Architecture of the human XPC DNA repair and stemcell coactivator complexElisa T. Zhanga,b,c, Yuan Hed,1, Patricia Groba,b, Yick W. Fonga,b,2, Eva Nogalesa,b,d, and Robert Tjiana,b,c,e,3

aDepartment of Molecular and Cell Biology, University of California, Berkeley, CA 94720; bHoward Hughes Medical Institute, Department of Molecularand Cell Biology, University of California, Berkeley, CA 94720; cLi Ka Shing Center for Biomedical and Health Sciences, CIRM Center of Excellence, Universityof California Berkeley, CA 94720; dLife Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94710; and eHoward Hughes MedicalInstitute, Chevy Chase, MD 20815-6789

Contributed by Robert Tjian, October 16, 2015 (sent for review August 20, 2015; reviewed by Montserrat Samso and Ning Zheng)

The Xeroderma pigmentosum complementation group C (XPC)complex is a versatile factor involved in both nucleotide excisionrepair and transcriptional coactivation as a critical component ofthe NANOG, OCT4, and SOX2 pluripotency gene regulatory net-work. Here we present the structure of the human holo-XPC com-plex determined by single-particle electron microscopy to reveal aflexible, ear-shaped structure that undergoes localized loss of orderupon DNA binding.We also determined the structure of the completeyeast homolog Rad4 holo-complex to find a similar overall architec-ture to the human complex, consistent with their shared DNA repairfunctions. Localized differences between these structures reflect anintriguing phylogenetic divergence in transcriptional capabilities thatwe present here. Having positioned the constituent subunits by tag-ging and deletion, we propose a model of key interaction interfacesthat reveals the structural basis for this difference in functional con-servation. Together, our findings establish a framework for under-standing the structure-function relationships of the XPC complex inthe interplay between transcription and DNA repair.

transcription | stem cells | DNA repair | structure | biochemistry

Genomes of living organisms serve two primary functions: asvehicles for hereditary information and as the template for

gene products involved in an organism’s development and re-sponses to environmental stimuli. Vital to maintaining the healthof genomes in the face of intrinsic and extrinsic sources of DNAdamage are a suite of DNA repair pathways, each dedicated tohandling specific lesions. Similarly, proper use and expression ofthis essential genomic information is regulated by a host of tran-scription factors, chromatin remodelers, and epigenetic modifiersand readers (1). The Xeroderma pigmentosum complementationgroup C (XPC) protein complex performs crucial roles in bothof these capacities by participating in nucleotide excision repair(NER) (2) and base excision repair (BER) (3), as well as transcrip-tional regulation (4) and other processes (5).The XPC complex is one of seven XP complementation groups

A–G and is composed of the 125-kDa XPC, the 58-kDa RAD23B(Rad23 homolog B; also known as HHR23B), and the 18-kDaCETN2 (Centrin2) subunits (2). RAD23B and CETN2 associatetightly with XPC and stabilize both its DNA repair (6–10) andstem cell coactivator functions (4). The XPC complex is the ini-tiator and main DNA damage sensor in global genome nucleotideexcision repair (GG-NER), one of two branches of the nucleotideexcision repair pathway that repairs a wide array of bulky, helix-distorting lesions (2, 11); the second form of NER or transcrip-tion-coupled repair (TC-NER) targets lesions blocking tran-scription to reactivate proper gene expression (2, 12). Defectsin GG-NER lead to photosensitivity and a predisposition to cer-tain cancers in animal models and in human patients with Xero-derma pigmentosum (13). In conjunction with the UV-damageDNA-binding protein (UV-DDB) (2, 12), the XPC complexrecruits >30 downstream factors, such as XPA (14), TFIIH (15,16), and the endonucleases XPF and XPG to remove these ad-ducts (2, 11). In addition to its role in GG-NER, XPC is also

involved in base excision repair (BER). BER is responsible forremoving primarily non-helix-distorting lesions from the ge-nome (2). In BER, the XPC complex helps repair oxidativedamage by stimulating the activities of DNA glycosylases suchas OGG1 and TDG (3) to target lesions including 8-oxoguanine,independently of other downstream GG-NER factors (17).More recently, the XPC complex has also been found to

perform crucial duties in the regulation of gene transcription, thesecond primary function of the genome. In embryonic stem cells(ESCs), the XPC complex acts as a coactivator to enhance theexpression of OCT4- and SOX2-driven pluripotency genes, mostnotably NANOG (4), buttressing the gene regulatory networkthat establishes and maintains the unique self-renewal and plu-ripotency properties of ESCs. The XPC complex performs itscoactivator functions independently of DNA binding (4, 18),presumably by bridging interactions between the sequence-spe-cific transcription factors OCT4 (octamer-binding transcrip-tion factor 4; also known as POU5F1) and SOX2 [SRY (sex-determining region Y)-box 2] and the general transcriptionalmachinery, such as TFIID and RNA pol II, thus following amechanism reminiscent of that of other coactivator complexes

Significance

Embryonic or pluripotent stem cells are unique in their ability toself-renew in culture and to generate all lineages of an adult or-ganism, making them valuable tools for modeling early de-velopmental processes and for developing regenerative medicinetechnologies. An important factor in controlling the expression ofpluripotency genes is the Xeroderma pigmentosum complemen-tation group C (XPC) DNA repair complex. This study presents, toour knowledge, the first complete structures of different XPCcomplexes by electron microscopy to establish an importantframework for a molecular understanding of XPC’s two primaryfunctions. In conjunction with our biochemical findings, we syn-thesize a model of how XPC performs both its evolutionarilyconserved DNA repair function and its evolutionarily noncon-served transcription function.

Author contributions: E.T.Z., E.N., and R.T. designed research; E.T.Z., Y.H., and Y.W.F.performed research; E.T.Z. and Y.W.F. contributed new reagents/analytic tools; E.T.Z.,Y.H., P.G., and Y.W.F. analyzed data; and E.T.Z., E.N., and R.T. wrote the paper.

Reviewers: M.S., Virginia Commonwealth University; and N.Z., University of Washington.

The authors declare no conflict of interest.

Data deposition: Atomic coordinates and structure factors have been deposited in theElectron Microscopy Data Bank (EMDB), https://www.ebi.ac.uk/pdbe/emdb/ (accessionnos. EMD-6495–EMD-6498).1Present address: Department of Molecular Biosciences, Northwestern University, Evanston,IL 60208.

2Present address: Brigham Regenerative Medicine Center, Cardiovascular Division, De-partment of Medicine Brigham and Women’s Hospital, Harvard Medical School, Cam-bridge, MA 02139.

3To whom correspondence should be addressed. Email: jmlim@berkeley.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1520104112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1520104112 PNAS | December 1, 2015 | vol. 112 | no. 48 | 14817–14822

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

020

such as Mediator and p300/CBP (19). In a separate study, XPCwas found to localize to active but not to inactive RNA pol II-dependent gene promoters in the absence of exogenous geno-toxic stress (11, 20).Although both the DNA repair and transcriptional functions

of mammalian XPC complexes have been characterized bio-chemically and genetically, 3D structural information of theholo-complex has been unavailable. The partial crystal structureof Rad4, the yeast homolog of the human XPC subunit, incomplex with the Rad4-interaction domain of yeast Rad23, hasprovided information helpful toward an understanding of Rad4/XPC’s biochemical behavior and of certain phenotypic outcomes(18, 21). However, given XPC’s low sequence homology with theyeast homolog Rad4 (21, 22), the absence of key domains in theavailable crystal structure, and the divergence of requirements fortranscriptional vs. repair activities (4, 15, 18, 23), structural in-formation of the complete, three-subunit, native human XPCcomplex is much needed for an understanding of the functionalversatility of the XPC complex. At present, no information hasbeen reported on the overall architecture of the XPC holo-complex, the possible large-scale conformational rearrangementsin XPC upon DNA-binding, or the extent of structural conser-vation of the human XPC complex with homolog complexes.Here we address the lack of structural data on human XPC

using 3D single-particle reconstruction by electron microscopy(EM) to characterize the overall architecture of XPC, as well asgenetic tagging and computational docking to locate the relativepositions of its constituent subunits. We also assess the confor-mational changes to the complex upon binding to DNA. Giventhe evolutionary conservation of GG-NER (24), we queried theextent of structural and functional conservation over evolution-ary time by solving the structure of the complete yeast homologRad4 complex and testing whether the OCT4/SOX2 transcrip-tional coactivation function is supported by the yeast complex.Together with existing biochemical data (14–16, 23, 25), wesought to identify the approximate regions of contact betweenthe XPC complex and its partner proteins OCT4, SOX2, XPA,and TFIIH.

ResultsReconstitution of the Human XPC Complex. We purified the com-plete, three-subunit human XPC-RAD23B-CETN2 complex (Fig.1A) expressed in Sf9 insect cells using a two-step affinity purifi-cation procedure (Fig. S1 A and B). SDS/PAGE analysis indicatedthat the purified complex was nearly homogeneous and stoichio-metric (Fig. S1B, Left). This result is consistent with previous datashowing that XPC and RAD23B interact in a 1:1 ratio (26) andthat XPC and CETN2 also interact in a 1:1 ratio (27, 28). Fur-thermore, a 1:1:1 stoichiometry is consistent with the size of the∼200-kDa product we observe for the cross-linked complex (Fig.S1B, Right).Initial attempts at negative stain EM data collection were

hampered by the extremely heterogeneous appearance of theparticles in both size and shape (data not shown). These resultssuggested that the complex was either unstable during EM samplepreparation and/or suffered from extreme conformational flexi-bility. To overcome these limitations, we optimized cross-linkingconditions across temperature, incubation time, and cross-linkerconcentration to identify the minimum requirements for achievingcomplete subunit incorporation as detected by Coomassie staining(for more details, see SI Materials and Methods, Expression andPurification of XPC/Rad4 Complexes). These particular conditionswere then used for subsequent negative stain sample preparation,followed by data collection and single particle analysis (Fig. S1B,Right and Fig. S2A).Two-dimensional reference-free class averages (Fig. S2B) show

C-shaped views, multilobed structures, and some very small,compact, globular shapes, with most of the particles having an

elongated appearance. Such elongated shapes would be consistentwith our observation that the XPC complex runs as a relativelybroad peak centered at 275 kDa on a size-exclusion column,slightly larger than its mass of ∼200 kDa (Fig. 1A and Fig. S1 Band C).

Ab Initio 3D Reconstruction of the Human XPC Complex by RandomConical Tilt and Subunit Localization. We used random conical tilt(29) (Fig. S1 D–G) to generate an ab initio 3D reconstruction ofthe human XPC complex (Fig. 1B and Fig. S2). Handedness androbustness of our EM reconstruction was supported by the resultsof the freehand test (30) using projection-matching of particlepairs at 0° and 30° tilt, which indicated that 40% of particles fallwithin 26 degrees of the expected tilt angle (Fig. S3D). The struc-ture is ∼170 Å by ∼100 Å by ∼70 Å and roughly resembles theshape of a human ear (Fig. 1B). To localize the position of theCETN2 subunit within the reconstruction, we followed two par-allel strategies: visualization of a complex that included a maltose-binding protein (MBP) tag at the N terminus of CETN2, andvisualization of a complex lacking the CETN2 subunit (Fig. 1 CandD and Fig. S2 F andG). As seen in the 3D difference maps, theMBP density is localized primarily outside the shorter end or ear-lobe of the XPC complex, whereas the CETN2 density is localizedprimarily inside the earlobe (Fig. 1 C and D). Consistent with thislocalization of CETN2, the crystal structure of Rad4/Rad23Rad4-BD

180° UbL

180°

C

D

180°

E

180°

170

Å

100 Å

B

XPC

R C

MBP

XPC

R C

RAD23B 419

UbL UBA1 UBA2 XPC-BD

1

172 EF EF

1 CETN2

RAD23B-BD1

940 1 TGD TGD BH1 BH2 BH3

DNA-BD

XPC OCT4-BD SOX2-BD

CETN2-BD TFIIH-BD2

TFIIH-BD1

XPA- & OGG1-BD 154

144

331 511 808

814

124

204 223 RAD23B-BD2

607 742

734 496

276 334

XPC-BD 105 166

847 863

UBA2 UBA1

A

Fig. 1. 3D reconstruction of the human XPC complex and localization ofsubunits. (A) Schematic representation of the subunits and domains of thehuman XPC complex. Transglutaminase homology domain (TGD), β-hairpindomains 1–3 (BH), ubiquitin-like domain (UbL), ubiquitin-associated domains 1and 2 (UBA), EF-hand domains (EF), and protein- and DNA-binding domains(BD) are indicated accordingly. (B) Front and back views of the XPC complexreconstructed in EMAN2 (49). Estimated dimensions are indicated. (C) Positive(yellow) and negative (purple) 3D difference densities at 4σ between thecomplex containing MBP-CETN2 and untagged complex. (D) Positive (pink) 3Ddifference density at 5σ between the full complex and the XPC-RAD23B sub-complex, indicating the likely position for the CETN2 subunit. No negativedifference density was observed at this threshold. (E) Docking of the yeastRad4/Rad23 [Protein Data Bank (PDB) ID 2QSF; cyan/green] into the modelwith the human CETN2 and XPC interaction peptide (PDB ID 2GGM; pink/cyan)by Situs (31) in a manner consistent with the difference density data in A andB. Shown are predicted approximate positions of the UbL, UBA1, and UBA2domains of RAD23B based on positional information of the Rad23Rad4-BD (darkgreen) N and C termini in the crystal structure.

14818 | www.pnas.org/cgi/doi/10.1073/pnas.1520104112 Zhang et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

020

docked (31) into the upper end of the “ear” in an orientation suchthat the C terminus of Rad4 points downward toward the “earlobe.”In further agreement with this subunit organization, docking ofthe CETN2 was placed in the earlobe by objective, automaticdocking using Situs (31) (Fig. 1E). Based on the orientation ofthe Rad23 termini observed in the Rad4/Rad23 crystal structure,we also indicate the predicted, approximate locations of theUbL, UBA1, and UBA2 subunits of RAD23B that are not ob-served in the Rad4/Rad23 crystal structure (Fig. 1E).To ensure that the cross-linking and the use of stain did not

significantly compromise the integrity of the structure obtained,we also analyzed the native, uncross-linked complex and usedcryo-electron microscopy (cryo-EM). The native complex inphosphotungstate stain and the cross-linked complex under cryoconditions were all consistent with the cross-linked complexunder negative stain conditions (Fig. 1B and Fig. S3 A–C). Al-though the native complex appears to have better definitionbetween the earlobe and the rest of the structure, this differencedensity is only present at 2σ and not at the statistically significant3σ (Fig. S3C).The low resolution of our reconstruction (∼25 Å) suggested

that the XPC complex may be flexible and adopt multiple con-formations. To gain an understanding of the possible range ofconformational states, we used a large data set of ∼210,000particles and performed 3D sorting and classification usingRELION (32) to produce three distinct structures representingthe range of XPC conformations (Fig. 2A), followed by Situsdocking using available crystal structures. The XPC complex ap-pears to be partitioned between a more elongated (Fig. 2A, Top)and more compact states (Fig. 2A, Middle and Bottom). Three-dimensional reprojections of these three models match reference-free 2D class averages (Fig. 2B).

Structural Changes Following DNA Binding. To visualize possiblestructural changes in the human XPC complex upon binding toDNA, we used a monomeric avidin and biotinylated-DNA af-finity purification strategy to isolate only DNA-bound XPC mol-ecules (Fig. S4 A and B). The 48-bp DNA bubble mismatch duplexwas chosen from a validated EMSA probe (22) that demonstrated

one of the strongest affinities for the XPC complex. Comparisonof the reconstructions obtained from the apo vs. DNA-boundsamples indicates that the addition of DNA primarily affectedthe density region immediately above the “earlobe” (Fig. 3A).Addition of DNA to XPC did not appear to lock the structureinto a single conformation because the 2D class averages (Fig.3B) and the resolution of ∼24 Å are similar to those we observedfor the apo complex. The changes observed are consistent with apossible movement of the BH domains of XPC toward the regionof DNA-binding. This assessment is based on comparisons betweenthe changes imposed by DNA binding in the Rad4 crystal structureand the position of the crystal structure docked into the EM densityeither as the intact Rad4/Rad23 crystal structure (Fig. 3A) or withthe C-terminal portion of the Rad4 TGD domain considered sep-arately from the N-terminal portion to better reflect structuralhomology with the human XPC, which contains an insertion in itsTGD domain (Figs. S4C and S5 A and B). A second interestingpossibility is that certain regions of RAD23B become disorderedupon DNA-binding, thus resulting in the observed loss of density;this is consistent with the finding that some regions of Rad23 thatare ordered in the apo Rad4/Rad23 crystal structure becomingdisordered upon binding DNA (21).The XPC complex has been demonstrated to bind other sub-

strates as well, such as single-stranded DNA (33). Therefore, wealso prepared ssDNA-bound XPC molecules using the samestrategy of biotinylated-DNA pull-down (Fig. S4 A and B). Three-dimensional difference density analysis indicates that this structureis nearly indistinguishable from the mismatch-DNA-bound XPCcomplex (Fig. S4D).

Conservation of Structure and Function. The human XPC complex(XPC, RAD23B, CETN2) and the yeast homolog Rad4 complex(Rad4, Rad23, Rad33; ref. 34) appear to function equivalently innucleotide excision repair, given their similar binding propertiesto damaged DNA (22). These two complexes are also thought tobe structurally similar based on strong sequence homology be-tween human RAD23B and yeast Rad23 (6) and between humanCETN2 and the purported yeast CETN2 homolog Rad33 (Fig.S5C). Despite low sequence conservation overall between XPCand Rad4, these two proteins share sequence homology of keydomains (35) and very similar predicted secondary structures

180°

180°

180°

A

2D

3D

2D

3D

2D

3D

B

Fig. 2. The XPC complex adopts highly flexible conformations. (A) Threemodelsof the XPC complex generated in RELION with Situs-based docking (31) of theyeast Rad4/Rad23 (PDB ID 2QSF; cyan/green) and the human CETN2/XPCCETN2-BD

(PDB ID 2GGM; pink/cyan) crystal structures. (B) Comparison of 3D reprojectionsof the models to their corresponding reference-free 2D class averages.

90°

A

2D

3D

90°

DNA-bound apo

[apo] - [bound]DNA-bound apo

Rad23 (apo)Rad23 (bound)

Docked structures:

[bound] - [apo]

EM Densities: B

Fig. 3. DNA binding by XPC is accompanied by large but localized confor-mational changes distal from the sites of presumed DNA contact. (A) Front,side, and back views of the XPC complex bound to a mismatch bubble sub-strate generated in RELION (32) (yellow) shown with the apo structure (meshgray), with the 3D [apo] − [DNA-bound] difference density at 3σ (cyan), andwith the [DNA-bound] − [apo] difference density (brown). Also shown is theSitus-based docking (31) of the yeast Rad4/Rad23 apo (PDB ID 2QSF; purple/orange) structure with the DNA-bound (PDB ID 2QSH; green/yellow) structurealigned to the apo via the Rad23 subunit. (B) Comparison of 3D reprojectionsof the model to their corresponding reference-free 2D class averages.

Zhang et al. PNAS | December 1, 2015 | vol. 112 | no. 48 | 14819

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

020

(Fig. S5 A and B). To examine this question of structural con-servation, we obtained a 3D reconstruction at ∼23 Å resolutionof the complete yeast Rad4 complex (Fig. 4 A and B). Theoverall architecture of the complexes from the two species isremarkably similar at both the 3D and 2D levels (Fig. 4 B–D,Right vs. Fig. 1B and Fig. S2 B and E), although there are smallareas of difference between the human and yeast complexes, asseen in the 3D difference maps (Fig. 4C). We posit that theregion of difference density in the earlobe may be due tostructural differences between the CETN2 and Rad33 homologs(Fig. 1E and Fig. S5C).The structural similarity between the human and yeast XPC/

Rad4 complexes suggested that other functions of the XPCcomplex, in particular its transcriptional roles in ES cells, mightalso be conserved. To our surprise, we observed that unlike thehuman and mouse XPC complexes, the Rad4 complex exhibitedno coactivator activity in our in vitro transcription assay (Fig. 4E)and was completely incapable of forming a stable interactionwith SOX2, the primary requisite activator for in vitro activationof NANOG gene transcription (4) (Fig. 4F). Using informationon XPC’s interaction domains with partner proteins (14, 15, 25,27, 36), sequence homology between yeast Rad4 and humanXPC (Fig. S5), as well as the docking of the Rad4/Rad23 crystalstructure, we were able to generate a model indicating the pre-dicted locations of the interaction domains on the XPC complex(Fig. 5A). These interaction domains are clustered on the top ofthe ear. Intriguingly, when we superimpose the regions of differ-ence density between the human and yeast EM maps with thepredicted interaction domains, we note that one such region islocated near the OCT4- and SOX2-binding interfaces but not the

DNA-binding residues in regions DNA-BD3-5 (Fig. 5B), sug-gesting that these regions of difference may underlie the phylo-genetic divergence in transcriptional activity between the humanand yeast complexes.

DiscussionOur single particle analyses reveal an ear-like shape for the humanXPC complex and indicate the existence of a range of conforma-tional states for this DNA repair and stem cell coactivator complex(Figs. 1 and 2). We show that the yeast homolog Rad4 holo-com-plex has a similar overall architecture but small regions of differencecompared with the human XPC complex that may reflect theirfunctional nonequivalence in biochemical assays (Fig. 4). Usinglabeling, mutational, and docking strategies, we localize the indi-vidual subunits of the complex within the structure (Fig. 1). Thebinding to the two distinct DNA substrates used in this study re-sulted in a similar overall conformational change in the left domainimmediately above the earlobe, which is consistent with previousobservations in the Rad4/Rad23 crystal structures (21) (Fig. 3).Our study reveals the apparent flexibility of the XPC complex,

in large part mirroring its functional versatility. The flexibility ofthe complex may stem, at least in part, from RAD23B, becausecertain regions of Rad23 were found to be disordered in the Rad4/Rad23 crystal structure (21). Part of the conformational hetero-geneity seen in our EM structures may be due to variations in theinteraction between the UbL and the equivalent UBA1 andUBA2 domains of RAD23B (Fig. 1A). Some of the conformational

IP: HA-SOX2 IP: HA-RFP

IB: His XPC/Rad4

IB: FLAG RAD23(B)

IB: HA

XPC Rad4 XPC Rad4IN IP IN IP IN IP IN IP

A

E F

180°

D 2D3D

180°

4xNanogCAT( )-140 O SO S

EE yeast mouse human

-

Rad4

Rad23

Rad33

+glutaraldehyde

-

250 150

100

75

50

25 20 15

37

10

B

C

Fig. 4. Comparative studies of the human and yeast XPC/Rad4 complexes re-veal divergence in function but not structure. (A) Purification and cross-linkingof the homologous yeast Rad4-Rad23-Rad33 complex. (B) Three-dimensionalmodel of the yeast Rad4 complex as solved by EMAN2 (49). (C) Three-dimensional difference density at 3σ between the human (mesh pink) and theyeast (solid gray) complexes. Positive difference density, or [human]-[yeast], isshown in cyan; negative difference density or [yeast] − [human] is shown inpurple. (D) Comparisons of 3D reprojections of the yeast Rad4 complex with 2Dclass averages. (E) Titrations over a fourfold concentration range of yeast,mouse, and human XPC homolog complexes in in vitro transcription reactions ofa NANOG promoter template engineered with four extra copies of the oct-soxcomposite binding element (bottom), performed in the presence of OCT4 andSOX2 protein. Transcripts are indicated with arrowheads. (F) Coimmunopreci-pitation of human and yeast XPC/Rad4 complexes with HA-tagged SOX2 or RFP.

A

180°

OCT4

XPASOX2

XPAPPOGG1

TFIIH

BD3 BD4

BD5

B

[human] – [yeast] apo

OCT4-BD (by homology) XPA- and OGG1-BD (by homology)

SOX2-BD (by homology)DNA-binding residues

EM densities: Docked crystal structures:

Fig. 5. Model of key interaction interfaces on the XPC complex. (A) Model ofpredicted interaction surfaces based on docking and sequence homology withthe yeast Rad4/Rad23 crystal structure (PDB ID 2QSF). Indicated are residuesinvolved in binding to XPA (14) (blue; yeast residues 101–296), OCT4 (25) (red;yeast residues 298–392), SOX2 (25) (orange; yeast residues 392–609), and TFIIH(15, 16, 44) (violet; human residues 847–863 and yeast residues 76–115 and610–631). (B) Top back view of the XPC complex showing an area of positivedifference density between the human and yeast structures (yellow) that co-incides with the predicted OCT4- (red) and SOX2-binding domains (orange)but not the DNA-binding residues (dark blue residues and circles; BD3-5). TheXPA binding domain has been omitted in the enlarged view for clarity.

14820 | www.pnas.org/cgi/doi/10.1073/pnas.1520104112 Zhang et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

020

variability may also originate from the XPC subunit. Structureprediction analysis (14), limited proteolysis (14), and NMR (37)identified several highly disordered regions: the N terminus (resi-dues 1–154), the C terminus (residues 816–940), and a loop in-serted into the TGD domain comprising residues 331–517. Finally,CETN2 may also contribute to this overall flexibility, because it canadopt different conformations depending on its metal-binding state(38); however, the resolution of the XPC-RAD23B subcomplexwas not markedly improved, suggesting that this contribution tocomplex flexibility is minor, as would be expected for its relativelysmall mass contribution to the complex. The recently describedrequirement of RNA for the XPC complex to interact with itstranscription partner SOX2 (25) invokes the idea of low-complexitydomains or regions, perhaps interspersed throughout XPC, linkingtheir inherent flexibility to a critical aspect of the XPC complex’sfunction. The possibility that the mammalian-specific insertionwithin the TGD domain (residues 331–517; Fig. 1A and Fig. S5B)could be partly responsible for some of the observed structuralheterogeneity is particularly interesting.With the use of an MBP-tag as a labeling strategy, as well as

the exclusion of CETN2 from the complex, we were able to lo-calize CETN2 to the earlobe of the structure (Fig. 1 C and D).Extending these findings, we used rigid-body docking in an un-biased manner to place the Rad4/Rad23 crystal structure at thetop of the ear and the CETN2 crystal structure in the earlobe(Fig. 1E). Attempts to tag the XPC and RAD23B subunits werenot successful for a variety of reasons. The absence of CETN2does not appear to impose large conformational rearrangements,as seen by the comparison between the full complex and theXPC-RAD23B subcomplex (Fig. 1 B and D). This observation isconsistent with the lesser functional consequence of removingCETN2 than that of removing RAD23B in transcriptional coac-tivation assays, as well as with the inconsequential removal of theC-terminal CETN2-interaction domain on XPC (residues 814–940) (4).The similarity between the dsDNA- and ssDNA-bound struc-

tures is consistent with the fact that the XPC complex is capable ofbinding a large suite of different DNA structures, including UV-induced thymine dimers (2), mismatch bubbles (22), ssDNA–dsDNA junctions (39), apurinic/apyrimidic (AP) sites (40), andeven undamaged duplex and certain single-stranded DNA substrates(33). This similarity between different DNA-bound structures is alsoconsistent with recent work describing a kinetic but not structuralmeans of discrimination between damaged and undamaged DNA byRad4 (41). The reduction of density in our DNA-bound structurescompared with the apo XPC complex reflects a conformationalchange consistent with two phenomena observed in the Rad4crystal structure: (i) the C-terminal portion of Rad4 shiftingtoward the DNA substrate, and in the context of our docking,away from the region of reduced density (Fig. 3A and Fig. S4 Cand E); (ii) portions of Rad23 originally contributing to ordereddensity in the apo crystal structure becoming disordered uponbinding to DNA (21). We are inclined to favor this latter ob-servation to explain the DNA-induced changes to the XPC com-plex because the primary EM density difference we observecannot easily be accounted for by the modest, ∼13- to 14-Å shiftin Rad4 observed in the crystal structure (Fig. S5E). Therefore,we propose that DNA binding by the XPC complex inducesspecific conformational changes and disorder of certain domains,possibly such as the UbL domain (Fig. 1E).The overall similarities between the human XPC structure and

the yeast Rad4 structure, despite their divergent amino acid se-quences, provide additional, indirect validation of the accuraciesof the 3D models. Indeed, it seems quite remarkable that withoutrequiring major changes to the overall evolutionarily conserved3D shape and structure of the mammalian XPC complex, it hasnevertheless adopted entirely new transcriptional coactivator func-tions in the context of ES cell regulatory pathways that are not

relevant in yeast. The extent of structural conservation is consistentwith their equivalence in repair (22). Regions displaying differ-ences may reflect the divergence in functional capabilities that weobserve in a transcriptional context (Fig. 4). Indeed, one of theseregions at the top portion of the ear corresponds to residues ho-mologous to those shown to interact with OCT4 and SOX2 (Fig.5B, yellow). Importantly, this region is well separated from DNA-binding residues (Fig. 5B, circled BD3-5; key residues in darkblue). The only close-by DNA-binding residues (BD3) are in factnot conserved and are not found in the human sequence (Fig.S5B). This structural separation-of-function suggests that thelocal degree of structural conservation, even at modest resolu-tion, is predictive of functional convergence or divergence. Froman evolutionary point of view, a conserved process, such asnucleotide excision repair, would be expected to exhibit func-tional conservation between the yeast and human XPC homologs,whereas a nonconserved process, such as regulating genesexpressed in mammalian embryonic stem cells, would not. Indeed,a number of XPC mutations that have differential effects in DNArepair vs. transcription support this idea. For instance, althoughdeletion of the N-terminal UbL domain of yRad23 (42) and theW690S mutation of XPC (18, 43) have adverse consequences onnucleotide excision repair capabilities, respective mutations inXPC did not affect the ability to coactivate transcription (4) (Y.W.F., unpublished). Similarly, although the N and C termini of XPCare critical for recruitment and stimulation of TFIIH at sites ofdamage for global nucleotide excision repair (15, 16, 44, 45), theremoval of the N- and C-terminal TFIIH-binding domains of XPC(residues 1–195 and 814–940, respectively) only impacts repair butnot transcriptional activity (4, 25).Reflecting the ever-expanding repertoire of reported XPC

roles is the number of known physical and functional interactorsof the XPC complex, e.g., TFIIH (15, 16), OGG1 (23), TDG (3),SOX2 (4, 25) (Fig. 4), and OCT4 (4, 25), among others (46). It ispossible that the XPC complex serves as a coactivator not just forOCT4 and SOX2, especially given that its transcriptional activ-ities do not appear to be cell-type-restrained (20); therefore, thelist of XPC’s functional and physical partners is likely to grow.Although the residues on XPC through which some of theseknown interactions occur have been mapped, there is a degree ofoverlap between some of these regions, suggesting the need formore fine-tuned characterization and structural elucidation inthe future (Fig. 1A). Our recent work describing the involve-ment of RNA in mediating the XPC-SOX2 interaction adds anadditional and potentially intriguing dimension to future struc-tural studies in this regard (25). Additionally, it would be in-teresting to explore whether structural changes are imposed onthe XPC complex upon binding to its partner proteins; furtherbiochemical and structural work to assemble such larger pro-tein assemblies is required. In summary, this work provides astructural framework for integrating biochemical and struc-tural information into a mechanistic understanding of the XPCcomplex’s undoubtedly complicated roles in DNA repair and tran-scriptional regulation.

Materials and MethodsDetailed methods can be found in SI Materials and Methods. XPC/Rad4complexes were affinity purified from Sf9 cells. DNA-bound XPC sampleswere affinity purified using biotinylated DNA substrates. EM was performedusing continuous carbon films and uranyl formate or phosphotungstatestain for negative stain. Leginon software (47) was used to collect images ina Tecnai F20 microscope equipped with a Gatan UltraScan 4000 camera.Data processing was performed primarily in the Appion pipeline (48). Three-dimensional reconstructions were performed using EMAN2 (49) and RELION(32). Coimmunoprecipitation assays were performed in HEK293T cells. Invitro transcription assays were performed essentially as described (50).

ACKNOWLEDGMENTS. We thank G. Kemalyan, R. Louder, S. Howes, andD. W. Taylor for microscope and data processing guidance; T. Houweling for

Zhang et al. PNAS | December 1, 2015 | vol. 112 | no. 48 | 14821

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

020

computer support; G. Dailey for help with expression constructs; and S. Zhengand C. Inouye for help with in vitro transcription components. We aregrateful to M. Iadanza and T. Gonen for help with initial negative stainanalysis. We thank R. Lesch and R. Schekman for yeast cells for cloning theyeast Rad4 complex. We thank C. Cattoglio, J. J. Ho, G. E. Katibah, D. C. Rio,A. Martin, and D. E. Wemmer for valuable discussion. This work was sup-ported by the California Institute for Regenerative Medicine (CIRM) Research

Grant RB4-06016 (to R.T.) and by the National Institute of General MedicalSciences (GM63072; to E.N.). E.T.Z. was a National Science Foundation Grad-uate Research Fellow and a University of California Berkeley DistinguishedFellow. Y.W.F. was a CIRM Scholar (Training Grant T1-00007). R.T. and E.N.are Howard Hughes Medical Institute (HHMI) Investigators. R.T. is the Pres-ident of HHMI and the Director of the Li Ka Shing Center for Biomedical andHealth Sciences.

1. Weake VM, Workman JL (2010) Inducible gene expression: Diverse regulatorymechanisms. Nat Rev Genet 11(6):426–437.

2. Marteijn JA, Lans H, Vermeulen W, Hoeijmakers JHJ (2014) Understanding nucleotideexcision repair and its roles in cancer and ageing. Nat Rev Mol Cell Biol 15(7):465–481.

3. Shimizu Y, Iwai S, Hanaoka F, Sugasawa K (2003) Xeroderma pigmentosum group Cprotein interacts physically and functionally with thymine DNA glycosylase. EMBO J22(1):164–173.

4. Fong YW, et al. (2011) A DNA repair complex functions as an Oct4/Sox2 coactivator inembryonic stem cells. Cell 147(1):120–131.

5. Krzeszinski JY, et al. (2014) XPC promotes MDM2-mediated degradation of the p53tumor suppressor. Mol Biol Cell 25(2):213–221.

6. Sugasawa K, et al. (1996) HHR23B, a human Rad23 homolog, stimulates XPC proteinin nucleotide excision repair in vitro. Mol Cell Biol 16(9):4852–4861.

7. Nishi R, et al. (2005) Centrin 2 stimulates nucleotide excision repair by interacting withxeroderma pigmentosum group C protein. Mol Cell Biol 25(13):5664–5674.

8. Xie Z, Liu S, Zhang Y, Wang Z (2004) Roles of Rad23 protein in yeast nucleotide ex-cision repair. Nucleic Acids Res 32(20):5981–5990.

9. Ortolan TG, Chen L, Tongaonkar P, Madura K (2004) Rad23 stabilizes Rad4 fromdegradation by the Ub/proteasome pathway. Nucleic Acids Res 32(22):6490–6500.

10. Araki M, et al. (2001) Centrosome protein centrin 2/caltractin 1 is part of the xero-derma pigmentosum group C complex that initiates global genome nucleotide ex-cision repair. J Biol Chem 276(22):18665–18672.

11. Le May N, Egly JM, Coin F (2010) True lies: The double life of the nucleotide excisionrepair factors in transcription and DNA repair. J Nucleic Acids 2010(9):1–10.

12. Petruseva IO, Evdokimov AN, Lavrik OI (2014) Molecular mechanism of global ge-nome nucleotide excision repair. Acta Naturae 6(1):23–34.

13. Melis JPM, Luijten M, Mullenders LHF, van Steeg H (2011) The role of XPC: Implica-tions in cancer and oxidative DNA damage. Mutat Res 728(3):107–117.

14. Bunick CG, Miller MR, Fuller BE, Fanning E, Chazin WJ (2006) Biochemical and struc-tural domain analysis of xeroderma pigmentosum complementation group C protein.Biochemistry 45(50):14965–14979.

15. Uchida A, et al. (2002) The carboxy-terminal domain of the XPC protein plays a crucialrole in nucleotide excision repair through interactions with transcription factor IIH.DNA Repair (Amst) 1(6):449–461.

16. Yokoi M, et al. (2000) The xeroderma pigmentosum group C protein complex XPC-HR23B plays an important role in the recruitment of transcription factor IIH todamaged DNA. J Biol Chem 275(13):9870–9875.

17. Menoni H, Hoeijmakers JHJ, Vermeulen W (2012) Nucleotide excision repair-initiatingproteins bind to oxidative DNA lesions in vivo. J Cell Biol 199(7):1037–1046.

18. Yasuda G, et al. (2007) In vivo destabilization and functional defects of the xerodermapigmentosum C protein caused by a pathogenic missense mutation. Mol Cell Biol27(19):6606–6614.

19. Näär AM, Lemon BD, Tjian R (2001) Transcriptional coactivator complexes. Annu RevBiochem 70(1):475–501.

20. Le May N, et al. (2010) NER factors are recruited to active promoters and facilitatechromatin modification for transcription in the absence of exogenous genotoxic at-tack. Mol Cell 38(1):54–66.

21. Min J-H, Pavletich NP (2007) Recognition of DNA damage by the Rad4 nucleotideexcision repair protein. Nature 449(7162):570–575.

22. Krasikova YS, et al. (2013) Human and yeast DNA damage recognition complexes bindwith high affinity DNA structures mimicking in size transcription bubble. J MolRecognit 26(12):653–661.

23. Bernardes de Jesus BM, Bjørås M, Coin F, Egly JM (2008) Dissection of the moleculardefects caused by pathogenic mutations in the DNA repair factor XPC. Mol Cell Biol28(23):7225–7235.

24. Shuck SC, Short EA, Turchi JJ (2008) Eukaryotic nucleotide excision repair: From un-derstanding mechanisms to influencing biology. Cell Res 18(1):64–72.

25. Cattoglio C, et al. (2015) Functional and mechanistic studies of XPC DNA-repaircomplex as transcriptional coactivator in embryonic stem cells. Proc Natl Acad Sci USA112(18):E2317–E2326.

26. Masutani C, et al. (1994) Purification and cloning of a nucleotide excision repaircomplex involving the xeroderma pigmentosum group C protein and a human ho-mologue of yeast RAD23. EMBO J 13(8):1831–1843.

27. Popescu A, et al. (2003) Xeroderma pigmentosum group C protein possesses a high af-finity binding site to human centrin 2 and calmodulin. J Biol Chem 278(41):40252–40261.

28. Thompson JR, Ryan ZC, Salisbury JL, Kumar R (2006) The structure of the humancentrin 2-xeroderma pigmentosum group C protein complex. J Biol Chem 281(27):18746–18752.

29. Radermacher M, Wagenknecht T, Verschoor A, Frank J (1987) Three-dimensional re-construction from a single-exposure, random conical tilt series applied to the 50S ri-bosomal subunit of Escherichia coli. J Microsc 146(Pt 2):113–136.

30. Rosenthal PB, Henderson R (2003) Optimal determination of particle orientation,absolute hand, and contrast loss in single-particle electron cryomicroscopy. J Mol Biol333(4):721–745.

31. Wriggers W (2010) Using Situs for the integration of multi-resolution structures.Biophys Rev 2(1):21–27.

32. Scheres SHW (2012) RELION: Implementation of a Bayesian approach to cryo-EMstructure determination. J Struct Biol 180(3):519–530.

33. Trego KS, Turchi JJ (2006) Pre-steady-state binding of damaged DNA by XPC-hHR23Breveals a kinetic mechanism for damage discrimination. Biochemistry 45(6):1961–1969.

34. Dulk den B, van Eijk P, de Ruijter M, Brandsma JA, Brouwer J (2008) The NER proteinRad33 shows functional homology to human Centrin2 and is involved in modificationof Rad4. DNA Repair 7(6):858–868.

35. Legerski R, Peterson C (1992) Expression cloning of a human DNA repair gene in-volved in xeroderma pigmentosum group C. Nature 360(6404):610–610.

36. Li L, Lu X, Peterson C, Legerski R (1997) XPC interacts with both HHR23B and HHR23Ain vivo. Mutat Res 383(3):197–203.

37. Miron S, Duchambon P, Blouquit Y, Durand D, Craescu CT (2008) The carboxy-terminaldomain of xeroderma pigmentosum complementation group C protein, involved inTFIIH and centrin binding, is highly disordered. Biochemistry 47(5):1403–1413.

38. Craig TA, et al. (2006) Metal-binding properties of human centrin-2 determined bymicro-electrospray ionization mass spectrometry and UV spectroscopy. J Am Soc MassSpectrom 17(8):1158–1171.

39. Sugasawa K, Shimizu Y, Iwai S, Hanaoka F (2002) A molecular mechanism for DNAdamage recognition by the xeroderma pigmentosum group C protein complex. DNARepair (Amst) 1(1):95–107.

40. Pestryakov P, et al. (2012) Effect of the multifunctional proteins RPA, YB-1, and XPCrepair factor on AP site cleavage by DNA glycosylase NEIL1. J Mol Recognit 25(4):224–233.

41. Chen X, et al. (2015) Kinetic gating mechanism of DNA damage recognition by Rad4/XPC. Nat Commun 6:5849.

42. Ortolan TG, et al. (2000) The DNA repair protein rad23 is a negative regulator ofmulti-ubiquitin chain assembly. Nat Cell Biol 2(9):601–608.

43. Maillard O, Solyom S, Naegeli H (2007) An aromatic sensor with aversion to damagedstrands confers versatility to DNA repair. PLoS Biol 5(4):e79.

44. Lafrance-Vanasse J, Arseneault G, Cappadocia L, Legault P, Omichinski JG (2013)Structural and functional evidence that Rad4 competes with Rad2 for binding to theTfb1 subunit of TFIIH in NER. Nucleic Acids Res 41(4):2736–2745.

45. Ziani S, et al. (2014) Sequential and ordered assembly of a large DNA repair complexon undamaged chromatin. J Cell Biol 206(5):589–598.

46. Lubin A, Zhang L, Chen H, White VM, Gong F (2014) A human XPC protein inter-actome–a resource. Int J Mol Sci 15(1):141–158.

47. Suloway C, et al. (2005) Automated molecular microscopy: The new Leginon system.J Struct Biol 151(1):41–60.

48. Lander GC, et al. (2009) Appion: An integrated, database-driven pipeline to facilitateEM image processing. J Struct Biol 166(1):95–102.

49. Tang G, et al. (2007) EMAN2: An extensible image processing suite for electron mi-croscopy. J Struct Biol 157(1):38–46.

50. Ryu S, Zhou S, Ladurner AG, Tjian R (1999) The transcriptional cofactor complex CRSPis required for activity of the enhancer-binding protein Sp1. Nature 397(6718):446–450.

51. Bradley DE (1965) The preparation of specimen support films. Techniques forElectron Microscopy (Blackwell Scientific Publications, Oxford, UK), pp 58–74.

52. Suloway C, et al. (2009) Fully automated, sequential tilt-series acquisition with Legi-non. J Struct Biol 167(1):11–18.

53. Voss NR, Yoshioka CK, Radermacher M, Potter CS, Carragher B (2009) DoG Picker andTiltPicker: Software tools to facilitate particle selection in single particle electronmicroscopy. J Struct Biol 166(2):205–213.

54. Ogura T, Iwasaki K, Sato C (2003) Topology representing network enables highlyaccurate classification of protein images taken by cryo electron-microscope withoutmasking. J Struct Biol 143(3):185–200.

55. van Heel M, Harauz G, Orlova EV, Schmidt R, Schatz M (1996) A new generation ofthe IMAGIC image processing system. J Struct Biol 116(1):17–24.

56. Frank J, et al. (1996) SPIDER and WEB: Processing and visualization of images in 3Delectron microscopy and related fields. J Struct Biol 116(1):190–199.

57. Mindell JA, Grigorieff N (2003) Accurate determination of local defocus and specimentilt in electron microscopy. J Struct Biol 142(3):334–347.

58. Mallick SP, Carragher B, Potter CS, Kriegman DJ (2005) ACE: Automated CTF esti-mation. Ultramicroscopy 104(1):8–29.

59. Hohn M, et al. (2007) SPARX, a new environment for Cryo-EM image processing.J Struct Biol 157(1):47–55.

60. van Heel M, Schatz M (2005) Fourier shell correlation threshold criteria. J Struct Biol151(3):250–262.

61. Pettersen EF, et al. (2004) UCSF Chimera–a visualization system for exploratory re-search and analysis. J Comput Chem 25(13):1605–1612.

62. Drozdetskiy A, Cole C, Procter J, Barton GJ (2015) JPred4: A protein secondarystructure prediction server. Nucleic Acids Res 43(W1):W389-94.

14822 | www.pnas.org/cgi/doi/10.1073/pnas.1520104112 Zhang et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

020