The chemotherapeutic drug CX-5461 is a potent mutagen in cultured human cells

Abstract

The chemotherapeutic agent CX-5461, or pidnarulex, has been fast-tracked by the United States Food and Drug Administration for early-stage clinical studies of BRCA1-, BRCA2- and PALB2-mutated cancers. It is under investigation in phase I and II trials. Here, we find that, although CX-5461 exhibits synthetic lethality in BRCA1-/BRCA2-deficient cells, it also causes extensive, nonselective, collateral mutagenesis in all three cell lines tested, to magnitudes that exceed known environmental carcinogens.

Fig. 1. CX-5461 induces heavy mutagenesis, leaving distinctive mutational signatures in hTERT-immortalized RPE1 cells.

a, Mutation accumulation experiment in which isogenic RPE1-BRCA1^–/–, -BRCA2^–/– and control cells were treated with compounds of interest (PDS, ETO, CX-5641) or vehicle control (DMSO) repeatedly, over ~35 days and allowed to recover. Subsequently, two to four independent subclones were isolated per treatment per genotype and expanded for WGS. b, De novo mutation counts. Bars are mean ± s.e.m., n = 2–4 independent subclones per treatment per genotype (Supplementary Table 2). Two-tailed Studentʼs t test was used to calculate P values. c, SBS, DBS and small indel signatures (InD) of CX-5461. d, Prevalence of signatures across different treatments and genotypes. SBS-HRd (substitution signature previously reported as SBS3 (associated with HRd) was averaged from SBS-BRCA1 and SBS-BRCA2 (Extended Data Fig. 1b); InD-HRd was averaged from InD-BRCA1 and InD-BRCA2 (Extended Data Fig. 1f).

Fig. 2. Diverse mechanisms underpin synthetic lethality of CX-5461 and its mutagenicity.

a, Depletion of CX-5461 mutations at and around predicted G4s. The gray line represents simulated mutations controlling for trinucleotide context and proximity to original mutation (within 10 kb); the red line shows depletion of actual mutations. b, Nucleosome density for SBS-CX-5461 mutations. the gray line shows the distribution predicted by simulation if mutations were distributed randomly; the dark blue line shows average nucleosome signal for real mutations. c, Normalized SBS-CX-5461 mutations across cell cycle, from early to late replication timing regions (separated into deciles, left to right). Purple dots and error bars represent the mean ± s.d. of predicted SBS-CX-5461 mutations from n = 100 bootstrapped replicates. Green bars represent the distribution of observed substitution mutations from n = 4 subclones treated with CX-5461. d, Transcriptional strand asymmetry of SBS-CX-5461 mutations (Supplementary Table 3). e, Percentage of possible stop gain, missense, synonymous and splice site mutations based on SBS-CX-5461 mutation contexts against COSMIC Cancer Gene Census Tier 1 and 2 cancer genes (Supplementary Table 5).

Extended Data Fig. 1. Mutational signatures of CX-5461 in isogenic hTERT-RPE1 cells.

a. Distinguishing de novo mutational profiles of experimental subclones from controls. Light green, pink, and blue error bars (left to right) depict the mean ± 3SD of cosine similarities between n = 100 bootstrapped control profiles and the control mutational profile aggregated from n = 4 DMSO-treated control subclones, of respective mutation types with increasing mutation counts. The x axis displays the mutation counts for respective mutation classes. See Methods for details. b. Single base substitution (SBS) signatures of gene knockouts and CX-5461 in different knockout backgrounds. Background signature was derived from untreated RPE1 cells. c. Heatmap showing cosine similarities between experimental SBS signatures. d. Cosine similarities comparing SBS-CX-5461 and SBS-HRd to reference SBS signatures. e. in silico permutation to assess whether DBS-CX-5461 is a chance occurrence due to high mutation burden given SBS-CX-5461 pattern. DBS, double base substitution. f. Small insertion and deletion signatures associated with homologous recombination deficiency (HRd), etoposide (ETO), and CX-5461 exposure. InD-BRCA1 and InD-BRCA2 were identical (cosine similarity, 0.99), and hence averaged as InD-HRd. Background signature was derived from untreated RPE1 cells. g. Heatmap showing cosine similarities between experimental indel signatures (InDs).

Extended Data Fig. 2. Drug sensitivity profiling by CellTiter-Glo cell viability assay.

Drug sensitivity profiling of isogenic RPE1-BRCA1^−/−, BRCA2^−/−, and control cells to CX-5461, topoisomerase II poison, etoposide (ETO), and G-quadruplex stabilising compound, pyridostatin (PDS) confirmed synthetic lethality of CX-5461 in RPE1-BRCA1^−/−, BRCA2^−/− cells. RPE1-LIG4^−/− was used as a positive control. Cells were also profiled against two other therapeutic agents commonly used for the treatment of BRCA1/2-mutated cancers, cisplatin, and olaparib. Data are mean ± standard errors (error bars), n = 3 independent biological replicates. All comparisons were made against WT. Two-tailed Wilcoxon signed-rank test was used to calculate P values.

Extended Data Fig. 3. Validation of CX-5461 signatures in alternative cellular models.

a. Aggregated whole-genome mutational profiles of CX-5461-treated HAP1 subclones (n = 2). b. Mutation frequencies normalized by the total duplex bases per sample across different compounds in HAP1 (left) and human induced pluripotent stem cells (hiPSC) (right) (n = 1 per treatment arm). Mutation frequency fold-increases were calculated against respective untreated control (bar top). BaP, benzo(a)pyrene; ETO, etoposide; PDS, pyridostatin. c. Trinucleotide spectrum plots for treated bulk cells by duplex sequencing. d. Unsupervised hierarchical clustering of mutational spectra (six mutation types) collapsed from c. using (1-cosine similarity) as distance matrix. BaP, benzo(a)pyrene; Cis, cisplatin; ETO, etoposide; PDS, pyridostatin.