DNA repair plays a critical, but imprecisely defined role in neuronal survival during cortical neurogenesis. We examined cortical development in mice deficient for the DNA end-joining protein, Ku70. At gestational day 14.5, corresponding to the peak of neurogenesis, the Ku70−/− embryonic cerebral cortex displayed 25- to 30-fold more cell death than heterozygous littermates, as judged by DNA breaks, pyknosis and active caspase-3. In Ku70−/− embryos only, large clusters of dying neurons were found in the intermediate zone. Cell death declined until P4, when the number of dying cells became comparable to that in heterozygous mice. Two groups of dying cells were evident: a GLAST+ neural progenitor population in the subventricular and ventricular zones, and a doublecortin+ immature neuron population in the intermediate zone, the latter exhibiting strong staining for oxidative DNA damage. Antioxidants and lower oxygen tension reduced the high levels of neuronal death in primary cortical cultures derived from Ku70−/− mice, but not the low levels of cell death in wildtype cortical cultures. Results indicate migrating cortical neurons undergo oxidative DNA damage, which is normally repaired by non-homologous end joining. Failure to repair oxidative damage triggers a form of apoptosis involving caspase-3 activation.
Neurons and glia are overproduced during development of the central nervous system (CNS). Critical to normal development, surplus cells are deleted by a mechanism termed programmed cell death (PCD), which is regulated by survival signals, including limited growth factor availability and synaptic activity (Oppenheim et al., 1982; Cowan et al., 1984; Hamburger, 1992; Haydar et al., 1999; de la Rosa and de Pablo, 2000; Raballo et al., 2000). PCD typically involves chromatin condensation, pyknosis and DNA fragmentation. Therefore, delineation of the role of DNA damage, its causes and its regulation by DNA repair mechanisms will enhance our understanding of normal embryonic brain development and provide insights into pathological forms of neuronal death in adulthood.
Over 125 human DNA repair genes have been identified. Dysfunction of certain DNA repair pathways could be expected to cause embryonic neurological deficits, and mutations in many of these genes have been linked to neurodegenerative diseases (Wood et al., 2001; Brooks, 2002). In particular, cortical and spinal cord neurodegeneration have been found to result from disruption of several of these genes, including polymerase β, XRCC2, XRCC4, DNA ligase IV, DNA-PKcs, Ku70 and Ku80/86 (Gu et al., 1997; Barnes et al., 1998; Frank et al., 1998; Gao et al., 1998a,b; Deans et al., 2000; Gu et al., 2000; Sugo et al., 2000; Vemuri et al., 2001), nearly all of which are involved in non-homologous end joining (NHEJ).
During initiation of NHEJ, Ku70 dimerizes with Ku80/86 and binds to broken DNA strands. Subsequently, the DNA-dependent protein kinase catalytic subunit binds to the Ku dimer and recruits additional components for DNA double strand repair (Smith and Jackson, 1999). To explain the requirement for NHEJ in cortical development, we hypothesized that NHEJ would be critical for neuronal survival when neuroblasts exit the cell cycle and begin migrating and differentiating. High metabolic activity at this stage of maturation could trigger oxidative stress and lead to DNA damage. To test this, we investigated the role of Ku70 in early cerebral cortex development, by comparing the kinetics of cortical cell death and characterizing the phenotypes of dying cells in Ku70 knockout mice versus that in both heterozygous and wildtype littermates.
In Ku70−/− mice, cortical cell death is highest on embryonic day 14–15 (E14.5), at the peak of neurogenesis, and progressively declines until postnatal day 8. By contrast, in wildtype mice, cell death is low at embryonic ages and peaks on postnatal day 4, declining thereafter. Regardless of genotype, most apoptotic cells were comprised of radial glia and migrating post-mitotic neurons. Excessive apoptosis in Ku70−/− cortex was characterized by prominent clusters of dying cells in the intermediate zone and lower cortical plate. At birth, Ku70−/− mice had smaller brains, lower neuron density and persistence of radial glia. Further characterization of E14.5 cortices revealed strong staining for oxidative damage in the dying cells specifically within the intermediate zone.
In primary cortical neuronal cultures, neurons derived from Ku70−/− mice displayed a high level of cell death, which was rescued by culture under conditions of low oxygen tension and/or the addition of antioxidants. In contrast, wildtype neurons displayed a much lower level of cell death, which was not affected either by adding antioxidants or by environmental oxygen reduction. Taken together, our findings suggest that during normal cortical development, oxygen radicals and consequent DNA damage may cull some neuronal precursors and migrating neurons through apoptotic pathways. Therefore NHEJ plays a developmental role in ensuring cortical neuron survival during maturation by limiting oxidative damage.
Materials and Methods
Ku70−/− embryos and pups were obtained from timed matings between Ku70 heterozygote (Ku70+/−) pairs maintained under pathogen-free conditions in the Wesleyan Animal Facility. Use of animals in this research was strictly compliant with the guidelines set forth by the US Public Health Service in their policy on Humane Care and Use of Laboratory Animals, and in the Guide for the Care and Use of Laboratory Animals [http://www.nap.edu/readingroom/books/labrats/contents.html]. All procedures received prior approval from the Institutional Animal Care and Use Committee.
DNA Extraction and PCR Genotyping
Mice were genotyped as described (Gu et al., 1997). Tail portions were digested overnight at 55°C using 20 μg proteinase-K (Sigma-Aldrich Corp, St. Louis, MO) in 100 μl buffer containing 50 mM KCl, 0.01 mg/ml gelatin, 0.045% NP-40, 0.045% Tween-20 in 10 mM Tris buffer, pH 8.3. The digested materials were boiled at 90°C for 10 min to eliminate proteinase-K activity, centrifuged and the supernatant used for PCR genotyping using a triple primer combination with the following primers: Ku70-5′: ACACGGTTCCTTAATGTGA; Ku70-3′: GGCTGGCTTTAGCACTGTCA; and LoxR1H3: ACGTAAACTCCTCTTCAGACCT. The wildtype mice yield a 425 bp band; Ku70-deficient mice yield a 300 bp band, while heterozygous mice yield both bands.
Tissue Collection and Sectioning
Tissue was collected, fixed and sectioned from mice of different ages, as described previously (Vemuri et al., 2001). Timed pregnancies were scored by the day of mating, designated E0. Pregnant dams were euthanized by exposure to CO2, followed by cervical dislocation. Uteri were dissected and embryos removed on E14.5 or E18. Heads were placed overnight in a fixative containing 4% paraformaldehyde in 0.1 M sodium phosphate buffer, pH 7.4. Neonatal (P0) and postnatal (P4, P8) pups were euthanized by an overdose of sodium pentobarbital (Nembutal, 50 mg/kg; Henry Schein Inc., Melville, NY). Postnatal brains were fixed by transcardiac perfusion with a rinse buffer containing 0.1 M sucrose and 0.1 U/ml heparin in 0.1 M sodium phosphate buffer, pH 7.4, followed by 4% paraformaldehyde in 0.1 M sodium phosphate buffer. Brains were postfixed by immersion in the fixative overnight at 4°C.
The tissue was then cryoprotected by sequential immersion in 10, 20 and 30% sucrose solutions in 0.1 M sodium phosphate buffer, pH 7.4, and embedded in the parasagittal or coronal planes in tissue freezing medium (Triangle Biomedical Sciences, Durham, NC). Embedded tissue was stored at −80°C. Twelve-micrometer-thick serial cryostat sections of the brain were mounted on glass microscope slides (Superfrost Plus; Fisher Scientific, Atlanta, GA) and stored frozen at −80°C.
Additional animals were euthanized for cytoarchitectonic studies of the brain. Animals were perfused with 4% paraformaldehyde and 1% glutaraldehyde in 0.1 M phosphate buffer at ages E14.5–P8. Heads were postfixed overnight in the same fixative. Brains were removed and stored in paraformaldehyde containing 10% sucrose for up to 2 weeks before they were embedded in gelatin–albumin and cured overnight in fixative containing 20–30% sucrose. Twenty-micrometer-thick coronal sections were cut on a cryostat and mounted onto glass slides coated with Sta-on (Surgipath, Richmond, IL). Sections were stained with cresyl violet, cleared in xylene, mounted and photographed.
Fluorescent terminal dUTP nick end labeling (TUNEL; Serologicals Corporation, Norcross, GA) was used to detect the presence of DNA strand breaks in the tissue sections, following manufacturer's instructions for frozen sections. Only cells with evident nuclei were scored as TUNEL-positive.
Immunostaining was performed on fixed dissociated cells, as well as cryostat sections of brain. TUNEL staining was performed alone as described above, or followed by immunolabeling with one of the following reagents: rabbit anti-active caspase-3 antibodies (1:250; Cell Signaling, Beverly, MA), rabbit anti-phosphohistone H3 (pH 3, 1:1000; Upstate Biotechnology, Charlottesville, VA) or rabbit anti-Ki67 (1:1000; Accurate Chemicals and Scientific Corporation, Westbury, NY). Additional sections were stained for oxidative DNA damage using mouse anti-8-OHdG (1:1000; QED Bioscience Inc, San Diego). Phosphohistone H3 antibody was used to identify cells in the G2–M phase of the cell cycle (Bradbury, 1992). Ki67 was used to identify cells in late G1 phase, S, G2 and M phases of the cell cycle (Scholzen and Gerdes, 2000; Birner et al., 2001). Antibody against PCNA (mouse anti-PCNA, 1:10 Calbiochem, San Diego, CA) was used to identify cells in S phase (Casiano et al., 1993). Antibody against GLAST (guinea pig anti-GLAST antibody, 1:2000; Chemicon, Temecula, CA) was used to identify radial glia (Malatesta et al., 2000; Hartfuss et al., 2001). Glial cells were identified using an antibody against glial fibrillary acidic protein (rabbit anti-GFAP antibody, 1:800; Sigma-Aldrich Corporation, St Louis, MO). Post-mitotic neurons were identified by expression of the following: neuron-specific β3-tubulin — a cytoskeletal protein expressed by newly differentiated neurons (mouse anti-β3-tubulin, 1:1000; Sigma-Aldrich Corporation) (Menezes and Luskin, 1994), doublecortin (DCX) — a marker for migrating neurons (guinea pig anti-doublecortin, 1:1000; Chemicon) (Gleeson et al., 1999; Walsh and Goffinet, 2000), and calbindin — a marker in mice for certain interneurons, subsets of Cajal-Retzius cells and some layer 5 pyramidal neurons (rabbit anti-calbindin, 1:3000; Swant, Switzerland) (Anderson et al., 1997; Hevner et al., 2003). Non-specific staining was blocked by incubating the tissue in phosphate-buffered saline (PBS), containing 5% normal goat serum (Vector Laboratories, Burlingame, CA) and 0.1% Triton X-100, for 30–60 min prior to staining. Tissue was incubated with primary antibodies overnight at 4°C and labeled with Alexa dye-conjugated secondary antibodies (Molecular Probes, Eugene, OR) for 1 h at room temperature before sections were counterstained with SYTOX Green or Hoechst 33342 (Molecular Probes, 1:10,000) and mounted in Prolong Antifade (Molecular Probes). Prior to 8-OHdG immunostaining in cell culture and PCNA immunostaining on sections, slides were boiled for 2 min in 0.01 M citrate buffer to unmask the antigen. The cells/sections were stained as described (Djebaili et al., 2001). Endogenous peroxidase was blocked by incubation in 1% hydrogen peroxide in methanol. Non-specific staining was blocked by incubation in 0.1 M Tris–HCl buffer, pH 7.5, containing 0.15 M sodium chloride, 0.05% Tween-20 (TNT buffer), 1% BSA and 10% normal goat serum. Tissue was incubated with primary antibody at room temperature overnight and washed extensively with TNT buffer prior to incubation for 1 h with a secondary antibody conjugated to biotin (biotinylated goat anti-mouse IgG; Vector Laboratories). Following washing, tissue was incubated with streptavidin-conjugated to horseradish peroxidase (Vector Laboratories). Staining was visualized using a Tyramide signal amplification kit (Perkin Elmer Life Sciences, Boston, MA).
Phenotypic Analysis of Dissociated Cells
The telencephalic hemispheres were dissected from neonates or embryonic mice (E14.5) in ice-cold Hanks' balanced salt solution (HBSS), then digested in HBSS with 0.125% trypsin (Invitrogen, Carlsbad, CA) for 20 min at 37°C. The enzymatic dissociation was stopped by the addition of soybean trypsin inhibitor (Sigma-Aldrich Corporation) and the cortex was dissociated by gentle trituration with a fire-polished pipette to obtain a single cell suspension. Cells were then washed in HBSS, pelleted, resuspended in Neurobasal medium containing B27 supplement (Invitrogen Life Technologies), and plated on poly-D-lysine-coated, four-well Permanox chamber slides (Nalge Nunc, Rochester, NY) at a density of 7.5 × 105 cells/well. The culture slides were incubated at 37°C in 20% oxygen and 5% CO2 for 2 h to allow the cells to adhere to the slides, then rinsed and incubated in fixative for 15 min at room temperature before immunocytochemical staining was performed, as described above. Maintaining the cultures at 20% or 3% oxygen did not significantly alter the composition of the cultures. This eliminates possible effects of differential integrin expression due to oxidative stress resulting in altered cell adhesion. Quantification of TUNEL/immunocytochemistry. Four to five cryostat sections were stained per case and approximately 1000 nuclei were counted for each section. With dissociated cells, a minimum of 1000 cells was counted per well in a four-well chamber slide. At least four animals per genotype per age were studied and littermates were compared. Quantification was carried out blind with respect to the genotype of the tissues.
Pyknotic Cell Counts
Counts of pyknotic cells were done in E14.5 embryonic brain regions, including anterior tectum (where the roof plate meets the midbrain), pons, ganglionic eminence, hypothalamus and caudate putamen, in 20 μm thick cresyl violet acetate-stained sections. Approximately eight sections were counted per brain. Pyknosis was determined by counting the number of pyknotic cells and the total number of cells in a field measuring ∼6400 μm2 per section, using a microscope equipped with brightfield illumination and a ×100 oil immersion objective containing a reticule (Klarmann Rulings, Litchfield, NH; 100 square grid; 20 mm diameter, 10 mm on each side of grid). Cells touching the upper or left boundary of the grid were included in the counts; cells touching the lower or right boundary of the grid were not counted.
Low and High Oxygen Tension Experiments
Telencephalic cells were isolated from neonates, and dissociated cultures were prepared as described previously (Chechlacz et al., 2001). Physiological oxygen tension in developing human and mouse CNS is ∼30 mm Hg, which translates to ∼2–5% oxygen concentration (Ganong, 2001; Stroka et al., 2001; Karanjawala et al., 2002). Hence, we grew the cultures under 3% oxygen versus standard atmospheric 20% oxygen. Cultures were placed in a modular incubator chamber (Billups-Rothenberg Inc., Del Mar, CA), and the chamber itself was placed in a 37°C incubator. The chamber was purged daily with a precertified mixture of 3% oxygen, 5% carbon dioxide and 92% nitrogen (Airgas Inc., Hartford, CT). For comparison with standard tissue culture conditions, additional cultures consisting of Ku70−/−, Ku70+/− or Ku70+/+ neonatal cortical neurons were maintained in the 37°C incubator with 20% oxygen, 5% carbon dioxide and 75% nitrogen. All cultures were incubated for 7 days in vitro (DIV). At 7 DIV, cultures were fixed with 4% paraformaldehyde. TUNEL and Hoechst staining were performed to quantify apoptosis.
Treatment with Antioxidants
Primary neuronal cultures were prepared from neonatal Ku70+/+, Ku70+/− and Ku70−/− mice as described previously. The cultures were immediately treated with 0.5 mM Tempol (Sigma-Aldrich Corporation) or vehicle (methanol). Tempol is a free radical scavenger (Reddan et al., 1992>) and neuroprotective at the concentration used (Lang-Rollin et al., 2003). One set of cultures was maintained at 20% oxygen in a tissue culture incubator and another set at 3% oxygen in the modular incubator chamber for 7 days, as described earlier. At 7 DIV, cultures were fixed with 4% paraformaldehyde and apoptosis was quantified as described earlier.
Microscopy and digital images were obtained using the following microscopes: Zeiss Axioplan with apotome module, Zeiss Axiovert with Axiovision software, or Zeiss LSM 510 confocal microscope. Figures were prepared using Adobe Photoshop software.
A two-tailed Student's t-test was used to determine significant differences between means, assuming unequal variances.
Ku70−/− Mice Have Smaller Bodies and Brains
Earlier studies reported that Ku70−/− mice are smaller than wildtype littermates and undergo premature senescence at ∼1–3 months after birth (Gu et al., 1997). Gross morphological examination of Ku70−/− pups versus their wildtype littermates confirmed that the knockout mice have smaller bodies and brains (Fig. 1A–D). No apparent cytoarchitectural differences were found when Ku70−/− brain sections were compared with heterozygous littermates (Fig. 1E,F). Cresyl staining of E18 brains showed that Ku70 deficiency resulted in a reduction in thickness of all cortical layers, particularly the cortical plate (Fig. 1G–J).
Pattern of Ku70 Expression in the Cerebral Cortex
Prior studies demonstrated expression of Ku80/86 in the developing cerebral cortex of human fetuses (Oka et al., 2000) and expression of Ku70 in the adult mouse brain (Koike et al., 1996). However, the pattern has not been examined in the developing rodent brain. To determine when and where Ku70 is expressed in the developing cerebral cortex, we performed immunostaining for Ku70 in combination with a nuclear stain to visualize cortical layers. Ku70 knockout mice exhibited no immunoreactivity for Ku70, as expected (Fig. 2A,B). Ku70 heterozygous mice exhibited robust intracellular staining throughout all layers of the cerebral cortex, including ventricular and subventricular zones (Fig. 2C–F). We observed prominent nuclear staining as well as immunoreactivity within apical dendrites.
Ku70−/− Mice Exhibit Increased Cerebro-cortical Cell Death in Prominent Clusters
Previous studies showed increased pyknosis and TUNEL in the CNS of Ku70−/− mice between E11.5 and E13.5 (Gu et al., 2000). To delineate developmental changes resulting from Ku70 deficiency, we analyzed cell death in the cerebral cortex at later ages. We initially compared the extent of cell death in the cerebral cortex by examining pyknosis, TUNEL staining and immunostaining for active caspase-3. At E14.5, the Ku70−/− cortex displayed significant increases in pyknosis compared with heterozygous littermates (Fig. 3A,B). At this age, apoptotic cells in Ku70−/− embryos were present throughout the ventricular zone, subventricular zone, intermediate zone, subplate and cortical plate of the developing cerebral wall. Only the knockout mice displayed prominent clusters of pyknotic cells in the intermediate zone, subplate, and rarely in the cortical plate (arrowheads in Fig. 3). Similarly, we observed numerous clusters of TUNEL-positive cells in cortices of Ku70−/− mice (Fig. 3C), but TUNEL-positive cells were rare in the wildtype or heterozygous embryos (Fig. 3D).
To define the underlying cell death pathways, we correlated TUNEL staining with the expression of active caspase-3. Many TUNEL-positive cortical cells expressed active caspase-3 (Fig. 3E,F) and were localized to the IZ and lower CP in Ku70−/− cortices at E14.5. Most cells in the cluster in the Ku70−/− cortex co-labeled for both TUNEL and active caspase-3 (inset, Fig. 3E).
To determine the laminar locations of dying cells in the Ku70−/− cortex, we further analyzed pyknotic clusters (Table 1). Although they were occasionally found in the cortical plate, ventricular and subventricular zone, by far the majority were located in the intermediate zone. On average, there were 10–12 clusters per brain section and each cluster contained 9–10 cells. The clusters were 100–150 μm in diameter. Collectively, these results suggest that Ku70 deficiency in embryonic cortical neurons can lead to DNA damage, caspase-3 activity and clusters of apoptotic neurons within the intermediate zone.
The location, number of clusters, and the average number of dying cells in a cluster in parasagittal sections of individual Ku70−/− mice at E14.5 were analyzed in six embryos. Clusters of dying cells were chiefly in the intermediate zone. Less frequently, clusters were seen in the ventricular/subventricular zone and rarely in the cortical plate.
Cell Death in Ku70−/− Mice is Elevated Compared with Wildtype Mice on E14.5, E18 and P0, but not P4 and P8
To further define the extent and time course for abnormal cell death in the cerebral cortex of Ku70−/− mice, we performed quantitative studies to compare knockouts with heterozygous and wildtype littermates (Fig. 4). On E14.5, Ku70−/− mice (solid squares) exhibit a 35- to 40-fold increase in TUNEL-positive cells, a 25- to 30-fold increase in pyknotic cells and a 20- to 30-fold increase in cells displaying active caspase-3, compared with heterozygous (gray triangles) or wildtype (open circles) littermates. On E18 and P0, similar trends were observed; however, the difference in levels of cell death was less striking. No statistically significant differences were observed between heterozygous and wildtype littermates at any age in vivo.
By P4, cortical cell death in the cerebral hemispheres peaked in heterozygous mice. At this age the magnitude of cell death was comparable in heterozygous and knockout littermates. By P8, apoptotic cells were rarely observed and no differences were found between wildtype, heterozygous, or Ku70−/− mice.
To determine whether Ku70 deficiency was linked to increased apoptosis in other brain areas, we next examined the striatum, ganglionic eminences, hypothalamus and pons at E14.5. Of these regions, the striatum and ganglionic eminences showed the largest increase in cell death in Ku70−/− mice (see Supplementary Material, Fig. 1). These additional findings demonstrate that there are striking regional variations in the magnitude of apoptosis triggered by Ku70 deficiency.
The quantitative measures of cell death show that Ku70 deficiency results in massive embryonic cell death in multiple brain areas. This effect is most pronounced at early stages of nervous system development, when levels of cell death in control mice are relatively low. By P4, when developmental death peaks in wildtype mice, we did not observe any differences in apoptosis. Because Ku70 deficiency-enhanced cell death is restricted to early cortical development, our data suggest that the increased embryonic cell death in Ku70−/− embryos is triggered by factors that are distinct from those that regulate cortical cell death postnatally.
Ku70 Deficiency Increases the Death of Progenitors and Immature Neurons
Because the preferred pathway for DNA repair in non-dividing cells utilizes NHEJ, a Ku70-dependent process, we predicted that many post-mitotic cells would be vulnerable to Ku70 deficits. To test this, we characterized apoptotic cells by combining TUNEL with immunostaining for other markers. The cell cycle phase of dying cycling cells was identified by co-labeling with Ki67, a marker for late G1–M phase (Fig. 5A–C), or PCNA, a marker for S phase (Fig. 5D–F). Similarly, the phenotypes of the dying cells were identified by co-labeling with GLAST, a marker for radial glia (Fig. 5G–I), or by co-labeling with the neuronal markers doublecortin or β3-tubulin (Fig. 5J–O). We confirmed co-localization of TUNEL with antigenic markers by optical sectioning of labeled cells using confocal microscopy. In Ku70−/− mice, each of the phenotypic markers showed some co-labeling with TUNEL. However, the most notable dual labeling was found in cells expressing radial glial or neuronal antigens (Fig. 5G–O). In Ku70−/− embryos, many of the apoptotic cells were clustered in the intermediate zone and dual-labeled for TUNEL and neuronal antigens β3-tubulin and doublecortin.
Prior studies have suggested that abnormal cell death in humans and mice may be preceded by abortive re-entry into the cell cycle (Herrup and Busser, 1995; Yang et al., 2001, 2003). To determine whether re-entry into the cell cycle may contribute to cell death, we quantified cells co-labeled by TUNEL and cell cycle markers (Fig. 6A). Quantification of TUNEL-positive cells in G2-M phase at E14.5 (TUNEL+/pH3+) revealed no significant differences between the Ku70−/− and Ku70+/− mice. By contrast, the number of apoptotic cells in mid-G1 through M phase (TUNEL+/Ki67+) showed a fivefold increase in Ku70−/− brains. Similarly, in the Ku70−/− cortex, the proportion of dying cells in the S phase of the cell cycle (TUNEL+/PCNA+) increased from 0.5 to 1.4%, a threefold increase. These results show that the number of dying cells in the G2–M phase does not change in Ku70-deficient mice, but cells dying in the S or G1 phase of the cell cycle are increased significantly. Our observations suggest that in the developing cerebral cortex, defective DNA repair results in some cells undergoing abnormal reentry into the cell cycle, S phase arrest, and death.
We next determined the proportions of the dying cells that were neural progenitors (GLAST-positive) or postmitotic neurons (β3-tubulin-positive, doublecortin-positive). In Ku70−/− brains from E14.5 embryos, ∼2% of the cells co-labeled for TUNEL and GLAST, representing a 19-fold increase in radial glial cell death above the levels observed in heterozygous embryos. Similarly, nearly 2% of the cortical cells co-labeled for TUNEL and β3-tubulin (Fig. 6B), a 25-fold increase in immature neuronal cell death above values obtained in heterozygous mice at this age. Finally, ∼0.6% of the cortical neurons underwent cell death during the course of migration, as indicated by co-expression of TUNEL and doublecortin, a fourfold increase over heterozygous mice at this age. When using a mixture of these antibodies, nearly all TUNEL+ cells were co-labeled, demonstrating that we have accounted for the phenotypes of most apoptotic cells.
By P0, cell death was strikingly reduced in Ku70−/− mice but significant differences were still observed when the brains were compared with heterozygous littermates. Less than 1% of the apoptotic cells were in the G2–M phase, levels comparable to heterozygous mice at this age. By contrast, there was a 10-fold increase in the fraction of dying cells that were in G1 phase. The Ku70−/− cortex also exhibited a 50-fold increase in radial glial cell death, a 16-fold increase in apoptotic neurons and a 20-fold increase in apoptotic migrating neurons compared with heterozygous or wildtype littermates (Fig. 6C). Therefore, despite a drop in the overall proportion of dying cortical cells between E14 and P0, Ku70 deficiency enhanced the death of radial glia and immature migrating neurons at P0, as well as E14.5.
Increased Cell Death in Ku70−/− Mice Alters the Composition of the Cortex
Our findings of a significant increase in the death of migrating neurons and radial glia at embryonic and neonatal ages, combined with our observation that the cortical layers are thinner in Ku70-deficient mice, led us to predict that the cellular composition of the cerebral cortex would be altered at birth. To test this prediction, we carried out a quantitative phenotypic analysis of the cortex, using freshly dissociated cortical cells from neonatal brains. The composition of Ku70−/− cerebral hemispheres was in fact quite distinct from the heterozygous mice at this age (Table 2). The most striking differences were found for the proportions of radial glial cells. GLAST-positive radial glia in the Ku70−/− cortex were increased by 15% at P0 despite the high levels of cell death in this population. Additionally, there was a slight increase in mitotic cells. The proportion of cortical neurons was decreased by 10%. Migrating neurons, which constitute a subset of the immature neurons, were also reduced in Ku70−/− mice by P0.
|Phenotypic marker||Marker-positive cells/1000 cells|
|Phenotypic marker||Marker-positive cells/1000 cells|
Significantly higher proportions of dividing cells (pH 3) and progenitors (GLAST) were present at P0 in Ku70−/− cortices. Depletion of immature migrating neurons was indicated by selective decreases in the β3-tubulin and doublecortin populations. The calbindin-expressing interneuron population remained constant. Dissociated cells from at least four neonatal mice were analyzed for each marker.
Cortical interneurons originate separately from projection neurons, and migrate tangentially into the cerebral cortex from the ventral forebrain during embryogenesis (Anderson et al., 1997; Nery et al., 2002). Calbindin is transiently expressed within these subsets at embryonic ages (Anderson et al., 1997; Hevner et al., 2003). To test whether Ku70 deficiency reduced the survival of calbindin-positive interneurons, we quantified these cells. Surprisingly, the proportion of calbindin-positive interneurons was comparable in Ku70-deficient, heterozygous or wildtype mice. These findings suggest that Ku70 deficiency may selectively target radially migrating cortical projection neurons, while sparing or compensating for cortical interneurons.
Oxidative DNA Damage
Spontaneous DNA breaks can occur in neurons due to reactive oxygen species, gamma rays, or malfunctions of endogenous DNA repair enzymes (Lieber et al., 2003). The presence of 8-OHdG is one of the hallmarks of oxidative DNA damage and identifies dying neurons (Klein et al., 2002). Increased oxidative damage is a strong trigger for neuronal death in animal models of neurological diseases (Klein et al., 2002; Klein and Ackerman, 2003). Moreover, studies have documented oxidative stress in vulnerable neurons in human neurodegenerative diseases, including Alzheimer's, Huntington's, Parkinson's, Cockayne syndrome, and ataxia telangiectasia (Davydov et al., 2003; Lieber et al., 2003; Tuo et al., 2003; Waters, 2003). To determine whether oxidative DNA damage was linked to the embryonic cell death observed in Ku70−/− brains, we performed 8-OHdG immunostaining (Fig. 7). Oxidative damage, as evidenced by strong perinuclear staining, was most prominent in cells located in the cortical intermediate zone of Ku70−/− mice at E14.5 (Fig. 7A). Heterozygous embryos exhibited only rare staining for oxidative damage (Fig. 7B). Within the dying cells of a cluster, heterogeneous 8-OHdG staining was observed; some cells showed oxidative DNA damage, while others did not, despite having pyknotic nuclear morphology. Interestingly, we did not observe 8-OHdG immunostaining in the VZ/SVZ or CP, where many TUNEL-positive cells were found. These data suggest that oxidative stress was most prominent in postmitotic neurons, particularly in neurons migrating through the intermediate zone.
Low Oxygen Conditions and Antioxidant Tempol Rescue Ku70-deficient Neurons from Cell Death
To test whether oxidative stress was a potential trigger for neuronal death, we used an in vitro paradigm. Lieber and colleagues have shown that increasing levels of reactive oxygen species in Ku80/86-deficient fibroblast cultures caused a marked increase in cell death (Karanjawala et al., 2002). To further investigate whether the oxidative DNA damage is linked to apoptosis of Ku70−/− neurons, we quantified levels of cell death in neuronal cultures grown under high (20%) or low (3%) oxygen concentrations. At 20% oxygen, cultures derived from Ku70−/− cortices had a fivefold higher rate of pyknosis after 7 DIV as compared with wildtype littermate controls. Unlike in vivo, the cortical cultures from Ku70 heterozygous littermates also showed intermediate levels of pyknosis between wildtype and knockout littermates (Fig. 7C). By contrast, maintaining the cultures in 3% oxygen resulted in a significant decrease in both pyknosis and TUNEL in the Ku70−/− and Ku70+/− cultures (Fig. 7D). Since lowering the environmental oxygen concentration rescued both Ku70-deficient and heterozygous neurons from apoptosis, our findings suggest that oxygen free radicals may trigger cell death induced by Ku70 deficiency. To confirm that rescue was mediated by lowered oxidative stress, we compared cell death at 20 and 3% oxygen with or without Tempol, a spin-trapping agent that is cell-permeable and scavenges superoxide ions by converting superoxide anions to O2 and H2O2, similar to superoxide dismutase (Reddan et al., 1992). In 20% oxygen, Tempol treatment significantly increased neuronal survival in Ku70−/− and Ku70+/− cultures to wildtype levels, whereas Tempol caused no additional rescue of cultures grown at 3% O2. Additionally, we stained cultures for 8-OHdG and found reduced staining in Ku70−/− primary neuronal cultures when maintained at 3% O2 or treated with antioxidants. Wildtype cultures displayed only very low levels of 8-OHdG staining which did not differ significantly with or without treatment (see Supplementary Material, Fig. 2). These data support our model that during development, oxidative stress is an important cause of neuronal death, an effect exacerbated by Ku70 deficiency.
We utilized Ku70-deficient mice to gain insights into why DNA repair is critical for neuronal survival during brain development. Results presented here show that Ku70 deficiency enhances cortical cell death chiefly during periods of neuronal migration and neurogenesis. Multiple pathways may trigger cell death within the developing brain, since activated caspase-3 was observed in migrating neurons in the IZ and CP, but not in dying progenitors within the VZ/SVZ. Ku70 deficiency-enhanced cell death was conspicuous from E14.5, the peak of neurogenesis, until birth. By P4, when cortical cell death peaked in wildtype littermates, cell death was comparable in the different genotypes. Therefore, Ku70 plays an important role in cell survival chiefly during the embryonic period of neural development.
In Ku70−/− embryos, TUNEL-positive cells were distributed in all layers of the developing cerebral cortex, but clusters of apoptotic neurons were found mainly in the intermediate zone (Table 1). These clusters of migrating neurons contained pyknotic cells exhibiting TUNEL with only some expressing active caspase-3 (Fig. 3) or oxidative DNA damage (Fig. 7). Similar clusters of dying cells have been reported in the developing retina, where it was shown that gap junctions permit diffusion of apoptotic signals into a neighboring cell, triggering apoptotic cascades (Cusato et al., 2003). Other examples of physical coupling have been shown to occur in the ventricular zone during early stages of cortical development (LoTurco and Kriegstein, 1991) and between migrating cortical neurons and radial glia (Nadarajah et al., 1997).
We further show that Ku70 deficits cause apoptosis of radial glia and migrating neurons, based on phenotypic analyses combining TUNEL with neuronal and radial glial markers. Enhanced apoptosis within these subsets results in proportionally more progenitors and fewer post-mitotic neurons in Ku70−/− brains by birth. That selective cell death occurs in cortical neurons during radial migration is supported by our demonstration that there is no change in the proportion of calbindin-positive neurons. Although a subset of interneurons originate in the cortex (Hevner et al., 2003), most interneurons are born outside of and migrate into the cortex along a tangential route originating in the medial and posterior ganglionic eminences (Anderson et al., 1997; Nery et al., 2002; Xu et al., 2004). Curiously, we observed enhanced cell death in the ganglionic eminences of Ku70−/− mice (supplementary data), but did not find differences in calbindin-expressing neurons, suggesting that, as in the cortex, projection neurons in subcortical regions are more vulnerable than interneurons. Although beyond the scope of this study, it would be of interest to know whether interneurons are more resistant to DNA damage or oxidative stress.
Results presented here demonstrating that Ku70 deficiency increases apoptosis of Ki67+ neuroblasts (late G1–M phase) and PCNA+ (S phase) but not pH3+ (late G2–M phase) neuroblasts suggest that Ku70 enhances survival when cells are in G1–S phase when NHEJ is the predominant pathway of DNA repair (Lieber, 1999; Essers et al., 2000; Jones et al., 2001; Barnes, 2002). This supports the interpretation that the Ku70 deficiency-enhanced cell death is specifically due to NHEJ deficiency.
The necessity for DNA double-strand break repair during neuronal migration can be explained by our demonstration that dying neurons in the intermediate zone had elevated 8-OHdG levels, a hallmark of oxidative DNA damage (Klein et al., 2002; Klein and Ackerman, 2003). These findings indicate that migrating neurons may produce reactive oxygen species (ROS), which cause DNA damage and trigger cell death. An involvement of ROS in developmental cell death is further supported by our in vitro studies demonstrating that that higher oxygen levels result in increased oxidative damage and cell death, whereas lower oxygen levels rescue Ku70-deficient neurons. These manipulations had no effect on survival of wildtype neurons in culture, indicating that these effects depend upon Ku70 deficiency. The elevated neuronal death in cortical cultures from Ku70+/− mice contrasts with our in vivo studies, where neuronal death in heterozygous animals was comparable to that in wildtype littermates. One explanation may be that in vitro conditions may exert more stress on cultured cells. The absence of exogenous factors that would normally promote cell survival in vivo may promote cell death in heterozygous cultures. Thus, we speculate that oxygen free radicals are a major factor linked to DNA damage and neural cell death in Ku70−/− embryos. In wildtype or heterozygous embryos, expression of Ku70 and intact NHEJ would result in the repair of oxidative damage occurring during neural differentiation and migration.
Additional mechanisms may also instigate cell death during cortical development. Lack of 8-OHdG immunoreactivity in the VZ/SVZ suggests that DNA damage may be triggered by additional factors in progenitors. Previous work suggested that recombination may be an active process in creating functional diversity within the ventricular zone (Chun and Shatz, 1999). Such recombination-dependent functional diversity could generate free DNA ends, which, if unrepaired, could trigger cell death within the ventricular or subventricular zones. This alternative form of cell death may account for our inability to completely rescue Ku70−/− neurons under low oxygen tension.
We conclude that during early embryonic development of the mammalian cortex, significant cell death occurs via apoptotic mechanisms involving oxidative stress, DNA damage, and caspase-3 activation. Ku70 deficiency reveals the prevalence of oxidative damage during neuronal migration and the consequences of DNA repair deficits in the developing cerebral cortex. Post-mitotic neurons with unrepaired DNA breaks may be forced to undergo apoptosis when DNA damage signals lead to re-entry into the cell cycle and progression is arrested (Herrup and Arendt, 2002; Yang et al., 2003). Several neurodegenerative disorders of childhood involve defective DNA repair and are accompanied by severe neurologic dysfunction (Fuller and Bohr, 1999). Future studies of DNA damage sensors and downstream apoptotic signals may provide insights into how defective DNA repair triggers nervous system cell death and lend a better understanding of the significance of DNA repair during normal and abnormal human brain development.
Supplementary material can be found at http://www.cercor.oupjournals.org/
The authors would like to thank Dr Paul J. Lombroso for helpful comments. A breeding pair of Ku70+/− mice was generously provided by Drs JoAnn Sekiguchi and Fredrick W. Alt. We also thank Ron Gordon and Greg Pare of the Wesleyan Animal Facility. This work was supported by NIH RO1 NS4286 (J.R.N.) and an HHMI undergraduate fellowship (A.T.).
1Program in Neuroscience and Behavior, Department of Biology, Wesleyan University, Middletown, CT 06459-0171, USA and 2Department of Psychiatry, UMDNJ-Robert Wood Johnson Medical School, 675 Hoes Lane, Piscataway, NJ 08854-5635, USA