|Home | About | Journals | Submit | Contact Us | Français|
Age-dependent accumulation of partially deleted mitochondrial DNA (ΔmtDNA) has been suggested to contribute to aging and the development of age-associated diseases including Parkinson’s disease. However, the molecular mechanisms underlying the generation and accumulation of ΔmtDNA have not been addressed in vivo. In this study, we have developed a mouse model expressing an inducible mitochondria-targeted restriction endonuclease (PstI). Using this system, we could trigger mtDNA double-strand breaks (DSBs) in adult neurons. We found that this transient event leads to the generation of a family of ΔmtDNA with features that closely resemble naturally-occurring mtDNA deletions. The formation of these deleted species is likely to be mediated by yet uncharacterized DNA repairing machineries that participate in homologous recombination and non-homologous end-joining. Furthermore, we obtained in vivo evidence that ΔmtDNAs with larger deletions accumulate faster than those with smaller deletions, implying a replicative advantage of smaller mtDNAs. These findings identify DSB, DNA repair systems and replicative advantage as likely mechanisms underlying the generation and age-associated accumulation of ΔmtDNA in mammalian neurons.
Deletions of mitochondrial DNA (mtDNA), which encodes essential components of oxidative phosphorylation (OXPHOS), have been observed in aged tissues of various species (1). The characterization of partially-deleted mtDNA (ΔmtDNA) in aged human muscle and brain tissues revealed that deletions occur preferentially between the origin of L-strand replication and the D-loop (2,3). Although the accumulation of ΔmtDNA has been proposed to have a role in aging (4), the levels of such mutated mtDNA in aged mammalian tissues were estimated to be low, and their causal role in aging processes has been questioned.
At the cellular level, however, a different picture has emerged. In muscle biopsy of normal elderly subjects, 0.1–5% of muscle fibers were found to be deficient of cytochrome c oxidase (COX), the terminal enzyme complex of mitochondrial respiratory chain (5). Nearly 7% of the COX-deficient fibers were found to contain a high level (>90%) of ΔmtDNA. Bua and colleagues estimated that ~6% of muscle fibers from a 49-year-old individual, ~22% of muscle fibers from a 67-year old individual, and ~31% from a 92-year old individual contain OXPHOS abnormalities somewhere along their length (2). In the central nervous system, the levels of ΔmtDNA molecules can exceed 50% of the total mtDNA in some neurons of the aged substantia nigra (3,6). Patients with Parkinson’s disease tend to contain an even higher amount of ΔmtDNA (6). A similar accumulation of ΔmtDNA at the cellular level has been also obverved in rodents (7,8). Therefore, the amount of ΔmtDNA can reach a substantial level in a subset of cells in aged tissues and may contribute to age-associated physiological and pathological changes through the impairment of oxidative phosphorylation. Understanding the molecular basis of generation and accumulation of ΔmtDNA during aging might lead to the development of novel interventions that prevents or delays the onset of age-associated diseases.
To better understand the molecular mechanisms underlying the generation and age-associated accumulation of ΔmtDNA in vivo, we developed a transgenic mouse model expressing a mitochondria-targeted bacterial PstI gene (mitoPstI) in a subset of CNS neurons in a doxycycline (Dox)-regulated manner. The advantage of our present model is the temporal and spatial regulation of mitoPstI expression, which allowed us to monitor the age-associated accumulation of ΔmtDNA upon the introduction of double-strand breaks (DSBs) by a transient expression of mitoPstI. DSB or closely spaced single-strand breaks can be induced by endogenous factors including reactive oxygen species and the following DNA repairing processes are responsible for the recombination of damaged nuclear DNA (9). Recently, it was proposed that DSB play an important role in generating deletions in mtDNA (10).
We have previously shown that constitutive expression of mitochondria-targeted restriction endonucleases can lead to efficient cleavage and degradation of mtDNA (11,12). To study the consequences of a limited exposure of mtDNA to DSB, without causing mtDNA depletion, we created a mouse model expressing an inducible mitochondria-targeted PstI (mitoPstI) by placing its gene under the control of an inducible (Tet) promoter. Prior to the generation of mitoPstI transgenic mice, we transiently transfected tet-off HeLa cells constitutively expressing the tetracycline transactivator (tTA) with a mitoPstI construct, which drives the expression of mitoPstI through a CMV minimal promoter regulated by a tetracycline-responsive element (TRE). In the absence of Dox, mitoPstI was strongly expressed in a subpopulation of tet-off HeLa cells (Fig. 1A). Co-staining of cells with Mitotracker Red confirmed that the expressed mitoPstI co-localized with mitochondria (Fig. 1A).
Mice expressing neuronal-specific mitoPstI were generated by crossing: (i) a mouse carrying mitoPstI controlled by TRE and (ii) a mouse carrying tTA gene controlled by the CamKIIα promoter (13). In the resulting double transgenic mice, tTA is expressed in a subset of neurons in the CNS, mostly cortex and hippocampus. In the absence of doxycycline (Dox), CamKIIα-tTA drives the expression of mitoPstI in these regions (Fig. 1B). Once expressed, mitoPstI targeted to the matrix of the mitochondria is expected to introduce DSB in the mouse mtDNA at nucleotide (nt) positions 8420 and 12 238 (Fig. 1B).
Four out of five transgenic lines obtained successfully transmitted mitoPstI transgene to their progenies and were crossed to CamKIIα-tTA mice. Double transgenic mice carrying both mitoPstI gene and tTA gene (mitoPstI/tTA) as well as littermate controls were grown in the absence of Dox from the beginning of pregnancy (E0) and were perfused at 10 weeks of age for the subsequent expression analysis of mitoPstI. Expression of mitoPstI in the cortex was most robust in Line 1, weaker in Line 2, and undetectable in Lines 3 and 4 (Fig. 1C). Therefore, we decided to use Line 1 in this study. Expression of mitoPstI in the neuronal population was validated by immunostaining of brain sections using anti-PstI antibody (Supplementary Material, Fig. S1). The expression of mitoPstI was observed in the cortex and all subregions of the hippocampus, especially the dentate gyrus and the CA1 region.
When the double transgenic mice were grown in the absence of Dox, they developed an abnormal ‘limb-clasping behavior’ at 2 months of age, a phenotype previously observed in mouse models of movement or neurodegenerative disorders (14–16) (Fig. 2A). However, when initially grown on Dox from E0 to P21, none of the double transgenic mice exhibited the abnormal limb-clasping behavior in their adulthood even in the absence of Dox, though the double transgenic mice progressively became less active after the removal of Dox and died before 100-day-old (n = 7) (Fig. 2B). None of the single transgenic mice (mitoPstI+ or tTA+ mice) and Dox-treated double transgenic mice exhibited any behavioral abnormalities or reduced lifespan (data not shown).
To confirm that mitoPstI is expressed, functional and capable of reducing the level of mtDNA, first we examined the level of mtDNA-encoded COX I protein, a catalytic subunit of COX (cytochrome c oxidase, or complex IV), in the brain tissues of the above 10-week-old double transgenic mouse (Line 1) grown in the absence of Dox. This experiment revealed that COX I protein is selectively reduced in the tissues expressing mitoPstI but not in the cerebellum (Fig. 2C), implying that mitoPstI is functional. To directly confirm that the expressed mitoPstI is functional, southern blots were performed to quantify the levels of total mtDNA in different brain regions of the double transgenic mouse and the control single transgenic mouse. Consistent with the COX I western blot, the levels of mtDNA (normalized by the levels of 18S rDNA, a nuclear DNA) were selectively reduced in the cortex, hippocampus and striatum of the double transgenic mouse, down to ~30% of the total mtDNA in control tissues (Fig. 2D), confirming that the expressed mitoPstI is functional. To study whether the reduced mtDNA has a functional consequence on the activity of OXPHOS, we performed in situ COX activity staining on parasagittal brain sections of the double and single (mitoPstI+) transgenic mice. This experiment revealed that COX activity is broadly reduced in the forebrain but not in the cerebellum of the double transgenic mouse (Supplementary Material, Fig. S2). Collectively, these experiments demonstrated that the expression of mitoPstI is functional, reducing the levels of mtDNA, mtDNA-encoded OXPHOS subunits and OXPHOS activity.
To study the regulation of mitoPstI expression by Dox, we initially raised double transgenic mice under a continuous exposure to Dox. During embryonic and neonatal stages, they received Dox through their pregnant and lactating mother consuming Dox-containing diet. When they reached 4 weeks of age, we allowed for a short induction of mitoPstI by removing Dox for 3 days and then re-suppressing it by re-introducing Dox in their diet (Fig. 3A). Mice were 5- or 14-week-old when they were analyzed. The 3-day period without Dox was sufficient to transiently induce mitoPstI expression (Supplementary Material, Fig. S3A), and the 1-week suppression following the 3-day transient induction was sufficient to reduce the mitoPstI to undetectable levels (Supplementary Material, Fig. S3B). Southern blots showed that, under these conditions, there was no significant reduction in mtDNA levels in cortical tissues at 14 weeks (Fig. 3B) or 5 weeks (data not shown). No overt behavioral defects were observed in any experimental groups at either age.
Southern analysis with a probe complementary to nt 3500–4200 region of mtDNA (Fig. 3B) showed the presence of a weak band of ~9.3 kb in one of the 14-week-old cortical samples that underwent the transient induction of mitoPstI. This band was absent in the control samples and 5-week-old induced samples (Fig. 3B). When the samples were analyzed with a probe complementary to nt 8931–10 660, the 9.3 kb band, which might have originated from ΔmtDNA was not discernible in induced samples (Fig. 3B), implying that the mtDNA species lack the 8931–10 660 region. To better characterize ΔmtDNA species, we performed PCRs using the 8300-forward and 12 812-backward primers, flanking the two PstI sites of mouse mtDNA. This PCR was expected to amplify 4.5 kb bands in control samples in which there was no recombination of mtDNA. However, if there was recombination in mtDNA somewhere between nt 8300 and 12 812, this primer set would preferentially amplify a shorter breakpoint region. As expected, we observed the amplification of ~4.5 kb bands in control cortical DNA from 14-week-old tTA+ single transgenic mice (Fig. 3C). On the other hand, this PCR yielded smaller bands of ~670 bp in mitoPstI-induced samples (Fig. 3C). Sequencing analysis of the ~670 bp bands revealed that these bands originated from a family of ΔmtDNA with ~3.8 kb deletions (Fig. 3D). One of the three deletions identified in this analysis (Deletion 2 in Fig. 3D) was characterized by the presence of 8 bp direct repeat at the breakpoint and was likely generated by homologous recombination. Another deletion (Deletion 3 in Fig. 3D) contained no direct repeats at the breakpoint. This deletion appeared to have been generated by non-homologous end-joining (NHEJ) of two free ends generated by PstI digestion following trimming of single-strand ends by endogenous 3′–5′ exonuclease activity. The other deletion (Deletion 1 in Fig. 3D) contained a 2 bp small repeat at the breakpoint region. Interestingly, the breakpoint of this deletion, formed by nt 8404 and 12 245 of wild type mtDNA, were slightly away from the original PstI-recognition sites (nt 8420 and 12 238).
To identify additional deletions in the induced cortical tissues, we designed a new set of primers, 5399-forward and 15 905-backward primers, located at the inner edges of what is called the mtDNA ‘major arc’, where most of the age-associated and disease-associated deletions have been observed (2,3,17). Under the PCR conditions used, this primer set was unlikely to amplify wild-type mtDNA because of the distance between them (~10.5 kb). As expected, this PCR set did not yield any DNA bands in control cortical DNA (Fig. 3E), but amplified DNA products comprising multiple bands in induced cortical samples (Fig. 3E), suggesting that the introduction of DSB at the PstI sites generated a wide variety of ΔmtDNA molecules different in size. Using the same primer set, we observed identical results in hippocampal DNA from 14-week-old control and induced mice (data not shown). Similarly, using different primer pairs (5399-forward and 15 905-backward and 8300-forward and 15 905-backward), we were able to amplify multiple PCR products in the induced cortical DNA but not in control DNA (Fig. 3F). Gel purification and sequencing analysis of those bands identified a series of deletion breakpoints (Fig. 4). Of the total 22 deletions identified in this study, 13 deletions contained no direct repeats at the breakpoints, seven deletions contained small/imperfect homologies and two deletions contained direct repeats of 7–8 bp (Fig. 4). The breakpoints involved nt 5445–8425 at one end and nt 12 243–15 453 at the other end. Some deletions shared the same breakpoints at one end but differed at the other end. Considering that our detection and characterization of ΔmtDNA are biased by the fact that PCR tends to preferentially amplify smaller products, it is likely that we characterized only a partial portion of ΔmtDNA molecules generated by mitoPstI. Nevertheless, this series of experiments demonstrated that DSB at fixed positions of mtDNA lead to the generation of a wide variety of ΔmtDNA in neurons, differing in sizes and mechanisms of recombination.
To test the relative rate of accumulation of different ΔmtDNAs, we performed semi-quantitative PCR for cortical DNA from the above 5-week-old and 14-week-old mice. This analysis revealed the presence of deletions even in 5-week-old tissues (Fig. 5A). However, the amplification of deletions in 14-week-old tissues was much higher, demonstrating that once generated, ΔmtDNA accumulates during aging (Fig. 5A).
If the smaller sizes of ΔmtDNA confer a replicative advantage to those mutant species over wild-type mtDNA, it would be also the case that smaller mtDNA with larger deletions replicate faster than larger mtDNA with smaller deletions. To test this possibility, we performed quantitative real-time PCR for ΔmtDNA with ~3.8 kb deletions (small deletions) and those with ~10 kb deletions (large deletions). By optimizing PCR conditions and primer pairs, we were able to amplify a cluster of ΔmtDNA with very similar sizes only in induced samples. With 8378-forward and 12 281-backward primers, we amplified a cluster of PCR products of ~80 bp originating from smaller deletions of ~3.8 kb (Supplementary Material, Fig. S4). With 5549-forward and 15 501-backward primers, on the other hand, we amplified a group of bands of ~100 bp from larger deletions of ~10 kb (Supplementary Material, Fig. S4). No amplification was observed in control tissues using these pairs of primers. Figure 5B shows the results of the qPCR for the small deletions in two independent 5-week-old mice (Animals 1 and 2) and in two independent 14-week-old mice (Animals 3 and 4). The relative amounts of deletions normalized by the amount of total mtDNA (ND1 region located at the minor arc of mtDNA) were similar in Animals 1 and 2. Compared with the average of control animals 1 and 2, Animal 3 showed a 10.7-fold higher relative levels of small deletions and Animal 4 had a 5.1-fold higher relative levels of deletions. Compared with the average levels in Animals 1 and 2, the relative levels of the larger deletions were 36-fold higher in Animal 3 and 12-fold higher in Animal 4 (Fig. 5C). In both Animals 3 and 4, the age-dependent increase in the levels of larger deletions was higher than that of smaller deletions, supporting the hypothesis that larger deletions have a replicative advantage over smaller deletions (and probably over the full length) during aging.
Accumulation of ΔmtDNA has been observed in (i) mitochondrial diseases, which are caused by non-Mendelian inheritance of maternal mutated mtDNA or inheritance of mutant nuclear genes associated with the maintenance of mtDNA and (ii) mammalian aging. In both scenarios, the characteristics of ΔmtDNAs are similar (2,3,18). Deletions often involve the major arc of mtDNA, between the major origins of replication. They are either flanked by direct repeats (one of which is removed by the deletion process) or no direct repeats. Therefore, similar molecular mechanisms may underlie the generation of deletions in these scenarios.
As discussed in a recent review, replication errors and repair of damaged mtDNA are the two potential mechanisms that may involve in the formation of ΔmtDNA (10). While recombination of mtDNA during replication still stands as a valid model, this model could only account for generation of deletions with direct repeats or homology-dependent recombination.
Krishnan et al. (10) proposed that 3′–5′ exonuclease activity following the introduction of DSB within the major arc exposes single-stranded regions of mtDNA, which subsequently undergoes homologous annealing at micro-domains containing direct repeats. Although this model remains attractive, it only accounts for homology-dependent recombination. In addition, previous work from our laboratory on muscle-specific constitutive mitoPstI transgenic mice showed that DSB-induced deletions contained short or no direct repeats (12). This work proposed that single-stranded 3′ free ends generated by DSB could anneal with the end of the unpaired D-loop strand and initiate recombination, through short homologous sequences.
One shortcoming of the previous model, however, was the constitutive activity of mitoPstI throughout the life of mice and the resulting severe depletion of mtDNA. This scenario compromised cell viability and was unlikely to be suitable for the accumulation of ΔmtDNA. In the present study, we generated an improved mouse model, in which the expression of mitoPstI and therefore the induction of DSB can be spatially and temporarily regulated. Three-day transient removal of Dox from young adult mitoPstI/tTA double transgenic mice successfully induced the formation of ΔmtDNA without a noticeable depletion of mtDNA. Because our identification of ΔmtDNA relied on the cloning of PCR products, it is likely that our technique identified only a subset of total ΔmtDNA species. Nevertheless, our detection and characterization of ΔmtDNA in neurons revealed the formation of a variety of deletions with various sizes (6.3–12.5 kb) with and without direct repeats, a pattern also found in naturally occurring age-related ΔmtDNA. These results suggest that DSBs in neuronal mtDNA lead to both homology-dependent and independent recombination. DSB in mtDNA can occur by several processes, including replication stalling (19,20). The 3′ and 5′ends of breakpoints were proximal to the PstI-recognition sites in some deletions, but were quite distal in others. The formation of deletions such as Deletions 20–22, which involved recombination of 3′ end proximal to the PstI site and the D-loop region, could be explained by a preferential recombination of free ends at the D-loop (12,21,22). However, the formation of other deletions is difficult to explain without the participation of exonuclease activity (10). We propose that NHEJ also plays a role in homology-independent recombination. Although NHEJ machineries have not been identified in neuronal mitochondria (23), Deletion 3 formed by blunted ends of two PstI-recognition sites is likely to be mediated by this mechanism. Exonucleases such as POLG (10) may degrade single-strand ends generated by DSB, and intermediates may undergo NHEJ in a stochastic manner, forming ΔmtDNA without the involvement of direct repeats. POLG is an mtDNA polymerase with proofreading 3′–5′ exonuclease activity and may be involved in this process. However, knockin mice carrying homozygous mutant POLG with proofreading defects also accumulate large-scale deletions with no direct repeats (24). Furthermore, it has not been confirmed that POLG has a capacity to degrade a large portion of mtDNA. Therefore, it is possible that other uncharacterized exonuleases participate in the formation of deletions. Identification and characterization of those exonucleases may provide an insight in preventing the formation and accumulation of ΔmtDNA molecules during aging.
In aged muscle, it has been shown that accumulation of ΔmtDNA occurs through clonal expansion of limited numbers of ΔmtDNA formed stochastically along the course of aging (10,21). However, the molecular mechanisms that account for this clonal expansion has been poorly understood. The proposed mechanisms include (i) random genetic drift (25) and (ii) replicative advantage of ΔmtDNA (26). Limitations of the former mechanism were recently discussed by Krishnan et al. (10). Although we have described ex vivo data suggesting a replicative advantage of ΔmtDNA (26), there has been no experimental in vivo evidence supporting the replicative advantage of ΔmtDNA molecules in post-mitotic cells. Our data, albeit limited to a few mice, suggest that after being formed by DSB, mtDNAs with large deletions accumulate preferentially when compared with mtDNAs with smaller deletions. This observation demonstrates that the accumulation of ΔmtDNA during neuronal aging is not merely due to mtDNAs defective properties, implying a replicative advantage of smaller mtDNAs.
A mammalian version of the bacterial PstI gene with a 5′ mitochondrial targeting sequence from human COX VIII (cytochrome c oxidase subunit VIII) (12) was cloned in pTRE2hyg vector (Clontech, Palo Alto, CA, USA). Subsequently, intervening sequence 8 (IVS8) was introduced between the TRE promoter sequence and the mitoPstI coding sequence. Microinjection of the DNA fragment into B6/SJLF1 fertilized oocytes was performed by the transgenic core facility of the University of Miami. Mice were subsequently crossed with C57BL/6J mice for colony maintenance. The nuclear background was mostly C57BL/6J, but contained genetic contributions from SJLF1 and CBA.
All mice procedures were performed according to a protocol approved by the University of Miami Institutional Animal Care and Use Committee. Mice were housed in a virus-antigen-free facility of the University of Miami Division of Veterinary Resources in a 12 h light/dark cycle at room temperature and fed ad libitum with a standard diet or doxycycline-containing food (200 mg/kg) (Bio-Serv, Frenchtown, NJ, USA). The CamKIIα-tTA mouse was purchased from Jackson Laboratories.
The tet-off HeLa cell line that constitutively expresses tTA was obtained from Clontech. Tet-off HeLa cells cultured in either tetracycline-free medium or doxycycline-containing (100 ng/ml) medium were transfected with 1 µg of the mitoPstI construct using FuGENE 6 Reagent (Roche Applied Science, Indianapolis, IN, USA) in a six-well plate. Transfected cells were cultured for 2 days and were pre-loaded with 200 nm Mitotracker Red (Molecular Probes) for 30 min. They were analyzed by immunocytochemistry as described (12).
Brain tissues were homogenized in PBS containing 1× protease inhibitor cocktail (Roche). Western blots were performed as described (27). Immunostaining of mitoPstI and in situ COX activity staining was performed on paraffin-embedded sections (10 µm) and frozen sections (20 µm) as described (28).
Total DNA was isolated from the cortex, hippocampus or cerebellum using phenol–chloroform extraction. PCR to detect deletions was performed using LA Taq polymerase (TAKARA, Japan). The following primer sets were used: 8300-forward–12 812-backward, 5399-forward–15 965-backward, 5399-forward–15 905-backward and 8300-forward–15 905-backward. Amplified PCR products were separated and visualized on agarose gels containing ethidium bromide. DNA markers were from Fermentas, Hanover, MD, USA (O’Gene 1 kb ruler).
For the characterization of deletions, PCR products unpurified or gel-purified were directly cloned into pJET2.1/Blunt vector using CloneJET PCR cloning kit (Fermentas) and sequenced. Southern blotting was performed as described (12). Unless otherwise indicated, 5 µg of total DNA was digested with MluI, separated on 0.7% agarose gel and transferred to a Zeta-probe GT membrane (Biorad). Probes to detect mtDNA were amplified from tail DNA from a C57BL/6J mouse with Mango Taq DNA polymerase (Bioline, Taunton, MA, USA) using the 3500-forward and 4222-backward primer set. A probe against the gene encoding 18S ribosomal RNA (18S rDNA) was amplified from the same tail DNA using m-18SrDNA-F and m-18SrDNA-B primers. Amplified DNA was purified, labeled with [α-32P]dCTP using Random primed DNA labeling kit (Roche Applied Science) and re-purified with G-50 Sephadex quick spin columns (Roche Applied Science). Detection and densitometric quantification of mtDNA and 18S rDNA signals was performed with the Cyclone Plus Phosphor Imager equipped with the Optiquat software (PerkinElmer Life and Analytical Sciences, Boston, MA, USA). Sequences of primers used in this study were summarized in Supplementary Material, Table S1.
We used 8378-forward and 12 281-backward primers to specifically amplify mtDNA with ~3.8 kb deletions (small deletions), 5549-forward and 15 501-backward primers to amplify mtDNA with 10 kb deletions (large deletions) and 2785-forward and 2869-backward primers to amplify a portion of minor arc or undeleted region of mtDNA (ND1). The amplicon size, in all cases was ~100 bp. All reactions were carried out in triplicate in 25 µl volume containing template DNA (10, 100 and 0.01 ng for small deletions, large deletions and ND1, respectively) and 1× QuantiTect SYBR Green PCR Master Mix (Qiagen, Valencia, CA, USA) on a 7300 PCR Sequence Detection System (Applied Biosystems, Foster City, CA, USA). Specificity of each reaction was evaluated by melting curve analysis as well as agarose-gel electrophoresis. Relative amounts of small deletions, large deletions and ND1 control region at different ages were estimated from the corresponding standard curves generated from serial dilutions of template DNA that contain mitoPstI-induced deletions. DNA from tTA single transgenic mice was used as negative controls. No amplification of deleted mtDNA was observed in any negative controls.
This work was supported by US Public Health Service Grant R01-NS41777, R01-EY10804, R01-CA85700 and by the University of Miami Neuroscience Center (to C.T.M.). H.F. was supported by the Lois Pope LIFE Fellowship.
We are grateful to Dr Sarika Srivastava for reagents and suggestions and to Dr Francisca Diaz for critically reading the manuscript. We are also in debt with New England Biolabs for reagents.
Conflict of Interest statement. Authors declare no conflict of interest.