Two-stage dynamic DNA quality check by Xeroderma Pigmentosum group C protein

Two-stage dynamic DNA quality check by xeroderma pigmentosum group C protein

Ulrike Camenisch '.4, Daniel Trautlein2.4, Flurina C Clement', Jia Fei', Alfred Leitenstorfer2, Elisa Ferrando-M ay

3

and Hanspeter Naegeli"*

'Institute of Pharmacology and Toxicology, University of ZUrich- Vetsuisse, ZUrich, Switzerland, 'Department of Physics and Center for Applied Photonics, University of Konstanz, Konstanz, Germany and 3Bioimaging Center, University of Konstanz, Konstanz, Germany

Xeroderma pigmentosum group C (XPC) protein initiates the DNA excision repair of helix-distorting base lesions.

To understand how this versatile subunit searches for aberrant sites within the vast background of normal genomic DNA, the real-time redistribution of fluorescent fusion constructs was monitored after high-resolution DNA damage induction . Bidirectional truncation analyses disclosed a surprisingly short recognition hotspot, com- prising - 15% of human XPC, that includes two p-hairpin domains with a preference for non-hydrogen-bonded bases in double-stranded DNA. However, to detect damaged sites in living cells, these DNA-attractive domains depend on the partially DNA-repulsive action of an adjacent p-turn extension that promotes the mobility of XPC molecules searching for lesions. The key function of this dynamic interaction surface is shown by a site- directed charge inversion , which results in increased affinity for native DNA, retarded nuclear mobility and diminished repair efficiency. These studies reveal a two- stage discrimination process, whereby XPC protein first deploys a dynamic sensor interface to rapidly interrogate the double helix, thus forming a transient recognition intermediate before th e final installation of a more static repair-initiating complex.

Subject Categories: genome stability &

dynamiCS

Keywords: DNA repair;

genome stab ility; protein dynamics

Introduction

Nucleotide exc ision repair (NER) is a fundamental protective system that promotes genome stability by eliminating a wide range of DNA les ions (G illet and Scharer, 2006). In add ition

to

(6-4) photoproducts and cyclobutane pyr imidine dimers (CPOs) ca used by ultraviolet (UV) li ght, the NER pathway removes DNA add ucts ge nerated by electrophilic chem icals

'Corresponding author. Institute of Pharmacology and Toxicology, University of Zurich-Vetsuisse, Winterthurerstrasse 260, Zurich 8057, Switzerland. Tel.:

+

41 446358763; Fax: -I 41 4463589 10;

E-mail: naegelih@vetpharm.uzh.ch

4These authors contributed equally to this work

as well as intrastra nd DNA cross-links, DNA-protein cross- links a nd a subset of oxidative les ions (Huang et

ai, 1994;

Kuraoka

et ai,

2000; Reardon and Sancar, 2006)

.

The NER system operates through the cleavage of damaged strands o n either side of injured sites, thus releas ing defective bases as the co mponent of oligomeriC DNA fragments (Evans

et ai,

199 7) . Subsequently, th e excised oligonucl eotides are re- placed by repair patch synth esis before DNA integrity is restored by

ligation.

Hereditary defects in this NER process cause devastating syndro mes such as xeroderma p igmento- su m (XP), a recessive disorder prese nting with photo- sensitivity, a > 1000-fold increased risk of sk in cancer and, occasionally, internal tumours and neurological complica

-

tions (Cleaver, 2005; Andressoo

et ai,

2006; Friedberg

et ai,

2006). XP patients are classified into seven repair-deficient co mplementation groups designated XP-A through XP-G (Clea ver et a

i, 1999;

Lehmann, 2003).

In the NER pathway, the initial detection of DNA damage occurs by two alternative mechanisms. One s ubpathway, referred to as tra nscr iption-coupled repa ir, takes place when the transc ription machinery is blocked by obstructi ng lesions in th e transcribed strand (Hanawalt and Spivak, 2008). The seco nd subpathway, known as global genome repair (GGR) , is triggered by the binding of a versatile recog nition complex, composed of XPC, Rad23B and centrin 2, to damaged DNA anywhere in the genome (Sugasawa et al, 1998; Nishi

et ai,

2005) . XPC protein, which is the ac tual damage sensor of this initiator complex, displays a ge neral preference for DNA substrates that contain helix-destabilizing lesions including (6-4) photoproducts (Batty et al, 2000; Sugasawa et ai,

2001).

In the particular case of CPOs, thi s recognition function depe nds on an auxili ary protein discovered by virtue of its character istic UV-da maged DNA-binding (UV-DDS) activity (Nichol s

et ai,

2000; Fitch

et ai,

2003) . The affinity of this accessory factor for UV-irradiated substrates is conferred by a DNA-binding subunit (DDB2) mutated in XP -E cells (Scrima

et ai,

2008).

To ac hieve its outstanding substrate versatility, XPC pro- tein interacts with an array of normal nucl eic acid residues surrounding th e lesion in a way that no direct contacts are made w ith the damaged bases themselves (Buterin

et ai,

2005; Trego and Turchi , 2006; Maillard

et ai, 2007)

. This exceptional bindin g strategy has bee n co nfirmed by structural analyses of Rad4 protei n, a yeast ort hologue that shares -40% similarit y with the human XPC sequence. In co- crys tals, Rad4 protein associates with DNA through a large transglutaminase-homology domain erGO) fl ank ed by the three

~-hairpin

domains BHDl, BHD2 and BHD3 (Supplementary Figure 1; Min and P avletich, 2007). In view of the position of these structural elements relative to the accom- panying model substrate, a recognition mechanism has been proposed in which BHD3 would 'sample the DNA's conforma- tional space to detect a les ion' (Min and Pavletich, 2007) .

These earlier studies describing tile features of an ultimately stable XPC/Rad4- DNA co mplex exp lain its ab ility

2387 First publ. in: The EMBO Journal ; 28 (2009), 16. - pp. 2387-2399

http://dx.doi.org/10.1038/emboj.2009.187

Konstanzer Online-Publikations-System (KOPS)

to serve as a molecular platform for the recruitment of transcription factor lIH (TFlIH) or other downstream NER players (Yokoi et ai, 2000; Uchida el ai, 2002). How ever, one of th e most chall enging issues in th e DNA repa ir fi eld is the ques tion of how a versatile sensor-like XPC/Rad4 examines the Watso n- Crick double helix a nd faces the task of actually finding base lesions among a large excess of native DNA in a typical mammali an genome (Scharer, 2007; Sugasawa and Hanao ka, 2007). To address this

long-standing

question , we exploited flu orescence-based im aging techniques (Houtsmull er et ai, 1999; Houtsmull er and Vermeulen, 2001; Politi el ai, 2005) to visuali ze the mobility of XPC protein at work in the chromatin context of living cells. Our res ults point to a two-stage discrimination process, in which the rapid DNA quality check driven by a dynamic sensor of no n-hydrogen-bond ed bases precedes th e fina

l

engage ment of BHD3 with

lesion

sites .

Results

Instantaneous recognition of DNA lesions in human cells

Damage-induced changes of molecul ar dynamics in the nu- clear compartment have been followed by C-termina

l

con- jugation of

the

human XPC polypeptide with green- fluoresce nt protein (GFP). The time-depend ent relocati o n of this fusion product was tested by transfection of repair- deficient X P-C fibrobl asts th at la ck fun ctional

XPC

because of a mutation leading to premature termination at codon 718 (Chavanne et ai, 2000).

Individual

nuclei containing

low levels

of XPC-GFP (simil ar to the X PC expression in wild- type fibroblasts) were id

entifi

ed on

the basis of

their overall fluorescence (Suppl ementary Figure 2). To indu ce lesions, the

nuclei

were subjected to near-infrared irradiation using a pulsed multiphoton laser, thereby generating spatially confined and clearly detecta ble pJttern s of DNA damJge with minim al collateral effects (Meldrum et ai, 2003).

A

BIW

B

(6-4) CPO

XPC XPC

The res ulting

laser

tracks contained (6-4) photoproducts (Figure 1A) and CPOs (Figure 1 B), representing the major UV

lesions

processed by the NER sys

tem. As expected,

wild- type XPC-GFP was rapidly co ncen trated at nuclear sites containing such photolesions (Figure 1A and

8).

As earlier studies showed that th e UV-induced accumulation of XPC is stimulated by DDB2 protein (Fitch et ai, 2003 ; Moser et ai, 2005) , we app

lied

the same procedure to XP-E cells, in which an R273H mutat

ion generates a

DDB2 product that is inactive in DNA binding and fails to be expressed to detectab

le levels

(Nichols et ai, 2000; Itoh et ai, 2001).

In

this XP-E back

-

gro und , X PC-GFP is nevertheless effectiv ely relocated to UV- irradiated tracks (Figure 1C), cons

istent with the

known abi

lity

of XPC protein to detect (6-4) photoproducts

in

the absence of UV-DDB activity (Batty et ai, 2000; Kusumoto el ai, 2001).

To determine the kinet

ics

of protein redi stribution, DNA photoproducts were fonned along a single

lO-~lm line cross-

ing the nucleus of XP-C cells. Maximal accumulation of XPC protein was detected after treatm ent with a near-infrared radiation of 300-360 GW . cm-

²

(Supp

lementary

Figure 3).

Subsequently, DNA damage was induced with 314 GW cm-

²

to generate

~

5000 UV

les

ions in each ce

ll

or, on the average, 1 UV lesion

in ~

1.6

x lOG base

pairs (see Materials and method s). Under th ese conditi ons,

the local

fluor escence in

irradiated

areas increased nea rly

instantaneously leading to

a cl early distingui shabl e relo cati on of XPC fu sion protein already 3 s after irradi ation (Supplementary Movie 1). With progressive accumulation of wild-type XPC, a half-maximal increase in local fluor escen ce intensity WJS reach ed after

~40

s (Figure 1D) . A plateau level of fluorescence in the irrad iati on tracks, refl ecting a steady-s tate situ ati on with constant turnover, was detected after

~

300 s.

Concordance of relocation and DNA-binding activity

Besides the truncating

XPC

mutation, th e XP-C fibrob

lasts

used in this study (GM16093) are characterized by a

BNJ

C

XP-E cells BNJ

1l

a: XPC

c

~

(!)

0

~ ^XPC-GFP

c~

'!IH:lHl!HllH.I!ffHll:lfull~lffi

.- Q) 60

Q)U

fH

⁴⁰

^««tll

^{W .}

g

~ 20

pt

- 0

;;:: ::J 0

0 100 200 300 400

Time (s)

Figure 1 In sta ntaneous recognition of DNA damage by XPC protein in

li

vin g cell s. (A) High-resolution patterns of DNA damage and XPC-GFP accumulation

. X

P-C fibrob

lasts expressing low levels of XPC-GFP were laser treated to generate - 5000

UV

lesions along each linear irradiati

on track. The cell

s were fixed after 61ll

in and

(6-4) photoproducts were detectecl by imJ11unochemica

l staining using the red dye Alexa 546.

B/W,

bl ack-and-white images illustrating the pattern of UV lesions

(upper panel)

and the accumul ation of XPC-GFP

(lower panel).

Merged, superimposed images in which the relocation of XPC-GFP matches the pattern of DNA damage. Hoechst, DNA staining vi sualizing the nuclei.

(B) Co

-lo

caliza tion of XPC-GFP and CPOs. (el Efficien t relocati on of XPC-GFP to UV irradiation tracks in XP-E cells devoid of UV-DDB activity.

(0) R ea

l-time kin

etics of DNA damage recogn

iti

on. A single lO-pm line of UV photoproducts was generated across each nucleus of XP-C cells.

The acc umulati on of XPC-GFP at different time poi nt s is plotted as a percentage of the average flu oresce nce before irradiati on

(n=7).

Error

bars, standard errors of the mea n.

comparably low level of DDB2 protein (Supplementary Fi gure 4). This

reduced

DDB2 expression suggested

that th

e CM

l6093

fibroblasts may prov ide a cellu

lar

co nt ex t in wh

ich. in

contrast to an earlier report (Yasuda et

al. 2007).

the dam age recognition defect of XPC mutants becomes evident without preceding DDB2 down-regu

lation.

This view was confirm ed by testing th

e

nuclear dynamics of a repair-deficient W690S mutant with minimal DNA-binding affinity (Bunick et ai, 2006; Maillard et al. 2007; Hoogstraten et

al. 2008).

In co njunction with the CFP fusion partner. this pathogen

ic

mutant is expressed in similar amounts as th e wi

ld-type control and also localizes to the nuclei. However, in

th e XP-C fibroblas ts of thi s stud y.

the single W690S

mutati on causes

> fiv

e-fold redu cti on in th e reloca ti on to

UV-damaged

areas (Figure 2A; Suppl ementary Movie 2). These findin gs were confirm ed when another tec

hn

ique was used to inflict geno tox ic stress. th at

is

by UV-C irradiation (254 nm wavelength)

th

rough

the

pores of polycarbonate filte rs (Mone

el

al. 2004). In fact. compared with wild-type XPC. the W690S mutant exhibits only a marginal tendency to accumulate in UV-C radiation-induced foci (data not shown).

Oligo nucleotide-b

inding assays with XPC

protein expressed

A

^BI'N

B

-g

¹⁰⁰

0::

l 80

"0 c

60

"

.0 0

C

cl: ⁴⁰

~ z

C!l 0

20

0

~-f ~v

C 60

~

50

Q) (J c

40

Q) (J

in in

sect

cells confi rmed

that

this W690S mutati o n and th

e

corresponding alan

ine substitution (W690A) abrogate the interaction with DNA (Fi

gure 2B) .

The same analys is was ex t.end ed to furth er repair-defi cient XPC mutants targeting conserved aromatic residues (Maill ard et

al.

2007). A nearly complete

loss of

DNA b

inding

is co nferred by th e F733A mutat ion. whereas the W531A and W542A substitutions are associated with more moderate defects (Figure 2B). When tested in CM 16093 fibroblasts as CFP fusions. the damage-dependent redist ribution of these different mutants correlates closely with the respective DNA

-binding

properties. In fact, the W690S. W690A and F733A deri vatives display a poor ability to concentrate at damaged sites . In contrast.

the residual DNA-binding activity

of W531A and W542A

leads

to an intermediary

level

of accumulation

in

areas co ntaining UV photoproducts (Figure 2C). From this tight corres pondence between DNA binding and nuclear redistribution. we concluded that the rapid reloca tion of XPC protei n to UV lesion sites reflects the intrinsic capacity of this sensor subu nit to detect DNA damage throu gh direct interactio ns with the nucleic acid substrate.

.... ~ t>"v~ ^r!>~ ~~ 1:)"3

~","J ~\., <f:'''J _{~ro ~roOJ}

•• • ^•• • ^{+ wt}

• •

W542A

'"

^Q)

0 " ³⁰ ~!!!."~iiI~~~:;'~~" • ^W531A

~~

;;::

.!;

Q) I/)

"'

^~

²⁰

(J

.£

10

0

100

..

200 Time (s)

•

300

• F733A A W690A + W690S

Figure 2 Dependence on intrinsic DNA-binding activity. (A) Representative image (in col our and black-and-white) showi ng the low residual accumulation of th e W690S mutant 6 min

after

irradiation . DNA lesions were counterstain ed by antibodies against CPOs. (8) DNA-binding ac ti vity determined by direct pull down. Wild

-type

(wI) XPC or mutants were expressed in S{9 cells as fu

sion

constructs with maltose-binding protein (MB P)

. Cell

Iysates con taining si mi

lar amou

nts of XPC protein (Maillard

e/ al.

2007) were in cubated with a sin gle-stranded l3S-mer oli gonucleo tide. Subsequently. racliolabelled DNA molecul

es

captured by XPC protein were separated from th e free probes using anti-MBP antibodies linked to magnetic beads, and th e radioac tivity in each fract

i

on was Quan

tified in a sci

ntillati on co unter. DNA binding is represented as the percentage of radioactivit y imm obilized by wt XPC protein after deduction of a background va

lue determined with

empty beads

(n=3).

Error bars. standard dev iati on.

(e)

Correlation between DNA bindin g and the kinetics of X PC accumulati on

in

XP-C cells

(n =

7). See legend to

Figure 10 for details.

Role of the transglutaminase-like domain

As

th

e

transglutaminase-like region maps to the N-terminal

part of

human

XPC (Figure 3A), we ge

nerated

N-terminal truncations

(XPCI18-9~O' XPC~27-9'IO

an d XPC607

-9>

1O)

to test h

ow

the TGD sequences contribute to

DNA damage recogn

i-

tion

in living cells. The

positions US and 607 were selected for

these truncation

s to allow for comparisons with an ear

lier

in vitro stu dy monitoring

the

DNA-, Rad23B- and TFlIH- binding activity of XPC fragments (Uchi da et ai, 2002).

Another

trun

cate

(XPCI_~~5)

was

inclu

ded as a

negative

cont ro

l that lacks th

e entire C-termi

nal half. The functionality

of these constructs, conjugated

to GFP at

their C-terminus, was compared in a host-cell reactivation assay that

has been

developed to measure the cellu

lar GGR

activit y (Carreau et ai,

1995). Briefl

y, XP-C fibroblasts were

transfected with

a dual

luciferase

reporter sys

tem along with

an expression vector codi

ng

for full

-length

or truncated XPC

fusions.

The reporter

A

1

185 337 520 632

940

N ~ I __ ~IT_GD~I __ ~lm_D~I __ ~lc

XPC118_940

XPC427_940

---

XPC607_ 940

c

_~0

^'*'

I>.<::J I>.<::J

0

'?J'*'

qjO)

I'./>J

~./O) 1>.<»\1>.'\,'0'0

~

..: ... '"

^{~ ~}

..: ..: ..:

II)

Cl 150 -^{75 -}

-- _- - -

^It) 0> ^{() a.}

'1

G-

F G

60

c~

#:

'i ~ 40

lll~

~II)

o Q)

20

c ~

pla

smid, which

carries a Photinus

luciferase gene,

wa

s

da-

maged by

exposure to

UV-C light

and supplement ed with an

und

amaged vector

that expresses the

Renilla

luciferase.

GGR

efficiency was assessed aft

er IS-

h incubati

ons by determ ining Photinus

lu

ciferase activit y

in

cell lysates, followed by

normalization

again

st the

Renilla control.

The full

-length

pro

tein

(XPC I

_940

) and an XPC I18

-9>

IO deriv ative,

isolated

by functional complementation (Legerski

an

d Peterson,

] 992), were

profici

ent in

correcting

the repa

ir defect of XP-C cells (Figure 3B), thus showing that gene reactivation is determined by

the ability of

the GGR pathway to excise offendi

ng UV lesions.

However,

this

repair activit y could not be rescued by XPC427-

940

and XPCG07-

940

(Figure 3B), implying that

th

e N-terminal part of XPC protein

is essenti

al for th e GGR reaction. All

tested fragments

w

ere

detected in transfected fibrob

lasts

in similar amo

unts

as the full

-length

control or

th

e

fun

ctional XPC118-

940

derivative

B C

100

.~

⁸⁰

~

₆₀

:5 ~ ⁴⁰

.3 20

o

8N1

CPDs

a1

0::

C

~

(!)

E

8N1

0

cr,

""

I C\I

....

Sf

0>

""

I

g

8N1

Red

- 0

607-940

~ Green

0

1-940 607-940

Figure 3 Mapping of

the

damage

sensor

domain

to the

C-terminal part of human XPC. (A) Scheme illustrating th

e positi

on of

the TCD seque

nces relati ve to the N-tenninal XPC trun

cates.

(B) CCR activity determined by host-cell reactivation

assay (n = 5; error

bars, standard

deviation).

(e)

Immunoblot analys

is of XP-C cells transfected

with expre

ssion vectors coding

for th

e indicated

fu

sions. The protein level was

probed

usin

g anti-CFP antibodies. C,

endogenous CAPDH control. (D) Representative image showing that an X

PC fragment

lacking the C- terminus (XPCI ⁴⁹⁵⁾

fa

ils to accllmulate in laser-damaged areas.

The XP-C fibrohlasts were fixed

6 min after irradiation. BjW,

black-and-white

images showing

that the tra

cks

of DNA damage

(upper panel)

do not

induce an accumulati

on of tru ncated

X

PC fu

sions (lower panels).

(E)

Representative images (in

colour and

black and wh

ite) showing

that XPC

427 ^9>\0

and XPC

607 ^<)40 accumulate

in damaged

areas

of XP-C

fibroblasts. The distribution of fluorescent fusion products was monito

red 6m

in after irrad

iation. (F)

Local increase of fluorescence resulting

from

the damage-induced redistri

bu

tion of full-length XPC or XPC607-9,IO

' A 10-llmline of

UV

photoproducts was generated across each nucleus

and the

resulting

accumulation

of fusion proteins (aft

er a 6-min incubation) is

plotted as a percentage of th

e average

fluoresce nce before

irradiation (n = 7).

E

rror

bars, standard errors of the means. (G) R

ep

resentative

image illustrating

that XPC607-

9'IO

accumulates

in f

oci

generated by

UV-C

irradiation

(lOOJ

m

2)

through th

e pores of polycarbonate filters. The XP-C cells were fixed 15 min after treatment

Jnd CPOs

were detected by

immun

ochemical stain

ing. The position of XPC6D7-9,1O

foci

is in

dica ted by

the arrows.

(Figure 3C), indicating that their repair defici ency do es not result from reduced expression or enhanced degradation.

Next, all GGR-deficient trunca tes were tested for th eir damage recognition profici ency in XP-C fibroblasts. Neither XPC I - 495 (Figure 3D) nor XPC I _ 7 1 8 (Supplementary Figure 4) were red i stributed to sites of photoproduct formation in the irradi ated nuclei of living cells, confirming that the C-term- inal half of XPC protein is necessary for lesion recognition.

However, unlike these C-terminal truncation s, fragment

XPC'127-9~0

retains the ability to co ncentrate in laser-in'a- diated areas (Figure 3E). Even mo re surprising was the observation that the sma ller fragment XPC 6 0 7

^-?,1O

read ily accumulates at sites containing UV photolesions (Figure 3E). The quantifi ca tion of defined

lO-~lm

tracks showed that XPC607- 9'IO is only -3 0% less effic ient than full -length XPC in relocating to damaged sites (Figure 3F) . Thu s, a large N-term inal part of human XPC (65 % of the full - length protein including its TGD reg ions) stimulates DNA damage recognition, but is not absolutely required for th e sensi ng process itself. This conclusion is confirmed by the accumulation of XPC 607 940 in UV-C foci generated by irra- di ati on th rough the pores of polyca rbonat e filters (Figure 3G) .

Differential contribution of p-hairpin domains

According to the Rad4 crysta l, three consecutive

~-hairpin

domains (BHD1, BHD2 and BHD3) mediate the interaction with damaged DNA (see Supplementary Fi gure I) . In the homologous XPC sequence, these structural elements range from res idue 637 (start of BHDI) to residue 831 (end of BHD3). To examine how each of these domains contributes to DNA damage recognition in living cells, we generated the C -terminal truncations XPC I _ 7 41 (comprising BHDl and BHD2) and XPC I _ 8 31 , which includ es all three BHDs (Figure 4A). Again, the trunca tion position 741 was chosen to allow for comparisons with an ea rli er

in vitro

study (Uchida

et

ai, 2002) . The constructs were conjugated to GFP at their C -term inus and tested for th eir ability to initi ate the GGR reaction. [n the case of XPC 1 7 41, th e repair funct i on is reduced to a background level observed with empty GFP vector (Figure 4B). However, the repo rter gene was reacti- vated to -40% of control in th e presence of X PC 1 831, indica ting that despite its C-terminal truncation , thi s l arge fragment reta ins in part th e ability to recruit NER factors to lesion sites. Although attempting to delineate the borders of a minimal sensor domain, we surprisingly found that essen- tia ll y the same GGR activity was induced by XPC I _ 766 , that is by adding only 25 amino acids to XPC I _ 7'11 (Figure 4B). A compariso n with the Rad4 orthologue indicates that these 25 amino acids (residues 742- 766) belong to an N-termin al extension of BHD3, which folds into a

~-turn

structure (see Figure 4A).

The UV-induced relocation of trunca ted XPC derivatives was tested in XP-C fibrobla sts ex pressing sim ilar low leve ls of each GFP construct (Supplementary Figure 5). Consistent with its distinctive functionality in th e GGR assay, we ob- served that XPC 1 766 accumulates more effectively than XPC I _ 7 '11 to the

lO-~lm

tracks of photolesions generated by laser irradiation (Figure 4C). An unequivocal pattern of XPC I - 766 accumulation along the radiation tracks was also recorded in XP-E fibrobl asts, that is ill th e absence of UV-DDB activity (Figure 40). A quantitative comparison in both XP-C and XP-E cells highlights the increase in damage recognition

when the truncation was introduced at residue 766 as com - pared with the truncation at position 741 (Figure 4E), thus show ing that th e dam age-s pecific accumulation of XPC trun- cates as well as the effect of the

~-turn

structure takes place in the absence of DDB2 protein. A cl ea r difference between XPC I - 766 and XPC I _ 7'1 1 was reprodu ced when foci of flu or- escence were monitored after UV-C irradi ation through the pores of polycarbonate filters (Figure 4F) . Taken together, thi s efficien t redistribution of XPC I- 766, irrespective of the cell type or techniqu e used to inflict DNA damag e, estab lishes for th e fi rst tim e that mo st of BIID3 is not required for th e initial damage-sensing process.

The p-turn structure enhances XPC dynamics

The GGR and re location assays of Figure 4 revealed a striking difference between XPC I _ 7 41 and XPC I - 7 6 6 because of the 25- amino-acid

~-turn

extension. To analyse the function of thi s

~-turn

structure, we compared th e nuclear mobility of differ- ent truncates using flu orescence recove ry after photobl each- ing (FRAP; Houtsmuller and V ermeulen, 2001). [n cells that express similarly low levels of GFP fu sion constructs, a nuclea r area of 4

~lm2

was bleached and, subsequently, protein movements were tested by recording the recovery of local fluorescence, which is dependent on the abilit y of the GFP fusions to move rapidly within the nuclea r compartment.

The control experiment of Figure SA shows how, in th e absence of a fusion partner, the GFP moiety moves freely inside the cells. Instead, the nuclear mobility of full -length XPC-GFP is restrained by its larger size and propensity to undergo macromolecular interactions, as reported earlier (Hoogstraten el ai, 2008). Surprisingly, in a direct comparison between XPC I _ 74I , XPC I _ 766 and XPC I - 83 1, a larger size correlated with increased nuclear mobility (Figure 5B) . The FRAP curves obtained with these different truncates were use d to ca lculate effecti ve diffusion coeffi cients

(Doff;

Supplementary Tab le

I). It

was un expected to find that. in undamaged ce ll s, XPC I _ 766 (containing BHDl, BHD2 and the

~-turn

structure) and XPC I - 831 (containing all three BHDs) move more rapidly inside the nucleus (Dorr= 0.44 and 0.49 j.1m

²

S - I, respectively) than th e shorter polypeptide XPC I _ 74 1 la cking th e

~-turn (Dcrr=

0.34j.1m

²

S- I) . We con- cluded that these C-terminal truncations di sclose the exis- tence of a dynamic interface, residing within the

~-turn

structure, which enhances th e constitutive nuclear mobility of XPC protein in the absence of ge notoxic stress.

Subsequently, the FRAP approach was used to assess th e

corres ponding responses to UV-C irradiation. [n accord with

its poo r accumu lation along DNA damage tracks (Figure 4C) ,

the mobility of XPC I _ 74 1 i s only marginally affected by the

induction of photoles ions (Figure 5C) . [n contrast, the diffu-

sion rates of XPC I - 766 (Figure 5D) and XPC I - 831 (Figure 5E) ,

whil:h accuJllulat e in UV les ion tracks, are significan tly

reduced (the respective

Dcrr

values are listed in

Supplementary Table

I).

In the case of XPC I _ 83 I , th e induc-

tion of DNA damage had a two -fold effect. First, UV lesion s

decreased the initial rate of protein diffusion exactl y as

observe d with XPC I _ 7 6G . Second, similar to th e response of

full -length XPC (Hoogstraten el ai, 2008), th e overall fluor-

escence recovery is less co mplete on UV irradiation

(Figure 5E), indicat ing that a fra ction of XPC I - 831 is immo-

bi lized in a damage-specific mann er. In summary, th ese

A

1

N ~

^__

~~~~~~

^__

~

^____

~C

II-tum

XPC'_74'

c

^{1-766 BIW 1-741}

E

XP-C cells XP-E cells

~ 100 '0

~ 80

~ I:

<I>

~ 60 e? 0 ::>

;::

.5 40

~

e?

..s

0 ₂₀

0

bo'

"rJ>

^{~, bo'}

"rJ>

^~,

;..

^}b

;..

^/'0

,

^~

_{" "}

^~

_"

B

~ 100

C

~ 80 .2 60

~

~ 40

$ 20 '0

.3

0 -1-1---....1..-

BfW

F

CPDs

1-766

CPDs

1-741

D

^XP-Ecells

1-766

BfW

BfW

Figure 4 BHD3 is not required for DN

A

damage detection. (A) Scheme illustrating the location of BHD

and

~-turn

sequences

relative to th

e C

- termin

al XPC trun cates of this study. (B) GGR ac

tivit

y determined by host-cell

reactivation assay in

XP-C fibroblasts

(n =

5; error bars, standard

deviation). (e) Representative images (taken 6 min

after irradiati on) comparing the accumulation of XPC'

_7GG

an

d XPC'_74' at damaged sites.

In the black-and-white representati

on, the li near irradia

tion tracks are surrounded by a dashed rectangle. (D) Representati

ve im age illustrating

the accumulation of XPC'_7GG along UV radiation tracks

generated

in

XP-E fib

roblasts

devoid of

UV-DDB activity.

(E) The local

increase in fl

uorescence, because of damage-induced redistribution

s of XPC truncates,

was measu

red in XP-C and XP-E cells and ploUed as

the percentages of

wt

control as outlined

in

Figure

10 (11 =

5;

error bars, standard errors

of the mea

n). (F)

XPC'

_7(,6 is also more efficient

than XPC'

_7<1

'

in

accuillu

lating in DNA damage foci

generated by

UV-C irradiation through the pores o1'polycarbonate filters

(see Figure 3G

for details). XPC'_7('(' [top)

and X

PC'_7

4'

foci (bollom)

arc indicated by

the arrows.

protein mobility studies

show th at

BHD3 indu

ces th e

forma- tion

of a stabl e

nucleoprotein complex once the lesion has been detected.

Antagonistic composition of the dynamic sensor domain

The truncation

studi

es

of Figures

4

and 5 suggeste

d that residues 607-766

may

be sufficient to find lesion sites in the

ge

nome. This hypoth

es

is was

co

nfirm

ed

by

expressing sho

rt protein fragments in

XP-C fibroblas

ts (Figure 6A). In the case

of

XPC607-

766 (consisting of BHD1/BHD2

and the

p-turn

structu

re).

a

clear

pa

ttern of damage-induced accumulation was detected imm

ediately after

laser irradiation (Figlll'e 6B).

In

contrast, XPCGQ7_7

4, (l

acking

the p-turn) failed to accumu- late in the tracks

o

f UV lesions.

XPC607-

741 was unable to

relocate

to damaged

areas

regardless of whether the GFP moiety was placed at the

C-

(Figlll'e 6C)

or at the N-term

inus (data not

shown) . These

results support the conclusion that

XPC607-

766 di

splays a

minimal

sensor surface

with damage recognition

activity

in living human

cells.

Immobile

fraction

XPC-GFP

Prebleach Bleach Postbleach recovery

t

Diffusion of mobile molecules

Recovery time Bleach

B

1.2

C

1.2

.~ 0.8 .~ 0.8

'"

c: c:

~ 0.6

. 1 - 831 ( -

UV) ^{Q) 0.6} •

1-741

(

-

UV)

.!:

:s

0.4 . 1-766 (

- UV)

^0.4 •

1-741 (+ UV)

0.2 . 1-741

(- UV)

^0.2

0 0

-5 0 5 10 15 20 25 30 35 -5 0 5 10 15 20 25 30 35

TIme(s)

Time (s)

0

^1.2

Retarded diffusion E

^1.2 Immobile fraction

"

.~ 0.8

~

^~ 0.8

'" '"

r:::

0.6 r::: 0.6

~

^0.4 _0.2

• _• ^{1- 1- 766 (- 766(+}

^UV)

^{UV) ~}

^0.4 _0.2

• ^{1- 831 ( - UV)}

• ^1- 831 (+ UV)

0 0

-5 0 5 10 15 20 25 30 35 -5 0 5 10 15 20 25 30 35 40

TIme(s) Time (s)

Figure 5 Identifi catio n of a dynamic co re and two-stage damage recog niti on. (A) Principle of FRAP analysis. An area of 4

~m2 in

th e nuclei of XP-C fibroblasts ex pressing a particular GFP

construct

i s bl eached with a 488

-nm

wavel ength

laser. The kin

etics and ex tent

of flu

orescence

recovery

(shown for GFP

and XPC-GFP)

depends on diffu sion rate. molecular interaction s as well as th e fr action

of

immobile molecules.

(B) R ecovery

plots of XPC truncates

normalized to prebleach intensity

(n

= 12). E rror bars. standard errors of the mea n. Th e difference between X PC ,

~66

and

XPC ,_83, is not significant. (e)

The nuclear mobility of XPC ,

741

remains unaffected by UV-C irradiation at a dose

of lOJ m-2

(n = 12) . (0) The initial diffusion of XPC ,-

₇₆₆

is reduced by UV

light (IOJ

m-

2

• n =

12).

reflecting transient molecular interactions durin g stage 1 of the damage recognition process. (E)

A fra

ction of XPC ,_

83,

i s stably immobi

li

zed after UV

irradiation

(lOJ cm

2

. n =

12).

reflectin g stage 2 of th e damage

recognition

process.

The fragments

XPC607 741. XPCG07 766

and

XPC607 831

have been

isolated to assess

their DNA-binding properti es using

13s-mer DNA substrates.

All three fragments were ex pressed and

purified

as so

luble polypeptides

without any signs of aggregation or prec ipitation that

would be

indicative

of

defective

protein

folding (Figure

60).

We

compared

their

binding

with three different DNA

conformations:

homodu- plexes. heteroduplexes

with

three contiguous base mis- matches or

single-stranded oligonucleo

tides of the

same length.

Although

XPC607 741 (containing BHDI

and

BHD2)

is unable

to

find DNA

lesio

ns in

living cell

s.

this

fra gmenl displays a preference for unpaired bases emb edded in double- strand ed DNA.

In

fact.

XPC607-741 binds

with higher

affinity

to

heteroduplex

DNA relative to homodupl exes or

single-

stranded oligo nucleotid es (Fi gure

6E).

A similar preference for hetero-

over

homodupl exes is retained

by XPC607-766.

which includes both BHDl / BHD2 and the

~-turn

structure

(Figure 6F).

thus supporting the notion that this minim al sensor is active in

living

cell s by sea rching for destabili zed base

pairs.

A side-by-side com- parison

of

do se-dependent DNA-binding ac tivities with

XPCG07-741 and XPCG07-76G showed that the ~-turn

structu

re leads

to

a

substantial

reduction

in nucleic

acid

binding

(Figure

6F).

In particular.

we found

that the aSSOCiatIOn constant representing the

interaction with homoduplex

DNA decreases nea rly lO-fold from

2.7 x 10⁹

M- ' for

XPC607_741

to

2.8 X

10

⁸

M

I

for

XPC607-766' This

drop in

binding to

the native double

helix

implies that the enhan ced nuclea r mobility co nferred by

amino

acids 742-766 (Figure sB) results from

an antagonistic DNA-repulsive effect.

Finally. to test the contribution of BHD3, the same 13s-mer substrat es were us ed to monitor the DNA-binding

properties of a longer fragm

ent

(XPC607 831)

co mprising all three

BHDs.

Figure

6G

shows that this

larger

fragment ha s the

character-

istics

of a single-stranded DNA-binding

protein, indicating that BHD3

itself confers a pronounced selectivity for

single-

stranded conformations. The characteri

stic DNA-binding pro-

file

of thi s larger fragment

XPC607-831

corresponds roughly to that detected when

identical reactions

were

carried out

with full

-length XPC

protein (Supp

lementary Figure 6).

Design of an XPC mutant with retarded nuclear mobility

We postulated that part of the DNA

-repulsive action mediat

ed

by the

~-turn stru

cture (Figure

6F) arises

from negatively

charged side

chains that clash with

the

phosphates of th e

nucl eic acid backbone. Thi s hypothes

is predi

cts that it s

hould

o

kOa

100 75

50 37

25

B

607-766

-

- -

75 50 37

25

BIW c

81W

6071 IBHOll BH02 I 741

6071 IBHOll BH021 1766

~-turn

E

80 • heteroduplex

80 • heteroduplex

IBHOll BH021 I BH03 1831

+

^(:I-turn

~

⁶⁰ • homoduplex

"0

c: ::l

• single stranded

-g

.8

⁴⁰

.8

^::l 40

<

Z

o

20

<

t;

20

o __

~--~--~~--~ O~~--~--~~--~

o

~;;;:=::!=~:::::!::::!-

o

50 100 150 200 250

o

50 100 150 200 250

o

50 100 150 200 250 Fragment 607-741 (nM) Fragment 607-766 (nM) Fragment 607-831 (nM)

Figure 6 Antagonistic composition of the minim al damage senso r. (A)

Immunobl

ot analysis of XP-C fibroblasts after transfecti on with vectors coding for the indicated XPC-GFP sequences. The expression was probed usin g anti-GFP antibodies. NTC, non-transfected cell s; GFP, cells transfected with

the

GFP sequence alone; G, GAPDH co ntrol. (B) R ep resentative image

illustrating that fragm

ent XPC('n7

7('('

readily accumutat

es

in damaged areas co ntaining DNA photolesions. The distri bution of fluoresce nt fusion produ cts was monitored Imin after

laser irradiation. B/W, black-a

nd-white image.

(e)

XP C607

-7'"

is unabl e to recog nize

UV

lesions in

li

ving cell s. Fibroblasts were subjected to fixa tion 1 min after irradiation and

(6-4)

photoproducts were detected by immunochemi cal staining. B/ W, black-a nd-w hite im ages

showing

th at

UV lesions (upper panel)

did not lead to accumulation of th e fu sion protein

(lower panel).

(0) Gel electrophoretic analysis of purified XPC fragments expressed as glutathion e-S-tran sfera se (GST) fusions in E. coli or with a histidine (Hi s) tag in S(9 cell s. (E) DNA binding of XPC607

_74

1 determined by oligonucleotide capture. The indica ted co ncentrations of XPC-GST fragments were incubated with radiolabelled 135- mer oli gonucleo

tides

(3-mi srnatch heteroduplexes, homoduplexes and single strands)

.

Thereafter, DNA molecules immobilized by XPC fragme nts were sepa rated from th e free oligonucleot ides using glutathionc-Sepharose beads. followed

by

th e quantifica tion of radioacti vity associated with the beads. DNA binding is represe nted as the percen tage of tot al input radioacti vit y cap tured by XPC fragmen ts after deducti on of a background value de termined with

empty beads (n=6; error bars,

standard deviation).

(F)

DNA-binding profi le of th

e minimal damage

sensor (XPC

607 -766) determined as described

in th e legen d to Figure 6E. (G) Con tribution of BHD3

. The DNA-binding profile of XPC607 ⁸³¹

was determined as outlined in the lege nd

to

Figure 6E, except that pu

ll

downs were performed with Ni-NTA agarose beads.

be possibl e to mitigate thi s DNA-repellent effect by replacing negatively charged am

ino acids with positiv

ely charged ana- logues. We identi fied a glutamat

e

moi

ety at

position 755 of th e human

~-turn

motif that

is conserved among

hi gher eukaryotes (Figure 7 A) and

inv

erted

the charge

of thi s particular side chain by substitution with lysine.

The consequence of this engineered charge inversion was first tested by co mpal'ing th

e interacti

o n with nat

ive

double- stranded DNA in biochemical assays. For th at purpose, the

lysin

e substitution was introduced into XPCG07-

76G,

thus generating a mutated fragm ent of 160 amino acids (E755K607-

766) that, similar to it

s wild-type counterpart (XPC607-

766),

is ame nab le to

ex

pression and pu rification

as

a so

luble

polypeptide. DNA homoduplexes of 135 base pairs were used to determine the DNA

-binding

capacity of this mutated fragm ent

in

relation

to

th e wild-type contro

l.

As

illustrated

in the comparison of Figure 7B, the E75SK muta- tion was ab

le

to partially reverse

the

drop in DNA binding

res ulting from the presence of the

~-turn

structure

in

XPC607-

766' Binding sa

turation studi es with homoduplex DNA indica tes that the association co nstant increased from 2.8

x 10⁸

M-

^I

for XPC607-

766

containing th e wild-type se- quence (determined

in the earlier

section) to 7.4 x 10

⁸

M-

¹

for the E7SSK607-

766

derivative, which carries th e sin gle charge inversion.

The

se

f i ndin gs led us

to

generate a mutant GFP fu sion

co nstruct to co nfirm that the effect of the

~-turn

structure in

enhancing the XPC dynamics, observed with truncated deri-

vatives (Figure 5B), is retained

in

the full -length protein

context. Unlike other repa

ir-defective

XPC mutants (W5 31A,

W542A, W690A , W690S and F733A), all of which di

splay

a higher nuclear mobility than the wi

ld-type co

ntrol

(I-1oogs traten et ai, 2008 and data not shown), the novel

E7SSI< mutant is characterized by a strikingly redu ced nucle-

ar mob ility (Fi gure 7C) accompanied by a signifi ca nt GGR

defec t (Figure 70)

.

Collec tively, these effects induced by a

A

Human Mouse Chicken Zebrafish

+

741-YQPPVAVDG-KVPRNEFG-757 726-YQPPIAVDG-KVPRNEFG-742 738-YQPPIAVDG-KVPRNEYG-754 676-YQPPIAVDG-KVPRNEFG-692 Worm 936-YRRPPLKNG-KIPHNEYG-952 Thale cress 656-LCLPPAVNG-IVPKNERG-672 Yeast 540-YIPPLASASGEITKNTFG-557

C

^1.2

.~ 0.8 c: 0.6

Q)

.E

0.4

0.2 0

•

^E755K

-5 0 5 10 15 20 25 30 35

nme(s)

D

~

100

~

f ⁸⁰

~ ⁶⁰

ll: ₄₀

~ Q)

13 20

..J ::J 0

.:l:-

*" ^{.: f:>}

^i:J'?-

~~':) ~'OC8 ~'OC8

«t

Figure 7 Analysis of the dynamic interface by site-directed mutagenesis. (A)

I

dentification of a co nserved glutama te

(arrow)

in the p-turn motif of higher eukaryotes

. This residue is not

conserved in the Rad4 sequence, suggesting that

the yeast orlhologue

may

have

different dynami c properties. (B) A sin gle E7SSK mutation red uces

the DNA-repell

ent effect of the p-tUI'll structure. The associat

i

on of XPC('1I7

7'''

, XPC(,t17 7(,(, and E7SSK('07

7(,(,

with homoduplex DNA was compared at a po

lypeptide concentrati

on of I S O nM, as outlined in the legend

to

Figure 6E. DNA bindin g i s

represented as

the percenta ge of total input

radi

oacti vity ca ptured by XPC

fragments (n =

6; error bars, standard deviati on). A control reacti on was carri ed out with empt y beads.

(e)

FRAP analys

i

s showing that, in undamaged cells,

the nuclear mobilit

y of the full

-length

E7SSK mutan t is retarded rel ative t o the wt contro

l (n = 12;

error bars, standard

error of the

mea n) . (D) !-Ios

t·cell

reactiva tion assay showi ng that the E7SSK mutation co nfers a signi fica nt GGR defect. All results we

re corrected

for the background ac tivity in XP·C cells

transfected with

th e GFP vector

(n =

5; error bars, st andard deviation)

.

single site-d irected mutation co nfirm

that th

e dynamic prop- erties of

its

minimal sensor surface, co

nferred

by

th

e

~-turn

struclure, are crilical for the abilit y of human XPC protein to act as a sensor of DNA damage.

Discussion

We elucidated the mechanism by which XPC protein

scruti-

ni zes DNA quality in

living

cells. The most outstanding finding

is

th e identifica ti on of a tw o-stage di scrimi nati on process

tri

ggered by a dyna

mi

c sensor

interfa

ce

that

detects DNA damage without the

involvement

of a prominent DNA- binding doma

in

(BHD3). which was thought to represen

t the

primary

lesion

recognition module on th e basis of

the

Rad4 crystal structure (Min and Pavletich, 2007). The new

ly iden-

tified

se

nso r interfa

ce serves

to

rapidly screen

th

e

double

helix

for

the

prese

nce

of unpaired bases, thus

localizi

ng damaged target sites

th

at are amenab

le to

the subsequent

installation of an ulti

mate repair-initiat

ing co

mplex.

Dynamic molecular dialogue with the DNA double helix Accord

in

g

to

the aforement

ioned

Rad4 structure, the TGD

region

cooperates with BHD

I to

associate with a po

rti

on of doub

le-stranded DNA

fl anking th e lesion (see Suppl ement

ary

Figure

1).

However, we observed that a

large N-terminal

segment (6S % of the hum an sequence

including

most TGD sequences) has a stimul atory role, but is not directl y requi red for the relocation of XPC protein to focus o

n

DNA

lesions

(Figure 3). [n the absence of

this

TGD segment , a stro

ng interaction

with the norm al duplex

is

neve rtheless med iated

by the ear

lier described (Uchi

da et ai, 2002) minimal DNA- binding fragment XPC607-

7

41 , which consists of BHDI and BHD2 (Figure 6E)

. Instead, a longer fragment

cover

ing all

three BHDs d

i

sp

lays a co

mparably low affi nil y for the norma l duplex (Figure 6G), indica ting that the doubl e-stra

nded DNA-

binding ac

ti

vity of BHDl/BHD2 is opposed by th e neighbour-

ing BHD3

sequence. Th e further dissection of this critical XPC region revealed th at a short

~-turn

extension of BHD3

is

suffici

ent

to mediate in part such

an

antagonis

tic

effect (Figure 6F).

Several observat

ions in li

ving ce

lls support the notion that

the add

ition

of this

~-turn

extension co

nveys

a

tru

e gain of function rath er

th

an ca usin g the des tabili zation of adjacent struct ural element s in th e respec tive XPC co nstructs. First, XPC

I-76(,

and XPC 1

_R31

di splay a residual GGR function that

is

miss

ing in

the case of XPC

I_7

4I, which

lac

ks the

~-turn

structure (Figure 48). The fact that XPC

I-766

and XPC

I_831

exert a similar

ly low complementing activity is likely because

of

th

e absenc'e of at

least

so me co mponents of th e TFIIH-

recruitin g doma

in

in

th

eir C- terminal region (Uchida el ai,

2002). Second, a side-by-side comparison of the same C-

terminal truncates shows that

the enhanced

nuclear mobility

co nferred by the

~-turn

structure (Figure SB) co rrelates with a

morc efficient relocati on

1.0

UV les ions (F

i

gure 4E). Thi rd, th e

nucl ea r mobility of XPC I

-766

, but not XPC

_I_7•_II, is retarded by UV

damage (Figure

SC

and D)

,

confirming that the fo rmer

detects DNA

lesions

more effecti vely. Fourth, in living ce

lls,

the dam

age-

induced accumul ati on of an earlier defined mini-

mal DNA-binding fragment (XPC607-

741) is stri

ctly depende

nt

on the presence of the

~-turn

structure (Figure 6B). Finally,

the critical role of this dynamic

~-turn

su bdomain is sup- ported by a site-d irected E755K substitution tha t reverts in part its DNA-repelle nt acti

on.

Th

e

in

creased

affinity of thi

s

novel mutant for the native doubl

e he

lix results in decreased nuclear mobility and markedly reduced repair activity (Figure 7). According to the Rad4 structure, the critical position 755 maps to an amino-ac id sequence tha t is in close contact with the DNA

substrate

(Min and Pavleti

ch, 2007).

Thus, our find ings indicate that the

~-turn

structure displays both DNA-a ttractive and DNA-repulsive forces that dictate the dynamic interplay with duplex DNA such that, in the full genome

co

ntext, this subdoma in facilitates damage recognition by providing

sufficie

nt mobility to the XPC molecules searching for les ions.

Identification of a transient recognition intermediate

On binding to damaged substrates, XPC protein indu ces loca l DNA melting and kinking (Evans et ai, 1997; Janicijevic et ai, 2003; Mocquet et ai, 2007). A structural basis for these rearrangements is again prov ided by the Rad4 crystal, in which th e

~-hairpin

of BHD3 is inser ted through the DNA dupl

ex, ca

using two ba

se

pa irs to

e

ntirely flip out of th

e

double helix (see Supplementary Figure 1)

.

In view of these features of the Rad4-DNA co mpl

ex, it was unexpec

ted to find that most of BHD3 includin g the protruding

~-hairpin

is

actually

not necessary to sense DNA damage in living cells.

In fa ct, an XPC fragme nt that conta ins the

~-turn

structure, but is devoid of the re maining BHD3 sequence because of a truncat ion at position 766 (XPC

I 766), accumulates in UV

foci with remarkabl

e efficiency (-

60% of th

e

full-length control;

Figure 4E), but without forming stable nucleopro te in com- plexes (Figure 50)

.

Similar to the W690S mutant, this trun- cated XPC

I-766

derivative is even able to induce GGR activity (Figure 4B), although to mode rate levels that are not

suffi-

cient

to

complement the repair defect of XP-C cells. A damage-specific accumulation of XPC

I_766

was also detected in DDB2-deficient XP-E fibroblasts (Fi

gure 40 a

nd E) a nd V79 hamster cells (data not shown), thus excluding that the BHD3-ind ependent relocation occurs in an indirect manner by association with UV-DDB. Finally, the conclusion th at XPC protein form

s

a transient damage recognition inte rmediate withou t th

e in

volvement of BHD3 is supported by th

e finding

that a small fragment (XPC607-

766) consisting only of BHDl/

BHD2 and the

~-turn struc

ture (together - 15% of th e human XPC seq uence) s till function

s

as a cellul ar DNA damage sensor (Figure 6B). This minimal

sensor surface

displays a binding preference for dupl

exes containing non-hydrogen-

bond

ed bases,

a generic feature of damaged DNA, and hence fun

ctions

as a mol

ecular

calipe r of thermodynami

c

bas

e-pair stability.

A two-stage quality-control inspection

Although the BHD3 segment (residu

es 767-83

1) a nd its

~-hairpin

are not required to attract XPC protein to lesion sites, thi

s

additiona l domain favours the subsequ

ent

forma- tion of stable nucl

eoprotein

comp lexes, reSUlting in an im- mobile fra

ction of

XPC prote in in respon

se

to DNA damage (Figure

5E). The

biochemica l

ana

lysis of p urified fragm

e

nts

shows

that, unlike the

BHD1/BI-lD2/~-turn

minimal

sensor,

which displays a preference for duplexes with unpaired bases, BHD3 co nfers an exquisite selectivity for single- stranded DNA conform a ti ons (Figure 6G). [n conjunction

with th

e

ea rli

er me

ntioned Rad4

structure,

th

ese

findin

gs

indicate that BHD3 does not pa rticipate in the

ear

ly and tra nsie nt recognition intermediate, but, in

stead,

facilitates the sub

sequent

stabilization of a repair-initi at ing complex using its Si ngle-stranded DNA-binding activity to encirc le the unda maged strand across les ion sit es.

To co nclud

e,

this is the first report providing ev id

ence for a

two

-stage

discrimina tion mechanism by w hich

XPC protein

carries out its versatile recognition function (Figure 8). This tw

o-stage

process obviates the diffi

culty

of probing

every

genomic base pair for it

s susceptibility

to undergo a BHD3

-

me diated

~-hairpin

insertion. Instead, the energetica lly less dema nding search conducted by the dynamic BHDl/BHD2/

~-turn

interface is like ly to precede more extensive BHD3- dependent struct ural adjustments. This initi al search leads to th

e detection of non-hydrogen-bonded residues tha

t are more prone than native base pa irs to be flipped out of th

e

doubl

e

helix and

,

hence, become an interaction partner for the single-stranded DNA-binding ac tivity of BHD3. A crit ical step of this two-stage quality-control process is the transition from

a

n ini tiall y labile senso r intermediate to the more stable ultimate recognition complex. Two constitutive interaction partners of XPC protein, Rad23B and centrin 2

, are thought to

exert an accessory function not only by inhibiting XPC degradation, but also by stimul atin g its DNA-binding activity (Ng et ai, 2003; Xie et ai, 2004; Nishi et ai, 2005). Such an

a

uxiliary role is supported for Rad23B by the observation that XPC607-

94

0, a fragment that fails to associate with Rad23B (Uchida et ai, 2002), has a reduced DNA damage recognition capacity in living cells (Figure 3F). In addition, the two-step discrimination process identified in thi s study raises the possibility that Rad23B

,

centrin 2 or other binding partners may facilitate the installat ion of an ultimate XPC-DNA com- plex by lowerin g the energetic cost of critical nucleoprotein rea rrangem

ents required for the

fin

a

l

~-hairpin

insertion.

B

I I I I I I I I I

Uttimate

1

recognition complex

Figure 8 Two-stage detection of DNA lesions by XPC protein.

Model depicting the switch from a dynamic damage sensor inter- mediate to the ultimate recognition complex. CA) This study iden- tifies a minimal sensor interface that rapidly scrutinizes base-pair integrity. This initial search, carried out by BHDl/BI-tD2 in con- junction with the ~-turn structure, results in the formation of a labile nucleoprotein intermediate. (B) The single-stranded DNA- binding activity of BHD3 promotes the subsequent transition to a stable recognition complex by capturing extruded nucleotides in the undamaged strand.