5.2. GENES AND ENZYMES FOR THE REMOVAL OF OXIDATIVE DNA BASE DAMAGE
The main repair mechanism for the handling of oxidative DNA damage is the base excision repair pathway (BER). The initial step of BER is catalyzed by a DNA glycosylase that distinguishes damaged from normal base residues in DNA and removes the modified base in a free form (for recent reviews, see 20-22). Base release is followed by recognition and incision of the abasic site by an AP-endonuclease or an AP-lyase making breaks 5' or 3' to the damaged residues, respectively. Several DNA glycosylases have an associated AP-lyase activity and are hence referred to as bifunctional. The remaining deoxyribosephosphate residue is removed by phosphodiesterase(s) and the DNA strand continuity restored by the sequential action of a DNA polymerase and DNA ligase. Several different enzymes from human cells have been implicated in AP-site repair e.g. the AP-endonuclease HAP1/APE (for review see 23), DNA polymerase beta (for review see 24).
DNA polymerase sigma (25), Flap endonuclease/EndonucleaseIV (FEN1; 7), XRCC1 (27), DNA ligase III (27, 28), and poly (ADP-ribose) polymerase (29). However, the present communication will deal primarily with the action of the DNA glycosylases responsible for the removal of the oxidative DNA base damage.
5.2.1. GENES FOR DNA GLYCOSYLASES IN E. coli AND HUMAN CELLS
Most of what we know about the basic properties of different DNA glycosylases comes from pilot studies of E. coli. In this organism, 8 different genes have been identified encoding DNA glycosylases with distinct or overlapping functions in the removal of different types of damaged base residues. One is specific for the repair of alkylation damage (Tag, 30, 31), one for uracil (Ung; 32, 33), two for the mismatches A/G (MutY; 34) and T/G (or U/G; Mug: 35), three for the removal of oxidative DNA damage (Endonuclease Ill/Nth ; 36, 37, Formamidopyrimidine/ 8-oxoguanine DNA glycosylase/Fpg ; 38, 39, and Endonuclease VIII/Nei ; 40) and one with a broad specificity for base residues of diverse structure, including alkylated residues, methyloxidized thymines, mustard damage, ethenopurines and hypoxanthine (AlkA; 31, 41, see also 42 for refs.). Functional counterparts of most if not all of these enzymes have more recently been identified in eukaryotic organisms.
However, in mammalian cells only 5 genes with proven glycosylase function have been identified to date, comprising those for the uracil DNA glycosylase UNG (43), the human counterpart of AlkA ANPG/AAG/MPG (44-46), the T/G mismatch enzyme TDG (47), the homologue of the bacterial endonucleaseIII Hnth1 (48, 49), and the functional homologue of Fpg hOGG1/hOGHl (50-55). In addition, a homologue of the bacterial mutY (A/G mismatch enzyme) has also been identified (56) although the glycosylase function remains to be confirmed by biochemical analysis. It is to be expected that more genes for DNA glycosylases will be identified in the near future. The avenue for the identification and cloning of the glycosylase genes have varied and the approaches include protein purification and sequencing (UNG), functional complementation of E. coli alkylation sensitive mutants (ANPG), and homology searches of expressed sequence tags with known sequences from bacteria or yeast.
Still another path for molecular cloning has been the trapping of DNA glycosylase molecules covalently bound to the DNA as reaction intermediates in the DNA glycosylase/AP-lyase reaction (49, 53). This approach is valid for the bifunctional DNA glycosylases with associated AP-lyase activity that introduce strand cleavages 3' to the AP-site by beta-elimination. During beta-elimination a Schiff's base intermediate is formed that can be trapped by borhydride reduction leaving the enzyme covalently bound to the DNA (57). Such intermediates can be isolated in a pure form by relatively simple means and the protein moiety can be analyzed for aminoacid sequence for production of oligonucleotide screening probes. It has become increasingly clear that endogenous cellular agents play a major role in the formation of DNA damage and that reactive oxygen species are of particular importance in this respect (58).
Repair systems dealing with the major forms of oxidative DNA damage are therefore likely to be essential for genome maintenance including avoidance of mutations, cancer and aging. Base excision repair mechanisms for oxidized base residues are likely to be key functions in this respect and studies with transgenic null-mutant mice have shown that a number of genes involved in the base excision repair pathway are essential functions in mammalian cells, including the major AP-endonuclease, XRCC1 and DNA polymerase-beta. On the other hand, the only glycosylase deficient mouse system studied in some detail so far, lacking ANPG, appears to have normal viability and is so far apparently without any phenotype (59, 60) except for an increase in the mutation frequency of peripheral T-lymphocytes in response to methylmethanesulfonate challenge (61), suggesting that back up functions exist that can replace the defect of ANPG in repair.
More detailed analysis of the biochemical properties of different DNA glycosylases have indicated an overlap in the specificity and we shall probably have to await the construction of double mutant knock-outs to assess the physiological importance of DNA glycosylases. In view of the common mechanism of DNA glycosylases for the cleavage of N-glycosylic bonds it might have been expected to find certain sequence similarities between the various enzymes of this class. However, the various types of DNA glycosylases appear to be rather different in primary structure with the exception of the EndoIII family and the A/G mismatch repair glycosylases that can be aligned throughout the entire sequences (34). Both of these contain an iron-sulfur cluster (4Fe-4S) involved in DNA binding. Nevertheless, a short sequence motif of about 20 amino acids termed the Helix-hairpin-Helix (HhH) motif has been identified in several different DNA glycosylases (20, 62).
The HhH motif includes conserved residues characteristic for each group of enzymes with similar specificity. The structure of this motif has become apparent from the crystal structures of EndonucleaseIII (62) and AlkA (63, 64) and from that of DNA polymerase beta, which has a similar motif in spite of not having any associated DNA glycosylase activity (24).
5.2.2. ENDONUCLEASE III AND HOMOLOGUES IN EUKARYOTIC CELLS
Endonuclease III (Nth) from E. coli was originally discovered as an endonuclease producing strand breaks in heavily UV-irradiated DNA (65). The break formation, at UV-lesions different from pyrimidine dimers, was later shown to be caused by the consecutive action of the DNA glycosylase and the AP-lyase reactions catalyzed by the same protein molecule. Endonuclease III has been shown to remove a wide variety of different oxidized pyrimidine residues including thymine glycol, urea, dihydroxyuracil, cytosine hydrates, dihydrothymine and hydroxycytosine (36, 37, 40). The mutant lacking Endonuclease III (nth) is largely without any detectable phenotype in terms of hypersensitivity to DNA damaging agents or showing an increased mutation frequency to such agents (40).
This could have been interpreted as Nth being without importance in repair, however the recent discovery of an alternative DNA glycosylase with similar specificity, endonuclease VIII, has proven the existence of alternative enzymes with back-up functions for one another.
The double mutant nth nei is hypersensitive to oxidizing agents such as hydrogen peroxide and show increased spontaneous mutation frequency relative to either single mutant or wild type (40). EndoVIII apparently does not belong to the helix-hairpin helix enzymes and is similar to the E. coli formamidopyrimidine DNA glycosylase Fpg in structure. With the known sequence from E. coli it has been possible to identify sequence homologues of EndoIII in yeast and human cells (48, 49, 66 69; see Table 5.I). Eukaryotic enzymes with properties similar to EndoIII had been reported previously but extensive purification was hampered by a low abundance in the eukaryotic cells. However, with the coding sequences available, the active eukaryotic enzymes have now been expressed in E. coli and purified and characterized extensively.
One major difference in the substrate specificity of the eukaryotic sequences relative to Endo III is their high activity for the removal of formamidopyrimidines (67, 69).
Two different homologues of EndoIII exist in yeast, Ntgl and Ntg2 (68, 69). Ntgl is lacking the iron-sulfur domain characteristic of all the other members of the EndoIII familiy. Mutant analysis has shown that both enzymes are required for repair of oxidative DNA damage as judged from the additive effect of each single mutation on the spontaneous and the hydrogenperoxide induced mutagenesis in S. cerevisiae (69). The need for two distinct enzymes can be explained by differences in the reaction mode and in the substrate specificity. Whereas Ntgl is quite effective on cytosine hydrates in UV-irradiated DNA, it appears rather inefficient in the removal of hydroxycytosines. On the other hand Ntg2 cannot remove all types UV-induced cytosine lesions and is more efficient towards hydroxycytosine removal. It could also be that Ntgl and Ntg2 participate differently in global repair and transcription coupled removal of thymine glycols. Both enzymes have high affinity against thymine glycols.
Interestingly, Ntgl appears to be sorted both to the mitochondria and to the nucleus whereas Ntg2 is localized exclusively to the nucleus as judged from expression analysis of GFP fusion proteins.
The mammalian EndoIII has also been studied in some detail and appears to remove all the ring fragmented and nonaromatic modifications of pyrimidines similarly as the E. coli enzyme. In addition, the human enzyme also appears to be quite active against formamidopyrimidines (67). Northern analysis has shown an even distribution of hNTH1 transcript in all tissues examined. Fusions to GFP have indicated sorting both to the mitochondria and to the nucleus and both mitochondrial and nuclear localization signals are found in the N-terminal part of the protein sequence (70). The human gene appears to have 5 different exons and is regulated in a cell cycle dependent manner with increased expression during early S-phase (70).
5.2.3. THE HUMAN ENZYME FOR 8-OXOGUANIN REMOVAL
In E. coli, the primary oxidation product 8-oxoguanine is removed by the DNA glycosylase encoded by the fpg/mutM gene. The enzyme was first recognized for its ability to remove imidazole ring-opened residues, formamidopyrimidines (38). Fpg/MutM is one of three gene functions in E. coli that participates in the mutation avoidance system preventing mutagenesis resulting from the formation of 8-oxoguanine, the others being MutY and MutT. Mutations in either one of these loci result in higher spontaneous mutation frequencies than normal and the effects are particularly pronounced in the double mutant mutM mutY or the triple mutant defective in all three-gene functions. MutY is a DNA glycosylase for removal of adenines from A/8-oxoG (or A/G) mispairs, thus preventing mutations arising from the frequent mispairing of template 8-oxoguanine with adenine during replication (34). MutT is a GTPase with special affinity for 8-oxo-dGTP (producing 8-oxo-dGMP),
thus avoiding 8-oxoguanine from being misincorporated in the newly replicated DNA strand (71). Counterparts of both MutT and MutY have been recognized in cDNA libraries from human cells by sequence analysis of expressed sequence tags (56, 72).
However, no sequence homologue of MutM/Fpg has been found except for a single EST sequence that represents a prokaryotic contaminant in a cDNA library and the Genbank database (Bjoras & Seeberg, unpublished). It was therefore a major advance in the understanding of the eukaryotic 8-oxoguanine repair system when a functional yeast homologue of Fpg was cloned by its ability to suppress the spontaneous mutation frequency of the E. coli fpg mutY mutant (73, 74). The yeast gene thus isolated, OGG1, appeared to be substantially different from Fpg in sequence, however, similar to other DNA glycosylases of the helix-hairpin-helix family. With the yeast sequence at hand it was a simple matter to identify the human or mammalian homologues and the cloning of cDNAs expressing the human 8-oxoguanine DNA glycosylase was reported simultaneously by several different laboratories (50-55).
In most studies the human enzyme has been termed hOGG1, whereas others have used hMMH (human mutM homologue; 55) or hOGHl (human Ogg1 Homologue; 50). Analysis of the gene structure of the human hOGHl gene has revealed eight different exons of which exon 7 and 8 appears to be present in alternatively spliced forms that is sorted to the nucleus or the mitochondria, respectively (70). Similarly as for hNTHI, the hOGHl appears to be cell cycle regulated with increased expression during early S-phase. The hOGHl gene is only weakly expressed and it is difficult to monitor any transcript in organ tissues by Northern analysis. However, two different transcripts are apparent from the upregulation during early S-phase of synchronized cell cultures. We have characterized the enzymatic properties of the human enzyme using highly purified preparations obtained after expression in E. coli (50).
In comparing the human enzyme hOGHl with the Fpg enzyme from E. coli, some important differences were observed. Firstly, the human enzyme has a DNA glycosyhisc activity that is uncoupled from its associated AP-lyase activity, implying that 8-oxoguanine opposite adenine is removed with a relatively high efficiency although without subsequent introduction of strand breaks. After base excision the enzyme remains associated with the AP-site and probably prevents other enzymes from processing the AP-site. The second observation was that the associated AP-lyase activity is specific for an orphan C in the opposite strand regardless of the prior removal of an 8-oxoguanine.
Fig. 5.2. Outline of the 8-oxoguanine mutation avoidance in human cells.
This means that repair will only be completed if 8-oxoguanine is situated across C in the opposite strand and ensures that the excision repair process will be completed without the formation of mutation (Fig. 5.2). In the case that excision occurs from an 8-oxoguanine/A pair it will be essential for the repair machinery to identify the parental strand as opposed to the newly synthesized strand, to make the right decision on how repair should be completed. If 8-oxoguanine has been inserted across A in the newly replicated strand, the AP-site should be removed and replaced with T as in the ordinary base excision repair process. If A is misincorporated, then a second duplex is required to restore G in place of the AP-site by some type of postreplication repair. This model might indicate a possible link between mismatch repair and base excision repair for the purpose of carrying out proper strand identification.
In view of the rather close similarity between E. coli and human cells with respect to sequence/and or functional homologues of all three genes of the mutation avoidance system it is somewhat surprising that obvious homologues of MutT or MutY have not been identified in S. cerevisiae. The entire genome sequence is available, but no obvious sequence homologue of MutY or MutT can be recognized. It is still possible that functional homologues exist which cannot be easily identified from sequence homology searches, although other interpretations are also possible.
5.2.4. BROAD SPECIFICITY DNA GLYCOSYLASES IN THE REPAIR OF OXIDATIVE DNA DAMAGE
Among the first DNA glycosylases to be characterized in any detail were those for the removal of N-alkylated base residues. Strong methylating agents such as methylmethanesulfonate or dimethylsulfate introduce different lesions among which N-3-methyladenine and N-7-methylguanine are the major products. E. coli possesses two DNA glycosylases for the removal of N-alkylated residues, i.e. Tag, which is specific for 3-methyladenine removal, and AlkA, which has a broad specificity against many different types of alkylation damage as well as several other rather diverse types of base modifications (31). These also include important oxidative modifications such as the thymine-derived products 5-formyluracil and 5-hydroxymethyluracil (75). Enzymes analogous to AlkA have been identified in all organisms investigated and comprise Mag in S. cerevisiae (76, 77) and ANPG/MPG in the mammalian cells (44-46; see Fig 5.3).
In an attempt to define the discrimination made by AlkA for damaged versus non-damaged residues we have made a comparison between the stability of the base-sugar attachment and enzyme activity (42). In doing so, we find that to the extent that we can judge from the available data, there is a qualitative correlation between the base-sugar bond instability and the efficiency with which the bases are being released. One of the best substrates for AlkA, N-3-methyladenine, has a half life in the DNA of only 25 h. N-7-methylguanine, which has a halflife of about 150 h, is still a good substrate for AlkA but is more slowly removed than N-3-methyladenine. Similarly, AlkA removes 5-formyluracil much more efficiently than 5- hydroxymethyluracil, again corresponding to the relative stability of the base attachment. Similar arguments can also be made for ethenopurines versus ethenocytosines and hypoxanthine versus guanine.
In view of the fact that most modifications of the base will destabilize the sugar attachment, AlkA and analogues may represent a general type of base excision repair system functioning by lowering the activation energy for the N-glycosylic bonds largely without specificity for the base involved. Excision efficiency will then be a function of the stability of the base attachment to the DNA. One implication of such a model is that because of the dynamic stability of the DNA, it might be expected that even normal bases will be removed with a certain frequency. This has been tested experimentally for AlkA from E. coli, Mag from S. cerevisiae and ANPG from human cells measuring guanine release (42). All of these enzymes release guanine with biologically significant frequencies and the excision appears to be strongly dependent on pH in a narrow range (6.5-8), almost proportional to the proton concentration.
Protonation of purines is known to cause destabilization of the N-glycosylic bond and the pH effect is therefore supporting the bond destabilization model. Other DNA glycosylases such as the narrow specificity alkylation repair enzyme Tag from E. coli, and the uracil DNA glycosylase from human cells have tested negative for normal base removal as also would be expected since these enzymes presumably cannot adapt a normal base in its active site. Another implication of the model is that normal base removal also might be occurring at a significant rate in vivo. This was tested by measuring the spontaneous mutation frequency in E. coli cells transformed by different alkA gene constructs producing different levels of AlkA overexpression. There was a significant increase in spontaneous mutation frequency correlated to the amount of enzyme overproduction. This can be explained by normal base removal yielding AP-sites as potential premutagenic lesions.
Studies with Mag overproduction in yeast have yielded a similar effect, which was particularly pronounced in a yeast mutant lacking AP-endonuclease activity (78). As more genome sequence data has become available with many smaller genomes being completed predictions can be made as to the extent and evolution of DNA glycosylases. Helix-hairpin helix enzymes appear to be widely distributed in all organisms thus far analyzed. One might speculate that the origin of such enzymes might represent broad specificity species with rather high activity for normal base residues in DNA. Such enzymes might be useful for organisms with smaller genomes for which a fairly high mutation frequency would be tolerated and could even be an advantage. As genomes became larger, the activity of the promiscuous base removing enzymes had to be reduced and other more specialized forms have evolved, e.g. those for the removal of 8-oxoguanine.
Since some frequent oxidative lesions like 8-oxoguanine have a rather stable attachment to the phosphodiester chain it became necessary to develop a stronger base release mechanism which resulted in the development of the bifunctional enzymes with associated AP-lyase activity and narrow specificity. This view might also elude to why the broad specificity enzymes are monofunctional rather than bifunctional. If the broad specificity enzymes were to produce strand breaks directly into DNA one would expect to get frequent introduction of double-strand breaks, which would be detrimental to the cells. The broad specificity enzymes without associated AP-lyase activities have a very high affinity for AP-sites in the DNA and probably remain attached to the abasic sites until it can be further processed by the subsequent enzymes in the base excision repair pathway.
This may be an important way to avoid excessive action and formation of unprotected AP-sites that are highly reactive and represent strong premutagenic lesions in DNA.
Fig. 5.3. DNA glycosylases for the removal of oxidative DNA base damage in E. coli, yeast and human cells. Enzyme symbols in bold all belong to the helix-hairpin-helix family of enzymes. The choice of the enzyme/gene symbol NTG in yeast rather than NTH was to avoid confusion with a previously assigned gene symbol of NTH for neutral trehalase in yeast. The gene symbol hOGHl used by Bjoras et al. (50) for the human 8-oxoguanine DNA glycosylase stands for human Ogg1 homologue by analogy with the nomenclature of the MSH/MLH proteins in human cells.