Over the past 40 years considerable emphasis has been placed on the development of accurate and sensitive methods for the detection and quantitation of DNA adducts. The formation of DNA adducts resulting from the covalent interaction of genotoxic carcinogens with DNA, derived from exogenous and endogenous sources, either directly or following metabolic activation, can if not repaired lead to mutations in critical genes such as those involved in the regulation of cellular growth and subsequent development of cancer. The major analytical challenge has been to detect levels of DNA adducts at the level of 0.1–1 adducts per 10 8 unmodified DNA bases using only low microgram amounts of DNA, and with high specificity and accuracy, in humans exposed to genotoxic carcinogens derived from occupational, environmental, dietary and life-style sources. In this review we will highlight the merits as well as discuss the progress made by liquid chromatography coupled to electrospray ionization mass spectrometry as a method for DNA adduct detection.
The formation of DNA adducts by genotoxic carcinogens is an essential stage in the process by which such compounds cause cancer, summarized schematically in Figure 1 . Some genotoxic carcinogens can directly react with DNA but the majority require metabolic activation to form the ultimate electrophilic reactive species that covalently binds to nucleophilic sites in DNA ( 1 , 2 ). The presence of DNA adducts does not in itself signify that mutations will result, as lesions may be repaired before cell division and in some cases the DNA adduct may not be promutagenic. However in the cases where adducts are not repaired, alterations in the DNA sequence may occur upon DNA replication. Alterations in the DNA sequence may also occur when adducts are subjected to erroneous repair. Some genotoxic carcinogens show a degree of specificity in the sites at which they produce such mutations, because of site-selectivity either in adduct formation or in repair of these adducts, and characteristic mutation spectra then result following exposure to these compounds. This phenomenon has been observed, for example, in humans exposed to aflatoxin B 1 whose hepatocellular tumours show a relatively specific mutation at codon 249 of the p53 tumour suppressor gene, which is believed to be associated with this genotoxic carcinogen ( 3 ). Furthermore the spectrum of mutations induced by polycyclic aromatic hydrocarbons (PAHs) in the p53 tumour suppressor gene in bronchial epithelial cells has been reported to be similar to the major mutational hotspots in human lung cancers, indicating that PAHs may be involved in lung carcinogenesis ( 4 , 5 ). Thus determination of the nature and extent of adduct formation, together with (if possible) the position of these adducts within the DNA structure, may play a role in the assessment of carcinogenic hazard and possibly of risk.
Humans are constantly being exposed to increasing numbers of genotoxic carcinogens in their everyday lives ( 6 , 7 ). Though occupational exposure to genotoxic carcinogens may be controlled due to safer work practices and regulation, exposure to genotoxic carcinogens through diet and environmental pollution may be on the increase, posing a risk to human health ( 8 ). Genotoxic carcinogens found in the diet include mycotoxins and N -nitroso compounds as well as compounds that can be generated by the cooking process, examples of which include heterocyclic aromatic amines, PAHs and more recently discovered acrylamide, which has come to prominence as a potential human genotoxic carcinogen that is formed following the heating of carbohydrate rich foods to high temperatures ( 9 – 14 ). Elevated levels of environmental air pollution are found in highly industrialized cities, with one of the major components being PAHs, which are established animal genotoxic carcinogens ( 15 , 16 ). Tobacco smoke represents a life-style route of human exposure to genotoxic carcinogens, containing ∼4000 chemicals of which ∼60 are classified as carcinogens in humans and animals ( 17 – 19 ). The presence of endogenous background levels of DNA damage, which is related to endogenous metabolic processes and oxidative stress, has been well documented and investigations have commenced into how this damage may impact on human carcinogenesis ( 6 , 20 , 21 ).
The process of carcinogenic risk assessment for an individual person on the basis of DNA adduct levels is so far not well advanced, as difficulties may be encountered in some cases because of the multi-stage nature of the carcinogenic process and the interindividual variation in many of the stages of this process. However on a population basis, exposure to increased levels of carcinogens does lead to increased DNA adduct levels and increased tumour incidence signifying the potential of using adduct determinations as part of the risk assessment process for these populations ( 22 , 23 ). It is generally accepted by many regulatory authorities that the dose–response relationship for genotoxic carcinogens does not have a threshold and thus the presence of any amount of exposure presents a carcinogenic risk. If one assumes that adduct formation is linearly related to dose (for which there is experimental support for several compounds, for example aflatoxin B 1 and aromatic amines), then any amount of adduct level must be presumed to be associated with some risk of carcinogenesis ( 24 , 25 ). The view that the dose–response relationship for genotoxic carcinogens is linear is not universally upheld however, as it is not always possible to detect an increase in mutations above background at administered doses of genotoxic agents which produce detectable adducts, which implies a non-linear dose–response relationship for these compounds.
The scientific community has yet to agree on whether there is a level of DNA adducts that may be acceptable in human DNA, because the risk level associated with it is so low as to be of little or no concern. It does however seem unlikely that a single ‘universal’ value could be given to such an acceptable level of adducts, as their mutational effectiveness varies according to the nature of the carcinogen and the chemical structure of the DNA adduct, for example N -7 alkylguanine adducts have considerably less mutagenic potential than O6 alkylguanine adducts, although it may be that such a process may be feasible for some specific carcinogens ( 26 ).
Thus for experimental and human molecular epidemiological studies it is important to have methodologies available to assess DNA adduct formation down to extremely low levels and, if available, to the ranges that can be considered acceptable with regard to human risk. Indeed it should also not be overlooked that the lack of adduct formation signifies that no mutation could occur, and demonstration of the inability of a chemical to produce adducts may be used to eliminate the possibility that the compound is a genotoxic carcinogen, which may have a major influence on regulatory decisions on such compounds (e.g. 2-phenylphenol and ochratoxin A as determined by accelerator mass spectrometry) ( 27 , 28 ). The demonstration of such a null result clearly requires analytical methodology of the utmost sensitivity and reliability.
Generally speaking the level of DNA adducts that we are normally concerned with analysing ranges from 0.1–1 adducts per 10 8 unmodified DNA bases (as found in humans at background levels or after low level exposure to genotoxic carcinogens) to 1 adduct per 10 4 unmodified DNA bases (as found in animals treated with carcinogenic doses of chemicals) ( 29 ). A wide variety of approaches for such analyses have been developed, which include the use of detection of fluorescence, radioactive decay, electrochemical properties, immunological response and mass spectrometric ion intensity, for the quantitation of carcinogen adducts. One limiting factor for DNA adduct determinations from human samples is the restricted availability of the analyte. Non-invasive access to DNA adducts can in some cases be possible, such as some alkylated purine adducts which are excreted in urine after their enzymatic removal from DNA ( 30 ). Though it should be noted that some urinary DNA adducts could potentially originate from intracellular nucleotide pools or dead cells. Urine also contains exfoliated urothelial cells from which DNA adduct levels can be determined ( 31 , 32 ). Otherwise DNA is normally prepared from white blood cells or from other accessible tissues such as placenta and buccal swabs. Typically 1 ml of blood will yield 20–40 μg of DNA. So the amount of DNA available for adduct analysis may be limited. There is no such problem encountered for animal studies, where the amount of tissue available is plentiful and typically the yield is ∼1 mg of DNA from 1 g of tissue. The choice of timing of the sample collection is also critical, and consideration should be given to the fact that adduct lifetimes may range from hours to weeks depending on their structure and the processes that are involved in their repair. The analytical challenge therefore is to be able to achieve sensitivities for adduct detection at or <1 adduct per 10 8 unmodified DNA bases, using only low microgram amounts of DNA, and with high specificity and accuracy. The purpose of this review is to illustrate the increasing importance of LC-MS with particular reference to electrospray ionization (ESI) to this field.
Mass spectrometry in the analysis of DNA adducts
Historically the role of MS in the determination of DNA adducts has been limited to providing information for the identification of new DNA adducts, or for the structural characterization of DNA adduct standards that have been utilized to determine adduct levels by other detection methods, such as 32 P-postlabelling ( 33 – 36 ). However in recent times, with the development of LC coupled to MS and on-going technological advances in ionization methods, in particular ESI as well as ion transmission and detection, MS has become a viable method for the quantitation of DNA adducts levels in animal and human samples obtained after exposure to exogenous and endogenous genotoxic compounds. Comprehensive reviews describing the application of LC coupled to MS for the study of DNA adducts have been published by Esmans et al. (1998) Andrews et al. (1999), Doerge et al. (2002), Koc and Swenberg (2002) and Watson et al. (2003) ( 36 – 40 ). More general reviews on the role of MS in biochemical research and the detection of DNA adducts have been published by Siuzdak (1994), Farmer and Sweetman (1995), Griffiths et al. (2001), Lim and Lord (2002) as well as Banoub et al. (2005) ( 41 – 45 ). Figure 2 highlights the guanine and phosphate backbone adducts formed in DNA following exposure to various genotoxic chemicals that have been studied using LC-ESI-MS. The development of LC-MS has progressed to such an extent that we are now at a stage to address the question of the feasibility of MS as an acceptable alternative to the other DNA adduct detection methods such as 32 P-postlabelling.
Since the early 1960s when Brookes and Lawley unequivocally established that DNA was a macromolecular target for chemical carcinogens, which could covalently bind to it forming DNA adducts, numerous methods have been developed for the detection of adducts in animal and human tissues ( 46 – 48 ). The merits and limitations of the different methods available have been reviewed by Phillips et al. (2000) and Esaka et al. (2003) ( 49 , 50 ). A similar review was published by De Kok et al. (2002) focusing on the comparison of methods available for the detection of PAH and heterocyclic aromatic amine DNA adducts formed following environmental and occupational exposures ( 51 ). A review by Turesky and Vouros (2004) summarizes the different methods used for the detection of heterocyclic aromatic amine DNA adducts in experimental animals and humans ( 52 ). DNA adducts can be described in terms of their hydrophobicity and divided into small or polar adducts and bulky or non-polar adducts. The amount of DNA sample available and the chemical nature of the DNA adduct normally influence the method of detection employed ( 49 ).
The 32 P-postlabelling method has been the most widely applied and one of the most sensitive procedures for the analysis of DNA adducts since its introduction in 1981 by Randerath et al. ( 53 ). The 32 P-postlabelling method appears to be the preferred method to study bulky/non-polar DNA adducts due to the ease of separation of the hydrophobic adducts from unmodified 2′-deoxynucleotides ( 34 , 51 ). The procedure involves the enzymatic digestion of the DNA sample to 2′-deoxynucleoside 3′-monophosphates and subsequent labelling of the DNA adduct at the 5′ position with [γ- 32 P] ATP, which is a reaction catalysed by T4 polynucleotide kinase. The labelled adducted 2′-deoxynucleotide is then separated using multi-dimensional thin layer chromatography or high performance liquid chromatography (HPLC). Normally adduct enrichment steps are employed to increase sensitivity, these being particularly applicable for small/polar DNA adducts due to their very similar chromatographic properties to unmodified 2′-deoxynucleotides following enzymatic digestion of the DNA ( 54 , 55 ). In the case of bulky/non-polar DNA adducts increased sensitivity can be obtained using nuclease P 1 enrichment which preferentially 3′-dephosphorylates unmodified 2′-deoxynucleotides thus preventing them from being substrates for the 32 P-phosphorylation by T4 polynucleotide kinase. Alternatively butanol extraction can be employed which selectively extracts bulky/non-polar DNA adducts from the unmodified 2′-deoxynucleotides ( 34 , 56 , 57 ).
Some DNA adducts exhibit an inherent ability to be oxidized following the application of a potential, and hence HPLC with electrochemical (ampometric) or coulometric detectors can be employed for their detection ( 58 – 60 ). Similarly HPLC with fluorescence detection may be used for those adducts which exhibit fluorescence properties ( 61 , 62 ). If a suitable antibody is available for a DNA adduct then immunological assays such as enzyme linked immunosorbent assay (ELISA), immunohistochemical or immunoslot-blot assays can be employed ( 63 – 66 ). Accelerator MS represents the most sensitive analytical method so far available for detecting DNA adducts with a limit of detection that may be as low as 1 adduct per 10 12 unmodified DNA bases ( 67 ). The accelerator mass spectrometer basically consists of two mass spectrometers separated by an electrostatic (Van de Graff) generator, which measures isotope ratios. For the detection of 14 C radioisotope incorporation into DNA, the sample is converted to graphite prior to introduction into the accelerator mass spectrometer. Atoms extracted from the sample in the source are negatively ionized and these undergo acceleration by a high voltage followed by separation according to their momentum, charge and energy ( 68 ). The main limitation of the technique is that it depends on the presence of an isotope such as 14 C or 3 H in the molecule of interest, which means that at present its application is restricted to experimental systems where labelled compounds may be used. However attempts have been made to develop 14 C-postlabelling procedures, which would allow the use of unlabelled carcinogens ( 69 , 70 )
Before the advent of LC-MS the main MS method available (and one which is still being used today) for studying DNA adducts was gas chromatography-mass spectrometry (GC-MS) utilizing electron ionization (impact) or chemical ionization ( 71 , 72 ). However GC-MS has been more extensively used for the determination of protein adducts, which act as surrogate markers of exposure to carcinogens ( 42 , 73 , 74 ). Numerous DNA adducts have been studied using gas chromatography-electron capture negative chemical ionization mass spectrometry (GC-EC-NCI-MS) and comprehensive reviews on electron capture MS have been published by Giese (2000) as well as Leis et al. (2004) ( 75 , 76 ). DNA base adducts which have been studied using this technique include malondialdehyde-guanine, N -7 (2-hydroxyethyl)guanine ( N -7(2-OHEt)G), N2 , N -3 ethenoguanine, N -1, N2 ethenoguanine and N -1, N6 ethenoadenine ( 77 – 80 ). Gas chromatography offers the advantage of greater peak resolution compared with conventional LC. However the main limitation is that only non-polar/volatile compounds can be analysed, and in the case of the majority of DNA adducts, which are non-volatile and/or polar, there is a requirement for derivatization at high temperature prior to the analysis. In the case of 8-oxoguanine analysis the derivatization step may cause artefactual formation of the adduct and thus hinder the accurate measurement of background levels of oxidative damage ( 81 ). The advent of the ESI source meant that the direct analysis of polar compounds could be realized without the need for derivatization, allowing the direct coupling of LC to the mass spectrometer.
The process of ESI occurs at atmospheric pressure where eluent from the LC apparatus enters the electrospray source via the electrospray probe. The probe consists of a stainless steel capillary, to the tip of which is applied a high voltage (typically between 2 and 5 kV) relative to a counter electrode on the cone, resulting in the generation of a potential gradient across the source. A flow of nitrogen gas (referred to as nebulizing or sheath gas) is applied co-axially to the stainless steel capillary, in combination with the application of heated nitrogen known as the drying gas, which results in the generation of an aerosol spray of charged droplets that eventually undergo solvent evaporation (desolvation), leading to the formation of gas-phase ions. The ions then pass through the source (sampling) cone into the high vacuum region of the mass spectrometer. The exact nature of the mechanism for the formation of charged ions from the droplet surface is not clear. Theories have been proposed involving either coulombic fission, where large droplets divide into smaller droplets and eventually become single ions, or coulombic repulsion where ions are released from the droplet surface by overcoming the surface tension ( 82 ). A more detailed explanation of the ESI process can be found in reviews published by Gaskell (1997) and by Bruins (1998) ( 83 , 84 ). ESI results in negligible dissociation of the molecule ion into fragment ions and may be termed a soft ionization technique, which forms protonated or deprotonated ions by application of a positive or negative potential to the tip of the stainless steel capillary, respectively. Acids such as formic or acetic can be added to the mobile phase to help promote ionization and protonate basic compounds, such as those containing amine groups. Conversely ionizable compounds containing hydroxyl, carboxylic, phosphate and sulphate groups can be deprotonated by the addition of a base such as ammonium hydroxide ( 84 ).
Although ESI is the main choice of ionization technique for studying DNA adducts, atmospheric pressure chemical ionization (APCI) has also been employed for their detection, albeit to a much lesser extent. APCI differs from ESI in that ions are generated by gas-phase ionization rather than liquid phase ionization. The APCI source creates a corona discharge by the application of an electrical current to a corona discharge pin positioned between the APCI probe tip and the source cone, which is used to ionize the analyte of interest. APCI also allows for the coupling of LC operating at higher flow rates compared with those used for ESI and is useful for analysing less polar compounds ( 85 ). Recently, using APCI the detection of etheno-2′-deoxyadenosine and malondialdehyde-2′-deoxyguanosine adducts was reported in human urine ( 86 , 87 ).
The most common design of instrument used for DNA adduct analysis is a triple quadrupole as shown in Figure 3 . This configuration allows for tandem mass spectrometry (MS/MS) to be performed in which the first (Q1) and third (Q3) quadrupoles are the mass analysers (filters) consisting of four parallel cylindrical rods to which voltages are applied generating oscillating electric fields. The second (q2) quadrupole is the collision cell filled with an inert gas (normally argon), in which analyte ions undergo collision induced dissociation (CID). Kinetic energy is transferred to the analyte ions by the application to the collision cell of what is termed as the collision energy (in eV), so that when the analyte ions collide with the argon atoms the kinetic energy is converted to internal energy resulting in dissociation of the analyte ion. A more detailed description of the process of CID can be found in the review by Sleno and Volmer (2004) ( 88 ). The mass analysers determine the mass ( m ) to charge ( z ) ratio ( m/z ), which is referred to as a thompson unit (Th). A different design of instrument, less often used in DNA adduct analysis, is the ion-trap mass spectrometer, which has a quadrupole ion-trap mass analyser consisting of three electrode arrangement in which a ring electrode separates two hemispherical electrodes to trap ions in a small volume. The main advantages of the ion-trap mass spectrometer include its compact size and the ability to trap and accumulate ions to increase the signal-to-noise ratio of a measurement providing MS n data ( n th generation MS/MS). Further information can be found in publications by Jonscher and Yates (1997) as well as March (1997) ( 89 , 90 ). Casale et al. used ion-trap MS for the detection of benzo[ a ]pyrene adducted DNA bases in the urine of women exposed to household coal smoke ( 91 , 92 ). A recent development has been the triple quadrupole linear ion-trap mass spectrometer, which allows for quantitative data to be obtained, and has been reviewed by Hopfgartner et al . (2004) ( 93 ).
In addition time-of-flight (TOF) instruments coupled with a quadrupole analyser (Q-TOF) have been used to detect DNA adducts. A TOF mass spectrometer uses the differences in transit time to separate ions of different masses having the same kinetic energy when they are accelerated by application of an electric field into a field-free drift region, which means that lower mass ions have a higher velocity than higher mass ions and so they reach the detector sooner. Modern TOF instruments generate higher resolution data (accurate mass data) and provide much more sensitive product ion spectra compared with quadrupole instruments ( 94 ). The use of TOF instruments has so far been limited to providing information for the characterization of DNA adducts, highlighted for example for adducts derived from the reaction of benzo[ a ] pyrene diol epoxide with calf thymus DNA and those formed in the lung tissue of mice exposed to asphalt fumes ( 95 , 96 ).
The two main detection methods used for determination of DNA adducts are single ion monitoring (SIM) and selected reaction monitoring (SRM), which may be referred to as multiple reaction monitoring (MRM) in some publications. In SIM the first quadrupole (Q1) is set to pass only the ions of interest to the detector ( Figure 3A ). In contrast SRM involves the selection of a precursor ion of the compound of interest in the first quadrupole (Q1) which then undergoes transition in the collision cell (q2) by CID to a specific product ion, which is then determined by the third quadrupole (Q3) ( Figure 3B ). For optimum sensitivity of quantitative analysis of DNA adducts by LC-ESI-MS/MS SRM, unique and abundant product ions need to be identified following CID. Typical examples of ESI-MS/MS CID product ion spectra obtained for different guanine nucleic acid base, 2′-deoxynucleoside and 2′-deoxynucleotide DNA adducts are shown in Figure 4 . It is interesting to note that the different 2′-deoxynucleoside DNA adducts share a common CID pathway where the major product ion corresponds to the adducted base, which is formed by the loss of 2′-deoxyribose. SRM offers the greatest sensitivity due to decreased solvent and matrix interferences, and increased specificity due to the monitoring of a characteristic dissociation in the collision cell of the compound under investigation ( 97 ). For example Ravanat et al. showed that for the analysis of 8- oxo -2′-deoxyguanosine standards using SIM the limit of detection was much higher at 5 pmol compared with 20 fmol when using SRM ( 98 ). A similar finding was observed by Liao et al. who showed that a lower limit of detection was obtained for the analysis of N -7(2-OHEt)G standards with calf thymus DNA as the matrix using SRM (16 fmol) when compared with SIM (1 pmol) ( 99 ).
Another MS/MS detection mode is constant neutral loss (CNL). This is used to a lesser extent than SRM for quantification of DNA adducts, but instead is used primarily to screen for specific compounds that share a common CID pathway with the loss of a specific neutral moiety or functional group. In CNL mode both quadrupoles Q1 and Q3 are set to scan for ions at specified mass range with a signal only being observed following dissociation in the collision cell if the precursor ion gives a product ion of the mass difference specified ( Figure 3C ). The sensitivity CNL of analysis is much lower than SRM, which was highlighted by Gangl et al. when they studied the formation of DNA adducts by the heterocyclic aromatic amine 2-amino-3-methylimidazo-[4,5- f ]quinoline (IQ) in vitro and in Cynomologus monkeys. The authors found that CNL gave a limit of detection of 1 adduct per 10 4 unmodified DNA bases compared with SRM where the limit of detection was 1 adduct per 10 7 unmodified DNA bases ( 100 ).
In terms of sample preparation for LC-MS similar problems apply that are encountered for 32 P-postlabelling methods, because the level of the DNA adduct of interest is considerably lower in comparison with the unmodified DNA bases, which may interfere with the analysis. Therefore enrichment of the DNA adduct is required following hydrolysis or enzymatic digestion of the DNA sample prior to analysis by LC-MS. A number of enrichment methods have been employed, such as immunoaffinity column purification, solid phase extraction, HPLC and column switching, which will be discussed in more detail later. Enrichment is also essential for the removal of contaminants such as inorganic ions, salts and buffers which may be present in the DNA/sample matrix and have the potential to reduce sensitivity by suppression of ionization of the DNA adduct of interest ( 101 – 103 ). The high level of unmodified bases present following the digestion of DNA may cause loss of sensitivity by signal suppression or in some cases by generating artefactual responses ( 104 , 105 ). For example an artefactual response was observed for the detection of 8- oxo -2′-deoxyguanosine in enzymatically digested DNA samples, which was due to an adduct being formed between the protonated molecule ([M + H] + ) 2′-deoxyadenosine and methanol. Methanol was a constituent of the mobile phase and the resulting [M + H] + 2′-deoxyadenosine and methanol adduct ion gave a SRM transition which was identical to that used to detect 8- oxo -2′-deoxyguanosine ( 106 ). During the detection of 8- oxo -2′-deoxyguanosine in DNA samples by SRM a peak corresponding to the retention time of 2′-deoxyguanosine has also been observed, which may be explained by the conversion of 2′-deoxyguanosine to 8- oxo -2′-deoxyguanosine in the electrospray source following elution from the LC column ( 39 , 98 ). A similar observation was made for the analysis of human urine samples purified by solid phase extraction, where a peak eluting at the retention time of 2′-deoxyguanosine was detected in the SRM channel for 8- oxo -2′-deoxyguanosine. The authors confirmed that 2′-deoxyguanosine was oxidized to 8- oxo -2′-deoxyguanosine in the electrospray source during the ionization process by obtaining a similar result with the analysis of a pure 2′-deoxyguanosine standard and concluded that chromatographic separation of 2′-deoxyguanosine from 8- oxo -2′-deoxyguanosine was essential for accurate quantitation ( 107 ).
For the 32 P-postlabelling method to function, the DNA adduct under investigation must be in the form of a 2′-deoxynucleoside 3′-monophosphate. However in terms of the analysis of DNA adducts by LC-MS the method of analysis is not restricted just to the 2′-deoxynucleotide since the DNA adduct can also be determined as the purine or pyrimidine base or 2′-deoxynucleoside, as illustrated below.
Mass spectrometric analysis of adducted nucleic acid bases
Alkyl base adducts (excluding methyl adducts) of the N -7 position of guanine analysed by positive ESI-MS/MS undergo a characteristic CID where the bond to the alkyl group is broken resulting in the transfer of a proton from the alkyl group to the guanine base, which is the major product ion formed. In contrast, for N -7 methylguanine the major product ion formed from the [M + H] + precursor ion at m/z 166 following CID is at m/z 149, which corresponds to the loss of the exocyclic –NH 2 group ( 108 ). The typical ESI-MS/MS CID product ion spectra obtained for N -7 ethylguanine ( N -7EtG) and N -7(2-OHEt)G are shown in Figure 4A and B , respectively. This transition of the [M + H] + precursor ion to guanine ( m/z 152) can be used as a generic method for monitoring N -7 alkylguanines. This approach has been used for the detection of N -7(2-OHEt)G, N -7 (1- or 2-hydroxybutyl)guanine, N -7 (trihydroxybutyl)guanine, N -7EtG and N -7 (2-carbamoyl-2-hydroxyethyl)guanine adducts by SRM ( 99 , 109 – 112 ).
Detection of the base adduct is particularly suited to the analysis of chemical carcinogens that form DNA adducts which are thermally labile, such as at the N -7 and N -3 positions of guanine and adenine, respectively. These labile base adducts can be released from the DNA by thermal hydrolysis and then separated from the depurinated DNA backbone by filtration using molecular weight cut-off filters. DNA adducts formed by ethylene oxide, ethylating agents and acrylamide have been studied using this approach ( 99 , 110 , 111 ). Doerge et al. detected the N -3 adenine and N -7 guanine adducts derived from glycidamide, the reactive epoxide metabolite of acrylamide, in mice and rats following neutral thermal hydrolysis. The analysis was performed using an online column switching approach where the initial flow was diverted to waste, which will be discussed in more detail later ( 112 ). For adducts that are not thermally labile, mild acid hydrolysis can be employed which selectively cleaves the purine bases. This approach was used to determine malondialdehyde-guanine in calf thymus DNA treated with malondialdehyde ( 113 ). Tretyakova et al. used both neutral thermal hydrolysis and mild acid hydrolysis to selectively detect the different N -7 guanine and N -3, N -1, N6 adenine adducts formed by 1,3-butadiene in vitro and in lung and liver DNA from rats and mice ( 109 ). Similarly using acid and thermal hydrolysis Lehner et al. detected the formation of N -7 guanine and adenine adducts following the treatment of calf thymus DNA with 7-sulfooxymethyl-12-methylbenz[ a ]anthracene, the reactive metabolite of 7,12-dimethylbenz[ a ]anthracene ( 114 ). More stable adducts can be released from the DNA by strong acid hydrolysis and thermal hydrolysis at high temperatures. Strong as well as mild acid hydrolysis combined with thermal hydrolysis also leads to the cleavage of unmodified DNA bases, which are present in vast excess relative to the adducted bases, and hence a purification step is normally required to prevent ionization and signal suppression resulting from the presence of the unmodified bases. Muller et al. could not detect any DNA adducts following the direct LC-ESI-MS/MS SRM analysis of calf thymus DNA treated with the reactive metabolite of vinyl chloride, 2-chloroethylene oxide, which had been subjected to acid and thermal hydrolysis. The authors concluded that the presence of large amounts of guanine and adenine that were also released by depurination under the acidic conditions used for hydrolysis led to the suppression of ionization, which was rectified by using HPLC to purify the DNA adducts from the unmodified bases ( 115 ). Nakao et al. used acid and thermal hydrolysis followed by an offline HPLC purification step before the LC-ESI-MS/MS SRM detection of C -8 (1-hydroxyethyl)guanine adducts in liver DNA and RNA in control rats and those dosed with ethanol ( 116 ). Yen et al. used cation exchange columns followed by C 18 solid phase extraction for the analysis of liver N2 , N -3 ethenoguanine adducts by SIM in rats dosed with 2-chloroethylene oxide as well as background adduct levels in unexposed human liver DNA ( 117 ). Inagaki et al. detected N2 ethylguanine and N -1, N2 propanoguanine adducts of acetaldehyde, the primary oxidative metabolite of ethanol, in calf thymus DNA and DNA from cultured cells treated with acetaldehyde following acid and thermal hydrolysis using LC-ESI-MS SIM ( 118 ).
Selective removal of base adducts can also be achieved by using DNA repair enzymes such as formamidopyrimidine glycosylase (Fpg), which cleave the glycosidic bond between the adducted base and the sugar. Rodriguez et al. have used this approach with Fpg to selectively release 8-oxoguanine, which was then determined by GC-MS following filtration from the depurinated DNA, but as yet this approach has not been applied to the detection of excised adducts by LC-MS ( 119 ).
The analysis of base adducts present in urine, which may result from spontaneous depurination or following DNA repair, requires purification using methods such as solid phase chromatography or immunoaffinity column purification. This is due to the complex nature of the urinary matrix, which may lead to ionization and signal suppression. Bhattacharya et al. used immunoaffinity columns for purifying excreted benzo[ a ]pyrene base adducts from human urine, formed by the reaction of benzo[ a ]pyrene radical cations with DNA following exposure to coal smoke ( 92 ). Egner et al. used solid phase extraction combined with immunoaffinity column purification for the identification of a hydroxylated aflatoxin N -7 guanine adducts in the urine and liver DNA of tree shrews and rats dosed with aflatoxin B 1 ( 120 ).
Mass spectrometric analysis of adducted 2′-deoxynucleosides
Numerous methods have been published for the determination of DNA adducts as 2′-deoxynucleosides by LC-ESI-MS/MS SRM, and this class of compounds all share a common [M + H] + precursor ion to adducted base [B + H 2 ] + product ion CID pathway in positive ESI. Analysis of adducts that are stable to spontaneous depurination once formed in DNA, as 2′-deoxynucleosides by SRM provides the most sensitive method for the LC-ESI-MS/MS determination of the majority of DNA adducts. This is due to the glycosidic bond between the 2′-deoxyribose and the adducted base readily undergoing cleavage following CID. Adducted 2′-deoxynucleosides undergo dissociation in the collision cell at relatively low collision energies when compared with the corresponding base or 2′-deoxynucleotide adducts, which are more stable and require higher energy for dissociation to occur ( 115 , 121 ). The net result is the transfer of a proton from the 2′-deoxyribose to the adducted base which is the most prominent product ion formed. The 2′-deoxyribose is lost as a neutral moiety (116 u) even though a low intensity product ion at m/z 117 is also observed that corresponds to the protonated 2′-deoxyribose. SRM analysis using the transition of the [M + H] + precursor ion to adducted base product ion has been utilized to study a number of DNA adducts. The typical ESI-MS/MS CID product ion spectra obtained for 8- oxo -, malondialdehyde- and benzo[ a ]pyrene diol epoxide-2′-deoxyguanosine adducts are shown in Figure 4C, D and E , respectively. It should be noted, for C-8 oxidized purine 2′-deoxynucleosides analysed in negative ESI, that the most abundant product ion formed following CID results from the loss of a neutral group of 90 mass units by the cleavage of two bonds of the 2′-deoxyribose ring rather than loss of the 2′-deoxyribose which occurs in positive ESI ( 122 ). Certain adducts like thymidine glycol and 5-hydroxy-2′-deoxyuridine are detected with greater sensitivity in negative ESI compared with positive ESI and undergo cleavage of the glycosidic bond following CID ( 122 , 123 ). Adducted 2′-deoxynucleosides can be subjected to in-source dissociation by the application of a high cone voltage resulting in fragmentation of the [M + H] + ion occurring in the electrospray source. Chaudhary et al. used this approach for the structural characterization of the malondialdehyde-2′-deoxyadenosine adduct formed in calf thymus DNA following treatment with malondialdehyde. The adducted 2′-deoxynucleoside was dissociated to the base adduct in the electrospray source which was then subjected to further analysis by MS/MS CID ( 124 ).
The analysis of 2′-deoxynucleoside adducts first requires the enzymatic digestion of the DNA sample to 2′-deoxynucleosides. For example this can be achieved by using enzymes such as deoxyribonuclease and snake venom phosphodiesterase, which are endo and exonucleases, respectively, and subsequent 5′-dephosphorylation by shrimp alkaline phosphatase. ( 106 ) The DNA adduct then requires enrichment from the unmodified 2′-deoxynucleosides for which a number of methods have been employed, including immunoaffinity column purification, solid phase extraction and also HPLC. An approach utilizing solid phase extraction purification was used by Paehler et al. to detect the C -8 and N2 2′-deoxyguanosine DNA adducts formed in rat liver following dosing with the food derived carcinogen 2-amino-3,8-dimethylimidazo[4,5- f ]quinoxaline (MeIQx) ( 125 ). An immunoaffinity column purification approach was used for the LC-ESI-MS/MS SRM detection of background levels of 8- oxo -2′-deoxyguanosine in calf thymus DNA, following enzymatic digestion to 2′-deoxynucleosides ( 106 ). Jones and Sabbioni used liquid–liquid extraction with water saturated ethyl acetate or butanol for the separation of DNA adducts formed by various arylamines and nitroarenes in calf thymus DNA and rat liver DNA, following enzymatic digestion to 2′-deoxynucleosides prior to LC-ESI-MS/MS SRM analysis ( 126 ). The importance of enrichment of the DNA adduct was highlighted by Crosbie et al. who directly analysed 2-amino-1-methyl-6-phenylimidazo[4,5- b ]pyridine (PhIP) derived 2′-deoxyguanosine adducts by LC-ESI-MS SIM following digestion to 2′-deoxynucleosides without prior enrichment, following the in vitro exposure of cells to PhIP. The authors noted that peak intensities were reduced and that there was a rapid decline in instrument sensitivity over time as samples were analysed ( 127 ). Otteneder et al. used a combined approach of solid phase extraction and immunoaffinity column purification for the determination of malondialdehyde-2′-deoxyguanosine adducts in urine of rats that had been dosed with carbon tetrachloride ( 128 ). Fang et al. used an immunoaffinity purification approach for the LC-ESI-MS/MS SRM detection of 2′-deoxyguanosine adducts in rat urine formed following the oral administration of the food derived heterocyclic amine, PhIP ( 129 ).
A recent innovation is the use of column switching, which allows for the direct analysis of the digested DNA sample, with any interfering compounds being diverted to waste by automated switching valves to which the LC columns are connected. In the simplest set-up the initial eluent from the LC column is diverted to waste with the mobile phase flow being maintained to the mass spectrometer by a second LC pump until the analyte of interest elutes and then the flow reverts back to the mass spectrometer. This simple approach was used by Marsch et al. to study the formation of DNA adducts resulting from the conjugation of dihalomethanes to glutathione by glutathione transferases in which the initial eluent from the LC column was diverted to waste to remove any inorganic ions and salts prior to analysis by the mass spectrometer ( 130 ). This approach was also used by Weimann et al. for the determination of oxidative damage derived DNA and RNA adducts in human and rat urine ( 131 , 132 ). A more typical column switching set-up uses a small trap column onto which the sample is loaded, with the eluent initially being directed to waste removing any interfering compounds that are not retained on the trap column. At the same time the mobile phase flow is maintained to the mass spectrometer by a second LC pump connected to an analytical column. The retained DNA adduct is then back flushed from the trap column onto the analytical column which is then eluted into the mass spectrometer. This approach has been successfully used to determine DNA adducts derived as a result of oxidative stress, lipid peroxidation as well as tamoxifen and melphalan exposure ( 38 , 133 – 137 ). Churchwell et al. reported the simultaneous analysis of four different lipid peroxidation and oxidative stress derived DNA adducts in untreated rat and normal human liver tissue. The method allowed for the simultaneous analysis of 8- oxo -2′-deoxyguanosine, malondialdehyde-2′-deoxyguanosine and the etheno adducts of 2′-deoxyadenosine and 2′-deoxycytidine from a single aliquot of an enzymatically digested DNA sample ( 133 ). Etheno DNA adducts of 2′-deoxyadenosine have been studied using online C 18 trap column chromatography purification coupled to LC-ESI-MS/MS SRM in normal human placental and untreated animal tissues as well as in mice treated with urethane ( 134 ). Roberts et al. observed an ∼100-fold greater sensitivity for the detection of etheno-2′-deoxycytidine DNA adducts by LC-ESI-MS/MS SRM using online immunoaffinity chromatography purification coupled to graphitized carbon-trap columns when compared with using C 18 trap columns. The authors encountered a problem with the residual 2′-deoxyadenosine remaining following purification of the DNA hydrolysate which interfered in the analysis of etheno-2′-deoxycytidine due to the SRM transition used to monitor for this adduct being identical to that of 2′-deoxyadenosine. The problem was resolved by conversion of 2′-deoxyadenosine to 2′-deoxyinosine using adenosine deaminase ( 138 ). A slightly more sophisticated approach utilizing two automated switching valves allowing online sample concentration and cleanup was used to determine 4-aminobiphenyl adducts in calf thymus DNA and mouse liver DNA resulting from treatment with N -hydroxy-4-aminobiphenyl in vitro and in vivo , respectively ( 139 ).
The loss of the neutral 2′-deoxyribose moiety (116 u) may be used to scan for 2′-deoxynucleoside related compounds using CNL. Although the sensitivity of CNL is much lower than SRM and may not be able to detect DNA adducts formed in vivo , it does provide information for the determination of the different DNA adducts formed, for example when a genotoxic carcinogen is reacted with DNA for the first time in vitro . Regulus et al. detected previously unidentified DNA adducts induced by ionizing radiation in aerated calf thymus DNA by LC-ESI-MS/MS using CNL following digestion to 2′-deoxynucleosides. The authors reported the detection of three novel DNA adducts in the irradiated calf thymus DNA that were not present in the untreated calf thymus DNA ( 140 ). Rindgen et al. used CNL to confirm the presence of the major C -8 adduct and two previously unidentified minor PhIP modified 2′-deoxyguanosine adducts detected by 32 P-postlabelling in calf thymus DNA that had been treated with N -acetoxy-PhIP in vitro ( 35 ). Chaudhary et al. used solid phase extraction in conjunction with LC-ESI-MS/MS CNL to study calf thymus that had been treated with malondialdehyde as well as to confirm unequivocally the presence of the malonaldehyde-2′-deoxyguanosine adduct in healthy human liver DNA ( 141 ).
Mass spectrometric analysis of adducted 2′-deoxynucleotides
Methods for the quantitative determination of DNA adducts as 2′-deoxynucleotides by LC-ESI-MS/MS SRM are less common. Due to the presence of the phosphate group negative ESI is used, which is in contrast to the base and 2′-deoxynucleoside adducts which are analysed using positive ESI. Siethoff et al. studied the formation of 2′-deoxynucleoside 5′-monophosphate DNA adducts following the treatment of calf thymus DNA with styrene oxide using LC-ESI-MS SIM ( 142 ). Zollner et al. characterized the different DNA adducts formed following the reaction of cisplatin and four structurally related platinum (II) complexes with 2′-deoxyguanosine 5′-monophosphate using ion exchange LC-ESI-MS SIM ( 143 ). Doerge et al. used LC-ESI-MS/MS to characterize the DNA adducts formed following the reaction of 2′-deoxyguanosine 3′-monophosphate with malondialdehyde, crotonaldehyde, 2-hexenal and 4-hydroxy-2-nonenal as well as malondialdehyde-2′-deoxyguanosine 3′,5′-bisphosphate formed following the postlabelling of malondialdehyde-2′-deoxyguanosine 3′-monophosphate using non-radioactive ATP ( 144 ). Deforce et al. showed that phosphotriester adducts as well as base adducts were formed following the reaction of phenyl glycidyl ether, a compound used in the paint and resin industry, with calf thymus DNA ( 145 ). The position of the phosphate group on the 2′-deoxyribose may affect the extent of fragmentation. Hernandez showed that 2′-deoxynucleoside 3′-monophosphates underwent in-source dissociation much more readily than the respective 2′-deoxynucleoside 5′-monophosphates ( 146 ). Analysis by ESI-MS/MS CID can provide useful structural information about 2′-deoxynucleotide adduct standards, for example the difference between alkylated base and phosphodiester adducts can be ascertained. The typical negative ESI-MS/MS CID product ion spectrum obtained for the ethylated phosphodiester adduct of 2′-deoxyguanosine 5′-monophosphate is shown in Figure 4F . The product ions at m/z 125 and m/z 223 are consistent with ethylation of the phosphate group and the 2′-deoxyribose 5′-monophosphate of the adducted 2′-deoxynucleotide, respectively. The presence of the product ion at m/z 150 corresponding to guanine confirms that ethylation is limited to the phosphate group and not the base ( 147 ). Haglund et al. have studied the formation of ethyl phosphotriester dinucleotide adducts following the in vitro treatment of calf thymus DNA with N -ethyl- N -nitrosourea. The ethyl phosphotriester dinucleotide adducts were selectively released from the DNA by using nuclease P 1 since the enzyme is not capable of hydrolysing internucleotide bonds adjacent to a completely esterified phosphate group. A column switching approach was used to remove the unmodified 2′-deoxynucleosides following a further digestion of the DNA sample by incubation with a 5′-phosphodiesterase and alkaline phosphatase. Ten different ethyl phosphotriester dinucleotide adducts were characterized by their CID product ion spectra ( 148 ).
Mass spectrometric analysis of adducted oligonucleotides and the consequences of DNA adducts
Thus far we have discussed the various LC-MS approaches that can be used for the detection of the levels of DNA adducts following exposure to genotoxic carcinogens. However such approaches do not provide any information regarding the site of the adducts within the DNA sequence, as the internucleotide linkages are cleaved during the sample preparation procedures. Knowledge of the sequence selectivity of adduct formation and the mutational consequences of such adduct formation is of paramount importance, to help us (i) develop biomarkers of effect associated with the carcinogen exposure, (ii) identify the nature of the most mutagenically potent constituents of the complex mixtures of carcinogens to which humans are exposed and (iii) determine the risk associated with the carcinogen exposures.
LC-ESI-MS has been used to provide sequence information in numerous model studies where oligonucleotides have been adducted with genotoxic carcinogens in vitro . For example a study by Glover et al. employed ESI-MS/MS CID to determine the sequence of a 7-mer oligonucleotide that had been treated with para-benzoquinone, a reactive metabolite of benzene ( 149 ). Harsch et al. used LC-ESI-MS/MS for the identification of adduct sites as well the levels of adducts at the different sites, following the treatment of a double-stranded oligonucleotide derived from the HRPT gene sequence, with benzo[ c ]phenanthrene diol epoxide ( 150 ). Tretyakova et al. used LC-ESI-MS/MS for the analysis of double-stranded oligonucleotide sequences derived from the K-ras proto-oncogene and the p53 tumour suppressor gene containing 15 N-labelled 2′-deoxyguanosine placed at defined positions within the sequence that were treated with benzo[ a ]pyrene diol epoxide. The authors determined the levels of benzo[ a ]pyrene diol epoxide-2′-deoxyguanosine adducts that were formed non-randomly within specific positions of the sequence using SRM following enzymatic digestion of the DNA to 2′-deoxynucleosides. They also observed that the formation of the benzo[ a ]pyrene diol epoxide-2′-deoxyguanosine adduct was enhanced by the presence of a 5-methyl substituent at the cytosine base paired with the target guanine ( 151 ). Using a similar stable isotope labelling approach Matter et al. showed that the four different diastereoisomer adducts of benzo[ a ]pyrene diol epoxide were preferentially formed at the methylated CG dinucleotide sites present in a double-stranded oligonucleotide representing the p53 gene lung cancer mutational hotspots and their surrounding DNA sequences ( 152 ).
In order to determine the DNA sequence in which an adduct is placed in human samples, one would clearly need an exceptionally sensitive DNA sequencing method. Despite the recent development of high throughput technology for DNA sequencing, these procedures in general require the incorporation of a PCR step into the assay in order to produce sufficient material to allow for sequencing. Thus the adduct itself, even if present at a high enough level to be detected in the original DNA sample, would be lost prior to sequencing. It is however much more easier to determine the mutational consequences of the adduct, i.e. the nature of the altered base that may result following replication of the DNA.
The use of MS for DNA sequencing for mutation detection is growing in popularity. However for high throughput sequencing there has been extensive use in the past, for example of capillary array DNA sequencing using laser-induced fluorescence, which has established itself as a highly successful approach ( 153 ). Nonetheless mass spectrometric approaches also offer some other advantages compared with fluorescence detection, such as their high accuracy and specificity. Most mass spectrometric approaches for DNA sequencing have tended to employ matrix assisted laser desorption ionization (MALDI), which is a method of producing ions from samples in the solid phase in the presence of small chromophore molecule containing matrices. The matrix absorbs energy from a laser beam, which is dissipated to the sample resulting in ionization of the sample without fragmentation. The fundamental principles of the ionization process are still poorly understood ( 154 ). Due to its wide mass range and ease of automation for high speed sample throughput a wide variety of approaches have been developed for genotyping (single nucleotide polymorphisms (SNPs) detection) and mutation detection using MALDI-MS. The use of MS for genotyping has been reviewed by Jackson et al. (2000) and Edwards et al. (2005) ( 155 , 156 ).
Recently a highly promising LC-MS approach has been developed for the detection of mutations in short DNA fragments. Referred to as short oligonucleotide mass analysis (SOMA), this procedure involves PCR amplification of the area of interest of the gene with primers containing a sequence for the restriction endonucleases BpmI, and digestion of the PCR product with BpmI liberating short (<20-mer) oligonucleotides that can be analysed by LC-ESI-MS/MS (or MALDI-TOF) ( 157 ). This approach has been used for detection of base changes in the codons of the APC tumour suppressor gene, codon 249 of the p53 gene, hepatitis B virus and codon 12 of the K- ras gene ( 157 – 163 ). In the case of the p53 studies the sensitivity of the mass spectrometric method was compared with that of restriction fragment length polymorphisms (RFLPs) by Qian et al. with the conclusion that the former was ∼2.5 times more sensitive, increasing to 15-fold more sensitive when an extra digestion step with the restriction enzyme HaeIII was carried out to cleave wild-type sequences ( 159 ). Mutations in the p53 gene have been detected in tumour and plasma samples from hepatocellular carcinoma patients in China and also in samples from the Gambia ( 158 , 160 , 161 ).
In the cases exemplified above the mutation in codon 249 of the p53 gene is thought to be associated with exposure to aflatoxin B 1 . However the mutational characteristics (site-specificity) of other carcinogens are mostly less well understood. SOMA also has the potential to identify such mutation characteristics, which could subsequently be used as indicators of human exposure and possibly in risk assessment for other carcinogens. For example specific adducts may be incorporated within the supF gene, contained within a plasmid, which is then used in a forward mutation assay. The plasmid is transfected into human cells, which are allowed to replicate, and the plasmid is then recovered and screened in a bacterial system for mutations in the gene ( 164 ). Mutated sequences can then be analysed by a modified SOMA assay, allowing for the determination of any base changes that have occurred due to the presence of the adduct. There is presently a lack of information on the mutational significance of low levels of DNA adducts, and it is envisaged that similar improvements will be made in methods in the future for the detection of these biological effects to complement the increased sensitivity that is now available for the detection of the adducts in DNA.
Quantitation of DNA adducts by LC-MS
The great advantage of MS compared with other analytical methods is that it provides information about the structural identity of the compound under investigation, thus confirming that the correct compound is being studied. Furthermore accurate quantitation can be achieved by the use of stable isotope internal standards. These standards are identical in every respect chromatographically to the DNA adduct of interest but differ in mass and hence can only be differentiated by MS. The stable isotope standards also allow for the confirmation of the structural identity of the product ions formed by CID of the unlabelled DNA adduct of interest ( 110 , 122 ). Typically the 14 N and/or 12 C atoms in purines or pyrimidines are substituted for 15 N and/or 13 C, respectively, resulting in a higher mass ( 106 , 117 , 123 ). Alternatively stable isotope internal standards containing deuterium atoms ( 2 H) replacing 1 H atoms can be used ( 165 ). However caution is required for their use due to the possible occurrence of exchange of the deuterium for hydrogen. In the case of 8- oxo -2′-deoxyguanosine a standard containing 18 O substituted for 16 O at the C -8 position has been utilized ( 166 ). A novel application was the generation of a DNA standard containing 18 O labelled 8- oxo -2′-deoxyguanosine following incubation with 18 O labelled singlet oxygen, which was produced by the thermal decomposition of an 18 O labelled endoperoxide naphthalene derivative. This standard was used to investigate the potential of various DNA extraction and preparation (such as DNA digestion) procedures for the artefactual formation of 8- oxo -2′-deoxyguanosine. The ratio of unlabelled 8- oxo -2′-deoxyguanosine to 18 O labelled 8- oxo -2′-deoxyguanosine (which is not affected by artefactual oxidation) as determined by LC-ESI-MS/MS SRM allowed for the unambiguous comparison between different procedures in terms of assessing the artefactual formation of 8- oxo -2′-deoxyguanosine ( 167 ). The ability of MS to differentiate between compounds containing stable isotopes and those that do not allows for the simultaneous determination of endogenously and exogenously formed DNA adducts in animal studies, which can be used to investigate pathways for DNA adduct formation. For example Morinello et al. exposed rats to 13 C-labelled vinyl chloride to study the relative contribution of endogenous and exogenous formation of adducts in the liver and brain. The authors concluded that the reactive metabolite of vinyl chloride was not transported or formed in brain tissue, since 13 C-labelled N2 , N -3 ethenoguanine adducts were not detected, while the endogenously formed level of the adduct remained unchanged ( 168 ). Ham et al . showed that N2 , N -3 ethenoguanine was formed by direct alkylation following the treatment calf thymus DNA with 13 C-labelled ethyl linoleate under peroxidizing conditions. The authors showed that the level of incorporation of the 13 C label was 86% suggesting that N2 , N -3 ethenoguanine was also being formed by an alternative mechanism ( 169 ).
The stable isotope internal standards are normally added at the start of the procedure thus accounting for any losses during sample manipulation and preparation. The stable isotope internal standards also account for any variations in the response of the mass spectrometer such as suppression of ionization due to matrix effects. As an example Ravanat et al. used isotope dilution MS for the accurate detection of 8- oxo -2′-deoxyguanosine in calf thymus DNA, pig liver DNA and human urine ( 98 ). Though not ideal if stable isotope internal standards are not available, then an external calibration line must be obtained before and after the sample set using standards of the DNA adduct of interest spiked into the sample matrix, which are processed in exactly the same way as the samples to be analysed ( 107 , 170 ).
Comparisons have been made between the results obtained using LC-MS and other detection techniques for the determination of DNA adducts. A study was conducted by Beland et al. where the authors compared LC-ESI-MS SIM, 32 P-postlabelling and dissociation-enhanced lanthanide fluoroimmunoassay (DELFIA) for the determination of 4-aminobiphenyl adduct levels in calf thymus DNA that was reacted in vitro , and in the liver DNA of mice treated with radiolabelled [ 3 H] N -hydroxy-4-aminobiphenyl. The results showed that 32 P-postlabelling underestimated adduct levels (3–5% relative level), DELFIA overestimated adduct levels (260–550% relative level) and that LC-ESI-MS provided levels of adducts which agreed most closely with those obtained for the specific binding of the [ 3 H] radiolabel to the mice liver DNA. Furthermore the authors found that the LC-MS method provided the most accurate and precise results following statistical evaluation of the data when compared with the other two methods ( 139 , 171 ). Soglia et al. observed a good correlation between LC-ESI-MS/MS SRM and 32 P-postlabelling for the detection of IQ adduct levels in rat liver DNA; however, the levels of adducts determined by 32 P-postlabelling were lower ( 172 ). For the detection of 8- oxo -2′-deoxyguanosine both LC-ESI-MS/MS SRM and LC-ESI-MS SIM were validated by comparison with HPLC-electrochemical detection and GC-MS, respectively. In each case similar results were obtained for the levels of 8- oxo -2′-deoxyguanosine adducts formed in DNA that had been exposed to ionizing radiation in vitro ( 123 , 166 ). Hu et al. found that the levels of 8- oxo -2′-deoxyguanosine adducts determined by using ELISA in the urine of coke oven plant workers was 2-fold higher compared with levels detected using LC-ESI-MS/MS SRM. There was a significant correlation between the levels of 8- oxo -2′-deoxyguanosine adducts detected using the two methods; however, LC-ESI-MS/MS was the only method that showed a significant difference in the levels of the adduct between exposed and control individuals ( 173 ). Chen et al. obtained similar values for the level of 1, N6 ethenoadenine adducts in human placental DNA using GC-MS, LC-ESI-MS/MS SRM and HPLC-fluorescence detection ( 174 ). The same author also detected lipid peroxidation derived etheno adenine adducts in the human urine samples from healthy individuals using LC-ESI-MS/MS SRM and GC-MS, with both methods showing a good correlation for the level of adducts observed ( 121 ). Jaruga et al. obtained identical levels for the determination of the hydroxyl radical derived tandem DNA adduct, 8,5′-cyclo-2′-deoxyguanosine in mammalian cells using both LC-ESI-MS SIM and GC-MS ( 175 ). A study in Cynomolgus monkeys dosed with tamoxifen for 30 days showed that the levels of tamoxifen-DNA adducts detected in various organs by LC-ESI-MS/MS SRM were comparable to the levels detected by using a chemiluminescence immunoassay ( 176 ). Beland et al. used both LC-ESI-MS/MS SRM and 32 P-postlabelling to show that tamoxifen-DNA adducts were not present in human endometrial tissue obtained from women that had been treated with tamoxifen ( 177 ).
The development of MS has reached a stage where for some DNA adducts it represents the method of choice in terms of sensitivity as compared with 32 P-postlabelling. This is particularly true in the case of small/polar DNA adducts or adducts that are unstable and prone to spontaneous depurination. For example Dennehy and Loeppky showed that LC-ESI-MS/MS SRM provided better sensitivity and a much more rapid analysis time compared with 32 P-postlabelling for the detection of glyoxal-2′-deoxyguanosine and O6 (2-hydroxyethyl)-2′-deoxyguanosine adducts in rat liver DNA following dosing with N -nitrosodiethanolamine ( 178 ). The disadvantage of 32 P-postlabelling is that in most cases it does not provide information regarding the chemical identity of the DNA adduct being detected unless standards of the DNA adducts are available allowing for a chromatographic comparison. The procedure can be quite labour intensive and there is a requirement of a dedicated radioactivity laboratory equipped with protective shielding. For the sensitive detection of DNA adducts very high levels of specific activity for the 32 P isotope are required such as 7000 Ci/mmol. The labelling efficiency of T4 polynucleotide kinase may also vary for different DNA adducts ( 179 ). However the main advantage, in particular for the postlabelling of bulky/non-polar DNA adducts, is the requirement of low amounts of DNA, usually <10 μg for the analysis, which is in contrast to LC-MS where typically 50–100 μg of DNA is the normal requirement at the moment. This is still a great improvement compared with the early LC-MS publications where higher amounts (1–2 mg of DNA) were required to achieve sufficient sensitivity ( 77 , 141 ). In both cases though the methods require sample enrichment for obtaining increased sensitivity. The typical detection limits of LC-ESI-MS that have been quoted for various DNA adducts range from 0.5–5 adducts per 10 8 unmodified DNA bases, which is less sensitive in comparison with postlabelling where the limit of detection for some adducts is quoted at 0.1–1 adduct per 10 9 unmodified DNA bases ( 57 , 110 – 112 ). The detection limits of LC-ESI-MS are still improving as exemplified by recent studies on tamoxifen and estrogen DNA adducts where limits of detection at 2–5 adducts per 10 9 unmodified DNA bases have been reported and on etheno-2′-deoxyadenosine and 2′-deoxycytidine adducts where the detection limit has been reported to be as low as 1–2 adducts per 10 10 unmodified DNA bases ( 134 , 136 , 180 ).
Over recent years improvements of the sensitivity of 32 P-postlabelling have been limited, which is in contrast to MS where technological advances in instrumentation have resulted in increased efficiency of ionization, transmission and detection of ions. For instance the introduction of the so called Z-spray™ electrospray source where ions pass orthogonally into the mass spectrometer has resulted in improvements of sensitivity and less time spent on maintenance. (Micromass, Waters Ltd., Manchester, UK) Niessen (1998) and (1999) has published two reviews on the advances that have been made in the LC-ESI-MS technology ( 181 , 182 ). The extent of the development of LC-MS for DNA adduct detection can be gauged by the increasing number of publications where DNA adducts have been determined in human tissues and urine ( 92 , 121 , 131 , 132 , 173 , 174 , 180 , 183 – 187 ).
In terms of future developments for increasing the sensitivity of LC-MS, the introduction of capillary and nano capillary LC (to replace the normally employed narrow bore and microbore LC) coupled to micro- or nanoelectrospray ionization-MS has the potential to increase detection limits and sensitivity of DNA adduct analysis ( 188 – 190 ). Vanhoutte et al. observed an improvement in sensitivity by a factor of 3300 for the detection of 2′-deoxyguanosine 5′-monophosphate adducts of bisphenol A diglycidyl ether using nano capillary LC when compared with conventional LC ( 189 ). Abian et al. (1999) have published an extensive review on the practical considerations and technical aspects that need to be taken into account for the implementation of capillary LC with ESI-MS ( 188 ). Typical capillary-bore columns of 0.3–0.8 mm and nanobore columns of 0.025–0.1 mm internal diameters (I.D.) operate at flow rates of 5–10 μl/min and 200–500 nl/min, respectively. This is in contrast to the higher flow rates encountered for narrowbore (2.1 mm I.D.) and microbore (1.0 mm I.D.) columns operating at flow rates of 100–200 μl/min and 25–50 μl/min, respectively. Capillary LC results in a more concentrated sample containing eluent passing into the electrospray source and also helps improve the resolution as well as reduce the levels of buffers and other contaminants entering the mass spectrometer. Increased sensitivity is achieved due to the concentration-sensitive nature of ESI (a principle similar to that observed for a UV detector), which is an exceptional mode of operation of mass spectrometers, as they are normally regarded as mass sensitive detectors and dependent on analyte mass flow. The explanation for this phenomenon is that ionization and ion sampling efficiency increases at lower flow rates in the electrospray source thus allowing more analyte ions to enter the mass spectrometer ( 84 , 188 , 191 ). Ricicki et al. used capillary LC coupled to microelectrospray ionization-MS for the quantitation of 4-aminobiphenyl DNA adducts in human pancreas tissue using only 13.3 μg of the processed DNA sample for the analysis ( 185 ). The technology has also been successfully applied to the detection of the heterocyclic aromatic amine IQ DNA adducts in both in vitro and in vivo studies ( 100 , 172 ). Hoes et al. determined 2′-deoxynucleotide adducts in calf thymus DNA derived from the nitrogen mustard bifunctional alkylating agent, melphalan, using capillary LC coupled to positive ESI-MS/MS ( 192 ). The coupling of nano capillary LC to nanoelectrospray ionization-MS should lead to further gains in sensitivity. This approach was used by Embrechts et al. for the detection of various oestrogen derived DNA adducts in human breast tumour tissue and adjacent normal tissue biopsy samples ( 180 ).
The coupling of capillary electrophoresis (CE) to MS has also been investigated to study DNA adducts ( 193 ). The most simple form of CE, termed as capillary zone electrophoresis (CZE), allows for the separation of charged or polar compounds using fused silica capillaries filled with a buffer solution, based on their different migration under the influence of an electric field generating electro-osmotic flow. Detailed reviews discussing the advances in CE-ESI-MS have been published by Von Brocke et al. (2001) and Willems et al. (2005) ( 194 , 195 ). Using CZE-MS to analyse 2′-deoxynucleoside 5′-monophophates, Willems et al. identified the formation of adducts with exocyclic -NH 2 groups of guanine, adenine and cytosine following the reaction of benzo[ a ]pyrene diol epoxide with calf thymus DNA. The authors used an online sample stacking technique for the pre-concentration of the negatively charged adducts and removal of sample buffer ( 196 ). A similar approach was used by Barry et al. to analyse enzymatically hydrolysed calf thymus DNA treated with benzo[ a ]pyrene diol epoxide for which the authors noted the presence of an adducted thymidine–guanine dinucleotide in addition to a 2′-deoxyguanosine 5′-monophosphate adduct of benzo[ a ]pyrene diol epoxide ( 197 ). Again the sample stacking approach in combination with CZE-ESI-MS was used by Deforce et al. to detect the formation of adducts following the reaction of phenyl glycidyl ether with calf thymus DNA ( 145 ). The sample stacking approach is ideally suited for the analysis of 2′-deoxynucleotide DNA adducts due to the presence of a negative charge on the phosphate group and is discussed in detail by Wolf and Vorous (1995) ( 198 ). Gennaro et al. used CZE-MS for the detection of benzo[ a ]pyrene diol epoxide derived 2′-deoxyguanosine 3′-monophosphate adducts in calf thymus DNA which were selectively extracted from unmodified 2′-deoxynucleosides using metal affinity solid phase extraction ZipTips® following incubation with nuclease P 1 ( 199 ). Another development of CE is capillary electrochromatography (CEC), which is a hybrid separation method that combines the high separation efficiency of CE with HPLC using capillary columns packed with HPLC stationary phases, and again uses an electric field rather than hydraulic pressure as used in HPLC, to propel the mobile phase through the capillary for the separation of compounds. A detailed review on the coupling of CEC with MS has been published by Barcelo-Barrachina et al. (2004) ( 200 ) This approach was used by Ding and Vouros for the separation and identification of PAH DNA adducts formed in vitro ( 201 ).
Alternative ways to improve sensitivity could involve chemically altering the structure of the DNA adduct being studied so that it is more readily ionized ( 202 ). This approach was used by Otteneder et al. who reduced the aromatic nitrogen of the malondialdehyde-2′-deoxyguanosine adduct using sodium borohydride, resulting in the generation of a 5,6-dihydro derivative containing a secondary amine which was more readily protonated following analysis by positive ESI ( 128 ). Jeong et al. have developed an aldehyde reactive probe method that results in the formation of stable conjugates of malondialdehyde-2′-deoxyguanosine, which provides the advantage that the conjugate is more hydrophobic, hence more easily separated from unmodified 2′-deoxynucleosides as well as being more readily ionized compared with malondialdehyde-2′-deoxyguanosine, leading to a lower limit of detection ( 203 ). To improve the sensitivity and enhance chromatographic separations for the detection of N -acetyl-(2-aminofluorene)-2′-deoxyguanosine 5′-monophosphate adducts Flarako et al. generated hydrophobic derivatives on the phosphate group with for example hexamethyleneimine resulting in increased ionization efficiency in positive ESI, when compared with underivatized 2′-deoxynucleotide adducts ( 204 ).
Mass spectrometry has made enormous gains in its applicability for the detection of DNA adducts since its first uses in the 1960s, when up to 50 mg of DNA was required for the analysis of adducts using the direct probe insertion technique ( 205 ). As described above, it is now possible to characterize and in most cases determine DNA adducts by LC-MS with a sensitivity comparable to that of the other analytical approaches available for adduct detection, such as 32 P-postlabelling or immunoassays. Highly accurate and reproducible data can be obtained by using stable isotope internal standards for the determination of adducted nucleic acid bases, 2′-deoxynucleosides or 2′-deoxynucleotides, with LC-ESI-MS/MS SRM providing the greatest sensitivity and specificity. To increase the sensitivity of detection there is a requirement for adduct enrichment prior to analysis by LC-MS to overcome problems encountered with suppression of ionization due to matrix effects. Analytical detection limits for adducted bases or 2′-deoxynucleosides as low as 0.2–2.0 adducts per 10 8 unmodified DNA bases have been obtained with these mass spectrometric techniques using 10–100 μg amounts of DNA, indicating that these methods should be applicable for use in biomonitoring of exposure to genotoxic carcinogens in human molecular epidemiological studies ( 110 , 111 , 133 , 177 , 190 , 203 ). Furthermore technological advances are on-going, highlighted by the recent concept of microfluidics utilized in conjunction with nanoelectrospray ionization-MS which has led to the development of chip/LC-MS. This technology consists of microchannels etched onto glass, silicon or polymer chips, which are packed with stationary phase and are interfaced to the ESI-MS via microelectrospray tips. The advantages offered by chip/LC-MS include allowing the processing of samples of limited volume, sample enrichment and separation of analytes as well as complete automation of the analysis procedure and the possibility of offering screening based applications ( 206 – 208 ). The future impact of this new technology on the analysis of DNA adducts remains to be ascertained, although there seems no doubt that in future MS will take on a larger role in DNA adduct determinations in human studies, and that lower detection limits will be obtainable for the measurement of structurally characterized DNA adducts formed in human populations exposed to genotoxic carcinogens.
The authors acknowledge support from the Medical Research Council (grant G0100873).
Conflict of Interest Statement : None declared.