Stabilization of mutant BRCA1 protein confers PARPinhibitor and platinum resistanceNeil Johnsona,b,1,2, Shawn F. Johnsona, Wei Yaoa, Yu-Chen Lia, Young-Eun Choic, Andrea J. Bernhardyd, Yifan Wangd,Marzia Capellettia, Kristopher A. Sarosieka, Lisa A. Moreauc,e, Dipanjan Chowdhuryc, Anneka Wickramanayakef,Maria I. Harrellf, Joyce F. Liua,b, Alan D. D’Andreac,e, Alexander Mirong, Elizabeth M. Swisherf,and Geoffrey I. Shapiroa,b,2

Departments of aMedical Oncology, cRadiation Oncology, and gCancer Biology, Dana–Farber Cancer Institute and Harvard Medical School, Boston, MA02215; bDepartment of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA 02215; dDevelopmental Therapeutics Program,Fox Chase Cancer Center, Philadelphia, PA 19111; eDepartment of Pediatrics, Children’s Hospital and Harvard Medical School, Boston, MA 02215;and fDepartments of Obstetrics and Gynecology and Medicine, University of Washington, Seattle, WA 98195

Edited by Stephen J. Elledge, Harvard Medical School, Boston, MA, and approved August 26, 2013 (received for review March 18, 2013)

Breast Cancer Type 1 Susceptibility Protein (BRCA1)-deficient cellshave compromised DNA repair and are sensitive to poly(ADP-ribose)polymerase (PARP) inhibitors. Despite initial responses, the devel-opment of resistance limits clinical efficacy. Mutations in the BRCA C-terminal (BRCT) domain of BRCA1 frequently create protein productsunable to fold that are subject to protease-mediated degradation.Here, we show HSP90-mediated stabilization of a BRCT domainmutant BRCA1 protein under PARP inhibitor selection pressure. Thestabilized mutant BRCA1 protein interacted with PALB2-BRCA2-RAD51, was essential for RAD51 focus formation, and conferredPARP inhibitor as well as cisplatin resistance. Treatment of resistantcells with the HSP90 inhibitor 17-dimethylaminoethylamino-17-demethoxygeldanamycin reduced mutant BRCA1 protein levelsand restored their sensitivity to PARP inhibition. Resistant cells alsoacquired a TP53BP1 mutation that facilitated DNA end resection inthe absence of a BRCA1 protein capable of binding CtIP. Finally,concomitant increased mutant BRCA1 and decreased 53BP1 proteinexpression occur in clinical samples of BRCA1-mutated recurrentovarian carcinomas that have developed resistance to platinum.These results provide evidence for a two-event mechanism bywhichBRCA1-mutant tumors acquire anticancer therapy resistance.

homologous recombination | cancer therapy

The breast cancer 1, early onset (BRCA1) gene is commonlymutated in hereditary breast and ovarian cancers. The BRCA1

protein has multiple domains that mediate protein interactions;BRCA1 gene mutations may produce truncated proteins that losethe ability to interact with associated proteins. Additionally,mutations in the BRCA C-terminal (BRCT) domain of BRCA1create protein folding defects that result in protease-mediateddegradation (1–3).Cells that contain dysfunctional BRCA1 proteins are hyper-

sensitive to DNA damaging agents (4). In particular, BRCA1-deficient cell lines are exquisitely sensitive to poly(ADP-ribose)polymerase (PARP) inhibition (5). Despite initial responses ofBRCA1-mutant cancers to PARP inhibitor treatment (6), acquiredresistance universally develops. Resistance may result from sec-ondary mutations in the BRCA1 gene that restore the readingframe and produce a functional BRCA1 protein (7, 8). In Brca1-mutated mouse mammary tumors, activation of p-glycoprotein orloss of p53 binding protein 1 (53BP1) expression resulting fromtruncating TP53BP1 mutations confers PARP inhibitor resistance(9). Loss of 53BP1 in BRCA1-deficient cells provides the C-terminal binding protein interacting protein (CtIP) with un-restricted access to DNA breaks, facilitating DNA end resection,an early step in homologous recombination (HR) (9–11).Following BRCA1-CtIP–mediated activation of DNA end

resection, eventual BRCA2-mediated assembly of the RAD51recombinase in nucleoprotein filaments is a critical step in HR. Arole for BRCA1 in RAD51 loading and the mechanisms by whichit participates have not been fully clarified. Of note, in PARP

inhibitor-resistant BRCA1- and 53BP1-deficient tumors andderived cell lines, RAD51 γ-irradiation–induced foci were detec-ted, although at a lower level than in BRCA1- and 53BP1-proficientcells (9). Previous studies demonstrated that RAD51 foci werepartially reduced in BRCA1- or partner and localizer of BRCA2(PALB2)-deficient cells reconstituted with BRCA1 or PALB2constructs carrying mutations that disrupt the BRCA1–PALB2 in-teraction (12, 13), suggesting that BRCA1 may enlist PALB2, whichin turn organizes the recruitment of BRCA2 and RAD51.To date, the described mechanisms of PARP inhibitor re-

sistance occur in only a fraction of the BRCA1 mutant patientpopulation or in PARP inhibitor-resistant Brca1-mutated mousemammary tumors (8, 10). Here, we used a human breast cancercell line that contains a BRCT domain BRCA1 mutation toidentify additional mechanisms of acquired PARP inhibitor re-sistance, and demonstrate that stabilization of the mutant BRCA1protein is critical for the restoration of RAD51 focus formation.

ResultsMDA-MB-436 Clones Are Resistant to PARP Inhibitors and Cisplatin.To study PARP inhibitor resistance, we cultured the triple-

Significance

Poly(ADP-ribose) polymerase (PARP) inhibitors have pro-duced responses in homologous recombination (HR) repair-deficient cancers, such as those with a mutated breast cancer 1,early onset (BRCA1) gene. We have delineated a two-eventmechanism of acquired resistance by using a BRCA1 BRCA C-terminal (BRCT) domain-mutated breast cancer cell line, in-volving heat shock protein (HSP)90-mediated stabilization ofthe mutant protein coupled with tumor protein p53 bindingprotein 1 (TP53BP1) gene mutation, which together restoreDNA end resection and RAD51 filament formation, critical stepsin HR. Similar events may occur in primary BRCA1-mutatedovarian cancers as cells develop resistance to platinum. Thedata demonstrate that, even though BRCA1 BRCT domainmutant proteins cannot promote DNA end resection, they re-tain partial function and can contribute to RAD51 loading andHR. Finally, HSP90 inhibition may prove useful for resensitizingresistant BRCA1-mutant cancer cells to drug treatment.

Author contributions: N.J., E.M.S., and G.I.S. designed research; N.J., S.F.J., W.Y., Y.-C.L., Y.-E.C.,A.J.B., Y.W., M.C., K.A.S., L.A.M., A.W., M.H., J.F.L., and A.M. performed research; N.J., D.C.,A.D.D., A.M., E.M.S., and G.I.S. analyzed data; and N.J. and G.I.S. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1Present address: Developmental Therapeutics Program, Fox Chase Cancer Center,Philadelphia, PA 19111.

2To whom correspondence may be addressed. E-mail: neil.johnson@fccc.edu orgeoffrey_shapiro@dfci.harvard.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1305170110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1305170110 PNAS | October 15, 2013 | vol. 110 | no. 42 | 17041–17046

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

June

11,

202

0

negative breast cancer cell line MDA-MB-436 in the presence ofthe PARP inhibitor rucaparib. MDA-MB-436 cells contain aBRCA1 5396 + 1G>Amutation in the splice donor site of exon 20that results in a BRCT domain-truncated protein (14). Drug-resistant clones, labeled rucaparib-resistant (RR) 1 through 6,emerged ∼2 to 4 mo after initial exposure. Clones were highlyresistant to rucaparib, and cross-resistant to olaparib, as well ascisplatin (Fig. 1A). Concentrations required to reduce colonyformation by 50% (lethal concentration 50, LC50) were 482- to590-fold (P < 0.0001), 254- to 492-fold (P < 0.0001), 150- to 173-fold (P < 0.0001), and 27- to 59-fold (P = 0.0056) greater thanthose for parental cells for rucaparib, rucaparib after a 6-moholiday from rucaparib selection, olaparib, and cisplatin, respec-tively. Additionally, MDA-MB-436–resistant clones had a marked

decrease in the number of aberrant chromosome structures aftertreatment with rucaparib compared with the parental cell line,with 10- to 20-fold (P < 0.0001) and 7- to 15-fold (P < 0.0001)fewer aberrations and radials per cell, respectively (Fig. 1B).To rule out drug efflux as a mechanism of PARP inhibitor

resistance, we measured the ability of rucaparib to inhibit thePARP enzyme by assessing cellular poly(ADP-ribose) (PAR)levels by Western blot in the absence of activated DNA. Ruca-parib reduced the levels of PAR to a similar degree in MDA-MB-436 parental cells and in all the resistant clones except forRR-1 (Fig. S1A). Of note, clones RR-5 and RR-6 had reducedbasal PAR levels. To assess if the lack of PARP inhibition in RR-1 cells accounted for drug resistance, we used siRNA to depletePARP-1 and PARP-2 levels. PAR levels were reduced aftersiRNA treatment (Fig. S1B); however, the colony forming po-tential of RR-1 cells was not significantly impacted (Fig. S1C).

Fig. 1. MDA-MB-436 clones are resistant to PARP inhibitors and cisplatin.(A) MDA-MB-436 clones RR-1 to RR-6 were significantly more resistant torucaparib than parental cells (red curve). Cells cultured in the absence ofdrug for 6 mo remained resistant to rucaparib (+6 mo). Cells were also cross-resistant to olaparib and cisplatin, as measured by colony formation assay(n = 3, mean ± SEM of colonies formed relative to DMSO-treated cells).(B) Metaphase spread analyses of chromosome aberrations and radial for-mations after treatment with rucaparib (1 μM) for 24 h (n = 3, mean ± SEM).(Inset) Representative metaphase spreads.

Fig. 2. Mutant BRCA1 protein is abundant in MDA-MB-436 resistant clones.(A) BRCA1, RAD51, histone H3, and tubulin levels were measured in cyto-plasmic (marked as “c”) and nuclear (marked as “n”) extracts from MCF7cells, MDA-MB-436 parental cells and resistant clones RR-1 to RR-6 byWestern blot. (B) MCF7 cells, MDA-MB-436 parental cells and resistant clonesRR-1, RR-5, and RR-6 were treated with DMSO (−) or 1 μM rucaparib (+)for 24 h, and BRCA1 protein levels were assessed by using BRCA1 N- or C-terminal–specific antibodies by Western blot. (C) Detection of BRCA1,RAD51, γ-H2AX, and DAPI by immunofluorescence in MDA-MB-436 parentaland resistant cells (n = 3, mean ± SEM percentage of cells containing morethan five foci). (Inset) Representative cells.

17042 | www.pnas.org/cgi/doi/10.1073/pnas.1305170110 Johnson et al.

Dow

nloa

ded

by g

uest

on

June

11,

202

0

We conclude that, although rucaparib did not inhibit PARP aseffectively in RR-1 cells, additional events may have contributedto rucaparib resistance.

Increased Mutant BRCA1 Protein in Resistant Clones. We next mea-sured BRCA1 and RAD51 protein levels by Western blot. MCF7cells express WT BRCA1 protein and were used as a positivecontrol. Mutant BRCA1 protein was undetectable in MDA-MB-436 parental cells, but was abundant in resistant clones. RAD51protein levels were similar in parental cells and resistant clones(Fig. 2A). To determine whether BRCA1 reversion mutation hadoccurred, we sequenced BRCA1 gene introns and exons. MDA-MB-436–resistant clones retained the original 5396+1G>A mu-tation, and did not harbor any additional mutations in BRCA1(Fig. S2A). Furthermore, the BRCA1 mRNA sequences of pa-rental cells and resistant clones were identical (Fig. S2 B and C).The BRCA1 protein detected in resistant clones by an N-terminal BRCA1 antibody was C-terminal truncated and conse-quently not recognized by a C-terminal–specific antibody (Fig.2B). The BRCA1 5396+1G>A mutation produces two splicevariants (14). We used siRNAs specific to each isoform to de-termine which variant accounted for the reexpressed protein.MDA-MB-436 parental cells stably expressing an exogenous WTBRCA1 protein (MDA-MB-436+WT) were used as a control fornonspecific BRCA1 protein knockdown. We demonstrated thatsiRNA specifically targeting the exon 20 deletion variant resultedin knockdown of mutant but not WT BRCA1 protein. Therefore,it is likely that the exon 20 deletion variant accounted for thereexpressed protein in resistant clones (Fig. S2 D and E).The mutant BRCA1 protein could be detected in association

with chromatin (Fig. S3). As expected, γ-irradiation–inducedBRCA1 foci were not detectable in MDA-MB-436 parental cells;in contrast, BRCA1 foci were readily detectable in resistantclones. Similarly, RAD51 foci were not detected in parentalcells, despite the abundance of RAD51 protein; however, re-sistant clones readily formed RAD51 foci following irradiation.Formation of γ-H2AX foci, a marker of DNA damage, was presentto the same degree in parental and resistant cells (Fig. 2C).

Protein Stability Accounts for Increased Mutant BRCA1 Protein. Wenext investigated factors that could contribute to changes inBRCA1 protein levels in PARP inhibitor-resistant clones. Therewere no changes in BRCA1 gene copy number (Fig. S4A); ad-ditionally, resistant clones demonstrated only a 1.5- to 2.7-fold(P = 0.0061) increase in BRCA1 mRNA by quantitative RT-PCRanalyses (Fig. S4B). To determine if increased BRCA1 proteinexpression was dependent on transcription or translation, wetreated parental and resistant clones with cycloheximide to in-hibit protein translation. We could detect a faint BRCA1 band inMDA-MB-436 parental cells when we increased protein loadingand film exposure time; however, BRCA1 protein was undetect-able at 6 h after cycloheximide treatment. In contrast, BRCA1protein levels were maintained for as long as 24 h after cyclo-heximide treatment in RR clones (Fig. S4C). These data suggestthat the increase in mutant BRCA1 protein in resistant cells waslikely a result of protein stabilization rather than hyperactivation ofBRCA1 transcription or translation.BRCT domain mutations often result in an inability of the

mutant protein to fold correctly; consequently, the unfoldedprotein is more susceptible to protease-mediated degradation(1–3). It is therefore possible that the mutant BRCA1 protein inMDA-MB-436 parent cells is undetectable as a result of an in-ability to correctly fold, with subsequent degradation by theproteasome. Consistent with this hypothesis, MDA-MB-436 pa-rental cells treated with the proteasome inhibitors MG132 orbortezomib had detectable levels of mutant BRCA1 protein,suggesting protein was being generated but rapidly degraded dueto folding defects (Fig. S4D).

HSP90 Stabilizes Mutant BRCA1 Protein. Because HSP90 has beenimplicated in the folding of cancer-related mutant proteins (15),

we investigated the dependency of BRCA1 mutant protein levelson HSP90 activity. First, we assessed the association of BRCA1proteins with HSP90 by determining levels of BRCA1 proteinin HSP90 immunoprecipitates from MDA-MB-436+WT cellsor PARP inhibitor-resistant clones. Mutant and ectopically ex-pressed WT BRCA1 protein from the parental cell line wereabsent or weakly in complex with HSP90. In contrast, mutantBRCA1 protein from resistant clones was readily found inassociation with HSP90 (Fig. 3A). Similarly, when we immuno-precipitated BRCA1 from MDA-MB-436+WT cells or RR cells,HSP90 could only be found in association with the mutantBRCA1 proteins (Fig. 3B). Next, we treated MDA-MB-436+WTBRCA1 cells, as well as RR cells, with the HSP90 inhibitor17-dimethylaminoethylamino-17-demethoxygeldanamycin (17-DMAG). WT BRCA1 protein remained at levels comparable tountreated cells at 72 h posttreatment. In contrast, RR cells hadreduced mutant BRCA1 protein levels by 48 h posttreatment.HSP70 levels increased in response to 17-DMAG, indicating

Fig. 3. HSP90 stabilizes mutant BRCA1. (A) HSP90 was immunoprecipitatedfrom MDA-MB-436 control (GFP) cells, MDA-MB-436+WT cells, and RR clones 1to 6, and HSP90 and BRCA1 protein levels were analyzed byWestern blot (WCE,whole cell extract). (B) BRCA1 was immunoprecipitated from MDA-MB-436control (GFP) cells, MDA-MB-436+WT cells, and RR clones 1 to 3, and BRCA1 andHSP90 protein levels were analyzed by Western blot. (C) MDA-MB-436+WT,RR-1, RR-5, and RR-6 were treated with 100 nM 17-DMAG for the indicatedtimes, and BRCA1, HSP70, and tubulin protein levels were measured by West-ern blot. (D) RR-1, RR-5, and RR-6 were treated with vehicle (marked as “V”) or50 nM 17-DMAG (marked as “D”) in the presence of vehicle (marked as “V”) or100 nM rucaparib (marked as “R”), and colony formation was assessed (n = 3,mean ± SEM of colonies formed relative to vehicle + vehicle-treated cells).

Johnson et al. PNAS | October 15, 2013 | vol. 110 | no. 42 | 17043

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

June

11,

202

0

HSP90 was inhibited to an equal degree in all cell lines (Fig. 3C).Furthermore, 17-DMAG treatment of resistant clones restoredsensitivity to rucaparib; compared with DMSO/rucaparib, clonalsurvival of RR-1, RR-5, and RR-6 in 17-DMAG/rucaparib wasreduced 4.7-fold (P < 0.0001), 13.1-fold (P = 0.0007), and 4.9-fold (P = 0.0023), respectively (Fig. 3D). Together, these datasuggest that HSP90 promotes mutant BRCA1 protein foldingand conformational stability in RR cells. Of note, 17-DMAGtreatment also sensitized MDA-MB-436+WT cells to rucaparibtreatment (Fig. S5A), likely mediated through a reduction inBRCA2 and RAD51 protein levels (16) (Fig. S5B).

Reduced 53BP1 Facilitates BRCA1-Independent DNA End Resection.We next analyzed the contribution of stabilized mutant BRCA1in RR cells to two critical steps of HR, DNA end resection andRAD51 loading. First, we investigated the ability of exogenousWT BRCA1 in parental cells and the C-terminal truncatedmutant BRCA1 protein in resistant clones to interact withproteins known to complex with BRCA1. Analyses of immuno-precipitated exogenous WT BRCA1 protein from MDA-MB-436+WT cells demonstrated that BARD1, PALB2, BRCA2, RAD51,CtIP, and RAP80 could all be detected in association with WTBRCA1. Similarly, BARD1, PALB2, BRCA2, and RAD51associated with endogenous mutant BRCA1 protein immu-noprecipitated from resistant clones. However, the BRCT do-main interacting proteins CtIP and RAP80 were not found tointeract with the C-terminal truncated BRCA1 protein fromMDA-MB-436 resistant cells (Fig. S6A).DNA end resection is dependent on the activities of BRCA1

and CtIP proteins (17, 18). We investigated the role of the mu-tant BRCA1 protein in DNA end resection by measuring theformation of RPA32 foci after γ-irradiation (Fig. S6 B and C).MCF7 and MDA-MB-436+WT cells express WT BRCA1 pro-tein and were used for comparison with mutant BRCA1 pro-teins. Depletion of WT BRCA1 by using three individualsiRNAs resulted in a fourfold (P = 0.0001) and 3- to 15-fold (P =0.0011) decrease in the formation of RPA32 foci compared withscrambled siRNA control-treated MCF7 and MDA-MB-436+WT cells, respectively. In contrast, depletion of mutant BRCA1protein from RR clones RR-1 and RR-5 did not affect theformation of RPA32 foci. Depletion of CtIP by using threeindividual siRNAs resulted in a three- to fivefold (P < 0.0001), 3-to 15-fold (P = 0.0042), 5- to 21-fold (P < 0.0001), and 13- to25-fold (P < 0.0001) decrease in the formation of RPA32 focicompared with scrambled siRNA control-treated MCF7, MDA-MB-436+WT, RR-1, and RR-5 cells, respectively (Fig. S6C).These data indicate that the truncated C-terminal BRCA1 pro-tein that did not interact with CtIP did not influence DNA endresection, and that CtIP activates DNA end resection independentlyof BRCA1 interaction in MDA-MB-436 resistant clones.TP53BP1 mutation and loss of function have been demon-

strated to confer PARP inhibitor resistance (9–11). We se-quenced TP53BP1 gene introns and exons in parental cells andresistant clones. MDA-MB-436 parental cells contained homo-zygous WT TP53BP1 gene sequences. In contrast, all resistantclones contained a heterozygous 3708 del 11 mutation locatedin exon 18 (Fig. S7A). This was a microhomology-mediated de-letion (Fig. S7B), a mechanism of deletion common to BRCA1/2-mutant cancers (19). The mutation creates a frameshift and anearly stop codon predicted to produce a truncated protein (p.P1235PfsX37). Consistent with a heterozygous TP53BP1 geneloss-of-function mutation, PARP inhibitor-resistant clones hadlower levels of 53BP1 protein compared with parental cells (Fig.S7C). 53BP1 protein was reduced four- to sevenfold comparedwith parental cells (by densitometry), a greater reduction thanexpected from loss of one allele, suggesting possible transcrip-tional silencing of the remaining WT allele.To investigate if loss of function of 53BP1 accounted for

PARP inhibitor resistance, we depleted 53BP1 from parentalMDA-MB-436 cells by using siRNA and measured PARP in-hibitor sensitivity. Consistent with a previous report (20), siRNA-

mediated depletion of 53BP1 conferred only a slight degree ofresistance to PARP inhibitor treatment in MDA-MB-436 cells(3.4-fold increase in rucaparib LC50 value vs. scrambled siRNAtreatment; P = 0.1295, unpaired t test; Fig. S7D). In contrast,ectopic expression of WT BRCA1 (MDA-MB-436+WT) con-ferred substantial rucaparib resistance (426-fold increase in ruca-parib LC50 value compared with GFP control cells; P < 0.0001,unpaired t test; Fig. S7E), similar to that seen in our RR clones(Fig. 1A). These data indicate that disruption of 53BP1 functionalone could not fully account for the resistance acquired by theMDA-MB-436 clones derived under rucaparib selection pressure.Because 53BP1 deletion was previously shown to provide

PARP inhibitor resistance in mouse Brca1 mutant cell lines, wefurther investigated the effect of 53BP1 depletion on additionalhuman BRCA1 mutated cancer cell lines, including SUM1315(185delAG) and HCC1395 (5251C>T). Consistent with the datain MDA-MB-436 cells, siRNA mediated-depletion of 53BP1 inSUM1315 and HCC1395 cells conferred a 5.1-fold and 5.7-foldincrease in rucaparib LC50 value compared with scrambledsiRNA treatment (P = 0.145 and P = 0.083, unpaired t test),respectively (Fig. S7 F and G).We hypothesized that the reduction in 53BP1 protein levels in

PARP inhibitor-resistant clones enabled CtIP to activate DNAend resection and RPA32 loading in the absence of CtIP–BRCA1 protein interaction. We demonstrated that a twofoldincrease (by densitometry) in 53BP1 protein levels in RR-1 cellsengineered to express ectopic WT 53BP1 resulted in a 1.5-fold(P = 0.005, unpaired t test) and 1.3-fold (P = 0.025, unpairedt test) reduction in the percentage of RPA32 and RAD51 foci-positive cells compared with control RR-1 cells, respectively(Fig. S8A). Furthermore, reexpression of 53BP1 increased RR-1sensitivity to PARP inhibitor treatment with a twofold decrease(P = 0.049, unpaired t test) in the LC50 value of rucaparibcompared with control cells (Fig. S8B).

RAD51 Focus Formation Is Dependent on Mutant BRCA1. To de-termine why decreased 53BP1 protein levels conferred onlymodest PARP inhibitor resistance in MDA-MB-436 cells, westudied RAD51 assembly following DNA damage in these cells.Of note, RNF8 and RNF168 have been implicated in RAD51loading during HR in the absence of BRCA1 and 53BP1 (21).However, levels of these proteins remained unchanged in re-sistant clones (Fig. S8C).We measured the effect of 53BP1 depletion on RPA32 and

RAD51 foci after γ-irradiation treatment in MDA-MB-436 cellsengineered to express GFP control or exogenous WT BRCA1(Fig. 4A). Depletion of 53BP1 increased RPA32 foci 3.4-fold(P < 0.001) and 4.9-fold (P < 0.001) in MDA-MB-436 control (+GFP) and MDA-MB-436+WT cells, respectively. Therefore, thepresence or absence of BRCA1 protein did not affect the in-crease in formation of RPA32 foci after 53BP1 depletion. Incontrast, depletion of 53BP1 resulted in a 3.3-fold increase (P <0.001) in RAD51 foci in MDA-MB-436+WT cells that containedWT BRCA1 protein; however, RAD51 foci remained completelyabsent in MDA-MB-436 control cells. Similar to MDA-MB-436+WT cells, depletion of 53BP1 in MCF7 cells expressing en-dogenous WT BRCA1 also resulted in a 2.2-fold (P < 0.005) and2.4-fold (P < 0.001) increase in RPA32 and RAD51 focus for-mation, respectively (Fig. S8 D and E). Therefore, in MDA-MB-436 cells that lack BRCA1 protein, reducing 53BP1 proteinlevels enabled CtIP to activate DNA end resection and enhanceRPA32 loading, but did not facilitate efficient RAD51 recruit-ment. Consequently, 53BP1 depletion did not afford dramaticPARP inhibitor resistance in MDA-MB-436 parental cells, incontrast to the more substantial effects of 53BP1 depletion inother model systems (10, 11).MDA-MB-436 resistant clones readily form RAD51 foci (Fig.

2C). We therefore investigated the role of the mutant BRCA1protein in restoring RAD51 focus formation. Depletion of BRCA1by using three individual siRNAs abolished formation of RAD51foci in MDA-MB-436 resistant clones, indicating that recruitment

17044 | www.pnas.org/cgi/doi/10.1073/pnas.1305170110 Johnson et al.

Dow

nloa

ded

by g

uest

on

June

11,

202

0

of RAD51 following DNA damage was dependent on mutantBRCA1 protein. siRNA-mediated depletion of BRCA1 had noeffect on the formation of γ-H2AX foci or RAD51 protein levels(Fig. 4B). Additionally, siRNA-mediated BRCA1 depletiondramatically resensitized resistant clones, but not already-sensitiveparental cells, to PARP inhibitor treatment. Following expres-sion of BRCA1 siRNA, rucaparib LC50 values were reduced 1- to3.6-fold (P = 0.3), 132- to 175-fold (P < 0.0001), 69- to 153-fold(P < 0.0001), and 33- to 115-fold (P < 0.0001) compared withscrambled siRNA treatment in MDA-MB-436 parental, RR-1,RR-5, and RR-6 cells, respectively (Fig. 4C).

Increased Mutant BRCA1 Protein in Human Cancers. We assesseda panel of carboplatin-treated recurrent BRCA1 mutant ovariancarcinomas for platinum sensitivity, secondary BRCA1 reversionmutations, and increased BRCA1 and decreased 53BP1 staining(Table S1). Increased BRCA1 protein expression in the absenceof BRCA1 reversion mutation occurred in two of four tumorscarrying BRCT domain mutations (5382insC to 5622C>T). Similarto RR MDA-MB-436 cells, the recurrent carcinoma from patient

149101 had an increase in BRCA1 staining combined with a re-duction in 53BP1 staining, and was resistant to platinum (Fig. 5).

DiscussionMutations in the BRCT domains of BRCA1 often preventproper protein folding, and misfolded proteins are subject toprotease-mediated degradation (1–3). In the present study, weshow that, under PARP inhibitor selection pressure, HSP90interacts with and stabilizes mutant BRCA1 proteins. The sta-bilized C-terminal truncated protein is semifunctional, as it isunable to interact with CtIP, but retains the protein domainsnecessary to mediate interactions with PALB2-BRCA2-RAD51(12, 22). Importantly, the mutant BRCA1 protein is capable ofpromoting RAD51 loading onto DNA following DNA damage.Because the BRCT domain-deficient BRCA1 protein is in-

capable of interacting with CtIP, cells require further geneticadaptations to survive selection pressure. In our model, PARPinhibitor-resistant MDA-MB-436 cells had reduced 53BP1 proteinlevels as a result of a heterozygous loss-of-function mutation, anevent that provided CtIP with unrestricted access to DNA ends

Fig. 4. Mutant BRCA1 protein promotes RAD51 focus formation. (A) MDA-MB-436 control (GFP) and MDA-MB-436+WT BRCA1 cells (WT) were treated withscrambled (Sc) or 53BP1 siRNA and fixed 6 h after γ-irradiation. RPA32 and RAD51 foci were detected by immunofluorescence. (Left) Western blot dem-onstrating 53BP1 knockdown and images of representative DAPI-stained cells. (Right) Quantification of RPA32 and RAD51 foci (n = 3, mean ± SEM per-centage of cells containing more than five foci). (B) MCF7 and MDA-MB-436 resistant clones RR-1, RR-5, and RR-6 were left untreated (−) or treated withscrambled (Sc) control or three individual BRCA1 siRNAs. After 72 h, cells were fixed 6 h after γ-irradiation treatment. (Left) BRCA1 and RAD51 proteinknockdown measured by Western blot and images of representative RR-1 cells after detection of BRCA1, RAD51, γ-H2AX and DAPI by immunofluorescence.(Right) Quantification of foci positive cells (n = 3, mean ± SEM percentage of cells containing more than five foci). (C) MDA-MB-436 parental cells or resistantclones were treated with scrambled (Sc) or three individual BRCA1 siRNAs, exposed to increasing concentrations of rucaparib for 72 h and replated for colonyformation. Colony formation was calculated as in Fig. 1A (n = 3, mean ± SEM of colonies formed relative to DMSO-treated cells).

Johnson et al. PNAS | October 15, 2013 | vol. 110 | no. 42 | 17045

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

June

11,

202

0

and facilitated BRCA1-independent DNA end resection (10, 11).Consistent with other studies, 53BP1 depletion alone contributedto PARP inhibitor resistance (10, 11), but conferred only a slightdegree of resistance in this BRCA1 mutant human cancer cell linemodel (20). Although 53BP1 depletion hyperactivated DNA endresection and RPA32 loading, without stabilization and increasedexpression of the mutant BRCA1 protein, RAD51 assembly couldnot occur following DNA damage.The present study shows inherent partial function of a BRCT

domain-mutated BRCA1 protein that can contribute to HR.Other studies have demonstrated functionality of N-terminalmissense mutations, and knock-in Brca1-deficient mouse modelsexpressing these mutants responded poorly to platinum drugs,mitomycin C, or PARP inhibition (23, 24). Additionally, althougha reduction in 53BP1 expression may facilitate DNA end resection

irrespective of the primary BRCA1 mutation, this event may beobligatory in BRCT domain-mutated cells so that end-resectioncan occur in the absence of a BRCA1 BRCT domain–CtIP in-teraction. Finally, our data suggest that HSP90 inhibition mayreverse PARP inhibitor resistance and may be a rational strategyparticularly germane in BRCA1 BRCT domain mutant cancers.

Materials and MethodsCell Culture. MDA-MB-436 cells were cultured in the presence of graduallyincreasing concentrations of rucaparib until resistant clones emerged thatgrew in 1 μM drug. Colonies were labeled RR 1 through 6. Cells werecultured in the absence of rucaparib for at least 1 mo before they wereused for experiments. Cells were routinely analyzed 6 h after 10 Gyγ-irradiation treatment.

Genomic Manipulations and Immunoprecipitation. Lentiviral generation andinfections and siRNA transfections were carried out according to standardprotocols. Protein knockdown or reexpression was routinely assessed 72 hafter transfection or 96 h after infection. BRCA1 or HSP90 complexes wererecovered from 2 mg of nuclear extract by using the Pierce Classic IP Kit(Thermo Scientific) according to the manufacturer’s instructions.

Preparation of DNA. Genomic DNA was isolated from cells using the DNeasytissue kit (Qiagen). BRCA1 gene sequencing was carried out as previouslydescribed (7). SNP chip array was carried out using the human SNP 6.0 arrayaccording to the manufacturer’s instructions (Affymetrix). Total RNA wasisolated from cell lines by using TRIzol (Invitrogen) and purified by using anRNeasy cleanup kit (Qiagen).

Statistics. Mean and SE values were compared by using unpaired t tests.For multiple comparisons we used one-way ANOVA. P < 0.05 was con-sidered statistically significant.

Further Details. Further details are provided in SI Materials and Methods.

ACKNOWLEDGMENTS. Rucaparib and olaparib were supplied by Clovis andAstraZeneca, respectively. This work was supported by US National Institutesof Health Grants R01 CA090687 (to G.I.S.), P50 CA089393 [Dana-Farber/Harvard Cancer Center (DF/HCC) Specialized Program of Research Excellence(SPORE) in Breast Cancer (to G.I.S.)], P50 CA090578 [DF/HCC SPORE in LungCancer (to G.I.S.)], P50 CA83636 [Pacific Ovarian Cancer Research ConsortiumSPORE in Ovarian Cancer (to E.M.S.)], P30 CA006927 [Fox Chase CancerCenter Developmental New Investigator Funds (to N.J.)], and R01 CA142698(to D.C.); Susan G. Komen Investigator Initiated Research Grant 12223953 (toG.I.S.); Susan G. Komen Career Catalyst Award CCR12226280 (to N.J.); theWendy Feuer Ovarian Cancer Research Fund (to E.M.S.); and AmericanCancer Society Research Scholar Grant RSG-12-079-01 (to D.C.).

1. Williams RS, Glover JN (2003) Structural consequences of a cancer-causing BRCA1-BRCT missense mutation. J Biol Chem 278(4):2630–2635.

2. Williams RS, et al. (2003) Detection of protein folding defects caused by BRCA1-BRCT

truncation and missense mutations. J Biol Chem 278(52):53007–53016.3. Lee MS, et al. (2010) Comprehensive analysis of missense variations in the BRCT do-

main of BRCA1 by structural and functional assays. Cancer Res 70(12):4880–4890.4. Kennedy RD, Quinn JE, Mullan PB, Johnston PG, Harkin DP (2004) The role of BRCA1

in the cellular response to chemotherapy. J Natl Cancer Inst 96(22):1659–1668.5. Farmer H, et al. (2005) Targeting the DNA repair defect in BRCA mutant cells as

a therapeutic strategy. Nature 434(7035):917–921.6. Fong PC, et al. (2009) Inhibition of poly(ADP-ribose) polymerase in tumors from BRCA

mutation carriers. N Engl J Med 361(2):123–134.7. Swisher EM, et al. (2008) Secondary BRCA1 mutations in BRCA1-mutated ovarian

carcinomas with platinum resistance. Cancer Res 68(8):2581–2586.8. Norquist B, et al. (2011) Secondary somatic mutations restoring BRCA1/2 predict

chemotherapy resistance in hereditary ovarian carcinomas. J Clin Oncol 29(22):

3008–3015.9. Jaspers JE, et al. (2013) Loss of 53BP1 causes PARP inhibitor resistance in Brca1-

mutated mouse mammary tumors. Cancer Discov 3(1):68–81.10. Bouwman P, et al. (2010) 53BP1 loss rescues BRCA1 deficiency and is associated with

triple-negative and BRCA-mutated breast cancers. Nat Struct Mol Biol 17(6):688–695.11. Bunting SF, et al. (2010) 53BP1 inhibits homologous recombination in Brca1-deficient

cells by blocking resection of DNA breaks. Cell 141(2):243–254.12. Sy SM, Huen MS, Chen J (2009) PALB2 is an integral component of the BRCA complex

required for homologous recombination repair. Proc Natl Acad Sci USA 106(17):

7155–7160.

13. Zhang F, Fan Q, Ren K, Andreassen PR (2009) PALB2 functionally connects the breastcancer susceptibility proteins BRCA1 and BRCA2. Mol Cancer Res 7(7):1110–1118.

14. Elstrodt F, et al. (2006) BRCA1 mutation analysis of 41 human breast cancer cell linesreveals three new deleterious mutants. Cancer Res 66(1):41–45.

15. Whitesell L, Lindquist SL (2005) HSP90 and the chaperoning of cancer. Nat Rev Cancer5(10):761–772.

16. Dungey FA, Caldecott KW, Chalmers AJ (2009) Enhanced radiosensitization of humanglioma cells by combining inhibition of poly(ADP-ribose) polymerase with inhibitionof heat shock protein 90. Mol Cancer Ther 8(8):2243–2254.

17. Sartori AA, et al. (2007) Human CtIP promotes DNA end resection. Nature 450(7169):509–514.

18. Yun MH, Hiom K (2009) CtIP-BRCA1 modulates the choice of DNA double-strand-break repair pathway throughout the cell cycle. Nature 459(7245):460–463.

19. Nik-Zainal S, et al.; Breast Cancer Working Group of the International Cancer GenomeConsortium (2012) Mutational processes molding the genomes of 21 breast cancers.Cell 149(5):979–993.

20. Oplustilova L, et al. (2012) Evaluation of candidate biomarkers to predict cancer cellsensitivity or resistance to PARP-1 inhibitor treatment. Cell Cycle 11(20):3837–3850.

21. Nakada S, Yonamine RM, Matsuo K (2012) RNF8 regulates assembly of RAD51 at DNAdouble-strand breaks in the absence of BRCA1 and 53BP1. Cancer Res 72(19):4974–4983.

22. Scully R, et al. (1997) Association of BRCA1 with Rad51 in mitotic and meiotic cells.Cell 88(2):265–275.

23. Drost R, et al. (2011) BRCA1 RING function is essential for tumor suppression butdispensable for therapy resistance. Cancer Cell 20(6):797–809.

24. Shakya R, et al. (2011) BRCA1 tumor suppression depends on BRCT phosphoproteinbinding, but not its E3 ligase activity. Science 334(6055):525–528.

Fig. 5. Increased mutant BRCA1 protein in a platinum-resistant ovarian car-cinoma. BRCA1 and 53BP1 protein levels measured by immunohistochemistryfrom patient 149101. Representative stains of biopsies taken from the plati-num-sensitive primary ovarian tumor and the recurrent resistant tumor.

17046 | www.pnas.org/cgi/doi/10.1073/pnas.1305170110 Johnson et al.

Dow

nloa

ded

by g

uest

on

June

11,

202

0