A first-in-class Polymerase Theta Inhibitor selectively targets Homologous-Recombination-Deficient Tumors

Abstract

DNA polymerase theta (POLθ) is synthetic lethal with Homologous Recombination (HR) deficiency and thus a candidate target for HR-deficient cancers. Through high-throughput small molecule screens we identified the antibiotic Novobiocin (NVB) as a specific POLθ inhibitor that selectively kills HR-deficient tumor cells in vitro and in vivo. NVB directly binds to the POLθ ATPase domain, inhibits its ATPase activity, and phenocopies POLθ depletion. NVB kills HR-deficient breast and ovarian tumors in GEMM, xenograft and PDX models. Increased POLθ levels predict NVB sensitivity, and BRCA-deficient tumor cells with acquired resistance to PARP inhibitors (PARPi) are sensitive to NVB in vitro and in vivo. Mechanistically, NVB-mediated cell death in PARPi-resistant cells arises from increased double-strand break end resection, leading to accumulation of single-strand DNA intermediates and non-functional RAD51 foci. Our results demonstrate that NVB may be useful alone or in combination with PARPi in treating HR-deficient tumors, including those with acquired PARPi resistance. (151/150).

Keywords: Fanconi Anemia; HRD cancer; Homologous Recombination; MMEJ; Novobiocin; PARP inhibitor resistance; Polymerase theta (POLθ).

Extended Data Fig. 1. Characterization of POLθ inhibitors obtained from the small-molecule screen, with biochemical assays and cell-based assays.Characterization of POLθ inhibitors obtained from the small-molecule screen, with biochemical assays and cell-based assays.

a, ADP-Glo assay for quantification of POLθ ATPase activity in various experimental conditions. N = 1 experiment. b, A secondary screen using the 72 initial hits in reducing POLθ ATPase activity (below 70 %, z-score < −4). The screen was done in the presence and absence of ssDNA. The four most promising hits advancing for further analysis are labeled (see Table S4 for analysis). Aurintricarboxylic acid (ATA), Reactive Blue 2 (RB), Suramin (SUR), and Novobiocin (NVB). Data shown are mean of n=2 independent experiments. c, Quantification of POLθ and SMARCAL1 ATPase activity with indicated small molecules. The negative control is an inert small molecule (Vandetanib). Data shown are mean ± SD, n=3 independent experiments. Statistics were performed using two tailed t-test with Welch’s correction. ***p < 0.001, **p < 0.01. d. Coomassie blue stained gels showing purified ATPase that were used in this study. e, Conjugation of NVB to epoxy-activated Sepharose-6B (Sigma). NVB conjugated beads showed light yellow color. The structure of NVB is shown. f, Pulldown experiments with NVB-conjugated beads and purified POLθ, SMARCAL1, CHD1, BLM, and RAD51. POLθ, SMARCAL1, CHD1, and BLM were detected by Western blot using anti-Flag antibody (Sigma #F1804) in this assay. RAD51 was detected by anti-RAD51 antibody (SantaCruz #398587). g, Pulldown experiments with NVB-conjugated beads and cell lysate from HEK293T cells expressing GFP-tagged full-length POLθ. Cell lysates were incubated with empty beads (EB), NVB-beads (NVB) or anti-GFP-beads (GFP) to assay for direct POLθ binding. GFP-POLθ bound to beads and input of each group were subjected to Western blot analysis, using an anti-GFP antibody. h, Competition assay between NVB-conjugated beads and free NVB when incubated with GFP-tagged full-length POLθ extracted from HEK293T cells. i, Thermal shift assay using GFP-POLQθ expressing HEK293T cell lysate incubated with NVB at the indicated temperatures. Supernatants after heat treatment were subjected to Western blot using anti-GFP and anti-Actin antibodies. J-l, Titration of indicated NVB (or DMSO) concentration with POLθ (j), BLM (k), or MRE11 (l) in the thermal shift assay to measure the effect of NVB on the protein stability. Average first derivative of the scattering profile is shown. Scattering peak max values are shown as mean ± SD, from n=3 independent experiments.

Extended Data Fig. 2. Novobiocin phenocopies POLθ depletion in human cells. Novobiocin phenocopies POLθ depletion in human cells.

a, Images and quantification of RAD51 foci in U2OS cells under increasing concentrations of NVB with or without gamma-irradiation (IR). b, Images and quantification of γH2AX foci in U2OS cells under increasing concentrations of NVB with or without IR. Data in (a) and (b) were pooled from two independent experiments. Each dot represents foci numbers

Extended Data Fig. 3. Efficacy of NVB in BRCA1^−/− GEMM and RAD51 PD study in TOV21G xenograft models. Efficacy of NVB in BRCA1^−/− GEMM and RAD51 PD study in TOV21G xenograft models.

a, NVB efficacy in the GEMM model (Tp53^−/−Brca1^−/− TNBC). Response of each tumor in each individual mouse is shown. Tumor chunks from GEMM mice were implanted in syngeneic FVB/129P2 mice, which were treated with PBS or 100 mg/kg NVB twice a day via IP for 5 weeks. b, Olaparib sensitivity of TOV21G (+ EV) or FANCF-complemented TOV21G (+ FANCF cDNA). c, Novobiocin sensitivity of TOV21G (+ EV) or FANCF-complemented TOV21G (+ FANCF cDNA). Mean ± SD of n=6 biologically independent samples are shown in b and c. d-f, Immunohistochemical (IHC) study of the pharmacodynamic biomarker RAD51 after NVB and/or olaparib treatment in TOV21G tumors. d, Representative images of RAD51 IHC staining in xenografted TOV21G tumor cells. Tumor bearing NU(NCr)-Foxn1nu mice were treated with indicated drugs for 18 days before tumors were taken. FFPE tissue sections of the tumors were stained using an anti-RAD51 antibody and representative images (40X) are shown. e, Zoom-in of the IHC images in (d) to show RAD51 foci in detail. f, Quantification of RAD51 foci positive cells in TOV21G tumors. Cells with three or more RAD51 foci were counted as positive cells. Tumor samples from each group were processed and analyzed, and Mean±SD are shown, n=4 for Olaparib and n=6 for other groups. The total number of cells counted were shown in the graph (N). Statistical analysis was performed using one-way ANOVA in Prism, *, p < 0.05; **, p < 0.01; ***, p < 0.001.

Extended Data Fig. 4. Effects of novobiocin in BRCA1^−/− and BRCA2^−/− cells. Effects of novobiocin in BRCA1^−/− and BRCA2^−/− cells.

a, Clonogenic survival of BRCA1^−/− and WT RPE1 cells under increasing concentrations of the PARPi rucaparib. b, Clonogenic survival of BRCA2^−/− and WT RPE1 cells under increasing concentrations of PARPi rucaparib. c, Clonogenic survival of BRCA2^−/− and WT RPE1 cells under increasing concentrations of POLθ inhibitor. The survival fraction in a-c is normalized to the untreated samples. Data shown are Mean ± SEM, n = 3 independent experiments. d, Cell viability assay (CellTiter-Glo) in BRCA1^−/− and WT RPE1 cells treated with indicated POLθ inhibitors (50 μM) or with the PARPi Olaparib as control. Cells were treated twice on days 1 and 4 and cell viability was measured on day 7. Data were mean ± SD, n = 3 biological replicates. Statistics analyses were two tailed t-test. *, p < 0.05; **, p < 0.01; ***, p < 0.001, ns, not significant (p > 0.05).

Extended Data Fig. 5. NVB induces apoptosis and chromosome aberrations in RPE1 BRCA1^-/−. NVB induces apoptosis and chromosome aberrations in RPE1 BRCA1^-/−.

a, Quantification of apoptosis in BRCA1^−/− and WT RPE1 cells under increasing concentrations of NVB. Mean ± SD, n = 3 independent experiments. b, Representative images of chromosomal aberrations (blue arrows) including radial figures (red arrows) in RPE1 (WT or BRCA1^−/−) cells treated with NVB. c-d, Quantification of radial figures (c) and total chromosome aberration (d) in BRCA1^−/− and WT RPE1 cells treated with NVB alone or in combination with mitomycin C (MMC). Data in c-d are mean ± SD of n = 3 independent experiments. Statistical significance was determined by one-way ANOVA multiple comparisons using uncorrected Fisher’s LSD test. ***, p < 0.001.

Extended Data Fig. 6. POLQ^−/− U2OS cells were resistant to NVB. POLQ^−/− U2OS cells were resistant to NVB.

a-d, Verification of the U2OS-POLQ^−/− cell line. a, Mapping the genetic alteration (deletions) in the U2OS POLQ knockout clone generated by CRISPR-Cas9. The targeted region was sequenced by Next Gen sequencing (number of reads are shown), and the deletions were mapped to POLQ locus of human genome on chromosome 3. b, A Western blot showing POLθ protein expression in WT and POLQ^−/− U2OS cells. c, RAD51 foci assay in WT or POLQ^−/− U2OS cells before and after (4 h) of 4 Gy IR. Mean of n=2 independent experiments is shown. d, IR sensitivity of WT and POLQ^−/− U2OS cells in clonogenic assays. Mean ± SD of n=3 independent experiments are shown. e, NVB sensitivity of WT and POLQ^−/− U2OS cells in clonogenic assays. Mean ± SD of n=6 biologically independent samples are shown.

Extended Data Fig. 7. NVB inhibits POLθ but not HSP90 or TOP2 in human cells. NVB inhibits POLθ but not HSP90 or TOP2 in human cells.

a, HSP90 client degradation assay in RPE1 and RPE1-BRCA1^−/− cells. Cells were treated with a potent HSP90 inhibitor PU-H71 or NVB for 48 hours, and then cells were collected for Western blot analysis of the HSP90 client AKT1. b-c, HSP90 client degradation assay in MCF7 cells. Cells were treated with the potent HSP90 inhibitor PU-H71 or NVB for 24 hours, and then cells were collected for Western blot analysis of the HSP90 client proteins AKT1, CDK6 (b) and BRCA1 (c). Levels of HSP90 and HSP70 were also analyzed after NVB or PU-H71 treatment (c). d-e, Combination effect of etoposide and novobiocin in killing TOV21G cells (d) and CAPAN1 cells (e), showing their non-epistatic effects. Mean ± SD of n=3 independent experiments are shown. f, CellTiter-Glo cell viability assay of empty vector (EV) and FANCF-complemented TOV21G cells, in the presence of olaparib, novobiocin or both. Mean ± SD, n = 4 biological replicates are shown. g, IC₅₀ values of olaparib in TOV21G cells with or without NVB, derived from data in f. Mean ± SD of IC₅₀ were shown, from n = 4 biological replicates.

Extended Data Fig. 8. Characterization of PARP inhibitor resistant clones derived from RPE1-BRCA1^-/−.Characterization of PARP inhibitor resistant clones derived from RPE1-BRCA1^-/−.

a-c, Olaparib (a-b) and cisplatin (c) sensitivity of WT and BRCA1^−/− RPE1 cells and the PARPi resistant clones of BRCA1^−/− RPE1. Mean ± SD of biological replicates are shown, with n=8 in A, n=3 in B, and n=8 in C. d. DNA fiber assay to measure the replication fork stability of the PARPi resistant clones. Mean±95%CI are shown from 3independent experiments are shown, with n = number of fibers scored labeled. e. RAD51 foci analysis in WT and BRCA1^−/− RPE1 cells and the PARPi resistant clones after irradiation (5Gy), to determine restoration of RAD51 foci in R clones. RAD51 was stained 4 hours after IR. Scale bar = 10 μm. f, A Western blot of WT and BRCA1^−/− RPE cells and the PARPi resistant clones using a REV7 antibody. g, Immunofluorescence staining of 53BP1 in WT and BRCA1^−/− RPE1 cells and the PARPi resistant clones after irradiation. Clone R4 showed much reduced 53BP1 foci after IR. h, Immunofluorescence staining of BRCA1 in WT and BRCA1^−/− RPE1 cells and the PARPi resistant clones after irradiation. No BRCA1 foci was observed except in WT RPE1 cells. i, CellTiter-Glo assay to determine survival of PARPi-resistant and parental BRCA1^−/− RPE1 cells upon CRISPR knockout of the POLQ gene. Data are mean ± SEM, with n = 4 biological replicates. Statistical analysis was t-test, *, p < 0.05; ns, not significant.

Extended Data Fig. 9. PARPi resistant MDA-MB-436 and UWB1.289 cells are sensitive to NVB, and NVB has synergy with PARPi in TOV21G cells. PARPi resistant MDA-MB-436 and UWB1.289 cells are sensitive to NVB, and NVB has synergy with PARPi in TOV21G cells.

a-b, Olaparib (a) and NVB (b) sensitivity of UWB1.289 and a PARPi resistant UWB1.289 clone (UWB1.289-YSR12) in CellTiter-Glo assays. Data shown are mean ± SD, n = 4 biological replicates. c. Western blot analysis of POLQ expression levels in in WT and BRCA1^−/− RPE1 cells and the PARPi resistant BRCA1^−/− clones. d. Olaparib sensitivity of PEO1 (BRCA2 mutated) and PEO4 cells (BRCA2 restored). e. NVB sensitivity of PEO1 and PEO4 cells. Data shown in d and e are Mean ± SD, n= 3 biological replicates. f. A Western blot shows POLQ expression in BRCA2-decient cells lines (PEO1) and their counterparts with BRCA2 reverted to wild type (PEO4).

Fig. 1.. A small-molecule screen identifies Novobiocin (NVB) as a specific POLθ ATPase inhibitor that kills HR-deficient tumors.

a, The domain structures of full length (FL) and ATPase domain (ΔPol, a.a. 1–987) of POLθ, a Coomassie-stained gel of the purified POLθ ATPase domain used for screen, and a schematic of the small molecule screen. b, Results of the small-molecule screen. Shown is POLθ ATPase activity in the presence of small-molecule libraries (enriched with bioactive compounds). Four top hits verified in the secondary screen (Extended Data Fig. 1b) were labeled: reactive blue 2 (RB), suramin (SUR), novobiocin (NVB) and aurintricarboxylic acid (ATA). Data are mean from n = 2 replicates. c, Quantification of the ATPase activity of POLθ, SMARCAL1, CHD1, BLM, TRIP13, RAD51 and HSP90AA1 with increasing concentration of NVB. ATPase activity was determined by ADP-Glo assay (Promega). Activities were normalized to DMSO (0.1%). Mean of 6 replicates from n = 2 two independent experiments are shown. d, The NVB binding tunnel in POLθ ATPase domain (shown in surface - PDB 5AGA) was predicted by extra precision glide docking and lowest binding free energy from prime MM-GBSA calculations to have multiple hydrogen bonds and close hydrophobic packing. Top and side views showing NVB docking into the tunnel. Binding modes of NVB (green sticks) and AMP-PNP (magenta sticks) in POLθ helicase domain (PDB 5AGA) with green sphere showing active site Mg²⁺ ion. e, Quantification and representative images of POLθ-GFP accumulation at sites of laser micro-irradiated DNA damage in RPE1-POLQ^−/− cells overexpressing GFP-tagged POLQ WT, polymerase mutant, ATPase/Helicase mutant, or double mutant. GFP-POLQ WT expressing cells treated with DMSO (0.1%) or 100 μM of NVB were shown. Mean ± SEM were shown. WT(+DMSO), n = 7 cells; WT(+NVB), n = 8 cells; Pol dead, n = 24 cells; ATPase dead, n = 13 cells; Pol/ATPase dead, n= 14 cells, from three independent experiments. GFP only, n = 4, from two independent experiments. f, MMEJ and HR repair reporter assays in U2OS cells treated with increasing concentration of NVB. Percentage of GFP positive cells are shown as pathway efficiency. Mean ± SD, from n = 6 (for MMEJ) and n = 3 (for HR) from three independent experiments, ordinary one-way ANOVA. g, Tumor growth of the GEMM model (Brca1^−/− TNBC) after treatment with vehicle (PBS) or NVB. Tumor bearing FVB/129P2 were treated with PBS or 100 mg/kg NVB via IP injection twice a day for 5 weeks. Mean ± SEM is shown, n = 6 mice for PBS group and n = 7 mice for NVB group. h, Survival plot of the experiment shown in (g). Median survival and p value are shown, Statistical analyses in g and h were two-tailed t-test. *, p < 0.05. **, p < 0.01.

Fig. 2.. NVB kills HR-deficient tumors in vivo and in vitro and synergizes with PARPi.

a, Tumor growth of the xenograft model with TOV21G (FANCF deficient ovarian cancer) or FANCF-complemented TOV21G cells. The tumor bearing mice were treated with vehicle or 100 mg/kg NVB via IP injection twice a day for 4 weeks. Fold change of tumor was calculated using the formula Fold = T/T0, where T is the tumor size at given time and T0 is the initial tumor size. Data are Mean ± SEM, from n = 10 mice in each group. Statistical analysis was Dunn’s multiple comparisons test. b, Representative images of TOV21G xenograft tumors at the end of NVB treatment. c, Clonogenic survival of BRCA1^−/− and WT RPE1 cells under increasing concentrations of the POLθ inhibitor NVB. Survival is normalized to the untreated sample. Data shown are mean ± SEM, n = 3 independent experiments. d, IR sensitivity of WT and POLQ^−/− RPE1 cells in clonogenic assays. e, NVB sensitivity of WT and POLQ^−/− RPE1 cells. Data shown in d and e are mean ± SD, from n = 6 biological replicates. Significance in c-e were determined by paired t-test, two-tailed. f, Clonogenic survival of BRCA1^−/− and WT RPE1 cells under increasing concentrations of the PARPi (Rucaparib) alone or in combination with NVB. Quantifications of survival fraction (mean from n=2 independent experiments), representative images and calculated IC50 values are shown.

Fig. 3.. NVB kills HR-deficient tumors and overrides PARPi resistance in vivo in PDX models.

a-b, Efficacy of NVB in PDX model DF83 (RAD51C deficient). NSG mice (n = 7 mice/group) bearing luciferized DF83 cells derived from an ovarian cancer patient were treated with 50 mg/kg of olaparib (daily, orally), 75 mg/kg NVB (via IP, b.i.d.), or both for 4 weeks. Tumor growth was monitored weekly by bioluminescence imaging. Representative images of tumor burden on day 0 and day 28 are shown, and quantifications shown are mean ± SEM, n = 8 mice. Log₂(fold change in tumor size) was calculated using the formula log₂(T/T0), where T is the tumor volume at given time and T0 is the initial tumor volume. c-d, Efficacy of NVB in PARPi resistant PDX model DF59 (BRCA1-deficient, 53BP1 loss), n = 6 mice/group. Same conditions and analysis were applied as in a-b. e-f, Efficacy of NVB in PARPi resistant PDX model DF149 (BRCA1-WT, HR proficient), n = 7 mice/group. Same conditions and analysis were applied as in a-b.

Fig. 4.. NVB induces excessive DSB end resection and non-productive RAD51 foci.

a, DSB end resection assay in U2OS cells with or without NVB. A scheme of the AsiSI resection assay is shown. Cells were incubated with 100 μM NVB for 18 h and then 4-OHT for 4 h. Genomic DNA were isolated to analyze %ssDNA flanking the AsiSI DSB site at indicated positions, using BrsGI digestion and qPCR quantification. Mean ± SEM, n = 3 independent experiments, two tailed unpaired t-test. b, Western blot analysis of phosphor-RPA32 in PD samples of the DF59 PDX model, 52 hours after first dose. Tumor bearing mice were dosed as in efficacy study in Fig. 3. c-d, Representative IHC images and quantifications of RAD51 foci in DF-59 and DF-83 PDX models. FFPE sections of the isolated tumor cells (as described in b) were stained with a RAD51 antibody. %RAD51-foci positive cells (> 4 foci/cell) in 3 random 40X fields was estimated in DF-59. No RAD51 focus was observed in the DF-83 model. Data shown are Mean±SD, n=3 tumors, ordinary one-way ANOVA Bonferroni’s multiple comparisons test. e-f, CellTiter-Glo cell viability assay in parental MDA-MB-436 (BRCA1-mutated) and PARPi-resistant MDA-MB-436-R cells under increasing concentration of Olaparib (e) or NVB (f). Mean ± SEM are shown, with n = 6 independent cultures. ns, no significant, two-tailed paired t-test. g, RAD51 foci assay with DMSO or NVB in MDA-MB-436 or MDA-MB-436-R. Cells with more than 5 foci were counted as positive. Data were mean from two independent experiments. h, A scheme shows the mechanism of NVB cytotoxicity in PARPi naïve and resistant tumor cells. In PARPi naïve HR-deficient settings, cells rely on MMEJ to repair DSBs. NVB inhibits MMEJ repair and causes unrepaired DSBs, leading to cell death. In HR-deficient but PARPi resistant settings, NVB inhibits the counteracting function of POLθ on RPA and RAD51, leading to enhanced DSB resection, accumulation of non-functional RAD51 foci, unrepaired DSBs and eventually cell death.

Fig. 5.. NVB overcomes multiple PARPi resistance mechanisms and POLQ expression level is a predictive biomarker for NVB sensitivity.

a. A schematic shows how the PARPi resistant clones R1-R4 were generated. b. A Western blot of WT and BRCA1^−/− RPE1 cells and the PARPi resistant clones, using an anti-BRCA1 antibody (Millipore #OP92). No BRCA1 reversion was observed in the clones. c-d. NVB sensitivity of WT RPE1, BRCA1^−/− RPE1, and PARPi-resistant BRCA1^−/− clones (R1-R4). R1 was selected by olaparib and R2-R4 were selected by Niraparib. R1 and R2 were tested in clonogenic survival assays (12–14 days) (c). R3 and R4 were tested in CellTiter-Glo cell viability assay (6 days) (d). Data are mean ± SEM, n = 3 independent cultures. e. POLQ expression at mRNA level and protein level in WT RPE1, BRCA1^−/− RPE1, and PARPi-resistant BRCA1^−/− clones (R1-R4). POLQ mRNA was measured by qRT-PCR and normalized to beta-Actin (ACTB). Mean ± SD of n=4 independent cultures. f. POLQ expression at mRNA level in parental (P), PARPi resistant (R), and empty vector (+EV) or BRCA1-cDNA (+BR1) complemented MDA-MB-436 cells. Mean ± SD of n=4 independent cultures. g. POLQ expression at mRNA level in PDX models DF53, DF83 and DF149. Mean of 4 technical replicates for each PDX model are shown. h. Olaparib sensitivity of CAPAN1 (BRCA2 mutated) and CAPAN1-CR cells (BRCA2 CRISPR edited back to wild type) in clonogenic survival assays. i. NVB sensitivity of CAPAN1 and CAPAN1-CR cells in clonogenic survival assays. Mean of n=2 independent experiments are shown in h and i. Statistical analysis in h and i were paired t-test, *p < 0.05, j. The expression levels of POLQ in CAPAN1 and CAPAN1-CR cells, analyzed by Western blotting.