REVIEW ARTICLE
Is there more to aging than mitochondrial DNA and
reactive oxygen species?
Mikhail F. Alexeyev
1,2
1 Department of Cell Biology and Neuroscience, University of South Alabama, Mobile, AL, USA
2 Institute of Molecular Biology and Genetics, Kyiv, Ukraine
Introduction
Aging is a multifactorial phenomenon characterized by
a time-dependent decline in physiological function [1].
This decline is believed to be associated with an accu-
mulation of defects in metabolic pathways. More than
50 years ago, Harman first proposed the Free Radical
Theory of Aging [2], which, over the years, has been
refined to include not only free radicals, but also other
reactive species such as hydrogen peroxide (H
2
O
2
) and
singlet oxygen. In 1972, Harman identified mitochon-
dria as both the main source of reactive oxygen species
(ROS) and a major target for their damaging effects
[3]. This development has identified mitochondrion as
a biological clock, but because the mitochondrion has
a complex biochemical composition, a question about
the molecular identity of this clock remained open.
RNA, proteins and other cellular macromolecules with
relatively short half-lifes are poor candidates for the
progressive accumulation of damage over a lifetime, as
would be expected of such ‘tally keepers’. For this rea-
son, even early studies on the molecular mechanisms
of aging have focused on DNA [4,5].
In mammalian cells, mitochondria are the only
organelles, besides the nucleus, that contain their own
genome, which led Miquel [6] to postulate that aging is
Keywords
antioxidants; lifespan extension;
mitochondria; mitochondrial DNA
degradation; mitochondrial DNA mutations;
mitochondrial DNA repair; mitochondrial
theory of aging; oxidative damage
Correspondence
M. Alexeyev, University of South Alabama,
Department Cell Biology and Neuroscience,
307 University Blvd., MSB1201, Mobile, AL
36688, USA
Fax: +1 251 460 6771
Tel: +1 251 460 6789
E-mail: malexeye@jaguar1.usouthal.edu
(Received 12 July 2009, revised 3 August
2009, accepted 11 August 2009)
doi:10.1111/j.1742-4658.2009.07269.x
With the aging of the population, we are seeing a global increase in the
prevalence of age-related disorders, especially in developed countries.
Chronic diseases disproportionately affect the older segment of the popula-
tion, contributing to disability, a diminished quality of life and an increase
in healthcare costs. Increased life expectancy reflects the success of contem-
porary medicine, which must now respond to the challenges created by this
achievement, including the growing burden of chronic illnesses, injuries and
disabilities. A well-developed theoretical framework is required to under-
stand the molecular basis of aging. Such a framework is a prerequisite for
the development of clinical interventions that will constitute an efficient
response to the challenge of age-related health issues. This review critically
analyzes the experimental evidence that supports and refutes the Free Radi-
cal ⁄Mitochondrial Theory of Aging, which has dominated the field of
aging research for almost half a century.
Abbreviations
BER, base excision repair; ESCODD, European Standards Committee on Oxidative DNA Damage; ETC, electron transport chain; GPx,
glutathione peroxidase; H
2
O
2
, hydrogen peroxide; mtDNA, mitochondrial DNA; MTA, mitochondrial theory of aging; nDNA, nuclear DNA;
8-oxodG, 7,8-dihydro-8-oxo-2¢-deoxyguanosine; Polg, DNA polymerase c; Prx, peroxiredoxin; RET, reverse electron transfer; ROS, reactive
oxygen species; Sod, superoxide dismutase.
5768 FEBS Journal 276 (2009) 5768–5787 ª2009 The Author Journal compilation ª2009 FEBS
caused by accumulation of damage to the mitochon-
drial DNA (mtDNA). This narrowed the focus of
the theory and resulted in the Mitochondrial Theory
of Aging (MTA). Several lines of evidence indirectly
implicate mtDNA in longevity. The Framingham
Longevity Study of Coronary Heart Disease found that
longevity is more strongly associated with age of mater-
nal death than with age of paternal death, suggesting
the cytosolic (mitochondrial) inheritance [7]. In addi-
tion, certain mtDNA polymorphisms have been associ-
ated with longevity. For example, male Italian
centenarians have an increased incidence of mtDNA
haplogroup J [8], while French centenarians have an
increased incidence of a G to A transition at mt9055
[9]. In a Japanese population, longevity was associated
with mtDNA haplogroups D4a, D4b2b and D5 [10,11].
However, a study of an Irish population failed to link
longevity to any particular mitochondrial haplotype,
indicating that factors other than mtDNA polymor-
phism also may play a role in aging [12]. Finally, Castri
et al. have found that while mtDNA variants can be
linked to both increased and decreased longevity, the
time period in which a person was born has a much
greater impact on longevity than the presence or
absence of a particular polymorphism [13].
Environmental genotoxins may facilitate preferential
mtDNA mutagenesis. Mitochondria accumulate high
levels of lipophilic carcinogens, such as polycyclic aro-
matic hydrocarbons [14,15]. When cells are exposed to
some of these compounds, mtDNA is damaged prefer-
entially [16]. Other mutagenic chemicals have also been
shown to preferentially target mtDNA [15,17–21].
Therefore, it is conceivable that lifelong exposure to
certain environmental toxins could result in a preferen-
tial accumulation of mtDNA damage, leading to
aging. However, aging can occur in the absence of
detectable exposure to environmental toxins, which
suggests that a role of these toxins in natural aging is
limited. At present, after many years of refinement,
there is no universally accepted definition of the MTA.
Nonetheless, most investigators agree that it contains
the following components.
lMitochondria are a major source of ROS in the
cell.
lMitochondrially produced ROS inflict oxidative
damage on mtDNA.
lOxidative mtDNA damage results in mutations that
lead to defective electron transport chain (ETC) com-
ponents.
lIncorporation of defective subunits into the ETC
causes a further increase in ROS production, leading
to a ‘vicious cycle’ of ROS production and mtDNA
mutations.
lmtDNA mutations, ROS production and cellular
damage by ROS eventually reach levels that are
incompatible with life.
Despite its intellectual appeal, the MTA was not
well received initially [22], but until recently it has
enjoyed almost universal acceptance. However, recent
years have seen an abundance of experimental evidence
that contradicts the MTA in its present form. This
article critically reviews the evidence in support of, and
against, the MTA, by addressing each of the compo-
nents listed above, in turn.
Mitochondria are a major source of
ROS in the cell
The premise that mitochondria produce substantial
amounts of ROS appears to be valid and is rarely dis-
puted. Some researchers in the field have taken this
argument further, however, claiming that mitochondria
are the primary source of ROS in cells. This is based,
at least in part, on early estimates of mitochondrial
production of H
2
O
2
under nonphysiological conditions
[23]. It is important to note in this regard that cells
possess multiple enzyme systems capable of generating
ROS, and the relative contribution of each system,
which will probably depend on the cell type and physi-
ological state, has not yet been determined. Therefore,
it is impossible to state, a priori, that mitochondria are
the main source of ROS in every cell type and under
all physiological conditions [24].
Mitochondria possess at least nine enzyme systems
that are capable of producing ROS under favorable
conditions [25]. However, in the context of aging, only
ROS production by ETC complexes I and III is usually
considered. This is mostly because early studies estab-
lished that 1–2% of oxygen consumed by mitochondria
can be converted to H
2
O
2
. Considering the constitutive
nature of respiration, such a leak corresponds to a large
quantity of ROS, establishes mitochondrial ETC as a
major cellular source of ROS and establishes ROS as
compulsory by-products of respiration [23]. These find-
ings, however, were subsequently challenged by Hans-
ford et al. [26] who found that active H
2
O
2
production,
which is an indirect measure of superoxide (O
2) gener-
ation, requires both a high fractional reduction of
complex I, as determined by the NADH ⁄(NADH +
NAD
+
) ratio and a high membrane potential (DW).
The authors state that these conditions are achieved
only with supraphysiological concentrations of the
complex II substrate succinate. With physiological
concentrations of the NAD
+
-linked substrates that are
the main source of reduced equivalents for oxidative
phosphorylation, H
2
O
2
-formation rates are much
M. F. Alexeyev mtDNA + ROS = Aging?
FEBS Journal 276 (2009) 5768–5787 ª2009 The Author Journal compilation ª2009 FEBS 5769
lower, at less than 0.1% of the respiratory chain elec-
tron flux. Staniek and Nohl [27,28] also reported that
when mitochondria use complex I and complex II sub-
strates for respiration, detectable H
2
O
2
is generated
only in the presence of the complex III inhibitor anti-
mycin. They suggest that the rates of mitochondrial
H
2
O
2
production reported by other studies were artifi-
cially high because of experimental design flaws, and
point out that because mitochondrial O
2formation
under homeostatic conditions has not yet been demon-
strated in situ, conclusions drawn from isolated mito-
chondria should not be overinterpreted [28].
St Pierre et al. capitalized on these findings and used
an improved experimental design to show that mito-
chondria do not release measurable amounts of O
2or
H
2
O
2
when respiring on complex I or complex II sub-
strates, but release significant amounts of O
2from
complex I when respiring on palmitoyl carnitine [29].
However, even at saturating concentrations of palmitoyl
carnitine, only 0.15% of the electron flow is estimated
to give rise to H
2
O
2
. These results were obtained under
resting conditions with a respiration rate of 200 nmol of
electrons per min, per mg of mitochondrial protein.
Under physiological conditions, the rate is predicted to
be even lower because the partial pressure of oxygen,
the concentration of palmitoyl carnitine and the
mitochondrial membrane potential are all lower. The
authors conclude [29] that under physiological condi-
tions ROS are produced by ETC in quantities that can
be efficiently scavenged by mitochondrial antioxidant
systems. They proposed that as long as cells have nor-
mal levels of antioxidants, an electron leak from the
ETC should not result in significant oxidative damage
to mitochondrial components, including mtDNA. This
conclusion is consistent with observations from trans-
genic animal models showing that overexpression of
ROS-scavenging enzymes generally does not extend life
span and can even be detrimental (discussed later).
The highest production of ROS by mitochondria
in vitro was observed under conditions of reverse elec-
tron transfer (RET) from complex II through complex
I, towards NAD
+
. This flow is thermodynamically
unfavorable and must be coupled to the expenditure of
the energy of membrane potential. This energy is maxi-
mal when ADP supply is limited (state 4 respiration),
or when electron flow through complex III is blocked
by antimycin. Under these conditions, the dependence
of ROS production on the membrane potential is so
great that a 10% drop in membrane potential results in
a 90% reduction in ROS production ([30,31]; reviewed
in [25]). Although the feasibility of RET in vivo remains
to be fully elucidated, this possibility cannot be com-
pletely excluded [32,33]. Nonetheless, even if RET
occurs physiologically, current evidence suggests that it
may occur only intermittently, under a narrow set of
conditions. While it is plausible that RET may generate
significant quantities of ROS in mitochondria under
certain circumstances, it is currently unclear whether or
not it can lead to a lifelong accumulation of mtDNA
mutations, as specified by the MTA.
Mitochondrially produced ROS inflict
oxidative damage on mtDNA
In vitro, DNA damage by ROS exposure is well docu-
mented [34–40], but in vivo, mitochondria possess mul-
tiple and redundant ROS scavenging systems. mtDNA
damage by ROS requires oxidative stress, an imbal-
ance between ROS production and ROS neutraliza-
tion. The mitochondrial pathways for ROS generation
and scavenging are briefly considered here.
Mitochondrial ROS generation
The proximal ROS generated by electron leak from
the ETC is O
2(Fig. 1 and Eqn 1), which is charged,
comparatively unstable and has relatively low reactiv-
ity. The negative charge has been proposed to render
O
2impermeable to membranes [41], and this hypothe-
sis is supported by results obtained from studies using
thylakoid and phospholipid liposome membranes
[42–44]. The permeability of the mitochondrial inner
membrane to O
2is one of the factors that determines
the accessibility of the agent to mtDNA. Therefore,
50% of O
2generated at complex III has no access
to mtDNA, while all O
2generated at complex I has
unimpeded access to it [41]. Although O
2permeates
erythrocyte ghost membranes through an anion chan-
nel [45], no evidence exists for a similar channel in the
inner mitochondrial membrane, which is probably
impermeable to this species.
Fig. 1. A major pathway for the detoxification of ROS in the mito-
chondrial matrix. O
2is formed by the reduction of O
2
with elec-
trons leaked from the ETC. O
2is efficiently converted to H
2
O
2
by
mitochondrial superoxide dismutase (Sod2). H
2
O
2
is then detoxified
to H
2
O either by mitochondrial glutathione peroxidase (GPx1) with
concomitant oxidation of glutathione (GSH), or by peroxiredoxins III
and V (PrxIII and PrxV). GSH, reduced glutathione; GSSG, oxidized
glutathione.
mtDNA + ROS = Aging? M. F. Alexeyev
5770 FEBS Journal 276 (2009) 5768–5787 ª2009 The Author Journal compilation ª2009 FEBS
In fact, however, the membrane permeability of O
2
may be of little consequence because it is unable to
react directly with DNA [46–50]. Reaction of O
2with
nonradicals is spin forbidden. In biological systems,
this means that the main reactions of O
2are with
itself (dismutation) or with another biological radical,
such as nitric oxide.
One important feature of O
2production by mito-
chondria is that it can be self-limiting through the
inactivation of mitochondrial aconitase. This inactiva-
tion can reduce NADH formation by the citric acid
cycle and, consequently, electron flow through the
ETC. The net effect would be a lowering of the
steady-state levels of reduction of complexes I and III,
which would diminish O
2production [51,52].
Mitochondrial ROS neutralization
The O
2generated by the ETC is quickly converted to
H
2
O
2
(Fig. 1 and Eqn 2), which is the principal cellular
mediator of oxidative stress. This conversion occurs
either spontaneously, with a second-order rate constant
of approximately 10
5
m
)1
s
)1
, or enzymatically, cata-
lyzed by superoxide dismutases, with a first-order rate
constant of 10
9
m
)1
s
)1
[53]. Mitochondria possess two
superoxide dismutases: Sod1 (Cu ⁄ZnSod) in the inter-
membrane space; and Sod2 (MnSod) in the matrix.
Intriguingly, Sod1 appears to exist in an inactive,
reduced form that can be activated by ETC-generated
O
2[54]. The relative stability and membrane perme-
ability of H
2
O
2
allows it free access to mtDNA, yet, like
O
2, it is also unable to react directly with DNA [46–
50]. However, in the presence of redox-active metal ions,
such as Fe
2+
,H
2
O
2
can undergo Fenton chemistry
(Eqn 3), generating the extremely reactive hydroxyl rad-
ical
•
OH that efficiently damages DNA [34,35]. To pre-
vent the potentially devastating consequences of the
Fenton reaction, H
2
O
2
is detoxified in the mitochondrial
matrix by glutathione peroxidase 1 (GPx1; Fig. 1 and
Eqn 4) and peroxiredoxins III and V (PrxIII and PrxV;
Fig. 1 and Eqns 5 and 6, respectively; [55]). At least
seven GPx enzymes have been described to date in
mammalian cells [56], and two – GPx1 and GPx4
(PHGPx4) – are ubiquitously expressed [56–58]. GPx1 is
found in both the cytosol and the mitochondrial matrix,
and its preferred substrate is H
2
O
2
. GPx4 is most effi-
cient at reducing lipid hydroperoxides. In addition to
direct inactivation of ROS, GPx enzymes indirectly pro-
tect the cell from damage by the O
2, by preventing per-
oxide-mediated inactivation of Sod1 [59]. Interestingly,
Sod itself protects GPx from inactivation by O
2[60].
Thus, Sod and GPx may participate in a cross-
protection that prevents their inactivation by ROS.
The family of mammalian Prxs has at least six mem-
bers, of which PrxIII and PrxV are mitochondrial.
PrxIII is found only in mitochondria and is about
30-fold more abundant than GPx1 in HeLa cell
mitochondria [61]. PrxV is expressed as a long and
short forms, which are found in the mitochondrion
and in peroxisomes, respectively [62–64]. Catalase has
been reported in rat cardiac mitochondria [65], but this
was not confirmed in a follow-up study [66]. Therefore,
GPx1, and PrxIII and V are the main, and probably
only, contributors to H
2
O
2
detoxification in the mito-
chondrial matrix (Fig. 1).
O2þe!O
2ð1Þ
2O
2þ2Hþ!H2O2þO2ð2Þ
Fe2þþH2O2!Fe3þþOH þOHð3Þ
H2O2þ2GSH !GS SG þ2H2Oð4Þ
H2O2þ2Pr xIII(SH)2!2H2OþPr xIII(SH)
SS(SH) Pr xIII ð5Þ
H2O2þPr xV(SH)2!2H2OþPr xV(S S) ð6Þ
7,8-Dihydro-8-oxo-2¢-deoxyguanosine as a marker of
oxidative mtDNA damage
The main pyrimidine product of oxidative DNA base
damage is thymine glycol [67] and the main purine prod-
uct is 7,8-dihydro-8-oxo-2¢-deoxyguanosine (8-oxodG)
[68–70]. The former has low mutagenicity, while the lat-
ter, upon replication, can cause characteristic G:T trans-
versions at a relatively low frequency [71]. Initial studies
revealed that mtDNA accumulates approximately 15
times more 8-oxodG than nuclear DNA (nDNA), thus
establishing extensive mtDNA damage by ROS under
physiological conditions [72]. These studies also sug-
gested potential causes for the increased sensitivity of
mtDNA to oxidative stress, which include the proximity
to the source of ROS, the lack of protective histones
and relatively inefficient mtDNA repair. Each of these
causes is examined in more detail below.
Proximity of mtDNA to the ETC and steady-state
oxidative damage
The hypothesis that mtDNA is at greater risk to oxida-
tive damage than nDNA because it is close to the source
of ROS was logical, especially when early reports sug-
gested that mtDNA contained higher levels of oxidative
M. F. Alexeyev mtDNA + ROS = Aging?
FEBS Journal 276 (2009) 5768–5787 ª2009 The Author Journal compilation ª2009 FEBS 5771
lesions than nDNA [72]. However, revision of the initial
data no longer supports this conclusion [73–75]. In any
case, oxidative damage resulting from proximity to the
ETC is only possible if protection by antioxidant
defenses and DNA repair are inadequate.
Lack of histones in mitochondria and susceptibility of
mtDNA to oxidative stress
Histone proteins are reported to protect DNA from a
variety of potentially dangerous reactive species, such as
•
OH [76–78]. Mitochondria lack histones, and this is
cited as a possible reason for the higher susceptibility of
mtDNA to ROS damage. However, nucleosome pack-
aging does not protect DNA from the damage caused
by charge transfer through base pair stacks [37,79].
Electron transfer occurs easily from histones to DNA,
leading to DNA damage [80]. In addition, damage
induced by Cu
2+
⁄H
2
O
2
is enhanced in nucleosomal
DNA compared with naked DNA [37], and some DNA–
peptide interactions can increase metal ⁄H
2
O
2
-induced
DNA breakage [81]. Therefore, histones are protective
under some, but not all, conditions. In addition, a
recent study demonstrated that protein components of
mitochondrial nucleoids show the same protection as
histones, under conditions in which histones protect
against oxidative stress [82]. This is in agreement with
a report that mitochondrial transcription factor A
(a DNA-binding protein and a major component of
mitochondrial nucleoids) is present in mitochondria in
quantities sufficient to completely cover mtDNA [83].
Repair of oxidative base lesions in mitochondria
The discovery that mitochondria are unable to repair
UV-induced pyrimidine dimers [84,85] and some types
of alkylating damage [18], demonstrated that they con-
tain a reduced complement of DNA-repair pathways.
However, Anderson and Friedberg [86] found uracil-
DNA glycosylase activity in mitochondrial extracts,
suggesting at least the presence of the base excision
repair (BER) pathway. This was confirmed by mito-
chondrial repair of O
6
-ethyl-2¢-deoxyguanosine, which
is also processed by BER in the nucleus [87,88]. Subse-
quently, repair of a variety of mtDNA lesions, includ-
ing those arising from oxidative damage, was
demonstrated [89–98]. Recently, long-patch BER of
oxidative DNA lesions [99–101], and mismatch repair
[102], have been discovered in mammalian mitochon-
dria, so to date, no specific defect in the mitochondrial,
as compared to nuclear, repair of oxidative damage
has been reported. BER, with its single-nucleotide and
long-patch subpathways, is the main pathway for
repairing oxidative base lesions in both the nucleus
and mitochondria, and 8-oxodG, the most prominent
oxidative base lesion, is repaired more efficiently in
mitochondria than in the nucleus [103].
Accumulation of oxidative damage in mtDNA
compared with nDNA
The report that mtDNA has a greater 8-oxodG content
than nDNA was quickly followed by the report of an
age-dependent increase of this lesion in cellular DNA
[104]. However, a decade later, the same group reduced
the estimates of cellular 8-oxodG by an order of magni-
tude, after finding that the isolation procedure used in
earlier studies resulted in the artificial oxidation of
DNA [105]. Nevertheless, the steady-state level of 8-ox-
odG in the DNA of old rats was almost three times
higher than that of young animals [105], and 8-oxodG
became widely accepted as a marker of oxidative DNA
damage. Reported values for the baseline 8-oxodG con-
tent of mtDNA span almost five orders of magnitude,
however, and the lowest reported values are not signifi-
cantly different from those reported for nDNA [106]. A
series of carefully designed studies established that the
endogenous oxidative damage of mtDNA is not greater
than that of nDNA [73–75], and one study showed that
some oxidative lesions (including 8-hydroxyguanine,
Fapy-adenine, 8-hydroxyadenine, 5,6-dihydroxyuracil,
5-hydroxyuracil, 5-hydroxycytosine and 5-hydroxym-
ethyluracil) are found less often in mtDNA [73].
Yakes and Van Houten [107] reported that the
mtDNA of mouse embryonic fibroblasts exposed to
H
2
O
2
had more polymerase-blocking lesions than
nDNA. These lesions are predominantly strand breaks
that are generated, either directly or indirectly, through
the action of mitochondrial apurinic and apyrimidinic
endonuclease at abasic sites, or through the action of
bifunctional glycosylases on oxidatively damaged
DNA bases. In any case, this apparent increase in the
susceptibility of mtDNA to oxidative damage may in
fact be part of a mitochondria-specific mechanism that
protects mtDNA integrity through the degradation of
severely damaged mtDNA molecules (discussed later).
Oxidative mtDNA damage results in
mutations that lead to defective ETC
components
The mitochondrial genome accumulates mutations
approximately one order of magnitude faster than
nDNA [108–110]. This could be caused by a variety of
factors, including an intrinsically lower fidelity of repli-
cation by mitochondria-specific DNA polymerase c
mtDNA + ROS = Aging? M. F. Alexeyev
5772 FEBS Journal 276 (2009) 5768–5787 ª2009 The Author Journal compilation ª2009 FEBS