ARTICLES

Structure and mechanism of human DNApolymerase gChristian Biertumpfel1, Ye Zhao1,2, Yuji Kondo3, Santiago Ramon-Maiques1{, Mark Gregory1, Jae Young Lee1{,Chikahide Masutani3, Alan R. Lehmann4, Fumio Hanaoka3,5 & Wei Yang1

The variant form of the human syndrome xeroderma pigmentosum (XPV) is caused by a deficiency in DNA polymerase g(Polg), a DNA polymerase that enables replication through ultraviolet-induced pyrimidine dimers. Here we reporthigh-resolution crystal structures of human Polg at four consecutive steps during DNA synthesis through cis-syncyclobutane thymine dimers. Polg acts like a ‘molecular splint’ to stabilize damaged DNA in a normal B-form conformation.An enlarged active site accommodates the thymine dimer with excellent stereochemistry for two-metal ion catalysis. Tworesidues conserved among Polg orthologues form specific hydrogen bonds with the lesion and the incoming nucleotide toassist translesion synthesis. On the basis of the structures, eight Polg missense mutations causing XPV can be rationalizedas undermining the molecular splint or perturbing the active-site alignment. The structures also provide an insight into therole of Polg in replicating through D loop and DNA fragile sites.

Life on Earth can scarcely escape the ultraviolet radiation of sunlight,which catalyses covalent linkages between adjacent pyrimidines inDNA1. The resulting pyrimidine dimers are barriers to most DNApolymerases, but the human POLH gene, defective in XPV, encodes aY-family DNA polymerase, Polg, that is specialized for translesionsynthesis (TLS) through cyclobutane pyrimidine dimers (CPDs)2,3.XPV is characterized by sunlight-induced pigmentation changes anda highly elevated incidence of skin malignancies. In contrast tohumans, yeast lives well without Polg when chronically exposed tolow doses of ultraviolet radiation4. In humans, TLS by Polg candecrease cellular sensitivity to anticancer DNA adducts, for example,cisplatin5. In addition to TLS, Polg is involved in somatic hyper-mutation6 and replication of D loops and DNA fragile sites7–10.

Polg is the only one of the fifteen human DNA polymerases inwhich defects are unequivocally associated with cancer11–13. Amonghuman Y-family polymerases (Rev1 and polymerases g, i and k), thelesion-bypass specificity and efficiency of Polg stand out. Polg bindsCPD-containing DNA better than undamaged DNA and extendsprimers more processively opposite and up to two nucleotides beyondCPDs14,15. However, Polg is completely blocked by ultraviolet-induced pyrimidine (6-4) pyrimidone photoproducts (6-4PP)16.6-4PPs occur less frequently than CPDs and distort the helical struc-ture of DNA more severely and are thus more efficiently recognizedand removed by nucleotide-excision repair proteins17,18.

Structural analyses of Y-family TLS polymerases reveal a spaciousactive site that accommodates DNA adducts and non-Watson–Crickbase pairs with little discrimination19,20. However, the molecular basisfor the specificity and efficiency of Polg in DNA synthesis throughCPDs has remained unclear despite many crystal structures, includ-ing those of the apo- and cisplatin-DNA-bound yeast Polg21,22.

We report here the crystal structures of the polymerase domain ofhuman Polg catalysing the four-step TLS through a cis-syn thyminedimer and in a ternary complex with normal DNA. Complementingthese structures, two residues conserved among Polg orthologues

were mutated, and the resulting defects in TLS assessed by functionalstudies.

Crystallization and features of human Polg

The catalytic domain of human Polg (1–432 amino acids) wasexpressed in Escherichia coli after codon optimization (see Methods).Crystals of ternary complexes were initially grown with normal DNAstrands, and the structure was solved and refined to 2.9 A resolution.However, a hydrophobic crystal-lattice contact between protein mole-cules appeared to distort Polg–DNA interactions as compared withknown structures of Y-family polymerases (Supplementary Fig. 1).Multiple amino-acid substitutions were engineered to break thislattice contact. Non-hydrolysable dNMPNPPs (29-deoxynucleoside-59-[(a,b)-imido]-triphosphate) were synthesized to replace dNTPsfor co-crystallization of an active human Polg, DNA and Mg21.Different DNA lengths were systematically sampled and proven mosteffective for breaking the undesirable lattice contacts. After many trials,two new crystal forms were grown that diffracted X-rays to ,3 A andsubsequently improved to 1.75–2.15 A (Supplementary Tables 1 and 2).

The crystal structure of wild-type human Polg in complex withnormal DNA and dAMPNPP (29-deoxy-adenosine-59-[(a,b)-imido]-triphosphate) opposite the template T (denoted Nrm) was determinedand refined to 1.83 A (see Methods) (Fig. 1a). The relevance of thestructure is validated by the appearance of a well-formed active sitecentred on two properly coordinated Mg21 ions and the 39-OH of theprimer strand 3.2 A from the a-phosphate of dAMPNPP poised for in-line nucleophilic attack (Fig. 1b). The single water ligand of the Mg21

ions (Fig. 1b) has been observed in several DNA polymerases and mayparticipate in catalysis23. Configured around the non-hydrolysabledNTP analogue, the active site is arguably the most reaction readyamong the homologous A-, B- and Y-family DNA polymerases crystal-lized so far. The incoming nucleotide, catalytic carboxylates and metalions are nearly superimposable with those of human Polb (X-family)despite dissimilar tertiary structures (Supplementary Fig. 2). This active

1Laboratory of Molecular Biology, NIDDK, NIH, 9000 Rockville Pike, Building 5, Room B103, Bethesda, Maryland 20892, USA. 2Institute of Nuclear-Agricultural Sciences, ZhejiangUniversity, Hangzhou 310029, China. 3Graduate School of Frontier Biosciences, Osaka University, Osaka 565-0871, Japan. 4Genome Damage and Stability Centre, University ofSussex, Falmer, Brighton BN1 9RQ, UK. 5Faculty of Science, Gakushuin University, 1-5-1 Mejiro, Toshima-ku, Tokyo 171-8588, Japan. {Present addresses: Spanish National CancerResearch Centre, Madrid 28029, Spain (S.R.-M.); Dongguk University-Seoul, Seoul 100-715, Korea (J.Y.L.).

Vol 465 | 24 June 2010 | doi:10.1038/nature09196

1044Macmillan Publishers Limited. All rights reserved©2010

site configuration is probably conserved among all DNA polymerasesand possibly all RNA polymerases as well.

Like all Y-family polymerases, human Polg contains four domains—palm, finger, thumb and little finger (LF)—with the active site in thepalm domain and DNA bound between thumb and LF. Unlike thestructures of yeast Polg–cisplatin complexes22, the finger domain ofhuman Polg is closed and contacts the replicating base pair (Fig. 1cand Supplementary Fig. 3a), and the LF domain interacts extensivelywith both template and primer in the major groove (Fig. 1d). A surfacearea of 2,200 A2 is buried between human Polg and DNA. The finger,palm and incoming dNTP in human Polg superimpose well withthose in Dpo4 and polymerases k and i (Supplementary Fig. 3a).Uniquely in human Polg, however, two bases of the template strandinstead of one are in the active site, and the 39 T is base-paired with theincoming dAMPNPP (Fig. 1c). Translocation of two template basesinto the active site was previously observed with Dpo4, but there the 59

base was paired with the dNTP, resulting in a misalignment24. Inhuman Polg a slight shift of LF relative to the finger domain enlargesthe active site, and allows it to accommodate two template baseswithout misalignment. Concomitantly, the thumb and LF domainsof human Polg are oriented differently relative to Dpo4 and poly-merases i and k (Supplementary Fig. 3a). The thumb domain onlycontacts the DNA primer (Fig. 1d).

Structures displaying TLS through a CPD

CPD-containing oligonucleotides were synthesized to place eachcrosslinked thymine at the template position or 1 or 2 base pairs(bp) upstream (Supplementary Table 1). Four ternary complexes werecrystallized and labelled as TT1, TT2, TT3 and TT4. All except TT2were crystallized in the same space group as the complex of normal

DNA (Nrm). These structures were determined by molecular replace-ment (see Methods). After refinement, the four CPD-containing com-plexes are virtually superimposable with Nrm (Fig. 2a). The root meansquared deviation values among proteins are 0.15–0.71 A over ,400pairs of Ca atoms.

In TT1 the 39 thymine of CPD serves as the template and forms aWatson–Crick base pair with dAMPNPP, much like an undamagedbase (Fig. 2b). The 59 thymine of CPD is turned and moved closer tothe finger domain than the undamaged base in Nrm. The tighterinteractions with the CPD may explain the better fidelity of humanPolg in TLS (Fig. 3a). Ser 62 forms van der Waals contacts with the 59

base in both cases (Fig. 1c), but Ser 62 is not conserved and replacingit with Gly increases the TLS efficiency of human Polg25. Even withthe S62G substitution, a 6-4PP would not fit in the active site(Supplementary Fig. 4a), which explains the selectivity of Polg.

In TT2, the 59 thymine of CPD also forms a Watson–Crick base pairwith dAMPNPP (Fig. 2c). The finger domain in TT2 is more open thanin the other human Polg complex structures. Despite being crosslinkedand rotationally constrained (Fig. 2a), the 39 thymine of the CPDmaintains a Watson–Crick base pair with the primer, which undergoesminimal changes relative to the Nrm and TT1 complexes. This differsfrom CPDs complexed with a stalled replicative polymerase, whereonly one crosslinked thymine is base paired26. In TT1 and TT2, thetemplate thymine adopts the position most similar to that of undam-aged DNA, and the crosslinked partners adjust. In TT2 the deoxyriboseof the 39 thymine moves by 2–3 A and the base is shifted towards themajor groove by ,2 A and tilted by ,20u (Fig. 2c).

Each template thymine in TT1 and TT2 is hydrogen bonded toGln 38 of human Polg via its O2 oxygen (Fig. 2b, c). Both cytosine

a

c

Palm

Thumb

LF

Finger

5´T

3´T

dAMPNPP

R93

R61S62

Y92

K86

L89

F18

Q38I48I48

Y39

D13

D115E116

M14(O)

dAMPNPP

P316

R351

R313R313

K317K317

R377

G259

K261

G260G260

N324N324

R383

b

d

Figure 1 | Structure of human Polg. a, The ternary complex of human Polgwith a normal DNA. Protein domains are shown in distinct colours andlabelled. The DNA template is coloured orange and the primer yellow.Oxygen and nitrogen atoms are coloured red and blue, respectively.dAMPNPP is shown in ball-and-stick representation; Mg21 is shown aspurple spheres. All structural figures were made using PyMOL (http://www.pymol.org). b, The active site. Mg21 coordination is indicated by paleyellow dashed lines. The 39-OH of the primer strand is 3.2 A from thea-phosphate as indicated by the red dashed line. c, Human Polg–DNAinteractions around the active site. Protein side chains from the finger andpalm domain are shown as light blue and pink sticks, respectively.d, Interactions with the upstream DNA. The LF domain (shown as lightpurple ribbon diagram) contacts both template and primer, and the thumbdomain (shown in green) only makes 3–4 hydrogen bonds with the primerstrand. Side chains that make DNA contacts are highlighted in stickrepresentation and labelled.

a

c

d

e

TT1 vs Nrm TT2 vs Nrm

TT3 vs Nrm TT4 vs Nrm

Nrm

TT1

TT2

TT4

Q38

Q38

Q38

Q38

R61

R61

R61

R61

F18

F18

F18

F18

TT3

CPD

CPD

b

Figure 2 | Structures of lesion DNAs. a, Overlay of TT1, TT2, TT3 and TT4DNA with the DNA in Nrm (undamaged). CPDs are shown as red sticks.DNA and Mg21 are coloured as in Fig. 1. b–d, The replicating base pairs infour ternary complexes (highly similar in TT3 and TT4) are each shown withhuman Polg residues that contact the bases and deoxyribose. Hydrogenbonds are shown as dashed lines. e, The CPD in TT3 and TT4 formsWatson–Crick-like hydrogen bonds but cannot base stack withneighbouring nucleotides.

NATURE | Vol 465 | 24 June 2010 ARTICLES

1045Macmillan Publishers Limited. All rights reserved©2010

and thymine are vulnerable to forming ultraviolet-induced CPDs,and both have the O2 to form the hydrogen bond with Gln 38 asobserved in TT1 to TT4 (Fig. 2d). In Nrm Gln 38 is instead hydrogenbonded with deoxyribose (Fig. 2b) owing to a slight shift of the DNAtemplate (Fig. 2a). Gln 38 is one of the two residues uniquely con-served in the Polg family (Supplementary Fig. 5). Ala substitution ofGln reduces the catalytic efficiency as expected (Fig. 3a). Notably,Q38A increases polymerase stalling after the CPD at the stage equi-valent to TT3 (Fig. 3b). This stalling probably occurs because thetemplate base cannot stack with the upstream CPD (Fig. 2e) anddepends on Gln 38, its sole contact with the polymerase, to align withthe incoming dNTP.

The second invariant residue in the Polg family is an Arg (Arg 61 inhuman Polg). In the crystal structures of the yeast Polg–cisplatinDNA complex, its equivalent (Arg 73) forms cation–p interactionswith the base and polar interactions with the phosphates of the incom-ing dNTP22. In the five human Polg structures, the closed fingerdomain prevents Arg 61 from stacking with the base. Arg 61 insteadadopts different rotamer conformations and is hydrogen bonded withthe N7 of purines (A or G) or the phosphates (Fig. 2b–d), thus favour-ing the anti-conformation of dNTP and preventing Hoogsteen basepairing27. Ala substitution of Arg 61 reduces the polymerase efficiencyas well as misincorporation of dG opposite a template T (Fig. 3a).However, dGTP would be the correct incoming nucleotide if a cyto-sine is in the CPD. Arg 61 is thus important for Polg to synthesizeDNA efficiently through cis-syn cytosine and thymine dimers, but atthe cost of potential misincorporation.

Human Polg is a molecular splint

The DNAs in all four CPD–Polg complexes maintain a straight B-formconformation (Fig. 2a) in spite of the inability of CPDs to base stack.This unperturbed structure is in contrast to the bent and unwoundstructures of CPD-containing DNAs alone or complexed with repairproteins (Supplementary Fig. 6)28–31. Poor base stacking and segmenta-tion of the DNA helices are the structural features of CPDs that arerecognized by repair proteins32. However, when complexed withhuman Polg, lesion-induced perturbations are absorbed by minoradjustments to torsion angles of the surrounding nucleotides.

Human Polg acts like a molecular splint to keep the damagedDNA straight and rigid owing to a continuous and highly positivelycharged DNA-binding surface that interacts extensively with the fourtemplate nucleotides immediately upstream of the active site(Fig. 4a). A b-strand in the LF domain (amino acids 316–324) isnearly parallel to the template strand, and every other main-chainamide donates a hydrogen bond to the template phosphates (Figs 1dand 4b). Each phosphate forms additional hydrogen bonds with sidechains of Arg, Lys, Tyr and Thr. In contrast, the template-bindingsurface in other Y-family polymerases has a gap or holes owing toseparations between LF and the catalytic core (palm, thumb andfinger) (Fig. 4a and Supplementary Fig. 3a). The gap is particularlylarge in Polk, and the divided protein requires an N-terminalappendage (N-clasp) for stabilization33. Polk and Dpo4 probablyuse the structural gaps (Fig. 4a) to accommodate bulky minor-grooveadducts during TLS34,35. The gaps in Dpo4 and Dbh also promotetemplate looping out as a means of lesion bypass36,37. In human Polgthe LF domain is connected to the catalytic core by hydrogen bonds,salt bridges and hydrophobic interactions (Supplementary Fig. 3b–c).The resulting DNA-binding surface constrains the template backboneand reinforces a B-form structure in spite of CPDs.

The side chains of Arg 93 and Arg 111 extend from the palm domainto the template strand and further strengthen this molecular splint.Arg 93 connects the template strand with the incoming nucleotide viaits neighbour, Tyr 92, which stacks with Phe 18 (the steric gate)38 and

b

- 3 6 9 12 15 18 - 3 6 9 12 15 18 min

ND CPD

3′T5′T+1C+2T

–2A–1C

3′T5′T

+1C+2T

–2A–1C

3′T5′T+1C+2T

–2A–1C

3′T5′T

+1C+2T

–2A–1C

ND CPD

dATP incorporationa dGTP misincorporation

NDCPD

0.04

4

0.120.

55

7.511

20

A–2 C–1 T3′ T5′ C+1 T+2 A–2 C–1 T3′ T5′ C+1 T+2

Term

inat

ion

pro

bab

ility

(%)

Term

inat

ion

pro

bab

ility

(%)80

706050403020100

80706050403020100

WT Q38A

NDCPD

5′ [32P] – CACTGACTGTATGA3′ GTGACTGACATACTACTTCTACGACTGCTC 5′

Kca

t/Km

(μM

–1 m

in–1

)

Kca

t/Km

(μM

–1 m

in–1

)

- 3 6 9 12 15 18 - 3 6 9 12 15 18 min

60

40

20

0

4

2

0WT Q38A R61A WT Q38A R61A

Figure 3 | Functional analyses of Q38A and R61A mutant Polg.a, Efficiencies (kcat/Km) of wild type (WT) and two mutants in correct andincorrect nucleotide incorporation on normal (ND) and CPD DNA. Thesevalues are derived from Supplementary Table 3. b, CPD bypass by wild-type(left) and Q38A mutant Polg (right). The DNA sequence for primerextension assays is shown. Magenta-coloured TT represents the CPD. Thesame DNA template with a two-nucleotide-longer primer was used for thekinetic assay shown in Fig. 3a. Reactions were carried out at asubstrate:enzyme molar ratio of 100:1 for the indicated incubation time.Termination probabilities ([In]/Sn2end[I], where In is band intensity at thenth position) on normal (ND) and CPD DNA are plotted below.

Dpo4a

c

Human Rev1

Human Polη

Human Polι Human Polκ

Dbh

R111

R93

F18 R61Q38Y39

K317

P316

N324

S322

F423T318

M421

R256

K261L378

F423

G259P316

R111 K317

N324K86

R351

R93

R313

b

Figure 4 | Human Polg is a molecular splint. a, The DNA binding surfaces ofvarious Y-family polymerases are shown with electrostatic surface potentials.Blue and red represent the positive and negative charge potential at the 1 and210 kT e21 scale, respectively. The thumb domain in all cases and the N-claspof Polk are removed for clarity. Five phosphorus atoms at position 21 to 25in the human PolgNrm (yellow) and TT4 (orange) structures are also shown.b, The extensive interactions between human Polg and the template strand inTT4 (as an example) are shown. Hydrogen bonds are indicated by dashedlines. The finger and LF domains are coloured in light purple and light green.Two residues forming hydrophobic interactions with the CPD bases areshown in teal. c, A view approximately 90u from b. Detailed humanPolg–DNA interactions are shown in Supplementary Fig. 9.

ARTICLES NATURE | Vol 465 | 24 June 2010

1046Macmillan Publishers Limited. All rights reserved©2010

the deoxyribose of dNTP (Fig. 1c). Although this Tyr 92/Arg 93 pair isfound in polymerases i and k, only in Polg is the DNA templatehydrogen bonded to Arg 93. Arg 111 is adjacent to the catalyticcarboxylate Asp 115 and interacts with both LF and the templatestrand (Fig. 5a), which explains why substitution of Arg 111 by Hisleads to the XPV phenotype11.

Dissociation of human Polg after CPDs

TLS polymerases are recruited to stalled replication forks by Rad6–Rad18 mediated ubiquitination of PCNA39. Dissociation after TLS,however, may be an intrinsic property of human Polg because ofreduced affinity for DNA 3 bp past the CPD14,15. In all five structures,the DNA-binding surface of human Polgmakes extensive interactionswith 4 bp (Fig. 4), but has reduced interactions with the templatestrand further upstream. Modelling a CPD 1 bp further upstream fromTT4 reveals steric clashes of its phosphates with human Polg in addi-tion to the loss of favourable hydrogen bonds (Supplementary Fig. 4b).

The crosslinked thymines are shifted 1–2 A into the major groove(Fig. 2). The C5 methyl groups form van der Waals contacts with LF(L378 and F423) in TT3 and TT4 (Fig. 4c). In Nrm, the space betweenthe DNA bases and the LF domain is filled with water and glycerolmolecules. These major-groove interactions are lost when the CPD is3 bp beyond the active site. The position-dependent CPD–Polginteractions provide a structural basis for polymerase switching afterlesion bypass.

XPV mutations and beyond

Many mutations in Polg have been identified in XPV patients, andfive missense mutations (R111H, A117P, T122P, G263V and R361S)are within the first 432 amino acids11,12. These mutations and threenewly identified XPV mutations (A264P, F290S and G295R; A.R.L.,unpublished data) are modelled in Fig. 5b. Ala 117 adjoins the catalytic

carboxylate Glu 116, and substitution by Pro would cause clashes withthe carbonyl oxygen of Val 12. Thr 122 occurs at the beginning of ana-helix and participates in the hydrogen-bond network surroundingall three catalytic carboxylates (Asp 13, Asp 115 and Glu 116) (Fig. 5c).T122P would probably perturb the active site as does A117P. Arg 111is a part of the molecular splint (see the previous section). Arg 361 ishydrogen bonded with the carbonyl oxygen of Pro 316 and anchorsthe b-strand (amino acids 316–324) for template binding (Fig. 5a, d).The R111H and R361S mutations are likely to relax the protein–DNAinterface and break the molecular splint.

The remaining four XPV mutations are in the thumb domain andaffect primer binding, which is essential to complete the molecularsplint. Gly 263 to Val and Ala 264 to Pro mutations would cause stericclashes with the carbonyl oxygens of Leu 258 and Lys 261, respec-tively (Fig. 5e). These clashes would disrupt primer binding by theGGK motif (amino acids 259–261) (Figs 1d and 4c). Phe 290 andGly 295 form the hydrophobic core of the thumb domain, and F290Sand G295R mutations would destabilize the structure. HomozygousW174C has been implicated in XPV13, and V99M and I272T havebeen found in melanoma patients40. However, they are unlikely toalter directly the structure or polymerase activity of Polg (Sup-plementary Fig. 7). The W174C mutation led to a reduced Polg levelprobably due to altered pre-mRNA processing and protein stability13.

Human Polg has been shown to synthesize through D loop andfragile sites, which contain hairpins and non-B-form structures7–10.Interestingly, in both crystal forms of human Polg the back of the LFdomain interacts with a neighbouring DNA (Supplementary Fig. 8).The two symmetry-related DNA molecules are reminiscent of D-loopand hairpin structures, as the template strands are nearly contiguous.Mimics of downstream DNA have not been observed previously inpolymerase–DNA co-crystals. Many Polg molecules appear to retainthis secondary DNA-binding site, although the interfacial residues(Arg 81, Arg 84 and Trp 339) are not strictly conserved (Supplemen-tary Fig. 5). We hypothesize that the back of human Polg may be anauthentic downstream DNA-binding site, and the LF domain serves asa wedge to separate non-B-form DNAs and aids the molecular splintto complete replication through D loop and fragile sites.

Concluding remarks

Human Polg is specifically optimized for bypassing the most ubiqui-tous DNA lesion. An enlarged active site accommodates CPDs, thecomplementary DNA-binding surface reinforces the B-form con-formation, and hydrophobic residues interact with crosslinked pyrimi-dines in the major groove. These features readily explain the specificityof human Polg in TLS through CPDs and reduced efficiencies for otherlesions including cisplatin41–44. The limited interactions with the DNAminor groove may enable Polg to misincorporate nucleotides duringsomatic hypermutation. The high-resolution structures reported herenot only provide a detailed mechanism of TLS but also generate testablemodels for investigating novel functions of Polg.

METHODS SUMMARY

The full-length human Polg gene codon-optimized for E. coli expression was

synthesized by GenScript. The catalytic core (amino acids 1–432, human Polg)

was cloned into modified pET28a45, expressed in E. coli and purified by Ni21-

affinity, MonoS and Superdex75 chromatography. The His tag was removed by

PreScission protease. Mutagenesis was performed using QuikChange (Stratagene).

Non-hydrolysable dNMPNPPs were purchased from Jena Bioscience, and phos-

phoramidites of CPD from Glen Research. CPD oligonucleotides were synthesized

and purified by TriLink Biotechnologies. Ternary complexes were prepared by

mixing wild-type or C406M mutant human Polg and annealed DNA at a 1:1.05

molar ratio and addition of 5 mM Mg21 and 1 mM dNMPNPPs. The final protein

concentration was 6–7 mg ml21. Crystals were grown in 0.1 M MES (pH 6.0),

19–21% (w/v) PEG 2K-MME and 5 mM MgCl2 after several rounds of microseed-

ing. Diffraction data were collected at sectors 22 and 23 of the APS. Phases were

determined by molecular replacement46 and multi-wavelength anomalous dis-

persion using selenomethionine-labelled human Polg47. Structures were refined

using CNS48 and interspersed with manual model building using COOT49. All

a

d

b

e

G263VG263V

A264PA264PK261K261

L258L258

Prim

erP

rimer

R361R361

D355D355

Template

Template

P316P316

E1E11616

D13D13

HelixHelixA1A11717

V12(O)V12(O) V8(N)V8(N)

T122T122

D1D11515

D1D11515E1E11616

RR111111

R361R361

P316P316

R313R313

TemplateTemplateD13D13

RR111111HH

R361SR361S

A1A117P17P

T122PT122P

G263V, A264PG263V, A264PF290SF290S

G295RG295R

c

Figure 5 | XPV mutations. a, Arg 111 contacts DNA template and stacks withPro 316 of the LF domain. b, Mapping of eight missense XPV mutations. Theprotein is represented by the Ca trace, DNA by tube-and-ladderrepresentation, and the altered residues are shown as cyan ball-and-stickrepresentation. c, Local interactions of Ala 117 and Thr 122 and their relationto the active site. Most side chains are removed for clarity. d, Arg 361 supportsPro 316 and theb-strand that forms an extensive hydrogen-bond network withthe DNA template. e, Gly 263 and Ala 264 are adjacent to the thumb–DNAprimer interface. G263V and A264P substitutions would clash with Leu 258and Lys 261. Electron densities are shown as meshes in panels d and e.

NATURE | Vol 465 | 24 June 2010 ARTICLES

1047Macmillan Publishers Limited. All rights reserved©2010

residues are in the most favourable (98%) and allowed (1.7%) regions of theRamachandran plot except for two that are well defined by electron densities.

For functional assays, the C-terminal truncated human Polg (amino acids

1–511), which has the same TLS activity as the full-length human Polg43, was

subcloned into pET21a and readily expressed in E. coli. Q38A and R61A mutations

were made using Mutant-K (TaKaRa BIO Inc.). Steady-state kinetic assays and

primer extension reactions were carried out as described43.

Full Methods and any associated references are available in the online version ofthe paper at www.nature.com/nature.

Received 9 April; accepted 21 May 2010.

1. Brash, D. E. Sunlight and the onset of skin cancer. Trends Genet. 13, 410–414 (1997).2. Masutani, C. et al. The XPV (xeroderma pigmentosum variant) gene encodes

human DNA polymerase g. Nature 399, 700–704 (1999).3. Johnson, R. E., Kondratick, C. M., Prakash, S. & Prakash, L. hRAD30 mutations in

the variant form of xeroderma pigmentosum. Science 285, 263–265 (1999).4. Hishida, T., Kubota, Y., Carr, A. M. & Iwasaki, H. RAD6–RAD18–RAD5-pathway-

dependent tolerance to chronic low-dose ultraviolet light. Nature 457, 612–615(2009).

5. Chaney, S. G., Campbell, S. L., Bassett, E. & Wu, Y. Recognition and processing ofcisplatin- and oxaliplatin-DNA adducts. Crit. Rev. Oncol. Hematol. 53, 3–11 (2005).

6. Saribasak, H., Rajagopal, D., Maul, R. W. & Gearhart, P. J. Hijacked DNA repairproteins and unchained DNA polymerases. Phil. Trans. R. Soc. Lond. B 364,605–611 (2009).

7. Kawamoto, T. et al. Dual roles for DNA polymerase g in homologous DNArecombination and translesion DNA synthesis. Mol. Cell 20, 793–799 (2005).

8. McIlwraith, M. J. et al. Human DNA polymerase g promotes DNA synthesis fromstrand invasion intermediates of homologous recombination. Mol. Cell 20,783–792 (2005).

9. Bugreev, D. V., Hanaoka, F. & Mazin, A. V. Rad54 dissociates homologousrecombination intermediates by branch migration. Nature Struct. Mol. Biol. 14,746–753 (2007).

10. Rey, L. et al. Human DNA polymeraseg is required for common fragile site stabilityduring unperturbed DNA replication. Mol. Cell. Biol. 29, 3344–3354 (2009).

11. Broughton, B. C. et al. Molecular analysis of mutations in DNA polymerase g inxeroderma pigmentosum-variant patients. Proc. Natl Acad. Sci. USA 99, 815–820(2002).

12. Tanioka, M. et al. Molecular analysis of DNA polymerase g gene in Japanesepatients diagnosed as xeroderma pigmentosum variant type. J. Invest. Dermatol.127, 1745–1751 (2007).

13. Inui, H. et al. Xeroderma pigmentosum-variant patients from America, Europe,and Asia. J. Invest. Dermatol. 128, 2055–2068 (2008).

14. Kusumoto, R., Masutani, C., Shimmyo, S., Iwai, S. & Hanaoka, F. DNA bindingproperties of human DNA polymerase g: implications for fidelity and polymeraseswitching of translesion synthesis. Genes Cells 9, 1139–1150 (2004).

15. McCulloch, S. D. et al. Preferential cis-syn thymine dimer bypass by DNApolymerase g occurs with biased fidelity. Nature 428, 97–100 (2004).

16. Yao, J., Dixon, K. & Carty, M. P. A single (6–4) photoproduct inhibits plasmid DNAreplication in xeroderma pigmentosum variant cell extracts. Environ. Mol.Mutagen. 38, 19–29 (2001).

17. Scrima, A. et al. Structural basis of UV DNA-damage recognition by theDDB1–DDB2 complex. Cell 135, 1213–1223 (2008).

18. Sugasawa, K. XPC: its product and biological roles. Adv. Exp. Med. Biol. 637, 47–56(2008).

19. Yang, W. & Woodgate, R. What a difference a decade makes: insights intotranslesion DNA synthesis. Proc. Natl Acad. Sci. USA 104, 15591–15598 (2007).

20. Broyde, S., Wang, L., Rechkoblit, O., Geacintov, N. E. & Patel, D. J. Lesionprocessing: high-fidelity versus lesion-bypass DNA polymerases. Trends Biochem.Sci. 33, 209–219 (2008).

21. Trincao, J. et al. Structure of the catalytic core of S. cerevisiae DNA polymerase g:implications for translesion DNA synthesis. Mol. Cell 8, 417–426 (2001).

22. Alt, A. et al. Bypass of DNA lesions generated during anticancer treatment withcisplatin by DNA polymerase g. Science 318, 967–970 (2007).

23. Wang, L., Broyde, S. & Zhang, Y. Polymerase-tailored variations in the water-mediated and substrate-assisted mechanism for nucleotidyl transfer: insightsfrom a study of T7 DNA polymerase. J. Mol. Biol. 389, 787–796 (2009).

24. Ling, H., Boudsocq, F., Woodgate, R. & Yang, W. Crystal structure of a Y-familyDNA polymerase in action: a mechanism for error-prone and lesion-bypassreplication. Cell 107, 91–102 (2001).

25. Glick, E., Vigna, K. L. & Loeb, L. A. Mutations in human DNA polymerase g motif IIalter bypass of DNA lesions. EMBO J. 20, 7303–7312 (2001).

26. Li, Y. et al. Nucleotide insertion opposite a cis-syn thymine dimer by a replicative DNApolymerase from bacteriophage T7. Nature Struct. Mol. Biol. 11, 784–790 (2004).

27. Ling, H., Boudsocq, F., Plosky, B. S., Woodgate, R. & Yang, W. Replication of acis–syn thymine dimer at atomic resolution. Nature 424, 1083–1087 (2003).

28. Park, H. et al. Crystal structure of a DNA decamer containing a cis-syn thyminedimer. Proc. Natl Acad. Sci. USA 99, 15965–15970 (2002).

29. Vassylyev, D. G. et al. Atomic model of a pyrimidine dimer excision repair enzymecomplexed with a DNA substrate: structural basis for damaged DNA recognition.Cell 83, 773–782 (1995).

30. Min, J. H. & Pavletich, N. P. Recognition of DNA damage by the Rad4 nucleotideexcision repair protein. Nature 449, 570–575 (2007).

31. Mees, A. et al. Crystal structure of a photolyase bound to a CPD-like DNA lesionafter in situ repair. Science 306, 1789–1793 (2004).

32. Yang, W. Poor base stacking at DNA lesions may initiate recognition by manyrepair proteins. DNA Repair (Amst.) 5, 654–666 (2006).

33. Lone, S. et al. Human DNA polymerase k encircles DNA: implications formismatch extension and lesion bypass. Mol. Cell 25, 601–614 (2007).

34. Bauer, J. et al. A structural gap in Dpo4 supports mutagenic bypass of a majorbenzo[a]pyrene dG adduct in DNA through template misalignment. Proc. NatlAcad. Sci. USA 104, 14905–14910 (2007).

35. Jia, L., Geacintov, N. E. & Broyde, S. The N-clasp of human DNA polymerase kpromotes blockage or error-free bypass of adenine- or guanine-benzo[a]pyrenyllesions. Nucleic Acids Res. 36, 6571–6584 (2008).

36. Ling, H., Boudsocq, F., Woodgate, R. & Yang, W. Snapshots of replication throughan abasic lesion; structural basis for base substitutions and frameshifts. Mol. Cell13, 751–762 (2004).

37. Wilson, R. C. & Pata, J. D. Structural insights into the generation of single-basedeletions by the Y family DNA polymerase dbh. Mol. Cell 29, 767–779 (2008).

38. Jarosz, D. F., Godoy, V. G., Delaney, J. C., Essigmann, J. M. & Walker, G. C. A singleamino acid governs enhanced activity of DinB DNA polymerases on damagedtemplates. Nature 439, 225–228 (2006).

39. Lehmann, A. R. et al. Translesion synthesis: Y-family polymerases and thepolymerase switch. DNA Repair (Amst.) 6, 891–899 (2007).

40. Di Lucca, J. et al. Variants of the xeroderma pigmentosum variant gene (POLH)are associated with melanoma risk. Eur. J. Cancer 45, 3228–3236 (2009).

41. Bassett, E. et al. Efficiency of extension of mismatched primer termini across fromcisplatin and oxaliplatin adducts by human DNA polymerases b and g in vitro.Biochemistry 42, 14197–14206 (2003).

42. Kokoska, R. J., McCulloch, S. D. & Kunkel, T. A. The efficiency and specificity ofapurinic/apyrimidinic site bypass by human DNA polymerase g and Sulfolobussolfataricus Dpo4. J. Biol. Chem. 278, 50537–50545 (2003).

43. Kusumoto, R., Masutani, C., Iwai, S. & Hanaoka, F. Translesion synthesis byhuman DNA polymerase g across thymine glycol lesions. Biochemistry 41,6090–6099 (2002).

44. Masutani, C., Kusumoto, R., Iwai, S. & Hanaoka, F. Mechanisms of accuratetranslesion synthesis by human DNA polymeraseg. EMBO J. 19, 3100–3109 (2000).

45. Wang, F. & Yang, W. Structural insight into translesion synthesis by DNA Pol II.Cell 139, 1279–1289 (2009).

46. McCoy, A. J. Solving structures of protein complexes by molecular replacementwith Phaser. Acta Crystallogr. D 63, 32–41 (2007).

47. Hendrickson, W. A., Horton, J. R. & LeMaster, D. M. Selenomethionyl proteinsproduced for analysis by multiwavelength anomalous diffraction (MAD): a vehicle fordirect determination of three-dimensional structure. EMBO J. 9, 1665–1672 (1990).

48. Brunger, A. T. et al. Crystallography and NMR system: a new software suite formacromolecular structure determination. Acta Crystallogr. D 54, 905–921 (1998).

49. Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development ofCoot. Acta Crystallogr. D 66, 486–501 (2010).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements We thank D. Leahy, M. Gellert and R. Craigie for criticalreading of the manuscript. The research was funded by the intramural researchprogram of NIDDK, NIH, and grants from the Ministry of Education, Culture,Sports, Science, and Technology of Japan. Y.Z. is the recipient of a Chinese Ministryof Education scholarship and joint PhD student in NIH-Zhejiang UniversityGraduate Partnership Program. S.R.-M. received a fellowship from the HumanFrontiers Science Program.

Author Contributions C.B. determined the five structures; Y.Z. prepared thesamples and grew the crystals; Y.K. did the kinetic and bypass assays; S.R.-M.determined the type 1 structure; M.G. prepared the clone and type 1 crystals; J.Y.L.made mutants; C.M. designed the functional assays; A.R.L. identified theunpublished XPV mutations; F.H. conceived the project; and W.Y. supervised thestructure determination. C.B., Y.Z., F.H. and W.Y. prepared the manuscript. C.B.and Y.Z. contributed equally to the study. All authors discussed the results andcommented on the manuscript.

Author Information Atomic coordinates and structure factors for the reportedcrystal structures have been deposited with the Protein Data Bank under accessioncodes 3MR2 (Nrm), 3MR3 (TT1), 3MR4 (TT2), 3MR5 (TT3) and 3MR6 (TT4).Reprints and permissions information is available at www.nature.com/reprints.The authors declare no competing financial interests. Readers are welcome tocomment on the online version of this article at www.nature.com/nature.Correspondence and requests for materials should be addressed to W.Y.(wei.yang@nih.gov) or F.H. (fumio.hanaoka@gakushuin.ac.jp).

ARTICLES NATURE | Vol 465 | 24 June 2010

1048Macmillan Publishers Limited. All rights reserved©2010

METHODSCloning and expression. Codon-optimized human Polg (residues 1–432) was

amplified by PCR and subcloned into modified pET28a plasmid with the

PreScission protease recognition sequence as described45. E. coli BL21 (DE3)

pLysS cells transfected with this plasmid were grown at 37 uC in LB medium

containing 50mg ml21 kanamycin to an optical density at 600 nm of 0.6–0.8.

Protein expression was induced at 16 uC by addition of 0.4 mM IPTG. After over-

night growth, cells were collected by centrifugation at 5,000g for 30 min at 4 uC and

stored at 280 uC. Human Polg mutants (W339K, W339S, W339Q, W339M,

W339R, V372I, C406M, W339Q/E335Q) were made using the QuikChange kit(Stratagene). Selenomethionine-labelled (SeMet) human Polg protein was

expressed using the methionine auxotrophic E. coli strain B834 (DE3)pLysS in a

defined medium47. For the translesion synthesis assays, the C-terminal truncated

human Polg (amino acids 1–511), which shows almost the same TLS activity as

the full-length human Polg43 (Supplementary Table 3), was subcloned into the

pET21a vector. The resulting human Polg (1–511) has the LEHHHHHH tag at its

C terminus. Human Polg mutants (Q38A, R61A) were made using the mutant-K

kit (TaKaRa). E. coli BL21 (DE3) cells transfected with these plasmids were grown

at 30 uC in LB medium containing 50mg ml21 carbenicillin to an optical density at

600 nm of 0.4–0.6. Protein expression was induced at 30 uC by the addition of

0.5 mM IPTG. After 3 h induction, cells were collected by centrifugation at 5,000g

for 15 min at 4 uC and stored at 280 uC.

Protein purification and oligonucleotides. All mutant and selenomethionine-

labelled human Polg proteins (amino acids 1–432) were purified similarly by

Ni21-affinity, MonoS and Superdex75 chromatography. Human Polg (1–511

amino acids) protein was purified using Hitrap DEAE Sepharose Fast Flow (GE

Healthcare), Ni21-NTA agarose (Quiagen) column, and MonoS PC1.6/5 column

(GE Healthcare). Purified human Polg (1–511) was stored in 20 mM sodiumphosphate (pH 7.5), 10% glycerol, 10 mM b-ME, 1 mM EDTA and NaCl.

Undamaged DNA oligonucleotides were synthesized and reverse-phase HPLC

purified. Template and primer pairs at a 1:1 molar ratio were annealed in HEN

buffer (10 mM HEPES (pH 8.0), 0.1 mM EDTA and 50 mM NaCl) by heating for

5 min at 85 uC and slow cooling to 25 uC.

Crystallization and data collection. Type 1 crystals were obtained after screen-

ing of wild-type human Polg complexed with 13–16 bp duplex DNA (see

Supplementary Table 1) and dNTP in the presence of Ca21. Type 2 and type 3

crystals were obtained with wild-type as well as C406M mutant human Polgcomplexed with 8–9 bp DNA, dNMPNPP and MgCl2 (Supplementary Table 1).

The ends of the DNA substrate were optimized to improve diffraction qualities

of these crystals. For crystallization, freshly purified protein (,2 mg ml21) and

DNA were mixed at 1:1.05 molar ratio and incubated in the final buffer at room

temperature for 10 min. After threefold dilution by addition of 5 mM MgCl2, the

complex was concentrated to the protein concentration of ,7 mg ml21 by ultra-

filtration. Non-hydrolysable nucleotides were added at last to form ternary

complexes. Initial crystallization hits were found using Hampton Classics and

Index screens. All crystals were grown by the hanging-drop vapour-diffusionmethod. The optimized reservoir buffer contained 0.1 M MES (pH 6.0), 5 mM

MgCl2, 19–21% (w/v) PEG 2K-MME. After several rounds of streak and micro-

seeding (6–16% PEG 2K-MME), diffraction-quality crystals of native protein

and SeMet ternary complexes grew to maximal dimensions of ,0.04 by 0.04 by

0.2 mm3 over 3–7 days. Crystals were flash frozen in liquid nitrogen after step-

wise soaking in 10%, 20% and 30% (v/v) glycerol plus 25% (w/v) PEG 2K-MME

and other buffer components. Diffraction data were collected on beamlines

22ID, 22BM, ID23B and ID23D at the Advanced Photon Source.

Structure determination. Data were processed with HKL200050 or XDS51 and

converted to structure factors by TRUNCATE52. The type 1 crystal structure was

solved by molecular replacement (MR) in PHASER53 using yeast Polg (Protein

Data Bank 1JIH and 2R8J) as a search model. DNA and dNTP were located, but

model building was difficult and sequence assignment uncertain. The type 2

crystal structure was solved by a combined MR with the type 1 structure as a

search model and ab initio phasing by SeMet substitution and multiwavelength

anomalous dispersion (MAD). The MAD data were measured at beamline

ID23B at three wavelengths corresponding to the edge and peak of the selenium

K edge absorption profile plus a remote high-energy point. SeMet crystals

showed severe radiation damage so MAD diffractions data at each wavelength

were collected from a fresh point along a single needle-like crystal (0.02 3 0.02 3

0.08 mm3) with a 20-mm beam focus. The positions of 11 selenium atoms were

determined by SOLVE54 and experimental phases were calculated using

SHARP55. With the resulting map human Polg, DNA and dNTP were readily

traceable. Models were manually built with COOT and refined with CNS. The

refined structures (Supplementary Table 2) include 432 amino acids of human

Polg plus three N-terminal residues from the cloning vector, DNA, incoming

dNMPNPP, 2 catalytic Mg21 ions and solvent molecules. Several residues were

modelled with alternate conformations. In particular, residues 178–185 could be

traced in two conformations. In TT1 and TT3, the CPDs also appear to have a

second conformation. In TT2 and TT4, the CPDs are in a single conformation.

All structures show poorly defined loop regions for residues 130–135, 152–159,

178–185 and 408–413. Two 59 nucleotides of the template as well as the 59

nucleotide of the primer strand are disordered in some of the described struc-

tures. Ramachandran plots show 98% of residues (397 amino acids) in preferred

and 1.7% (7 amino acids) in allowed regions. Residues C16 and Y39 are outliers

(0.7%). Structure analysis was performed in COOT, Molprobity56 and PyMOL

(http://www.pymol.org). All structural figures were rendered in PyMOL and

electrostatic surface potentials were calculated using GRASP57.

Translesion synthesis assays. Primer extension reactions with CPD-containing

template (Fig. 3b) were performed as described43.

Steady-state kinetics. Steady-state kinetic parameters Km and kcat were mea-

sured using the template/primer as described in Fig. 3b, except that the primer

was two nucleotides longer at the 39 end. Reactions were performed as

described43. Kinetic constants were derived as described58.

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

51. Kabsch, W. XDS. Acta Crystallogr. D 66, 125–132 (2010).52. CCP4. The CCP4 suite: programs for protein crystallography. Acta Crystallogr. D

50, 760–763 (1994).53. McCoy, A. J. et al. Phaser crystallographic software. J. Appl. Crystallogr. 40,

658–674 (2007).54. Terwilliger, T. C. & Berendzen, J. Automated structure solution for MIR and MAD.

Acta Crystallogr. D 55, 849–861 (1999).55. Vonrhein, C., Blanc, E., Roversi, P. & Bricogne, G. Automated structure solution

with autoSHARP. Methods Mol. Biol. 364, 215–230 (2007).

56. Davis, I. W. et al. MolProbity: all-atom contacts and structure validation forproteins and nucleic acids. Nucleic Acids Res. 35, W375–W383 (2007).

57. Nicholls, A., Sharp, K. A. & Honig, B. Protein folding and association: insights fromthe interfacial and themodynamic properties of hydrocarbons. Proteins 11,281–296 (1991).

58. Creighton, S., Bloom, L. B. & Goodman, M. F. Gel fidelity assay measuringnucleotide misinsertion, exonucleolytic proofreading, and lesion bypassefficiencies. Methods Enzymol. 262, 232–256 (1995).

doi:10.1038/nature09196

Macmillan Publishers Limited. All rights reserved©2010