| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Oregon State University, Department of Biochemistry and Biophysics, Corvallis, Oregon, USA
1 Correspondence: Department of Biochemistry and Biophysics, Oregon State University, 2011 Agricultural & Life Sciences Bldg., Corvallis, OR 97331-7305, USA. E-mail: mathewsc{at}onid.orst.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONNOVEL ASPECTS OF dNTP...dNTP POOL ASYMMETRIES AS...dNTP METABOLISM AND...dNTP METABOLISM AND MUTAGENESIS...CHEMICAL MODIFICATION OF...RELATIONSHIPS BETWEEN dNTP...REFERENCES |
|---|
Key Words: mutagenesis • deoxyribonucleotides • nucleotide pools • ribonucleotide reductase • mitochondria • oxidative DNA damage
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONNOVEL ASPECTS OF dNTP...dNTP POOL ASYMMETRIES AS...dNTP METABOLISM AND...dNTP METABOLISM AND MUTAGENESIS...CHEMICAL MODIFICATION OF...RELATIONSHIPS BETWEEN dNTP...REFERENCES |
|---|
. Through studies of DNA synthesis in vitro, primarily during the 1980s, the most significant mutagenic mechanisms were shown to be replication errors caused by 1) misinsertion, in which a non-Watson-Crick base pair is formed as a result of a dNTP excess or deficiency, or 2) inhibition of proofreading (next-nucleotide effect), in which a dNTP in excess drives DNA chain extension past the site of a mismatch before the polymerase-associated editing nuclease can correct the error. Later it was shown that dNTP excess can also force frameshift mutations by forming a correct base pair at a slipped or dislocated 3' terminus (2)
.
A major question unanswered in 1994, when this field was last reviewed comprehensively (1)
, is whether natural dNTP pool asymmetries can be mutagenic, i.e., are intracellular dNTP pools regulated so as to maximize replication accuracy? An answer to that question has come through recent work on human mitochondrial diseases, in which accumulation of mitochondrial gene mutations results from several defined abnormalities of nuclear or mitochondrial dNTP metabolism (3)
. Other recent developments include discovery of novel mechanisms in dNTP metabolic regulation, including, in mammalian cells, a p53-dependent form of ribonucleotide reductase (4)
, and in yeast a DNA damage-induced translocation of a ribonucleotide reductase subunit across the nuclear membrane (5)
. In addition, recent work indicates that the selection of eukaryotic origins from which replication initiates is influenced by intracellular nt concentrations (6)
. Moreover, our laboratory recently presented evidence that balanced, as well as unbalanced, dNTP accumulations are mutagenic, probably due to indirect inhibition of proofreading (7)
, and this raises the question of whether loss of control leading to abnormal dNTP accumulation might be one of the events leading to a mutator phenotype, postulated to be involved in carcinogenesis (8)
. Other contemporary research focuses on the concept of damaged nucleotides as premutagenic lesions, i.e., the idea that mutagenesis by reactive oxygen species (ROS) proceeds in part via oxidation of free nucleotides followed by their incorporation into DNA (9)
.
As background to our discussion, Fig. 1
summarizes reactions and major allosteric control mechanisms in dNTP biosynthesis in mammalian cells. In addition to the regulation of biosynthetic pathways, there is good evidence that dNTP pools are also controlled at the concentration of nucleotide catabolism (10)
, with hydrolysis of deoxyribonucleoside monophosphates being the key regulated step.
|
Although Fig. 1
does not show salvage pathways leading to dNTPs, their significance must not be overlooked. Many mutagens are nucleobase or nucleoside analogs whose actions require their conversion, via salvage pathways and nucleotide kinases, to the respective dNTPs. For example, 2-aminopurine is anabolized via adenine phosphoribosyltransferase and 5-bromodeoxyuridine by thymidine kinase. Thus, a change in intracellular activities of one or more salvage enzymes will alter the sensitivity of those cells to the genotoxicity of the respective analog. Numerous examples are tabulated in ref 1
. Two additional studies are mentioned here. Meuth (11)
reported that a Chinese hamster ovary cell line lacking deoxycytidine kinase activity became sensitized to the mutagenic and cytotoxic actions of DNA-alkylating agents. These cells contain greatly increased dTTP pools and decreased dCTP, conditions shown in previous studies to sensitize cells to alkylating agents. Although the basis for this sensitization is not known, the significance of this study is its demonstration of a function for this salvage enzyme in maintaining normal dNTP pools even when salvage substrates are not present; in this study the culture medium contained dialyzed serum, so there was no ready source of nucleosides. To my knowledge, this function has not yet been identified.
A more recent report (12)
involves attempts by Bouzon and Marlière to generate organisms that produce DNA of novel base composition. To this end, they cloned the human gene for deoxycytidine kinase into Escherichia coli and found that it becomes a "conditional mutator," meaning that cells expressing the mutant enzyme readily incorporate many nucleoside analogs into their DNA at high frequency. E. coli lacks deoxycytidine kinase, and the human enzyme has broad specificity for both pyrimidine and purine nucleoside analogs.
|
NOVEL ASPECTS OF dNTP POOL REGULATION |
|---|
|
TOPABSTRACTINTRODUCTIONNOVEL ASPECTS OF dNTP...dNTP POOL ASYMMETRIES AS...dNTP METABOLISM AND...dNTP METABOLISM AND MUTAGENESIS...CHEMICAL MODIFICATION OF...RELATIONSHIPS BETWEEN dNTP...REFERENCES |
|---|
, 13)
show that mutations affecting allosteric control sites lead both to unbalanced dNTP pools and to increases in the spontaneous mutation rate. Work from our laboratory (14)
suggests ribonucleoside diphosphate (rNDP) substrate concentrations as another significant regulatory parameter, because all four rNDPs compete for binding to the same catalytic RNR site. When mouse RNR in vitro was provided a mixture of rNDP substrates and nucleoside triphosphate effectors, all at their estimated intracellular concentrations, the four deoxyribonucleoside diphosphate products (dNDPs) were synthesized at rates corresponding to the proportions of the four deoxyribonucleotides in the mouse genome. On the other hand, an equimolar rNDP mixture led to unbalanced rates of dNDP synthesis.
It is well established that intracellular RNR activity is cell cycle dependent, with activities rising at the G1/S border and contributing to dNTP pool expansions during S phase. The principal target for regulation is the small protein (R2). Eukaryotic RNR is a heterotetramer (10)
consisting of a large protein, R1, and a small protein, R2, with subunit compositions 
2 and ß2, respectively (although the subunit composition of R1 has not been established absolutely, ref 15
). Whereas R1 levels remain relatively constant and high, through the mammalian cell cycle, R2 fluctuates cyclically, with control exercised at the levels both of protein synthesis and degradation; thus, regulation of activity is exerted by control of R2 levels. The principal transcriptional effect is repression, brought about through action of the E2F factor, which controls a number of cell cycle-specific genes. Chabes et al. (16)
have described an activator region, upstream from the mouse R2 gene, that is involved in release from repression at the beginning of S phase. When cells complete S phase, R2 levels are diminished by selective protein degradation. Chabes et al. (17)
have described the process in mouse cells. Late in S phase a cell cycle regulator called Cdh1 undergoes phosphorylation, which leads to its release from a ubiquitin ligase called APC (anaphase-promoting complex). APC recognizes a specific sequence on the R2 protein, leading to ubiquitination and subsequent proteasomal degradation of the R2 protein. This effectively shuts off dNTP synthesis in G2, although residual dNTP pools remain, as shown in most studies of synchronized cells. Ke et al. (18)
have shown that APC controls thymidine kinase and thymidylate kinase similarly, and that when these control mechanisms are shut off, dTTP accumulates in unbalanced fashion, causing an increase in spontaneous mutations. Thus, the post-S-phase shutoff of dNTP synthesis evidently plays a role in maintaining optimal genomic stability.
Other genetic regulatory mechanisms control dNTP pools, although the specific mechanisms are not yet fully understood. For example, the RB tumor suppressor gene blocks cell cycle progression, one of its effects being to down-regulate enzymes of dNTP synthesis (19)
.
When DNA-damaging events occur outside of S phase, dNTPs must be available for DNA repair. As shown by Tanaka et al. (4)
, mammalian cells respond by synthesizing a novel R2 protein under the control of the tumor suppressor p53. This novel protein, p53R2, associates with R1 to form an active ribonucleotide reductase (4
, 20)
. Somewhat unexpectedly, however, Håkansson et al. (21)
reported that expression of this protein in response to DNA damage in mammalian cells does not lead to major dNTP pool expansion. In this respect mammalian cells differ from yeast (see below). This finding suggests that the role of p53R2 is more complex than originally thought, and Håkansson et al. (21)
suggest functions in basal DNA repair processes and in maintaining dNTP pools within mitochondria (see below).
A comparable situation exists in yeast, Saccharomyces cerevisiae, which presents some interesting parallels and distinctions in comparison with mammalian cells. Yeast contains four RNR proteins, called Rnr1 and Rnr3 ( subunits), and Rnr2 and Rnr4 (ß subunits). The standard R1 protein in yeast is thought to be an Rnr1 homodimer. Rnr3 is a large subunit protein that is normally present at low levels, but whose accumulation is activated by genotoxic stress (22)
, generating an alternative R1 protein. The small RNR protein is actually a heterodimer containing one molecule each of Rnr2 and Rnr4 (23)
. Cell cycle regulation is fairly well understood and involves the checkpoint kinases Mec1, Rad53, and Dun1, all of which have homologs in higher cells. The Mec1/Rad53/Dun1 pathway is activated during S phase and in response to DNA damage, with a key regulated parameter being the concentration of a Sml1, a protein inhibitor of RNR that binds to the large subunit (24)
. During a normal cell cycle, Rnr1 and Rnr3 are localized predominantly in the cytoplasm, whereas Rnr2 and Rnr4 have a primarily nuclear localization. DNA-damaging events activate RNR activity within the cell by stimulating translocation of the Rnr2 and Rnr4 proteins from nucleus to cytoplasm (5)
. Proteins involved in anchoring R2 within the nucleus have recently been identified (25
, 26)
.
Of considerable interest is a recent report (27)
that activation of the Mec1/Rad53 checkpoint pathway doubles the mitochondrial DNA copy number in yeast. Whether this results from increased dNTP supplies within the mitochondrion as a result of RNR activation remains to be determined, but this study describes the first known involvement of a conserved signaling pathway in regulation of mitochondrial DNA copy number.
As seen in mammalian cells, dNTP levels rise as yeast cells enter S phase. Although one might expect dNTP accumulation and replication initiation to be jointly controlled, Koç et al. (28)
showed that in synchronized cells each process occurs independently of the other. dNTPs also accumulate in yeast cells as the result of DNA-damaging events—by some 6- to 8-fold, as shown by Chabes et al. (29)
. An accumulation this great was somewhat unexpected because of the complex allosteric regulation of RNR by nucleotides, particularly the inhibition of all four activities by dATP. Chabes et al. found that when the dATP-dependent allosteric site (the "activity site") was inactivated by site-directed mutagenesis, dNTP levels rose an additional 2-fold, with a consequent increase in the mutation rate. The authors proposed that yeast RNR has a "relaxed feedback inhibition" of RNR, which allows sufficient accumulation of dNTPs to permit efficient DNA repair but prevents the additional accumulation that could be genotoxic.
Could RNR transport through the nuclear envelope be involved in regulating dNTP synthesis in mammalian cells? There have been reports over the years of RNR translocation into the nucleus, but most analyses of subcellular localization show RNR and other enzymes of dNTP synthesis to be localized in the cytosol. Recently Liu et al. (30)
reported that in a human tumor cell line both R2 and p53R2 proteins undergo translocation from cytosol to nucleus coincident with activation of DNA synthesis, and they proposed that this translocation could be a regulatory mechanism activating dNTP biosynthesis. Because comparable experiments were not reported on R1 localization, it is not yet clear whether or not RNR activation does result from R2 nuclear translocation in mammalian cells.
|
dNTP POOL ASYMMETRIES AS MUTAGENIC DETERMINANTS |
|---|
|
TOPABSTRACTINTRODUCTIONNOVEL ASPECTS OF dNTP...dNTP POOL ASYMMETRIES AS...dNTP METABOLISM AND...dNTP METABOLISM AND MUTAGENESIS...CHEMICAL MODIFICATION OF...RELATIONSHIPS BETWEEN dNTP...REFERENCES |
|---|
), the ability of a mismatched nucleotide to force a replication error, whether through misinsertion or a next-nucleotide effect, is clearly a function of the concentrations of the correct and incorrect dNTPs. Therefore, over evolutionary time we might expect the DNA base composition of an organism to reflect the composition of its dNTP pool at replication sites. In fact, in nearly all organisms analyzed, dNTP pools are highly asymmetric. The most striking general feature is underrepresentation of dGTP, which usually comprises just 5 to 10% of the total dNTP pool (31)
. For example, we estimated the concentrations of the four dNTPs in synchronized S-phase HeLa cells to be dATP, 60 µM; dTTP, 60 µM; dCTP, 30 µM; dGTP, 10 µM, with dGTP making up just 6% of the total (32)
. Although estimates of this type do not take into account possible compartmentation effects, they do provide a useful first approximation.
Genetic consequences of natural dNTP pool asymmetry
It is generally thought that cells are organized to maximize DNA replication accuracy, because spontaneous mutation rates are sometimes lower than expected from the fidelity of replication complexes, even when mismatch repair is taken into account (33)
. However, there is reason to speculate that cells and organisms strike a balance between replication velocity and accuracy, allowing rapid reproduction while at the same time allowing cells to respond to stress conditions, and ultimately permitting desirable evolutionary variation (34)
. Might dNTP pool asymmetries contribute by increasing in vivo replication error rates relative to what would be seen if the pools were balanced? In our laboratory, Stella Martomo (35)
approached this question by running replication reactions in vitro in the presence of dNTPs either at their estimated in vivo concentrations or equimolar. She found that a 3- to 4-fold underrepresentation of dGTP, comparable to that seen in cells, was not strongly mutagenic. However, if dGTP was increased she found the error rate to increase in direct proportion to the concentration of the nucleotide in excess. This latter result probably deserves further investigation with in vitro systems. For example, Darè et al. (36)
used dCMP deaminase deficiency and addition of specific nucleosides to artificially boost pools of single dNTPs in V79 hamster cells, and they reported that dGTP accumulation in vivo is more strongly mutagenic than that of either dCTP or dTTP, a finding without an obvious explanation. Like Darè et al., we have observed a dCMP deaminase deficiency, which causes a specific accumulation of dCTP, to be nonmutagenic (unpublished results).
The limited data available, as just discussed, suggest that natural dNTP pool asymmetries are not strongly mutagenic, at least for nuclear DNA. However, the absolute values of dNTP pools may contribute to mutagenesis, even when those pools are balanced. A review of the literature, plus new data, revealed that mammalian cell lines that have undergone oncogenic transformation contain dNTP pools 3- to 4-fold larger than those of normal diploid cells (35)
. When Martomo followed DNA replication in cell-free extracts programmed with dNTPs at transformed or normal diploid levels, she found that at normal diploid levels the error frequency was indistinguishable from background levels, whereas at the higher levels characteristic of transformed cells, the error frequency exceeded background values by 2- to 5-fold. This finding suggested that balanced accumulation of dNTPs could be mutagenic. Linda Wheeler tested this in E. coli by expressing recombinant ribonucleotide reductase. As shown in Table 1
, this treatment, which caused the dNTP pools to expand by some 3- to 5-fold, caused an increase in the spontaneous mutation frequency out of proportion to the pool accumulation—up to 40-fold (7)
. We propose that this results from decreased proofreading efficiency at high dNTP levels, resulting from enhanced DNA chain extension from a mismatched 3' terminal nt. From kinetic properties of replicative DNA polymerases, it seems clear that the enzyme is saturated with dNTPs at physiological levels for DNA chain extension from a correctly matched 3' terminus, but the rate of extension from a mismatch is far below saturation; published values for different polymerases give KM values for mismatch extension in the millimolar range (7)
. Thus, as dNTP levels increase above normal values, the rate of normal elongation will not increase, but mismatch extension will increase, possibly in direct proportion to dNTP concentrations, with consequent inhibition of proofreading. These relationships are summarized in Fig. 2
.
|
|
Frameshift mutations induced by dNTP pool imbalances
What kinds of mutations are induced by dNTP imbalances? Early studies with in vitro systems, and later in cultured cell systems, suggested that most mutations were substitutions, caused by competition between correctly and incorrectly matched dNTPs for insertion. However, Bebenek et al. (2)
showed that dNTP imbalances in vitro can also stimulate frameshift mutagenesis. Their explanation was that the template-product duplex can undergo slippage, with a correct base-pairing stabilizing a slipped structure or misalignment after incorporation of a mismatched nucleotide. Both processes are shown in Fig. 3
. Bebenek et al. found this kind of replication error to occur at appreciable rates in vitro at dNTP imbalances as low as 20-fold, an imbalance that could well exist in living cells.
|
Do dNTP imbalances promote frameshift mutagenesis in vivo? Bebenek et al. presented preliminary evidence in the affirmative. Later evidence came from analysis of a mutator mutation in E. coli. Lu et al. (37)
used gene disruption to inactivate E. coli gene ndk, the structural gene for nucleoside diphosphate kinase. The mutant cells survived because they can use adenylate kinase to carry out the essential synthesis of nucleoside triphosphates from the corresponding diphosphates. Of interest is the fact that the mutant cells contain unbalanced dNTP pools—a large dCTP accumulation and smaller, but still significant, expansions of the dATP and dGTP pools. Although the biochemical basis for these pool effects is still unknown, the mutant cells have a mutator phenotype, as might be expected from these pool abnormalities. However, Roel Schaaper (personal communication) found that most of the mutations arising spontaneously in these cells were A·T
T·A transversions, an event not readily explained by the known dNTP imbalances.
Later Miller et al. (38)
carried out a similar analysis in cells containing both the ndk mutation and a mutS mutation affecting mismatch repair, so that replication errors leading to mutation, which might otherwise be corrected, could now be detected. In this background 32 of 33 sequenced mutations leading to rifampicin resistance were A·TG·C transitions, the pathway expected to predominate under the dNTP pool alterations created by the ndk mutation, whereas only one was an A·TT·A transversion. By use of strains carrrying lacZ mutations designed to revert to wild-type (WT) by single-base insertions or deletions within monotonous runs (e.g., from 6 to 7 Gs), Miller et al. could detect stimulation of frameshifts by an ndk mutation, with or without an accompanying mutS mutation. However, the important question of whether natural dNTP imbalances stimulate frameshift mutations to a significant extent relative to substitutions has not yet been definitively answered.
|
dNTP METABOLISM AND HYPERMUTAGENESIS OF RETROVIRAL GENOMES |
|---|
|
TOPABSTRACTINTRODUCTIONNOVEL ASPECTS OF dNTP...dNTP POOL ASYMMETRIES AS...dNTP METABOLISM AND...dNTP METABOLISM AND MUTAGENESIS...CHEMICAL MODIFICATION OF...RELATIONSHIPS BETWEEN dNTP...REFERENCES |
|---|
, 40)
, pointed to nt metabolism as part of the basis for the extraordinarily high mutation rate of HIV, the human immunodeficiency virus. HIV reverse transcriptase, like that of other lentiviruses, lacks a proofreading activity. The resulting infidelity of nucleic acid synthesis contributes to viral pathogenicity by creating variants that are drug resistant or can escape immune surveillance. Normally reverse transcription generates substitution errors at very high rates—0.05 to 1 per genome per replication cycle. Occasionally there is a complete breakdown of fidelity mechanisms, leading to hundreds of GA transitions within a single genome. This phenomenon of hypermutation can be reproduced by reverse transcription of viral RNA molecules in vitro (40)
, but only at dTTP to dCTP ratios of several thousand, which would favor mispairing of T with template G. Natural dNTP pool ratios are rarely seen to vary by more than 10- to 50-fold (31
, 32)
. Although one cannot rule out compartmentation effects or extreme pool bias in individual cells, it seemed wise to consider other mechanisms for hypermutation.
Such a mechanism materialized through studies of a family of nucleic acid cytidine/deoxycytidine deaminases called the Apobec family (41
, 42)
. An early observation was the finding that the mRNA for apolipoprotein B, the protein component of LDL, undergoes RNA editing by deamination of a specific cytidine residue to uridine, with subsequent change in coding. The enzyme is called Apobec1 (for APO B Editing Complex). Subsequently a related protein, AID (activation-induced deaminase), was found to participate in somatic cell hypermutation, a late phase in diversification of the repertoire of immunoglobulin molecules, as part of the immune response. Transfer of the AID gene into E. coli generated a mutator phenotype, with G·CA·T transitions favored (43)
. These recombinant bacteria contained normal levels of dNTPs, including dUTP (44)
, suggesting that if deamination of cytosine was responsible for mutagenesis, the deamination did not occur at the nucleotide concentration. Isolation of the AID gene product showed it to be a DNA-cytosine deaminase (42
, 45
, 46)
; conversion of a DNA-dCMP residue to dUMP would convert a G·C base pair to G·U, leading after replication to A·U, then to A·T.
What about HIV? A related protein, Apobec3, has been shown to be packaged with virions, where it deaminates C to U in minus-strand reverse transcripts (47
, 48
, 49)
, leading to extensive replacement of G by A in virion RNA molecules—100 or more substitutions per genome. This may represent a cellular antiviral defense mechanism, leading to massive mutagenesis with consequent inactivation of the virus. In support of this idea, the HIV genome encodes a protein called Vif (viral infectivity factor), which binds to Apobec3 and induces its degradation (50)
—a viral offense mechanism, as it were, against the cellular defense.
Although these findings indicate that HIV hypermutagenesis occurs at the macromolecular concentration rather than involving nucleotide metabolism directly, recent developments spotlight an aspect of nucleotide metabolism that deserves attention. Neuberger et al. (51)
have pointed out that somatic cell hypermutation occurs in two phases. The first phase, targeted at G·C base pairs, involves DNA-dCMP deamination by AID. The second phase is targeted at A·T base pairs and could involve polymerase errors. However, Neuberger and colleagues speculate that this later phase could involve intracellular accumulation of dUTP, with its correctly base-paired incorporation opposite template A. This in turn could trigger action of DNA-uracil N-glycosylase (Ung) as the first step in uracil-initiated base excision repair (BER). The resultant abasic site would lead to nontemplated DNA synthesis at that site. In support of this idea, Neuberger et al. point out that dUTPase, which helps to exclude uracil from DNA, is cell cycle regulated, like ribonucleotide reductase, with maximal activity in S phase. Therefore, dUTP might accumulate at other cell cycle phases and present an antiviral defense mechanism, with its incorporation into invading viral genomes contributing to lethal mutagenesis. Consistent with this idea, several lentiviruses are known to encode a dUTPase whose activity is essential for full infectivity (52)
. All of these speculations point up the importance of direct measurements of intracellular dUTP pools in synchronized cell populations, a goal that unfortunately is difficult to achieve because of the similarity of dUTP to dTTP. However, it is of interest that the HIV/hypermutagenesis story, which began with dNTP-based mutagenesis and moved toward a macromolecular mechanism, may still require considerations of nucleotide metabolism for its complete understanding.
|
dNTP METABOLISM AND MUTAGENESIS IN MITOCHONDRIAL GENOMES |
|---|
|
TOPABSTRACTINTRODUCTIONNOVEL ASPECTS OF dNTP...dNTP POOL ASYMMETRIES AS...dNTP METABOLISM AND...dNTP METABOLISM AND MUTAGENESIS...CHEMICAL MODIFICATION OF...RELATIONSHIPS BETWEEN dNTP...REFERENCES |
|---|
. To what extent might nucleotide metabolism be a factor? Second, mitochondria contain pools of dNTPs that are physically and metabolically distinct from the much larger pools supplying the nuclear genome (53)
. Third, the mitochondrial pools must be regulated quite differently from the pools supplying nuclear DNA, because mitochondrial DNA replicates throughout the cell cycle (55)
, meaning that dNTPs must be continuously available within the organelle. Fourth, among the large number of human genetic diseases known to affect mitochondrial function (3
, 56
, 57)
are several (see Table 2
) that involve abnormalities of nucleotide metabolism, and some of these diseases offer windows into relationships between dNTP synthesis and mitochondrial DNA replication. Finally, because mitochondrial gene mutations accumulate with aging (67
68
69
70
71)
, it is important to understand whether age-related changes in nucleotide metabolism contribute to these mutations.
|
Pathways of dNTP synthesis in mitochondria
In principle, the precursors to dNTPs within the mitochondrion could be either 1) deoxyribonucleotides that are transported inward from the cytosol; 2) deoxyribonucleosides that undergo passive or facilitated diffusion into the organelle and are phosphorylated within the mitochondrion; or 3) ribonucleotides that undergo reduction within the organelle. These routes are summarized in Fig. 4
. Evidence supports the existence of all three pathways, and it is possible that mitochondria in different cells vary in the pathways used. Bridges et al. (72)
have described an active transport system for dCTP in human mitochondria. Activity of this system with other dNTPs was not reported, but resistance of this system to inhibition by atractyloside indicated that the system differs from the well-known adenine nucleotide translocator.
|
Later, Dolce et al. (73)
described a mammalian protein called DNC, the deoxynucleotide carrier (74)
. As reconstituted into proteoliposomes, the protein was shown to transport a variety of nucleotides, with deoxyribonucleoside diphosphates being the best substrates, followed by ribonucleoside diphosphates. There is some question of whether the physiological role of DNC is actually nucleotide transport. In a collaboration involving the laboratories of L. Biesecker, F. Palmieri, and C. K. Mathewss Lindhurst et al. (unpublished results) generated cell lines deficient in DNC (Table 2)
. These authors found mitochondrial dNTP pools to be normal in these cell lines. However, the mitochondria from these individuals were deficient in thiamine pyrophosphate, suggesting that the protein serves primarily for transport of the vitamin or its phosphorylated forms. Palmieri and colleagues (75)
have reported that in yeast most of the mitochondrial dNTP pools arise through direct transport of dNTPs from the cytosol, where they are synthesized, and they have described a guanine nucleotide transporter (76)
and a countercurrent exchanger that works optimally with pyrimidine nucleoside di- and triphosphates (75)
. Because yeast nucleoside diphosphate kinase is concentrated in the intermembrane space, it seems likely that triphosphates are the preferred substrates for transport into the matrix.
Although the yeast genome encodes several proteins predicted or shown to function as nucleotide transporters, it is of some interest that two laboratories (77
, 78)
have described a yeast mitochondrial isoform of NADH kinase; one of those groups (78)
showed that loss of this enzyme increased the mutation rate for mitochondrial genes. Since NADPH is the ultimate electron source for both ribonucleotide reductase and thymidylate synthase, it is intriguing to consider whether this means that the mutator effect results from altered dNTP pools, which in turn would suggest that de novo pathways of dNTP synthesis might be active in the mitochondrion.
Salvage pathways beginning with nucleosides certainly exist within mitochondria, because two of the four human deoxyribonucleoside kinases—thymidine kinase 2 and deoxyguanosine kinase—are located within mitochondria (79)
. These enzymes have sufficiently broad substrate specificity to account for synthesis of the four canonical deoxyribonucleotides. In addition, it is well established that the toxicity of several nucleoside analogs used in antiviral chemotherapy results from their conversion to the corresponding triphosphates within the mitochondrion, leading to interference with mtDNA replication or gene expression (80
, 81)
. Another potential pathway for mitochondrial dNTP synthesis comes from evidence in our laboratory indicating the presence of ribonucleotide reductase activity within mitochondria (82)
, and suggesting in turn that dNTPs could be synthesized de novo within the organelle. Shiwei Song in our laboratory (83)
found the specific activity of RNR in mammalian liver to be severalfold higher in mitochondrial than in cytosolic extracts, arguing against the idea that mitochondrial RNR is a cytosolic contaminant. Regardless of the pathways used to synthesize dNTPs, there is evidently active turnover of deoxyribonucleotides within the organelle, as shown by the presence of intramitochondrial 5'-nucleotidase activities (84)
. Roles of these enzymes in pool size regulation have not yet been defined.
Induction of long deletions by dNTP pool imbalance
MNGIE is a devastating disease characterized by accumulation of point mutations and long mtDNA deletions in heart and skeletal muscle (see Table 2
). Hirano and colleagues (58)
showed that this autosomal recessive condition results from deficiency of thymidine phosphorylase, an enzyme encoded by a nuclear gene. Later the same group (59)
found that patients with this condition have abnormally elevated levels of circulating thymidine, consistent with the postulated role of this reversible enzyme in nucleoside catabolism, viz., thymidine + orthophosphate thymine + deoxyribose-1-phosphate. They speculated that the thymidine is taken into cells and into mitochondria, and that for reasons unknown it generates more of a dNTP imbalance within mitochondria than within cytosol or nucleus. Song et al. (85)
tested this prediction by direct analysis of mitochondrial and whole-cell dNTP pools within HeLa cells that were exposed to thymidine in the culture medium for several hours. The effects on dTTP, dATP, and dGTP pools were similar in mitochondria and whole cells; dTTP and dGTP pools more than doubled whereas dATP rose slightly. However, in mitochondria the dCTP pool shrank under these conditions to nearly undetectable levels.
Can dNTP imbalances of the type seen here account for the long mtDNA deletions seen in MNGIE patients? To answer this question, Song et al. subjected HeLa cells to long-term culture in medium either containing or not containing thymidine at 50 µM. mtDNA was isolated from the cultures at intervals and checked for deletions by Southern blot and by polymerase chain reaction (PCR). After 8 months, long deletions were seen in mtDNA from the thymidine-treated cells, but not the control. This is the first indication that unbalanced dNTP pools can cause deletions to occur. Because the deletions were several kilobase pairs long, the mechanisms proposed earlier for frameshift mutagenesis cannot work. A plausible mechanism is shown in Fig. 5
(see also ref 56
), and is based on the premise that mtDNA replicates asymmetrically, so that template DNA is single-stranded in advance of the replication fork. Although that long-established model for mtDNA replication has been challenged (86)
, recent evidence suggests that both modes of replication occur under different conditions (87)
. At any rate, if DNA replication is asymmetric, then DNA polymerase stalling during replication may give the product DNA strand a chance to partially dissociate from the template. Then that strand may pair inappropriately with a downstream homologous template sequence that would lead, upon resumption of DNA synthesis, to deletion of all product DNA between the two direct repeat sequences. The important question, then, is whether the dCTP depletion caused by thymidine is sufficient to limit the activity of DNA polymerase 
, causing it to stall. In thymidine-treated cells, dCTP represents just 2% of the total dNTP pool, and its estimated concentration is in the low micromolar range, possibly in the range to limit DNA polymerase activity in the organelle.
|
MNGIE patients show point mutations as well as deletions in mtDNA, and Nishigaki et al. (88)
have sequenced a number of these. The most common point mutations seen were TC transitions, where T in the template is followed by at least two As. As shown in Fig. 6
, a likely mutagenic pathway is favored by the pool changes we observed; G misinsertion opposite T is favored by dGTP expansion, and dTTP accumulation allows for next-nucleotide effects at the two As 5' to the site of misinsertion. Thus, both the point mutations and deletions observed can be explained in terms of the effects of thymidine uptake on mitochondrial dNTP pools, making this perhaps the first demonstration that physiological changes in DNA precursor pools can influence specific mutational pathways. Processes of this type may be easier to see in mitochondria, where mismatch repair activities, which could obscure these results, are certainly lower than in the nucleus (89)
, if indeed they occur there at all.
|
While the work in our laboratory was being carried out, the Hirano laboratory (90)
reported that MNGIE patients show accumulation of deoxyuridine, another thymidine phosphorylase substrate, to an even greater extent than thymidine. This finding suggests that the effects of deoxyuridine on mitochondrial dNTP pools and mutagenesis patterns should be investigated. These experiments will be technically more difficult, because deoxyuridine is expected to cause dUTP to accumulate, as well as dTTP. Whereas dUTP can be assayed in the presence of dTTP, the small amounts of dNTPs available in mitochondrial extracts could make these assays problematic.
Mitochondrial dNTP pools and mutagenesis in animal tissue mitochondria
Experiments on the MNGIE model suggested that mutation rates and mutational spectra could be influenced by dNTP levels in mammalian mitochondria. Since mitochondrial mutations accumulate with age (67
68
69
70
71)
, it is of interest to know whether the mitochondrial mutation rate increases with age and, if so, whether changes in mitochondrial dNTP pools are partly responsible. This study involved analysis of mitochondrial dNTP pools in organs of young and old rats, and yielded some surprises (91)
. Chief among these was the finding that dGTP, the least abundant of the four dNTPs in whole-cell extracts of virtually every organism tested (31)
, is the predominant dNTP in mitochondria from some tissues, particularly heart and skeletal muscle. In these tissues dGTP comprised as much as 90% of the total dNTP pool, while dTTP pools were so low in some instances as to be nearly undetectable. Table 3
shows estimated molar concentrations of the dNTPs in mitochondria from several tissues of young rats. In other experiments we found these levels not to be significantly changed in mitochondria from aged rats.
|
To determine whether these highly asymmetric pool sizes affect replication fidelity, Dr. Zac Pursell carried out in vitro DNA synthesis experiments with recombinant human DNA polymerase (91)
. These assays involved gap-filling reactions run at estimated physiological dNTP concentrations, using a DNA template containing a lacZ mutational target. When compared with reaction mixtures containing equimolar dNTPs, the reactions run at quasi-physiological concentrations showed stimulation of replication errors of 
3-fold. When the reactions were run with a 3' exonuclease-deficient polymerase mutant, defective in proofreading, the degree of stimulation of errors was as much as 6-fold. Sequence analysis of the mutants confirmed that most of the mutations were substitutions and that the mutation stimulated most dramatically by mitochondrial pool asymmetry was A · TG · C transitions, where T in the template strand is followed by C, giving a strong next-nucleotide effect in the presence of high dGTP—as expected from the high dGTP and low dTTP concentrations in mitochondria. Among the spectrum of spontaneous mutations in mitochondrial DNA the A · TG · C transition predominates. In three studies of mice and humans this mutational pathway represented from 55 to 62% of the total mutations analyzed (88
, 91)
. These findings suggest that in mitochondria the natural dNTP asymmetry is mutagenic and stimulates specific mutational events. However, the metabolic significance of the extreme asymmetries seen, particularly in heart and skeletal muscle mitochondria, remains unknown.
|
CHEMICAL MODIFICATION OF NUCLEOTIDES AS A PREMUTAGENIC EVENT |
|---|
|
TOPABSTRACTINTRODUCTIONNOVEL ASPECTS OF dNTP...dNTP POOL ASYMMETRIES AS...dNTP METABOLISM AND...dNTP METABOLISM AND MUTAGENESIS...CHEMICAL MODIFICATION OF...RELATIONSHIPS BETWEEN dNTP...REFERENCES |
|---|
showed that treating cells with a DNA-methylating agent led to far more methylation of free nucleotides than of nucleotide residues in DNA. Whether those modified nucleotides were subsequently incorporated into DNA in significant amounts has still not been established.
8-Oxodeoxyguanosine triphosphate and other oxidized nucleotides
The more recent work on genotoxicity of ROS began with analysis of three E. coli genes known to protect against mutagenesis caused by oxidative stress—mutM, mutY, and mutT. All three genes have homologs in eukaryotic cells. Oxidative damage to DNA creates several oxidized DNA bases, of which the most prominent, and most strongly mutagenic, is 7,8-dihydro-8-oxoguanine (8-oxoG). This base efficiently pairs with adenine, and the resultant purine-purine base pair is an intermediate in transversion mutagenesis (93)
.
MutM and MutY are DNA glycosylases, known to initiate BER pathways (93)
. However, MutT is a nucleotidase, shown originally to cleave dGTP to dGMP and pyrophosphate (94)
. Subsequently, Maki and Sekiguchi (95)
demonstrated that the MutT protein cleaves much more efficiently the oxidized dGTP derivative, 8-oxodeoxyguanosine triphosphate (8-oxo-dGTP). This important finding suggested that the genotoxicity of ROS results significantly from incorporation of the oxidized nt into DNA, as shown in Scheme 1. According to this model, MutT acts to "sanitize" the nucleotide pool by removing a DNA polymerase substrate whose incorporation into DNA would be strongly mutagenic. The bacterial MutT protein differs from its relative, hMTH1 (human MutT homolog), in several significant respects. The bacterial enzyme cleaves the oxidized ribonucleotide 8-oxo-GTP as well as the deoxyribonucleotide 8-oxo-dGTP (96)
, while the human enzyme is active on two oxidized adenine nucleotide substrates—2-hydroxy-dATP and 8-oxo-dATP (97
, 98)
. Although the physiological substrates for these enzymes have not yet been identified, both 8-oxo-dGTP and 2-hydroxy-dATP have been shown to be strongly mutagenic during DNA replication in vitro (99
, 100)
.
|
The E. coli MutT protein is a member of a family of proteins termed Nudix hydrolases by Bessman et al. (101)
. These proteins, which share a signature amino acid sequence motif, have in common the ability to cleave a nucleoside diphosphate-X substrate, where X is any functional group, such as an additional phosphate or a sugar residue. Other nudix proteins have been shown to possess nucleoside triphosphatase activities, and preliminary evidence (102)
has been presented supporting a role for one such protein, E. coli Orf135, in counteracting mutagenesis induced by oxidized dATP derivatives. However, the important question of whether disrupting this gene generates a mutator phenotype has not yet been addressed.
The E. coli mutT gene was the first mutator gene to be discovered, and is one of the most powerful known mutators. This finding plus the 8-oxo-dGTPase activity of the protein, plus the strong mutagenic effect of the nt after incorporation of DNA in vitro, have led to widespread acceptance of the view that 8-oxo-dGTP does represent a true premutagenic lesion. In addition, the finding that MTH1 knockout mice show high rates of spontaneous tumorigenesis (103)
is consistent with the idea that the MTH1 plays a role in mammals similar to that of the bacterial enzyme. However, several observations are difficult to reconcile with the idea of 8-oxo-dGTP as a significant premutagenic lesion.
First, although 8-oxo-dGTP is incorporated into DNA in vitro, where it stimulates transversion errors (100)
, in those studies the oxidized nt was present in reaction mixtures at concentrations comparable to those of the four canonical dNTPs—in other words, at far higher concentrations than the nt is apt to exist in living cells. Second, Fowler et al. (104)
reported that under certain anaerobic growth conditions, where nt oxidation should not be occurring, mutT mutant bacteria retained the mutator phenotype. Third, Einolf and Guengerich (105
, 106)
reported that 8-oxo-dGTP is a poorer substrate than dGTP, by several orders of magnitude, for several different DNA polymerases tested.
Our laboratory’s contribution to the problem has been to develop an assay for 8-oxo-dGTP sufficiently sensitive to detect it at the low concentrations at which it was assumed to exist in vivo. Then, if the nucleotide is a significant mutagenic intermediate, we expect to find it in higher abundance in mutT mutant cells, which lack the nucleoside triphosphatase activity, than in WT cells, and we expect its levels in WT cells to increase after oxidative stress. Accordingly, Mary Lynn Tassotto (107)
developed an HPLC assay that used electrochemical detection to quantitate 8-oxo-dGTP. Surprisingly, she was not able to detect the nucleotide, even in mutT mutant bacteria, where it should have accumulated. We estimated that the highest concentration that could have gone undetected in our system was 0.3 µM. When an vitro replication system was programmed with 8-oxo-dGTP at this concentration and the four canonical dNTPs at their estimated physiological concentrations, there was no discernible effect of the oxidized nucleotide on replication errors. These results present us with a paradox that cries out for resolution. One possibility is that the true physiological MutT substrate is a different damaged nucleotide.
However, 8-oxo-dGTP deserves further attention. For one thing, the very high dGTP levels we found in some mammalian mitochondria, plus the highly oxidizing conditions within this organelle, suggest that dGTP might present a suitable oxidation target and that its oxidation product(s) contributes to the high mutation rate within the mitochondrial genome. If 8-oxo-dGTP can be detected in mammalian mitochondria, do its levels correlate with the abundance of dGTP among the dNTPs in mitochondria from different tissues?
More recent work with mammalian cells from Bignami’s laboratory (108
109
110)
supports the idea that 2-hydroxy-dATP is a physiologically significant premutagenic lesion in mammalian cells. First, Colussi et al. (108)
reported that the mismatch repair (MMR) system in mammalian cells can act at sites containing oxidized bases, and Macpherson et al. (109)
reported that in some sequence contexts MutS, the MMR protein involved in mismatch recognition, does recognize misincorporated oxoG. Since mismatch repair closely tracks the replication machinery, these findings imply that the oxidized bases arose through incorporation of the corresponding oxidized nucleotides. Russo et al. (110)
enlarged on this work, showing that expression of hMTH1 in MMR-defective mouse embryo fibroblasts attenuated mutagenesis caused by the MMR defect. Sequence analysis of revertants at the hprt locus showed that hMTH1 expression most strongly reduces rates of G · CT · A and A · TT · A transversions and A · TG · C transitions. Whereas G · CT · A transversions could result from incorporation of either 8-oxo-dGTP or 2-hydroxy-dATP, it is far more likely that A · TT · A transversions and A · TG · C transitions would have arisen from base pairs containing 2-hydroxyadenine, because this base can pair with either A or C. These experiments cast a strong spotlight on 2-hydroxy-dATP, which is a substrate for hMTH1, and point up the importance of developing an assay for quantitating pools of this oxidized nt.
dITP, dXTP, and genomic stability
Although technically not a "damaged" nt, dUTP offers some interesting parallels with 8-oxo-dGTP in that both are hydrolyzed to the corresponding monophosphate as a means of excluding the undesired nt from DNA. dUTPase enhances genomic stability, partly because misincorporated dUMP can be mutagenic and partly because massive dUMP incorporation on both DNA strands can cause repair reactions to proceed past one another and lead to irreparable double-strand breaks. Recent work from Kuzminov’s laboratory (111
, 112)
suggests a similar role for E. coli RdgB (RecA-dependent growth). The rdgB gene encodes a putative nucleotidase (113)
active against the ribo- and deoxyribonucleoside triphosphates of hypoxanthine and xanthine, the deamination products of adenine and guanine nucleotides, respectively. rdgB mutants are inviable in a recA background, possibly because misincorporated deoxyinosine or deoxyxanthosine nucleotides (dITP or dXTP) lead to DNA breakage events that require RecA protein in order for repair to be completed. Testing this model will require development of methods for analysis of intracellular pools of dITP and dXTP. However, it is noteworthy that studies of mutT and rdgB, though incomplete, complement our understanding of the dUTPase reaction and suggest that dNTP hydrolysis is a general mechanism in maintaining genomic stability.
Inflammation and mutagenic halogenated nucleotides
A different source of mutagenic nucleotides has been described by Henderson et al. (114)
. In chronic inflammation, eosinophils generate oxidants that lead to mutagenesis in neighboring cells. Henderson et al. showed that one such oxidant, hypobromous acid (HOBr), could, under the influence of eosinophil peroxidase, brominate deoxycytidine, and that this product underwent deamination and was incorporated into DNA as the highly mutagenic bromodeoxyuridine. 5-Chlorouracil compounds are formed by a similar mechanism involving hypochlorous acid (115)
, although the pathway by which the mutagenic 5-chlorouracil arises in DNA has not been established.
|
RELATIONSHIPS BETWEEN dNTP METABOLISM, GENOME STABILITY, AGING, AND CANCER |
|---|
|
TOPABSTRACTINTRODUCTIONNOVEL ASPECTS OF dNTP...dNTP POOL ASYMMETRIES AS...dNTP METABOLISM AND...dNTP METABOLISM AND MUTAGENESIS...CHEMICAL MODIFICATION OF...RELATIONSHIPS BETWEEN dNTP...REFERENCES |
|---|
, it was evident that induction of nt pool imbalances, either by mutations or by exogenous compounds, affected mutation patterns as expected from dNTP pool measurements and results of in vitro DNA synthesis experiments. It was not clear that similar mutagenic processes are influenced by natural dNTP pool asymmetries. Our more recent in vitro data (35)
suggest that natural pool asymmetries are not strongly mutagenic, at least, for nuclear gene mutations. The mitochondrial dNTP pools, which are more highly skewed, do appear to be mutagenic, at least for some types of replication errors (91)
. It is also possible that low mismatch repair activities in mitochondria magnify the effects of replication errors induced by pool asymmetry, contributing to elevated mutation rates in the mitochondrial genome.
Our recent finding that proportional dNTP accumulations are mutagenic (7)
was unexpected, although perhaps it should not have been. PCR reactions are routinely run at low dNTP concentrations to maximize polymerase proofreading and minimize replication errors. Should it be apparent, then, that at high concentrations dNTPs would enhance chain extension from mismatches, thereby maximizing the next-nucleotide effect? In this context it would be of great interest to know whether the high dNTP pools seen in transformed cells in culture, relative to normal diploid cells, are related to the accumulation of mutations that occurs in oncogenesis. Loeb, among others (8)
, has proposed that a mutator phenotype must be established in cells destined for oncogenic transformation to account for the large number of genetic differences between a tumor cell and the normal cell from which it originated (116)
. Could the mutagenic accumulation of dNTPs be a contributing factor or is it an accidental by-product of other changes more directly on the path to carcinogenesis?
The recently described relationships between mismatch repair and oxidized DNA bases make it critically important to understand the extent to which oxidized nucleotides contribute to mutagenesis, since defective mismatch repair is associated both with genomic instability and with cancer. The relationship of oxidized nt metabolism to aging also needs attention. Many studies have described aging-related increases in 8-oxoguanine residues in DNA (e.g., ref 117
) although some of the earlier studies are flawed by DNA oxidation occurring during sample workup (118)
. Recent investigations, which take this into account, show a strong relationship, particularly with mitochondrial DNA (e.g., ref 119
). Despite our inability so far to detect 8-oxo-dGTP in E. coli, these findings suggest the importance of analyzing age-related changes in oxidized nt metabolism, which might help explain the significance of these findings.
|
ACKNOWLEDGMENTS |
|---|
Received for publication February 7, 2006. Accepted for publication March 3, 2006.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONNOVEL ASPECTS OF dNTP...dNTP POOL ASYMMETRIES AS...dNTP METABOLISM AND...dNTP METABOLISM AND MUTAGENESIS...CHEMICAL MODIFICATION OF...RELATIONSHIPS BETWEEN dNTP...REFERENCES |
|---|
) 3' terminus as the concentrations of the four dNTPs are increased proportionately, based on the assumptions that KM for values for extension from matched and mismatched termini are 1 µM and 1000 µM, respectively (7)
