Exploration of novel strategies to extend the benefit of PARP inhibitors beyond BRCA-mutant cancers is of great interest in personalized medicine. Here, we identified EGFR amplification as a potential biomarker to predict sensitivity to PARP inhibition, providing selection for the glioblastoma (GBM) patient population who will benefit from PARP inhibition therapy.
Selective sensitivity to the PARP inhibitor talazoparib was screened and validated in two sets [test set (n = 14) and validation set (n = 13)] of well-characterized patient-derived glioma sphere-forming cells (GSC). FISH was used to detect EGFR copy number. DNA damage response following talazoparib treatment was evaluated by γH2AX and 53BP1 staining and neutral comet assay. PARP–DNA trapping was analyzed by subcellular fractionation. The selective monotherapy of talazoparib was confirmed using in vivo glioma models.
EGFR-amplified GSCs showed remarkable sensitivity to talazoparib treatment. EGFR amplification was associated with increased reactive oxygen species (ROS) and subsequent increased basal expression of DNA-repair pathways to counterelevated oxidative stress, and thus rendered vulnerability to PARP inhibition. Following talazoparib treatment, EGFR-amplified GSCs showed enhanced DNA damage and increased PARP–DNA trapping, which augmented the cytotoxicity. EGFR amplification–associated selective sensitivity was further supported by the in vivo experimental results showing that talazoparib significantly suppressed tumor growth in EGFR-amplified subcutaneous models but not in nonamplified models.
EGFR-amplified cells are highly sensitive to talazoparib. Our data provide insight into the potential of using EGFR amplification as a selection biomarker for the development of personalized therapy.
EGFR amplification occurs in approximately 50% of glioma patients. A total of 27 patient-derived glioma sphere-forming cells (GSC) were characterized and screened for the PARP inhibitor talazoparib sensitivity. GSCs harboring EGFR amplification showed significantly higher sensitivity to talazoparib treatment. EGFR amplification was associated with elevated reactive oxygen species (ROS) levels, and gene enrichment analysis revealed DNA-repair pathways were upregulated to counterelevated oxidative stress, and thus rendered vulnerability to PARP inhibition. Inhibition of PARP by talazoparib in EGFR-amplified cells led to unrepaired double-strand break and increased PARP–DNA trapping. The selective sensitivity was further supported by the in vivo results showing that talazoparib significantly suppressed tumor growth in two EGFR-amplified subcutaneous models but not in the two nonamplified models. Hence, our results demonstrated that EGFR amplification is a novel predictive biomarker of response to PARP inhibitor talazoparib therapy in GBM.
Glioblastoma (GBM) is one of the most aggressive and lethal human cancers. Despite advances in surgery, radiotherapy, and chemotherapy, the median overall survival is less than 15 months (1–3). Mutations and copy-number aberrations of EGFR have been identified as one of the most frequent genetic events in GBM by The Cancer Genome Atlas (TCGA; refs. 4, 5). EGFR amplification/mutation was associated with significant elevations in total EGFR expression and phosphorylation and activation of multiple oncogenic pathways.
Glioma sphere-forming cells (GSC) are derived from primary GBM and cultured in minimal medium containing growth factors to select the cells with sphere-forming capacity. These sphere-forming cells are often termed “glioma stem–like cells” as they have the capacity for self-renewal and multipotency (6, 7), and the capacity to form gliomas that reflect the histopathologic heterogeneity of the parental tumors. These cells are thought to be responsible for tumor progression, recurrence, and resistance to radiation and conventional chemotherapy (8, 9), and thus represent a vital target for cancer therapy. A growing body of evidence demonstrated, although still controversial, that GSCs had higher activation of DNA damage response (DDR) and single-strand break response (SSBR) at both the basal level and after ionizing radiation to tolerate DNA damage stress and oxidative stress (8, 10–12). The reliance on DNA damage repair to handle additional genotoxic stress in GSCs suggested that targeting key DNA damage repair pathway molecules would be an effective therapy for GBM.
Poly-ADP-ribose polymerase 1 (PARP1) is an enzyme that catalyzes the transfer of ADP-ribose polymers to substrates, including numerous DNA-repair enzymes to sense DNA lesions, activates DDRs, and facilitates DNA damage repairs. The central role of PARP in DNA damage response, particularly in SSBR, makes it a promising therapeutic target. Multiple PARP inhibitors have been developed in preclinical and clinical studies in various tumors, including GBM (13). Combination of PARP inhibitors olaparib and veliparib with TMZ has been studied in a phase I trial and a phase II/III trials, respectively. However, excessive toxicity in combination with TMZ limited its efficacy and led to clinical trial failure (14, 15). Talazoparib, a novel oral PARP inhibitor with greater in vitro activity than any other PARP inhibitor currently in development (16, 17), showed promising single-agent lethality in advanced ovarian and breast cancers harboring deleterious BRCA1/2 mutations. However, BRCA mutations are rare in other cancers, including GBM (18, 19), which limited the application of talazoparib therapy. Talazoparib is currently in clinical trials for GBM (NCT02116777). Exploration of novel strategies and identifying predictive biomarkers to select GBM patients who are most likely to benefit from talazoparib treatment is urgently needed.
In the current study, using two sets of GSCs, we demonstrated that talazoparib monotherapy selectively inhibited the proliferation of EGFR-amplified GSCs in vitro and suppressed EGFR-amplified tumor progression in GSC xenograft models. Our data suggested that EGFR amplification may be a biomarker to predict sensitivity to talazoparib treatment in GBM, providing selection of patient populations who will benefit from talazoparib treatment.
Materials and Methods
Cell lines and reagents
GSC lines were established by isolating neurosphere-forming cells from fresh surgical specimens of human GBM tissue from 2005 through 2008, as described previously (20). The study was approved by the Institutional Review Board of MD Anderson Cancer Center, and informed consent was obtained from all subjects. GSC lines were cultured in DMEM/F12 medium containing B27 supplement (Invitrogen), basic fibroblast growth factor, and epidermal growth factor (20 ng/mL each). All GSCs used in the study were fewer than 15 passages. Cells were authenticated by testing short tandem repeats using the Applied Biosystems AmpFISTR Identifier kit. The last authentication test was performed in July 2017. All cell lines were tested negative for Mycoplasma contamination by using the MycoAlert Detection Kit (Lonza). Talazoparib was from WuXi AppTec. Olaparib, veliparib, and pamiparib were from Selleckchem. MnTBAP was from Abcam. For in vitro use, all inhibitors were dissolved in dimethyl sulfoxide (Sigma-Aldrich).
Cell proliferation assay
Cells were treated with talazoparib in triplicate for 5 days, and cell proliferation was determined using the CellTiter-Blue viability assay (Promega). The IC50 value was calculated as the mean drug concentration required to inhibit cell proliferation by 50% compared with vehicle-treated controls using GraphPad software (GraphPad Software, Inc.).
Subcellular fractionation assay
To detect chromatin-bound PARP, cells were collected and cytoplasmic, nuclear soluble, and chromatin-bound proteins were fractionated using a subcellular protein fractionation kit from Thermo Scientific (#78840), following the manufacturer's instructions (17). Immunoblotting was carried out using standard procedures. Actin was blotted for cytoplasmic fraction marker, Lamin B for nuclear soluble marker, and Histone H3 for chromatin-bound marker.
Fluorescence in situ hybridization (FISH)
Exponentially growing cells were treated with colcemid (0.04 μg/mL) for 2 hours, and chromosomes were prepared by conventional fixation. FISH assay was performed on the slides using EGFR FISH probe from Empire Genomics following the manufacturer's instructions with slight modifications. Briefly, the probe was applied on the slide and covered with a glass coverslip and sealed with rubber cement. The slides were then denatured at 70°C using ThermoBrite system (Abbott Laboratories) and incubated at 37°C overnight. The slides were then washed using 2× saline sodium citrate at 45°C for 1 to 2 minutes, rinsed in 0.05% Tween 20 in PBS counterstained with DAPI, and analyzed under Nikon ECLIPSE 80i fluorescent microscope. A minimum of 50 cells were examined to score the number of fluorescent signals.
Cells were seeded onto Lab-Tek II tissue culture slides (Thermo Fisher) and treated with talazoparib and washed out for indicated time. Cells were fixed with 4% paraformaldehyde, permeabilized with 0.2% Triton X-100, and blocked with 5% BSA in PBS and then stained overnight at 4°C with anti-53BP1 antibody (Santa Cruz) and anti-gamma H2AX (phospho S139) antibody (Abcam). Cells were washed with PBS and stained with secondary antibodies (Alexa Fluor 594 donkey anti-rabbit IgG; Alexa Fluor 488 donkey anti-goat IgG; Invitrogen) for 1 hour. The cells were counterstained with Vecta shield sealant containing 4′,6-diamidino-2-phenylindole (DAPI; Vector Laboratories). The percentage of cells displaying foci was quantified by counting 5 random fields.
Measurement of ROS activity
To measure ROS activity, GSCs were stained with a CM-H2DCFDA probe, according to the manufacturer's instructions (Molecular Probes, Invitrogen). In brief, cells were dissociated with Accutase to single cells, suspended in prewarmed HBSS containing a 10 μmol/L CM-H2DCFDA probe, and incubated at 37°C for 30 minutes. Cells were then acquired with FACS (BD Biosciences) and analyzed using FlowJo software (Tree Star, Inc.).
Neutral comet assay
Neutral comet assays were performed using an OxiSelect Comet Assay Kit (Cell Biolabs) per the manufacture's protocol. Briefly, cells were dissociated with Accutase and washed with PBS, and replicates were suspended in OxiSelect comet agarose (Cell Biolabs). Neutral electrophoresis was conducted at 30 V for 30 minutes. Data were collected with a fluorescent microscope with an FITC filter and analyzed using the OpenComet software (21). All steps after agarose treatment were conducted in the dark to prevent additional DNA damage.
All animal studies were approved by the institutional review board of The University of Texas MD Anderson Cancer Center. Cells were transfected with MSCV-Luciferase-EF1α-copGFP-T2A-Puro BLIV 2.0 Lentivector (SBI System Biosciences) to generate luciferase expressing cells for in vivo imaging. To create subcutaneous tumor model, luciferase expressing cells (5 × 106 cells) were implanted into the hind flanks of 4 to 6 weeks old male nu/nu mice. For intracranial models, luciferase expressing cells (0.5 × 106) were implanted intracranially into nude mice using a previously described guide-screw system (22). Starting on day 14 after tumor cell implantation, mice were treated with 0.33 mg/kg talazoparib by oral gavage. The treatment frequency was twice a day for 5 days, with 2 days off between treatments, for a total duration of 6 weeks. Tumor growth and development was visualized and quantified using the IVIS Spectrum in vivo imaging system. Mice were monitored daily and euthanized when they became moribund. The whole brains from intracranial models and tumors from subcutaneous models were collected and preserved in formaldehyde solution for future use.
Mutations and copy-number alterations of GSCs were determined by exome sequencing (Illumina HiSeq) and Affymetrix OncoScan FFPE V2 arrays. In this study, we focused on the prominent driver genes p53, EGFR, and PTEN that also showed frequent mutations and copy-number alterations in GBM. Gene set enrichment analysis (GSEA) was performed with a preranked list of genes. To rank genes, EGFR-amplified was compared with nonamplified GSCs to obtain differentially expressed genes using limma (23). Then, genes were ranked according to P value and fold change (FC) multiplying −log10 (P value) by the FC sign. We ran GSEA using the gene sets from the Canonical Pathways compendia (C2) version 6.1 from MsigDB (24).
The statistical comparison was performed using the Student t test or Fisher exact test, as appropriate. The results are presented as the mean of at least 3 independent experiments. All tests were two-sided. Statistical analyses were carried out using GraphPad Prism software. A P value of <0.05 was considered statistically significant. Survival curves were plotted using the Kaplan–Meier method, and log-rank tests were used to compare survival curves between groups.
PARP inhibition by talazoparib selectively suppresses the proliferation of EGFR-amplified GSCs
To test the antiproliferative activity of talazoparib, we first used a test set of 14 GSCs to study a dose–response profile to talazoparib (Fig. 1A). The dose of talazoparib required to inhibit 50% of cell growth (IC50) varied widely from 4.3 nmol/L to over 1500 nmol/L (Fig. 1B). Of 14 GSCs, 7 were sensitive to talazoparib, with IC50 less than 300 nmol/L, and the other 7 were relatively resistant, with IC50 higher than 300 nmol/L.
To identify potential markers of talazoparib sensitivity, we studied genetic status of most frequently mutated genes in GBM. The mutation, amplification/deletion of p53, EGFR, PTEN, PDGFRA, and MET were determined by whole-genome sequencing as well as OncoScan array. As shown in Fig. 1B, EGFR amplification was one of the main determinants in 6 of 7 sensitive GSCs (85.7%); 0 out of 7 resistant GSCs (0%) showed amplification of EGFR (P = 0.005, Fisher exact test). We then used a validation set of 13 GSCs to determine the dose response to talazoparib (Fig. 1C). Six GSCs showed an IC50 less than 300 nmol/L. Similar to the test set, 5 of the 6 sensitive GSCs in the validation set were EGFR amplified (83.3%; Fig. 1D); 1 out of 7 resistant GSCs (14.3%) was EGFR amplified (P = 0.029, Fisher exact test). Collectively, the results from 27 cell GSC lines revealed that EGFR-amplified cells have significantly lower IC50 than do EGFR nonamplified cells (P < 0.05; Fig. 1E). In contrast, the sensitivity was not associated with alteration of PTEN, p53, or CDKN2A, the other three most frequent genetic alterations in GBM (Fig. 1F–H).
EGFR amplification was evaluated by copy-number variation (CNV) as determined by OncoScan array and stratified according to a threshold of CNV ≥ 2 (Fig. 1I). EGFR amplification was further validated by FISH (Fig. 1J). At least 50 cells for each line were randomly examined for EGFR FISH signals. Consistent with CNV results, EGFR-amplified cells displayed higher FISH signals (average signals ranges from 20 to ≥ 30 spots per cell), with a majority of the cells displaying ≥ 10 signals (Fig. 1K). Furthermore, we performed Western blot to detect EGFR protein expression and the canonical downstream signaling. EGFR protein expression is indeed higher in the EGFR-amplified group compared with the nonamplified group. pStat3-Y705, and pAKT-T308 to a less extent, was elevated in amplified group, confirming that EGFR signaling is activated in the amplified group (Fig. 1L; Supplementary Fig. S1).
Talazoparib selectively depleted the subpopulation with high EGFR copy number
EGFR amplification in GBM tumors is heterogeneous (25–27). This heterogeneity is also noted in our GSCs. For example in GSC262, 76% of cells displayed very high EGFR signal (>35 signals/cell) and 24% of cells have a low EGFR signal less than 10. For GSC11, 68% of cells displayed very high EGFR signal and 32% of cells have a low EGFR signal less than 10 (Supplementary Fig. S2A). Therefore, we ask if talazoparib preferentially kills the subpopulation of cells with high EGFR copy number and, if so, will talazoparib-treated cells lose EGFR amplification. To answer this question, we treated GSC262 and GSC11 cells with talazoparib and the posttreatment living cells were sorted by FACS and the EGFR copy number were evaluated by FISH. As expected, only low EGFR copy-number cells (<10 signals/cell) survived after talazoparib treatment (Supplementary Fig. S2A–S2C) and the cells with greater than 35 signals were completely lost (Supplementary Fig. S2C). Further, we showed that the total EGFR protein level was also decreased after talazoparib treatment for 6 days (Supplementary Fig. S2D), confirming that talazoparib selectively depleted the subpopulation of cells with high EGFR copy number.
Depletion of EGFR renders resistance to talazoparib
To further examine the role of EGFR in talazoparib sensitivity, we knock out EGFR in an EGFR-amplified GSC line GSC262 using the CRISPR gene editing method (refs. 28, 29; Fig. 2A). Two clones were isolated and sequenced to confirm that the EGFR gene was edited. Knockout of EGFR slowed down cell proliferation by 20% (data not shown). As shown in Fig. 2A, expression of EGFR and subsequent p-EGFR-Y1068 was completely depleted, and pSTAT3-Y705 was markedly decreased, confirming EGFR expression and signaling were inhibited by EGFR knockout. Further, we showed that knockout of EGFR renders cells resistant to talazoparib, as shown by the increased IC50 (Fig. 2B and C). We also knocked down EGFR in amplified GSC274 and GSC11 using two shRNAs targeting EGFR. Although shRNA being less effective to deplete EGFR expression in comparison with CRISPR/Cas9/sgRNA, it is noteworthy to observe a moderate resistance to talazoparib (Fig. 2D–I), further confirming EGFR-mediated sensitivity to talazoparib. In contrast, we overexpressed wild-type EGFR (EGFR-wt) and kinase-inactive EGFR (EGFR-KI) in a nonamplified/resistant cell line GSC272. pEGFR-Y1068 and pSTAT3-Y705 were increased in EGFR-wt cells but not in EGFR-KI cells, confirming that EGFR signaling was activated by EGFR-wt overexpression but not by EGFR-KI overexpression (Fig. 2J). Overexpression of EGFR-wt sensitized cells to talazoparib (Fig. 2K and L). However, overexpression of EGFR-KI failed to sensitize cells to talazoparib (Fig. 2K and L), suggesting that kinase activity is required for EGFR-mediated inhibition of cell proliferation by talazoparib.
Talazoparib sensitivity was correlated with increased ROS production in EGFR-amplified GSCs
Elevated ROS levels has been reported to be associated with EGFR activity leading to sensitivity to PARP inhibition (30). To determine if EGFR amplification induce ROS in GSCs and resultant sensitivity to talazoparib, we tested ROS levels in 10 GSCs. EGFR-amplified GSCs (n = 5) contain higher basal ROS levels than nonamplified GSCs (n = 5; P = 0.026; Fig. 3A). EGFR amplification appears to be important for ROS production and sensitivity to talazoparib as knockout of EGFR in amplified cells reduced ROS levels (Fig. 3B; P = 0.024) and decreased sensitivity to talazoparib (Fig. 2C; P < 0.0001).
To determine whether ROS contributes to EGFR-mediated talazoparib sensitivity, we treated EGFR-amplified cells with ROS-specific inhibitor MnTBAP. Our data showed that MnTBAP treatment desensitized GSCs to talazoparib (Fig. 3C). In contrast, MnTBAP failed to desensitize EGFR-nonamplified GSCs (Fig. 3C). More importantly, whereas MnTBAP significantly desensitized GSC262 vector cells, it has very slight effect on CRISPR EGFR knockout cells regarding talazoparib response (Fig. 3D), indicating that EGFR amplification mediated ROS accumulation and sensitivity to talazoparib.
To evaluate if the excess ROS induces oxidative DNA damage, we evaluated the level of 8-hydroxy-2′-deoxyguanosine (8-OHdG), the most representative product of oxidative modifications of DNA and predominant form of free radical–induced oxidative lesions in a panel of EGFR-amplified and nonamplified GSCs. Interestingly, the basal level of 8-OHdG pre-talazoparib treatment showed no difference between two groups (Fig. 3E). However, talazoparib treatment induced 8-OHdG in EGFR-amplified cells but not in nonamplified cells (Fig. 3E). These observations suggested that the ROS-induced oxidative lesions were countered by enhanced DNA damage repair, conferring a likely reliance on the key SSBR mediator PARP in EGFR-amplified GSCs.
EGFR-amplified cells have increased DNA damage repair capacity
Next, we evaluated whether the increased oxidative stress induced DDR components in EGFR-amplified GSCs. To gain better understanding of the relationship between EGFR and DNA damage repair, we analyzed the RNA-sequencing data of 27 GSC lines (12 EGFR-amplified and 15 nonamplified). A GSEA revealed several DDR and DNA-repair pathways were significantly and positively correlated with EGFR amplification, including G2–M checkpoints (NES = 1.97, P = 0.000, rank = 2), homologous recombination (NES = 1.86, P = 0.0034, rank = 11), double-strand break (DSB) repair (NES = 1.80, P = 0.0036, rank = 14), global genome nucleotide excision repair (GG-NER; NES = 1.56, P = 0.026, rank = 36), activation of ATR in response to replication stress (NES = 1.97, P = 0.000, rank = 6; Fig. 4A). These data were further substantiated by analysis of gene-expression profiling data from the TCGA GBM database that showed GG-NER genes were also enriched in EGFR-amplified patient tissue samples (Supplementary Fig. S3), thus suggesting the clinical relevance of this correlation in clinical settings. These results support the hypothesis that the EGFR-amplified group has more active DNA-repair machinery through coordinated overexpression of several key genes in the pathway.
Next, we tested if blocking DNA repair with a PARP inhibitor would result in DSB accumulation in EGFR-amplified cells. We used immunofluorescence to detect γH2AX foci and 53BP1 foci as indicator of DNA breaks. We showed that talazoparib treatment induced γH2AX foci and 53BP1 foci in EGFR-amplified GSC262 and GSC274 cells but not in the nonamplified GSC23 and GSC20 cells (Fig. 4B). Talazoparib-induced DSBs in EGFR-amplified cells are further confirmed by comet assay. As shown in Fig. 4C, comet tail moments were induced by talazoparib in GSC262 vector cells as early as 24-hour time point and persists even at 72-hour time point, whereas less induction was observed in EGFR knockout cells even at 72-hour time lapse. The DNA in the tail is represented as a percentage of moments at 0 to 72 hours span increasing from 5% to 80% in GSC262-amplified cells, and the percentage of DNA moment is decreased in EGFR knockout cells in comparison with parent cells.
To further prove that the EGFR-amplified cells have increased repair capacity to counter DNA damage, we challenged the cells with talazoparib for 72 hours to block DNA repair and allowed DNA damage to accumulate, after which talazoparib medium was replaced with fresh medium to allow cells to recover from damage for 24, 48, and 72 hours, and DNA damage was assessed by γ-H2AX and 53BP1 foci formation. As shown in Fig. 4D, in GSC262 (EGFR-amplified cells), talazoparib treatment resulted in accumulation of DNA damage as indicated by the colocalization of γ-H2AX and 53BP1 foci. After talazoparib washout, both the foci number and the percentage of positive cells decreased in a time-dependent manner, indicating the high repair capacity of the EGFR-amplified cells once the repair inhibition was removed. In EGFR knockout cells, block of repair by talazoparib treatment resulted in a weak DNA damage as indicated by the less percentage of positive cells with less foci number. After talazoparib washout, the cells recovered in a much slower way.
Similarly, in the EGFR-nonamplified GSC272 cells, blocking of DNA repair by talazoparib treatment resulted in a weak DNA damage as indicated by the low percentage of positive cells with few foci number (Fig. 4E). After talazoparib washout, the cells recovered slowly. In contrast, in GSC272 EGFR-overexpressed cells, talazoparib treatment resulted in a strong DNA damage as indicated by the increased foci number and the percentage of positive cells. After talazoparib washout, the foci number and the percentage of positive cells decreased rapidly, indicating the DNA damage was repaired efficiently.
These findings prompted us to conclude that EGFR-amplified GSCs have increased DNA damage repair capacity and are reliant on PARP to tolerate oxidative base damage and maintain genomic stability. This is further supported by the observation that PARP inhibition by talazoparib resulted in remarkable apoptosis in EGFR-amplified GSCs (Supplementary Fig. S4) due to increased DNA damage and inhibition of DNA repair.
Talazoparib is efficient at trapping PARP–DNA complex in EGFR-amplified GSCs
We next investigated the mechanism of response to talazoparib in EGFR-amplified and nonamplified GSCs. We first tested the potency of talazoparib in inhibiting total cellular poly (ADP-ribosyl)ation (PARylation) by Western blot against PAR. As shown in Supplementary Fig. S5, talazoparib reduced total PARylation levels at 10 nmol/L concentrations in both sensitive (GSC262, EGFR-amplified, IC50 = 133 nmol/L) and resistant cells (GSC272, EGFR-nonamplified, IC50 = 6930 nmol/L). Although the IC50 of talazoparib varies between the two groups, the concentration needed to fully inhibit PARylation was similar and was much lower than the IC50, indicating that the differential response to talazoparib is not purely mediated by the enzymatic inhibition of PARP.
In addition to inhibiting the enzymatic activity of PARP, trapping PARP on damaged DNA has been identified as an important and major mechanism that accounts for the cytotoxicity of PARP inhibitors (31). Because EGFR-amplified cells have more ROS-induced DNA damage accumulated than nonamplified cells following PARP inhibition, and as PARP selectively binds to damaged DNA, we determined whether talazoparib trapped more PARP–DNA complex in EGFR-amplified GSCs than in nonamplified cells as a mechanism to account for enhanced activity in amplified cells. An analysis of the subcellular fraction revealed that talazoparib treatment trapped significant chromatin-bound PARP in GSC262 (Fig. 5A) and GSC274 (Fig. 5B) cells, which are both talazoparib sensitive and EGFR amplified. About 42.1% of the total nuclear PARP was trapped at DNA at day 5 in GSC262 (Fig. 5A), and about 54.6% of the total nuclear PARP was trapped at DNA at day 5 in GSC274 (Fig. 5B). However, chromatin-bound PARP was not induced by talazoparib in resistant and nonamplified GSC272 cells (Fig. 5C). We further showed that knockout of EGFR in GSC262 cells blocked talazoparib-induced PARP–DNA complex formation (Fig. 5D), whereas overexpression of EGFR in GSC272 enhanced talazoparib-induced PARP–DNA complex formation (Fig. 5E). Collectively, these results suggested that talazoparib was able to trap more PARP–DNA complexes in EGFR-amplified GSCs, eventually inducing cell death in these GSCs.
EGFR-amplified GSCs showed selective sensitivity to pamiparib but not to olaparib or veliparib
To test if the EGFR-associated sensitivity can be observed with other PARP inhibitors and if PARP–DNA trapping capability of PARP inhibitors is important for cellular toxicity, we determined the efficacy of three other clinical PARP inhibitors—pamiparib, veliparib, and olaparib—in 14 GSCs. Pamiparib showed selective sensitivity in EGFR-amplified GSCs (Supplementary Fig. S6A–S6C). In contrast, olaparib and veliparib did not show any selective response in EGFR-amplified GSCs (Supplementary Fig. S6D–S6I). Consistent with previous reports (17, 31), our study also observed that pamiparib and talazoparib have stronger PARP–DNA trapping capacity than olaparib and veliparib (Supplementary Fig. S6J), suggesting a role for PARP trapping on damaged DNA in toxic effects induced by talazoparib and pamiparib.
Talazoparib selectively inhibits EGFR-amplified tumors in xenograft models
The toxicity of talazoparib was evaluated in nude mice after administering a daily dose of 0.33 mg/kg (b.i.d., Monday–Friday) for 6 weeks. Body weight was monitored every 2 weeks, and facial vein sampling was performed at day 42 to collect blood samples for hematology analysis. Talazoparib was well tolerated with no significant decrease in body weight (Supplementary Fig. S7A), white blood cell count (Supplementary Fig. S7B), red blood cell count (Supplementary Fig. S7C), or hemoglobin (Supplementary Fig. S7D).
To determine whether there is an EGFR amplification–dependent response to talazoparib in vivo, we established two EGFR-amplified models (GSC262 and GSC11) and two nonamplified models (GSC272 and GSC23). We observed talazoparib treatment induced a significant reduction in tumor size in the GSC262 model (FC = 12.4 at day 43; P = 0.0018; Fig. 6A) and in the GSC11 model (FC = 5.2 at day 34; P = 0.0018; Fig. 6B). In contrast, talazoparib did not suppress tumor growth for GSC23 and GSC272 (Fig. 6C and D), further confirming in vivo antitumor activity of talazoparib in EGFR-amplified tumors.
We evaluated the DNA damage and apoptosis in xenograft tumor tissues by analyzing expression of DNA DSB marker γH2AX and apoptosis marker cleaved caspase-3. Consistent with in vitro data, talazoparib treatment resulted in significant γH2AX and cleaved caspase-3–positive cells in the EGFR-amplified GSC262 and GSC11 xenograft tumors, but not in the nonamplified GSC272 or GSC23 xenograft tumors (Fig. 6E). In line with this observation, talazoparib treatment resulted in an increase of TUNEL-positive cells in EGFR-amplified GSC262 and GSC11, but not in the nonamplified GSC23 or GSC272 (Fig. 6F). Taken together, our data suggest EGFR-amplified tumor cells exhibited enhanced DNA damage and apoptosis after talazoparib treatment.
In vivo therapeutic efficacy of talazoparib in intracranial models
The in vivo efficacy of talazoparib was further tested in orthotropic intracranial xenografts. Nude mice with intracranial implantation of GSC262 cells (EGFR-amplified) and GSC272 cells (EGFR-nonamplified) were treated with vehicle or talazoparib at 0.33 mg/kg b.i.d. for 6 weeks. In the GSC262 model, talazoparib moderately inhibited tumor growth at day 36 (P = 0.044) and day 42 (P = 0.04) compared with vehicle-treated mice (Fig. 6G and H). The median survival of talazoparib-treated mice was 62 days compared with 51.5 days for vehicle-treated mice (P = 0.0039, log-rank test; Fig. 6I), indicating that treatment was able to delay the survival time by 20.4%. In contrast, talazoparib did not suppress tumor growth (Fig. 6J and K) nor extended animal survival in the EGFR-nonamplified GSC272 intracranial model (Fig. 6L, 49.5 days for control vs. 47 days for talazoparib treatment).
The efficacy is relatively moderate in the intracranial model compared with the subcutaneous model; therefore, we tested if the moderate intracranial efficacy is due to limited blood–brain penetration of talazoparib. After a single p.o. dose (0.33 mg/kg or 0.5 mg/kg) of talazoparib, brain tissues and blood samples were harvested at different time intervals, and talazoparib levels were measured. Peak concentrations in brain and whole blood were achieved 2 hours after dosing (Supplementary Fig. S6A and S6B) and decreased very quickly at hour 4. The peak concentration is as high as ∼400 nmol/L in blood (Supplementary Fig. S6A), but ∼15 nmol/L in brain tissue (Supplementary Fig. S6B), which is much lower than optimal concentrations associated with cell proliferation inhibition in vitro, suggesting that the moderate efficacy of the intracranial model is in fact due to the limited blood–brain penetration of talazoparib.
We demonstrated EGFR-dependent sensitivity to the PARP inhibitor talazoparib in a large set of clinically relevant GSC lines. The EGFR-dependent talazoparib sensitivities were profound, as indicated by 30-fold difference in the median IC50 values of talazoparib in the EGFR-amplified and nonamplified groups. Knockout of EGFR using CRISPR desensitized EGFR-amplified cells to talazoparib, and overexpression of EGFR sensitized cells to PARP inhibitor, further showing that EGFR amplification is a predictor of response to talazoparib. Given that approximately 50% of GBM patients are detected with EGFR amplification (32, 33), it is likely that a large subset of patients may benefit from talazoparib. EGFRVIII variations occurred in 30% of EGFR-amplified GBM (32, 33). It is notable that the EGFRVIII mutation rate in GSC samples is somehow lower than that detected in GBM patients, with only two cell lines (GSC280 and GSC248) out of 27 GSC lines that showed EGFRVIII mutation. Nonetheless, GSC280 and GSC248 are among the most sensitive lines to talazoparib despite their EGFR gene copy number and EGFR expression is moderate, suggesting EGFR activation is important for talazoparib sensitivity, which is further confirmed by the result that overexpression of kinase-inactive EGFR failed to sensitize cell to talazoparib.
PTEN mutation was reported to be associated with PARP inhibitors KU0058948 and KU0059436 sensitivity in colorectal tumor cell and prostate tumor cells (34). However, in the current study, we find no significant association between PTEN mutation and talazoparib sensitivity in GSC cell lines. A possible explanation for this discrepancy is that PTEN mutations are different in different tumor types, further studies comparing PTEN mutation among the different tumor types are warranted. In addition, the characteristic difference of PARP inhibitors, including PARP–DNA trapping capacity and target specificity, may also cause discrepancy. KU0058948 is the strongest inhibitor of PARP-3 activity (35), whereas talazoparib is more specific for PARP-1 and PARP-2.
In addition, IDH1 mutation and MYC amplification were also reported to confer sensitivity to PARP inhibitor in brain tumors (36–40). IDH1 mutation is a major event in low-grade gliomas (LGG) but is quite rare in GBM. IDH1 mutation occurs in 77% LGG samples but in only 4% GBM samples based on TCGA database, and we did not find any IDH1 mutation in 27 cell lines used in the current study. MYC amplification occurs in 0.2% primary GBM according to TCGA database. We observed 4 out of 27 cell lines have MYC amplification, and notably 3 cell lines are sensitive and 1 cell line is resistant to talazoparib, suggesting a possible association, but due to limited cell lines, statistical significance was not observed for MYC amplification.
Numerous studies have suggested that EGFR activation induces multiple cellular stresses, including increased oxidative stress, proliferation stress, and replication stress, all leading to DNA damage (30), requiring increased DNA-repair capacity to maintain genomic stability, and thus providing vulnerability to DNA-repair targeted therapy. Here, we showed that ROS level is higher in the EGFR-amplified cells compared with that in the EGFR-nonamplified cells. Knockout of EGFR by CRISPR decreased ROS level, further confirming that EGFR amplification leads to ROS production. ROS inhibitor MnTBAP, which suppressed ROS production, desensitized EGFR-amplified cells, suggesting that the increased ROS level is associated with talazoparib sensitivity. Our finding is consistent with previous report showing correlation between increased EGFR activity, ROS, and PARP inhibitor sensitivity by testing six primary GBM cell lines with varying EGFR expression (30). However, due to the limited number of cell lines used in this study, a robust association between EGFR expression and PARP inhibition sensitivity cannot be reached. In the current study, we used a total of 27 cell lines to show that EGFR amplification renders selective sensitivity to PARP inhibitor. Our study indicated EGFR amplification could serve as a genetic marker, which is an easier selection marker than the detection of EGFR activity and ROS level, for PARP inhibitor–responsive patients.
Although accumulating evidence supports that EGFR activation results in elevated ROS, the underlying mechanisms vary and remain to be further explored. EGFR has been reported to activate NADPH oxidase to generate excessive ROS production by activating multiple possible molecules/pathways (41–43). Furthermore, ROS functions not only as a mediator of the EGFR signaling pathway, but also as a regulator of EGFR protein activity. EGFR-mediated signaling resulted in H2O2 production and oxidation of downstream proteins (44); meanwhile, H2O2-induced sulfenylation of cysteine (Cys797) can enhance EGFR tyrosine kinase activity (45), and high ROS levels can trigger overoxidation of the Met residue of EGFRT790M and shut down the EGFR downstream survival pathway (46). The cross-talk between ROS and EGFR suggested that modulating ROS levels is a feasible strategy to overcome drug resistance. In this study, we have clearly shown that EGFR amplification induced excessive ROS production, and ROS scavenger MnTBAP desensitized cells to talazoparib treatment.
GSEA showed EGFR-amplified GSCs as well as TCGA GBM tumors have a more active DNA-repair machinery as a result of the coordinated upregulation of several DNA-repair pathways. We further assessed the DNA-repair capacity of EGFR-amplified and -nonamplified cells by washout experiments. DNA damage accumulated in EGFR-amplified or -overexpressed cells after talazoparib treatment and decreased rapidly following removing DNA damage repair inhibition by washing out of talazoparib, further demonstrating the increased DNA-repair capacity of EGFR-amplified cells in response to the increased DNA damage in these cells.
We found that more PARP–DNA complexes were trapped by talazoparib in EGFR-amplified cells than in nonamplified cells, confirming that more DNA breaks resulted from excessive ROS induced by EGFR amplification. Overall, our results suggest that EGFR-associated and drug-specific sensitivities are associated with the PARP–DNA trapping in EGFR-amplified cells.
EGFR-associated sensitivity was further evaluated in our in vivo studies. The significant inhibition of tumor growth in the two EGFR-amplified but not in the two nonamplified subcutaneous models confirmed the EGFR amplification–dependent selective response to talazoparib in vivo.
Collectively, we reported that EGFR amplification leads to elevated ROS levels, which resulted in oxidative DNA damage. To tolerate the constitutive DNA damage caused by ROS, EGFR-amplified GSCs exhibited enhanced DNA-repair activity. Talazoparib acts in two steps. First, it inhibits PARP enzymatic activity which results in accumulated unrepaired DNA breaks in EGFR-amplified GSCs. The accumulated unrepaired DNA then recruits more PARP as PARP selectively binds to damaged DNA, and DNA–PARP complex is trapped and accumulated in EGFR-amplified cells, resulting in augmented cytotoxicity.
The efficacy of talazoparib in an intracranial model is moderate, probably due to the limited blood–brain barrier penetration of talazoparib, a common issue for application of PARP inhibitors in GBM, thus limiting its therapeutic implications (47–49). Improving brain delivery by either modifying the compound or using the nanoparticle delivery system would maximize the therapeutic efficacy in GBM patients and offers an immediate opportunity for clinical translation.
Disclosure of Potential Conflicts of Interest
E.P. Sulman is an employee/paid consultant for Novocure, Blue Earth Diagnostics, AbbVie, and Merck, and reports receiving speakers bureau honoraria from Merck, Novocure, Zai Lab, and Physican's Education Resource. T.P. Heffernan is an employee/paid consultant for Cullgen Inc. W.K.A. Yung reports receiving speakers bureau honoraria from DNATrix, and is an advisory board member/unpaid consultant for DNATrix, Quadriga, and ICLT. No potential conflicts of interest were disclosed by the other authors.
Conception and design: S. Wu, E.P. Sulman, W.K.A. Yung, D. Koul
Development of methodology: E.P. Sulman, J.F. de Groot, T.P. Heffernan
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Wu, F. Gao, R. Ezhilarasan, X. Li, N. Feng, A. Multani, E.P. Sulman, D. Koul
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Wu, S. Zheng, C. Zhang, E. Martinez-Ledesma, R.G. Verhaak, J.F. de Groot, W.K.A. Yung, D. Koul
Writing, review, and/or revision of the manuscript: S. Wu, J.F. de Groot, W.K.A. Yung, D. Koul
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Ding, N. Feng, E.P. Sulman, W.K.A. Yung
Study supervision: W.K.A. Yung, D. Koul
This study was funded by a National Brain Tumor Society (Defeat GBM) grant, National Foundation for Cancer Research (NFCR) to W.K.A. Yung, a SPORE grant (P50 CA127001 to F.F. Lang), and a Cancer Center Support Grant (CA016672). The authors would like to thank Verlene Henry and Caroline Carrillo for performing animal studies and Ann Sutton in Scientific Publications department for manuscript editing.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.