DNA REPAIR MECHANISMS FOR THE RECOGNITION AND REMOVAL OF DAMAGED DNA BASES

Transcription

1 Annu. Rev. Biophys. Biomol. Struct : Copyright c 1999 by Annual Reviews. All rights reserved DNA REPAIR MECHANISMS FOR THE RECOGNITION AND REMOVAL OF DAMAGED DNA BASES Clifford D. Mol, Sudip S. Parikh, Christopher D. Putnam, Terence P. Lo, and John A. Tainer Department of Molecular Biology and the Skaggs Institute for Chemical Biology, The Scripps Research Institute, La Jolla, California 92037; jat@scripps.edu KEY WORDS: DNA base excision repair, DNA glycosylases, minor groove recognition, nucleotide flipping ABSTRACT Recent structural and biochemical studies have begun to illuminate how cells solve the problems of recognizing and removing damaged DNA bases. Bases damaged by environmental, chemical, or enzymatic mechanisms must be efficiently found within a large excess of undamaged DNA. Structural studies suggest that a rapid damage-scanning mechanism probes for both conformational deviations and local deformability of the DNA base stack. At susceptible lesions, enzyme-induced conformational changes lead to direct interactions with specific damaged bases. The diverse array of damaged DNA bases are processed through a two-stage pathway in which damage-specific enzymes recognize and remove the base lesion, creating a common abasic site intermediate that is processed by damage-general repair enzymes to restore the correct DNA sequence. CONTENTS PERSPECTIVES AND OVERVIEW DEAMINATION Uracil-DNA Glycosylase G T/U Mismatch-Specific DNA Glycosylase OXIDATION MutY Mismatched-Adenine Glycosylase /99/ $

2 102 MOL ET AL Endonuclease III Oxidized-Pyrimidine Glycosylase/Lyase ALKYLATION AlkA 3-Methyladenine DNA Glycosylase ULTRAVIOLET RADIATION T4 Endonuclease V ABASIC SITES AP Endonuclease FUTURE DIRECTIONS PERSPECTIVES AND OVERVIEW The genetic integrity of cells depends upon DNA repair enzymes that detect, recognize, and remove mutagenic lesions from DNA. The dominant form of DNA damage in cells is damage to DNA bases, which must be recognized and removed by specific DNA repair enzymes (49). Structural analyses, coupled with elegant mutagenesis and biochemical experiments, have recently furthered our understanding of how DNA repair enzymes recognize and remove damaged DNA bases. The phenomenon of nucleotide flipping, in which a target DNA nucleotide is flipped out of the DNA base stack, has been proven to occur in some damage-specific DNA repair enzymes and postulated to be used by many other DNA repair enzymes. Nucleotide flipping was first observed directly in cytosine-5-methyltransferases bound to their target DNA sequences; however, the methyltransferases are sequence-specific DNA-modifying enzymes that bind to a particular DNA sequence and flip a normal base out of the DNA base stack (9), whereas DNA repair enzymes must recognize damaged bases regardless of DNA sequence context. Recent results are beginning to reveal intricate methods for the initial detection of damaged DNA bases and suggest that the mechanisms used by DNA repair enzymes to locate, recognize, and bind their target nucleotides are likely different from those used by sequence-specific enzymes. Nucleotide flipping provides a general mechanism for the recognition of potentially mutagenic lesions that arise in DNA and has been structurally characterized in enzymes of the base excision repair (BER) pathway (70). BER, which is the primary defense against all major forms of DNA base damage, occurs in two stages: an initial, damage-specific stage carried out by individual DNA glycosylases targeted to distinct base lesions, and a damage-general stage that restores the correct DNA base sequence (Figure 1). The function of DNA BER enzymes is to detect and remove a variety of specific, individual base lesions within a large pool of undamaged DNA. The elegance of nucleotide flipping for DNA base repair is that efficient initial lesion detection, ease of damaged DNA base flipping, complementarity of interactions with damaged base partners, and stabilization of the extrahelical nucleotide, all improve specificity beyond

3 DNA DAMAGE RECOGNITION & REPAIR 103 Figure 1 Schematic outline of DNA base excision repair (BER) showing the steps of the separate damage-specific and damage-general stages. Many distinct DNA glycosylases (see text) recognize and remove bases damaged by exogenous and endogenous agents to generate a central AP site intermediate. The AP site products of the glycosylases and the 3 0 termini generated by AP lyases are processed by the damage-general 5 0 -AP endonuclease to provide a free 3 0 OH for repair synthesis. DNA polymerase inserts the correct nucleotide and DNA ligase seals the phosphodiester backbone to complete repair. direct recognition of DNA base damage. Thus, within duplex DNA in vivo, the activity of a DNA repair enzyme is due to both its chemical specificity for a particular damaged base and its nucleotide-flipping specificity for a particular damaged base mispair in DNA. This review describes current knowledge of how DNA repair enzymes may accomplish the recognition and removal of the five major types of DNA base damage: deamination, oxidation, alkylation, ultraviolet radiation, and

4 104 MOL ET AL depurination/depyrimidination. The structure and biochemistry of enzymes for each major type of DNA base damage are examined in terms of the structural chemistry and mechanisms for damage detection and removal. DEAMINATION The DNA bases adenine, guanine, and cytosine have exocyclic amino groups and are deaminated by hydrolytic attack, the action of chemical agents, and the abortive action of DNA methyltransferases. Hypoxanthine, arising from adenine deamination, can mispair with cytosine to give AT! GC transition mutations and can be excised from DNA by a hypoxanthine-dna glycosylase (16, 35, 63) or by the broad-specificity enzyme 3-methyladenine DNA glycosylase II (AlkA) (78), which is described in the section on the repair of alkylated bases. Xanthine in DNA, arising due to guanine deamination, can mispair with thymine (23), and may also be repaired by AlkA (4). Cytosine deamination, however, is particularly common and genotoxic in that it creates U G mispairs in DNA that cause GC! AT transition mutations if left unrepaired prior to the next round of replication (25). To counteract these cytotoxic effects, cells contain specific DNA repair enzymes that efficiently recognize and remove uracil from DNA; these enzymes have been characterized both structurally and biochemically. Uracil-DNA Glycosylase Uracil-DNA glycosylase (UDG) is the prototypical DNA repair glycosylase that cleaves the N-C1 0 glycosyl bond between a target uracil base and the deoxyribose sugar. UDG removes uracil from both U G mispairs and from U A base pairs in double-stranded DNA, as well as uracil in single-stranded DNA, but will not remove uracil from RNA (84). X-ray crystal structures of human (58) and viral (79) UDG show that it has a classic, single-domain / fold with a central, four-stranded parallel sheet. Residue conservation (60), mutation data, and structures of UDG complexes with uracil (79) and 6-aminouracil (58) reveal the basis for uracil specificity in an active-site pocket within a positively charged groove at the C-terminal edge of the central sheet. The uracil stacks with a conserved phenylalanine side chain and forms specific hydrogen bonds with its Watson-Crick atoms, effectively precluding cytosine binding, whereas thymine binding is blocked by van der Waals contacts with a tyrosine at the uracil C5 position (Figure 2). These interactions demonstrate that for uracil to bind in the active-site pocket, it must be extrahelical, providing the first direct evidence that DNA repair enzymes flip their target bases out of the DNA base stack. Crystal structures of human UDG (57) and viral UDG (80) bound to the uracil glycosylase inhibitor (Ugi) protein from Bacillus subtilis bacteriophages

5 DNA DAMAGE RECOGNITION & REPAIR 105 Figure 2 Stereo view of the chemical specificity of the human UDG active-site pocket. Uracil specificity comes from exquisite complementarity by the active-site pocket (gray tubes with polar atoms as spheres) with the flipped-out uracil nucleotide (thin dark lines), which prevents recognition of other pyrimidines. Specific hydrogen bonds (small black dots) discriminate against cytosine bases, and the close approach of a Tyr side chain (upper center) sterically excludes thymine from the pocket. Based on UDG-DNA crystal structures, this model of the transition state shows the catalytic water, which is coordinated by a Pro carbonyl and a His side chain (bottom), activated by the catalytic Asp (bottom center) and poised for attack on the uridine C1 0 atom (center left). PBS1/2 revealed an interesting molecular mimicry of DNA by protein. The human UDG:Ugi complex was interpreted to identify a conserved Leu residue, which protrudes from an enzyme loop above the uracil-recognition pocket, as playing a key role in nucleotide flipping. Ugi inhibits a wide range of UDGs by targeting the DNA-binding surface of the enzyme, thereby enveloping the exposed enzyme Leu within a hydrophobic pocket. Human UDG enzyme kinetic and DNA binding studies showed that mutation of this Leu significantly altered activity, with a Leu! Arg substitution increasing K M for double-stranded DNA substrates (85). The enhanced DNA binding of the UDG Leu! Arg mutation, along with an activity decreasing active-site Asp! Asn mutation, was exploited to yield a crystal structure of a human UDG double-mutant enzyme bound to doublestranded uracil-containing DNA (85). This first structure of UDG bound to substrate DNA confirmed flipping of the damaged nucleotide, clarified the catalytic mechanism, and suggested the major steps involved in damaged base recognition. UDG orients its positively charged, active-site groove along DNA, with the protruding enzyme loop inserted in the DNA minor groove, and binds the sugar-phosphates flanking the uridine to compress the DNA backbone. The target uracil, deoxyribose, and 5 0 phosphate are flipped out through the DNA major groove and into the enzyme active-site pocket through a push-pull mechanism: The push of the Leu into the base stack is accompanied by the

6 106 MOL ET AL pull of specific recognition of the uracil nucleotide by complementary activesite residues (85). Concerted UDG loop movements around the bound uridine exclude bulk solvent from the active site, insert a side chain into the DNA base stack, and bring a His into position to deliver a charged hydrogen bond to uracil O2. This and other active-site hydrogen bonds to uracil polarize the N-C1 0 bond, making it susceptible to attack by a water nucleophile activated by an Asp residue (Figure 2). Mutational analysis of human UDG verified the roles of these key UDG residues in uracil recognition and catalysis (58), and designed single-mutant enzymes exhibited cytosine and thymine DNA glycosylase activity, respectively (39). A Leu! Ala human UDG mutant DNA co-crystal structure reveals further structural aspects of UDG catalysis (69). This structure contains only a flipped-out AP site, and the uracil-recognition pocket contains two ordered water molecules rather than a uracil base, suggesting that the enzyme had cleaved the uracil and disengaged the products prior to rebinding the AP DNA. DNA-binding kinetics show that the Leu! Ala mutant UDG dissociates from substrate DNA faster than the wild-type enzyme, and also that the mutant and wild-type UDG associate more rapidly with, and bind more tightly to, AP DNA than with uracil-containing substrate DNA (69). Thus, an important function of human UDG in vivo may be to protect cells from the cytotoxic effects of AP sites. High-resolution X-ray crystal structures of wild-type UDG bound to DNA containing either a U A base pair or a U G mispair confirm that the protruding Leu does intercalate into the DNA base stack (69). These structures are trapped product complexes that exhibit cleavage of the glycosyl bond and bind free uracil and an extrahelical abasic deoxyribose. The DNA, both 5 0 and 3 0 of the flipped-out nucleotide, is B-form, and the distance between the phosphates flanking the uracil nucleotide is compressed by 4 Å by three rigid enzyme loops, which causes the DNA to kink by 45. These rigid loops contain Ser, Pro, and Gly residues that allow a close approach of the polypeptide chain to the DNA backbone and would clash with the DNA in an initial UDG-DNA complex, modeled by superimposing straight, undamaged B-DNA onto the uncomplexed UDG structure. Thus, free UDG cannot bind B-DNA without pinching and thereby bending the DNA backbone. However, initial lesion detection and specific uracil recognition are likely distinct events, with detection via the loop pinch preceding complete nucleotide flipping, and allow for an efficient and rapid scanning of double-stranded DNA for uracil lesions. This rigid loop pinch mechanism differs from other models for initial lesion detection by DNA glycosylases, which include serendipitous capture of spontaneously flippedout bases (68), DNA strand separation (94), assisted flipping by an accessory protein (10), and extrahelical migration of flipped-out nucleotides (95).

7 DNA DAMAGE RECOGNITION & REPAIR 107 In the pinch-push-pull model, the coupling of initial uracil damage detection by DNA backbone compression to nucleotide flipping explains the preferential excision of U G mispairs through an enhanced lesion-detection mechanism. In the UDG:U G-DNA complex, the position of the orphan guanine is displaced into the DNA major groove, whereas in the UDG:U A-DNA structure, orphan adenine remains in its normal position in the DNA base stack. In U G wobble mispairs, the uracil is shifted into the major groove and the guanine is displaced from the helix axis into the DNA minor groove, but this purine displacement does not occur in U A pairs. The U G wobble displacement would clash with the Leu in the minor groove and stop UDG from precessing along the DNA. The branched Leu side chain may promote uracil nucleotide flipping from U G mispairs by pushing both the guanine slightly into the major groove and the uracil completely out through the DNA major groove and toward the enzyme active site (Figure 3). Interactions with the DNA minor groove via a reading head formed by the protruding UDG Leu loop residues may provide additional specificity. Human UDG reads the DNA minor groove through water-mediated hydrogen bonds between polypeptide backbone atoms, an Arg and a Tyr side chain, and the N3 atoms of purine DNA bases. Different interactions observed with U G mispairs Figure 3 The mechanism whereby uracil arises in DNA through the deamination of cytosine residues in C G base pairs (left) and generation of wobble mispairs. The U G wobble mispair (right) causes the uracil to shift toward the DNA major groove and the guanine to fall toward the DNA minor groove. The shifted guanine exocyclic amino group may be detected in the DNA minor groove by the UDG, or MUG, protruding leucine-containing loop. The leucine side-chain push on the guanine likely initiates flipping of the uracil nucleotide out through the DNA major groove and into the enzyme active site.

8 108 MOL ET AL and U A base pairs likely influence the mispair and sequence preference of UDG (84). The Tyr side chain is wedged in and widens the DNA minor groove, likely facilitating full Leu side-chain insertion into the DNA base stack and flipping of the uracil nucleotide. The loop conformation and minor groove interacting residues in human UDG are not conserved in HSV-1 UDG, which may account for the reduced activity on U A pairs (96), stronger binding to U G mispairs (68), and lack of AP DNA binding (1) of the viral enzyme. Efficient UDG detection and flipping of uracil in DNA involves a pinchpush-pull mechanism: Rapid scanning by DNA backbone compression (pinch) slightly bends the DNA, which becomes fully kinked upon uracil nucleotide flipping by penetration of the Leu loop into the DNA base stack (push) and the attraction of the complementary, uracil-specific recognition pocket (pull). The UDG pull is so effective that any uracil base in DNA will be bound and cleaved, and UDG acts effectively even on single-stranded DNA, where the pinch and push components are negligible. These coupled structural, mutational, and biochemical results have significantly increased our understanding of how UDG might perform its functions, accounting for its extreme efficiency and absolute specificity for uracil in DNA. Aspects of the UDG pinch-push-pull method are common themes likely used by other damage-specific DNA repair enzymes. G T/U Mismatch-Specific DNA Glycosylase In cells, cytosine is often methylated at its 5 position to control gene expression, and these 5-methylcytosine bases can undergo deamination to yield 5-methyluracil, or thymine, and thus give rise to G T mismatches in DNA. Because UDG will not bind and cleave thymine residues, these mispairs must be repaired by the G T/U mismatch-specific DNA glycosylase (MUG). MUG will cleave thymine or uracil from mispairs with G, C, and T, but is inactive on A T/U pairs and on single-stranded DNA (66). The crystal structure of MUG from Escherichia coli (1) revealed unexpected structural homology to UDG, including the central sheet, C-terminal helices, rigid pinch loops, and a loop analogous to the protruding UDG leucine loop above a similar base-specificity pocket. The crystal structure of a MUG:DNA complex, however, revealed that flipped-out nucleotide binding by MUG does not involve any base-specific interactions, with the enzyme able to accommodate either thymine or uracil in its active site, and suggested a possible basis for mismatch recognition through interactions with the orphan guanine base partner of the flipped-out target nucleotide (1). Despite the very low catalytic efficiency of MUG, 6000 times lower than UDG (1), the MUG:DNA complex contains a flipped-out abasic deoxyribose and the base binding pocket is empty. The residues that line this pocket differ from those in the UDG pocket and do not provide the specific complementary hydrogen bonds and van der Waals contacts to discriminate

9 DNA DAMAGE RECOGNITION & REPAIR 109 Figure 4 Stereo view of the chemical specificity of the MUG active site illustrating the lack of strict complementarity, compared to the high complementarity seen in the UDG active site. To accommodate both thymine and uracil bases, MUG (gray tubes with polar atoms as spheres) replaces the discriminating Tyr in UDG with a Gly (lower center). Based on MUG and UDG-DNA crystal structures, this model of a flipped-out uncleaved thymidine (thin dark lines) shows that thymine stacks with the Phe side chain (top right) and receives an uncharged hydrogen bond from an Asn (top left). The catalytic water, positioned by another weakly basic Asn side chain (bottom center), is oriented for attack on the thymidine C1 0 (center left). between pyrimidines (Figure 4). Therefore, MUG can bind either uracil or thymine but lacks the charged residues and specific hydrogen bonds to the flipped-out nucleotide seen in UDG and is thus less effective at polarizing and cleaving the N-C1 0 bond (Figure 4). Above the MUG active site, a leucine-containing loop binds the DNA minor groove, inserting a Leu into the DNA base stack and forming specific backbone hydrogen bonds with the orphan guanine N1 and exocyclic amino group. While these interactions would likely not discriminate against the smaller pyrimidine orphan bases in C T/U and T T/U mispairs, they could not be formed with an adenine orphan base. Adenine lacks the exocyclic amino group of guanine, and the adenine N1 atom cannot donate a hydrogen bond to the MUG backbone carbonyls, suggesting that MUG mismatch specificity may reflect negative selectivity that actively discriminates against binding and flipping at A T/U base pairs but accommodates thymine or uracil in mispairs. As in the UDG:DNA complexes, the DNA in the MUG:DNA structure is kinked, suggesting that MUG also employs a DNA backbone pinch for initial damage detection. The phosphodiester bond 3 0 of the orphan guanine is broken, however, and thus the amount of DNA kinking seen may not be as great as in an intact MUG:DNA complex. DNA damage detection and recognition by MUG would appear to be a pinch-push mechanism that lacks the pull of a

10 110 MOL ET AL complementary base recognition pocket. Lacking chemical specificity for its target base, MUG relies on a nucleotide-flipping specificity, which ensures MUG will remain bound to only G T and G U mispairs, but not normal base pairs, long enough for glycosylic bond cleavage to occur by the relatively inefficient MUG catalytic site. OXIDATION All eukaryotic life requires the efficiency of aerobic metabolism (32), yet these oxidative processes generate reactive oxygen species that attack the cellular infrastructure (13), leading to mutagenesis, carcinogenesis, and aging (27). Hydroxyl radicals will rapidly react with the C5-C6 double bond of pyrimidines, creating radical intermediates that react with O 2 to yield thymine glycol and other oxidatively damaged pyrimidines, which are excised by the broadspecificity DNA glycosylase/ap lyase, named endonuclease III (Endo III) due to its AP lyase activity. Hydroxyl addition also occurs frequently at the susceptible guanine C8 position, yielding 7500 mutagenic 8-oxoguanine ( 8 OG) lesions per human cell (2). While MutM (55) in Escherichia coli, and OGG1 in eukaryotic systems (64, 65, 91), excise 8 OG from 8 OG C DNA base pairs, it is the mismatched-adenine specific glycosylase MutY that recognizes and removes adenine from the mutational intermediate 8 OG A mispair (30, 55). MutY and Endo III are founding members of a DNA repair protein superfamily (89), including AlkA and OGG1 (64), that have similar protein architectures and related chemical mechanisms (17). The basis for substrate specificity and catalysis revealed by recent high-resolution structures of MutY shows how this conserved molecular scaffold may detect, recognize, and remove diverse types of DNA base damage, and suggests how the specificity for other members of this enzyme superfamily might be achieved. MutY Mismatched-Adenine Glycosylase MutY is a 39-kDa iron-sulfur cluster [4Fe-4S] protein that specifically removes adenine mispaired with 8 OG, guanine, or cytosine. The DNA repair activity of MutY is localized to the N-terminal 26-kDa region of the protein (28, 52). The crystal structure of this region (31) shows that it belongs to the Helix-hairpin- Helix (HhH) superfamily of DNA repair enzymes, with an all- helical fold comprised of two domains: a six- helix barrel domain and a four- helix domain that is structurally coordinated by the [4Fe-4S] cluster. These domains are connected by two short hinge loops and have an electrostatically positive surface for binding DNA. This two-domain protein architecture is characteristic of the canonical Endo III HhH glycosylase fold that is found in a large superfamily of DNA repair glycosylases (64). The HhH structural motif, which was first

11 DNA DAMAGE RECOGNITION & REPAIR 111 noted in Endo III (89) and since found in several DNA-binding protein families (20), consists of two closely packed, nearly antiparallel helices connected by a four-residue type II hairpin turn. In the turn, two conserved glycine residues direct their amide nitrogens into solvent to interact with DNA backbone phosphates, as directly observed in DNA polymerase (73) and RuvA (34) crystal structures bound to DNA. The primary MutY HhH motif is located on the six-helix barrel domain at one end of a deep groove formed at the domain interface, while at the other end of this groove there is a second possible HhH motif. Conserved residues and residues whose mutation affects MutY activity cluster around the interdomain cleft, implicating this region as the MutY active site, and the MutY HhH motifs likely orient the DNA for mispair recognition and catalysis. The structure of MutY in complex with its adenine substrate (31) provided the first direct experimental observation of how a target base binds to an HhH glycosylase active site. The position of the adenine within the interdomain cleft reveals that this enzyme superfamily uses nucleotide flipping and that target purine bases can also be flipped out from the DNA helix into an enzyme active site similarly to the way the smaller pyrimidines are flipped by UDG and MUG. The chemical specificity of MutY for adenine derives from a highly complementary binding pocket that forms specific hydrogen bonds between the adenine Watson-Crick N1 and N6 atoms and a glutamine side chain, and from the adenine Hoogsteen N6 and N7 atoms to a glutamic acid residue (Figure 5). These interactions discriminate against binding of other normal DNA and RNA bases. Figure 5 Stereo view of the chemical specificity of the MutY active-site pocket demonstrating strong hydrogen bonding and shape complementarity for adenine bases. Binding of the flippedout adenosine (thin dark lines), based on the MutY:adenine complex structure, shows the specific hydrogen bonds (small black dots) from a MutY (gray tubes with polar atoms as spheres) Gln side chain (upper right) to adenine N1 and N6 and from a MutY Glu side chain (bottom right) to adenine N6 and N7. A water is positioned by the catalytic Asp (bottom left) for nucleophilic attack on the deoxyribose C1 0 position (center left).

12 112 MOL ET AL Although some associated DNA lyase activity exists (53), MutY functions primarily as a pure glycosylase, cleaving the glycosyl bond of flipped-out adenine nucleotides by the attack of a water molecule activated by an aspartic acid residue (Figure 5). Several of the residues involved in adenine specificity, including the catalytic Asp, emanate from the hinge loops that connect the two MutY domains. The nature of these loop residues can alter the active-site environment and hence the specificity of MutY as well as other HhH glycosylase enzymes. These loops may also act as hinges to allow the two domains to close around a bound adenine and facilitate catalysis by excluding bulk solvent, as seen for UDG, which changes from an open to a more closed conformation upon binding substrate (69). MutY mismatched-adenine recognition cannot be imparted solely by the chemical specificity of the MutY recognition pocket, which can accommodate adenine from any base pair, but must also derive from the nucleotide-flipping specificity imparted by the adenine base partner. The constraints of MutY HhH motif interactions with DNA phosphates, flipped-out adenine binding, and the requirement that the catalytic Asp be correctly positioned for catalysis, establish the orientation of DNA binding (31). In this MutY:DNA complex model, the DNA is bent by phosphate compression between the two HhH motifs, and a conserved motif is positioned to bind the DNA minor groove. Mutations in this MutY minor-groove reading motif eliminate MutY activity (31), supporting a role for it in nucleotide flipping. In A G and A 8OG mispairs, the guanine or 8 O-guanine preferentially assume syn configurations (7, 54). In this orientation, the 8-oxo group of 8 OG is 1 Å deeper in the minor groove than the O2 group of thymine in a normal A T base pair (Figure 6). This structural aberration may clash sterically with the MutY minor-groove binding loop, which contains two consecutive, conserved Gln s. These Gln s are positioned to make specific interactions with the adenine base partner, and provide the push on the A 8OG mispair to function like the Leu s seen in UDG and MUG, stabilizing the extrahelical conformation of adenine nucleotide by inserting in the hole in the DNA base stack. These mutational and structural results suggest that MutY employs a pinchpush-pull mechanism similar to UDG for DNA damage detection and recognition. In this model, rapid damage scanning by compression of the DNA intrastrand phosphate distance between the MutY HhH motifs leads to DNA bending and the initiation of nucleotide flipping. A 8OG mispairs are detected in the DNA minor groove through the syn conformation of the 8 OG and through specific interactions of adenine with the MutY pocket. The nucleotide-flipping specificity of the push is likely a key determinant of MutY activity in concert with the chemical specificity of the pull afforded by the adenine-binding pocket. Thus, MutY only cleaves adenine and is one thousand times more active

13 DNA DAMAGE RECOGNITION & REPAIR 113 Figure 6 Generation of adenine mispairs with 8-oxoguanine ( 8 OG) in DNA during DNA replication. 8 OG bases in DNA can assume the syn conformation (right), with the 8-oxo group in the DNA minor groove. DNA polymerase will insert an adenine opposite syn 8 OG on the template strand. Detection of the resulting A 8OG mispairs is likely accomplished by a MutY Gln residue in the DNA minor groove, prompting adenine nucleotide flipping through the DNA major groove and into the MutY interdomain recognition pocket. A normal A T base pair is shown for comparison (left). for adenine opposite 8 OG than for adenine opposite AP-sites or in A C mispairs (7). Endonuclease III Oxidized-Pyrimidine Glycosylase/Lyase Endo III is a [4Fe-4S] cluster enzyme that is ubiquitous in the three kingdoms of life and was discovered as an activity in Escherichia coli that nicked UVirradiated DNA (75). This 23-kDa enzyme is both a broad-substrate DNA glycosylase that cleaves oxidized pyrimidine bases, and an AP lyase that nicks DNA at the resultant abasic sites (19, 38). Oxidatively damaged pyrimidines, including thymine glycol and urea residues (15), cytosine and uracil hydrates (6), and 5-hydroxypyrimidines (36), are produced by ionizing radiation (86), ultraviolet light (75), and chemical oxidizing agents (26), and are all removed by Endo III. The crystal structure of E. coli endonuclease III (43) founded the HhH glycosylase enzyme family and first revealed the versatile HhH structural motif and its implications for binding DNA backbone phosphates (89). Endo III is bilobal with two distinct domains: a six- helix barrel domain, containing the conserved HhH motif and a minor groove reading motif, and a four- helix [4Fe-4S] cluster domain. Comparisons with the equivalent domains in MutY reveal variability in the interdomain angle, resulting in the Endo III interdomain cleft being 5 Å wider than the MutY cleft (31). This relative movement

14 114 MOL ET AL suggests the two domains pivot about the hinge loops linking them and clamp down on bound substrate, to form the catalytically competent complex. The groove at the Endo III domain interface contains a hydrophilic pocket identified as the enzyme active-site pocket by mutagenesis experiments (89). Whereas the larger MutY pocket is highly specific for adenine, the Endo III pocket is lined with polar residues and filled with a hydrogen-bonded network of water molecules, which may provide the structural plasticity required to accommodate a wide range of damaged pyrimidines. The Endo III active-site chemistry also differs from that of MutY. The conserved Asp activates a lysine N nucleophile rather than a water molecule (Figure 7). The Lys attacks C1 0 and forms a covalent enzyme-dna Schiff s base intermediate, which breaks the 3 0 C-O phosphodiester bond through -elimination and hydrolysis and releases an, -unsaturated aldehyde, imparting Endo III with AP lyase strand nicking activity. Mutation of either the Asp or the Lys severely impairs Endo III activity but not DNA binding (89). A common feature of thymine glycol and other oxidized pyrimidines recognized by Endo III is that they are no longer planar and aromatic and tend to be predominantly extrahelical in solution (37, 42). Thus, damage recognition by Endo III may not require a significant push to flip the damaged nucleotide, and the nucleotide-flipping specificity is not affected by the interactions with the damaged base partner on the complementary strand. The pull afforded by Figure 7 Stereo view of the broad chemical specificity of the Endo III active-site pocket. Based on the MutY-adenine co-crystal structure, this model for binding of thymine glycol (thin dark lines) in the Endo III active site (gray tubes with polar atoms as spheres) shows that hydrophilic and charged residues in concert with water molecules (not shown) make the pocket capable of binding a variety of oxidized pyrimidines. In contrast to pure glycosylases, Endo III uses a Lys N (bottom center) activated by an Asp as a nucleophile (bottom left) to attack the C1 0 atom of the damaged nucleotide.

15 DNA DAMAGE RECOGNITION & REPAIR 115 the polar Endo III damage-recognition pocket provides the general shape and chemical environment needed to accommodate a wide range of damaged pyrimidines, and importantly, its size provides negative selectivity that discriminates against larger, hydrophobic purine bases. Endo III and its homologs are apparently general-purpose enzymes that guard against the most common form of damage inflicted on DNA pyrimidine bases. Alkylation damage, which is more commonly incurred by purine bases, is removed by another broad-specificity HhH family glycosylase, AlkA. ALKYLATION Alkylating agents are strong electrophiles that attack biological macromolecules, including the 20 nucleophilic centers of DNA bases. Alkylated DNA bases are positively charged and have altered hydrogen bonding patterns, which can lead to mispairs during DNA replication and produce mutations. Major alkylation sites in DNA include the N3 atoms of purines, in the DNA minor groove, and the O4 and O6 positions of thymines and guanines, respectively. Of these alkylated bases, O6-methylguanine, which induces GC! AT transition mutations, is strongly mutagenic and is directly repaired in an irreversible stoichiometric reaction by the suicide DNA repair protein alkyl-dna alkyltransferase (reviewed in 72). A second major cytotoxic and mutagenic alkylated DNA base lesion is 3-methyladenine, which in E. coli is removed by two separate enzymes. The Tag protein, also called 3-methyladenine DNA glycosylase I, is expressed constitutively and specifically excises 3-methyladenine and, at lower efficiencies, 3-methylguanine (5), whereas the expression of 3-methyladenine DNA glycosylase II, or AlkA, is induced by alkylating agents (50). AlkA excises not only 3-methyladenine but a wide range of both alkylated purines and pyrimidines, oxidative lesions, and even undamaged normal bases from DNA (4). AlkA 3-Methyladenine DNA Glycosylase The crystal structures of E. coli AlkA, determined independently by two groups (44, 102), reveal a compact three-domain construction that demonstrates surprising homology of the two C-terminal domains to Endo III despite insignificant sequence similarity and the lack of an AlkA iron-sulfur cluster. The N-terminal domain, which is unique to AlkA, consists of a five-stranded, antiparallel sheet and two helices and is topologically similar to the conserved tandem repeat of the TATA-binding protein (44). The second AlkA domain is equivalent to the MutY and Endo III six-helix barrel domains, and contains the conserved HhH motif. The third AlkA domain lacks a [4Fe-4S] cluster and the portion of the domain that would ligate the iron atoms, yet otherwise has a

16 116 MOL ET AL similar helix packing and topology to the iron-sulfur cluster domains of Endo III and MutY. Unlike Endo III, however, AlkA is a pure glycosylase that cleaves the glycosyl bond of target bases by the hydrolytic attack of a water molecule activated by a conserved Asp. Mutagenesis (44) and residue sequence conservation (102) locate the AlkA active site in an extremely hydrophobic cleft between the second and third domains, analogous to the active site locations of MutY and Endo III. The ability of the two AlkA domains to adjust the size of the hydrophobic cleft by rotation about the hinge loops was observed in crystal soaking experiments, suggesting that such motion may allow flexible binding of, and domain closure around, a range of different alkylated DNA bases. The AlkA damaged base recognition pocket is hydrophobic and rich in electron-donating aromatic residues that can stack favorably with an extrahelical, electron-deficient, substrate base (44). AlkA Leu and Trp residues replace the Gln and Glu that form adenine-specific hydrogen bonds in MutY, whereas binding of positively charged alkylated bases may be favored by electron stacking interactions with an AlkA Trp residue (Figure 8). These features are likely crucial to the broad substrate specificity of AlkA, allowing both smaller pyrimidines and larger purine bases to bind but preventing favorable interactions with the nonaromatic and polar oxidized pyrimidine damaged bases recognized by Endo III. Thus, nature appears to have evolved two structurally related, broadly specific HhH glycosylase enzymes to Figure 8 Stereo view of the broad chemical specificity of the AlkA active-site pocket. The position and orientation of 3-methyladenine (thin dark lines) in the AlkA pocket (gray tubes with polar atoms as spheres) is based on the position of the adenine base in the MutY co-crystal structure. Unlike MutY, AlkA recognizes a diverse array of alkylated bases that fit within a hydrophobic pocket that has few potential hydrogen-bonding partners. The catalytic Asp (center left) is shown activating a nucleophilic water molecule poised to attack the C1 0 atom (upper left) of the bound nucleotide.

17 DNA DAMAGE RECOGNITION & REPAIR 117 recognize and remove the two most common forms of damage that occur to pyrimidine and purine bases: oxidation and alkylation. AlkA also has a minor-groove reading motif for damaged base detection that is equivalent to the MutY and Endo III motifs but has a Leu residue rather than the Gln s seen in the latter enzymes. Leu intercalation into the DNA base stack is a shared feature of the UDG:DNA and MUG:DNA co-crystal structures. The size and shape of the branched side chains of Leu and Gln may be optimal for DNA damage detection in the DNA minor groove and for facilitating nucleotide flipping. For AlkA, the Leu may insert into the DNA and interact favorably with the hydrophobic DNA bases flanking the flipped-out nucleotide, resulting in slower enzyme dissociation from susceptible target base lesions, as seen for UDG:DNA complexes in solution (69). Both AlkA and Endo III provide a general chemical specificity that is favorable for substrate binding, but they lack the strong attractive pull of a complementary damaged base recognition pocket. Damage detection by AlkA is likely influenced by the nucleotideflipping specificity afforded by the push of the AlkA Leu and minor-groove binding motif. ULTRAVIOLET RADIATION Ultraviolet (UV) light generates DNA base damage by many different mechanisms, including the production of oxygen radicals that lead to oxidative damage, and the generation of photoaddition products, particularly at adjacent pyrimidines, which readily form a cyclobutane pyrimidine dimer (CPD) across their C5-C6 double bonds. CPDs in DNA, as well as other forms of UV-induced DNA damage, can be directly repaired photochemically by DNA photolyases (reviewed in 77), and there is substantial evidence that the CPD is flipped out of the DNA base stack by these enzymes (71, 88, 103). CPDs in DNA can also be repaired by BER enzymes that avoid the difficult chemical steps of direct damage reversal. Crystal structures of the bacteriophage T4 Endonuclease V (Endo V), which initiates BER of thymine dimers in DNA, reveal a different mechanism for detecting and removing these cytotoxic UV-generated lesions. T4 Endonuclease V T4 Endo V initiates repair of cis-syn CPD lesions commonly caused by UV damage to DNA. Like Endo III, this enzyme is a combined DNA glycosylase/ap lyase that attacks the C1 0 atom of the 5 0 pyrimidine in the CPD with an activated primary amine nucleophile, forming a covalent Schiff s base intermediate. Decomposition of the intermediate by -elimination causes DNA strand scission at the phosphate between the pyrimidines of the CPD. Elegant biochemical experiments (18, 82, 83) proved that the primary amino group in

18 118 MOL ET AL the Endo V reaction mechanism is not a lysine residue, as is the case with Endo III, but is the amino group of the N-terminal Thr residue. Crystal structures of free T4 Endo V (61, 62) and of an inactive enzyme mutant in complex with a CPD-containing DNA substrate (93) indicate that the protein structure is preformed for binding its DNA substrate and does not change upon DNA binding. This 138-residue enzyme is comprised of three packed helices, which all run roughly parallel to the DNA helical axis in the substrate complex. The N terminus threads between two helices at the center of the molecule to reach into the active site where the N-terminal amino group acts as the active-site nucleophile (Figure 9). Endo V interacts extensively with the phosphodiester backbone on both DNA strands: 3 0 of the CPD with the CPD-containing strand and 5 0 of the CPD with the complementary strand. These interactions force the DNA substrate to conform to the enzyme surface, kinking the DNA by 60 at the CPD site. Nucleotide flipping by Endo V is unique in that the undamaged adenine opposite the 5 0 pyrimidine of the CPD is flipped and not the CPD lesion itself. Flipping of the undamaged adenine from the DNA allows protein side-chain and main-chain atoms to insert into the DNA to perform glycosylase and AP lyase reactions on the intrahelical CPD substrate (Figure 9). The flipped-out adenine is accommodated in a hydrophobic enzyme pocket but does not form specific interactions with the enzyme, and no specificity is observed for the DNA base opposite the 3 0 pyrimidine of the CPD (93). Thus, nucleotide-flipping by Endo V does not place the target DNA substrate in an enzyme pocket but Figure 9 Stereo view of the unique Endo V recognition of cyclobutane pyrimidine dimers (CPDs). Recognition of the lesion involves both extensive backbone interactions (not shown) around the CPD site (thin dark lines) and intercalation of active-site residues (gray tubes with polar atoms as spheres) into the DNA base stack at the position of the flipped-out adenine nucleotide. The amino group of the N-terminal Thr (center left) is positioned for nucleophilic attack beneath the 5 0 pyrimidine of the CPD, which would have paired with the flipped nucleotide.

19 DNA DAMAGE RECOGNITION & REPAIR 119 rather creates a pocket in DNA that allows enzyme residues to be positioned for catalysis. DNA damage detection and recognition by Endo V may also employ aspects of the UDG pinch-push-pull mechanism. Endo V scans double-stranded DNA by one-dimensional diffusion, which is mediated by specific, charged residues on the DNA binding surface (22, 67). Specific recognition and formation of active complexes apparently involves distortion of the DNA to fit the enzyme surface. Weakened stacking interactions of the adenines opposite the CPD lesion moderately kink ( 10 ) the DNA (40, 47, 97), with full DNA kinking accompanied by the push of Endo V catalytic residues into the DNA. The specificity likely derives from the CPD-containing DNA substrate being able to accept the enzyme-induced kinking and also from the insertion of enzyme residues at the lesion site, and thus Endo V can bind AP sites in duplex DNA (46). The flipped-out adenine binding pocket has little attractive pull, but likely provides additional nonspecific binding. In Endo V, the overall bend enforced upon substrate DNA by the preformed enzyme-binding site generates local phosphate compression at the site of nucleotide flipping. For UDG, phosphate compression is generated through steric clash augmented by concerted clamping of rigid proline-rich loops, which generate the overall bend in the bound DNA. Thus, these enzymes apply different forces to the coupled phosphate compression and DNA kinking at the active site. UDG locally compresses phosphates, which globally kinks DNA; and Endo V globally kinks DNA, which locally compresses phosphates. ABASIC SITES The removal of damaged bases by DNA glycosylases creates an abasic site that is recognized and cleaved by AP endonucleases. Abasic sites also arise continuously in DNA through spontaneous depurination and depyrimidination, caused by direct attack of oxygen radicals. As the repair of damaged bases created by deamination, reactive oxygen species, alkylating agents, and ionizing radiation all proceed through AP sites, these AP sites are a central damagegeneral intermediate in BER. The efficient repair of these AP sites is critical because they are unstable and cause mutagenesis and cell death (14, 48, 51). In E. coli, the enzymes that process AP sites and remove unusual 3 0 termini, such as the, unsaturated aldehydes that result from AP lyase activity (12, 14), are Exonuclease III (Exo III), which is the major constitutive AP endonuclease, and Endonuclease IV (Endo IV), which is induced by superoxide anion generators. In humans, the AP endonuclease that is structurally homologous to Exo III is referred to by various names (APE, REF-1, HAP-1), and its expression is induced by oxidating agents (76).

20 120 MOL ET AL AP Endonuclease AP endonucleases are divalent metal-ion-dependent metalloenzymes that specifically recognize AP sites in double-stranded DNA. They cleave the DNA phosphodiester backbone, creating a deoxyribose 5 0 -phosphate and a 3 0 -hydroxyl nucleotide to prime DNA repair synthesis. The Exo III structure consists of a four-layered -sandwich fold (59) similar to the fold of bovine pancreatic DNase I (87) and highly similar to the crystal structure of human APE (29). The divalent metal-ion binding site and active site are located at one end of the two six-stranded -sheets. These three enzymes share a conserved Asp-His pair that deprotonates a water molecule for nucleophilic attack on the DNA phosphate 5 0 of the abasic site (Figure 10). The crystal structure of APE (29) builds on the established structural framework for AP endonuclease function and suggests that detection and recognition of AP sites in DNA may also proceed through a nucleotide-flipping mechanism. Exo III and APE have extensive protruding loops on their DNA-binding surfaces that likely bind the DNA minor and major grooves (60). Crystal structures of DNase I DNA complexes (45, 98) show the enzyme binding the DNA minor groove 5 0 of the active site, but the DNA does not bind to the regions corresponding to the AP endonuclease loops. An APE-DNA model complex, therefore, suggests that the role of these protruding loops may be to induce structural distortions that prompt the abasic site to flip out (29). APE in solution binds 3 base pairs 3 0 and 4 base pairs 5 0 of the AP site, and the phosphate Figure 10 Stereo view of the Exo III active site and model for binding of a flipped-out abasic sugar. The flipped-out AP site (thin dark lines) can bind in the enzyme active site (gray tubes with polar atoms as spheres) with the scissile 5 0 phosphate positioned by the Mg 2+ ion bound to a conserved Glu side chain (lower right). A His-Asp pair (lower left) is implicated by mutagenesis and homology in the activation of a nucleophilic water for attack on the 5 0 phosphate.

21 DNA DAMAGE RECOGNITION & REPAIR to the AP site is readily cleaved by chemical agents (99, 100). These data support a model for APE binding to flipped-out AP sites, and the proposed abasic sugar-binding pocket has conserved hydrogen bonding groups able to interact with both the - and -anomers of an abasic site (Figure 10). APE is known to interact directly with the DNA repair polymerase (Pol ) (3), and Pol interacts with DNA ligase (8). This linkage of the damage-general BER enzymes, possibly within a multiprotein complex (74) that may involve the scaffolding protein XRCC1 (8, 41), is an efficient way to convey reaction products to the next enzyme in the BER pathway. A possible means of coupling these damage-general steps of BER to the base damage-specific stage involves similar minor groove interactions and binding to a flipped-out AP site by the DNA glycosylases and AP endonucleases. A growing body of experimental data supports this hypothesis. The UDG:DNA (69, 85) and MUG:DNA (1) co-crystal structures demonstrate binding of these glycosylases to flippedout AP sites, and the UDG DNA-binding kinetics (69) show that UDG binds AP sites with nanomolar affinity. AP sites in double-stranded DNA inhibit a number of DNA glycosylases, including UDG (21), MutY (7), and AlkA (81). Thus, binding to and inhibition by flipped-out AP sites may be inherent to all nucleotide-flipping enzymes. As an efficient means of initiating and coordinating AP site processing, the damage-specific glycosylases may remain bound to, or rapidly rebind, AP sites, thereby marking them for repair completion. Interestingly, UDG activity on double-stranded substrate DNA is increased by AP endonuclease, likely through a mechanism that induces UDG dissociation from AP sites (69). Coupling of the damage-specific and damage-general stages of BER through competition for binding to AP sites suggests that scanning of the DNA minor groove, nucleotide flipping, and binding of extrahelical AP sites is of paramount importance in the detection of base damage for BER (Figure 11). Aberrations resulting from base damage, such as syn base conformations or wobble mispairs, may be more easily detected from the DNA minor groove, which is more structurally uniform than the major groove. Nucleotide flipping combines specific damaged base detection with general AP site recognition and effectively couples the numerous damage-specific DNA glycosylases to the common AP endonuclease through the central abasic site intermediate. FUTURE DIRECTIONS The detection of single-site DNA base damage in the context of a complete genome remains poorly understood and yet is one of the most important processes for DNA excision repair proteins. The dynamic nature of the transient interactions of DNA repair enzymes with DNA makes these states challenging

22 122 MOL ET AL