**Background:** RAD52 is a human DNA repair protein that is a coveted target for anticancer drug discovery because its pharmacological inhibition is synthetically lethal with defects in BRCA1, BRCA2, and PALB2—mutations found in approximately 25% of breast and ovarian cancers. Unlike many DNA repair proteins, RAD52 inactivation is well tolerated in normal cells, making it an attractive therapeutic target. However, RAD52 lacks enzymatic activity and its functions depend on binding single-stranded DNA (ssDNA), making it challenging to develop small molecule inhibitors that disrupt protein–DNA interactions. Previous work identified epigallocatechin (EGC) as a selective RAD52 inhibitor, but natural products like EGC have poor drug-like properties. This study aimed to perform scaffold hopping from EGC into drug-like chemical space to develop synthetically tractable RAD52 inhibitors with improved potency and selectivity.
**Methods:** The authors employed a multi-tiered computational workflow using the Enamine REAL database of drug-like compounds. Shape similarity screening based on the bound EGC–RAD52 complex narrowed 3 million compounds to ~7,700, which were docked into the EGC binding pocket on RAD52. Two scoring approaches were used: (1) a classic docking/scoring function (London dG + MM/GBSA) and (2) Protein–Ligand Interaction Fingerprints (PLIF) based on EGC contacts. The top three compounds from each method were synthesized. For scaffold expansion, the program InfiniSee searched ~3.2 × 10^10 virtual products for compounds with >97% chemical similarity to Z99, yielding six derivatives. Compounds were tested using FRET-based assays to measure disruption of RAD52–ssDNA, RPA–ssDNA, and RAD52–RPA–ssDNA complexes. Direct binding was confirmed by intrinsic tyrosine fluorescence quenching and mass photometry. Cell viability was assessed using CellTiter-Glo and clonogenic survival assays in BRCA1-mutant (MDA-MB-436), BRCA2-mutant (EUFA423F, Capan1), BRCA2-complemented (EUFA423F HA), and BRCA-proficient (MCF10a) cell lines. RAD52-specific cellular functions were evaluated using neutral comet assays (DSB formation) and native IdU assays (nascent ssDNA exposure) in MRC5SV40 fibroblasts with and without RAD52 depletion.
**Key Results:** Six initial compounds were identified with IC50 values for RAD52–ssDNA inhibition ranging from ~23 to 1200 μM. Z56 (docking/scoring hit) showed IC50 of 35.1 ± 2.6 μM for RAD52–ssDNA and did not inhibit RPA–ssDNA (no IC50 measurable), while Z99 (PLIF hit) showed IC50 of 50.5 ± 5.8 μM for RAD52–ssDNA but also inhibited RPA–ssDNA (IC50 14.0 ± 1.8 μM). Both compounds directly bound RAD52 as confirmed by intrinsic fluorescence quenching. In cell viability assays, Z56 selectively killed BRCA1-mutant (MDA-MB-436) and BRCA2-mutant (EUFA423F, Capan1) cells with no toxicity to BRCA2-complemented or MCF10a cells, even at 250 μM. Z99 also killed BRCA-mutant cells but showed toxicity toward BRCA-proficient cells at 100 μM. In neutral comet assays, both Z56 and Z99 (at 10 μM) significantly reduced HU-induced DSBs in RAD52-proficient cells but failed to further reduce DSBs in RAD52-depleted cells, confirming target specificity. In native IdU assays, both compounds induced nascent ssDNA exposure at stalled forks in RAD52-proficient but not RAD52-depleted cells. In situ PLA showed Z56 suppressed RAD52 binding to parental ssDNA. Scaffold expansion of Z99 yielded six derivatives with improved RAD52–ssDNA inhibition (IC50 ~2–8 μM, a 6–40-fold improvement over Z99). Importantly, these derivatives showed reduced or similar RPA inhibition compared to Z99. At 10 μM, compounds Z58-54, Z58-72, Z58-83, and Z58-87 selectively killed BRCA1-mutant and BRCA2-mutant cells while showing no toxicity to BRCA2-complemented or MCF10a cells. Z58-87 showed the most consistent activity across all assays.
**Clinical Implications:** This study demonstrates the feasibility of developing drug-like small molecule inhibitors that selectively disrupt the RAD52–ssDNA interaction, a challenging protein–DNA target. The optimized Z99 derivatives (particularly Z58-54 and Z58-87) show low micromolar potency, excellent selectivity for RAD52 over RPA, and selective toxicity toward BRCA1/BRCA2-mutant cancer cells with minimal effects on normal cells. These compounds provide a roadmap for next-generation cancer therapeutics targeting the synthetic lethality between RAD52 and BRCA deficiency. Given the emergence of resistance to PARP inhibitors, these RAD52 inhibitors could expand the pharmacopeia for treating BRCA-mutant cancers, including breast, ovarian, and pancreatic cancers. The computational workflow—from natural product pharmacophore to drug-like inhibitors via ultra-large virtual chemical space screening—represents a broadly applicable strategy for targeting other challenging protein–nucleic acid interactions.