Main

The APOBEC (apolipoprotein B editing enzyme catalytic polypeptide-like) family of polynucleotide cytidine deaminase enzymes fulfil various physiological functions in humans, varying from antibody gene diversification to innate immunity (Harris and Liddament, 2004; Conticello, 2008). APOBECs function in innate immunity by deaminating single-stranded DNA (ssDNA) replication intermediates of viral pathogens (retro-, hepadna-, papilloma-viruses), inhibiting the retrotransposition of L1 and Alu elements and mediating the clearance of foreign DNA through deamination-dependent mechanisms (Jarmuz et al, 2002; Burns et al, 2015). In addition to these functions in normal physiology, APOBEC enzymes have also been implicated in cancer pathogenesis. The expression levels of several APOBEC enzymes, including APOBEC3B, are altered in tumours (Ding et al, 2011; Nik-Zainal et al, 2012a; Burns et al, 2013a; Alexandrov et al, 2013; Leonard et al, 2013; Henderson et al, 2014; Rebhandl et al, 2014; Saraconi et al, 2014; The Cancer Genome Atlas Research Network, 2014). Upregulation of APOBEC3B in tumours has been associated with localised clusters of strand-specific cytosine to thymine (C-to-T) or cytosine to guanine (C-to-G) mutations (kataegic mutations) that tend to occur within TCW sequence motifs (W=adenine or thymine) and at rearrangement breakpoints (Nik-Zainal et al, 2012a; Roberts et al, 2012; Burns et al, 2013a, 2013b; Drier et al, 2013). This kataegic mutational signature has been identified in multiple cancers, including bladder, breast, cervix, head/neck, thyroid, lung, and several of the blood (Nik-Zainal et al, 2012a; Alexandrov et al, 2013). It seems likely that the ability of APOBEC3B to deaminate cytosine residues to uracil (C-to-U) in ssDNA contributes to this kataegic process (Helleday et al, 2014; Chan and Gordenin, 2015; Swanton et al, 2015). It has been suggested that APOBEC3B-induced mutagenesis occurs primarily within lagging strand templates arising during DNA replication (Bhagwat et al, 2016; Haradhvala et al, 2016; Hoopes et al, 2016; Seplyarskiy et al, 2016) or on stretches of ssDNA exposed during the resection stage of double-strand break repair or at stalled or collapsed replication forks (Nik-Zainal et al, 2012a; Roberts et al, 2012). The subsequent repair of uracil residues can lead to C-to-T transition mutations or C-to-G transversions, depending on the DNA repair pathway used to process the uracil lesions (Simonelli et al, 2005; Roberts et al, 2012). Kataegic clustering of C-to-T and C-to-G mutations in the APOBEC3B-preferred TCW motif has been recapitulated in yeast cells expressing human APOBEC3B (Nik-Zainal et al, 2012a; Taylor et al, 2013) and C-to-T mutations accumulate in breast tumour cell lines with upregulated APOBEC3B expression (Burns et al, 2013a; Akre et al, 2016).

The causes of elevated APOBEC3B expression are not fully understood, but it is now established that in human papillomavirus (HPV)-positive cancers of the head/neck and cervix, the HPV E6/E7 oncoprotein causes upregulation of APOBEC3B both at the mRNA and enzymatic activity level (Ohba et al, 2014; Vieira et al, 2014). BK polyomavirus has also been shown to upregulate APOBEC3B in kidney cells, which might explain elevated APOBEC3B levels in urothelial cancers (Verhalen et al, 2016). Furthermore, amplification of the ERBB2 oncogene and loss of PTEN tumour-suppressor function in breast tumours has been proposed to cause replicative stress, which in turn causes an increase in APOBEC3B transcription via an ATR/CHEK1-dependent pathway (Kanu et al, 2016). Finally, the protein kinase C/NFKB pathway can also induce transcriptional upregulation of APOBEC3B (Leonard et al, 2015).

Here we show that, while induction of APOBEC3B expression elicits a DNA damage response (DDR) and G2/M cell cycle arrest, the cell cycle arrest can be abrogated by loss of p53 function, formally identifying p53 dysfunction as a molecular route by which tumour cells might tolerate APOBEC3B activity. Using whole-genome sequencing (WGS), we also show that, in p53-defective cells, APOBEC3B expression drives an increase in the genomic uracil load and causes kataegic-like mutational signatures. Finally, we show that high-level APOBEC3B expression causes hypersensitivity to small-molecule inhibitors that target the DDR, suggesting that APOBEC3B overexpression can impart targetable vulnerabilities upon cells.

Materials and methods

Cell lines

293-A3B and 293-GFP cell lines have been previously described (Akre et al, 2016). Parental 293-A3B cells were infected with GIPZ TP53 shRNA lentivirus (Dharmacon, GE Healthcare, Little Chalfont, UK) to generate constitutive silencing of TP53 mRNA in the cells. Seventy-two hours after viral infection, cells were selected in 1 μg ml−1 puromycin to remove non-transduced cells, and clones were isolated after single-cell sorting with BD FACSAria II SORP (BD Biosciences, San Jose, CA, USA).

CRISPR/Cas9-mediated gene knockout

A single guide RNA (sgRNA) was designed to target the 5′-ACTTCCTGAAAACAACGTTC-3′ site in exon 3 of TP53 (Supplementary Figures S3B–E). The sgRNA was ligated into a lentiviral CRISPR vector, which allows for dual expression of the sgRNA and Cas9 endonuclease using the protocol from the Zhang Laboratory (Sanjana et al, 2014). This construct was cloned by annealing oligos 5′-CACCGACTTCCTGAAAACAACGTTC-3′ and 5′-GAACGAACGTTGTTTTCAGGAAGTC-3′. This vector was used to produce transducing virus and create a 293-A3B-Δp53 clone following a previously described protocol (Anderson and Harris, 2015).

Clonogenic survival assay

293-A3B and 293-A3B-Δp53 cells were seeded into six-well plates (500 cells per well) (Supplementary Figures S3C and D). The cells were treated with either mock (H2O) or doxycycline (dox; 1 μg ml−1) 8 h after plating. Once colonies of >50 cells formed in the 293-A3B mock-treated wells (∼9 days), the plates were washed with PBS and stained with 50% methanol and 0.5% crystal violet, and the colonies were counted by hand.

Western blotting

Cells were collected, washed with PBS and resuspended in lysis buffer containing 1 mM sodium vanadate, 20 mM sodium fluoride, 1% Igepal CA-630, 0.0225% Triton X, 1 × cOmplete ULTRA EDTA-free protease inhibitor cocktail tablet (Roche, Welwyn Garden City, UK) and 1 × PhosSTOP phosphatase inhibitor cocktail tablet (Roche). To enhance cell lysis, sonication was used. The insoluble fraction was removed by centrifugation. Protein concentration was measured with Bio-Rad Protein Assay (Bio-Rad, Cambridge, UK) to ensure equivalent total protein load per lane. Proteins were denatured by boiling for 10 min and resolved under reducing conditions on 4–12% Bis-Tris gels (NuPAGE, Thermo Fisher Scientific, Paisley, UK). Proteins were transferred to nitrocellulose membrane (Amersham Protran, GE Healthcare, Little Chalfont, UK). Membranes were blocked with either 5% (w/v) non-fat dried milk or mixture of 1% (w/v) non-fat dried milk and 1% (w/v) bovine serum albumin. Primary antibodies used were: γH2AX (Ser139, Cell Signaling, Danvers, MA, USA; used at 1 : 3000 (v/v) dilution), pRPA32 (S4/S8, Bethyl Laboratories, Montgomery, TX, USA; 1 : 1000), APOBEC3B (5210-87-13, (Leonard et al, 2015), 1 : 1000 and Abcam, Cambridge, UK; 1 : 5000), GFP (Cell Signaling, 1 : 5000), pCHK1 (Ser345, Cell Signaling, 1 : 1000), pCHK2 (Thr68, Cell Signaling, 1 : 1000), CHK1 (Santa Cruz, Santa Cruz, CA, USA 1 : 1000), CHK2 (Cell Signaling, 1 : 1000), PARP1 (Santa Cruz, 1 : 1000), p-p53 (Ser15, Cell Signaling, 1 : 1000), p53 (DO-I clone, Sigma, 1 : 1000), RPA1 (RPA-70, Abcam, 1 : 1000), Histone H3 (Cell Signaling, 1 : 1000), Actin (Sigma, Poole, UK; 1 : 10 000), or β-tubulin (1 : 10 000, Sigma). After applying horseradish peroxidase-coupled secondary antibodies (Amersham Pharmacia Biotech, Little Chalfont, UK), immunoreactive bands were visualised by ECL chemiluminescence reagent (Novex, Thermo Fisher Scientific, Little Chalfont, UK), and images were developed onto Amersham Hyperfilm (GE Healthcare, Little Chalfont, UK). Expression levels were quantified by using the Image J 1.50i software (NIH, Bethesda, MD, USA). In Supplementary Figure S3B, cells were collected, washed with PBS, and resuspended in 100 ml reducing sample buffer per one million cells (0.5M Tris-HCl pH 6.8, 1% 2-mercaptoethanol, 10% sodium dodecyl sulfate, 50% glycerol). Proteins were denatured by boiling samples for 20 min, and resolved by SDS–polyacrylamide gel electrophoresis. Proteins were transferred to a PVDF-FL membrane (Millipore, Watford, UK) and blocked in 5% milk in phosphate-buffered saline (PBS). Primary antibodies were incubated in PBS.0.1% Tween 20, and secondary antibodies were incubated in PBS.0.1% Tween 20.0.2% SDS. The membrane was imaged using a LI-COR Odyssey instrument. The primary antibodies used were: p53 (DO-1 clone, Santa Cruz, 1 : 1000) and b-actin (Cell Signaling, 1 : 2000). The secondary antibodies used were: IRDye 800CW goat anti-rabbit (LI-COR, Cambridge, UK; 1 : 10 000) and Alexa Fluor 680 goat anti-mouse (Molecular Probes, Paisley, UK, 1 : 10 000).

Uracil incorporation assay

DNA was purified using the MasterPure Complete DNA and RNA Purification Kit and the DNA Clean & Concentrator-5 Kit (Zymo Research, Irvine, CA, USA) from 292-A3B and 293-A3B-p53 cells exposed to 100 and 1000 ng ml−1 ml−1 dox (72 h). DNA was prepared according to the manufacturer’s instructions. Uracil quantifications for the 293-A3B and 293-A3B-p53 DNA samples were performed as in described in Rona et al (2016).

Cell cycle analysis

293-A3B and 293-A3B-p53 cells were either exposed to 0 or 100 ng ml−1 dox for 48 h before fixing in 70% (v/v) EtOH. Cells were stained for DNA synthesis using the Click-iT EdU Alexa Fluor 647 Flow Cytometer Assay Kit (Molecular Probes, Thermo Fisher Scientific, Loughborough, UK), according to the manufacturer’s protocol. In addition, the cells were stained with anti-phospho-Histone H3 (Ser10) antibody that specifically recognises M phase cells (1 : 200, Merck Millipore, Watford, UK). Fluorescence labeling was done with a Per-CP conjugated antibody (1 : 30, Stratech, Newmarket, UK). The nucleotide analogue 5-ethynyl-2′-deoxyuridine (EdU) was conjugated with Alexa Fluor 647 azide and DNA content was measured by addition of 4′,6-diamino-2-phenylindole (DAPI, 1 : 10 000, Molecular Probes, Thermo Fisher Scientific). Measurements took place on a BD LSR II SORP flow cytometer (BD Biosciences) equipped with a 404 nm violet laser, a 488 nm blue laser and a 633 nm red laser. Cell population was gated in a FSC/SSC dot plot and doublets were gated out based on a DAPI area/width dot plot. The single-cell populations were further analysed regarding its cell distribution. G1, S and G2/M phase cell populations were defined in a DAPI/EdU-Alexa Fluor 647 dot plot and G2/M phase cells were further separated in a DAPI/Per-CP dot plot. For quantification, BD FACSDiva software (BD Biosciences) was used.

Whole-genome sequencing

293-A3B-p53 cells were exposed to either 0 or 1000 ng ml−1 dox for 14 days followed by single-cell sorting with BD FACSAria III (BD Biosciences). Clones were expanded in normal growth media. DNA from pelleted cells was isolated using the Puregene Cell and Tissue DNA Isolation Kit (Qiagen, Manchester, UK) and WGS of DNA samples was performed by the Beijing Genome Institute (BGI). Libraries were sequenced using a HiSeq X Ten sequencer (Illumina, San Diego, CA, USA), acquiring approximately 600 million 150 bp paired-end reads per sample. Sequences were aligned to the human reference genome (GRCh37) using bwa-mem (http://arxiv.org/abs/1303.3997). PCR duplicates were removed prior to further processing and variant detection. Median depth of coverage ranged from 27 reads to 32 reads and >97.9% of the mappable genome was covered by five reads or more. Base quality recalibration and realignment was performed using the GATK (v3) (https://software.broadinstitute.org/gatk/). Mutation calling was performed using Mutect (https://software.broadinstitute.org/gatk/). Data were deposited on NCBI Short Read Archive (http://www.ncbi.nlm.nih.gov/Traces/sra/sra.cgi) using the accession number SRP090739.

Small-molecule inhibitor assays

293-A3B-p53 cells were plated on six-well plates (500 cells per well) 24 h before starting small-molecule inhibitor exposure and inducing APOBEC3B with dox. Medium containing 0 or 1000 ng ml−1 dox was added with inhibitors. Small-molecule inhibitors were supplied by Selleckchem, Munich, Germany, with the exception of CCT241533 (Tocris Bioscience, Bristol, UK), SAR020106 (the kind gift of Professor Ian Collins, ICR) and cisplatin (Sigma). Inhibitors and dox-containing media was changed every 3 days, and cell viability was estimated after 14 days using CellTiter-Glo reagent (Promega, Madison, WI, USA).

RT-qPCR analysis

RNA was isolated using Qiagen RNeasy Mini Kit (Qiagen, Manchester, UK) according to the manufacturer’s instructions, and cDNA was generated by reverse transcription reaction using the Omniscript Kit (Qiagen) according to the manufacturer’s instructions. TaqMan gene expression assay was used to analyse relative APOBEC3B mRNA expression levels. APOBEC3B expression was normalised to GAPDH mRNA expression (Applied Biosystems, ThermoFisher Scientific, Loughborough, UK).

Statistical analysis

Two-sided t-test was used for single inhibitor tests and two-way ANOVA for the dose/response curves. Statistics were calculated using GraphPad Prism version 6.00 for Mac OS X (GraphPad Software, La Jolla, CA, USA; http://www.graphpad.com).

Results

Inactivation of TP53 abrogates the cell cycle checkpoints elicited by APOBEC3B expression

To explore the molecular phenotypes associated with elevated APOBEC3B expression, we used human (HEK) 293 cells encompassing a chromosomally integrated dox-inducible APOBEC3B-GFP fusion cDNA construct (293-A3B cells (Akre et al, 2016)). Consistent with previous reports (Burns et al, 2013a; Lackey et al, 2013; Taylor et al, 2013; Akre et al, 2016), we found that APOBEC3B expression caused induction of a DDR characterised by γH2AX and RPA phosphorylation, biomarkers of replication fork stress, which was absent from dox-exposed 293-GFP cells lacking APOBEC3B construct (Figure 1A). We also noted increased levels of CHEK2 Thr68 phosphorylation normally associated with the G2/M cell cycle checkpoint (Reinhardt and Yaffe, 2009), as well as CHEK1 Ser345 phosphorylation upon induction of APOBEC3B expression (Supplementary Figures S1, S2A and B). Consistent with these observations, we found that induction of APOBEC3B reduced EdU incorporation (Figure 1B; 10.6% EdU-negative cells in APOBEC3B-positive cells vs 1.8% in EdU-negative cells), suggesting that replication in S phase was impaired. Overexpression of APOBEC3B also resulted in G2/M cell cycle arrest (Figure 1B); in the absence of APOBEC3B 7.6% of cells were in G2 cells and 1.4% in M phase, while the expression of APOBEC3B caused these proportions to rise to 21% (G2) and 5.7% (M phase) (Figure 1B). This induction of APOBEC3B expression also impaired the proliferative capacity of cells (Figure 1C).

Figure 1
figure 1

TP53 silencing abrogates APOBEC3B-induced cell cycle arrest and cell death. (A) Induced APOBEC3B expression results in elevated phosphorylation of γH2AX(Ser139) and RPA32(S4/8) in 293-A3B cells. 293-GFP cells were included as a negative control and express GFP but not APOBEC3B upon dox exposure. (B) Induced APOBEC3B expression results in replication stress and arrests cells in G2/M phase. FACS profiles in cells labelled with EdU (left panels) or anti-phospho histone H3 antibody (‘histone H3’, right panels) are shown. Fraction (%) of cells in non-active S phase are shown in the left-hand panels. Fraction (%) of anti-phospho histone H3 +ve and −ve cells are shown in the right-hand panels. (C) APOBEC3B induction drives cell inhibition in vitro, which can be rescued by TP53 silencing. Growth curves of 293-A3B and 293-A3B-p53 cells treated with DMSO and 100 or 1000 ng ml−1 of dox, was measured using CellTiter-Glo reagent are shown. Cell growth was analysed on the indicated day after inducing APOBEC3B-GFP expression. Each data point represents eight replicates. Two-way ANOVA was used to calculate P-values. Error bars indicate s.d. for each measured group. (D) Induced APOBEC3B expression activates the apoptotic pathway. APOBEC3B induction results in an increase in cleaved PARP1 and phosphorylated p53 on Ser15 levels 48–72 h after dox induction in 293-A3B cells as shown by western blotting. (E) TP53 silencing abolishes p53 expression while maintaining inducible APOBEC3B and GFP expression, leading to elevated phosphorylation of H2AX (Ser139, γH2AX) and RPA32 (S4/8) in 293-A3B-p53 cells 48–72 h and also 5 days after inducing APOBEC3B with 0.1–1000 ng ml−1 of dox. (F) TP53 silencing alleviates APOBEC3B-driven replication stress and G2/M arrest in 293-A3B-p53 cells as shown by FACS profiling. Legend as per panel (B). (G) APOBEC3B-GFP induction increases genomic uracil (P=0.0009) in 293-A3B-p53 cells when compared with non-induced 293-A3B-p53 cells. Each data point represents the mean of at least six replicates. A two-sided Student’s t-test was used to estimate statistical significance. Error bars indicate s.d. for each measured groups. Method described in (Rona et al, 2016). In panels (A, B and D), APOBEC3B was induced with 1 and 100 ng ml−1 of dox for 48 or 72 h, DMSO was used as control (0 ng ml−1).

Full size image

DDR often results in the induction of p53 activity and apoptosis (Lakin et al, 1999; Tibbetts et al, 1999). We found that APOBEC3B induction in 293-A3B cells led to elevated p53 levels and increased phosphorylation of p53 at residue Ser15 and PARP1 cleavage, indicative of an apoptotic response (Kaufmann et al, 1993) (Figure 1D, Supplementary Figures S2C and D). To assess whether inhibition of cell proliferation caused by APOBEC3B expression was mediated by p53 activity, we depleted p53 in 293-A3B cells using constitutively expressed shRNA, generating a 293-A3B-p53 clone. In 293-A3B-p53 cells, the induction of APOBEC3B did not cause a proliferative defect, as it did in cells expressing p53 (Figure 1C), but still caused a DDR, as measured by induction of γH2AX and phosphorylation of RPA (Figure 1E). The level of APOBEC3B and γH2AX expression induction was similar over a range of concentrations of dox used (10–1000 ng ml−1 of dox, Supplementary Figure S3A). We also found that inactivation of p53 in 293-A3B cells, via CRISPR-Cas9-mediated mutagenesis, also reversed the proliferative defect caused by APOBEC3B induction (Supplementary Figures S3B–E), confirming the observations made in 293-A3B-p53 cells. Consistent with the absence of a proliferative defect, APOBEC3B expression in 293-A3B-p53 cells did not induce G2/M arrest (Figure 1F). However, we found that induction of APOBEC3B expression caused similar levels of uracil incorporation into genomic DNA in 293-A3B and 293-A3B-p53 cells (Figure 1G), suggesting that the failure to observe cell cycle arrest in 293-A3B-p53 cells was not due to reduced APOBEC3B activity. Taken together, this data formally demonstrated that loss of p53 function allows cells to tolerate elevated APOBEC3B expression in the face of an on-going DDR.

APOBEC3B expression increases APOBEC signature mutations and a kataegis-like process in p53-defective cells

293-A3B-p53 cells continued to proliferate in the presence of high-level APOBEC3B expression, despite displaying elevated levels of genomic uracil incorporation and on-going activation of DDR biomarkers. This led us to assess whether APOBEC3B expression in p53-defective cells would ultimately lead to a kataegis-like DNA mutational pattern. To do this, we exposed 293-A3B-p53 cells to 1000 ng ml−1 of dox for 14 continuous days to chronically upregulate APOBEC3B and then isolated three individual subclones (clones 1–3) from single cells as well as a control subclone isolated from 293-A3B-p53 cells cultured for 14 days in the absence of dox/APOBEC3B induction (control clone). Each of these clones was subjected to WGS (average depth of 30 ×). The genomic sequence in the control clone was used as the reference genome. Using these data, we identified mutations that had arisen in 293-A3B-p53 clones 1–3 but were absent in the control clone; variant calls for clones 1–3 are presented in Supplementary Tables S1–S3. We then estimated the extent of kataegis by calculating intermutational distances (IMD; Nik-Zainal et al, 2012a) for each clone, illustrating this data as rainfall plots (Figure 2A and Supplementary Figure S4). A computational definition of a kataegic mutational cluster is where six or more consecutive mutations occur with an IMD of ≤1000 bp (Nik-Zainal et al, 2012b). Our IMD analysis identified a kataegic-like signature in all three clones derived from long-term APOBEC3B expression, when compared with the control clone (Figure 2A, Supplementary Figure S4). In general, we observed an average ninefold increase in kataegic mutational clusters in clones 1–3 when compared with the control clone. Altogether 348–422 kataegic events (IMD≤1000 bp) with mean IMD 155.50–175.34 bp were analysed for clones 1–3 (clone 1: 369 events with mean IMD 155.50 bp; clone 2: 348 events with mean IMD 175.34 bp; clone 3: 422 events with mean IMD 166.17 bp). These kataegic mutations represented 4.6–5.6% of all mutated cytosines. The most frequent cytosine mutation events in clones 1–3 were C-to-T transitions (>45% in Figure 2B; Supplementary Figure S5A) characteristic of APOBEC3B overexpression. This was more pronounced in kataegic mutational clusters, (IMD≤1000 bp) where >57% of the mutated cytosines were C-to-T transitions (Figure 2C; Supplementary Figures S5B and S6).

Figure 2
figure 2

Induced APOBEC3B expression drives a kataegis-like mutation signature and leads to an increase in overall C-to-T mutation load in the TCW sequence context. (A) APOBEC3B induction causes genome-wide mutation clustering that was present only in dox-treated 293-A3B-p53 cells. Mutation clusters were measured by their IMD (upper panel). Each dot represents one mutation. y Axis represents intermutation distance, that is, the genomic distance of each mutation from the previous one. x Axis represents position in human genome. The horizontal line (1000 bp) shows the cutoff for mutations with IMD≤1000 bp. (B) Cytosine mutations present only in dox-treated 293-A3B-p53 cells (clone 1) were mainly C-to-T transitions (red). (C) Cytosine mutations in kataegic clusters (mutations with IMD≤1000 bp) were mostly C-to-T transitions (red). (D) Cytosine (C-to-G and C-to-A) mutations in dox-treated 293-A3B-p53 cells were enriched in TCW trinucleotide contexts (marked with asterisk). Panels (A–C) show clone 1 as an example; other clones shown in Supplementary Figure S5.

Full size image

We also noted that chronic APOBEC3B expression in 293-A3B-p53 cells resulted in enrichment of cytosine mutations in the APOBEC3B-preferred TCG motif (Burns et al, 2013a; Burns et al, 2013b; Leonard et al, 2013) and kataegic TCW trinucleotide motif (Nik-Zainal et al, 2012a). Surprisingly, an enrichment of C-to-A mutations in these motifs was also seen. This is not characteristic of APOBEC3B-induced mutations and could be explained by other mutational processes occurring, such as transcription-coupled nucleotide excision repair, which has been associated with mutational signatures dominated by C-to-A mutations (Alexandrov et al, 2013). It is therefore possible that other mutational processes are active in the APOBEC3B-induced 293-A3B-p53 cells. Most of the mutated cytosines in clones 1–3 were C-to-T transitions (45.4–46.6% of C-to-T mutations compared with 35.2–36.1% of C-to-A and 17.3–18.8% of C-to-G mutations; Figure 2B, Supplementary Figure S5A) enriched in TCG motifs, a characteristic of APOBEC-mediated mutagenesis (Fisher’s exact test for mutated cytosines in TCG context: clone 1, P=1.1 × 10−3, OR=2.42; clone 2, P=1.0 × 10−7, OR=3.31; clone 3 P=9.0 × 10−5, OR=2.28). Furthermore, C-to-G and C-to-A mutations were enriched in kataegic TCW trinucleotides (Fisher’s exact test for mutated cytosines in TCW motif: C-to-A, P=4.70 × 10−9–5.57 × 10−13; C-to-G, P=3.65 × 10−4–8.95 × 10−8; C-to-T P=1.00 for all DOX1–3; Figure 2D, Supplementary Figure S7 and Supplementary Table S4). Taken together, our data suggested that prolonged APOBEC3B expression in p53-defective cells results in a kataegic-like mutational signature.

APOBEC3B expression sensitises cells to small-molecule DDR inhibitors

Although inactivation of p53 allowed cells to tolerate APOBEC3B expression, we hypothesised that the ongoing DDR caused by APOBEC3B (Figure 1E) might make cells vulnerable to small-molecule inhibitors that further perturb replication or inhibit processes connected to DNA repair and/or replication fork integrity. Hydroxyurea (HU) stalls replication forks and targets cells with existing replicative stress by inhibiting deoxyribonucleotide production and delaying replicative DNA polymerase progression (Zeman and Cimprich, 2014). We exposed 293-A3B-p53 cells to 50 μM HU and found that APOBEC3B expression caused a fourfold increase in the sensitivity to HU (Figure 3A; P<0.0001, two-sided Student’s t-test). We also found that induction of APOBEC3B in 293-A3B-p53 cells caused sensitivity to previously validated small-molecule inhibitors of either ATR (VX-970 (Hall et al, 2014) or AZD6738 (O’Connor, 2015), CHEK1 (SAR020106 (Walton et al, 2010)), WEE1 (AZD1775 (O’Connor, 2015)) or CHEK2 (CCT241533 (Anderson et al, 2011)) (Figures 3B–F, ANOVA P<0.0001 in each case). We also titrated dox-induced APOBEC3B expression and confirmed that even lesser degrees of induction of APOBEC3B expression caused sensitivity to the CHEK2 inhibitor CCT241533 (1–1000 ng ml−1 dox; Figure 3G; ANOVA P<0.0001 in each case). Many clinical PARP inhibitors cause cytotoxicity by ‘trapping’ PARP1 on DNA (Murai et al, 2012; Pettitt et al, 2013; Hopkins et al, 2015), an effect that likely impairs the progress of replication forks. We found that the expression of APOBEC3B caused moderate sensitisation to two chemically distinct clinical PARP inhibitors, olaparib (Figure 4A; ANOVA P<0.0001) and talazoparib (aka BMN 673, Figure 4B; ANOVA P<0.0001).

Figure 3
figure 3

APOBEC3B expression sensitizes cells to inhibitors targeting DDR-related kinases. APOBEC3B induction in 293-A3B-p53 cells causes sensitivity to: (A) 50 μM HU (Student’s t-test, P<0.0001), (B) the ATR inhibitor VX-970 (ANOVA, P<0.0001), (C) the ATR inhibitor AZD6738 (ANOVA, P<0.0001), (D) the CHEK1 inhibitor SAR020106 (ANOVA, P<0.0001), (E) the WEE1 inhibitor AZD1775 (ANOVA, P<0.0001), and (F and G) the CHEK2 inhibitor CCT241533 (ANOVA, P<0.0001). In panels (A–F), 14-day inhibitor exposures were used with or without 1000 ng ml−1 dox to induce APOBEC3B. In panel (G), the indicated concentrations were used. Cell viability was assessed using CellTiter-Glo reagent at the end of each treatment course. Each data point represents the mean of three replicates. Error bars represent s.d.

Full size image
Figure 4
figure 4

APOBEC3B expression sensitizes cells to PARP inhibitors. APOBEC3B induction in 293-A3B-p53 cells causes sensitivity to the clinical PARP inhibitors, olaparib (A, ANOVA, P<0.0001) and (B) BMN-673 (ANOVA, P<0.0001). Fourteen-day inhibitor exposures were used with or without 1000 ng ml−1 dox to induced APOBEC3B. Cell viability was assessed using CellTiter-Glo reagent at the end of each treatment course. Each data point represents the mean of three replicates. Error bars represent s.d.

Full size image

We next assessed whether drug combination strategies designed to target cells with replicative stress might also selectively target cells with elevated APOBEC3B expression. To do this, we assessed whether elevated APOBEC3B might sensitise cells to a platinum salt, cisplatin used in combination with an ATR inhibitor (Wang et al, 2012) or a PARP inhibitor used in combination with either an ATR or WEE1 inhibitor. We found that cells with induced APOBEC3B expression exhibited an enhanced level of sensitivity to each of these combinations. For example, the induction of APOBEC3B caused a >4000-fold increase in the sensitivity to a cisplatin/ATR inhibitor combination (Figures 5A and B; ANOVA P<0.0001 in each case), >40 000-fold increase in the sensitivity to an ATR/PARP inhibitor combination (Figure 5C; ANOVA P<0.0001 in each case) and 148-fold change in the sensitivity to PARP inhibitor/WEE1 inhibitor combination (Figure 5D; ANOVA P<0.0001 in each case). The profound sensitivity to the ATR/PARP inhibitor combination caused by APOBEC3B expression in 293-A3B-p53 cells led us to assess this combination in HCC1143 breast tumour cells, which exhibit a high level of APOBEC3B expression (Supplementary Figure S8). Compared with non-tumour breast epithelial MCF10A cells, HCC1143 cells were significantly more sensitive to ATR/PARP inhibitor combination (Supplementary Figure S9; ANOVA P<0.001 in each case). We do note, however, that although APOBEC3B expression was significantly elevated in HCC1143 cells compared with MCF10A cells (Supplementary Figure S8), factors other than APOBEC3B expression might explain the differences in ATR/PARP inhibitor sensitivity. Taken together, our observations suggested that elevated APOBEC3B expression might cause vulnerabilities in tumour cells that can be exploited by drugs targeting DDR kinase enzymes.

Figure 5
figure 5

APOBEC3B induction in 293-A3B-p53 cells causes sensitivity to DDR inhibitor combinations. (A and B) ATR inhibitors (VX-970, 10 nM and AZ20, 100 nM) potentiate the sensitivity of cisplatin in the presence of APOBEC3B induction (ANOVA, P<0.0001 as shown). (C) ATR inhibitor (10 nM) potentiates the sensitivity of olaparib (0–1000 nM) in the presence of APOBEC3B induction (ANOVA, P<0.0001 as shown). (D) Olaparib (100 nM) potentiates the sensitivity of WEE1 inhibitor in the presence of APOBEC3B induction (ANOVA, P<0.0001 as shown). Fourteen-day inhibitor exposures were used with or without 1000 ng ml−1 dox to induce APOBEC3B. Cell viability was assessed using CellTiter-Glo reagent at the end of each treatment course. Each data point represents the mean of three replicates. Error bars represent s.d.

Full size image

Discussion

Here we show that the proliferative defects imparted upon a cell by high-level APOBEC3B expression can be abrogated by inactivation of p53. It seems reasonable to suggest that this might explain why cancers with high levels of APOBEC3B expression and/or an APOBEC3-mutational signature frequently exhibit TP53 mutations (Burns et al, 2013a; Henderson et al, 2014; Silwal-Pandit et al, 2014). We also show that once the G2/M cell cycle arrest caused by elevated APOBEC3B is reversed via loss of p53, cells with prolonged APOBEC3B expression develop a kataegic mutational signature. Akre et al (2016) recently described the mutational consequences of intermittently induced APOBEC3B expression in 293 cells with active p53, rather than continuously expressed APOBEC3B in p53-defective 293-A3B-p53 cells, described here. In 293-A3B cells, Akre et al (2016) found that intermittent APOBEC3B expression caused an increase in C-to-T mutations in TCG motifs and an increase in the frequency of mutations at A/T base pairs. These observations are consistent with the hypothesis that genomic uracil residues caused by APOBEC3B activity might be repaired by error-prone, and thus mutation-inducing, processes. Furthermore, Akre et al (2016) found an enrichment for C-to-G mutations within TCW motifs in cells with intermittent APOBEC3B expression. Each of these observations are consistent with the mutational spectrum we observed in 293-A3B-p53 cells with induced APOBEC3B expression, notably the increase in C residue mutations within TCW motifs (Figure 2, Supplementary Figures S4–S7). In our 293-A3B-p53 cells with chronic APOBEC3B expression, we observed an increase in a kataegic mutational pattern, with a ninefold increase in clustered mutations (IMD≤1000 bp), the majority of these being C-to-T transitions (Figure 2, Supplementary Figure S4). In comparison, while pulsed APOBEC3B expression in 293-A3B cells clearly caused an increase in mutations likely caused by APOBEC3B, the kataegic clustering of mutations was not as profound. It is possible that this difference might be due to the loss of p53 in 293-A3B cells, with the loss of p53 either allowing cells to tolerate clustered mutations or the processes that cause these. However, we also note that there are important differences in experimental design in the work described in Figure 2, compared with the experiments described in Akre et al (2016) that might also explain the somewhat distinct mutational profiles. For example, in Akre et al (2016), the mutational spectrum in cells was assessed after 10 rounds of dox exposure; in the work presented here, we describe mutational consequences after continuous 14-day dox exposure.

Despite the proliferative capacity of p53-inactivated cells, we also found that p53-defective cells with elevated APOBEC3B expression are vulnerable to small-molecule DDR inhibitors, including ATR, WEE1, CHEK1 and CHEK2 inhibitors. The CHEK1 inhibitor sensitivity caused by APOBEC3B expression seen in this work is consistent with a correlation between APOBEC3B mRNA expression and sensitivity to the CHEK1 inhibitor CCT244747 in tumour cell lines, observed elsewhere (Kanu et al, 2016). The most notable drug vulnerabilities associated with APOBEC3B expression were DDR inhibitor combinations, including a combination of ATR plus PARP inhibitor now being assessed clinically (NCT02264678).

One simple model to explain our observations is that, while p53 inactivation allows cells to proliferate in the face of high-level APOBEC3B expression, APOBEC3B still causes a form of DNA damage, which under certain conditions such as the presence of DDR inhibitors might impair the fitness of cells. It seems possible that the γH2AX and RPA phosphorylation we observe in cells with elevated APOBEC3B, both when cells are p53 proficient or p53 defective (Figure 1), is indicative of replication fork stress caused by the excessive APOBEC3B activity, possibly on lagging strand DNA (Hoopes et al, 2016). Recent work has suggested that APOBEC3B activity is induced by replication fork stress (Kanu et al, 2016); our work suggests that APOBEC3B expression/activity might also contribute to replication fork stress as well as being caused by it. Replication fork stress, especially in the absence of p53 function, might generate an enhanced reliance upon DNA repair kinases such as ATR and CHEK1 that instigate signalling cascades that stabilise and repair replication forks, as well as inhibiting of processes such as new fork origin firing or the further accumulation of ssDNA that would otherwise exacerbate replication fork stress (Shechter et al, 2004; Petermann et al, 2010). This reliance on DNA repair kinases might extend to WEE1 and CHEK2, given their roles in preventing premature entry through the cell cycle in the face of a deleterious level of DNA damage (O'Connell et al, 1997; Reinhardt and Yaffe, 2009).

Whether these DNA repair kinase vulnerabilities extend to human tumours with elevated APOBEC3B expression remains to be seen, but we note that some of the inhibitors we describe here are already undergoing clinical trial assessment, including the ATR inhibitors AZD6738 (e.g., clinical trial identifiers NCT01955668, NCT02223923, NCT02264678), VX970 (NCT02567409, NCT02487095, NCT02567422) and the WEE1 inhibitor AZD1775 (e.g., NCT01164995 and others summarised in O’Connor, 2015), suggesting that retrospectively assessing tumoural APOBEC3B expression in these trials might be worthwhile. We also note that, although we have identified DDR inhibitor vulnerabilities associated with high-level APOBEC3B expression in 293-A3B-p53 cells, subsequent validation is required to extend these observations and to optimise how DDR inhibitors might be used to deliver profound, long-lasting antitumour responses in a significant proportion of patients with tumours with elevated APOBEC3B expression. Such a validation process might involve modulating APOBEC3B levels in relatively large panels of tumour cell lines and in in vivo models of cancer and then assessing drug sensitivity. Given the central role of p53 in controlling the cell cycle and DDR, it is also possible that the lack of p53 in 293-A3B-p53 cells influences some of the DDR inhibitors responses we have observed; it remains to be seen whether tumour cells without p53 dysfunction but which tolerate high-level APOBEC3B expression by other means also exhibit similar drug sensitivities.