The O⁶-Methyguanine-DNA Methyltransferase Inhibitor O⁶-Benzylguanine Enhanced Activity of Temozolomide + Irinotecan Against Models of High-Risk Neuroblastoma

Abstract

DNA-damaging chemotherapy is a major component of therapy for high-risk neuroblastoma, and patients often relapse with treatment-refractory disease. We hypothesized that DNA repair genes with increased expression in alkylating agent resistant models would provide therapeutic targets for enhancing chemotherapy. In vitro cytotoxicity of alkylating agents for 12 patient-derived neuroblastoma cell lines was assayed using DIMSCAN, and mRNA expression of 57 DNA repair, 3 transporter, and 2 glutathione synthesis genes was assessed by TaqMan Low Density Array with further validation by qRT-PCR in 26 cell lines. O⁶-methylguanine-DNA methyltransferase (MGMT) mRNA was upregulated in cell lines with greater melphalan and temozolomide (TMZ) resistance. MGMT expression also correlated significantly with resistance to TMZ + irinotecan (IRN) (in vitro as the SN38 active metabolite). Forced overexpression of MGMT (lentiviral transduction) in MGMT non-expressing cell lines significantly increased TMZ+SN38 resistance. The MGMT inhibitor O⁶-benzylguanine (O6BG) enhanced TMZ+SN38 in vitro cytotoxicity, H2AX phosphorylation, caspase-3 cleavage, and apoptosis by terminal deoxynucleotidyl transferase dUTP nick end labeling. TMZ+IRN+O6BG delayed tumor growth and increased survival relative to TMZ+IRN in 2 of 7 patient-derived xenografts established at time of death from progressive neuroblastoma. We demonstrated that high MGMT expression was associated with resistance to alkylating agents and TMZ+IRN in preclinical neuroblastoma models. The MGMT inhibitor O6BG enhanced the anticancer effect of TMZ+IRN in vitro and in vivo. These results support further preclinical studies exploring MGMT as a therapeutic target and biomarker of TMZ+IRN resistance in high-risk neuroblastoma.

Introduction

Neuroblastoma is the most common pediatric extracranial solid tumor, accounting for 15% of pediatric cancer deaths, and high-risk neuroblastoma patients receive multimodal therapy consisting of surgery and induction chemotherapy, dose-intensive myeloablative chemotherapy with autologous peripheral blood stem cell support [1–3], followed by maintenance therapy with isotretinoin, dinutuximab, and cytokines [4]. The multi-agent chemotherapeutic regimens used for both induction and myeloablative therapy rely heavily on alkylating agents [1,3].

Despite intensive treatment many children diagnosed with high-risk neuroblastoma have treatment-refractory disease progression, with long-term overall survival of about 50% [2]. Among metastatic patients who relapse, 5-year event-free survival (EFS) is a dismal 8% [5]. Chemoresistance is associated with p53 loss-of-function [6,7], alternative lengthening of telomeres (ALT) phenotype [8], ALK mutation [9], overexpression of HDAC1 [10], overexpression of MDR1 [11], and expression of MRP1 [12]. We hypothesized that increased DNA repair capacities contribute to resistance to DNA-damaging therapies and that targeting such pathways could reverse drug resistance.

The bifunctional alkylating agent melphalan (LPAM) is a component of myeloablative chemotherapy [1,3,13], inducing primarily N⁷-guanine and N³-adenine interstrand crosslinks and monoadducts [14]. Temozolomide (TMZ) is an alkylating agent which methylates primarily N⁷-guanine, N³-adenine, and O⁶-guanine, with O⁶-guanine methylation being the driver of cytotoxicity [15]. According to the futile cycling hypothesis, unrepaired O⁶-methylguanines base pair like adenines, giving rise to ^meG-T mispairs upon DNA replication and inducing loci of futile mismatch repair (MMR) cycling, single and double strand breaks, and replication fork collapse as the thymine and surrounding nucleotides are repeatedly excised and erroneously resynthesized [15,16]. The maximum clinically achievable peak plasma concentration for TMZ is around 100 μM, but peak plasma concentrations as dosed for neuroblastoma are ~20–50 μM [17–19]. Irinotecan (IRN) is a topoisomerase 1 inhibitor that is metabolized in the liver into SN38, a product 100-fold to 1000-fold more cytotoxic than IRN [20]. Following intravenous dosing of 20 mg/m², peak IRN plasma concentrations were ~50–200 ng/ml (~85–340 nM) with SN38 levels of ~1.5–20 ng/ml (~4–50 nM) [21]. Temozolomide in combination with irinotecan (TMZ+IRN) is active against progressive disease high-risk neuroblastoma [19], especially when combined with the anti-GD2 antibody dinutuximab [22], and future relapsed neuroblastoma studies will test the addition of other agents to improve this regimen [23].

O⁶-methylguanine-DNA methyltransferase (MGMT) is a DNA repair protein that directly removes a single O⁶-guanine alkylation before being ubiquitinated and targeted for proteasomal degradation [24]. The number of possible alkylation repairs is stoichiometrically proportional to the amount MGMT protein in the cell, and further repairs following MGMT depletion require de novo MGMT synthesis [25]. MGMT gene silencing by promoter cytosine methylation at CpG islands is associated with a favorable response to TMZ +/− radiotherapy in glioblastoma, with MGMT-methylated cohorts surviving nearly twice as long as unmethylated cohorts in many trials [26].

O⁶-benzylguanine (O6BG) and lomeguatrib are pseudosubstrate inhibitors of MGMT that have been evaluated as chemosensitizing drugs for alkylating agents in various cancers, including glioma [27], multiple myeloma [28], and melanoma [29]. O6BG covalently transfers its benzyl group onto the active site cysteine-145 of MGMT, leading to MGMT degradation [25,30]. The clinically achievable plasma concentration of O6BG is 25 μM [30], with toxicities being primarily myelosuppressive when used with alkylating agents [27]. Lomeguatrib has a mechanism similar to O6BG, covalently transferring a bromothenyl group onto the active site cysteine of MGMT [29].

In approaching the hypothesis that upregulated DNA repair contributes to neuroblastoma chemoresistance we screened for DNA repair genes with higher expression in cell lines with higher resistance to melphalan and TMZ. This screening yielded MGMT as a possible target, for which clinical-stage inhibitors exist. We then evaluated MGMT as a contributor to chemoresistance to the clinically employed re-induction regimen TMZ+IRN. Finally, we evaluated the efficacy of MGMT inhibitors using in vitro and in vivo neuroblastoma models established at the time of death from disease progression.

Materials and Methods

Tissue culture.

All CHLA- series and COG-N- series cell lines were cultured without antibiotics in IMDM with 1X ITS (Corning cat# 354350), 4 mM L-glutamine, and 20% FBS (IMDM+FBS) in 20% O₂ (room air + 5% CO₂) or in 5% O₂ (bone marrow level) as described [31]. SMS-, SK-N-, and LA-N- series cell lines were also grown in IMDM+FBS, except during the TaqMan Low Density Array screening and qRT-PCR validation, where they were grown in 10% FBS-supplemented RPMI-1640 [6]. Cell lines were established from blood, bone marrow, or tumor taken at diagnosis (Dx), progressive disease following induction therapy (PD), progressive disease following myeloablative therapy (PD-BMT), or progressive disease leading to death (post-mortem, PM) (see Table S1, Supplemental Digital Content, which shows cell line characterization). Adherent cells were suspended using Puck’s saline A + EDTA [32]. Cell line identity was verified by short tandem repeat (STR) matching using the AmpFLSTR Identifiler Kit (Thermo) validated against the Children’s Oncology Group database at www.CCCells.org. Cell line and PDX identities were re-confirmed by STR at time of experimentation. Cell lines were verified free of mycoplasma using the MycoAlert Mycoplasma Detection Kit (Lonza), and verified EBV negative [33]. Tyrosine hydroxylase (TH) mRNA expression was used to confirm adrenergic neural origin as described [31]. MYCN genomic amplification was defined as ≥10 gene copies (by PCR) per diploid genome [31]. Functionality of p53 was defined by p21 induction in response to melphalan [6].

TaqMan Low Density Array (TLDA) screening and qRT-PCR.

RNA was extracted using Qiagen’s RNeasy kit and quantified using NanoDrop2000. cDNA was synthesized at 100 ng/μl using a High Capacity cDNA Reverse Transcription Kit (Thermo).

TLDAs for mRNA screening of 57 DNA repair, 3 transporter, and 2 glutathione synthesis genes (see Table S2, Supplemental Digital Content, which shows TLDA genes) were loaded with 100 ng cDNA per reservoir and analyzed using an ABI 7900HT Fast Real-Time PCR System and RQ manager 1.2 software.

We carried out qRT-PCR in 96-well plates using the ABI 7900HT or Bio-Rad CFX96 Real-Time PCR Detection System using TaqMan Gene Expression Assays (Thermo, GAPDH 4352934E, MGMT Hs01037698_m1) with TaqMan Gene Expression Master Mix (Thermo) or custom-designed primers (Integrated DNA Technologies; MGMT Fwd: 5’-ACCGTTTGCGACTTGGTACT-3’, MGMT Rev: 5’-TGCTCACAACCAGACAGCTC-3’, GAPDH Fwd: 5’-TGGAAGGACTCATGACCACAG-3’, GAPDH Rev: 5’-CAGCTCAGGGATGACCTTGC-3’, HPRT1 Fwd: 5’-AACCTCTCGGCTTTCCCG-3’, HPRT1 Rev: 5’-CACTAATCACGACGCCAGG-3’) with Power SYBR Green PCR Master Mix (Thermo). Primers were designed using NCBI Primer-BLAST and verified by gel electrophoresis to amplify a single product. Relative expression was calculated by the ΔΔCt method, normalized by GAPDH or HPRT1.

DIMSCAN cytotoxicity assay.

DIMSCAN was used to assay cytotoxic drug response [34,35]. With melphalan, cells were seeded in 150 μl medium for 24 hours, then exposed to melphalan (50 μl addition) for four days [35]. With TMZ and TMZ+SN38+/−O6BG, cells were seeded in 96-well plates in 150 μl IMDM+FBS for 24 hours, followed by +/−24 hour pretreatment of 25 μM steady-state O6BG (50 μl addition), and then TMZ and/or SN38 in a two-fold serial dilution curve with 8–9 days incubation at a fixed ratio of 100 μM to 5 nM (50 μl addition). Lomeguatrib was dosed identically to O6BG, but at 5 μM. TMZ+SN38 chemosensitization testing with O6BG and lomeguatrib was carried out in bone marrow hypoxia (5% O₂), and all other cytotoxicity studies were carried out in 20% O₂.

MGMT protein by western blot.

Protein was extracted using RIPA buffer (Pierce RIPA buffer, Thermo) with protease inhibitor (Protease Inhibitor Cocktail Set III, Calbiochem), prepared in Laemmli buffer, and separated by SDS-PAGE. Transfer was by semi-dry method (BioRad Trans-Blot SD). Blots were probed using commercial antibodies in 5% BSA in 0.1% TBST and imaged on a VersaDoc imaging system (Bio-Rad). Antibodies were diluted 1:1000 and included: anti-MGMT MT3.1 clone (Thermo, MS-470-P1), anti-B-actin (Santa Cruz, sc-47778), anti-mouse IgG HRP-linked (Cell Signaling, 3076S), and anti-rabbit IgG HRP-linked (Cell Signaling, 7074S).

MGMT forced overexpression.

The coding sequence for the MGMT gene was spliced out of Origene’s TrueClone MGMT Human cDNA Clone (pCMV6-AC vector, cat# SC322190) and inserted into a pLenti-IRES-Puro plasmid for expression under the CMV promoter. The resulting plasmid construct was transformed into E. coli for amplification under ampicillin selection and extracted by Qiagen Plasmid Maxi Kit. The MGMT overexpression plasmid was transfected along with lentiviral packaging plasmids into HEK-293FT cells using Origene’s Lenti-vpak packaging kit (cat# TR30022) in Opti-MEM medium, which was replaced with IMDM+FBS after 16 hours. Medium was collected at 48, 72, and 96 hours and used to transduce CHLA-225 and COG-N-269 cells in the presence of hexadimethrine bromide. Transduced neuroblastoma cells were selected with 1.5 μg/ml puromycin in IMDM+FBS. MGMT protein expression was confirmed by western blot.

Cleaved caspase-3 and phospho-histone H2AX assays for apoptosis and double-strand breaks.

Neuroblastoma cells cultured in IMDM+FBS at 5% O₂ were exposed to 20 μM TMZ + 1 nM SN38 +/− 24 hour pretreatment with 25 μM O6BG. Control flasks were dosed with drug vehicles. Cells were collected 72 hours following TMZ+SN38 dosing. Pellets were extracted in RIPA buffer (Pierce) with 1 mM sodium orthovanadate, 1 mM sodium fluoride, and protease inhibitor cocktail (Set III, Calbiochem). Cleaved caspase-3 and p-gamma-H2AX were assessed by immunoblot and imaged on a Bio-Rad VersaDoc imaging system (CHLA-225, COG-N-269, and COG-N-452h) or using Amersham Hyperfilm ECL (COG-N-561h). Anti-cleaved caspase-3 (Cell Signaling, 9664S) and anti-phosphoserine-139 gamma-H2AX (Novus, NB100–384) antibodies were diluted 1:1000.

Flow cytometric TUNEL assay for apoptosis.

TUNEL (APO-DIRECT kit, BD Biosciences) was used to evaluate apoptotic DNA fragmentation. Neuroblastoma cells in IMDM+FBS at 5% O₂ were cultured with 20 μM TMZ + 1 nM SN38 +/− 24 hour pretreatment with 25 μM O6BG. Control flasks were dosed with drug vehicles; 2–3 million cells were collected 120 hours following TMZ+SN38 dosing, fixed in 1.0% w/v paraformaldehyde in PBS for one hour on ice, and stored in 70% ethanol at −20°C for 24 hours. Cells were then washed with PBS and processed by APO-DIRECT kit. Cells were incubated in 50 μL of DNA labeling solution at 37°C for three hours, washed twice in Rinse Buffer, and suspended for 30 minutes in 500 μL of PI/RNase Staining Buffer. Cells were analyzed on a LSR-II flow cytometer (BD Biosciences) with data analysis by FlowJo software.

Neuroblastoma xenografts.

The TTUHSC Institutional Animal Care and Use Committee (IACUC) approved all animal protocols. Four cell line-derived xenografts (CDX), CHLA-79, CHLA-119, CHLA-171, CHLA-136, and seven patient-derived xenografts (PDX), Felix-PDX, COG-N-421x, COG-N-452x, COG-N-470x, COG-N-519x, COG-N-561x, and COG-N-564x, were used for this study. PDXs were established from blood obtained from patients at time of death from progressive disease after multiagent chemotherapy (see Table S3, Supplemental Digital Content, which shows PDX characterization). Tyrosine hydroxylase mRNA expression and H&E staining of formalin-fixed paraffin-embedded sections were used to verify neuroblastoma origin of new PDX models. After establishment in NSG mice, each PDX was verified to match the patient of origin by STR profiling and was passaged in NSG or NU/NU mice. Strained cells from a previous tumor were prepared for injection by suspension in RPMI-1640 then mixed 50/50 with Matrigel (Corning), drawn into a 1 ml insulin syringe, and kept on ice. NU/NU mice were injected subcutaneously between the shoulder blades with 200 μl of cell preparation containing 5 to 20 million viable cells. Mice were randomized to treatment groups when tumors reached 150–250 mm³, with tumor volumes calculated as: volume = 0.5 * height * width * length [36]. CDX experiments were carried out similarly, as described [37].

The xenograft dosing for TMZ, IRN, and O6BG was as established by Cai, et al. 2010 to mimic clinical dosing [38]. O6BG (Sigma) was dissolved in DMSO, then diluted with a 60:40 mix of 0.9% saline and PEG-400. TMZ was prepared as a slurry in water; 20 mg/ml irinotecan HCl (Sagent) was diluted 10x in 0.9% saline. The 21-day cycle consisted of five days of treatment followed by 16 days off. O6BG was injected intraperitoneally at 30 mg/kg, followed one hour later by TMZ by oral gavage at 25 mg/kg, followed one hour later by IRN by tail vein injection at 7.5 mg/kg. Control mice were treated with appropriate drug vehicles. Tumors were measured at least once per week. Tumor progression endpoint was defined as 1500 mm³, and other adverse event endpoints included lethargy and failure to groom, infections, abdominal distension, and >20% weight loss. Progressive disease was defined as failure to reduce tumor volume by 50% relative to the volume at randomization. Partial response was defined as >50% reduction of tumor volume, without complete clearance. Complete response was defined as no detectable tumor for at least one measurement. Maintained complete response was defined as no evidence of tumor at 200 days.

Tumor cleaved caspase-3 and MGMT protein were assessed by western blot in COG-N-452x. Mice were sacrificed and tumors snap frozen on day 5, one hour after treatments.

Statistical analyses.

IC₅₀ and IC₉₀ values were calculated using linear interpolation. Comparisons of mean IC₅₀, mean expression, and mean weight loss were by Student’s t-test with significance at P < 0.05. Correlations were by Spearman’s r with significance at P < 0.05. Statistical analysis of flow cytometric TUNEL was by one-way ANOVA with Tukey posttest (P < 0.05). Xenograft EFS differences were evaluated by the log-rank test, with significance at P < 0.05. Excel was used for Student’s t-test calculations, and GraphPad Prism was used for all other statistical calculations.

Results

MGMT mRNA and MGMT protein expression correlated with alkylating agent resistance.

To identify DNA repair genes potentially related to alkylator resistance, we began by comparing expression between cell lines grouped by in vitro response to alkylating agents. TMZ cytotoxicity and melphalan cytotoxicity were assessed in 12 and 14 neuroblastoma cell lines, respectively (Fig. 1A, which shows TMZ cytotoxicity, and Fig. S1A, Supplemental Digital Content, which shows melphalan cytotoxicity), and mRNA expression of 62 genes was screened by TLDA (see Tables S2 and S4, Supplemental Digital Content, which show the TLDA genes and the screening cell lines, respectively). TMZ resistance groups were defined by whether the IC₅₀ was greater or less than 200 μM, a threshold that divided the screening cell lines into two clusters (Fig. 1A). Of note, TMZ generally shows low in vitro activity, and doses above the clinical maximum plasma concentration (Cmax) are common for in vitro studies [38,39]. The melphalan resistance groups were defined by IC₉₀ greater or less than 30 μM, as previously described [40]. We observed that O⁶-methylguanine-DNA methyltransferase (MGMT) upregulation was associated with greater TMZ and melphalan resistance (P < 0.05, Fig. 1B, which shows MGMT expression in the TMZ panel and Fig. S1B, Supplemental Digital Content, which shows MGMT expression in the Melphalan panel).

Figure 1. MGMT correlated with alkylator resistance in neuroblastoma cell lines in vitro.

A) Temozolomide (TMZ) cytotoxicity curves for cell lines used in TLDA screening. Cell lines with TMZ IC₅₀ > 200 μM are colored black, and cell lines with TMZ IC₅₀ < 200 μM are colored gray. The 50% survival fraction is demarcated by the dotted line. B) Cell lines with greater TMZ resistance had higher MGMT mRNA expression by TLDA (P < 0.05). C) TMZ cytotoxicity curves for an expanded validation panel of relapse neuroblastoma cell lines. D) Expression of MGMT mRNA was assayed by qRT-PCR in the validation panel. MGMT was upregulated in the cell lines with greater TMZ resistance (P < 0.05). Refer to Tables S4 and S5 in Supplemental Digital Content for the list of cell lines used in the TLDA screening assay and in the qRT-PCR validation assay, respectively. E-K) Correlation was evaluated between resistance to the combination of TMZ + SN38 in vitro, MGMT protein expression, and MGMT mRNA expression across 14 progressive disease neuroblastoma cell lines with a range of MGMT expression. E) Representative western blot probed with anti-MGMT MT3.1 clone (Thermo, MS-470-P1) antibody in the 14 progressive disease neuroblastoma cell lines F) Bar graph showing densitometric quantitation of MGMT from western blots as shown in E (mean of duplicates with SEM). G) Bar graph of MGMT mRNA expression (mean of triplicates with SEM, ΔΔCt HPRT1-normalized). H) Bar graph of TMZ+SN38 IC₅₀ corresponding to cytotoxicity curves shown in Fig. S2 in Supplemental Digital Content (mean of triplicates with SEM). TMZ and SN38 were dosed as a fixed ratio of 100 μM TMZ: 5 nM SN38 with arbitrary units (AU) set equal to TMZ concentration for graphing purposes. I) Spearman correlation of MGMT mRNA and MGMT protein expression (P < 0.05). J) Spearman correlation of MGMT mRNA and TMZ+SN38 IC₅₀ (P < 0.05). K) Spearman correlation of MGMT protein and TMZ+SN38 IC₅₀ (P < 0.05). I-K) Curves were fit by least squares linear regression. Dotted line shows 95% confidence interval. The CHLA-70 IC₅₀ data point (open circle) was removed as an outlier by GraphPad Prism’s automatic outlier elimination.

Using expanded validation panels of cell lines cultured at 20% O₂, we assessed TMZ cytotoxicity (26 cell lines) and melphalan cytotoxicity (27 cell lines) by DIMSCAN, along with MGMT mRNA expression by qRT-PCR (see Tables S5 and S6, Supplemental Digital Content, which show the expanded cell line panels for TMZ and Melphalan, respectively). As expected, the 11 cell lines with TMZ IC₅₀ > 200 μM showed higher mean MGMT mRNA expression than the 15 cell lines with TMZ IC₅₀ < 200 μM (P < 0.05, Fig. 1C and 1D). Similarly, the 14 cell lines with melphalan IC₉₀ > 30 μM showed higher mean MGMT mRNA expression than the 13 cell lines with IC₉₀ < 30 μM. (P < 0.05, Fig. S1C and S1D, Supplemental Digital Content, which show melphalan cytotoxicity and MGMT expression, respectively).

The combination of temozolomide and irinotecan (TMZ+IRN) was selected for further evaluation in the context of MGMT due to the high relevance of TMZ+IRN to relapse neuroblastoma salvage therapies and due to the importance of MGMT in repairing TMZ-induced methyl lesions [15,23]. As MGMT does not confer melphalan resistance in HeLa and CHO cells [41], melphalan was not pursued for further studies. We evaluated MGMT protein expression (Fig. 1E and 1F), MGMT mRNA expression (Fig. 1G), and cytotoxic response to the TMZ+SN38 combination (Fig. 1H, which graphs the IC₅₀ values and Fig. S2, Supplemental Digital Content, which plots the cytotoxicity curves) in 14 progressive disease neuroblastoma cell lines representing a range of MGMT expression from non-expression to high expression. MGMT protein expression positively correlated with MGMT mRNA expression (Spearman’s r= 0.92, P < 0.05, Fig. 1I), and TMZ+SN38 IC₅₀ positively correlated with both MGMT mRNA and MGMT protein expression (Spearman’s r = 0.73, P < 0.05 and Spearman’s r = 0.59, P < 0.05, respectively, Fig. 1J and 1K). CHLA-70 (open circle) was censored from IC₅₀ correlations by GraphPad Prism’s automatic outlier elimination feature.

MGMT expression and response to TMZ+IRN correlated in vivo.

We evaluated the correlation of cell line MGMT mRNA expression against in vivo CDX response to temozolomide (TMZ) + irinotecan (IRN). Progressive disease cell lines CHLA-136, CHLA-171, CHLA-119, and CHLA-79, previously reported as multidrug resistant [10,38], were subcutaneously xenografted in NU/NU mice. Each received three 5-day cycles (initiated at T = 0, 21, and 42 days, gray arrows in Fig. 2A and 2B) of TMZ+IRN (25 mg/kg TMZ by oral gavage, 7.5 mg/kg IRN by tail vein). MGMT mRNA expression of the same cell lines from in vitro cultures (Fig. 2C) showed an inverse correlation with the median EFS of xenografted mice (Spearman’s r = −1.0, P < 0.05, Fig. 2D).

Figure 2. MGMT expression and TMZ+IRN response correlated in vivo.

A) Tumor growth curves and B) corresponding event-free survival of scapular subcutaneous cell line xenografts in NU/NU mice. TMZ+IRN treated mice are shown in black and controls are shown in gray. Each mouse received three 21-day cycles (gray arrows) of TMZ+IRN (TMZ at 25 mg/kg by oral gavage and IRN at 7.5 mg/kg by tail vein injection, given on days 1–5 of the cycle). C) In vitro relative MGMT mRNA expression (mean of duplicates with SEM, GAPDH-normalized). D) There was an inverse trend between MGMT mRNA expression and survival following treatment with TMZ+IRN (P < 0.05, Spearman’s r = −1.0), suggesting that tumors with low MGMT may respond well to TMZ+IRN based regimens.

Single agent and combination TMZ and IRN treatment were carried out for the CHLA-79 and CHLA-171 xenografts (Fig. S3, Supplemental Digital Content, which plots tumor growth and survival for controls, TMZ, IRN, and TMZ+IRN). Interestingly, for CHLA-79, which had very low MGMT mRNA expression, the xenograft response to TMZ+IRN treatment appeared to be driven by TMZ, while for CHLA-171, which displayed high MGMT mRNA expression, the TMZ+IRN response appeared to be dominated by IRN.

Induced over-expression of MGMT in MGMT non-expressing cell lines increased resistance to TMZ+SN38 in vitro.

CHLA-225 and COG-N-269 are neuroblastoma cell lines established from progressive disease with no detectable MGMT protein expression. Lentiviral overexpression of MGMT in CHLA-225 and COG-N-269 (Fig. 3A) proved protective against TMZ+SN38-induced cytotoxicity, resulting in significantly higher IC₅₀ values relative to empty vector-transduced controls (Fig. 3B, P < 0.05, a 12-fold and a 5-fold increase in IC₅₀, respectively). Representative cytotoxicity curves are shown in Fig. 3C.

Figure 3. Forced expression of MGMT increased in vitro resistance to TMZ+SN38.

A) Western blot of MGMT showing two MGMT-nonexpressing cell lines (CHLA-225 and COG-N-269) transduced with MGMT overexpression plasmid or empty vector (EV) plasmid. B) TMZ+SN38 IC₅₀ for MGMT overexpression models and EV models. Asterisks(*) indicate P < 0.05 by Student’s t-test (mean of triplicates with SEM). For each MGMT overexpression model, TMZ+SN38 IC₅₀ was compared to each of the two corresponding empty vector models. TMZ and SN38 were dosed as a fixed ratio of 100 μM TMZ : 5 nM SN38 with arbitrary units (AU) set equal to TMZ concentration for graphing purposes. C) Representative TMZ+SN38 cytotoxicity dose-response curves for MGMT overexpression and EV models.

MGMT inhibitor O6BG sensitized MGMT expressing neuroblastoma cell lines to TMZ+SN38 in vitro.

To evaluate the potential for inhibition of MGMT to enhance TMZ+IRN, a commonly used salvage chemotherapy regimen for neuroblastoma, TMZ+SN38 +/− O⁶-benzylguanine (O6BG) was evaluated in bone marrow level hypoxia (5% O₂) for 10 neuroblastoma cell lines established from 9 progressive disease patients at time of death (COG-N-452 and COG-N-453 are from the same patient). All cell lines expressed moderate to high levels of MGMT in vitro (Fig. 4a), and all showed in vitro sensitization of clinically achievable TMZ+SN38 dosing following 24-hour pretreatment with 25 μM O6BG (Fig. 4B). TMZ+SN38 IC₅₀ values were reduced by 2.6-fold to 27-fold (see Table S7, Supplemental Digital Content, which shows TMZ+SN38 IC₅₀ with and without O6BG), and mean TMZ+SN38 IC₅₀ across the models was significantly lower when combined with O6BG (P < 0.05, Fig. 4C). Notably, O6BG induced multi-fold reductions in TMZ+SN38 IC₅₀ in both p53 functional (Felix-h, COG-N-440h, −452h, −453h, −470h, and −561h) and p53 non-functional models (COG-N-421h, −471hnb, and −519h).

Figure 4. MGMT inhibitor O6BG chemosensitized MGMT-expressing post-mortem (PM) neuroblastoma cell lines to TMZ+SN38 in vitro.

A) Basal MGMT in 10 neuroblastoma cell lines established post-mortem from 9 patients (COG-N-452h and COG-N-453h are isogenic) for which corresponding patient-derived xenograft (PDX) models exist. Cell lines were established and cultured in IMDM+FBS at 5% O₂, except for COG-N-471hnb which was established in neurobasal medium and acclimated to IMDM+FBS. B) 24 hour pre-treatment with O6BG caused in vitro chemosensitization of MGMT-expressing PM cell lines to clinically achievable TMZ+SN38 dosing. The TMZ+SN38 cytotoxicity curves are shown in white, and the TMZ+SN38+O6BG curves are shown in black. TMZ and SN38 were dosed as a fixed ratio of 100 μM TMZ : 5 nM SN38. C) Plot of TMZ+SN38 IC₅₀ values without (left) and with (right) O6BG pretreatment, corresponding to the curves shown in B. Addition of O6BG significantly reduced the mean TMZ+SN38 doses needed to reach IC₅₀ (P < 0.05, Student’s t-test). Arbitrary units (AU) were set equal to TMZ concentration for graphing purposes. D) In vitro MGMT depletion shown 24 hours following dosing of 25 μM O6BG.

The TMZ+SN38 and TMZ+SN38+O6BG cytotoxicity curves converged at higher doses for all cell lines except COG-N-440h and Felix-h (Fig. 4B). Treatment with 25 μM O6BG for 24 hours substantially, but not completely, depleted MGMT protein (Fig. 4D). Combining O6BG with TMZ alone or SN38 alone showed that O6BG chemosensitizes through TMZ and not SN38 (Fig. S4, Supplemental Digital Content, which shows in vitro responses to TMZ, SN38, and TMZ+SN38 with and without O6BG).

As a single agent, O6BG induced only limited cytotoxicity in sixteen progressive disease cell lines that were treated with O6BG from 6.25 to 50 μM. The maximum clinically achievable concentration of 25 μM O6BG reduced cell counts on average by 10%, while only four cell lines showed >20% loss of cells at 25 μM O6BG (Fig. S5, Supplemental Digital Content).

We also evaluated chemosensitization by the second-generation MGMT inhibitor lomeguatrib. We assessed 8 of the neuroblastoma cell lines established from patients with progressive disease at time of death and found that lomeguatrib or O6BG, in combination with TMZ+SN38, yielded nearly identical cytotoxicity profiles (Fig. S6, Supplemental Digital Content).

MGMT depletion with O6BG increased DNA double-strand breaks, apoptosis, and DNA fragmentation following in vitro TMZ+SN38.

To investigate the effects of O6BG on DNA damage and apoptosis, we selected two cell lines, COG-N-452h and COG-N-561h, which express MGMT and show in vitro chemosensitization to TMZ+SN38 with O6BG and two cell lines, CHLA-225 and COG-N-269, which express no MGMT and showed no chemosensitization (Fig. 5A and 5B). We analyzed the following: 1) Phospho-histone H2AX (a marker for DNA double-strand breaks) [42]. 2) Cleaved caspase-3 (caspase-3 is activated by cleavage and is an executioner caspase common to the intrinsic and extrinsic apoptotic pathways) [43]. 3) DNA fragmentation (characteristic of end-stage apoptosis and quantifiable by TUNEL) [36]. We found that phospho-H2AX and cleaved caspase-3 levels (Fig. 5B), as well as TUNEL positivity (Fig. 5C) were increased in MGMT-expressing (but not in MGMT non-expressing) cell lines treated with O6BG in combination with TMZ+SN38, relative to TMZ+SN38 alone.

Figure 5. MGMT depletion following the addition of O6BG to TMZ+SN38 in vitro increased DNA damage (p-H2AX) and apoptosis (cleaved caspase-3 and TUNEL).

A) In vitro TMZ+SN38 +/− O6BG dose-response curves for two MGMT-expressing models (COG-N-452h and COG-N-561h, as seen in Figure 4) are shown alongside two MGMT-nonexpressing models which show no chemosensitization with the addition of O6BG (CHLA-225 and COG-N-269). B) Cells cultured in IMDM+FBS at 5% O₂ were drugged with 20 μM TMZ + 1 nM SN38 +/− 24 hour pretreatment with 25 μM O6BG. Cells were collected 72 hours following TMZ+SN38 dosing. Increased p-H2AX and increased cleaved caspase-3 are seen in the MGMT-expressing cell lines but not in MGMT-nonexpressing cell lines. C) Cells cultured in IMDM+FBS at 5% O₂ were drugged with 20 μM TMZ + 1 nM SN38 +/− 24 hour pretreatment with 25 μM O6BG. Cells were collected 120 hours following TMZ+SN38 dosing. The addition of O6BG to TMZ+SN38 in vitro increased apoptosis (DNA fragmentation via TUNEL) in MGMT-expressing cell lines (P < 0.05, Tukey) but not in MGMT-nonexpressing cell lines.

Addition of O6BG to TMZ+IRN improved EFS in two of seven post-mortem neuroblastoma patient-derived xenografts (PDX).

We evaluated the combination of O6BG with TMZ+IRN in seven PDX models (Fig. 6, Table 1). All models were established at time of death after failure of multi-agent chemotherapy, and all models expressed MGMT (Fig. 6B). Mice were given two cycles of TMZ+IRN +/− O6BG except for Felix-PDX, which was given one cycle. Two models, COG-N-452x and COG-N-564x, showed improvement in EFS with the O6BG combination therapy (Fig. 6A, blue vs red lines, P < 0.05, log-rank test, Table 1). For COG-N-564x, there was improvement in the objective responses, with all mice treated with the combination of O6BG and TMZ+IRN (blue lines) achieving a complete response, while mice treated with TMZ+IRN alone (red lines) obtained partial responses at best (Table 1). The other five models saw no significant improvement with the TMZ+IRN+O6BG combination over treatment with TMZ+IRN. In the PDX models analyzed here, there was no apparent relationship between higher MGMT expression and greater O6BG responsiveness. Consistent with our in vitro activity data and previously published data [38], O6BG, evaluated as a single agent in four of the seven models, showed no differences in EFS compared to controls.

Figure 6. TMZ+IRN +/− O6BG in vivo.

A) Subcutaneously xenografted NU/NU mice were given one or two 5-day cycles (gray arrows) of TMZ+IRN +/− O6BG. Xenograft models were post-mortem PDXs. O6BG was given 30 mg/kg intraperitoneal, followed one hour later by 25 mg/kg TMZ by oral gavage, followed one hour later by 7.5 mg/kg IRN by tail vein. The O6BG-responsive models are only moderate MGMT expressers, casting doubt on high MGMT expression level as an effective biomarker for O6BG responsiveness. B) Western blot of basal MGMT for the PDX models. C) Addition of O6BG to TMZ+IRN depleted MGMT and increased cleaved caspase-3 in the O6BG-responsive COG-N-452x xenograft model, indicating increased apoptosis due to MGMT depletion. Tumors were collected one hour after the completion of cycle 1 dosing.

Table 1.

Summary of PDX treatment outcomes by model.

Groups	Cycles Given	N	PD (%)	PR (%)	CR (%)	MCR (%)	Adverse Event	Median EFS (days)	EFS T/C
Felix-PDX
Control	1	5	5 (100)	-	-	-	-	19	1.00
TMZ+IRN	1	5	-	3 (60)	2 (40)	-	-	41	2.16
TMZ+IRN+O6BG	1	5	-	5 (100)	-	-	-	41	2.16
COG-N-421x
Control	2	4	4 (100)	-	-	-	-	24.5	1.00
O6BG	2	4	4 (100)	-	-	-	-	19	0.78
TMZ+IRN	2	4	-	-	4 (100)	-	-	115.5	4.71
TMZ+IRN+O6BG	2	4	-	-	4 (100)	1 (25)	-	113	4.61
COG-N-452x
Control	2	5	5 (100)	-	-	-	-	15	1.00
O6BG	2	5	5 (100)	-	-	-	-	17	1.13
TMZ+IRN	2	5	-	-	5 (100)	-	-	87	5.80
TMZ+IRN+O6BG	2	5	-	-	5 (100)	1 (20)	-	147b	9.80
COG-N-470x
Control	2	3	3 (100)	-	-	-	-	20	1.00
O6BG	2	3	3 (100)	-	-	-	-	33	1.65
TMZ+IRN	2	4	-	-	4 (100)	1 (25)	-	134.5	6.73
TMZ+IRN+O6BG	2	4	-	-	4 (100)	1 (25)	2	146.5	7.33
COG-N-519x
Control	2	3	3 (100)	-	-	-	-	9	1.00
TMZ+IRN	2	3	-	1 (33)	2 (67)	-	-	86	9.56
TMZ+IRN+O6BG	2	4	-	-	4 (100)	1 (25)	-	74	8.22
COG-N-561x
Control	2	5	5 (100)	-	-	-	-	22	1.00
O6BG	2	5	5 (100)	-	-	-	-	20	0.91
TMZ+IRN	2	5	-	-	5 (100)	2 (40)	1	>200	>9.09
TMZ+IRN+O6BG	2	5	-	-	5 (100)	2 (40)	1	>200	>9.09
COG-N-564x
Control	2	3	3 (100)	-	-	-	-	12	1.00
TMZ+IRN	2	3	1 (33)	2 (67)	-	-	-	42	3.50
TMZ+IRN+O6BG	2	4	-	-	4 (100)	-	-	96b	8.00
Total (2 cycles)
Control	2	23	23 (100)	-	-	-	-	18	1.00
O6BG	2	17	17 (100)	-	-	-	-	18	0.9a
TMZ+IRN	2	24	1 (4)	3 (13)	20 (83)	3 (13)	1	110	6.11
TMZ+IRN+O6BG	2	26	-	-	26 (100)	6 (23)	3	143.5	7.97

^a:Median EFS of aggregated controls for models with corresponding O6BG cohorts was 20 days.

^b:Significant increase in EFS (P < 0.05 by log-rank test) with TMZ+IRN+O6BG relative to TMZ+IRN.

PD: Progressive disease, treatment failed to reduce tumor volume by 50%. PR: Partial response, treatment reduced tumor volume by more than 50%, without complete clearance. CR: Complete response, no detectable lesion. MCR: Maintained CR, no evidence of tumor at 200 days. EFS: Event-free Survival. EFS T/C: Ratio of treatment median EFS to control median EFS. Adverse Event: Mouse sacrificed due to event other than tumor progression to 1500 mm³.

In the O6BG-responsive PDX COG-N-452x, TMZ+IRN+O6BG treatment resulted in tumor MGMT depletion and increased cleaved caspase-3, relative to TMZ+IRN treatment (Fig. 6C).

The combination of O6BG with TMZ+IRN was marginally more toxic than TMZ+IRN in NU/NU mice.

Using weight loss as an indicator of toxicity, the combination of O6BG with TMZ+IRN marginally increased toxicity within acceptable levels (20% weight loss) as defined by the TTUHSC IACUC (See Fig. S7, Supplemental Digital Content). Of the 24 mice receiving two cycles of TMZ+IRN, two showed weight loss greater than 10% during either cycle 1 or cycle 2, but none had weight loss greater than 15%. Of the 26 mice receiving two cycles of TMZ+IRN+O6BG, 8 lost over 10% body weight, of which 4 lost more than 15%. Thus, TMZ+IRN+O6BG in these studies was employed at the maximally tolerated dose for NU/NU mice.

Discussion

In this study, we demonstrated that higher MGMT expression correlated with increased resistance to the re-induction chemotherapy regimen TMZ+IRN (TMZ+SN38) in neuroblastoma cell lines in vitro and in cell line xenografts. Forced overexpression of MGMT increased TMZ+SN38 resistance, while addition of the MGMT inhibitor O6BG enhanced the cytotoxicity of TMZ+SN38 for neuroblastoma cell lines cultured in bone marrow-level hypoxia, a common site of neuroblastoma progressive disease. However, in testing the ability of O6BG to reverse resistance to TMZ+IRN in neuroblastoma PDXs established from seven patients at the time of death from disease progression, only two PDXs showed chemosensitization to TMZ+IRN with O6BG.

Both the in vitro TMZ+SN38 and in vivo TMZ+IRN response data suggest that neuroblastomas with low MGMT expression were more sensitive to TMZ+IRN than neuroblastomas expressing high MGMT. Thus, MGMT expression does provide a potential biomarker of resistance to TMZ+IRN in relapse neuroblastoma. Most studies evaluating MGMT for the purpose of patient-subset designation have assessed promoter methylation to infer gene silencing or assessed protein via immunohistochemistry [26], however it has been reported that assessment of MGMT protein by immunohistochemistry in glioblastoma gives highly discordant results when cross-referenced against promoter methylation assays and mRNA expression by qRT-PCR [44]. As our data suggest that basal MGMT mRNA provided a good surrogate for basal MGMT protein levels, and as RNA sequencing is being obtained on primary tumors for many neuroblastoma patients, the relationship of MGMT mRNA expression to response and outcome should be examined in future clinical trials, though such analyses may be limited by failing to assess the induction of expression by exposure to cytotoxic agents [45].

For this work, we chose an extended-incubation dosing scheme to evaluate TMZ+SN38+/−O6BG responses in vitro. Previously the TMZ+SN38+/−O6BG combination was evaluated with fixed ratio TMZ+SN38 in a serial dilution curve with high doses of 250 μM TMZ and 50 nM SN38 for 72 hours [38]. In that study, the in vitro cytotoxicity data showed TMZ+SN38+/−O6BG curves dominated by SN38 with little-to-no effect of adding TMZ and/or O6BG. According to the futile cycling hypothesis, DNA replication is necessary to convert the ^meG-C pairs into the ^meG-T mispairs at the core of a futile cycling complex [15]. The doubling time for many neuroblastoma cell lines is > 72 hours (see Table S1, Supplemental Digital Content), limiting the cycling of those lines when using a 3-day in vitro therapeutic exposure. For our studies we incubated TMZ+SN38+/−O6BG for 8 days, enabling the longer duration drug exposure to better mimic tumor exposures to TMZ+SN38 in vivo and to allow more time for DNA replication to occur.

A feature common to the TMZ+SN38+/−O6BG cytotoxicity curves for most in vitro models as dosed here is that, despite causing a several-fold reduction of IC₅₀, the addition of O6BG adds no greater depth of cell kill at the maximum dosing. Thus, as TMZ+SN38 doses increase, the depletion of MGMT with O6BG offers diminishing returns in cytotoxicity gains over TMZ+SN38 alone. The presumed cytotoxic species resulting from TMZ are the foci of futile MMR cycling allowed by ^meG–T mispairs. While methylation burden delivered to the genome could continue increasing proportionally to TMZ exposures far beyond the point of clinical tolerability, the finite MMR machinery should be saturable. Once all MMR machinery is engaged, additional ^meG–T mispairs (and hence additional methylations) cannot lead to additional MMR futile cycling complexes. Saturability may be a reason for the low concordance in O6BG-responsiveness between cell lines in vitro and the matching PDX models, but currently the reasons for the observed differences in in vitro vs in vivo O6BG activity are not understood.

The generation of futile cycling complexes through MMR is a mechanism that may not be operative in nonresponsive cancers. It was observed that MMR-deficient neuroblastoma cell line xenografts showed greater temozolomide resistance relative to MMR-proficient models [46]. Further, studies using medulloblastoma and colorectal cancer in vitro models have found that loss of MMR proficiency abolishes O6BG responsiveness in addition to TMZ responsiveness [47,48].

Repletion of MGMT following clearance of O6BG may also contribute to the low rate of O6BG-responsiveness seen in the PDXs [16]. Futile MMR cycling complexes depend on O⁶-methylguanine for the cytotoxic cycles of excision and subsequent flawed resynthesis [16]. It has been noted in MMR-competent glioblastoma SF767 cells that repletion of MGMT protein can reverse G2/M arrest up to 5 days following TMZ exposure, and cells are able to escape cytotoxicity if repletion occurs within 24 hours following the onset of G2/M arrest [49]. O6BG activity may be increased with higher doses of the agent; however, this is not likely to be feasible in PDX models as several TMZ+IRN+O6BG treated mice were already near the limit of acceptable weight loss. It is also possible that enhancing DNA damage alone is insufficient to overcome multiple resistance mechanisms likely to have developed in some neuroblastomas due to multiple exposure/selection events over the course of therapy.

Previous clinical trials using O6BG to sensitize cancers to temozolomide or carmustine have found that reduced doses of alkylating agent were required to avoid toxicities, and responses were not improved by the combination, thus the therapeutic window was not improved [27,50]. However the majority of clinical trial patients have had glioblastoma or other CNS tumors, and a trial of an MGMT inhibition strategy has not been conducted for neuroblastoma. To our knowledge MGMT inhibition strategies also have not been explored in a myeloablative setting, where myelosuppressive toxicity contributed by O6BG may be inconsequential and higher O6BG doses may be attainable. However the diminishing in vitro sensitization by O6BG at higher TMZ+SN38 concentrations would caution that MGMT inhibition may not increase the anticancer activities of TMZ or other drugs at high myeloablative doses.

After finding higher MGMT expression in models with greater TMZ and TMZ+SN38 resistance, we hypothesized that the highest MGMT expression may accompany greater O6BG responsiveness, possibly indicative of an expanded therapeutic window for that subset of neuroblastomas. However we found no indications of such a pattern. Our data suggest that a subset of neuroblastomas may show enhanced response to TMZ+IRN if combined with O6BG, but it is unclear how to identify patients whose tumors may be responsive to O6BG. While the in vitro experiments reliably resulted in multi-fold reductions in TMZ+SN38 IC50 with pretreatment of O6BG, in vivo benefit was seen for 2 of 7 post-mortem PDX models. It does not appear that high basal MGMT expression is required for O6BG-responsiveness since the responsive PDXs COG-N-452x and COG-N-564x showed only moderate basal MGMT expression. However, basal MGMT protein and mRNA expression may not reflect MGMT levels after exposure to genotoxic stresses. Functionality of p53 was neither required for O6BG responsiveness in vitro nor predictive of responsiveness in vivo.

In summary, higher MGMT expression is associated with resistance to alkylating agent-based therapy in neuroblastoma cell lines and xenografts. The cytotoxic response to TMZ+SN38 in vitro can be readily enhanced by O6BG, but only a subset of neuroblastoma PDX models show enhanced response to TMZ+IRN with O6BG. A more complete understanding of all mechanisms of resistance to DNA damaging agents in neuroblastoma may in the future enable developing biomarkers that can identify tumors responsive to O6BG combined with TMZ+IRN.