1,N6-α-hydroxypropanoadenine (HPA) is an exocyclic DNA adduct of acrolein – an environmental pollutant and endocellular oxidative stress product. Escherichia coli AlkB dioxygenase belongs to the superfamily of α-ketoglutarate (αKG)- and iron-dependent dioxygenases which remove alkyl lesions from bases via an oxidative mechanism, thereby restoring native DNA structure. Here, we provide in vivo and in vitro evidence that HPA is mutagenic and is effectively repaired by AlkB dioxygenase. HPA generated in plasmid DNA caused A → C and A → T transversions and, less frequently, A → G transitions. The lesion was efficiently repaired by purified AlkB protein; the optimal pH, Fe(II), and αKG concentrations for this reaction were determined. In vitro kinetic data show that the protonated form of HPA is preferentially repaired by AlkB, albeit the reaction is stereoselective. Moreover, the number of reaction cycles carried out by an AlkB molecule remains limited. Molecular modeling of the T(HPA)T/AlkB complex demonstrated that the R stereoisomer in the equatorial conformation of the HPA hydroxyl group is strongly preferred, while the S stereoisomer seems to be susceptible to AlkB-directed oxidative hydroxylation only when HPA adopts the syn conformation around the glycosidic bond. In addition to the biochemical activity assays, substrate binding to the protein was monitored by differential scanning fluorimetry allowing identification of the active protein form, with cofactor and cosubstrate bound, and monitoring of substrate binding. In contrast FTO, a human AlkB homolog, failed to bind an ssDNA trimer carrying HPA.

Introduction

Acrolein (ACR), a highly reactive α,β-unsaturated aldehyde, is a ubiquitous environmental contaminant formed during incomplete combustion and overcooking of various food products, it is also used for the synthesis of diverse organic compounds, including methionine and the refrigerant, methyl chloride. The general population is exposed to ACR via smoking, second-hand tobacco smoke inhalation, and exposure to wood or plastics smoke. The distribution of the ACR–DNA adduct in the TP53 gene has been shown to coincide with the p53 mutational pattern in cigarette smoke-related lung cancer [1]. Firefighters and populations living or working in areas with heavy automotive traffic are exposed to higher levels of ACR by inhalation of ambient air. Ample data indicate that ACR causes irritation of many organs (for review see [2]). It is formed during the metabolism of the anticancer drug cyclophosphamide, and appears to be involved in the bladder toxicity of this drug [3,4], and has been shown to initiate tumor growth in the rat bladder [5]. Acrolein can also be generated endogenously as a by-product of lipid peroxidation [6]. Its interactions with diverse cellular molecules, including amino acids, proteins, and nucleic acids, have been extensively studied (reviewed in [7]). ACR is strongly mutagenic in the DNA-repair-deficient xeroderma pigmentosum fibroblasts but not in normal human fibroblasts and has also been found to cause an inhibitory effect on DNA repair. It seems that this effect is through its direct interaction with DNA repair proteins [8].

Acrolein mainly forms exocyclic adducts to DNA [9], mostly at guanine residues 1,N2-α-hydroxypropanoguanine and 1,N2-γ-hydroxypropanoguanine (HPGs) [10]. The C5 cytosine methylation at CpG sequences greatly enhances the ACR–dG adduction at those sites. Escherichia coli (E. coli) nucleotide excision repair (NER) enzymes, UvrA, UvrB, and UvrC, working as UvrABC nuclease, can incise HPGs specifically and quantitatively [1].

In addition to the guanine adducts, 1,N6-α-hydroxypropanoadenine (HPA) (see scheme in Figure 3 for structure) and 3,N4-α-hydroxypropanocytosine (HPC) are minor cyclic acrolein adducts found in DNA. To the best of our knowledge, the only work describing the possible mutagenicity of HPA is reference [11], where 22% of base substitutions induced by ACR in human fibroblasts were found at A·T sites. The formation of this adduct was also shown in rat liver epithelial RL34 cells exposed to acrolein and in rats exposed to Fe3+NTA, a carcinogenic iron chelate that specifically induces oxidative stress in the kidney of rodents [12].

Exposure of prokaryotic cells to sublethal doses of DNA alkylating agents induces expression of four genes, ada, alkB, alkA, and aidB, forming an adaptive response system. Among the proteins induced in this system are AlkA glycosylase and AlkB dioxygenase, both displaying broad substrate specificity [1316]. E. coli AlkB is one of the most studied members of the superfamily of α-ketoglutarate (αKG)- and iron-dependent dioxygenases [17]. Its homologs are present in almost all organisms including humans [18,19]. A previous bioinformatic analysis of protein sequence databases identified 1943 AlkB sequences with eight new AlkB subfamilies [20]. AlkB-like DNA repair dioxygenases remove alkyl lesions from DNA bases via an oxidative mechanism, restoring the native bases in the DNA. They require oxygen (O2) and non-heme Fe(II) as cofactors and αKG as a cosubstrate to initiate oxidative demethylation of DNA bases, resulting in the formation of succinate and CO2. E. coli AlkB acts preferably on ssDNA effectively oxidizing its substrates in short oligodeoxynucleotides such as pentamers and trimers [21]. Methylated dsDNA and RNA can also be repaired by it, as well as by other proteins of the AlkB family [2224]. AlkB is a promiscuous enzyme [25]. Although m1dAMP and m1dATP have been shown to be its minimal substrates, it doesn't appear to remove cyclic adducts from modified deoxynucleoside triphosphates [21,26]. Its primary substrates are 1-methyladenine (m1A) and 3-methylcytosine (m3C) [27,28], it can also revert bulkier DNA adducts such as ethyl, propyl, and hydroxyalkyl ones [21,29] as well as exocyclic etheno (ε) and ethano adducts including 1,N6-ethenoadenine (εA) and 3,N4-ethenocytosine (εC) [3036]. 1-methylguanine (m1G), 3-methylthymine (m3T), and 1,N2-ethenoguanine are repaired much less efficiently [37,38]. It should be noted that a large number of experimental structures of AlkB and its homologs are available in PBD records, some of which can be used as a reliable template in the modeling of complexes of AlkB with new substrates. However, in most of these structures Co(II) or Mn(II) are present as a catalytically silent analog of Fe(II), both binding in exactly the same way [25,39,40].

Previously, we found that AlkB repairs yet another chemically distinct group of substrates – exocyclic saturated adducts: the five-membered chloroacetaldehyde (CAA) adduct 3,N4-α-hydroxyetanocytosine (HEC) [34] and the six-membered acrolein adduct 3,N4-α-hydroxypropanocytosine (HPC) [35]; 1,N6-α-hydroxypropanoadenine (HPA) is a new member of this group. In 2013, we provided experimental evidence for the molecular mechanism of action confirming that E. coli AlkB dioxygenase preferentially repairs substrates in cationic form and showed that AlkB efficiently removes the hydroxypropano ring from the acrolein adduct to cytosine (3,N4-α-hydroxypropanocytosine) [35,36]. Singh et al. later demonstrated that ACR adducts to guanine, which are neutral at physiological pH, are also processed by AlkB; however, reactions are substantially less efficient [41]. Since the balance between cytotoxicity and genotoxicity of exogenous acrolein is still under debate [42], we are using the assay in which the plasmids modified in vitro with acrolein were transferred to cells untreated with acrolein. This approach diminished the challenging problem of the cytotoxity of exogenous ACR. Herein, we present results of a combined in vivo, in vitro, and in silico analysis of the mutagenic properties of HPA and molecular mechanisms of its repair by AlkB, and confirm that the cationic form of HPA is preferentially repaired by AlkB. Moreover, for the first time, we evidenced the limited turnover number of AlkB-mediated repair, which for HPA was estimated to be 38 ± 4 cycles.

Materials and methods

ACR modification of plasmid DNA – pIF101, pIF105, and pIF106

Plasmids were isolated according to a standard protocol [43] and then treated with ACR for 15 min at 37°C. The incubation mixture in the volume of 0.1 ml contained: 200 µg of plasmid DNA, 0.5 M sodium acetate buffer pH 4.5, and 0.5–5 µl of 1 M ACR (Fluka, final concentration 5–50 mM, respectively). After modification, DNA was precipitated twice with 2.5 volumes of ice-cold ethanol in the presence of 0.3 M sodium acetate buffer pH 5.2, washed with cold 70% ethanol, redissolved in an appropriate volume of TE buffer and then frozen at −20°C. Mock-treated plasmids were prepared in the same way except that ACR was absent from the incubation mixture.

Mutagenicity assay, preparation of electrocompetent cells and electrotransformation

The mutagenicity assay and the preparation of electrocompetent cells and electrotransformation were performed according to [34]. Briefly, modified and mock-treated plasmids were introduced by electroporation into wt and alkB electrocompetent E. coli cells. Appropriate dilutions of the transformation mixture were spread on LB plates containing chloramphenicol (30 µg/ml) to count the total number of transformed cells and on lactose minimal plates for selection of Lac+ revertants. Mutation frequency (MF) was calculated as the number of mutants per 104 transformed cells.

Statistics

Since some distributions were found to be non-Gaussian according to the Anderson–Darling test [44], the statistical significance of the observed differences was assessed according to the non-parametric Kruskal–Wallis H test [45]. All analyses were performed using Statistica 10 (StatSoft, Inc., 2011. STATISTICA, version 10. www.statsoft.com). Null hypotheses that given distributions do not differ from each other were tested at the significance levels of α = 0.05, 0.01, and 0.001.

Preparation of HPA-containing deoxyoligomeric substrates

To a solution of 150 nmol of the TTATT deoxypentanucleotide or TAT deoxytrinucleotide (Metabion, Martinsried, Germany) in 340 µl of 0.65 M sodium acetate pH 4.5, 15 µl of 15 M ACR (95% aq. solution, Fluka) was added and the mixture was incubated for 40 min at 37°C. The reaction mixture containing completely modified oligomers was purified from ACR and reaction buffer using preparative HPLC according to [35], lyophilized and redissolved in TE buffer. The identity of the modified oligomers was ascertained by HPLC and MS analysis.

AlkB dioxygenase purification, HPLC chromatography, and electrospray ionization-time-of-flight mass spectrometry

AlkB dioxygenase purification, HPLC chromatography, and electrospray ionization-time-of-flight (ESI-TOF) mass spectrometry were performed exactly as previously described [34,35].

Size exclusion chromatography coupled with inductively coupled plasma mass spectrometry

Determination of the amount of Mn, Fe, Ni, Cu, and Zn in FTO protein was carried out using an Agilent 7500a ICP Mass Spectrometer (Agilent Technologies, Tokyo, Japan) with 10 ng ml−1 of yttrium (89Y) as an internal standard for analyzed samples and a standard solution was used for instrument calibration. Solutions containing the protein were more stable in 5 mM ammonium acetate pH 6.0 for the external calibration curve. Curves were linear in the investigated range from 0.5 ng ml−1 to 100 ng ml−1 with r2 above 0.999. Limit of detection (LOD) was calculated for standard deviations of 10 measurements of blank and was found to be 0.2–0.7 ng ml−1. Inductively coupled plasma mass spectrometry (ICP-MS) measurement conditions (nebulizer gas flow, rf power, and lens voltage) were optimized daily using a standard built-in procedure.

Temperature dependence of protein fluorescence – differential scanning fluorimetry

Temperature dependence of protein fluorescence – differential scanning fluorimetry (DSF) assay [46]. Fluorescence data were collected on a Varian Cary Eclipse spectrofluorometer equipped with a variable-temperature cell holder, using 10 mm path-length cuvettes. Emission was monitored at 345 nm (excitation at 280 nm) at temperatures increasing in the range of 20–85°C (95°C for FTO) with a 1°C/min heating rate. The protein sample was diluted to a final concentration of 1.2 µM with 10 mM HEPES buffer pH 7.5 containing 100 mM NaCl and 2 mM DTT. Thermal unfolding of the protein was monitored either in the presence or absence of 100 µM αKG and 50 µM Mn(II) and with increasing concentrations of a potential substrate (0.6–6 µM).

DSF data analysis

The simplest model for thermal unfolding of a protein assuming a two-state process associated with temperature-dependent equilibrium of the folded, F, and unfolded, U, protein states was applied in the data analysis [47], together with a linear approximation of temperature-dependent fluorescence intensity for both protein forms (FF(T) and FU(T), respectively). Thermodynamic parameters were determined by fitting the model to the experimentally observed temperature-dependent intensities of fluorescence, F(T), according to the following formulae:

 
formula
1
 
formula
2
 
formula
3
 
formula
4

where: xU is the relative population of the unfolded protein, ΔGunf is the free energy change for unfolding transition, ΔCp,unf is the heat capacity change associated with the protein unfolding process (assumed to be temperature independent), and and represents the van't Hoff heat and entropy of unfolding at the arbitrarily selected reference temperature To. When the melting temperature, Tm, is selected as the reference, , Equation (4) can be expressed in the form:

 
formula
5

A numerical model based on Equation (1), combined with Equations (2), (3), and (5) was fitted to the experimental data using the optimization algorithms implemented in the Origin 9.0 package (www.originlab.com). Initial low- and high-temperature asymptotes (AF/U and BF/U parameters) were roughly pre-estimated using restricted datasets (T < 30°C and T > 70°C, respectively), and and Tm were then optimized keeping all other parameters constrained (ΔCp,unf fixed to 0). All data were carefully curated, and those which disagreed with neighboring ones were removed from the analyses. Finally, the whole sets of six, and afterwards seven parameters, were optimized, albeit in some cases the latter procedure led to strongly biased estimates of ΔCp,unf.

AlkB assay

The reaction mixtures (20 µl) contained: appropriate 50 mM buffer, 1 mM DTT, Fe(NH4)2(SO4)2 (tested 5 µM–1 mM, optimal 150 µM), αKG at varying concentrations (tested 5 µM–2 mM, optimal 50 µM) and 10–50 µM TT(HPA)TT ssDNA substrate. Reactions were started by the addition of purified AlkB protein (0.1–0.5 µM for kinetics and 1 µM for optimization) or the component studied, allowed to proceed at 37°C, stopped by adding 230 µl of ice-cold water and frozen at −20°C to deactivate AlkB, and then analyzed by HPLC.

Quantum mechanical calculations

The initial geometry of 1,N6-α-hydroxypropano-N9-methyl-adenine, was adopted from the PDB structure of adenosine in which the sugar moiety was replaced by a methyl group. This was followed by ab initio analysis with the aid of the Firefly version 8.1.1 program [48]. The density functional theory calculations were performed with the B3LYP functional [49] using the 6–31G(d,p) basis set [50] previously found applicable in the analysis of similar systems [35,51,52]. The resulting geometry, together with ESP-derived atomic charges, was used to parameterize the force field used in further molecular modeling.

Molecular modeling of AlkB - T(HPA)T complexes

Initial co-ordinates were adopted from the structure of T(m1A)T bound to AlkB (pdb3i2o) [25]. Structural calculations were performed with the aid of simulated annealing followed by molecular dynamics in the presence of explicit water molecules (both implemented in the YASARA Structure package) using the standard YASARA2 force field [53] extended for the nonstandard ligand by adding ab initio-derived geometry and ESP-derived charge distribution. In accordance with the quantum mechanical (QM) calculations, R and S isomers of the cyclo-hydroxypropyl group were tested in either axial or equatorial conformation of the hydroxyl group (see Figure 6 for details). For each of these four structural forms, the syn and anti conformers of the HPA glycosidic bond were analyzed.

Results and discussion

In vivo repair of HPA by E. coli AlkB

HPA is one of the minor acrolein adducts to DNA bases. Apparently, the only report considering its possible mutagenicity was published in 1998 [11]. In an analysis of mutations induced by ACR in normal human fibroblasts using supF shuttle vector plasmids, it was found that the majority of the ACR-induced base substitutions took place at guanine or cytosine residues. However, to the authors' surprise, 22% of the base substitutions were at A·T sites. Employing a dsDNA plasmid treated with acrolein at pH 7.4, which allows modification of all DNA bases, the authors were unable to identify which of the possible G·C and T·A pair derivatives were responsible for the observed substitutions. Here, for the first time, we provide evidence that the acrolein adduct to adenine is mutagenic and can be effectively repaired by AlkB dioxygenase.

To examine the in vivo mutagenic potency of HPA we employed the test system elaborated and successfully used in our previous studies on chloroacetaldehyde mutagenicity [33,34]. The system involves a series of pIF plasmids [54] carrying alleles of the lactose operon of the CC101–106 strain [55] allowing monitoring of the occurrence of Lac+ revertants arising by substitution mutation of a particular base. Acrolein originated from exogenous sources is known to be cytotoxic to bacteria even in relatively low doses due to its impact on cellular proteins [56]. To avoid this toxic effect, plasmid DNA was modified with acrolein in vitro. The modification reactions were performed in acidic conditions, where thymine is not modified by acrolein [57,58], so base substitutions caused by HPA could be observed exclusively. After electrotransformation of the ACR-modified plasmids into bacteria (wt and alkB) their repair in vivo was studied. Three plasmids: pIF101, pIF105, and pIF106 indicatory for AT → CG, AT → TA, and AT → GC mutations, respectively, were used. The transformation efficiency with mock-treated (control) plasmids varied between 1 × 106 and 2 × 106 cfu/µg DNA depending on the batch of competent cells and decreased with the ACR concentration used for plasmid modification, down to 2 × 104 for the alkB strain and 50 mM ACR. The decreased plasmid survival results from ACR-induced lesions in the whole plasmid DNA, as previously discussed in detail [34].

The mutation frequencies are shown in Figure 1 and in Supplementary Table S1 (see Supplementary Materials). They increased in a dose-dependent manner for both strains and all the plasmids tested and were 1.5–9 times higher for the alkB strain than for the wt. HPA mainly caused transversions (A → T slightly more often than A → C) and less frequently, A → G transitions. Background (spontaneous) mutation frequency varied depending on the plasmid used, being the highest for the A → G substitutions, whereas, the difference between the wt and alkB strain was the lowest.

Mutation frequency in the lacZ gene induced by acrolein.

Figure 1.
Mutation frequency in the lacZ gene induced by acrolein.

Frequencies of A → C, A → T, and A → G mutations of the pIF101 (A), pIF105 (B), and pIF106 (C) plasmids, respectively, replicated in wild type (black circles) and AlkB-deficient (empty circles) E. coli. Data are represented by medians (circles) and quartiles, while min-max ranges are denoted by vertical lines. The statistical significances of the differences between groups are denoted by asterisks (*P < 0.05, **P < 0.01, ***P < 0.001). Please note that the scale for A → G substitutions (C) differs from the others.

Figure 1.
Mutation frequency in the lacZ gene induced by acrolein.

Frequencies of A → C, A → T, and A → G mutations of the pIF101 (A), pIF105 (B), and pIF106 (C) plasmids, respectively, replicated in wild type (black circles) and AlkB-deficient (empty circles) E. coli. Data are represented by medians (circles) and quartiles, while min-max ranges are denoted by vertical lines. The statistical significances of the differences between groups are denoted by asterisks (*P < 0.05, **P < 0.01, ***P < 0.001). Please note that the scale for A → G substitutions (C) differs from the others.

AlkB dioxygenase is induced as part of the system of adaptive response to alkylating agents in E. coli. It is worth emphasizing that the observed HPA repair (Figure 1) occurred without prior Ada response induction in the studied bacterial cells, thus indicating that AlkB dioxygenase at the constitutive level exerted observable repair activity. The protective effect of the AlkB protein was previously observed for uninduced E. coli cells treated with CAA, or bearing a CAA-modified plasmid, by the Essigmann group [32,59] and us [33,34]. Moreover, neither CAA [60] nor ACR (Dylewska et al., in preparation) induce the Ada response. Interestingly, induction of the system in the studied bacteria decreased the mutation frequency not only in the wt, but also in the alkB strain (Dylewska et al., in preparation). This could indicate an involvement of another inducible protein in HPA repair and, consequently, suggests that HPA may be even more mutagenic than is shown here.

Assembly of protein/cosubstrate/cofactor–substrate complexes monitored by DSF

DSF is commonly used in screening for ligand binding to protein, whenever ligand binding affects the protein fluorescence, its thermal stability, or heat of unfolding [46].

The DSF data recorded for AlkB in the apo form and in the presence of a cofactor (Mn(II) used as a non-active Fe(II) analog) and/or cosubstrate (αKG) demonstrated that AlkB minimally bound Mn(II) in the absence of αKG, while the binding of αKG alone substantially decreased the protein fluorescence (Figure 2A).

DSF data for AlkB (A) and FTO (B) dioxygenases.

Figure 2.
DSF data for AlkB (A) and FTO (B) dioxygenases.

Difference between apo and Mn(II)/αKG loaded forms, represented by empty and black circles, clearly demonstrates the formation of AlkB and FTO active complexes. Solid lines represent models fitted to DSF data, the parameters of which are listed in Table 1. Distribution of gray markers in panel A indicates that Mn(II) alone does not bind to apo AlkB (diamonds) while αKG alone does (triangles). Distribution of gray diamonds in panel B, representing DSF data recorded for FTO/Mn(II)/αKG in the presence of T(HPA)T in a 1:1 or 5:1 molar ratio, shows that T(HPA)T does not bind to the protein.

Figure 2.
DSF data for AlkB (A) and FTO (B) dioxygenases.

Difference between apo and Mn(II)/αKG loaded forms, represented by empty and black circles, clearly demonstrates the formation of AlkB and FTO active complexes. Solid lines represent models fitted to DSF data, the parameters of which are listed in Table 1. Distribution of gray markers in panel A indicates that Mn(II) alone does not bind to apo AlkB (diamonds) while αKG alone does (triangles). Distribution of gray diamonds in panel B, representing DSF data recorded for FTO/Mn(II)/αKG in the presence of T(HPA)T in a 1:1 or 5:1 molar ratio, shows that T(HPA)T does not bind to the protein.

Table 1.
Thermodynamic parameters of thermal unfolding of AlkB and FTO dioxygenases in the presence of various ligands
Form Tm (°C) ΔHm (kJ/mol) ΔCp (kJ/mol/K) ΔSm (J/mol/K) 
AlkB 
 apo 54.1 ± 0.2 431 ± 32 20 ± 2 1.3 ± 0.1 
 Mn 55.0 ± 0.2 427 ± 32 19 ± 3 1.3 ± 0.1 
 αKG 54.6 ± 0.3 770 ± 140 39 ± 11 2.4 ± 0.4 
 αKG/Mn 61.2 ± 1.8 290 ± 90 12 ± 4 0.9 ± 0.3 
AlkB/αKG/Mn 
 T(HPA)T 1:1 58.8 ± 0.3 420 ± 50 18 ± 3 1.3 ± 0.2 
 T(HPA)T 5:1 57.7 ± 0.3 630 ± 110 28 ± 5 1.9 ± 0.3 
FTO 
 apo 63.1 ± 0.6 185 ± 20 5 ± 2 0.5 ± 0.1 
 αKG/Mn 81 ± 9 110 ± 60 0 (fixed) 0.3 ± 0.2 
Form Tm (°C) ΔHm (kJ/mol) ΔCp (kJ/mol/K) ΔSm (J/mol/K) 
AlkB 
 apo 54.1 ± 0.2 431 ± 32 20 ± 2 1.3 ± 0.1 
 Mn 55.0 ± 0.2 427 ± 32 19 ± 3 1.3 ± 0.1 
 αKG 54.6 ± 0.3 770 ± 140 39 ± 11 2.4 ± 0.4 
 αKG/Mn 61.2 ± 1.8 290 ± 90 12 ± 4 0.9 ± 0.3 
AlkB/αKG/Mn 
 T(HPA)T 1:1 58.8 ± 0.3 420 ± 50 18 ± 3 1.3 ± 0.2 
 T(HPA)T 5:1 57.7 ± 0.3 630 ± 110 28 ± 5 1.9 ± 0.3 
FTO 
 apo 63.1 ± 0.6 185 ± 20 5 ± 2 0.5 ± 0.1 
 αKG/Mn 81 ± 9 110 ± 60 0 (fixed) 0.3 ± 0.2 

αKG content, assessed with ESI-TOF mass spectrometry, was undetectable both in all protein samples and in the supernatant remaining after protein thermal degradation. Since the peak at 144.99 Da, corresponding to the αKG monoanion, was clearly visible for the equimolar αKG solution, samples of both proteins were considered to be αKG free.

Since the AlkB/αKG form was inactive in biochemical assays [27], the content of Fe(II) in the purified protein sample could be neglected, and the concentration of Mn(II) driving the formation of protein structure was almost identical with that of the active (i.e. Fe(II)-bound) form [25].

Based on the known structures of the protein with ligands (cofactor and cosubstrate) bound, the αKG-induced changes in AlkB fluorescence should be assigned directly to quenching of Tyr 122 and Trp 178 by the proximal αKG, separated from each other by less than 4.3 Å, while the next-nearest Trp/Tyr side-chain (Trp 69) is separated by more than 6.5 Å (see Supplementary Figure S1A). In contrast with the apo form, the fluorescence of AlkB/αKG was visibly affected by Mn(II), thus confirming [17,27,28] that co-coordination by αKG is crucial for metal binding (Figure 2A). In this case Trp 178 is mostly affected, and the protein fluorescence is decreased further. The high temperature part of the DSF curves remain unaffected by Mn(II) or αKG, clearly demonstrating that neither of the ligands can interact with the unfolded protein. The thermodynamic parameters extracted from the DSF data, shown in Table 1, confirm that Mn(II) does not bind to the apo form of the protein. Interestingly, although αKG binding causes a substantial increase in the van't Hoff heat of unfolding (770 ± 140 kJ/mol vs. 431 ± 32 kJ/mol for the apo form estimated at ∼55°C), no thermal stabilization of the protein is observed. The latter results from an increase in the entropy of unfolding (2.4 ± 0.4 vs. 1.3 ± 0.1 J/mol/K) associated with a restricted flexibility of the bound αKG. However, further addition of Mn(II) results in a strong increase in Tm by 7.2 ± 1.8°C, thus confirming the formation of an active form of AlkB.

More importantly, the DSF curves recorded for the AlkB/αKG/Mn(II) complex change visibly upon further addition of a putative substrate, deoxytrinucleotide T(HPA)T. Analysis of the AlkB crystal structure (pdb3khb) [39] (Supplementary Figure S1A) and model of the AlkB complex with T(HPA)T (Figure 7) indicates that Trp 69, together with Tyr 76 and Tyr 78, may sense ligand binding. It should be noted, however, that the fluorescence intensity observed for both folded and unfolded AlkB forms may also be affected by absorption at 280 nm by the added deoxytrinucleotides decreasing the effective excitation. Despite that, the estimated thermodynamic parameters (Tm, , Cp,unf, and ) consistently vary with the substrate concentration. Using the simplest model in which changes of each observable follow the equilibrium of substrate–protein binding, the observed trends allow the T(HPA)T–AlkB dissociation constant to be estimated at 1.8 ± 1.1 µM (Supplementary Figure S2). This demonstrates that even a limited set of DSF experiments allows a rough estimation of substrate affinity.

Among nine homologs of AlkB identified in the human genome, ABH2 and ABH3 are considered to be involved in DNA repair [22]. Interestingly, FTO dioxygenase, some of whose alleles are associated with obesity [19], has also been shown to have some DNA repair activity [61]. Since no suitable biochemical assay has yet been established, the content of some bivalent metal cations (Fe, Mn, Ni, Cu, Zn) was measured using size exclusion chromatography (SEC) coupled with the inductively coupled plasma mass spectrometry (SEC-ICP-MS) technique. The concentration of any of these ions was found to be substantially lower than that of the protein (see Supplementary Table S2), so the FTO sample may be considered iron and manganese free.

As a preliminary step to determine the FTO substrate specificity, we carried out DSF experiments analogous to those with AlkB described above (Figure 2B). According to the crystal structure of FTO (pdb3lfm) [40], Tyr 295 is the residue whose florescence is most strongly affected by the proximal αKG, while Tyr 106, Tyr 108, and Trp 270 should sense substrate binding. It must be noted, however, that due to the small difference between the fluorescence of the folded and unfolded states, Tm could only be reasonably precisely determined for the apo form of FTO, while other DSF data allowed a qualitative analysis only. Such analysis showed that FTO binds Mn(II) in the presence of αKG, while T(HPA)T is not bound by FTO/αKG/Mn(II) (Figure 2B). The same results were obtained for other deoxynucleotide trimers, indicating that none of them can be regarded as an FTO substrate (results not shown).

All these negative results are consistent with the structural alignment made for FTO (pdb3lfm) [40] and the AlkB–DNA complex (pdb4nid) [62], which showed that the six-residue loop of FTO (84RIQGKD89) cannot sample the DNA major groove as efficiently as the short β-turn 52PGG54 in the structure of AlkB does. Moreover, long loop 208D-G224 in FTO precludes recognition of the DNA minor groove (Supplementary Figure S3).

Altogether, the results presented here show that DSF analysis can be used to monitor the assembly of protein–cofactor–cosubstrate complexes and to screen for potential substrates or competitive inhibitors, and even for a rough estimation of their binding affinities.

Repair of TT(HPA)TT by purified AlkB protein

The best AlkB substrates, m3C and m1A, are in a cationic form at physiological pH, whereas m3T and m1G, which are neutral at pH 7, are repaired very poorly [37,63,64]. As we have shown previously [35], efficient repair of the exocyclic adducts εC, εA, HEC, and HPC occurs at a pH favoring their protonated forms. Interestingly, in acrolein-treated TTGTT pentadeoxynucleotide we were not able to identify AlkB-mediated repair of acrolein adducts to G (ACR-Gs), which may be a direct consequence of a predominance of the neutral form of ACR-Gs (see Supplementary Figure S4 for HPLC chromatograms, reactions details, and comment). We observed a general negative correlation between the pH optimal for repair of a given substrate and the logarithm of iron concentration required for maximal efficiency of this process. We reason that, at a lower pH, the imidazole ring of the histidine residues involved in Fe(II) co-ordination (His 131 and His 187) is protonated, which has to be compensated for by an increase in Fe(II) concentration. In contrast, the optimal αKG concentration did not depend on pH since even at the most acidic conditions screened, the cosubstrate maintained its active doubly dissociated anionic form.

Here, we used a HPA-containing pentadeoxynucleotide, TT(HPA)TT, as a substrate for AlkB. Using HPLC to separate the modified and unmodified (repaired) oligonucleotides we confirmed the in vivo observation that the modification is indeed efficiently repaired by E. coli AlkB dioxygenase. The optimal pH for HPA repair was found to be 7.5, slightly more than one pH unit below the pKa of HPA (8.77 ± 0.06, see Supplementary Figure S5). This clearly indicates that the cationic form of the HPA base is preferably repaired. The optimal iron concentration was 147 ± 5 µM for repair at pH 7.5, while the optimal αKG concentration was 50 ± 7 µM (Supplementary Figure S6). Supplementary Figure S7 shows combined data on the interdependence between pH and Fe(II) concentration optimal for repair of m3C, HPC, HEC, εA, and εC adducts by the AlkB protein taken from [35], together with the present result for HPA at the optimal pH of 7.5. The slope of −0.54 ± 0.03 estimated for log(Fe) vs. pH relationship indicates that iron binding by AlkB is at a lower pH associated with the deprotonation of the imidazole ring of both His residues involved in metal co-ordination (His 131 and His 187).

Mass spectrometry analysis of HPA repair by AlkB

AlkB usually acts by a direct reversal mechanism leading to regeneration of the natural base, but some exceptions have been noted. A cyclic alkanoadduct to adenine, 1,N6-ethanoadenine, was found to be oxidatively hydroxylated by AlkB, but the final product of this reaction was ethano-ring-opened N6-(2-oxoethyl)-adenine [31]. AlkB repair reactions of acrolein and malondialdehyde adducts to guanine (HPGs) – α-hydroxypropano-dG (α-OH-PdG), γ-hydroxypropano-dG (γ-OH-PdG), and m1dG, respectively, are complex, involving multiple intermediates, overlapping pathways with multiple convergence points, and only for α-OH-PdG, the formation of a small amount of fully dealkylated product (dG) was observed [41]. Bearing in mind the above observations, we verified full HPA repair by AlkB via mass spectrometry (MS). As shown in Supplementary Table S3, after incubation with AlkB the pentadeoxynucleotide containing HPA was converted to its unmodified (i.e. repaired) form. Our data indicate that the adduct to adenine studied here is repaired completely, as are other saturated hydroxyalkanoadducts, described by us previously, HEC and HPC [34,35].

Based on the ACR reaction with adenine described in [65], the general mechanism of AlkB action [25,27,28,32] and the finding that the final product released during AlkB-mediated transformation of ACR-modified guanine is malondialdehyde [41], in Figure 3 we present a reaction of the formation and foregoing AlkB-mediated repair of HPA in DNA.

Acrolein adduct to adenine.

Figure 3.
Acrolein adduct to adenine.

Chemical structures and proposed pathway for the formation and AlkB-mediated repair of HPA.

Figure 3.
Acrolein adduct to adenine.

Chemical structures and proposed pathway for the formation and AlkB-mediated repair of HPA.

HPA repair by AlkB differs qualitatively from that observed for εA

In our previous work [35], we analyzed the AlkB-directed repair of the chloroacetaldehyde adduct to adenine – non-isomeric εA. We found that the reaction followed pseudo-first-order kinetics during the initial 10 min, while further reaction progress was visibly slower than expected. Here, the repair of HPA by AlkB was followed in reaction mixtures of various composition to improve the reliability of the estimated rates using the same approach as for εA.

Initially, the TT(HPA)TT to TTATT conversion follows the general trend previously observed by us for TT(εA)TT → TTATT [35], as presented in Figure 4, for the reaction carried out by 0.5 µM AlkB at conditions optimized for either substrate. Interestingly, for the lowest substrate concentration (10 µM) the HPA decay during the first minute agrees perfectly with the reaction progress observed for the same concentration of εA, clearly indicating that HPA is as good a substrate for AlkB as is εA (empty triangles vs. empty circles in Figure 4). However, with the reaction progress the two curves diverge, and the concentration of εA decreases steadily accordingly to pseudo-first-order kinetics (dashed arrow) while the concentration of HPA reaches a plateau. The same tendency is observed at higher initial substrate concentrations, for which the higher the HPA concentration is, the slower the repair relative to that of εA. Furthermore, the fraction of HPA remaining unprocessed increases with the initial substrate concentration. Taken together, these results indicate that there is a qualitative difference between the two reactions. Although εA is smaller than hydroxypropano-ringed HPA, both molecules fit into the catalytic site, and the initial rates of their repair are close. However, the substantial difference concerns the size of the released by-product – glyoxal and malondialdehyde for εA and HPA, respectively [32,41].

Repair of HPA and εA by AlkB dioxygenase (0.5 µM).

Figure 4.
Repair of HPA and εA by AlkB dioxygenase (0.5 µM).

Both the rate of repair and the fraction of processed deoxypentamers correlate negatively with initial substrate concentration. It should be noted that the initial rate of HPA repair is at the lowest substrate concentration, the same as that of εA [35], albeit the amount of unprocessed HPA is higher than that of εA.

Figure 4.
Repair of HPA and εA by AlkB dioxygenase (0.5 µM).

Both the rate of repair and the fraction of processed deoxypentamers correlate negatively with initial substrate concentration. It should be noted that the initial rate of HPA repair is at the lowest substrate concentration, the same as that of εA [35], albeit the amount of unprocessed HPA is higher than that of εA.

The apparent initial rate of the repair of TT(HPA)TT at a low concentration is very high and virtually identical with that determined by us previously for TT(εA)TT [35], but with increasing substrate concentration the repair rate falls, and, in parallel, increases the amount of substrate remaining unprocessed. Moreover, for all the substrate concentrations the rate decreases substantially as the reaction proceeds. Altogether, it implies that each AlkB molecule is capable of processing only a limited number of substrate molecules and, consequently, the amount of active protein decreases continuously, until its full deactivation.

The data presented in Table 2 give a semi-quantitative insight into the HPA repair by AlkB. Thus, the initial first-order rate (v0) decreases substantially with increasing TT(HPA)TT concentration. This clearly implies that HPA inhibits AlkB activity. It should be noted that the reaction progress extrapolated to the initial reaction time shows that a substantial amount of substrate is processed in the burst phase, during the first 20 s. However, the amount of substrate processed during this pre-steady-state period exceeds a 7–10 fold AlkB concentration, clearly indicating that the initial rate of AlkB-driven TT(HPA)TT repair must be higher than the observed rate of ‘late’ first-order kinetics (i.e. after 20 s).

Table 2.
Kinetic parameters estimated for the AlkB-directed repair of HPA in pentadeoxynucleoside
TT(HPA)TT (µM) HPA processed during 90 min (%) v0 (min−1‘burst' turnovers 
0.1 µM AlkB 0.25 µM AlkB 0.5 µM AlkB 
45 87 87 15.4. ± 1.5 9.9 ± 1.5 
14 40 51 88 11.2 ± 1.3 6.6 ± 1.8 
23 32 19 66 3.0 ± 2.1 6.6 ± 1.4 
34 36 26 2.6 ± 1.0 – 
TT(HPA)TT (µM) HPA processed during 90 min (%) v0 (min−1‘burst' turnovers 
0.1 µM AlkB 0.25 µM AlkB 0.5 µM AlkB 
45 87 87 15.4. ± 1.5 9.9 ± 1.5 
14 40 51 88 11.2 ± 1.3 6.6 ± 1.8 
23 32 19 66 3.0 ± 2.1 6.6 ± 1.4 
34 36 26 2.6 ± 1.0 – 

Limited amount of HPA can be processed by AlkB

Inspection of substrate concentration at very long reaction times clearly confirms that, unlike for most previously studied substrates [35], even after 90 min a significant fraction of HPA remains unprocessed. However, a general relationship between the AlkB to substrate concentration ratio and apparent limit of turnover number could be deduced. Thus, the higher the protein to substrate ratio is, the larger amount of HPA is repaired (Figure 5). Interestingly, even at the highest ratio, only approximately 86% of TT(HPA)TT is converted to TTATT. Moreover, the intercept calculated for the relationship observed between % of the repair represented as a function of enzyme to substrate ratio indicates that each AlkB molecule carries out 38 ± 4 enzymatic cycles before its total inactivation. The most likely candidate for the toxic entity is malondialdehyde – the chemically active by-product, which is apparently much more reactive than glyoxal.

Dependence of the final fraction of HPA repaired on the initial AlkB/HPA molar ratio.

Figure 5.
Dependence of the final fraction of HPA repaired on the initial AlkB/HPA molar ratio.

The fraction of unprocessed HPA decreases with an increasing enzyme : substrate ratio, but even at the highest ratio tested, less than 90% of TT(HPA)TT can be repaired. The asymptotic trend observed for the low enzyme : substrate ratio indicates that the limited number of HPA molecules is processed by a single AlkB molecule.

Figure 5.
Dependence of the final fraction of HPA repaired on the initial AlkB/HPA molar ratio.

The fraction of unprocessed HPA decreases with an increasing enzyme : substrate ratio, but even at the highest ratio tested, less than 90% of TT(HPA)TT can be repaired. The asymptotic trend observed for the low enzyme : substrate ratio indicates that the limited number of HPA molecules is processed by a single AlkB molecule.

Taken together, kinetic data supports the following scenario. AlkB repairs TT(HPA)TT extremely efficiently, however, the by-products are toxic for the protein. Therefore, after 7–10 enzymatic cycles the protein loses most of its activity, but carries on slowly over the next 30 cycles, until its full deactivation.

Ab initio sampling of 1,N6-α-hydroxypropano-N9-methyl-adenine low-energy states

Conformational equilibrium of HPA and relative energy of the most stable conformers were analyzed using an ab initio approach. 1,N6-α-hydroxypropano-N9-methyl-adenine was used as a general model for the base of HPA deoxynucleoside. The hydroxyl group can be attached to Cα to form either the R or S stereoisomer (Figure 6), both of which display identical properties alone, however, become non-equivalent when (deoxy)ribose is attached at N9, and can even be substantially differentiated in an asymmetric environment (e.g. in a protein binding site). Moreover, the additional six-member ring formed by α-hydroxypropane can adopt two alternative conformations with the hydroxyl group oriented either equatorially or axially, the equatorial form being favored by ∼1.8 kcal/mol. Similar calculations were performed for cationic forms of both stereoisomers, which were found to be protonated preferentially at N6. BCJ-2016-1008F6 

Possible stereoisomers/conformers of HPA.

Figure 6.
Possible stereoisomers/conformers of HPA.

R (top) and S (down) configuration at Cα with either equatorial (left) or axial (right) orientation of the hydroxyl group.

Figure 6.
Possible stereoisomers/conformers of HPA.

R (top) and S (down) configuration at Cα with either equatorial (left) or axial (right) orientation of the hydroxyl group.

HPA in the binding site of AlkB.

Figure 7.
HPA in the binding site of AlkB.

R-anti (A) and S-syn (B) HPA was modeled into the AlkB active site using the structure of m1A bound to AlkB (pdb3i2o) [25] as the template. Both conformers are susceptible to the action of radicals formed close to the metal center. In complexes with the alternative conformers (R-syn and S-anti), the exocyclic hydroxyl group points towards the metal center, thereby shielding the cyclohydroxypropane adduct from a direct radical attack. HPA with the axial orientation of the hydroxyl group (see Figure 6) does not fit the AlkB binding site. Asp135, postulated to be crucial for recognition of substrates in their cationic forms, is shown in stick representation, and Fe(II) is represented as a pink sphere of appropriate radius.

Figure 7.
HPA in the binding site of AlkB.

R-anti (A) and S-syn (B) HPA was modeled into the AlkB active site using the structure of m1A bound to AlkB (pdb3i2o) [25] as the template. Both conformers are susceptible to the action of radicals formed close to the metal center. In complexes with the alternative conformers (R-syn and S-anti), the exocyclic hydroxyl group points towards the metal center, thereby shielding the cyclohydroxypropane adduct from a direct radical attack. HPA with the axial orientation of the hydroxyl group (see Figure 6) does not fit the AlkB binding site. Asp135, postulated to be crucial for recognition of substrates in their cationic forms, is shown in stick representation, and Fe(II) is represented as a pink sphere of appropriate radius.

Molecular modeling of HPA location in the AlkB substrate-binding pocket

Since no structures of AlkB with the substrate studied here had been published, we modeled the structure of T(HPA)T bound to AlkB. Initial co-ordinates of the complex were obtained by a stepwise simulated-annealing procedure in which 1-methyladenine from a complex of AlkB with DNA dodecamer (pdb3bi3) [66] was modified sequentially to 1,N6-dimethyladenine, 1,N6-propanoadenine, and finally to R and S stereoisomers of HPA. The two latter structures, each in either syn or anti conformation of the glycosidic bond, were further subjected to 1 ns molecular dynamics in the presence of explicit water molecules (see Methods for details). Representative structures obtained are shown in Figure 7. Interestingly, both isomers pack perfectly inside the binding pocket, when the α-hydroxyl group adopts the equatorial conformation (R-anti and S-cis). In both cases, the modified base makes stacking interactions with the side-chain of proximal Trp 69, accompanied by hydrophobic contacts with Tyr 76 and Leu 118. In both structures, the proximal iron cation makes the base susceptible to a radical attack that initiates the dealkylation process. Additionally, the side-chain carboxyl group of the proximal Asp 135 recognizes the electron-deficient region at the N4 side, of which the interaction is in accordance with the experimentally observed preference towards a cationic form of the substrate. However, the Asp 135 side-chain is moved slightly away from the (S)-HPA form due to steric hindrance with its α-hydroxyl oxygen, which may result in preferential activity towards the R isomer of HPA. In general, despite the increased size of the α-hydroxypropano derivative compared with the original 1-methyladenine, neither stereoisomer of HPA significantly disturbs the network of intermolecular interactions observed in the crystal structure of AlkB with the DNA oligomer [66]. Thus, the presented models show the molecular mechanism of HPA recognition, including Trp 69 involved in a stacking interaction with the modified base and Asp 135 acting as a gate-keeper. Moreover, it should be highlighted that the AlkB pocket is filled with the tricyclic ring of HPA almost completely, so the diffusion of the reaction by-product (malondialdehyde [41]) may be slowed down markedly as long as the product (adenine) occupies the binding site. Such hardly mobile reactive species could attack on fly an AlkB residue(s) essential for enzymatic activity, leading to its gradual inactivation, which was observed here by us to be much faster for HPA than found previously for εA [35].

Conclusions

In the present study, we provide in vivo and in vitro evidence that, the acrolein adduct to adenine, HPA is mutagenic and is effectively repaired by AlkB dioxygenase. As the lesion is efficiently formed in vivo after exposure to acrolein [12], it is likely to be harmful to living organisms.

The molecular mechanism of AlkB action was characterized using kinetic experiments and in silico modeling. The analysis demonstrated that a protonated form of HPA is preferentially repaired, the reaction is stereoselective, and AlkB exhibits a highly limited turnover number towards its substrate.

Additionally, a differential scanning fluorimetry assay was elaborated, providing a useful tool for discrimination of protein active forms, with cofactor and cosubstrate bound, and for monitoring substrate binding.

Abbreviations

     
  • ACR

    acrolein

  •  
  • αKG

    alpha-ketoglutarate

  •  
  • CAA

    chloroacetaldehyde

  •  
  • DSF

    differential scanning fluorimetry

  •  
  • ε

    etheno

  •  
  • HEC

    3,N4-α-hydroxyetanocytosine

  •  
  • HPA

    1,N6-α-hydroxypropanoadenine

  •  
  • HPC

    3,N4-α-hydroxypropanocytosine

  •  
  • HPG

    1,N2-α-hydroxypropanoguanine and 1,N2-γ-hydroxypropanoguanine

  •  
  • m

    methyl

  •  
  • QM

    quantum mechanical

  •  
  • wt

    wild type

Author Contribution

A.M.M., J.P., and J.T.K. conceived the study and designed experiments; M.D., A.M.M., and T.P. performed the experiments; J.P. performed M.M. and Q.M. calculations; M.D., A.M.M., J.P., and J.T.K. analyzed the data; A.M.M. and J.P. wrote the paper.

Funding

This work was supported by National Science Centre grant [UMO-2012/04/M/NZ1/00068] and partially by Polish-Norwegian Research Fund [POL-NOR/196258/59/2013]. The equipment used was sponsored in part by the Centre for Preclinical Research and Technology (CePT), a project co-sponsored by the European Regional Development Fund and Innovative Economy, The National Cohesion Strategy of Poland. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Acknowledgments

We would like to thank researchers from our Institute, Professors Iwona Fijałkowska and Piotr Jonczyk, for kindly providing the pIF plasmids for mutagenesis tests, Professor Elżbieta Grzesiuk for a sample of FTO protein for DSF screening, and Jacek Olędzki Msc, from the Laboratory of Mass Spectrometry for the MS analysis and Dr Katarzyna Pawlak and Wioletta Jakubczak Msc, from Warsaw University of Technology for SEC-ICP-MS, and for Dr Konrad Kosicki, from Warsaw University for sharing some preliminary data.

Competing Interests

The Authors declare that there are no competing interests associated with the manuscript.

References

References
1
Feng
,
Z.
,
Hu
,
W.
,
Hu
,
Y.
and
Tang
,
M.-S.
(
2006
)
Acrolein is a major cigarette-related lung cancer agent: preferential binding at p53 mutational hotspots and inhibition of DNA repair
.
Proc. Natl Acad. Sci. U.S.A.
103
,
15404
15409
doi:
2
Faroon
,
O.
,
Roney
,
N.
,
Taylor
,
J.
,
Ashizawa
,
A.
,
Lumpkin
,
M.H.
and
Plewak
,
D.
(
2008
)
Acrolein health effects
.
Toxicol. Ind. Health
24
,
447
490
doi:
3
Crook
,
T.R.
,
Souhami
,
R.L.
and
McLean
,
A.E.
(
1986
)
Cytotoxicity, DNA cross-linking, and single-strand breaks induced by activated cyclophosphamide and acrolein in human-leukemia cells
.
Cancer Res.
46
,
5029
5034
PMID:
[PubMed]
4
Ramu
,
K.
,
Fraiser
,
L.H.
,
Mamiya
,
B.
,
Ahmed
,
T.
and
Kehrer
,
J.P.
(
1995
)
Acrolein mercapturates: synthesis, characterization, and assessment of their role in the bladder toxicity of cyclophosphamide
.
Chem. Res. Toxicol.
8
,
515
524
doi:
5
Cohen
,
S.M.
,
Garland
,
E.M.
,
St John
,
M.
,
Okamura
,
T.
and
Smith
,
R.A.
(
1992
)
Acrolein initiates rat urinary-bladder carcinogenesis
.
Cancer Res.
52
,
3577
3581
PMID:
[PubMed]
6
Chung
,
F.-L.
,
Chen
,
H.-J.
and
Nath
,
R.G.
(
1996
)
Lipid peroxidation as a potential endogenous source for the formation of exocyclic DNA adducts
.
Carcinogenesis
17
,
2105
2111
doi:
7
Minko
,
I.G.
,
Kozekov
,
I.D.
,
Harris
,
T.M.
,
Rizzo
,
C.J.
,
Lloyd
,
R.S.
and
Stone
,
M.P.
(
2009
)
Chemistry and biology of DNA containing 1, N2-deoxyguanosine adducts of the α,β-unsaturated aldehydes acrolein, crotonaldehyde, and 4-hydroxynonenal
.
Chem. Res. Toxicol.
22
,
759
778
doi:
8
Curren
,
R.D.
,
Yang
,
L.L.
,
Conklin
,
P.M.
,
Grafstrom
,
R.C.
and
Harris
,
C.C.
(
1988
)
Mutagenesis of xeroderma pigmentosum fibroblasts by acrolein
.
Mutat. Res.
209
,
17
22
PMID:
[PubMed]
9
Hang
,
B.
(
2004
)
Repair of exocyclic DNA adducts: rings of complexity
.
Bioessays
26
,
1195
1208
doi:
10
Zhang
,
S.
,
Villalta
,
P.W.
,
Wang
,
M.
and
Hecht
,
S.S.
(
2007
)
Detection and quantitation of acrolein-derived 1, N2-propanodeoxyguanosine adducts in human lung by liquid chromatography-electrospray ionization-tandem mass spectrometry
.
Chem. Res. Toxicol.
20
,
565
571
doi:
11
Kawanishi
,
M.
,
Matsuda
,
T.
,
Nakayama
,
A.
,
Takebe
,
H.
,
Matsui
,
S.
and
Yagi
,
T.
(
1998
)
Molecular analysis of mutations induced by acrolein in human fibroblast cells using supF shuttle vector plasmids
.
Mutat. Res.
417
,
65
73
doi:
12
Kawai
,
Y.
,
Furuhata
,
A.
,
Toyokuni
,
S.
,
Aratani
,
Y.
and
Uchida
,
K.
(
2003
)
Formation of acrolein-derived 2′-deoxyadenosine adduct in an iron-induced carcinogenesis model
.
J. Biol. Chem.
278
,
50346
50354
doi:
13
Teo
,
I.
,
Sedgwick
,
B.
,
Demple
,
B.
,
Li
,
B.
and
Lindahl
,
T.
(
1984
)
Induction of resistance to alkylating-agents in Escherichia coli: the ada+ gene product serves both as a regulatory protein and as an enzyme for repair of mutagenic damage
.
EMBO J.
3
,
2151
2157
PMID:
[PubMed]
14
Lindahl
,
T.
,
Sedgwick
,
B.
,
Sekiguchi
,
M.
and
Nakabeppu
,
Y.
(
1988
)
Regulation and expression of the adaptive response to alkylating agents
.
Annu. Rev. Biochem.
57
,
133
157
doi:
15
Krwawicz
,
J.
,
Arczewska
,
K.D.
,
Speina
,
E.
,
Maciejewska
,
A.
and
Grzesiuk
,
E.
(
2007
)
Bacterial DNA repair genes and their eukaryotic homologues: 1. Mutations in genes involved in base excision repair (BER) and DNA-end processors and their implication in mutagenesis and human disease
.
Acta Biochim. Pol.
54
,
413
434
PMID:
[PubMed]
16
Mielecki
,
D.
,
Wrzesiński
,
M.
and
Grzesiuk
,
E.
(
2015
)
Inducible repair of alkylated DNA in microorganisms
.
Mutat. Res. Rev. Mut. Res.
763
,
294
305
doi:
17
Aravind
,
L.
and
Koonin
,
E.V.
(
2001
)
The DNA-repair protein AlkB, EGL-9, and leprecan define new families of 2-oxoglutarate- and iron-dependent dioxygenases
.
Genome Biol.
2
,
RESEARCH0007.1
doi:
18
Kurowski
,
M.A.
,
Bhagwat
,
A.S.
,
Papaj
,
G.
and
Bujnicki
,
J.M.
(
2003
)
Phylogenomic identification of five new human homologs of the DNA repair enzyme AlkB
.
BMC Genomics
4
,
48
doi:
19
Gerken
,
T.
,
Girard
,
C.A.
,
Tung
,
Y.-C.L.
,
Webby
,
C.J.
,
Saudek
,
V.
,
Hewitson
,
K.S.
et al. 
(
2007
)
The obesity-associated FTO gene encodes a 2-oxoglutarate-dependent nucleic acid demethylase
.
Science
318
,
1469
1472
doi:
20
Mielecki
,
D.
,
Zugaj
,
D.L.
,
Muszewska
,
A.
,
Piwowarski
,
J.
,
Chojnacka
,
A.
,
Mielecki
,
M.
et al. 
(
2012
)
Novel AlkB dioxygenases—alternative models for in silico and in vivo studies
.
PLoS ONE
7
,
e30588
doi:
21
Koivisto
,
P.
,
Duncan
,
T.
,
Lindahl
,
T.
and
Sedgwick
,
B.
(
2003
)
Minimal methylated substrate and extended substrate range of Escherichia coli AlkB protein, a 1-methyladenine-DNA dioxygenase
.
J. Biol. Chem.
278
,
44348
44354
doi:
22
Aas
,
P.A.
,
Otterlei
,
M.
,
Falnes
,
P.Ø.
,
Vågbø
,
C.B.
,
Skorpen
,
F.
,
Akbari
,
M.
et al. 
(
2003
)
Human and bacterial oxidative demethylases repair alkylation damage in both RNA and DNA
.
Nature
421
,
859
863
doi:
23
Falnes
,
P.Ø.
,
Bj
ø
rås
,
M.
,
Aas
,
P.A.
,
Sundheim
,
O.
and
Seeberg
,
E.
(
2004
)
Substrate specificities of bacterial and human AlkB proteins
.
Nucleic Acids Res.
32
,
3456
3461
doi:
24
Ougland
,
R.
,
Zhang
,
C.-M.
,
Liiv
,
A.
,
Johansen
,
R.F.
,
Seeberg
,
E.
,
Hou
,
Y.-M.
et al. 
(
2004
)
Alkb restores the biological function of mRNA and tRNA inactivated by chemical methylation
.
Mol. Cell
16
,
107
116
doi:
25
Yu
,
B.
and
Hunt
,
J.F.
(
2009
)
Enzymological and structural studies of the mechanism of promiscuous substrate recognition by the oxidative DNA repair enzyme AlkB
.
Proc. Natl Acad. Sci. U.S.A.
106
,
14315
14320
doi:
26
Sikora
,
A.
,
Maciejewska
,
A.M.
,
Poznański
,
J.
,
Pilżys
,
T.
,
Marcinkowski
,
M.
,
Dylewska
,
M.
et al. 
(
2015
)
Effects of changes in intracellular iron pool on AlkB-dependent and AlkB-independent mechanisms protecting E.coli cells against mutagenic action of alkylating agent
.
Mutat. Res.Fundam. Mol. Mech. Mutag.
778
,
52
60
doi:
27
Falnes
,
P.Ø.
,
Johansen
,
R.F.
and
Seeberg
,
E.
(
2002
)
AlkB-mediated oxidative demethylation reverses DNA damage in Escherichia coli
.
Nature
419
,
178
182
doi:
28
Trewick
,
S.C.
,
Henshaw
,
T.F.
,
Hausinger
,
R.P.
,
Lindahl
,
T.
and
Sedgwick
,
B.
(
2002
)
Oxidative demethylation by Escherichia coli AlkB directly reverts DNA base damage
.
Nature
419
,
174
178
doi:
29
Duncan
,
T.
,
Trewick
,
S.C.
,
Koivisto
,
P.
,
Bates
,
P.A.
,
Lindahl
,
T.
and
Sedgwick
,
B.
(
2002
)
Reversal of DNA alkylation damage by two human dioxygenases
.
Proc. Natl Acad. Sci. U.S.A.
99
,
16660
16665
doi:
30
Mishina
,
Y.
,
Yang
,
C.-G.
and
He
,
C.
(
2005
)
Direct repair of the exocyclic DNA adduct 1,N6-ethenoadenine by the DNA repair AlkB proteins
.
J. Am. Chem. Soc.
127
,
14594
14595
doi:
31
Frick
,
L.E.
,
Delaney
,
J.C.
,
Wong
,
C.
,
Drennan
,
C.L.
and
Essigmann
,
J.M.
(
2007
)
Alleviation of 1, N6-ethanoadenine genotoxicity by the Escherichia coli adaptive response protein AlkB
.
Proc. Natl Acad. Sci. U.S.A.
104
,
755
760
doi:
32
Delaney
,
J.C.
,
Smeester
,
L.
,
Wong
,
C.
,
Frick
,
L.E.
,
Taghizadeh
,
K.
,
Wishnok
,
J.S.
et al. 
(
2005
)
Alkb reverses etheno DNA lesions caused by lipid oxidation in vitro and in vivo
.
Nat. Struct. Mol. Biol.
12
,
855
860
doi:
33
Maciejewska
,
A.M.
,
Sokolowska
,
B.
,
Nowicki
,
A.
and
Kusmierek
,
J.T.
(
2011
)
The role of AlkB protein in repair of 1,N6-ethenoadenine in Escherichia coli cells
.
Mutagenesis
26
,
401
406
doi:
34
Maciejewska
,
A.M.
,
Ruszel
,
K.P.
,
Nieminuszczy
,
J.
,
Lewicka
,
J.
,
Sokołowska
,
B.
,
Grzesiuk
,
E.
et al. 
(
2010
)
Chloroacetaldehyde-induced mutagenesis in Escherichia coli: the role of AlkB protein in repair of 3,N4-ethenocytosine and 3,N4-α-hydroxyethanocytosine
.
Mutat. Res.
684
,
24
34
doi:
35
Maciejewska
,
A.M.
,
Poznanski
,
J.
,
Kaczmarska
,
Z.
,
Krowisz
,
B.
,
Nieminuszczy
,
J.
,
Polkowska-Nowakowska
,
A.
et al. 
(
2013
)
Alkb dioxygenase preferentially repairs protonated substrates: specificity against exocyclic adducts and molecular mechanism of action
.
J. Biol. Chem.
288
,
432
441
doi:
36
Mielecki
,
D.
,
Saumaa
,
S.
,
Wrzesinski
,
M.
,
Maciejewska
,
A.M.
,
Zuchniewicz
,
K.
,
Sikora
,
A.
et al. 
(
2013
)
Pseudomonas putida AlkA and AlkB proteins comprise different defense systems for the repair of alkylation damage to DNA – in vivo, in vitro, and in silico studies
.
PLoS ONE
8
,
e76198
doi:
37
Delaney
,
J.C.
and
Essigmann
,
J.M.
(
2004
)
Mutagenesis, genotoxicity, and repair of 1-methyladenine, 3-alkylcytosines, 1-methylguanine, and 3-methylthymine in AlkB Escherichia coli
.
Proc. Natl Acad. Sci. U.S.A.
101
,
14051
14056
doi:
38
Zdzalik
,
D.
,
Vagbo
,
C.B.
,
Kirpekar
,
F.
,
Davydova
,
E.
,
Puscian
,
A.
,
Maciejewska
,
A.M.
et al. 
(
2014
)
Protozoan ALKBH8 oxygenases display both DNA repair and tRNA modification activities
.
PLoS ONE
9
,
e98729
doi:
39
Holland
,
P.J.
and
Hollis
,
T.
(
2010
)
Structural and mutational analysis of Escherichia coli AlkB provides insight into substrate specificity and DNA damage searching
.
PLoS ONE
5
,
e8680
doi:
40
Han
,
Z.
,
Niu
,
T.
,
Chang
,
J.
,
Lei
,
X.
,
Zhao
,
M.
,
Wang
,
Q.
et al. 
(
2010
)
Crystal structure of the FTO protein reveals basis for its substrate specificity
.
Nature
464
,
1205
1209
doi:
41
Singh
,
V.
,
Fedeles
,
B.I.
,
Li
,
D.
,
Delaney
,
J.C.
,
Kozekov
,
I.D.
,
Kozekova
,
A.
et al. 
(
2014
)
Mechanism of repair of acrolein- and malondialdehyde-derived exocyclic guanine adducts by the α-ketoglutarate/Fe(II) dioxygenase AlkB
.
Chem. Res. Toxicol.
27
,
1619
1631
doi:
42
Eckl
,
P.M.
and
Bresgen
,
N.
(
2017
)
Genotoxicity of lipid oxidation compounds
.
Free Radical Biol. Med.
doi:
43
Sambrook
,
J.
,
Fritsch
,
E.
and
Maniatis
,
T.
(
1989
)
Molecular Cloning: A Laboratory Manual
,
Cold Spring Harbor Laboratory Press, Cold Spring Harbor
,
NY
44
Anderson
,
T.W.
and
Darling
,
D.A.
(
1952
)
Asymptotic theory of certain ‘goodness of fit’ criteria based on stochastic processes
.
Ann. Math. Stat.
23
,
193
212
doi:
45
Kruskal
,
W.H.
and
Allen Wallis
,
W.
(
1952
)
Use of ranks in one-criterion variance analysis
.
J. Am. Stat. Assoc.
47
,
583
621
doi:
46
Niesen
,
F.H.
,
Berglund
,
H.
and
Vedadi
,
M.
(
2007
)
The use of differential scanning fluorimetry to detect ligand interactions that promote protein stability
.
Nat. Protoc.
2
,
2212
2221
doi:
47
Eftink
,
M.R.
(
1994
)
The use of fluorescence methods to monitor unfolding transitions in proteins
.
Biophys. J.
66
,
482
501
doi:
48
Schmidt
,
M.W.
,
Baldridge
,
K.K.
,
Boatz
,
J.A.
,
Elbert
,
S.T.
,
Gordon
,
M.S.
,
Jensen
,
J.H.
et al. 
(
1993
)
General atomic and molecular electronic structure system
.
J. Comput. Chem.
14
,
1347
1363
doi:
49
Hertwig
,
R.H.
and
Koch
,
W.
(
1997
)
On the parameterization of the local correlation functional: what is becke-3-LYP
?
Chem. Phys. Lett.
268
,
345
351
doi:
50
Francl
,
M.M.
,
Pietro
,
W.J.
,
Hehre
,
W.J.
,
Binkley
,
J.S.
,
Gordon
,
M.S.
,
DeFrees
,
D.J.
et al. 
(
1982
)
Self-consistent molecular orbital methods. XXIII. A polarization-type basis set for second-row elements
.
J. Chem. Phys.
77
,
3654
3665
doi:
51
Poznański
,
J.
,
Najda
,
A.
,
Bretner
,
M.
and
Shugar
,
D.
(
2007
)
Experimental (13C NMR) and theoretical (ab initio molecular orbital calculations) studies on the prototropic tautomerism of benzotriazole and some derivatives symmetrically substituted on the benzene ring
.
J. Phys. Chem. A
111
,
6501
6509
doi:
52
Wasik
,
R.
,
Lebska
,
M.
,
Felczak
,
K.
,
Poznański
,
J.
and
Shugar
,
D.
(
2010
)
Relative role of halogen bonds and hydrophobic interactions in inhibition of human protein kinase CK2α by tetrabromobenzotriazole and some C(5)-substituted analogues
.
J. Phys. Chem. B
114
,
10601
10611
doi:
53
Krieger
,
E.
,
Joo
,
K.
,
Lee
,
J.
,
Lee
,
J.
,
Raman
,
S.
,
Thompson
,
J.
et al. 
(
2009
)
Improving physical realism, stereochemistry, and side-chain accuracy in homology modeling: four approaches that performed well in CASP8
.
Proteins
77
(
Suppl 9
),
114
122
doi:
54
Fijalkowska
,
I.J.
,
Jonczyk
,
P.
,
Tkaczyk
,
M.M.
,
Bialoskorska
,
M.
and
Schaaper
,
R.M.
(
1998
)
Unequal fidelity of leading strand and lagging strand DNA replication on the Escherichia coli chromosome
.
Proc. Natl Acad. Sci. U.S.A.
95
,
10020
10025
doi:
55
Cupples
,
C.G.
and
Miller
,
J.H.
(
1989
)
A set of lacZ mutations in Escherichia coli that allow rapid detection of each of the six base substitutions
.
Proc. Natl Acad. Sci. U.S.A.
86
,
5345
5349
doi:
56
Parent
,
R.A.
,
Caravello
,
H.E.
and
San
,
R.H.C.
(
1996
)
Mutagenic activity of acrolein in S. typhimurium and E. coli
.
J. Appl. Toxicol.
16
,
103
108
doi:
57
Chenna
,
A.
,
Rieger
,
R.A.
and
Iden
,
C.R.
(
1992
)
Characterization of thymidine adducts formed by acrolein and 2-bromoacrolein
.
Carcinogenesis
13
,
2361
2365
doi:
58
Chenna
,
A.
and
Iden
,
C.R.
(
1993
)
Characterization of 2′-deoxycytidine and 2′-deoxyuridine adducts formed in reactions with acrolein and 2-bromoacrolein
.
Chem. Res. Toxicol.
6
,
261
268
doi:
59
Kim
,
M.Y.
,
Zhou
,
X.
,
Delaney
,
J.C.
,
Taghizadeh
,
K.
,
Dedon
,
P.C.
,
Essigmann
,
J.M.
et al. 
(
2007
)
Alkb influences the chloroacetaldehyde-induced mutation spectra and toxicity in the pSP189 supF shuttle vector
.
Chem. Res. Toxicol.
20
,
1075
1083
doi:
60
Mroczkowska
,
M.M.
,
Kolasa
,
I.K.
and
Kusmierek
,
J.T.
(
1993
)
Chloroacetaldehyde-induced mutagenesis in Escherichia coli: specificity of mutations and modulation by induction of the adaptive response to alkylating agents
.
Mutagenesis
8
,
341
348
doi:
61
Jia
,
G.
,
Yang
,
C.-G.
,
Yang
,
S.
,
Jian
,
X.
,
Yi
,
C.
,
Zhou
,
Z.
et al. 
(
2008
)
Oxidative demethylation of 3-methylthymine and 3-methyluracil in single-stranded DNA and RNA by mouse and human FTO
.
FEBS Lett.
582
,
3313
3319
doi:
62
Zhu
,
C.
and
Yi
,
C.
(
2014
)
Switching demethylation activities between AlkB family RNA/DNA demethylases through exchange of active-site residues
.
Angew. Chem. Int. Ed. Engl.
53
,
3659
3662
doi:
63
Koivisto
,
P.
,
Robins
,
P.
,
Lindahl
,
T.
and
Sedgwick
,
B.
(
2004
)
Demethylation of 3-methylthymine in DNA by bacterial and human DNA dioxygenases
.
J. Biol. Chem.
279
,
40470
40474
doi:
64
Falnes
,
P.O.
(
2004
)
Repair of 3-methylthymine and 1-methylguanine lesions by bacterial and human AlkB proteins
.
Nucleic Acids Res.
32
,
6260
6267
doi:
65
Sodum
,
R.S.
and
Shapiro
,
R.
(
1988
)
Reaction of acrolein with cytosine and adenine derivatives
.
Bioorg. Chem.
16
,
272
282
doi:
66
Yang
,
C.-G.
,
Yi
,
C.
,
Duguid
,
E.M.
,
Sullivan
,
C.T.
,
Jian
,
X.
,
Rice
,
P.A.
et al. 
(
2008
)
Crystal structures of DNA/RNA repair enzymes AlkB and ABH2 bound to dsDNA
.
Nature
452
,
961
965
doi: