase-AMP bound at a nick (right, with DNA omitted) were superimposed with respect to their NTase domains (cyan) and then

offset laterally to reveal the large movement of the OB domain (beige) triggered by nick recognition. The movement entails a rigid-body rotation about the

hinge segment connecting the NTase and OB domains (indicated by arrow in free ligase). The surface peptide loop destined to become the latch in the DNA-

bound ligase is disordered in the free ligase. A sulfate ion is bound on the surface of free ligase-AMP in a position that mimics the 5¢-PO4 of the nick.

(b) Stereo view of Chlorella virus ligase (shown as ribbon trace with stick models of selected amino acid residues) bound to nicked DNA (rendered as

transparent surface over a stick model). The image highlights the penetration of the latch (magenta) into the major groove opposite the nick and the numerous

main chain and side chain contacts between the latch and the phosphate backbone of both strands lining the major groove. The kissing contacts of the latch

and the NTase domain (cyan), as well as the DNA contacts of the motif Ia loop (41-SRTFKP-46) of the NTase domain, are shown and discussed in the text.

ART IC L E S

77 2 VOLUME 14 NUMBER 8 AUGUST 2007 NATURE STRUCTURAL & MOLECULAR BIOLOGY

©20

07 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nsm

b

from 205 to 231 with a pentapeptide Gly-Ser-Gly-Ser-Gly linker. Therecombinant DLatch protein had the expected smaller size in an SDS-PAGE analysis, and it retained adenylyltransferase activity, as assayedby reaction with [a-32P]ATP to form a covalent ligase-[32P]AMPadduct (Fig. 3a). Thus, the latch deletion did not globally affectfolding of the NTase domain or of the OB domain, as the C terminusof the OB domain is required for the ligase autoadenylylationreaction24. However, loss of the latch decreased specific activity innick sealing by ten-fold, compared with wild-type ligase (Fig. 3b).Also, the DLatch mutant was much more sensitive than wild-typeligase to inhibition of nick sealing by salt (Fig. 3c). These data implyan important role for the ionic and polar contacts of the latch to theDNA seen in the crystal structure. To measure directly the effects oflatch deletion on nick sensing by Chlorella virus ligase, we exploited aDNA-binding assay in which ligase is incubated with radiolabelednicked DNA in the absence of ATP or a divalent cation and theproducts are analyzed by nondenaturing gel electrophoresis27. Becauseonly the adenylylated form of ligase binds nicked DNA in this assay28,we first determined the fraction of ligase-AMP in the recombinantprotein preparations, from the extent of ligation as a function of inputprotein in the presence of magnesium and the absence of exogenousATP. We found that the fraction of adenylylated ligase was similar forwild-type ligase (11%) and the DLatch variant (15%). The instructivefinding was that deletion of the latch virtually abolished the formationof a stable complex between the enzyme and the nicked DNA ligand

(Fig. 3d). Thus, the latch, by virtue of completing the protein clampabout the duplex, makes a major contribution to nick sensing.

DNA interface of the OB and NTase domainsThe suite of atomic contacts of Chlorella virus ligase with the nickedDNA spans a 14-bp segment, covering 6 bp on the 3¢-OH side of thenick and 8 bp on the 5¢-PO4 side (Fig. 1a and Supplementary Fig. 2).The asymmetric footprint in the crystal agrees with solution foot-printing studies of the ligase–DNA complex with exonuclease III,which revealed a larger region of protection on the 5¢-PO4 side of thenick than on the 3¢-OH side29. Ligase binding induces a 121 bend inthe DNA centered at the nick. The DNA helix has typical B-formsecondary structure throughout, except at the two base pairs on eitherside of the nick, which adopt an A-like conformation.

The OB domain is a five-stand antiparallel b-barrel that occupiesthe minor groove flanking the nick and interacts extensively with a5-nucleotide segment of the 5¢-PO4 strand (Fig. 4a and Supplemen-tary Fig. 2). The Lys274 side chain and Ala253 main chain amidecoordinate the third and fifth phosphates of the 5¢-PO4 strand,

WT

97

Size(kDa)

66

45

31

22

97

0.4

0.3

0.2

0.1

00 2 4

Ligase (ng)

Liga

tion

(pm

ol)

0.4

0.3

0.2

0.1

00 25 50 75 100

NaCI (mM)125 150

Liga

tion

(pm

ol)

6 8 10

55

36

29

21

14

∆ WT

WT∆

WT

– –

∆

WT

∆

∆

a b

c d

a b

5′-P

3′-

5′-P

3′-O

Figure 4 DNA interface of the OB and NTase domains. (a) Stereo view of the OB domain (beige ribbon) bound to the nicked DNA, which is rendered as a

transparent surface over a stick model. Shown are the insertion of the OB domain into the minor groove and the hydrophilic contacts to the DNA backbone

phosphates lining the minor groove. (b) Stereo view of the NTase domain (cyan ribbon) bound to the nick. This image highlights an a-helix of the NTase

domain (75-FQDTTSAVMTG-85) that inserts into the minor groove on the 3¢-OH side of the nick (at top right) and the contact to the phosphodiesterbackbone across the major groove on the 5¢-PO4 side of the nick (at bottom right).

Figure 3 Deleting the latch module suppresses nick recognition and sealing.

(a) Ligase adenylylation. Left, aliquots (2 mg) of wild-type ligase and the

DLatch mutant (D) were analyzed by SDS-PAGE and stained with Coomassie

blue. Right, wild-type ligase and the DLatch mutant both reacted with

[a-32P]ATP to form the covalent ligase-[32P]AMP intermediate. Positions of

size-marker proteins are indicated next to the gels. (b) The extents of nick

sealing by wild-type ligase and by the DLatch mutant are plotted as a

function of input enzyme. (c) Effect of salt. The extents of nick sealing

by 1.25 ng wild-type ligase and by 5 ng DLatch mutant are plotted as a

function of added NaCl concentration. (d) Nick sensing. Binding reaction

mixtures contained 32P-labeled nicked 36-bp DNA (depicted at bottom with

the radiolabeled nick 5¢-PO4 indicated by �) and increasing amounts of

wild-type ligase or DLatch mutant, as specified in Methods. Free DNA and

ligase–DNA complexes of retarded mobility were resolved by electrophoresis

and visualized by autoradiography. Minus signs indicate control reactionsfrom which ligase was omitted.

ART IC L E S

NATURE STRUCTURAL & MOLECULAR BIOLOGY VOLUME 14 NUMBER 8 AUGUST 2007 7 7 3

©20

07 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nsm

b

respectively. Phe286 packs against the 5¢-teminal deoxynucleosidesugar and forces the terminal base pair into an RNA-like A-helicalconformation. Val288 makes van der Waals contacts with the secondand third deoxynucleoside sugars of the 5¢-PO4 strand. Thr249 makesa bifurcated hydrogen bond to the deoxyribose O4¢ atom of the thirdnucleoside and to the adenine N3 atom of the second nucleoside ofthe 5¢-PO4 strand (Fig. 4a). The latter is the sole hydrogen bondbetween the ligase and a nucleoside base. The OB domain engagesfour phosphate groups of the template DNA strand on the 3¢-OH sideof the nick (Fig. 4a and Supplementary Fig. 2). The contacts includehydrogen bonds to the phosphate oxygens from Lys281 Nz, Cys283 Sg,Ser235 Og and the main chain amides of Asp282 and Ser246.

The NTase domain binds along the face of the DNA helix overlyingthe nick. It engages the major groove on the 5¢-PO4 side of the nick.The outer limit of the ligase footprint on the 5¢-PO4 side of the nickis demarcated by contacts from Thr178 Og and Lys173 Nz to twovicinal phosphates along the major groove of the template DNAstrand (Fig. 4b and Supplementary Fig. 2). The outermost phos-phate also receives a water-mediated contact from the main chainamides of Leu179 and Lys180. The nick’s terminal 5¢-PO4 and thepenultimate phosphate are coordinated by Arg176 and Lys188,respectively, and Lys186 makes a bifurcated contact with both the5¢-terminal and penultimate phosphates (Fig. 4b). The contacts ofthe NTase and latch domains in and across the DNA major grooveover a 7-bp segment on the 5¢-PO4 side of the nick might helpenforce the requirement for B-form helical conformation on the5¢ side, as inferred from the fact that a nicked duplex with a 5¢-PO4

RNA strand is not sealed by Chlorella virus ligase36, or by mostother DNA ligases15.

An a-helix of the NTase domain (75-FQDTTSAVMTG-85) insertsinto the minor groove on the 3¢-OH side of the nick, making contactsto the phosphates and nucleoside sugars (Fig. 4b). At the proximalend of the a-helix, Phe75 makes van der Waals contacts to the3¢-terminal sugar that help impart an A-like conformation to theterminal base pair. Gln76, Thr79, Met83 and Thr84 also make van derWaals interactions with the sugar-phosphate backbone of the strandsacross the minor groove (Fig. 4b). The a-helix donates hydrogenbonds from Ser80 Og and the Gly85 amide to two adjacent phos-phates of the template DNA strand (Fig. 2b and Fig. 4b). The Arg48side chain, which flanks motif Ia and is situated near the distal end ofthe a-helix, projects into the minor groove and makes a van der Waalscontact to a template-strand sugar (Fig. 4b). We cited in the precedingsection the contacts of the motif Ia loop 41-SRTFKP-46 with thesecond and third phosphates of the 3¢-OH strand flanking the nick.Additionally, the phosphate of the terminal nucleotide on the3¢-OH side of the nick is engaged by Arg42 in the major groove, bythe main chain amide of Arg32 and through a water-mediatedhydrogen bond to the Arg32 main chain carbonyl (Fig. 4b).

The NTase and OB domains form an interface through contactsemanating from the loop connecting the fourth and fifth b-strands ofthe OB domain. These include a salt bridge from Arg285 to Asp29

(Fig. 5), a hydrogen bond from between the Gly279 main chaincarbonyl and Gln76 Nd, and van der Waals interactions betweenGlu161 (motif III) and both Phe276 and Met278. An equivalent of theChlorella virus ligase Arg285-Asp29 ion pair that stabilizes the NTase-OB domain arrangement enveloping the DNA is also present in thestructure of human LIG1, between Arg871 and Asp570 (ref. 15).

Insights into nick recognition and nucleotidyl transferThe active site of Chlorella virus ligase in the DNA complex and thecontacts to the 3¢-OH and 5¢-PO4 nick termini are depicted inFigure 5b. A comparison with the active site in the free ligase-AMPintermediate (Fig. 5a) shows the same ligase before and after nickedDNA is recognized. The adenine base of the covalently attached AMPis stacked on Phe98, an essential residue in motif IIIa14, and theadenosine nucleoside is in the anti conformation in the DNA complex,unchanged from the conformation of the free ligase-adenylate. Glu67,an essential constituent of motif III37, contacts the ribose O4¢ of theAMP adduct via Oe1 (Fig. 5b) and accepts hydrogen bonds to Oe2from the main chain amides of Gly30 and Ile31 in motif I(27-KIDGIR-32). Glu67 also interacts via waters with the terminalphosphate of the 3¢-OH strand and with Arg32, an essential motif Iside chain28 that makes direct and water-mediated hydrogen bonds tothe ribose O2¢ and O3¢ atoms of the covalent adenylate (Fig. 5b). Itappears that subtle remodeling of the adenosine nucleoside interac-tions upon nick recognition results in a net gain of atomic contactscompared with the free enzyme.

A sulfate anion in the active site of the free enzyme (Fig. 5a) hasbeen posited to mimic the docking of the 5¢-PO4 of nicked DNA14,and, indeed, we found that the terminal phosphate at the nick isbound in a similar position (Fig. 5b). The nick 5¢-PO4 and thelysyl-AMP adduct are the two most crucial determinants of nick

a

b

c

Figure 5 Architecture of the ligase active site for DNA adenylylation.

(a) Stereo view of the active site of the free ligase-AMP intermediate, with a

sulfate anion bound in a position analogous to that of the nick 5¢-PO4 in b.

Waters are depicted as red spheres; the water proposed to mimic a divalent

cation cofactor for step 2 catalysis is colored pink. (b) Stereo view of theatomic interactions at the 3¢-OH–5¢-PO4 nick. (c) Stereo view of the active

site after exposure of a ligase-AMP–nicked DNA crystal to 5 mM Mn2+.

Provision of the metal ion triggered catalysis of phosphodiester formation

at the former nick.

ART IC L E S

77 4 VOLUME 14 NUMBER 8 AUGUST 2007 NATURE STRUCTURAL & MOLECULAR BIOLOGY

©20

07 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nsm

b

recognition28. In the ligase–DNA cocrystal, the nick 5¢-PO4 is coordi-nated by Arg176 and Lys186, and the lysyladenylate is itself acomponent of the nick 5¢-PO4 coordination complex, via a water-mediated hydrogen bond from the ribose O3¢ to a phosphate oxygen(Fig. 5b). These interactions are functionally relevant, insofar asalanine substitutions for Arg176 and Lys186 reduce nick-sealingactivity to 12% and 7% of the wild-type ligase’s specific activity,respectively14. Moreover, the adenylylated R176A mutant is defectivein nick sensing14.

The structure of the ligase-AMP–nick complex determined in theabsence of a divalent cation is construed to approximate that of theMichaelis complex for the DNA adenylylation step (step 2) of theligation pathway. Step 2 chemistry entails the attack of a nick 5¢-PO4

oxygen on the AMP phosphorus to expel Lys27 and form the activatedAppDNA phosphoanhydride intermediate. This type of reactionoccurs with inversion of configuration at the AMP phosphorus20

and is presumed to proceed through a pentacoordinate intermediate.In the nick complex of Chlorella virus ligase, one of the nonbridgingAMP phosphate oxygens is coordinated jointly by essential motif Vresidues Lys186 and Lys188 (Fig. 5b), implying that these lysinesstabilize the step 2 transition state by neutralizing the extra negativecharge developed on the AMP phosphate. The salient mechanisticinsights stem from the contacts of the other nonbridging AMPphosphorus oxygen, which receives a hydrogen bond from theterminal 3¢-OH of the nick (located 2.2 A away) and is alsocoordinated by a water molecule (colored pink in Fig. 5b) that wepropose mimics the position of a catalytic metal ion. This water is atthe center of a pseudo octahedral coordination complex (as would beexpected for a divalent cation) with five of the apical positionsoccupied by Glu161 Oe1 and Oe2, Asp29 Od1, an adenylate phos-phate oxygen and the nick 3¢-OH (Fig. 5b). A divalent cationoccupying the water position would be poised to stabilize the transi-tion state of the AMP phosphate during step 2 catalysis. This model isquite consistent with functional studies of the putative metal-bindingresidues. The Glu161 side chain (motif IV) is essential for nick sealingby Chlorella virus ligase37. Glu161 coordinates a water that bridges theadenylate phosphate in the free ligase structure (Fig. 5a, cyan), andthis water ligand is replaced by lutetium in an isomorphous heavy-metal derivative used to solve the structure of free ligase-AMP14. Theequivalent motif IV glutamate coordinates a divalent cation in theactive sites of ATP-dependent RNA ligases and mycobacterial ATP-dependent DNA ligase D16,20,21. A potential metal-binding site invol-ving the motif IV glutamate has also been inferred for human LIG1(ref. 15). Chlorella virus ligase Asp29 (in motif I) is essential for overallnick sealing; this residue has a catalytic role during steps 2 and 3 of theligation pathway but is dispensable for ligase adenylylation and nicksensing29,37. Finally, the nick 3¢-OH, though not required for nicksensing by ligase-adenylate, is crucial for catalysis of the DNAadenylylation reaction29, a step in which the nick 3¢-OH is not itselfchemically transformed. We can now refine the proposed ligasemechanism in which catalysis of step 2 is assisted by a nonreactivemoiety of the DNA substrate, by suggesting that the nick 3¢-OHparticipates in binding the catalytic metal for step 2. The same metalligand would be poised to act as a catalyst of step 3 by activating the3¢-OH nucleophile for attack on DNA-adenylate.

Catalysis of strand joining in crystalloA caveat to the assumption that the nick complex of Chlorella virusligase-AMP reflects the state of the enzyme before step 2 chemistry isthat the distance from the AMP phosphorus to the nearest nick 5¢-PO4

oxygen is 5.3 A, which is not optimal for impending nucleophilic

attack. It is likely that electrostatic repulsion of the two phosphatescontributes to this gap, which might be shortened in the presence of abridging metal ion during the ligase reaction in solution. Also, theArg42 side chain, which is required for ligation and for nick sensing14,is near but not within direct contact distance of the nick 5¢-PO4 in thepresent crystal structure. Thus, we envision that the 5¢-PO4 mustundergo a small movement toward the AMP phosphate for the step 2reaction to proceed.

We obtained evidence that the nick-bound structure is a bona fideintermediate along the ligation pathway by exposing the preformedcrystals to 5 mM Mn2+ immediately before cryoprotection. Thistreatment reduced the symmetry of the crystal lattice and affectedthe unit cell constants, transforming the space group from P212121

with one protein–DNA complex in the asymmetric unit to spacegroup P1 with four complexes in the asymmetric unit. The structureof the Mn2+-treated complex was determined at 3.0-A resolution. Thesimulated annealing omit maps showed a continuous electron densityover the sealed phosphodiester in all four complexes of the unit cell,with clear separation of the density corresponding to the Lys27 step 2leaving group (Supplementary Fig. 1b), demonstrating that catalysisof steps 2 and 3 had occurred in situ in the crystal upon provision of adivalent cation. Electron densities for AMP and manganese were notclearly discernible in the active site of the NTase domain, so we did notinclude AMP and divalent cation in the refined model of the productcomplex. We surmise that the complex might have ‘breathed’ duringcatalysis in situ, concomitant with the change in space group, so thatAMP and manganese could have escaped from the pocket before theligase settled back onto the sealed DNA. Alternatively, the AMP mightbe present but mobile with the binding pocket. Superposition of theligase protomers of the nick complex and each of the four productcomplexes revealed r.m.s. deviations of 0.5, 0.6, 0.4 and 0.4 A,respectively, for all Ca positions.

The state of the active site in the product complex is depicted inFigure 5c. Comparison to the nick complex (Fig. 5b) highlights a netloss of atomic contacts from ligase to DNA after completion ofphosphodiester synthesis: for example, Lys186 and Arg176 haverelinquished their interactions with the former nick 5¢-PO4. Inessence, the nick-sensing interactions disappear once the nick isrepaired, thereby facilitating dissociation of the ligase from the sealedDNA product in solution.

DISCUSSIONThe Chlorella virus ligase–DNA structure affords new insights to themechanism of nick recognition and catalysis of DNA adenylylation byDNA ligases. We found that the covalent lysyl-AMP adduct is a keyparticipant in nick sensing, via contacts (direct and indirect) betweenthe AMP sugar and phosphate moieties and both the 3¢-OH and5¢-PO4 termini. This mechanism ensures that ligase-adenylate, whichis poised to catalyze phosphodiester formation, is not competing withligase apoenzyme for binding to sites in need of repair.

The inference that the nick 3¢-OH is a component of the metal-coordination complex of the nick-bound ligase provides an explana-tion for the earlier proposal of substrate-assisted catalysis by the3¢-OH during the nick 5¢ adenylylation reaction29,38. This feature ofenzymatic ligation can be viewed as a checkpoint that helps ensurethat 5¢ end activation is confined to sites where a reactive 3¢-OH is inplace and the ligation reaction can continue through the step ofphosphodiester synthesis. Otherwise, ligases would run the risk ofadenylylating 5¢-PO4 termini inappropriately at gaps in duplex struc-tures or at single-stranded ends. Such AppDNA ends would bedifficult to seal directly, given that (i) ligases are predominantly in

ART IC L E S

NATURE STRUCTURAL & MOLECULAR BIOLOGY VOLUME 14 NUMBER 8 AUGUST 2007 7 7 5

©20

07 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nsm

b

the adenylylated state at physiological ATP concentrations and(ii) ligase-adenylate can neither seal nor deadenylate an AppDNAterminus. The checkpoint is not perfect, insofar as AppDNA doesaccumulate when DNA ligases encounter 3¢-OH and 5¢-PO4 ends at a1-nucleotide gap or at a 3¢-OH–5¢-PO4 nick in which there is a basemismatch at the 3¢-OH terminus36,39. Although the AppDNA endwould be resistant to processing by most known 5¢ exonucleases,recent studies show that eukaryotes resolve this roadblock to DNArepair via the action of aprataxin, an enzyme that specifically hydro-lyzes the AMP from AppDNA to restore a ligatable 5¢-PO4 end40.

Chlorella virus ligase comprises the minimal DNA ligase, onedevoid of the large structural domains that decorate all cellularanalogs. We have proposed a model of ligase evolution based on thefollowing two hypotheses. (i) A small pluripotent ligase exemplified byChlorella virus ligase is the progenitor of the large ATP-dependentligases found currently in eukaryotic cells. (ii) The fusion of newdomains to the core ligase structure (with concomitant acquisition ofnew protein-protein interactions), along with variations within thecore domains, served to enhance some of the functional repertoire ofthe ancestral ligase while disabling other components of that reper-toire33. The structure of Chlorella virus ligase at a DNA nick providesstrong support for our evolutionary model by revealing the latch asthe minimized structural solution to circumferential envelopment ofduplex DNA flanking a nick. Indeed, it is noteworthy that the latchmodule is missing from the OB domain of human LIG1 (whichdisables the intrinsic nick-sensing function of its NTase-OB core). InLIG1 (and probably in the other mammalian ligases), the DNA-clamping function is restored by fusion to the large globular DBD.

Not all DNA ligases have a well-developed nick sensor. Ligaseswithout such a sensor include mycobacterial DNA ligase C (LigC) andDNA ligase D (LigD), which are ATP-dependent ligases specializing innonhomologous end joining in mycobacteria41. LigC and LigD haverelatively weak intrinsic activity in sealing DNA nicks, and theyaccumulate AppDNA in the presence of ATP because the ligasesreadily dissociate from the nick after catalysis of step 2 (ref. 42).The crystal structure of the ligase domain of Mycobacteriumtuberculosis LigD, which is composed of the NTase-OB core, isremarkable for the absence of a latch in its OB domain16. Instead,the surface loop is truncated, so that LigD would not form acircumferential clamp around DNA. We suspect that the same istrue of LigC. The loss of nick sensing by bacterial LigC and LigDis apparently compensated by interactions with bacterial Ku, which isessential for mycobacterial nonhomologous end joining in vivo andwhich assists in recruiting LigD to DNA41,43.

Although the latch seen in Chlorella virus ligase is one of thesimplest solutions to DNA encirclement, it might not be the onlysimple solution that has evolved. Thus, it will be of interest to explorewhether, and how, other small DNA ligases sense nicks and todetermine the structures of such small ligases (for example, T7) inthe nick-bound state. Absent such a DNA-bound structure, a super-position of the individual domains of the free T7 ligase12 on thestructure of DNA-bound Chlorella virus ligase shows that the inter-strand loop of the OB domain from which the Chlorella virus ligaselatch emanates is truncated in the T7 enzyme (see SupplementaryFig. 3 online). Although the truncated T7 loop might well interactwith DNA, it seems unlikely that T7 ligase could use this short loop toclose a fully circumferential clamp around the DNA, as seen in theChlorella virus ligase–DNA cocrystal. It is intriguing that T7 ligase hastwo inserts in its primary structure, which are absent in Chlorella virusligase14 and which comprise surface loops with disordered gaps inthe T7 ligase tertiary structure12 (Supplementary Fig. 3). The

superposition onto DNA-bound Chlorella virus ligase suggests thatthe T7-specific insert in the NTase domain (T7 residues Trp112–Pro136) might be poised to interact with the DNA major groove onthe 3¢-OH side of the nick. This loop is the principal site of sensitivityto limited proteolysis in the native T7 ligase44. By analogy to theprotease-sensitive latch in Chlorella virus ligase, we speculate that theunique T7 NTase domain loop becomes ordered upon DNA bindingand contributes directly (and perhaps in a structurally novel manner)to the ligase-DNA interface.

METHODSLarge-scale ligase purification for crystallography. Cultures (4 l) of E. coli

BL21 (DE3) with a plasmid expressing His-tagged Chlorella virus ligase (pET-

His-ChVLig) were grown at 37 1C in LB medium containing 0.1 mg ml–1

ampicillin until A600 reached B0.6. The cultures were adjusted to 0.4 mM

IPTG and incubation was continued at for 18 h at 17 1C. Cells were harvested

by centrifugation and resuspended in a solution containing 50 mM Tris-HCl

(pH 7.5), 1.2 M NaCl, 15 mM imidazole, 10% (v/v) glycerol, 1 mg ml–1

lysozyme, 0.2 mM PMSF, 1 mM benzamidine and 0.2% (v/v) Triton X-100.

The lysate was sonicated to reduce viscosity and insoluble material was

removed by centrifugation. His10-tagged Chlorella virus ligase was isolated

from the soluble extract by adsorption to nickel–nitrilotriacetic acid agarose

and elution with 200 mM imidazole. The protein was purified further by gel

filtration through a Superdex-75 column equilibrated in 20 mM Tris-HCl

(pH 8.0), 350 mM NaCl and 1 mM b-mercaptoethanol. The peak fractions

from the gel-filtration column were dialyzed against buffer containing 10 mM

Tris-HCl (pH 8.0), 100 mM NaCl and 1 mM DTT. Ligase was concentrated to

11 mg ml–1 by centrifugal ultrafiltration and stored at –80 1C.

Nicked DNA ligand for crystallography. Oligonucleotides purchased from

Oligos Etc. were purified by Protein-PAK anion exchange HPLC (Waters) and

then recovered by ethanol precipitation. To anneal the component strands of

the nicked duplex, an equimolar mixture of the 11-nucleotide 5¢-PO4 DNA

strand (5¢-pCACTATCGGAA-3¢), the 10-nucleotide 3¢-OH strand (5¢-ATTGC

GAC-mC-C-3¢, mC denoting 2¢-O-methylcytidine) and the 21-nucleotide

template strand (5¢-TTCCGATAGTGGGGTCGCAAT-3¢) was heated to 85 1C

in buffer containing 10 mM Tris-HCl (pH 8.0), 50 mM NaCl and 1 mM EDTA,

then cooled to 22 1C over 8 h. The DNA was stored at –20 1C and thawed

before use in crystallization trials.

Crystallization and structure determination. A mixture of Chlorella virus

ligase (225 mM), nicked duplex DNA (222 mM) and 2 mM EDTA was added to

an equal volume of a well solution containing 100 mM Bis-Tris (pH 6.5),

30 mM ammonium acetate and 22% PEG 4,000. Crystals were grown at 22 1C

by sitting drop vapor diffusion. Crystals appeared after 3 d. The crystals were

transferred to buffer containing 100 mM Bis-Tris (pH 6.5), 30 mM ammonium

acetate, 22% (w/v) PEG 4,000 and 15% (v/v) glycerol before flash-freezing in

liquid nitrogen. X-ray diffraction data to 2.3-A resolution were collected at a

wavelength of 0.9792 A at 100 K at the Advanced Photon Source (NE-CAT,

beamline 24ID). The data were processed using DENZO, SCALEPACK and

CCP4 (refs. 45,46). The crystals belonged to the P212121 space group with unit

cell dimensions a ¼ 66.3 A, b ¼ 81.3 A and c ¼ 96.4 A.

Molecular replacement in CNS47 using the free Chlorella virus ligase-AMP

intermediate (PDB 1FVI) as the search model yielded solutions for the

nucleotidyltransferase and OB domains of the Chlorella virus ligase–DNA

complex. The model was then subjected to rigid-body and all-atom refinement

using REFMAC48. The difference maps showed unambiguously the position

of the nick, the presence of a covalent lysyl-AMP adduct in the active site,

and the nucleic acid backbone. The DNA was built manually with O49, and

the structure of the complex was refined using CNS. The final model at

2.3-A resolution (Rfree ¼ 26.1, R ¼ 21.4) consisted of the complete DNA ligand

and a continuous Chlorella virus ligase polypeptide from residue 1 to 293

(Table 1). Electron density was not observed for the N-terminal His tag or the

C-terminal pentapeptide (294–298). The ligase protein model had excellent

geometry, with only one residue (Thr37) in the disallowed regions of the

Ramachandran plot.

ART IC L E S

77 6 VOLUME 14 NUMBER 8 AUGUST 2007 NATURE STRUCTURAL & MOLECULAR BIOLOGY

©20

07 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nsm

b

Nick sealing in crystallo after soaking in manganese. Crystals of ligase in

complex with nicked DNA (grown as described above) were transferred to a

solution containing 100 mM Bis-Tris (pH 6.5), 100 mM ammonium acetate,

22% (w/v) PEG 4,000 and 5 mM MnCl2 for 5 min, then placed into a solution

containing the same components plus 15% (v/v) glycerol before flash-freezing

of the crystals in liquid nitrogen. Diffraction data for the manganese-soaked

crystals were collected (to 3.0-A resolution) at a wavelength of 0.9795 A at

100 K at the National Synchrotron Light Source (beamline X9A) and processed

as described above. The crystals belonged to space group P1 with unit

cell dimensions a ¼ 66.5 A, b ¼ 92.9 A, c ¼ 89.6 A, a ¼ 70.31, b ¼ 78.51

and g ¼ 89.91.

Phases for the structure determination of this crystal form were obtained via

molecular replacement in AMORE50 using the ligase-AMP–nicked DNA

complex as the search model. The initial best rotation function was fixed and

subsequent protomers were added to assemble a complete model of four

protomers in the asymmetric unit that gave a correlation coefficient on

intensities of 72.6% with an overall crystallographic residual of 36.3% for data

from 15.0 to 3.6 A. Initial difference density maps revealed that the DNA in all

four complexes had been sealed. The connectivity of the DNA was confirmed by

higher-quality difference and composite-omit simulated annealing maps. No

clear density was observed for AMP, and it was not included in the final

structure. The initial molecular-replacement solution was rebuilt by hand using

O and refined in CNS using a combination of rigid-body, positional, restrained

noncrystallographic symmetry and restrained individual B-factor refinement

(Table 1). Water molecules were added when evinced clearly by composite-

omit map density, positive Fo – Fc density and the presence of a corresponding

water molecule in the higher-resolution ligase-AMP–nicked DNA complex.

The final model at 3.0-A resolution had Rfree ¼ 28.3 and R ¼ 24.9 (Table 1).

Ligase activity assays. Adenylyltransferase reaction mixtures (20 ml) containing

50 mM Tris HCl (pH 7.5), 5 mM DTT, 5 mM MgCl2, 20 mM [a-32P]ATP and

2 pmol wild-type ligase or DLatch mutant were incubated for 5 min at 37 1C.

The products were resolved by SDS-PAGE and detected by autoradiography.

Nick-sealing reaction mixtures (10 ml) containing 50 mM Tris-HCl (pH 7.5),

5 mM DTT, 10 mM MgCl2, 1 mM ATP, 0.5 pmol 32P 5¢ end–labeled nicked

36-bp DNA substrate (shown in Fig. 4d) and wild-type ligase or DLatch

mutant as specified were incubated at 22 1C for 10 min. The products were

resolved by electrophoresis through an 18% (w/v) polyacrylamide gel contain-

ing 7 M urea in 90 mM Tris-borate and 2.5 mM EDTA. The extent of ligation

was determined by phosphorimaging. To gauge the effects of salt on nick

sealing, reaction mixtures (10 ml) containing 50 mM Tris-HCl (pH 7.5), 5 mM

DTT, 10 mM MgCl2, 1 mM ATP, 0.5 pmol 32P 5¢ end–labeled nicked 36-bp

DNA substrate, either 1.25 ng wild-type ligase or 5 ng DLatch mutant, and 0,

25, 50, 100 or 150 mM NaCl were incubated at 22 1C for 10 min. To assay

binding of ligase to nicked DNA, reaction mixtures (20 ml) containing 50 mM

Tris-HCl (pH 7.5), 5 mM DTT, 250 fmol 32P-labeled nicked 36-bp DNA, and

2.7, 5.4 or 10.9 pmol wild-type ligase or DLatch mutant (corresponding to 0.3,

0.6 or 1.2 pmol of wild-type ligase-AMP and 0.4, 0.8 or 1.6 pmol of DLatch-

AMP) were incubated for 10 min at 22 1C. Glycerol was added to 5% (v/v) and

the samples were analyzed by electrophoresis through a 15-cm nondenaturing

6% (w/v) polyacrylamide gel in 90 mM Tris-borate and 2.5 mM EDTA (110 V

for 3 h at 4 1C). Free DNA and ligase–DNA complexes of retarded mobility

were visualized by autoradiography.

Accession codes. Protein Data Bank: coordinates of the ligase-AMP–nick

complex and the ligase-sealed DNA product complex have been deposited

with accession codes 2Q2T and 2Q2U, respectively.

Note: Supplementary information is available on the Nature Structural & MolecularBiology website.

ACKNOWLEDGMENTSThis paper is inspired by the 40th anniversary of the discovery of viral ATP-dependent DNA ligases by Charles Richardson and Jerry Hurwitz. This workwas supported by US National Institutes of Health grants GM63611 (S.S.) andGM61906 (C.D.L.). S.S. is an American Cancer Society Research Professor.COMPETING INTERESTS STATEMENTThe authors declare no competing financial interests.

Published online at http://www.nature.com/nsmb/

Reprints and permissions information is available online at http://npg.nature.com/

reprintsandpermissions

1. Gellert, M. Formation of covalent circles of lambda DNA by E. coli extracts. Proc. Natl.Acad. Sci. USA 57, 148–155 (1967).

2. Zimmerman, S.B., Little, J.W., Oshinsky, C.K. & Gellert, M. Enzymatic joining of DNAstrands: a novel reaction of diphosphopyridine nucleotide. Proc. Natl. Acad. Sci. USA57, 1841–1848 (1967).

3. Little, J.W., Zimmerman, S.B., Oshinsky, C.K. & Gellert, M. Enzymatic joining of DNAstrands, II: an enzyme-adenylate intermediate in the DPN-dependent DNA ligasereaction. Proc. Natl. Acad. Sci. USA 58, 2004–2011 (1967).

4. Olivera, B.M. & Lehman, I.R. Linkage of polynucleotides through phosphodiester bondsby an enzyme from Escherichia coli. Proc. Natl. Acad. Sci. USA 57, 1426–1433(1967).

5. Olivera, B.M. & Lehman, I.R. Diphosphopyridine nucleotide: a cofactor for thepolynucleotide-joining enzyme from Escherichia coli. Proc. Natl. Acad. Sci. USA 57,1700–1704 (1967).

6. Olivera, B.M., Hall, Z.W. & Lehman, I.R. Enzymatic joining of polynucleotides, V:a DNA-adenylate intermediate in the polynucleotide-joining reaction. Proc. Natl. Acad.Sci. USA 61, 237–244 (1968).

7. Weiss, B. & Richardson, C.C. Enzymatic breakage and joining of deoxyribonucleicacid, I: repair of single-strand breaks in DNA by and enzyme system from Escherichiacoli infected with T4 bacteriophage. Proc. Natl. Acad. Sci. USA 57, 1021–1028(1967).

8. Weiss, B. & Richardson, C.C. Enzymatic breakage and joining of deoxyribonucleic acid,III: an enzyme-adenylate intermediate in the polynucleotide ligase reaction. J. Biol.Chem. 242, 4270–4278 (1967).

9. Gefter, M.L., Becker, A. & Hurwitz, J. The enzymatic repair of DNA, I: formation ofcircular DNA. Proc. Natl. Acad. Sci. USA 58, 240–247 (1967).

10. Becker, A., Lyn, G., Gefter, M. & Hurwitz, J. The enzymatic repair of DNA, II:characterization of phaged-induced sealase. Proc. Natl. Acad. Sci. USA 58,1996–2003 (1967).

11. Tomkinson, A.E., Vijayakumar, S., Pascal, J.M. & Ellenberger, T. DNA ligases:structure, reaction mechanism, and function. Chem. Rev. 106, 687–699 (2006).

12. Subramanya, H.S., Doherty, A.J., Ashford, S.R. & Wigley, D.B. Crystal structure of anATP-dependent DNA ligase from bacteriophage T7. Cell 85, 607–615 (1996).

Table 1 Data collection and refinement statistics

Ligase–DNA complex Ligase–DNA Mn2+ soak

Data collection

Space group P212121 P1

Cell dimensions

a, b, c (A) 66.30, 81.29, 96.42 66.48, 92.90, 89.63

a, b, g (1) 90, 90, 90 70.32, 78.45, 89.87

Resolution (A) 50–2.3 (2.38–2.3) 50–3.0 (3.11–3.0)

Rmerge 8.4 (33.2) 7.8 (45.5)

I / sI 14.2 (3.2) 7.3 (1.4)

Completeness (%) 96.9 (95.7) 95.8 (87.7)

Redundancy 3.6 (3.1) 1.9 (1.7)

Refinement

Resolution (A) 39.0–2.3 24.9–3.0

No. reflections 23,077 38,001

Rwork / Rfree 21.4 / 26.1 24.9 / 28.3

No. atoms

Protein 2,383 9,444

Ligand/ion 858 3428

Water 145 56

B-factors

Protein 42.5 48.3

Ligand/ion 42.5 36.9

Water 44.8 22.2

R.m.s. deviations

Bond lengths (A) 0.006 0.007

Bond angles (1) 1.2 1.1

One crystal was diffracted for each structure. Values in parentheses are for the highest-resolution shell.

ART IC L E S

NATURE STRUCTURAL & MOLECULAR BIOLOGY VOLUME 14 NUMBER 8 AUGUST 2007 7 7 7

©20

07 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nsm

b

13. Lee, J.Y. et al. Crystal structure of NAD+-dependent DNA ligase: modular architectureand functional implications. EMBO J. 19, 1119–1129 (2000).

14. Odell, M., Sriskanda, V., Shuman, S. & Nikolov, D.B. Crystal structure of eukaryoticDNA ligase-adenylate illuminates the mechanism of nick sensing and strand joining.Mol. Cell 6, 1183–1193 (2000).

15. Pascal, J.M., O’Brien, P.J., Tomkinson, A.E. & Ellenberger, T. Human DNA ligase Icompletely encircles and partially unwinds nicked DNA. Nature 432, 473–478 (2004).

16. Akey, D. et al. Crystal structure and nonhomologous end joining function of the ligasedomain of Mycobacterium DNA ligase D. J. Biol. Chem. 281, 13412–13423 (2006).

17. Pascal, J.M. et al. A flexible interface between DNA ligase and PCNA supportsconformational switching and efficient ligation of DNA. Mol. Cell 24, 279–291(2006).

18. Nishida, H., Kiyonari, S., Ishino, Y. & Morikawa, K. The closed structure of an archaealDNA ligase from Pyrococcus furiosus. J. Mol. Biol. 360, 956–967 (2006).

19. Nandakumar, J., Nair, P.A. & Shuman, S. Last stop on the road to repair: structure ofE. coli DNA ligase bound to nicked DNA-adenylate. Mol. Cell 26, 257–271 (2007).

20. Nandakumar, J., Shuman, S. & Lima, C.D. RNA ligase structures reveal the basisfor RNA specificity and conformational changes that drive ligation forward. Cell 127,71–84 (2006).

21. El Omari, K., Ren, J., Bird, L.E., Bona, M.K., Klarman, G., Legrice, S.F. & Stammers,D.K. Molecular architecture and ligand recognition determinants for T4 RNA ligase.J. Biol. Chem. 281, 1573–1579 (2006).

22. Hakansson, K., Doherty, A.J., Shuman, S. & Wigley, D.B. X-ray crystallography reveals alarge conformational change during guanyl transfer by mRNA capping enzymes. Cell89, 545–553 (1997).

23. Shuman, S. & Lima, C.D. The polynucleotide ligase and RNA capping enzymesuperfamily of covalent nucleotidyltransferases. Curr. Opin. Struct. Biol. 14,757–764 (2004).

24. Sriskanda, V. & Shuman, S. Mutational analysis of Chlorella virus DNA ligase: catalyticroles of domain I and motif VI. Nucleic Acids Res. 26, 4618–4625 (1998).

25. Sriskanda, V. & Shuman, S. Conserved residues in domain Ia are required for thereaction of Escherichia coli DNA ligase with NAD+. J. Biol. Chem. 277, 9695–9700(2002).

26. Gajiwala, K.S. & Pinko, C. Structural rearrangement accompanying NAD+ synthesiswithin a bacterial DNA ligase crystal. Structure 12, 1449–1459 (2004).

27. Ho, C.K., Van Etten, J.L. & Shuman, S. Characterization of an ATP-dependent DNAligase encoded by Chlorella virus PBCV-1. J. Virol. 71, 1931–1937 (1997).

28. Sriskanda, V. & Shuman, S. Chlorella virus DNA ligase: nick recognition and mutationalanalysis. Nucleic Acids Res. 26, 525–531 (1998).

29. Odell, M. & Shuman, S. Footprinting of Chlorella virus DNA ligase bound at a nick induplex DNA. J. Biol. Chem. 274, 14032–14039 (1999).

30. Sekiguchi, J. & Shuman, S. Nick sensing by DNA ligase requires a 5¢ phosphate atthe nick and occupancy of the adenylate binding site on the enzyme. J. Virol. 71,9679–9684 (1997).

31. Nandakumar, J. & Shuman, S. How an RNA ligase discriminates RNA versus DNAdamage. Mol. Cell 16, 211–221 (2004).

32. Sriskanda, V., Schwer, B., Ho, C.K. & Shuman, S. Mutational analysis of E. coli DNAligase identifies amino acids required for nick-ligation in vitro and for in vivo

complementation of the growth of yeast cells deleted for CDC9 and LIG4. NucleicAcids Res. 27, 3953–3963 (1999).

33. Odell, M., Malinina, L., Sriskanda, V., Teplova, M. & Shuman, S. Analysis of the DNAjoining repertoire of Chlorella virus DNA ligase and a new crystal structure of the ligase-adenylate intermediate. Nucleic Acids Res. 31, 5090–5100 (2003).

34. Pascal, J.M. & Ellenberger, T. RNA ligase does the AMP shuffle. Nat. Struct. Mol. Biol.13, 950–951 (2006).

35. Marchetti, C. et al. Identification of a novel motif in DNA ligases exemplified by DNAligase IV. DNA Repair (Amst.) 5, 788–798 (2006).

36. Sriskanda, V. & Shuman, S. Specificity and fidelity of strand joining by Chlorella virusDNA ligase. Nucleic Acids Res. 26, 3536–3541 (1998).

37. Sriskanda, V. & Shuman, S. Role of nucleotidyltransferase motifs I, III and IV in thecatalysis of phosphodiester bond formation by Chlorella virus DNA ligase. NucleicAcids Res. 30, 903–911 (2002).

38. Nandakumar, J. & Shuman, S. Dual mechanisms whereby a broken RNA end assiststhe catalysis of its repair by T4 RNA ligase 2. J. Biol. Chem. 280, 23484–23489(2005).

39. Shuman, S. Vaccinia DNA ligase: specificity, fidelity, and inhibition. Biochemistry 34,16138–16147 (1995).

40. Ahel, I. et al. The neurodegenerative disease protein resolves abortive DNA ligationintermediates. Nature 443, 713–716 (2006).

41. Gong, C. et al. Mechanism of nonhomologous end-joining in mycobacteria: a lowfidelity repair system driven by Ku, ligase D and ligase C. Nat. Struct. Mol. Biol. 12,304–312 (2005).

42. Gong, C., Martins, A., Bongiorno, P., Glickman, M. & Shuman, S. Biochemical andgenetic analysis of the four DNA ligases of mycobacteria. J. Biol. Chem. 279,20594–20606 (2004).

43. Weller, G.R. et al. Identification of a DNA nonhomologous end-joining complex inbacteria. Science 297, 1686–1689 (2002).

44. Doherty, A.J., Ashford, S.R. & Wigley, D.B. Characterization of proteolytic fragments ofbacteriophage T7 DNA ligase. Nucleic Acids Res. 24, 2281–2287 (1996).

45. Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected in oscillationmode. Methods Enzymol. 276, 307–326 (1997).

46. Collaborative Computational Project. Number 4. The CCP4 suite: programs for proteincrystallography. Acta Crystallogr. D Biol. Crystallogr. 50, 760–763 (1994).

47. Brunger, A.T. et al. Crystallography & NMR system: a new software suite for macro-molecular structure determination. Acta Crystallogr. D Biol. Crystallogr. 54, 905–921(1998).

48. Murshudov, G.N., Vagin, A.A. & Dodson, E.J. Refinement of macromolecular structuresby the Maximum-Likelihood Method. Acta Crystallogr. D Biol. Crystallogr. 53,240–255 (1997).

49. Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjeldgaard, M. Improved methods for buildingprotein models in electron density maps and the location of errors in these models.Acta Crystallogr. A 47, 110–118 (1991).

50. Navaza, J. AMoRe: an automated package for molecular replacement. Acta Crystallogr.A 50, 157–163 (1994).

51. DeLano, W.L. The PyMOL Molecular Graphics System (DeLano Scientific, San Carlos,California, USA, 2002).

ART IC L E S

77 8 VOLUME 14 NUMBER 8 AUGUST 2007 NATURE STRUCTURAL & MOLECULAR BIOLOGY

©20

07 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nsm

b