Double-strand DNA breaks recruit the centromerichistone CENP-ASamantha G. Zeitlina,b,1, Norman M. Bakerc, Brian R. Chapadosd, Evi Soutogloue,2, Jean Y. J. Wangf, Michael W. Bernsg,h,and Don W. Clevelanda,b,f,1

aLudwig Institute for Cancer Research, Departments of bCellular and Molecular Medicine, cElectrical and Computer Engineering, fMedicine andgBioengineering, University of California at San Diego, La Jolla, CA 92093; dDepartment of Molecular Biology, Scripps Research Institute, La Jolla, CA 92037;eNational Cancer Institute, Bethesda, MD 20892; and hDepartment of Biomedical Engineering, University of California, Irvine, CA 92612

Contributed by Don W. Cleveland, July 24, 2009 (sent for review June 4, 2009)

The histone H3 variant CENP-A is required for epigenetic specificationof centromere identity through a loading mechanism independent ofDNA sequence. Using multiphoton absorption and DNA cleavage atunique sites by I-SceI endonuclease, we demonstrate that CENP-A israpidly recruited to double-strand breaks in DNA, along with threecomponents (CENP-N, CENP-T, and CENP-U) associated with CENP-A atcentromeres. The centromere-targeting domain of CENP-A is bothnecessary and sufficient for recruitment to double-strand breaks.CENP-A accumulation at DNA breaks is enhanced by active non-homologous end-joining but does not require DNA-PKcs or Ligase IV,and is independent of H2AX. Thus, induction of a double-strand breakis sufficient to recruit CENP-A in human and mouse cells. Finally, sincecell survival after radiation-induced DNA damage correlates withCENP-A expression level, we propose that CENP-A may have a func-tion in DNA repair.

chromatin � DNA repair

DNA repair in chromatin is thought to occur in a stepwisemanner. Cells must recognize a damage event, and recruit

repair and chromatin machinery to the site of damage. Damagesignaling includes phosphorylation of ATM, which phosphorylatesChk2, H2AX, Nbs1, and many other proteins [reviewed in (1)].Chromatin remodeling provides access to the damaged DNA[reviewed in (2, 3)]. According to current models, chromatin isrestored some time after repair to the DNA is completed (4).

The reported kinetics of chromatin remodeling at sites of DNAdamage span minutes to hours. Histone H2AX phosphorylation isabsent from up to 6 kb on either side of a double-strand break, butspreads outward at least 40 kb on both sides (5, 6). After UVdamage, new histone H3.1 appears approximately 30 min afterdamage (4), presumably due to reassembly after DNA repair. Incontrast, bound H2B was detected surrounding double-strandbreaks for up to 4 h, then replaced 10 h later (7). Finally, despitetriggering DNA damage signaling, unprotected telomere free DNAends do not induce detectable chromatin turnover at all (8). Thus,the extent and kinetics of histone turnover and replacement to sitesof DNA damage are currently unclear.

Centromere protein A (CENP-A), a component of centromericchromatin, is an essential histone H3 variant in all eukaryoticspecies examined to date. CENP-A is known to be required for (9,10), and may be sufficient to promote (11, 12), centromere identityand assembly of the associated kinetochore protein complex, whichmediates chromosome segregation during cell division. Since cen-tromeric DNA sequences are not conserved in metazoans [re-viewed in (13)], CENP-A presumably exerts its role in centromerespecification via a sequence-independent mechanism. Previously,we observed that widespread DNA damage induced assembly ofXenopus CENP-A onto sperm DNA in cell-free egg extracts (14).Here we tested the hypothesis that localized DNA damage issufficient to recruit human and mouse CENP-A in vivo.

ResultsEndogenous CENP-A Recruited to DNA Damage Sites. To generatedouble-strand breaks in DNA in living human cells, we used the second

harmonic (532nm)ofapulsed(12ps)Nd:YAGlaser (15).This strategyavoids local heating (15), and differs from approaches that use higherenergy UV (337 nm) light, which produce damage not confined to thenuclear interior. We deliberately avoided presensitizing analogues suchas BrdU, which can disrupt chromatin (16). After laser targeting alonga 0.4 �m-wide line (Fig. 1A), phosphorylation of H2AX [reviewed in(17)] was routinely detected. As expected, no changes in nuclearstructure were detected [DAPI, Fig. 1A and (15)]. Although H2AXphosphorylation can occur without DNA damage (18), an antibodyagainst activated Chk2 (19, 20) detected a DNA damage-dependentepitope coincident with phosphorylated H2AX (Fig. 1A). Activated(Ser-1981 phosphorylated) ATM and Rad51, both thought to be boundprimarily at double-strand breaks, were also detected (Fig. S1), as werephosphorylated Nbs1 and 53BP1 (Fig. 1B). CENP-A signals in lasertargeted lines of interphase cells were observed in almost all cells (87%of 143b and 85% of HeLa, n � 30 and 20, respectively). In all cases,endogenous CENP-A was still detectable at centromeres (smaller foci;grayscale or green in Fig. 1), demonstrating that CENP-A is notremoved from centromeres in response to DNA damage.

Rapid Accumulation of GFP-CENP-A at Sites of DNA Damage. Toexamine the kinetics of CENP-A targeting to sites of DNA damage,laser targeting was performed on two clonal human Hek293 cell linesinducibly expressing a GFP-CENP-A fusion protein (10) after FRT-mediated integration at a defined locus. Similar results were obtainedwith both lines. CENP-A mRNA increased approximately 7-fold within24 h of induction (Fig. S2A), and the 44 kDa GFP-CENP-A proteinaccumulated to approximately the initial level of endogenous CENP-A,which was in turn reduced to about one third of its earlier level (Fig.S2B). The decrease in endogenous CENP-A may be due to competi-tion between transfected and endogenous CENP-A for stabilization byCENP-A binding factors, as seen previously (3, 21–24).

Within 4 hours of induction, GFP-CENP-A (Fig. S2 C and D,green) colocalized with endogenous centromeres (detected us-ing human anti-centromere autoantisera Fig. S2C, red), and wasdistributed throughout nuclei (Fig. S2D), consistent with priorreports (25, 26). After longer induction times (�24 h), all cellsexhibited centromeric foci surrounded by a generalized nuclearsignal. GFP signal was not removed by extraction with non-ionicdetergents before fixation (Fig. S2, compare fixed cells in C withlive cells in D). The generalized nuclear signal (Fig. S2D)represented full-length GFP-CENP-A, not a degraded form justcontaining GFP, since the majority of protein migrated at the

Author contributions: S.G.Z., N.M.B., E.S., J.Y.J.W., M.W.B., and D.W.C. designed research;S.G.Z., N.M.B., and E.S. performed research; B.R.C. contributed new reagents/analytic tools;S.G.Z., N.M.B., B.R.C., E.S., J.Y.J.W., M.W.B., and D.W.C. analyzed data; and S.G.Z. andD.W.C. wrote the paper.

The authors declare no conflict of interest.

1To whom correspondence should be addressed. E-mail: szeitlin@sciencegeeks.org ordcleveland@ucsd.edu.

2Present address: Institut de Genetique et de Biologie Moleculaire et Cellulaire, B.P. 1014267404 Illkirch Cedex, France.

This article contains supporting information online at www.pnas.org/cgi/content/full/0908233106/DCSupplemental.

15762–15767 � PNAS � September 15, 2009 � vol. 106 � no. 37 www.pnas.org�cgi�doi�10.1073�pnas.0908233106

expected molecular weight for GFP-CENP-A (�45 kDa). En-dogenous CENP-A (�80%) was retained in the pellet afterwashing with high salt and detergent (consistent with assemblyinto nucleosomes). In addition, while half the GFP-CENP-A wasretained under these stringent conditions (Fig. S2E), the re-mainder was removed, consistent with a proportion of CENP-Aoutside of centromeric chromatin (11, 14, 21, 23, 26).

Timing of CENP-A recruitment to sites of DNA damage wasdetermined after induction of GFP-CENP-A expression in thetwo clonal cell lines. Cells were visualized with phase-contrastand fluorescence microscopy, and areas were chosen for lasertargeting (boxes in Fig. 2 A–C). Laser firing produced an initialapproximately 1.5 �m2 photobleached area (Fig. 2 C and E),whereas GFP-CENP-A accumulated in a smaller (�0.6 �m2)spot in the center, consistent with H2AX phosphorylation (seeFig. 4). This was expected since the multiphoton effect thatcreates double-strand breaks is limited to a smaller volume:comparable photobleaching was observed at approximately6-fold lower laser doses, but this was insufficient to induce DNAdamage, as reported earlier (15). Fluorescence recovery withinthe photobleached zone required at least 1 h (e.g., cell 8 in Fig.2 E and F), much longer than the �1 s that would be expectedfor a soluble approximately 45-kDa protein (27), confirming thatmost nuclear GFP-CENP-A was not freely diffusible.

Damage consistently induced CENP-A foci in 71% of targetedcells (�10%, n � 176 interphase cells; for example, eight of 10 cellsshown in Fig. 2), within an average of 5 min (�2 min, n � 82interphase cells), including �30 experiments on separate days (n �100 cells per experiment). Once formed, each CENP-A focusremained stationary, and increased in intensity for about 1 h (Fig.2F). Foci formed with identical frequency and kinetics at both roomtemperature (25 °C) and 37 °C. However, at 37 °C, GFP-CENP-Aaccumulations at targeted sites appeared, became brighter, andthen were abruptly lost (e.g., between 63 min and 68 min in Fig. 2G).Foci were not lost at 25 °C, as some were still visible 18 h later. Cellswith foci were never observed to enter mitosis.

ThehighfrequencywithwhichCENP-Aaccumulatedat sitesof laserexposure in asynchronous samples suggested that this can occurthroughout the majority of interphase, as seen for endogenousCENP-A. Nuclear cross sectional area is known to correlate with cellcycle stage (28). Comparing the nuclear areas of cells that formed fociafter laser exposure (n�31, Fig. S3, red) with the areas of the randomlycycling cells surrounding them (n � 450; Fig. S3, green) confirmed thatCENP-A focus formation was not restricted to a subset of interphase

cell cycle stages. Conversely, no cell cycle stage (aside from mitosis) wasrefractory to CENP-A focus formation.

Finally, DNA damage induced focus formation was not a phenom-enon limited to immortalized cells. GFP-tagged human CENP-A wastransiently transfected into primary human fibroblasts, and laser-induced damage produced CENP-A foci with rapid kinetics (Fig. S4B).

Rapid Accumulation of GFP-CENP-A at Sites of I-SceI Cleavage. Tofurther test whether a double-strand break is sufficient to recruitCENP-A to DNA, we used the site-specific endonuclease I-SceI (29) ina mouse NIH2/4 cell line (TM815 cells) which carries a single I-SceItarget site flanked by LacI repeats (31). A GFP-tagged version ofmouse CENP-A (GFP-mCENP-A) was constructed and expressed bytransfection in these cells. As expected, GFP-mCENP-A was recruitedto sites of laser-induced DNA damage with kinetics similar to GFP-tagged human CENP-A (GFP-hCENP-A) in human cells (Fig. S4A).After transient expression of I-SceI, a single double strand break,marked by the presence of phosphorylated histone H2AX, was gener-ated at the lacI array [visualized using mCherry-lacR (red)] (Fig. 3A).In 47% of cells, GFP-mCENP-A (green) was recruited along withphosphorylated histone H2AX (blue) (Fig. 3A). (The less than 100%efficiency is expected in this triple transfection experiment.) Addition-ally, TM815 cells were co-transfected with GFP-mCENP-A and anRFP-tagged fusion of I-SceI with the glucocorticoid receptor (30).Before addition of the synthetic glucocoricoid triamcinolone acetonide(TA), RFP-I-SceI-GR was cytoplasmic, and GFP-mCENP-A wasdetected throughout nuclei (Fig. 3B). Within 1 h after TA addition,RFP-I-SceI-GR translocated into nuclei, and in 57% of the cells (n �100) phosphorylated histone H2AX appeared at the I-SceI cleavage site

1 2

t = 90 min

12

t = 0

H2AXP

CENP-A

Chk2T68P H2AXP Chk2T68P

DNA

CENP-A

CENP-A CENP-ANbs1P

53BP1Nbs1P 53BP1

AB

Fig. 1. Endogenous CENP-A localizes to sites of laser-induced damage, alongwith DNA repair markers phosphorylated H2AX, Chk2, Nbs1, and 53BP1. (A) (Topleft) Phase contrast image of human osteosarcoma (143b) cells just before lasertargetingalongaline(red). (Otherpanels)Matchedconfocal immunofluorescentimaging 90 min (at 25 °C) after laser targeting. (Red) Phospho-histone H2AX;(green) Thr-68 phospho-Chk2; (blue) DNA detected with DAPI; (grayscale) en-dogenous CENP-A signal. Note the small foci in the CENP-A image are centro-meres. Note that the cells are motile: the orientation of cell #2 changed in the 90min after targeting. (B) HeLa cell 90 min. (at 25 °C) after laser targeting along asingle line. (Blue) DNA, (red) 53BP1 or Nbs1 phospho-Ser-343, and (green)CENP-A. Other foci in the CENP-A image are centromeres.

Bef

ore

Las

er

Aft

er L

aser

5 min. after laser 85 min. after laser

1

2

34

5

6

7

8

9

1010

1

2

34

5

6

7

8

9

1010

1

2

34

5

6

77

88

9

1010

1

2

3

4

5

6

77

88

9

1010

A B

C

E F

DLaser targeted areas

10µm

hr:min:sec

00:00:00 00:00:00 00:00:00 00:00:00

00:05:02 00:15:31 00:21:34 00:35:08

00:41:35 00:55:12 01:03:53 01:08:50

Aft

er L

aser

Exp

osu

re

37°C

, 5%

CO

2

G

roo

m t

emp

(25

°C)

Bef

ore

Las

er

Fig. 2. Rapid GFP-CENP-A accumulation at sites of DNA damage. (A–F) GFP-CENP-A cells before and after laser targeting at 25 °C. Phase contrast (A) beforeand (D) 5 min. after targeting the areas boxed in red. Epifluorescence images ofGFP-CENP-A, immediately (B) before and (C) 4 or (E) 5 min. after initiating laserexposure. Inmostcells (numbered inwhite),GFP-CENP-Aaccumulatedat thesitesof targeting. [Cells #7, 8 and 10 (numbered in yellow) were bleached during lasertargeting,] (F) Within 85 min, GFP-CENP-A formed foci within targeted regions.(G) Laser targetingas in (A–F),maintainedat37 °Cafter targeting:aCENP-Afocusappears within �5 min after laser exposure, reaches its peak intensity �1 h afterlaser exposure, and then disappears approximately 15 min later. Timestamprepresents hours:minutes:seconds.

Zeitlin et al. PNAS � September 15, 2009 � vol. 106 � no. 37 � 15763

CELL

BIO

LOG

Y

(marked with CFP-LacR) along with a large GFP-mCENP-A focus(Fig. 3B).

GFP-hCENP-A was also recruited to double stranded DNA breaksin a diploid human cell line carrying I-SceI sites at two loci (31).Transient co-transfection (Fig. 3C) was used to express GFP-hCENP-Aand RFP-I-SceI-GR (30). Before addition of TA, RFP-I-SceI-GR wasdetected in the cytoplasm, and GFP-hCENP-A was detected through-out the nucleus, with visible foci at centromeres (Fig. 3C). Afteraddition of TA, RFP-I-SceI-GR became nuclear within 1 h andGFP-hCENP-A formed one or two new foci in 75% of the cells (n �100 per experiment, repeated three times). Cells with two foci displayedone larger focus and one smaller focus (Fig. 3D), as expected for this celllinewithmultiple target sites forcleavageonchromosome6andasinglesite on chromosome 10 (31). Remarkably, GFP-hCENP-A formed afocus as rapidly as 1 min after addition of TA (Fig. 3E). Taken together,these results demonstrate that CENP-A is rapidly recruited to definedDNA double-strand breaks.

Histones H3.1 and H2B Do Not Accumulate at Double-Strand Breaks.Next, we tested whether other histones accumulated at sites of DNAdamage. Despite CENP-A accumulation and H2AX phosphorylationat sites of laser targeting (Fig. 1), neither YFP-H2B (n � 72 cells per

experiment, repeated 3 times) (Fig. 4) nor YFP-H3.1 (n � 75 cells perexperiment; see Fig. S5) accumulated in targeted areas in live or fixedcells. Instead, both YFP-H2B (Fig. 4 and Fig. S6) and YFP-H3.1 (Fig.S5) expressing cells gradually recovered fluorescence in the lasertargeted areas, with similar kinetics (3–4 h, n�10 cells each), consistentwith a previous study of chromatin reassembly of fluorescently taggedcore histones (32). Neither YFP-H2B nor YFP-H3.1 focal accumula-tions were observed, even after fixation and staining with anti-GFPantibodies to detect photobleached YFP-tagged histones. Moreover,even using transient transfection to produce a majority of histone H3.1as a GFP-tagged protein, no GFP-H3.1 foci were ever observed afterlaser targeting (n � 63 cells; Fig. S7A).

The Centromere-Targeting Domain of CENP-A (the CATD) Can DriveHistone H3 to Sites of DNA Damage. It was reported previously thatsubstitution into histone H3.1 of the CENP-A centromere Tar-geting Domain, or CATD, the central 31 aa portion of CENP-A(the last 6 residues of � helix1, all of loop 1, and all of � helix 2)is sufficient to promote assembly of chimeric histone H3.1 tocentromeres (33). Since CENP-A accumulated at sites of DNAdamage in a majority of cells, but histone H3.1 never did, wetested whether this centromere targeting domain would cause

1:00 1:30 2:00

merge RFP-SceI-GR

GFP-hCENP-A

-TA-TA -1:00 +TA 0:00 0:30

merge RFP-SceI-GR

GFP-hCENP-A

+TA 1 hr

C

D

E

RGB merge H2AX(P) mCherry-LacR DAPI GFP-mCENP-A

-SceI+SceI

RFP-SceI-GR GFP-mCENP-A

-TA+TA

RGB merge H2AX(P) CFP-LacR

A

B

min:sec

Fig. 3. Rapid GFP-CENP-A accumulation at double-strand breaks induced by I-SceI cleavage in human and mouse cells. (A) Mouse cells carrying an I-SceI target siteco-integratedwith lacOrepeats (30)after transient transfectiontoexpress (green)GFP-mCENP-Aand(red)mCherry-lacRand(top)withoutor (bottom)withHA-taggedI-SceI. (Blue) �-H2AX; (grayscale) DNA detected with DAPI. (B) The same mouse cell line as in (A), transiently co-transfected to express (green) GFP-mCENP-A, (red)CFP-lacR and (grayscale) RFP-I-SceI-GR and (top) without TA or (bottom) after TA addition for 1 h. (Blue) �-H2AX. (C–E) Human cells carrying a single SceI target site onchromosome 10 and multiple SceI target sites on chromosome 6 were transiently co-transfected to express GFP-CENP-A and RFP-I-SceI-GR. (C and D) Cells imaged for(red) RFP-SceI-GR or (green) GFP-hCENP-A and (C) without or (D) 1 h. after addition of TA, which induces nuclear accumulation of RFP-I-SceI-GR to cleave the DNA. (E)Timelapse images of a single cell transfected as in (C and D) immediately before and after addition of TA. Timestamp is minutes:seconds.

0:00:00

0:00:00 0:13:32 3:41:32

Before laser

0:00:00

Targetedregions

0:01:17

Afterlaser

H2AXP

After laser 0:02:24

Before laser 0:00:00

A

GFP-H2B

90 minutes

F

B C D E

G H I J

90 minutes

Fig. 4. Histone H2B never accumulates in areas of laser-induced DNA damage. (A–D) Epifluorescence images of HeLa cells stably expressing YFP-H2B (A) beforeand (B–D) after laser exposure (red lines). (C) �-H2AX and (D) YFP-H2B 90 min. after laser exposure. (E) Phase contrast image of HeLa cells stably expressingYFP-H2B before laser targeting. (F–J) YFP-H2B epifluorescence of the cells in (E) and (F and G) before or (H–J) after laser targeting. Red squares in (G and H) denotelaser targeted areas. Timestamp is hours:minutes:seconds.

15764 � www.pnas.org�cgi�doi�10.1073�pnas.0908233106 Zeitlin et al.

accumulation of chimeric H3.1 at sites of DNA damage. Genesencoding a panel of CENP-A:H3.1 chimeric proteins (26, 33)were tagged with GFP and expressed in HeLa and HCT116 cellsusing transient transfection. The CATD was able to driverecruitment of histone H3.1 to sites of DNA damage, albeit atlower frequency than wild-type CENP-A (Fig. 5 A and B). Thecomplementary mutations in CENP-A, replacing parts of theCATD with the analogous sequences of histone H3.1, reduced orabolished targeting to sites of laser exposure (Fig. 5A, HSA,HH2, H2.1, and H2.3), collectively identifying the �2 helix(mutant HH2) as the most critical region. Thus, recruitment tosites of DNA damage and assembly at centromeres utilizes acommon targeting motif within CENP-A.

Other Centromeric Proteins Are also Recruited to Sites of DNADamage. Since CENP-A recruitment to sites of DNA damage andcentromeres requires the same domains, we examined whetherother centromeric proteins were also recruited to sites of DNAdamage. GFP-tagged expression constructs for CENP-N,CENP-U, CENP-T, and CENP-M (22, 34, 35) were each tran-siently transfected into HCT116 cells, and these cells weresubjected to laser exposure. In most (80%) targeted cells,CENP-N-GFP accumulated very rapidly (within 2 min) at sitesof DNA damage (Fig. 5C). CENP-U-GFP was recruited to thesesites of DNA damage with comparable kinetics (2 min) and asfrequently (75%), although the foci were consistently smallerthan those formed by CENP-N-GFP (Fig. 5D). CENP-T-GFPaccumulated less often (44%) at sites of DNA damage, and onlybecame visible later (30 min on average; Fig. 5E). In contrast,CENP-M-GFP did not accumulate at sites of DNA damage, andits recovery within the bleached areas was very rapid (Fig. S7B).,consistent with freely diffusing protein (27). Taken together,these results demonstrate that, in addition to CENP-A, compo-nents of the CENP-ANAC, including CENP-N and CENP-U, arerapidly recruited to sites of DNA damage, while others (e.g.,CENP-T) assemble after CENP-A, consistent with the order ofassembly at centromeres (34–36).

CENP-A Recruitment to Double-Strand Breaks Is Enhanced by NHEJ butIndependent of H2AX. In mammalian cells, non-homologous endjoining (NHEJ) is thought to be the predominant pathway fordouble-strand break repair (37). Early steps in NHEJ involvebinding of the Ku86/Ku70 heterodimer to free DNA ends,

followed by the DNA-dependent protein kinase catalytic subunit(DNA-PKcs), and eventual ligation by DNA ligase IV. Tomeasure CENP-A accumulation after laser-induced damage,GFP-CENP-A was transiently transfected into HCT116 cell lineslacking various components of the NHEJ pathway (38, 39). Laserexposure induced GFP-CENP-A foci at high frequency in pa-rental (WT) HCT116 cells (57 � 10%; Fig. 6A), whereas focioccurred approximately 5-fold less frequently in HCT116 deriv-atives deficient in Ligase IV�/� or DNA-PKcs�/� (10%; Fig.6A). In HCT116 cells heterozygous for Ku86 (Ku86�/�),CENP-A accumulation was detected at an intermediate fre-quency (19%; Fig. 6A). Together, these data suggested that thehighest frequency of CENP-A accumulation at sites of damagecorrelates with activity of the NHEJ DNA repair pathway.

To further test this hypothesis, GFP-tagged Ligase IV wastransiently transfected into WT or Ligase IV�/� HCT116 cells (Fig.S8 A and B). In the absence of DNA damage, GFP-Ligase IVlocalized throughout the nucleus, with weak GFP signal in thecytoplasm, consistent with a previous report of Ligase IV localiza-tion by immunofluorescence (40). After laser targeting, the major-ity of GFP-Ligase IV was bleached, indicating that it is highlydynamic. GFP-Ligase IV rapidly formed a bright focus at the siteof laser exposure in almost all WT (Fig. S8A; 96%, n � 55) andLigase IV�/� cells (Fig. S8B, 88%, n � 57). Similar foci weredetected with a GFP-tagged mutant, R278H, reported to havereduced (5–10% of WT) catalytic activity (40, 41), in both WT (Fig.S8C, 94%, n � 36) and Ligase IV�/� cells (Fig. S8D, 94%, n � 50).

Next, constructs encoding wild-type or mutant Ligase IVfused to mCherry (instead of GFP) were transiently co-transfected along with GFP-CENP-A into HCT116 WT andLigase IV�/� cells, and double-positive cells were laser targeted(Fig. 6B). As expected, expressing mCherry-Ligase IV did notincrease the frequency of GFP-CENP-A foci in WT cells afterlaser exposure (57%; n � 20), but did increase the frequency inLigase IV�/� cells (28%; n � 29), while mCherry-Ligase IV-R278H produced an intermediate effect (17%; n � 12) (Fig. 6B).

The observation that CENP-A is recruited to sites of damage sorapidly, and apparently even in the absence of Ligase IV activity,raised the possibility that CENP-A recruitment might participate inDNA repair. Since CENP-A depleted cells or cells unable to loadCENP-A are non-viable (24, 42), to test whether CENP-A couldpromote cell survival after DNA damage, inducible GFP-hCENP-A cells were exposed in triplicate to three doses of ionizing

GFP-H3

00:00:00 00:00:00 00:02:08

00:22:26 01:08:49 01:09:57

CENP-N-GFP 80%, 2 min, (n = 25)

CENP-U-GFP 75%, 2 min, (n = 24)

60-80%

60-70%

5-10%

BaA

CENP-T-GFP 44%, 30 min, (n = 27)D

D

C

E

Fig. 5. CENP-A recruitment to sites of DNA damage requires its centromere-targeting domain (CATD) and recruits centromeric nucleosome-associated factorsCENP-N and CENP-U. (A) Schematic of CENP-A/H3 chimeric proteins and frequencies at which each is recruited to DNA damage foci. (B–E) Human HCT116 cellswere transiently transfected. DNA damage was induced by laser targeting, and frequencies of focal accumulation at site of DNA damage were measured. (B)GFP-H3CATD. (C) GFP-CENP-N. (D) GFP-CENP-U. (E) GFP-CENP-T.

Zeitlin et al. PNAS � September 15, 2009 � vol. 106 � no. 37 � 15765

CELL

BIO

LOG

Y

radiation (0 Gy, 2 Gy, and 10 Gy), with or without induction ofGFP-hCENP-A. Parental 293 cells (Fig. 6C, open boxes) displayed26% hypercondensed nuclei, indicative of cell death, within 24 h of2 Gy, whereas cells induced to express GFP-hCENP-A (triangles)were protected (only 2% hypercondensed nuclei). Uninduced cells

exhibited intermediate sensitivity at 2 Gy (5% hypercondensednuclei), probably due to basal GFP-hCENP-A expression (seebelow). Both the parental and uninduced GFP-hCENP-A cellsexhibited extensive cell death after 10 Gy (Fig. 6C, boxes).

To test whether cells could continue dividing after radiation in aCENP-A-dependent manner, clonogenic survival assays were per-formed. After 8 Gy, parental 293 cells were uniformly killed, butinduction of GFP-hCENP-A promoted cell survival and sustainedcolony growth (Fig. 6D). Intermediate numbers of surviving col-onies were observed with the two independent clonal lines in theabsence of tetracycline-mediated induction of GFP-hCENP-A.Survival was accompanied by increased levels of CENP-A (de-tected as increased GFP fluorescence), (Fig. S9) presumably fromtransient stabilization of GFP-hCENP-A.

Finally, to test whether CENP-A recruitment requires histoneH2AX (5, 6), GFP-mCENP-A was transiently transfected intomouse embryonic fibroblasts (MEFs) from H2AX null mice (43)and cells were laser targeted after 48 h (Fig. 6E and F). Absenceof H2AX did not affect GFP-mCENP-A focus formation, withsimilar frequencies of rapid recruitment in laser targeted WT(43 � 3%; n � 65) and H2AX null (32 � 4%; n � 72) MEFs.

DiscussionOur observations using laser-induced DNA damage and I-SceI cleav-age now establish that CENP-A accumulates at double strand DNAbreaks in normal and immortalized human and mouse cells. Recruit-ment of CENP-A, CENP-N and CENP-U occurs within 1–2 min, atimescale as fast as that seen for phosphorylation of H2AX or othercomponents of DNA repair [e.g., frequencies at which each is recruitedto DNA damage foci. (B–E) Human HCT116 cells were transientlytransfected. DNA damage was induced by laser targeting, and frequen-cies of focal accumulation at site of DNA damage were measured. (B)GFP-H3CATD. (C) GFP-CENP-N. (D) GFP-CENP-U. (E) GFP-CENP-T.ATM, Nbs1; (44)]. CENP-A recruitment is transient, disap-pearing again within approximately 1 h, and it is independent of H2AX.A comparison with CENP-A spot size and intensity at centromeres,estimated to contain approximately 5,000–15,000 CENP-A nucleo-somes (24, 45), suggests that CENP-A spreads outward surrounding adouble-strand DNA break.

Mammalian CENP-A assembles into nucleosomes in vitro (24,46, 47) and has been found in nucleosomes of asynchronous cells(22, 42, 48, 49). It has also been proposed that CENP-N interactswith CENP-A in nucleosomal form (36). We would caution,however, that it has not been established whether DNA damage-dependent CENP-A loading represents assembly into nucleosomes.At least three models for CENP-A binding are plausible (Fig. 6G).First, CENP-A could be assembled into nucleosomes (Fig. 6G, left).If so, the failure of H2B-GFP to re-accumulate at these damage fociwould require that CENP-A must be assembled along with H2B,H2A, and H4 from the original (photobleached) nucleosomes or anadjacent (photobleached) pool. A specific possibility is the selectiveremoval of H3 (with or without H4) from individual nucleosomesand their replacement with CENP-A or CENP-A/H4. Second,CENP-A (and CENP-N and CENP-U) could bind on top ofexisting chromatin (Fig. 6G, middle). Third, a final alternative (Fig.6G, right) is that DNA repair occurs in a nucleosome-free zone, asproposed previously (6). This would require removing histoneH3-containing nucleosomes and recruiting CENP-A bound in anon-nucleosomal form that lacks H2A and H2B, as has beenproposed for budding yeast centromeres (50–52).

While it is clear that CENP-A recruitment to double-strandbreaks occurs in both human and mouse cells, including primarycells, the function of CENP-A at double-strand breaks has not beendefined. Our demonstration that CENP-A is recruited to double-strand breaks independent of H2AX suggests that it is not part ofthe damage checkpoint. In contrast, the cell-cycle independence ofCENP-A recruitment, rapid kinetics, and correlation with LigaseIV catalytic activity, suggest that CENP-A may be recruited to

A

Double-strand breaks(laser or Sce1)

(~1 hour)

replace bulk chromatin

CENP-A is removed

NA

U

CENP-A recruitment within 1 minute:

DNA damagenucleus

A

NU

AA A A

NU

NU N

U

(3) CENP-A binding to damaged DNA

(with nucleosome eviction)

(1) CENP-A chromatinassembly/remodeling

(+/- nucleosome eviction)

H4

H2A

H2B

NA

U

H4

H2A

H2B

NA

U

H4

H2A

H2B

NA

U

H4

H2A

H2B

NA

U

(2) CENP-A associationwith damaged chromatin

(without nucleosome eviction)

H4

H2A

H2B

NAU

H3

H4

H2A

H2B

N A

H3

U

H4

H2A

H2B

NAU

H3

H4

H2A

H2B

N A

H3

UU

B HCT116 cells + mCherry-LigIV

% h

yper

cond

ense

d nu

cleiC D

% c

ells

with

GF

P-C

EN

P-A

foci

% c

ells

with

GF

P-C

EN

P-A

foci

HCT116 cells

colo

nies

(f

old

incr

ease

)

10 d. after 8 Gy (IR)IR dose (Gy)

% c

ells

with

GF

P-C

EN

P-A

foci

E

G

00:00:00 00:00:00

00:03:20 00:08:47

F

Fig. 6. CENP-A in function in DNA repair independent of H2AX, and possiblemodels for CENP-A assembly at sites of DNA repair. (A) HCT116 cells generatedby targeted disruption of NHEJ genes were transiently transfected to expressGFP-CENP-A; frequencies of focus formation were measured after laser expo-sure. [WT (n � 47 cells); DNA-PKcs�/� (n � 39 cells); LigIV�/� (n � 57 cells);Ku86�/� (n � 38 cells); P � 0.0202 (one-way ANOVA)]. (B) Frequencies ofGFP-CENP-A recruitment to sites of laser-induced DNA damage in Ligase IV�/�

cells transfected to express mCherry-LigaseIV or mCherry-LigaseIV-R278H. (C)Percentages of cell death following 0 Gy, 2 Gy, and 10 Gy irradiation (done intriplicate) in cells with a stably integrated, GFP-CENP-A gene, either with orwithout tetracyline-mediated gene induction. Cell death was scored by small,bright nuclei [using Hoechst] and quantified 24 h later by automated micros-copy (n � 2,500–5,000 cells per sample). Error bars show the standard deviation. (D) Cell survival in a colony formation assay after 8 Gy irradiation ofparental 293 cells or inducible GFP-CENP-A cells. (E) Frequency of (transientlytransfected) GFP-mCENP-A recruitment to sites of laser-induced DNA damagein wild-type and H2AX null MEFs. (F) GFP-mCENP-A recruitment to focal DNAdamage (red box) in an H2AX null cell. (G) Model for CENP-A recruitment tosites of DNA damage. Laser-mediated damage or Sce1 cleavage creates DNAdouble-strand breaks. CENP-A, CENP-N and CENP-U are recruited within 1–2min. Three possible modes of CENP-A binding are shown (see Discussion).

15766 � www.pnas.org�cgi�doi�10.1073�pnas.0908233106 Zeitlin et al.

double-strand breaks along with components of the NHEJ pathway,and raises the possibility that CENP-A participates in DNA repair.

Finally, when added to our earlier demonstration that CENP-Aassembly at centromeres of Xenopus sperm requires DNA repairactivities stockpiled in egg cytosol (14), CENP-A recruitment withsome of its centromeric partners to double strand DNA breakssuggests a mechanism for the formation of new centromeres(neocentromeres), which are only rarely detected as stable events(51). Since the timing of CENP-A disappearance from double-strand breaks correlates with the timing of DNA repair, ourevidence provides the initial support for a model in which CENP-Ais recruited along with repair machinery, and removed as repair iscompleted. CENP-A retention at these sites along with CENP-Nand CENP-U would provide the nucleus for a new centromere, butonly after an extremely rare convergence of permissive conditions.These conditions could include (1) failure to complete DNA repairand remove CENP-A, (2) coincidental timing of DNA damage andcentromeric chromatin replication, the latter of which has beenargued to occur only during G1 (53), or (3) defective DNA damagecheckpoint signaling, allowing cells to enter mitosis prematurely.

Materials and MethodsGFP-CENP-A stable inducible cell lines were generated based on a publishedplasmid (10) and commercial Flp-In Hek293 cells (Invitrogen). Transient trans-

fections were performed in various cell lines (22, 31, 38, 39) using Lipo-fectamine2000 (Invitrogen). Laser methods were essentially as recently de-scribed (15). Additional materials and methods are discussed in SI Text.

ACKNOWLEDGMENTS. We thank the M.W.B. laboratory (University of Cal-ifornia, San Diego) and Beth Baber (University of California, San Diego andSalk Institute) for laser sofeware and equipment support.; Annette Ochoa,David Chung and the Willert Laboratory for cell culture reagents andtechnical assistance (University of California, San Diego); Ainhoa Mielgoand Jeff Lindquist for reagents (University of California, San Diego); BethWeaver (University of Wisconsin-Madison) for cells; Dan Foltz (University ofVirginia), Ben Black (University of Pennsylvania) and Xiaodong Huang(University of California, San Diego) for cells and plasmids; David Chen (UTSouthwestern), Scott Stuart (University of California, San Diego) and KevinSullivan (National University of Ireland, Galway) for plasmids and antibod-ies; Kevin Harvey (EMD Biosciences, San Diego), Matt Weitzman (SalkInstitute) and Kyoto Yokomori (University of California, Irvine) for anti-bodies; Eric Hendrickson and York Marahrens (University of Minnesota),Matt Thayer (Oregon Health & Science University) and Andre Nussenzweig(National Institutes of Health) for cell lines. We thank Samarendra Mo-hanty and Suzanne Genc, Beckman Laser Institute, University of California,Irvine, for help with laser dosimetry and the dual objective method. Thiswork has been supported by a grant from the National Institutes of HealthGrant GM 074150 (to D.W.C.). Partial support for M.W.B. came from a grantfrom the Air Force Office of Scientific Research Grant FA9550-04-1-0101,and from the Beckman Laser Institute Foundation. D.W.C. receives salarysupport from the Ludwig Institute for Cancer Research.

1. Shrivastav M, De Haro LP, Nickoloff JA (2008) Regulation of DNA double-strand breakrepair pathway choice. Cell Res 18:134–147.

2. Bao Y, Shen X (2007) SnapShot: Chromatin remodeling complexes. Cell 129:632.3. Osley MA, Tsukuda T, Nickoloff JA (2007) ATP-dependent chromatin remodeling

factors and DNA damage repair. Mutat Res 618:65–80.4. Polo SE, Roche D, Almouzni G (2006) New histone incorporation marks sites of UV

repair in human cells. Cell 127:481–493.5. Berkovich E, Monnat RJ, Jr, Kastan MB (2008) Assessment of protein dynamics and DNA

repair following generation of DNA double-strand breaks at defined genomic sites.Nat Protoc 3:915–922.

6. Tsukuda T, Fleming AB, Nickoloff JA, Osley MA (2005) Chromatin remodeling at a DNAdouble-strand break site in Saccharomyces cerevisiae. Nature 438:379–383.

7. Berkovich E, Monnat RJ, Jr, Kastan MB (2007) Roles of ATM and NBS1 in chromatinstructure modulation and DNA double-strand break repair. Nat Cell Biol 9:683–690.

8. Wu P, de Lange T (2008) No overt nucleosome eviction at deprotected telomeres. MolCell Biol 28:5724–5735.

9. Howman EV, et al. (2000) Early disruption of centromeric chromatin organization incentromere protein A (Cenpa) null mice. Proc Natl Acad Sci USA 97:1148–1153.

10. Sugimoto K, Fukuda R, Himeno M (2000) Centromere/kinetochore localization ofhuman centromere protein A (CENP-A) exogenously expressed as a fusion to greenfluorescent protein. Cell Struct Funct 25:253–261.

11. Heun P, et al. (2006) Mislocalization of the Drosophila centromere-specific histone CIDpromotes formation of functional ectopic kinetochores. Dev Cell 10:303–315.

12. Van Hooser AA, et al. (2001) Specification of kinetochore-forming chromatin by thehistone H3 variant CENP-A. J Cell Sci 114:3529–3542.

13. Murphy TD, Karpen GH (1998) Centromeres take flight: Alpha satellite and the questfor the human centromere. Cell 93:317–320.

14. Zeitlin SG, Patel S, Kavli B, Slupphaug G (2005) Xenopus CENP-A assembly into chro-matin requires base excision repair proteins. DNA Repair (Amst) 4:760–772.

15. Kong X, et al. (2009) Comparative analysis of different laser systems to study cellularresponses to DNA damage in mammalian cells. Nucleic Acids Res 37(9):e68.

16. Miki K, Shimizu M, Fujii M, Hossain MN, Ayusawa D (2008) 5-Bromouracil disruptsnucleosome positioning by inducing A-form-like DNA conformation in yeast cells.Biochem Biophys Res Commun 368:662–669.

17. Fillingham J, Keogh M-C, Krogan NJ (2006) GammaH2AX and its role in DNA double-strand break repair. Biochem Cell Biol 84:568–577.

18. Chadwick BP, Lane TF (2005) BRCA1 associates with the inactive X chromosome in lateS-phase, coupled with transient H2AX phosphorylation. Chromosoma 114:432–439.

19. Ahn JY, Li X, Davis HL, Canman CE (2002) Phosphorylation of threonine 68 promotesoligomerization and autophosphorylation of the Chk2 protein kinase via the fork-head-associated domain. J Biol Chem 277:19389–19395.

20. Tsvetkov L, Xu X, Li J, Stern DF (2003) Polo-like kinase 1 and Chk2 interact andco-localize to centrosomes and the midbody. J Biol Chem 278:8468–8475.

21. Collins KA, Furuyama S, Biggins S (2004) Proteolysis contributes to the exclusivecentromere localization of the yeast Cse4/CENP-A histone H3 variant. Curr Biol14:1968–1972.

22. Foltz DR, et al. (2006) The human CENP-A centromeric nucleosome-associated complex.Nat Cell Biol 8:458–469.

23. Moreno-Moreno O, Torras-Llort M, Azorin F (2006) Proteolysis restricts localization ofCID, the centromere-specific histone H3 variant of Drosophila, to centromeres. NucleicAcids Res 34:6247–6255.

24. Black BE, et al. (2007) Centromere identity maintained by nucleosomes assembled withhistone H3 containing the CENP-A targeting domain. Mol Cell 25:309–322.

25. Shelby RD, Monier K, Sullivan KF (2000) Chromatin assembly at kinetochores is uncou-pled from DNA replication. J Cell Biol 151:1113–1118.

26. Shelby RD, Vafa O, Sullivan KF (1997) Assembly of CENP-A into centromeric chromatinrequires a cooperative array of nucleosomal DNA contact sites. J Cell Biol 136:501–513.

27. Dundr M, Misteli T (2003) Measuring dynamics of nuclear proteins by photobleaching.Curr Protoc Cell Biol 13:13.15.

28. Fidorra J, Mielke T, Booz J, Feinendegen LE (1981) Cellular and nuclear volume ofhuman cells during the cell cycle. Radiat Environ Biophys 19:205–214.

29. Watabe H, Iino T, Kaneko T, Shibata T, Ando T (1983) A new class of site-specificendodeoxyribonucleases. Endo.Sce I isolated from a eukaryote, Saccharomyces cerevi-siae. J Biol Chem 258:4663–4665.

30. Soutoglou E, et al. (2007) Positional stability of single double-strand breaks in mam-malian cells. Nat Cell Biol 9:675–682.

31. Breger KS, Smith L, Thayer MJ (2005) Engineering translocations with delayed repli-cation: evidence for cis control of chromosome replication timing. Hum Mol Genet14:2813–2827.

32. Kimura H, Cook PR (2001) Kinetics of core histones in living human cells: Little exchangeof H3 and H4 and some rapid exchange of H2B. J Cell Biol 153:1341–1353.

33. Black BE, et al. (2004) Structural determinants for generating centromeric chromatin.Nature 430:578–582.

34. Hellwig D, et al. (2008) Live-cell imaging reveals sustained centromere binding ofCENP-T via CENP-A and CENP-B. J Biophotonics 1:245–254.

35. Hori T ,et al. (2008) CCAN makes multiple contacts with centromeric DNA to providedistinct pathways to the outer kinetochore. Cell 135:1039–1052.

36. Carroll CW, Silva MC, Godek KM, Jansen LE, Straight AF (2009) Centromere assemblyrequires the direct recognition of CENP-A nucleosomes by CENP-N. Nat Cell Biol11:896–902.

37. Friedberg EC, et al. (2006) in DNA Repair and Mutagenesis (ASM Press, Washington, D.C.).38. Li G, Nelsen C, Hendrickson EA (2002) Ku86 is essential in human somatic cells. Proc Natl

Acad Sci USA 99:832–837.39. Ruis BL, Fattah KR, Hendrickson EA (2008) The catalytic subunit of DNA-dependent

protein kinase regulates proliferation, telomere length and genomic stability in hu-man somatic cells. Mol Cell Biol 28:6182–6195.

40. O’Driscoll M, et al. (2001) DNA ligase IV mutations identified in patients exhibitingdevelopmental delay and immunodeficiency. Mol Cell 8:1175–1185.

41. Riballo E, et al. (1999) Identification of a defect in DNA ligase IV in a radiosensitiveleukaemia patient. Curr Biol 9:699–702.

42. Okada M, Hori T, Fukagawa T (2006) The DT40 system as a tool for analyzing kineto-chore assembly. Subcell Biochem 40:91–106.

43. Celeste A, et al. (2003) Histone H2AX phosphorylation is dispensable for the initialrecognition of DNA breaks. Nat Cell Biol 5:675–679.

44. Kruhlak MJ, et al. (2006) Changes in chromatin structure and mobility in living cells atsites of DNA double-strand breaks. J Cell Biol 172:823–834.

45. Cleveland DW, Mao Y, Sullivan KF (2003) Centromeres and kinetochores: From epige-netics to mitotic checkpoint signaling. Cell 112:407–421.

46. Conde e Silva N, et al. (2007) CENP-A-containing nucleosomes: Easier disassemblyversus exclusive centromeric localization. J Mol Biol 370:555–573.

47. Yoda K, et al. (2000) Human centromere protein A (CENP-A) can replace histone H3 innucleosome reconstitution in vitro. Proc Natl Acad Sci USA 97:7266–7271.

48. Obuse C, et al. (2004) Proteomics analysis of the centromere complex from HeLainterphase cells: UV-damaged DNA binding protein 1 (DDB-1) is a component of theCEN-complex, while BMI-1 is transiently co-localized with the centromeric region ininterphase. Genes Cells 9:105–120.

49. Palmer DK, O’Day K, Wener MH, Andrews BS, Margolis RL (1987) A 17-kD centromereprotein (CENP-A) copurifies with nucleosome core particles and with histones. J CellBiol 104:805–815.

50. Camahort R, et al. (2007) Scm3 is essential to recruit the histone h3 variant cse4 tocentromeres and to maintain a functional kinetochore. Mol Cell 26:853–865.

51. Mizuguchi G, Xiao H, Wisniewski J, Smith MM, Wu C (2007) Nonhistone Scm3 andhistones CenH3–H4 assemble the core of centromere-specific nucleosomes. Cell129:1153–1164.

52. Stoler S, et al. (2007) Scm3, an essential Saccharomyces cerevisiae centromere proteinrequired for G2/M progression and Cse4 localization. Proc Natl Acad Sci USA104:10571–10576.

53. Jansen LET, Black BE, Foltz DR, Cleveland DW (2007) Propagation of centromeric chroma-tin requires exit from mitosis. J Cell Biol 176:795–805.

Zeitlin et al. PNAS � September 15, 2009 � vol. 106 � no. 37 � 15767

CELL

BIO

LOG

Y

Supporting InformationZeitlin et al. 10.1073/pnas.0908233106SI TextLaser Sample Preparation. For laser microirradiation experiments,cells were cultured in 35-mm glass-bottom dishes (Mattek cat-alog# P35G-1.5–14-C), coated by incubation with 10 �g/mLfibronectin in PBS (30–60� at 37 °C). After coating, fibronectinwas aspirated, and cells were plated in the presence of tetracy-cline 18–24 h before laser microirradiation.

Laser Induction of DNA Damage. A short-pulsed green wavelengthNd: YVO4 laser (532 nm, repetition rate: 76 MHz, 12 ps;Vanguard Laser System, Spectra-Physics, Inc.) was used in thesestudies. The beam was expanded and relayed to the backaperture of the microscope objective (63�, NA � 1.4) via theepi-f luorescence port of the Zeiss inverted microscope (Axiovert200). The pulse energy at the focused spot was varied by theorientation of a Glan-Thompson polarizer (mounted on a mo-torized rotational stage). After passing through the Glan-Thompson polarizer, the laser beam passed through the micro-scope to the back aperture of the microscope objective. Laserpower at the back aperture of the objective was measured witha power meter/detector (S 120 UV, Thorlabs). The transmissionof the Zeiss Plan-Neofluar 63�/1.4 NA objective was measuredusing the dual objective method to correct for total internalreflection losses at the objective-oil-glass-water interfaces, anddetermined to be 0.68.

The laser pulse energy at the object focal plane was deter-mined by measuring the input energy at the back aperture of theobjective multiplied by the transmission factor (0.68) of theobjective. Exposure time in the focal spot was controlled by acomputerized mechanical shutter. The exposure time of 30 msrepresents an accumulation of 2.28 � 106 pulses (12 ps each) perfocused spot of an area 0.17 �m2. The scanning pattern of thefocal spot in the nucleus was generated by a rapid scanningmirror (FSM-300, Newport Inc.), controlled by in-house soft-ware on a LabView platform and National Instrument’s dataacquisition and control board (2). From the measured power atthe back aperture of the objective, individual 12 ps pulse energyin the focal spot was 0.098 nJ/pulse.

A series of control experiments were performed to determinethe average laser power to consistently create detectable DNAdamage. Low power (5.5 mW before objective) created DNAdamage detected by immunofluoresence for H2AX and otherdamage markers in fixed cells [Fig. 1; (3)]. 11.2 mW was used asthe maximum dose for cells expressing GFP-tagged proteins. Inevery experiment with GFP-tagged proteins, stable inducibleGFP-CENP-A expressing cells were used to verify consistentfrequency of focus formation at this dose.

The average power measured before the objective was 5.5, 11,and 16.5 mW. Thus, the average power after the objective was3.75, 7.5, and 11.2 mW in the focused spot. Since the 63� (1.4NA) microscope objective focuses the beam to a diffractionlimited spot diameter of 464 nm, peak irradiances of 0.25 � 1010,0.49 � 1010, and 0.735 � 1010 W/cm2 were achieved in thefocused spot. For one line scan of 5 �m (with �10 spots/line),2,234.4 �J of total energy (energy/pulse � no. of pulses perspot � no. of spots in the line) was delivered to each cell. Weestimate that this amount of energy will create on average 1–10breaks per targeted box.

Images were originally collected manually at varying intervals;longer movies later used a time-lapse parameter with regularcollection frequency. Temporal resolution was thus limited byboth the frequency of manual collection and the time required

for laser targeting (typically 1–3 min per field, depending on thenumber of cells targeted). Timestamps are indicated for theimages shown. Targeted areas were generally chosen to belocated near or overlapping with nucleoli, to help ensure thatlaser exposure occurred internal to the nucleus.

Controlled Environment Stage Holder. A Basic Water Jacket mi-croscope stage incubator (MSI) (Okolabhttp://www.oko-lab.com/182.page) was mounted on the Axiovert 200M Zeissmicroscope for temperature control (37 °C, via water jacket) andCO2 (injecting module for CO2 and air mixture, 5% final CO2).

CENP-A-GFP Stable Inducible Cell Line Construction and Induction.GFP-tagged CENP-A plasmid (gift from Kevin Sullivan, Na-tional University of Ireland, Galway) encoding GFP ligated tothe amino terminus of human CENP-A (1), was cut with NheIand NotI, and the pcDNA5/FRT/TO plasmid (Invitrogen) wascut with KpnI and NotI. A double-strand oligonucleotide adaptor(cat#15000, EZ Clone Systems) used to make the KpnI sitecompatible with XbaI, was added directly to ligations (NEBQuick Ligase Kit). The final plasmid was verified by sequencing(Eton Bioscience) with BGH reverse and CMV promoter prim-ers (Invitrogen).

Human Flp-In Hek293 cells (Invitrogen) were co-transfectedwith FRT recombinase plasmid pOGG4 and GFP-CENP-A inplasmid pcDNA5/FRT/TO, selected transiently for the recom-binase with blasticidin (15 �g/mL), and maintained under se-lection for the GFP-CENP-A gene with 200 �g/mL hygromycinB. Colonies were isolated using glass cylinders, and expanded inselective media. Aliquots of low-passage cells from each clonewere frozen; freshly thawed aliquots were used for 6–12 pas-sages. GFP-CENP-A expression was induced with tetracycline(final concentration, 5 �g/mL).

RNA Preparation and Gene Expression Analysis. RNA (�250 ng/�Lfinal) was prepared from approximately 600,000 cells per sample,using Qiashredder and RNeasy (Qiagen). RNA was eluted in 30�L nuclease-free water, and quantitated on a Nanodrop ND-1000 Spectrophotometer. cDNA was generated using 50 �LRNA at 60 ng/�L with the Applied Biosystems high-capacitycDNA archive kit. Taqman assays were performed on an Ap-plied Biosystems Instrument 7000, in quadruplicate, normalizedagainst actin RNA, and repeated twice. Quantitation curves(serial dilutions) were generated for all samples. Data wereanalyzed using Microsoft Excel and graphs were generated usingCricketGraph.

Cellular Fractionation. GFP-CENP-A expressing cells were grownto confluence in 15-cm dishes with tetracycline for 24 h. Mediawas removed and cells were scraped into a 1.5-mL hypotonicbuffer dish (1 mM EDTA, 20 mM HEPES, pH 7.4, 5 mM KCl,and 2 mM MgCl2, with beta-mercaptoethanol and proteaseinhibitor mixture (Roche or Sigma). Nonidet P-40 (IgepalCA-630, Sigma) was added to a final concentration of 0.1% andcells were incubated on ice 10 min. To isolate nuclei, lysate waslayered on a 50% sucrose cushion and spun, 1 min at 10,000 rpmin an Eppendorf tabletop minicentrifuge. The supernatant wastaken (cytosolic fraction) and pellets were resuspended in thesame hypotonic buffer (without Nonidet P-40) and spun throughsucrose cushions twice more. The resulting nuclear pellet wasresuspended in 75 �L nuclear extract buffer (hypotonic buffer �5 M NaCl to 400 mM final concentration and 20% Triton X-100

Zeitlin et al. www.pnas.org/cgi/content/short/0908233106 1 of 12

to 1% final concentration). The sucrose cushion procedure wasrepeated three times as before, and the final pellet represents thechromatin fraction.

Transiently Transfected GFP-Tagged Expression Constructs. CENP-A:H3.1 chimeras and mutants were obtained in the originalvectors (4), except H3.1CATD (5). ORFs were amplified usingPCR (Pfu Ultra High Fidelity, Stratagene cat#600380–52; withmatching buffer and 10% DMSO). Oligonucleotide primerswere designed to preserve the reading frame and linker from theGFP-CENP-A construct (N terminus), and to insert a stopcodon to eliminate the C-terminal HA tags in the originalvectors. PCR products were band purified (USA ScientificX-tracta and Qiaquick) and ligated into PCR-Blunt II Topoplasmid (Invitrogen Zero Blunt PCR Cloning Kit), transformed,purified and sequenced (Eton Bioscience). Inserts were digested,purified, and quick-ligated (NEB, cat#M2200S) into the eGFPvector. Resulting clones were verified by digestion and sequenc-ing. DNA was purified (Qiagen spin miniprep kit, cat#27106),and concentration and purity were determined (Nanodrop)before use in transient transfections.

Cell Culture. Stable YFP-H2B expressing HeLa cells (gift fromJagesh Shah, Harvard Medical School), YFP-H3.1 expressingHeLa cells (gift from Ben Black, University of Pennsylvania),HeLa cells (gift from Beth Weaver, University ofWisconsin),143b osteosarcoma cells (gift from Haig Kazazian,University of Pennsylvania), H2AX wild-type and null MEFs[gifts from Andre Nussenzweig, NIH, (6)] and GM3491 normalhuman diploid skin fibroblasts (gift from York Marahrens,University of Minnesota) were cultured in DMEM � nonessen-tial amino acids, penicillin-streptomycin, 10% FBS at 37 °C with5% CO2.

P175 and P175-NH cells carrying SceI sites (gift from MattThayer, Oregon Health Sciences University) were maintained asdescribed (7). TM815 (mouse NIH2/4) cells carrying SceI siteswere maintained as described (8). Parental and NHEJ-deficient(Ku86�/� and DNA-PKcs�/� cell lines) HCT116 cells (gift fromEric Hendrickson,University of Minnesota) were maintained asdescribed (9, 10). LigaseIV�/�cells were a generous gift fromSehyun Oh and Eric Hendrickson, generated using a Neo-containing rAAV targeting to disrupt the first 302 bp of codingsequence in the first allele, followed by treatment with Crerecombinase and a second round of targeting, as described (11).

Transient Transfection of GFP-Tagged Expression Constructs. Cellswere transfected in suspension using Lipofectamine 2000 inOptimem (Invitrogen), plated on fibronectin-coated glass-bottom dishes (see Sample Preparation) and incubated overnight.Media with serum was added (McCoy’s 5A for HCT116; DMEMfor HeLa) a minimum of 1 h before laser exposure. Plasmidsencoding GFP-tagged NAC proteins CENP-N, CENP-U,CENP-T, and CENP-M were a gift from Dan Foltz (12).GFP-Ligase IV was a gift from David Chen’s lab (University ofTexas, Southwestern). The R278H mutation was introduced intothis construct using Quickchange (Stratagene), verified by di-gestion [the mutation creates an NdeI site, (13)] and confirmedby sequencing. mCherry-Ligase IV WT and R278H were con-structed by replacing the GFP ORF with that of mCherry. Theseconstructs were also verified by digestion and sequencing.

Induction of Double-Strand Breaks by I-SceI. The synthetic glucocor-ticoid ligand triamcinolone acetonide (TA, Sigma) was dissolved

in DMSO. To induce cleavage, human P175 or P175-NH cells (7)were transiently transfected with RFP-SceI-GR using Lipo-fectamine LTX in Optimem (Invitrogen). Twenty-four hoursafter transfection, TA was diluted in Optimem (final concen-tration 10�7 M), and this media was added to the transfected cellson the microscope stage (Olympus FV-1000 confocal micro-scope).

The mouse NIH 2/4 stable cell line (8) was transfected usingthe Nucleofector kit (Amaxa, Nucleofector R) with the appro-priate DNA plasmids, and the fluorescent proteins were visu-alized 16 h later (Zeiss Meta confocal microscope). Transfectionwith the RFP–ISceI–GR chimera was performed in cells cul-tured in DMEM supplemented with 10% stripped serum (Tetsystem approved; Clontech) and the concentration of triamcin-olone acetonide (TA, Sigma) added was 10�7 M. For immuno-fluorescence detection, cells were fixed and H2AX phosphory-lation was detected as described (8).

Colony Formation Assays. Cells were plated in triplicate (500,000cells per well) in 6-well dishes (without fibronectin), and cellswere irradiated at 0 Gy, 2 Gy, and 8Gy. Immediately afterirradiation, each well was trypsanized and replated into a 10-cm2

dish. Cells were incubated 10 days under normal culture condi-tions. Cells were fixed in 100% methanol and stained with 0.2%crystal violet (Sigma cat# C-2886), washed with tap water andallowed to dry; colonies were counted by visual inspection.Standard deviations were calculated using Prism (GraphPadSoftware).

Immunofluorescence and Confocal Image Collection. After laserexposure, cells in glass-bottom dishes were fixed with 4%formaldehyde for 10� at room temperature. Mouse anti-H2AX-ser139-P (Upstate, 1:1,000), Chk2-T68P (Cell Signaling,cat#2661, 1:200), hACA (83JD, a gift from Kevin Sullivan,1:500), anti-Nbs1-Ser-343-P and anti-53BP1 (Calbiochem, a giftfrom Kevin Harvey, 1:100). Highly cross-absorbed secondaryantibodies labeled with Tritc or Cy5 (Jackson); Alexa 488(Invitrogen) were used at 1:100. Antibodies were diluted inblocking solution (5% serum in PBS � 0.1% Triton X-100) andstaining was performed as described (14). Samples were post-fixed with formaldehyde and stored in PBS � 0.1% azide.Confocal microscopy was performed on an Olympus FV-1000.

Automated Microscopy Analysis of Cell Cycle and Cell Death. Livecells in 96-well dishes were exposed to ionizing radiation andthen stained with 5 �M Hoechst 33342 (Molecular Probes,cat#H-3570, 3,000� stock solution) diluted in PBS, and imageswere collected using the Cellomics ArrayScan (ThermoFisher)Target Activation module with a 20� Zeiss lens, five fields perwell (a minimum of 2,500 cells per well), three wells per sample.Object identification used automated segmentation of Hoechst.Analysis used custom software (CellFinder, based on the Data-Graph framework,http://www.visualdatatools.com/DataGraph/)to generate scatterplots for subset quantitation. For apoptosisquantitation (Fig. 6C), cells in the brightest subset of Hoechstsignal (total intensity �100,000) were counted relative to totalcell numbers, and percentages and standard deviations werecalculated using Prism (GraphPad Software). Similar methodswere used to measure nuclear area (detected with DAPI) for cellcycle analysis shown in Fig. S3, and further validated by analysisusing ImageJ.

1. Sugimoto K, Fukuda R, Himeno M (2000) Centromere/kinetochore localization ofhuman centromere protein A (CENP-A) exogenously expressed as a fusion to greenfluorescent protein. Cell Struct Funct 25:253–261.

2. Botvinick EL, Berns MW (2005) Internet-based robotic laser scissors and tweezersmicroscopy. Microsc Res Tech 68:65–74.

3. Kong X, et al. (2009) Comparative analysis of different laser systems to study cellularresponses to DNA damage in mammalian cells. Nucleic Acids Res.

Zeitlin et al. www.pnas.org/cgi/content/short/0908233106 2 of 12

4. Shelby RD, Vafa O, Sullivan KF (1997) Assembly of CENP-A into centromeric chromatinrequires a cooperative array of nucleosomal DNA contact sites. J Cell Biol 136:501–513.

5. Black BE, et al. (2004) Structural determinants for generating centromeric chromatin.Nature 430:578–582.

6. Celeste A, et al. (2003) Histone H2AX phosphorylation is dispensable for the initialrecognition of DNA breaks. Nat Cell Biol 5:675–679.

7. Breger KS, Smith L, Thayer MJ (2005) Engineering translocations with delayed repli-cation: evidence for cis control of chromosome replication timing. Hum Mol Genet14:2813–2827.

8. Soutoglou E, et al. (2007) Positional stability of single double-strand breaks in mam-malian cells. Nat Cell Biol 9:675–682.

9. Li G, Nelsen C, Hendrickson EA (2002) Ku86 is essential in human somatic cells. Proc NatlAcad Sci USA 99:832–837.

10. Ruis BL, Fattah KR, Hendrickson EA (2008) The catalytic subunit of DNA-dependentprotein kinase regulates proliferation, telomere length and genomic stability in hu-man somatic cells. Mol Cell Biol 28:6182–6195.

11. Kohli M, Rago C, Lengauer C, Kinzler KW, Vogelstein B (2004) Facile methods forgenerating human somatic cell gene knockouts using recombinant adeno-associatedviruses. Nucleic Acids Res 32:e3.

12. Foltz DR, et al. (2006) The human CENP-A centromeric nucleosome-associated complex.Nat Cell Biol 8:458–469.

13. Riballo E, et al. (1999) Identification of a defect in DNA ligase IV in a radiosensitiveleukaemia patient. Curr Biol 9:699–702.

14. Zeitlin SG, Barber CM, Allis CD, Sullivan KF (2001) Differential regulation of CENP-A andhistone H3 phosphorylation in G2/M. J Cell Sci 114:653–661.

Zeitlin et al. www.pnas.org/cgi/content/short/0908233106 3 of 12

Rad51ATM (Ser1981P)

a b

1

2

1

2

1

2

1

2

DNA alone

DNA alone

Fig. S1. Laser targeting activates ATM and recruits Rad51. Immunofluorescence detection of markers for double-strand breaks. HeLa cells were exposed (cell#1) or not exposed (cell #2) to laser targeting, fixed and incubated with antibodies to detect activated (phosphorylated at Ser-1981) ATM kinase (a) or for Rad51(b), shown in green. DNA (detected with DAPI) is shown in blue).

Zeitlin et al. www.pnas.org/cgi/content/short/0908233106 4 of 12

Fig. S2. Characterization of inducible GFP-CENP-A cell lines. (A) Relative levels of CENP-A mRNA before (uninduced) and after addition of tetracycline (induced)were measured using quantitative RT-PCR. (B) Relative levels of tagged and endogenous CENP-A protein were measured by western analysis with an antibodythat detects both proteins with and without tetracycline induction. Tubulin was detected as a loading control. (C) Colocalization of GFP-CENP-A fluorescence[previously characterized (1)] with endogenous centromeres (detected with human autoimmune anti-centromere antisera, hACA, in red) was assessed usingsingle confocal slices. DNA is detected with DAPI and shown in blue. Since GFP-CENP-A and hACA levels are not always equal, a Z-projection is shown (far right)where levels were equalized to emphasize overlap (yellow). (D) Variable pattern of GFP-CENP-A localization at early timepoints after induction is shown in anexample confocal image. Similar results were obtained with multiple clonal cell lines. (E) The majority of endogenous CENP-A (anti-CENP-A) and core histonefraction (anti-H3 antibody) are tightly bound to DNA, since these remain in the chromatin subfraction (chr) after washing with high salt and detergent.Approximately equal quantities of GFP-CENP-A (anti-GFP) are detected in both a nuclear fraction extracted by high salt and detergent (nuclear sup) and aninsoluble fraction (chromatin pellet).

Zeitlin et al. www.pnas.org/cgi/content/short/0908233106 5 of 12

G1 S G2 >G2Fig. S3. Nuclear size analysis of cell cycle stage distribution of cells displaying CENP-A focus formation. A dish of cells exposed to laser targeting was fixed withformaldehyde, and DNA was detected with DAPI. Images were collected on a population of cells, and nuclear areas were calculated for this population (green�s). The nuclear areas of cells with GFP-CENP-A foci were mapped onto this distribution (red triangles).

Zeitlin et al. www.pnas.org/cgi/content/short/0908233106 6 of 12

0:00:00

0:01:00

0:03:23 0:03:45

0:00:00

Before laser

0:00:44

After laser

0:00:00

0:11:18

0:00:00

Before laser

0:01:56

After laser

a

bGFP-hCENP-A in primary skin fibroblasts

GFP-mCENP-ANIH2/4 cells

Fig. S4. GFP-tagged mouse CENP-A accumulates in areas of laser-induced DNA damage in mouse cells, and GFP-tagged human CENP-A behaves similarly inprimary human cells. (a) GFP-tagged mouse CENP-A was transiently transfected into NIH 2/4 cells, and these cells were subjected to laser exposure. (b) GFP-taggedhuman CENP-A was transiently transfected into primary human GM3491 cells, and these cells were subjected to laser exposure. Timestamp representshours:minutes:seconds.

Zeitlin et al. www.pnas.org/cgi/content/short/0908233106 7 of 12

00:00:00 00:00:00 00:00:00

00:01:00 00:21:21 00:43:47

01:31:34 02:33:26 04:28:49

YFP-H3.1 (HeLa stable cell line)

Fig. S5. Example extended timelapse of YFP-H3.1 recovery after laser targeting. Targeted areas are shown in red. Timestamp is shown in hours:minutes:seconds.YFP is shown in grayscale.

Zeitlin et al. www.pnas.org/cgi/content/short/0908233106 8 of 12

00:00:00 00:00:07 00:00:39

00:23:57 00:53:58 01:45:10

02:31:50 03:22:00 03:53:57

YFP-H2B (HeLa stable cell line)

00:00:47 00:01:58 00:02:07

Fig. S6. Example extended timelapse of YFP-H2B recovery after laser targeting. Targeted areas are shown in red. Timestamp is shown in hours:minutes:seconds.YFP is shown in grayscale.

Zeitlin et al. www.pnas.org/cgi/content/short/0908233106 9 of 12

00:00:00 00:00:00 00:00:00

00:02:32 00:03:08 00:12:54

CENP-M-GFP no foci (n = 31)

GFP-H3.1 no foci (n=63) a

b

Fig. S7. No focal accumulations in cells transiently transfected to overexpress GFP-H3.1 or GFP-CENP-M. (a) Example images from transient transfections ofGFP-H3.1 expression plasmid in HCT116 cells. No accumulations were observed in areas of laser targeting (red boxes). (b) GFP-CENP-M nuclear fluorescencebleached extensively but recovered rapidly, although accumulation at the site of laser targeting was not observed (n � 31 cells). Timestamp representshours:minutes:seconds.

Zeitlin et al. www.pnas.org/cgi/content/short/0908233106 10 of 12

a bBefore laser

0:00:00 0:00:00

After laser

WT HCT116 cells

After laser

0:00:00

0:00:38 0:00:55

0:00:00

Before laser

LigIV-/- HCT116 cells0:01:35 0:01:43

c d

After laser

0:00:00 0:00:00

0:00:59 0:01:06

Before laser

After laser

0:00:00 0:00:00

0:01:09 0:01:18

Before laser

GFP-LigaseIV GFP-LigaseIV

GFP-LigaseIV-R278H GFP-LigaseIV-R278H

WT HCT116 cells LigIV-/- HCT116 cellsFig. S8. Ligase IV recruitment to sites of laser-induced DNA damage. GFP-LigaseIV focus formation after laser exposure in WT (a) or LigaseIV�/� (b) HCT116 cells.GFP-LigaseIV-R278H focus formation after laser exposure in WT (c) or LigaseIV�/� (d) HCT116 cells. Timestamp represents hours:minutes:seconds.

Zeitlin et al. www.pnas.org/cgi/content/short/0908233106 11 of 12

GFP-CENP-A total intensity

cell

cou

nt

(b

inn

ed b

y in

ten

sity

)

GFP-CENP-A uninduced

GFP-CENP-A induced

0 Gy

2 Gy

10 Gy

400

200

0

400

200

0

400

200

0

5 10 5 10 15 x 105

5 10 5 10 15 x 105

5 10 15 x 105

Fig. S9. GFP-CENP-A-positive cells accumulate, and GFP fluorescence levels increase, transiently after radiation-induced DNA damage. Population analysisdemonstrated that GFP-CENP-A signals increased transiently after irradiation in a dose-dependent manner, suggesting post-translational stabilization ofGFP-CENP-A protein in response to DNA damage.

Zeitlin et al. www.pnas.org/cgi/content/short/0908233106 12 of 12