Introduction

Glioblastoma multiforme (GBM) is the most aggressive primary brain tumor in adults [1]. GBM is notorious for the high degree of intrinsic and acquired resistance to current standard of care which involves surgical resection followed by radiotherapy with concurrent or subsequent mono- or combined chemotherapy (for review [1-7]). The poor efficacy of standard therapies necessitates intense research to identify novel molecular targets and define more effective therapeutic approaches [8].

Selected methylating or chloro-ethylating agents are used for glioblastoma chemotherapy. Their mode of action is based on modifications of DNA bases (TMZ, procarbazine), or interstrand cross-link formation (lomustine, nimustine) that lead downstream to secondary cytotoxic DSB (for review [5, 9]). Among the potential therapeutic strategies, modulation of the DNA repair pathways that are involved in the innate or acquired tumor resistance, is frequently under consideration. In human cells, there are two main mechanisms for DSB repair: Homologous Recombination (HR) which is error-free and the error-prone Non-Homologous End Joining (NHEJ) pathway. HR is the most important repair pathway for chemically-induced, replication-dependent DSB [10-12]. Therefore, HR is an important target to overcome tumor resistance to DNA alkylating drugs [13, 14]. In the absence of HR, unrepaired DSB become substrates for PARP [poly(ADP-ribose) polymerase-1] -dependent NHEJ as a back-up mechanism [15]. Actually, PARP plays a role in numerous pathways involved in repair of DNA single-strand and DSB induced by alkylating drugs (for a review [16]). It is well established that cells with functional defects of key HR factors, the so-called BRCAness [17] or HRDness are hypersensitive to PARPi [18, 19]. This mechanism established the clinically interesting principle of synthetic lethality [20]. Recently, it has been suggested that BRCAness in GBM is associated with mutations / deletions in IDH1/2, EGFR, PTEN, MYC proto/ oncogene and ERß which are to be considered as biomarkers for responsiveness to PARPi therapy [21]. Among the pointed BRCAness biomarkers, the tumor suppressor PTEN (phosphatase and tensin homolog) is a key regulator of the phosphatidylinositol 3-OH kinase and is known to be frequently mutated or aberrant in GBM [9, 22]. Several studies report radio-sensitization of glioblastoma stem-like cells after PARPi treatment [23, 24], others show modulating effects of PARPi on TMZ response in xenograft models [25, 26]. Based on such findings, several phase I-III clinical trials were approved to treat glioblastoma patients with TMZ in combination with different PARPi [for review [16]]. Key HR proteins like RAD51 can also be targeted by small molecule inhibitors [27, 28] to induce BRCAness and to modulate the responsiveness of GBM cells to standard chemotherapy. For example, we have previously demonstrated that a combination with a RAD51 inhibitor (RAD51i), reduced significantly the growth of subcutaneous glioma xenoplants treated with the chloronitrosourea lomustine [13].

In the present study we used the p53- and PTEN-wild-type (wt) LN229 glioblastoma cell line to determine the sensitivity to the PARPi pamiparib and Rad51i RI-1, applied either as single or combined treatments. Furthermore, we used an orthotopic GBM mouse model [29, 30] to test a combined therapy scheme with a PARPi and a RAD51i for a synthetic lethality-like response in vivo. For our study we selected pamiparib, since at least one biochemical study has demonstrated that this PARPi penetrates through intact blood–brain barrier of GBM patients and can achieve pharmacologically active drug concentrations in the tumor areas [31]. We compared this therapeutic approach to the combination of TMZ with PARPi.

Materials and Methods

Cell Culture and Treatments

The human glioblastoma cell line LN229 was used (see Table S1 Key Resources). All used drugs are listed in Table S1. For the colony formation assay, TMZ was diluted in DMSO at a concentration of 10 mmol/L and stored at -80°C. Upon requirement, aliquots were thawed and added to the cell culture medium or additionally diluted in medium to give a final concentration of 2.5 µmol/L. RAD51i RI-1 was diluted in DMSO to concentration of 25 mmol/L, PARPi pamiparib was diluted also in DMSO to concentration of 5 mmol/L and stored at -80°C. For animal experiments, concentrated solutions of TMZ, RAD51i and PARPi were prepared in DMSO to final concentrations of 10 g/L, 40 g/L and 15 g/L, respectively, and stored at -80°C until required. 

Clonogenic Survival Assay

Cells growing in the log phase were used to test colony formation as described [13, 32]. Briefly, cells (n = 400) were seeded in duplicate in cell culture dishes. They were allowed to attach, exposed to 2.5 µmol/L TMZ for 2 hours, and then the medium was changed with or without continuous inhibitor treatment (inhibitor concentrations shown in Fig.2). After 10-day incubation, formed colonies were fixed with 70% ethanol for 10 minutes, stained in crystal violet solution (1 g/L dH2O) for 10 minutes and colonies containing at least 40 cells were counted and presented graphically as a percentage of untreated cells (control). All clonogenic assays were repeated at least 3 times. Statistical significance was evaluated by the two-sided paired t-test in MS Excel.

Table S1: Key Resources Table.

*Fahrer J, Wagner S, Burkle A, Konigsrainer A. Rapamycin inhibits poly(ADP-ribosyl)ation in intact cells. Biochem Biophys Res Commun 2009; 386: p. 232-6.

Animal Experiments

An orthotopic intracranial murine model [30, 33] was used to evaluate the tumorigenic potential of LN229 GBM PTEN-wt cells in immuno-deficient NMRI-Fox-1nu/nu mice (Charles River). Animal experiments were conducted at the Translational Animal Research Center (TARC) of the University Medical Center Mainz with an approval by the State Office of chemical investigations of Rhineland-Palatinate, Mainz, Germany (permission #23 17707/G19-1-014). Immunodeficient (nude) mice (strain NMRI, sixweek old females) were commercially obtained (Charles River Europe) and housed under SPF or GF conditions in type II EU (European Union) cages inhabited by 2 to 5 companions. The mice were maintained at the standard lab diet for nude mice with water ad libitum, by 22±2°C room temperature and a 12-hour light/dark cycle. All animal groups consisted of female mice that were age-matched and free of clinical symptoms. For intracranial implantation, single cell suspensions of LN229 cells were washed twice in PBS and re-suspended at 15000 cells/µL. Cell suspension (3 µL=45000 cells) was injected into the caudato-putamen of the right hemisphere using a stereotactic frame (TSE Systems, Bad Homburg, Germany) with the following coordinates: 1 mm (anteroposterior axis), 3 mm (lateromedial axis), and 2.5 mm (vertical axis), in reference to the bregma. TMZ treatment started 12 days post implantation. Mice (n=9-10 per group) were injected intraperitoneally for the duration of 21 days. For injections, concentrated TMZ solution (10 g/L) was diluted 1:10 in sterile 0.9% NaCl and injected at a dose of 10 mg/kg body weight each other day for 3 weeks [modified after [29]]. The PARPi was diluted as suggested by the manufacturer in 0.9% NaCl supplied with 10%DMSO and 20% captisol, then injected at a dose of 12 mg/ kg body weight each other day [modified after [34]], RAD51i was dissolved in 50 µl corn oil and injected once a week [50 mg/kg body weight as described [13]]. Test animals were terminated at the manifestation of neurological symptoms related to the brain tumor or body weight reduction of >20%, otherwise 28 weeks post implantation at the latest as approved by the authorities. KaplanMeier curves were generated based on the survival time of each mouse and survival rates of the treatment groups were compared by the log-rank (Mantel-Cox) test using GraphPad Prism software (version 6.01 for Windows).

Immunohistochemistry on paraffin-embedded glioblastoma sections

We followed a previously described procedure [30, 33, 35, 36]. Briefly, the brains of sacrificed mice were fixed in 4% paraformaldehyde in PBS, sectioned in five parts, which were ordered and embedded together in paraffin. Paraffin blocks were further cooled down, sliced at 1 µm thickness and transferred to slides. For immunohistochemistry, brain sections on slides were deparaffinized by pre-heating at 60°C for 30 min and afterwards incubated in xylene (3×5 min), followed by ethanol series (100/100/96/90/80/70%), and finally rinsed in PBS to rehydrate. Further, specimens were incubated in pre-heated citrate buffer in a steamer for 20 min to retrieve the antigens, then allowed to cool down at room temperature for 20 min. Further, sections were rinsed in PBS then, after blocking, incubated overnight at 4°C with primary antibodies (disclosed in Key Resources Table S1). Incubation with secondary antibodies was performed for 2 h at room temperature: goat Alexa488- or Cy3-conjugated antirabbit or anti-mouse (1:600; Jackson ImmunoResearch Eu-rope). Nuclei were counter-stained with TO-PRO–3 (1:100) or DAPI and mounted with Fluoromount fluorescence mounting medium. For To-Pro-3 stained slides, images were taken using the 63x oil objective of LSM710 system equipped with ZEN 2009 software. For tumor area measurements, images were taken on DAPI stained slides using laser micro-scope DMIRB (Leica) equipped with LAS software Version 4.0.0. Images were exported and PCNA, Ki67 or γH2AX positive cells were counted using the Cell Counter function of ImageJ software. Treatment groups (n=3-4 animals per group) were compared by unpaired t-test in MS Excel. Tumor area was measured in square µm in ImageJ and statistical evaluation was performed in MS Excel by unpaired t-test.

Immunofluorescence staining of cells grown on coverslips (Figures S1C and S2A) were performed as previously described [11]. Intranuclear γH2AX foci per nucleus were scored using ImageJ software. Cell clones and treatment variants were compared by unpaired t-test in MS Excel.

TUNEL assay on paraffin embedded sections

In order to detect apoptotic cells in the tumor mass, sections were stained with InSitu Cell Death Detection Kit (Roche) according to a modified version of the manufacturer’s protocol. Briefly, reaction mixture was prepared containing 15U of the enzyme terminal deoxynucleotidyl transferase (TdT), 1xBuffer TdT, 1 mM dUTP labeled with biotin, then sections were incubated with 100 µl mixture for 60 min at 37°C. After washing with PBS and blocking, samples were further stained for Nestin to recognize glioblastoma cells. The primary anti–Nestin antibody was recognized by an anti-mouse Alexa488-coupled secondary antibody whereas biotin was tracked by Streptavidin conjugated with Cy3 (1:200). Nuclei were counterstained with To-Pro-3. Counting of TUNEL+ cells and statistical evaluation were performed similar to proliferating cells using ImageJ software and MS Excel.

Results

Increased responsiveness of GBM cells to combined DNA repair inhibitor treatment with PARPi and or RAD51i in cellulo

In order to test the sensitivity of LN229 GBM cells to DNA-damaging tumor treatments including RAD51 inhibitior (RAD51i), PARPi and TMZ we aimed to perform colony formation assays. First, we validated the inhibitory activity of the PARPi pamiparib on formation of poly-ADP-ribose (PAR) polymer foci in the cells upon treatment with H2O2. As expected, PARPi efficiently abolished PAR foci formation indicating its functionality in cells (Figure S1).

Supplementary Figure S1: PARPi Pamiparib inhibits accumulation or PAR-specific modifications. Immunofluorescence detection of poly-ADP-ribose (PAR)-specific modifications in LN229 cells using a PAR-specific antibody after pretreatment with PARPi for 24 h and treatment with hydrogen peroxide for 5 min

In general, analysis of the colony formation assays indicated that the responsiveness of the GBM cells was relatively low to the used drug concentration for the single treatments, which was intended to be able to determine potential additive or synergistic effects of the drug combinations (Figure 1A, B). Along these lines, significantly higher sensitivity was found to combinatorial drug treatments (Figure 1A, B), which suggests synthetic lethality. In addition, combined treatment of TMZ along with PARPi had a stronger growth inhibitory effect compared to the respective mono treatments. Remarkably, RAD51i, which did not reduce cell survival upon single treatment at the used concentration, significantly potentiated the killing effect of TMZ and of PARPi (Figure 2A, B). Collectively, our results identified drug combinations that show synthetic lethality-like effects and chemosensitization, and identities the combination of RAD51i with TMZ or PARPi as efficient drug combinations.

Figure 1: Sensitivity of human LN229 glioblastoma cells to TMZ, PARPi, RAD51i and combined treatments in vitro. (A) Representative images of the stained colonies for each treatment variant.  Cells (n=400) were seeded in duplicate, after adherence treated for 2 h with TMZ, then inhibitors were added continuously till harvest. Cultures were incubated for 10 days, formed colonies were fixed, stained and counted. (B) Clonogenic survival was determined in the colony formation assay. Each measure point represents the mean of three independent experiments ± standard deviation (SD). (C) p-values of the comparisons of single treatment variants to combined exposure.

Figure 2: Survival of mice harboring orthotopic GBM after inhibitor treatments. (A) Experimental scheme: Glioblastoma cells (n=40,000) were injected into the brains of NMRI-Fox-1nu/nu mice. Twelve days post implantation, treatment with PARPi (pamiparib), RAD51i (RI-1) or combination of these drugs started. The untreated controls (SHAM) are mice injected with solvent only. Generation of Kaplan-Meier curves and statistical analysis are described in Material and Methods. (B) Survival of SHAM-treated mice vs treated with RAD51i. (C) Survival of SHAM-treated mice vs treated with PARPi. (D) Significantly improved survival of mice treated with RAD51+PARPi (p=0.0361) compared to SHAM-exposed. (E) Median survival times (MST) for each treatment group calculated using GraphPad Prism software.

In vivo analysis of DNA repair inhibitor treatments in mice with orthotopic GBM xenoplants

Our cell culture-based data of the colony formation assays suggested synthetic lethality-like effects in glioblastoma cells after combined exposure to RAD51i and PARPi. Since this drug combination to our best knowledge has not been reported so far for GBM treatment, we aimed to validate our results in vivo using an orthotopic glioblastoma murine model based on intracranial implantation of GBM cells for subsequent tumor growth. In brief, mice were anesthetized and GBM cells were injected into the caudato-putamen of the right hemisphere. Subsequently, mice were housed for 12 days to allow brain tumor growth, then cancer treatments were performed as indicated (Figure 2A). Test animals were sacrificed at the manifestation of neurological symptoms related to the brain tumor growth or body weight reduction over 20%. Survivors were terminated 28 weeks post implantation. Differences in survival rates between the groups were determined on Kaplan-Meier curves using the log-rank (Mantel-Cox) test.

Our results show that, treatment with a combination of PARPi and RAD51i caused significant increase in survival time when compared to the SHAM control group (p=0.03, Figure 2D, 2E). There was also a tendency of prolonged survival following RAD51i treatment alone but the difference did not reach the level of significance (p=0.15, Figure 2B).

Measurements of the tumor areas in the mouse brains confirm the tumor-retarding effects of the combined inhibitor therapy for GBM implants, since in mice treated with both inhibitors it was significantly smaller if compared to those of the SHAM exposed group (p=0.0284, Figure 3B, representative images in Figure 3A). Taken together, our results suggest that also in vivo the combined RAD51i and PARPi treatment is of benefit in a cancer mouse model.

Figure 3: Measurements of the tumor area in brains of mice harboring orthotopic GBM implants. Mouse brains were fixed in 4% paraformaldehyde, sectioned in five parts and paraffin-embedded. 1 µM slices were generated. (A) Representative images of tumor slices from SHAM- or inhibitor treated mice are shown (tumor areas are delineated in white colour). Slide preparation, staining and tumor area measurement are described in Material and Methods. (B) Quantification of tumor area presented in square mm. Statistical evaluation was done using unpaired t-test in MS Excel.

Combined PARPi and Rad51i treatment significantly deminishes GBM proliferation in vivo

Considering that the implanted human glioblastoma cells are the only rapidly dividing cells in the host brain, we used immunofluorescence analysis on brain sections to assess expression of cell proliferation markers. We performed immune-histochemistry double-labeling for the proliferation markers PCNA or Ki67 and the neural stem cell (NSC) marker Nestin, which is known to be also expressed in glioblastoma stem cells [35] and GBM [37]. We counted the PCNA positive cells among the population of Nestin positive glioblastoma cells (representative LSM images in Figure 4A) on brain slices. Of note, the percent of PCNA expressing cells was significantly reduced in the tumors after PARPi or combined PARPi+RAD51i therapy (Figure 4B). Furthermore, we determined the expression of the clinically relevant proliferation marker Ki67 [38-40] (shown in Figure 5A). Cell counts confirmed significantly reduced numbers of Ki67 positive, proliferating tumor cells after combined inhibitor therapy (Figure 5B), and correlated well with the results on PCNA expression. This significant decrease of proliferating tumor in mouse brain slices cells is in accordance with recent evidence that PARPi pamiparib penetrates the blood-brain barrier and is well retained in the tumor tissue [31]. Taken together, these results confirm the tumor growthretarding effects of the drug combination PARPi +RAD51i.

Figure 4: Immunofluorescence-based detection of PCNA in GBM xenoplants in mouse brain tumor sections. (A) Representative images of nestin/PCNA labeling on brain sections from the different treatment groups. Brains of terminated mice (n=3-4) were processed as described in Material and Methods. Sections were incubated with specific primary antibodies against Nestin and PCNA, followed by incubation with secondary antibodies coupled with fluorochromes (red = Cy3, green=Alexa488). Nuclei were counterstained with To-Pro-3. (B) Quantification of PCNA positive cells detected among the population of nestin positive human glio-blastoma cells after SHAM, PARPi, RAD51i or combined inhibitor exposure. Significant decrease of proliferating GBM cells was found after PARPi or combined inhibitor exposure.

Figure 5: Immunofluorescence-based detection of Ki67 in GBM xenoplants in mouse brain tumor sections. (A) Representative images of Nestin/Ki67 labeling on brain sections from the different treatment groups. Brains were processed as described in Material and Methods. Deparaffinized brain sections were blocked, incubated with specific primary antibodies against NSC marker nestin and proliferation marker Ki67, followed by incubation with secondary antibodies coupled with fluorochromes (red=Alexa555, green=Alexa488) and nuclear counterstaining with DAPI (blue). (B) Quantification of Ki67 positive cells from mice exposed to the vehicle (SHAM), PARPi, RAD51i or inhibitor combination. Statistical evaluation showed significantly less proliferating cells after combined inhibitor exposure compared to SHAM-treated or treated with PARPi alone (*p=0.05).

DNA repair inhibitors lead to increased DSBs in GBM cells in vivo

Further, we expected that the administered DNA repair inhibitors would cause accumulation of DSBs [16] after penetration into the glioblastoma cells. To detect DSBs, we applied a specific antibody against the DSB marker histone 2AXSer139 (γH2AX) (representative images Figure 6A).  Despite the quite variable counts of γH2AX foci in the animals from one and the same group, the group treated with the inhibitor combination differed with high significance from SHAM exposed animals (Figure 6B). At least in part, the observed variability might be explained by the fact that the mice were sacrificed at different time points after treatment in accordance with the termination criteria for this pre-clinical animal model.

Figure 6: Immunofluorescence-based detection of γH2AX in GBM xenoplants in mouse brain tumor sections. Detection of the DNA damage marker γH2AX on mouse brain sections. (A) Representative images of double staining for Nestin/DNA damage marker γH2AX. Sections were incubated with specific primary antibodies against nestin and γH2AX, followed by incubation with secondary antibodies coupled with fluorochromes (red = Cy3, green=Alexa488). Nuclei were counterstained with To-Pro-3. (B) Quantification of γH2AX positive cells detected among the population of nestin positive human glioblastoma cells after SHAM, PARPi, RAD51i combined inhibitor exposure. γH2AX positive cells were scored using ImageJ software and counts were compared by t-test in MS Excel. Statistical evaluation showed highly significant increase in the number of cells with DNA damage after combined inhibitor exposure compared to SHAM exposure (**p=0.01).

Combined PARPi and Rad51i treatment inceases apoptosis of GBM cells in vivo

Apoptosis is a prominent cell death mechanism activated in response to DSBs induction [41]. Therefore, we investigated whether the observed antitumor effects might be mediated by induction of apoptosis in the orthotopic GBM cancer xenoplants. To this end, we performed TUNEL assays to detect apoptotic cells on brain sections of mice sacrificed at the end of the therapy. To identify the tumor area, we performed co-stainings of Nestin along with TUNEL (Figure 7A). Of note, quantification of the results showed a significant increase of apoptotic TUNEL positive cells after combined RAD51 inhibitor and PARPi treatment (p=0.018) (Figure 7B). A tendency, however, not of significance was also observed after PARPi monotherapy (p=0.0673), whereas RAD51i monotherapy showed no increase in TUNEL positivity (Figure 7B). In association with the decreased tumor areas reported (Figure 3B) and the reduced rate of proliferating cells after these treatments (Figure 4B and Figure 5B), these findings provide an explanation for the (modestly) prolonged median survival after combined inhibitor treatment of mice harboring human glioblastoma-derived tumors. Taken together, our results suggest that combined treatment with RAD51i and PARPi is a promising novel treatment strategy, that might be considered for treatment of GBM in the future.

Figure 7: Detection of apoptotic cells by TUNEL assay on mouse brain sections. (A) Representative images of double staining for Nestin/TUNEL. Sections were stained with InSitu Cell Death Detection Kit (Roche), blocked and subsequently stained for nestin. Nestin antibody was recognized by Alexa488-coupled secondary antibody (green), biotin-labeled dUTP (used in the TUNEL reaction) was recognized by Cy3-coupled streptavidin (red) and nuclei were counterstained with To-Pro-3 (blue). (B) Quantification of TUNEL positive cells after SHAM, PARPi, RAD51i or combined inhibitor exposure. Significantly more apoptotic cells were observed after combined inhibitor exposure compared to SHAM exposure (*p=0.05).