Proteome-Wide Identification of O⁶-Methyl-2’-deoxyguanosine-Binding Proteins

Abstract

DNA is subjected to damage from various endogenous and exogenous sources of alkylating agents, resulting in alkylated DNA lesions. Among these lesions, O⁶-methyl-2′-deoxyguanosine (O⁶-Me-dG) is highly mutagenic, and it can be repaired by O⁶-alkylguanine DNA alkyltransferase and mismatch repair pathway. It, however, remains unclear whether O⁶-Me-dG in DNA can be recognized by other cellular proteins. Here, we employed a quantitative mass spectrometry-based approach to uncover reader proteins of O⁶-Me-dG in DNA when it is paired with a 2′-deoxycytidine (dC) or thymidine (dT). We were able to identify 67 and 31 candidate reader proteins for duplex DNA harboring O⁶-Me-dG:dC and O⁶-Me-dG:dT base pairs, respectively. In addition, genetic ablation of CDKN2AIP, a.k.a. CARF, one of those proteins that can recognize both the O⁶-Me-dG:dC and O⁶-Me-dG:dT base pairs, in HEK293T cells conferred augmented tolerance to N-nitroso-N-methylurea (NMU), an alkylating agent that can induce O⁶-Me-dG in DNA. Accordingly, our LC-MS/MS quantification results revealed that the loss of CDKN2AIP led to diminished accumulation of NMU-induced O⁶-Me-dG in genomic DNA. Together, we explored the damage recognition proteins of O⁶-Me-dG using a quantitative mass spectrometry-based approach, and our results revealed an unexpected role of CDKN2AIP in sensitizing cultured cells toward a DNA methylating agent.

Graphical Abstract

INTRODUCTION

DNA is subjected to damage from exposure to endogenous and exogenous sources of alkylating agents, e.g., S-adenosyl-l-methionine, N-nitrosamines in food and tobacco products, and cancer chemotherapeutic agents.^1–4 Alkylating agents can induce the formation of different DNA adducts that can impede DNA replication and transcription as well as elicit mutations in these cellular processes.^5,6 O⁶-methyl-2’-deoxyguanosine (O⁶-Me-dG) is a major DNA adduct induced upon exposure to methylating agents.⁷ The presence of O⁶-Me-dG can result in G:C→A:T transition mutation during DNA replication, which arises from preferential misincorporation of dTTP opposite O⁶-Me-dG.⁸

Cells are equipped with different mechanisms to sense the existence and initiate the repair of alkylated DNA lesions.^5,9 In this vein, O⁶-Me-dG can be repaired by mismatch repair¹⁰ or through direct removal of the methyl group by the DNA repair protein O⁶-methylguanine DNA methyltransferase (MGMT).⁵ We reason that there may exist additional repair proteins sensing the presence and modulating the repair of O⁶-Me-dG in DNA. O⁶-Me-dG can pair with thymidine (dT) and 2′-deoxycytidine (dC) in pseudo-Watson–Crick and wobble geometries, respectively (Figure 1a). On the grounds of the recent observation that the presence of DNA mismatches modulates the binding of transcription factors,¹¹ we hypothesize that O⁶-Me-dG:dC and O⁶-Me-dG:dT pairs in DNA may be recognized by different cellular proteins.

Figure 1.

Base pairing configurations of O⁶-Me-dG when paired with dC (wobble pair) or dT (pseudo-Watson–Crick pair) (a) and forward SILAC-based interaction screening for identifying novel damage recognition proteins of O⁶-Me-dG (b). The “B” in red circle indicates 5′-biotin labeling.

Quantitative proteomics-based affinity screening has been employed for the identification of new cellular proteins that can recognize modified nucleosides in DNA and RNA, as well as unique DNA secondary structures.^12–18 In this study, we set out to identify comprehensively cellular proteins that are involved in the recognition and repair of O⁶-Me-dG. We employed a quantitative proteomics method, based upon stable isotope labeling by amino acids in cell culture (SILAC),¹⁹ to discover reader proteins of duplex DNA harboring an O⁶-Me-dG paired with a dC or dT (Figure 1b). We were able to identify common and distinct candidate reader proteins for O⁶-Me-dG when paired with a dC or dT. We also found that CRISPR-mediated ablation of the gene encoding one of these reader proteins, CDKN2AIP, in HEK293T cells led to diminished accumulation of O⁶-Me-dG in genomic DNA and rendered cells more resistant toward a DNA methylating agent, N-methyl-N-nitrosourea (NMU).

MATERIALS AND METHODS

Synthesis of Oligodeoxyribonucleotide (ODN) with a Site-Specifically Inserted O⁶-Me-dG.

O⁶-Me-dG phosphoramidite (Glen Research, Sterling, VA) was employed for the solid-phase synthesis of 5′-d(ACGGAGXGACAT)-3′, where X = O⁶-Me-dG, on a Beckman Oligo 1000M DNA Synthesizer following standard procedures. The ODN was deprotected by incubating in 10% 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU) in methanol at room temperature for 5 days, and purified using HPLC with a solvent system of 50 mM triethylammonium acetate (TEAA) (Solution A) and 30% acetonitrile in A (Solution B). The gradient was as follows: 0–10% B for 5 min, 10–40% B for 40 min, 40–100% B for 3 min, 100% B for 3 min, 100–0% B for 3 min. The purified ODN was characterized by ESI-MS and MS/MS.

The purified 12mer O⁶-Me-dG-bearing ODN was ligated with an 8mer, 5′-d(GATCCTAT)-3′, to yield a 20mer in the presence of a 28mer complementary strand ODN in the ligase buffer at 16 °C overnight. The resulting 20mer was purified by denaturing polyacrylamide gel electrophoresis (PAGE) and characterized again by ESI-MS and MS/MS (Figure S1). We chose this 20mer sequence for the affinity pull-down experiments because the same sequence was recently employed for identifying N²-nBu-dG-binding proteins,²⁰ thereby allowing us to reveal the similarities and differences in cellular proteins involved in the recognition of major- and minor-groove alkylated dG lesions.

Cell Culture.

HeLa cells (ATCC) were cultured in SILAC DMEM medium containing 10% dialyzed fetal bovine serum (Invitrogen) and 1% penicillin and streptomycin (Invitrogen). The SILAC media were prepared by supplementing arginine- and lysine-depleted DMEM media with unlabeled l-arginine and l-lysine (Sigma, for the light media), or [¹³C₆]-l-arginine and [¹³C₆,¹⁵N₂]-l-lysine (Cambridge Isotope Laboratories, for the heavy media). The cells were maintained at 37 °C in a humidified incubator containing 5% CO₂ and cultured in heavy SILAC media for at least 10 cell doublings to ensure nearly complete heavy isotope incorporation.

Preparation of Nuclear Lysate.

Upon reaching 80% confluency, HeLa cells were harvested by treating with trypsin-EDTA (Invitrogen) and pelleted by centrifugation. The cell pellet was washed with phosphate-buffered saline. The nuclear proteome was prepared from the heavy and light-labeled cells using the NE-PER nuclear and cytoplasmic extraction reagents (Thermo Fisher Scientific) following the manufacturer’s guidelines. In this vein, our previous affinity pull-down experiments with the nuclear lysate prepared using the same procedure led to the identification of guanine quadruplex DNA- and damage DNA-binding proteins that were later validated to be capable of binding directly with these DNA substrates.^13,16,17,20 Hence, the procedure for nuclear lysate generation should not perturb the folding of nuclear proteins. The protein concentration of the nuclear fraction was measured using Bradford Quick Start Protein Assay kit (Bio-Rad), and the nuclear lysate was stored at −80 °C until further use.

Affinity Pull-Down of O⁶-Me-dG-Binding Proteins.

The 20mer ODN containing a site-specific O⁶-Me-dG and the corresponding unmodified ODN (100 pmol each) were annealed individually with a 5′-biotin-labeled complementary strand (50 pmol) with a dC or dT being opposite the lesion (purchased from Integrated DNA Technologies) in a buffer containing 20 mM Tris-HCl (pH 7.5), 10 mM KCl, 10 mM MgCl₂ and 0.5 mM EDTA. The resulting duplex DNA was incubated with high-capacity streptavidin agarose beads (Thermo Pierce) at room temperature for 60 min. The beads were then washed three times with a buffer containing 20 mM Tris-HCl (pH 7.5), 10 mM KCl, and 0.5 mM EDTA to remove any single-stranded or unbound DNA.

The DNA-bound beads were then incubated with nuclear lysates (500 μg) at 4 °C for 2 h in a buffer containing 20 mM Tris-HCl (pH 7.5), 10 mM KCl, 10% glycerol and 0.5 mM EDTA. By assuming an average molecular weight of 50 kDa for nuclear proteins, the molar ratio of DNA probe/nuclear protein is estimated to be 1:100. The lesion-containing DNA and the corresponding lesion-free DNA were incubated respectively with the heavy- and light-isotope labeled lysates in the forward SILAC experiments. On the other hand, the lesion-containing DNA and the corresponding lesion-free DNA were incubated with the light- and heavy-isotope labeled lysates, respectively, in the reverse SILAC experiments. The beads were washed sequentially with a buffer containing 20 mM Tris-HCl (pH 7.5), 10 mM KCl, 0.5 mM EDTA, and increasing concentrations of NaCl (i.e., 50, 100, and 200 mM).

After washing, the beads harboring the lesion and control ODNs were combined and boiled for 10 min in 30 μL of 3× SDS-PAGE loading buffer (Bio-Rad). After centrifugation, the supernatant was removed, and proteins were eluted from the beads. The eluted proteins were separated on a 4%/15% stacking SDS-PAGE gel until the 10 and 15 kDa bands in the protein ladder were resolved from each other. The gel was stained for 15 min with Coomassie Blue, and destained overnight. The protein band above the streptavidin band was excised, cut into 1 mm² pieces, and incubated sequentially with rotation in 50 mM ammonium bicarbonate containing 25% and 50% acetonitrile. Gel pieces were dehydrated with acetonitrile, and incubated in a solution containing 10 mM dithiothreitol in 50 mM ammonium bicarbonate at 37 °C for 1 h. An aliquot of 2.5 volumes of acetonitrile was added to the solution to wash the gel pieces, and removed. Gel pieces were subsequently incubated in 55 mM iodoacetamide (IAA, Sigma) at room temperature for 30 min. Gel pieces were then incubated for 15 min each with 50 mM ammonium bicarbonate, 50% acetonitrile in 50 mM ammonium bicarbonate, and 100% acetonitrile. Proteins were digested in-gel with trypsin at 37 °C for 16 h, and peptides were eluted from gel pieces by incubating sequentially in 5% formic acid, 50% acetonitrile in 5% formic acid, and 70% acetonitrile in 5% formic acid for 15 min each. After elution, the peptide fractions were pooled, evaporated to dryness, and desalted using OMIX C₁₈ Tips (Agilent).

Mass Spectrometry.

The tryptic digestion mixtures were subjected to LC-MS/MS analysis on a Fusion Lumos Orbitrap Tribrid mass spectrometer (Thermo) coupled with an Easy nLC 1000 (Thermo) and a Flex nanoelectrospray ion source (Thermo). The mass spectrometer was equipped with a high-field asymmetric-waveform ion mobility spectrometry (FAIMS, Thermo), where the incorporation of FAIMS improves proteome coverage.²¹ The compensation voltages (CV) for FAIMS were set at −40, −60, and −80 V, each for 1/3 of a 3-s cycle, whereas the carrier gas flow rate was 4.2 L/min. The analytical column was packed in-house using 3 μm Reprosil-Pur C18-AQ resin (Dr. Maisch GmbH HPLC) in a ~25 cm long, 75 μm i.d. fused silica column. The trapping column was also prepared in-house using 5 μm Reprosil-Pur C18-AQ resin (Dr. Maisch GmbH HPLC) in a 4 cm long, 150 μm i.d. fused silica column. The LC gradient included 6–43% of buffer B (0.1% formic acid in 80% acetonitrile) at a flow rate of 0.3 μL/min. The spray voltage was 2 kV, the RF lens setting was 30%, and the ion transfer tube temperature was 320 °C. The mass spectrometer was operated in the data-dependent acquisition mode, where the 25 most abundant ions observed in MS were isolated sequentially in the quadrupole at an isolation window of 2.0 Th, fragmented in the higher-energy collisional dissociation (HCD) cell at a normalized collision energy (NCE) of 30. The MS/MS were acquired in the Orbitrap at a resolution of 50 000, where the maximum injection time was 50 ms and the AGC setting was 4 × 10⁵. The raw files were split based on CV using FAIMS_MzXML_Generator (https://github.com/coongroup/FAIMS-MzXML-Generator) and searched using Maxquant.

CRISPR/Cas9-Mediated Genome Editing of HEK293T Cells.

Genome editing with the CRISPR/Cas9 system^22,23 was conducted following the previously reported protocols, where the single guide RNAs (sgRNAs) were designed using the online sgRNA tool (http://www.broadinstitute.org/rnai/public/analysis-tools/sgrna-design).²⁴ ODNs corresponding to target sequences were obtained from Integrated DNA Technologies and inserted into the hSpCas9 plasmid pX330 (Addgene, Cambridge, MA). The guide sequence was 5′-GG GAA CTC AGC TCG GAG CTC TGG-3′, where the protospacer adjacent motif (PAM) sequence is underlined. The constructed plasmids were then transfected into HEK293T cells using Lipofectamine 2000 (Invitrogen) in a 6-well plate. After the transfection, individual cells were plated by dilution and cultured for further analysis. Genomic DNA was extracted from individual clonal cell lines and the specific DNA regions surrounding the targeted sites were screened by nested PCR, followed by agarose gel electrophoresis to assess the modification efficiency. Sanger sequencing was employed to identify the deletion loci. A set of clones with both alleles being successfully cleaved by Cas9 were isolated, and the successful deletion of the CDKN2AIP gene was also validated by Western blot analysis.

O⁶-Me-dG Repair Assay.

O⁶-Me-dG repair assay followed a previously published method.^20,25 In this vein, HEK293T cells, and the isogenic CDKN2AIP^−/− cells were seeded in 6-well plates and incubated for 1 h with 200 μM NMU, which was previously shown to induce O⁶-Me-dG in DNA.^10,25 The cells were subsequently changed to fresh medium without NMU and harvested immediately or 22 h later. Genomic DNA was extracted using the Qiagen DNeasy Blood & Tissue Kit.

1 μg of cellular DNA was digested with 1 unit of nuclease P1 from Penicillium citrinum (Sigma, N8630) and 0.00125 unit of phosphodiesterase II from bovine spleen (Sigma, P9041) in a 45-μL buffer containing 30 mM sodium acetate (pH 5.6), 1 mM ZnCl₂, and 2.5 nmol of erythro-9-(2-hydroxy-3-nonyl)-adenine (EHNA, Thermo Fisher Scientific, 126110) at 37 °C for 24 h. To the mixture were then added one unit of Quick CIP from Pichia pastoris (New England Biolabs, M0525S) and 0.0025 unit of phosphodiesterase I from Crotalus adamanteus venom (Sigma, P3243) in 60 μL of 500 mM Tris-HCl (pH 8.9), and the solution was incubated at 37 °C for 4 h. 100 fmol of [D₃]-O⁶-Me-dG and 16.2 pmol of [¹⁵N₅]-dG were added to each sample as internal standards. Enzymes were removed from the digestion mixtures by chloroform extraction, and the aqueous layer was dried by Speed-Vac, desalted by precipitation with 90% acetonitrile, and finally reconstituted in 20 μL of water for LC-MS/MS analysis.

LC-MS/MS analyses were conducted on a TSQ-Altis Plus triple-quadrupole mass spectrometer equipped with a Nano-spray Flex source and coupled with a Vanquish Neo UHPLC system for separation (Thermo Fisher Scientific, San Jose, CA). The samples were loaded onto a Thermo Fisher μ-Precolumn (C18 PepMap 100, 300 μm × 5 mm, 5 μm particle size, 100 Å in pore size) and injected into an in-house packed Zorbax SB-C18 (5 μm in particle size, 200 Å in pore size, Michrom BioResource, Auburn, CA) analytical column (75 μm × 20 cm). Formic acid (0.1%, v/v) in water and formic acid (0.1%, v/v) in 80% acetonitrile were used as mobile phases A and B, respectively. A 14 min linear gradient of 1–85% B at a flow rate of 300 nL/min was used for the separation.

The TSQ-Altis Plus triple-quadrupole mass spectrometer was operated in the positive-ion multiple-reaction monitoring (MRM) mode, where the neutral losses of a 2-deoxyribose (116 Da) from the [M + H]⁺ ions of dG (m/z 268 → 152), [¹⁵N₅]-dG (m/z 273 → 157), O⁶-Me-dG (m/z 282 → 166), and [D₃]-O⁶-Me-dG (m/z 285 → 169) were monitored. The voltage for electrospray was set at 1.5 kV, RF lens voltage was 29.0 V, and the ion transfer tube temperature was 350 °C. The resolutions of Q1 and Q3 were set at 0.7 and 1.2 Th full-width at half-maximum (FWHM), respectively. Fragmentation of precursor ions in Q2 was conducted with 1.5 mTorr argon and a collision energy of 20 V. The calibration curves of O⁶-Me-dG and dG were constructed by mixing different amounts of analytes with fixed amounts of [D₃]-O⁶-Me-dG and [¹⁵N₅]-dG. The numbers of moles of O⁶-Me-dG and dG in each sample were calculated from the peak area ratios of the analytes over the corresponding stable isotope-labeled standards, the calibration curves, and the amounts of stable isotope-labeled standards added. The levels of O⁶-Me-dG are reported as the number of lesions per 10⁶ dG.

Clonogenic Survival Assay.

Clonogenic survival assay was performed based on published procedures.²⁶ HEK293T and the isogenic CRISPR-engineered CDKN2AIP^−/− cells were plated in triplicate in six-well plates at a concentration of 300 cells per well. The cells were treated with NMU at concentrations of 0, 0.125, 0.25, 0.50, 1.0, and 2.0 mM. Subsequently, the cells were incubated at 37 °C in a 5% CO₂ environment for 7 days. The culture medium was removed, and the cells were rinsed with 6.0% glutaraldehyde in PBS. The rinsing solution was removed, and cell colonies were fixed and stained with an aqueous solution containing 6.0% glutaraldehyde and 0.5% crystal violet. The plates were subsequently rinsed with water, and dried at room temperature in air. The surviving fraction (SF) was calculated with the following equation:

$$SF=\frac{\left(\frac{N_{\text{colonies,NMU}}}{N_{\text{seeded,NMU}}}\right)}{\left(\frac{N_{\text{colonies,control}}}{N_{\text{seeded,control}}}\right)}$$

RESULTS

Quantitative Proteomic Screening of O⁶-Me-dG:dC- and O⁶-Me-dG:dT-Binding Proteins.

Exposure to endogenous and exogenous alkylating agents can lead to the formation of DNA adducts, including O⁶-Me-dG. DNA polymerases incorporate dTTP opposite O⁶-Me-dG, leading to G → A transition mutation.⁸ To understand how cells respond to O⁶-Me-dG in DNA when paired with a dT and dC, we set out to employ a quantitative proteomic approach to identify cellular proteins involved in binding to duplex DNA containing an O⁶-Me-dG:dC or O⁶-Me-dG:dT pair. For this purpose, we synthesized a 20mer ODN containing a site-specifically inserted O⁶-Me-dG, where the identities and purities of the O⁶-Me-dG-carrying ODN were confirmed by LC-MS and MS/MS analyses (Figure S1). We then annealed the O⁶-Me-dG-containing ODN with a 5′-biotin-conjugated complementary strand carrying a dC or dT opposite the O⁶-Me-dG. We chose to use a 20mer DNA probe due to its relative ease of preparation and purification as well as its use in previous proteomic studies to detect reader proteins of N²-nBu-dG as well as O²- and O⁴-n-butylthymidine.^13,20 In addition, the damage site is located 13 base pairs away from the biotin label in the complementary strand, which minimizes potential interference of streptavidin beads on protein–DNA interactions.

Using these two different O⁶-Me-dG-containing duplex ODNs and the corresponding lesion-free ODNs (dA:dT for O⁶-Me-dG:dT and dG:dC for O⁶-Me-dG:dC) as baits, we developed a SILAC-based quantitative proteomic method to discover cellular proteins that can bind to O⁶-Me-dG paired with a dC or dT. To remove experimental bias, we conducted both forward and reverse SILAC experiments.

The results from the quantitative proteomic experiments facilitated the identification of proteins that bind preferentially to duplex DNA harboring an O⁶-Me-dG:dC or O⁶-Me-dG:dT pair over the corresponding duplex with a dG:dC or dA:dT pair (Figure 2a,b). The criterion for assigning a protein as a reader was an average SILAC enrichment ratio of at least 1.5 of the damaged over the undamaged DNA probe. With this criterion, 67 and 31 proteins exhibited preferential binding toward O⁶-Me-dG:dT and O⁶-Me-dG:dC pairs, respectively, where 14 proteins were commonly identified for the two DNA probes (Figure 2c, Tables S1 and S2). A comparison of these results with our recently reported binding proteins for duplex DNA harboring an N²-nBu-dG:dC pair in the same sequence context showed only one common protein, i.e., YBX1,²⁰ underscoring that the major-groove lesion O⁶-Me-dG and the minor-groove lesion N²-nBu-dG are recognized by distinct cellular proteins.

Figure 2.

SILAC-based interaction screening led to the identification of candidate recognition proteins for duplex DNA containing an O⁶-Me-dG:dC or O⁶-Me-dG:dT base pair. Scatter plots showing the proteins identified from forward and reverse SILAC pull-down assays using an O⁶-Me-dG:dT DNA probe relative to a dA:dT probe with nuclear protein lysates (a), O⁶-Me-dG:dC DNA probe relative to the dG:dC probe with nuclear protein lysates (b), and a Venn diagram displaying the overlap in interacting proteins of O⁶-Me-dG:dC and O⁶-Me-dG:dT (c).

MutSα, which is a component of the DNA mismatch repair system and a heterodimer comprising DNA mismatch repair proteins 2 and 6 (MSH2 and MSH6), was previously shown to bind preferentially to O⁶-Me-dG paired with dT over dC.²⁷ Consistently, we found that the average SILAC ratios for MSH2 were 1.7 ± 0.5 and 1.2 ± 0.4 for DNA probes harboring the O⁶-Me-dG:dT and O⁶-Me-dG:dC base pairs, respectively. Representative ESI-MS of a tryptic peptide from MSH2 is shown in Figure 3.

Figure 3.

ESI-MS ((a, c) forward SILAC, (b, d) reverse SILAC) showing the [M + 2H]²⁺ ions of the light and heavy lysine-containing peptide, LTSLNEEYTK derived from MSH2 in O⁶-Me-dG:dT versus dA:dT pulldown (a, b) and O⁶-Me-dG:dC versus dG:dC pulldown (c, d). The calculated m/z for the monoisotopic peak of the [M + 2H]²⁺ ion of the light lysine-containing peptide is 599.3035.

CDKN2AIP was detected as a common reader protein of O⁶-Me-dG in both base pairing contexts, with the average SILAC ratios being 1.9 ± 0.6 and 1.6 ± 0.3 for O⁶-Me-dG:dT/dA:dT and O⁶-Me-dG:dC/dG:dC, respectively. Representative ESI-MS and MS/MS of a tryptic peptide from CDKN2AIP are displayed in Figures 4 and S3. We then chose to investigate further the role of CDKN2AIP in cellular response and repair of O⁶-Me-dG because it is known to interact with p53, possibly mediating p53-induced DNA damage response.^28–30

Figure 4.

ESI-MS showing the [M+2H]²⁺ ions of light and heavy lysine-containing peptides, PSSETASSGLTSK from CDKN2AIP in the O^`-Me-dG:dT versus dA:dT pulldown (a, b), and light and heavy lysine-containing peptide GISSSNEGVEEPSK from CDKN2AIP in the O⁶-Me-dG:dC versus dG:dC pulldown (c, d). The calculated m/z values for the monoisotopic peaks of the [M + 2H]²⁺ ions of the light lysine-containing peptides of PSSETASSGLTSK and GISSSNEGVEEPSK are m/z 626.3068 and 710.3335, respectively.

Genetic Ablation of CDKN2AIP in HEK293T Cells Led to Increased Resistance to NMU and Diminished Accumulation of NMU-Induced O⁶-Me-dG.

To assess the role of CDKN2AIP in cellular response and repair of O⁶-Me-dG, we ablated the CDKN2AIP gene in HEK293T cells using CRISPR-Cas9. Sanger sequencing showed out-of-frame deletion in CDKN2AIP gene, and Western blot analysis confirmed the absence of CDKN2AIP protein in CDKN2AIP^−/− cells (Figure S4). We then examined the survival rates of HEK293T and CDKN2AIP^−/− cells exposed to various concentrations of NMU, an alkylating agent able to induce O⁶-Me-dG.^10,25 The results from the clonogenic survival assay revealed that ablation of CDKN2AIP in HEK293T cells led to diminished sensitivity toward NMU (Figure 5a). For instance, 25% of HEK239T cells survived upon exposure to 1 mM NMU, whereas 52% of CDKN2AIP^−/− cells survived under the same conditions.

Figure 5.

Loss of CDKN2AIP rendered HEK293T cells more resistant to NMU and conferred attenuated accumulation of NMU-elicited O⁶-Me-dG in genomic DNA. (a) Clonogenic survival assay results for HEK293T and the isogenic CDKN2AIP^−/− cells upon exposure to different concentrations of NMU (n = 3). (b) LC-MS/MS quantification results of O⁶-Me-dG in the genomic DNA of HEK293T and the isogenic CDKN2AIP^−/− cells harvested immediately and at 22 h after a 1.0-h exposure to 200 μM NMU. The p values were calculated by using two-way ANOVA with Sidak’s multiple comparisons test: *p < 0.05; ns, p > 0.05.

We also conducted an LC-MS/MS-based assay to examine how loss of CDKN2AIP modulates the accumulation of NMU-induced O⁶-Me-dG in genomic DNA. To this end, we extracted genomic DNA from HEK293T and the isogenic CDKN2AIP^−/− cells at 0 and 22 h following a 1.0-h exposure to 200 μM NMU, digested the DNA to single nucleosides, and subjected the digestion mixture to LC-MS/MS analysis with the stable isotope-dilution method. Our results revealed that genetic ablation of CDKN2AIP gave rise to diminished accumulation of O⁶-Me-dG in genomic DNA (Figures 5b and S5). For instance, the levels of O⁶-Me-dG were 107 ± 3 and 66 ± 23 lesions per 10⁶ dG in HEK293T and CDKN2AIP^−/− cells, respectively, immediately following NMU exposure (Figure 5b). These data corroborate with the above clonogenic survival assay results and suggest the binding of CDKN2AIP may hinder the repair of O⁶-Me-dG.

DISCUSSION

Alkylation represents an important class of DNA damage.^5,31 Among the alkylated DNA lesions, O⁶-Me-dG is mutagenic, and it directs a preferential misincorporation of thymidine during DNA replication.⁸ Owing to its deleterious effects, cells are equipped with pathways to repair O⁶-Me-dG, including MGMT and mismatch repair.⁵ However, there remain to be discovered additional cellular proteins that can recognize O⁶-Me-dG. Here, we used an unbiased SILAC-based quantitative proteomics approach to identify reader proteins of O⁶-Me-dG paired with dC or dT without a priori knowledge.

Our quantitative proteomic experiments led to the identification of proteins exhibiting preferential binding to O⁶-Me-dG:dC- and O⁶-Me-dG:dT-containing DNA compared to the corresponding unmodified DNA (Figure 2). The ratios obtained from the forward and reverse SILAC experiments are consistent (Figure 2a,b), demonstrating the robustness of the method. Aside from proteins that can bind to both O⁶-Me-dG-bearing duplex DNA sequences, we also identified more candidate recognition proteins of the O⁶-Me-dG:dC pair than the O⁶-Me-dG:dT pair, suggesting that cellular proteins can differentiate the wobble and pseudo-Watson–Crick base pair configurations assumed by the O⁶-Me-dG:dC and O⁶-Me-dG:dT pairs, respectively (Figures 1a and 2c).

We were able to identify proteins known to be involved in DNA damage repair. These proteins include FANCI for only the O⁶-Me-dG:dC-containing duplex DNA, and MSH2 exhibits preferential binding toward O⁶-Me-dG:dT over O⁶-Me-dG:dC-bearing duplex DNA (Tables S1 and S2). In particular, MSH2 was detected with a higher SILAC ratio for O⁶-Me-dG when it is paired with dT than dC in our proteomic experiments, which is in keeping with the previous finding that MSH2 displayed preferential binding toward O⁶-Me-dG:dT over O⁶-Me-dG:dC in vitro.²⁷ Our quantitative proteomic experiments, however, did not lead to the identification of MGMT as a reader protein for O⁶-Me-dG, which could be attributed to its relatively low level of expression; it is also possible that MGMT repairs the lesion and thus loses its ability to bind with O⁶-Me-dG-containing DNA.

Among the newly identified O⁶-Me-dG-binding proteins, CDKN2AIP is particularly interesting. CDKN2AIP is an interaction partner of p53, and activation of p53 by CDKN2AIP occurs through pathways that are dependent or independent of ADP-ribosylation factor (ARF). CDKN2AIP collaborates with ARF and MDM2 E3 ubiquitin ligase to activate p53 in an ARF-dependent pathway.^32,33 On the other hand, CDKN2AIP interacts directly and stabilizes p53 in an ARF-independent pathway.³⁴ We assessed the role of CDKN2AIP in DNA damage response upon exposure to an alkylating agent that can induce O⁶-Me-dG. Our results from clonogenic survival assay revealed an augmented resistance of CDKN2AIP^−/− cells to NMU than parental HEK293T cells (Figure 5a). In addition, our LC-MS/MS data showed that loss of CDKN2AIP led to attenuated accumulation of NMU-elicited O⁶-Me-dG in genomic DNA (Figure 5b).

A possible explanation for the increased survival of CDKN2AIP^−/− cells is that CDKN2AIP may bind to O⁶-Me-dG, thereby shielding it from DNA repair proteins, e.g., the mismatch repair machinery and/or MGMT. Human upstream binding factor (hUBF) and MSH2 were shown to bind to cisplatin-induced DNA adducts at a K_d in the picomolar range.^35–37 The binding of hUBF to cisplatin-induced DNA adducts led to diminished rRNA synthesis.³⁵ In addition, the TATA box-binding protein (TBP)/TFIID binds to DNA adducts induced by UV irradiation and cisplatin.³⁸ Moreover, the binding of YBX1 to 8-oxo-2′-deoxyguanosine was found to suppress the repair of the lesion.³⁹ Therefore, in addition to inducing mutations, O⁶-Me-dG may also exert its cytotoxic effects partly through elevated engagement with DNA-binding proteins such as CDKN2AIP.

CONCLUSIONS

In this work, by using an unbiased quantitative proteomics approach, we identified multiple candidate DNA damage-recognition proteins for O⁶-Me-dG paired with dC or dT. We revealed a new function of one of these proteins, i.e., CDKN2AIP, in conferring increased cellular sensitivity to NMU by eliciting elevated accumulation of NMU-induced O⁶-Me-dG. Hence, this study expanded the functions of CDKN2AIP, and generated a list of candidate readers of O⁶-Me-dG that can be further explored in future biological studies.

We anticipate that the method will have broad applicability for the identification of proteins recognizing other types of DNA lesions. In addition, by inserting damage-containing ODNs with an affinity handle (e.g., a biotin tag) into a double-stranded plasmid, we envisage the method can be adapted for characterizing protein-damaged DNA interactions in live cells.

Supplementary Material

ASSOCIATED CONTENT

Supporting Information

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.analchem.5c03119.

A list of all candidate O⁶-Me-dG-binding proteins, LC-MS and MS/MS results, and calibration curves (PDF)