DNA damage regulates the mobility of Brca2 withinthe nucleoplasm of living cellsAnand D. Jeyasekharana, Nabieh Ayouba,1,2, Robert Mahena,2, Jonas Riesb, Alessandro Espositoa, Eeson Rajendraa,Hiroyoshi Hattoria, Rajan P. Kulkarnia,3, and Ashok R. Venkitaramana,4

aThe Medical Research Council Cancer Cell Unit, Hutchison/Medical Research Council Research Centre, Cambridge CB2 0XZ, United Kingdom; andbEidgenössiche Technische Hochschule Zurich, Laboratory of Physical Chemistry, CH-8093 Zurich, Switzerland

Edited by Stephen C. Kowalczykowski, University of California, Davis, CA, and approved October 26, 2010 (received for review July 13, 2010)

How the biochemical reactions that lead to the repair of DNAdamage are controlled by the diffusion and availability of proteinreactants within the nucleoplasm is poorly understood. Here, weuse gene targeting to replace Brca2 (a cancer suppressor proteinessential for DNA repair) with a functional enhanced green fluo-rescent protein (EGFP)-tagged form, followed by fluorescence cor-relation spectroscopy to measure Brca2-EGFP diffusion in thenucleoplasm of living cells exposed to DNA breakage. Before dam-age, nucleoplasmic Brca2 molecules exhibit complex states of mo-bility, with long dwell times within a sub-fL observation volume,indicative of restricted motion. DNA damage significantly enhan-ces the mobility of Brca2 molecules in the S/G2 phases of the cellcycle, via signaling through damage-activated protein kinases.Brca2 mobilization is accompanied by increased binding withinthe nucleoplasm to its cargo, the Rad51 recombinase, measuredby fluorescence cross-correlation spectroscopy. Together, theseresults suggest that DNA breakage triggers the redistribution ofsoluble nucleoplasmic Brca2 molecules from a state of restricteddiffusion, into a mobile fraction available for Rad51 binding. Ourfindings identify signal-regulated changes in nucleoplasmic pro-tein diffusion as a means to control biochemical reactions in thecell nucleus.

DNA damage response | protein dynamics | fluorescence spectroscopy |single-molecule imaging

The error-free repair of DNA breaks in the cell nucleus isessential for genome integrity (1). Signals emanating from

broken DNA trigger the recruitment of diffusing proteins fromthe nucleoplasm into microscopically visible “foci” (1) at thepositionally stable damage sites (2), where repair reactions arecarried out. Whether these events involve signal-initiatedchanges in the nucleoplasmic protein pool is unclear. The dif-fusion and availability of protein reactants in the nucleoplasm isinfluenced by molecular crowding and specific intermolecularinteractions that incorporate them into macromolecular com-plexes of varying size (3). However, changes in nucleoplasmicprotein diffusion are inaccessible to traditional biochemicalmethods, and so their involvement in DNA repair (or other site-specific DNA transactions) remains largely unexplored.To address this problem, we have studied DNA double-strand

break (DSB) repair by homologous recombination (HR), a fun-damental biological process mediated in vertebrate cells by theassembly in repair foci of the cancer suppressor protein, Brca2,with its cargo, the recombination enzyme Rad51 (4). We havecombined somatic gene targeting with fluorescence correlationspectroscopy (FCS) [a method successfully applied to the anal-ysis of fluorescent particles in solution (5) or within cells (6)] todirectly measure the diffusion of EGFP-tagged Brca2 expressednatively in the nucleoplasm of living cells. Here, we show that thesignals triggered by DNA breakage initiate significant changes innucleoplasmic Brca2 diffusion, suggestive of release from a stateof restricted mobility, accompanied by an increase in its in-teraction with Rad51 in the same compartment detected byfluorescence cross-correlation spectroscopy (FCCS) (7). Ourfindings provide fresh insight into the control of HR reactionsmediated by Brca2 and Rad51 and demonstrate signal-initiated

changes in the diffusion of nucleoplasmic proteins during a bio-logical process.

ResultsReplacement of Endogenous Brca2 with a Functional GFP-TaggedForm by Gene Targeting in DT40 Cells. We used gene targeting toreplace the endogenous Brca2 protein with a functional, fluo-rescently labeled form. This procedure overcomes the limitationsof conventional approaches where fluorophore-tagged proteinsare heterologously expressed in vertebrate cells for spectroscopicanalysis (8), retaining the endogenous, untagged protein. Wedeleted one Brca2 allele in the avian DT40 cell line (9), a widelyused experimental model for HR (10), and “knocked in” the geneencoding enhanced green fluorescent protein (EGFP) to the 3′end of the coding sequence of the second allele, yielding a finalgenotype ofBrca2.EGFP/− (Fig. 1A–C, SIMethods, and Table S1).In cell extracts, anti-EGFP reacts with only a single species inWestern blots corresponding to Brca2-EGFP (Fig. S1A, Fig. 1D);alternative EGFP species that might confound fluorescencemeasurements are absent. An antibody against Gallus gallus (gg)Brca2 detects a band with slightly retarded mobility relative tountagged Brca2, but not the untagged protein (Fig. 1D). NuclearBrca2-EGFP at endogenous levels can be distinguished from thepredominantly cytoplasmic autofluorescence background, but thetypical fluorescence intensities are insufficient for image-basedanalyses (Fig. 1E, Fig. S1 B–D). Staining with an anti-GFP anti-body in fixed cells confirms that Brca2-EGFP is mainly nuclearand expressed at varying levels in the asynchronous cell pop-ulation, as expected (Fig. S1 E and F). Several lines of evidenceconfirm that Brca2-EGFP functionally replaces endogenousBrca2. First, Brca2-EGFP forms foci after DNA damage, whichcolocalize with Rad51 (Fig. S2A). Second, Brca2.EGFP/− cellsare as efficient as parental cells in supporting Rad51 focus for-mation after DNA damage (Fig. S2B). Finally, Brca2.EGFP/−cells have no measurable deficit in DNA repair by HR comparedwith control parental cells. They are neither hypersensitive to theDNA cross-linking agent, mitomycin C (MMC) (Fig. S2C), nordefective in sister chromatid exchange (SCE), a direct measure ofHR (10), before or after exposure to MMC (Fig. S2 D and E).Thus, we have created a unique experimental model to study

Author contributions: A.D.J. and A.R.V. designed research; A.D.J., N.A., R.M., E.R., R.P.K.,and H.H. performed research; N.A. and A.D.J. contributed new reagents/analytic tools;A.D.J., J.R., A.E., and A.R.V. analyzed data; and A.D.J., A.E., and A.R.V. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available through the PNAS open access option.1Present address: Department of Biology, Technion-Israel Institute of Technology, Haifa,Israel, 32000.

2N.A. and R.M. contributed equally to this work.3Present address: David Geffen School of Medicine, University of California, Los Angeles,CA 90095.

4To whom correspondence should be addressed. E-mail: arv22@cam.ac.uk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1009577107/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1009577107 PNAS | December 14, 2010 | vol. 107 | no. 50 | 21937–21942

SYST

EMSBIOLO

GY

changes in Brca2 diffusion during physiological reactions, usingquantitative fluorescence imaging.

Diffusion of Soluble Nucleoplasmic Brca2-EGFP Measured byFluorescence Correlation Spectroscopy. We used FCS to dissectthe diffusion of molecular complexes containing Brca2-EGFPwithin the nucleus of living cells. FCS measures (5) the temporalfluctuations in fluorescence arising when fluorophores diffusethrough a small confocal volume (Fig. S3 A and B). Their corre-lation over time is described by the autocorrelation function(ACF) or G(τ) (Fig. S3A; Methods). The ACF depends on theaverage concentration of fluorescent species and their dwell time(τD) within the observation volume, permitting an estimate of thediffusion coefficient andnumber offluorescently taggedmolecules(for further information, see ref. 8, SI Methods). Two features ofFCS make the technique ideal to study the diffusion of Brca2.First, FCS is particularly suited to fluorescent molecules like en-dogenous Brca2-EGFP expressed at low concentrations. Second,FCS preferentially reports the properties of soluble Brca2 mole-cules in the nucleoplasm (11), a fraction that has not been pre-viously characterized.Although the observation volume for our instrument (∼0.25 fL)

is not small enough to entirely exclude chromatin domains or smallsubnuclear structures, any fluorescent molecules stably bound tothese domains or structures are photobleached very rapidly whenobservation commences (e.g., the fluorescence trace in Fig. S3C).However, initial photobleaching is absent from the majority ofFCS traces (e.g., Fig. S3B) and is likely related to the occasionalpresence of foci of stablemolecules close to the FCSmeasurementsite (Fig. S3 legend andMethods). This result is consistent with theobservation that in the few cells expressing sufficient Brca2-EGFPto enable fluorescence recovery after photobleaching (FRAP),>60% of nuclear Brca2 is mobile (Fig. S4 A and B). Nevertheless,most cells are not amenable to FRAP analysis because the limitedfluorescence signal emitted by the low concentration of endoge-nously expressed Brca2-EGFP was too small to permit robustquantification (Fig. S4C). We therefore used FCS to focus on theanalysis of mobile Brca2-EGFP molecules, excluding the immo-bile fraction in the initial segment of the fluorescence traces.Mobile Brca2-EGFP molecules are collectively termed the “sol-uble nucleoplasmic pool.” This fraction includes molecules thatundergo transient binding for times less than the period requiredfor fluorophore bleaching during observation (8). The fluores-

cence of diffusing Brca2-EGFP molecules in the nucleoplasm canbe clearly distinguished from the autofluorescent background(Fig. S3D), showing higher signals and autocorrelation comparedwith measurements performed on parental lines.

Modeling of Brca2-EGFP FCS Measurements. The mobility of fluo-rescent molecules in solution can be characterized by fitting theACF to appropriate models (12). The simplest diffusion processthat can be described by FCS measurements is random 3D diffu-sion (12, 13) for a single component (Methods). This simple modelfits the ACF of a biologically inert tracer, EGFP (Fig. 2A), but notthe complex diffusion pattern of Brca2-EGFP (Fig. 2 A–C).Nucleoplasmic Brca2-EGFP diffusion may be altered by its

participation in protein complexes or hindered by molecularcrowding, transient intermolecular interactions, or obstructionsto free diffusion created by chromatin and other immobile struc-tures (14). We therefore applied two typical models: the two-component 3D diffusion model, incorporating a second diffusingspecies (12, 15), or alternatively amodel accounting for anomalousor “subdiffusive” processes (16) (Methods). Subdiffusive processesrepresent situations where a particle in a random walk spendsa higher-than-average time at each point before moving on(reflected mathematically as the anomaly parameter), which mayoccur within the nucleoplasm (14). Indeed, the goodness of fit(the residual sum of squares) for FCS measurements from a pop-ulation of Brca2.EGFP/− cells was improved relative to a single-component model by including either anomalous diffusion ora second diffusive species (Fig. S5 A and B).In the first case, fitting of the ACF yields an anomaly pa-

rameter, α = 0.67 ± 0.01, similar to that obtained for othernuclear proteins (14), and a dwell time for anomalous diffusionof τDα = 3.9 ± 0.2 ms (Fig. S5 C and D). However, the stan-dardized residuals for the anomalous diffusion fit show system-atic deviations that are not present when the same data areanalyzed with a two-component model (Fig. 2C). Thus, a singleanomalously diffusing component is insufficient to describeBrca2-EGFP behavior. Indeed, a multicomponent model is morebiologically intuitive for DNA repair proteins like Brca2, knownto form soluble complexes involving different binding partners(17). We used ATP depletion to dissect the contribution ofbinding events to Brca2-EGFP mobility (SI Discussion). ATPdepletion decreases the average time of anomalous diffusionfor Brca2-EGFP, whereas the anomaly parameter itself is un-

A

B

C D E

Fig. 1. Construction and visualization of an endoge-nously tagged fluorescent Brca2 protein. (A) Generationof the Brca2.EGFP/− DT40 cell line. One Brca2 allele wasexcised to create Brca2/− heterozygote cells (11). EGFPwas then targeted to the 3′ end of the wild-type allele togenerate Brca2.EGFP/− cells. (B) The targeting constructs.Primers For-B2GFP and Rev-B2GFP (arrows) amplify byPCR (C) a 4.5-kb band only from the EGFP-targeted allelein Brca2.EGFP/− cells (lane 2) or a 0.6-kb band from thewild-type allele in Brca2/− cells (lane 3). (D) Brca2.EGFP/−cells express a full-length Brca2-EGFP fusion protein(black arrow) detected by anti-GFP or anti-Gg Brca2 asa band with slightly retarded gel mobility comparedwith Brca2 (gray arrow). (E) Brca2.EGFP localization ina living cell imaged with a laser scanning microscope (10-μm optical slice, averaged ×16).

21938 | www.pnas.org/cgi/doi/10.1073/pnas.1009577107 Jeyasekharan et al.

affected. This result suggests that energy-dependent bindingevents occur within the diffusing pool, regardless of the modelchosen (Fig. S6 and SI Discussion). Together, these consid-erations indicate that in our biological system a two-component3D-diffusion model is the best minimal choice for data analysis.

A Pool of Nucleoplasmic Brca2-EGFP with Restricted Mobility. Thetwo-component 3D-diffusion model identifies a faster-diffusingBrca2-EGFP component accounting for ∼50% (47.9 ± 1.2%,mean ± SE, n= 70 cells) of the total pool (Fig. 2D), which has anaveragediffusion coefficient,D=14.8± 1.0μm2/s (Fig. 2E). This is∼2.1-fold lower than that of freeEGFP (D=30± 2 μm2/s, n=22).Although the diffusion coefficient of proteins in vivo cannot bereliably predicted (13, 14), the expected difference in mass be-tween EGFP (27 kDa) and Brca2-EGFP (411 kDa) suggests thatthe fast component can be interpreted as free, or minimally boundBrca2-EGFP protein (SI Discussion). The remaining ∼50% of thediffusing Brca2-EGFP pool is represented by a significantly slowercomponent with an average dwell time of τD= 22.1± 1.4 ms (n=70) (Fig. 2E). The slow component can be interpreted as either thediffusion of gigantic complexes (∼8 GDa) of Brca2-EGFP (D =0.45± 0.03 μm2/s) or, alternatively, transient immobilization of theprotein due to binding events (koff = 1/τD ∼ 45 s−1) and likelyrepresents a combination of complex formation and transientbinding events (Fig. S7 A and B). Satisfactory fitting to the two-component model does not exclude the presence of anomalousdiffusion within the nucleus; further dissection is not currentlypossible by FCS alone. Importantly, however, bothmodels suggest

that a significant fraction of nucleoplasmic Brca2-EGFP mole-cules reside within pools of restricted mobility (Figs. S6 and S7 Aand B and SI Discussion).Fit estimates from the analysis of the FCS measurements show

significant scatter between cells, which could represent biologicaldifferences between cells endogenously expressing Brca2-EGFPunder native control. However, a similar distribution of values isalso observed among multiple readings taken within a single cell,suggesting that our sampling accurately represents the averagebehavior of functional Brca2-EGFP molecules (Fig. S7 C–F).

DNA Damage Mobilizes Brca2-EGFP, but Not Inert Fluorescent Tracers.DNAdamage induces significant changes in the diffusion of Brca2-EGFP, suggestive of its release from pools with restrictedmobility.This conclusion is apparent from a comparison of the populationmean± SEof amplitude-normalizedACFcurves from cells before,and 1–3 h after (Fig. 3A), exposure to 10 Gy of ionizing radiation(IR). The leftward shift reflects a shorter average dwell time forfluorescentmolecules in the observation volume, demonstrating anoverall increase in the diffusional mobility of Brca2-EGFP afterDNA damage. In contrast, two sets of measurements 4 h apart onpopulations of undamaged cells are identical (Fig. 3B). The ob-served ACF change after DNA damage (noting the logarithmicscale used in Fig. 3A) is consistent with a substantial reorganizationof Brca2-EGFP–containing macromolecular complexes in vivo.This change is unlikely to be caused by the mobilization of chro-matin-bound Brca2 because prior studies using classical bio-chemical fractionation show that the overall amount of chromatin-bound Brca2 increases rather than decreases after DNA damage(18). Instead, the statistically significant increase we observe in thefraction of Brca2-EGFP molecules corresponding to the fast-diffusing component (Fig. 3 C and D, t test, P < 0.01) suggestsa reorganizationofBrca2-EGFP–containingmacromolecular com-plexes visible under FCS into simpler forms with higher mobility.Damage-induced mobilization of Brca2-EGFP could arise ei-

ther directly, from alterations in Brca2-EGFP’s physical prop-erties, or indirectly, through alterations in nuclear viscosity (SIMethods). To discriminate between these possibilities, we in-troduced inert molecular tracers to report changes in the nuclearmicroenvironment that might be induced by DNA damage. Nodetectable alterations were found. There is no shift in the meanACF for free EGFP (Fig. 3D), a globular, compact molecule; fora 10-kDa TMR-dextran polymer that exhibits a greater hydro-dynamic radius (SI Methods, Fig. 3E); or for Cy3-labeled oligo(dT) (Fig. 3F), a tracer incorporated into massive ribonucleo-protein complexes of the order of Brca2-EGFP in molecularmass (19). Thus, DNA damage neither triggers global alterationsin nuclear viscosity nor grossly alters the shape of the observationvolume within the cell. This result confirms that damage-inducedBrca2-EGFP mobilization arises from specific changes in itsphysical properties. Collectively (and irrespective of the modelchosen for FCS analysis), our experiments show that DNAdamage changes nucleoplasmic Brca2 molecules, suggestive ofredistribution from restricted mobility to a more mobile state.

Brca2-EGFP Mobilization Coincides with Active HR. Several observa-tions suggest that damage-induced Brca2-EGFP mobilizationcorrelates with active HR. First, the altered mobility of Brca2-EGFP increases over time following IR damage, peaking 3–4 hafterward (Fig. 4A), similar to the time course for the assemblyof HR foci at DSBs (20). Second, Brca2-EGFP mobilizationdoes not occur if cells are exposed to IR during the G1 phase ofthe cell cycle (Fig. 4B), a period when HR foci do not form, andHR may be inactive (21). By contrast, there is a pronouncedmobilization in cells exposed to IR during the S and G2 phases(Fig. 4C), corresponding to the period when HR is most active.Finally, the clastogenic agents VP16 and MMC, which generateDNA lesions that induce HR foci formation and engage the HRrepair machinery (22), also trigger the mobilization of Brca2-EGFP (Fig. 4 D and E, respectively). These findings suggest that

A B

C D

E

Fig. 2. A pool of nucleoplasmic Brca2-EGFP with restricted mobility. (A) Rep-resentativeACFs forBrca2-EGFP (blue squares)or freeEGFP (openblack circles),each fitted to a single-component 3D diffusion model (red lines). (B) Distribu-tionofamplitude-normalizedACFs from70Brca2.EGFP/− cells. Eachcurve is theaverage of ten 5-s readings in a single cell. The residuals from fitting to threedifferent diffusion models are shown in C. Error bars represent the SEM. (D)Distributionof the fast-diffusing Brca2-EGFP component described by the two-component model in 70 Brca2.EGFP/− cells. (E) Comparison of the measureddwell time (τD) of the fast- and slow-diffusing Brca2-EGFP components (graybars) in 70Brca2.EGFP/− cellswith freeEGFP (blackbars;n=30). Thepopulationaverage was used to calculate the apparent D values noted in the text.

Jeyasekharan et al. PNAS | December 14, 2010 | vol. 107 | no. 50 | 21939

SYST

EMSBIOLO

GY

Brca2-EGFP mobilization is a biological response to DNAbreakage that coincides with HR activity during the cell cycle.

Brca2-EGFP Mobilization Is Triggered by Damage-Activated ProteinKinase Activity. Protein kinases of the phosphoinositide-kinase–like (PIK) family, like ATM, ATR, and DNA-PK, signal DSBs toinitiate repair by HR. Preexposure of cells to wortmannin (apotent inhibitor of these kinases) (23) inhibits H2AX phos-phorylation, a known PIK-dependent response to DSBs. Strik-ingly, it also suppresses Brca2-EGFP mobilization after exposureto IR (Fig. 4F), suggesting that this change is initiated by DSB-activated protein kinase activity.Excessive DNA damage may trigger apoptosis. However, ap-

optotic cells marked by propidium iodide incorporation afterexposure to 10 Gy IR (Fig. S8A) show no autocorrelation forBrca2-EGFP (Fig. S8B), suggesting they do not contribute to thedamage-induced mobilization seen by FCS. Moreover, irradia-tion of asynchronous cells with a lower dose of IR, 5 Gy, induceda moderate but statistically significant mobilization of Brca2-

EGFP (Fig. S8C). When specifically analyzing cells synchronizedin the S-G2 phase, where Brca2 activity is maximal, we find thatthese cells exposed to 5 Gy IR exhibit a response similar to thatof 10 Gy irradiated S-G2 cells, although the extent of apoptosiswas considerably lower (Fig. S8 D–G). Together, these resultsargue that damage-induced Brca2-EGFP mobilization is un-related to apoptosis. This conclusion is also supported by theobservation that wortmannin, which enhances IR-induced celldeath (24), nonetheless suppresses mobilization.

Brca2-EGFP Mobilization Is Accompanied by an Increase in Binding toRad51. Collectively, our findings provide several lines of evidencethat DNA damage induces the mobilization of Brca2-EGFP froma slowly diffusing state in the nucleoplasm. To test the functionalsignificance of this biological response, we used FCCS (Fig. S9A)with the well-characterized EGFP/mCherry fluorophore pair (25)to examine the interaction of Brca2-EGFP with its cargo, Rad51,in the nucleoplasm before and after DNA damage. When twopopulations of spectrally distinct fluorophores diffuse together

A B C D

E F G

Fig. 3. DNA damage increases the mobility ofBrca2-EGFP. (A) Mean amplitude-normalizedACF curves for Brca2-EGFP before (black) or 3 hafter (red) exposure of Brca2.EGFP/− cells to 10Gy IR; note the leftward shift of the red curve.Each data point is the average ± SEM of readingsfrom 50 cells. Insets show Brca2-EGFP localiza-tion. (B) Mean amplitude-normalized ACF curves± SEM for Brca2-EGFP from undamaged Brca2.EGFP/− cells at similar time points. (C) Per-cellvalues for the amount of fast-diffusing Brca2-EGFP in a two-component fit (an index of mo-bility derived from the ACF). Each circle repre-sents the average from a single cell, and the redline shows the population mean ± SEM. DNAdamage induces statistically significant changes(t test, P < 0.01, n ∼ 50 cells each). (D) Dot plot ofthe same dataset analyzed by globally linkingthe τD values, which shows a similar difference infast-diffusing Brca2-EGFP. Numbers in paren-theses show the τD (in μs) for the fast and slowBrca2-EGFP components from the linked fit.Mean amplitude-normalized ACF curves ± SEMfor free EGFP (E, n = 20 cells each), for 10 kDa TMR-dextran (F, n = 25), and for Cy3-Oligo(dT) (G, n = 25) are shown before (black) and 1–3 h after (red) 10 Gy IR.

A B C

D E F

Fig. 4. Brca2-EGFPmobilization coincides with active HR. (A)Average amount ± SEM of fast-diffusing Brca2-EGFP in a two-componentfit from15 cells exposed to 10Gy IR (binned at 1-hintervals). Timepoints>6harenot showndue to increased celldeath. (B) Mean amplitude-normalized ACF curves ± SEM forBrca2-EGFP from Brca2.EGFP/− cells arrested in G1 usingmimosinebefore (black) or 2h after (red) exposure to10Gy IR.Inset shows DNA content measured by propidium iodidestaining and flow cytometry in asynchronous (gray), G1-arrested (black), or G1-arrested and IR-exposed (red) cells. Asimilar experiment is depicted in C, except that the cells aresynchronized in S-G2 after release from mimosine arrest. (D)Percentage of fast-diffusing Brca2-EGFP in two-componentfits from Brca2.EGFP/− cells before and after exposure to 20μMEtoposide (VP16) for 2h (P< 0.01,n=20each). (E) A similarplot for G1-synchronized cells released into 100 ng/mL MMCand followed for 6 h (P < 0.01, pre- vs. post-MMC 2 h and 4 h;n ∼ 50 each). Experimental design is depicted below the data.(F) A similar plot for cells before or 3 h after 10 Gy IR, with orwithoutpreexposure to1μMwortmannin for30min (P<0.01,n∼ 25 cells each). Eachdatapoint represents the averageof 10measurements from within a cell, with error bars at the 95%confidence interval. Wortmannin also inhibits γH2AX forma-tion after 10 Gy IR, with β-actin serving as a loading control.

21940 | www.pnas.org/cgi/doi/10.1073/pnas.1009577107 Jeyasekharan et al.

within the observation volume, the amplitude of their cross-cor-relation is directly proportional to the number of interactingmolecules (26). We generated a Brca2.EGFP gene-targeted cellline coexpressing an mCherry-Rad51 fusion protein (which wehave previously shown to functionally replace endogenousRad51) (27). mCherry-Rad51 in Brca2.EGFP cells was stablyexpressed at levels similar to those of the endogenous protein,(Fig. S9B) and colocalized with Brca2-EGFP in damage-inducedHR foci (Fig. S9C). FCCS analysis in these cells (corrected forpossible errors due to background autofluorescence, spectralcross-talk, or chromatic aberrations as previously described in ref.26) (SI Methods) showed a significant cross-correlation signal, incontrast to measurements on Brca2-EGFP and free mCherry(Fig. 5A). To quantify the extent of interaction between Brca2-EGFP and mCherry-Rad51 in its simplest case (SI Methods), wecalculated the ratio of the cross-correlation amplitude to thelower autocorrelation amplitude (Gx/Gr, because in this case theGr <Gg, whereGx is the cross-correlation amplitude, andGg andGr are the autocorrelation amplitudes of EGFP- and mCherry-tagged proteins, respectively) (26). Interestingly, we observea significant increase in the Gx/Gr ratio (P < 0.01; t test, n = 40each) after DNA damage (Fig. 5B). This observation reflects anincrease in the extent of interaction between Brca2-EGFP andmCherry-Rad51 within the diffusible pool of nucleoplasmic pro-teins, accompanying damage-induced Brca2-EGFP mobilizationin this pool. Thus, our results suggest that DNA damage triggersthe redistribution of soluble nucleoplasmic Brca2 molecules froma state of restricted diffusion into a mobile fraction available forRad51 binding.

DiscussionHere, we have combined gene targeting with FCS to study thediffusion of functional, EGFP-tagged Brca2 molecules expressednatively within the nucleoplasm of living cells. In the absence ofendogenous untagged Brca2, FCS measurements account for allfunctional mobile pools of the protein, enabling correlation withthe biological process under study. Our work highlights the useof gene targeting in DT40 cells to study the diffusion of fluo-rescent proteins (28) and provides a template for future FCSstudies on DNA transactions in the cell nucleus.We find that Brca2 molecules exist in complex states of mo-

bility in the nucleoplasm. Their behavior can be interpreted bi-ologically to imply the existence of transient or stable complexes(using the two-component model for 3D diffusion) (8) or as anobstruction to free diffusion through crowding, compartmental-ization, or transient binding events (using the anomalous diffu-sion model) (16). In either case, a significant fraction of Brca2

within the nucleoplasm of living cells resides in a pool with re-stricted mobility (SI Discussion). Strikingly, DNA damage sig-nificantly enhances Brca2 mobility, without evidence of a globalchange in nucleoplasmic diffusion, implying that it triggers theredistribution of Brca2 from slow-moving forms into a moremobile state (Fig. 5C). Several lines of evidence indicate that theobserved changes in Brca2-EGFP diffusion are biologically rel-evant. They are not only statistically significant across severalindependent observations, but also correlate well with the knownbiological activity of Brca2. Moreover, these changes coincidewith active HR and are dependent on signaling through DNAdamage-activated kinases of the PIK family, characteristic ofa specific, signal-initiated cellular response to DNA breakage.Further, damage-induced increases in Brca2 mobility are ac-companied by an increase in binding to its cargo, Rad51, in thenucleoplasm. Thus, release from functional sequestration intoa freely diffusing nucleoplasmic pool may increase the avail-ability of Brca2 for its interaction with Rad51. This change couldinvolve Brca2 release from an intermolecular interaction thatprevents Rad51 binding and/or an increased chance of encounterwith Rad51 molecules in the nucleoplasm.Our findings suggest that Brca2 diffusion, a basic physico-

chemical property of the protein, is dynamically regulated bybinding to partner molecules in the cell nucleus. Biochemicalfractionation of cell lysates shows that a proportion of detergent-extractable Brca2 exists in stable macromolecular complexesof >1 MDa (17). Whether these complexes represent biologicallyrelevant states that exist in living cells is uncertain, but theirexistence is consistent with models for Brca2 diffusion that in-clude a slow-diffusing component. The sheer size of these Brca2-containing complexes, and/or the presence of elements thattransiently tether them to structures like the nuclear matrix,could help explain their restricted mobility. In this light, ourfinding that DNA damage induces the redistribution of Brca2into simpler forms with higher mobility could reflect release fromthis type of functional sequestration. The identification ofbinding partners that mediate sequestration and release will beimportant in understanding the regulation of this critical tumorsuppressor pathway. Potential candidates include the FANCcomplex, PALB2, EMSY, and DSS1 (4), known binding partnersof Brca2 that participate in HR.Regulated protein redistribution across microscopically visible

nuclear compartments has previously been described (29). Thenoninvasive nature of FCS and its selectivity for the solublefraction of a protein have permitted us to reveal a mechanism ofsequestration of a nuclear protein into slow-moving complexeswithin the diffusing pool (rather than, for example, into staticstructures visible under conventional light microscopy), from

A

C

B Fig. 5. Brca2-EGFP mobilization is accompanied byincreased association with mCherry-Rad51 in the nu-cleoplasm. (A) Cross-correlation between Brca2-EGFPand mCherry-Rad51 (black) vs. Brca2-EGFP and freemCherry (red). (B) Cross-correlation index for Brca2-EGFP and mCherry-Rad51, defined as the ratio of cross-correlation to the lower autocorrelation value (Gx /Gr),plotted before (black) and after (blue) exposure to 10Gy IR (P < 0.01, t test, n = 40 each). (C) Signal-initiatedchanges in nucleoplasmic Brca2 diffusion: a hypotheti-cal model. In undamaged cells (Left), soluble nucleo-plasmic Brca2 molecules (blue-crossed yellow circles)visible under FCS can be represented as a fast-diffusingcomponent (Upper Left) or a slow-diffusing fractionshowing restricted mobility (Upper Right). ImmobileBrca2 (e.g., stably bound to DNA) is invisible under FCS(Lower, gray background). DNA damage (Right) mobi-lizes Brca2 from dynamic sequestration in the nucleo-plasm, increasing its availability for binding to Rad51(red triangles, Upper Left and Right). These changescould enhance the delivery of Brca2/Rad51 complexesto positionally stable DSBs (Lower).

Jeyasekharan et al. PNAS | December 14, 2010 | vol. 107 | no. 50 | 21941

SYST

EMSBIOLO

GY

which it can be released, in a regulated fashion, during a bi-ological process. Such a mechanism may be particularly relevantin the case of nuclear proteins that assemble at specific sites onDNA, when diffusion limits the efficiency of DNA scanning (30,31). Thus, the increase in Brca2 mobility may augment theprobability of encounters with immobile DNA breaks (2) withinthe nuclear volume. It is tempting in this light to speculate thatsignal-initiated changes in the diffusion of nucleoplasmic pro-teins may provide a general means to control other essentialDNA transactions, including replication, repair, or transcription.

MethodsFCS Calibration and Acquisition. Fluorescence correlation spectroscopymeasurements were performed using the confocor2 system as describedpreviously (8). The pinhole of the FCS detector was aligned using 10 nMRhodamine 6G (Rh6G) solution excited at 488 nm and imaged over the 505-to 550-nm spectral band. The same setup was used to measure fluctuationsin the fluorescence emission of EGFP. Using a 70-μm pinhole for the FCSdetector, the optimal excitation intensity for Brca2.EGFP cells was estimatedat 75 μW by preliminary experiments. Ten sequential readings of 5–10 s eachwere performed per cell; the first sequence was discarded to exclude arti-facts generated by immediate photobleaching of immobile molecules.Fluctuation traces exhibiting gross deviations in the count rate (suggestiveof cell movement) or a poor signal-to-noise ratio (less than three times thebackground count rate from DT40 nuclei not expressing fluorophore-taggedproteins) were discarded. For FCS analysis of TMR-dextran and Cy3-Oligo(dT), the system was calibrated using a solution of free tetramethyl rhoda-mine (TMR) (10 nM), for an optical setup of 543 nm excitation and 560LP emission.

Cross-Correlation Acquisition. Fluorescence cross-correlation measurementswere performed using the Zeiss Confocor2 system, using an optical ar-rangement for study of EGFP and mCherry—i.e., excitation by 488- and 543-nm lasers, appropriate beamsplitters, and detection bandpass filters 505–530for EGFP [channel (Ch)2] and 600–650 for mCherry (Ch1). System setup andcalibration were performed using a modification of the protocols describedin ref. 7. Briefly, 10 nM Rh6G solution was used to calibrate the optimalposition of the collimator and pinholes for maximum (max) count rate inboth EGFP and mCherry channels. Then, the pinhole for the red channel(Ch1) was adjusted (diameter and x, y, z position), using excitation of Rh6G

by the 488-nm laser, to optimize for maximum overlap. Cellular readingswere performed on Brca2.EGFP Cherry.Rad51 cells, similar to FCS acquisition.A single LSM image (max speed, no averaging) was acquired to check forcells that expressed sufficient quantity of the green and red fluorophore.The crosshair for FCCS measurement was positioned on the basis of thebrightfield image, at the center of the nucleus. Ten 5-s readings wererecorded from each cell.

Data Analysis. The fluorescence fluctuations (δF) are converted within theconfocor software into the autocorrelation function (G(τ)) by the equation

GðτÞ ¼ hδFðtÞ · δFðtþ τÞihFðtÞi2

and into the cross-correlation function (Ggr(τ)) by the equation

GgrðτÞ ¼�δFgðtÞ · δFrðtþ τÞ��FgðtÞ

�hFrðtÞi;

where g and r represent the green and red channels, respectively. The av-erage curves from multiple cells were compiled within the confocor softwareand exported to GraphPad Prism 5.0 for graphical depiction. Models andanalysis regimens used for fitting the autocorrelation and cross-correlationfunctions are discussed in detail in SI Methods.

ACKNOWLEDGMENTS. We thank Drs. S. Takeda (Kyoto University, Kyoto)and K. J. Patel (Medical Research Council Laboratory of Molecular Biology,Cambridge, UK) for providing DT40 reagents and protocols, Drs. K. Sato andM. Lee for technical assistance, Dr. Enrico Gratton (University of California,Irvine, CA) for critical input on the data analysis, and Drs. V. Wickramasingheand M. Garnett for thoughtful comments. A.D.J. thanks Klaus Weisshart(Carl Zeiss GmBH, Jena, Germany) and the instructors of the European Mo-lecular Biology Organization course (advances in high-resolution microscopy;Buenos Aires, 2006) for assistance with FCS methods. A.D.J. received a GatesCambridge Scholarship and a Career Development Fellowship from theUnited Kingdom Medical Research Council, E.R. and R.M. received student-ships from the Medical Research Council and the Wellcome Trust, respec-tively, and A.E. received funding from the Engineering and Physical SciencesResearch Council (EP/F044011/1). The Medical Research Council supportswork in A.R.V.’s laboratory.

1. Friedberg ECWG, Siede W, Wood RD, Schultz RA, Ellenberger T (2006) DNA Repairand Mutagenesis (Am Soc Microbiol, Washington, DC).

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

3. Rippe K (2007) Dynamic organization of the cell nucleus. Curr Opin Genet Dev 17:373–380.

4. Moynahan ME, Jasin M (2010) Mitotic homologous recombination maintains genomicstability and suppresses tumorigenesis. Nat Rev Mol Cell Biol 11:196–207.

5. Magde D, Elson EL, Webb WW (1974) Fluorescence correlation spectroscopy. II. Anexperimental realization. Biopolymers 13:29–61.

6. Brock R, Jovin TM (1998) Fluorescence correlation microscopy (FCM)—fluorescencecorrelationspectroscopy(FCS) takenintothecell.CellMolBiol (Noisy-le-grand)44:847–856.

7. Schwille P, Meyer-Almes FJ, Rigler R (1997) Dual-color fluorescence cross-correlationspectroscopy for multicomponent diffusional analysis in solution. Biophys J 72:1878–1886.

8. Bacia K, Schwille P (2003) A dynamic view of cellular processes by in vivo fluorescenceauto- and cross-correlation spectroscopy. Methods 29:74–85.

9. Ayoub N, et al. (2009) The carboxyl terminus of Brca2 links the disassembly of Rad51complexes to mitotic entry. Curr Biol 19:1075–1085.

10. YamazoeM, SonodaE, HocheggerH, TakedaS (2004) Reversegenetic studies of theDNAdamageresponse inthechickenB lymphocyte lineDT40.DNARepair (Amst)3:1175–1185.

11. Schmiedeberg L, Weisshart K, Diekmann S, Meyer Zu Hoerste G, Hemmerich P (2004)High- and low-mobility populations of HP1 in heterochromatin of mammalian cells.Mol Biol Cell 15:2819–2833.

12. Schwille P, Haupts U, Maiti S, Webb WW (1999) Molecular dynamics in living cellsobserved by fluorescence correlation spectroscopy with one- and two-photonexcitation. Biophys J 77:2251–2265.

13. Pack C, Saito K, Tamura M, Kinjo M (2006) Microenvironment and effect of energydepletion in the nucleus analyzed by mobility of multiple oligomeric EGFPs. Biophys J91:3921–3936.

14. Bancaud A, et al. (2009) Molecular crowding affects diffusion and binding of nuclearproteins in heterochromatin and reveals the fractal organization of chromatin. EMBOJ 28:3785–3798.

15. Digman MA, Brown CM, Horwitz AR, Mantulin WW, Gratton E (2008) Paxillindynamics measured during adhesion assembly and disassembly by correlationspectroscopy. Biophys J 94:2819–2831.

16. Weiss M, Hashimoto H, Nilsson T (2003) Anomalous protein diffusion in living cells asseen by fluorescence correlation spectroscopy. Biophys J 84:4043–4052.

17. Marmorstein LY, et al. (2001) A human BRCA2 complex containing a structural DNAbinding component influences cell cycle progression. Cell 104:247–257.

18. WangX,AndreassenPR,D’AndreaAD(2004)Functional interactionofmonoubiquitinatedFANCD2 and BRCA2/FANCD1 in chromatin.Mol Cell Biol 24:5850–5862.

19. Politz JC, Browne ES, Wolf DE, Pederson T (1998) Intranuclear diffusion andhybridization state of oligonucleotides measured by fluorescence correlationspectroscopy in living cells. Proc Natl Acad Sci USA 95:6043–6048.

20. Scully R, et al. (1997) Dynamic changes of BRCA1 subnuclear location andphosphorylation state are initiated by DNA damage. Cell 90:425–435.

21. Tashiro S, et al. (1996) S phase specific formation of the human Rad51 protein nuclearfoci in lymphocytes. Oncogene 12:2165–2170.

22. Helleday T, Lo J, van Gent DC, Engelward BP (2007) DNA double-strand break repair:From mechanistic understanding to cancer treatment. DNA Repair (Amst) 6:923–935.

23. Sarkaria JN, et al. (1998) Inhibition of phosphoinositide 3-kinase related kinases bythe radiosensitizing agent wortmannin. Cancer Res 58:4375–4382.

24. Price BD, Youmell MB (1996) The phosphatidylinositol 3-kinase inhibitor wortmanninsensitizes murine fibroblasts and human tumor cells to radiation and blocks inductionof p53 following DNA damage. Cancer Res 56:246–250.

25. Maeder CI, et al. (2007) Spatial regulationof Fus3MAPkinaseactivity througha reaction-diffusion mechanism in yeast pheromone signalling. Nat Cell Biol 9:1319–1326.

26. Bacia K, Kim SA, Schwille P (2006) Fluorescence cross-correlation spectroscopy in livingcells. Nat Methods 3:83–89.

27. Rajendra E, Venkitaraman AR (2010) Two modules in the BRC repeats of BRCA2mediate structural and functional interactions with the RAD51 recombinase. NucleicAcids Res 38:82–96.

28. Lippincott-Schwartz J, Patterson GH (2003) Development and use of fluorescentprotein markers in living cells. Science 300:87–91.

29. Gorski SA, Dundr M, Misteli T (2006) The road much traveled: Trafficking in the cellnucleus. Curr Opin Cell Biol 18:284–290.

30. Stanford NP, Szczelkun MD, Marko JF, Halford SE (2000) One- and three-dimensionalpathways for proteins to reach specific DNA sites. EMBO J 19:6546–6557.

31. Elf J, Li GW, Xie XS (2007) Probing transcription factor dynamics at the single-molecule level in a living cell. Science 316:1191–1194.

21942 | www.pnas.org/cgi/doi/10.1073/pnas.1009577107 Jeyasekharan et al.