Fig. 3. Generation of ROS by IR: IR can directly ionize DNA or protein in its path generating DNA or protein radicals. Indirectly ionizing events include the ionization of water molecules surrounding the DNA or protein generating the reactive hydroxyl radical ( OH), which can then react with DNA or protein, rendering it reactive. The dashed circle represents a spur (Section 4.1) [30]. The shaded globules represent proteins.

Early studies by Fornace and Little [32] and [33] using alkaline elution demonstrated the induction of DPCs in aerated human cells exposed to very high doses of X-rays. They also showed an increase in DPC induction efficiency under hypoxic conditions. A similar observation was made by Meyn et al. [34] and [35] using Chinese hamster ovary (CHO) cells and by Radford [36] using mouse L cells, again using alkaline elution, and by Xue et al. [11] in V79 hamster cells using a filter binding assay. Zhang et al. [37], [38] and [39] suggested that negligible levels of DPCs are formed at oxygen concentrations above 1%, that there is maximal DPC induction at oxygen concentrations below 0.1%, and that oxygenated cells are 10–100-fold less susceptible to forming DPCs than hypoxic cells. Similarly, vanAnkeren, Murray, and Meyn (unpublished data) examined the relationship between oxygenation and DPC induction in CHO cells exposed to γ-radiation and found that the yield of DPCs decreases as oxygen levels increase (Fig. 4). Several other studies have also shown a marked increase in cellular DPCs induced by IR under hypoxic conditions [40], [41], [42], [43] and [44].

Fig. 4. Oxygen dependence of DPCs (squares) and survival (triangles) in γ-irradiated AA8 CHO cells. DNA–protein crosslinks were measured by the alkaline elution assay and cell survival was measured by colony-forming assay. The x-axis represents the percent oxygen in the gassing mixture. Single-cell suspensions were stirred at 4 °C while being gassed with a mixture of 5% CO₂, varying concentrations of O₂, balance N₂, for 3 h prior to irradiation.

Zhang et al. [38] showed that pH, nutrient depletion, temperature, and growth phase did not significantly influence the yield of IR-induced DPCs in aerated normal and tumor cells as measured by alkaline elution. Similarly, pH and nutrient status had no effect on cellular DPC induction when oxygen was absent [45]. Importantly, Zhang et al. [38] pointed out that it is difficult to compare DPC studies as the various techniques used to measure DPCs differ in their detection limits.

Given that the yield of DPCs in cells decreases markedly as oxygen is introduced, whereas the effect of oxygen on IR-induced cell killing goes in the opposite direction, and because the yield of other types of DNA damage such as DSBs closely parallels cell killing under these conditions, the role of DPCs in the biological effects of IR has been largely disregarded. However, as will be discussed in Section 9, these lesions may contribute to the radiosensitivity of hypoxic cells if their repair is compromised.

The situation with respect to DPCs and high linear energy transfer (LET) radiation has received some theoretical consideration. One unresolved question is whether DPCs, either alone or in association with clustered lesions, might differentially contribute to cell killing induced by radiations of differing LET. The thinking is that higher LET tracks will generate more complex clustered lesions, possibly with a higher probability of involving a DPC. Putative high-LET “specific” lesions could include complex clustered-damaged sites wherein DSBs are associated with DPCs [46], [47] and [48]. Some experimental studies have addressed the issue of whether the yields and/or repair of DPCs might differ with LET. Blakely et al. [49] showed that the initial DPC yields in normal hamster cells were similar for X-rays and high energy Ne-ions of 32, 100, and 183 keV/μm at low doses, although N-ions (120 keV/μm) generated a lower DPC yield. Another study suggests that a high-LET beam of N-ions appeared to induce higher levels of residual (6 h post-IR) DPCs per unit dose than low-LET X-rays in human melanoma cells (data from Eguchi et al. [50], re-calculated by Frankenberg-Schwager [51]). This difference may be attributable to the above-mentioned induction of more complex lesions at higher LET, rendering DPCs more difficult to repair.

4.2. Radiation-induced DPC structures

To understand the cellular consequences of DPCs and to investigate their possible repair pathways, it will be important to delineate the chemistries of these linkages (Fig. 1). Extensive work with cell-free models has demonstrated the covalent nature of IR-induced DPCs, and the chemical structure of some DPCs has been determined using gas chromatography/mass spectrometry (GC/MS) analyses [52], [53] and [54]. These reports examined γ-irradiated aqueous mixtures of thymine and amino acids (lysine, glycine, alanine, valine, leucine, isoleucine, tyrosine, and threonine) and demonstrated that particular DNA–amino acid crosslinks exist as several isomers [52], [53] and [54]. The involvement of these amino acids in DPCs was also shown in vitro in isolated irradiated mixtures of calf thymus nucleohistone [52], [53] and [54].

The GC/MS experiments were extended to analyze the formation of DPCs in vivo using cultured mammalian cells [44] and [55] and rat renal tissue [56]. These samples were treated with ferrous ions, hydrogen peroxide, or IR, and the chromatin was isolated, subjected to acid hydrolysis, and analyzed by GC/MS. Crosslinking of DNA to protein through a thymine–tyrosine linkage was detected in these samples. In both the in vitro and in vivo studies, the induction of DNA–amino acid complexes and DPCs increased linearly with IR dose. Hydrogen peroxide treatment of cultured cells also resulted in the concentration-dependent induction of DPCs in chromatin [44]. Addition of radical scavengers/metal chelators (dimethylsulfoxide or o-phenanthroline) partially inhibited DPC formation [44].

Dizdaroglu et al. [54] have proposed that the OH radical is involved in the formation of the crosslink whether these DPCs are induced by ferrous ions, hydrogen peroxide or IR. Free radicals/ROS are also generated through biological redox reactions and under conditions causing oxidative stress, such as malnutrition, numerous disease states, exposure to particular drugs and environmental pollution. The crosslinking mechanism involves H-atom abstraction from the methyl group of thymine by OH, addition of the resultant thymine radical to the carbon-3 position of the tyrosine ring, and oxidation of the resulting adduct radical [54].

Electrospray–ionization mass spectrometry (ESI–MS) analysis of an irradiated solution containing angiotensin and thymine demonstrated the formation of a covalent bond between the methyl group of thymine and C3 of the angiotensin tyrosine ring [57] and also indicated C2 of tyrosine as another major site of bond formation. Crosslinks between thymine and tyrosine were detected at IR doses as low as 0.1 Gy, and the yield of crosslinks was linear up to 100 Gy. Reaction of OH with thymine most frequently resulted in addition to the C5–C6 double bond ( 60% and 30%, respectively, at the 5 and 6 positions), and abstraction of an H-atom from the methyl group occurred only 10% of the time.

It will be of interest to determine if specific proteins found to be covalently crosslinked to DNA in vivo will prove to be linked through any of these identified target residues. Additionally, this information may be of use in predicting which proteins are likely targets for DPC formation because of their amino acid composition and their contact with the DNA. Identifying a crosslinked protein and the residue through which the linkage forms may also provide information on molecular geometry because the DNA and protein must be in close proximity during free radical generation.

4.3. Protein radicals and DPCs

DNA is not the only site of free radical generation or the only target for free radical attack following IR exposure (Fig. 3). Proteins and amino acids are also susceptible to attack by ROS. Indeed, an alternative mechanism for DPC induction involves an initial protein radical created by abstraction of an H-atom by OH from the amino acid, followed by addition of the amino acid radical to the C6 position of thymine and oxidation of the adduct radical [52]. ESI–MS studies by Weir Lipton et al. [57] show that OH adds to the tyrosine ring at C3 50% of the time and at C2 35% of the time. The C3 tyrosine adduct radical then loses water to generate a phenoxyl radical, which can then react with DNA. Thus, a DPC may be formed by the addition of a protein radical to DNA or vice versa, or from a combination of two radicals.

Exposure of proteins to ROS can generate protein hydroperoxides or other reactive protein species as well as additional free radicals. An in vitro study [58] used several purified proteins (insulin, α-casein, apotransferrin, and bovine serum albumin (BSA)) irradiated in aqueous solution in the presence of oxygen or nitrous oxide to generate protein hydroperoxides, and tested these for DPC formation with plasmid DNA based on the retardation of DNA migration on an agarose gel. The observation that inclusion of anti-oxidants did not reduce the yield of DPCs suggested that these lesions were not generated from long-lived radical species produced at the irradiation step. However, the formation of DPCs was reduced by including metal chelators in the reaction, suggesting that at least some of the DPCs were dependent on metal atoms associated with the DNA. Other reports have indicated that proteins that do not bind to DNA (e.g., BSA [59]) cannot generate DPCs in vitro, so there is some question as to whether or not non-DNA binding proteins can be involved in DPCs. It is likely that the conflicting reports reflect variations in in vitro experimental parameters such as DNA and/or protein concentrations, presence of radical scavengers, and presence of salts or metals or reductants that would interfere with the DPC formation reaction.

Further work examining the role of reactive protein species in DPC formation used hypochlorous acid (HOCl), an oxidant that is produced by normal metabolic processes such as phagocyte activity [60]. HOCl can react with protein amino groups, generating chloramines that decompose to protein radicals, which can react with DNA. HOCl can also interact with DNA to form chloramines. Hawkins et al. [60] investigated the formation of DPCs by HOCl in nucleosomes of eukaryotic-cell nuclei using electron paramagnetic spin resonance spectroscopy. The reaction of protein radicals with pyrimidine nucleosides was observed to yield nucleobase radicals, which could result in covalent crosslinking of DNA to protein. These authors [60] suggested that reaction of HOCl occurs predominantly with the protein and not the DNA, and that 50–80% of these reactions are with lysine or histidine residues. The finding that adduct formation was decreased in the presence of radical scavengers suggested that a radical is involved in this reaction.

Similar steps in DPC formation were suggested by analysis of malondialdehyde-induced DPCs in vitro [59]. These investigators reacted malondialdehyde with either protein or DNA in aqueous solution, purified away non-reacted material, and then attempted the second half of the DPC reaction (by introducing DNA or protein). For the formation of a DPC, it was apparent that the malondialdehyde must first react with the protein to generate an adduct that subsequently reacts with the DNA to form the crosslink.

Although both DNA- and protein-radical formation have been suggested as the first step in DPC formation in vitro, it remains to be seen which mechanism predominates in vivo. Both mechanisms are probably operative for various agents, and other factors may influence the levels of each type of radical produced. For example, in the case of IR, the spatial distribution of DNA and proteins in the radiation track may be critical in this regard [48].

5. Stability of DPCs in vitro

Different types of DPCs appear to have very different chemical stability. Aldehyde-induced DPCs are reversed by spontaneous hydrolysis and are also reversible by incubation at elevated temperatures (discussed in [24]). Acetaldehyde-induced DPCs are hydrolytically unstable, and in in vitro experiments only 25% of these DPCs remained after 8 h at 37 °C [61] and [62]. By comparison, malondialdehyde-induced DPCs formed in vitro using purified DNA and histone protein had a much longer half-life of 13.4 days at 37 °C [59].

The lifetime of formaldehyde-induced DPCs in vitro was investigated by Quievryn and Zhitkovich [24] using purified DNA and histone H1. Inclusion of either SDS or 0.8 M sodium chloride with the formaldehyde during the crosslinking reaction reduced crosslinking of histone H1 by preventing its binding to DNA. Addition of SDS after the formaldehyde crosslinking reaction decreased the lifetime of the histone H1–DNA DPC from 26.3 to 18.3 h at 37 °C, suggesting that if the protein is allowed to stay associated with the DNA, the crosslinks can reform under physiological conditions.

6. Biological consequences of DPCs

The covalent crosslinking of proteins to DNA is expected to interrupt DNA metabolic processes such as replication, repair, recombination, transcription, chromatin remodeling, etc. Indeed, the effect of agents that cause DPCs on DNA replication has been widely investigated ([63], [64] and [65] and others). DPCs are expected to act as bulky helix-distorting adducts and would therefore be likely to physically block the progression of replication or transcription complexes and/or prevent access of proteins required either for synthesis along the template strand, for transcription, or for repair recognition and/or incision. They may also affect all of these processes by anchoring the chromatin and preventing its remodeling.

Unfortunately, our understanding of the biological consequences of DPCs is hampered by the fact that no agent exclusively induces these lesions in genomic DNA (although studies using plasmid DNA have provided some insight into the processing of these lesions by cells; see below). Thus, all known DPC-inducing agents generate other forms of DNA damage in addition to DPCs, and direct attribution of any observed effect such as mutagenesis or carcinogenesis to DPCs is inevitably confounded by the concomitant impact of these other lesions. Nonetheless, several studies have reported that the induction of DPCs by many agents correlates with genetic damage such as sister chromatid exchanges (SCEs), transformation, and cytotoxicity [66], [67], [68], [69] and [70]. Thus, DPCs may contribute to the genotoxic effects of many different DNA-damaging agents, some of which are discussed below.

6.1. Nickel

Various types of chromosome damage (DNA gaps and breaks, SCEs and others) have been shown to persist in lymphocytes of nickel workers for years after exposure [71] and [72]. Earlier studies demonstrated an increased incidence of alveolar/bronchial/adrenal medulla neoplasms in rats exposed to nickel compounds [73].

6.2. Chromium

Chromium exposure has been associated with an increased incidence of respiratory cancers (reviewed in [74]). Voitkun et al. [75] used amino acid–chromium–DNA adducts (model DPCs) in a shuttle vector to show that processing of these lesions by human cells can result in mutagenesis. Plasmids containing DNA–Cr(III)–glutathione or DNA–Cr(III)–amino acid adducts were transfected into human fibroblasts, re-isolated after a 48 h incubation, and sequenced. The types of mutations caused by the DPCs were mainly single base substitutions at G:C base pairs, with G:C → A:T transitions and G:C → T:A transversions being induced with similar frequency. Chromium–DNA complexes also resulted in sequence mutations, although this effect was weaker.

The feasibility of using DPCs as biomarkers for exposure to chromium in human cells has been investigated [74]. Higher levels of DPCs were detected in lymphocytes of individuals exposed to chromium compounds than in non-exposed individuals, although the DPC level was found to plateau in individuals exposed to high levels of chromium.

6.3. Arsenic

Arsenic has been implicated in the induction of skin, lung, bladder, and liver cancers [76], [77] and [78]. Although it is carcinogenic, arsenic has not been found to be mutagenic. Earlier studies suggested that arsenic only induces DNA damage at high concentrations; however, a recent study [79] suggests that different cell types differ in their sensitivity to arsenic. Arsenic does in fact induce DNA damage at concentrations that are biologically relevant, the major forms of arsenic-induced DNA damage being oxidative DNA adducts and DPCs [79]. As well, multiple pathways have been proposed for arsenic-induced cytotoxicity [79]. Treatment with arsenite may result in DNA damage through the production of HOCl because there is an activation of NADH oxidase and an increase in superoxide production after NADH addition in arsenite-treated human vascular smooth muscle cells [80]. This pathway can result in DNA damage because superoxide is converted to hydrogen peroxide by superoxide dismutase, and the resulting hydrogen peroxide can react with chloride ions to form HOCl or with transition metal ions to produce OH [80], [81] and [82].