A Short BRCA2-Derived Cell-Penetrating Peptide Targets RAD51 Function and Confers Hypersensitivity toward PARP Inhibition

Double-strand breaks (DSB) are highly detrimental DNA lesions because, if left unrepaired or misrepaired, they can trigger cell death and genomic instability, ultimately causing cancer (1). To circumvent this threat, cells are equipped with diverse DSB repair mechanisms, including nonhomologous end joining (NHEJ) and homologous recombination (HR) as the two major pathways (2). Furthermore, recent work has established that in response to DNA replication stress, several key HR factors play a crucial role in protecting stalled DNA replication forks from nucleolytic degradation (3). Because rapidly dividing cancer cells rely on efficient DSB repair and fork protection mechanisms for their survival, inhibiting HR represents an attractive strategy for the development of novel therapeutic drugs, in particular when used in combination with DNA-damaging agents (4, 5).

The human BRCA2 protein plays an essential role in HR by promoting homology search and stimulating strand invasion into the sister chromatid (6). Specifically, following DNA-end resection, BRCA2 directs RAD51 filament nucleation onto RPA-coated single-stranded DNA (ssDNA; ref. 7). RAD51 interacts with two distinct regions in BRCA2, the BRC repeat motifs and a C-terminal domain (8–10). Importantly, the eight evolutionarily conserved BRC repeats, each consisting of about 35 amino acids, significantly differ in their capacity to bind RAD51 with BRC4 displaying the highest affinity (11, 12). Consequently, it was proposed that BRC repeats 1–4 facilitate nucleation of RAD51 by binding monomeric RAD51 and reducing its ATPase activity (11, 13). Structural analysis of the BRC4 repeat identified residues 1523-GFHTASG-1529 of BRCA2 to structurally mimic the self-oligomerization motif of RAD51 (14). In addition to the FHTA motif, a second consensus tetrameric module in BRC4, denoted as LFDE motif, was shown to bind to a distinct pocket in RAD51 distant from the oligomerization interface (15). In contrast to the BRC repeats, the C-terminal domain does not bind monomeric RAD51 but instead stabilizes RAD51 nucleoprotein filaments (9, 10). Taken together, compounds that selectively and efficiently block BRCA2–RAD51 interaction could advance into the clinic as bona fide HR inhibitors for both monotherapy and add-on therapy with DNA-damaging agents.

The physical nature of protein–protein interaction (PPI) interfaces often renders them unable to support binding of small-molecule inhibitors (SMI). Instead, peptide therapeutics offer an alternative way to target PPIs with key advantages over SMIs, including their direct similarity to protein fragments and the coverage of extensive PPI interfaces (16). However, poor membrane permeability has previously limited their use to extracellular targets (17). Thus, hydrophilic peptides are reliant on a permeation-enhancing strategy that facilitates targeting of intracellular molecules (18). Recently, cell-penetrating peptides (CPP) have been developed to enhance the cellular uptake and nuclear translocation of membrane-impermeable cargo molecules (19). They comprise a highly diverse class of short, primarily cationic peptides that combine a limited cytotoxicity and the ability to mediate receptor-independent transport of cargoes across cell membranes (20). Notably, the nona-arginine (R9) peptide is one of the most potent CPPs, giving a high transduction efficiency combined with low cytotoxicity (21).

Here, we design a CPP comprised of a 16-amino acid stretch of the BRCA2 BRC4 repeat able to inhibit BRCA2–RAD51 interaction. Our detailed functional analysis reveals that an R9-fused BRC4 CPP prevents RAD51 loading onto ssDNA, resulting in defective homology-mediated repair of DSBs as well as increased MRE11-dependent degradation of stalled DNA replication forks. Consequently, peptide incubation renders cells hypersensitive to the PARP inhibitor olaparib, providing a potential use for BRCA2-derived peptides in the treatment of certain types of cancer.

HeLa, U2OS, RPE1, MRC5 (all from ATCC), and HeLa DR-GFP were cultured in DMEM (Gibco) supplemented with 10% FCS (Sigma-Aldrich), 100 U/mL penicillin, and 100 μg/mL streptomycin (Life Technologies). PEO1 and PEO4 cells were purchased from the Health Protection Agency Culture Collections and cultured in RPMI medium (Gibco) supplemented with 10% FCS, 2 mmol/L sodium pyruvate (Gibco), and penicillin/streptomycin. MCF10A cells were purchased from ATCC and cultured in DMEM/F12 (Gibco) containing 5% Horse Serum (Gibco), 20 ng/mL human EGF (Sigma-Aldrich), 0.5 mg/mL hydrocortisone (Sigma-Aldrich), 10 μg/mL insulin, and penicillin/streptomycin. Stable U2OS cells expressing GFP-RAD51 (22) were grown in DMEM supplemented with 10% Tet system approved FCS (GIBCO) and penicillin/streptomycin. To induce GFP-RAD51 expression, cells were treated with 1 μg/mL doxycycline (Sigma-Aldrich) for 24 hours. All cell lines were confirmed to be free of mycoplasma contamination on a regular basis (PCR Mycoplasma Test Kit, AppliChem). Cells were passaged for no longer than 2 months after thawing of early-passage stocks. For cells that have been received from secondary sources, no cell line authentication was performed. Irradiation was performed using a Faxitron X-ray machine.

Camptothecin, RAD51 inhibitor B02 (23), cycloheximide, hydroxyurea (HU), and mirin (24) were purchased from Sigma-Aldrich. Olaparib (AZD2281) was provided by Selleck Chemicals. Thymidine analogs CIdU, IdU, and EdU were purchased from Sigma-Aldrich and Life Technologies, respectively. Custom-designed peptides were purchased from Bachem AG and, if not specified, synthesized according to standard practice (l-amino acids, N-terminal tag or acetylation, C-terminal amidation). Lyophilized peptides were dissolved in PBS at 1 mg/mL.

A detailed list of all primary and secondary antibodies can be found in Supplementary Tables S1 and S2, respectively.

A detailed list of siRNA oligonucleotide sequences used in this study can be found in Supplementary Table S3. siRNA oligos were used at a final concentration of 10 nmol/L and transfected using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer's instructions.

BRCA2 GST-fusion plasmids (GST-BRC 1-2, GST-BRC3-5, GST-BRC6-8, GST-C-term) have been described before (25). BRCA2 GST-fusion proteins were expressed in BL21-CodonPlus-RIL Escherichia coli (E. coli) by growing them overnight at 18°C using 100 μmol/L isopropyl β-d-thiogalactoside. Recombinant full-length RAD51 was prepared as described previously (26).

If not specified otherwise, cells were lysed in Laemmli buffer (4% SDS, 20% glycerol, 120 mmol/L Tris-HCl pH 6.8) and resolved by Tris-glycine SDS-PAGE. To probe for BRCA2, 3%–6% NuPAGE Tris-Acetate gels (Thermo Fisher Scientific) were run according to the manufacturer's instructions. After transfer to nitrocellulose membranes, immunoblotting was performed with indicated primary antibodies overnight at 4°C and secondary antibodies for 1 hour at room temperature. Stained proteins were visualized using the Advansta WesternBright ECL reagent and the VilberLourmat Fusion Solo S imaging system.

For peptide pull-downs, 30 μL streptavidin-coupled Dynabeads (Life Technologies) were incubated with 5 μg (2.7 nmol) of biotinylated BRC4 peptides or biotin analogue d-desthiobiotin (Sigma-Aldrich) in 1 mL PBS-T (0.1% Triton X-100) for 1 hour at 4°C. Beads were washed three times with PBS-T and blocked for 30 minutes with 0.3% BSA in PBS at 4°C. A total of 50 ng (1.35 pmol) recombinant RAD51 together with 2 μmol/L ATP was added to the beads and incubated for 2 hours in 700 μL PBS-T. The beads were washed four times with NTEN300 (20 mmol/L Tris pH 7.4, 0.1 mmol/L EDTA, 300 mmol/L NaCl, 0.5% NP-40) and once with TEN100 (20 mmol/L Tris pH 7.4, 0.1 mmol/L EDTA, 100 mmol/L NaCl) before complexes were boiled in 2× SDS sample buffer (10 mmol/L Tris pH 6.8, 20% glycerol, 3% SDS, 200 mmol/L DTT, 0.04% bromophenol blue) and subjected to immunoblotting. For GST pull-down assays, glutathione sepharose beads (GE Healthcare) were incubated for 1 hour at 4°C with equalized amounts of BL21 E. coli soluble extracts expressing one of the four GST-BRCA2 fusion constructs in TEN100. Beads were washed three times with NTEN300 buffer and once with TEN100 before adding either 1 mg HeLa nuclear extracts (CilBiotech) or 50 ng purified RAD51 supplemented with varying amounts of BRC4^wt or BRC4^mut peptides filled up to 1 mL with TEN100. After 2 hours of incubation, beads were washed twice with NTEN500 (20 mmol/L Tris pH 7.4, 0.1 mmol/L EDTA, 500 mmol/L NaCl, 0.5% NP-40), twice with NTEN300, and twice with TEN100 buffer before boiling in SDS sample buffer and protein analysis by immunoblotting.

Cell extracts were prepared using NP-40 extraction buffer [50 mmol/L Tris-HCl pH 7.5, 120 mmol/L NaCl, 1 mmol/L EDTA, 6 mmol/L EGTA, 15 mmol/L sodium pyrophosphate and 1% NP-40 supplemented with phosphatase inhibitors (20 mmol/L NaF, 1 mmol/L sodium orthovanadate), and protease inhibitors (1 mmol/L benzamidine and 0.1 mmol/L PMSF)]. After benzonase (Novagen) digestion for 30 minutes at 4°C, cell extracts were cleared by centrifugation. Lysates (2 mg) were supplemented with increasing amounts of peptides filled up to 1 mL with NP-40 extraction buffer and incubated for 1 hour at 4°C before adding 20 μL GFP-Trap agarose beads (ChromoTek) for 1 hour at 4°C. Beads were subsequently washed three times with GFP-IP buffer (100 mmol/L NaCl, 0.2% NP-40, 1 mmol/L MgCl₂, 10% glycerol, 5 mmol/L NaF, 50 mmol/L Tris-HCl pH 7.5) and boiled in SDS sample buffer for analysis by immunoblotting.

Cells were seeded either into 8-well chamber imaging slides (μ-Slide 8 Well, ibidi), 24-well plates, 6-well plates, or 6-cm culture dishes (Sarstedt) and grown to around 80% confluence at day of peptide transfection. Cells were washed at least once with PBS to remove residual FCS and incubated with indicated peptide concentrations in appropriate serum-free medium for 1 hour at 37°C. If not specified otherwise, the following incubation volumes were used: 0.3 mL for 8-well chamber imaging slides and 24-well plates, 0.5 mL for 12-well plates and 2 mL for 6-well plates and 6-cm culture dishes.

A total of 4 × 10⁴ cells were seeded into 8-well chamber imaging slides and grown overnight. After 30 minutes of staining with 0.5 μg/mL Hoechst 33342 (Life Technologies), cells were washed twice with PBS and incubated with indicated peptide concentrations. Cells were washed twice with PBS and imaged in Live Cell Imaging solution (Thermo Fisher Scientific). Images were taken with CLSM SP5 Mid UV-VIS Leica with 63× objective at 37°C at ambient CO₂ concentrations.

EdU incorporation was analyzed using the Click-it EdU technology (Thermo Fisher Scientific) according to the manufacturer's instructions. For peptide uptake studies, 1 × 10⁵ cells were seeded into 12-well plates. The day after, peptide transfection was performed and cells were harvested by trypsinization to remove membrane-bound peptides. After one wash with PBS, cells were resuspended in PBS and subjected to flow cytometry analysis. To quantify intracellular peptide stability, same cells were released for indicated time points in DMEM + 10% FCS and fixed with 4% formaldehyde (w/v) in PBS for 15 minutes at room temperature. To measure the TAMRA fluorescence intensity, the LSR II Fortessa equipped with a 561 nm laser line and a 586/15 band-pass filter was used. Of note, the fluorescence intensity of TAMRA-labeled R9-BRC4^mut peptides was corrected for quenching by multiplying measured TAMRA intensity with a quenching factor. The quenching factor was calculated by loading 10 pmol of freshly solubilized peptides on Tricine SDS-PAGE gel and quantifying TAMRA intensity (see Fig. 2C, lane 5). Fluorescein intensity was measured with Attune Nxt Flow Cytometer equipped with 488 laser and 530/30 band-pass filter. For each condition, 20,000 events were recorded. MACS Quant Calibration Beads (MACS Miltenyi Biotec) were applied for voltage standardization to exclude any machine-dependent variations between measurements.

To resolve low molecular weight peptides, Tricine SDS-PAGE was performed as described previously (27). For peptide separation, Laemmli lysates were loaded onto 16% Tricine SDS-PAGE gel containing 6 mol/L urea. For peptide detection via fluorescence, gels were scanned using a Typhoon FLA 9500 FluorImager.

HR frequency was measured as described previously (28, 29). Briefly, following siRNA transfection, 1 × 10⁵ HeLa cells containing a stably integrated DR-GFP reporter construct were seeded into 12-well plates. The day after, cells were either mock-transfected or transfected with 0.6 μg I-SceI expression plasmid (pCBASce) using jetPrime transfection reagent (Polyplus transfection). Four hours later, medium was exchanged and either a 1-hour peptide incubation or second siRNA transfection was performed. Peptide incubations were repeated 24 and 34 hours post–I-SceI transfection. After each peptide incubation, 0.5 mL of DMEM + 20% FCS was directly added to the peptide/DMEM mix. Forty-eight hours after I-SceI transfection, cells were harvested and directly analyzed for GFP expression by flow cytometry using an Attune Nxt Flow Cytometer.

Twenty-four hours after siRNA transfection, 8 × 10⁴ cells were seeded on coverslips in 24-well plates. The day after, cells were treated either with 100 nmol/L camptothecin for 1 hour or irradiated and incubated for another hour with the peptides, before releasing them for 3 hours by directly adding 1 mL of DMEM + 14% FCS. Alternatively, cells were transfected with the peptides, followed by 1-hour camptothecin treatment and direct processing. Cells were preextracted for 5 minutes on ice (25 mmol/L HEPES pH 7.4, 50 mmol/L NaCl, 1 mmol/L EDTA, 3 mmol/L MgCl₂, 300 mmol/L sucrose, 0.5% Triton X-100), fixed with 4% formaldehyde (w/v) in PBS for 15 minutes at room temperature, before incubating them with indicated primary and appropriate secondary antibodies for 1 hour. Afterwards, coverslips were mounted with Vectashield (Vector Laboratories) containing DAPI and sealed. Images were acquired on Leica DMI6000 widefield fluorescence microscope with a 63× objective.

DNA fiber analyses were carried out as described previously (30). In brief, U2OS cells were seeded into 6-well plates at a confluence of 40%. Twenty-four hours later, cells were pulse-labeled with 33 μmol/L CIdU for 30 minutes, followed by 340 μmol/L IdU for 30 minutes prior to incubation with 2 mmol/L HU and peptides (10 μmol/L) for 4 hours in serum-free medium. Cells were lysed (200 mmol/L Tris-HCl pH 7.4, 50 mmol/L EDTA, 0.5% SDS) and DNA fibers were stretched onto glass slides before fixation in methanol–acetic acid (3:1, Merck) overnight. Rehydration in PBS was followed by denaturation in 2.5 mol/L HCl for 1 hour, a PBS wash, and blockage (2% BSA (w/v) PBS, 0.1% Tween 20) for 40 minutes. CIdU and IdU staining was performed using anti-BrdU primary and secondary antibodies for 2.5 hours. Coverslips were mounted using Antifade Gold (Invitrogen). Images were acquired on Olympus microscope IX81 with ×60 magnification, and analysis was carried out using ImageJ software.

Indicated cell lines were plated in poly-l-lysine (Sigma-Aldrich) coated 24-well plates at low cell dilutions of 200 cells/well in technical triplicates. PEO1 and PEO4 cells were seeded at 500 and 1,000 cells/well, respectively. Twenty-four hours later, cells were washed once with PBS and incubated with olaparib in presence or absence of peptides. After 1 hour, 1 mL of appropriate medium containing 14% FCS with indicated olaparib concentrations was directly added to the cells without removing the peptide solution. For MCF10A cells, FCS concentration of culture medium was increased to 7%. Alternatively, HeLa cells were treated for 1 hour with 1 μmol/L camptothecin, washed twice with PBS, and peptide transfection was carried out for 1 hour before directly adding 1 mL of DMEM + 14% FCS. Cells were grown for 10 days before fixation with crystal violet solution [0.5 % crystal violet, 20% ethanol (w/v)]. For analysis, plates were scanned and analyzed with the ImageJ Plugin ColonyArea using the parameter colony intensity, integrating the percentage of the covered area and staining intensity (31).

All results were confirmed in at least two independent experiments. Quantitative data are displayed as mean ± SD, and statistical analyses were performed using GraphPad Prism 7. P values <0.05 were considered statistically significant.

It is well established that BRCA2 binds RAD51 through its BRC repeats composed of two highly conserved tetrameric motifs (Supplementary Fig. S1A). Among the eight BRC repeats, BRC4 was reported to display the highest affinity for RAD51, mainly using its FHTA sequence to bind the RAD51 oligomerization motif. To specifically target the BRCA2–RAD51 protein interaction interface, we therefore synthesized a 16-mer peptide mimicking the N-terminal half of BRC4 comprising the FHTA hydrophobic motif (Fig. 1A). In addition to the “wild type” BRC4 peptide (BRC4^wt), we included a “mutated” BRC4 peptide (BRC4^mut) harboring an inverted FHTA sequence (Fig. 1A). Employing N-terminally biotinylated peptides, recombinant RAD51 was efficiently pulled down by BRC4^wt but to a much lesser extent by BRC4^mut (Fig. 1B). In agreement with previous reports (32), we observed that GST-tagged BRCA2 fusion proteins spanning BRC 1-2, BRC 3-5, or the C-terminal (C-term) domain were able to interact with RAD51 (Supplementary Fig. S1B; Fig. 1C). Remarkably, BRC4^wt was able to outcompete each of these individual interactions in a concentration-dependent manner (Fig. 1D). Interestingly, the BRC4 peptide was more effective in outcompeting RAD51 binding to the BRCA2 C-term and BRC repeats 1-2 than to BRC repeats 3-5 (Fig. 1D), indicating that the BRC4 repeat of BRCA2 exhibits the highest binding affinity for RAD51. Similar GST pull-down results were obtained using HeLa nuclear extracts as a source for RAD51 (Supplementary Fig. S1C). Given that the BRCA2 C-terminal region exclusively binds to assembled RAD51 oligomers, we reasoned that the BRC4 peptide is able to disrupt RAD51 multimers present in solution, which is in agreement with binding of the FHTA cluster to the RAD51 oligomerization motif (9, 10, 14). Most importantly, coimmunoprecipitation experiments in U2OS cells inducibly expressing GFP-tagged RAD51 demonstrated that BRC4^wt, but not BRC4^mut, is capable of interfering both with BRCA2 binding to RAD51 and RAD51 oligomerization (Fig. 1E; Supplementary Fig. S1D and S1E). Taken together, our results indicated that a short, synthetic BRC4-derived peptide is proficient in blocking BRCA2–RAD51 protein–protein interaction.

Native peptides do not readily cross cell membranes. To enhance cellular uptake, we decided to conjugate BRC4 peptides with a nona-arginine (R9) CPP. In addition, a red fluorescent dye (TAMRA) was N-terminally attached to R9-BRC4 peptides to analyze cellular uptake. Using confocal microscopy, we observed robust cytoplasmic and nuclear TAMRA signals in HeLa and U2OS cells upon transfection with R9-fused peptides (Fig. 2A). Moreover, we found that the concentration threshold for efficient R9-BRC4 cell penetration was above 10 μmol/L (Supplementary Fig. S2A). Flow cytometry analyses further confirmed that BRC4 peptide delivery reached a transduction efficiency of almost 100% when fused to R9 (Fig. 2B; Supplementary Fig. S2B). Importantly, when replacing the TAMRA label with a green fluorescent dye (Fluorescein), we observed very similar subcellular localization patterns and fluorescent intensities of the R9-BRC4 peptides (Supplementary Fig. S2C and S2D). As peptides are prone to proteolytic degradation upon cellular uptake, we next determined the amount of intact peptides being delivered to the cells. To differentiate between full-length and degraded peptides, whole-cell lysates of HeLa and U2OS cells incubated with fluorescently labeled peptides were subjected to SDS-PAGE designed for resolving very low molecular weight protein species (27). Using this approach, we detected significant amounts of intact TAMRA- and Fluorescein-labeled R9-BRC4 peptides being effectively delivered to HeLa and U2OS cells (Fig. 2C; Supplementary Fig. S2E). Comparing the band intensity of TAMRA signals between 10 pmol of freshly solubilized peptides directly loaded onto the gel and those of peptide-transduced cell lysates, we calculated a delivery rate of approximately 10⁷ peptides per cell, yielding an estimated intracellular peptide concentration of around 1 to 10 μmol/L (Fig. 2C).

Next, to more precisely determine the intracellular residence time and stability of our BRC4 CPPs, we modified a previously established method (33) and monitored the Fluorescein signal intensity in HeLa cells over a time course of 24 hours after peptide incubation. SDS-PAGE analysis revealed that BRC4^wt as well as BRC4^mut cargo peptides were gradually degraded with an approximate half-life of 2 hours (Fig. 3A). Strikingly, using the same experimental setup, flow cytometry analysis of Fluorescein-R9-BRC4^wt indicated a rather heterogeneous degradation pattern, which was most pronounced after 8 hours with a large proportion of cells still showing moderate to high fluorescent intensities (Fig. 3B).

Collectively, we concluded that intracellular uptake of R9-BRC4^wt and R9-BRC4^mut was efficient and comparable as determined by confocal microscopy, flow cytometry, and SDS-PAGE analysis, thus providing a solid basis for further mechanistic investigations.

Conditional overexpression of full-length BRC4 was previously shown to inhibit DNA damage–induced RAD51 foci formation in MCF7 (breast cancer) and chicken DT40 cells (12, 34). Therefore, we wanted to investigate whether the 16-mer BRC4 peptide fused to R9 was able to mimic this phenotype, indicative of effective disruption of RAD51 binding to BRCA2. To this end, HeLa or U2OS cells were first incubated with our peptides and subsequently treated with the DNA topoisomerase I poison camptothecin to induce replication-dependent DSBs. Importantly, we did not detect any significant change in camptothecin-induced ATM and CHK2 phosphorylation as well as γH2AX foci formation upon peptide addition, indicating regular activation of apical DNA damage response kinases (Fig. 4A; Supplementary Fig. S3A and S3B). In contrast, camptothecin-induced RAD51 foci formation in both cell lines was specifically compromised in the presence of BRC4^wt but not BRC4^mut CPPs (Fig. 4B and C; Supplementary Fig. S3C and S3D). In line with this observation, BRC4^wt significantly suppressed RAD51 foci formation following ionizing radiation (Fig. 4D). Notably, intrinsic RAD51 protein stability was not affected by BRC4^wt cellular uptake (Supplementary Fig. S3E). To directly examine the impact of BRC4 peptides on DSB repair by HR, we performed repair reporter assays (DR-GFP) in HeLa cells and observed a significant decrease in HR frequency upon repetitive BRC4^wt transfections (Fig. 4E). Throughout all experiments reported in this section, we observed that siRNA-mediated BRCA2 depletion conferred much stronger phenotypes compared with delivery of the BRC4^wt peptide.

In summary, our findings suggested that delivery of a synthetic BRC4 peptide potently inhibits RAD51 foci formation and HR, most likely as a result of defective BRCA2-mediated RAD51 loading onto resected DSBs.

BRCA2-dependent RAD51 loading is not only critical for HR, but has also been established to protect stalled replication forks from nucleolytic degradation by MRE11 (35, 36). Thus, we examined a potential effect of the BRC4 peptide on replication fork protection by performing dual-labeling DNA fiber assays in the presence of HU, which stalls fork progression. Strikingly, we found that the BRC4^wt CPP resulted in shortening of nascent DNA tracts, indicative of increased fork degradation (Fig. 5A). BRC4 intracellular uptake did not interfere with global replication rates, as we could not observe any differences in EdU incorporation (Supplementary Fig. S4). Similar to what has been shown for BRCA2-deficient cells, fork degradation in BRC4 peptide–transduced cells was completely rescued both by mirin, an MRE11 SMI (Fig. 5B), and by siRNA-mediated MRE11 depletion (Fig. 5C). Of note, BRC4 delivery did not result in additive or synergistic effects when delivered into BRCA2-depleted cells, indicating that peptide-mediated fork degradation resulted most likely from specific targeting of the BRCA2–RAD51 interaction (Fig. 5C).

On the basis of our molecular analyses providing robust evidence of the inhibitory effect of the BRC4 peptide on BRCA2 functions in HR and fork protection, we next sought to determine whether it also sensitizes cells to DNA-damaging agents. Profound hypersensitivity of BRCA2-mutant cells to PARP inhibitors has become an emerging therapeutic paradigm known as synthetic lethality (37). Therefore, we performed clonogenic survival assays in peptide-transfected HeLa cells treated with the PARP inhibitor olaparib (38). Indeed, targeting of the BRCA2–RAD51 interaction by BRC4^wt CPPs resulted in a significant reduction in cellular viability in response to chronic PARP inhibition (Fig. 6A and B). Consistent with impaired RAD51 foci formation, delivery of BRC4^wt CPPs also sensitized HeLa cells to camptothecin (Supplementary Fig. S5A). Upon treatment with olaparib, we estimated the IC₅₀ value of BRC4 to be around 10 μmol/L for HeLa cells (Fig. 6C). Importantly, R9-BRC4^wt alone did not decrease cell survival in otherwise undamaged cells (Fig. 6C). Moreover, despite its relatively short half-life, the efficacy of BRC4 CPPs in sensitizing HeLa cells to PARP inhibition was comparable with that of the B02 small-molecule compound, which specifically inhibits RAD51 binding to DNA (Fig. 6D; refs. 23, 39). To further exclude potential off-target effects of the BRC4 CPP in stimulating olaparib-induced cytotoxicity, we employed PEO1 (BRCA2 null) and PEO4 (BRCA2 revertant to wild type) isogenic ovarian cancer cell lines (40). Notably, we observed a strong synergy between PARP inhibition and R9-BRC4 peptides in PEO4 cells, whereas PEO1 cells did not exhibit BRC4-mediated sensitivity toward olaparib (Supplementary Fig. S5B). Differential peptide uptake could be excluded as both cell lines displayed comparable uptake efficiency and intracellular peptide localization (Supplementary Fig. S5B). Finally, in addition to HeLa cells, we observed that BRC4 CPPs elicited olaparib-induced cell death of U2OS human osteosarcoma cells but not of noncancerous cell lines, including hTERT-immortalized human retinal pigment epithelial cells (hTERT-RPE1), SV40-immortalized human fetal lung fibroblasts (MRC5) and human breast epithelial cells (MCF10A; Fig. 6E; Supplementary Fig. S6A). This finding could at least in part be explained by an overall low cellular uptake efficiency and predominant endosome trapping of BRC4 peptides in RPE1, MRC5, and MCF10A nontumorigenic cell lines (Supplementary Fig. S6B and S6C).

PPIs represent attractive but at the same time challenging targets for pharmacologic intervention in cancer therapy. The emergence of successful peptide-based PPI inhibitors seemed promising, yet drug development was faced with the low membrane translocation ability of native peptides. The identification of a wide range of carrier molecules, commonly termed as CPPs or protein transduction domains, provided a possible solution for this major limitation (19).

Here, we designed a 16-amino acid short peptide derived from the N-terminal half of the human BRCA2 BRC4 repeat able to occupy the RAD51 oligomerization interface and thereby potently inhibiting the interaction between BRCA2 and RAD51. Covalent fusion of the BRC4 peptide with the cationic polyarginine R9 CPP resulted in efficient uptake into both cytoplasm and nucleus, causing defects in HR repair, increased fork degradation, and, ultimately, hypersensitivity to DNA-damaging agents.

Detailed biochemical studies have demonstrated that the full-length BRC4 repeat covering both FHTA and LFDE tetrameric cluster motifs selectively interacts with RAD51 and inhibits RAD51–DNA complex formation (15, 41). To increase the likelihood of intracellular peptide uptake (42), we reduced the length of the native BRC4 peptide sequence from 35 to 16 amino acids, including only the first FHTA module, which was shown to mimic the RAD51 oligomerization motif (14). In line with our findings, a similar 17-mer N-terminal BRC4 peptide was reported to compete with full-length BRC4 for RAD51 binding, albeit with modest potency (15). We further corroborated the importance of the FHTA sequence by showing that BRC4 16-mer peptides harboring a mirrored ATHF sequence display greatly reduced RAD51 binding affinity.

Importantly, intracellular BRC4 peptide uptake in the low micromolar range was sufficient to impair BRCA2–RAD51 complex formation. A high uptake efficiency was likely achieved through a combination of the potent CPP R9 with the beneficial biochemical properties of the BRC4 peptide, including its positive net charge, hydrophobicity, and three-dimensional conformation (18). CPP internalization involves both endocytosis and direct translocation, whereby direct cell penetration is preferred because endocytosis can lead to the trapping of peptides in endosomes and their ultimate degradation (18). Multiple factors, including the CPP, cargo identity, concentration, cell type, and differentiation status, influence the balance between both uptake mechanisms (18). We primarily observed a diffuse fluorescent staining and only minor punctuate patterns, indicative of direct translocation as the preferred uptake mechanism. The substantial number of peptide molecules reaching the intracellular space might compensate for the short half-life. Even though only 5% of intact peptides are present 24 hours after transfection, the absolute peptide number per cell at this stage still adds up to 10⁶ functional peptides. According to a recent proteomics study in the U2OS cell line, most proteins involved in DSB repair processes, including BRCA2, were shown to be of low to moderate abundance (10²–10⁴ copies/cell; ref. 43). On the basis of these numbers, we speculated that BRC4 peptides are present in high molar excess over their endogenous BRCA2 target protein even 24 hours after uptake.

RAD51 and its property to self-assemble are under dynamic control to enable faithful homology-directed DSB repair. Yu and colleagues proposed the presence of three RAD51 fractions in the cell: a mobile fraction of RAD51 monomers, an immobile oligomerized fraction, and an immobile BRCA2-bound fraction (44). Strikingly, they found the BRCA2-bound fraction to be selectively mobilized upon DNA damage and suggested a dual BRCA2 function of RAD51 sequestration and mobilization. Importantly, the other fractions did not change, suggesting that only the BRCA2-bound RAD51 fraction is involved in DNA repair. Notably, R9-BRC4^wt did not affect viability under unstressed growth conditions of conventionally cultured cancer cell lines. Thus, we hypothesize that under normal, undamaged conditions, RAD51 is stably bound to BRCA2 and the peptide's affinity for RAD51 is not high enough to disassemble the protein complex. Accordingly, the BRC4 mimetic peptide is only able to disassemble mobilized BRCA2–RAD51 complexes. We speculate that upon mobilization and ultimate RAD51 loading onto ssDNA, the protein complex becomes increasingly unstable due to an equilibrium change during DNA loading. In this context, the peptide is potent enough to sequester RAD51 transiently dissociated from BRCA2. We further presume that at low levels of DNA damage, requiring only a small fraction of BRCA2-bound RAD51, repair processes are not significantly disturbed by the BRC4 peptide. However, upon increasing DNA damage load and replication stress, when there is a high demand of functional BRCA2–RAD51 complexes, a major pool of RAD51 is sequestered away from BRCA2 by the BRC4 peptide.

Despite its potentially rather high dissociation constant compared with SMIs, BRC4 peptides provided an unexpectedly high degree of specificity. CPPs have often been reported to cause unspecific membrane disruption, which has led to false conclusions regarding their specificity for a number of peptide therapeutics. Because we do not observe a decrease in cell viability upon BRC4 peptide uptake, we can exclude such unspecific CPP-mediated toxicity (45). Furthermore, our finding that MRE11 depletion fully rescued the peptide-induced replication fork degradation phenotype and the lack of synergy between peptides and PARP inhibition in BRCA2-deficient cancer cells argue against potential off-target activity of the BRC4 CPP. Nevertheless, when comparing the outcomes of siRNA-mediated BRCA2 depletion versus R9-BRC4^wt peptide delivery, depletion is commonly yielding much more pronounced phenotypes. A possible explanation for this relates to the high susceptibility of peptides to extra- and intracellular protease attacks (17). To improve their pharmacokinetic potential, possible peptide optimizations include various backbone modifications and cyclization (17, 46). Moreover, the BRCA2–RAD51 interaction might not be completely abolished by the BRC4 peptide as it only comprises the first FHTA but lacks the second LFDE tetrameric cluster motif.

The lack of cell-type specificity is commonly reported to present another major limitation of CPP-mediated drug delivery (19, 20). Strikingly, we could not find a synergistic relationship between the PARP inhibitor olaparib and the R9-BRC4 peptide in normal cell lines as compared with cancer cell lines. Interestingly, our peptide uptake data suggest an overall decreased intracellular BRC4 concentration. It is reasonable to speculate that efficient intracellular peptide delivery relies on specific membrane components or lipid compositions that are more uptake responsive in cancer versus noncancerous cell lines. Numerous reports classify lipid metabolic reprogramming as a major source of cell transformation (47). Specifically, the sphingolipid metabolism, which was suggested to influence R9-mediated peptide uptake, was reported to significantly alter during transformation (48). In addition, cancer cells possess a more negatively charged membrane than normal cells, which is partly caused by a loss in membrane symmetry and the exposure of anionic phosphatidylcholine on the outer leaflet (49). This feature, in combination with higher membrane fluidity (50), might favor the uptake of the cationic R9-BRC4 peptide.

Ultimately, the BRC4 peptide could emerge as a potent radio- and chemosensitizer. Specifically, BRC4 peptide–induced HR deficiency could represent a promising strategy for expanding the utility of PARP inhibitors, successfully applied in breast and ovarian cancer patients with a germline BRCA1 or BRCA2 mutation (51, 52), to BRCA-proficient cancers. Similarly, the BRC4 peptide could resensitize BRCA1/2-mutated tumors that acquired chemoresistance to PARP inhibitors by restoring HR (53). Moreover, malignant cells are frequently compromised in genome stability maintenance pathways that are synthetically lethal with HR deficiency (54). Consequently, monotherapy with the identified BRCA2 peptide inhibitor could provide a promising option for the treatment of these tumors. The BRCA2–BRC4 peptide described in this study could also be applied in combination with proton irradiation, which was shown to be highly advantageous over conventional photon therapy. Notably, it was found that HR-deficient tumor cells exhibit an enhanced susceptibility toward proton versus photon irradiation (55). Drug-induced HR deficiency with the BRC4 peptide inhibitor could make this advantageous effect accessible for patients with HR-proficient tumors. Finally, we reveal a straightforward approach to study distinct PPIs in a biological context without the need of elaborate mutagenesis methodologies and provide a potent research tool to study BRCA2-dependent RAD51 loading to ssDNA in different biological contexts.

No potential conflicts of interest were disclosed.

Conception and design: A. Trenner, A.A. Sartori

Development of methodology: A. Trenner, A.A. Sartori

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Trenner, J. Godau,

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Trenner, A.A. Sartori

Writing, review, and/or revision of the manuscript: A. Trenner, J. Godau, A.A. Sartori

Study supervision: A.A. Sartori

The authors would like to thank Roland Brock (Institute of Molecular Life Sciences, Radboud University Nijmengen, the Netherlands), Martin Pruschy (Clinic of Radiation Oncology, University Hospital Zurich, Zurich, Switzerland), and Stefano Ferrari (Institute of Molecular Cancer Research, University of Zurich, Zurich, Switzerland) for scientific input. We are grateful to Pavel Janscak (Institute of Molecular Cancer Research, University of Zurich) for providing purified, recombinant RAD51 protein and anti-RAD51 rabbit polyclonal antibody and to Fumiko Esashi (Sir William Dunn School of Pathology, University of Oxford, Oxford, United Kingdom) for providing the GST-tagged BRCA2 fusion plasmids. Furthermore, we are grateful to Ross Chapman (Nuffield Department of Medicine, University of Oxford) for providing hTERT-RPE1 cells, Minoru Takata (Radiation Biology Center, Kyoto University, Kyoto, Japan) for the U2OS GFP-RAD51 cell line, Marcel van Vugt (Medical Oncology Department, University Medical Centre Groningen, Groningen, the Netherlands) for HeLa DR-GFP cells, and Steve Jackson (Gurdon Institute, Cambridge, United Kingdom) for MRC5 cells. We also want to thank Antonio Porro (Institute of Molecular Cancer Research, University of Zurich) for critical reading of the manuscript. This study was supported by research grants from the Krebsliga Schweiz (KFS-3845-02-2016, to A.A. Sartori), Swiss National Science Foundation (31003A_156023 and 31003A_176161, to A.A. Sartori), Promedica Stiftung (1317/M to A.A. Sartori), and Novartis Foundation for Medical-Biological Research (#17C155 to A.A. Sartori).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.