Therapeutic Targeting of BRCA1-Mutated Breast Cancers with Agents That Activate DNA Repair

The breast cancer susceptibility gene 1 (BRCA1) encodes a tumor suppressor that acts to maintain genetic stability. Individuals who inherit a mutation in BRCA1 are predisposed to developing breast, ovarian, and other cancers. In fact, carriers of BRCA1 mutations have a 44% to 75% lifetime risk for developing breast cancer (1, 2) and up to an approximate 45% risk for developing a second primary breast cancer (3–6). Germline mutations in BRCA1 are responsible for approximately half of hereditary breast cancer cases (7). BRCA1-mutated breast cancers are particularly aggressive and exhibit poor outcomes compared with other subtypes of breast cancer. They frequently contain mutations in TP53 and tend to be “triple-negative” due to their lack of expression of estrogen receptor, progesterone receptor, and HER2/NEU oncogene, thereby rendering them relatively resistant to existing anticancer strategies that rely on p53-dependent apoptosis or use endocrine-based or HER2-directed therapy (2, 8–11). Notably, the selective estrogen receptor modifiers (SERM) tamoxifen and raloxifene are the only FDA-approved chemoprevention agents for breast cancer. Tamoxifen has displayed mixed results in studies that evaluated its preventive benefit in carriers of BRCA1 mutations (12–16); however, due to the lack of expression of estrogen and progesterone receptors in BRCA1-mutated tumors, SERMs are thought to be relatively ineffective in this setting. Furthermore, raloxifene is not currently FDA-approved for the chemoprevention of breast cancer in premenopausal women, and carriers of BRCA1 mutations tend to develop breast cancer before menopause (17, 18). Therefore, BRCA1-associated malignancies are prevalent, aggressive, and in need of an effective chemoprevention strategy.

Cellular characteristics that contribute to tumorigenesis, such as defects in the defense mechanisms against oxidative DNA damage (ODD), are important for the discovery of cancer prevention strategies. ODD is generated by reactive oxygen species due to normal metabolism and other biologic processes. A single human cell experiences approximately 10⁴ oxidative lesions per day (19), making ODD the most common form of DNA damage. ODD is typically repaired by the base-excision DNA repair (BER) pathway. When left unrepaired, ODD leads to mutagenesis, genetic instability, and ultimately the initiation and progression of breast and other cancers (20–22). BRCA1 has been shown to play a role in BER of ODD (23, 24), and BRCA1-mutated breast cancers displayed a compromised ability for BER of ODD (23). Therefore, in pursuit of a targeted chemoprevention strategy for BRCA1-associated malignancies, we sought to identify small molecules that enhanced the repair of ODD in the presence of mutant BRCA1, which, in turn, may prevent tumorigenesis.

Human breast cell lines included those that represent mutant BRCA1 breast cancer (SUM149, SUM1315, and HCC1937), wild-type BRCA1 breast cancer (BT474, MCF7), and the normal breast (MCF10A). SUM149 and SUM1315 were purchased and characterized by Asterand plc using short tandem repeat (STR) polymorphism analysis. All other cell lines were purchased and characterized by ATCC using STR DNA profiling. All cell lines tested negative for mycoplasma, were passaged for fewer than 6 months after receipt or resuscitation from frozen stocks created within 2 weeks of purchase, and were then cultured as recommended by the manufacturer. Human breast cancer cell lines isogenic for BRCA1 included MCF7-shCTRL and MCF7-shBRCA1. Nontargeting control shRNA (5′-GGAGATCAGCCATTAATAT-3′) and BRCA1 shRNA (5′-TGCCAAAGTAGCTAATGTA-3′) were cloned into pSUPER.retro.puro (Oligoengine) according to the manufacturer's instructions and then transduced into MCF7 cells. Stable selection was carried out with puromycin.

The reagents, compound library, screening protocol, and post-screen analysis as depicted in Fig. 1 are further described in Supplementary Methods.

Benserazide, acetohexamide, tamoxifen, and carbidopa were purchased from Sigma-Aldrich and prepared as described by the manufacturer. Treatment with drug occurred for 24 hours unless specified otherwise.

Adenovirus coding GFP (ad-GFP, Clontech) was incubated with methylene blue and exposed to visible light for 0 (undamaged control) or 2 minutes to induce ODD. Cells were then infected with undamaged or damaged ad-GFP and, after 24 hours, analyzed for GFP expression. Percentage repair is the percentage of GFP expression (normalized to cell number) from damaged virus relative to undamaged virus.

Detection of 8oxoG was measured by flow cytometry using the OxyFLOW Kit (HemoGenix) according to the manufacturer's instructions. Detection of H₂O₂-induced ODD was measured by the alkaline comet assay modified for detection of oxidative damage. Cells were treated with 25 μmol/L H₂O₂ for 10 minutes at 4°C and then allowed to repair for 4 hours under normal culture conditions. Cells were then subjected to the alkaline comet assay using the CometAssay HT Reagent Kit (Trevigen) as described by the manufacturer with the following modifications. Before denaturation, DNA was treated ± FPG enzyme (NE Biolabs) at 37°C and 5% CO₂ for 45 minutes to nick DNA at sites of ODD. Comet heads indicate undamaged DNA and comet tails indicate DNA damage; therefore, comets were analyzed for the “% DNA in tail” using CometScore software (TriTek Corporation). ODD was calculated by subtracting “% DNA in tail” in FPG⁻ samples from “% DNA in tail” in FPG⁺ samples. Detection of DNA strand breaks was also carried out by the alkaline comet assay using the CometAssay HT Reagent Kit (Trevigen) as described above; however, DNA strand breaks were calculated as the “%DNA in tail” in FPG⁻ samples. Detection of serum levels of 8oxodG was carried out using the HT 8-oxo-dG ELISA Kit (Trevigen) according to the manufacturer's instructions.

Single-cell suspensions were cultured in soft agar supplemented with medium containing drug or vehicle control for up to 14 days. Colonies were counted and expressed as a percentage of the control.

Single-cell suspensions were cultured in regular growth media with drug or vehicle control. After 2 days, cells were washed and subjected to trypan blue dye. The number of blue/nonviable and white/viable cells was then counted to calculate the percentage of viable cells.

Animal experiments were carried out in compliance with AAALAC Accreditation guidelines and were approved by the Institutional Animal Care and Use Committee at Stanford University (Stanford, CA). NOD/SCID female mice, 4 to 6 weeks of age, were obtained from Jackson Laboratories and were maintained under specific pathogen-free conditions. Mice were inoculated by subcutaneous injection with 3 × 10⁶ HCC1937 human breast cancer cells confirmed to be pathogen-free in 100 μL of PBS and Matrigel (1:1). Subcutaneous injection into the mammary fat pad occurred under 0.5% to 3% isoflurane for animal comfort. One day before tumor inoculation, mice were divided into three groups for study: vehicle control (saline), 50 mg/kg benserazide, or 250 mg/kg benserazide. Mice were administered treatment in 0.05 mL saline daily by intraperitoneal injection until procedure endpoint (60 days). Palpable tumor burden was determined weekly beginning at day 30. At the study endpoint, blood was collected for analysis of ODD, tumor was excised and weighed, and anatomy was analyzed for metastasis. To decipher between treatment-mediated effects and the lack of tumor uptake, mice without a palpable tumor were monitored for late tumor formation for 6 additional weeks without treatment after the study endpoint. Late tumor formation indicated that the lack of tumor at the study endpoint was due to treatment-mediated effects rather than a lack of tumor uptake; tumor and metastasis data collected at the study endpoint were then included in the analyses. Conversely, the absence of late tumor formation suggested the lack of tumor uptake, and data were thereby excluded from the analysis.

All statistical analyses were carried out using the Student t test with the following exceptions. The relationship between the average serum levels of 8oxodG and tumor weights was determined using the Pearson correlation coefficient (r). P values for in vivo data over time were relative to the vehicle control and determined by ANOVA.

We first established a high-throughput screening protocol to identify small molecules that activate repair of ODD. Our protocol consisted of screening a compound library using a DNA repair assay that evaluated BER of ODD while maintaining an intact cellular environment (23). Briefly, cells are delivered an ODD-containing GFP reporter gene via adenoviral-mediated gene transfer, allowed sufficient time to carry out repair, and then analyzed for fluorescence. In this system, ODD interferes with the transcription and subsequent expression of GFP. Therefore, cells with defective BER do not fluoresce. However, small molecules that facilitate repair of the ODD-containing GFP reporter gene may enable the expression of GFP in cells with defective BER and thereby produce a fluorescent signal. We used the BRCA1-mutant SUM149 human breast cancer cell line due to its susceptibility to adenoviral infectivity (required for delivery of the ODD reporter gene) and its inability to effectively repair ODD (required for detection of compound-facilitated BER; Supplementary Fig. S1; ref. 23). We chose to screen the LOPAC¹²⁸⁰ library (Sigma-Aldrich), which consisted of 1,280 compounds with known pharmacologic activity, in a quantitative high-throughput (qHT) format (25) starting at a 20 μmol/L final concentration of compound in each well (0.2% DMSO) followed by up to 6 (2-fold) dilutions. DMSO was the negative control and 18 μmol/L bromodeoxyuridine (BrdUrd) was the positive control. We identified BrdUrd in a preliminary screen (Supplementary Fig. S2) due to the lack of an existing but necessary small-molecule positive control.

We next adapted and validated our protocol for high-throughput screening (Supplementary Figs. S3 and S4). The protocol, as depicted in Fig. 1A, produced a Z′-factor = 0.6, coefficient of variance (CV) = 7%, S/B = 10, and no significant edge or drift effects (Supplementary Fig. S4C).

Finally, we carried out the validated protocol in 384-well qHT format. BRCA1-mutant SUM149PT cells were treated with up to 7 doses of compound (or control), subjected to the DNA repair assay, and assessed for GFP expression by 2 different means (i.e., imaging and plate reading). Greater GFP expression indicated greater enhancement of repair.

We then developed a scoring system to identify and prioritize active compounds based on their ability to enhance BER of ODD. The scoring system consisted of assigning “priority scores” based on the level of significance of each method of GFP detection (imaging > plate reading) and definition of an active compound (EC₅₀ > Activity_[High]) and then calculating a “summed priority score” that ranged from 1 to 10 for each molecule; higher scores indicated higher priority (Fig. 1B). Figure 1C and D depict the number of molecules for each assigned “priority score” and “summed priority score,” respectively. In general, the number of molecules increased as the score decreased. The high prevalence and low priority of molecules with a “summed priority score” of 1 resulted in their elimination; remaining molecules were defined as hits.

We identified 94 molecules that potentially enhanced BER of ODD (hit rate = 7.3%; Fig. 1D). Evaluation of the functional classes of these compounds revealed the greatest percentage of hits as having activity in DNA metabolism or DNA repair (Fig. 1E). Given that cytotoxic compounds may be undesirable for chemoprevention, we used our screening data acquired by imaging to identify compounds that decreased the number of Hoechst 33342–positive cells (i.e., living cells) by ≥20% compared with the negative control. Elimination of these potentially cytotoxic compounds resulted in a subset of 49 hits (hit rate = 3.8%; Fig. 1D). Detailed results for all 94 hits have been provided in Supplementary Table S1, including the EC₅₀ and% BER activity at high concentrations (for each method of GFP detection), the “summed priority score,” percentage of Hoechst-positive cells, and functional class. These molecules have been classified as “DNA repair–activating agents.”

We further analyzed 2 hits, acetohexamide and benserazide (Fig. 2), because they carried the highest summed priority score with the least amount of cytotoxicity (based on screening data). Both acetohexamide and benserazide carried summed priority scores of 7 and resulted in 87.8% and 85.9% of cells staining positive for Hoechst 33342 live cell dye, respectively. To validate activity, acetohexamide and benserazide were reordered from the original vendor and analyzed for BER activity and ODD levels. Acetohexamide produced a dose–response increase in repair of an oxidatively damaged GFP reporter gene in the mutant BRCA1-containing cell line SUM149PT (Fig. 2B). Likewise, benserazide produced a dose–response increase in repair of an oxidatively damaged GFP reporter gene in SUM149PT cells but not in wild-type BRCA1 MCF7 breast cancer cells. Additional breast cancer cell lines displayed similar results (Fig. 2D and E) using the same assay. Treatment of mutant BRCA1-containing cell lines with 20 μmol/L acetohexamide, that is, the highest concentration used in the HT screen, produced a statistically significant increase in BER (HCC1937, P = 0.01; SUM149, P = 0.007), whereas the same treatment in wild-type BRCA1-containing cell lines cells exhibited no effect on BER (BT474, P = 0.4; MCF7, P = 0.1; Fig. 2D). Similarly, treatment of mutant BRCA1-containing cell lines with 20 μmol/L benserazide produced a statistically significant increase in BER (HCC1937, P = 0.007; SUM1315, P = 0.01), whereas the same treatment in wild-type BRCA1-containing cell lines cells resulted in no effect on BER (BT474, P = 0.3; MCF7, P = 0.5; Fig. 2E). Acetohexamide and benserazide also decreased basal levels of 8oxoG lesions, that is, the most common form of ODD, in a time-dependent manner in SUM149 cells (Fig. 2F and G) and decreased ODD induced by 25 μmol/L H₂O₂ to levels similar to those of the undamaged control in 2 breast cancer cell lines with mutant BRCA1 (Fig. 2H and I). Taken together, these data suggest that acetohexamide and benserazide are 2 DNA repair-activating agents that function by enhancing BER of ODD in the presence of mutant, but not wild-type BRCA1.

Under normal conditions, DNA damage activates mechanisms of DNA repair (26); therefore, we tested whether or not acetohexamide or benserazide functioned by inducing DNA damage. When mutant BRCA1-containing cell lines were treated with 0.4 μmol/L acetohexamide or 20 μmol/L benserazide and then analyzed for DNA strand breaks using the alkaline comet assay, we found that neither drug induced DNA strand breaks relative to the vehicle control (Fig. 2J and K). Therefore, DNA repair–activating agents, acetohexamide and benserazide, increase BER directly, rather than indirectly through induction of DNA damage.

For purposes of comparison, we also tested the effect of tamoxifen on BER (due to it being the current FDA-approved chemoprevention agent for breast cancer). We used the highest concentration of tamoxifen that did not produce a statistically significant decrease in cell viability in breast cancer cell lines under our experimental conditions (i.e., 1 μmol/L). Unlike DNA repair–activating agents, treatment with tamoxifen did not significantly increase repair of an oxidatively damaged GFP reporter gene in SUM149 cells (P = 0.28; Fig. 2L).

Anchorage independence is a characteristic of transformed cells that often indicates tumorigenic potential. We asked whether or not DNA repair–activating agents could reduce anchorage-independent growth. Therefore, we treated breast cancer cells with benserazide or vehicle control and measured colony formation in soft agar. Benserazide significantly decreased colony formation in a mutant BRCA1-containing cell line (HCC1937, P = 0.0003) as well as a wild-type BRCA1-containing cell line (MCF7, P = 0.03), with the effect being significantly greater for the mutant BRCA1-containing cell line than for the wild-type cell line (P = 0.001; Fig. 3A). Similarly, benserazide reduced colony formation in MCF7 cells stably transduced with shRNA to BRCA1 (P = 0.03) or a nontargeting control (P = 0.02), with the effect being significantly greater in those cells containing shRNA to BRCA1 compared with the control (P = 0.005; Fig. 3B). We next assessed whether the reduction in colony formation was due to a decrease in anchorage-independent growth or a decrease in cell viability. Therefore, we carried out the soft agar colony formation assay in parallel to the trypan blue exclusion assay for the wild-type and mutant BRCA1-containing cell lines and found that benserazide did not significantly affect cell viability in either cell line (MCF7, P = 0.2; HCC1937, P = 0.4; Fig. 3C).

To compare benserazide with the existing FDA-approved chemoprevention agent for breast cancer, we also examined the effect of tamoxifen on anchorage-independent growth using the soft agar colony formation assay. Tamoxifen showed no effect on colony formation in mutant BRCA1-containing cells (HCC1937, P = 0.7) but significantly decreased colony formation in wild-type BRCA1-containing cells (MCF7, P = 0.0003; Fig. 3A). Likewise, tamoxifen did not significantly decrease colony formation in MCF7 cells transduced with shRNA to BRCA1 (P = 0.2) but significantly decreased colony formation in cells with the nontargeting control (P = 0.03; Fig. 3B). A parallel experiment using the trypan blue exclusion assay revealed that tamoxifen did not affect cell viability in either mutant (HCC1937, P = 0.3) or wild-type (MCF7, P = 0.6) BRCA1-containing cells (Fig. 3C) but rather reduced colony formation in wild-type BRCA1-containing cells due to a decrease in anchorage-independent growth.

Taken together, benserazide was more effective than tamoxifen at decreasing the tumorigenic potential of cell lines with mutant or deficient BRCA1 but not wild-type BRCA1.

Tumor progression is the final step of tumorigenesis that allows cells to develop and maintain malignant properties. Therefore, we carried out an in vivo study to analyze characteristics of tumor progression, including tumor formation, growth, and metastasis. We used a xenograft mouse model and began treatment with DNA repair–activating agent 24 hours before inoculation of mutant BRCA1 HCC1937 cells into the mammary fat pad. Treatment with vehicle control, 50 mg/kg benserazide, or 250 mg/kg benserazide continued daily for up to 60 days while monitoring for tumor formation by palpitation. After 60 days (i.e., study endpoint), we collected serum for analysis of 8oxodG levels, excised and weighed tumors (if present), and analyzed metastasis by necropsy. If no tumor was present, mice remained under surveillance for tumor formation in the absence of treatment for up to 6 weeks after the study endpoint to distinguish between unsuccessful tumor inoculation and drug-mediated effects. Mice that did not form tumors were eliminated from the data pool presumably due to the lack of tumor uptake following inoculation rather than an antitumor effect by the drug. This model system feasibly allowed for completion of an in vivo study that mimicked chemoprevention rather than therapeutic conditions.

We found that benserazide delayed the rate of tumor formation in mice in a dose–response manner (Fig. 3D). Compared with the vehicle control, treatment with 50 and 250 mg/kg benserazide increased the percentage of tumor-free mice over time (P = 0.27 and 0.047, respectively). We also discovered that benserazide reduced tumor growth as evidenced by a dose–response decrease in the average weight of tumors at the study endpoint (Fig. 3E). Interestingly, benserazide simultaneously increased serum levels of 8oxodG (Fig. 3F), a byproduct of DNA repair of ODD. Indeed, the average tumor weight negatively correlated with the average serum levels of 8oxodG among treatment groups (r = −0.82), indicating that the reduction in tumor growth was proportional to benserazide-mediated BER. Finally, unlike the vehicle control, both concentrations of benserazide prevented metastasis (Fig. 3G).

Benserazide and its analogue, carbidopa (Fig. 4A), are both known to function as DOPA decarboxylase inhibitors. Benserazide and carbidopa were also both included among the compounds screened for BER-enhancing activity (Fig. 1). However, our screen did not identify carbidopa as a hit (Fig. 4B). Therefore, we first validated our screening data by subjecting the mutant BRCA1-containing SUM149 cell line to the BER assay following pretreatment with vehicle control or carbidopa. Given that benserazide is typically administered at approximately half the dose of carbidopa (27), we used carbidopa at the dose equivalent to the effective dose of benserazide (10 μmol/L) and at the highest screened concentration or twice the effective dose of benserazide (20 μmol/L). Indeed, carbidopa exhibited no effect on repair activity compared with the vehicle control (10 μmol/L, P = 0.4; 20 μmol/L, P = 0.3; Fig. 4C). We next asked whether carbidopa had an effect on tumorigenic potential by measuring anchorage-independent growth. The soft agar colony formation assay revealed that compared with the vehicle control, treatment with 10 or 20 μmol/L carbidopa had no significant effect on colony formation of MCF7 cells transduced with either shRNA to BRCA1 (10 μmol/L, P = 0.09; 20 μmol/L, P = 0.1) or a nontargeting control (10 μmol/L, P = 0.7; 20 μmol/L, P = 0.4; Fig. 4D). Taken together, unlike benserazide, its carbidopa analogue does not function as a DNA repair–activating agent or decrease anchorage-independent growth. These data suggest that the benserazide-mediated effects on BER of ODD and tumorigenic potential are independent of DOPA decarboxylase inhibition.

BRCA1 mutation or deficiency leads to defective BER and excessive ODD (23, 24), which, in turn, leads to tumorigenesis. We have discovered small molecules that enhance BER of ODD and reduce malignant properties associated with tumorigenesis, including anchorage-independent growth and tumor progression. While previous studies have reported small molecules that inhibit DNA repair (28), this study, to our knowledge, reports small molecules that activate repair for the first time. These DNA repair–activating agents may potentially be used for the chemoprevention of BRCA1-associated malignancies.

A high-throughput chemical screen identified a novel class of drugs termed “DNA repair–activating agents.” We pursued two of these agents, acetohexamide and benserazide, because of their predicted activity, lack of potential cytotoxicity, and clinical potential. Both acetohexamide and benserazide enhanced BER (Fig. 2B and E) as well as decreased basal levels and H₂O₂-induced levels of ODD (Fig. 2F–I). These effects were observed in a dose–response and/or time-dependent manner and were specific to mutant BRCA1-containing cells. Neither drug induced DNA damage, including single- or double-strand breaks (Fig. 2J and K), suggesting that the ability to enhance DNA repair was due to direct means as opposed to indirect means of damage-induced repair. Both acetohexamide and benserazide also exhibited minimal cytotoxicity at concentrations that enhanced BER of ODD (Fig. 3C and Supplementary Fig. S5) and lacked chemical, toxicologic, and environmental health concerns according to the U.S. National Library of Medicine TOXNET database. Furthermore, we derived DNA repair–activating agents by screening compounds with known pharmacologic activity, thereby facilitating their translation into the clinic. For example, acetohexamide and benserazide have already been through human trials and have been approved for clinical use in the United States for diabetes and in the United Kingdom and Canada for Parkinson disease, respectively. Taken together, DNA repair–activating agents may potentially be repurposed to reverse a defect in DNA repair associated with BRCA1-mediated tumorigenesis.

The DNA repair–activating agent benserazide decreased malignant properties associated with tumorigenesis. Our data showed that compared with vehicle controls, benserazide reduced anchorage-independent growth in vitro as well as delayed tumor formation and decreased the average weight of tumors in vivo when administered before tumor implantation (Fig. 3A–D). We observed a statistically significant decrease in the rate of tumor formation for mice treated with the higher dose (250 mg/kg) but not the lower dose (50 mg/kg) of benserazide. The lack of a statistically significant difference in the latter case may have been due to small sample sizes. Nonetheless, the effects of benserazide on tumor formation and growth occurred in a dose–response manner. Importantly, an increase in levels of a serum marker for repair of ODD (8oxodG) correlated with the decrease in tumor weight in mice treated with benserazide (r = −0.82; Fig. 3E and F). On the other hand, carbidopa, a benserazide analogue, did not exhibit BER activity (Fig. 4B and C) and did not reduce in vitro tumorigenic potential (Fig. 4D). Together, these data provide evidence for a link between enhancing BER and preventing tumorigenesis. Interestingly, treatment with benserazide also prevented metastasis (Fig. 3G), suggesting that DNA repair–activating agents may not only be useful before a cancer diagnosis for chemopreventive purposes but also after a diagnosis for the treatment of BRCA1-mutated breast cancers. Through its role in maintaining genetic stability, BRCA1 has been implicated in both the initiation and progression steps of tumorigenesis (21, 29, 30). However, we only captured data for tumor progression because our assays used cancer cell lines that have already undergone tumor initiation. It remains possible that benserazide or other DNA repair–activating agents compensate for mutant BRCA1 function at the earlier step of tumor initiation. A model system that uses primary cultures with a single BRCA1 mutation or a mouse model with a conditional Brca1 mutation (31) will better serve to answer this question, but unfortunately, these models are limited by uncertainties of tumor formation and long latency periods. Furthermore, we avoided data that were skewed in favor of treatment-mediated effects due to our use of established cell lines by monitoring for tumor formation after the study endpoint and eliminating data obtained from mice that did not successfully experience tumor uptake upon inoculation.

Our study suggests that DNA repair–activating agents may provide a better option than tamoxifen for the chemoprevention of BRCA1-associated breast cancers. While benserazide significantly enhanced BER of ODD (Fig. 2C, E, G, and I) and decreased malignant properties associated with tumorigenesis (Fig. 3A, B, and D–F), tamoxifen did not enhance BER of ODD in BRCA1-mutant cells (Fig. 2L), nor did it effectively reduce tumorigenic potential in BRCA1-mutant or -deficient cells (Fig. 3A and B). However, tamoxifen significantly reduced tumorigenic potential in wild-type BRCA1-containing cells (Fig. 3A and B), which is consistent with published studies that examined its effectiveness in breast cancer populations (14). The use of tamoxifen for breast cancer prevention is often avoided by patients because it increases the risk for endometrial carcinoma (32–36). While tamoxifen is a well-established carcinogen, neither acetohexamide nor benserazide has been reported to exhibit carcinogenic potential (37, 38). Therefore, DNA repair–activating agents provide an added advantage to carriers of BRCA1 mutations who are naturally susceptible to carcinogen-induced DNA damage.

The molecular target(s) of DNA repair–activating agents that are responsible for enhancing BER remain unknown. Given that the chemical screen used to identify these agents used a library of compounds with known pharmacologic activity, DNA repair–activating agents have well-established molecular targets. However, the molecular targets responsible for BER-enhancing activity are independent of their established targets because functional analogues of DNA repair–activating agents do not enhance BER or prevent tumorigenic potential. Both benserazide (used in the United Kingdom and Canada) and carbidopa (the equivalent used in the United States) target DOPA decarboxylase in the periphery. There is no significant difference in the therapeutic effects for Parkinson disease between these two drugs (27). However, unlike benserazide (Figs. 2C, E, G, I and 3), carbidopa did not enhance BER, nor did it reduce tumorigenic potential (Fig. 4). Likewise, the chemical screen used to identify DNA repair–activating agents (Fig. 1A and B) tested 6 sulfonylureas that stimulate insulin release via ATP-dependent K+ channels, but only two of these antidiabetic molecules, acetohexamide (strong hit) and glipizide (weak hit), showed BER-enhancing activity (Figs. 2B, D, F, H and Supplementary Table S1). Furthermore, if acetohexamide functioned through its known/antidiabetic mechanism, one would predict all sulfonylureas to reduce the risk for breast cancer, which is not the case (39). Examination of an NCBI GEO dataset for MCF7 cells treated with benserazide revealed an upregulation of BER genes previously shown to be transcriptionally activated by BRCA1, including OGG1 by 35-fold, NTH1 by 130-fold, and APE1 by 1049-fold (24, 40). Therefore, DNA repair–activating agents may function by increasing the expression of one of these molecules, although it is unlikely to be OGG1 given that we found no significant difference in the effect of benserazide or acetohexamide on BER activity between shOGG1-containing and nontargeting control MCF7 cells (data not shown). Because these data reflect activity in the presence of wild-type BRCA1, extensive studies are underway to reveal the molecular targets of DNA repair–activating agents in mutant BRCA1-containing cells. DNA repair–activating agents are expected to directly or indirectly target a required enzyme(s)/protein for BER rather than an accessory protein because these agents were identified using an endpoint (not rate-determining) assay for BER (ref. 23 and Fig. 1A). They may inhibit a negative regulator of BER, as most drugs function through inhibition, or they may activate any positive regulator of BER. For example, evidence suggests that altering protein–protein interactions or the acetylation state of APEX1 endonuclease may stimulate BER activity (41). Although DNA repair–activating agents were identified using an assay that enriched for molecules that alter BER activity, ODD may convert to single- or double-strand breaks, and thus DNA repair–activating agents may alter other repair processes in which BRCA1 also plays a role, such as homologous recombination, nonhomologous end-joining, or nucleotide excision repair (42–46). However, this possibility is unlikely due to our finding that both benserazide and acetohexamide reduced levels of ODD (Fig. 2F–I) and not strand breaks (Fig. 2J and K).

In summary, this study describes a novel class of drugs that target a defect in repair associated with mutations in BRCA1. Our data suggest that these drugs, that is, DNA repair–activating agents, may be used to prevent or slow the progression of primary or secondary breast tumors or even delay metastasis. Therefore, these findings support further studies to evaluate the use of DNA repair–activating agents for the chemoprevention of BRCA1-mutated cancers. Clinical studies are best suited following the development of better chemoprevention models for BRCA1-mutated cancers and upon a better understanding of the molecular targets of these drugs. However, benserazide is typically administered with l-DOPA and has little effect on its own and thus is primed for clinical translation. Other DNA repair–activating agents should first undergo structure–activity relationship studies to identify an appropriate analogue that maintains its DNA repair activity without any adverse effects. For example, acetohexamide should undergo modification to eliminate its glucose-lowering activity to minimize the possibility of inducing hypoglycemia in healthy patients being treated for cancer prevention. Finally, these drugs may be effective for other malignancies associated with oxidative stress, including other cancers, degenerative diseases, and sclerotic disorders.

E. Alli and J. Ford have ownership interest in a pending patent. No potential conflicts of interest were disclosed by the other authors.

Conception and design: E. Alli, J.M. Ford

Development of methodology: E. Alli, D. Solow-Cordero, J.M. Ford

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Alli, D. Solow-Cordero, S.C. Casey, J.M. Ford

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Alli, D. Solow-Cordero, S.C. Casey, J.M. Ford

Writing, review, and/or revision of the manuscript: E. Alli, J.M. Ford

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Solow-Cordero, S.C. Casey

Study supervision: J.M. Ford

The authors thank Allison Kurian at the Stanford Women's Cancer Center for helpful discussion on the article and Jason Wu at the Stanford High-Throughput Bioscience Center for technical assistance with the chemical screen. They also thank Shaveta Vinayak and Kedar Hastak for advice and technical assistance with the in vivo protocol.

Support for this study includes a Susan G. Komen for the Cure Postdoctoral Fellowship (E. Alli), Mary Kay Ash Charitable Foundation Grant, and U.S. NIH R21 grant (1R21NS061674-01A2).

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.