Formation and repair of unavoidable, endogenous interstrand cross-links in cellular DNA

Abstract

Genome integrity is essential for life and, as a result, DNA repair systems evolved to remove unavoidable DNA lesions from cellular DNA. Many forms of life possess the capacity to remove interstrand DNA cross-links (ICLs) from their genome but the identity of the naturally-occurring, endogenous substrates that drove the evolution and retention of these DNA repair systems across a wide range of life forms remains uncertain. In this review, we describe more than a dozen chemical processes by which endogenous ICLs plausibly can be introduced into cellular DNA. The majority involve DNA degradation processes that introduce aldehyde residues into the double helix or reactions of DNA with endogenous low molecular weight aldehyde metabolites. A smaller number of the cross-linking processes involve reactions of DNA radicals generated by oxidation.

Unavoidable DNA damage and its repair

DNA provides the blueprint and operating instructions for every living organism on earth [1]. The genetic information encoded in the sequence of DNA nucleobases must be maintained for the entire lifetime of a cell, many years in some cases [2]. Some elements of DNA structure are exceptionally stable. For example, the half-life for spontaneous hydrolysis of the phosphodiester linkages in the DNA backbone is 3 × 10⁷ y [3]. However, there are a number of chemical “weak spots” that render the DNA in cells susceptible to reactions that alter its structure and function. As a result, cellular DNA continually sustains low levels of unavoidable damage induced by reactions with water, reactive oxygen species, and aldehyde metabolites [4–6].

For example, cytosine and 5-methylcytosine residues undergo spontaneous and enzyme-catalyzed hydrolytic deamination to yield mispaired uracil and thymine residues in duplex DNA [7–12], a process that leaves a common mutational signature in the genetic code of living organisms [13]. The generation of reactive oxygen species in cells[14, 15] leads to DNA strand breaks via reactions with the sugar-phosphate backbone[16, 17] and oxidatively-damaged nucleobases such as 8-oxoguanine and formamidopyrimidines (Fapy) [18–20]. It is important to recognize that, given the large number of nucleotides in a genome (e.g. 12.8 billion in a diploid human cell) [21] even very inefficient reactions can produce significant numbers of DNA lesions. For example, the spontaneous hydrolytic loss of purine residues from DNA (depurination) occurs with a half-life of 730 y, yet this process generates more than 10,000 apurinic (AP) sites per day in each human cell [5, 22, 23]. As an interesting aside, a number of unavoidable DNA degradation processes (possibly including some that generate interstrand cross-links) [24] continue after cell death and limit the ability to recover sequence information from ancient DNA samples [25–27].

Error-prone replication of damaged DNA can introduce mutations in the genetic code [13, 28, 29]. Mutations contribute to the evolution of species [12, 30–33] but, generally, are either silent or detrimental to individual organisms [34–36]. For example, in mammals, mutations and genomic instability arising from DNA damage may contribute to cancer, aging, and neurodegeneration [37–43].

The relentless nature of endogenous DNA damage combined with requirements for genetic stability in living organisms, creates a selection pressure that has driven the evolution and retention of DNA repair systems across all forms of life [36, 44–52]. The human genome contains at least 130 genes encoding proteins involved in DNA repair [53]. DNA is the only molecule in the cell that benefits from such an extensive repair network. Logically, repair systems evolved to handle the common types of unavoidable DNA damage that block replication or cause genetic instability (e.g. mutations) [44–46, 48, 52]. In other words, endogenous DNA damage and replication errors [54] generate the substrates that explain the existence of most known DNA-repair systems. For example, the base excision repair enzyme uracil DNA glycosylase (UDG) excises uracil residues that arise from the unavoidable hydrolytic deamination of cytosine residues or the misincorporation of 2’-deoxyuridine in cellular DNA [12, 45, 55–58].

Cross-link repair pathways exist – but what are the endogenous substrates?

Interestingly, organisms ranging from bacteria to plants to humans possess the capacity to repair interstrand DNA-DNA cross-links (ICLs), in which the two complementary strands of the double helix are covalently bonded to one another [59–61]. The removal of ICLs from cellular DNA was first studied more than 50 years ago in the context of xenobiotics such as sulfur mustards, nitrogen mustards, mitomycin C, psoralens, and cisplatin [62–65]. However, it seems unlikely that ICL repair pathways evolved to protect cells from such external threats, given that agents of this type are not ubiquitous in the environment. Rather, the existence of ICL repair pathways across many diverse forms of life suggests that the generation of ICL(s) in genomic DNA is an unavoidable consequence of cellular life. However, the identity of these naturally-occurring ICLs remains uncertain. In this review, we consider endogenous chemical processes with the potential to generate ICLs in the DNA of living organisms.

Repair of Interstrand Cross-links in Cellular DNA.

Before considering the formation and structure of potential endogenous ICLs, it may be useful to briefly review the repair of ICLs in vertebrates. ICLs are exceptionally toxic [66] because they block the strand separation required for read-out and replication of genetic information embedded in the nucleotide sequence of DNA [67–69]. Cells really have “no good options” for dealing with cross-links in their DNA. Unrepaired cross-links may lead to cell senescence, cell death, and tissue dysfunction [37, 70–72]. On the other hand, the repair of cross-links can be error-prone, with the potential to generate genetic mutations and deletions [73–77]. In humans, endogenous ICLs may contribute to aging, neurodegeneration, and cancer [37–43].

The majority of ICL repair occurs during DNA replication [78–80]. When replication forks collide with an ICL, complex DNA damage response pathways are activated [67, 80, 81]. The Fanconi anemia pathway is critical for vertebrate cross-link repair [40, 42, 82, 83]. Association of the FANCI-FANCD2 proteins with a stalled replication fork is followed by ubiquitylation of this protein complex [67, 84–86]. This, in turn, leads to recruitment of nucleases that make incisions to “unhook” the cross-link (these incisions are depicted by lightning bolts in Scheme 1A) [67, 87, 88]. The endonuclease XPF-ERCC1 and the scaffolding protein SLX4 are critical in the incision process [42, 89, 90]. A number of other endonucleases that may be involved in cross-link unhooking have been identified, including MUS81-EME1, SLX1, FAN1, and SNM1A, although the circumstances under which they become involved remains uncertain [42, 89, 91–93]. There is evidence that the 5’→3’ exonuclease activity of SNM1A may digest through the ICL after the initial incision, thereby alleviating the need for a second incision [93– 96]. After unhooking, the resulting ICL remnant – now simply a large DNA adduct – is bypassed by error-prone translesion synthesis (TLS) polymerases such as pols η, κ, ι, and ν (eta, kappa, iota, and nu).[97–100] Subsequent extension after bypass often involves Rev1 and pol ζ (zeta) [67, 101–103]. The double-strand break induced by the incision(s) is repaired by RAD51-dependent homologous recombination [67, 73, 77, 104–106].

Scheme 1.

Simplified views of ICL repair pathways. Panel A: Dual incision, Fanconi- Anemia pathway (DSB = double strand break, HR = homologous recombination, Bypass involves translesion synthesis polymerases). Panel B: NEIL3-dependent unhooking of the dA-AP ICL (Repl. indicates normal error-free replication). Newly synthesized strands are shown in blue (right side).

Fanconi anemia is a clinically heterogeneous disease caused by mutations in one of twenty two proteins involved in replication-coupled ICL repair [107, 108]. Patients display symptoms that can include bone marrow failure, congenital skeletal anomalies, and increased risk of acute myeloid leukemia and other cancers [107–110]. A defining feature of the disease is that patient’s cells display genomic instability and hypersensitivity to interstrand cross-linking agents [111]. Thus, the symptoms of Fanconi anemia are thought to arise from failure to resolve endogenous ICLs during replication; however, as noted above, the nature of the unavoidable ICLs in cellular DNA that drive disease progression is not known [40, 112, 113]. The molecular events underlying the symptoms experienced by individuals with Fanconi anemia may extend beyond DNA repair deficits [114].

Several other ICL repair pathways have been described. Semlow et al. recently discovered an unprecedented repair pathway in which the base excision repair enzyme NEIL3 unhooks ICLs derived from AP sites in DNA and from psoralen (Scheme 1B) [115–117]. This pathway is discussed below in the context of AP- and acetaldehyde-derived ICLs. Seidman’s group discovered a FANCM-dependent ICL traverse that defers repair until after replication [118, 119]. Finally, replication-independent repair is important in post-mitotic and rarely dividing cells [120–23].

The three-dimensional structure of an ICL in duplex DNA may be important in replication-independent recognition and repair, with more distorted ICLs recognized and repaired more readily than those that retain normal B-DNA-like structures [120–123]. On the other hand, the three-dimensional structure adopted by cross-linked DNA may not be directly relevant to replication-dependent repair in which stalling of advancing replication forks generates strand-separated, Y-shaped, or X-shaped structures at the cross-link [89]. The capacity of the BER glycosylase NEIL3 to unhook various ICLs likely depends on the chemical stability of glycosidic bonds or imine linkages at the cross-linking site [117].

The role of aldehydes in the formation of endogenous ICLs is rooted in their unique chemical properties.

Subsequent sections of this review describe reactions of DNA with a variety of endogenous, low molecular weight aldehydes. As a result, it may be useful to examine some general properties of the aldehyde functional group that are relevant to its DNA-modifying properties. Aldehyde functional groups are electrophiles that can covalently modify both proteins and nucleic acids [124–128]. The reactivity of aldehydes (and their corresponding iminium ions) is most commonly associated with toxicity in biological systems [124–133], but has been exploited in the development of covalent enzyme inhibitors [134–137], probes for the identification of protein kinase substrates [138], preparation of protein conjugates,[139–141] characterization of macromolecular interactions in cells [142–144], ribosome-targeting antibiotics [145], and methods for fixing biological samples [143, 146].

Nucleophiles readily attack the electrophilic carbon atom of the aldehyde carbonyl group (Scheme 2) [147, 148]. However, unlike most classical DNA-modifying electrophiles, such as epoxides or alkylsulfonates, the reactivity of aldehydes is not irreversibly quenched by reaction with nucleophiles such as water. Rather, water and many other nucleophiles including thiols add rapidly, but reversibly, to the aldehyde (Scheme 2A) [147, 149–152]. The presence of thiols does not prevent the reaction of an aldehyde with nitrogen nucleophiles, though thiols may depress the rates of these reactions (see Figure 4 in ref [153]). Thus, aldehydes can diffuse through biological systems, attaching reversibly to water, thiols, and other biomolecules, until they come to rest at thermodynamically-favored locations.

Scheme 2.

Reversible addition of nucleophiles to aldehydes. Panel A: Reversible hydration of an aldehyde. Panel B: The reaction of an amine with an aldehyde reversibly generates an imine. When the reaction is conducted in the presence of a water-compatible hydride donor such as NaBH₃CN, the equilibria are drawn irreversibly forward to generate good yields of the corresponding alkylamine. This type of reaction is known as a “reductive amination of an aldehyde”.

Nucleobases react readily with aldehyde functional groups to generate DNA adducts [6, 129, 131, 154–162]. Indeed, most of the processes considered as candidates for the generation of endogenous ICLs in this review involve reversible reactions of aldehydes with nucleophilic nitrogen atoms in the DNA bases. Aldehydes react reversibly with both endocyclic and exocyclic nitrogen atoms in the nucleobases [154, 155]. Generally, adducts on the exocyclic NH₂ groups are formed more slowly, but display greater kinetic stability and are more readily characterized [154]. Attack of an amine group on an aldehyde residue generates a carbinolamine (also sometimes referred to as a hemiaminal, Scheme 2B) [154, 163, 164]. This can be followed by dehydration to produce the corresponding imine (also known as a Schiff base, Scheme 2B) [163, 164]. While the mechanism and rate of imine formation have been studied in detail [163, 164], relatively little is known about the factors that control the equilibrium yields of imine resulting from the reactions of various amines and aldehydes [165].

Imines can be irreversibly converted to the corresponding alkylamine adduct by hydride (H:–) reducing agents, such as sodium cyanoborohydride (NaBH₃CN), sodium borohydride, or sodium triacetoxyborane (Scheme 2B) [157, 166–173]. Biological agents such as glutathione, NAD(P)H, and ascorbate also have the potential to reduce some nucleobase-imine adducts [174–176]. Imines derived from enolizable aldehydes also have the potential to generate enamines.[177]

The nucleobases of DNA also react readily with α,β-unsaturated aldehydes [156, 160, 178–182]. The addition of nucleophiles to α,β-unsaturated aldehydes can occur either at the aldehyde (1,2-addition) or the double bond (conjugate, 1,4-addition, Scheme 3) [156, 160, 178–182]. Cyclic adducts and cross-links can result from a sequence of reactions involving initial 1,4-addition of a nucleophile, followed by 1,2-addition of a second nucleophilic center to the aldehyde unit (Scheme 3B) [130, 156, 160, 178–182].

Scheme 3.

Addition of amines to α,β-unsaturated aldehydes. Panel A: Reversible 1,2-addition to the aldehyde group. Panel B: Sequential 1,4-addition and 1,2-addition reactions can generate cyclic DNA adducts or DNA ICLs in which two nitrogen atoms are bridged by a 3-carbon, “propano” linker. There are a number of stereochemical considerations in these reactions: the configuration of the alkene in the α,β-unsaturated aldehyde can be E or Z (cis or trans), 1,4- addition has the potential to generate a new stereocenter, carbinolamine formation generates a new stereocenter, and imines can exist in E or Z configurations.

ICLs derived from abasic sites in cellular DNA.

Abasic (apurinic, AP) sites are generated by the spontaneous [5, 183, 184] and enzymatic [10, 55, 56, 185, 186] release of nucleobases from the DNA backbone and are among the most common unavoidable lesions present in cellular DNA [23]. The AP sugar residue in DNA exists as an equilibrating mixture of the ring-closed hemiacetal (99%) and the ring-opened aldehyde (1%, Scheme 4) [187]. Thus, an AP site in DNA positions an electrophilic aldehyde group on the interior of the double helix. Reactions of the AP aldehyde with adenine or guanine residues on the opposing strand of duplex DNA generate ICLs [188–191].

Scheme 4.

Panel A: Formation mechanism and chemical structure of native and reduced dG-AP ICLs. Panel B: DNA sequence motifs that support formation of dG-AP and dA-AP ICLs.

The dG-AP ICL has been observed in sequences where a guanine residue is offset one nucleotide to the 5’-side of the AP site on the opposing strand (Scheme 4) [188, 189]. Typical equilibrium yields of this ICL formed in duplex DNA are in the range of 2–3% [189]. Significantly higher (approximately 20%) yields of the corresponding chemically stable alkylamine ICL (dG-AP_red) can be generated when the reaction is carried out in the presence of the water-compatible hydride reagent, NaBH₃CN (Scheme 4) [188, 189, 192–194]. This reductive amination reaction draws equilibria forward via irreversible reduction of the imine and has been exploited for the preparation of DNA duplexes containing structurally-defined, chemically-stable, site-specific ICLs [193, 194]. Significantly higher yields of dG-AP_red have been observed in sequences containing mispaired bases near the AP site [195]. It remains to be seen whether biological agents such as glutathione, ascorbate, or NAD(P)H can stabilize the dG-AP ICL by reduction.

The dA-AP ICL has been observed in sequences where an adenine residue is offset one nucleotide to the 3’-side of the AP site on the opposing strand (Schemes 4B and 5) [190, 196]. Typical equilibrium yields of the dA-AP ICL in duplex DNA range from 15–70% depending upon the flanking sequences [190]. Reduction of the dA-AP ICL by chemical or biological reagents has not been described. In some cases, ICL yields are dramatically increased by the presence of mispaired bases near the AP site [195, 196].

Scheme 5.

Formation mechanism and chemical structure of the dA-AP ICL.

The formation of AP-derived ICLs can be detected by gel electrophoretic analysis within minutes of AP site generation in duplex DNA though the reactions require approximately 48–100 h to reach final equilibrium yields [189–191]. Once formed, the AP-derived ICLs are quite stable. The dissociation of the isolated dG-AP ICL occurs with a half-life of about 160 h and dissociation of the dA-AP ICL occurs with a half-life of 60–90 h, depending upon the sequence context (37 ˚C, pH 7) [192, 197].

Importantly, the dA-AP ICL is sufficiently stable to block DNA replication by a highly processive, strand-displacing polymerase [198]. Similarly, in single molecule experiments, the dA-AP ICL is stable against dissociation when the flanking DNA on one side is unzipped in a protein nanopore [196]. The kinetic stability of the AP-derived ICLs may stem from the fact that they exist as cyclic aminoglycosides, rather than as the hydrolytically-labile imine form [192, 197, 199]. Literature precedents indicate that cyclic nucleobase adducts of this type, including N⁶-glycosyladenine, N²-tetrahydropyranoyl-dG, N²-tetrahydrofuranosyl-dG, N²-glucopyranosyl-dG and N²-ribosyl-dG derivatives, possess substantial stability [200–204].

The preferred sequence motifs for the dA-AP and dG-AP ICLs reflect an interplay of the right-handed helical twist of DNA and the groove location of the exocyclic NH₂ groups involved in the cross-links [194, 205]. These structural issues are illustrated by the recently solved NMR structure (pdb code 6xah) of the dA-AP ICL located in a 5’CXT/5’AAG sequence (where X=AP and the underlined A is the site of ICL attachment) [206]. The NMR structure shows that the dA-AP cross-link in duplex DNA exists in the ring-closed (aminoglycoside) form and, at least in this sequence context, is exclusively in the β configuration. Juxtapositioning of the N⁶-dA amino group and C1 of the AP site within bonding distance, while maintaining the flanking unpaired adenine and thymine bases stacked within the duplex is achieved by a modest unwinding of the DNA at the site of the ICL [206]. Otherwise, the duplex retains a normal B-DNA structure. The relatively undistorted nature of the dA-AP ICL may limit recognition and removal of this lesion by replication-independent repair [120, 121, 123].

Recent studies by Walters’ group identified an unprecedented pathway for repair of the dA-AP ICL. They showed that the dA-AP ICL blocks DNA replication in Xenopus egg extracts and is unhooked at the stalled replication fork by the DNA glycosylase NEIL3 (Scheme 1B) [116, 117]. Unhooking of the dA-AP ICL by NEIL3 regenerates the native adenine residue and an AP site. Translesion synthesis across the AP site is error-prone [207], but the NEIL3 unhooking pathway evades double-strand breaks, homologous recombination, and bypass of a bulky adduct that can introduce point mutations and deletions. NEIL3 was shown to be the “front line” mechanism for unhooking of the dA-AP ICL – engaging before the Fanconi anemia pathway is activated by the stalled replication fork. When the NEIL3 unhooking pathway was inhibited by various means, the classical Fanconi anemia pathway engaged to resolve the dA-AP ICL [116].

In biochemical experiments, the glycosylase domain of NEIL3 is competent to unhook the dA-AP ICL selectively in splayed-arm substrates that mimic the strand-separated DNA at the leading edge of a replication fork [116, 208]. In Xenopus extracts, replication-dependent unhooking of dA-AP by NEIL3 requires both the glycosylase domain and the C-terminal domains of the enzyme [115]. NEIL3 is recruited to the dA-AP ICL by association of the tandem GRF1 and GRF2 zinc finger domains with regions of single-stranded DNA at the leading edge of the stalled replication fork[209] and binding of the NPL4-type zinc-finger (NZF) domain to ubiquitin chains that are added to the stalled CMG helicase by the E2-ubiquitin ligase TRAIP [115]. The GRF domains also regulate NEIL3 activity on splayed substrates such as the dA-AP ICL.[209] The discovery of this specialized NEIL3 repair pathway for unhooking of the dA-AP ICL may provide circumstantial evidence that this ICL is important in living systems [115, 116, 208].

Complex ICLs arising from strand breaks at abasic sites in duplex DNA.

As noted above, AP sites are common, unavoidable lesions in cellular DNA [5, 22, 23]. AP sites can be converted to DNA strand breaks by β-elimination of the 3’-phosphate group (Scheme 6) [183]. This reaction can be catalyzed by biogenic polyamines such as spermine, spermidine, and putrescine [168, 183, 210, 211] and by amine residues on proteins and peptides [171, 212]. The amine-catalyzed elimination reaction proceeds via iminium ion intermediates to produce an α,β-unsaturated sugar remnant (referred to here at the 3’dRP group) on the 3’-terminus of the resulting strand break [213]. Mass spectroscopic studies have detected the 3’dRP lesion in cellular DNA [214, 215].

Scheme 6.

Elimination of the 3’-phosphoryl group from an abasic (AP) site generates electrophilic α,β-unsaturated iminium ion (3’dRP-iminium) and α,β-unsaturated aldehyde (3’dRP) sugar remnants in duplex DNA.

Recent work has shown that the electrophilic α,β-unsaturated iminium ion or aldehyde generated by cleavage of an AP site in duplex DNA can lead to formation of an ICL at the strand break. Specifically, spermine-mediated strand cleavage at an abasic site located in a 5’CXT/5’AAG sequence (where X = AP site and the ICL attachment is to the underlined A) leads to cross-link formation by reaction with an adenine residue on the opposing strand (Scheme 7) [117]. The ICL is chemically-stable and is a complete block to DNA replication by the strand-displacing ϕ29 DNA polymerase [117]. Chemical structure(s) of the ICL consistent with the data are shown in Scheme 7. Chemical precedents gleaned from reactions of the low molecular weight α,β-unsaturated aldehyde acrolein with 2’-deoxyadenosine and polydeoxyadenylic acid suggest that ICL formation in this case proceeds via conjugate addition of N1-dA to the 3’dRP end group [182, 216]. An alternative N⁶-dA attachment (indistinguishable on the basis of mass spectrometric data) could arise either via direct N⁶-attack or Dimroth rearrangement of an initial N1-adduct [182].

Scheme 7.

Spermine-catalyzed generation of a strand break in duplex DNA, followed by attack of an adenine residue from the opposing strand on the 3’dRP sugar remnant, produces a complex (clustered) lesion composed of an ICL adjacent to a DNA strand break.

In light of the propensity for various nucleobases to form adducts with α,β-unsaturated aldehydes [126, 180–182, 216–220], it seems likely that strand cleavage at abasic sites can give rise to a family of ICLs derived from reactions of various nucleobases with the 3’-α,β-unsaturated sugar remnant. The general architecture of the dA-dRP lesion is analogous to that of the ICLs generated by 5’-dioxobutane lesions and C4’-oxidized abasic sites described below.

Acetaldehyde-derived ICLs.

Acetaldehyde is mutagenic and carcinogenic in cellular and animal experiments [221, 222]. Exposure to acetaldehyde comes from a variety of sources including various combustion processes, bacterial metabolism of glucose, and its presence in foods [126, 223, 224]. Most human exposure to acetaldehyde is, strictly speaking, not from endogenous sources. Rather, acetaldehyde arises primarily from the metabolism of ethanol to acetaldehyde [221, 222]. The acetaldehyde generated by ethanol oxidation is further metabolized to acetate by the enzyme ALDH2 [221, 222]. Individuals with genetic variations that compromise the function of ALDH2 have an increased risk of alcohol-related cancers, presumably due to elevated levels of acetaldehyde-derived DNA damage experienced following ethanol consumption [221, 222].

Acetaldehyde generates a variety of DNA adducts [157, 175, 221, 222, 225–227]. Most important in the context of this review, early work suggested that acetaldehyde generates ICLs in cellular DNA [228, 229]. Although a single equivalent of acetaldehyde can generate an intrastrand cross-link between two guanine residues in DNA (Figure 1) [226, 230], there is no evidence for the generation of interstrand DNA cross-links involving reaction of a single acetaldehyde unit with nucleobases on the opposing strands of the double helix, despite the fact that such an ICL is structurally reasonable, given the existence of the analogous formaldehyde-derived ICLs (see below). Rather, the acetaldehyde-derived ICLs that have been characterized arise from initial generation of an 1,N²-α-methyl-γ-hydroxypropano-2’-deoxyguanosine adduct (αMeγOHPdG, abbreviated as PdG in Scheme 8) in duplex DNA (Scheme 8) [157, 231, 232]. This guanine adduct is formed either via an aldol condensation of two equivalents of acetaldehyde in solution to generate crotonaldehyde [175, 233] or via reaction of a second equivalent of acetaldehyde with the N²-ethylidene-2’-deoxyguanosine adduct on DNA (Scheme 8) [157]. This ICL-precursor adduct has been detected in the DNA of cultured cells treated with acetaldehyde[234] and in DNA isolated from the blood of ALDH2-deficient alcoholics [235]. Biogenic amines and histones may catalyze the formation of the αMeγOHPdG adduct [236–239].

Figure 1.

Acetaldehyde generates intrastrand G-G cross-links.

Scheme 8.

Condensation of two equivalents of acetaldehyde gives rise to a propano-2’- deoxyguanosine (PdG) adduct in cellular DNA. The propano-2’-deoxyguanosine lesions can generate ICLs (Scheme 9).

The αMeγOHPdG adducts are stable and, in nucleosides and in single-stranded DNA, exist predominantly in the ring-closed form (Scheme 8) [218, 232]. On the other hand, in duplex DNA, when the dG adduct is paired with a cytosine residue, the adduct undergoes partial ring-opening that enables the aldehyde group at the γ-position to react with the exocyclic NH₂-group of a guanine residue on the opposing strand to generate ICLs at 5’CG sequences (Scheme 9). Starting from the α-R-isomer of the αMeγOHPdG adduct installed in a 5’CG sequence in duplex DNA, ICL formation reaches an equilibrium yield of approximately 40% over 21 days at pH 7 and 37 ˚C [157, 231, 232, 240]. In duplex DNA this ICL does not disrupt base pairing (PDB code: 2hmd) [241]. The nucleosidic ICL remnant obtained by digestion of duplexes containing this cross-link exists in the cyclic pyrimidopurinone form and is quite stable (Scheme 9) [157, 231, 232, 240]. In duplex DNA, however, the ICL exists as the ring-opened carbinolamine (Scheme 9) [160, 232, 240]. This cross-linked DNA duplex is stable enough to be analyzed by HPLC, isolated, and subjected to ligation reactions [231, 240, 242], but the ICL is completely dissociated by thermal denaturation of the duplex [240]. Treatment of duplex DNA containing the αMeγOHPdG ICL with NaBH₃CN for 24 h in pH 7 buffer generated the corresponding, chemically-stable trimethylene-type ICL (Scheme 9) [240]. It is not known whether biological agents such as NAD(P)H, ascorbate, or glutathione have the capacity to reduce this ICL.

Scheme 9.

Reversible generation of imine and carbinolamine ICLs from propano-2’- deoxyguanosine (PdG) adducts at 5’CG sequences in duplex DNA and conversion to chemically-stable trimethylene-type ICLs by the reducing agent sodium cyanoborohydride (NaBH₃CN).

Replication-dependent repair of the αMeγOHPdG ICL has been examined in a Xenopus egg extract [242]. The evidence indicates involvement of two distinct unhooking pathways for this ICL. The dual-incision, Fanconi anemia pathway is activated and contributes to unhooking and bypass of the ICL [242]. Interestingly, a second, faster, Fanconi-independent unhooking pathway also was observed. Both unhooking pathways are dependent upon the convergence of two replication forks on the ICL [242]. The corresponding reduced (chemically stable) αMeγOHPdG_red ICL was unhooked exclusively by the Fanconi pathway, suggesting that the Fanconi-independent pathway exploits the reversible nature of this imine/carbinolamine cross-link [242]. The Fanconi-independent pathway did not involve the glycosylase NEIL3 or generation of an AP site [242]. Overall, the evidence is consistent with a Fanconi-independent unhooking reaction that regenerates native dG on one strand and the αMeγOHPdG adduct on the opposing strand [242]. Bypass of the adduct generated by the Fanconi-independent unhooking process is error-prone[243, 244], requires Rev1 and pol ζ [242, 244], and shows a mutation signature identical to that of the αMeγOHPdG adduct [242]. The Fanconi-independent unhooking reaction could be catalyzed by some as-yet-unidentified protein at the replication fork. However, it seems equally possible that the faster, Fanconi-independent unhooking process observed for the αMeγOHPdG-ICL results from spontaneous dissociation of the ICL induced by strand separation of the duplex at the stalled replication forks. The idea that this PdG ICL could dissociate spontaneously in the strand-separated DNA at a stalled replication fork on the timescale of the replication-coupled repair assay is supported by earlier experiments showing that the structurally-related ICL derived from the 1,N²-γ-hydroxypropano-2’-deoxyguanosine adduct readily dissociated (within 1 h) under conditions that disrupt the duplex structure (unbuffered water) [160, 240].

The observed engagement of the Fanconi anemia pathway in the repair of acetaldehyde-induced ICLs in a Xenopus egg extract is consistent with several previous observations: 1. cells deficient in Fanconi anemia proteins are hypersensitive to acetaldehyde [245, 246], 2. acetaldehyde activates the Fanconi anemia response in cells, and 3. acetaldehyde induces replication-dependent double-strand breaks in mammalian cells [246–248]. Recent work showed that Aldh2^–/–Fancd2^–/– knockout mice show profound health deficits including susceptibility to lymphoblastic leukemia, dramatic decreases in their hematopoietic stem cell pool, and ethanol-induced bone marrow failure [249–251]. This meshes with results showing that genetic deficiencies in ALDH2 activity in Fanconi anemia patients correlate with more rapid progression toward bone marrow failure [252]. Curiously, Aldh2^–/–Rev^–/–mice do not display the same phenotype as Aldh2^–/–Fanca^–/– animals despite apparent involvement of Rev1 repair of the acetaldehyde-derived ICLs via the Fanconi anemia pathway.

Overall, these results suggest acetaldehyde (and perhaps other endogenous ALDH2 substrates) generates ICLs that require the Fanconi pathway for repair [249–251]. It is important to note that a number of different biogenic aldehydes are substrates for ALDH2 including acetaldehyde, 4-hydroxynonenal, and acrolein [253–255]. While the ICL precursor lesion αMeγOHPdG has been found in cellular DNA [235], the actual ICLs derived from this precursor have not been detected in cellular DNA. It seems possible that acetaldehyde-derived intrastrand cross-links[226, 230] could be relevant to some of the observations surrounding Fancd2^–/–Aldh2^–/^– cell lines and animals, considering evidence that FANCD2 is involved in cellular tolerance of intrastrand cross-links induced by UV-light [256, 257].

ICLs derived from α,β-unsaturated aldehydes, acrolein, crotonaldehyde, and 4-hydroxy-2-nonenal.

A number of α,β-unsaturated aldehydes including acrolein, crotonaldehyde, and 4-hydroxy-2-nonenal (HNE) are generated by endogenous metabolism of amino acids and polyamines, and via lipid peroxidation [125, 126, 156, 217, 258–261]. These compounds are also ubiquitous in foods and the environment [126, 156, 217, 258–261].

These α,β-unsaturated aldehydes have the potential to react with all of the DNA nucleobases, but the best characterized adducts are those with 2’-deoxyguanosine [126, 180–182, 216–220]. As described above in the context of the crotonaldehyde adducts derived from the condensation of acetaldehyde, α,β-unsaturated aldehydes generate PdG-type adducts on guanine residues in cellular DNA (Schemes 8 and 9) [218, 259, 262]. PdG adducts generated by α,β-unsaturated aldehydes are removed by nucleotide excision repair[263] and potentially by ALKB [264], but these lesions have been detected in cellular DNA [217, 244, 258, 265]. The PdG adducts derived from acrolein, crotonaldehyde, and HNE serve as precursors for the generation of ICLs in duplex DNA [160, 232, 240, 266–268]. The properties of these ICLs are similar to the crotonaldehyde-derived ICLs described above in the context of acetaldehyde. Starting from the corresponding R-isomers of the PdG adducts in DNA, the equilibrium yields of ICLs (pH 7, 37 ˚C) are approximately 50% after 7 days for acrolein, 40% after 21 days for crotonaldehyde, and 85% after 2 months for HNE [160]. To date, ICLs derived from α,β-unsaturated aldehydes have not been detected in cellular DNA.

Acrolein, crotonaldehyde, and HNE are detoxified in cells by ALDH enzymes including ALDH2, ALDH1A1 and ALDH7A1 [253–255, 269, 270]. Thus, the cellular levels of these α,β-unsaturated aldehydes are expected to increase in ALDH2-deficient individuals and in Aldh2^–/– cell lines [235]. As a result, ICLs derived from acrolein, crotonaldehyde, and HNE could contribute to the genetic toxicity induced by mutations in, or knockout of, the Aldh2 gene [112, 249, 250]. It is noteworthy, however, that chicken DT-40 cells deficient in FANCD2 are not hypersensitive to acrolein or crotonaldehyde (in contrast, these cell lines are hypersensitive acetaldehyde or formaldehyde) [246]. These results suggest that acrolein and crotonaldehyde do not generate significant levels of ICLs in DT-40 cells, or that the ICLs derived from these chemicals are unhooked by a Fanconi-independent pathway, or that the ICLs dissociate spontaneously during DNA replication.

Malondialdehyde-derived ICL.

Malondialdehyde is an endogenous product of thromboxane biosynthesis and lipid peroxidation[271–273] that exists as a mixture of the dialdehyde, the enol, and (predominantly) β-alkoxyacrolein at physiological pH (Scheme 10). This compound is a bifunctional electrophile that forms cyclic adducts with nucleobases [274–279].

Scheme 10.

Formation an ICL from the malondialdehyde-2’-deoxyguanosine adduct (M₁G) in duplex DNA.

The major lesion formed by the reaction of malondialdehyde with DNA is the 2’-deoxyguanosine adduct, commonly referred to as M₁G [274, 276]. The M₁G adduct has been detected in the DNA of healthy humans [276]. Interestingly, base propenal residues generated by oxidative DNA strand cleavage serve as malondialdehyde equivalents, capable of generating the M₁G lesion [280, 281].

Similar to the PdG adducts described above, the M₁G adduct paired with a complementary cytosine residue in duplex DNA exists in the ring-opened form that has the potential to generate cross-links [282]. Indeed, gel electrophoretic analysis provided evidence that malondialdehyde generates an ICL at 5’CG sequences [283]. This ICL has not been well characterized because it is generated in low yields and is unstable [283, 284], but the proposed[283] structure shown in Scheme 10 is consistent with the chemical expectations and the data [283]. Interestingly, it was not possible to generate the malondialdehyde ICL from the M₁G lesion using the same approaches employed with great success to characterize the PdG-type ICLs [285]. The three-dimensional structure of the malondialdehyde ICL in duplex DNA has been modeled by a dG-dG trimethylene cross-link (lower structure, Scheme 9) [286].

Malondialdehyde is a substrate for ALDH2[255] and, as a result, has the potential to contribute to the genotoxicity observed in Aldh2^–/– cell lines and individuals with ALDH2 deficiencies. Malondialdehyde also can generate DNA-protein cross-links (DPCs) [287]. No direct evidence has been reported supporting the presence of malondialdehyde-derived ICLs in cellular DNA.

Formaldehyde-derived ICLs.

Formaldehyde is ubiquitous in the environment and in foods [288]. In addition, formaldehyde is unavoidably generated by endogenous metabolic processes including demethylation of histones in the cell nucleus [125, 126]. Formaldehyde concentrations in the range of 30–300 μM have been reported in mammalian tissue and blood [289–291]. Formaldehyde has been classified as a human carcinogen [292–294].

Formaldehyde forms reversible covalent adducts with both DNA and proteins [154, 155, 295]. These chemical properties underlie the use of formaldehyde as an embalming agent and a fixative in molecular and cell biology [146]. Formaldehyde is also used to cross-link proteins to their DNA-binding sites in chromatin immunoprecipitation (ChIP) assays [144, 296]. Formaldehyde can condense with other aldehydes to generate complex mixtures of DNA adducts [157, 297].

Formaldehyde reversibly forms adducts with all of the DNA bases [154, 155]. Adducts on the endocyclic nitrogens of the nucleobases typically form faster but are less stable, while adducts with the exocyclic nitrogens (–NH₂) form more slowly but are more stable [154, 155]. The hydroxymethyl (carbinolamine) adducts generated by the reaction of the exocyclic amino groups of dGMP and dAMP with formaldehyde decompose with half-lives in the range of 2–6 h at neutral pH and 23 ˚C [146, 298].

Formaldehyde generates cross-links between nucleobases in duplex DNA [159, 299, 300]. Analysis of hydrolysates of DNA treated with formaldehyde provided evidence for dG-CH₂-dG, dA-CH₂-dA, dG-CH₂-dA, dC-CH₂-dG and dC-CH₂-dA cross-links [159, 299, 300]. The dA-CH₂-dA cross-link has been detected in the hepatic DNA of rats treated with formaldehyde-releasing chemicals such as N-nitrosodimethylamine [301]. These studies also detected background levels of the cross-link in untreated animals, presumably arising from endogenous formaldehyde [301]. In general, it is not possible to know whether cross-link “remnants” obtained by the digestion of DNA arise from interstrand or intrastrand cross-links. However, in this case, there is good in vitro evidence that formaldehyde generates ICLs in duplex DNA [299, 300, 302–305]. Hopkins’ group rigorously characterized the structure and properties of the dA-CH₂-dA ICL formed at 5’AT sequences in duplex DNA (Scheme 11) [300, 302]. This ICL is formed slowly and in low yields in vitro, with <1% yield generated over the course of 9 days when a oligomeric DNA duplex was treated with formaldehyde (25 mM) in pH 6 phosphate buffer (50 mM) containing NaCl (25 mM) at 25 ˚C [300, 302]. Once formed, the dA-CH₂-dA ICL is quite stable both in duplex DNA and as the dinucleoside ICL remnant, a feature that enables detection and characterization of the lesion [146, 299, 300, 302].

Scheme 11.

Formaldehyde generates an ICL at 5’AT sequences in duplex DNA.

There is indirect evidence suggesting that formaldehyde has the capacity to generate intrastrand G-G and G-A cross-links in DNA.[306] This process is analogous to the better-characterized generation of intrastrand cross-links by acetaldehyde (Figure 1) [226, 230]. In addition, formaldehyde-mediated DNA-protein cross-links have been detected in cells [128, 307].

In humans, formaldehyde is detoxified by the glutathione-dependent formaldehyde dehydrogenase activity of the enzyme alcohol dehydrogenase 5 (ADH5) [308–310]. Other enzymes involved in the removal of formaldehyde include ADH1, ALDH2, and cytochrome P450 2E1 (CYP2E1) [308, 309]. Formate resulting from oxidation of formaldehyde may be shunted into the one-carbon metabolic cycle for use in the biosynthesis of nucleosides [310, 311]. Vertebrate cell lines deficient in Fanconi anemia proteins such as FANCD2 or FANCB are hypersensitive to formaldehyde [245, 246]. Furthermore, simultaneous knockout of ALD5 and FANCL or FANCD2 is lethal to vertebrate cells [312]. Adh5^–/–Fancd2^–/– mice suffer from bone marrow failure, liver and kidney dysfunction, and fatal malignancies [72]. Together, the results suggest that aldehyde detoxification and cross-link repair pathways work together to protect the mammalian genome from formaldehyde-derived damage.

Glyoxal-derived ICLs.

Glyoxal is generated in cells by a variety of processes including carbohydrate metabolism, amino acid metabolism, lipid oxidation, and DNA oxidation [126, 223]. This dialdehyde generates DNA adducts with dA, dC, and dG [313]. Similar to acrolein and MDA, glyoxal forms a cyclic adduct with dG [314, 315]. The cyclic dG adduct is stable, decomposing with half-life of 595 h in duplex DNA (pH 7.4, 37 ˚C) [316]. Glyoxal-DNA adducts may be removed by the enzyme DJ-1/Park7 although this issue remains controversial [317, 318].

Reactions in the context of nucleosides indicate that glyoxal has the potential to generate DNA cross-links involving dG/dG, dC/dC, dG/dC, and dG/dA attachments [313, 319]. For example, reaction of the glyoxal-2’-deoxyguanosine nucleoside adduct with 2’-deoxyguanosine (300 mM) in pH 7 buffer at 60 ˚C for 6 days produced a 4–5% yield of the GgxG linkage (Scheme 12) [319]. Higher yields (71%) of the GgxC dinucleoside were obtained in DMSO (60 ˚C, 6 d) [319]. While these dinucleoside cross-link attachments are formed slowly, they are chemically stable, showing no dissociation over 15 h at pH 7.4, 37 ˚C [316]. The GgxC, GgxA, and GgxG cross-link remnants have been detected in human placental DNA hydrolysates at levels of 3.50, 2.49, and 1.26 per 108 nucleotides, respectively [320]. Similarly, the GgxC and GgxA adducts were detected in leukocyte DNA hydrolysates at levels of 2.10 and 1.94 per 108 nucleotides [321, 322]. It is uncertain whether the cross-linked nucleobases detected after DNA digestion[316, 320–322] arise from intrastrand cross-links, interstrand cross-links, or both. In general, we are not aware of any biochemical studies (e.g. gel electrophoretic or mass spectrometric analyses of intact duplexes) characterizing glyoxal-derived ICLs in duplex DNA. Experiments in chicken DT-40 cells suggest that the Fanconi anemia ICL repair pathway does not play a significant role in mitigating the cytotoxicity of glyoxal [246].

Scheme 12.

Studies with nucleosides suggest that the glyoxal-dG adduct has the potential to generate ICLs.

Other endogenous 1,2-diketo compounds such as methylglyoxal and 4,5-dioxovaleric acid form DNA adducts and have the potential to generate cross-links [131, 323–329].

ICLs derived from nitric oxide and nitrous acid.

The endogenous signaling agent nitric oxide decomposes under cellular conditions to generate reactive oxygen and reactive nitrogen species including the nitrosating agent N₂O₃ [330, 331]. Nitrosating agents generate nucleobase diazonium ions that can be attacked by water to generate deaminated bases [332]. Alternatively, nucleobase diazonium ions can react with the exocyclic amino group of a purine residue on the opposing strand of DNA to generate an ICL (Scheme 13) [332–335]. The dG-dG ICL predominates over the dG-dA ICL (formally, these cross-links might be more accurately described as inosine-guanine, inosine-adenine, or guanine-nebularine ICLs). Treatment of duplex DNA with nitrous acid (500 mM, pH 4.15, 100 min) generates the dG-dG ICL (2% yield) selectively at 5’CG sequences[333] in DNA and the resulting connection of the two purine residues by a single nitrogen atom at these sites in duplex DNA is accommodated by extrusion of the two cytosine base pairing partners from the double helix [336]. Deamination products predominate over the dG-dG ICL when DNA is treated with nitric oxide under physiologically-relevant conditions (a 6:94 ratio of dG-dG to 2’-deoxyxanthosine) [337, 338]. Nitrosative DNA damage has been characterized in a mouse model of inflammation, but nitrosation-derived ICLs were not among the lesions analyzed in these in vivo experiments [331].

Scheme 13.

The endogenous signaling agent nitric oxide (NO) can decompose inside cells to yield nitrosating agents that generate nucleobase diazonium ions that, in turn, produce ICLs in duplex DNA.

An ICL derived from the 5-(2’-deoxyuridinyl)methyl radical.

The generation of reactive oxygen and nitrogen species is an unavoidable consequence of aerobic life [14, 15]. These include superoxide radical (O₂•−), hydrogen peroxide (H₂O₂), hydroxyl radical (HO•), nitric oxide (NO), and peroxynitrite (ONOO–) [14, 15, 330, 331]. Ionizing radiation is also unavoidable in the environment and generates oxygen radicals in cells [339]. Hydroxyl radicals derived from oxidative stress or ionizing radiation inflict a wide variety of damages on the nucleobases and deoxyribose backbone of DNA [16, 17, 20, 340, 341]. The 5-(2’-deoxyuridinyl)methyl radical (T•, Scheme 14) can arise by hydrogen atom abstraction from the methyl group of thymine residues or by one-electron oxidation of the nucleobase [342, 343]. Greenberg’s group showed that the 5-(2’-deoxyuridinyl)methyl radical can generate an ICL via reaction with the directly opposing adenine residue [344, 345]. The cross-linking reaction can take place in aerobic solution (unlike some radical processes that are inhibited by O₂) and involves rate-limiting rotation of the thymidine radical to the syn-glycosidic conformation that positions the radical for addition to the N1 position of the opposing adenine residue [346]. The initial C5-N1 ICL subsequently isomerizes to the C5-N⁶ ICL via a Dimroth rearrangement [346]. This ICL is chemically stable and would be expected to block replication and transcription [344– 346].

Scheme 14.

ICL formation involving reaction of the 5-(2’-deoxyuridinyl)methyl radical (T•) with the directly-opposing adenine residue in duplex DNA.

It is worth noting that the 5-(2’-deoxyuridinyl)methyl radical and the analogous radical derived from 5-methylcytosine can generate intrastrand cross-links via addition to the C8-position of an adjacent guanine residue [347, 348].

ICL derived from 2’-deoxyguanosine radical cation.

Under conditions of inflammation, reactions of superoxide radical (O₂•−), nitric oxide (NO•), and CO₂ can give rise to the carbonate radical anion (CO₃•−) via nitrosoperoxycarbonate [349]. In addition, carbonate radical anion is generated under Fenton reaction conditions in bicarbonate buffer [350]. In duplex DNA, one-electron oxidation of guanine residues by CO₃•− produces the guanine radical cation (G•⁺) [343]. Ionizing radiation and high intensity UVC light also can generate G•⁺ in DNA [343]. Nucleophilic attack of water on G•⁺ can lead to either 8-oxoguanine or formamidopyridine (Fapy) lesions [343]. Attack of the lysine amine residue on the G•⁺ intermediate generates DNA- protein cross-links [343, 351]. An analogous reaction of G•⁺ with an amine residue of a nucleobase on the opposing strand of duplex DNA has the potential to generate an ICL (Scheme 15). Indeed, exposure of duplex DNA to UVC light leads to products that migrate slowly in denaturing gels, consistent with the generation of an ICL(s) derived from the G•⁺ intermediate [352]. These slowly-migrating products display partial lability to piperidine workup [352]. It is proposed that this ICL is generated by nucleophilic attack of the exocyclic N⁴-amino group in the opposing cytosine residue on G•⁺ [343]; however, the formation, structure, and properties of these putative ICLs remains largely uncharacterized. It is worth noting that the G•⁺ intermediate also can give rise to intrastrand G-T cross-links [343, 353].

Scheme 15.

Generation of the guanine radical cation in duplex DNA may lead to ICLs.

A complex ICL generated by the oxidative DNA strand-cleavage product, 5’-dioxobutane.

Hydrogen atom abstraction from the C5’ position of deoxyribose sugars in DNA by agents such as hydroxyl radical can generate a strand break with a 5’-dioxobutane end group (DOB) [17, 354]. The DOB lesion exists as an equilibrating mixture of the ring-closed and the electrophilic dialdehyde forms. The DOB end group generates ICLs in sequences where an adenine residue is offset one nucleotide to the 3’-side of the lesion on the opposing strand (Scheme 16) [355]. The cross-linking reaction generates 30–40% yields of the ICL over the course of 15–20 h (k₁ = 2.6 × 10⁻⁵ s⁻¹, pH 7.2, 37 ˚C) [355]. The reaction is reversible and the isolated ICL dissociates with a half-life of approximately 11 h (pH 7.2, 37 ˚C). A slower reaction eliminates the 5’DOB sugar remnant from the duplex altogether (k₂ = 5.6 × 10⁻⁶ s⁻¹) [355]. Treatment with NaBH₃CN does not stabilize the ICL against dissociation [355], a feature that may hinder detection and quantitation of this lesion in cellular DNA. Nonetheless, the ICL was stable enough for mass spectrometric analyses results of which the ICL structure shown in Scheme 16 [355].

Scheme 16.

Oxidative strand cleavage induced by abstraction of a C5’-hydrogen atom can generate a strand break with a 5’-dioxobutane (5’DOB) terminus on one side of the break. Reaction of the 5’DOB group with an adenine residue on the opposing strand of duplex DNA can generate a complex lesion composed of an ICL adjacent to a strand break.

Studies employing a chemically-stable synthetic analog of the DOB-derived ICL[356] provided evidence that the bacterial NER complex UvrABC generates a deleterious double-strand break via incision on the 3’-side of the cross-linked adenine residue (Scheme 16) [357]. The results of these studies highlight the challenges presented to cellular repair systems by this type of complex (clustered) lesion consisting of an ICL adjacent to a single-strand break.

ICLs derived from the C4’-oxidized abasic site in duplex DNA.

The C4’-oxidized abasic site (C4AP) can arise from abstraction of the C4’-hydrogen atom from the sugar-phosphate backbone of DNA by agents such as hydroxyl radical [17]. This lesion is generated by exposure of DNA to ionizing radiation or antibiotics such as bleomycin [17]. Greenberg’s group provided rigorous characterization of two different types of ICL that can arise from this lesion [358, 359]. The C4AP lesion, similar to the 5’DOB lesion described above, exists as an equilibrium mixture of the ring-closed form and an electrophilic, ring-opened keto-aldehyde form (Scheme 17). The C4AP lesion can generate an ICL via reaction with an adenine residue located one base to the 3’side of the lesion on the opposing strand (Scheme 17) [358, 359]. The reaction generates 15– 40% yield of the ICL (depending on local sequence) over the course of 10–12 h (k = 5.5 × 10-⁵ s⁻¹, pH 7.2, 37 ˚C) [358, 359]. The isolated ICL is unstable, dissociating with a half-life of approximately 3 h [358, 359]. Treatment with NaBH₃CN does not stabilize the ICL [358, 359]. The proposed structure of the C4AP ICL is shown in Scheme 17. As an aside, Tris buffer has the potential to quench the reactivity of C4AP by formation of a tridentate adduct [360].

Scheme 17.

Oxidative strand cleavage arising from abstraction of the C4’-hydrogen atom can generate the C4’-oxidized abasic site (C4AP). The C4AP lesion can generate an ICL by reaction with an adenine residue on the opposing strand. Elimination of the 3’-phosphoryl group from the C4AP lesion produces the unsaturated C4AP lesion that can react with adenine or cytosine residues on the opposing strand to generate a complex lesion consisting of an ICL adjacent to a DNA strand break.

Elimination of the 3’-phosphoryl group from the C4AP lesion generates a strand break containing the α,β-unsaturated keto-aldehyde sugar remnant on the 3’-terminus of the break (unsatd-C4AP, Scheme 17). Interestingly, this strand cleavage reaction is catalyzed by an adenine residue directly opposing the C4AP lesion [358, 359]. The α,β-unsaturated sugar remnant derived from C4AP generates ICLs in 1–30% yields (depending upon local sequence) via reactions with cytosine or adenine residues on the opposing strand (unsatd-C4AP ICL, Scheme 17). The elimination reaction that generates unsatd-C4AP is rather slow and the nicked ICLs grow in over the course of 40–50 h (pH 7.2, 37 ˚C) [359]. The C4AP ICLs are the fast-forming (kinetic) products and the unsaturated-C4AP ICLs are the late-forming (thermodynamic) products.

The ICLs generated by the unsaturated C4AP are stable and the chemical structure of the C4AP-dA cross-linkage was rigorously determined by independent chemical synthesis combined with mass spectrometric analyses [359]. ICL formation by the unsaturated C4AP residue is inhibited by the presence of the dithiothreitol (1 mM) in the reaction mixture, presumably due to conjugate addition of the thiol to the α,β-unsaturated keto-aldehyde functional group. When this nicked ICL lesion is processed by the bacterial NER complex UvrABC, 15% of the incisions involved a dysfunctional repair pathway that generates a double-strand break [361].

A cross-link remnant arising from the reaction of 2’-deoxycytidine with the unsatd-C4AP residue has been detected in DNA hydrolysates of cultured human lymphocytes treated with either bleomycin or ionizing radiation [362]. Following exposure to these DNA-damaging agents, this lesion was removed from cellular DNA more slowly than the repair of typical oxidative nucleobase lesions, with a half-life of 10 h [362]. It remains uncertain whether the dC-unsaturated-C4AP cross-link remnant detected in the DNA hydrolysates arises from an interstrand cross-link, intrastrand cross-link, or both. Chemical synthesis of a duplex containing the unsaturated C4AP-dA ICL at a defined location has been reported and may prove useful in future DNA repair studies [363].

ICL generated by furan-modified adenine residues in DNA.

Furfural is generated by the hydroxylation of the C5’-position of deoxyribose residues in DNA [364]. Furfural has the potential to generate DNA adducts. Specifically, there is some evidence for the presence of N⁶-furfuryladenine residues in cellular DNA [365]. The formation of this lesion would involve condensation of furfural with the exocyclic amino group on adenine followed by reduction of the N⁶-furfuryl imine adduct. To our knowledge the reduction of N⁶-imine adduct in DNA is unprecedented.

Madder’s group provided gel electrophoretic evidence that oxidation of N⁶-furfuryladenine residue by N-bromosuccinimide in DNA can generate a low yield of an unstable ICL in a duplex where the furan lesion is opposed by cytosine residues [366]. The proposed structure for the resulting ICL is shown in Scheme 18. This reaction is analogous to the oxidation of furan to a 1,4-dialdehyde product that can modify DNA [367].

Scheme 18.

Possible mechanism for generation of an ICL from an N⁶-furfuryladenine residue in duplex DNA.

ICLs generated by ultraviolet light.

For many organisms, UV-light presents an unavoidable threat to their genetic material. Indeed, DNA damage associated with increased levels of UV-B flux resulting from depletion of the ozone layer has been proposed as the mechanism of a terrestrial mass extinction event on earth 359 million years ago [368]. UV-light generates a number of DNA lesions including, most famously, intrastrand DNA cross-links involving cyclobutane dimer formation [369]. Several early studies provided evidence that UV-irradiation also generates ICLs in duplex DNA [370–372]. The idea that ICLs can be generated by ultraviolet irradiation of DNA was further supported by more recent work [373, 374]. Together, the studies suggest that non-B-DNA conformations or increased duplex dynamics associated with local melting favor UV-induced ICL formation [370–374]. Otherwise, nothing is known regarding the yields of these ICLs, the sequence(s) in which they form, their chemical structure(s), or their biological significance.

ICLs generated by psoralen and other natural products.

The psoralens are a group of furocoumarin natural products derived from plants such as celery, limes, lemons, and cow parsley [375]. These compounds form DNA photoadducts including ICLs [375]. Psoralen-derived ICLs are not endogenous lesions but have been widely employed in studies of ICL repair. Psoralens bind weakly to DNA by intercalation[376, 377] and irradiation of psoralen-DNA complexes with 300–400 nm light induces reactions between the pi-bonds in the natural product and the 5,6-double bonds of thymine and cytosine residues (Scheme 19) [375, 378–381]. Psoralen ICLs are generated preferentially at 5’-TA sequences[382] via sequential photoinduced [2+2] cycloadditions of thymine residues with the 4’,5’-double bond in the furan ring of the psoralen, followed by reaction of the 3,4-double bond of the pyrone system in the natural product (Scheme 19) [375, 380]. Duplexes containing a psoralen ICL display some local unwinding at the cross-link site but, overall, retain a largely B-form structure [378–384].

Scheme 19.

Psoralens are plant-derived (not endogenous) natural products that generate ICLs when the DNA-bound complex is exposed to UV-irradiation.

There are a number of other plant-derived and microbial natural products that can induce ICLs. These include colibactin [385–387], azinomycin B [388–391], the mitomycins [392], and pyrrolizidine alkaloids such as FR900482 and dehydromonocrotaline [205, 393–395]. It is not clear that environmental cross-linking agents could represent a universal genomic threat of the type required to explain the evolution of ICL repair pathways.

Conclusions.

DNA repair pathways evolved to handle unavoidable damage to cellular DNA. In most cases, the endogenous lesions that drove the evolution of known DNA repair pathways can be inferred. On the other hand, the identity of the endogenous lesion(s) underlying the evolution of ICL repair pathways found in most life forms remains uncertain. In this review, we described more than a dozen processes by which ICLs might be unavoidably introduced into the DNA of living cells. Given the large number of processes by which endogenous ICLs plausibly can be generated, it seems likely that multiple reaction pathways contribute to the total load of ICLs in cellular DNA. The relative importance of different DNA cross-linking pathways may vary depending upon many factors including species, growth conditions, tissue-type, and pathophysiological conditions.Much work remains to define the formation, properties, and cellular occurrence of ICLs. In humans, endogenous ICLs may contribute to cancer, neurodegeneration, and aging [37–43]. Molecular understanding of the formation, consequences, and repair of endogenous ICLs – including how these features are affected by individual variations in the human genome – may enable preventive, predictive, personalized medical approaches that mitigate disease and increase healthspan.