Targeting O⁶-methylguanine-DNA methyltransferase deficiency in preclinical models of extracranial human cancers with a tumor-selective DNA modifying agent

Abstract

Temozolomide (TMZ) is an established therapy for gliomas with silencing of the DNA repair protein O⁶-methylguanine-DNA methyltransferase (MGMT) but is hindered by the frequent development of resistance via loss of mismatch repair (MMR) proteins. To overcome this resistance, a novel 2-fluoroethylating agent, KL-50, was designed to generate toxic DNA interstrand crosslinks specifically in MGMT-deficient cells, bypassing MMR. KL-50 has shown efficacy in preclinical MGMT-silenced, TMZ-resistant glioma models. MGMT loss also occurs in a wide range of non-glioma cancers, but the ability of KL-50 to effectively treat these tumors with minimal off-target toxicity remains untested. Here, we utilize a large-scale cell line screen, focused panels of cancer cell lines, and tumor xenograft mouse models, including patient-derived xenografts, to interrogate the therapeutic potential of KL-50 in treating intracranial and extracranial tumors. We find that KL-50 is highly efficacious in a wide range of MGMT-deficient human tumor models, including colon cancer, melanoma, and lung cancer in vivo models. In addition, we demonstrate that KL-50 is impervious to loss of MMR and remains effective in tumors with induced TMZ resistance. Finally, we determine that KL-50 possesses higher MGMT selectivity in vitro compared to chloroethylating agents and is associated with less in vivo systemic toxicity in preclinical models. These results establish KL-50 as the first agent to maintain strong selectivity for MGMT loss while inducing tumor toxicity by an MMR-independent mechanism and support further development of 2-fluoroethylating agents for use in a tumor type-agnostic, biomarker-based strategy targeting MGMT-deficient tumors.

Introduction

DNA repair defects are prevalent in human cancer and can be exploited therapeutically by agents that induce DNA damage dependent upon the loss of a specific DNA repair pathway or function [1]. This approach of generating DNA damage specifically in DNA repair-deficient tumors has been validated clinically, as exemplified by the use of temozolomide (TMZ, 1a; the following compounds 1a–4 are shown in Fig. 1A) in O⁶-methylguanine-DNA methyltransferase (MGMT)-silenced tumors or PARP inhibitors in homologous recombination (HR)-deficient tumor [2,3]. However, as with nearly all targeted agents, resistance mechanisms limit the effectiveness of compounds that exploit DNA repair defects [4,5]. While reversion of the sensitizing genetic defect can sometimes drive resistance, alterations in other DNA repair proteins often instead underlie resistance mechanisms, for example loss of mismatch repair (MMR) proteins leading to TMZ (1a) resistance or loss of 53BP1 or PARG conferring PARP inhibitor resistance [6-8]. Thus, there is an opportunity to identify novel agents that are less likely to succumb to acquired resistance.

Fig. 1. MGMT expression is a strong predictor of KL-50 (2a) activity in vitro across solid tumor types.

(A) Proposed mechanisms of action of DNA methylating, fluoroethylating, and chloroethylating imidazotetrazine compounds. Scheme 1: The imidazotetrazine analogs TMZ (1a), KL-50 (2a), and MTZ (3a) hydrolyze under aqueous conditions to form the triazene compounds 1b, 2b, or 3b, which generate the diazonium species 1c, 2c, or 3c, which alkylate DNA generating O⁶MeG (1d), O⁶FEtG (2d), or O⁶ClEtG (3d), respectively, among other lesions. Scheme 2: O⁶MeG (1d) is rapidly reversed by MGMT in MGMT+ cells. In MGMT− cells, the lesion persists and mispairs with thymidine during replication. The mismatch is recognized by MMR proteins, leading to futile cycles of excision and resynthesis of the mismatch, generating replication stress and DNA DSBs which lead to cytotoxicity. In the absence of MMR, the mismatch is tolerated, and cells display TMZ (1a) resistance. Scheme 3: O⁶FEtG (2d) is also rapidly repaired by MGMT with an estimated half-life of 2.9 hours based on known rate of repair of O⁶-ethylguanine [27]. In MGMT− cells, O⁶FEtG (2d) persists and slowly converts to a G-C ICL through the N1,O⁶EtG (4) intermediate, with a half-life of ~18 h [26]. The ICLs lead to cytotoxicity in MGMT− cells, independent of MMR. Scheme 4: O⁶ClEtG (2d) is thought to be repaired by MGMT with an estimated half-life of 2.9 hours similar to O⁶-ethylguanine [27]. More rapidly, O⁶ClEtG (3d) converts to N1,O⁶EtG (4) with a half-life of ~18 min, and thus N1,O⁶EtG (4) forms in both MGMT+ and MGMT− cells. N1,O⁶EtG (4) can yield G–C ICLs or, in the presence of MGMT, MGMT-DNA protein crosslinks, which together lead to cytotoxicity independent of MGMT. (B) MGMT expression across cancer types in TCGA PanCancer Atlas. Left, scatter dot plots of MGMT mRNA expression modified Z-scores for individual tumor samples, with values <1 SD below the median demarcating “MGMT−” samples. Right, percentage of tumors that are MGMT− (<1 SD below the median). (C) Correlation analysis between MGMT expression and drug responses in the multiplexed cell line PRISM screen. Points represent the correlation between MGMT mRNA expression levels and cell viability with KL-50 (2a), MTZ (3a), TMZ (1a), or CCNU at individual dose levels or aggregate dose response as measured by AUC (area under curve) or IC₅₀. (D) Viability response to 22 μM KL-50 (2a) or 22 μM TMZ (1a) in the PRISM screen across cell lines categorized by MGMT and MMR status. MGMT+/− status was determined by MGMT mRNA expression using <1 SD below the median as a cut point. MMR+/− status was determined by occurrence of damaging mutations in MSH2, MSH6, MLH1, or PMS2. Points indicate log₂ fold-change in viability of individual cell lines; lines indicate mean; error bars indicate SD; significance determined by ordinary one-way ANOVA with Tukey’s multiple comparisons test. (E) Log₂ fold-change in viability to 22 μM KL-50 (2a) in the PRISM screen across solid tumor types. Cell lines were categorized as MGMT+/− by MGMT mRNA expression using <1 SD below the median as a cut point. Tumor types represented by 2 or more MGMT− cell lines were included. Boxes indicate 25^th to 75^th percentiles with line at median; whiskers indicate min to max; significance determined by unpaired two-tailed t test.

TMZ (1a) is a monofunctional alkylating agent which undergoes hydrolytic cleavage to MTIC (1b) and decomposition to methyldiazonium (1c), the active species which deposits methyl groups at various DNA base sites [9]. The most abundant methylated bases, N7-methyl-guanine and N3-methyl-adenine, accounting for ~70% and ~9% of lesions respectively, are readily corrected by base excision repair enzymes [9,10]. In contrast O⁶-methyl-guanine (O⁶MeG, 1d), an endogenous substrate of MGMT, accounts for ~5% of alkylated sites, but is an important source of cytotoxicity in biological systems lacking MGMT. In the absence of MGMT, O⁶MeG (1d) mispairs with thymidine (T) during replication, leading to processing of the lesion by the MMR pathway. The mismatch is recognized by the MSH2-MSH6 scanning complex, the surrounding bases are degraded upon recruitment of the MLH1-PMS2 endonuclease complex, and the gap is filled by polymerases resulting in resynthesis of the O⁶MeG–T mismatch. This “futile cycling” of mismatch excision and resynthesis is believed to lead to replication stress, DNA double-strand break (DSB) formation, and apoptosis [11]. Loss of MMR activity confers resistance to TMZ (1a) in cell culture and is a common mechanism of clinical resistance identified in recurrent gliomas [6,12-16].

We previously reported the discovery of a novel class of compounds that can overcome TMZ (1a) resistance mediated by MMR loss in glioma, while still exploiting tumor intrinsic MGMT deficiency [17]. The first-generation compound, KL-50 (2a), is derived from TMZ (1a) by substitution of the methyl group at the N3 position of the imidazotetrazine core with a 2-fluoroethyl group. In a manner analogous to TMZ (1a), KL-50 (2a) decomposes to a 2-fluoroethyl diazonium ion (2c), which deposits 2-fluoroethyl lesions on DNA (Fig. 1A). The O⁶-fluoroethylguanine (O⁶FEtG, 2d) lesion is a substrate for direct reversal by MGMT. However, in the absence of MGMT, the initial O⁶FEtG (2d) slowly evolves to a G-C DNA interstrand crosslink (ICL) through an N1,O⁶-ethanoguanine intermediate (N1,O⁶EtG, 4), with a half-life of ~80 hours in vitro [18]. Because the formation of a DNA ICL bypasses the requirement for MMR-mediated futile cycling to induce cell death, KL-50 (2a) confers MGMT-selective, but MMR-independent, cytotoxicity.

Consistent with the proposed mechanism of 2-fluoroethylating agents, we previously demonstrated in vitro and in vivo that KL-50 (2a) is effective in MGMT-deficient glioma models regardless of their MMR status, including those that are TMZ (1a)-resistant. Supporting a mechanism involving slow DNA-crosslinking, KL-50 (2a) induces delayed formation of DNA ICLs and markers of replication stress and subsequent DNA double-strand break formation, ca. 24–48 hours post-treatment in cells. Preliminary studies showed an acceptable toxicity profile in mice, with limited myelosuppression at supratherapeutic doses. In addition, KL-50 (2a) was expected to be CNS-penetrant given its structural similarity to TMZ (1a) and was effective in treating TMZ (1a)-resistant intracranial orthotopic cell-derived xenograft (CDX) glioma models [17].

Chloroethylating agents, such as lomustine (CCNU), have been in clinical use for gliomas for several decades but are known for dose-limiting side effects, most notably hematologic toxicity [19]. Mitozolomide (MTZ, 3a), the exact 2-chloroethyl-substituted derivative of TMZ (1a), was also tested in early human trials, but failed due to severe myelosuppression [20-24]. At the molecular level, the increased leaving group ability of chlorine (relative to fluorine) present in the O⁶-chloroethylguanine (O⁶ClEtG, 3d) lesions generated by MTZ (3a) or CCNU accelerates the cyclization to N1,O⁶EtG (4) (half-lives of 3d and 2d are 18 min and 18 h, respectively) [25,26]. These reactions are in competition with MGMT-mediated repair of O⁶FEtG (2d) or O⁶ClEtG (3d). While the rates of repair of these lesions have not been directly determined, O⁶-ethylguanine undergoes repair by MGMT with a half-life of 2.9 h, and MGMT reversal of O⁶FEtG (2d) or O⁶ClEtG (3d) is expected to be slower as β halogen substituents are known to decrease the rate of reversal [18,27]. Accordingly, because cyclization of O⁶ClEtG (3d) to N1,O⁶EtG (4) is competitive with MGMT reversal, chloroethylating agents form DNA ICLs in MGMT+ cells, and their selectivity for MGMT-deficient tumors is diminished. Additionally, MGMT can act on the N1,O⁶EtG (4) intermediate leading to MGMT-DNA protein crosslinks (Fig. 1A) [18]. Altogether, these data suggest that fluoroethylating agents will possess greater MGMT dependency than chloroethylating agents and will be better tolerated due to reduced toxicity in MGMT-expressing normal human cells.

While MGMT promoter silencing is best studied in gliomas, a broad range of additional cancer types display significant rates of MGMT-silencing, highlighting the potential for use of MGMT-targeted therapies in a wide array of malignancies [28,29]. For example, MGMT promoter silencing has been reported in ~20–25% of malignant melanomas and ~30–40% of colon carcinomas [30-34]. TMZ (1a) has been studied in these malignancies, historically being of use for treating brain metastases due to its CNS penetration. In melanoma, TMZ (1a) remains a recognized option for patients with advanced disease who are not candidates for immunotherapy or targeted inhibitors [35]. In colon cancer, numerous phase II trials have been conducted with TMZ (1a) in advanced disease, alone or in combination therapies [36]. In these settings, TMZ (1a) response rates and durations are limited, likely due to intrinsic or acquired resistance, and emerging data in colon cancer suggests that MMR mutations can arise in response to TMZ (1a) treatment [37]. Thus, an agent that is highly selective for MGMT-deficient tumors and which could overcome TMZ (1a) resistance would be of significant therapeutic benefit in these malignancies. Here, we present evidence supporting the expanded therapeutic potential of 2-fluoroethylating agents in both gliomas and non-glioma malignancies.

Materials and Methods

Cell culture.

Human cancer cell lines A101D (RRID:CVCL_1057), CACO2 (RRID:CVCL_0025), SKCO1 (RRID:CVCL_0626), SW48 (RRID:CVCL_1724), and SW480 (RRID:CVCL_0546) were obtained from ATCC; KM12 (KM12-Luc, RRID:CVCL_J258) were obtained from JCRB; DLD1 (DLD1 BRCA2+/−, RRID:CVCL_HD56) were obtained from Horizon Discovery; and SW620 (RRID:CVCL_0547) were a gift from P. Glazer. DLD1 cells were maintained in RPMI (Thermo Fisher) with 10% FBS (Sigma Aldrich). All other cell lines were maintained in DMEM (Thermo Fisher) with 10% FBS. All established cell lines were tested and confirmed negative for mycoplasma via the MycoAlert Mycoplasma Detection Kit (Lonza). Established cell lines were authenticated by short tandem repeat profiling and confirmed to be >90% match to database profiles using CLASTR (RRID:SCR_024863). Cell lines were obtained between the years of 2021 and 2023. Patient-derived melanoma cells YUHEF, YUDOSO, YUMUT, YUSEEP, YUNIGE, and YUROL were obtained from the Specimen Resource Core of the Yale SPORE in Skin Cancer in 2022 and maintained in DMEM + 10% FBS. Melanoma cells were tested and confirmed negative for mycoplasma via qPCR. Melanoma tumor samples were collected with written informed patient consent and were approved by the Yale University IRB (HIC #0609001869) in accordance with the Declaration of Helsinki.

Chemicals.

KL-50 (2a) was synthesized as previously described [17] at Yale University or at WuXi Apptec. Temozolomide (TMZ, 1a), lomustine (CCNU), Val-083 [38], talazoparib, and olaparib were purchased from Selleck Chemicals. Mitozolomide (MTZ, 3a) was purchased from Enamine. Mitomycin C (MMC) was purchased from Sigma. Cisplatin was purchased from Cayman Chemicals. The compounds were dissolved and stored in DMSO at the following stock concentrations: 200 mM KL-50 (2a), 150 mM TMZ (1a), 100 mM CCNU, 200 mM Val-083, 10 mM talazoparib, 200 mM olaparib, 200 mM MTZ (3a), and 10 mM MMC. Cisplatin was dissolved in H₂O at 4 mM stock concentration. All compounds were stored at −20 °C or −80 °C.

TCGA database analysis.

MGMT mRNA expression values (RSEM, Batch normalized from Illumina HiSeq_RNASeqV2) available for 10,071 samples representing 29 cancer types from the TCGA PanCancer Atlas were downloaded from cBioPortal (RRID:SCR_014555) and used to calculate modified Z-scores of log₂(RSEM+1). The percentage of samples for each cancer type which were MGMT− was determined using a cutoff of <1 SD below the median. MGMT promoter methylation data was obtained from cBioPortal and analyzed according to the model developed by Bady et al. [39]. Briefly, β values for two methylation probes associated with gene silencing and outcomes in glioma, cg12981137 and cg12434587, derived from merged HM27 and HM450 methylome data, were downloaded for 10,013 samples from the TCGA PanCancer Atlas. β values were converted to M-values according to the equation M = log₂(β/(1−β)) [40], and methylation probability was determined using the logit function described by Bady et al. [39]. A cut-off of 0.358 was used to determine MGMT methylation status.

Cell line viability screen.

Multiplexed cell line screening was performed using the PRISM (Profiling Relative Inhibition Simultaneously in Mixtures) platform as previously described [41,42]. KL-50 (2a), TMZ (1a), MTZ (3a), CCNU, Val-083, MMC, and talazoparib were screened across a barcoded library of ~900 cell lines using 8-point 3-fold serial dilutions in triplicate for a 5-day treatment. The maximum doses of KL-50 (2a), MTZ (3a), CCNU, and Val-083 were 200 μM, the maximum dose of TMZ (1a) was 67 μM due to solubility limitation at 200 μM, and the maximum doses of MMC and talazoparib were 10 μM. Bortezomib (20 μM) and DMSO were used as positive and negative controls, respectively. Cell lines with less than 2 passing replicates based on error rate above 0.05 or dynamic ranges less than −log₂(0.3) were filtered out, resulting in 881 unique cell lines passing quality control. Normalized log₂ fold-change cell viability values were calculated with respect to DMSO negative control. A four-parameter logistic curve was fit to the response of each cell line to compute AUC and IC₅₀ values for each dose-response curve. Univariate association between MGMT expression and viability was determined by calculation of Pearson correlations and associated q-values from the Benjamini-Hochberg algorithm for each dose level, AUC, and IC₅₀ values. For subgroup analysis of viability in relation to MGMT and MMR status and cancer type, MGMT mRNA Expression, MMR gene Damaging Mutations, and the Model file containing cell lineage information were downloaded from the Cancer Dependency Map (DepMap) Portal (RRID:SCR_017655) Public 23Q4 release dataset. Cell lines were classified as MGMT+ or MGMT− using a cutoff of MGMT log₂-transformed mRNA expression of <1 SD below the median. Cell lines with a damaging mutation in any of the primary MMR genes (MSH2, MSH6, MLH1, PMS2) were classified as MMR−.

Genomic characterization of cell panels.

Bulk RNA sequencing of patient-derived melanoma cells obtained from the Specimen Resource Core of the Yale SPORE in Skin Cancer has been previously described [43]. Log₂-transformed batch-normalized TPM expression values from 105 melanoma samples were downloaded from the Gene Expression Omnibus (GEO, RRID: SCR_005012) with the accession code GSE190113 and were used to calculate modified Z-scores. For colon cancer cell lines, MGMT mRNA expression log₂(TPM+1) values were downloaded from DepMap Portal (RRID:SCR_017655) Public 23Q4 release dataset. MMR alterations have been reported previously [44-46], and were confirmed by Western blot analysis.

Generation of MGMT+/−, MMR+/− isogenic A101D cells.

Stable A101D MGMT+ cells were generated using a ready-to-use lentiviral preparation encoding human MGMT obtained from the Yale Functional Genomics Core Facility. The lentivirus was produced using expression plasmid pGenLenti-Neo-hMGMT (GenScript, CloneID OHu44348C) and second-generation packaging plasmids psPAX2 (RRID:Addgene_12260) and pMD2.G (RRID:Addgene_12259). A101D cells were transduced following the Addgene protocol for generating stable cell lines with lentivirus [47]. Briefly, reverse transduction was performed by seeding 50,000 cells in DMEM (Gibco) supplemented with 10 μg/mL polybrene (EMD Millipore) into wells of a 6-well plate (Costar, Corning) that were preloaded with varying ratios of lentiviral supernatant, followed by a 72-hour incubation. After infection, the medium was replaced with DMEM with 10% FBS (Gibco), and the cells were allowed to recover for 48 hours before undergoing antibiotic selection. Cells were then selected with 400 μg/mL G418 sulfate (Geneticin, Gibco) for 10–14 days until non-transduced control cells were eliminated, and stable populations were maintained in 200 μg/mL G418 sulfate. MGMT overexpression was verified by Western blot analysis prior to conducting downstream assays. A101D MMR− cells were generated via MSH6 shRNA knockdown as previously described [48]. Briefly, A101D MGMT− and MGMT+ cells were infected with MSH6 shRNA lentiviral media, as previously described, and 8 μg/mL polybrene. After 48 hours, cells were selected with 1 μg/mL of puromycin for 3–4 days before use. Cells were harvested as a polyclonal population, and protein knockdown was confirmed by Western blot analysis.

Western blot analysis.

For Western blots presented in Fig. 2A and Fig. 3A, cells were lysed in RIPA buffer (50 mM HEPES, 250 mM NaCl, 5 mM EDTA, 1% NP-40 in DI H₂O) supplemented with 1X cOmplete EDTA-free Protein Inhibitor Cocktail (Roche) and 1 mM dithiothreitol (DTT, American Bio). Lysates were sonicated for 10 seconds on and 10 seconds off cycles for 1 minute, then centrifuged for 10 min at 13K rpm, 4°C. Cleared lysates were subjected to gel electrophoresis on NuPAGE^™ Bis-Tris Mini Protein Gels, 4–12% (Thermo Fisher), and protein was transferred to a nitrocellulose membrane (Sigma). Membranes were blocked for 1h at RT in each target protein’s respective blocking buffer, followed by blotting with primary and secondary antibodies (Supplemental Table S1). Blots were imaged on a ChemiDoc XRS+ Molecular Imager (Bio-Rad), and densitometry for protein quantification was performed using ImageJ (RRID:SCR_003070).

Fig. 2. KL-50 (2a) displays MGMT-dependent, MMR-independent activity in primary melanoma cells and an engineered isogenic cell line model.

(A) Characterization of patient-derived primary melanoma lines by mRNA expression (top) and Western blot analysis (bottom) of MGMT, MLH1, MSH2, MSH6, and PMS2. GAPDH serves as Western blot loading control. (B–E) Left panels: IC₅₀ values from short-term cell viability assays in melanoma cell lines treated with TMZ (1a), KL-50 (2a), MTZ (3a), or cisplatin. Points indicate replicate IC₅₀ values; bars indicate geometric mean; error bars indicate geometric SD; n = 2 (B–D) or 1 (E) biological replicates. Right panels: IC₅₀ values of cell lines aggregated by MGMT and MMR phenotype. Points indicate geometric mean of IC₅₀ values from individual cell lines; bars indicate geometric mean; error bars indicate geometric SD. Representative dose-response curves are shown in Supplemental Fig. S2B. (F) Western blot analysis of MGMT, MLH1, MSH2, and MSH6 in isogenic A101D lines with comparison to MGMT+/MMR+ CACO2 cells. ß-actin serves as loading control. (G) IC₅₀ values from short-term cell viability assays in isogenic A101D cell lines treated with TMZ (1a), KL-50 (2a), MTZ (3a), or cisplatin. Points indicate replicate IC₅₀ values; bars indicate geometric mean; error bars indicate geometric SD; n = 2–3 biological replicates. Representative dose-response curves are shown in Supplemental Fig. S2C.

Fig. 3. KL-50 (2a) displays MGMT-dependent, MMR-independent activity in colon cancer models in vitro.

(A) Characterization of colon cancer cell lines by MGMT mRNA expression values and known MMR alterations (top) and Western blot analysis of MGMT, MLH1, MSH6, and MSH2 (bottom). GAPDH serves as Western blot loading control. (B–E) Left panels: IC₅₀ values from short-term cell viability assays in colon cancer cell lines treated with TMZ (1a), KL-50 (2a), MTZ (3a), or cisplatin. Points indicate replicate IC₅₀ values; bars indicate geometric mean; error bars indicate geometric SD, n = 3 biological replicates. Right panels: IC₅₀ values of cell lines aggregated by MGMT and MMR phenotype. Points indicate geometric mean of IC₅₀ values from individual cell lines; bars indicate geometric mean; error bars indicate geometric SD. Representative dose-response curves are shown in Supplemental Fig. S3. (F–H) Clonogenic survival assays with TMZ (1a), KL-50 (2a), and MTZ (3a) in CACO2 (F), KM12 (G), and SW620 (H) cell lines. Points indicate mean; error bars indicate SEM; n ≥ 3 technical replicates.

For Western blots presented in Fig. 2F and Fig. 4H, cells were lysed in Pierce RIPA buffer (ThermoFisher Scientific), supplemented with 1X cOmplete Protease Inhibitor Cocktail (Millipore Sigma), 1X PhosSTOP Phosphatase Inhibitor Cocktail (Roche), 1 mM Phenylmethanesulfonyl fluoride (PMSF; Sigma-Aldrich), and 0.1 mM DTT (Sigma-Aldrich). Following incubation on ice for 20–30 minutes, lysates were centrifuged at 14,000 x g for 15 minutes, and total protein was quantified. Equal amounts of protein were resolved on Mini-PROTEAN 4–15% gels (Bio-Rad) and transferred to PVDF membranes (Bio-Rad). Membranes were blocked for 5–10 min in EveryBlot Blocking Buffer (Bio-Rad) and incubated with primary and secondary antibodies (Supplemental Table S1). Chemiluminescent detection was performed on a ChemiDoc^™ XRS+ Molecular Imager (Bio-Rad) or a ChemiDoc Go Imaging System (Bio-Rad) using Clarity Western ECL Substrate (Bio-Rad) or Clarity Max Western ECL Substrate (Bio-Rad).

Fig. 4. KL-50 (2a) effectively treats MGMT-deficient colon cancer xenograft tumors and overcomes induced drug resistance to TMZ (1a).

(A) Tumor growth curves of SW48 CDX tumors in nude mice after treatment with Vehicle (10% cyclodextrin, P.O.), TMZ (1a, 10 mg/kg, P.O.), or KL-50 (2a, 10 mg/kg, P.O.) for 5 days for 1 cycle (n = 7–8 mice per group). (B) Tumor growth curves of SKCO1 CDX tumors in nude mice after treatment with Vehicle (10% cyclodextrin, P.O.) for 5 days, TMZ (1a, 20 mg/kg, P.O.) for 3 days, or KL-50 (2a, 10 mg/kg, P.O.) for 5 days for 1 cycle (n = 5 mice per group). (C) Tumor growth curves of CACO2 CDX tumors in nude mice after treatment with Vehicle (10% cyclodextrin, P.O.), TMZ (1a, 10 mg/kg, P.O.), or KL-50 (2a, 10 mg/kg, P.O.) for days 1–5 of 7-day cycles, for 3 cycles (n = 6–7 mice per group). In panels A–C, points indicate group mean tumor volumes, error bars indicate SEM; significance is indicated relative to vehicle control as determined by ordinary one-way ANOVA with Tukey’s multiple comparisons test. (D–F) Mouse body weights from the xenograft studies presented in panels A–C. Points indicate mean % change in individual mouse body weight normalized to day 0; error bars indicate SEM. (G) IC₅₀ values from short-term cell viability assays in SKCO1 TMZ-R cells treated with TMZ (1a) or KL-50 (2a). Points indicate replicate IC₅₀ values; bars indicate geometric mean; error bars indicate geometric SD; n = 3 biological replicates. Representative dose-response curves are shown in Supplemental Fig. S5A. (H) Western blot analysis of MGMT, MLH1, MSH2, MSH6, and PMS2 in SKCO1 parental and TMZ-R cells. ß-actin serves as loading control and CACO2 cells serve as a positive MGMT expression control. (I–J) Tumor growth curves of SKCO1 TMZ-R #552 (I) or SKCO1 TMZ-R #554 (J) tumors in nude mice after treatment with vehicle control (10% cyclodextrin, P.O.) for 5 days, TMZ (1a, 20 mg/kg, P.O.) for 3 days, or KL-50 (2a, 10 mg/kg, P.O.) for 5 days for 1 cycle (n = 2–5 mice per group). Individual tumor volumes were normalized to starting tumor volume. Points indicate geometric mean of the fold change in tumor volume; error bars indicate geometric SD; significance is indicated relative to vehicle control as determined by lognormal ordinary one-way ANOVA with Tukey’s multiple comparisons test.

For Western blots presented in Fig. 6A, cells were lysed in kinase buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 1 mM EGTA, 1 mM sodium fluoride, 1 mM sodium orthovanadate, 10 mM β-glycerophosphate) supplemented with cOmplete Mini Protease Inhibitor Cocktail tablet (Roche). Cells were lysed on ice for 10 min followed by centrifugation at 14K rpm, 4 °C. Cleared lysates were subjected to gel electrophoresis on NuPAGE^™ Bis-Tris Mini Protein Gels, 4–12% (Thermo Fisher), and protein was transferred to Immobilon-P PVDF membrane (Millipore). After incubation with primary and secondary antibodies (Supplemental Table S1), blots were imaged on an Azure Biosystems c600 imager.

Fig. 6. KL-50 (2a) is effective in TMZ (1a)-resistant intracranial and extracranial PDX models.

(A) Western blot analysis of GBM12 TMZ (1a)-resistant sublines. GBM12TMZ-8023 demonstrates loss of MSH6 relative to placebo-treated subline GBM12P-5199. GBM12TMZ-3080 serves as positive control for MGMT expression. β-actin serves as loading control. (B) Kaplan-Meier survival analysis of GBM12TMZ-8023 intracranial tumors in athymic nude mice treated with KL-50 (2a, 25 mg/kg, P.O.), TMZ (2a, 25 mg/kg, P.O.) or Vehicle (Ora-Plus, P.O.) daily for 5 days, repeated every 28 days, for 3 cycles (n = 10 mice per group). Significance was determined by Log-rank Mantel-Cox test. Adapted from Huseman E. et al., ChemRxiv, 2023 [Preprint]. DOI: https://doi.org/10.26434/chemrxiv-2023-zwj94. (C–D) Tumor growth curves of MGMT−/MSH2− melanoma PDX ME12145 (C) and MGMT−/MLH1− lung adenocarcinoma LU2033 (D) flank tumors in BALBc/nude mice treated with KL-50 (2a, 10 mg/kg, P.O.), TMZ (1a, 10 mg/kg, P.O.) or Vehicle (10% cyclodextrin, P.O.) for 3 cycles of every other day dosing for 3 total doses, repeated every 14 days (n = 10 mice per group). Points indicate group mean tumor volumes; error bars indicate SEM; significance is indicated relative to vehicle control as determined by ordinary one-way ANOVA with Tukey’s multiple comparisons test. (E–F) Kaplan-Meier survival analysis of ME12145 (E) and LU2033 (F) PDX cohorts shown in panels (C) and (D), respectively. In (E), survival endpoint was tumor size >3000 mm³, death, or sacrifice due to weight loss. In (F), survival endpoint was tumor size >1500 mm³, death, or sacrifice due to weight loss. Significance was determined by Log-rank Mantel-Cox test.

For Western blots presented in Supplemental Fig. S7A, tumor fragments were chopped and mechanically dissociated using a disposable pestle in an Eppendorf tube and then lysed in RIPA buffer (50 mM Tris-HCl [pH 7.5], 150 mM NaCl, 0.1% SDS, 1% NP-40, 0.5% sodium deoxycholate, 1 mM β-glycerophosphate, 2.5 mM sodium pyrophosphate, 10 mM sodium fluoride, 1 mM sodium orthovanadate, 2.5 μM pepstatin A, 1 mM EDTA, and 1 mM EGTA) supplemented with 1X EDTA-free cOmplete Protease Inhibitor Cocktail (Roche). Lysates were sonicated for 10 seconds on and 10 seconds off cycles for 1 minute, then centrifuged for 10 min at 13K rpm, 4°C. Cleared lysates were subjected to gel electrophoresis on a NuPAGE^™ Bis-Tris Mini Protein Gel, 4–12% (Thermo Fisher), and protein was transferred to a PVDF membrane (Millipore). Membranes were blocked in Intercept-TBS Blocking Buffer (LI-COR Biosciences), incubated with primary and secondary antibodies (Supplemental Table S1), and then imaged on an Odyssey Imaging system (LI-COR Biosciences).

Individual Western blots were repeated at least two times, with representative images shown (excluding Fig. 6C, where only a single Western blot was run due to limited PDX material availability).

Short-term cell viability assays.

Cells were cultured, counted, and plated into 96-well plates in a volume of 100 μL/well. After overnight adherence, the inner 60 wells were treated with a 9-point 2-fold or 3-fold serial dilution of TMZ (1a), KL-50 (2a), MTZ (3a), or cisplatin. DMSO concentration was held at 1% for all drug concentrations and used as a negative control. The plates were incubated for 6 days and then prepared for analysis by gently washing with Dulbecco’s Phosphate Buffered Saline (DBPS), fixing in 3.7% Formaldehyde (Thermo Fisher) in DPBS for 30 minutes, washing with DPBS, and staining with 1 μg/mL Hoechst 33342 dye (Thermo Fisher) for 30 minutes before imaging on a Cytation 3 Cell Imager (Biotek). Cell counts were measured using CellProfiler Image Analysis Software (RRID:SCR_007358) and normalized to untreated control wells. A non-linear regression curve, [Inhibitor] vs. normalized response with variable slope, was used to determine IC₅₀ values.

Clonogenic survival assays.

Cells were trypsinized, counted, and plated in 10-cm plates. After overnight adherence, the plates were treated with TMZ (1a), KL-50 (2a), or MTZ (3a) in a series of 2-fold dilutions with maximum doses of 100 μM. The DMSO concentration was held at 0.1% for all drug concentrations and used as a negative control. After 72 hours, cells from each 10-cm plate were counted, suspended into single-cell solutions, and replated into triplicate wells of a 6-well plate at cell densities ranging from 50 to 200,000 cells/well based on expected surviving fractions. The 6-well plates were left to incubate for 14 days then fixed, stained, and counted as previously described [17].

Generation of TMZ-R SKCO1 cells.

SKCO1 cells were seeded at a density of 1.5 × 10⁵ cells per 10-cm dish (Corning #430167) and allowed to adhere for 48 hours under standard culture conditions (37°C, 5% CO₂). Treatment with TMZ (1a) was initiated at 30 μM, and cells were maintained in media containing the drug for 5–7 days. After each treatment cycle, surviving cells were allowed to recover in drug-free media until they reached ~80–90% confluence. Recovery was monitored visually, based on adherence, morphology, cell viability, and proliferation. The dose of TMZ (1a) was incrementally increased every 1–2 passages (e.g., 30 μM to 50 μM to 75 μM to 100 μM), until a final concentration of 500 μM was achieved. This stepwise escalation was continued over 8–10 weeks to promote adaptation and selection of resistant clones. The established TMZ (1a)-resistant (TMZ-R) line was expanded and cryopreserved for downstream analysis. Control cells were passaged in parallel without TMZ (1a).

Animal studies.

1. Colon cancer cell-derived xenograft (CDX) flank efficacy studies.

Athymic nude mice (Hsd:Athymic Nude-Foxn1^nu) were purchased from Envigo. Human SW48, SKCO1, CACO2, and SW620 cells were used to establish tumor models by subcutaneous implantation as previously described [17]. Briefly, tumor cells (5 × 10⁶ SW48, SKCO1, or SW620 or 10⁷ CACO2) were diluted in 50 μL Matrigel (Corning) and were injected into the flanks of female nude mice. Treatment was initiated when the average tumor volume reached approximately 100–300 mm³, at which time mice were randomized and assigned to treatment groups. SW48 tumor-bearing mice were dosed by oral gavage daily with vehicle control (10% cyclodextrin), TMZ (1a, 10 mg/kg), or KL-50 (2a, 10 mg/kg) for 5 days for 1 cycle. SKCO1 tumor-bearing mice were dosed by oral gavage daily with vehicle control (10% cyclodextrin) for 5 days, TMZ (1a, 20 mg/kg) for 3 days, or KL-50 (2a, 10 mg/kg) for 5 days for 1 cycle. CACO2 tumor-bearing mice were dosed by oral gavage daily with vehicle control (10% cyclodextrin), KL-50 (2a, 10 mg/kg), or TMZ (1a, 10 mg/kg) on days 1–5 of 7-day cycles, for 3 cycles. SW620 tumor-bearing mice were dosed by oral gavage daily with vehicle control (10% cyclodextrin), KL-50 (2a, 20 mg/kg), or MTZ (3a, 7.5 mg/kg or 10 mg/kg) for 5 days for 1 cycle. Tumors were measured using calipers and tumor volume was calculated using the formula V = 0.52 × Length × Width². Mice were euthanized if body weight loss exceeded 20% or if tumor volume exceeded ~2000 mm³. All experimental procedures were approved by the Yale University Institutional Animal Care and Use Committee (IACUC).

2. TMZ (1a) Resistant (TMZ-R) SKCO1 tumor establishment and drug efficacy testing.

The mice in the SKCO1 flank efficacy study treated with an initial dosing cycle of TMZ (1a, 20 mg/kg) for 3 days were utilized for generation of TMZ-R sublines. Upon tumor regrowth, mice were dosed with a second cycle of an increased dose of TMZ (1a, 40 mg/kg) for 3 days, followed by a final third cycle of increased dose of TMZ (1a, 60 mg/kg) for 3 days, prior to sacrifice and tumor harvest. Two TMZ-R tumors (#552 and #554) were reimplanted subcutaneously into the flanks of NSG mice. Once tumor volume reached approximately 2000 mm³, tumors were harvested for tertiary reimplantation. Excised tumor tissues were minced into small fragments and further homogenized using a handheld Fisherbrand^™ Pellet Pestle^™ Motor (Fisher Scientific), followed by repeated passage through an 18-gauge PrecisionGlide^™ needle (BD) to obtain a uniform suspension. The resulting tumor homogenate was mixed 1:1 with ice-cold Matrigel (Corning), and 100 μL of the mixture was subcutaneously injected into the flanks of athymic female nude mice using a 30-gauge PrecisionGlide^™ needle (BD). Treatment was initiated when the average tumor volume reached approximately 400–500 mm³, at which time mice were randomized and assigned to treatment groups. Mice were dosed by oral gavage daily with vehicle control (10% cyclodextrin) for 5 days, TMZ (1a, 20 mg/kg) for 3 days, or KL-50 (2a, 10 mg/kg) for 5 days for 1 cycle. Tumors were measured using calipers and tumor volume was calculated using the formula V = 0.52 × Length × Width². Due to large variation in initial tumor volume, tumor growth was normalized to starting volume. Mice were euthanized if body weight loss exceeded 20% or if tumor volume exceeded ~3000 mm³. All experimental procedures were approved by the Yale University IACUC.

3. Mouse toxicity studies.

Female athymic nude mice (Hsd:Athymic Nude-Foxn1^nu, 3 weeks old, ~15 g) were purchased from Envigo. In the first study, mice were weighed and randomized to treatment by oral gavage with vehicle control or 5, 10, or 20 mg/kg MTZ (3a), given on days 1–5 of 7-day cycles, for 2 cycles. In the second study, mice were dosed by oral gavage with MTZ (3a, 7.5 mg/kg), TMZ (1a, 20 mg/kg), KL-50 (2a, 20 mg/kg), or vehicle control for 5 days. Mice were sacrificed on day 6 for blood and bone marrow analysis. In both studies, mouse body weights were measured and normalized to baseline weights. All experimental procedures were approved by the Yale University IACUC.

4. Glioma patient-derived xenograft (PDX) intracranial efficacy study.

GBM12 TMZ-resistant subline GBM12TMZ-8023 was previously established by dosing the parental GBM12 line with 50 mg/kg TMZ (1a) on days 15–17, 44–46, and 71–74 [49]. Female athymic nude mice (NCr-nu/nu, strain code: 553, age 6–7 weeks) were purchased from Charles River. Intracranial xenografts were established with the GBM12TMZ-8023 subline as previously described [50]. Mice were randomized and dosed by oral gavage daily with KL-50 (2a, 25 mg/kg), TMZ (1a, 25 mg/kg) or vehicle control (Ora-Plus) on days 1–5 of 28-day cycles, for 3 cycles (Study Days 4–8, 32–36, and 60–64). Mice were observed daily and euthanized when moribund, as determined by body condition score, circling, hunched posture, and/or neurological deficits. Presence or absence of brain tumor was noted as grossly seen by eye. All animal use and procedures were approved by the Mayo Clinic IACUC (Protocol #A00006634-22).

5. Non-glioma PDX flank efficacy studies.

BALB/c nude mice were obtained from GemPharmatech Co., Ltd (Changzhou), maintained in individual ventilation cages at up to 5 mice per cage at 20–24 °C and 40–70% humidity, and provided food and water ad libitum. Tumor fragments measuring 2–3 mm in diameter of lung cancer (LU2033) and melanoma (ME12145) primary patient-derived xenograft tumors were harvested and used for inoculation into the right upper flanks of 6–8-week-old female BALB/c nude mice. When mean tumor sized reached ~175 mm³, mice in each tumor model group were randomized and dosed by oral gavage daily with KL-50 (2a, 10 mg/kg), TMZ (1a, 10 mg/kg) or vehicle control (10% cyclodextrin) on days 1, 3, and 5 of 14-day cycles, for 3 cycles. Body weights and tumor volumes were measured twice per week after randomization. For the ME12145 model, diet gel was supplied to all mice beginning on Day 7. For the LU2033 model, diet gel was supplied to all mice in a group once an individual mouse displayed body weight loss >10%. A dosing holiday was given to individual mice with body weight loss >15% on the day of dosing. Mice were euthanized if body weight loss exceeded 20% or if tumor volume exceeded 3000 mm³. All procedures involving care and use of animals were approved by the IACUC of CrownBio and were conducted in accordance with the regulations of the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC).

Blood and bone marrow analysis.

Mice were sacrificed after 5 days of treatment with MTZ (3a), TMZ (1a), or KL-50 (2a) and compared to vehicle-treated mice. Blood was collected via cardiac puncture and mixed with EDTA so that the final EDTA concentration was approximately 5 mM. Whole blood counts were performed using a HemaVet 950FS (Drew Scientific). Bone marrow was extracted by cutting open both ends of the femur and tibia bones, placing them in an Eppendorf tube, and centrifuging at 5000 x g for 1 minute. The pellet containing the bone marrow was resuspended in 1 mL PBS and counted on a TC20 Automated Cell Counter (BioRad). Bone marrow cells counts were normalized to total bone weight.

Statistical analysis.

All statistical analyses were performed using GraphPad Prism (RRID:SCR_002798). For PRISM screen analyses, statistical analysis was performed by ordinary one-way ANOVA with Tukey’s multiple comparisons test for comparison of 3 or more groups (Fig. 1D) or by unpaired two-tailed t test for comparison of 2 groups (Fig. 1E). For in vivo growth curve and mouse body weight analyses, statistical testing was performed at the indicated time points by ordinary one-way ANOVA with Tukey’s multiple comparisons test for comparison between all groups (Fig. 4A-C, 5B, 5D, 6C-D), by lognormal ordinary one-way ANOVA with Tukey’s multiple comparisons test for comparison of fold changes in tumor volume between all groups (Fig. 4I-J), or by ordinary one-way ANOVA with Dunnett’s multiple comparisons test for comparison to control group only (Fig. 5A, Supplemental Fig. S6A-C). Significance testing of Kaplan-Meier survival curves was by Log-rank Mantel-Cox test (Fig. 6B, 6E-F). In all figures, statistical significance is designated as *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001; n.s., not significant.

Fig. 5. KL-50 (2a) has less mouse toxicity compared to MTZ (3a) at doses equivalent for tumor control.

(A) Body weights of athymic nude mice following P.O. administration of Vehicle (10% cyclodextrin) or MTZ (3a) at 5, 10, or 20 mg/kg daily for 2 cycles of 5 days on 7 days off (n = 3 mice per group). Points indicate group mean % change in weight normalized to individual mouse baseline weight; error bars represent SEM; significance is indicated relative to vehicle control as determined by ordinary one-way ANOVA with Dunnett’s multiple comparisons test. (B) Tumor growth curves of SW620 colon cancer CDX tumors in athymic nude mice treated with a single cycle of 5 days of P.O. treatments with Vehicle (10% cyclodextrin), MTZ (3a, 7.5 or 10 mg/kg), or KL-50 (2a, 20 mg/kg) (n = 9–10 mice per group). Bars indicate group mean; error bars indicate SEM; significance determined by ordinary one-way ANOVA with Tukey’s multiple comparisons test. (C) Mouse body weights from the xenograft study presented in (B). Points indicate group mean % change in individual mouse body weight normalized to day 0; error bars indicate SEM. (D) Mouse weights at Day 5 from the xenograft study presented in (B). Data points indicate % change in individual mouse body weight normalized to day 0; bars indicate group mean; error bars indicate SEM; significance determined by ordinary one-way ANOVA with Tukey’s multiple comparisons test.

Data availability.

The TGCA PanCancer MGMT mRNA expression and methylation data analyzed in this study were obtained from cBioPortal. The processed melanoma mRNA expression data analyzed in this study were obtained from Gene Expression Omnibus (GEO) at GSE190113. The established cell line mRNA expression data analyzed in this study were obtained from the DepMap Portal Public 23Q4 release. All other raw data generated in this study are available upon request to the corresponding authors.

Results

Large tumor databases and cell line screening reveal the wide potential of KL-50 (2a) in targeting MGMT-deficient tumors.

Use of MGMT deficiency as a biomarker for alkylator sensitivity is best-established in human gliomas. However, loss of MGMT expression by immunohistochemistry and MGMT promoter silencing has been reported in significant fractions of extracranial solid tumors, including melanoma, colon cancer, lung cancer, esophageal cancer, neuroendocrine tumors, and sarcomas, and has been correlated with alkylator response [28-34,36,51-56]. This observation led us to investigate the potential for using 2-fluoroethylating agents in a tumor type-agnostic manner. We performed an updated analysis of The Cancer Genome Atlas (TCGA) Pan-Cancer Database to identify the rates of MGMT loss across cancer types (Fig. 1B). A cutoff of <1 SD below the median MGMT expression level was used to identify MGMT-deficient (MGMT−) tumors. Using this cutoff, 77% of lower grade gliomas and 42% of glioblastomas were categorized as MGMT−, consistent with well-established rates of MGMT loss in these tumors [2,57]. Overall, 14.4% of tumors in the TCGA PanCancer dataset were MGMT−. The other solid cancer types with the highest percentage of MGMT− tumors included cutaneous melanoma (37.9%), esophageal carcinoma (34.3%), head and neck squamous cell carcinoma (21.9%), lung squamous cell carcinoma (21.3%), colorectal adenocarcinoma (17.1%), and sarcoma (14.6%).

We also analyzed MGMT promoter methylation across the TCGA PanCancer dataset using the model developed by Bady et al., [39] and correlated the results with MGMT mRNA loss of expression on a cancer type and individual tumor level basis (Supplemental Fig. S1). Overall, 14.7% of tumors in the TCGA PanCancer dataset were MGMT promoter silenced. Across all cancer types, 9.1% of samples displayed both MGMT promoter silencing and MGMT mRNA loss, while 10.7% of samples displayed either MGMT promoter silencing or MGMT mRNA loss, but not both, and the remaining 80.2% of samples had no MGMT promoter silencing and no loss of mRNA. Altogether, these results demonstrate the frequent loss of MGMT across cancer types.

We previously subjected KL-50 (2a) to PRISM, a multiplexed barcoded cell line screen containing approximately 900 cell lines. The viability of these cell lines was correlated with gene expression profiles, and MGMT expression was found to be the strongest correlated feature with KL-50 (2a) response [18]. To benchmark this activity against comparator agents, we repeated this screen with KL-50 (2a), TMZ (1a), the chloroethylating agents MTZ (3a) and CCNU, the non-specific crosslinking agents Val-083 [38] and mitomycin C (MMC), and the PARP inhibitor talazoparib. We again identified MGMT mRNA expression as the strongest correlated variable with response to KL-50 (2a), as well as TMZ (1a), MTZ (3a), and CCNU. In comparing the magnitude and significance of the MGMT correlations among agents, we found that KL-50 (2a) displayed the strongest dose-dependent correlation between MGMT expression and various dose and curve metrics, followed by MTZ (3a), TMZ (1a), and finally CCNU (Fig. 1C). As expected, agents that do not generate O⁶-alkylguanine lesions, such as Val-083, MMC, and talazoparib showed no MGMT dependence (Pearson correlations of −0.013, 0.003, and 0.015; p > 0.5).

To investigate the impact of MMR loss on MGMT dependence, we utilized damaging mutations in any of the 4 primary MMR genes (MSH2, MSH6, PMS2, MLH1) contained in the DepMap dataset to identify tumors with likely loss of MMR function. MGMT mRNA expression <1 SD below the median was utilized as a cutoff for MGMT loss, and we divided cell lines into four subgroups based on MGMT and MMR status. KL-50 (2a) demonstrated significant activity in both MGMT−/MMR+ and MGMT−/MMR− subgroups compared to MGMT+ cell lines (Fig. 1D). In contrast, TMZ (1a) was active in the MGMT−/MMR+ subgroup but showed no activity in MGMT−/MMR− cell lines compared to MGMT+ cell lines. We then divided the cell lines by cancer type and analyzed KL-50 (2a) response in MGMT− and MGMT+ subgroups. Across nearly all solid cancer types, we observed a statistically significant decrease in cell viability with KL-50 (2a) in the MGMT− compared to MGMT+ subgroups (Fig. 1E).

KL-50 (2a) displays MGMT-selective, MMR-independent activity across panels of in vitro cancer models.

To validate the results observed from the large-scale cell line screen, we obtained representative cell line panels of MGMT+/− and MMR+/− phenotypes. Examining the PanCancer TCGA data, we identified cutaneous melanoma as having a high percentage of tumors with low MGMT mRNA expression and pursued it as a model cell type in vitro (Fig. 1B). Patient-derived melanoma cell lines from the Yale SPORE in Skin Cancer biorepository were identified as putatively MGMT− using MGMT mRNA expression modified Z-score cutoff of approximately less than −1. Putative MMR− models were similarly chosen by identification of cell lines with low MSH2, MSH6, MLH1, or PMS2 mRNA levels (Fig. 2A). Western blotting confirmed the presence of MGMT protein expression in the YUHEF and YUDOSO cell lines while the YUMUT, YUSEEP, YUNIGE, and YUROL cell lines lacked detectable expression of MGMT. The YUROL cell line was identified as having low MMR gene mRNA expression and displayed reduced MLH1 and barely detectable MSH2 and MSH6 protein expression, consistent with a deficiency in the MMR pathway (Fig. 2A, Supplemental Fig. S2A). In addition, prior whole-exome sequencing has identified MLH1 nonsense and missense and MSH6 missense mutations in the YUROL cell line [43]. Moreover, mutational signature analysis demonstrated that the YUROL mutational signature is characterized predominantly by Signature 11 (79.5%) [43], which has a proposed etiology of TMZ treatment in combination with MMR deficiency [15,58]. Together, these data strongly support functional MMR deficiency in the YUROL cell line secondary to MLH1 and/or MSH6 mutations, and further suggest that the cell line may derive from a tumor with induced MMR deficiency due to prior alkylating chemotherapy.

In short-term cell viability assays, MGMT−/MMR+ cell lines (YUMUT, YUSEEP, and YUNIGE) displayed sensitivity to TMZ (1a) while the MGMT−/MMR− cell line (YUROL) was resistant. Likewise, MGMT+ cells (YUHEF and YUDOSO) were considerably more resistant to TMZ (1a) than those with a MGMT−/MMR+ phenotype (Fig. 2B). Aggregating these results by MGMT and MMR status, the MGMT−/MMR+ cells had a 23.4-fold increase in MGMT-dependent sensitivity to TMZ (1a) whereas the MGMT−/MMR− cell line, YUROL, had a 0.7-fold sensitivity. When these melanoma cell lines were treated with KL-50 (2a), a striking shift in the MMR− YUROL cells could be seen with an MGMT-dependent sensitivity of 34.6-fold (Fig. 2C). A strong MGMT-dependent sensitivity of 15.3-fold was maintained in the aggregated YUMUT, YUSEEP, and YUNIGE cell lines (Fig. 2C). In comparison, upon treatment with MTZ (3a) the MGMT-dependent sensitivity dropped to 7.1-fold in the MMR− YUROL and to 8.9-fold in the aggregated MMR+ lines (Fig. 2D). To ensure the MGMT-dependence of the trends seen, we also tested the differential sensitivities to the platinating agent cisplatin. The MMR− YUROL displayed mild resistance to cisplatin, consistent with previous reports of MMR based cisplatin resistance [59], but there was no dependence on MGMT expression (Fig. 2E).

To confirm the MGMT and MMR dependences observed above in an isogenic setting, we performed overexpression of MGMT and shRNA-mediated knockdown of MSH6 in the A101D melanoma cell line, which displays baseline loss of MGMT (Fig. 2F). As expected, MGMT overexpression resulted in a substantial increase in resistance to TMZ (1a) and KL-50 (2a), and to a lesser extent, MTZ (3a), while MSH6 knockdown conferred resistance only to TMZ (1a) (Fig. 2G).

Like melanoma, colorectal cancer was identified as commonly having decreased MGMT expression in the TCGA database (Fig. 1B). A representative panel of established colon cancer cell lines was chosen based on MGMT mRNA expression and MMR gene mutations and mRNA expression in the Broad Institute’s DepMap Consortium (Fig. 3A). Western blotting confirmed CACO2, SW480, and DLD1 as MGMT+ cell lines while SW620, SKCO1, SW48, and KM12 were confirmed to be MGMT− (Fig. 3A). SW48 and KM12 displayed loss of MLH1 and DLD1 displayed loss of MSH6, indicative of MMR deficiency while the remaining cell lines had intact MMR protein expression (Fig. 3A).

We again performed short-term cell viability assays comparing the activity of TMZ (1a), KL-50 (2a), and MTZ (3a). MGMT−/MMR+ cell lines exhibited increased sensitivity to TMZ (1a) relative to MGMT+ or MGMT−/MMR− cell lines (Fig. 3B). This resulted in an MGMT-dependent sensitivity of 71.8-fold in the aggregated MGMT−/MMR+ cell lines whereas there was no increase in sensitivity in the MGMT−/MMR− cell lines. With KL-50 (2a), the MGMT+ cells remained resistant while resistance in the MMR− cells was reversed, with an aggregated MGMT-dependent therapeutic index of 24.7-fold in MGMT−/MMR+ lines and 31.8-fold in MGMT−/MMR− lines (Fig. 3C). Treatment with MTZ (3a) resulted in diminished MGMT-dependence, with 16.1-fold MGMT-dependent sensitivities in both MMR+ and MMR− cells (Fig. 3D). With cisplatin, there was a trend toward elevated resistance in the MMR− lines, but no discernible pattern of sensitivity between MGMT+ and MGMT− cells (Fig. 3E).

To further assess the strength of the MGMT-dependence of KL-50 (2a) compared to MTZ (3a) in vitro, we performed clonogenic survival assays in representative colon cancer cell lines. In the MGMT+ CACO2 cell line, a striking decrease in surviving fraction was observed with MTZ (3a) at 50 and 100 μM whereas there was no significant decrease with KL-50 (2a) or TMZ (1a) at the same doses (Fig. 3F). As expected, TMZ (1a) had no effect on MGMT−/MMR− KM12 cells while both KL-50 (2a) and MTZ (3a) greatly decreased the surviving fraction at 50 and 100 μM (Fig. 3G). The MGMT−/MMR+ SW620 cell line demonstrated similar sensitivities to TMZ (1a), KL-50 (2a), and MTZ (3a) at all doses (Fig. 3H). Altogether, these data demonstrate the superiority of KL-50 (2a) as a potential therapeutic for MGMT−/MMR− tumors compared to TMZ (1a) and MTZ (3a) in vitro.

KL-50 (2a) effectively treats MGMT-deficient non-glioma xenograft tumors and overcomes TMZ (1a) resistance due to MMR loss.

We next tested whether the activity of KL-50 (2a) would be recapitulated in vivo. We grew several colon cancer cell lines as cell-derived xenograft (CDX) tumors in mice and evaluated the growth inhibitory capability of KL-50 (2a) compared to TMZ (1a). Consistent with our in vitro studies, TMZ (1a) demonstrated no activity in the MGMT−/MMR− SW48 CDX model, while KL-50 strongly inhibited tumor growth (Fig. 4A, Supplemental Fig. S4A). In the SKCO1 MGMT−/MMR+ CDX model, TMZ (1a) and KL-50 (2a) displayed similar efficacy, though we observed earlier tumor regrowth following 1 cycle of TMZ (1a) (Fig. 4B, Supplemental Fig. S4C-E). Finally, CACO2 MGMT+ tumors were completely resistant to both TMZ (1a) and KL-50 (2a) despite multiple cycles of drug dosing (Fig. 4C, Supplemental Fig. S4B). In all studies, we saw that TMZ (1a) and KL-50 (2a) were well tolerated as assessed by mouse body weights (Fig. 4D-F).

We next developed an in vitro TMZ (1a)-resistant (TMZ-R) cell model by treating SKCO1 cells with increasing doses of TMZ (1a). After ~2 months of treatment, SKCO1 TMZ-R cells displayed a 36-fold increase in their TMZ (1a) IC₅₀ compared to untreated cells passaged in parallel. In contrast, they remained highly sensitive to KL-50 (2a) (Fig. 4G, Supplemental Fig. S5A). Western blot analysis of these TMZ-R cells revealed loss of expression of MSH6 with a corresponding reduction in MSH2 expression (Fig. 4H).

Having shown the ability of KL-50 (2a) to kill TMZ-R SKCO1 cells in vitro, we undertook an analogous study in vivo. As noted previously, SKCO1 CDX tumors in mice treated with TMZ (1a, 20 mg/kg) for 3 days for a single cycle responded initially but exhibited tumor regrowth in 3 of 5 mice after approximately 60 days and in the remaining 2 mice after approximately 90 days (Fig. 4B, Supplemental Fig. S4E). In contrast, tumors in mice treated with KL-50 (1a, 10 mg/kg) for 5 days for a single cycle showed no evidence of regrowth after >100 days (Fig. 4B, Supplemental Fig. S4D). To establish bona fide TMZ-R tumors, mice in the TMZ (1a) treatment arm were dosed with an increased dose of TMZ (1a, 40 mg/kg) for 3 days. Tumors showed variable response to this cycle and were allowed to resume growth prior to treatment with a final cycle of increased dose of TMZ (1a, 60 mg/kg) for 3 days. All 5 tumors eventually showed regrowth (Supplemental Fig. S4E). Two TMZ-R tumors were passaged into additional mice for rechallenge with TMZ (1a, 20 mg/kg) for 3 days or KL-50 (2a, 10 mg/kg) for 5 days (Fig. 4I-J, Supplemental Fig. S5B-E). Tumor growth curve analysis demonstrated sustained resistance to TMZ (1a) and continued robust response to KL-50 (2a), confirming the ability of KL-50 (2a) to overcome induced TMZ (1a) resistance in vivo.

KL-50 (2a) has durable control of tumor growth while mitigating mouse toxicity observed with MTZ (3a).

In cell-based in vitro assays, KL-50 (2a) displays significantly enhanced MGMT-dependence compared to MTZ (3a), as evidenced by the higher MGMT expression correlation coefficient in the PRISM screen and the higher therapeutic indices across melanoma and colon cancer cell line panels (Fig. 1C, 2, 3). This effect is driven primarily by increased cytotoxicity of MTZ (3a) in MGMT+ cells relative to KL-50 (2a), while the effective cytotoxicity in MGMT− cells is similar between MTZ (3a) and KL-50 (2a). Given the history of unacceptable toxicity with MTZ (3a) in clinical trials [20-24], we sought to compare the differences between these agents in vivo.

Previous work by our group demonstrated that KL-50 (2a) is well-tolerated in murine models with <3% decrease in mouse body weight at 25 mg/kg in a single 5-day cycle of daily treatment and could be given in single doses up to 100 mg/kg [17]. In comparison, the lethal dose (LD₁₀) of MTZ (3a) in mice was previously reported at 45 mg/kg when given as a single dose [60]. To determine the toxicity and maximum tolerability of repeated doses of MTZ (3a) in vivo, we performed a pilot dose escalation study to observe the effects of MTZ (3a) treatment on mouse body weight as a general marker for tolerability. Mice were treated once daily with 5, 10, or 20 mg/kg MTZ (3a) P.O. for two cycles of 5 days on, 7 days off (n = 3 per group). Following the first cycle, dose-dependent body weight loss was seen across all groups, with significant decreases in average body weight at day 5 of 7.2% and 11.3% in groups receiving 10 mg/kg or 20 mg/kg MTZ (3a) respectively (Fig. 5A). All 3 mice treated with 20 mg/kg MTZ (3a) lost >10% body weight. Following the second cycle, signs of gastrointestinal toxicity (i.e., loose stools) were observed in MTZ (3a)-treated mice. Further weight loss leading to an average decrease of 10.9% was observed in the group receiving 10 mg/kg, with one mouse displaying 18% body weight loss. All 3 mice treated with 20 mg/mg displayed >30% body weight loss. Thus, we estimate the maximum tolerable dose of MTZ (3a) in mice to be approximately 10 mg/kg when delivered in 5-day cycles.

To directly compare KL-50 (2a) and MTZ (3a), an in vivo tumor efficacy study was performed utilizing SW620 MGMT−/MMR+. Doses of 7.5 and 10 mg/kg of MTZ (3a) and 20 mg/kg of KL-50 (2a) were chosen to maximize efficacy and tolerability based on the pilot study. Xenograft-bearing mice were treated for 5 consecutive days and monitored for both toxicity based on body weight loss and tumor growth control. One-way ANOVA at the termination of the study demonstrated that all treatment groups significantly reduced tumor growth rates compared to control with no statistically significant difference between treatment groups (Fig. 5B). Consistent with our prior experiments, 20 mg/kg KL-50 (2a) caused no body weight loss compared to the control group, whereas both 7.5 and 10 mg/kg MTZ (3a) resulted in significant body weight loss, which peaked five days post treatment start, with a subset of mice experiencing >10% body weight loss (Fig. 5C-D). We assessed mouse complete blood counts and bone marrow cellularity following a 5-day treatment cycle with 7.5 mg/kg MTZ (3a), 20 mg/kg KL-50 (2a), or 20 mg/kg TMZ (1a). We observed slight trends toward decreased white blood cell (WBC) counts and decreased hemoglobin with MTZ (3a) treatment, but overall, there were no significant decreases in hematologic parameters relative to vehicle-treated mice (Supplemental Fig. S6A-B). Significant weight loss was again observed with MTZ (3a) but not KL-50 (2a) or TMZ (1a) (Supplemental Fig. S6C). Overall, these results indicate that KL-50 (2a) displays reduced toxicity in vivo at significantly higher doses than MTZ (3a) while maintaining comparable tumor control.

KL-50 (2a) has potent in vivo activity against MGMT-deficient, TMZ (1a)-resistant patient-derived xenografts.

Finally, we turned to patient-derived xenograft (PDX) models to investigate the in vivo anti-tumor activity in intracranial and extracranial MGMT-deficient cancer models. KL-50 (2a) efficacy was recently demonstrated in a post-TMZ (1a), MMR-deficient glioblastoma (GBM) PDX [61]. We first confirmed this observation using a separate TMZ (1a)-resistant glioblastoma (GBM) PDX model (GBM12TMZ subline 8023), previously developed through repeated cycles of in vivo TMZ (1a) exposure of the parental MGMT-silenced GBM12 PDX [49]. Western blot analysis demonstrated that this TMZ (1a)-resistant subline remained MGMT-deficient but displayed loss of MSH6 protein (Fig. 6A). In an orthotopic intracranial survival study in mice harboring this subline, TMZ (1a), yielded no survival benefit, whereas KL-50 (2a) produced a >6-fold increase in median overall survival (mOS) (Fig. 6B). Mice in control and TMZ (1a)-treated groups had visible brain tumor at sacrifice. In contrast, only one mouse in the KL-50 (2a)-treated group had obvious tumor.

We next selected a panel of commercially available non-glioma PDX models predicted to be MGMT deficient or proficient based on MGMT mRNA expression and MMR deficient or proficient based on microsatellite instability status. Western blot analysis across this panel of PDX models confirmed several MGMT− tumors with loss of MLH1 or MSH2 (Supplemental Fig. S7A). We selected two models, one MSH2-deficient melanoma and one MLH1-deficient lung adenocarcinoma PDX model, for in vivo efficacy studies. In both models, KL-50 (2a) displayed robust anti-tumor activity and significantly improved the survival outcome (Fig. 6C-F, Supplemental Fig. S7B-C). Tumors uniformly regressed and were monitored for up to 7 weeks post-treatment with no evidence of tumor recurrence. In contrast, both tumors were highly resistant to TMZ (1a).

The dose of KL-50 (2a) selected in these studies, while not expected to induce toxicity based on prior reported studies in non-tumor bearing mice [17], did result in an average of 9.3% and 11.5% body weight loss in the lung cancer and melanoma xenograft-bearing mice, respectively, following the first dosing cycle (Supplemental Fig. S7D-E). In subsequent dosing cycles, weight loss was less substantial. One mouse in the lung cancer PDX cohort was euthanized due to body weight loss, whereas weights recovered in all remaining mice.

Discussion

The frequent loss of MGMT expression in tumors creates a therapeutic window for selectively targeting MGMT-deficient tumors with agents that alkylate the guanine O⁶ position, such as TMZ (1a) or CCNU. However, emergence of drug resistance and off-target normal tissue toxicity often limit the efficacy of these agents. As previously reported, KL-50 (2a), through its kinetically-controlled, MGMT-dependent interstrand crosslinking capability, overcomes MMR-based resistance to TMZ (1a) in glioma cell lines and orthotopic models [17,18]. Here we demonstrate that KL-50 (2a) is highly efficacious in a variety of MGMT-deficient non-glioma tumor models both in vitro and in vivo. We confirm that KL-50 (2a) remains active independent of MMR status and overcomes induced resistance to TMZ (1a) in a colon cancer model. We further establish this fluoroethylating agent to have enhanced MGMT− tumor selectivity compared to chloroethylating agents, as evidenced by higher in vitro therapeutic indices and comparable efficacy in human CDX tumor studies with better tolerability in mice. While our studies focus largely on colon cancer and melanoma models due to model availability and the high rates of MGMT loss in these tumor types, this work builds support for development of 2-fluoroethylating agents as potential drugs for exploiting tumor loss of MGMT, agnostic of tumor type.

According to our analysis of TCGA and consistent with multiple prior studies [28,29], MGMT promoter silencing and loss of expression is prevalent in numerous cancer types outside of glioma, including cutaneous melanoma, esophageal cancers, head and neck cancer, lung cancer, and colorectal cancer. Past literature supports the notion that MGMT promoter hypermethylation is a valuable clinical biomarker in both gliomas and extracranial malignancies [2,31,32,54,55]. Loss of MMR is well-established as a resistance mechanism to TMZ (1a) in gliomas [6,13-16]. While TMZ (1a) resistance is less studied outside of glioma, MMR loss has been reported upon treatment with TMZ (1a) in patients with colorectal cancer [37], consistent with our findings of loss MMR loss upon treatment of SKCO1 cells with TMZ (1a). Baseline MMR deficiency is also seen across multiple cancer types, including >15% of colon cancer, potentially accounting for intrinsic resistance to TMZ (1a) [62]. Furthermore, MMR deficiencies are clinically readily identified in tumor specimens by immunohistochemistry of MMR proteins or by microsatellite instability testing [63]. Given these established clinical biomarkers, the evidence provided in this paper suggests the potential for a biomarker-based strategy using 2-fluoroethylating agents in tumors identified as MGMT-silenced and MMR-deficient as second-line therapy in the case of tumor progression on TMZ (1a) or as first-line therapy in MGMT-silenced tumors to bypass this known resistance mechanism altogether.

TMZ (1a) has a long history of use in primary brain tumors due to favorable CNS penetration, with a CSF-to plasma ratio of ~0.2 in humans [64]. We found previously that KL-50 (2a) displays anti-tumor activity in intracranial glioma tumor models [17], as expected based on its imidazotetrazine scaffold, and confirmed here in a TMZ-R glioma PDX. Interestingly, several tumor types we identified as being commonly MGMT-silenced frequently metastasize to the brain. For example, 30%–43% of NSCLC cases develop brain lesions of which 22% present upon initial diagnosis [65]. Similarly, brain metastases occur in 10–40% of cutaneous melanoma [66]. Together, lung cancer, melanoma, and gastrointestinal cancers account for over half of all brain metastatic lesions [65]. These brain lesions significantly increase the difficulty of patient management and are a leading cause of morbidity. Stereotactic or whole brain radiotherapy and surgery are standard treatments for intracranial metastasis with various chemotherapies available. TMZ (1a) has historically been utilized given its high oral bioavailability and blood-brain barrier penetration [9,67]. However, efficacy of TMZ (1a) in brain metastases has been highly variable, likely due to a lack of testing for the presence of MGMT or MMR proteins [68]. Moch et. al described frequent MGMT promoter methylation and MGMT loss by immunostaining in brain lesions of lung, breast, kidney, and melanoma cancers with 32% of samples having homogeneous loss of MGMT protein staining and 43% of these samples having methylated MGMT promoters [69]. Interestingly, reports of increased genomic instability and/or increased mutations in DNA damage repair (DDR) proteins in brain metastases have been described [70-73]. Deficiencies in homologous repair and MMR, specifically, have been suggested to occur more often in instances of distant metastasis to the brain [73], which would be expected to abrogate TMZ (1a) activity. Thus, KL-50 (2a) may hold promise for the treatment MGMT-silenced brain metastases, especially when additional DDR pathways, i.e. MMR, are lost, and warrants further investigation.

Our data indicate KL-50 (2a) has similar efficacy as MTZ (3a) in treating MGMT− xenograft tumors without equivalent weight loss in murine models. The systemic toxicity associated with MTZ (3a) is likely related to the short half-life of O⁶ClEtG (3d) and subsequent formation of DNA ICLs and DNA-protein crosslinks in healthy MGMT+ cells, thus degrading the MGMT-dependence of MTZ (3a) [18]. Furthermore, two other mechanisms potentially contribute to the toxicity of chloroethylating agents in MGMT+ cells: (1) chloroethyl lesions at other non-O⁶G sites may progress to DNA crosslinks via direct attack from the opposing base [74]; and (2) MGMT-DNA protein crosslinks may arise following MGMT-mediated repair of O⁶ClEtG through an episulfonium ion intermediate [18,75]. Importantly, these mechanisms are not likely to occur with fluoroethyl adducts. We note that weight loss was observed in a subset of tumor xenograft models treated effectively with KL-50 (2a), despite it not causing weight loss in non-tumor bearing mice or in certain other xenograft models. These results suggest a model-specific interaction between drug treatment and the cachexia-inducing propensity of certain xenograft models, potentially mediated by inflammatory or metabolic signaling factors released by tumor cells [76]. Finally, while both TMZ (1a) and MTZ (3a) are associated with hematologic adverse effects in humans [20-23,77], we did not detect significant differences in blood counts and bone marrow cellularity following a single cycle of drug dosing. It is possible that additional dosing cycles may lead to detectable hematologic toxicity, though these experiments may also be limited by the intrinsic limitations of mouse models to assess bone marrow toxicity with DNA-damaging agents [78].

Biomarker-based, tumor type-agnostic strategies have emerged over the past several years as a promising and viable approach toward more effective personalized medicine [79]. The data presented in this study validate, through extensive in vitro screening and rigorous in vivo models, a novel drug class in targeting the loss of MGMT as a pan-cancer biomarker and in an MMR-independent manner. Altogether, these preclinical data build a strong foundation for investigating 2-fluoroethylating agents in appropriate clinical settings in both GBM and selected extracranial malignancies.