From the Division of Biochemistry and Molecular Biology, University of California, Berkeley, California 94720-3202
Although oxygen is a powerful oxidant, the
triplet ground state of dioxygen constitutes a kinetic barrier for
oxidation of biological molecules, which are mostly singlet state (1).
However, the unpaired orbitals of dioxygen can sequentially accommodate single electrons to yield O
Although DNA is a biologically important target for reactive oxygen
species, free O
Most transition metals have more than one oxidation state besides
the ground state, and their valence electrons may be unpaired (13),
allowing one-electron redox reactions. As such, transition metals can
react with H2O2 to produce ·OH and
related oxidants. In 1894, Fenton (14) described the oxidation of
tartaric acid by Fe2+ and H2O2, and
the stoichiometry of Fe2+ and H2O2
consumption was subsequently shown to be consistent with that of
Reaction 5 (15). This review will focus on the iron-mediated Fenton
reactions; those with other transition state metals are discussed
elsewhere (1, 16).
Iron has five oxidation states in aqueous solution, Fe(II)-Fe(VI).
Fe(II) and Fe(III) are the most common, and their reactions with oxygen
and its reduced forms are well documented (1, 17). More recently,
however, reactions with Fe(IV) have been implicated in biological
processes and proposed to be involved in damage to cellular components
(1, 16, 18-20). For example, in the case of Fe2+ chelates
with ADP, ortho-phosphate, or EDTA, the oxidant formed from
H2O2 behaves differently than expected for
·OH, and it has been proposed to be the ferryl radical (Fig. 1, Reaction 6). Alternately, a caged or bound ·OH, often
denoted as [Fe-H2O2]2+ or
[FeOOH]+, might account for the noted differences (18).
This distinction might be arbitrary, however, as this bound ·OH
might be an intermediate of Reactions 5 or 6 (16) and the ferryl
radical itself could give rise to ·OH via Reaction 7 (1,
18).
Both Fe2+ and Fe3+ may have open orbitals that
can share outer sphere electrons with ligands for iron coordination.
The chemical properties of such complexes are determined by the ligands
(1, 16, 18). For example, Luo et al. (21) found that in the presence of the very complex ligand, DNA, there are three kinetically distinguishable oxidants formed that cause DNA strand breakage. One of
these is easily scavengable, consistent with it being a freely
diffusible ·OH, whereas the other two vary in their scavenging
susceptibilities, and one or both of these might be an iron(IV)
species.
Therefore, after 100 years, the basic nature of the Fenton oxidant(s)
is still undefined so that "·OH" may be regarded as a symbol
representing the stoichiometric equivalent of the univalent oxidation
agents produced by the Fenton reaction. However, it is clear that
whatever the oxidant, hydroxylations and hydrogen abstractions are the
two most common modifications of organic substrates by Fenton oxidants
(17, 22, 23).
It is noteworthy that H2O2 can also react with
Fe3+ to form O The Generation of Reactive Species by
H2O2 and O H2O2 and O Singlet dioxygen is not spin-restricted from oxidizing organic
compounds as is triplet state oxygen (1) and was once proposed to be
the product of dioxygen-producing reactions involving either H2O2 or O O A substantial portion of H2O2 lethality
involves DNA damage by oxidants generated from iron-mediated Fenton
reactions (4, 12). It would appear that NADH can drive the process by
replenishing Fe2+ from Fe3+ in bacteria and
in vitro (4). Moreover, NADH enhances iron-DNA association
(32).
A large portion of H2O2-dependent
DNA damage appears not to be due to diffusible hydroxyl radicals (4,
21, 33). Instead, DNA-damaging Fenton oxidants are produced on
Fe2+ atoms associated with DNA, and it would appear that
the location of iron binding may determine the substrate and nature of
attack. There appear to be at least two distinguishable classes of
iron-mediated Fenton oxidants of DNA (4, 21). Type I oxidants are
moderately sensitive to H2O2 and ethanol and
appear to cleave DNA preferentially within the sequences
RTGR, TATTY, and
CTTR (the bold, underscored nucleotides are the
sites of cleavage); Type II oxidants, on the other hand, make
preferential cleavages in the sequence
NGGG.2 The
distinguishing characteristics of these radicals may be predominantly due to localization of the iron that gives rise to them. However, it
may be that the sites of nicking are not necessarily the iron-binding sites. The NGGG sites in particular may be sinks for radical electrons, which are formed elsewhere on the helix and travel through the base
stack (34). Whether there are differences in the spectrum of base
damages by the two types of oxidants has not been reported.
Damage by Fenton oxidants may occur at the DNA bases or sugars. Sugar
damage is initiated by hydrogen abstraction from one of the deoxyribose
carbons, and the predominant consequence is eventual strand breakage
and base release (35, 36). In approximately half of these alterations,
a 5 Another alteration at the sugar moiety is a Radical attack on the bases results primarily in OH addition to the
electron-rich double bonds, particularly the purine N-7-C-8 bond and
the pyrimidine 5,6 bond (22, 23).3 Hydrogen
abstraction from thymine-methyl groups also occurs (35).3
In general, radical attack on the base moieties of DNA does not give
rise to altered sugars or strand breaks except when base modifications
labilize the N-glycosyl bond, allowing the formation of
baseless sites that are subject to One source of the difference between products formed by Fenton oxidants
versus ionizing radiation could be the participation of iron
ions directly in product formation. DNA-bound iron may interact with
nascent DNA radicals and thereby qualitatively and quantitatively alter
the products (45). In the absence of O2, Fe3+
can react with reducing DNA radicals (Fig. 1, Reaction 12).
In the presence of O2, DNA peroxyl radicals are formed,
which can react with Fe2+ (Fig. 1, Reactions 13 and 14). These reactions affect the product spectrum and
thereby obfuscate identification of the initial oxidants through
product analyses.
Among the oxidized purines, formamidopyrimidines and
7,8-dihydro-8-oxoguanine (8-oxo-Gua)4 have
received widespread study, whereas among the pyrimidines, thymine
glycol and its spontaneous hydrolysis products have been actively
studied, most likely because of the ubiquitous presence of enzymes for
the excision of all of these products (see below). 8-oxo-Gua is also
the object of much study because of its highly mutagenic nature (it
base pairs relatively well with adenine) (46) and the relative ease of
its isolation and quantitation.
Another type of DNA damage mediated by iron in vivo is
DNA-protein cross-links, e.g. thymine-tyrosine (47). DNA
interstrand cross-links have not been shown to be formed by oxygen
radicals.
Elimination of Fenton Oxidants The fidelity of the metabolic redox reactions (2) and the
sequestering of iron in ferritin and transferrin (9, 48) generally
minimize the burden from reactive oxygen species. Moreover, compartmentalization of free iron and superoxide and the impediment for
iron binding to DNA by histones (49) diminish the occurrence of Fenton
reactions on DNA.
Active oxygen species produced by iron/H2O2 are
also removed by superoxide dismutases (SODs) (Reaction 1), catalases
(Fig. 1, Reaction 15), and peroxidases that catalyze the
reduction of H2O2 by organic reductants (RH)
such as glutathione, ascorbate, and cytochrome c (Fig. 1,
Reaction 16). The major source of protection would appear to
be SOD. Mammalian cells produce a mitochondrial Mn-SOD, a cytoplasmic
Cu,Zn-SOD that is also found in peroxysomes (50), and an extracellular
Cu,Zn-SOD (51). Fe-SOD is additionally found in some bacteria and in
chloroplasts (52). Since superoxide dismutation forms
H2O2, the detoxifying effect of SOD is most likely a result of preventing the accumulation of free Fe2+
(Reactions 2 and 3) and peroxynitrite production (Reaction 11).
Catalase does not appear to be nearly so important as SOD, judging from
the weak phenotypes of cells that lack it (4) and persons with
acatalasemia (53). In mammalian cells catalase is largely contained in
peroxysomes (54) and to a lesser extent it is secreted (55).
Escherichia coli contains two catalases, one regulated by
stationary phase and the other by H2O2 exposure (56).
In eukaryotes, glutathione peroxidases are found in the mitochondria,
cytoplasm, and peroxysomes (50). These enzymes, especially the selenium
glutathione peroxidase, are more effective in removing H2O2 than catalase (52). Peroxidases are less
specific than catalase and can also reduce organic hydroperoxides that
can react in Fenton-like reactions. Oxidized glutathione is reduced by
NADPH-dependent glutathione reductase, an auxiliary enzyme
for this antioxidant function.
The relative levels of SOD, catalase, and glutathione peroxidase are
important. For instance, an increase in SOD would deplete the cell of
superoxide but would increase H2O2 production,
which might be deleterious unless sufficient catalase and/or
glutathione peroxidase were available. Likewise, excess glutathione
peroxidase could unnecessarily deplete glutathione and/or NADPH
reserves even though sufficient catalase was present (52, 57).
Eukaryotes also contain a thiol-specific antioxidant enzyme that acts
as a thiol-dependent peroxidase, at least at low
H2O2 concentrations (~50 µM)
(58). At high concentrations of H2O2 (~10
mM), thiol-specific antioxidant enzyme is reported to
protect DNA against damage by thiol/metal-catalyzed oxidation (59); however, this protection does not appear to be mediated by the peroxidase activity.
The only effective means of detoxification of ·OH is to scavenge
it non-enzymatically. Histones and the compact structure of chromatin
protect the DNA by this means (60). As yet, an enzymatic apparatus for
singlet oxygen removal has not been detected; rather the cell appears
to employ scavengers such as carotenoids (61).
Raising NADH levels exacerbates H2O2 toxicity
in E. coli (4), and NADH increases
iron/H2O2-mediated DNA damage in
vitro (22, 23). However, NADPH is at least an order of magnitude slower in reducing Fe3+ and competes very effectively with
NADH for iron binding (32). It may therefore be important that in
E. coli an O Repair of DNA Damage from Reactive Oxygen Species Direct enzymatic reversion of any
oxidative DNA damage product has not been described. However, under
some conditions, carbon-centered radicals formed on the DNA backbone by
·OH attack may be restituted to undamaged DNA by hydrogen
donation from a sulfhydryl (65). O2,
H2O2, and iron may interfere in this
"chemical restitution," and sulfhydryls may in fact exacerbate DNA
damage by iron/H2O2 (66).
Once DNA nucleoside damage is manifested, enzymatic mechanisms are
necessary to correct the alteration. The damage must be recognized,
removed, and replaced with normal nucleotides, and DNA ligase must seal
all strand breaks (67).
Base excision repair is manifested
through a DNA glycosylase, which recognizes the damaged base and
cleaves its glycosylic bond (68, 69). An enzyme recognizing
hydroxymethyluracil is present in eukaryotes (70) but apparently not in
bacteria (71). Its role would appear to be to avoid mutations due to
the formation of hydroxymethyluracil upon oxidation of 5-methylcytosine
in DNA. Most organisms appear to contain formamidopyrimidine (FAPy)
glycosylases (Fpg protein) and several pyrimidine hydrate DNA
glycosylases (e.g. E. coli endonucleases III and VIII). The
former recognizes formamidopyrimidines and 8-oxopurines. The latter
recognizes thymine glycols, pyrimidine hydrates, and their degradation
products. Saccharomyces cerevisiae is exceptional in having
an enzyme that recognizes both pyrimidine hydrates and
formamidopyrimidines but not 8-oxo-Gua (72). These enzymes also
catalyze a In addition, DNA deoxyribosephosphodiesterase (drPase) activities exist
that utilize hydrolytic mechanisms for removing 5 A mismatch repair DNA glycosylase in E. coli (MutY) (46) and
in human cells (73) is an adenine DNA glycosylase that removes adenine
when it is paired to 8-oxo-Gua. After the adenine is replaced by
cytosine, the 8-oxo-Gua is then excised by the Fpg protein, which does
not act on 8-oxo-Gua:A mismatches.
This process is catalyzed by
large enzyme complexes that ultimately result in the excision of an
oligonucleotide of roughly 13 nucleotides in procaryotes or 28 nucleotides in eukaryotes (69). Undoubtedly a subset of oxidative
damage is removed by this pathway. A unique pathway for repair of
8-oxo-Gua lesions has been reported in human cell extracts in which an
8-oxo-Gua endonuclease recognizes the lesion and makes incisions
immediately 3 Exonucleases also take part in nucleotide excision repair in the
capacity of nick translation, removal of unpaired damaged termini, or
in the removal of abnormal 3 Double strand breaks and DNA-protein
cross-links formed by oxygen radicals are repaired either by homologous
recombination or by non-homologous end joining. In homologous
recombination, double strand breaks are initially processed
by degrading the 5 By being mutagenic, reactive oxygen species have been implicated
in cancer and other degenerative diseases. However,
p53-dependent apoptosis seems to be mediated by reactive
oxygen species (79), so these agents have diametrical effects; they
cause undesirable cellular alterations but also prevent undesirable
consequences of DNA damage by helping to eliminate damaged cells.
A final consideration is DNA damage to mitochondrial DNA. Clearly
mitochondrial DNA is damaged by reactive oxygen species, and pyrimidine
hydrate DNA glycosylases (80), AP endonuclease (81), and a
recombination (82) exist in mitochondria. Whether repair of the
multiple mitochondrial genomes of the cell is sufficient to prevent an
accumulation of ineffective mitochondrial genomes and hence an
age-related "error-catastrophe" is an active area of interest.
2, H2O2,
the very reactive ·OH, and water (Fig. 1,
Reaction 1). The oxidative potential of atmospheric oxygen
is maintained by the non-alignment of electron spins, and aerobic life
is based upon harnessing energy via the catalytic spin pairing of
triplet oxygen by the electron transport chain (2). The latter process
occasionally errs, however, giving rise to O
2 and other
reactive oxygen species (3) that cause cellular and genetic damage
(4-7). Moreover, catabolic oxidases such as xanthine oxidase, anabolic
processes such as nucleoside reduction, and defense processes such as
phagocytosis also produce oxygen radicals.
Fig. 1.
Reactions cited in the text. Reaction
1 shows the univalent reductions of aqueous oxygen to water at pH
7. Potentials are for 1 M aqueous oxygen at pH 7 (1). The
negative potential for the O2/O2 couple and the
positive potential for the O
2/H2O2 couple indicate that O
2 is redox ambivalent and
disproportionation (Reaction 2) is favorable. The large
potential for ·OH reduction and its radical nature allows it to
oxidize organic molecules almost indiscriminately.
[View Larger Version of this Image (26K GIF file)]
2 is relatively unreactive with DNA (8). However, O
2 dismutates (via spontaneous or enzyme-catalyzed
reactions) to produce H2O2 (Fig. 1,
Reaction 2). O
2 can also reduce and liberate
Fe3+ from ferritin (9) (Fig. 1, Reaction 3) or
liberate Fe2+ from iron-sulfur clusters (10) (Fig. 1,
Reaction 4); subsequently very reactive oxygen species can
form via the Fenton reaction (Fig. 1, Reaction 5). Thus, the
cytotoxic effects of O
2 (as well as of iron and
H2O2) have been linked to DNA damage by way of the Fenton reaction (4, 11, 12) (Fig. 2).
Fig. 2.
Cellular reactions leading to oxidative
damage of DNA via the Fenton reaction.
H2O2 is formed by endogenous metabolism or is
available exogenously. Superoxide is produced as a by-product of
O2 reduction in the electron transport chain. Superoxide
dismutation and release of protein-bound iron by superoxide form
H2O2 and Fe2+, respectively, which
in turn can react to form ·OH-type oxidant(s). These oxidant(s)
may cause DNA damage. Fe3+ produced by the Fenton reaction
is thought to be reduced by available NADH, thus replenishing
Fe2+. H2O2 can be depleted by
catalase or by peroxidases, which utilize reduced glutathione, other
thiols, cytochrome c, ascorbate, etc.
[View Larger Version of this Image (22K GIF file)]
2, presumably via Reaction 8 (24),
and that if H2O2 is in excess, the
Fe2+ which is thus formed can subsequently generate
reactive oxygen species via the Fenton reaction.
2 in the Absence of
Metals
2 may participate in
the production of singlet oxygen and peroxynitrite. The generation of
these species may be concurrent with reactions involving iron, and
under some circumstances they might be important contributors to
H2O2 toxicity (25, 26).
2 (27). However, it now
appears that singlet oxygen is not generated via Fenton/Haber-Weiss
chemistry. Instead, for example, OCl
produced by the
reaction of Cl
with H2O2 (Fig. 1,
Reaction 9) might react with H2O2 to
generate singlet oxygen (Fig. 1, Reaction 10). Reaction 9 is
facilitated by chloroperoxidases, which generate singlet oxygen from
H2O2 and chloride in vitro (28), and
singlet oxygen is produced in neutrophils, which contain abundant
H2O2 and chloroperoxidases (29).1
2 reacts rapidly with nitric oxide (Fig. 1, Reaction
11) to form peroxynitrite anion (26) (Reaction 11), the protonated form of which, peroxynitrous acid (pKa = 6.7),
reacts well with biological molecules. Alternatively,
ONOO
might form singlet oxygen from
H2O2 (30). Consequently, NO· production
by nitric oxide synthase may render cells vulnerable to
superoxide-mediated damage (31).
-phosphate end group is located 3
to the cleavage, a
3
-phosphoglycolate is located 5
to the cleavage, and a base propenal
is released, which subsequently decomposes to the free base and
malondialdehyde (37, 38). The majority of other sugar damages yield 5
-
and 3
-phosphomonoesters flanking a one-nucleoside gap. Some sugar
alterations, such as the
-lactone, do not give this product
immediately but do so after adequate time or treatment (35).
to
inversion at the
1
-carbon, which disrupts the B-DNA structure (39). Simultaneous
alteration of a sugar and base moiety of a DNA nucleoside to yield
5
-8-cyclodeoxyribopurines has been reported after ionizing radiation
(40), and 5
-8-cyclodeoxyguanosine was observed as a product of dGMP
subjected to Fe2+ and H2O2,
although it has not been shown to occur for
H2O2-mediated damage of DNA (22).
-elimination (41). Attack at the
DNA bases leads to as many as 50 base alterations (22, 23,
42-44).3 The spectrum of damages due to
iron/H2O2 is quite similar to (but is not
congruent with) that caused by ionizing radiation (22,
23).3
2 challenge induces glucose-6-phosphate
dehydrogenase and hence raises NADPH levels (62) and that a
H2O2 challenge increases the ratio of NADPH to
NADH.5 Moreover, in mammalian cells DNA
strand breaks result in the depletion of nuclear NAD+ (and
NADH) by forming poly(ADP)-ribose (63). Finally, E. coli aconitase is inactivated by superoxide (64), thus shutting down the
Krebs cycle and NADH production.
-lyase activity that cleaves a 3
-phosphodiester of a
baseless sugar (AP site), leaving a nick with an unsaturated sugar at
the 3
terminus and a 5
-phosphomonoester group (Fig. 1, Reaction
17). The function, if any, of the
-lyase activities is
unknown.
or 3
sugar residues
or sugar fragments such as glycolyate residues. Moreover, ubiquitous
class II AP endonucleases initiate sugar removal by hydrolyzing the
5
-phosphodiester bond of an AP site (Fig. 1, Reaction 18).
The resulting 5
-terminal deoxyribose phosphate is a substrate for the
drPase or
-lyase activity of the DNA glycosylases. Once the baseless
sugars or sugar fragments are removed, the small gap is filled, most
likely by DNA polymerase I in bacteria or DNA polymerase
in higher
eukaryotes and then sealed by DNA ligase.
and 5
to it to form a 1-nucleotide gap (74).
termini such as phosphomonoesters or
phosphoglycolates (75, 76).
-ends to reveal 3
-OH single strand overhangs. These
single strands associate with undamaged homologous DNA, which acts as a
scaffold and template for resynthesis of the 5
-degraded ends from the
3
-OH overhangs (77). In mammalian cells, double strand breaks are
predominantly repaired by non-homologous end joining (illegitimate
recombination), and it seems that this mode of repair is mediated by
the V(D)J system, which rejoins blunt double strand breaks (78). Since
homologous DNA does not act as a scaffold, nucleotides may be lost and
ends from different molecules may be joined resulting in gross
chromosomal rearrangements.