The aflatoxin B₁-induced formamidopyrimidine adduct is repaired by transcription-coupled nucleotide excision repair in human cells

Abstract

The mycotoxin, aflatoxin B₁ (AFB₁), is a potent mutagen that contaminates agricultural food supplies. After ingestion, AFB₁ is oxidized into a reactive electrophile that alkylates DNA, forming bulky lesions such as the genotoxic formamidopyrimidine lesion, AFB₁-Fapy dG. This lesion is mainly repaired by nucleotide excision repair (NER) in bacteria; however, in humans the picture is less clear. We report a plasmid-based host cell reactivation assay containing a site-specific AFB₁-Fapy dG lesion and present evidence that this lesion is mainly repaired by transcription-coupled NER (TC-NER) in human cells. Using a combination of isogenic knockout cell lines and immortalized fibroblasts from xeroderma pigmentosum and Cockayne syndrome patients, we show that the TC-NER factors CSA, CSB, and UVSSA are required for efficient AFB₁-Fapy dG repair, while the global-genome NER protein, XPC, is dispensable. Furthermore, knockout of CSB or UVSSA impairs AFB₁-Fapy dG repair to a similar degree as knockout of the core NER nuclease, XPF. Our data indicate that TC-NER is the major repair pathway for AFB₁-Fapy dG adducts in human cells.

Graphical Abstract

Graphical Abstract.

Introduction

Aflatoxin B₁ (AFB₁) is a carcinogenic natural product produced by fungal species including Aspergillus flavus and Aspergillus parasiticus. These fungi contaminate agricultural food supplies such as corn and peanuts, especially in warm, humid climates. In developed countries, strict limits are placed on AFB₁ levels in food supplies; however, exposure to AFB₁ is a significant public health burden throughout Southeast Asia and sub-Saharan Africa where the climate favors growth of aflatoxin-producing Aspergillus species. Acute exposure to high levels of AFB₁ causes aflatoxicosis, which is primarily characterized by liver injury [1]. Although outbreaks of aflatoxicosis occasionally occur in endemic regions, chronic exposure to low levels of AFB₁ is far more common and is the major public health concern [2]. Epidemiological studies have linked AFB₁ exposure to the development of hepatocellular carcinoma (HCC), and this risk is compounded by other mutagenic agents that target the liver, including alcohol, hepatitis B, and hepatitis C [3, 4]. Over 50% of HCC cases in AFB₁-exposed individuals possess a characteristic G→T transversion mutation at codon 249 of the TP53 gene [5].

The primary route of AFB₁ exposure is ingestion, following which the toxin is transported to the liver and oxidized by cytochrome P450 enzymes to the exo-8,9-epoxide (Fig. 1A and Fig. 2A) [6, 7]. This epoxide alkylates DNA at the N7 position of guanine, forming the cation, AFB₁-N⁷-dG [8–10]. This lesion is unstable and either depurinates to form an abasic site or decomposes via hydrolysis of the imidazole ring into the stable, ring-opened 8,9-dihydro-8(2,6-diamino-4-oxo-3,4-dihydropyrimid5-yl-formamido)-9-hydroxyaflatoxin B₁ (AFB₁-Fapy dG; Fig. 1A). This adduct equilibrates between the α- and β-deoxyribose anomers in solution, although the β-anomer predominates in double-stranded DNA, where it intercalates and stabilizes the duplex via base stacking interactions [11, 12]. It is likely that the N⁷-AFB₁ adduct, abasic sites, and the AFB₁-Fapy dG lesion all contribute to the mutagenicity of AFB₁ [13]; however, several lines of evidence indicate that AFB₁-Fapy dG is the major mutagenic lesion. Rats treated with a single sublethal dose of AFB₁ rapidly accumulate AFB₁-N⁷-dG lesions in liver DNA; however, this adduct decays with a half-life of ∼7.5 h, while AFB₁-Fapy dG persists at high levels for at least 72 h [14]. Finally, AFB₁-Fapy dG induces G→T transversion mutations at a rate of 33% in Escherichia coli shuttle vectors, a mutation frequency that is 6 times greater than that of AFB₁-N⁷-dG [12]. The mutation frequency in primate cells is even higher, where AFB₁-Fapy dG induces G→T transversions at a frequency of 86% [15].

Figure 1.

Construction of a fluorescence-based assay for measuring repair of a site-specific AFB₁-Fapy dG adduct in mammalian cells. (A) Schematic showing preparation of AFB₁-Fapy dG adduct by oxidation of AFB₁ by dimethyldioxirane. (B) Cartoon illustration of BFP_AFB₁-Fapy reporter assay created in BioRender. Orange arrow, cytomegalovirus (CMV) promoter; larger blue and purple arrows, blue fluorescent protein (BFP) and mPlum reporter genes. (C) Analysis of BFP_AFB₁-Fapy by gel electrophoresis. Lane 1: ladder; lane 2: closed circular pMax_BFP; lane 3: pMax_BFP nicked with Nb.BtsI; lane 4: BFP_AFB₁-Fapy; lane 5: pMax_BFP incubated with five-fold molar excess of Neil1 for 1 h; lane 6: BFP_AFB₁-Fapy incubated with five-fold molar excess of Neil1 for 1 h. (D) Flow cytometry density plots of nucleotide excision repair (NER)-deficient XP2OS cells transfected with undamaged plasmid cocktail (pMax_BFP and pMax_mPlum; top three plots) or damaged plasmid cocktail (BFP_AFB₁-Fapy and pMax_mPlum; bottom three plots). The leftmost plot for each sample shows forward scatter area plotted against side scatter area. The gate on this plot encompasses the cells, which are then displayed on the other two plots. The middle plot shows the gating strategy used for identifying BFP-positive cells, and the rightmost plot shows the gating strategy for identifying mPlum-positive cells. A detailed strategy and gating hierarchy is provided in Supplementary Fig. S11. (E) Expression of BFP_AFB₁-Fapy in human cell lines. Each data point represents an independent experiment and error bars show the standard deviation.

Figure 2.

Immortalized human skin fibroblasts harboring mutations in NER genes are defective in repair of the BFP_AFB₁-Fapy reporter plasmid. (A) Illustration of FM-HCR plasmids used in Fig. 2 created in BioRender. Undamaged cocktail: pMax_GFP transfection control, pMax_BFP_2, and pMax_mOrange. Damaged cocktail: pMax_GFP transfection control, BFP_AFB1-Fapy, and mOrange_UV (randomly distributed UV damage induced by irradiation with 800 J/m² UV light). (B) Representative flow cytometry scatter plots of FM-HCR experiments in XP2OS (top four plots) and XP2OS + XPA (bottom four plots) cells transfected with damaged plasmid cocktail. The leftmost plot for each sample (forward scatter area versus side scatter area) is gated to separate cells from debris. Cells are then displayed on the other three plots and gated for expression of BFP, mOrange, or GFP. The number of cells within each gate is displayed on the plot. A detailed gating strategy is provided in Supplementary Fig. S12. (C) FM-HCR in XP2OS/XP2OS + XPA cell line pair. (D) FM-HCR in XP2YO (ERCC4 mutant) cells and XP2YO complemented with wild-type XPF or catalytically inactive mutant, XPF D676A. (E) FM-HCR in XP3BRSV (ERCC5 mutant) and cells complemented with XPG_WT or catalytically inactive mutant, XPG E791A. In panels (B)–(D), each data point is from an independent experiment (n = 3), error bars show the standard deviation, and P-values are from unpaired two-tailed t-test.

Survival of AFB₁-treated mouse embryonic fibroblasts is dependent upon Rev3l, which encodes the catalytic subunit of the translesion synthesis polymerase Pol ζ [16]. Pol ζ bypasses AFB₁-Fapy dG in vitro and primarily inserts dA opposite the lesion, which would lead to a G→T transversion in vivo [15]. These data suggest that translesion DNA synthesis across unrepaired AFB₁-Fapy dG lesions, although protective against cell death, causes mutations that eventually lead to HCC. The ability of human cells to repair AFB₁-Fapy dG is therefore critical for preventing mutagenesis in exposed individuals; however, the repair of a site-specific AFB₁-Fapy dG adduct has not been investigated in human cells. The NER pathway, which excises bulky DNA adducts including ultraviolet light (UV)-induced pyrimidine dimers, cisplatin intrastrand crosslinks, and benzo[a]pyrene adducts [17–19], is a likely candidate for AFB₁-Fapy repair [20]. Studies in model organisms lend support to this hypothesis: AFB₁-Fapy dG is repaired exclusively by NER in E. coli, while the base excision repair (BER) glycosylase Fpg is inefficient at removing AFB₁-Fapy dG in vitro and is not required for repair in bacterial cells [21]. Additionally, knockout of NER genes in Caenorhabditis elegans increases AFB₁-induced mutational burden, especially G→T transversions [22]. However, in AFB₁-treated mice, knockout of the Neil1 glycosylase increases AFB₁-Fapy dG levels in liver DNA and increases cancer risk [23], suggesting that Neil1-dependent BER may be a major contributor to AFB₁-Fapy dG repair in mammals.

Using the fluorescence-based multiplexed host cell reactivation (FM-HCR) method [24, 25], we report the first assay for measuring repair of a site-specific AFB₁-Fapy dG lesion in human cells. We find that NER-deficient cell lines are inefficient at repairing AFB₁-Fapy dG, with knockout of XPA or XPF almost completely ablating expression of the AFB₁-Fapy dG reporter assay, BFP_AFB₁-Fapy. Interestingly, the transcription-coupled NER (TC-NER) factors CSA, CSB, and UVSSA are required for efficient expression of BFP_AFB₁-Fapy, while the global-genome NER (GG-NER) factor, XPC, is not. Our data suggest that, in human cells, AFB₁-Fapy dG evades recognition by GG-NER and is mainly repaired by TC-NER.

Materials and methods

Cell lines

TK6 wild type and XPA-knockout cells were purchased from the TK6 Consortium in Kyoto (https://www.nihs.go.jp/dgm/tk6.html). XP2OS and XP2OS + XPA cells were generously shared by Walter Chazin. XP2YO and XP3BRSV cell lines were reported previously and generously shared by Orlando Scharer [26]. U2OS cells were purchased from ATCC, and XPA and XPF knockouts were generated from these cells (data presented in Fig. 3) and validated by western blot. CSB, XPC, and UVSSA-knockout cells (and corresponding wild-type cells) were previously reported [27] and validated by sequencing. These cells, along with previously reported XPF-KO cells [28], were used for the experiments in Fig. 4. CS3BE and CS1AN cell line pairs (experiments in Fig. 4) were generously shared by Vilhelm Bohr, GM15983 cell line pair was purchased from Coriell Institute.

Figure 3.

Knockout of NER factors XPA and XPF suppresses AFB₁-Fapy dG repair in human cell lines. (A) BFP_AFB₁-Fapy expression in TK6 wild-type and XPA-knockout (XPA^−/−) cells. (B) Western blot for XPA in U2OS wild type or XPA^−/− cells stably expressing either empty vector or XPA. (C) FM-HCR in U2OS cells from panel (B) transfected with undamaged cocktail (pMax_BFP, pMax_mOrange, and pMax_mPlum) or damaged cocktail (BFP_AFB1, mOrange_UV, and pMax_mPlum). (D) Representative flow cytometry plots from cells transfected with the damaged cocktail from panel (C). These plots show three different representations of data from the same sample to illustrate the gating strategy. The leftmost plot gated for BFP-expressing cells and the rightmost plot gated for mPlum-expressing cells. The title above the plot indicates the population being gated. (E) Western blot for XPF in wild-type U2OS cells transfected with non-targeting sgRNA (sgNT) or U2OS XPF-knockout cells (two independent clones). (F) FM-HCR in cells from panel (E). In panels (A), (C), and (F), each data point is from an independent experiment (n = 3) and error bars show the standard deviation. P-values are from an unpaired two-tailed t-test with comparison to wild type unless otherwise indicated.

Figure 4.

TC-NER but not GG-NER is required for efficient AFB₁-Fapy dG repair in human cells. Repair efficiency of BFP_AFB₁-Fapy (A) or mOrange_UV (B) reporters in U2OS-knockout cell lines transfected by electroporation with either undamaged cocktail (pMax_BFP, pMax_mOrange, and pMax_mPlum) or damaged cocktail (BFP_AFB1-Fapy dG, mOrange_UV, and pMax_mPlum). (C) Representative flow cytometry plots for U2OS cells transfected with the damaged cocktail. In the leftmost plot, BFP (VL1 laser) is plotted on the X-axis and mPlum (YL3 laser) on the Y-axis. In the middle plot, mOrange (YL1 laser) is on the X-axis, mPlum on the Y-axis. In the rightmost plot, mPlum is on the X-axis with BFP on the Y-axis. The name of the fluorescent protein marked inside each gate identifies the population being gated. Repair efficiency in (D) CS3BE (ERCC8 mutant), (E) CS1AN (ERCC6 mutant), and (F) GM15983 (XPC mutant) immortalized fibroblasts complemented with the respective wild-type protein. In all panels except panel (C), each data point is from an independent experiment. Error bars show the standard deviation and P-values are from unpaired two-tailed t-test.

TK6 cells were cultured in RPMI (Thermo Fisher, cat. no. 11875093) with 10% fetal bovine serum (FBS). All other cell lines were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Thermo Fisher, cat. no. 11965092) with 10% FBS. All cell lines were confirmed negative for mycoplasma by Lonza MycoAlert testing kit at least once during the study period.

DNA and RNA

DNA oligonucleotides were from Integrated DNA Technologies. Single guide RNAs (sgRNAs) used for generating CRISPR–Cas9 knockouts were from Thermo Fisher or Dharmacon. All DNA and RNA sequences are provided in the Supplementary data.

Enzymes

All enzymes were purchased by New England Biolabs (NEB), except for Neil1, which was generously provided by Walter Chazin.

Antibodies

XPA Cell Signaling Technologies Rabbit monoclonal clone D9U5U, catalog #14607.

XPF Cell Signaling Technologies Rabbit monoclonal clone D3G8C, catalog #13465.

GAPDH Santa Cruz Biosciences clone 0411, catalog sc-47724.

Chemicals

AFB₁ was purchased from Millipore Sigma.

Caution: AFB₁ is a potent liver toxin and is a human carcinogen. Crystalline AFB₁ is particularly hazardous due to its electrostatic nature and therefore should be handled using appropriate containment procedures in a well-ventilated hood and wearing a respiratory mask to prevent inhalation. AFB₁ can be destroyed by oxidation with NaOCl. It should be presumed that AFB₁-exo-8,9-epoxide is also toxic and a human carcinogen; hence, manipulations should be carried out with suitable containment procedures.

Preparation of oligonucleotide with AFB₁-Fapy dG modification

The AFB₁-Fapy dG-containing oligonucleotide was generated as described in detail previously [29]. In brief, an HPLC-purified 11-mer oligodeoxynucleotide with the sequence 5′-d(CCATCGCTACC)-3′ was annealed to a short complementary scaffold oligonucleotide with the sequence 5′-d(TAGCGA) to form a partial double-stranded DNA structure. This was reacted with AFB₁-exo-8,9-epoxide (generated by oxidizing AFB1 with dimethyldioxirane) to create the AFB₁-N⁷-dG cationic adduct at the unique dG. The modified oligonucleotide was purified by reverse-phase HPLC, dissolved in 100 mM sodium carbonate buffer (pH 11), and incubated overnight at 4°C to convert the labile cationic adduct to the imidazole ring-opened AFB₁-Fapy dG adduct. The modified oligodeoxynucleotide was purified by reverse-phase HPLC and concentrated by lyophilization, and the presence of a single AFB₁-Fapy dG adduct was confirmed by MALDI-TOF mass spectrometry (theoretical mass: 3583.3; observed mass: 3582.69).

Preparation of BFP_AFB₁-Fapy reporter plasmid

The AFB₁-Fapy dG oligonucleotide has the sequence 5′-d(CCATCGCTACC)-3′. To place this oligonucleotide in the transcribed strand, we created a modified plasmid, pMax_BFP_2, containing the complementary sequence [5′-d(GGTAGCGATGG)-3′] in the non-transcribed strand between the CMV promoter and the start codon of the BFP reporter gene by removing the green fluorescent protein (GFP) gene from our previously reported pMax_GFP [25] using NheI and SacI (NEB), using T4 DNA ligase to ligate in an appropriately designed gBlock (see Supplementary Fig. S1 for sequence) that had been digested with NheI/SacI, and transforming into NEB5-alpha E. coli. Plasmid DNA was first isolated by Mini prep and confirmed by Sanger sequencing (Azenta) and then amplified by Endo-free Giga prep kit (Thermo Fisher) to yield 12 mg of plasmid from 2 l of culture.

BFP_AFB₁-Fapy was generated similarly to other FM-HCR reporters [25]. The transcribed strand was nicked with Nb.BtsI and digested with ExoIII to generate single-stranded DNA (ssDNA) in the same fashion as described [25]. In a departure from the Piett et al. method, which utilizes chemically phosphorylated oligonucleotides, the unphosphorylated AFB₁-Fapy dG-containing oligonucleotide (350 pmol) was annealed to 100 μg of ssDNA in 120 μl of 1× NEB 2.1 buffer by heating to 90°C and slow cooling to 30°C. A one-pot phosphorylation/extension/ligation reaction was conducted by adding the following solutions in this order: 10× NEB 2.1 buffer (48 μl), 10 mM dNTPs (60 μl), 10 mM ATP (60 μl), 30% polyethylene glycol 8000 (150 μl), H₂O (113.9 μl), DMSO (24 μl), T4 polynucleotide kinase (PNK) (5 μl/50 units), T4 polymerase (16.6 μl), and T4 DNA Ligase (2.5 μl), with mixing after the addition of each component. The reaction was incubated overnight at 37°C, and the closed circular BFP_AFB₁-Fapy plasmid was isolated by T5 exonuclease digestion of incomplete reaction products, inactivation of the T5 exonuclease with proteinase K, purification of the plasmid by polyethylene glycol PEG precipitation, and removal of the proteinase K by phenol-chloroform extraction/ethanol precipitation as previously described [25].

Analytical digestion of BFP_AFB₁-Fapy plasmid with Neil1

BFP_AFB₁-Fapy (500 ng, 230 fmol) was incubated with purified Neil1 glycosylase (1 nmol) in a 10 μl reaction in NEB CutSmart buffer for 1 h at 37°C. Proteinase K (0.5 μl) was added, followed by 6× NEB Gel loading dye (2 μl). The sample was mixed and resolved on 1% agarose gel. Conversion to open circular DNA confirmed the presence of the DNA lesion.

Generation of U2OS XPF-knockout cell line

U2OS cells were seeded into six-well plates at a density of 125 000 cells per well. The following day, cells were transfected with CRISPR–Cas9 ribonucleoprotein complexes (RNPs) by mixing TrueCut Cas9 v2 (Thermo Fisher) with the appropriate sgRNA (Supplementary Table S3) and Lipofectamine CRISPRMax according to the manufacturer’s instructions. Seventy-two hours later, cells were collected by trypsinization, counted, and serially diluted to a density of 10 cells/ml in DMEM with 10% FBS. The cell suspension was triturated to ensure a single cell suspension, and four 96-well plates were seeded with 100 μl (an average of 1 cell) per well. After 5 days, plates were checked each day to monitor the growth of colonies. Any well that visibly had multiple colonies was excluded. After 10–12 days, 24 wells that contained colonies were washed with phosphate buffered saline (PBS), trypsinized, and seeded into a 24-well plate. Once cells were at ∼70% confluence, they were passaged to six-well plates and then expanded to T25 flasks. Clones were screened by western blot to confirm absence of the target protein, and successful knockouts were then tested for mycoplasma to confirm negativity.

Generation of U2OS XPA-knockout cell line

Initial attempts to knock out XPA using sgXPA-1 or sgXPA-2 were unsuccessful, so we performed a modified version of a published CRISPR deletion method [30] using both sgRNAs (Supplementary Table S3). U2OS cells in a six-well plate were treated with 500 nM AZD7648 (Selleckchem, cat. no. S8843) by adding 1 μl of 10 mM solution in DMSO to the well containing 2 ml of media. Immediately after, RNPs were assembled in separate tubes by mixing 20 pmol sgRNA, 75 μl Opti-MEM, 0.7 μl of TrueCas9 v2, and 6.7 μl Cas9 Plus reagent in one tube. A separate tube contained 12 μl of CRISPRMax reagent and 150 μl of Opti-MEM. Seventy-five microliters from this tube was added to the tubes containing Cas9 and either sgRNA, incubated for 15 min, and then both tubes were added to the well. After 72 h, cells were trypsinized, counted, and seeded into 96-well plates as in the above section, and clones were screened by western blot.

Measurement of AFB₁-Fapy dG repair efficiency by FM-HCR

Adherent cells were transfected with reporter plasmids using Lipofectamine 3000 (Thermo Fisher, cat. no. L300015) by seeding 40 000–50 000 cells per well into a 12-well plate and adhering overnight. One well was transfected with wild-type pMax_BFP_2, pMax_mOrange, and pMax_GFP plasmids (undamaged plasmid cocktail), while another well was transfected with BFP_AFB₁-Fapy, mOrange_UV, and pMax_GFP plasmids (damaged cocktail) by mixing P3000 reagent (2 μl) with plasmid cocktail (1.5 μl) in Opti-MEM medium (50 μl). This mixture was combined with Lipofectamine 3000 reagent (1.8 μl) in Opti-MEM (50 μl) and incubated for 5 min at room temperature before gently pipetting the transfection mixture (100 μl) into the appropriate well. After 20–24 h, cells were collected by trypsinization and analyzed by flow cytometry. Compensation and gating were established using single color controls as described previously [25, 31]. Fluorescent reporter proteins were detected using the following Attune NxT parameters (channel-excitation wavelength, filter): BFP (VL1-405, 450/40); mOrange (YL1-561, 620/15); GFP (BL1-488, 530/30); and mPlum (YL3-561, 780/60). Experiments were repeated thrice on separate days.

Suspension cell lines (TK6) were transfected with reporter plasmids using Neon NxT transfection system (Thermo Fisher). Cells were counted, collected by centrifugation, washed with PBS, and resuspended in the provided R buffer. A portion of the cells (13.5 μl) was mixed with plasmid cocktail (1.5 μl), and 10 μl of this mixture was transfected by electroporation according to the Neon protocol using default settings (1400 V, 20 ms pulse width, 1 pulse).

Most experiments were conducted using BFP, GFP, and mOrange; although in some cases, pMax_mPlum was substituted for pMax_GFP as the transfection control. In all cases, a single transfection utilized 1.5 μl volume in TE buffer and contained the following amounts of each reporter plasmid:

Undamaged plasmid cocktail: 10 ng BFP, 10 ng GFP, 10 ng mOrange, 1000 ng pddCMV carrier.

Damaged plasmid cocktail: 10 ng BFP_AFB₁-Fapy, 10 ng mOrange_UV, 10 ng GFP, 1000 ng pddCMV carrier.

Repair efficiency was calculated as described in detail in [25]:

BFP_AFB₁-Fapy dG reporter expression:

mOrange_UV reporter expression:

Stable overexpression of XPA

XPA was overexpressed in U2OS XPA^−/− cells using piggyBac transposase and plasmids ordered from VectorBuilder (see Supplementary data for plasmid maps and sequences). Cells were seeded into six-well plates at 150 000 cells per well and incubated overnight. The next day, cells were transfected with the following plasmids: a plasmid expressing PiggyBac transposase (200 ng), a carrier plasmid (pddCMV) to enhance transfection efficiency (500 ng), and 500 ng of a plasmid containing either empty vector control (ev_puro) or XPA (500 ng) flanked by the appropriate inverted repeats for PiggyBac gene integration. Three days after transfection, cells were selected with puromycin (1 μg/ml) for 1 week and then maintained in DMEM containing 1 μg/ml puromycin. XPA overexpression was confirmed by western blotting.

Preparation of CRIPSR plasmids

PB-CRISPR and LentiCRISPRv2-Blast were obtained from Addgene (plasmids #160047 and #83480, respectively), amplified by Midi prep kit (Qiagen), and 5 μg of plasmid was linearized with BsmBI-v2 in a 50 μl reaction at 55°C for 1.5 h. The reaction was incubated at 80°C for 20 min to inactivate the enzyme, resolved on a 1% agarose gel, and the desired band was excised and purified by Monarch DNA gel extraction kit (NEB). 5′-phosphorylated oligonucleotide duplexes (Supplementary Table S4 for sequences) containing the appropriate overhangs were assembled by mixing 1 μl of a 100-μM solution of each oligonucleotide with 6.5 μl H₂O, 1 μl of 10× T4 DNA ligase buffer, and 1 μl T4 PNK (NEB), incubating at 37°C for 30 min, heating to 95°C for 5 min, and cooling to 20°C at a rate of 2°C every 2 min. One microliter of this duplex was ligated into 25 ng linearized backbone using T4 DNA ligase in a 20 μl reaction, and 1 μl from the ligation was transformed into NEB 5-alpha (PB-CRISPR) or NEBStable (LentiCRISPRv2-Blast) E. coli, plated on LB containing carbenicillin (100 μg/ml), and single colonies were amplified by Mini prep (NEB) and confirmed by sequencing.

Generation of polyclonal knockout cell lines

U2OS cells were seeded at 125 000 cells per well into six-well plates. The following day, cells were transfected with 400 ng of the appropriate PB-CRISPR donor plasmid (sgNT, sgXPA, sgXPC, sgERCC6, sgERCC8, sgNEIL1-sg1, orsgNEIL1-sg2) and 80 ng of PBase expression plasmid (purchased from VectorBuilder) using Lipofectamine 3000 (2 μl P3000, 1.8 μl L3000, 100 μl Opti-MEM). After 3 days, cells were subjected to puromycin selection (2 μg/ml) for 10 days; successful gene editing was confirmed as described below. Cells were maintained in medium containing 1 μg/ml puromycin for the duration of experiments.

Polyclonal knockouts using LentiCRISPRv2-Blast were prepared by seeding HEK293T cells in DMEM with 10% FBS at a density of 500 000 cells per well in six-well plates, transfecting the next day with 1 μg LentiCRISPRv2-Blast sgNT, sgNEIL1-sg3, or sgNEIL1-sg4 and 500 ng each of the following packaging plasmids: pRSV-Rev (Addgene, #12253), pCMV-VSV-G (Addgene ,#8454), and pMDLg/pRRE (Addgene, #12251) using Lipofectamine 3000 (5 μl P3000, 4.75 μl L3000 in 250 μl Opti-MEM). Medium was removed after 6 h and replaced with 2 ml DMEM with 10% FBS. After 48 h, virus-containing medium was collected, filtered through a 0.45-μm filter, and 300 μl viral medium was used to infect XP2OS cells that had been seeded the previous day at 150 000 cells per well in a six-well plate. After 7 days, blasticidin selection (5 μg/ml), gDNA was isolated and used to confirm on-target editing (described below).

Validation of U2OS XPC, CSB, and UVSSA-knockout cells

One million cells were collected by trypsinization, and genomic DNA was isolated by QiaAMP DNA Mini Kit (Qiagen) according to the manufacturer’s protocol. The locus targeted by CRISPR–Cas9 was amplified by PCR (see Supplementary Table S2 for primer sequences) using Q5 Hot Start 2X Master Mix (New England Biolabs) and 100 ng genomic DNA. The amplicon was purified by Monarch PCR cleanup kit (New England Biolabs) and submitted for Nanopore sequencing (Plasmidsaurus) for CSB and UVSSA, which showed homozygous knockout (Supplementary Fig. S2) or Sanger sequencing followed by DECODR analysis (decodr.org) for XPC [32].

Validation of XPC, CSA, and CSB-deficient fibroblast cell lines and polyclonal knockout cells

Genomic DNA was isolated from NER-deficient fibroblast lines, amplified by PCR as in the above section, and analyzed by Sanger sequencing by Azenta/Genewiz (see Supplementary Table S1 for primer sequences) to confirm the presence of an inactivating mutation (Supplementary Fig. S3). For polyclonal knockout cells, see Supplementary Table S5 for primer sequences. Sanger traces were analyzed by DECODR v3.

Western blotting

Protein extracts were prepared in NETN lysis buffer (100 mM NaCl, 20 mM Tris, pH 8.0, 0.5 mM ethylenediaminetetraacetic acid, 0.5% NP-40) containing protease inhibitor cocktail, and protein concentration was determined by BCA assay kit (Invitrogen). Proteins were separated on pre-cast 12% polyacrylamide gels from Bio-Rad and transferred to a nitrocellulose membrane (Bio-Rad) in cold Tris glycine buffer with 10% methanol at 100 V for 1 h. Membranes were blocked using 7.5% milk in phosphate buffered saline with 0.1% Tween (PBST) for 1 h and then incubated at 4°C overnight with primary antibody against XPA or XPF (1:1000 dilution) in PBST with 5% nonfat milk. The next day, blots were washed with PBST (three 10-min washes) and incubated with HRP-conjugated goat secondary antibodies (Bio-Rad) at 1:20 000 dilution in PBST at room temperature for 1 h followed by six 5-min washes and detection with SuperSignal Plus West Pico Reagent (Thermo Fisher, cat. no. 34577) and imaging using a Thermo Fisher iBright gel imager. Blots were stripped with stripping buffer (50 mM Tris HCl, pH 6.8, 8% sodium dodecyl sulfate, 100 mM beta-mercaptoethanol) by incubating at 55°C for 15 min and then shaking at room temp for 15 min. This was followed by six 10-min washes with PBST and then incubation with primary antibody against GAPDH (1:2000 dilution in PBST with 5% milk), overnight at 4°C. Blots were washed three times with PBST, incubated with HRP-conjugated secondary antibody (1:2000 dilution) for 1 h at room temperature with shaking, washed three times with PBST, and visualized by chemiluminescence with Clarity Western ECL Substrate (Bio-Rad, cat. no. 1705061).

Results

Construction and validation of BFP_AFB₁-Fapy reporter assay

Plasmid-based host cell reactivation assays are regularly used to study repair of bulky DNA adducts in human cells [31, 33, 34]. We created an FM-HCR assay, BFP_AFB₁-Fapy, containing a site-specific AFB₁-Fapy dG on the transcribed strand between the CMV promoter and BFP gene in the pMax_BFP reporter plasmid. When the plasmid is transfected into cells, the BFP reporter gene is transcribed; however, AFB₁-Fapy dG blocks transcription and prevents BFP fluorescence. Lesion repair reactivates BFP fluorescence, which is used as a readout for repair efficiency by normalizing BFP signal to the signal from a transfection control plasmid, in this case pMax_mPlum (Fig. 1B). We first generated an oligonucleotide containing a site-specific AFB₁-Fapy dG adduct at the unique dG (underlined) in the oligonucleotide 5′-d(CCATCGCTACC)-3′ as described previously [29]. We purified the modified oligonucleotide by reverse-phase HPLC (Supplementary Fig. S4) and confirmed the presence of the AFB₁-Fapy dG adduct by mass spectrometry (Supplementary Fig. S5).

We incorporated the AFB₁-Fapy dG oligonucleotide into the reporter plasmid in a similar fashion as other FM-HCR plasmids [25]. In brief, we created a modified pMax_BFP plasmid, pMax_BFP_2, which contains the appropriate complementary sequence, 5′-d(CCATCGCTACC)-3′, for placing the AFB₁-Fapy dG oligo in the 5′-untranslated region (5′-UTR) of the reporter gene (Supplementary Fig. S1). We nicked and digested the transcribed strand of pMax_BFP_2, annealed the AFB₁-Fapy dG oligonucleotide to its complementary sequence in the 5′-UTR, and conducted a one-pot phosphorylation/polymerization/ligation (see the “Materials and methods” section). We purified the closed circular plasmid according to established methods [25] and confirmed that BFP_AFB₁-Fapy migrated identically to a closed circular control plasmid on an agarose gel (compare lane 4 with lane 2 in Fig. 1C). Furthermore, incubation with purified human Neil1 glycosylase cleaved BFP_AFB₁-Fapy (lane 6, Fig. 1C) but not a lesion-free control (lane 5, Fig. 1C), confirming the presence of the lesion (Fig. 1C).

We assessed AFB₁-Fapy dG repair efficiency as previously described in detail for other FM-HCR plasmids (also see the “Materials and methods” section) [25]. For each sample, two transfections were conducted: one with “undamaged” plasmids and one with “damaged” plasmids (Fig. 1D). The undamaged mixture contained wild-type pMax_BFP_2 and a transfection control plasmid, in Fig. 1D, pMax_mPlum. The damaged plasmid mixture contained BFP_AFB₁-Fapy and the same transfection control plasmid, pMax_mPlum. Cells were analyzed by flow cytometry 18–24 h after transfection, and fluorescent reporter expression was calculated by multiplying the number of BFP-positive cells in the damaged plasmid transfection by the mean BFP fluorescence intensity. This value was divided by the corresponding value calculated using the same approach for mPlum. The resulting ratio was then divided by the same ratio calculated for the undamaged plasmid transfection. This ratio of ratios, denoted “% Reporter expression, ” represents the repair efficiency for the BFP_AFB₁-Fapy reporter. We note that % Reporter expression is linearly related to the amount of plasmid DNA transfected into cells (Supplementary Fig. S6), suggesting that this method quantitatively measures cellular DNA repair capacity.

The BFP_AFB₁-Fapy reporter measures repair efficiency in cells by taking advantage of the long-known ability of aflatoxin adducts to block transcription [35]. An unrepaired AFB₁-Fapy dG lesion was expected to block transcription and thereby reduce expression of the BFP reporter gene. To confirm this, we transfected BFP_AFB₁-Fapy into the NER-proficient human cell lines TK6 and U2OS, as well as the NER-deficient cell line XP2OS—an immortalized fibroblast line harboring an inactivating mutation in the XPA gene [36]. In all three cell lines, BFP_AFB₁-Fapy was expressed at lower levels than the lesion-free control plasmid (Fig. 1E), and this was most pronounced for the NER-deficient XP2OS cell line (5.3% expression), where the number of BFP-positive cells was greatly reduced as was the BFP fluorescence intensity (Fig. 1D). These data confirm that the AFB₁-Fapy lesion strongly interferes with transcription, particularly in NER-deficient cells, and thus enables the use of FM-HCR for measuring AFB₁-Fapy dG repair efficiency.

Immortalized fibroblasts from xeroderma pigmentosum patients are defective in AFB₁-Fapy repair

To assess whether NER repairs AFB₁-Fapy dG in human cells, we transfected BFP_AFB₁-Fapy into immortalized fibroblasts from xeroderma pigmentosum (XP) patients, which are deficient in NER. We co-transfected pMax_GFP, which served as a transfection control, and mOrange_UV, [39], which was generated by irradiating pMax_mOrange with 800 J/m² UV-C light (Fig. 2A). We have previously shown that this plasmid robustly measures NER activity in human cells [37]; and although the UV photoproducts are randomly distributed throughout the plasmid, we note that—in NER-null cells—expression of mOrange_UV is ∼1% as efficient as wild-type pMax_mOrange, implying that at least 99% of plasmids contain at least one UV photoproduct per reporter gene. In most experiments, we utilized the combination of BFP, GFP, and mOrange, allowing us to measure expression of all three fluorescent proteins within the same sample. These fluorescent proteins have minimal overlap in their emission spectra that is easily overcome with appropriate compensation and gating (Supplementary Fig. S7) [25].

Consistent with our previous studies, XPA-deficient XP2OS cells were nearly completely defective for repair of the NER control plasmid, mOrange_UV (0.6% reporter expression), and stable expression of XPA rescued reporter expression by over 30-fold, indicating that NER activity is functionally restored in these cells. Similarly, XP2OS cells expressed BFP_AFB₁-Fapy only 6% as efficiently as the damage free pMax_BFP_2 control, consistent with AFB₁-Fapy dG strongly blocking transcription in the absence of functional NER. Importantly, complementation with wild-type XPA enhanced reporter expression to 36% (Fig. 2B and C).

We extended this analysis to two additional immortalized XP cell lines: XP2YO cells, which harbor a mutation in ERCC4 (XPF protein), and XP3BRSV cells, which harbor a mutation in ERCC5 (XPG protein) [26]. Interestingly, XP2YO cells showed greater baseline expression of BFP_AFB₁-Fapy and mOrange_UV reporters (31% for BFP_AFB₁-Fapy and 5% for mOrange_UV; Fig. 2D) when compared with XP2OS (Fig. 2C). The residual repair capacity likely reflects a partially functional XPF protein in the XP2YO cells. Expression of wild-type XPF markedly enhanced expression of both reporters, while expression of catalytically inactive XPF (XPF D676A) had no effect. In XP3BRSV cells, BFP_AFB₁-Fapy was expressed inefficiently (9.5% of lesion-free control), and expression of wild-type XPG enhanced repair by over 10-fold to 102.9% (P= .01; Fig. 2E). Interestingly, expression of a catalytic mutant (XPG E791A) also enhanced expression of BFP_AFB₁-Fapy to a similar degree. Similar results were obtained for mOrange_UV, suggesting that the presence of XPG but not necessarily its catalytic activity is required for NER. These data mirror previous results in the XPG-deficient cell line, XPCS1RO, where expression of XPG E791A rescued UV-induced unscheduled DNA synthesis, albeit incompletely, contrasting with the full rescue observed here [26]. Overall, our data indicate that the core NER factors XPA, XPF, and XPG are required for efficient AFB₁-Fapy dG repair in immortalized fibroblast cell lines.

Knockout of XPA or XPF ablates AFB₁-Fapy dG repair in U2OS cells

XP patient fibroblasts are a valuable tool for assessing the contribution of NER to lesion repair; however, patient mutations do not always completely ablate protein function. For instance, despite the absolute requirement for both XPF and XPA for NER, XPF-deficient XP2YO cells are markedly more efficient at expressing BFP_AFB₁-Fapy and mOrange_UV than XPA-deficient XP2OS cells. We therefore employed multiple isogenic knockouts of NER proteins to further assess the contribution of NER to AFB₁-Fapy dG repair. We measured BFP_AFB₁-Fapy expression in a previously reported TK6 XPA-knockout cell line [38], where we observed significantly decreased BFP_AFB₁-Fapy expression (18% in wild type, 3.1% in XPA^−/−; Fig. 3A).

We extended our analysis of NER in AFB₁-Fapy dG repair by knocking out XPA in human U2OS cells. We transfected U2OS cells with Cas9-sgRNA RNPs, isolated single-cell clones, and assessed XPA protein expression by western blotting. Initial attempts to knock out XPA using a single sgRNA were unsuccessful, as all isolated clones expressed a small amount of residual XPA protein detectable by western blot (not shown). We therefore combined two independent sgRNAs targeting exons 3 and 6 (see the “Materials and methods” section for full details) and isolated a clone with essentially undetectable levels of XPA (Fig. 3B) and a nearly complete inability to repair the mOrange_UV reporter (Fig. 3C). We then stably re-expressed XPA in these cells using a PiggyBac transposase system, with an empty vector serving as a negative control in wild-type or XPA^−/− cells. U2OS XPA^−/− cells were markedly deficient in BFP_AFB₁-Fapy repair and, importantly, expression of XPA completely reversed this phenotype (Fig. 3C and D). Finally, we knocked out the core NER nuclease, XPF, in U2OS cells. We confirmed a lack of XPF expression in two independent clones by western blotting (Fig. 3E) and found that both clones were defective for repair of UV damage and AFB₁-Fapy dG (Fig. 3F). Taken together, these data indicate that NER makes a major contribution to repair of the AFB₁-Fapy dG lesion in human cells.

CSA, CSB, and UVSSA are required for efficient AFB₁-Fapy repair, while XPC is not

In humans, NER occurs via two subpathways: GG-NER and TC-NER, which differ mainly in the lesion recognition step. GG-NER is initiated by XPC, which recognizes helix distorting lesions throughout the genome [39], while TC-NER is initiated by CSA, CSB, and UVSSA when transcription complexes stall at DNA damage [27, 40, 41]. Sarmini et al. showed that a fluorescent reporter plasmid containing a 5′-8-cyclopurine lesion is repaired by TC-NER while a reporter plasmid containing the N-(deoxyguanosin-8-yl)-2-acetylaminofluorene lesion is repaired exclusively by GG-NER [33], indicating that reporter plasmids can be used to interrogate NER subpathway dependency.

To determine which NER subpathway repairs AFB₁-Fapy dG lesions, we electroporated BFP_AFB₁-Fapy into a panel of previously reported U2OS NER-knockout cell lines, including CSB^−/−, UVSSA^−/−, and XPC^−/−, reported by van der Weegen et al. [27] and XPF^−/−, reported by Sabatella et al.[28]. We confirmed the identity of XPC, CSB, and UVSSA-knockout cell lines by sequencing the Cas9 cut site (Supplementary Fig. S2). Parental U2OS cells efficiently expressed BFP_AFB₁-Fapy (48.1% of lesion-free pMax_BFP_2), while CSB^−/− and UVSSA^−/− cell lines displayed a nearly four-fold reduction in BFP_AFB₁-Fapy expression (13.4% in CSB^−/− and 13.2% in UVSSA^−/−; Fig. 4A). Strikingly, XPC^−/− cells showed no significant change in expression of BFP_AFB₁-Fapy (48.1% in wild type and 47.9% in XPC^−/−; P= .9). The UV-irradiated mOrange plasmid, serving as an internal NER control, showed that all U2OS NER knockouts displayed a functional reduction in NER activity (Fig. 4B), consistent with previous findings indicating that repair of UV-irradiated plasmids is impaired in both TC-NER-deficient cells and GG-NER-deficient cells [42]. Importantly, XPC knockout severely impaired mOrange_UV expression, while having no effect on BFP_AFB₁-Fapy expression. We repeated this experiment but instead transfected cells with Lipofectamine 3000, to determine whether the transfection method influences repair outcome. CSB^−/−, UVSSA^−/−, and XPF^−/− cells displayed a marked reduction in BFP_AFB1-Fapy expression that was nearly identical to that observed when electroporation was used (Supplementary Fig. S8). Interestingly, mOrange_UV also displayed nearly identical results when comparing Lipofectamine 3000 to electroporation, except for CSB^−/− cells, which surprisingly showed no significant decrease in repair efficiency compared to wild type. However, taken together, these data suggest that TC-NER is the dominant repair pathway for BFP_AFB₁-Fapy.

We further corroborated our finding that TC-NER is the dominant NER subpathway responsible for repairing AFB₁-Fapy adducts by transfecting BFP_AFB₁-Fapy into immortalized fibroblasts from Cockayne syndrome patients. CS3BE fibroblasts, which harbor compound heterozygous mutations in the ERCC8 gene (37G>T and 479C > T; Supplementary Fig. S3B), inefficiently expressed BFP_AFB₁-Fapy. Complementation with CSA (ERCC8 gene) markedly rescued repair efficiency, from 5% to 43% (Fig. 4D), mirroring results in XP2OS cells complemented with XPA (Fig. 2A). Interestingly, expression of mOrange_UV was only slightly increased with CSA complementation, increasing from 7% to 13%, in contrast with XP2OS cells, which showed 0.6% repair efficiency, which increased to 20% with XPA complementation. Similar results were obtained in CS1AN fibroblasts, which we found to contain a homozygous 1009A > T mutation in the ERCC6 gene, which introduces a premature stop codon. This is in contrast with a previous report that this cell line contains an additional mutation (2569C > T) (Supplementary Fig. S3A) [43]. These cells displayed 9% BFP_AFB₁-Fapy expression that was increased to 36% by CSB expression (Fig. 4E), further supporting a role for TC-NER in the repair of AFB1-Fapy dG. Finally, we assessed BFP_AFB₁-Fapy expression in the immortalized fibroblast line, GM15983, which harbors a homozygous 2-nt deletion in exon 9 of the XPC gene (Supplementary Fig. S3C). These cells inefficiently expressed the NER control plasmid, mOrange_UV, and complementation with XPC markedly enhanced expression from 6.3% to 24% (P= .043), indicating that NER is functionally restored in XPC-complemented cells (Fig. 4E). Notably, BFP_AFB₁-Fapy was expressed more efficiently in GG-NER-deficient GM15983 than in CS1AN and CS3BE, and complementation with XPC had no significant effect (28.9% in GM15983 and 29.9% in GM15983 + XPC). We conclude that TC-NER is the major repair pathway for AFB₁-Fapy dG in U2OS cells, with GG-NER making no major contribution.

Previous reports identified a role for Neil1 glycosylase in the repair of AFB1-Fapy dG both in vitro and in vivo. Indeed, incubation with excess Neil1 glycosylase fully converted the BFP_AFB1 plasmid to an open circular form (Fig. 1). To determine the contribution of Neil1 to AFB1-Fapy dG repair within the actively transcribed BFP_AFB1-Fapy plasmid, we generated polyclonal Neil1-knockout U2OS cells by integrating an expression cassette containing Cas9, a puromycin resistance gene, and two independent sgRNAs targeting NEIL1. We also generated cells expressing sgXPA, sgXPC, sgERCC6, sgERCC8, or an sgNT. We had difficulty detecting Neil1 by western blot in multiple cell types, so we confirmed successful CRISPR targeting by Sanger sequencing (Fig. 5A). Analysis of each sequencing trace by DECODR confirmed editing at the expected Cas9 cut site, with a large proportion of frameshift mutations in each polyclonal knockout (Fig. 5B and Supplementary Fig. S9). We transfected BFP_AFB1-Fapy or mOrange_UV into each cell line and assessed repair by FM-HCR. Interestingly, sgNEIL1 cells also showed no change in BFP_AFB1-Fapy expression (Fig. 5B), suggesting that Neil1 does not make a major contribution to repair of BFP_AFB1-Fapy. Cells expressing sgXPA, sgERCC6, or sgERCC8 showed a significant reduction in BFP_AFB1-Fapy expression, while sgXPC cells showed no significant change, similar to our results in clonal U2OS knockouts (Fig. 4). The mOrange_UV reporter confirmed functional depletion of NER in polyclonal knockout cells, with sgXPA and sgXPC cells, showing a significant reduction in repair relative to sgNT cells (Fig. 5C). Interestingly, sgERCC6 and sgERCC8 did not show any change in mOrange_UV expression, similarly to our observation that clonal U2OS CSB (ERCC6)-knockout cells show no decrease in repair of this reporter (Supplementary Fig. S8) when using lipofection. However, taken together, these data indicate that Neil1 does not make a major contribution to BFP-AFB1-Fapy repair in U2OS cells.

Figure 5.

Neil1 does not make a major contribution to BFP_AFB1-Fapy repair in U2OS and XP2OS cells. (A) Sanger sequencing traces from U2OS polyclonal knockout cells stably expressing Cas9 and the sgRNA indicated at the left of the plot (see Supplementary Fig. S3 for editing efficiency). (B, C) FM-HCR data from U2OS polyclonal knockouts transfected by Lipofectamine 3000 with the same reporter plasmids used in Fig. 4. (D, E) FM-HCR in XP2OS polyclonal knockout cells stably expressing Cas9 and the indicated sgRNA (see Supplementary Fig. S9 for editing efficiency). Each data point is from an independent experiment. Error bars show the standard deviation and P-values are from unpaired two-tailed t-test.

Finally, we considered the possibility that efficient NER in U2OS cells masks the contribution of Neil1 to AFB1-Fapy dG repair. We disrupted NEIL1 in the NER-deficient XP2OS cell line using CRISPR–Cas9. We also used two additional sgRNAs, one targeting exon 5 (sgNEIL1-3) and one targeting the second codon in exon 1 of the NEIL1 gene (sgNEIL1-4), which encodes the catalytic proline for Neil1 glycosylase activity [44] (Supplementary Fig. S10). For these experiments, we lentivirally integrated a Cas9/sgRNA expression cassette with a blasticidin-resistance gene and selected a polyclonal population of blasticidin-resistant cells. Sanger sequencing of the targeted region confirmed on-target Cas9 activity for both sgRNAs (Fig. 5D and Supplementary Fig. S10), with sgRNA4 showing 84% frameshift mutation efficiency; however, we again observed no significant change in BFP_AFB1-Fapy expression in sgNEIL1-sg3 or sgNEIL1-sg4 cells compared to non-targeting control cells (sgNT), nor did we observe a change in expression of mOrange_UV. Although we cannot exclude the possibility that residual Neil1 activity exists in these cells, our data suggest that Neil1 does not make a major contribution to BFP_AFB1-Fapy repair.

Discussion

AFB₁, a natural product produced by fungi that contaminate human food supplies, is bioactivated by cytochrome P450 enzymes in the liver, and subsequently reacts with DNA to form bulky adducts. AFB₁ exposure is associated with G→T transversion mutations and eventual development of HCC. Although AFB₁ contamination is less prevalent in developed countries, and strict limits on AFB₁ levels in food supplies reduce exposure even further, it remains important to understand the molecular mechanisms by which AFB₁ causes carcinogenesis in humans.

We investigated repair of the major AFB₁ DNA adduct, AFB₁-Fapy dG, in human cells using a fluorescence-based host cell reactivation plasmid, BFP_AFB₁-Fapy. We found that AFB₁-Fapy dG reduces expression of a BFP reporter gene, presumably by blocking transcription, and that reactivation of reporter gene expression is strictly dependent upon TC-NER. Knockout or mutation of CSA, CSB, or UVSSA impairs BFP_AFB₁-Fapy expression to a similar level as knockout of the NER core factor XPF. Conversely, knockout or mutation of XPC has no significant effect on expression of BFP_AFP₁-Fapy. Our data, which are consistent with recent results using translesion excision repair-sequencing in AFB₁-treated cells [45], suggest that the persistence of AFB₁-Fapy dG lesions in liver cells of exposed animals is due to inefficient recognition and repair of AFB₁-Fapy dG outside of transcribed genes. Interestingly, XPC initiates GG-NER by inserting a β-hairpin into the DNA duplex and flipping out the damaged bases, a process that is facilitated when the duplex is thermodynamically destabilized [46]. Despite its bulky size, the AFB₁-Fapy dG adduct stabilizes DNA duplexes, increasing the melting temperature by up to 15°C depending on sequence context. In contrast, the AFB₁-N⁷-dG adduct, which disappears from liver DNA more rapidly, stabilizes comparable duplexes by only 3°C [29, 47]. GG-NER efficiency often correlates with the extent of duplex destabilization by the lesion [48], suggesting that AFB₁-Fapy dG would be refractory to repair by this pathway, as we observed in our reporter assay.

Although we did not observe GG-NER of AFB₁-Fapy dG in our assay, we cannot completely rule out a contribution of this pathway to AFB₁-Fapy dG repair. Knockout of XPC in C. elegans increases AFB₁-induced mutations [22] and an epidemiological study found that polymorphisms in the XPC gene were associated with an increased risk of HCC in the Guangxi region of China, where AFB₁ exposure was high at the time of the study [49]. We propose that GG-NER is inefficient at repairing AFB1-Fapy dG lesions, and within the context of active transcription, RNA Pol II stalls at the lesion prior to lesion recognition by XPC, triggering TC-NER.

The human Neil1 glycosylase excises the AFB₁-Fapy dG lesion from duplex DNA in vitro, and mice lacking Neil1 show dramatically increased risk of AFB₁-induced liver cancer [23, 29]. Interestingly, both Neil1-knockout mice and XPA-knockout mice display increased G→T transversions following AFB₁ exposure; however, only XPA-knockout mice display a transcriptional bias in mutational signature [50]. Taken together with our observation that polyclonal NEIL1 knockout U2OS and XP2OS cells show no major change in BFP_AFB₁-Fapy expression, this could suggest that Neil1 does not make a major contribution to AFB₁-Fapy dG repair in highly transcribed genes such as the BFP_AFB₁-Fapy reporter. However, additional work is needed to fully understand the role of Neil1 in AFB₁-Fapy dG repair. We note that sequence context influences Neil1 activity in vitro, with greater Neil1 activity observed on sequences where the AFB₁-Fapy dG adduct is most destabilizing [29]. The 5′-CXC-3′ sequence context employed in this study (where X = AFB₁-Fapy dG) is a relatively poor Neil1 substrate in vitro, so we caution against generalizing our results to additional sequence contexts. Additionally, our investigation into the role of Neil1 in AFB₁-Fapy dG repair was limited to U2OS and XP2OS cells. We struggled to detect Neil1 by western blot in U2OS (data not shown), suggesting that Neil1 expression level is low in these cells. It will be important to determine how Neil1 knockout or depletion affects AFB₁-Fapy dG repair in cells with high Neil1 expression. Finally, Neil1 expression is markedly induced by treatment with H₂O₂ in MDA-MB-231 cells and is nearly undetectable in untreated cells [51]. It will therefore be interesting to determine how treatment with AFB₁ influences Neil1 expression and whether pre-treatment with AFB₁ changes repair of BFP_AFB₁-Fapy. Taken together, our results provide additional information into the role of Neil1 in AFB1-Fapy dG repair but also suggest that further work is needed to understand how local sequence context, cell type, and Neil1 expression level modulate AFB₁-Fapy dG repair and how GG-NER, TC-NER, and Neil1-dependent BER are balanced to repair this lesion.

Taken together, our data suggest that TC-NER is the major repair pathway for AFB₁-Fapy dG. Cancer driver mutations are greatly overrepresented in protein-coding regions of the genome [52, 53], and the observation that many cases of AFB₁-induced HCC harbor a G→T transversion at codon 249 of the TP53 gene suggests that AFB₁ induces mutations within transcribed genes, where TC-NER operates. In HCC cases with AFB₁ exposure, there is a significant transcriptional strand bias in the mutational signature, suggesting that TC-NER plays an important role in suppressing AFB₁-mediated carcinogenesis [54]. It will therefore be interesting to determine whether interindividual differences in TC-NER efficiency modulate susceptibility to aflatoxin toxicity or HCC in AFB₁ endemic regions.

Unrepaired AFB₁-Fapy lesions induce G→T transversions, likely formed by translesion synthesis polymerases that synthesize through replication-blocking lesions at the cost of mutagenesis. Recent evidence suggests that mutagenic translesion synthesis plays an important role in suppressing UV-induced carcinogenesis in mice, possibly because point mutations induced by translesion synthesis are less carcinogenic than the complex rearrangements that occur when replication forks collapse at blocking lesions [55]. Rev3 knockout reduces G→T transversion frequency in AFB₁-treated C. elegans but increases the frequency of structural variant mutations [56]. It will therefore be interesting to determine whether translesion synthesis suppresses carcinogenesis in response to AFB₁ exposure—and whether polymerases other than Pol zeta are involved—particularly in individuals with inefficient TC-NER.

Supplementary data

Supplementary data is available at NAR Cancer online.

Conflict of interest

None declared.

Funding

National Institute of Environmental Health Sciences (Grant/award numbers R01ES029357 and U01ES029520-01), National Cancer Institute (Grant/award number P01CA092584).

Data availability

All data are available within the manuscript and in the source data file.