Drug combination screen identifies GSK3i as acting synergistically with PARPi
To explore whether small-molecule inhibitors can sensitize cancer cells to PARPi, we performed a drug combination screen in BRCA1-deficient breast cancer cell line of HCC1937 and BRCA2-deficient CRC cell line of HCT-15, which express mutant-type BRCA1 or BRCA2 protein but modestly respond to PARPi. FDA-approved PARPi (Olaparib and Niraparib) and 99 well-characterized anticancer drugs targeting fifty classes of proteins belonging to indicated different kind of signaling pathway were chosen for initial screen (Table S1). Strikingly, a strong synergistic effect of GSK3i (CHIR99021 HCl and LY2090314) and PARPi (Olaparib and Niraparib) was observed in HCT-15 cells (Figure 1A). Unsurprisingly, ATR inhibitors and CHEK1 inhibitors showed synergistic effects with PARPi (Olaparib and Niraparib) in HCC1937 and HCT-15 cells, which had been reported that both ATR and CHEK1 inhibitors increased the sensitivity to PARPi in a BRCA1-independent way [9, 10]. Moreover, prior studies have demonstrated that inhibitor of BET, CDK1, HDAC, Protease, PI3K, and VEGFR could all decrease BRCA1 and other HRR factors at the protein level, thereby increasing the sensitivity of the cancer cell lines to PARP inhibition [24-29]. Consistent with the above conclusion, we found that these inhibitors displayed a synergistic effect with Olaparib and Niraparib in HCT-15 cells (Figure 1A). As reported [30-33], we also observed that Olaparib and Niraparib showed a synergistic effect in combination with inhibitors of DNMT, DNA-PK, mTOR, and HDM in HCT-15 cells (Figure 1A).
To further confirm the accuracy of our screening results, we validated the above results in the combination of Olaparib and CDK1 inhibitor (RO-3306) or ATR inhibitor (VE821) using the CalcuSyn model. Both combinations (i.e., olaparib + RO-3306 and olaparib + VE-821) caused obvious synergistic effects (CI < 0.7) in BRCA2-deficient HCT-15 cells, while only the olaparib and VE-821 combination produced synergistic effect (CI < 0.6) in the BRCA1-deficient HCC1937 cells (Figure 1B and 1C). These data were consistent with the observation shown in Figure 1A.
GSK3 inhibition sensitizes BRCA-proficient CRC cells to PARPi
We next sought to validate the observed interactions between GSK3 activity and PARPi. To further investigate the effect of GSK3 activity on cellular response to PARPi, two specific GSK3i, LY2090314 (LY) and CHIR99021 HCl (CHIR), were used in combination with five PARPi, including olaparib, niraparib, rucaparib, talazoparib and simmiparib. To exclude the synergistic effects were simply due to cell cycle arrest, we chose the concentrations of GSK3i (CHIR ≤ 10 μM; LY ≤ 5 μM) that had no discernible effect on cell proliferation (Figure S1A) and cell cycle phasing. Cells were treated with PARPi at eight concentrations, with or without LY2090314 or CHIR99021 HCl. The data showed that GSK3 inhibition strongly synergized with simmiparib (SP), talazoparib (TP), olaparib (OP), rucaparib (RP) and niraparib (NP) in HCT-15 cells (Figure 2A). The synergistic effect decreased in the order of simmiparib (sensitive fold: up to ~4463-fold), talazoparib (~185-fold), olaparib (~10-fold), niraparib (~4-fold) and rucaparib (~3-fold) when combined with LY2090314. Thus, simmiparib, a potent and selective PARP inhibitor currently in phase I clinical trials in China, was the most strongly perturbed following GSK3 inhibition (No. CTR20160475). Moreover, the presence of GSK3i led to a decrease IC50 of simmiparib in a concentration-dependent manner in HCT-15 cells (Figure 2B and Figure S1B). In line with the synergistic effects between simmiparib and GSK3i, we observed enhanced G2/M arrest and apoptotic cell death induced by simmiparib when combined with LY2090314 (Figure 2C-E) or CHIR99021 HCl (Figure S1C-E). The protein levels of cleaved PARP1 (p85) and cleaved-Caspase 3 increased accordingly (Figure 2F and Figure S1F). The results indicated that simmiparib and GSK3i combination treatment significantly suppressed tumor cell growth, caused cells to accumulate in G2/M of the cell cycle and induced remarkably apoptotic response.
To determine whether these synergies extend across other tumor cells, we used additional BRCA-proficient CRC cell lines (RKO, HCT-116, SW480, SW620, and HT-29). The data showed that GSK3 inhibition strongly synergized with simmiparib in all the BRCA-proficient CRC cells (CI < 0.6), as well as HCT-15 cells (Figure 2G and Figure S1G). Consistently, no combination activity (CI > 1) was observed in BRCA1-deficient HCC1937 cell lines (Figure S1H). This finding suggested a broader benefit of PARPi combined with GSK3i in BRCA-proficient CRC cells.
GSK3β depletion selectively sensitizes cancer cells to PARP and topoisomerase (Top) I inhibitors
There are two highly homologous forms of GSK3 in human, GSK3α and GSK3β, that have different tissue-specific functions and substrates [34, 35]. As GSK3i (LY2090314 and CHIR99021 HCl) block both GSK3α and GSK3β activity, we next generated GSK3α null and GSK3β null cells lines using CRISPR/Cas9 technique in HCT-15 and RKO cells, respectively (Figure 3A and 3B). Relative to the parental cells, the GSK3β KO cells (#KO1 and #KO2) displayed up to 60-fold increased sensitivity to the PARPi, simmiparib (Figure 3C and 3D). However, GSK3α depletion did not affect the cellular sensitivity to PARPi (Figure 3E). These results indicated that depletion of GSK3β selectively sensitized cancer cells to PARPi.
To investigate the possible involvement of GSK3β in sustaining genomic stability, we examined whether GSK3i, LY2090314, synergized with different DNA-damaging agents known to generate different forms of DNA lesions in HCT-15 cell line. The results revealed that GSK3i synergized with irinotecan (CPT-11, Top I inhibitor; CI < 0.4), but not adriamycin (ADR, Top II inhibitor; CI > 1) or etoposide (VP-16, Top II inhibitor; CI > 1) (Figure S2A and S2B). Similarly, GSK3β depletion, but not GSK3α, significantly increased the cellular sensitivity to CPT-11 (Figure S2C and S2D).
GSK3β is required for the HRR of DSBs
Although PARPi and Top I inhibitor cause different forms of DNA lesions, both agents are known to selectively kill proliferating cancer cells by causing replication-dependent DSBs [36, 37]. For this reason, we compared the occurrence of drug-induced DSBs in GSK3β KO and parental cells, using γH2AX as a marker. Higher level of γH2AX protein accumulated in GSK3β KO cells compared to the parental cells (Figure 4A and Figure S3A). This result was further supported by the enhanced γH2AX protein level in cells treated with a combination PARPi and GSK3i (Figure 4B and Figure S3B); and the observation was recapitulated using an immunofluorescence assay to stain nuclear γH2AX foci (Figure 4C and Figure S3C). However, the level of DSBs was similarly induced in GSK3α null cells and parental cells (Figure S3D). These results indicated that a defect in DSBs repair was caused by the knockout of GSK3β, but not GSK3α.
Replication-dependent DSBs lesions are known to be predominantly repaired by HR, a repair process requiring homologous DNA sequence as a template. To test whether GSK3β inhibition and knockdown cells were defective in HRR, we chose a well-characterized reporter assay using the DR-U2OS, a human osteosarcoma cell line with chromosomally integrated HR reporter gene containing an I-SceI recognition sequence . In this cell line, HRR using a direct repeat within the reporter cassette as a template results in an intact GFP gene, which can be detected by flow cytometry. The data showed that GSK3β knockdown using two independent siRNAs remarkably decreased the HR efficiency triggered by I-SceI (Figure 4D). Consistently, the GSK3i, CHIR99021 HCl and LY2090314, significantly reduced the capacity of HRR, in which ATR inhibitor, VE821, was used as a positive control (Figure 4E). However, GSK3α silencing had no impact on HR efficiency (Figure S3E). Additionally, we observed impaired RAD51 foci formation in GSK3β KO cells or GSK3i-treated cells which further strengthened the deficiency in HRR (Figure 4F and Figure S3F). Together, these data identified a previously unappreciated role of GSK3β in HRR, which echoed our findings that GSK3β inhibition and depletion affected cell sensitivity to PARP and Top I inhibitors.
GSK3β depletion represses the expression of BRCA1
To understand how GSK3β is involved in HRR, we analyzed the protein level of the key factors involved in the HR pathways using western blotting. GSK3β KO cells showed a marked reduction in BRCA1 protein levels, whereas the levels of Mre11, CtIP, RPA32 and RAD52 were not affected (Figure 5A and Figure S4A). Similarly, inactivation of GSK3β by CHIR99021 HCl and LY2090314 treatment led to a marked decrease in BRCA1 protein level in a concentration- and time-dependent manner (Figure 5B and 5C; Figure S4B and S4C). Furthermore, we found that GSK3β depletion and inhibition reduced RAD51 protein level in HCT-15 cells but not in RKO cells (Figure5A and 5B, Figure S4A and S4B). Therefore, we assessed the effect of LY2090314 on BRCA1 and RAD51 protein levels in other CRC cells (HCT116, HT29, SW480 and SW620). LY2090314 modestly decreased RAD51 level in HT-29 cells, while it consistently decreased BRCA1 protein in all the lines assessed (Figure S4D). We thus focused on BRCA1 as a likely mediator of the GSK3i effect. We transfected WT-GSK3β (WT) or a kinase-inactive mutant GSK3βY216F (Y216F) cDNA into HCT-15 KO cells, and obtained the corresponding variants that expressed the WT or Y216F GSK3β proteins. As expected, reconstitution with WT-GSK3β, but not Y216F-GSK3β, partially restored the BRCA1 protein level (Figure 5D), suggesting that GSK3β enzymatic activity was required to retain protein. Ectopically expressed FLAG-GSK3β also resulted in an increase in BRCA1 protein in the parental HCT-15 cells, which further suggested a strong association between GSK3β and BRCA1 (Figure 5E).
The reduction in BRCA1 protein level appeared to be a result of transcriptional repression, as RT-PCR revealed that the GSK3β KO cells had reduced BRCA1 mRNA expression (Figure 5F and Figure S4E). In addition, cells treated with GSK3i (CHIR99021 HCl and LY2090314) showed a reduced mRNA expression of BRCA1 in a time- and concentration-dependent manner (Figure 5G and S4F). However, BRCA1 protein levels were not affected by MG132 treatment in GSK3β KO or GSK3i-treated cells (Figure 5H). Collectively, these data implied that GSK3β may repress BRCA1 transcription and protein expression in an enzyme-dependent manner.
PARPi and GSK3β inhibition are synergistic in vivo
Our data thus far indicated that GSK3 inhibition strongly synergizes with PARPi in BRCA2-deficient and BRCA1/2-proficient cancer cells in vitro. We further validated this therapeutic potential using xenograft mice models. BRCA2-deficient HCT-15 cells and BRCA-proficient RKO cells were subcutaneously injected into nude mice, and once tumor volume reached ~70 mm3, either simmiparib or LY2090314, alone or in combination, was injected every other day for 14 d. Notably, the combination of these two agents significantly inhibited the growth of the tumor in the HCT-15 and RKO xenograft mouse model, although the tumor growth in the single agent groups was not affected following simmiparib or LY2090314 treatments (Figure 6A and 6B). Consistently, the tumor burden was significantly reduced as measured by the weight of dissected tumors (Figure 6C and 6D). The increased response to the combination treatment was associated with increased number of DSBs lesions (as indicated by γH2AX levels), as well as increased the levels of cleaved-Caspase3 in the combined treatment group (Figure 6E and 6F). In support of the mechanism identified in this study, the GSK3i group showed decreased BRCA1 protein level (Figure 6E and 6F). All the tested compounds caused no obvious loss of weight of the nude mice (Figure 6A and 6C) and were well tolerated during the drug administration.
To further validate the impact of GSK3β on in vivo sensitivity to PARPi, we used the HCT-15 GSK3β KO cells and parental cells to establish xenograft models in nude mice. As expected, administration of simmiparib significantly inhibited the growth of GSK3β KO tumor xenografts, but not the parental tumor xenografts. (Figure S5A and S5B). Consistently, there was a significant decrease in BRCA1 protein level and increase in γH2AX level in the GSK3β KO tumor xenografts treated with simmiparib (Figure S5C). These data demonstrated that inhibition or depletion of GSK3β could enhance the in vivo sensitivity to simmiparib without toxicity.