Research ArticleCANCER

Targeting dePARylation selectively suppresses DNA repair–defective and PARP inhibitor–resistant malignancies

See allHide authors and affiliations

Science Advances  10 Apr 2019:
Vol. 5, no. 4, eaav4340
DOI: 10.1126/sciadv.aav4340


While poly(ADP-ribosyl)ation (PARylation) plays an important role in DNA repair, the role of dePARylation in DNA repair remains elusive. Here, we report that a novel small molecule identified from the NCI database, COH34, specifically inhibits poly(ADP-ribose) glycohydrolase (PARG), the major dePARylation enzyme, with nanomolar potency in vitro and in vivo. COH34 binds to the catalytic domain of PARG, thereby prolonging PARylation at DNA lesions and trapping DNA repair factors. This compound induces lethality in cancer cells with DNA repair defects and exhibits antitumor activity in xenograft mouse cancer models. Moreover, COH34 can sensitize tumor cells with DNA repair defects to other DNA-damaging agents, such as topoisomerase I inhibitors and DNA-alkylating agents, which are widely used in cancer chemotherapy. Notably, COH34 also efficiently kills PARP inhibitor–resistant cancer cells. Together, our study reveals the molecular mechanism of PARG in DNA repair and provides an effective strategy for future cancer therapies.


Poly(ADP-ribosyl)ation (PARylation) is a unique posttranslational modification for maintaining genome stability via different molecular pathways, especially DNA repair (1, 2). Poly(ADP-ribose), known as PAR, is composed of repeating adenosine diphosphate (ADP)–ribose units covalently linked via a glycosidic ribose–ribose bond. Using nicotinamide adenine dinucleotide (NAD+) as the donor, PARylation is catalyzed by poly(ADP-ribose) polymerases (PARPs) (36). Current PARP inhibitors primarily suppress PARP1 and PARP2 enzymatic activities, which inhibits PARP1/2-dependent DNA repair. However, there are a total of 17 members in the PARP family. Besides PARP1 and PARP2, other PARPs, such as PARP3 and PARP10, may also be involved in DNA repair (7, 8). Therefore, PARP inhibitors may not be able to suppress all the ADP-ribosylation–dependent DNA repair. Perhaps this is one of the major reasons that not all the BRCA (Breast Cancer) mutation tumors respond well to PARP inhibitors. Moreover, recent clinical trials show that long-term PARP inhibitor treatment induces chemoresistance in patients with cancer (9, 10). Thus, it is urgent to develop a novel therapeutic strategy for DNA repair–defective cancers.

PARylation is a transient posttranslational modification and is quickly degraded by dePARylation enzymes, such as poly(ADP-ribose) glycohydrolase (PARG) (6). PARG specifically hydrolyzes the glycosidic bonds between ADP-ribose units in PAR chains and is the major dePARylation enzyme that accounts for ~90% of dePARylation activity (11, 12). Notably, similar to PARPs, PARG also facilitates both DNA double-strand break (DSB) and single-strand break (SSB) repair (13, 14). Following DNA damage, a number of DNA damage response factors recognize PARylation and are recruited by PARylation to the proximity of DNA lesions (15). However, PARylation has to be digested so that DNA damage machinery directly recognizes DNA lesions and repairs lesions (16). Otherwise, the repair machineries will be trapped by PARylation at the vicinity of DNA lesions. Thus, dePARylation is an immediate downstream step of PARylation in DNA repair, and suppression of dePARylation will affect PARylation-dependent DNA repair. It indicates that targeting dePARylation, similar to targeting PARPs, may selectively kill tumor cells with DNA repair defects.

Loss of PARG leads to cell death and embryonic lethality in mice (17), suggesting that it could be a potential pharmacological target (12). A number of biochemical studies have shown that cancer cells with genetic depletion or RNA interference silencing of PARG have increased susceptibility to γ irradiation (18, 19), DNA-alkylating agents (17), and other chemotherapeutic agents that induce DNA damage, such as cisplatin and epirubicin (20). Moreover, apoptosis-inducing factor–mediated cell death is specifically activated after ultraviolet treatment of PARG-null cells (21). On the basis of the potential clinical implications, great efforts have been made for targeting PARG over the past decade. To date, several small-molecule inhibitors have been discovered, which inhibit PARG activity with half-maximal inhibitory concentration (IC50) values in the submicromolar range (2233). However, low cell permeability limits the application of these PARG inhibitors (e.g., ADP-HPD and nobotanin K) in the biological context. More recently, one potent and cell-active PARG inhibitor, PDD00017273, was identified and applied to tumor cell suppression studies (3436). However, it still could not be used for in vivo studies because of its poor metabolic stability. Thus, discovering a small molecule that potently and specifically inhibits PARG both in vitro and in vivo is a great challenge. Here, we report the identification and characterization of a potent and specific small-molecule PARG inhibitor, COH34, both in vitro and in vivo. Treatment with COH34 impairs DNA repair by trapping DNA repair factors and significantly suppresses the growth of tumor cells with DNA repair defects in cell-based assays, as well as in xenograft mouse cancer models. Notably, this small molecule also efficiently suppresses PARP inhibitor–resistant cancers. Together, this study not only reveals the molecular mechanism and biological function of PARG in DNA repair but also facilitates the development of a potential therapeutic strategy for future cancer treatments.


PARG plays an important role in DNA repair

To examine whether PARG is important for DNA repair, we generated the U2OS cell line with PARG stably down-regulated by short hairpin RNA (shRNA) (PARGKD) (fig. S1A) and performed time course comet assays. To examine SSB repair, such as base excision repair (BER) and nucleotide excision repair (NER), we treated cells with 0.5 mM H2O2 or 10 J/M2 ultraviolet C (254 nm wave length) and analyzed them via alkaline comet assays (fig. S1B). In addition, we also treated cells with 1 mM methyl methanesulfonate (MMS) or 5 grays (Gy) of ionizing radiation (IR) to generate DNA DSBs indirectly or directly. We carried out neutral comet assays for the analysis of DSB repair (fig. S1C). Using parental U2OS cells as the control, we found that loss of PARG induced longer tail moments in alkaline comet assays and neutral comet assays, suggesting that PARG plays an important role in both SSB and DSB repair. To further confirm these data, we examined γ-H2AX foci formation in a time course analysis using PARGKD cells with 1 mM MMS or 2.5 Gy of IR. In agreement with the neutral comet assay results, γ-H2AX foci formation in the PARGKD cells was much more prolonged compared to that in control cells (figs. S1D and S2).

Accumulating evidence suggests that PARylation is involved in both SSB and DSB repair (7). Since PARG is the major dePARylation enzyme, we performed immunofluorescence staining to examine the DNA damage–induced PARylation. We treated the control cells and PARGKD cells with 0.5 mM H2O2. In control cells, PARylation started to disappear at around 10 min. However, in the PARGKD cells, high level of PARylation was still observed at the prolonged time points (fig. S1E).

Collectively, these results demonstrate that loss of PARG suppresses dePARylation and DNA repair (fig. S1F). Thus, targeting dePARylation could be a potential strategy to induce cancer cell apoptosis, especially for those cancer cells with defective DNA repair (35, 36).

Identification of a potent and cell-active PARG inhibitor

Although PARG could be a key therapeutic target, potent inhibitors of PARG both in vitro and in vivo have not been identified yet. Here, we used in silico strategy with siMMap (site-moiety map) (37) to discover novel PARG inhibitors by screening the National Cancer Institute (NCI) database (~260,000 small molecules). The site-moiety map of a predicted inhibitor binding site (Fig. 1A and fig. S3A) consists of one H-bond anchor (H1) and two van der Waals (vdW) anchors (V1 and V2) (Fig. 1B). For each anchor, we identified a set of chemically related entities, supporting the concept that a given hotspot shares a unique chemical-physical binding environment with conserved binding residues (Fig. 1B). Among the 1000 top-ranking compounds, 254 form hydrogen bonds with residues N869 and F900 in the anchor H1; 894 form steric interactions with residues F738, Q754, and F902 in the V1 anchor; and 573 form vdW interactions with residues F738, V753, and G873 in the V2 anchor. An anchor in a site-moiety map can describe the interacting preference between a binding pocket (forming by several residues) and its preferred moieties. Thus, a compound that highly agrees with the anchors of an inhibitor site is able to potently suppress PARG activity. On the basis of siMMap analysis and predicted chemical properties, we selected 40 candidates for PARG inhibition analysis to estimate the efficacy of inhibition activity. We found that COH5 and COH34 strongly suppressed PARG-dependent PARylation digestion at a concentration of 100 μM (fig. S3B). To further accurately examine the inhibitory activity, we carried out PARylation digestion assays in a dose course. The results show that COH34 (NSC191252) is a very potent PARG inhibitor with an IC50 value of 0.37 nM, whereas the IC50 value of COH5 was more than 10 μM (Fig. 1C and fig. S3C).

Fig. 1 COH34 is a potent and cell-active PARG inhibitor.

(A) Predicted inhibitor site (magenta area) of PARG. (B) Site-moiety map and docked conformation of COH34 (magenta) in inhibitor site. (C) Chemical structure and formula of COH34. The IC50 value of COH34 was measured by dot blotting with PAR antibody in a dose course of COH34 (n = 3 independent experiments). (D and E) HCT116 cells were pretreated with or without COH34 (0.1 μM) for 1 hour before treatment with 0.5 mM H2O2 at 37°C for 15 min. HCT116 cells without H2O2 treatment and HCT116-PARG knockdown (HCT116-PARGKD) cells with H2O2 treatment are negative control and positive control, respectively. The extent of PAR was determined by dot blotting with anti-PAR antibody. The time course data are shown in the histograms from three independent experiments. ***P < 0.001. (F) A microscope-coupled laser scissors system was used to generate DNA damage in nucleus. PAR at DNA lesions in U2OS cells with or without 100 nM PARG inhibitor (COH34) treatment was immunostained with PAR antibody (red dots) after laser scissors. The kinetics of the accumulation of PAR at DNA damage sites in a time course was shown as mean ± SD from 50 cells (n = 3 independent experiments). ***P < 0.001.

Next, to examine the efficacy of COH34 in cells, we preincubated HCT116 cells with or without 100 nM COH34 for 1 hour before the treatment with 0.5 mM H2O2. After recovery at 37°C for 15 min, compared to the control, a ~10-fold increase of endogenous PARylation was observed by dot blotting when cells were preincubated with COH34 (Fig. 1D). In addition, a time course analysis shows that COH34 treatment did not increase the initial PARylation level. Instead, it suppressed the PARG-dependent dePARylation process (Fig. 1E). Moreover, we validated the DNA damage–induced PARylation kinetics using immunofluorescence staining. PARylation was detected by immunofluorescence immediately following laser microirradiation, and the level of PARylation was almost undetectable after 10 min. However, when cells were pretreated with 100 nM COH34, PARylation was prolonged (Fig. 1F and fig. S4). Collectively, our results demonstrate that COH34 is a potent PARG inhibitor both in vitro and in cells.

COH34 specifically binds to PARG

We generated the glutathione S-transferase (GST)–tagged catalytic domain of PARG (PARGCD; residues 451 to 976) (fig. S5) and carried out isothermal titration calorimetry (ITC) assays. The binding curve with a dissociation constant (Kd) value of 0.547 μM was observed following titrations of COH34 into PARGCD protein solution, indicating that COH34 strongly binds to the catalytic domain of PARG. The stoichiometric 1:1 binding was observed (Fig. 2A). On the basis of the structure of PARG [Protein Data Bank (PDB): 4BLI; (38)], we performed docking analysis for the possible binding mode. The results show that COH34 fits well into the catalytic pocket of PARG. The nitrogen and sulfur atoms of COH34 may form hydrogen bonds with Asn869 (Fig. 2B); the backbone of Phe900 may interact with the hydroxyl group of COH34 via hydrogen bonds as well. To validate the possible binding mode, we generated the Asn869-to-Ala (N869A) variant. This mutant significantly reduced the binding affinity to COH34 with Kd ≈ 64.37 × 10−6 M (Fig. 2C). Notably, the N869A mutant still retained the enzymatic activity to hydrolyze PAR in vitro, but it cannot be suppressed by COH34 at the concentration of 10 nM (Fig. 2D). The results indicate that COH34 targets PARG with Asn869 as a key binding residue.

Fig. 2 COH34 specifically binds to the catalytic site of PARG.

(A) The affinity between COH34 and the recombinant catalytic domain of PARG was measured by ITC. Titration of COH34 into a solution containing the purified protein was performed at 25°C using a Nano ITC instrument. The binding isotherm shows the fit of the data to an equilibrium-binding isotherm. The fit provides an equilibrium Kd for the binding of COH34 to the catalytic domain of PARG. (B) A predicted binding mode of COH34 at the catalytic site of PARG (PDB: 4BLI). COH34 (magenta) fits well in the catalytic domain (orange) and interacts with two residues (green), N869 and F900, through hydrogen bonds (red dots). (C) ITC parameters between the wild-type PARG (WT) or the N869A mutant and COH34 are summarized in the table. (D) The N869A mutant retains the enzymatic activity, and COH34 is unable to suppress the N869A mutant. PAR digestion assays were performed with or without COH34 (10 nM). Results were analyzed using dot blotting with anti-PAR antibody (n = 3 independent experiments). Control means PAR only. (E) Target selectivity assay was carried out using PARG, PARP1, and TARG1 with indicated concentrations of COH34. COH34 against PARG and PARP1 activity was analyzed by dot blotting with anti-PAR antibody. TARG1 inhibition results were determined by Western blot with anti–ADP-ribose antibody. Average inhibition of targets in a dose course of COH34 is shown in the histograms (n = 3 independent experiments). ***P < 0.001

To test the specificity of COH34, we examined the activity of COH34 on other dePARylation enzymes. Besides PARG, O-acyl-ADP-ribose deacylase 1 (also known as terminal ADP-ribose protein glycohydrolase 1, OARD1/TARG1) is also able to remove PAR chains from PARP1 (39). Similar to the catalytic macro domain of PARG, the macro domain of TARG1 mediates the hydrolysis of ADP-ribose. However, slightly different from PARG, TARG1 mainly cleaves the ester bond between terminal ADP-ribose and Glu or Asp residues. Thus, it can remove either PARylation or MARylation [mono(ADP-ribosyl)ation] from PARP1 (39). In addition, ADP-ribose also fits into the catalytic cage of PARP1 during PAR chain elongation. However, at 1 μM, COH34 did not suppress the activity of PARP1 and had little impact on TARG1 as well (Fig. 2E). Together, these results suggest that COH34 is a specific PARG inhibitor with strong binding affinity to the active site of PARG.

COH34 suppresses DNA repair in cells

As dePARylation is suppressed by COH34, we asked whether COH34 was able to disrupt DNA repair. In response to DNA damage, PARylation is mainly catalyzed by PARP1 at DNA lesions (40). The DNA damage–induced PARylation is recognized by a group of DNA repair factors and mediates the early and fast relocation of these DNA repair factors to DNA lesions (7). Thus, we examined several known PAR-binding DNA repair machineries, including XRCC1 (x-ray repair cross-complementing protein 1), CHFR (checkpoint with forkhead and RING finger domains protein), and APLF (Aprataxin- and PNK-like factor) (4143). We measured the relocation kinetics of these factors with laser microirradiation assays. In the mock-treated cells, similar to the kinetics of PARylation, these factors were recruited to DNA lesions within a few seconds but were also released from DNA lesions within ~10 min when PAR was hydrolyzed (Fig. 3A). Notably, in the presence of 100 nM COH34, these repair factors stayed at DNA lesions for a prolonged time (at least 1 hour) because dePARylation was suppressed (Fig. 3A and fig. S6). Moreover, similar results were observed in the PARGKD cells (Fig. 3A). Together, these findings suggest that extended PARylation induced by COH34 treatment traps DNA repair factors at DNA lesions.

Fig. 3 COH34-dependent trapping mechanism affects DNA damage repair.

(A) The relocation kinetics of XRCC1, CHFR, and APLF to DNA damage sites. GFP-XRCC1, CHFR, or APLF was expressed in U2OS cells. Cells were with or without pretreatment of 0.1 μM COH34 for 1 hour, and the relocation kinetics was monitored in a time course following laser microirradiation. Results are shown as mean ± SD from 50 cells (n = 3 independent experiments). ***P < 0.001. (B and C) KillerRed is a light-induced system that generates reactive oxygen species–driven DNA damage in cells. Cells expressing KillerRed protein were pretreated with or without COH34 (0.1 μM) for 1 hour before treatment with white light at 25°C for 10 min. KillerRed signal (DNA damage site, red dot) was detected with KillerRed antibody following light treatment. The relocation kinetics of endogenous PAR and XRCC1 was examined with PAR and XRCC1 antibodies. (D) The relocation kinetics results are summarized and presented as mean ± SD from 50 cells (n = 3 independent experiments). ***P < 0.001. (E) Time course of the accumulation of γ-H2AX foci after IR (1 Gy) treatment. U2OS cells were with or without pretreatment of COH34 (0.1 μM) for 1 hour, and the γ-H2AX foci were analyzed in a time course following IR treatment. Immunofluorescence was performed with γ-H2AX antibody. (F) Scheme of suppression of PARG traps DNA damage repair factors at the vicinity of DNA lesions and affects DNA damage repair.

Laser microirradiation induces a mixture of DNA lesions, such as SSBs and DSBs. To validate these results, a light-inducible BER repair system, KillerRed (44), was carried out to determine the enrichment kinetics of endogenous PARylation and XRCC1 at DNA lesions. U2OS cells expressing KillerRed were mock-treated or pretreated with 100 nM COH34 for 1 hour before exposing them to white light to induce KillerRed-mediated oxidative damage. At DNA lesions indicated by the KillerRed, PARylation was ready to be observed within 1 min following DNA damage and could not be detected after 10 min of recovery (Fig. 3B). However, with COH34 treatment, PARylation was prolonged at DNA lesions and easily detected 30 min after DNA damage (Fig. 3B). Along with the extended PARylation, XRCC1, a key regulator in the BER pathway, was trapped at DNA lesions for a longer duration when cells were treated with COH34 (Fig. 3, C and D).

Next, we examined the biological significance of COH34 treatment in the context of IR-induced DSB repair. Since γ-H2AX is a surrogate marker for DSBs, we examined the irradiation-induced foci (IRIF) of γ-H2AX and found that most of the DSBs were repaired after 4 hours of recovery from 2.5 Gy of IR treatment (Fig. 3E). However, with the pretreatment of 0.1 μM COH34, the IRIF of γ-H2AX was prolonged, indicating that COH34 treatment suppressed DSB repair (Fig. 3E and fig. S7C). In addition, we examined the fragmented genomic DNA using comet assays. The U2OS cells were treated with various DNA-damaging agents to induce SSBs or DSBs. In the presence of COH34, both SSB and DSB repair were suppressed, as suggested by longer tail moments in either neutral comet assays or alkaline comet assays (fig. S7, A and B), which were caused by blocking dePARylation (fig. S7D). Moreover, to examine DSB repair pathways that are regulated by PARG inhibition, we examined the classic nonhomologous end joining (c-NHEJ), alternative NHEJ (a-NHEJ), and homologous recombination (HR). Suppression of PARG by COH34 treatment impaired all these DSB repair pathways (fig. S8), which is consistent with earlier studies showing that PARylation is involved in these repair pathways (45). Together, our results suggest that although COH34 treatment does not affect the recruitment of DNA repair machinery, it suppresses dePARylation at DNA lesions and traps PAR-binding DNA repair factors for a prolonged time (Fig. 3F).

COH34 exhibits lethality in BRCA-mutant and PARP inhibitor–resistant cancer cells

Since dePARylation is the subsequent step following PARylation in DNA repair, targeting dePARylation may have a similar effect to targeting PARylation in tumor suppression. It has been shown that suppression of PARylation by PARP inhibitors selectively kills tumor cells with DNA repair defects, especially tumor cells with BRCA mutations. Thus, we next sought to test whether COH34 also induces BRCA-mutant cancer cell lethality. To examine the therapeutic potential, we treated BRCA1-mutant ovarian cancer cells (UWB1.289) and BRCA2-mutant ovarian cancer cells (PEO-1) with COH34 and performed colonogenic assays. As shown in Fig. 4 (A and B), we found that both UWB1.289 and PEO-1 were hypersensitive to COH34 treatment [half-maximal effective concentration (EC50) = 2.1 and 0.8 μM, respectively]. However, BRCA1-reconstituted UWB1.289 (UWB1.289 + BRCA1) and BRCA2-reconstituted PEO-1 [also known as PEO-4; (46)] cells were resistant to the COH34 treatment, indicating that COH34 treatment selectively kills BRCA-mutant tumor cells. Consistently, PARylation-mediated DNA repair was suppressed in these BRCA-mutant tumor cells.

Fig. 4 COH34 selectively kills BRCA-mutant and PARP inhibitor–resistant cancer cells.

(A and B) Colony formation assays of the BRCA1-mutant ovarian cancer cells (UWB1.289) and BRCA2-mutant ovarian cancer cells (PEO-1) following 14 days of treatment with DMSO and indicated concentrations of COH34. BRCA1/2-reconstituted cells (UWB1.289 + BRCA1 and PEO-4) are used as controls. Cells were stained with crystal violet. Average cell viability is presented as mean ± SD. (C and D) Combination treatments with a series of concentrations of PARG inhibitor COH34 and DNA-damaging agents (cisplatin, doxorubicin, temozolomide, or camptothecin) in UWB1.289 or PEO-1 cells. The combination condition shows the best synergistic effect to kill cancer cells, and data are shown in the histograms. ***P < 0.001. (E) Colony formation assay of the olaparib-resistant BRCA1-mutant ovarian cancer cells (SYr12) following 14 days of treatment with DMSO and indicated concentrations of COH34 or olaparib. Cells were stained with crystal violet. Average cell viability is shown as mean ± SD. (F) Combination treatments with a series of concentrations of PARG inhibitor (COH34) and DNA-damaging agents [cisplatin, doxorubicin, temozolomide (TMZ), or camptothecin (CPT)] in SYr12 cells. The combination condition, which is the best synergistic effect to kill cancer cells, is shown in the histograms. ***P < 0.001. (G) Cell viability assay in a panel of triple-negative breast cancer (TNBC) cell lines treated with 0.625 to 20 μM COH34 at 37°C for 14 days. The average EC50 value of each cell line is presented as mean ± SD. (H) Annexin V/PI apoptosis analyses of HCC1395 and HCC1937 cells following COH34 or olaparib treatment. Cells were treated with 5 μM PARG inhibitor COH34 or 10 μM olaparib at 37°C for 72 hours and then collected for annexin V and PI staining. Results were analyzed by flow cytometry. The percentages of annexin V–positive (lower right quadrant) and annexin V/PI–double-positive (upper right quadrant) cells from these experiments are shown in the relevant quadrants. (A to H) Data were from three independent experiments.

To test whether COH34 targets PARG in these cells, we used shRNA to down-regulate the expression of PARG in PEO-1 cells that were insensitive to COH34 treatment. Next, we reconstituted the cells with wild-type PARG or the N869A mutant. Since the N869A mutant is not recognized by COH34, the cells expressing wild-type PARG but not the N869A mutant restored their sensitivity to the COH34 treatment (figs. S9 and S10). In all, these results demonstrate that COH specifically targets PARG to induce the lethality of BRCA-mutant cancer cells.

Next, we further performed clonogenic assays to examine whether COH34 was able to act synergistically with other DNA-damaging agents, such as cisplatin, doxorubicin, temozolomide, and camptothecin, to induce tumor cell lethality. In particular, earlier studies have shown that temozolomide and camptothecin treatment are highly synergized with PARP inhibitor treatment for suppression of human cancer cell growth (47). We treated PEO-1 and UWB1.289 cells with COH34 together with cisplatin, doxorubicin, temozolomide, or camptothecin. Compared with individual efficacy of each compound under nontoxic concentration, combining cisplatin, doxorubicin, temozolomide, or camptothecin treatment with low micromolar COH34 remarkably induced BRCA-mutant cancer cell lethality (Fig. 4, C and D). This synergistic effect is likely due to the blocking of DNA repair by COH34, which prevented DNA repair factors participating in DNA repair during cisplatin-, doxorubicin-, temozolomide-, and camptothecin-induced DNA damage in BRCA-mutant cells.

Moreover, despite high initial response rates to PARP inhibitor chemotherapy, most of the patients develop resistance and relapse, leaving few options for further treatment and a dismal prognosis. Thus, we ask whether targeting dePARylation could kill PARP inhibitor–resistant cancer cells. To test the efficacy of COH34 in PARP inhibitor–resistant cancer cells, we used olaparib-resistant UWB1.289 (SYr12) cells for cell viability assays (48). It has been shown that SYr12 cells were resistant to PARP inhibitor treatment owing to loss of 53BP1 (48). Loss of 53BP1 promotes HR repair by partial restoring of RAD51 in BRCA1-mutant cells, thus inducing cancer cells resistant to PARP inhibitor treatment (49). This cell line was hypersensitive to the COH34 treatment (EC50 = 1.98 μM) (Fig. 4E). It suggests that this partial-restored HR is insufficient to rescue BRCA-mutant cancer cells from the disruption of DNA repair by COH34-induced trapping of DNA repair factors (Fig. 3 and fig. S6). In addition, combining 0.25 μM COH34 with either nonlethal dose of cisplatin, doxorubicin, temozolomide, or camptothecin significantly enhanced lethality in SYr12 cells (Fig. 4F). These findings suggest that targeting PARG may provide a potential therapeutic strategy for PARP inhibitor–resistant cancer cells.

PARP inhibitors (e.g., olaparib, rucaparib, niraparib, and talazoparib) have been identified as promising targeted therapeutics for the treatment of BRCA-mutant breast and ovarian cancer (50). However, not all the clinical trials were successful with PARP inhibitors (51). In particular, it has been shown that several triple-negative breast cancer (TNBC) cell lines are not sensitive to olaparib treatments (52). Therefore, we performed colonogenic assays to examine the efficacy of COH34 in 3 TNBC cell lines with BRCA1 mutations (MDA-MB-436, SUM149PT, and HCC1937), 1 TNBC cell line with mutated BRCA1/2 [HCC1395; (52, 53)], and 10 other TNBC cell lines without any known BRCA1/2 mutation. All these BRCA-mutant TNBC cells were more sensitive to the COH34 treatment than to the olaparib treatment (Fig. 4G and fig. S11). In particular, HCC1937 is known to be resistant to olaparib treatment (54) but was hypersensitive to COH34 treatment. Like BRCA-mutant ovarian cancer cells, selective lethality was observed between BRCA-mutant TNBC cells (HCC1395 and HCC1937) and BRCA–wild-type (MDA-MB-231) or BRCA-reconstituted TNBC (HCC1937 + BRCA1) cells (fig. S12, A and B). Combining with a nonlethal dose of COH34 and cisplatin or doxorubicin also significantly induced synergistic lethality in BRCA-mutant TNBC cells (fig. S12, C and D).

To further validate the results of colonogenic assays and investigate the role of COH34 on cell apoptosis, we carried out annexin V/propidium iodide (PI) apoptosis assays using HCC1395 and HCC1937 cells. As expected, the percentage of early apoptotic cells (annexin V positive) increased in these two DNA repair–defective TNBC cells treated with COH34 compared to the mock-treated group (Fig. 4H and fig. S13). However, much fewer early apoptotic cells were detected following olaparib treatment. These data suggest that COH34 promotes apoptosis in cancer cells with DNA repair defects.

Together, COH34 treatment not only efficiently and selectively kills DNA repair–defective and PARP inhibitor–resistant cancers but also exhibits powerful synergistic effect with DNA-damaging drugs in cancer treatment.

COH34 exhibits antitumor activity in vivo

Next, we sought to test whether COH34 is able to treat tumors in vivo. To test toxicity in mice, we intraperitoneally injected COH34 in 30% solutol into 8-week-old NOD (nonobese diabetic)/SCID (severe combined immunodeficient) gamma (NSG) mice at 10 mg/kg (n = 3) or 20 mg/kg (n = 3) daily for 10 days. Mice were weighed daily and observed for any adverse effect. With vehicle (30% solutol, n = 3) as control, no significant changed body weight and signs of pain and distress were observed in mice treated with COH34, suggesting that this small molecule is a nontoxic agent at 20 mg/kg for further in vivo testing (Fig. 5A).

Fig. 5 COH34 exhibits antitumor activity in PARP inhibitor–resistant and DNA damage repair–defective cell line xenografts.

(A) Toxicity assay of COH34 in mice: Female NSG adult mice were dosed intraperitoneally with 30% solutol (vehicle, n = 3) or COH34 in 30% solutol at 10 mg/kg (n = 3) or 20 mg/kg (n = 3) once daily for 10 days. They were weighed daily and observed for signs of pain and distress. (B and C) PAR analyses of tumor samples taken from NSG mice of HCC1395 and PEO-1 xenografts by dot blotting. Samples were taken after a single-dose treatment at the indicated time from each of the five mice receiving either COH34 (20 mg/kg) or vehicle dosed intraperitoneally. (D) Effect of COH34 in an olaparib-resistant UWB1.289 xenograft. Eight-week-old female NSG mice were injected with 8 million SYr12 cells. After tumors reached an average size of ~70 mm3, they were treated with vehicle (n = 6) or COH34 (20 mg/kg, n = 6) through intraperitoneal injections for 2 weeks. ***P < 0.001. (E) Effect of COH34 in a BRCA2-mutant ovarian cancer xenograft. Eight-week-old female NSG mice were injected with 10 million PEO-1 cells. Mice were randomized into two treatment groups of six mice. Mice were treated with COH34 (20 mg/kg) or vehicle intraperitoneally once daily for 2 weeks when tumors reached an average size of ~90 mm3. ***P < 0.001. (F and G) Effect of COH34 in BRCA-mutant TNBC xenografts. Eight-week-old female NSG mice were injected with 8 million HCC1395 or HCC1937 cells. After tumors reached an average size of ~85 mm3, they were treated with vehicle (14 days of treatment, n = 6) or COH34 (20 mg/kg, 14 days of treatment; n = 6) through intraperitoneal injections. ***P < 0.001. (H) Analysis of apoptosis, detected by immunohistochemistry. Shown are images from representative sections of tumor samples from COH34 and vehicle-treated mice. (I) Lysates of HCC1395 xenografts (~60 mg) from COH34 and vehicle-treated mice were subjected to Western blotting with γ-H2AX, PARP1, cleaved caspase-3, or β-Actin antibody. β-Actin was used as a control of protein loading.

To evaluate whether COH34 is stable in vivo, we examined the pharmacokinetics by measuring the concentration of COH34 in plasma following mice dosed at 20 mg/kg intraperitoneally at different time points (0.25, 0.5, 1, 2, 4, 8, 12, and 24 hours; three mice per time point). After analyzing by ultra–high-performance liquid chromatography–tandem mass spectrometry (UPLC-MS/MS), the terminal half-life of COH34 was determined to be approximately 3.9 hours (fig. S14), indicating that COH34 is stable for the in vivo studies. Next, we aimed to test whether COH34 inhibits PARG activity in the xenografted tumors; the pharmacodynamics of COH34 were examined by analyzing xenografted tumors (HCC1395 and PEO-1) from female NSG adult mice that were administered intraperitoneally at 20 mg/kg. After dosing at 0.5, 2, 8, 24, 48, and 72 hours, we harvested tumor samples (five mice per time point) and analyzed PARylation levels using dot blotting with anti-PAR antibody. Much higher PARylation levels were observed shortly after a single dose of COH34 (20 mg/kg) in two cancer mouse models. After 72 hours, PARylation levels returned to those of the vehicle-treated control (Fig. 5, B and C). Collectively, these data suggest that it would be feasible to test the efficacy of COH34 in mouse cancer models.

The in vivo efficacy of COH34 was assessed in xenograft mouse tumor models using SYr12, PEO-1, HCC1395, and HCC1937 cell lines in 8-week-old female NSG mice. COH34 was administered daily via intraperitoneal injections at 20 mg/kg for 2 weeks. Treatment of mice with COH34 led to a significant inhibition of tumor growth compared with vehicle control groups (Fig. 5, D to G). To test whether COH34 inhibited PARG and affected DNA repair in these mouse tumors, we compared vehicle- and COH34-treated tumor samples for markers of PARG inhibition. In line with the previous results, we observed an increase for apoptosis detected by TUNEL (terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling) assays after COH34 administration (Fig. 5H). Moreover, a modest increase in γ-H2AX, cleaved PARP1, and cleaved caspase-3 was also observed by Western blotting analysis in a dose-dependent manner (Fig. 5I), showing that COH34 is able to induce apoptosis of tumor cells with DNA repair defects via blocking the dePARylation mechanism. Together, our results strongly suggest that COH34 is a potential therapeutic agent for cancers with DNA repair defects.


In this study, we have demonstrated that targeting dePARylation is an effective approach to suppressing tumor growth. PARylation is a very dynamic posttranslational modification that plays a key role in DNA repair. However, each PAR chain is able to extend to more than 100 units. Thus, although PAR-binding DNA repair factors are recruited by PARylation to the proximity of DNA lesions, these factors may not be able to access DNA lesions without dePARylation. With COH34 treatment, PAR chains are kept at DNA lesions, and PAR-binding DNA repair factors are trapped at PAR chains close to DNA lesions but not directly at the sites of DNA damage. Thus, COH34 treatment suppresses DNA repair.

Our results also indicate that dePARylation does not merely antagonize PARylation at DNA lesions but also is an immediate downstream step following PARylation. Accumulating evidence suggests that targeting PARylation-dependent DNA repair is able to selectively kill tumor cells with DNA repair defects (55). Thus, dePARylation could be an important target for cancer therapy. Here, we found that COH34 specifically inhibited enzymatic activity of PARG, the major dePARylation enzyme. Moreover, like PARP inhibition, COH34 treatment selectively kills BRCA-mutant tumors. Not all BRCA-mutant tumors are hypersensitive to olaparib treatment (55). In particular, several known TNBC cells with BRCA mutations, such as HCC1937, are insensitive to olaparib treatment (52). In contrast, these TNBC cells, including HCC1937, are hypersensitive to COH34 treatment. In addition, SYr12, an olaparib-resistant ovarian cancer cell line with BRCA1 mutations (48), is also sensitive to COH34 treatment. Thus, targeting dePARylation by COH34 is able to complement PARP inhibitors for the treatment of tumors with DNA repair defects.

Targeting dePARylation suppresses DNA repair. Notably, many first-line chemotherapeutic drugs are DNA-damaging agents, such as cisplatin, doxorubicin, camptothecin derivatives, and temozolomide. In particular, camptothecin derivatives and temozolomide act synergistically with PARP inhibitors in tumor cell growth suppression (47). Combining both COH34 and DNA-damaging agents synergistically induces cancer cell lethality because not only is the genomic DNA damaged but the repair system is also compromised. Thus, targeting dePARylation by COH34 is able to act as a sensitizer for conventional cancer chemotherapy.

Recent evidence suggests that PARP inhibitors have been identified as promising molecularly targeted therapeutics to treat multiple types of cancer, such as familial breast and ovarian cancers (56). In addition, these PARP inhibitors are currently in phase III clinical trials for various other types of cancers with low survival rates, such as lung cancer (56). It has been shown that BRCA-deficient breast and ovarian cancers overexpress POLQ, which drives DNA repair toward non–HR-dependent pathways including alternative end joining (57, 58). Like these cancers, non–small cell lung cancer (NSCLC) tumors (e.g., NCI-H1299) overexpress POLQ and RAD54L compared to normal lung (59), indicating that a set of NSCLC may be lacking BRCA1-dependent HR repair and may be sensitive to PARG inhibitors. We found that NCI-H1299 cells were hypersensitive to COH34 treatment both in vitro and in vivo (fig. S15). Thus, targeting dePARylation by COH34 may have broader clinical implications in the treatment of other types of cancer.

In addition, Caldecott’s group recently reported that endogenous PARylation is detected during S phase at sites of DNA replication when PARG is suppressed (60). In agreement with their findings, COH34-mediated PARG inhibition also led to the increased PARylation during S phase at replication sites (fig. S16), indicating that in addition to regulating DNA repair, COH34 treatment may also affect PARylation-mediated DNA replication, which may also affect tumor cell viability.


Virtual screening

To discover novel small-molecule inhibitors of PARG, an inhibitor site was identified on the basis of the distribution of 20,000 docking poses of 100 random compounds (molecular weight, 150 to 600) in a catalytic site of PARG [PDB: 4BLI; (38)]. The Bioinformatics core of City of Hope used multiple-stage full-coverage (MSFC-VS) algorithm to screen the NCI Developmental Therapeutics Program library (260,071 compounds; the NCI chemical database is at for small molecules that can be fit into the inhibitor site of PARG (Fig. 1A and fig. S3A). Top-ranking compounds (1000), based on docking scores of MSFC-VS, were picked up for generation of siMMap (37). Forty candidates were selected for PARG inhibition assay on the basis of the siMMap analysis and predicted chemical properties.


COH34 (NSC191252) was obtained from NCI and resynthesized by WuXi AppTec. Olaparib, temozolomide, and camptothecin were purchased from Sigma-Aldrich. Cisplatin and doxorubicin were purchased from Abcam.

Cell culture

Sf9 cells were grown in Grace’s Insect medium supplemented with 10% fetal bovine serum (FBS) and incubated at 27°C. U2OS, HCT116, HCC1937, HCC1937 + BRCA1, PEO-1, PEO-4, MDA-MB-436, SUM149PT, MDA-MB-157, Hs578T, BT-549, HCC1187, BT-20, HCC1569, MDA-MB-435, MDA-MB-468, MDA-MB-134, MDA-MB-231, and WI-38 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% FBS and incubated with 5% CO2 at 37°C. HCC1395 and NCI-H1299 cells were maintained in RPMI-1640 medium supplemented with 10% FBS and incubated with 5% CO2 at 37°C. UWB1.289, UWB1.289 + BRCA1, and olaparib-resistant UWB1.289 (SYr12) (48) cells were maintained in RPMI-1640 (American Type Culture Collection) and MEGM (mammary epithelial cell growth medium) BulletKit (1:1; Lonza) with 3% FBS and incubated with 5% CO2 at 37°C.

shRNA transfection

U2OS, HCT116, or PEO-1 cells were seeded in a six-well plate with 2 ml of complete medium (DMEM supplemented with 10% FBS) and incubated overnight at 37°C. On day 2, the culture medium was removed and replenished with a mixture of complete medium (2 ml) with Polybrene (Santa Cruz Biotechnology) at a final concentration of 2 μg/ml. The cells were then infected with 20 μl of the shRNA lentiviral particle stock (Santa Cruz Biotechnology) or control shRNA lentiviral particle stock (Santa Cruz Biotechnology). The infection was performed at 37°C overnight. On day 3, the culture medium was aspirated and replaced with 2 ml of fresh complete medium. Stable clones expressing the PARG shRNA were selected by puromycin at 2.5 μg/ml and validated by Western blotting with anti-PARG antibody (Cell Signaling Technology).

Generation of stable cell lines

The complementary DNA of human PARG or the N869A mutant was cloned into the pcDNA3.1-hygromycin vector. Cells were transfected with plasmids using Lipofectamine 2000 (Invitrogen, CA, USA) and were selected by hygromycin treatment. Cells stably expressing wild-type or mutated PARG constructs were validated by Western blotting using anti-PARG antibody (Cell Signaling Technology). β-Actin was examined by anti–β-Actin antibody (Sigma-Aldrich) as the protein loading control.

Comet assays

Following treatments, cells were collected and rinsed twice with ice-cold phosphate-buffered saline (PBS). A total of 1 × 105 cells/ml were combined with 1% LMAgarose at 37°C at a ratio of 1:10 (v/v) and immediately pipetted onto slides. For neutral comet assay, the slides were immersed in the neutral lysis solution [2% sarkosyl, 0.5 M EDTA, and proteinase K (0.5 mg/ml) (pH 8.0)] overnight at 25°C in the dark, followed by washing in rinse buffer [90 mM tris buffer, 90 mM boric acid, and 2 mM Na2EDTA (pH 8.5)] for 30 min with two repeats. Then, the slides were subjected to electrophoresis at 20 V for 25 min (0.6 V/cm) and stained in PI (2.5 μg/ml) for 20 min in the dark.

For alkaline comet assay, the slides were immersed in the alkaline lysis solution [2.5 M NaCl, 100 mM EDTA, 10 mM tris-HCl, 1% sarkosyl, and 1% Triton X-100 (pH 10.0)] for 30 min at 4°C. After lysis, the slides were washed in distilled water three times and immersed in fresh alkaline electrophoresis solution [300 mM NaOH and 1 mM EDTA (pH 13.0)] for 10 min at 4°C. An electric field was then applied at 20 V (1 V/cm) and 300 mA for 20 min. The slides were neutralized to pH 7.5 in 0.4 mM tris buffer and stained with PI (2.5 μg/ml) for 20 min in the dark. Images were taken with a fluorescence microscope and analyzed by OpenComet.

Protein expression and purification

Full-length PARG or the N869A mutant was expressed in Sf9 cells using the “Bac-To-Bac” Baculovirus Expression System (Invitrogen). The GST-tagged PARG was purified from the baculovirus-insect cell system and confirmed by SDS–polyacrylamide gel electrophoresis (PAGE). The recombinant GST-tagged catalytic domain of PARG or the N869A mutant was expressed in Escherichia coli strain BL-21. A 10-ml overnight culture of a single transformant was used to inoculate 1 liter of fresh LB medium containing ampicillin (100 μg/ml). The cells were grown at 37°C to A600 (absorbance at 600 nm) = 0.6 and induced with 1 mM isopropyl-β-d-thiogalactopyranoside for 20 hours at 18°C. The cells were harvested by centrifugation at 5000 rpm for 10 min, and the pellet was suspended in NETN100 buffer (20 mM tris-HCl at pH 8.0, 100 mM NaCl, 1 mM EDTA, and 0.5% nonidet P-40). A sonicator instrument was used to disrupt the cells and centrifuged at 12,000 rpm for 20 min to discard the debris. The cell-free extract was incubated with glutathione Sepharose 4B (Amersham Biosciences) for 2 hours at 4 C. The beads were washed with NETN100 buffer, and the protein was eluted with 100 mM tris buffer (pH 8.5) containing 10 mM reduced glutathione. The eluted proteins were dialyzed against elution buffer to remove reduced glutathione. The protein concentrations were determined by the Bradford method (Bio-Rad, USA). The purified proteins were confirmed by SDS-PAGE and stored in PBS at −80°C.

PAR digestion assay and dot blotting

Recombinant PAR was purified from a biochemical assay using PARP1 according to previous studies (61). The concentration of PAR was calculated as the ADP-ribose unit. Recombinant full-length PARG (0.25 nM) was incubated with 10 μM PAR in the presence of dimethyl sulfoxide (DMSO) (negative control) or small-molecule compounds in a 10-μl reaction for 20 min at room temperature. Positive control only contains PAR in PBS. For dot blotting analysis, samples (1 μl) were spotted onto a nitrocellulose membrane. Then, the membrane was baked for 30 min at 60°C and blocked with TBST buffer (0.15 M NaCl, 0.01 M tris-HCl at pH 7.4, and 0.1% Tween 20) supplemented with 5% milk for 30 min at room temperature. After washing with TBST, the membrane was incubated with anti-PAR monoclonal antibody (Trevigen) overnight at 4°C. Following standard Western blot method, the signals were visualized by chemiluminescent detection.


Cells were treated with 1 mM MMS or 2.5 Gy IR for γ-H2AX foci formation assay and with 0.5 mM H2O2 or microirradiation for the analysis of PARylation. H2O2 and MMS were removed after treatments, and cells were recovered in fresh media at 37°C. Treated cells were fixed in 3% paraformaldehyde for 15 min at room temperature and permeabilized in 20 mM Hepes (pH 7.4), 50 mM NaCl, 3 mM MgCl2, 300 mM sucrose, and 0.5% Triton X-100 (Sigma-Aldrich) for 15 min at 4°C. Thereafter, coverslips were washed in PBS before immunostaining. Primary antibody incubations were performed overnight at 4°C at 1:200 dilutions (1:500 for anti–γ-H2AX) in PBS supplemented with 5% goat serum (Sigma-Aldrich) and followed by washing in PBS. Incubations with anti-mouse fluorescein isothiocyanate (FITC) or with anti-rabbit FITC secondary antibodies (Sigma-Aldrich) were performed at room temperature at 1:500 dilutions in 5% goat serum for 1 hour in the dark. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (Sigma-Aldrich) for 10 min at room temperature in the dark. Coverslips were mounted in VECTASHIELD (Vector Laboratories; Peterborough, United Kingdom). Results were analyzed using a fluorescence microscope.

Isothermal titration calorimetry

The recombinant catalytic domain of wild-type PARG or its mutant (N869A) (10 μM, 200 μl) and COH34 (200 μM, 40 μl) was loaded into the Sample Cell and Titration Syringe, respectively, of a Nano ITC instrument (TA Instruments; New Castle, DE). The recombinant catalytic domain of PARG was titrated with 17 injections (2 μl per injection) of COH34 into the Sample Cell. Heat of binding (ΔH), the stoichiometry of binding (n), and Kd were calculated from plots of the heat evolved per mole of ligand injected versus the molar ratio of ligand to receptor using the analytic software provided by the vendor.

Target selectivity assays

For PARP1 inhibition assay, 0.2 μg of recombinant PARP1 was incubated with DMSO (control) or COH34 in a 10-μl reaction mixture [10 mM tris-HCl (pH 8.0), 10 mM MgCl2, 125 μM NAD+, 10 mM dithiothreitol, and octameric oligonucleotide GGAATTCC (5 ng/μl)] for 30 min at room temperature. Samples were analyzed by dot blotting with anti-PAR monoclonal antibody (Trevigen). For TARG1 inhibition assay, mono–ADP-ribosylated PARP10 was generated in a 10-μl incubation mixture comprising 50 ng of GST-tagged PARP10, 100 μM tris-HCl (pH 8.0), 100 μM MgCl2, 320 μM NAD+, and 100 μM dithiothreitol for 30 min at 37°C. Following incubation, GST beads were added to conjugate mono–ADP-ribosylated PARP10 for 2 hours at 4°C and then washed with Mg2+-contained tris-HCl buffer (pH 8.0) three times to remove residual NAD+. Beads (10 μl) were incubated with 0.5 μM TARG1 and DMSO (control) or COH34 in a 15-μl reaction for 30 min at 37°C. The samples were boiled in the SDS sample buffer at 95°C, and eluates were analyzed by Western blot with anti-adenosine diphosphate ribose antibody (Millipore).

Laser microirradiation and imaging of cells

U2OS cells with transfection of green fluorescent protein (GFP)–tagged XRCC1, CHFR, and APLF were plated on glass-bottomed culture dishes (MatTek Corporation) and pretreated with or without 100 nM COH34 at 37°C for 1 hour before treatment with laser microirradiation. Laser microirradiation was performed using an IX 71 microscope (Olympus) coupled with the MicroPoint laser illumination and ablation system (Photonic Instruments Inc.). A 337.1-nm laser diode (3.4 mW) was transmitted through a specific dye cell and then yielded a 365-nm-wavelength laser beam that was focused through 603 UPlanSApo/1.35 oil objective to yield a spot size of 0.5 to 1 mm. The duration of cell exposure to the laser beam was ~3.5 ns. The pulse energy was 150 mJ at 10 Hz. Images were taken by the same microscope with the cellSens software (Olympus). GFP fluorescence at the laser line was converted into a numerical value using ImageJ. Normalized fluorescent curves from 50 cells from three independent experiments were averaged. The error bars represent SD.

KillerRed assay

The KillerRed system was a gift from L. Lan (University of Pittsburgh School of Medicine, PA). Activation of KillerRed followed the method as previously described (62). Briefly, U2OS tet response element (TRE) cells with transfection of pBROAD3/tetR-KR were pretreated without or with 0.1 μM COH34 for 1 hour before exposure to a 15-W white fluorescent bulb for 10 min (height to light is 15 cm). Following 1 or 10 min of recovery, samples were examined by immunofluorescence with anti-KillerRed antibody (Evrogen), anti-PAR monoclonal antibody (Trevigen), or anti-XRCC1 antibody (GeneTex). Images were taken from a fluorescence microscope and analyzed by ImageJ.

HR and NHEJ assays

The NHEJ reporter construct was provided by V. Gorbunova (University of Rochester) (63). For the c-NHEJ assay, the NHEJ reporter construct was digested overnight with Hind III at 37°C and gel-purified. The predigested construct was transfected into U2OS cells treated with DMSO or 1 μM COH34. To measure the repair frequency of HR and a-NHEJ, DR-GFP U2OS (for HR) or EJ2-GFP U2OS (for a-NHEJ) cells were plated into six-well plates for 24 hours. Both DR-GFP and EJ2-GFP cells were transfected with pCBASceI plasmid using Lipofectamine 2000 (Invitrogen) and treated with DMSO or 1 μM COH34. After 72 hours of incubation, cells were harvested for fluorescence-activated cell sorting analysis. The percentage of GFP-positive cells (successful repair) was calculated.

Clonogenic assay

Cells were seeded into six-well plates (~1000 cells per well) and then treated by 0.625 to 20 μM COH34 or olaparib. For synergistic efficacy analyses, compounds dissolved in DMSO were added to each well to make various concentrations (for COH34, 0.05 to 1 μM; for cisplatin, 0.625 to 10 μM; for doxorubicin, 1.25 to 12.5 nM; for temozolomide, 0.625 to 20 μM; and for camptothecin, 0.125 to 2 nM). Cells were treated with COH34, cisplatin, doxorubicin, temozolomide, or camptothecin alone or combination treatments as indicated. After a 14-day culture, the viable cells were fixed by methanol and stained with crystal violet. The number of colonies (>50 cells for each colony) was calculated.

Annexin V/PI apoptosis assay

Cells (1 × 105) were treated with DMSO (mock), COH34 (2.5, 5, or 10 μM), or olaparib (10 or 20 μM) for 72 hours at 37°C. After treatment, cells were washed with PBS twice and resuspended in 100 μl of annexin V binding buffer [10 mM Hepes/NaOH (pH 7.4), 140 mM NaCl, and 2.5 mM CaCl2]. The mixture was then incubated with annexin V–FITC (BD Biosciences) for 15 min at room temperature in the dark. Then, 400 μl of annexin V binding buffer and PI was added to each sample for 15 min at room temperature in the dark. Results were analyzed by flow cytometry within 1 hour to differentiate between viable (annexin V–negative and PI-negative), early apoptotic (annexin V–positive and PI-negative), and late apoptotic (annexin V–positive and PI-positive) cells. The extent of apoptosis was quantified as percentage of annexin V–positive cells.

Toxicity test

Eight-week-old female NSG mice (The Jackson Laboratory) were assigned into three groups (n = 3 per group), including vehicle (30% solutol), COH34 (10 mg/kg in 30% solutol), and COH34 (20 mg/kg in 30% solutol). Mice were intraperitoneally dosed daily for 10 days. Mice were weighed daily and observed for signs of pain and distress.

Xenograft mouse cancer models

All animal experiments were conducted under a protocol approved by the Institutional Animal Care and Use Committee (IACUC) of City of Hope. Immunodeficient female NSG mice (8 weeks old; strain 005557, The Jackson Laboratory) were inoculated with olaparib-resistant UWB1.289 (8 × 106), PEO-1(1 × 107), HCC1395 (8 × 106), HCC1937 (8 × 106), or NCI-H1299 (2 × 106) cells subcutaneously in 100 μl of serum-free medium containing 50% Matrigel (Corning). After inoculation for 5 to 7 days, mice were assigned randomly into two groups (n = 6 per group) that received vehicle control (30% solutol) or COH34 (20 mg/kg in 30% solutol, 0.1% DMSO) daily by intraperitoneal injection. Mice received 14 days of treatment. Tumor size was measured by an electronic caliper every 3 or 4 days and calculated using a standard formula: 0.5 × length × width2.


NSG female mice (6 to 7 weeks of age and weighing 23 to 25 g; The Jackson Laboratory) were acclimated after shipping for 7 days. Mice were divided into groups of three mice per cage. COH34 was dissolved in 30% solutol. Mice were dosed at 20 mg/kg intraperitoneally, and samples were collected at time points of 0.25, 0.5, 1, 2, 4, 8, 12, or 24 hours (three mice per time point). Mice were euthanized by CO2, and 0.5 ml of blood was collected via cardiac puncture. Blood was immediately put into tubes with K2 EDTA/ascorbic acid and placed on ice. Samples were centrifuged for 10 min at 4°C (2000g), and then plasma was transferred to a clean microfuge tube with 10 mg of ascorbic acid. Plasma samples were flash-frozen in liquid nitrogen and stored at −80°C. In addition, six mice were used as a control and not treated with compound. Plasma from control mice was used to calibrate LC-MS and correct for matrix effects. COH34 concentration was determined in mouse plasma samples using UPLC-MS/MS. COH34 was extracted from plasma by mixing 100 ml of plasma samples with 900 ml of ethyl acetate. After mixing for 5 min, the mixture was centrifuged for 10 min, and the organic layer was transferred to a UPLC-MS vial, dried under a nitrogen stream, and reconstituted in 100 ml of 75% methanol. Calibration standards and quality control samples were prepared spanning a range of 10 to 250 ng/ml and extracted in the same manner as the samples.


Female NSG mice (8 weeks old; The Jackson Laboratory) were inoculated with 8 million HCC1395 or 10 million PEO-1 cells subcutaneously. Mice were separated into treatment groups (n = 30 per group) of roughly equal tumor size (100 mm3) and dosed with COH34 (20 mg/kg) in 30% solutol or vehicle (30% solutol) once via intraperitoneal injection. Tumor size was measured by an electronic caliper using the standard formula: 0.5 × length × width2. After dosing at 0.5, 2, 8, 24, 48, and 72 hours, mice were euthanized using a CO2 gas chamber, and tumors (n = 5 per time point) were harvested and frozen by dry ice. Tumors were suspended in 500 μl of 1% SDS lysis buffer [1% SDS, 10 mM Hepes (pH 8.5), 500 U of Benzonase (Millipore), and 2 mM MgCl2] and then homogenized by a Tissuemizer. To remove SDS, sample solution was centrifuged at 1000g at 4°C for 10 min after adding 25 μl of SDS precipitation reagents (Thermo Fisher Scientific). The samples were then analyzed by dot blotting with anti-PAR monoclonal antibody (Trevigen).

Pharmacodynamic studies

Eight-week-old female NSG mice (The Jackson Laboratory) were inoculated with 8 million HCC1395 or 2 million NCI-H1299 cells subcutaneously. Mice were separated into treatment groups of roughly equal tumor size (150 mm3) and population and dosed with indicated doses of COH34 in 30% solutol or vehicle (30% solutol) daily via intraperitoneal injections for 6 days. Tumors were harvested and homogenized in lysis buffer using a Tissuemizer. Lysate proteins of tumors (~60 mg) were probed with the following antibodies: γ-H2AX (1:1000, Millipore), PARP1 (1:1000, Cell Signaling Technology), cleaved caspase-3 (1:1000, Cell Signaling Technology), and β-actin (1:3000, Sigma-Aldrich).

TUNEL assay

In situ detection of apoptotic cells was carried out using the TUNEL assay kit according to the manufacturer’s protocol (Millipore). Briefly, sections were deparaffinized and washed in PBS, followed by proteinase K pretreatment and endogenous peroxidase activity quenching. Labeling was performed by adding terminal deoxynucleotidyl transferase enzyme mix to the tissue sections, and the reaction was stopped by immersing slides in stop buffer followed by three PBS washes. After adding anti-digoxigenin conjugate to the slide, the color was developed using a diaminobenzidine peroxidase substrate kit (Vector Laboratories Inc.). Last, the slides were counterstained with hematoxylin (Ventana) and coverslipped. Images were captured with an Olympus DP70 camera attached to an Olympus BX51 microscope using a ×20 objective.


Supplementary material for this article is available at

Fig. S1. Loss of PARG impairs DNA damage repair.

Fig. S2. Kinetics of γ-H2AX foci formation in the PARG knockdown cells.

Fig. S3. Identification of novel PARG inhibitors.

Fig. S4. COH34 prolongs PARylation in cells.

Fig. S5. SDS-PAGE and Coomassie blue staining analysis of PARG catalytic domain.

Fig. S6. COH34 traps DNA damage repair factors for a prolonged time.

Fig. S7. COH34 treatment impairs DNA damage repair.

Fig. S8. COH34 treatment impairs HR, c-NHEJ, and a-NHEJ pathways.

Fig. S9. COH34 targets the catalytic site of PARG in cells.

Fig. S10. COH34 treatment induces γ-H2AX foci formation and increases the level of PARylation in BRCA-mutant and PARP inhibitor–resistant cancer cells.

Fig. S11. Efficacy of olaparib in TNBC cell lines.

Fig. S12. COH34 treatment selectively suppresses BRCA-deficient tumor cell growth.

Fig. S13. HCC1395 and HCC1937 cancer cells are more sensitive to COH34 treatment compared to olaparib treatment.

Fig. S14. Half-life of COH34 in mice.

Fig. S15. COH34 induces apoptosis of DNA repair–defective NSCLC.

Fig. S16. COH34 treatment increases PARylation level in S phase cells.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank L. Zou for providing SYr12 cells; L. Lan for providing the KillerRed system; V. Gorbunova for providing the NHEJ reporter construct; J. Stark for providing DR-GFP and EJ2-GFP U2OS cells; and the Bioinformatics, Analytical Cytometry and Pathology cores of City of Hope for technical support. Funding: This work was supported by the National Institutes of Health (CA132755, CA130899, and CA187209 to X.Y.). X.Y. is a recipient of the Leukemia and Lymphoma Society Scholar Award and the Tower Foundation Award. S.-H.C. is a recipient of the Ann Schreiber Mentored Investigator Award from the Ovarian Cancer Research Fund Alliance (OCRFA). Author contributions: X.Y. designed the research project; S.-H.C. performed the experiments; X.Y. and S.-H.C. analyzed the data and wrote the manuscript. Competing interests: S.-H.C. and X.Y. are inventors on a patent application related to this work filed by the City of Hope National Medical Center (no. PCT/US2018/039053, filed 22 June 2018). The authors declare no other competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article