Establishing Linkages Among DNA Damage, Mutagenesis, and Genetic Diseases

Abstract

DNA damage by chemicals, radiation, or oxidative stress leads to a mutational spectrum, which is complex because it is determined in part by lesion structure, the DNA sequence context of the lesion, lesion repair kinetics, and the type of cells in which the lesion is replicated. Accumulation of mutations may give rise to genetic diseases such as cancer and therefore understanding the process underlying mutagenesis is of immense importance to preserve human health. Chemical or physical agents that cause cancer often leave their mutational fingerprints, which can be used to back-calculate the molecular events that led to disease. To make a clear link between DNA lesion structure and the mutations a given lesion induces, the field of single-lesion mutagenesis was developed. In the last three decades this area of research has seen much growth in several directions, which we attempt to describe in this Perspective.

Graphical Abstract

1. INTRODUCTION

Thirty-five years ago, we were asked by Larry Marnett, founding Editor of this journal, to write a paper on what was then a novel technology that provided a linkage between the structure of a piece of damaged DNA and a mutation that the lesion in DNA might cause when processed in living cells. The technology was new and, indeed, we called our original paper in Chemical Research in Toxicology a Perspective rather than a Review because the field was only four years old and there was not much literature at the time on this fledgling field.¹ We explained the rationale behind the construction of site-specifically modified genomes and emphasized the need for chemical rigor in the preparation of substrates, because we were working with trace amounts of often chemically fragile DNA lesions that were to be introduced into the vastness of space inside cells. Moreover, we projected ahead to the challenges of DNA repair, which could easily erase the lesion a chemist had so carefully inserted into a genome for biological evaluation. That paper was more foretelling than we thought when we wrote it as evidenced by the observation that the field developed very much as predicted in that early document. Yet, at the time we wrote the original paper, we could not anticipate the discovery of an entire family of lesion bypass polymerases (the Y-family DNA polymerases), the impressive progress in mass spectrometry and high-throughput DNA sequencing (NextGen sequencing) as well as other advances that occurred to bring the field to its present state.^2–5 Today, more than three decades later, it is timely to look back at how this field was born and developed, although we admit at the outset that the field now is too big to review competently in a short paper. We refer the reader to the work of several colleagues in the field who have written comprehensively on this topic.^4,6–10 We shall try to hit the highlights and apologize in advance for spotlighting some of our own work, but it is the work we know best and follows the template of our original 1988 paper. Most importantly, however, as in our paper of 1988, we shall offer some advice to the current generation of chemists who may be starting their careers and looking for a path that matches their skills and ambitions.

2. MUTAGENESIS

2.1. Origins of the Field of Mutagenesis from a Chemical Perspective.

New fields are rarely the products of spontaneous generation, springing into existence from the void. As examples relevant to this Perspective, the DNA damage and mutagenesis field went through a series of distinct stages in its development, and it is useful to touch upon a few of these milestones. Elizabeth and James Miller sought to find the chemical roots of cancer and showed that many dangerous chemicals, especially chemicals that cause cancer, often are chemically inert and need to be converted by metabolism, usually oxidative metabolism, to electrophiles that go on to damage cellular macromolecules as an essential first step on the pathway toward malignancy (Figure 1).¹¹ It is interesting that early in their careers, the Millers focused on electrophile–protein interactions because at the time it was a widely held view that proteins, because of their central roles as the purveyors of function in the cell, were likely the primary targets of carcinogens.^12,13 Once the work of Watson, Crick, Avery, MacLeod, Fraenkel-Conrat, and Schramm became mainstream,^14–18 the Millers and the balance of the field pivoted to embrace nucleic acids, especially DNA,¹⁹ as the target of electrophilic reagents that had the potential to change the coding specificity of information-carrying molecules.^20,21 A large field evolved in which a cohort of chemists revealed the pathways by which hundreds of compounds became bonded to DNA in ways that posed threats to genomic informational integrity.^22–26 In some cases, the chemistry of DNA adduct formation led to the use of adducts in accessible tissues as biomarkers of prior exposure,^27,28 and sometimes the presence of a particular DNA adduct correlated with mutagenic or carcinogenic end points. While useful, this developing repository of chemical information needed more in the way of a biological foundation as a next step. Definitive studies needed to be done to connect the DNA damages identified by chemists with the biological end points of mutation and, possibly, cancer.

Figure 1.

Chemical damage to the genome and its role in genetic diseases. There are many ways that the genome can become damaged. Inflammation and endogenous metabolic processes generate a host of reactive oxygen (e.g., hydroxyl radical), nitrogen (e.g., nitric oxide), and halogen (e.g., HOCl) species that directly damage DNA. Some of these agents damage DNA indirectly by reacting with lipids in biological membranes, to yield electrophilic lipid-derived products, or by damaging the nucleotide pool to form pool-originated mutagens such as the deoxynucleoside triphosphate of 7,8-dihydro-8-oxoguanine (8-OxoG). DNA lesions, also referred to here as adducts, are depicted as L1, L2, etc. Ionizing and nonionizing radiations similarly form many lesions, as do a host of organic and inorganic agents that are either inherently able to attack DNA directly or do so via electrophiles produced by oxidations mediated by enzymes such as the cytochrome P450s. The collection of lesions formed can be repaired but, if repair fails, polymerases attempt to replicate them. In some cases, DNA synthesis is blocked, potentially leading to a lethal outcome. In some other cases, lesion bypass occurs but occurs at the expense of fidelity, leading to a collection of mutations. Mutations that activate protooncogenes, such a RAS, or inactivate tumor suppressor genes, such as TP53, are particularly dangerous as they may propel the cell along the pathway to malignant transformation.

2.2. Convergence of Chemical Mutagenesis with Classical Genetics.

The chemical and biochemical advances mentioned occurred against a background seemingly unrelated genetic studies aimed initially at mapping the positions of genes and mutations within genes in the genomes of T-even bacteriophages. Seymour Benzer developed tools that provided ever-higher resolution mapping of genetic landmarks in bacteriophage T4.²⁹ As part of that work, he, Ernst Freese, Francis Crick, and others used chemicals that could change the coding properties of DNA, including chemicals that affected base tautomerism, chemicals that forced base rotation about the deoxyglycosidic bond and chemicals with a planar structure that favored intercalation, to cause transition and transversion, and frameshift mutations, respectively; this work helped, among other things, establish the triplet nature of the genetic code.³⁰ In essence, their efforts provided early examples of mutational spectra, i.e., the arrangement of mutations along a genome. More sophisticated systems followed, including the use of the Lac system elegantly developed to elucidate mechanisms of gene regulation.³¹ Using the Lac genetic system, Jeffrey Miller showed that a host of different DNA damaging agents produced distinctive mutational spectra offering hope that examination of the sequence of mutations in damaged DNA might reflect fundamental features underlying the mutagenic properties of DNA damaging agents.^32–35 A mutational spectrum, as exemplified in the work of Miller and defined conventionally, is the linear presentation of variances from the canonical wild type sequence of the genome that occurs either spontaneously or as provoked by chemical or physical agents. Variances can be point mutations (single base substitutions), and small or large deletions, insertions and recombination events. The early studies of Jeffrey Miller, which spawned the modern field of mutational spectrometry, provided essential information in that they showed that different chemical agents or radiations produced distinctive mutational spectra that sometimes suggested the chemical nature of the DNA lesion or lesions that engendered the mutational patterns observed.³⁶

The discovery of mutational spectra of chemical and physical mutagenic agents was a good first step, but there was always uncertainty regarding the specific DNA lesion that gave rise to a specific mutation that appeared in the full spectrum. That limitation was addressed by the construction of defined-sequence oligonucleotides, synthesized by unambiguous routes, and rigorously characterized, that could be inserted into the genomes of viruses or plasmids (Figure 2). These vectors could be introduced into cells and replicated either extra- or intrachromosomally along with host DNA. Inside the cell, the specific DNA lesions would be exposed presumably in a normal way to the natural repair and replication systems of the host. Hosts of varying repair and replication statuses were used. Ultimately, the type, amount and genetic requirements of mutations were characterized, thus providing the missing link between chemical damage to the genome and the genetic events triggered by those types of damage.

Figure 2.

Technologies discussed in this paper. (A) Mutational spectra classically are produced by damaging the genome of a cell or vector and replicating the damaged piece of DNA in cells that have normal or disabled DNA repair status or normal or altered replicative status. Mutations are determined and plotted along the sequence of the piece of DNA as shown. (B) Often it is of value to test the hypothesis that a given mutation might have been caused by a specific DNA adduct. In that case, an oligodeoxynucleotide is synthesized with the adduct at a specific site. The oligonucleotide is spliced into the genome of a vector, usually a plasmid or viral genome, and replicated in living cells. Mutant progeny are sequenced. The type, amount and genetic requirements for mutagenesis are thereby determined, along with the potential genotoxicity of the lesion. Most of the lesions discussed in the text were evaluated by this site-specific mutagenesis strategy.

2.3. The Dichotomy of Genetic Mutation.

As indicated, mutation, a change in DNA sequence in an organism, may include base substitutions, single or multiple base deletions and insertions.³⁷ DNA sequence alterations in a genome are inevitable, and sometimes they may provide an advantage to the welfare of the species. It is a harsh reality, however, that even a single mutation may give rise to a devastating disease outcome. As one example, sickle cell anemia can result from a single mutation (A→T) in the β-globin gene, which results in valine substituted for glutamate at position 6 and leads to the phenotypic manifestations of this blood disease.³⁸ Mutations can be either spontaneous or induced by chemical damage to the genome. Spontaneous genetic changes are usually the consequence of rare polymerase errors during replication.^39,40 The mutation rate because of polymerase errors is dependent on both the rate of copy errors and the number of copies made in a given amount of time, and it appears that these factors are directed by the inherent biology of each organism or species. In many vertebrates, species with shorter generation times have faster rates of evolution, presumably because they copy their germline DNA more often and rapidly.⁴¹ The average number of mutations in DNA per generation also varies among species. The average mutation rate in humans is estimated to be ~2.5 × 10⁻⁸ mutations per base or 175 mutations per diploid genome per generation.⁴² Certain hotspots for mutations exist; for example, the rates of mutation at CpG dinucleotides are one order of magnitude higher than mutation rates at other sites, and single nucleotide substitutions are also at least 10 times more frequent than any other types of mutations, including deletions and insertions of multiple bases.

When the genome is under stress, the rates of mutation often very substantially increase. As species vary in their rates of evolution, mutagenic frequencies in these organisms also vary with the size of the organism, species population dynamics, lifestyle, and environmental milieu. With regard to this last point, evolution rates in tropical regions are double those in more temperate environments.⁴³ Evolution rates are important because a low rate of mutation may restrict adaptation of the species and, when rapid environmental change happens, the organism may die, preventing its genetic material from being passed along to future generations. By contrast, too high a rate of mutation could be particularly deleterious to the organism by preventing the organism to generate functional gene products. Successful organisms strike a balance in mutation rates, producing rates that generate a desirable amount of diversity but not such a high rate so that the organism exceeds its error catastrophe limit.^44–46

3. DNA ADDUCTS/LESIONS AND DNA POLYMERASES: CHEMICAL AND BIOLOGICAL PURVEYORS OF GENETIC CHANGE

As shown in Figure 1, DNA is continuously damaged by exogenous agents, e.g., chemicals or radiation (UV, ionizing radiation), and endogenous agents, e.g., reactive oxygen species formed during metabolic processes or by an activated immune system.^21,47–50 The chemical structures of many DNA damages were elucidated mainly in the latter part of the twentieth century, leading to speculation as to how certain DNA lesions might provoke genetic changes. Alteration of base pairing modes, three-dimensional architecture, as well as binary and ternary polymerase complexes during replication of these DNA lesions have been studied by NMR and crystallography.^51–57 The respective mutagenic outcomes resulting from the presence of these DNA damages has been investigated in vitro using purified DNApolymerases and in prokaryotic and eukaryotic cells.^7,58–60 Great advances have been made in these areas.

The central players of replication are the DNA polymerases, although other enzymes, including primase, helicase, ligase, and topoisomerase, are necessary for complete cellular replication. Inaccurate bypass of a DNA lesion by a polymerase is the most frequent cause of mutations in cells that have experienced damage. Based on sequence homology, DNA polymerases have been divided into seven families (A, B, C, D, X, Y, and RT).^2,61,62 In Escherichia coli, the DNA polymerase III holoenzyme is a complex, multisubunit enzyme of the C family, which is responsible for the synthesis of most of the bacterium’s chromosome, whereas the most abundant polymerase, DNA polymerase I of the A family, is involved in DNA repair and other genome maintenance operations.⁶² A second repair polymerase, pol II, belongs to the B family. Translesion synthesis (TLS) polymerases pol IV and pol V are members of the Y family and involved in DNA lesion bypass.⁶³ In contrast to prokaryotes, the B-family enzymes are important in eukaryotic cells, as pol ε and pol δ of this family perform most nuclear DNA replication, and initiation and priming is carried out by pol α.^64–66 The high fidelity of DNA replication is maintained by polymerase selectivity for complementary nucleotide incorporation, exonucleolytic proofreading of mismatches, and removal of left-over mismatches via DNA mismatch repair.⁶⁷

Additional DNA polymerases with specific functions have been identified. For example, DNA polymerase μ is an X family member that participates in repairing DNA double strand breaks by nonhomologous end joining.^68,69 Likewise, DNA polymerase θ, a member of the A family, mediates a microhomology-mediated, error-prone, double strand break repair pathway.⁷⁰ DNA polymerase ν, of the A family, plays an active role in homology repair during cellular responses to cross-links.⁷¹ DNA lesion bypass is carried out primarily by the Y-family TLS DNA polymerases, though X- and B-family polymerases also frequently play a role.² Similar to the replicative polymerases, these lesion-bypass polymerases possess right-handed topology with the active site located in the “palm” domain; the active site in the palm domain of the bypass polymerases is larger than with replicative enzymes so that it can accommodate DNA lesions. However, whereas the finger and thumb domains of the replicative polymerases ensure correct pairing with the incoming nucleotide, in the bypass polymerases they are shorter and make little interaction with the template and the incoming dNTP, thereby reducing the polymerase’s ability to discriminate the accuracy of nucleotide insertion. The Y-family polymerases are assisted by the little finger domain of the polymerases to stabilize on DNA. Additionally, the Y-family polymerases and pol ζ of the B-family lack a 3′–5′ proofreading function, making them error-prone but allowing them to carry out TLS.

What are the fundamental requirements of DNA replication by the DNA polymerases? The discovery of the structure of duplex DNA by Watson and Crick in 1954 suggested the role of complementary purine:pyrimidine pairing during replication and further suggested that the pairing was mediated by a hydrogen bonding pattern that allowed favorable binding of guanine with cytosine and adenine with thymine. While in general hydrogen bonding in a Watson–Crick sense is important, other factors including shape and fit in the helix were more recently appreciated.^72–74 As one example, to probe the relative importance of Watson–Crick hydrogen bonding, a difluorotoluene isostere (dF) of thymine was incorporated in DNA (Figure 3). Despite its inability to form hydrogens bonds, which reduces the thermal melting temperature in a DNA duplex, adenine is incorporated with high efficiency and high selectivity opposite dF by the 3′–5′ exofree Klenow fragment of DNA polymerase I.⁷⁵ Similarly, dF triphosphate was incorporated opposite adenine with high selectively and high efficiency.⁷⁵ Subsequent studies showed that this nonpolar isostere is also an efficient substrate for the B-family enzyme RB69 and reverse transcriptases. However, dF is a poor substrate for pol α (B family), pol β (X family), and the Y-family enzymes Dpo4, pol η, and pol κ^.76–79 A rationale suggested for the bypass polymerases’ inability to use dF as a substrate is that hydrogen bonding is more important for enzymes with larger active sites to properly align the nucleotide for bonding. Using dF and the adenine isostere dQ (Figure 3), it was shown that these unnatural bases can be bypassed with moderate to high efficiency in E. coli cells and with very high efficiency when SOS (a typically error-prone survival response triggered by some forms of DNA damage) is induced.⁸⁰ The fidelity of nucleotide incorporation opposite dF is high, as 95% of the time it is replaced by dT in the daughter strand, whereas dQ is replaced 80% and 20% of the time by dA and dG, respectively. Taken together, these results indicate that accurate DNA replication can occur without Watson–Crick hydrogen bonds. Thus, recognition of DNA base shape might be the critical factor to maintain fidelity of base incorporation during replication, though hydrogen bonding of the base pairs is deemed essential for some DNA polymerases to replicate effectively. Additional investigation of a series where the fluorine atoms at the 2 and 4 positions of dF were replaced with either hydrogens or another halogen showed that the size of these isosteres is very important, as the dF analog with hydrogen, bromine, or iodine exhibit poor efficiency of nucleotide incorporation by the Klenow fragment and T7 DNA polymerase, whereas the dichloro derivative, whose shape is nearly identical to dT, is nearly as good as dT.⁷⁴ Replication of these isosteres in E. coli gave similar results.⁸¹ In summary, many DNA polymerases can effectively insert and extend base pairs without any hydrogen bonds, suggesting the importance of shape in these base pairings. Steric effects are also important for the recognition of these non-natural bases. Importantly, however, Watson–Crick hydrogen bonding is vital to achieve a very high activity by the DNA polymerases, and it is a requirement for the Y-family polymerases.

Figure 3.

Structures of dF and dQ, which lack Watson—Crick hydrogen-bonding ability but possess shapes similar to dT and dA, respectively.

There is evidence of the replicative bypass of DNA lesions where no hydrogen bonds can be formed. Most notable of such lesions is the abasic site (apurinic/apyrimidinic or AP-site) (Figure 4) that cannot form any hydrogen bond; in duplex DNA, its partner base is stacked with the neighboring base pairs.⁸² Though AP-sites stall replication, they are bypassed by the cellular replication apparatus,⁸³ which will be discussed later. Certain bulky DNA adducts also cannot form any hydrogen bond in a DNA duplex. For example, NMR and thermal melting of the C8-2′-deoxyguanosine adduct formed by 1-nitropyrene (C8-AP-dG) (Figure 4) show that it is unable to form any hydrogen bond, as the aminopyrene moiety of C8-AP-dG resides in an intercalated base-displaced orientation opposite either C or A in a DNA duplex.^84–86 Yet, the error-prone bypass of this adduct occurs in both E. coli and mammalian cells incorporating either C or A.^87–89 We note the caveat, however, that hydrogen bonds may be forming in the active site of the DNA polymerase as it bypasses the DNA adduct.

Figure 4.

Structures of some of the DNA lesions discussed in this paper.

4. ALKYLATION OF DNA AS A PRELUDE TO GENETIC CHANGE

DNA is a poly nucleophile and, therefore, it reacts with many electrophilic species, including alkylating agents, free radicals, epoxides, and carbocations. DNA alkylation generates both stable and unstable adducts. Alkylation at N1, N², and O⁶ of G, N⁶ of A, N⁴ of C, and N1 and O⁴ of T generate chemically stable adducts, whereas alkylation at the N7 and N3 positions of G and A, N1 of A, and N3 of C destabilizes the bases, facilitating base deglycosylation or base ring-opening.⁹⁰ The chemically preferred alkylation sites in the DNA bases by low molecular weight alkylating agents are N7 of G followed by N3 of A. Nevertheless, from a functional point of view, the minor alkylation products O⁶-alkylG and O⁴-alkylT are the most mutagenic lesions of DNA alkylating agents.²¹ One descriptor of the reactivity of direct-acting alkylating agents is the Swain–Scott substrate coefficient s, which quantifies the selectivity of alkylation in reactions with different nucleophiles.⁹¹ Swain and Scott developed a two-parameter equation to correlate the relative rates of reactions of various nucleophilic reagents with a series of alkylating agents in water: log k_a/k₀ = sn, where k_a represents the rate constant for reaction with the nucleophile relative to the rate constant k₀ for reaction with water, n is a reactivity constant (“nucleophilic strength”) of the nucleophilic center to be alkylated, and s is a constant related the nature of the alkylating agent.⁹¹ The value of s is measured experimentally on the basis of the relative reaction rates of methyl bromide with an assigned s value of 1.0. High s values typically indicate the more S_N2 type alkylating agents. The utility of the Swain-Scott equation can be exemplified as follows. The s value for N-ethyl-N-nitrosourea (ENU) and methylmethanesulfonate (MMS) are 0.26 and 0.83, respectively, suggesting that ENU mainly alkylates by an S_N1 mechanism; with its low s value, ENU can alkylate O⁶ of G and O⁴ of T better than MMS can. MMS, with its high s value, alkylates more by an S_N2 mechanism and generates relatively more N7G and N3A adducts. Indeed, the O⁶-/N7-alkylG ratio was determined to be 0.6–0.7 and 0.004 for ENU and MMS, respectively, indicating more than a 150-fold difference between the two alkylating agents.⁹² The general idea of S_N1 versus S_N2 alkylating agents was attractive, as it seemed to predict the mutagenicity of these molecules, even though it was applied to a small subset of alkylating agents. Loechler suggested that rather than S_N1 and S_N2 type alkylating agents, it would be more appropriate to name them “high oxyphilic” and “low oxyphilic”, respectively.⁹³ Importantly, however, the ratio of alkylation at different sites in DNA also depends on the size of the alkyl group and how electrophilic the reaction site is. When the O⁶G to N7G ratio is normalized to 1 for the methyldiazonium ion, the relative ratio of alkylation is 1, 8, and 79 by methyldiazonium ion, ethyldiazonium ion, and isopropyl cation, respectively, suggesting that the more reactive electrophile is essentially more “oxyphilic”.⁹⁴

Hydrophobic chemical carcinogens are usually metabolized by the cytochrome P450 enzymes to epoxides, which also can act as alkylating agents to react with DNA.²¹ Examples of such derivatives are 3,4-epoxy-1-butene and 1,2,3,4-diepoxybutane generated from the human carcinogen, 1,3-butadiene.^95,96 Each of these epoxides has the ability to enter cell nuclei and react with genomic DNA to form DNA adducts, most frequently with N7 of G.⁹⁷ Alkylation of DNA by epoxy derivatives of polycyclic and other bulky alkylating agents is more complex and highly dependent on the structure and conformation of the agent as well as DNA sequence context, because these larger, mostly planar, alkylating agents often form an intercalative complex prior to covalent binding. Although a number of polycyclic aromatic hydrocarbons (PAHs) are found to be carcinogenic, Cook, Kennaway, and co-workers’ groundbreaking isolation and identification of benzo[a]pyrene (B[a]P) as a potent carcinogen present in soot and coal tar in the 1930s fueled numerous investigations on it in the 1960s and beyond.^98,99 Particularly, since the discovery that sterically hindered “bay region” epoxides are the more mutagenic and carcinogenic metabolites of the PAHs, research on B[a]P intensified.^24,100,101 After it was established that 7,8-dihydroxy-9,10-epoxy-7,8,9,10-tetrahydrobenzo[a]pyrenes are the ultimate mutagenic and carcinogenic forms of the carcinogen B[a]P,¹⁰² in the late 1970s there were intense efforts by Jerina, Harvey, Weinstein, and others^25,103–111 to synthesize the various epoxy derivatives and identify the DNA adducts formed by them. Mechanistically, it was found that the metabolically activated PAHs such as benzo[a]pyrene diol epoxides (BPDEs) first noncovalently bind to DNA prior to covalent adduct formation (alternatively, the epoxides will hydrolyze to tetraols).^112–115 Two different types of binding geometry were discovered by Geacintov: site I in which the BPDE molecule is intercalated between adjacent base pairs of DNA and site II corresponding to an external type of binding site of BPDE on DNA^{112,113,115,116} This DNA–BPDE complex undergoes rate-determining protonation to yield an intercalated triol carbocation, more than 90% of which is hydrolyzed to yield tetraols. The minor but biologically more important reaction of the carbocation involves covalent bonding to DNA, predominantly at the exocyclic amino group (N²) of G. Although both the syn and anti diastereomers of BPDE form intercalative physical complexes with DNA, following covalent bond formation with N² of G, the pyrene ring system is reoriented in the case of (±)-anti-BPDE to an external conformation with the aromatic ring residing in the minor groove with its long axis inclined at an ~35° angle to the DNA helix axis. By contrast, the syn diastereomer remains in the intercalated conformation. With self-complementary duplex oligonucleotides, it was shown that the formation of (+)-anti-BPDE-N²-dG adducts is enhanced at guanines with adjacent C and G residues.¹¹⁷ As in the case of B[a]P, noncovalent binding, and specifically intercalation, is often a prerequisite for covalent DNA adduct formation by many other carcinogens, including other PAHs, aflatoxin B₁ (AFB₁), nitroaromatic compounds, and aromatic amines.^118–121 Alkylation may also occur with other types of metabolically activated chemicals. Estrogen, for example, may be metabolically oxidized to catechol estrogen quinones, which form DNA adducts.¹²² Likewise, α-hydroxylation followed by sulfation of tamoxifen leads to DNA adduct formation.¹²³ Although the focus of DNA adduct research has shifted toward investigation of three-dimensional architecture, base pairing modes as well as their in vitro and in vivo effects, the reactivity and details of DNA alkylation by the mutagen/carcinogen is also important for a firm understanding of their biological effects.

5. SITE-SPECIFIC LESION-DERIVED MUTAGENESIS

5.1. Examples of Site-Specific DNA Lesion-Derived Mutagenesis.

Structural studies on DNA lesions, both small alkyl adducts and larger lesions such as those of PAHs, led logically to strategies to link the chemical structures of damaged DNA to biological end points, the foremost of which is mutation. As indicated, DNA lesions can involve the addition of only a few atoms or they can be large chemical moieties; moreover, chemistry on the primary lesion can result in internal rearrangement of structural elements and/or fragmentation of the DNA base, DNA–DNA inter- or intrastrand cross-links (including tandem DNA lesions), DNA strand breaks, and clustered DNA lesions. Most DNA lesions can be subject to enzymatic or nonenzymatic repair and, therefore, the efficiency of lesion removal plays a key role in its eventual genotoxicity and mutagenicity. A prime example of the interplay of replication and repair as determinants of the biological effect of a DNA lesion is provided in Figure 5. O⁶-Methylguanine (O⁶-mG), a minor DNA lesion formed by methylating agents, is efficiently repaired by O⁶-alkyl-DNA alkyltransferase. While quantitatively a minor adduct, it is highly mutagenic and any unrepaired O⁶mG will induce G:C→A:T transitions in both prokaryotic and eukaryotic cells.^124,125

Figure 5.

(A) O⁶-mG, a minor lesion formed by methylating agents, is efficiently repaired by DNA alkyltransferases (depicted here as a Pacman symbol), but any unrepaired O⁶-mG will give rise to G:C→A:T mutations. The lollipop symbol designates a DNA methylating agent, which forms O⁶-mG. As shown, the repair protein transfers the methyl group to itself, rendering the repair protein inactive. If the adduct evades repair, it can be replicated in an error-free manner with C inserted as the opposing base (panel B, left) or in an error-prone manner with T inserted opposite the lesion (panel B, right).The pairing of O⁶-mG with T results, after the next round of replication, in a G:C→A:T mutation. The base pairing modes of O⁶-mG with C and T were determined from NMR and crystal structure studies (panel B).^126,127

O⁶-mG is a mispairing lesion, and DNA polymerases efficiently incorporate T opposite it, which results in the observed G→A transition. Its ability to mispair has been investigated by NMR and crystal structure analyses, which show that O⁶-mG can pair with either C or T (Figure 5), though at low pH a protonated C can maintain Watson–Crick base pairing (not shown). Although structural studies indicate that O⁶-mG can pair with C or T in a test tube, evidence from in vivo replication shows that mispairing with T is approximately 100%.¹²⁴

Similar mispairing lesions include O⁴-methylthymine (O⁴-mT), 2-aminopurine, and uracil (U), the latter resulting from deamination of cytosine (C). These altered bases have the ability to hydrogen bond with a wrong or noncomplementary base, leading to a mutation. In contrast, the mutagenic mechanisms of bulky DNA adducts, DNA–DNA cross-links, and fragmented DNA bases are more complex, in terms of their ability to induce various types of mutations as well as in the ways that repair proteins interact with them.

As, an example, the aflatoxin B₁-derived AFB₁-Fapy-dG adduct (Figure 4) induces 14–32% G→T transversions in E. coli,¹²⁸ depending on the availability of bypass polymerases, which are needed to traverse the bulky lesion. This adduct is even more mutagenic in simian kidney cells, inducing 86% G→T and 8% G→A.¹²⁹ It is suggested that lesion bypass polymerases such as pol V in E. coli and pol ζ in the primate cells are likely responsible for the observed mutations.

5.2. DNA Base Sequence and Local Chemical Factors May Favor Reaction in Some Local Contexts and Disallow Reaction in Others Leading to Complex Mutational Spectra.

DNA damage is nonrandom, as the interaction of an electrophile with a local base context varies because of both the primary and secondary structures of the helix. Certain DNA sequences are known to be hotspots of DNA damage, and the distribution of hotspots varies with the structure and reactivity of the DNA damaging agent.^130–133 Making things even more complex, biochemical repair of specific DNA lesions is also nonrandom.¹²⁴ The erroneous replication of DNA following nonrandom DNA damage and their nonrandom repair give rise to a complex, visibly rugged mutational spectrum, which can be unique to each mutagen and the DNA sequence of the template. Figure 6 shows a schematic representation of mutagen induced lesion formation, repair of the lesion, and generation of a unique mutational spectrum. Several assays have been helpful in determining the mutational spectra of a variety of mutagenic agents (see ref 134 for a review).

Figure 6.

Simplified scheme showing a pattern of DNA damage from an agent that, either directly or upon metabolic activation, reacts with various positions in the genome. The blue bars represent the level of damage at the bases, and the red bars specify the sites where mutations occurred. The height of the red bars represents the mutational frequency at the specified base. The asterisks (at positions 1, 3, and 5) indicate hotspots of lesion formation in panel A. Repair of these lesions is also nonrandom, and lesions at certain sites (e.g., 3, 4, and 5) are repaired more efficiently than at other sites as shown in panel C. Mutagenic consequences are nonrandom as well, as, for example, mutations are detected only in positions 2, 4, 5, and 7, as shown in panel B when replication occurrs in the absence of repair. If replication takes place after DNA repair, a much lower level of mutations is detected (shown in panel D). In many cases, replication of the damaged DNA may take place partly before repair and partly after repair, and the relative proportions of these two competitive events depend on a variety of factors. It should also be noted that most mutagens form multiple types of lesions, which further diversifies the mutational spectrum, and both the efficiency of repair and mutagenesis may be different for the leading strand and the lagging strand of the DNA helix.

DNA sequence specificity of lesion formation depends on a number of factors. As, for example, 8-OxoG formation is governed by the susceptibility of guanines to be oxidized at a specific sequence, which is determined by its ionization potential. Guanines with lower ionization potential are more prone to be oxidized, and theoretical studies have established that the ionization potential of guanines is affected by its sequence context. The 5′ G in a GG sequence has the lowest ionization potential, whereas guanines in the NGT and NGC sequences have much higher ionization potentials (where N represents any DNA base).^135,136 Nonrandom 8-OxoG formation on a template appears to be consistent with this specificity, although DNA secondary structure is another factor that plays a critical role. The importance of secondary structure was shown in a study revealing that 8-OxoG occurrence at G-quadruplexes is lower than at other sites.¹³⁷ However, only certain guanines are prone to oxidation in the G-quadruplexes, as determined in the coding strand of the promoter of the PCNA gene, which forms a parallel-stranded G-quadruplex.¹³⁸ The latter is biologically important, because the guanines oxidized to 8-OxoG in a loop position of the PCNA G-quadruplex can turn on transcription and gives nearly 4-fold greater expression than a sequence lacking the loop. Nonuniform DNA adduction by chemical agents was also illustrated when the diaqua form of the anticancer drug cisplatin (Figure 4) is allowed to react with G-rich G-quadruplexes. The latter is platinated twice as fast as two adjacent GGs within a duplex, which are themselves platinated three times as fast as a GG within a single strand.¹³⁹

A number of approaches has been used in an attempt to characterize the adduct-formation spectrum of DNA damaging agents. Detection and quantification of DNA lesions have been carried out by ³²P-postlabeling, immunochemical methods, and mass spectrometry, which typically provide the types and levels of the DNA lesions but are less useful at revealing high-resolution information on sequence-specific bonding. One early approach that successfully generated the DNA adduct/lesion spectrum is ligation-mediated PCR (LM-PCR).¹³⁰ In this approach, cells are treated with a DNA-reactive chemical (such as benzo[a]pyrene diol epoxide) and the DNA is isolated. The adduct-containing DNA is then incubated with the UvrABC repair complex from E. coli, which makes precise dual incisions 5′ and 3′ to the adduct-containing nucleotide. These precisely known break positions are visualized by LM-PCR in which specific oligonucleotide primers (e.g., TP53-specific) are used. Strong signals for the lesion hotspots are visualized and compared with mutational hotspots. The LM-PCR method offers excellent mapping of certain types of damages, giving insight as to the location of bulky DNA lesions that are mappable using UvrABC. Using this approach, the BPDE adduct distribution in the TP53 gene in BPDE-treated HeLa and bronchial epithelial cells was mapped. Preferential adduct formation in the TP53 gene was observed in guanines in codon 157, 248, and 273, which are also mutational hotspots observed in human cancer.¹³⁰ However, this approach is not specific for all DNA damages, as it relies on the specificity of the repair protein that excises the DNA fragment containing the DNA adduct. Lesion-specific sequencing methods were subsequently developed that use antibodies to bind to lesion sites followed by NextGen sequencing (NGS), high throughput microarray or other approaches to map the adduct spectrum. A review of sequence-specific mapping of DNA damages has recently been published by Sturla and co-workers,¹⁴⁰ so we shall not discuss it in detail here. Even so, given the importance of this field, it is appropriate to mention a couple of recent DNA lesion sequencing approaches. Burrows and co-workers determined 8-OxoG locations in the mouse genome using a chemical labeling method called “OG-Seq” to detect the lesion sites by NGS at ~0.15 kb resolution.¹⁴¹ The genome is first fragmented to smaller pieces of ~150 mers, which are subjected to enrichment of the 8-OxoG containing strands. 8-OxoG is selectively oxidized to an electrophilic intermediate that can be trapped with a primary-amine to form a stable amine-conjugate. Magnetic beads are used to extract the DNA and, after removal of the complementary strands by base-treatment, the single-stranded DNA is sequenced by NGS. Using this approach, ~10 000 regions of OG (8-OxoG) enrichment in wild type mouse embryonic fibroblasts and ~18 000 regions when Ogg1 was knocked out were mapped. In another approach, Sturla and her team developed a strategy that combines DNA lesion removal by repair proteins, incorporation of a novel alkynylated nucleoside with a DNA polymerase and, subsequently, the damage sites are labeled with a code sequence through click chemistry.¹⁴² Thus, the DNA lesion sites are replaced by a synthetic oligonucleotide with a tag (for affinity enrichment), an adaptor (for PCR amplification) and code sequence (for marking damage locations during sequencing). Using this approach, they were able to provide a high-resolution map of 8-OxoG distribution in the yeast genome.

Although less precise than the previously mentioned mapping methods, several experimental systems have been developed to study the positional relationship of DNA lesions with mutagenesis at a lower resolution vantage point. Such mutagenesis studies typically analyze mutations in a single gene, such as HPRT, lacZ, cII, TP53, as identified in human or animal tumors or from cells following treatment with a carcinogen. Because only a very small number of mutations in a particular gene is usually detected, mutational patterns are catalogued by combining data collected from many samples. Also, even if a DNA lesion is detected at a specific site, there is no guarantee that a mutation will be detected because many mutations are silent. That point notwithstanding, the most commonly mutated gene in human cancers is the tumor suppressor gene TP53,^143–146 and a large number of TP53 mutations identified in human tumors has been catalogued in the IARC TP53 mutation database. This catalog allowed identification of specific mutation patterns in human cancers, often as a result of exposure to specific carcinogens. UV-induced C→T and CC→TT mutations in squamous carcinomas and G:C→T:A mutations in tobacco smoking associated lung cancer are examples of specific carcinogen-derived mutations.¹⁴⁷ In the following section we shall describe more modern methods of cataloging mutations in diseased tissues, and sometimes mutagen-treated normal sequences of animal tissues and of cultured mammalian cells.^148,149

5.3. A DNA Lesion May Exhibit Different Types of Mutations in Different Organisms and Different Types of Cells.

Unlike O⁶-mG, which almost exclusively induces G→in all cell types, there are lesions that can give different mutagenic outcomes in different cells. It was established more than 50 years ago that the mutagenicity of replication blocking bulky DNA lesions and the abasic site (also known as the apurinic/apyrimidinic or AP-site) is dependent on the induction of the SOS functions in E. coli. Much later it was shown that the SOS response involves the induction of a promiscuous bypass polymerase pol V (UmuD′₂C complex). In E. coli cells, AP-site bypass is largely SOS-dependent and dAMP is most commonly inserted opposite it, which led to the so-called “A-rule”, which posits that blocked or hindered polymerases insert adenine nucleotides as a default option.^150,151 Site-specific mutagenic studies on AP-sites in E. coli confirmed that dAMP insertion occurs preferentially with SOS induction.¹⁵² While dAMP insertion is the favored event, a targeted single-base deletion also occurs in substantial frequency in E. coli, particularly when a chemically stable analog of the AP-site (tetrahydrofuran) is used.¹⁵³ In the eukaryote Saccharomyces cerevisiae, on the other hand, dCMP is inserted (i.e., there is a “C-rule”) when the AP-site is located in the single-stranded gap of a gapped duplex plasmid,¹⁵⁴ but in duplex DNA dAMP is inserted preferentially opposite the AP-site in both the leading and lagging strands.¹⁵⁵ In simian kidney cells, several investigations show a lack of specificity in nucleotide insertion opposite AP-sites,¹⁵⁶ whereas in human cells in single-stranded DNA, the majority of the bypass involves insertion of dAMP opposite AP-sites (Figure 7).^153,157

Figure 7.

Nucleotide incorporation patterns opposite an AP-site differs in different types of cells or organisms.

Attempts have been made to rationalize the predominant dAMP incorporation opposite a noninstructional lesion such as the AP-site. Thermodynamic investigations show that both A and T are enthalpically unfavorable opposite an AP-site, but an A is more stable than a T, the latter being the next preferred base incorporated opposite AP-site in human cells.¹⁵⁸ Goodman and co-workers showed that the specificity of nucleotide insertion is 6–11 times greater for A over G and about 20–50 times greater for A over C and T by Drosophila DNA polymerase α. ¹⁵⁹ The base stacking partners adjacent to the AP-site have up to a 4-fold effect on nucleotide insertion specificity. Later, these workers also showed that Pol V of E. coli incorporates dAMP efficiently opposite an AP-site, whereas pol III and pol IV cannot bypass it. ⁶³ By contrast, pol IV can incorporate a nucleotide opposite the AP-site, and it is possible that another polymerase carries out the extension step. Both yeast and human pol η can efficiently bypass AP-sites preferentially inserting purine nucleotides.^155,160 The yeast pol η also induces a high frequency of −1 and −2 frameshifts.¹⁶¹ Human pol κ can also bypass AP-sites with reduced efficiency, but accessory proteins PCNA, RFC, and RPA increase efficiency by more than an order of magnitude.¹⁶² Human pol ι also bypasses AP-sites inserting either dGMP or dTMP with 10-fold reduced efficiency.¹⁶³ Pol δ, in the presence of pol ζ, can bypass AP-sites, suggesting that pol δ probably inserts dAMP opposite the AP-site, which is extended by pol ζ,¹⁶⁴ but arguments against this model have also been presented.¹⁶⁵ Thermodynamic studies could not rationalize the preferential dAMP insertion opposite AP-site followed by its successful extension by several DNA polymerases,¹⁶⁶ but exclusion of water molecules from the active sites of a polymerase–DNA complex may differentially increase the free energy changes (ΔG°) in favor of adenine incorporation.¹⁶⁷ Adenine incorporation also occurs opposite another seemingly noninstructional lesion, urea, both in vitro by the exofree Klenow fragment of DNA polymerase I and in E. coli cells.^168,169 But, interestingly, both dAMP and dGMP were incorporated at approximately equal efficiency in SOS-induced E. coli,¹⁶⁹ suggesting a different pattern of nucleotide incorporation for the urea residue compared to the AP-site.

Another example of different pattern of mutagenesis in different types of cells is as follows. In SOS-induced wild type E. coli, O²-Me-dT and O²-POB-dT (Figure 4) formed by tobacco-specific nitrosamines induce 21% and 56% mutations, respectively.¹⁷⁰ For O²-POB-dT, the major type of mutation was T→G followed by T→A, whereas for O²-Me-dT, T→G and T→A occur in equal frequency. Wang and co-workers later showed that both pol IV and pol V are necessary for dCMP misincorporation opposite O²-alkyldT, whereas only Pol V is essential for the T → A transversions.¹⁷¹ O²-Me-dT and O²-POB-dT are also highly mutagenic in HEK293T cells.¹⁷² While the bulkier O²-POB-dT was a stronger replication block compared to O²-Me-dT, the latter was more mutagenic. The major type of mutations induced by these lesions was T→A, but low frequencies of T→G, T→C, and semitargeted mutations also were detected. Using a series of O²-alkyl-dT lesions, it was later shown that T→A mutations in human cells require both pol η and pol ζ.¹⁷³

As already indicated, bulky DNA adducts typically show more diversity in the types of mutations they induce, though in most cases one or two major types predominate. Loechler has shown that [+trans anti]-B[a]P-N²-dG induces 95% G→T mutations in a 5′-TGC-3′ sequence context and ~80% G→A mutations in a 5′-CGT-3′ sequence context.^174,175 The [+trans anti]-B[a]P-N²-dG is shown to induce principally G→A mutations (>90%) either without or with SOS induction in 5ȃ-AGA-3′ sequence context.¹⁷⁶

The primary reason for different mutational outcomes in different species and different conditions in the cell is attributable to the specifics of the DNA polymerase that bypasses the lesion. Structural and conformational alterations, which may occur in different DNA sequence contexts, are other reasons for variances in mutational outcomes.

5.4. DNA Lesion-Derived Mutagenesis Can Be Untargeted.

In most cases, DNA lesion-derived mutagenesis is targeted, which means that in the progeny the adduct or modified base is either replaced by another base or it triggers a more severe alteration, such as a deletion at the lesion site. However, some lesions, in addition to targeted mutations, have been reported to induce either off-target or semitargeted mutations. The term off-targeted mutation implies mutation at any base position other than the lesion site, whereas the semitargeted mutation can be defined as mutations in one or more sites near the lesion (e.g., within 10–12 bases from the lesion).

Moschel reported that O⁶-ethylguanine and O⁶-benzylguanine induce semitargeted as well as targeted mutations, in contrast to 0⁶-mG, which induces only targeted mutations in Rat4 embryo fibroblasts.¹⁷⁷ At the first and second position of H-ras codon 12 (GGA), semitargeted mutations of G→A at the base 3′ to a position 1 adduct or 5′ to a position 2 adduct were detected, in addition to targeted mutations. It was suggested that a bulky O⁶-substituent on guanine would cause significant backbone distortion irrespective of the base opposite it, which may be the cause of semitargeted mutations.

2-Nitropropane, a widely used industrial solvent and potent hepatocarcinogen in rats, forms 8-OxodG and 8-aminoguanine (8-NH₂G) (Figure 4). The latter lesion, in a CXC sequence, induces primarily semitargeted mutations in E. coli upon induction of SOS functions.¹⁷⁸ It is noteworthy that in COS-7 cells only targeted mutations are detected in a TXG sequence, (where X represents 8-NH₂G) but in a GXC sequence, in addition to targeted mutations, semitargeted mutations at the 5′-G are observed.¹⁷⁹

Kamiya and co-workers found that siRNA knockdown of Werner syndrome protein (WRN), in a double-stranded plasmid containing an 8-OxoG:C pair, significantly increased the total mutation frequency in the supF gene.¹⁸⁰ While the G:C→T:A transversion at the 8-OxoG:C site (position 122) did not increase, mutations were enhanced in frequency at G:C sites other than position 122. Strikingly, a significant frequency of mutations could be detected as far as 27 bases away from the lesion site. They also showed that the knockdown of DNA polymerase λ enhances the mutant frequency at G:C sites other the one at which the 8-OxoG was located.¹⁸¹ Asubsequent study using a double knockdown of both WRN and pol λ, by contrast, did not reveal a synergistic effect of the two mutations on these off-targeted mutations, suggesting that these two proteins do not work independently to produce the off-targeted mutations.¹⁸² The authors hypothesize that the formation of 8-OxoG radicals followed by radical cation migration on DNA, owing to a prooxidant state in the WRN-reduced cells, maybe a reason for the off-targeted mutation, which is suppressed by WRN-pol λ.

Several intrastrand cross-links have been reported to cause high levels of semitargeted mutations. Gentil was the first to report that the UV lesion T(6,4)T (Figure 8) is highly mutagenic (mutation frequency ~60%) in COS-7 cells inducing primarily G→T transversions that are semitargeted to the 5′ neighboring base of the 5′-T of the photoproduct.¹⁸³ Livneh reported that in pol η-deficient human cells 14–18% mutations occur with a site-specific cyclobutane–pyrimidine dimer (CPD) (Figure 8), of which about three-quarters are targeted mutations opposite the 3′-T of the CPD and nearly one-quarter are semitargeted mutations at the nearest nucleotides flanking the CPD.¹⁸⁴ In contrast, T(6,4)T is highly mutagenic (27–63%) in human cells, with little effect of the presence or absence of pol η and approximately half of the mutations are semitargeted, of which 84–93% are because of the insertion of an A opposite the template G 5′ to the T(6,4)T cross-link.¹⁸⁴

Figure 8.

UV and hydroxyl radical-induced tandem DNA damages.

The mutational spectra for the tandem lesions, G[8,5-Me]T and T[5-Me,8]G (Figure 8), in both COS-7 and HEK 293T cells show high frequencies of semitargeted mutations in addition to targeted mutations.¹⁸⁵ In fact, G[8,5-Me]T induces a higher frequency of semitargeted mutations than targeted mutations in both types of cells, whereas for T[5-Me,8]G in HEK 293T cells, targeted and semitargeted single-base substitutions occur at approximately the same frequency (Figure 9).¹⁸⁵

Figure 9.

Types and frequencies of single-base substitutions induced by G[8,5-Me]T (G^T; left panel) and T[5-Me,8]G (T^G; right panel) detected in HEK 293T cells.¹⁸⁵ The colors used in the bar graph represent T (red), A (green), G (blue), and C (yellow).

A significant frequency of semitargeted mutations has also been reported following replication of DNA–peptide cross-links. In human cells, replication of DNA containing a 10-mer Myc peptide covalently attached to 7-deaza-dG gives 20% targeted G→A and G→T plus 15% semitargeted mutations at a guanine 5 bases 3′ to the lesion site.¹⁸⁶ siRNA knockdown experiments suggest that pol η, pol κ, and pol ζ participate in the targeted mutations, but only pol ζ is involved in the semitargeted mutations. In another investigation, replication of a plasmid construct containing an 11-mer peptide cross-linked to DNA at the epigenetic marker, 5-formylcytosine, in human cells, was shown to result in 9% targeted C→T substitutions and a C deletion, and 5% semitargeted mutations composed of a number of different base substituents and deletions near the lesion site.^187,188 Many other lesions, including the DNA adducts formed by 1-nitropyrene, 6-nitrochrysene, 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK), and others have been reported to generate semitargeted mutations, in addition to a larger percentage of targeted mutations.^89,172,189

The mechanism of off-targeted and semitargeted mutations has not been widely investigated. In some cases, these mutations have been suggested to occur as a result of helix destabilization by the lesion. A possible scenario involves replacement of a replicative polymerase by a bypass polymerase as a result of stalled replication, and the bypass polymerase, which is more error-prone, may introduce mutations a few bases before and after the lesion site, even when it bypasses the lesion correctly. The increased error rate of the bypass polymerase may be the result of a lower thermodynamic stability of the local DNA duplex. This model cannot rationalize the off-targeted mutations far away from the lesion site and perhaps time is now ripe to investigate these effects systematically. Off-target mutations complicate the task of linking lesion structure in DNA to patterns of mutations.

5.5. Sequence Context-Dependent Mutagenesis.

As indicated, mutational spectra can be very complex. Three factors determine the overall mutational landscape.¹⁹⁰ First, DNA damaging agents react differentially with target bases depending on the chemical environment, including flanking bases, of the target base for adduct formation. Second, lesions in some contexts evade repair whereas the same lesion in another context may be vulnerable to removal. Third, polymerases may misreplicate an adduct in a context-dependent manner. While each of the mentioned points has been discussed, it is noteworthy that no studies have been done, for any individual lesion, in which a comprehensive analysis has been done of adduct formation, evasion of repair, and context-dependent polymerase-mediated mutagenesis. A few studies have been conducted in which a few chosen DNA sequence contexts were examined for either repair or mutagenesis. For example, repair of the tobacco-specific adduct O⁶-[4-oxo-4-(3-pyridyl)butyl]-guanine by human O⁶-alkylguanine-DNA alkyltransferases is influenced by the DNA sequence context.¹⁹¹ Nucleotide excision repair susceptibility of the (+)-trans-anti-[BP]-N²-dG adduct in several sequences shows that dynamic periodic denaturation of Watson–Crick base pairing on the 5′ flank of the lesion provides a strong recognition signal for repair in these sequences.^192,193 Biophysical studies showing marked local thermodynamic destabilization are consistent with this notion. Mutagenesis of the (+)-trans-anti-[BP]-N²-dG adduct also is context dependent.^174,194 Mutagenesis of the oxidative DNA damages 8-OxoG and Fapy-dG in human cells is also strongly dependent on DNA sequence context.^195,196 However, these and similar studies with other DNA lesions have been carried out in a limited number of DNA sequences.^{193,197–202} Even so, a combination of structural (NMR and crystal structure analyses) and computation studies carried out by Geacintov, Patel, Stone, Broyde, Wetmore and others has provided much needed links between the structure and conformation of a lesion in different sequences and the biological effects of that lesion.^{51,192,202,203}

There are only two studies in which context-dependent repair and replication have been studied in all three-base contexts. Delaney and Essigmann constructed a panel of 16 phage genomes in which O⁶-mG was placed in each of the 16 possible three-base contexts (e.g., 5′-NXN-3′, where Xis O⁶-mG and N is any base).¹²⁴ The vectors were introduced into E. coli cells that were either wild type, expressing only Ada and expressing only Ogt (Ada and Ogt are methyltransferases that are coexpressed in E. coli and remove the methyl group from O⁶-mG). The results show that Ada exhibits context dependence in repair; the 5′ preference of Ada decreased in the order GXN > CXN > TXN > AXN and its 3′ preference decreased as NX(T/C) > NX(G/A), with mutation frequencies for O⁶-mG in Ada-deficient cells ranging from 35% to 90%. Cells lacking Ogt display mutation frequencies for O⁶-mG ranging from 10 to 25%, with the context AXN being particularly refractory to repair; AXN is also the weakest context repaired by Ada. Cells in which both Ada and Ogt are deficient show mutation frequencies of nearly 100% in all contexts. In the complete absence of repair, therefore, the replicative polymerase of E. coli does not display context dependence in the generation of mutations. The polymerase always inserts TMP opposite O⁶-mG, regardless of the three-base context of the lesion. Taken together, the DNA repair enzymes for O⁶-mG show significant context dependencies, but the replicative complex does not.

The second study investigating possible context-dependent repair and mutation focused on 7,8-dihydro-8-oxo-1,N⁶-ethenoadenine, a hybrid lesion with features of 8-oxoadenine and 1,N⁶-ethenoadenine.²⁰⁴ Sixteen genomes with the adduct in all possible three-base contexts were replicated in E. coli cells that were either wild type or lacking AlkB, a dioxygenase that repairs 1,N⁶-ethenoadenine. Unlike the results of the study on O⁶-mG, no context dependence in repair or mutagenesis is observed for 7,8-dihydro-8-oxo-1,N⁶-ethenoadenine. The contrasting results of the study on O⁶-mG and that on 7,8-dihydro-8-oxo-1,N⁶-ethenoadenine underscore the importance of context in the process of mutagenesis. Moreover, they might explain why investigators, who typically use single contexts in site-specific mutagenesis studies, sometimes see unexpected results when the same DNA lesion is studied in different genetic systems. Although laborious, studying mutagenesis in all possible sequence contexts is advised if the goal is to provide a complete profile of the mutagenic properties of a given lesion. When a sequence context effect is observed, such as that seen with O⁶-mG, it is often challenging to provide a molecular rationale. Structural and modeling studies carried out in parallel with the mutagenesis and/or repair experiments may be one way to approach this problem.

5.6. DNA Lesion-Derived Mutagenesis in Chromosomes.

Most of the investigations using site-specific DNA lesions in mammalian cells has been carried out using plasmids or other extrachromosomal probes. Unlike the vectors typically used in site-specific mutagenesis investigations, the DNA of chromatin is engaged in a complex assembly of macromolecules composed of DNA, RNA, and protein, and it is likely that its higher-order structure affects the distribution of DNA damage, repair efficiency and subsequent mutational outcomes. Moreover, during the time that chromatin undergoes condensation to form a chromosome, DNA lesion formation, repair, and mutagenesis are likely to be even more varied than they are in a static chromosome. Only a limited number of studies on DNA lesion biology have been conducted on chromosomes.

A second issue that many site-specific mutagenic studies has not addressed is that translesion synthesis is only one of two major strategies for DNA damage tolerance. The other mode of DNA damage tolerance is homology-dependent repair (HDR), which is accomplished by copying the undamaged strand of the sister chromatid. As the latter has correct genetic information, HDR is essentially error-free. There are several ways HDR by template switching can be accomplished, and Figure 10 shows two of them.

Figure 10.

Model for DNA damage tolerance by HDR. A replication-stalling DNA lesion (indicated by the red filled circle) can be bypassed by either TLS or HDR. The latter involves a template switching mechanism, which can be initiated by strand invasion followed by making a copy of the complementary strand from the sister chromatid. Template switching can also happen by regression of the stalled fork, synthesis of the DNA complementary to the damaged site, and reversion. Various other models of HDR have been proposed. The structures in the boxes depict the bypassed lesion, by whichever mechanism the cell uses.

Fuchs and co-workers developed a methodology to insert a single C8-dG-AAF adduct (Figure 4) or a product of UV damage (Figure 8) at a particular chromosomal location in E. coli.²⁰⁵ This approach is based on the mechanism of phage λ site-specific recombination and includes the following components: a recipient E. coli strain with a single attR site, a nonreplicating plasmid construct containing the DNA lesion and an attL site plus an ampicillin resistance gene. The recombination reaction between attL and attR leads to the integration of the lesion-containing vector into the chromosome. Successful integration can be selected by the resistance of the integrant to ampicillin, which also restores a functional β-galactosidase gene (lacZ), which is then used as a target for mutagenesis. The undamaged opposite strand with a short sequence difference inactivates the lacZ gene and acts as a genetic marker allowing strand discrimination. Using this methodology, it was shown that for common replication-blocking lesions, such as UV-induced photoproducts or C8-dG-AAF adducts, damage avoidance events, which rely on the information present in the undamaged sister chromatid, massively outweigh TLS events (97% versus 1–3%) in non-SOS induced cells.

In an investigation of O⁶-alkylguanine mutagenesis, plasmid vectors containing O⁶-mG and O⁶-ethylguanine (O⁶-EtG) were constructed and replicated in Chinese Hamster Ovary (CHO) cells.²⁰⁶ The vectors integrated into and replicated in the host genome. After intrachromosomal replication, the DNA sequence surrounding and including the originally adducted site of each integrated vector was amplified from the host genome by PCR and was analyzed for mutations. Significant levels of mutation were observed from the O⁶-mG- and O⁶-EtG-containing vectors replicated in mex⁻ cells, which lack the enzyme O⁶-alkylguanine DNA-alkyltransferase (a mutation frequency of 19% for O⁶-mG and 11% for O⁶-EtG) inducing almost exclusively G→A transitions. However, no mutagenesis was detected when the adduct-containing vectors were introduced into mex⁺ cells, reflecting the powerful role of the O⁶-alkylguanine-DNA alkyltransferase in the repair of O⁶-mG and O⁶-EtG in CHO cells.

Recently, it was demonstrated by Livneh that DNA transposon plasmids could be engineered to include DNA lesions, which can be integrated into mammalian cells to study how cells bypass DNA lesions encountered during normal DNA replication.^207–209 The approach is as follows. A shuttle vector containing a site-specific lesion was constructed, which could integrate into chromosomes of mammalian cells by an ectopically expressed phage recombinase.²⁰⁷ A DNA vector containing a puromycin-resistant marker is cotransfected with the lesion-containing vector. In this system, replication occurs after the lesion-containing plasmid has integrated in the chromosomal locus. Following cell division, puromycin selection enables recovery of the cells in which the plasmid was integrated. Analysis of the DNA isolated from puromycin-resistant colonies indicated the mutations, if any. This method was used to investigate tolerance of the UV induced T(6,4)T photoproduct and benzo[a]pyrene-derived [+trans anti]-B[a]P-N²-dG adduct in human cells, which showed that they are both tolerated by TLS and HDR. It was determined that 89% TLS occurred in the construct with the photoproduct and that T(6,4)T is highly mutagenic, inducing semitargeted G→T at the 5′ neighbor of the cross-link (i.e., GT(6,4)T→TTT). For [+trans anti]-B[a]P-N²-dG, TLS gave 76% accurate and 6% G→T mutations, whereas HDR accounted for 18% of the progeny. Subsequently, a more advanced version of this method was developed, which allows a more accurate distinction between the HDR and TLS pathways.²⁰⁸ This approach involves the use of a piggyBac transposition-based system for the chromosomal integration of DNA lesions. The piggyBac-based assay is a robust chromosomal replication/repair assay that allows simultaneous investigation of lesion bypass, mutagenesis, and HDR in a mammalian chromosome. This method has the advantage of very efficient vector integration. Using this system, it was demonstrated that TLS is dominant for T(6,4) in mouse embryo fibroblasts. In contrast, [+trans anti]-B[a]P-N²-dG is bypassed primarily by HDR in murine cells. Furthermore, TLS and nucleotide excision repair were observed to work competitively, with 40% TLS occurring for the T(6,4)T in repair-proficient cells. The same pattern of mutagenic TLS was observed in mouse embryonic stem cells.

6. MUTATIONAL SIGNATURES AND HIGH-RESOLUTION MUTATIONAL SPECTRA: PROBING THE MUTATIONAL PATTERNS IN CANCER

About ten years ago, two advances were made that significantly impacted the field of mutagenesis. The first advance was by Stratton, Alexandrov, and Nik-Zainal, who, among others, used non-negative matrix factorization to computationally extract what they defined as “mutational signatures” from tens of thousands of sequenced human cancer genomes.^146,148,210 One novel feature of this approach was the way they presented single base substitution data (in duplex DNA, there are only 6 possible types of base substitutions (e.g., GC→AT, GC→TA, etc.) in all 16 possible three-base contexts (5′-NXN-3′, where N = T, C, G, or A, and X is the position of the mutation). As a consequence, mutational data appear as a 96-point spectrum (16 × 6 = 96), providing information not only about the type of mutations but also their sequence-context dependence. This additional layer of information can often distinguish between mutational processes that generate the same qualitative type of base substitution. For example, hydrolytic deamination of 5-methylcytosine, APOBEC enzymatic activity, and exposure to methylating agents are three distinct processes that generate C→T mutations. However, when considering the three-base sequence-contexts in which these mutations occur, they can be readily distinguished from one another: 5-methylcytosine-dependent mutations occur in 5′-NCG-3′ contexts, APOBEC-induced mutations occur in 5′-TCN-3′ contexts, and methylating agents that act via a methyldiazonium reactive intermediate induce mutations in 5′-NCY-3′ contexts where Y is a pyrimidine. In all these examples, the central C is the position of the C→T mutation.

The underlying hypothesis behind factorization approaches is that mutational data in human cancer genomes reflect the activity of a relatively small number of fundamental mutational processes, which operate to various degrees in many different cancers. Mathematical analysis can extract a unique pattern, termed “mutational signature” for each of these putative mutational processes, and then reconstruct the complex mutational spectra of individual tumors as linear combinations of mutational signatures. While many of these signatures were subsequently validated experimentally and associated with a defined biological process, they remain primarily computational constructs that often depend on the quality, size and complexity of the data set under analysis.²¹¹ Currently, there are about 120 single base substitution (SBS) mutational signatures proposed,²¹² and a large fraction of them has as of yet no known molecular explanation.

The following example showcases the utility of mutational signatures as hypothesis generators and illustrates how they can be validated experimentally. Studying liver cancer, Schulze et al. analyzed the computationally derived mutational signatures of human hepatocellular carcinomas.²¹³ They hypothesized that one of the mutational signatures identified, termed SBS24, may reflect exposure to the fungal toxin and human carcinogen aflatoxin B₁. This deduction was based on the mutational data (dominated by G→T transversions) and epidemiological data that suggested that members of the human population studied that exhibited this signature were exposed to AFB₁. One could not be certain, of course, that humans in the studied data set of SBS24 were actually exposed to AFB₁. Their prediction, however, was subsequently supported experimentally in works by Chawanthayatham et al.²¹⁴ and Huang et al.,²¹⁵ who administered AFB₁ to mice and showed that the experimentally derived mutational spectrum that occurs is strikingly similar to the computationally derived human signature SBS24.

As indicated, mutational signatures are mathematical constructs extracted by non-negative matrix factorization of large human tumor sequencing data sets. They often reflect a fundamental mutational process with a defined etiology, but they require careful experimental validation. By contrast, a mutational spectrum is a reproducible, experimental data point, obtained by sequencing the genome of an organism that was knowingly exposed to a known mutagenic/carcinogenic agent. Occasionally, if a mutational process is dominant, a mutational spectrum closely resembles a mutational signature; this is the case for AFB₁-induced spectra (see the work of Chawanthayatham et al. and Huang et al.).^214,215 However, more commonly, the mutational spectrum of a tumor, both in human and in animal models is very complex, reflecting contributions from multiple mutational signatures that arise late in the process of tumor development.²¹⁴ For convenience and to aid in comparison and mechanistic interpretation, mutational spectra and mutational signatures are usually both plotted as 96-point spectra, displaying point mutations as a function of three-base sequence contexts.

At about the same time as the work on mutational signatures, a second technological advance occurred, this time in the DNA sequencing field. This second advance allowed highly accurate mutational spectra to be observed, enabling comparisons with the rapidly growing catalog of computationally derived mutational signatures. As indicated in the beginning of this Perspective, mutational spectra are the linear presentation of mutations across a target sequence in the genome. As also indicated, the assays by which mutational spectra are constructed usually do not detect all mutations in the target region. For example, if an assay relies on a phenotype conferred by expression of a protein (e.g., the LacZ product, β-galactosidase), many mutations will be silent and therefore lost in the constructed mutational spectrum. As a major advance, Loeb and Vogelstein developed next-generation sequencing tools that enabled all mutations to be observed and these techniques have built-in ways to provide very high accuracy.^216,217 As one example, duplex consensus sequencing developed in the Loeb laboratory sequences each strand of DNA independently and then uses a computational error-correction algorithm to remove the false “mutations” artifactually created during the polymerase chain reaction steps, and during the sequencing reaction itself, which is typically error-prone.²¹⁸ This technology is at least 4 orders of magnitude more accurate than conventional next-generation sequencing.

7. CONCLUDING COMMENTS AND FUTURE PERSPECTIVE

Our goal in this Perspective was to provide a current overview of the relationship between DNA damage and mutagenesis, and how these processes relate to genetic diseases, particularly to the etiology of cancer. Some pertinent topics were omitted to sharpen the focus and make the length of this Perspective manageable. We have also deliberately skipped areas where significant advances in the chemistry of DNA lesions have been made but the biological effects of those lesions have not yet been pursued. To give one example, recent investigations of DNA–DNA cross-link formation by AP-sites is a stimulating area,^219–223 which we think will pave new avenues for future research. DNA–protein cross-links^224,225 and clustered DNA lesions^226,227 are also ripe areas of recent interest, which will likely expand. The relationship of DNA damage with epigenetics is another rapidly developing area that has significant potential.^228,229 Computational approaches that bridge the gap among DNA damage, DNA repair, and mutation are becoming more powerful tools,^230–232 which has thus far been applied only in a limited number of investigations. We believe that this area of research will continue to grow as more early stage chemists with interest in toxicology enter the arena and invigorate the field in the coming years.

We also see a need for expanded development and application of next-generation sequencing techniques that reveal mutational spectra that detect all mutations, even those that are phenotypically silent. The few studies done so far have revealed much novel detail regarding mutational processes.^214,215 In an ideal world, these high-resolution, unbiased mutation sequencing tools will be paired with application of the new lesion mapping methods;^141,142 this convergence of technologies will provide a highly granular picture of lesion formation and biological consequences in single well-controlled experiments. In parallel, advanced computational tools will be needed to mathematically deconvolute the very complex mutational patterns that we now know occur, especially the patterns that manifest late in tumorigenesis. Such highly diverse spectra can occur because of, among other things, the establishment of a mutator phenotype caused, perhaps, by oxidative stress from a triggered immune system in and around a tumor. Such studies may allow us to address “Holy Grail” issues, such as defining the extent to which inflammation contributes to the complex mutational spectra seen in late-stage tumor development.

Lastly, as indicated, the now-known complexity of mutational spectra is because of the following three factors: (a) the selective binding of mutagens to certain nucleotide targets,¹⁴⁰ (b) the resistance of adducts in certain contexts to DNA repair,^198,200,233 and (c) the property of a given polymerase to cause a specific adduct-driven mutation in a specific DNA sequence context.^7,234,235 To complement application of rigorous lesion mapping methods ((a) above), future site-specific mutagenesis experiments with single lesions should be designed wherein the mutagenic properties of lesions are studied in all sequence contexts and in cells that are deleted for, or expressing, all relevant repair proteins and polymerases. Lesion mapping will help define the context specificity underlying the chemical reactions of DNA adduction; the parallel biological evaluation of lesions in all repair and replication backgrounds will provide information that, combined with the complementary lesionmapping data, will help to bring the mutagenesis field to its next stage of evolution.

With regard to the future evolution of the field, we note that startlingly impressive advances have been made in the past 35 years, but the best days of this field are still ahead of it.

ABBREVIATIONS

TLS

translesion synthesis

8-OxoG

7,8-dihydro-8-oxoguanine

AP-site

apurinic/apyrimidinic site or abasic site

NGS

NextGen sequencing

C8-AP-dG

N-(deoxyguanosin-8-yl)-1-aminopyrene

ENU

N-ethyl-N-nitrosourea

MMS

methylmethanesulfonate

PAH

polycyclic aromatic hydrocarbon

B[a]P

benzolpyrene

BPDE

benzo[a]pyrene diol epoxides

AFB₁

aflatoxin B₁

O⁶-mG

O⁶-methylguanine

O⁶-EtG

O⁶-ethylguanine

O⁴-mT

O⁴-methylthymine

LM-PCR

ligation-mediated PCR

NNK

4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone

POB

4-(3-pyridyl)-4-oxobutyl

O⁶-POB-G

O⁶-[4-oxo-4-(3-pyridyl)-butyl]guanine

8-NH₂G

8-aminoguanine

WRN

Werner syndrome protein

CHO

Chinese Hamster Ovary

HDR

homology-dependent repair

SBS

single base substitution

Biographies

Biographies

Ashis Basu received his Ph.D. in Chemistry from Larry Marnett’s lab at Wayne State University, Detroit, MI and did postdoctoral research with John Essigmann at the Massachusetts Institute of Technology, Cambridge, MA. In 1990 he joined the University of Connecticut where he currently is a professor of chemistry. The research focus of the Basu laboratory is determination of the consequences of DNA damaged by antitumor drugs, chemical carcinogens, oxidation, or radiation. This research at the interface of Chemistry and Biology involves introduction of specific lesions in DNA by organic synthesis, investigation of the structural effects of the lesions, and studying their repair and replication.

John Essigmann received his Ph.D. in Toxicology with Gerald Wogan at the Massachusetts Institute of Technology, Cambridge, MA. His early work focused on identification of DNA adducts caused by chemical carcinogens, such as aflatoxin B₁. Since 1981, as a professor at MIT in the Departments of Chemistry and Biological Engineering, his laboratory developed a technology for probing the biological effects of single DNA lesions in living cells. The properties of over 100 lesions have been studied in his laboratory. In addition, he has worked on synthesis and mechanism of action of novel anticancer agents.