Characterization of Age-Dependent and Progressive Cortical Neuronal Degeneration in Presenilin Conditional Mutant Mice

Presenilins are the major causative genes of familial Alzheimer's disease (AD). Our previous study has demonstrated essential roles of presenilins in memory and neuronal survival. Here, we explore further how loss of presenilins results in age-related, progressive neurodegeneration in the adult cerebral cortex, where the pathogenesis of AD occurs. To circumvent the requirement of presenilins for embryonic development, we used presenilin conditional double knockout (Psen cDKO) mice, in which presenilin inactivation is restricted temporally and spatially to excitatory neurons of the postnatal forebrain beginning at 4 weeks of age. Increases in the number of degenerating (Fluoro-Jade B+, 7.6-fold) and apoptotic (TUNEL+, 7.4-fold) neurons, which represent ∼0.1% of all cortical neurons, were first detected at 2 months of age when there is still no significant loss of cortical neurons and volume in Psen cDKO mice. By 4 months of age, significant loss of cortical neurons (∼9%) and gliosis was found in Psen cDKO mice. The apoptotic cell death is associated with caspase activation, as shown by increased numbers of cells immunoreactive for active caspases 9 and 3 in the Psen cDKO cortex. The vulnerability of cortical neurons to loss of presenilins is region-specific with cortical neurons in the lateral cortex most susceptible. Compared to the neocortex, the increase in apoptotic cell death and the extent of neurodegeneration are less dramatic in the Psen cDKO hippocampus, possibly in part due to increased neurogenesis in the aging dentate gyrus. Neurodegeneration is also accompanied with mitochondrial defects, as indicated by reduced mitochondrial density and altered mitochondrial size distribution in aging Psen cortical neurons. Together, our findings show that loss of presenilins in cortical neurons causes apoptotic cell death occurring in a very small percentage of neurons, which accumulates over time and leads to substantial loss of cortical neurons in the aging brain. The low occurrence and significant delay of apoptosis among cortical neurons lacking presenilins suggest that loss of presenilins may induce apoptotic neuronal death through disruption of cellular homeostasis rather than direct activation of apoptosis pathways.


Introduction
Presenilins (Psen 1 and 2) are the major causative genes of earlyonset familial Alzheimer's disease (FAD) and harbor ,90% of the identified FAD-linked mutations. Presenilins are broadly expressed and play essential roles during embryonic development [1,2,3,4]. Specifically, presenilins are required for maintenance of neural progenitor population as well as neurogenesis and neuronal migration [4,5,6,7,8]. To circumvent the requirement of presenilins in development, we previously generated a presenilin conditional double knockout (Psen cDKO) mouse, in which presenilin inactivation is restricted spatially and temporally to excitatory neurons of the postnatal forebrain using the Cre/loxP technology [9,10]. Thus, Psen cDKO mice permit assessment of direct consequences of presenilin inactivation in excitatory pyramidal neurons of the adult cerebral cortex, where presenilins are normally expressed highly and AD pathogenesis occurs. Analysis of these mutant mice demonstrated that loss of presenilins in mature neurons of the cerebral cortex results in progressive impairment in synaptic plasticity and learning and memory, followed by age-dependent neurodegeneration [10,11].
In Psen cDKO mice at 2 months of age, approximately one month after presenilin inactivation, memory impairment as well as specific presynaptic and postsynaptic defects were found in the absence of significant loss of cortical neurons or volumes [10]. By 6 and 9 months of age, 18% and 24% of cortical neurons were lost, respectively [10]. These results were further supported by an independent study using a similar Cre line, which found elevated levels of gliosis and decreased cortical volume at 10 months of age [12]. Presenilins promote memory and neuronal survival in a c-secretase-dependent manner, as conditional inactivation of another component of the c-secretase complex, nicastrin, results in similar patterns of memory impairment and age-related neurodegeneration [13]. However, the precise time of the onset of neuronal degeneration and the mode of neuronal death were less clear in Psen cDKO mice.
In the current study, we show how loss of presenilin function in the adult brain leads to age-dependent neurodegeneration. Cell death induced by loss of presenilins begins at 2 months of age in the cerebral cortex via apoptosis. Remarkably, only a very small percentage of cortical neurons undergo apoptotic cell death at any given time point, though over time an increasingly higher percentage of cortical neurons are lost. Interestingly, hippocampal neurons are less vulnerable to cell death induced by loss of presenilins, whereas lateral cortical neurons are particularly vulnerable, suggesting brain subregion specificity for presenilindependent neuronal survival.

Tissue preparation
To prepare brain samples for immunostaining, Nissl and Fluoro-Jade B staining, and TUNEL labeling, mice were euthanized with carbon dioxide and perfused with 4% paraformaldehyde in phosphate-buffered saline (PBS). After perfusion, brains were removed and postfixed 2 hours at 4uC in the same fixative. Each brain was bisected sagittally, such that one hemisphere was prepared for paraffin embedding while the other was cryoprotected by immersion in 30% sucrose/PBS overnight at 4uC and embedded in OCT for frozen sectioning. For Nissl staining following by stereological neuron counting, 10 mm thick paraffin sections were collected; for immunostaining, Fluoro-Jade B, and TUNEL analysis, 20 mm thick sections were collected on a cryostat.

Neuron count
10 mm thick parasagittal paraffin sections were stained with cresyl violet, and neuron number was estimated using the fractionator and optical dissector methods under an Olympus BX51 light microscope equipped with a CCD camera connected to a computer running Bioquant image analysis software. Cortical volume was also estimated using the same software and equipment. The experimenter was blind to the genotypes of the mice. Statistical significance was determined using Student's t test.
Immunostaining and quantification 20 mm thick frozen sections were rinsed in PBS followed by antigen retrieval in boiling citrate buffer. Sections were blocked with 5% normal goat serum (NGS), 0.03% TritonX-100 in PBS for 1 hour at room temperature, and incubated in primary antibody diluted in blocking buffer overnight at 4uC. The following antibodies were used: rabbit anti-cleaved caspase-9 (Asp353) (Cell Signaling Technology, 1:200), and rabbit anticleaved caspase-3 (Asp175) (Cell Signaling Technology, 1:100). Primary antibodies were detected with fluorescent secondary antibodies (Alexa Fluor 488 or 594, Molecular Probes) diluted 1:300 in 5% NGS/PBS. Images of stained sections were captured on a Zeiss LSM510 laser-scanning confocal microscope. Quantification was performed blind to genotype on a series of sections from medial to lateral in each brain, and then the average number of positive cells per section was determined for each.

TUNEL labeling and quantification
For labeling of brain sections, a series of 10 frozen sagittal sections at approximately 300 mm intervals were stained and quantified per brain. TUNEL labeling was performed using the Roche In Situ Cell Death Detection Kit as per the manufacturer's recommended protocol, with the following modifications: antigen retrieval was performed using boiling citrate buffer; TdT enzyme was diluted 1:20 in labeling mix; and slides were blocked for 30 min at room temperature prior to TUNEL labeling with buffer containing 10% normal goat serum, 3% BSA in 0.1 M Tris pH 7.5. All positive cells were counted per section in the neocortex and hippocampus by an experimenter blind to genotype. The number of positive cells per 20 mm-thick section was determined, and then averaged per genotype.

Electron microscopy and quantitative analysis
Three pairs of age-matched mice (3 control, 3 PS cDKO, 5 months old) were perfused with 1% glutaraldehyde and 2% paraformaldehyde in 0.1 M phosphate buffer. Brains were removed and immersion fixed in 2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 M cacodylate buffer overnight at 4uC. The 150 mm-thick sections were made by a vibratome and postfixed in 1% OsO 4 , 0.8% potassium ferricyanide in the same buffer for one hour at room temperature. Specimens were stained en bloc with 2% aqueous uranyl acetate for 15 min, dehydrated in a graded series of ethanol to 100% and embedded in Poly/bed 812 (Polysciences Inc., Warrington, PA). Thin sections (60 nm) were made by a Leica Ultracut microtome and post-stained with uranyl acetate and lead citrate. The sample grids were examined with an FEI Tecnai transmission electron microscope at 120 kV of accelerating voltage, and the digital images were captured with a SIS Morada CCD camera.
For quantification of mitochondria, 13,0006 magnification images were collected and imported into ImageJ. The scale of magnification/pixel density was calibrated, and each mitochondrion per image was counted and its area quantified (ImageJ). The total number of mitochondria per image per genotype was scored, and the average number per genotype was calculated and compared. The size distribution of mitochondria per image was determined by assignment to arbitrary bins based on area. Three pairs of mice age 5-6 months were analyzed.

Time course of neurodegeneration in the cerebral cortex of Psen cDKO mice
To evaluate age-related neurodegeneration caused by loss of presenilins, we performed histological analysis of postnatal forebrain-specific Psen cDKO (fPsen1/fPsen1; Psen22/2; aCaM-KII-Cre) and control (fPsen1/fPsen1) mice from 2 to 22 months of age (Fig. 1A). We previously showed that expression of Psen1 mRNAs and proteins are unaltered in fPsen1/fPsen1 mice relative to wild-type mice, and that presenilin-1 inactivation begins at 3-4 weeks of age postnatally [14]. Comparison of comparable Nissl stained brain sections revealed a subtle thinning of the cerebral cortex at 4 and 6 months of age, and a striking reduction in the size of the cortex by 22 months of age (Fig. 1A). Quantitative stereological analysis showed similar cortical volume in Psen cDKO mice at 2 months and a significant reduction at 4 months  Table 1). Thus, inactivation of presenilins causes age-related, progressive loss of cortical volume with no significant loss at 2 months, ,22% loss at 4 months, ,35% loss at 6-9 months and ,53% loss at 22 months (Table 1).
We further quantified the number of neurons in the cerebral cortex of Psen cDKO mice at 4 months of age, as we previously found no significant reduction in the number of cortical neurons at 2 months, and loss of 18% and 24% of cortical neurons at 6 and 9 months, respectively [10]. Neuron count using stereological methods identified a small but significant decrease in cortical cell number in Psen cDKO at 4 months (8.7%, p,0.01; Fig. 1C; Table 1). These data indicate progressive neuronal cell loss in the absence of presenilins in the cerebral cortex. To assess neurodegeneration further in Psen cDKO mice, we used Fluoro-Jade B staining, which detects degenerating neurons [15,16,17,18], to quantify the number of degenerating neurons in Psen cDKO mice at 2 and 4 months of age ( Fig. 1D; Table 1). In the neocortex of Psen cDKO mice, the number of degenerating neurons was significantly increased at 2 months (7.6-fold; p,0.00001) and 4 months of age (9.0-fold; p,0.00001), compared to the control (Fig. 1D). Thus, while the total number of cortical neurons and the size of the cortical volume are not significantly altered at 2 months of age in Psen cDKO mice, the number of degenerating neurons is dramatically increased. However, despite the large increases in degenerating neurons in Psen cDKO mice, the absolute number of degenerating neurons is still rather small (,15-20 cells in the 20 mm sagittal section in the neocortex), representing ,0.1% of neurons at 2 months of age (Table 1).

Apoptotic cell death in the cerebral cortex of Psen cDKO mice
To investigate how neurons die in the absence of presenilins in the adult cerebral cortex, we performed TUNEL analysis to label apoptotic cells ( Fig. 2A, B). Indeed, significant increases in TUNEL+ cells were found in the Psen cDKO neocortex (2 months: 7.4-fold increase, p,0.00001; 4 months: 15.0-fold increase, p,0.00001; Fig. 2B). Interestingly, no increase was found in TUNEL+ cells in the neocortex of Psen cDKO mice at 6 weeks of age (p.0.05; Fig. 2B), despite the fact that presenilin inactivation occurs by postnatal 4 weeks of age, indicating a delay in the onset of apoptotic neuronal death caused by loss of presenilin function.
To confirm further the presence of apoptotic cells, we performed immunostaining using antibodies specific for active (cleaved) forms of caspases 9 and 3, which are excellent markers for apoptosis [19,20,21]. At 2 months of age, more active caspase 9-positive cells (7.160.3 per 20 mm sagittal section) were found in the Psen cDKO neocortex, compared to the control (4.160.1 per section; p,0.002; Fig. 2C). Similarly, at 4 months of age, more

Mild neurodegeneration in the hippocampus of Psen cDKO mice
We previously reported that presenilins are essential for neurotransmitter release, NMDAR-mediated functions and longterm potentiation in the hippocampus [10,22]. We further examined whether loss of presenilins causes age-related neurodegeneration in the hippocampus, similar to the neocortex. Stereological measurement revealed unchanged hippocampal volumes in  Fig. 3A, B), which is much smaller than the ,53% reduction in the volume of the neocortex.
Contrary to the loss of cortical volume in Psen cDKO mice, visual inspection of brain sections suggested an enlargement of the dentate gyrus in the aged Psen cDKO hippocampus (Fig. 3A). Stereological quantification confirmed an age-dependent increase in the neuron number in the dentate gyrus (2 mo: control, 2.  Fig. 3C). By 19 months of age, Psen cDKO mice had 41.5% more neurons in the dentate gyrus of the hippocampus relative to controls, which could be due to an increase in adult neurogenesis in the these brains.
To determine whether Psen cDKO hippocampal neurons share a common death mechanism with Psen cDKO cortical neurons, we analyzed brain sections with Fluoro-Jade B, TUNEL labeling, and active caspases. In contrast to the sizeable increase in the number of degenerating neurons in the neocortex of Psen cDKO mice, a smaller increase is observed in the hippocampus (Fig. 3D-G). Quantitative analysis of Fluoro-Jade B staining revealed no increases in degenerating neurons in the Psen cDKO hippocampus at 2 months (p.0.05; Fig. 3D), but a 2.3-fold increase in degenerating neurons at 4 months (p,0.002; Fig. 3D). Similar to the neocortex, no significant increase in TUNEL-positive cells was observed in the hippocampus of Psen cDKO mice at 6 weeks (p. 0.05; Fig. 3E), but more TUNEL-positive cells were found in the Psen cDKO hippocampus by 2 months (1.6-fold increase, p,0.01; Fig. 3E) and 4 months (2.8fold increase, p,0.001; Fig. 3E). Consistent with these results, in the Psen cDKO hippocampus we observed only mild increases in the number of cells positive for active caspase-9 (Fig. 3F) and active caspase-3 (Fig. 3G). Combined, these data describe a milder apoptotic cell death phenotype in the hippocampus of Psen cDKO mice as compared to the neocortex.
Neurons in the lateral cortex are more susceptible to loss of presenilin function To determine whether excitatory neurons of the cerebral cortex lacking presenilins exhibit region-specific selective vulnerability, we examined the distribution of degenerating and apoptotic neurons in a series of sagittal brain sections. As shown in Figure 4, the number of Fluoro-Jade B-or TUNEL-positive cells per section in 3 medial (M) or lateral (L) sections was averaged and compared between the genotypic groups. Striking increases were observed in the number of degenerating neurons in the lateral area of the Psen cDKO cerebral cortex, comprised of somatosensory and visual cortices, at 2 and 4 months of age ( Fig. 4 and Table 2). An average of 20.5-fold increase in Fluoro-Jade B-positive cells was found in the lateral sections of the Psen cDKO cortex at 2 months (cDKO: 34.163.1 per section; control: 1.760.7 per section; Fig. 4B). In the medial area of the Psen cDKO cerebral cortex, which consists of motor, anterior cingulate, and retrosplenial cortices, no significant increase in Fluoro-Jade B-positive cells was found at 2 months of age (cDKO: 3.460.3 per section; control: 2.560.3; p.0.05; Fig. 4B and Table 2). At 4 months of age, Psen cDKO mice showed increases (22.6-fold) in Fluoro-Jade B-positive cells in the lateral cortex similar to Psen cDKO mice at 2 months, but in contrast to 2 months, the medial cortex of Psen cDKO mice at 4 months also showed increased numbers of FJB+ cells (5.5-fold increase; Fig. 4D and Table 2). We further performed the TUNEL assay to identify apoptotic cells and immunohistochemical analysis for active caspases 3 and 9. We found a 16.1-fold increase in TUNEL+ cells in the lateral cortex of Psen cDKO mice at 2 months (cDKO: 54.868.0 per section; control: 3.460.4; p,0.03), but the increase in TUNEL+ cells in the medial cortex was not statistically significant (cDKO: 4.960.5; control: 3.060.3; 1.6-fold increase; p = 0.05; Fig. 4C). Similarly, at 4 months, a 14.3-fold increase was observed in the lateral cortex of Psen cDKO mice (cDKO: 32.964.1; control: 2.360.5; p,0.05), but the increase in the medial cortex was not statistically significant (cDKO: 11.962.3; control: 1.360.3; 5.2-fold increase; p = 0.05; Fig. 4E). Similar results were obtained for active caspases (Table 2). These results suggest that compared to medial cortical neurons, lateral cortical neurons are more susceptible to loss of presenilins and are more likely to undergo apoptotic cell death in the absence of presenilins.

Mitochondrial defects in the Psen cDKO cortex
Since Presenilins are present in the mitochondrial/lysosomal fraction of the brain [22], and normal mitochondrial function is essential to neuronal viability, we examined the consequence of presenilin inactivation in mitochondria. We quantified the density of mitochondria in the neocortex at 2 and 6 months of age using electron microscope (EM) images (Fig. 5A). Quantitative analysis revealed that numbers of mitochondria per 100 mm 2 in the Psen cDKO neocortex are normal at 2 months (control: 62.564.1; cDKO: 68.464.1; p.0.05) but reduced at 6 months (control: 61.764.8; cDKO: 44.264.6; p,0.001; Fig. 5A). More detailed analysis of mitochondrial size, or quantification of the area of each mitochondrion and assignment into arbitrary bins of 0.1 mm 2 (Fig. 5B, C), revealed normal size distribution of Psen cDKO mitochondria at 2 months (Fig. 5B), but decreased numbers of smaller mitochondria (,0.2 mm 2 ) in the neocortex of Psen cDKO mice at 6 months of age (Fig. 5C). In addition to the decrease in smaller mitochondria in the Psen cDKO neocortex, we observed an increase in larger mitochondria (Fig. 5D). Quantitative analysis of mitochondria revealed a higher percentage of larger mitochondria (.0.3 mm 2 ) in the Psen cDKO neocortex (9.8%63.6), compared to controls (4.7%61.8; p,0.05; Fig. 5D) at 6 months of age. Combined, the reduced mitochondrial density and altered size distribution in the aging Psen cDKO cortex may contribute to neuronal degeneration observed in these mice.

Loss of presenilins causes a delayed onset of apoptotic cell death in the cerebral cortex
In Psen cDKO mice, Presenilin inactivation is dependent on expression of Cre recombinase under the control of the aCaMKII promoter, Cre-mediated recombination of the floxed Psen1, and turnover of Psen1 mRNAs and proteins. Our in situ hybridization and western analyses have shown that expression of Cre mRNAs and loss of Psen1 mRNAs and proteins begins at 3 weeks of age in the cerebral cortex ( [14]; also MWS and JS, unpublished data). A striking feature of Psen cDKO mice is the delayed onset of apoptosis and degeneration relative to the timing of presenilin inactivation. As opposed to mouse models of hypoxia-ischemia, or hypoglycemia, or exposure to excitotoxins (e.g. glutamate) [23,24,25] or other neurotoxic compounds such as ethanol or NMDA antagonists [26,27,28,29], where the brain has a rapid, dramatic increase in dying cells, we failed to detect any increases in TUNEL+ cells in the cerebral cortex of Psen cDKO mice at 6 weeks of age (Figure 2), despite the presence of presynaptic and postsynaptic defects at this age (D. Zhang and JS, unpublished results). By 2 months of age, 5 weeks after the onset of presenilin inactivation, significant increases in apoptotic and degenerating neurons were found in the cerebral cortex of Psen cDKO mice (Figures 1 and 2). However, no significant reduction of cortical volume, which reflects the loss of neuronal processes as well as neuronal numbers, and cortical neuron number was found at 2 months of age, because of the low absolute number of cells that are undergoing apoptosis at this age (see below).

Apoptotic cell death occurs in a very small percentage of cortical neurons lacking presenilins
Cortical neurons lacking presenilins degenerate at a slow steady rate of approximately 0.1% of total neurons, measured by TUNEL+ and Fluoro-Jade B+, beginning at 2 months of age (Table 1). This rate corresponds to the percentage of cells undergoing apoptotic death at the time of the detection, since the time course of apoptosis is very short and dying cells are usually removed within 12-24 hours [30,31]. Interestingly, this low rate of cortical neurons undergoing apoptosis we have identified in Psen cDKO mice is similar to the observed rate of apoptotic cell death (0.02-0.09%) in AD brains [30,32,33,34]. Progressive loss of neurons and neuronal processes in the cerebral cortex of Psen cDKO mice Although only a very small percentage of cortical neurons lacking presenilins would undergo apoptotic cell death, the number of cells lost accumulates over time, leading to significant loss of cortical neurons in the aging cerebral cortex. For example, at 2 months of age, dramatic increases in apoptotic cells were detected in Psen cDKO mice, compared to controls, but no significant loss of cortical volume and neurons was found (Figure 1). However, by 4 months of age, significant loss of cortical volume (,22%) and neurons (,9%) was present (Figure 1). This degeneration process continues, leading to loss of ,18% and 24% of cortical neurons at 6 and 9 months of age, respectively ( Table 1). This slowly progressive nature of cortical neuronal degeneration observed in Psen cDKO mice further highlights the relevance of this model to the study of similarly slow, progressive neuronal degeneration in AD.
The delay of onset in cell death and the low rate of apoptotic cell death among cortical neurons lacking presenilins suggest that loss of presenilins may sensitize cells to die rather than immediately triggering apoptosis. While the identity of the crucial death trigger is unclear, the apoptotic cell death could be induced by a disruption of cellular homeostasis caused by loss of presenilins, which could in turn affect the survival of cortical neurons to varying degrees.
Region specificity of vulnerability of cortical neurons to loss of presenilins A common characteristic of Alzheimer's disease and other neurodegenerative diseases such as Parkinson's and Huntington's is the region specificity of neuronal degeneration in the brain. In the case of AD patients, neuronal loss is most prominent in the hippocampus, the temporal cortex (lateral), and the frontal cortex. Recent functional brain imaging (MRI) has found that the earliest signs of altered brain activity in families with AD are decreased activity in the cingulate cortex and decreased connectivity between the hippocampus and cingulate cortex [35,36]. Interestingly, the lateral cortex of Psen cDKO mice, which includes visual and cingulate cortices, indeed shows the earliest and most striking neuronal death ( Figure 4). However, while loss of presenilins causes hippocampal dependent learning and memory deficits as well as impaired synaptic functions in hippocampal neurons [10,14,22], the number of hippocampal neurons was not significantly decreased (CA Saura, MWS & JS, unpublished data). Increases in apoptotic cell death were detected in the hippocampus of Psen cDKO mice at 2 and 4 months of age (Figure 3), but the increase was less dramatic than that in the neocortex of Psen cDKO mice (Figures 1 and 2). It is not entirely clear why the total number of pyramidal neurons in the hippocampus of Psen cDKO mice is not significant reduced, though loss of dendritic complexities [10] and loss of hippocampal volumes ( Figure 3) were seen in aged Psen cDKO mice. One likely possibility is increased adult neurogenesis occurring in the hippocampus, which would compensate for the age-related loss of hippocampal neurons in the absence of presenilins. The fact that we indeed found a 41% increase in the number of neurons in the dentate gyrus of Psen cDKO mice at 19 months, the site of adult neurogenesis, lends further support to this interpretation. The exact mechanism underlying increased neurogenesis in the aging dentate gyrus of Psen cDKO mice is unknown. Since presenilin inactivation is targeted specifically to postmitotic excitatory neurons, rather than neural stem cells, this increase in neurogenesis is likely to be secondary to functional defects and neurodegenerative processes occurring in the hippocampus of Psen cDKO mice. Furthermore, loss of presenilin in neural progenitor cells during development results in decreases, rather than increases, in the number of neurons generated, due to premature depletion of neural progenitor pool in the absence of presenilins [5,6], further arguing against a cell autonomous direct effect.
In summary, we found that loss of presenilins in mature excitatory neurons in the mouse cerebral cortex results in increases in apopototic cell death among a very small percentage of cortical neurons at any given time. Over time, this low rate of cell death leads to significant loss of cortical neurons in an age dependent and region specific manner, reminiscent of the slow and progressive neurodegeneration process in AD. Since mutations in presenilins likely cause AD at least in part via a partial loss of function mechanism [37], this mouse model provides a unique experimental tool to study further the molecular mechanisms underlying neurodegeneration caused by loss of presenilin function and in FAD.