Introduction
Evidence from many paths of
gerontological research suggests that mitochondria are one of the key players
in aging. Not only do the number of mitochondria decrease in postmitotic
tissues like heart, skeletal muscle and brain during aging [1,2], but a
number of age-associated structural changes of mitochondria has been reported
as well [1,3]. The most
common aging-associated structural changes of mitochondria are alterations of
cristae structure, but also matrix vacuolization or densification and a general
enlargement of mitochondria have been found to increase with age [4-6].
Mitochondrial DNA (mtDNA) point mutations and deletions have been found in
aging humans, and in many cases, have been correlated to
mitochondrial dysfunction. Higher levels of mtDNA deletions were detected in
postmitotic tissues e.g., skeletal muscle, heart and brain in contrast to
kidney, spleen, skin and liver that are composed of dividing cells, indicating
a tissue-specific "pattern" of deletion accumulation [7]. However,
their possible causative effects have been intensely debated because of their
low abundance (under 1% of deleted mtDNA molecules in whole tissue homogenate)
and purely correlative connection with aging [8]. A real
breakthrough occurred when it was discovered that mitochondrial dysfunction is
unevenly distributed in tissues and that mtDNA point mutations and deletions
accumulate in some single cells. For instance, the analysis of five tRNA
genes showed high levels of clonally expanded mtDNA point mutations (49%-94%)
in COX-deficient skeletal muscle fibers isolated from aged individuals [9]. In
addition, point mutations associated with the major regulatory region for
replication of mtDNA were found in a high proportion (up to 50%) in skin
fibroblasts or skeletal muscles of individuals above 65 years of age [10,11].
To test the causative role of mtDNA mutations in aging
we have developed the mtDNA mutator mouse that accumulates high levels of point
mutations due to a proofreading deficiency of the mitochondrial DNA polymerase
(POLG) [12]. Subsequently,
a very similar model was developed by another group [13]. The two
mouse models show basically the same phenotypes, differing only in the time of
onset of phenotypes. In our hands, mtDNA mutator mice are born in Mendelian
proportions, without any visible defects, but after 6-7 months they start to
display a range of premature aging phenotypes, such as a weight loss, reduced
subcutane-ous fat, alopecia, kyphosis, osteoporosis, anaemia, reduced
fertility, heart disease, sarcopenia, progressive hearing loss and decreased
spontaneous activity [12]. Their
lifespan is also greatly reduced compared with wild-type littermate controls
and they die at around 46 weeks of age [12].
In addition to the standard circular chromosome, mtDNA
mutator mice harbour large linear mtDNA molecules caused by replication
stalling [14]. These
molecules are around 11-12 kb in length, and encompass the region between the
origins of replication for the heavy (OH) and the light strands (OL).
Around 25-30% of mtDNA consists of these linear molecules with deletion, and
this ratio does not change as the animal age [12]. As these
molecules exist at quite high levels, it has been proposed that this may also
contribute to the progeroid phenotype of the mtDNA mutator mouse [12,14]. The
POLG exonuclease activity may be involved in the resolution of replication
intermediates at OL, which can explain why exonuclease deficiency
will lead to replication stalling, thus leaving the mtDNA molecule susceptible
to breakage at the stall site [14]. It has
also been suggested that the time needed for the replication of mtDNA could be
prolonged in the mtDNA mutator mouse, as evidenced by a high amount of
replication intermediates [14]. This
prolonged replication is proposed to exhaust resources in the mitochondria.
Although plausible, this hypothesis still requires direct experimental
verification [14].
The main consequence of the mitochondrial
proof-reading deficiency in mtDNA mutator mice is the progressive, random
accumulation of mtDNA mutations during the course of mitochondrial biogenesis.
Table 1. The number of mtDNA point mutations/10000 base pairs detected by different groups using the cloning and sequencing (CS) and the random mutation capture approach (RMC).
Genotype
|
Brain ~8 weeks
|
Heart ~25 weeks
|
WT
|
0,33(CS)[12]
0,0066 (RMC)[15]
|
4,58 (CS)[12]
2,11 (CS)[13]
|
HZ
|
1,33 (CS) [12]
3,3 (RMC) [15]
|
-
-
|
MUT
|
9 (CS)[12]
14 (RMC)[15]
|
14 (CS)[12]
10 (CS)[13]
|
Different approaches have been used to measure levels
of somatic mtDNA mutations and the results often vary considerably depending on
the method used, e.g. cloning-sequencing (CS) or random mutation capture (RMC)
(Table 1.). Both of these methods seem to detect similar differences in levels
of mutations in mice heterozygous (Polg+/mut) or homozygous (Polgmut/mut)
for the mtDNA mutator allele. However, the RMC method detects a 500-fold less
mtDNA mutation load in control mice, i.e. wild type animals obtained after
intercrossing mice het erozygous for the mtDNA mutator allele (Table 1.). The
CS method is the most commonly used procedure to estimate the level of mtDNA
mutations and involves PCR amplification of mtDNA, followed by cloning and
sequencing of the obtained fragment. A problem with this method is that PCR
might introduce mutations in the initial step, which are then carried through
to the sequencing reaction. The RMC was developed by Vermulst et al. [15] and is
based on the fact that mtDNA mutations will alter a given restriction enzyme
recognition site. After restriction enzyme digestion only mutated molecules
should be amplified by PCR and those are subsequently quantified and sequenced.
However, the target sequence analyzed by this method is only 4 bases long (Taq
I recognition site) and may therefore not represent the rate of mutations
over the entire molecule [15]. The RMC
assay appears to work very well on large sample sizes when the mutations are
evenly distributed throughout the genome [16]. However,
when existing mutations are few but have been amplified, they might be missed.
MtDNA mutations can clonally expand in certain tissues, i.e. when a stem cell
gives rise to many cells. This can also happen
randomly when mtDNA molecules are segregate-ed or replicated unevenly. In
human colonic mucosa (a tissue known to clonally expand mtDNA mutations) the
RMC routinely detects between 10-1000 times less mutations than the CS approach
and around 5 times less than with single-molecule-PCR based method [16].
The analysis of mtDNA mutation
load in wild-type animals will be complicated when the animals are derived from
a mother heterozygous for the mtDNA mutator allele. The reason is that the
primordial germ cells of a heterozygous mother contain one copy of the mutant
POLG allele and this could cause accumulation of a small amount of mtDNA
mutations during their proliferation and maturation into oocytes. A genetic
bottleneck occurs after the oocyte is fertilized whereby a small subset of all
mtDNA molecules are partitioned into the primordial germ cells of the next
generation [17]. Subsequent rounds of
replication in the wt embryo are then carried out by only wt POLG, whereas in
the mtDNA mutator embryo proofreading-deficient POLG is used. As a result, the
embryo with the mutator allele will continue to introduce new mutations,
whereas the wt embryo will amplify only the pre-existing mutations with high
fidelity. This may explain why wt littermates of mtDNA mutator mice may have a
moderately increased mutation load in comparison with other wild-type mice whose
mitochondria have never come in contact with the mtDNA mutator allele.
Heterozygous animals from the same cross might continue to accumulate random
mtDNA mutations, but to a much lesser extent than mtDNA mutator mice. The
pedigree of the analyzed wild-type mice is therefore an important factor to
consider when determining mtDNA mutation loads.
In order to resolve this issue, we
believe that a direct cloning and sequencing approach should be used. This
would involve the cloning of large fragments or an entire mtDNA molecule into a
vector, without the use of an intermediate PCR step [18]. In this
way PCR-induced errors will be avoided and a more complete picture of the
distribution of mutations on the population of mtDNA molecules will be
achieved. Direct cloning and sequencing has 3 major drawbacks. First, it is
labour intensive. Second, it would require a large sequencing effort to sample
the same magnitude of bases as a RMC assay, making sequencing costs
prohibitive. Finally, cloning of small amounts of sample DNA is often
inefficient and studying single cells is therefore not possible at the moment.
Alternatively, the cloning step could be circumvented by usage of newly
developed deep sequencing techniques that could be quantitative simply because
the same region of the mtDNA would be sequenced over and over again. However,
many deep sequencing methods have inherent error rates, which may obscure
somatic mtDNA mutation loads.
In a recently published paper,
Loeb and co-workers suggest that a third type of mtDNA mutation, besides the
very abundant point mutations and linear deletions we have reported [12] are the main
driving force behind the shortened lifespan in mtDNA mutator mice [19]. This third
type of mutation is circular mtDNA molecule with large deletions of several
thousand base pairs. The main argument for this conclusion is that the RMC
method detects a 500-fold increase of mtDNA point mutations in heterozygous,
Polg+/mut in comparison with wild-type mice, despite the fact the
heterozygous mice have a normal lifespan [15]. The
authors also reported that circular mtDNA molecules with deletions are highly
increased in frequency throughout the lifetime of the mtDNA mutator mouse.
Similar molecules are also found to accumulate in the wt and Polg+/mut
mice , but to a much lesser extent (10- 100 times lower). In light of this
evidence they conclude that the circular mtDNA molecules with deletions
correlate much better with the phenotype and therefore must be the driving
force behind the premature aging of mtDNA mutator mice [19]. However,
their study is purely correlative and it does not take into account threshold
levels that are the imperative in mitochondrial genetics. In their study, they
have estimated only the relative amount of circular mtDNA molecules with
deletions between genotypes. To estimate the importance of these circular mtDNA
molecules with deletions it is necessary to know what fraction of the total
mtDNA they represent.
We recently attempted to detect these deletions using
long extension PCR technique. As controls, we used two strains of mice with
known amounts of deletions, the mito-mouse and the deletor mouse [20,21]. These
mouse strains carry circular mtDNA molecules with single (mito-mice) or
multiple (deletor-mice) deletions in their mitochondria. Although we have
robustly detected deleted mtDNA molecules in serial dilutions of these control
samples we were not able to detect any deletions in mtDNA mutator heart or
liver samples. Recently, Kraytsberg et. al used single molecule long extension
PCR to detect deletions in mtDNA mutator samples provided by Loeb and
co-workers [22]. They did
not detect deleted mtDNA among more than 320 amplified molecules from duodenum
and 144 molecules from heart of mtDNA mutator mice [22]. This
completely agrees with our data obtained from heart and liver of mtDNA mutator
mice. Kraytsberg et. al detected some mtDNA deletions in brain tissue, but
estimate the relative levels to be around 1% and they argue that this could be
an overestimation [22].
It is unclear how these rare circular molecules with
deletions could be the major driving force behind the shortened lifespan in the
mutator mice. The mitomice with deletions in up to 30% of all mtDNA molecules
do not die early, whereas those with deletions in 70% suffer from renal failure
[23].
The deletor mouse, although generating multiple deletions of the same range as
proposed for mtDNA mutator mice, show no premature aging phenotype. The
deletor mice have normal lifespan but develop mitochondrial myopathy late in
life [21].
Furthermore, deletions are not related to Parkinson's disease until they reach
a threshold of around 52% in dopaminergic neurons of humans [24]. The same
is true for most diseases caused by mutations in mtDNA. There is always a
threshold level of mutation that has to be reached before the mutation causes
respiratory chain dysfunction. This threshold level will vary depending on
mutation, but is generally between 60% (for deletions) - 90% (point mutations) [25]. In the
mtDNA mutator mouse the mutations are spread out over the genome and a very
high threshold level of a particular mutation might not be reached. However, if
every molecule accumulates approximately 20 mutations, many of these will be in
protein coding genes. For COXI for example there is an 88% chance that the gene
contains at least 1 mutation in a specific mtDNA molecule and a 61% chance that
it contains at least 2 mutations of which 40% will cause amino acid change.
Then there are another 12 genes for which numbers are slightly different
depending on the size of the genes. Assuming random distribution of the
mutations, a threshold level will be reached easily, not by one specific
mutation, but by the sum of all different mutations. Deletions, though
affecting many base pairs, still need to be at significant levels to cause
mitochondrial dysfunction. It has been estimated that there is about 1000 point
mutations for every single deleted mtDNA molecule in mtDNA mutator mice (if
there is >0.02 deletion per mtDNA molecule) [22]. This also means that mtDNA mutator mice have about 800
additional point mutations for every additional deletion when compared to Polga+/mut
animals [22].
Loeb and co-workers
do not report increased amounts of deletions in COX-negative cells (cytochrome
c oxidase deficient cells - indicating mitochondrial dysfunction), only an
increase in point mutations [19]. Large circular mtDNA deletions
typically span over several protein coding genes and include a number of tRNA
genes. Therefore, consequences of the large deletions could be detected as a
reduction in mitochondrial transcript levels or as a decrease in translation of
mitochondrially-encoded proteins. This, however, does not seem to be the case.
We found that the pattern of transcriptional change was more consistent with
increased activity from the heavy strand promoter, rather than a loss of
components via deletions. Translation was also normal, but we did observe
increased protein turnover [26]. So the mechanism by which these
rare, circular mtDNA molecules with deletions would be causing the premature aging
phenotype remains unclear.
So what have we learned from the mtDNA mutator mouse?
Are mtDNA mutations relevant in normal aging? Clearly wt mice never accumulate
the same level of point mutations as mtDNA mutator mouse. They do, however,
acquire mitochondrial dysfunction with old age, as do the mtDNA mutator mice.
This indicates mitochondrial dysfunction as not only being correlated to aging,
but also causative. Results from mtDNA mutator mice clearly show that mutations
in mtDNA can cause problems that resemble premature aging. Of these
mutations, the amino acid changing mutations in protein coding genes seem to be
the most deleterious [26]. This is in
agreement with a recent finding than non-synonymous mutations that change amino
acids in respiratory chain subunits are strongly selected against and very
rarely emerge in future generations, whereas tRNA and rRNA mutations seemed to
have less severe functional consequences and could easily be inherited [27]. In
addition to loss of function mutations there could be dominant deleterious
mutations, like the C>T mutation at nucleotide 5545 of mtDNA that cause
pathology at very low levels [28]. One
attractive candidate for the failure of the organism to thrive comes from the
finding that stem cell function in the mtDNA mutator mice seems to be altered
(Ahlquist et al. unpublished results). On the organismal level, it remains to
be determined what type of changes and in which cells have the most important
effects. Additional studies on mtDNA mutator mice in conjunction with old
control mice might be able to dissect out which changes are important and which
are just correlative in the way mitochondria affect the aging process.
A.T. is supported by grants from the Deutsche Forschungsgemeinschaft, Vetenskaprådet, Åke Wiberg Foundation, and
Loo&Hans Ostermans Foundation
The authors declare that
they have no competing financial interests related to this manuscript.