Processing of N⁵-substituted formamidopyrimidine DNA adducts by DNA glycosylases NEIL1 and NEIL3

Abstract

A variety of agents cause DNA base alkylation damage, including the known hepatocarcinogen aflatoxin B₁ (AFB₁) and chemotherapeutic drugs derived from nitrogen mustard (NM). The N7 site of guanine is the primary site of alkylation, with some N7-deoxyguanosine adducts undergoing imidazole ring-opening to stable mutagenic N⁵-alkyl formamidopyrimidine (Fapy-dG) adducts. These adducts exist as a mixture of canonical β- and unnatural α-anomeric forms. The β species are predominant in double-stranded (ds) DNA. Recently, we have demonstrated that the DNA glycosylase NEIL1 can initiate repair of AFB₁-Fapy-dG adducts both in vitro and in vivo, with Neil1^−/− mice showing an increased susceptibility to AFB₁-induced hepatocellular carcinoma.

Here, we hypothesized that NEIL1 could excise NM-Fapy-dG and that NEIL3, a closely related DNA glycosylase, could excise both NM-Fapy-dG and AFB₁-Fapy-dG. Product formation from the reaction of human NEIL1 with ds oligodeoxynucleotides containing a unique NM-Fapy-dG followed a bi-component exponential function under single turnover conditions. Thus, two adduct conformations were differentially recognized by hNEIL1. The excision rate of the major form (~ 13.0 min⁻¹), presumed to be the β-anomer, was significantly higher than that previously reported for 5-hydroxycytosine, 5-hydroxyuracil, thymine glycol (Tg), and AFB₁-Fapy-dG. Product generation from the minor form was much slower (~ 0.4 min ⁻¹), likely reflecting the rate of conversion of the α anomer into the β anomer. Mus musculus NEIL3 (MmuNEIL3Δ324) excised NM-Fapy-dG from single-stranded (ss) DNA (turnover rate of ~0.4 min⁻¹), but not from ds DNA. Product formation from ss substrate was incomplete, presumably because of a substantial presence of the α anomer. MmuNEIL3Δ324 could not initiate repair of AFB₁-Fapy-dG in either ds or ss DNA. Overall, the data suggest that both NEIL1 and NEIL3 may protect cells against cytotoxic and mutagenic effects of NM-Fapy-dG, but NEIL1 may have a unique role in initiation of base excision repair of AFB₁-Fapy-dG.

1. Introduction

Cellular DNA is susceptible to attack by alkylating agents that could originate from either normal metabolic processes or as result of exposure to various environmental toxicants. The latter compounds include aflatoxin B₁ (AFB₁), the product of Aspergillus flavus and related fungi that contaminate food crops. Chronic dietary exposure to AFB₁ represents an increased etiological risk for developing hepatocellular carcinoma (HCC) [1,2]. DNA alkylation is also induced by many chemotherapeutic drugs, such as temozolomide and nitrogen mustards (NMs), and is largely responsible for therapeutic effectiveness of these agents [3,4].

The guanine N7 atom is the most nucleophilic site within DNA bases [5] and is the primary target of alkylation (Fig. 1). Initial N7-deoxyguanosine (N7-dG) adducts can undergo secondary reactions because of the positive charge on the imidazole ring, yielding either an abasic site or the N⁵-substituted formamidopyrimidine (Fapy-dG) adducts [6]. The alkyl-Fapy-dGs are more chemically stable than the corresponding N7-alkylated dGs, leading to much longer half-lives of the alkylated lesions in cellular DNA [6–9]. The analyses of DNA modifications in livers of Fischer rats that were intraperitoneally injected with a single AFB₁ dose demonstrated rapid disappearance of the original N7-dG adducts, with an apparent half-life of 7.5 h [8]. Approximately 20% of these adducts were converted to AFB₁-Fapy-dGs that became the dominant lesions in 24 h and persisted during the 72-h period studied. When DNA repair-proficient mice were intraperitoneally injected with a single, non-lethal AFB₁ dose, the AFB₁-N7-dG and AFB₁-Fapy-dG adducts were present in liver DNA 6 h post-injection at comparable levels, approximately 50 and 68 pmoles/mg, respectively [9]. By 48 h, the initial N7-dG adducts essentially disappeared, while the level of AFB₁-Fapy-dGs was reduced less than 2-fold. Although kinetics of conversion have not been addressed in vivo for the NM-induced adducts, N7-dG and Fapy-dG modifications were detected in cultured cancer cells treated for 24 h with bis(2-chloroethyl)amine at the levels of 970 and 180 adducts per 10⁷ nucleotides, respectively [7]. As expected, this bifunctional agent also induced the formation of DNA crosslinks, including both NM-N7-dG-N7-dG (240 per 10⁷ nucleotides) and NM-Fapy-dG-N7-dG (6 per 10⁷ nucleotides).

Fig. 1.

Formation of N⁶-substituted formamidopyrimidine-dG adducts.

Structural and biochemical investigations on Fapy-dG adducts are complicated by the fact that these adducts exist as a mixture of interconverting α- and β-anomeric forms (Fig. 1). The β species are predominant in double-stranded (ds) DNA. This has been shown for the temozolomide-induced methyl-Fapy-dG (Me-Fapy-dG) [10], NM-Fapy-dG [11], and AFB₁-Fapy-dG [12–15].

The N⁵-substituted Fapy-dG adducts manifest a strong mutator phenotype. As demonstrated by analyses of sequence alterations following replication of site-specifically modified vectors in primate (simian kidney COS7) cells, the spectra of mutations induced by these lesions were dominated by base substitutions, with G to T transversions being most common [16–18]. The overall mutation frequencies measured for Me-Fapy-dG ranged from ~9 to 21%, depending on the local sequence context [16], and ~ 11–12% for NM-Fapy-dG [18]. The AFB₁-Fapy-dG adduct caused mutations at an unprecedentedly high ~97% frequency, including ~86% G to T transversions [17]. This spectrum parallels the mutational signature of AFB₁ that was determined using DNA from exposed cultured cells or experimental animals, or from AFB₁-associated HCCs [19–23].

AFB₁-Fapy-dG has been long recognized as a substrate for nucleotide excision repair [24,25]. However, we recently demonstrated that this lesion is removed by base excision repair. We showed that DNA glycosylase NEIL1, a member of the Fpg/Nei glycosylase family, could excise AFB₁-Fapy-dG from site-specifically modified oligodeoxynucleotides in vitro and that Neil1^−/− mice accumulated significantly more of these lesions in liver DNA following a single AFB₁ injection than Neil1 ^+/+ mice. Furthermore, Neil1 ^−/− mice showed an increased susceptibility to AFB₁-induced HCC versus control mice, implicating the AFB₁-Fapy-dG adduct as a major contributor to AFB₁-induced carcinogenesis [9].

In addition to AFB₁-Fapy-dG, NEIL1 recognizes unsubstituted Fapy-dG and Fapy-dA, Me-Fapy-dG, thymine glycol (Tg), psoralen-induced adducts, and oxidation products of 8-oxo-deoxyguanosine, guanidino-hydantoin and spiroiminohydantoin (Sp) [26–37]. NEIL3, another member of the Fpg/Nei DNA glycosylase family, recognizes a broad class of lesions that significantly overlaps with that of NEIL1 [37–42]. Germane to the present study, the shared substrates include unsubstituted Fapy-dG and Me-Fapy-dG. Thus, we hypothesized that NEIL3 would recognize AFB1-Fapy-dG and that both NEIL1 and NEIL3 would be able to initiate repair of NM-Fapy-dG. To address these possibilities, we measured the kinetics of product formation for the reaction of site-specifically modified oligodeoxynucleotides containing these or control lesions with human NEIL1 (hNEIL1) [32] or a truncated form of Mus musculus NEIL3 (MmuNEIL3Δ324) [38,42].

2. Materials and methods

2.1. DNA substrates

The oligodeoxynucleotides containing Me-Fapy-dG, NM-Fapy-dG, or AFB₁-Fapy-dG were synthesized as previously described [10,11,43]. The non-adducted and Tg-containing oligodeoxynucleotides were obtained from Integrated DNA Technologies, Inc. The sequences of modified DNAs were 5′-ACCACGCTAGCXAGTCCTAACAAC-3′ (where X is Me-Fapy-dG or NM-Fapy-dG) or 5′-ACCACTACTATXATTCATAACAAC-3′ (where X is AFB₁-Fapy-dG or Tg). The adducted oligodeoxynucleotides were labeled at the 5′ termini using [γ−³²P]ATP (PerkinElmer Life Sciences) and T4 polynucleotide kinase (New England BioLabs). The ds substrates were prepared as described in our prior study [9].

2.2. Enzymes

The hNEIL1 and MmuNEIL3Δ324 glycosylases were purified as described in Roy et al. [32] and Liu et al. [42], respectively. MmuNEIL3Δ324 was the generous gift of Dr. Sylvie Doublié (University of Vermont). Human apurinic/apyrimidinic (AP) endonuclease, APE1, was purchased from New England BioLabs Inc.

2.3. DNA glycosylase assays

The glycosylase reactions were conducted at 37 °C in buffers previously optimized and used for hNEIL1 (20 mM Tris-HCl, pH 7.4, 100 mM KCl, and 100 μg/ml BSA) [9,32] or MmuNEIL3Δ324 (20 mM HEPES-Na, pH 7.0, 50 mM NaCl, 100 μg/ml BSA, 0.01% Triton X-100, and 1 mM dithiothreitol) [38,42]. In order to analyze glycosylase activity independently of the AP lyase activity, the reactions were terminated by addition of equal volume of 0.5 N NaOH, followed by incubation at 90 °C for 5 min. Enzyme concentrations, substrate concentrations, and incubation times varied and are specified in the figures or figure legends. The rate constants were measured under single turnover conditions as previously described [9]. Following separation by polyacrylamide gel electrophoresis, the reaction products were visualized using a Personal Molecular Imager™ System (Bio-Rad). The product was quantified from a phosphor screen image by the Quantity One Software (Bio-Rad). The percent of non-specific products, amplitude of reaction, and rate constants were calculated using Kalei-daGraph software. All experiments were repeated at least three times, with representative gel images and curves shown. The data were expressed as average ± standard deviation.

3. Results and discussion

3.1. hNEIL1 can initiate repair of NM-Fapy-dG

A site-specifically modified ds oligodeoxynucleotide (24-mer) containing a NM-Fapy-dG at the 12^th position opposite deoxycytidine was used to examine hNEIL1 ability to initiate repair of NM-Fapy-dG; the same construct, except with a Tg adduct opposite deoxyadenosine was used as a control. Since the release of NEIL1 from its products is rate limiting, the reactions were conducted under single turnover conditions. The kinetics of product formation for the control Tg substrate followed a single exponential rise (Eq. (1)), with a strong correlation between the experimental points and the curve (R² > 0.99), (Fig. 2A & C). The observed excision rate constant (k_obs) calculated from this fit was 1.2 min⁻¹, which was in good agreement with results of our prior study (1.4 min⁻¹) [9] and the observed rate constant reported for Tg by Krishnamurthy and colleagues (1.3 min⁻¹) [34].

Fig. 2.

hNEILl-catalyzed incision at NM-Fapy-dG. The reactions were conducted at 37 °C in the presence of hNEIL1 (750 nM) using ³²P-labeled ds oligodeoxynucleotides (20 nM) that contained either Tg (A) or NM-Fapy-dG (B). Aliquots were removed at the indicated times, and following separation by gelelectrophoresis, DNA was visualized using a phosphor screen and a Personal Molecular Imager™ System (Bio-Rad). Representative gel images are shown. The product formation (P) was plotted as a function of time (t) (C), and the nonspecific product (P_ns), amplitude of substrate used (S), and observed rate constant (k_obs) values were obtained from the best fit of the data to a single (Eq. (1)) or bi-component (Eq. (2)) exponential function using KaleidaGraph software.

The experimental points from the same analysis of the NM-Fapy-dG-containing DNA significantly deviated from a single exponential rise curve, but fitted with strong correlation (R² > 0.99) to a bi-component exponential rise function (Eq. (2)) (Fig. 2B & C). We calculated rate constants of 13.0 ± 6.4 min⁻¹ (k_obs1) and 0.4 ± 0.2 min⁻¹ (k_obs2) from three independent experiments; the corresponding extrapolated amplitudes were 43.0 ± 13.0% (S₁) and 36.0 ± 1.0% (S2) of the initial substrate. The non-specific cleavage product accounted for 12.5 ± 2.0%.

To address the nature of non-specific cleavage, DNA was incubated in the absence of enzyme, with or without subsequent treatment with NaOH. In addition, DNA was reacted with APE1. These tests demonstrated that DNA was stable under the glycosylase reaction conditions, and that preexisting nicks and abasic sites accounted for only a small fraction of non-specific product (< 3% of the initial DNA); the non-specific cleavage largely occurred during alkali treatment (data not shown).

The DNA substrate existing as two or more structures that were differentially recognized by hNEIL1 could explain the complex kinetics of product formation described above. In support of this explanation, NM-Fapy-dG is known to exist as a slowly equilibrating mixture of α- and β-anomeric forms (Fig. 1), with the β form being predominant in ds DNA [11]. In addition, various configurations of the modified base are possible based on analogy with Me-Fapy-dG [44] and AFB₁-Fapy-dG [12–15], including geometrical isomers of the formamide and atropisomers at the C5-N⁵ bond. Since the β anomer is a natural state of nucleosides in DNA, we inferred that hNEIL1 incised β-anomeric adducts more efficiently. This interpretation would be consistent with our prior finding that only the β anomer of AFB₁-Fapy-dG is a substrate for hNEIL1 [9].

We considered several possibilities concerning the slower reacting fraction. In particular, hNEIL1 could discriminate between different base configurations of β-NM-Fapy-dG (e.g., formamide geometric isomers or atropisomers), as was previously observed for diastereomers of Sp [34]. Structural data regarding the isomeric composition of NM-Fapy-dG in ds DNA that would either support or dispute this hypothesis are currently unavailable. However, the ratio of differentially processed species, as estimated from the extrapolated amplitudes, was ~ 1.2 to 1, with the dominating species being a preferred NEIL1 substrate. Since this ratio is close to the previously estimated β/α anomeric composition of NM-Fapy-dG in ds DNA [11], we hypothesize that the products formed in the slower reaction originated from the α form. The probability of hNEIL1 directly acting on the α form is low, as a major rearrangement of DNA substrate would be required for proper positioning of the glycosidic bond in the active site. Thus, we favor the hypothesis that product formation through this route was the result of slow, rate-limiting conversion of the α anomer into the β anomer, followed by fast incision at the β anomer. This scenario requires that the enzyme environment accelerates the rate of interconversion between two anomers, since the equilibration of anomers in free ds DNA occurs over a period of days [11].

Interestingly, product formation of the AFB₁-Fapy-dG followed a single exponential rise under identical conditions (Eq. (1)) [9]. A likely reason for this result is that AFB₁-Fapy-dG exists in ds DNA almost exclusively in the β anomeric form [12–15]. The stabilization of the overall structure has been attributed to the stacking interactions of the AFB₁ moiety with the 5′ neighboring base. Since the NM moiety is expected to be more solvent exposed, the observed heterogeneity of this adduct can be easily explained. Thus, collectively with our prior analyses on hNEIL1-catalyzed excision of AFB₁-Fapy-dG [9], the present data on NM-Fapy-dG provide further evidence for structural differences between these two adducts.

The observed rate constant of hNEIL1-catalyzed excision of β-NM-Fapy-dG (13.0 ± 6.4 min⁻¹) was ~ 75-fold higher than that estimated for p-AFB₁-Fapy-dG (0.17 ± 0.03 min⁻¹) [9], which can be attributed to both size and structural differences, especially in regard to intercalation as discussed above. The excision rate of β-NM-Fapy-dG by hNEIL1 was much faster than the excision rates of 5-hydroxycytosine (0.24 min⁻¹), 5-hydroxyuracil (0.14 min⁻¹), and Tg (1.2–1.4 min⁻¹), but significantly slower than that of guanidinohydantoin (104 min⁻¹), Sp1 (117 min⁻¹), and Sp2 (139 min⁻¹) in their respective canonical base pairs [9,34,36]. Further analyses are needed to compare NM-Fapy-dG and the unsubstituted Fapy-dG regarding both the incision rates and kinetics.

3.2. MmuNEIL3Δ324 can initiate repair of NM-Fapy-dG but notAFB₁-Fapy-dG

Human and mouse NEIL3 orthologs share 74% sequence identity.

Prior analyses have demonstrated that hNEIL3Δ324 and MmuNEIL3A324 exhibit similar activity on a number of DNA substrates, including Tg and Sp [42]. The direct comparison of MmuNEIL3Δ324 with MmuNEIL3 showed that the preferred substrates, such as Sp, guanidinohydantoin, Me-Fapy-dG, and an abasic site in single-stranded (ss) DNA, were recognized by these enzymes equally well [38]. Thus, we utilized MmuNEIL3Δ324 as a model for characterization of NEIL3 glycosylase activity.

An initial assessment for the ability of MmuNEIL3Δ324 to initiate repair of NM- and AFB₁-induced Fapy-dG lesions was performed with ds oligodeoxynucleotides using a 100:1 ratio of enzyme to substrate (Fig. 3A). DNA substrates were constructed as above and included NM-Fapy-dG, AFB₁-Fapy-dG, and two reference adducts, Me-Fapy-dG and Tg. In control reactions with NEIL1, all four substrates were cleaved to near completion (Fig. 3A). In agreement with previous results [38], recognition of Me-Fapy-dG and Tg by MmuNEIL3Δ324 in the context of ds DNAs was poor, with only 4.0 ± 1.1 and 3.5 ± 1.1% product formed, respectively. No evidence was found for MmuNEIL3Δ324-catalyzed excision of either NM-Fapy-dG or AFB₁-Fapy-dG from ds DNA.

Fig. 3.

MmuNEIL3Δ324-catalyzed incision at substituted Fapy-dG adducts. The reactions were conducted using ³²P-labeled ds (A) or ss (B) oligodeoxynucleotides (5 nM) for 30 min at 37 °C. The kinetic analysis (C, D) was performed at 37 °C in the presence of MmuNEIL3Δ324 (750 nM) using ³²P-labeled ss oligodeoxynucleotides (20 nM); aliquots were removed at the indicated times. Following separation by gelelectrophoresis, DNA was visualized using a phosphor screen and Personal Molecular Imager™ System (Bio-Rad). Representative gel images are shown (A–C). N1 and N3 denote hNEIL1 and MmuNEIL3Δ324, respectively. The product formation (P) was plotted as a function of time (t) (D), and the non-specific product (P_ns), amplitude of substrate used (S), and observed rate constant (k_obs) values were obtained from the best fit of the data to a single (Eq. (1)) exponential function using KaleidaGraph software.

NEIL3 prefers ss DNA [38–42]. Thus, the MmuNEIL3Δ324-initiated repair of substituted Fapy-dG adducts was examined using the same set of modified oligodeoxynucleotides but in ss form (Fig. 3B). These data demonstrated that MmuNEIL3Δ324 recognized NM-Fapy-dG and Me-Fapy-dG equally and both were better substrates than Tg. When the product accumulation was not limited by substrate availabilities, at a 1:1 enzyme to DNA ratio, the amount of products, corrected for nonspecific cleavage, was 19.3 ± 2.6, 22.4 ± 4.0, and 4.1 ± 0.8%, respectively. A similar trend was also observed at 10-fold excess of MmuNEIL3Δ324, with 36.0 ± 1.6, 35.5 ± 2.7, and 22.2 ± 1.3% product formation, respectively. In contrast, no incision was detected at these enzyme concentrations on AFB₁-Fapy-dG-containing ss DNA. We reexamined the ability of MmuNEIL3Δ324 to initiate repair of AFB₁-Fapy-dG in ss DNA using 100-fold excess enzyme concentration over substrate concentration but observed no products (data not shown).

To further elucidate the MmuNEIL3Δ324-catalyzed repair of NM-Fapy-dG from ss DNA, reactions were conducted under single turnover conditions (Fig. 3C). The kinetics of product accumulation strongly correlated with a single exponential rise function (Eq. [1], R² > 0.99). The observed rate constant (k_obs) was 0.4 ± 0.1 min⁻¹ and the extrapolated amplitude (S) was 33.5 ± 2.0% of the initial substrate. We assume that these values reflect the excision rate constant of MmuNEIL3Δ324 for the β-anomeric adduct and the fraction of these species in ss DNA, respectively. Consistent with this interpretation, the β anomer was estimated to constitute ~ 35–40% NM-Fapy-dG when the lesion was positioned at the primer/template junction on template DNA [18]. The extrapolated non-specific cleavage product accounted for 19.0 ± 1.0%, which was notably higher relative to that observed in hNEIL1-catalyzed reactions using ds DNA (12.5 ± 2.0%). The latter difference probably reflects a decreased stability of NM-Fapy-dG in ss versus ds DNA, likely because of a higher fraction of the α-anomeric form.

Relative to Sp1, which is an excellent substrate for MmuNEIL3Δ324, the rate of excision of β-NM-Fapy-dG in ss DNA was slow. The k_obs values reported for Sp1 in ss and ds DNA were ~184 and 3.2 min⁻¹, respectively [38]. However, excision of NM-Fapy-dG from ss DNA under multiple turnover conditions (Fig. 3B) was more efficient than that of Tg and comparable with the excision of Me-Fapy-dG.

In conclusion, our study has contributed to the characterization of substrate specificities of NEIL1 and NEIL3 and provides new insights into the conformation of the NM-Fapy-dG adduct. No evidence was found for the ability of MmuNEIL3Δ324 to act on AFB₁-Fapy-dG, suggesting that the role of NEIL1 in base excision repair of this lesion could be unique. Consistent with the idea that there is no backup glycosylase to compensate for this NEIL1 function, Neil1^−/− mice accumulated significantly more AFB₁-Fapy-dG adducts in liver DNA than Neil1^+/+ mice and showed an increased susceptibility to AFB₁-induced HCC [9].

The present study established that hNEIL1 and MmuNEIL3Δ324 both could initiate repair of NM-Fapy-dG. Except for the oxidation products of 8-oxo-deoxyguanosine, NM-Fapy-dG was excised as well as other, previously known substrates of these glycosylases, or better. Thus, NEIL1 and NEIL3 may protect cells against cytotoxic and mutagenic effects of NM-Fapy-dG. The kinetics of NM-Fapy-dG excision showed that the adduct existed as a complex mixture of species that are differentially recognized by hNEIL1 and MmuNEIL3Δ324. We infer that only the β-anomeric form of the adduct is a substrate for these glycosylases. The fraction of the incision-resistant α anomer was substantial in both ss and ds DNA. The contribution of individual anomers to biological effects of DNA adducts is an important but poorly elucidated problem. Specifically with regard to the NM-Fapy-dG adduct, we recently demonstrated that although the α anomer could contribute to induction of base substitutions during replication of ss vector DNA in primate cells, it was a significantly stronger block for DNA synthesis than its β counterpart [18]. Thus, the α anomer is anticipated to be more cytotoxic and in the context of genomic DNA, is hypothesized to cause gross DNA alterations rather than point mutations. In addition, our study demonstrated that kinetic analyses conducted under single turnover conditions can be informative regarding the anomeric composition of Fapy-dG adducts in DNA. Using hNEIL1 and MmuNEIL3Δ324 for evaluation of the anomeric ratio in ds and ss DNA, effects of the sequence context, substituent nature, and other parameters on this ratio can be addressed.

Abbreviations:

AFB₁

aflatoxin B₁

NM

nitrogen mustard

N7-dG

N7-substituted deoxyguanosine

Fapy-dG

formamidopyrimidine deoxyguanosine adduct

ds

double-stranded

ss

single-stranded

HCC

hepatocellular carcinoma

Me

methyl

Tg

thymine glycol

Sp

spiroiminohydantoin

AP

apurinic/apyrimidinic