Abstract
Age-related cognitive decline and neurodegenerative diseases are a growing challenge for our societies with their aging populations. Accumulation of DNA damage has been proposed to contribute to these impairments, but direct proof that DNA damage results in impaired neuronal plasticity and memory is lacking. Here we take advantage of Ercc1Δ/− mutant mice, which are impaired in DNA nucleotide excision repair, interstrand crosslink repair, and double-strand break repair. We show that these mice exhibit an age-dependent decrease in neuronal plasticity and progressive neuronal pathology, suggestive of neurodegenerative processes. A similar phenotype is observed in mice where the mutation is restricted to excitatory forebrain neurons. Moreover, these neuron-specific mutants develop a learning impairment. Together, these results suggest a causal relationship between unrepaired, accumulating DNA damage, and age-dependent cognitive decline and neurodegeneration. Hence, accumulated DNA damage could therefore be an important factor in the onset and progression of age-related cognitive decline and neurodegenerative diseases.
Introduction
Accumulated DNA damage is thought to be an important factor underlying aging (Hoeijmakers, 2009). Several studies show that aging is accompanied by accumulation of neuronal DNA damage in rodents and humans (Sohal et al., 1994; Hamilton et al., 2001; Dorszewska and Adamczewska-Goncerzewicz, 2004; Gedik et al., 2005). Furthermore, the brain is particularly vulnerable to oxidative stress since it exhibits very high oxygen metabolism, has abundant lipid content and relatively low levels of antioxidants compared to other organs (Cai et al., 1996; Leutner et al., 2001; Serrano and Klann, 2004; Møller et al., 2010). Therefore, neurons may be especially prone to DNA lesions resulting from oxidative stress.
Evidence linking DNA damage to cognitive impairment follows from animals and patients receiving genotoxic chemotherapeutic drugs (Ahles and Saykin, 2007; Konat et al., 2008). Additionally, mutations in DNA repair genes may cause neurological impairments, progressive neurodegeneration, and segmental accelerated aging as in a variety of progeroid conditions like xeroderma pigmentosum, Cockayne syndrome, and trichothiodystrophy, which are caused by defective nucleotide excision repair (NER) (Nance and Berry, 1992; Lehmann, 2003; Kraemer et al., 2007). However, in above examples, it cannot be ruled out that the cognitive dysfunction is secondary to the large impact of chemotherapy or progeroid syndrome on overall health.
Interestingly, increased oxidative DNA damage has been observed in subjects with mild cognitive impairments as well as late-Alzheimer's Disease (Keller et al., 2005; Wang et al., 2005; Lovell and Markesbery, 2007), suggesting a correlation between age-related accumulation of DNA damage and cognitive decline. However, a direct method to reliably quantify physiological levels of heterogeneous types of DNA damage in mammalian organs and tissues is extremely difficult (and controversial) and only possible for a very limited subset of lesions (Dizdaroglu et al., 2002). Moreover, there is a chicken and egg problem in the interpretation of studies trying to correlate neuronal degeneration with signs of DNA damage.
To circumvent these problems, we took advantage of a well established mouse mutant with a mutation in the excision repair cross-complementing group 1 (Ercc1) gene. ERCC1 is involved in multiple DNA repair pathways: nucleotide excision repair (Houtsmuller et al., 1999), interstrand crosslink repair (Bergstralh and Sekelsky, 2008; Bhagwat et al., 2009), and double-strand break repair (Zhu et al., 2003; Ahmad et al., 2008). Consistently, ERCC1-deficient cells show increased sensitivity to agents or treatments that damage DNA, and hence it is expected that these animals accumulate DNA damage at a higher rate than control animals (Sijbers et al., 1996; Muñoz et al., 2005; Niedernhofer et al., 2006; Ahmad et al., 2008; Hoeijmakers, 2009). Recently it was shown that these animals have age-related neuronal changes in the spinal cord as well as in neuromuscular junctions of the skeletal muscle (de Waard et al., 2010). Here we show that global Ercc1 mutants as well as neuron-specific Ercc1 mutants exhibit an age-dependent decrease in neuronal plasticity, and progressive neuronal pathology, suggestive of neurodegenerative processes. These results suggest a causal relationship between unrepaired, accumulating DNA damage and age-dependent cognitive decline and neurodegeneration.
Materials and Methods
Generation and breeding of mutant mice.
The generation and characterization of nucleotide excision repair-deficient Ercc1Δ/+ and Ercc1−/+ mice has been previously described (Weeda et al., 1997). Ercc1Δ/− mice were obtained by crossing Ercc1+/− (in the C57BL6J background) with Ercc1Δ/+ mice (in the FVB background) to yield Ercc1Δ/− with an F1 C57BL6J/FVB hybrid background. Wild-type littermates were used as controls. All animals used in the studies described in this paper were of the same F1 C57BL6J/FVB hybrid background. Typical unfavorable characteristics, like blindness in an FVB background or deafness in a C57BL6J background, do not occur in this hybrid background.
For the conditional mutant studies, we made use of a mouse line that has loxP sites inserted into its Ercc1 gene (floxed Ercc1, Ercc1f/f) (Doig et al., 2006). To achieve Ercc1 gene inactivation, we used a transgenic line with Cre recombinase under the control of the αCaMKII promoter (Tsien et al., 1996a,b). Expression from this promoter is specific for postmitotic excitatory neurons (Madisen et al., 2010). Ercc1f/−CaMKII-Cre+ mice were obtained by crossing Ercc1f/f (in the FVB background) with Ercc1+/−CaMKII-Cre+ mice (in the C57BL6J background), to yield hybrid Ercc1f/−CaMKII-Cre+ mice. Ercc1f/−CaMKII-Cre+ mice are heterozygous for Ercc1 in their entire body, except for the excitatory postmitotic neurons in the forebrain, which are homozygous for Ercc1 after Cre excision of the floxed allele. These mice will be referred to as Ercc1f/− in the remainder of the study. As controls, we used Ercc1f/+CaMKII-Cre+ littermates (referred to as Ercc1f/+), which are wild type in their entire body, except for the excitatory postmitotic neurons in the forebrain, which are heterozygous. For water maze learning and fear conditioning, we also included Ercc1f/−CaMKII-Cre− control mice, which are heterozygous for Ercc1 and not different from Ercc1f/+CaMKII-Cre+ mice with respect to water maze learning and fear conditioning (data not shown).
Animals were screened for discomfort and weighed once a week. Animals were maintained in a controlled environment (19–24°C, 12 h light:12 h dark cycle), received standard rodent maintenance chow [CRM(P), Special Diets Services] and water ad libitum, and were housed in individual ventilated cages under SPF conditions. Ercc1Δ/− mice received liquefied food when they were not able to reach the food due to movement disabilities. For all experiments, we used mice of both sexes. All animal experiments were approved by the Dutch Ethical Committee, and in accordance with Dutch animal care and use laws.
Antibodies.
Primary antibodies [supplier; applications: immunohistochemistry (IHC); immunofluorescence (IF); and dilutions] reported in this study are as follows: rabbit anti-ATF3 (Santa Cruz Biotechnology; IHC, 1:1000); rabbit anti-GFAP (DAKO; IHC; 1:10,000; IF, 1:5000); rabbit anti-cleaved caspase 3 (Asp175, Cell Signaling Technology; IHC, 1:1000); mouse anti-MAP2 (clone AP20, Millipore; Sigma, IF, 1:200), rabbit anti-p53 (Leica; IHC, 1:1000), and mouse anti-NeuN (clone A60, Millipore; IF, 1:1000).
For avidin–biotin–peroxidase immunocytochemistry biotinylated secondary antibodies from Vector Laboratories, diluted 1:200 were used. FITC-, Cy3-, and Cy5-conjugated secondary antibodies raised in donkey (Jackson ImmunoResearch) diluted at 1:200 were used for confocal immunofluorescence.
Immunohistochemical and histopathological procedures.
For immunocytochemistry and immunofluorescence, mice were anesthetized with pentobarbital and perfused transcardially with 4% paraformaldehyde. The brain was carefully dissected out and postfixed overnight in 4% paraformaldehyde. Routinely, brain tissue was embedded in gelatin blocks (Jaarsma et al., 2000) and sectioned at 40 μm with a freezing microtome, and sections were processed, free-floating, using a standard avidin–biotin–immunoperoxidase complex method (ABC, Vector Laboratories) with diaminobenzidine (0.05%) as the chromogen, or single-, double-, and triple-labeling immunofluorescence (Jaarsma et al., 2000; Vlug et al., 2005). In addition, a selected number of frozen sections were processed for a silver or thionine staining procedure that selectively labels dying neurons and their processes (Jaarsma et al., 2000).
Immunoperoxidase-stained sections were analyzed and photographed using a Leica DM-RB microscope and a Leica DC300 digital camera. Sections stained for immunofluorescence were analyzed with a Zeiss LSM 510 confocal laser scanning microscope. ATF3-, p53-, and cleaved caspase 3-positive cells were quantified in the neocortex of serial parasagittal sections [1–1.5 mm lateral from the midline; corresponding to plates 109–116 in the stereotaxic atlas of Paxinos and Franklin (2001)] using an Olympus microscope fitted with a Lucivid miniature monitor and Neurolucida software (MicroBrightField). Cells were plotted in an area of 3–6 mm2 (depending on the age and genotype of the mice) extending from the visual cortex to the frontal cortex and including all cortical laminae, and the number of labeled cells per square millimeter cortical area was determined.
Water maze.
To test spatial memory, we used the water maze. Before the test, the mice were handled extensively (2 min/d; 5 d). Our pool is 1.2 m in diameter and has an 11 cm diameter platform submerged 1 cm below the surface. The water is painted milk-white with nontoxic paint and water temperature is kept constant at 26°C. We used dimmed lighting, and mouse tracking is performed using SMART version 2.0 (Panlab). Mice were given 2 trials/d, with 30 s intertrial interval for 5 consecutive days. At a training session, the mouse was placed on the platform for 30 s. Then it was placed in the water at a pseudorandom start position, and it was given a maximum of 60 s to find the platform. If the mouse did not find the platform in 60 s, it was placed back on the platform. After 30 s on the platform, this training procedure was performed once more. The platform remained at the same position during all trials.
One hour after the training on day 5, a probe trial was given to test spatial learning. Mice were placed on the platform for 30 s, after which the platform was removed and the mice were placed in the pool at the opposite side of the former platform location. The mice were then allowed to search for the platform for 60 s.
Fear conditioning.
Fear conditioning was performed in a conditioning chamber (Med Associates) equipped with a grid floor via which the foot shock could be administered. For context conditioning, each mouse was placed inside the conditioning chamber for 180 s. A foot shock (2 s, 0.4 mA) was delivered 148 s after placement in the chamber. Twenty-four hours later, context-dependent freezing was measured during 3 min.
Electrophysiology.
After the animals had been killed, sagittal slices (400 μm) were made and submerged in ice-cold artificial CSF (ACSF) using a vibratome, and hippocampi were dissected out. These sagittal hippocampal slices were maintained at room temperature for at least 1.5 h to recover before experiments were initiated. Then they were placed in a submerged recording chamber and perfused continuously at a rate of 2 ml/min with ACSF equilibrated with 95% O2, 5% CO2 at 31°C. ACSF contained the following (in mm): 120 NaCl, 3.5 KCl, 2.5 CaCl2, 1.3 MgSO4, 1.25 NaH2PO4, 26 NaHCO3, and 10 d-glucose. Extracellular recording of field EPSP (fEPSPs) were made in CA1 stratum radiatum with platinum/iridium electrodes (Frederick Haer). A bipolar Pt/Ir was used to stimulate Schaffer-collateral/commissural afferents with a stimulus duration of 100 μs. Stimulus response curves were obtained at the beginning of each experiment, 20 min after placing the electrodes, with 10, 20 40, 60, 80, and 100 μA stimuli. For analyses, the data from the strongest stimulation were used. LTP was evoked using two different tetani: (1) 10 Hz (1 train of 10 s at 10 Hz) and (2) 100 Hz (1 train of 1 s 100 Hz). The 10 Hz protocol was performed at two-thirds and the 100 Hz protocol at one-third of the maximum fEPSP. fEPSP measurements were done once per minute. Potentiation was measured as the normalized increase of the mean fEPSP slope for the duration of the baseline. Only stable recordings were included, and this judgment was made blind to genotype. Average LTP was defined as the mean last 10 min of each protocol. We used maximally two slices per mouse per protocol.
Results
Ercc1Δ/− mice develop mild age-related neurodegenerative changes
To examine whether a DNA repair defect is indeed sufficient to induce age-related neurodegeneration and cognitive decline, we made use of the Ercc1Δ/− mouse mutant. This mutant carries a null mutation in one allele, whereas the protein derived from the second allele shows reduced activity, due to a 7 aa C-terminal truncation (Weeda et al., 1997). This hypomorphic mutation results in increased sensitivity to DNA damaging treatments such as gamma and UV radiation and Mitomycin C treatment, due to severely impaired nucleotide excision repair, interstrand crosslink repair, and double-strand break repair (Weeda et al., 1997; Niedernhofer et al., 2006; Ahmad et al., 2008). Ercc1Δ/− mutants and their wild-type littermates were analyzed at 1 and 4 months of age. The age of 4 months was chosen since at this age animals display clear signs of premature aging, without significant mortality. The maximum lifespan of these mice in the genetically homogeneous F1 C57BL/6J-FVB/N hybrid genetic background used here is ∼6 months (de Waard et al., 2010).
Macroscopically, the brains of Ercc1Δ/− mice are smaller, but proportional to their reduced body size (de Waard et al., 2010). The gross histological organization analyzed in thionine stained sections appeared normal in all brain areas, including hippocampus (Fig. 1A), indicating that there is no evident developmental perturbation or massive neuronal degeneration. Immunostaining for glial fibrillary acidic protein (GFAP), which is known to be upregulated in response to neuronal injury and degeneration, showed increased GFAP staining, indicative of reactive astrocytosis, throughout the brain in 4-month-old Ercc1Δ/− mice, while at 1 month GFAP staining was the same as in Ercc1+/+ mice (Fig. 1A). These data indicate the occurrence of neurodegenerative changes in the Ercc1Δ/− brain at 4 months of age but not at 1 month. Further analysis of thionine-stained sections revealed sporadic cells with pyknotic nuclei, indicative of cells undergoing apoptotic cell death, in hippocampus (Fig. 1B) as well as other brain areas of 4-month-old Ercc1Δ/− mice. To further examine cell death, we immunostained for activated caspase 3, a final executioner caspase in multiple cell death pathways (Logue and Martin, 2008), and we used a silver staining method that selectively labels degenerating neurons and their axons. Caspase 3 staining revealed a low but distinctive number of labeled cells throughout the brain of 4-month-old Ercc1Δ/− mice. In many occasions, labeled cells showed a clear neuronal morphology, while a subset of labeled cells showed a glial morphology (Fig. 1B). Occasionally, caspase 3-positive cells also occurred in the brain of 1-month-old Ercc1Δ/− mice, albeit at a considerable lower frequency (Fig. 1C). In contrast, Ercc1+/+ mice did not show any caspase 3-positive cells (Fig. 1C). Silver staining, similar to active caspase 3 immunostaining revealed a low level of neurons in the process of dying in multiple brain areas of 4-month-old Ercc1Δ/− mice, including the hippocampal CA1 region and the dentate gyrus (Fig. 1D). The silver staining also produced argyrophilic staining of degenerating axons in white matter areas such an the corpus callosum, the fimbria-fornix, and the capsula interna bundles coursing through the striatum (Fig. 1D). The relatively more prominent argyrophilic staining of axons than of neuronal cell bodies may be explained by slower removal of degenerated axons resulting in cumulative argyrophilic staining in axonal trajectories (Jaarsma et al., 1992) even in conditions of a low level of ongoing neuronal death. Importantly, no, or very limited, axonal degeneration occurred in the white matter of 1-month-old Ercc1Δ/− mice (Fig. 1D), consistent with the much lower levels of neuronal degeneration at this age as indicated by thionine-, active caspase 3-, and GFAP-stained sections. Together, our analyses of cell death and neuronal degeneration indicate that Ercc1Δ/− mice show very sporadic cell death throughout the brain at 1 month, and considerably higher (albeit still modest) levels of ongoing cell death at 4 months of age. Finally, to further validate that the level of neuronal degeneration is limited and that a large proportion of neurons are preserved in hippocampus of 4-month-old Ercc1Δ/− mice, we stained for the somatodendritic neuronal marker MAP2. Indeed, as shown in Figure 1E, MAP2 staining in 4-month-old Ercc1Δ/− hippocampus appeared normal.
Young adult Ercc1Δ/− mice display reactive astrocytosis, mild neuronal degeneration, and signs of genotoxic stress. A, Coronal brain slices stained with thionine (upper panels) or processed for GFAP immunoperoxidase histochemistry (lower panels) illustrating normal gross histoarchitecture of the dorsal hippocampus and surrounding brain structures in Ercc1Δ/− mice, and increased GFAP staining throughout the brain of 4-month-old Ercc1Δ/− mice (scale bar, 500 μm). B, Light photomicrographs illustrating dying cells in hippocampus (CA1), cortex (NCx), and corpus callosum of 4-month-old Ercc1Δ/− mice. Dying cells are identified by their pyknotic nucleus in thionine-stained sections (white arrow and inset in second image), or by active caspase 3 staining. In many occasions, caspase 3-positive cells can be easily differentiated in neurons (arrows) or glial cells (arrowheads; scale bar, 25 μm). C, Quantification of cortical cell density positive for ATF3, p53 or caspase 3 (y-axis on left indicates values for ATF3 and p53; y-axis on the right indicates values for caspase 3). All data are reported as mean ± SEM. Two-way ANOVA revealed a significant effect for genotype, age and their interaction for ATF3 (all p < 0.0001), p53 (all p < 0.0001), and caspase 3 (all p < 0.01). D, Silver staining showing argyrophilic somatodendritic neuronal profiles indicative of dying neurons in 4-month-old Ercc1Δ/− hippocampus (arrows in upper row; scale bar, 50 μm), and argyrophilic degenerating axons in fimbria-fornix (ff) (arrow in middle row; scale bar, 250 μm) and striatal capsula interna (ci) bundles (arrow in lower row; scale bar, 50 μm) of 4-month-old Ercc1Δ/− brain. Note the absent and highly sporadic argyrophilic fiber degeneration in Ercc1+/+ mice and 1-month-old Ercc1Δ/− brain, respectively. E, Confocal immunofluorescent images of CA1 area in the hippocampus showing unaltered level of MAP2 and increased GFAP immunoreactivity in 4-month-old Ercc1Δ/− mice (scale bar, 100 μm). F, Confocal image showing p53-NeuN double-labeled cells in 4-month-old Ercc1Δ/− cortex and hippocampus (scale bar, 50 μm). G, ATF3-immunoperoxidase histochemistry illustrating multiple ATF3-positive cells in 4-month-old Ercc1Δ/− cortex. Inset shows enlargement of neuron with flattened eccentric nucleus (scale bar, 100 μm). Th, Thalamus; Str, striatum; Am, amygdala; cc, corpus callosum; DG, dentate gyrus; py, pyramidal layer; rad, stratum radiatum; ml, molecular layer; gr, granule layer.
The asynchronous sporadically distributed neuronal degeneration in the Ercc1Δ/− brain as outlined above is compatible with a model where neurons accumulate stochastic DNA lesions that eventually disrupt cell function. To obtain evidence that Ercc1Δ/− neurons experience genotoxic stress, we have studied the expression of the transcription factor p53, which is known to be activated by multiple types of DNA damage and to mediate neuronal degeneration (Levine et al., 2006), and ATF3, which also is induced following genotoxic stress via p53-dependent and -independent pathways (Hai et al., 1999; Fan et al., 2002; Turchi et al., 2009). Already at 1 month of age, Ercc1Δ/− mice showed cells with intensely p53-labeled nuclei throughout the brain, and this number was significantly increased at 4 months of age (Fig. 1C,F). Instead, no p53-positive cells were observed in the nervous systems of Ercc1+/+ mice (Fig. 1C). Double labeling with the neuronal marker NeuN showed that in cortex and hippocampus >95% of the p53-positive cells were neurons (Fig. 1F). ATF3 labeling, like p53, was absent in most brain areas of Ercc1+/+ mice except for weakly labeled neuronal nuclei in the dentate gyrus, olfactory bulb, and pyriform cortex. However, significant amounts of cells with intensely labeled ATF3 nuclei occurred in the brain of Ercc1Δ/− mice (Fig. 1C,G). Similar to p53, double labeling with NeuN indicated that the vast majority of ATF3-positive cells were neurons (data not shown). The frequencies of ATF3-positive and p53-positive cells were ∼10-fold the frequency of caspase 3-positive cells (Fig. 1C). Some p53- and ATF3-positive neurons displayed morphological abnormalities such as eccentric flattened nuclei, suggestive of compromised health (Fig. 1G). Together, these neuropathological data indicate that 4-month-old Ercc1Δ/− mice show a low number of neurons in the process of dying while a greater number of neurons show signs of genotoxic stress and poor condition.
Ercc1Δ/− mice show reduced hippocampal synaptic plasticity
We investigated basal synaptic transmission properties of Ercc1Δ/− mice, using fEPSP measurements of the Schaffer-collateral pathway. The ratio between the fEPSP slope and the fiber volley, which is a measure of the efficacy of the synapses, did not differ significantly between genotypes at either 1 month or at 4 months of age, suggesting no changes in basal synaptic transmission [ratio 2.9 ± 0.10 (n = 41 slices from 9 animals) and 2.7 ± 0.11 (n = 35 slices from 8 animals) for 1-month-old Ercc1+/+ and Ercc1Δ/−, respectively; ratio 3.5 ± 0.33 (n = 66 slices from 20 animals) and 2.9 ± 0.28 (74 slices from 20 animals) for 4-month-old Ercc1+/+ and Ercc1Δ/−, respectively; F(1,74) = 2.5, p = 0.11; F(1,138) = 2.0, p = 0.16, one-way ANOVA for 1 month and 4 months old, respectively].
Synaptic plasticity can be assessed in vitro by measuring the ability of synaptic connections to become potentiated upon a train of high-frequency stimulation, a process known as long-term potentiation (LTP). LTP was induced at the Schaffer-collateral synapse with a 10 Hz stimulus. Although LTP was observed in all groups, 4-month-old Ercc1Δ/− mice showed significantly less LTP than their controls, whereas no significant difference was observed in 1-month-old mice, indicating normal initial functioning and accelerated, progressive decline of plasticity with age (Fig. 2A,B; F(1,24) = 0.6, p = 0.40; F(1,35) = 5.3, p < 0.05, one-way ANOVA for 1 and 4 months old, respectively). A reduction of LTP was also observed when using a strong (100 Hz) stimulus; 4-month-old Ercc1Δ/− mice, but not 1-month-old Ercc1Δ/− mice, displayed significantly reduced LTP compared to age-matched controls (Fig. 2C,D; F(1,27) = 2.7 p = 0.10; F(1,33) = 13.4, p < 0.001, one-way ANOVA for 1 month and 4 months old, respectively). These results show that Ercc1Δ/− mice develop an age-related impairment of synaptic plasticity.
Ercc1Δ/− mice show reduced synaptic plasticity at 4 months of age. A, LTP (10 Hz) in 1-month-old mice shows no difference between Ercc1Δ/− and Ercc1+/+ mice (n = 14 slices from 9 animals and n = 12 slices from 8 animals for Ercc1+/+ and Ercc1Δ/−, respectively). B, LTP (10 Hz) in 4-month-old mice shows reduced LTP in Ercc1Δ/− mice [n = 19 slices from 10 animals and n = 18 slices from 11 animals (4 males and 6 females) for Ercc1+/+ and Ercc1Δ/−, respectively]. C, LTP (100 Hz) in 1-month-old mice shows no difference between Ercc1Δ/− and Ercc1+/+ mice (n = 17 slices from 9 animals and n = 12 slices from 8 animals for Ercc1+/+ and Ercc1Δ/−, respectively). D, LTP (100 Hz) in 4-month-old mice shows reduced LTP in Ercc1Δ/− mice (n = 18 slices from 10 animals and n = 17 slices from 11 animals and for Ercc1+/+ and Ercc1Δ/−, respectively). All data are reported as mean ± SEM. Filled circles represent Ercc1+/+. Open circles represent Ercc1Δ/−. *Significantly different (p < 0.05) from age-matched Ercc1+/+ mice.
A DNA repair defect exclusively in excitatory postnatal forebrain neurons causes gradual neuronal degeneration and reactive astrocytosis
Although the aforementioned results showed that the global Ercc1Δ/− mutation significantly affects neuronal health and plasticity, it cannot be ruled out that these outcomes are secondary to the reduced fitness of these animals due to liver, kidney, and other pathology (McWhir et al., 1993; Weeda et al., 1997; Selfridge et al., 2001; Lawrence et al., 2008; de Waard et al., 2010). To examine the direct effect of a DNA repair defect in neurons, and to rule out possibly confounding effects of systemic aging pathology, we made use of the Cre-loxP system to generate mutant mice with a neuron-specific ablation of Ercc1.
Ercc1f/f mice containing a floxed Ercc1 gene were crossed with Ercc1+/−CaMKII-Cre mice resulting in the desired Ercc1f/− CaMKII-Cre+ mice (hereafter referred to as Ercc1f/− mice), which are homozygous Ercc1 knock-outs in αCaMKII-expressing cells (mostly excitatory postmitotic neurons of the hippocampus and cortex) and heterozygous for Ercc1 in the remainder of their body. As controls we used Ercc1f/− CaMKII-Cre− mice, which are heterozygous for the Ercc1 gene, and Ercc1f/+ CaMKII-Cre+ littermate controls, which are heterozygous for the Ercc1 gene in the αCaMKII-expressing cells, and wild type in the rest of their body (hereafter referred to as Ercc1f/+; see Materials and Methods). This breeding strategy was chosen instead of the preferred Ercc1+/f with Ercc1+/f CaMKII-Cre breeding, because the latter breeding strategy frequently results in germ-line deletion of the floxed allele. Importantly, mice that are either heterozygous for Ercc1 (Ercc1+/− mice or Ercc1f/−mice without Cre) do not show a phenotype with respect to lifespan, body weight, general appearance, brain pathology, and synaptic plasticity (Weeda et al., 1997 and data not shown).
Like Ercc1Δ/− mice, the obtained Ercc1f/− mice showed an age-related increase in GFAP staining. However, increased GFAP staining was restricted to forebrain areas, in particular hippocampus and cortex (Fig. 3A), consistent with the Cre expression-dependent restricted ablation of Ercc1 in these mice. Changes in GFAP staining in hippocampus and cortex were detectable at the age of 4 and 6 months, but not yet at 2 months (Fig. 3A). Accordingly, immunohistochemistry for active caspase 3, p53, and ATF3 revealed no labeled cells in cortex and hippocampus of Ercc1f/− mice at 2 months of age, while sporadic active caspase 3-positive cells and higher levels of ATF3- and p53-positive cells occurred at 4 and 6 months of age (Fig. 3B,C). Positive cells only occurred in forebrain areas, again consistent with the restricted ablation of Ercc1. Regarding active caspase 3-labeled cells, we only observed cells with neuronal or undefined morphologies (Fig. 3B), but no labeled cells with clear glial morphologies such as observed in the brain of Ercc1Δ/− mice (Fig. 1B). Also ATF3- and p53-positive cells were neurons, as >95% was NeuN positive in double labeling experiments. The occurrence of neuronal degeneration in 4 and 6-month-old Ercc1f/− forebrain was further confirmed by silver degeneration staining producing sporadic argyrophilic neurons in the cortex and hippocampus, as well as argyrophilic fiber staining in the fimbria-fornix, capsula interna, and corpus callosum (data not shown).
Ercc1f/− mice show astrocytosis, genotoxic stress transcription factor expression, and mild age-related neuronal degeneration. A, GFAP-immunoperoxidase staining in sagittal brain section of Ercc1f/− and control (Ercc1f/+) mice. Note the strongly increased GFAP staining in forebrain regions, including neocortex (NCx), striatum (Str), and hippocampus, of 6-month-old Ercc1f/− mice while GFAP staining in other parts of the brain, including thalamus (Th), mesencephalon (Mes), and cerebellum (Cb) is the same as in control. The lower panels show that the increase GFAP staining in Ercc1f/− hippocampus and cortex starts after 2 months of age and is progressive. B, Active caspase 3, p53, and ATF3-immunoperoxidase histochemistry illustrating positive cells in 6-month-old Ercc1f/− hippocampal CA1 area (upper row) and neocortex (lower row, scale bar 100 μm). C, Bar graph illustrating the density of ATF3- and p53- (y-axis on left) and caspase 3- (y-axis on the right) positive cells in neocortex of Ercc1f/− and Ercc1f/+ mice. Data represent means ± SEM. Two-way ANOVA revealed a significant effect for genotype, age, and their interaction for ATF3 (all p < 0.001), p53 (all p < 0.001), and caspase 3 (genotype p = 0.01, age p = 0.01 and genotype × age p = 0.01). Tukey HSD post hoc tests revealed significant effects between all age groups for ATF3 (all p < 0.001) and for p53 (all p < 0.001). For caspase 3, Tukey HSD post hoc test revealed a significant difference between 2 and 6 months and 4 and 6 months (p < 0.001 and p = 0.002, respectively).
In conclusion, these immunohistochemistry data show that the Ercc1f/− mice have pathologies in hippocampus and cortex similar to those of the Ercc1Δ/− mice, albeit that they develop at a somewhat slower time course, establishing a cell intrinsic cause of the neuronal phenotype and demonstrating that a DNA repair defect in forebrain neurons is sufficient to cause increased expression of markers that are indicative for genotoxic stress and neuronal degeneration.
A DNA repair defect in excitatory neurons causes a reduction of LTP
We studied the electrophysiological properties of the Ercc1f/− mouse hippocampus to determine whether neuronal DNA damage is sufficient to cause the aging-like phenotype at the level of synaptic plasticity. Similar to Ercc1Δ/− mice, Ercc1f/− mice show no detectable impairment in synaptic transmission. The ratio between fEPSP slope and fiber volley does not differ at both 3 and 6 months of age [ratio 2.9 ± 0.16 (n = 36 slices from 5 animals) and 2.8 ± 0.16 (n = 55 slices from 12 animals) for 3-month-old Ercc1f/+ and Ercc1f/−, respectively; ratio 2.1 ± 0.13 (n = 34 slices from 5 animals) and 1.9 ± 0.06 (63 slices from 9 animals) for 6-month-old Ercc1f/+ and Ercc1f/−, respectively; F(1,89) = 2.9, p = 0.09; F(1,95) = 2.2, p = 0.14 one-way ANOVA for 3 and 6 months old, respectively]. Next we investigated the ability to induce LTP at the Schaffer-collateral synapse using a 10 Hz stimulus. Although LTP was observed in all four groups, 6-month-old Ercc1f/− mice showed significantly less LTP than their littermate Ercc1f/+ controls, whereas no significant difference was observed for the 3-month-old mice (Fig. 4A,B; F(1,33) < 0.001, p = 1.0; F(1,23) = 5.8, p < 0.05, one-way ANOVA for 3 and 6 months old, respectively). Similar results were obtained when a strong 100 Hz stimulus was applied: 6-month-old Ercc1f/− mice but not 3-month-old mice showed significantly reduced LTP as compared to their age-matched controls (Fig. 4C,D; F(1,29) = 1.6, p = 0.20; F(1,30) = 5.6, p < 0.05, one-way ANOVA for 3 and 6 months old, respectively). Hence, by restricting the defect in DNA repair to the excitatory neurons of the hippocampus, the mice are overall healthy, but hippocampal plasticity is still affected in an age-dependent fashion.
Ercc1f/− mice show reduced synaptic plasticity at 6 months of age. A, LTP (10 Hz) in 3-month-old mice shows no difference between Ercc1f/− and Ercc1f/+ mice [n = 16 slices from 5 animals, n = 19 slices from 8 animals for Ercc1f/+ and Ercc1f/−, respectively]. B, LTP (10 Hz) in 6-month-old mice shows reduced LTP in Ercc1f/− mice (n = 14 slices from 4 animals and n = 11 slices from 5 animals for Ercc1f/+ and Ercc1f/−, respectively). C, LTP (100 Hz) in 3-month-old mice shows no difference between Ercc1f/− and Ercc1f/+ mice (n = 13 slices from 4 animals, n = 18 slices from 10 animals for Ercc1f/+ and Ercc1f/−, respectively). D, LTP (100 Hz) in 6-month-old mice shows reduced LTP in Ercc1f/− mice (n = 15 slices from 5 animals and n = 17 slices from 7 animals for Ercc1f/+ and Ercc1f/−, respectively). All data are reported as mean ± SEM. Filled circles represent Ercc1f/+. Open circles represent Ercc1f/−. *Significantly different (p < 0.05) from age-matched Ercc1f/+ mice.
A DNA repair defect exclusively in neurons causes impaired cognitive function
Having shown that the Ercc1f/− mice are physically in good condition, but with regard to the hippocampus still display an accelerated aging phenotype similar to the Ercc1Δ/− mice, both in cellular pathology and in vitro synaptic plasticity, we tested the ability of the mutants to learn by using the Morris water maze test. In this test, which relies on a functional hippocampus, animals are trained over several days to locate a submerged platform in a circular pool filled with opaque water using distal visual cues. All four groups showed a significant reduction of their latency times to find the platform across training days (Fig. 5A,B; effect of training: F(4,148) = 28.5, p < 0.001; F(4,124) = 21.4, p < 0.001, repeated-measures ANOVA for 3 and 6 months old, respectively). There was no significant effect of genotype in both age groups (Fig. 5A,B; effect of genotype: F(1,37) = 0.2, p = 0.66; F(1,31) = 3.7, p = 0.06, repeated-measures ANOVA for 3 and 6 months old, respectively), and swim speed was not different between genotypes (18.1 ± 1.0 and 16.5 ± 0.5 cm/s for 3-month-old Ercc1f/+ and Ercc1f/−, respectively; 15.9 ± 0.3 and 16.3 ± 0.7 cm/s for 6-month-old Ercc1f/+ and Ercc1f/−, respectively; F(1,37) = 3.3, p = 0.08; F(1,34) = 0.4, p = 0.51 one-way ANOVA for 3 and 6 months old, respectively). These data indicate that mice in all four groups were able to execute this task and motivated to escape. To test whether the mutants had indeed learned the spatial location, a probe trial in which the platform was removed, was given after 5 d of training. This probe trial showed that all four groups of mice spent significantly more time in the target quadrant than in any of the other quadrants, indicating that the mutants had indeed learned the platform location at both 3 and 6 months of age (Fig. 5C,D). However, aged (but not young) mutants spend significant less time in the target quadrant than Ercc1f/+ littermate control mice (F(1,34) = 4.8, p = 0.03, one-way ANOVA) (Fig. 5E,F). This deficit could be overcome by 2 d of additional training (data not shown), indicating that the mutants are able and motivated to learn the task, but it takes more training to reach a similar level of performance.
Ercc1f/− mice show impaired fear conditioning and impaired water maze performance at 6 months of age. A, B, Escape latency of 3-month-old (A) and 6-month-old (B) mice shows no difference between Ercc1f/− and Ercc1f/+ mice (n = 9, n = 15, n = 11, and n = 17 for 3- and 6-month-old Ercc1f/+ and Ercc1f/−, respectively). Filled circles represent Ercc1f/+. Open circles represent Ercc1f/−. C, D, Quantification of quadrant occupancy in 3-month-old (C) and 6-month-old (D) mice, during the probe trial performed at day 5. Although all groups searched significantly more in the target quadrant, at 6 months old, Ercc1f/− mice spent significantly less time in the target quadrant than their Ercc1f/+ littermates. Black bar represents target quadrant. E, F, Visual representation of all search tracks during the probe trial of the 3-month-old (E) and 6-month-old (F) mice. The color of the heat plots indicate the mean time spent at a certain location. The white and black dashed lines indicate quadrants and former platform location, respectively. G, H, Contextual fear conditioning in 3-month-old mice (n = 9 and n = 20 for Ercc1f/+ and Ercc1f/−, respectively) (G) and 6-month-old mice (n = 10 and n = 13 for Ercc1f/+ and Ercc1f/−, respectively) (H). Six-month-old Ercc1f/− mice show normal baseline freezing before the shock but significantly reduced freezing when placed in the context 24 h after the shock. All data are reported as mean ± SEM. *Significantly different (p < 0.05) from age-matched Ercc1f/+ mice.
As an additional paradigm to test cognitive function, which is sensitive to hippocampal dysfunction, we made use of contextual fear conditioning. In this single-trial learning task, mice are conditioned to associate a certain context with a mildly aversive foot shock. This learning paradigm has been shown to require both a functional hippocampus as well as amygdala (Fanselow, 1990). Learning is assessed by measuring freezing behavior, i.e., the cessation of all movement except respiration, which is a natural expression of fear in mice. At 3 months of age, the freezing response of the Ercc1f/− mice was indistinguishable from that of their Ercc1f/+ littermate controls (Fig. 5G; F(1,42) = 0.9, p = 0.36, one-way ANOVA). However, at 6 months the Ercc1f/− mice froze significantly less than their littermate Ercc1f/+ controls (Fig. 5H; F(1,40) = 5.7, p < 0.05, one-way ANOVA). Neither the 3-month-old groups nor the 6-month-old groups showed a difference in baseline freezing behavior (Fig. 5G,H; F(1,242) = 0.5, p = 0.47; F(1,32) = 0.09, p = 0.77, one-way ANOVA for 3 and 6 months old, respectively), suggesting that reduced freezing is a consequence of the failure to form a contextual representation of the fear conditioning box, and not of reduced fear in general. The combination of reduced learning in the water maze, reduced levels of freezing after context conditioning, and the impaired hippocampal LTP strongly indicate impaired hippocampal function. However, we want to emphasize that the cognitive deficit is most likely not restricted to the hippocampus, but probably affects all the αCaMKII-expressing cells of the forebrain, as is also suggested by the pathological findings.
These results show that mice, in which the defect in DNA repair is limited to neurons, develop an age-related impairment of context-dependent fear learning.
Discussion
Although accumulation of DNA damage has been put forward as a potential cause for cognitive decline, there is no direct causal proof that a defect in DNA repair can induce cognitive deficits. This study shows that defective DNA repair, either in the entire body or in neurons alone, causes an accelerated aging-like phenotype of the brain with respect to both cellular pathology and synaptic plasticity deficits. In addition, we show that in otherwise healthy animals, homozygous deletion of Ercc1 restricted to αCaMKII-expressing neurons (mostly excitatory neurons of the forebrain) is sufficient to affect learning.
Our pathological data indicate that both the global and neuron-specific Ercc1 mutant mice, while showing respectively minimal and no degenerative changes at juvenile age, develop signs of genotoxic stress and mild neuronal degeneration as well as astrocytosis during young adult life. The increased expression of GFAP, indicative of reactive astrocytosis, in Ercc1 mutant mice is also seen in the brains normally aging rats, mice, and humans (O'Callaghan and Miller, 1991; Nichols et al., 1993; Takahashi et al., 2006). In addition, the increased p53 expression we found is observed in normal aging rat brain (Chung et al., 2000; Dorszewska and Adamczewska-Goncerzewicz, 2004). Together, these results show that (neuronal) DNA damage results in brain pathology that shares characteristics with normal aging. However, it is important to point out that although the Ercc1 mutants show global functional and structural deterioration; some aspects of neuroaging are not modeled, such as accumulation of lipofuscin and the aggregation of proteins.
The pattern and time course of neuronal degeneration is compatible with a model where neurons are afflicted by stochastic DNA lesions that can either cause neuronal degeneration by blocking or deregulating expression of essential genes, or trigger genotoxic stress response pathways, for instance, by stalling of RNA polymerase II (de Waard et al., 2010; Garinis et al., 2009; Mitchell et al., 2003; Hoeijmakers, 2009; Nouspikel, 2007; Brooks, 2008). Ercc1 mutants show impaired nucleotide excision repair, interstrand crosslink repair, and double-strand break repair, and hence may accumulate different types of DNA lesions (Houtsmuller et al., 1999; Zhu et al., 2003; Ahmad et al., 2008; Bergstralh and Sekelsky, 2008; Bhagwat et al., 2009). Oxidative base damage such 7,8-dihydro-8-oxoguanine (8-oxoG) is repaired by the base excision repair pathway and may not accumulate in Ercc1 mutant mice. However, other oxidative lesions such as malondialdehyde adducts and 8,5′-cyclopurine-2′-deoxynucleotides are both NER substrates and potential threats to transcription (Fishel et al., 2007; Nouspikel, 2007; Brooks, 2008). However, directly determining physiological levels of heterogeneous types of DNA damage in mammalian organs and tissues in a quantitatively reliable manner is extremely challenging and reliable for a very limited subset of lesions (Dizdaroglu et al., 2002; Himmelstein et al., 2009). Hence, the precise lesions that accumulate in Ercc1-deficient neurons remain to be determined.
In addition to the neuropathological changes, we observed an age-dependent impairment in synaptic plasticity in the global Ercc1 mutant and in synaptic plasticity as well as learning in the neuron-specific Ercc1 mutant. The strong resemblance of phenotypes in the global and the neuron-specific Ercc1 mutants indicate that synaptic plasticity and learning defects in the global Ercc1Δ/− mutant result from the Ercc1 deficiency in principal forebrain neurons, rather than from indirect effects in other tissues.
How does the lack of a fully functional DNA repair system cause a reduction of LTP and impaired learning? There are several possibilities. For instance, it could be a direct result of the observed neuronal degeneration. However this seems unlikely, because as judged by histology, the overall loss of excitatory neurons appears to be very limited. This is in agreement with the electrophysiology, which showed no significant change in the presynaptic fiber volley (a measure for the number of stimulated fibers) and no change in the size of the fEPSP as a function of stimulation strength (a measure for both the number of activated neurons and the efficacy of synaptic transmission). In addition, since LTP is a relative measure of change of synaptic strength, neurons will only contribute to this measure as long as they are able to show synaptic transmission. Hence, LTP will not be affected by a decreased number of neurons. It is therefore more likely that unrepaired DNA damage interferes with transcription of genes necessary for normal neuronal functioning, which results in decreased neuronal plasticity. In this respect, it is interesting to note that the induction phase of LTP appears to be normal in the Ercc1 mutants but that the later phase is affected. This impairment coincides with the start of the mRNA and protein synthesis-dependent phase of LTP (called late phase LTP; L-LTP) (Kelleher et al., 2004). Notably, L-LTP has been shown to be affected in aged animals (Bach et al., 1999).
Alternatively, the deficits in plasticity and learning could arise from impaired insulin-like growth factor 1 (IGF1) signaling. Both DNA damage and aging result in a reduction of the somatotroph axis by downregulating growth hormone (GH)/IGF1 signaling (Niedernhofer et al., 2006; van de Ven et al., 2006; van der Pluijm et al., 2007; Garinis et al., 2009), which appears to operate in a cell autonomous way, presumably via DNA damage-induced stalling of RNA polymerase II (Garinis et al., 2009). It has been shown that IGF1 regulates synaptic plasticity in the adult CNS (Torres-Aleman, 1999; Sonntag et al., 2000) and that age-related behavioral impairments can be alleviated by IGF1 (Markowska et al., 1998; Shi et al., 2005), thereby suggesting that IGF1 reduction could be a factor in age-related reduction of synaptic plasticity. In addition, it has recently been shown that the related family member IGF2 facilitates the stability of LTP and is critical for memory consolidation (Chen et al., 2011).
Finally, we cannot exclude that the plasticity deficits are caused by reactive glia, because even though the homozygous Ercc1 gene deletion was restricted to αCaMKII-expressing neurons, we observed a robust age-dependent increase of GFAP staining, indicative of reactive astrocytosis. GFAP is known to be upregulated in response to neuronal injury and degeneration, and indeed, increased GFAP staining was not observed in wild-type and heterozygous Ercc1 mutants. To what extend reactive astrocytes can contribute to LTP and learning impairment remains, however, to be investigated (Wenker, 2010).
The observed plasticity deficits and behavioral phenotype of the Ercc1 mouse, parallel findings in aged animals, as impairments in LTP, contextual conditioning, and spatial learning are commonly observed in aged rodents (Bach et al., 1999; Foster et al., 2001; Liu et al., 2002; Watson et al., 2002; Barnes, 2003; Blalock et al., 2003; Blank et al., 2003; Erickson and Barnes, 2003; Foster et al., 2003; Rosenzweig and Barnes, 2003; Verbitsky et al., 2004; Moyer and Brown, 2006; Kaczorowski and Disterhoft, 2009). In addition, some AD mouse models also show reduced LTP and learning (Lambert et al., 1998; Chapman et al., 1999; Walsh et al., 2002; Wang et al., 2002; Jacobsen et al., 2006; Laurén et al., 2009). Also, humans show an age-related deficit in performance on a virtual Morris water maze (Driscoll et al., 2003).
Together, our results show that unrepaired DNA damage is sufficient to cause progressive neuronal pathology, neuronal plasticity deficits, and cognitive decline. The phenotype of the Ercc1 mutants is reminiscent of the global structural and functional deterioration associated with aging. Therefore, we propose that the mouse model(s) mimic(s) aspects of neuroaging. Further experiments are required to more precisely determine which types of toxic DNA lesions accumulate in the Ercc1 nervous system, to which extent these lesions occur in the aging brain, and by which mechanisms these lesions affect neuronal function. Nevertheless, at this point the mice represent a unique model system to identify conditions that accelerate or prevent the accumulation of toxic DNA lesions. The results of these studies may prove to be important for the development of better therapeutic strategies to battle age-related cognitive decline or to prevent the devastating effects of neurodegenerative diseases.
Footnotes
This work was supported by a Top Institute Pharma grant to G.C.M.Z., J.H.J.H., and Y.E., and by an NWO-ZoNMW grant to Y.E. and G.T.J.v.d.H. Additional support is acknowledged from EU FP 7 (MarkAge 200880; Department of Genetics and DNage) and EU FP6 (Lifespan, EC-LSHG-CT-2007-036894; Department of Genetics) as well as grants from NIH (1PO1 AG17242-02) and NIEHS (1UO1 ES011044). J.H.J.H. and G.T.J.v.d.H. have a financial interest in DNage BV. The research was supported in part by DNage via the Top Institute Pharma, a public–private partnership consortium. G.C.M.Z. and I.v.d.P. are employed by DNage BV.
The authors declare no competing financial interests.
- Correspondence should be addressed to either Ype Elgersma or Dick Jaarsma, Department of Neuroscience, Erasmus MC, P.O. Box 2040, NL-3000 CA Rotterdam, The Netherlands, y.elgersma{at}erasmusmc.nl or d.jaarsma{at}erasmusmc.nl
