Cognitive Deficits and Disruption of Neurogenesis in a Mouse Model of Apolipoprotein E4 Domain Interaction*

Background: Domain interaction may be the principal pathogenic feature of apoE4 in Alzheimer disease. Results: Young and old mice with mutations causing apoE4 domain interaction showed impaired cognition and a disruption in neurogenesis. Conclusion: Domain interaction mediates apoE4 neuro-pathologies and cognitive phenotype. Significance: Domain interaction is a viable prophylactic target against apoE4 cognitive phenotype and increased susceptibility to Alzheimer disease. Apolipoprotein E4 (apoE4) allele is the major genetic risk factor for sporadic Alzheimer disease (AD) due to the higher prevalence and earlier onset of AD in apoE4 carriers. Accumulating data suggest that the interaction between the N- and the C-terminal domains in the protein may be the main pathologic feature of apoE4. To test this hypothesis, we used Arg-61 mice, a model of apoE4 domain interaction, by introducing the domain interaction feature of human apoE4 into native mouse apoE. We carried out hippocampus-dependent learning and memory tests and related cellular and molecular assays on 12- and 3-month-old Arg-61 and age-matched background C57BL/6J mice. Learning and memory task performance were impaired in Arg-61 mice at both old and young ages compared with C57BL/6J mice. Surprisingly, young Arg-61 mice had more mitotic doublecortin-positive cells in the subgranular zone; mRNA levels of brain-derived neurotrophic factor (BDNF) and TrkB were also higher in 3-month-old Arg-61 hippocampus compared with C57BL/6J mice. These early-age neurotrophic and neurogenic (proliferative) effects in the Arg-61 mouse may be an inadequate compensatory but eventually detrimental attempt by the system to “repair” itself. This is supported by the higher cleaved caspase-3 levels in the young animals that not only persisted, but increased in old age, and the lower levels of doublecortin at old age in the hippocampus of Arg-61 mice. These results are consistent with human apoE4-dependent cognitive and neuro-pathologic changes, supporting the principal role of domain interaction in the pathologic effect of apoE4. Domain interaction is, therefore, a viable therapeutic/prophylactic target for cognitive impairment and AD in apoE4 subjects.

Apolipoprotein E (apoE) 2 is a 34-kDa protein with 299 amino acids expressed in various organs including the brain, where the apoE level is very high, second only to levels in the liver (1). ApoE performs critical functions including cholesterol metabolism, transport, and redistribution of lipids, synaptic remodeling, and inflammatory responses among many others (2)(3)(4). ApoE exists as three isoforms referred to as apoE2, apoE3, and apoE4 in humans. ApoE3 is considered the "normal/wild type" isoform based on its prevalence in the general population (78%) as the most common followed by apoE4 (14%) and the least prevalent, apoE2 (8%), which is considered protective in AD (5,6). The risk of developing AD in relation to apoE genotype is in the order apoE4 Ͼ apoE3 Ͼ apoE2. The risk associated with apoE4 is gene dose-dependent as the frequency of AD in apoE4 homozygous subjects is 91% versus 47% and 20% in apoE4 heterozygous and non-apoE4 subjects, respectively. The age of onset of AD (68 years) in apoE4 homozygous subjects is at least 8 and 16 years earlier than in apoE4 heterozygous and non-apoE4 subjects, respectively (7,8). In the same vein, healthy apoE4 carriers exhibit some cognitive deficits, and their age-dependent cognitive decline is more rapid compared with non-apoE4 subjects (9 -12).
ApoE isoforms differ only by amino acids in two positions (112 and 158) of the protein. Although apoE3 has cysteine and arginine in positions 112 and 158, respectively, apoE2 has Cys and apoE4 has Arg in both positions (1). It has been proposed that the numerous pathologies associated with apoE4 are due to its biophysical structure, the most important of which is domain interaction. Domain interaction in the apoE4 protein is a result of the attraction between the Arg-61 in the N-terminal receptor binding domain and glutamine 255 in the C-terminal lipid binding domain. ApoE2 and ApoE3 lack this domain interaction (14,15). This interaction impairs apoE4 function and apoE4 levels as it is less stable than apoE2 and apoE3 and is targeted for degradation (16). The mouse apoE has an equivalent of Glu-255 but lacks the corresponding human Arg-61 amino acid and, therefore, behaves like apoE3. To eliminate the potentially confounding effect of a foreign gene in the different apoE4 mice available, the native mouse apoE was rendered apoE4-like in terms of domain interaction using site-directed mutagenesis. The threonine codon in position 61 was replaced with an arginine codon, thus creating the apoE4-like domain interaction found in human apoE4 (17). Therefore, features of the Arg-61 mice that are different from the C57BL/6J background mice are as a result of the domain interaction in the wild type mouse apoE.
Many mouse models show human AD-like pathologies and cognitive deficits only at older ages. Arg-61 mice also showed cognitive deficits at 12 months (18), but no amyloidosis has been reported. Given that healthy apoE4 carriers show cognitive deficits and AD starts early in them, we hypothesize that the (negative) effect of apoE4 on cognition may be independent of old age or AD-like hallmark pathologies and that domain interaction may be sufficient to cause this age-independent, apoE4-induced cognitive deficits. Therefore, we carried out hippocampus-dependent learning and memory tests on 12and 3-month-old Arg-61 and C57BL/6J mice, and we explored cellular and molecular mechanisms that may explain these deficits.

MATERIALS AND METHODS
Mice-Arg-61 mice were obtained from the Gladstone Institute courtesy of Drs. Weisgraber and Raffai. The colony was maintained in University Mississippi Medical Center. Control (wild type) C57BL/6J mice were purchased from Taconic Labs (Hudson, NY). All mice were maintained in a temperature-controlled environment with free access to food and water and kept on a 12-h light/dark cycle from 7 a.m. to 7 p.m. each day. The experiments were performed when the animals were either 3 or 12 months old. All animal procedures were in compliance with University of Mississippi Medical Center Institutional guidelines and were approved by the UMMC Animal Care and Use Committees (protocol #1155A).
Radial-arm Water Maze (RAWM)-We studied spatial memory and flexibility in learning using the RAWM protocol that includes a reversal learning component. The procedure was performed according to Alamed et al. (19), with some modifications as described herein for the younger mice. The apparatus consisted of a 30-cm-deep tank filled up to 16.5 cm with water made opaque with white-tempera paint. The tank was divided into 5 alleys by 28-cm high plastic dividers as shown in Fig. 1a. We used a five-alley tank to avoid symmetry because we observed that the animals tend to swim straight ahead and may, therefore, inadvertently enter the right or wrong arm (depending on the position of the platform relative to the start arm). The five-arm maze ensures that the animal makes at least a necessary angular turn to enter an arm. The tank also included a 16-cm high, 10-ϫ 10-cm platform that can be moved into any of the arms. Whenever the platform was used, it was placed deep into the arm, 5 cm away from the wall. The mouse was gently placed in one of the four arms (apart from the one in which the platform was placed) and allowed to find the platform within a maximum time of 60 s or gently guided to the platform. The start location for each animal was different for each trial, but location of the platform was the same for each cohort of five animals of which all groups were represented. A total of 10 training trials were done per day for each animal. This consisted of five blocks of 2 consecutive trials with ϳ10 min inter-block intervals. This was done for two consecutive days. In all trials the numbers of wrong arm entries or failure to find the target in the correct arm (errors) and the time to find the platform were recorded. On the third day, the test ended with a "reversal learning" variant in the old mice where the animals were trained with a new platform location. On the other hand, the test continued on days 3 and 5 with "probe trials" for the young mice during which the animals were allowed to explore the tank with the platform removed. This was recorded with a digital camera and analyzed offline with Noldus Ethovision XT 8.5 (Noldus Information Technology, Wageningen, The Netherlands). The data obtained were used to calculate the percentage correct arm entries and percentage of time in the correct arm. This percentage was compared with a 20% chance level by a one-sample t test. The schedules of the probe trial were as follows: day 1, 1 h after last training trial on day 1; day 2, 1 h after last training trial on day 2; day 3, 24 h after last training trial on day 2; day 5, 72 h after last training trial on day 2.
Novel Arm Discrimination-To study hippocampus-dependent spatial memory, we used the novel arm discrimination (spatial recognition memory) task. This test is based on the innate preference of rodents to explore a novel environment more than a familiar one (20 -22). The apparatus is a Y-Maze with three identical arms that are 10 cm wide, 40 cm long, and evenly separated at 120 degrees from each other. Different objects visible to the mice from within were placed at various distances around the maze to serve as extra-maze cues. The test consisted of two trials; in the first (acquisition) trial, one arm was blocked, whereas the animal was placed into one of the two remaining arms of the maze. The mouse was then allowed to explore both the "start arm" (S) and the "other arm" (O) for 5 min. One hour later, the blocked arm was opened ("novel arm" (N)), and the animal was allowed to explore again for 3 min (retention trial). A different set of durations was used for the older mice to make the test sensitive to the deficits: 10-min acquisition trial and 4-h inter-trial interval. The floor of the maze was covered with cage bedding that was mixed after each animal to avoid odor cues, and the walls were wiped down with 70% alcohol. Apart from the start arm that was kept constant for all animals, each of the two remaining arms were used evenly as novel arms among the animals. The experiment was recorded with a Panasonic digital camera, and the video was analyzed offline using the Noldus EthoVision XT 8.5 software. The percentage time spent and/or the percentage number of entries into N and O were calculated as follows: %N ϭ (100% ϫ N)/(N ϩ O) and %O ϭ (100% ϫ O)/(N ϩ O), respectively. %N and %O were compared with 50% chance level using a onesample t test to determine whether there was any discrimination/preference for the novel arm over the other (familiar) arm.
Spontaneous Alternation in Y-maze-We studied working memory using the spontaneous alternation task in the y-maze apparatus. The floor of the y-maze apparatus was covered with sawdust/cage bedding that was mixed after each animal to eliminate odor cues. Animals were allowed 5 min to explore the maze during which their spontaneous alternations were recorded for later offline extraction using the Microsoft Excelbased Zone transition tool kit that was supplied with the Noldus Ethovision XT 8.5 software. The measure of the animal working memory is the percentage of the possible complete triplets made during the set time (triplets/(total arm entries minus 2) ϫ 100) (23,24). For example, in a sequence such as ABCACBCABA the triplets that start with underlined letters are complete alternations, whereas the others are not. The animal scores (5/(10 Ϫ 2) ϫ 100) ϭ 62.5%. We compared the alternation percentage of each genotype (Arg-61 and C57Bl/6J) to 50% chance level using the one-sample t test (25).
Tissue Harvesting-Animals were anesthetized with isoflurane and transcardially perfused with cold normal saline. Brains were quickly removed from the cranium and then divided into the left and right half along the longitudinal fissure. The hippocampi from the right halves of the brains were rapidly dissected on ice, wrapped in pre-labeled aluminum foils, snapfrozen in liquid nitrogen, and stored at Ϫ80°C until ready for RNA and protein extraction. The left halves were kept in 4% paraformaldehyde at 4°C for 16 h after which they were stored in PBS containing 0.09% sodium azide until ready for sectioning.
Protein and RNA Extraction-Protein and RNA were extracted from the same hippocampus sample as previously described (26) with modifications. Briefly, samples were homogenized at 4°C in 270 l of cold DEPC water. 180 l of the homogenate was immediately transferred into a tube already containing 180 l of RIPA-Halt protease inhibitor mixture mix (1:100) (Halt was from Thermo-scientific; Rockford, IL). Denaturing solution (450 l) from the Totally RNA kit (Life Technologies) was added to the remaining homogenate and set apart for mRNA extraction. The homogenate-RIPA mix was centrifuged at 11,000 rpm at 4°C in a refrigerated centrifuge for 12 min. The supernatant was transferred into a fresh tube and centrifuged again at 11,000 rpm at 4°C for 15 min. The supernatant was taken out, and aliquots of 100 l were made in 500-l tubes and stored at Ϫ20°C until use. 10 l of the homogenates were diluted 1 in 5 for protein assay by the BCA method. The Totally RNA kit (Life Technologies) was used for RNA extraction according to manufacturer instructions. RNA pellets were eluted in 40 l of elution buffer and quantified with NanoDrop spectrophotometer (Thermoscientific; Wilmington, DE).
SDS-PAGE and Western Blotting-Aliquots were diluted to 2 g/l using RIPA-Halt mix; this was further diluted to 1.5 g/l using 4ϫ Laemmli buffer volume equal to ͓1/3͔ that of the 2 g/l-diluted aliquot of the homogenate. The samples were then boiled for 5 min at 95°C and kept on ice for at least 1 min before being loaded on the (10%) acrylamide gel. 20 -30 g (equal to 13.3-20 l) of protein were loaded in the respective wells along with the precision protein plus Western blot ladder (Bio-Rad). This was run at 120 V for 10 min and then at 90 V until a good resolution of the bands was achieved (ϳ2.5 h). The protein separation on the gel was transferred unto a PVDF membrane using the Turbo Transfer unit (Bio-Rad). The membrane was washed in the washing buffer before incubating with the primary antibody (doublecortin; catalog #4604; 1:500), cleaved caspase-3 (catalog #9661; 1:500), or BDNF (catalog #SC-20981; 1:250) using the Fast Western Kit (Thermoscientific; Wilmington, DE) according to manufacturer's instructions. The membranes were incubated in the ECL solution 1 and 2 (1:1 mix) of the fast Western kit HRP-linked secondary antibody. The membranes were placed in a thin transparent plastic film viewed and imaged with the optimum exposure time in a Chemidoc XRSϩ imager with the Image Lab 3.0 software (Bio-Rad). The membranes were stripped with the OneMinute stripping buffer (Genescript; Piscataway, NJ), and ␤-actin bands were detected as above using ␤-actin monoclonal antibody (Abcam, catalog #ab6276, 1:10,000). Relative protein expression of each lane was obtained by dividing the adjusted optical density of the protein by the adjusted optical density of the actin band of the respective lane (relative optical density).
Reverse Transcription and Real Time Polymerase Chain Reaction-RNA (4 g each) was reverse-transcribed with the Superscript III First strand kit (Invitrogen) according to manufacturer's instructions. Two l of the cDNA sample was used for the real-time quantitative PCR using (gene-specific) forward and reverse primers (Table 1) and RT 2 SYBR Green mastermix (Qiagen Sciences) according to the manufacturer's instructions. The samples were run in an iQ Mastercycler (Bio-Rad) with a program that allows for melting curve analysis to ascertain single product formation. GAPDH run along with the individual genes of interest was used as the housekeeping gene to normalize the gene expression. Relative gene expression was obtained from the threshold cycle (CT) values using the ⌬CT method according to Livak and Schmittgen (27). Primer sequences used were as follows in Table 1 (nci.nih.gov; Ref. 28).
Immunohistochemistry-The left hemibrain was transferred into 30% sucrose and left overnight at 4°C or until the brain sank to the bottom of the solution. The brain was embedded in OCT Embedding Matrix and sectioned coronally at 40-m thickness throughout the rostral-caudal length of the brain using a Leica CM3050 S Cryostat. Sections were collected in 24-well plates containing PBS with 0.99% sodium azide. The sections were stored at 4°C until they were used for immunohistochemical staining. All the 12th sections throughout the

real time PCR primer pairs
Primer sequences were obtained from nci.nih.gov and purchased from invitrogen.

Domain Interaction Is the Principal Pathogenic Feature of ApoE4
rostral-caudal length of the brain were used for each antibody staining. Sections were first washed in PBS and then blocked for 1 h 30 min at room temperature in blocking buffer (1ϫ PBS; 0.3% Triton X-100) containing 5% normal goat serum. The sections were then incubated overnight with gentle shaking at 4 ºC in primary antibody (rabbit anti-mouse Doublecortin; Cell Signaling Biotechnologies) that was diluted in antibody dilution buffer (1ϫ PBS, 1% BSA, 0.3% Triton X-100) at 1:200. On the next day sections were washed three times for 5 min with PBS and then incubated for 60 min in the dark at room temperature in secondary antibody (Goat anti-rabbit FITC; 1:5000 in dilution buffer). Sections were washed three times again as described above but in the dark. Sections were mounted on glass slides with mounting medium containing DAPI. Unbiased Stereology-The stereology was performed as described previously (26). Briefly, montage images of all sections that contain the hippocampus were captured using a 3i microscope with Slidebook 5.5 software (Intelligent Imaging Innovations, Inc, Denver, CO). A region of interest that included the dentate gyrus (granular layer and molecular layer) without the innermost potions of the hilus was selected. Stereology counting frames 100 ϫ 100 m in a 200 ϫ 200-m matrix were placed on the image by a semiautomatic stereology system (Zeis Axiovert 200M fluorescent microscope; 3I digital microscopy). The dissector height was 20 m. Only cell bodies that met the counting criteria according to the optical dissector principle within the counting frame were counted; cells within the counting frame that touch the red (lower and left) boundaries were excluded. The total number of cells counted was multiplied by 68.6 (1/12 for sections analyzed; 1 ⁄ 4 counting area and 70/100 tissue shrink factor (29). In all cases optimum sampling was achieved as the coefficient of error (CE) is lower than the coefficient of variation (30). The doublecortin-positive cells were categorized into the mitotic and post mitotic types as previously described by Plümpe et al. (31). Dendrites arising from doublecortin-positive cells were followed by scrolling through the z axis of the captured three-dimensional images to ascertain their branching and extent of growth in order to accurately categorize them.

Impaired Spatial Recognition Memory and Cognitive Flexibility in 12-Month-old Arg-61
Mice-It has been reported that domain interaction in apoE4 has an effect on spatial memory (18) using the Morris water maze method. We performed a Radial-arm water maze test that involved 2 days of training strictly using the procedure described by Alamed et al. (19) (including the alternation of visible and hidden platforms). This test did not expose any impairment in the Arg-61 mice as there were no significant differences between C57BL/6J and Arg-61 mice throughout the two-day training period. Also, comparing the performance of the animals during the first and the fourth block of training on day 2 using a paired sample t test, both C57BL/6J and Arg-61 mice showed evidence of having learned the target in terms of latency or time-to-target (11.5 Ϯ 1.7 versus 6.2 Ϯ 0.7 s, p ϭ 0.007 and 10.2 Ϯ 0.9 versus 7.4 Ϯ 1.0 s, p ϭ 0.004 for C57BL/6J and Arg-61, respectively) (Fig. 1b). The number of errors made also showed that they had learned the location of the platform (1.8 Ϯ 0.3 versus 0.6 Ϯ 0.1, p ϭ 0.004 and 1.2 Ϯ 0.2 versus 0.6 Ϯ 0.2, p ϭ 0.002 for C57BL/6J and Arg-61 respectively) (Fig. 1c). However, after modifying the protocol to test cognitive flexibility by reversal learning, we found out that right after the first trial, C57BL/6J mice were able to quickly learn that the platform location had changed, whereas the Arg-61 did not (Fig. 1, d-f). With respect to latency and errors, the performance of C57BL/6J mice had become asymptotic from the second trial during which all parameters measured (time-to-target, total errors and old-target errors; Fig. 1, d-f, respectively) were lower than those of Arg-61 mice. Although Arg-61 mice spent significantly more time on the average to find the target (26.0 Ϯ 2.9 versus 18.7 Ϯ 1.8 s; p ϭ 0.0315), they seemed to show signs of learning the new location as their time-to-target decreased with training. These data suggest that domain interaction in the mouse apoE protein is able to cause impairment in certain domains of spatial learning and memory as assessed by the RAWM reversal learning task.
To explore this subtle spatial memory deficit further, we performed a novel arm discrimination test that has been proven to test hippocampus-dependent spatial memory without the stress associated with Morris water maze task (20,22). In our pilot experiments, when we used a 10-min acquisition trial and a 4-h inter-trial interval in the novel arm discrimination test, 12-month-old Arg-61 mice showed an impaired spatial recognition memory (Fig. 1, g and h). Although the C57BL/6J mice entered the novel arm with a frequency that is significantly higher than (50%) chance level (61.6 Ϯ 3.3%; one sample t test p ϭ 0.0117), percent entry into the novel arm was not significantly different from chance level in the Arg-61 mice (50.6 Ϯ 3.9%; p Ͼ 0.05; Fig. 1h). The percentage of time spent in the novel arm was close among the C57BL/6J and Arg-61 mice, but perhaps due to a large intra-group variation, the 58.4 Ϯ 3.9% in Arg-61 mice was not quite significantly different from the 50% chance level (one sample t test p ϭ 0.0767). However, the percentage of the time in the novel arm was significantly different from chance level in the C57BL/6J mice (58.3 Ϯ 3.2%; p ϭ 0.0433; Fig. 1g).
Impaired Spatial Learning and Memory in Young Arg-61 Mice-We made some modifications to the original RAWM protocol (19) in terms of repeated but brief probe trials on days 1, 2, 3, and 5 of the test. In the training phase of the task, using the mean of all the time points/trial as presented in Fig. 2a, Arg-61 mice spent significantly more time on the average to find the target on day 1 during the trials (34.6 Ϯ 4.4 versus 18.3 Ϯ 2.2 s; p ϭ 0.004) even though the number of errors made were not significantly different (Fig. 2b). To understand the reason why Arg-61 mice had a significantly higher time-to-target but not a higher number of errors, we did a series of analysis on the locomotor function of the animals using a full 1-min video recorded during the probe trial. Arg-61 mice were neither (statistically) swimming slowly nor did they stay longer at the center; time moving, distance covered, and latency to enter the center were also not statistically different. However, Arg-61 mice made significantly fewer total arm entries (10.6 Ϯ 1.0 versus 14.2 Ϯ 0.7 for Arg-61 and C57BL/6J mice, respectively; p ϭ 0.010).
On day 2, there was no difference between C57BL/6J and Arg-61 mice in both the time-to-target and the number of errors made during the trials (data not shown). Also, both genotypes discriminated between the target arms and the other arms, as their percentage time-in-target and entry-into-target were significantly greater than (20% chance level; Fig. 2, c and  d). These suggest that although C57BL/6J mice were able to learn the target location within the first day, Arg-61 mice needed much longer training to acquire it, and they did on the second day. Probe trials on the third and fifth days, however, revealed impairment in the Arg-61 mice as they "lost" the preference for the target, whereas the C57BL/6J mice still retained it. Percentage time-in-target and percentage entry-into-target were significantly higher than (20%) chance level in C57BL/6J mice but not in Arg-61 mice on both days (Fig. 2, c and d; days 3 and 5). This suggests that long term spatial memory may be impaired in the young Arg-61 mice. Taken together with the data in older animals, it suggests that domain interaction in apoE protein not only causes impairment in older animals but may actually work independent of age as significant impairment were found also in young animals.
Spatial Recognition and Working Memory Impairments in Young Female Mice-Because there is major evidence for increased susceptibility to AD in females and the possibility that females may be more sensitive to the detrimental effects of apoE4 (32), we further analyzed the RAWM data separately for males and females. The analysis shows that the main effect may be with females as there was a significant difference in the timeto-target among females (Arg-61 versus C57BL/6J) but not among males (35.5 Ϯ 7.0 versus 15.2 Ϯ 2.3 s, p ϭ 0.025 and 33.7 Ϯ 6.1 versus 21.5 Ϯ 3.5 s, p ϭ 0.120 for females and males, respectively (Fig. 3a).
We saw a hint of a gender-dependent effect in spatial memory impairment in the Arg-61 mice (Fig. 3a), so we further tested this with other spatial memory tests. Spontaneous alternation in Y-maze shows that although all other groups (C57BL/6J males, C57BL/6J females, and Arg-61 males) performed above (50%) chance level (69.6 Ϯ 5.5%, p ϭ 0.028; 61.3 Ϯ 3.1%, p ϭ 0.022; 72.0 Ϯ 4.0%, p ϭ 0.005, respectively),  t test). There was no significant difference between C57BL/6J and Arg-61 mice in both measures during both of the trial blocks. Shown are time to locate the target (d), total errors or wrong arm entries made (e), and number of entries made into the previous target arm during the reversal learning (a test of cognitive flexibility) when the target with which they were trained for 2 days was changed to another one (f). Shown are percent time (g) and percent of total entries (h) in the Novel arm during a place memory/novel arm discrimination test. During the retention trial, mice were allowed to explore the three arms of a Y-maze that they had explored 4 h earlier during a 10-min acquisition trial when one of the arms (The Novel arm) was blocked. C57BL/6J mice discriminated between the novel and familiar arms, as the percent time and entry in the novel arm were significantly higher than the 50% chance level (dotted) line. Although the percent duration in the novel arm was not significantly different (p Ͼ 0.05), the percent entry into the novel arm was significantly higher in C57BL/6J mice than the Arg-61 mice (p ϭ 0.0498; §). For a and b, n ϭ 7 for both groups. Data shown are the mean Ϯ S.E. One sample t test (versus 50%). *, p Ͻ 0.05; **, p Ͻ 0.01. For c-e, n for C57BL/6J was 18; n for Arg-61 was 14. Independent sample t test (C57BL/6J versus Arg-61). *, p Ͻ 0.05; **, p Ͻ 0.01.

Increased Total and Mitotic Doublecortin-positive Cells in the Hippocampal Subgranular Zone of 3-Month Arg-61 Mice-
Doublecortin-positive cells were found in the neurogenic regions of the brain, namely the subventricular zone along the walls of the lateral ventricles and the subgranular zone of the hippocampus where they were found lining the junction of the hilus and granular layer. Some cells were found within the granular cell layer where they are located mostly closer to the hilus than the molecular layer. The cells were categorized based on their stage of development as either mitotic (early and intermediate, Fig. 4, f-g) or post-mitotic (late, Fig. 4e) types based on the extent and morphology of their dendritic outgrowths as previously described by Plümpe et al. (31) and Paus et al. (33). Thus, DCX-positive cells with either no processes or with short or medium processes in the granular cell layer were categorized as mitotic. The other DCX-positive cells with rather long dendrites, which branches within the granular cell layer and molecular layer, were categorized as post-mitotic.
The stereology estimation shows a significantly higher total of doublecortin-positive cells in the subgranular zone of the dentate gyrus of Arg-61 mice compared with C57BL/6J mice (3308.6 Ϯ 367.5, CE ϭ 0.111 versus 2142.9 Ϯ 238.2, CE ϭ 0.111; p value ϭ 0.0374), a 54.4% increase compared with C57BL/6J (Fig. 4a). When the data were further analyzed based on categorizing the cell types by mitotic or post-mitotic properties as described earlier, Arg-61 mice showed a significantly higher number of cells than C57BL/6J mice in the mitotic categories (54.8% higher than C57BL/6J) but not in the post mitotic category (mitotic: 2760.0 Ϯ 283.2, CE ϭ 0.103 versus 1782.9 Ϯ 160.8, CE ϭ 0.090; p ϭ 0.024; post-mitotic: 548.6 Ϯ 108.4, CE ϭ 0.198 versus 360.0 Ϯ 81.0, CE ϭ 0.225; p ϭ 0.213) (Fig. 4b). The percentage of mitotic cells and postmitotic cells as well as the ratio of mitotic to post-mitotic cells was not significantly different between Arg-61 and C57BL/6J (data not shown). These results suggest that there is an increased proliferative activity in the dentate gyrus of the Arg-61 mice hippocampus compared with the C57BL/6J mice hippocampus.

FIGURE 2. Impaired spatial learning and memory in 3-month-old Arg-61 mice.
Shown are latency/time-to-target (a) and total number of errors/wrong arm entries (b) made on day 1 of the radial-arm water maze test. No significant difference in number of errors but latency was significantly higher in Arg-61. Analysis of the total distance swum and swimming speed in the subsequent probe trial shows no significant difference in both locomotor measures (data not shown; see "Results"). Independent sample t test (C57BL/6J versus Arg-61). *, p Ͻ 0.05; **, p Ͻ 0.01. n for both groups was 10 (5 males, 5 females). Training blocks were two consecutive trials with two different start arms for each mouse. Shown are percent of time (c) and percent of entry in the target arm (d) during a series of probe trials in the radial arm water maze when the target was removed. Data shown are the mean Ϯ S.E. One-sample t test versus 20% chance (dotted line) level. *, p Ͻ 0.05; **, p Ͻ 0.01. n for both groups is 10 (5 males and 5 females). No significant difference between C57BL/6J and Arg-61 mice (independent t test p Ͼ 0.05) was found on any of the days in either percent duration (c) or frequency (d) in target. However, there was a trend to a higher percent time in target c in C57BL/6J mice on day 5 (p ϭ 0.0693). Day 1 ϭ 1 h after the last trial on day 1; day 2 ϭ 1 h after the last trial on day 2; day 3 ϭ 24 h after the last trial on day 2; day 5 ϭ 72 h after last trial on day 2.

Increased BDNF and TrkB Receptor mRNA Expression Young but Not Old Arg-61 Mice and Increased Cleaved Caspase-3 Protein in Both Young and Old Arg-61
Mice-BDNF, which is highly expressed in the hippocampus, plays important roles in regulating neurogenesis through its TrkB receptor signaling (34,35). We used RT-quantitative PCR to measure BNDF and TrkB receptor mRNA expression level in the hippocampus. Our data show a significantly higher expression of BDNF (41.0% higher; p ϭ 0.0318) and its receptor TrkB (21.3% higher; p ϭ 0.0291) in 3-month-old Arg-61 mice hippocampi compared with those of the C57BL/6J mice (Fig. 5, a and b). This suggests that the increase BDNF and TrkB receptor expression may act to facilitate the increase in cell proliferation in the Arg-61 mouse hippocampus. However, there were no significant differences in the BDNF protein level between the young C57BL/6J and Arg-61 mice (Fig. 5e). In contrast to the mRNA increases found in the young Arg-61 mice, there were no significant differences in BDNF and TrkB mRNA (p Ͼ 0.05) (Fig. 5, c  and d), neither was there a significant difference in the BDNF protein levels between old C57BL/6J and Arg-61 mice (p Ͼ 0.05) (Fig. 5f).
We also used Western blot to measure cleaved caspase-3 protein expression in the hippocampus of the 3-month-old mice. We found almost a 2-fold (87%) increase in expression of cleaved caspase-3 in the hippocampus of the young Arg-61 mice (p value ϭ 0.006; Fig. 5g). Similarly, we found an even greater level of increase in cleaved caspase-3 in the 12-monthold Arg-61 mice versus C57BL/6J mice. Arg-61 mice had almost 3 times (172.4% higher) the level of cleaved caspases-3 in C57BL/6J mice (Fig. 5h). These suggest that the apoptotic activity in young Arg-61 mice persists and may exacerbate with age.
Decreased Doublecortin Expression in Hippocampus of Old Arg-61 Mice-Unlike in young animals, doublecortin-positive cells are very sparse in the hippocampus of older mice (36), and this creates a challenge in stereological quantification. Therefore, we analyzed doublecortin expression by Western blotting in the hippocampus of 3-and 12-month-old C57BL/6J and Arg-61 mice. This method was found to be sensitive in characterizing age-dependent changes in neurogenesis in certain mouse models of Alzheimer disease (37) (38), in dogs (39), and in human (40). Therefore, levels of doublecortin immunoreactivity may give an indication of the level of neuronal proliferation in mice. Results show a significant age effect in doublecortin expression (F(1,18) ϭ 22.34; p ϭ 0.0002; interaction effect F(1,18) ϭ 6.584; p ϭ 0.0194) that seemed more pronounced in the Arg-61 mice than the C57BL/6J mice (65.3% reduction versus 22.2% reduction in Arg-61 and C57BL/6J, respectively) (Fig.  6). There was no significant difference in doublecortin expression in the young Arg-61 and C57BL/6J mice, but in the older mice, Arg-61 mice had a 48.8% lower expression compared with FIGURE 3. Impairment in young mice is gender-dependent; females more affected. a, average time-to-target over the five training blocks among 3-monthold male and female C57BL6J and Arg-61 mice. There was no significant difference among males, but average time to target was significantly different among females p ϭ 0.025. b, spontaneous alternation percentage score for 3-month-old male and female mice. All groups performed beyond chance level (50%; dotted line) except the female Arg-61 mice. Two-way analysis of variance shows a significant gender effect (F(1,16) ϭ 6.070; §, p ϭ 0.0255); although both t test and post hoc test did not reveal a significant male versus female difference in either genotype (p Ͼ 0.05), the t test showed a trend to a significantly lower alternation percentage in female Arg-61 mice versus males (p ϭ 0.0591) but not in C57BL/6J mice (p ϭ 0.226). Shown is percent of time (c) and percent of total entries (d) in the Novel arm during a place memory/Novel arm discrimination test. During the retention trial mice were allowed to explore the three arms of a Y-maze, which they had explored 1 h earlier during a 5-min acquisition trial when one of the arms (novel arm) was blocked. Female C57BL/6J mice discriminated between the novel and familiar arms, as the percent time and entry in novel arm were significantly higher than 50% chance level (dotted) line, whereas female Arg-61 mice did not; the percent time in novel arm was not significantly different between C57BL/6J and Arg-61 (p Ͼ 0.05), but there was a trend to a higher percent entry into novel arm in C57BL/6J mice versus Arg-61 mice (p ϭ 0.0524). Data shown are the mean Ϯ S.E. One-sample t test versus 50% chance (dotted line) level. *, p Ͻ 0.05; **, p Ͻ 0.01. n for all groups is 5.
C57BL/6J mice (p ϭ 0.002). Taken together with the stereology results, these data suggest that neuronal cell proliferation in Arg-61 mice is not sustained later in life, during which it may actually be reduced.

DISCUSSION
In this study we demonstrated that Arg-61 mice showed spatial learning and memory impairments at both old and younger age. We observed an abnormal increase of DCX-positive subgranular progenitor cells in 3-month-old Arg-61 mice, whereas the DCX protein expression was not increased at 3 months of age and was even highly reduced at 12 months old. These changes at 3 months of age seemed to be an inadequate compensatory mechanism by the Arg-61 mice brain to "repair" itself. This is because the increased proliferation in young mice was not sustained until adulthood and may even be counterbalanced in the young mice by increased apoptosis of these excess cells. More importantly, these cellular and molecular compensatory attempts did not prevent cognitive deficits either earlier or later in life. We have thus shown here that domain interaction in the apoE4 protein may be sufficient to precipitate a cognitive phenotype even at younger ages.
Cognitive Deficits in Older Arg-61 Mice-The most important clinical feature that negatively impacts the quality of life of Alzheimer disease patient is the learning and memory deficits. Except for mice carrying multiple AD-related mutations (which rarely occur together in a single individual), most animal models of AD that have been developed show cognitive deficits only at older ages. This is consistent with the situation in humans as old age is the most important risk factor for developing Alzheimer disease. Several studies have also demonstrated impairments in spatial learning and memory in various types of apoE4 mice at 12 months of age and older (41) (15-18 months) (42). Therefore, it is not quite surprising that our 12-month old Arg-61 mice displayed some level of cognitive deficits as previously reported by Zhong et al. (18). This cognitive deficit has been attributed, among other factors, to certain pathologies associated with early stage AD in human subjects, namely synaptic degeneration (43). This may be as a result of astrocyte dysfunctions in the Arg-61 mouse (18). However, it is worth pointing out that no deficit was found in 12-month-old Arg-61 mice during the training phase. Deficits were seen only after the task had been made more challenging during the reversal learning, a test of cognitive flexibility (44). This is in contrast to detectable deficits on the first day training in 3-month-old mice. As further discussed below, although, apoE4-like domain interaction negatively impact learning and memory in old mice, it seemed that the effect of apoE4 on cognition wane with age as concluded in several human studies (45)(46)(47). This does not contradict the fact that age and apoE4 gene are the top two risk factors for AD and, therefore, should interact to amplify the risk; it only means that even though the risk of developing AD increases with age, the presence of apoE4 increases the risk more at younger ages than it does at older ages.
Cognitive Deficits in Younger Arg-61 Mice and Gender Effect-Cognitive deficits in AD mouse models are usually observed in mice carrying multiple AD-related mutations; for example, the 3xTgAD and the 5xFAD mice. These animals show learning and memory deficits as early as 2-3 months. However, what differentiates them from the Arg-61 mice studied here is that both the 3xTgAD and 5xFAD begin to show classical hallmarks of AD-neuropathology around the same time the cognitive deficits begin to appear. In terms of classical AD pathologies leading to cognitive deficits, apoE4 is associated with increased amyloid ␤ plaque formation and decreased clearance (48 -53). ApoE4 is also associated with increased activity of GSK3␤ (54,55), a kinase that phosphorylates tau protein (56,57). Consequently, apoE4 also leads to increased phosphorylation of Tau and the formation of neurofibrillary tangles (58 -61). Other AD pathologies that may explain the effect of apoE4 on cognition includes increased synaptic degeneration, neuro-inflammation, and neuronal death (62). However, Arg-61 mice do not show any thioflavin S stained amyloid ␤ plaques nor neurofibrillary tangles even up to 24 months old (data not shown), yet their cognitive deficits are observable as early as 3 months of age. Furthermore, the deficits seem to be most pronounced in FIGURE 5. Potential/possible neurotropic and apoptotic processes in Arg-61 mice. Shown is hippocampal expression of BDNF (a and c) and TrkB (b and d) mRNA in the hippocampus of 3-and 12-month-old C57BL/6J and Arg-61 mice, respectively. Although young Arg-61 mice had significantly higher mRNA levels of both BDNF (a) and TrkB (b) in their hippocampi (*, p ϭ 0.0318 and 0.0291 respectively; n ϭ 4 for C57BL/6J and 5 for Arg-61), there were no significant differences in the mRNA levels of both genes between 12-month old C57BL/6J and Arg-61 mice (p Ͼ 0.05 for both BDNF and TrkB; n ϭ 5 for both C57BL/6J and Arg-61). Despite the higher mRNA levels of BDNF in young Arg-61 mice, protein levels of BDNF in both young (e) and old mice (f) showed no significant differences between C57BL/6J and Arg-61 mice (p Ͼ 0.05 in both cases. n ϭ 4 for both genotypes of the young animals and 5 for both genotypes of the old animals). However, Arg-61 mice had higher levels of cleaved caspase-3 at both ages with an 87.7% increase (**, p ϭ 0.006; n ϭ 5 per group) in the young (g) and a much higher 172.4% increase (*, p ϭ 0.0145; n ϭ 5 per group) in the old (h) animals. All data are presented as the mean Ϯ S.E. In all figures white bars represent C57BL/6J and black bars represent Arg-61 mice. the female mice, in line with the greater risk in women to develop AD and an apparently greater effect of apoE4 in impairing spatial memory performance in females (32,63,64). This observation is consistent with numerous studies in apoE4 subjects showing that the role of apoE4 in cognitive function extends beyond the AD disease state, which apoE4 gene clearly impacts. However, studies in healthy apoE4 subjects or those with mild cognitive impairment suggest that the effect of apoE4 on cognition may be independent of the presence of classical/ hallmark AD neuro-pathologies (65,66).
ApoE4 genotype is the most important genetic risk factor for developing AD (susceptibility gene) even though, unlike the disease-associated mutations in the amyloid precursor protein (APP), Tau protein, and the presenilins (PS1 and PS2), apoE4 is not a deterministic AD-causing gene. Age is the most important risk factor for AD, but surprisingly, apoE4 genotype impacts cognition even in healthy, non-demented, and young individuals (9,10). Also, whereas longitudinal cognitive decline is steeper in apoE4 carriers (11,12,68), numerous reports have shown that the effect of apoE4 on cognition is actually greater in younger than in older subjects (45)(46)(47). Thus, our results showing cognitive deficits in young Arg-61 mice are consistent with the apoE4 cognitive phenotype hypothesis (69). This hypothesis suggests that the cognitive deficit associated with apoE4 is a direct effect of the allele on neuronal function (69) or their response to certain perturbations or "cognitive toxic effects" (70). Interestingly, a very recent report showed similar cognitive impairment in 3-month-old apoE4-targeted replacement mice (71). Further strengthening this hypothesis is another interesting study that shows evidences of spatial memory deficits in apoE4-positive children as young as 7 years old (72). Therefore, observing such a cognitive phenotype in young Arg-61 mice suggests that domain interaction feature of apoE4 may be sufficient to cause cognitive impairment in young apoE4 subjects in the absence of AD pathology. This may thus contribute to the earlier onset and faster progression of AD in apoE4 subjects who happen to eventually develop AD.
Increased Proliferation and Compensatory Attempts-The use of single-labeled doublecortin-positive cells to assess the proliferative activity of neuronal cells in the dentate gyrus of rodents has previously been characterized. DCX-positive cells in the adult rodent hippocampus are neurons born within 12 days before sacrificing (73). There is also much evidence that some doublecortin-positive cells may have multipotent precursor cell functions and are able to proliferate and differentiate into other cell types (74). Therefore, taking this information together with the characterization of DCX-positive cells based on the length, location, and complexity of their dendritic outgrowths (31), we were able to use single-labeled DCX immunostaining to differentiate between the mitotic and post-mitotic newly born neurons of the mice, not as a one-day snapshot obtained with BrdU co-labeling but over the ϳ12-day period before the animals were sacrificed. The percentage of proliferative (mitotic) DCX-positive cells that were obtained (based on dendritic length and complexity; Ϸ84%) confirmed a convergence of these methods, as using DCX/BrdU co-labeling reported 90% (73). This method may be less disruptive as it obviates the toxic effects of BrdU birth-dating. Despite the importance of this method, at this point the increased proliferation observed in the Arg-61 mice may not be necessarily "neurogenesis" in its strict sense. Neurogenesis had been defined as "the process of generating functional neurons from progenitor cells, including proliferation and neuronal fate specification of neural progenitors and maturation and functional integration of neuronal progeny into neuronal circuits." (75). Thus, further studies will be needed to understand the fate of the higher number of the mitotic DCX neurons in Arg-61 mice.
We consider the higher number of DCX-positive cells in the Arg-61 mice proliferation rather than frank neurogenesis. In fact, despite the significantly higher mitotic subgroup in Arg-61 mice, our data show no difference in the post-mitotic types of DCX-positive neurons or in the percentage of mitotic neurons (83.7 Ϯ 1.8 versus 83.9 Ϯ 2.4% for Arg-61 and C57BL/6J mice, respectively).
ApoE4 has been shown to activate a cell type-specific, lipoprotein receptor-related protein-mediated apoptotic effect that may contribute to its role in AD and related dementia (76). Furthermore, Michikawa and Yanagisawa,(77) demonstrated that apoptotic cell death in neurons induced by apoE4 may be as a result of reduced endogenous cholesterol secretion; interestingly, Arg-61 mice showed reduced cholesterol secretion (18). Thus, taken together with the 87% increase in cleaved caspase-3 and its role as the main executioner caspase in apoptosis (78), it is tempting to speculate that a higher percentage of the excess proliferating cells in the Arg-61 mice, as in normal conditions, become apoptotic or die at one point or another without becoming functional or getting integrated into the local circuit (79). This may, therefore, mitigate the positive role that immature DCX-positive cells in the hippocampus play in learning (rather than remembering) (80). Interestingly, most of the impairments seen in the Arg-61 mice are related more to remembering than to learning (Figs. 2, c and d, and 3, a-c). Furthermore, previous data seem to favor the loss of the excess proliferating cells via apoptosis since; even though some DCXpositive cells may retain their multipotentiality (74) and thus are likely to differentiate into glia, Zhong et al. (18) reported that there was neither gliosis nor difference in NeuN-positive cell counts in the hippocampus of Arg-61 versus C57BL/6J mice. Meanwhile, the fact that we observed even a higher level of cleaved caspases-3 in older animals seems contrary to our earlier speculation about potential apoptosis of most of the mitotic DCX cells in young Arg-61 mice. This is because cell proliferation is drastically reduced to barely detectable levels in older animals (38), and therefore, such newborn cells cannot account for the higher magnitude of cleaved caspase-3 increase in the older mice. However, another plausible conjecture is that different groups of cells account for the cleaved caspase-3 expression in old versus young animals. A study dedicated to unravel caspase activity, vis à vis the identity of apoptotic cells in young versus old mice is, therefore, indicated and will be addressed in-depth in future studies.
Moreover, the emerging roles of caspases in the neuropathology of AD and recent studies suggest that the presence of active caspase may not necessarily lead to cell death in an acute or short term manner. In contrast, this chronic, non-apoptotic activation of caspase-3 is an early sign of AD pathology, which contributes to cleavage of Tau to form soluble species that are more neurotoxic than the fibrillar/tangle forms. This activated caspase stage corresponds to the onset of memory decline in Tg2576 AD mouse model. It is also worth noting that the said caspase activation is driven by soluble A␤ species (81-83) (for review, see Ref. 83), which we have not ruled out in our Arg-61 mice as the oligomers have been found in young apoE4 mice (84). Thus, thioflavin-negative A␤ oligomers may lead to an increase in activated caspase-3 in both the young and old Arg-61 mice, which may not necessarily cause cell death but which in turn leads to the formation of soluble Tau species that are synaptotoxic and may mediate the observed impairment in cognitive functions at both young and old ages.
Disruption in neurogenesis is a well known pathology of AD; however, increases and decreases of neurogenesis reported in human AD brains and several mouse models may not be contradictory, but rather, both may contribute to the disease condition (85). Although some previous reports show that functional neurogenesis is reduced in the 3xTgAD mouse models of AD (86,87), the increased proliferation reported here is consistent with data obtained either by DCX positive alone or by DCX/BrdU cell counts in a transgenic mouse model expressing human APP isoforms with the Swedish (K670N/M671L) and Indiana (V717F) mutations (88) and in other models (APP/PS1 double transgenic) (89). Interestingly, a similar increase in adult hippocampal cell proliferation (measured by multiple endogenous newborn neuron markers) was found in the subgranular zone of human subjects with AD (88). It has therefore been suggested that this increase may be a compensatory repair mechanism (88,90).
This compensatory attempt is in line with the role that adultborn hippocampal neurons play not only in learning but also in memory functions. Several AD mouse models that displayed impairment in learning and memory also show reduced adult neurogenesis phenotype (85). Similarly, neurogenesis drastically reduce with age, partly explaining why cognitive functions may decrease in aging (36,91). For this reason also, several types of AD therapeutic intervention attempts had been directed toward improving neurogenesis. Interestingly, such interventions including drugs, enriched environment, and exercise, which all increase neurogenesis, has also been shown to improve the learning and memory deficits in AD mice models and aged rodents (86,(92)(93)(94). Furthermore, treatments that decrease neurogenesis negatively impact both learning and memory (95), and the level of neurogenesis has been shown to correlate with cognitive function in both rodents (86,91,96) and humans (97), Thus it is clear that newborn neurons studies with endogenous DCX expression and other such markers play important roles in learning and memory. Several explanations have been put forward including the fact that adult-born neurons have a lower threshold for induction of long-term potentiation (95) and immediate early gene expression compared with the matured dentate-gyrus neurons. Computational models called pattern separation also suggest that immature adultborn neurons act as pattern integrators by encoding new memories to preserve old memories and prevent interference (98).
A study more relevant to ours showed that proliferation is also increased in the hippocampus of apoE4 mice relative to apoE3 mice, whereas maturation is delayed, and overall survival is decreased (99). Taken together, these data suggest that apoE4 domain interaction modeled in the Arg-61 mice may be sufficient to cause the disruption in neurogenesis as seen in the apoE4 mice.
BDNF (neurotrophic factor) signaling is a widely known, important player in the process of neurogenesis. We also observed a significant increase in BDNF and TrkB receptor mRNA in the Arg-61 mice even though this mRNA increase did not translate at protein levels from our Western blot assay (Fig.  5e). We believe this increase is another arm of the compensatory attempt in the Arg-61 mice against latent or subtle pathologies including chronic excitotoxicity and endoplasmic reticulum stress (100). This BDNF mRNA increase may also contribute in part to the increase in proliferation (34). Interestingly, no such increases were observed in the older Arg-61 mice (Fig. 5, c, d, and f) in which neurogenesis has actually decreased relative to the C57BL/6J mice (Fig. 6). The significance of increased proliferation-compensation, as seen in the Arg-61 mice, has been discussed in human AD brains and other AD mice models (40, 88 -90). However, the later effect of this compensatory mechanism seems to be detrimental, as highlighted in the J20 mouse model, a transgenic mouse model of AD bearing both the APP Swe and APP Ind , suggests that this early age increase may be consistent with what an increase in neurogenesis has been proposed to do in humans during early adulthood before the time the disease sets in. The authors posited that the initial increase may lead to a significant decrease in neurogenesis later in life (101), probably due to the depletion of neuronal stem/progenitor cells at early ages. In this hypothesis, the driving force of this early age stimulation of proliferation is amyloid-␤. However, amyloid-␤ plaques have not been observed in Arg-61 mice, and no (putative) driving force in humans has been identified either. Interestingly, consistent with this hypothesis, we saw a significant reduction in DCX expression in the older Arg-61 mice versus C57BL/6J (Fig. 6). This suggests that the effect of Arg-61 apoE on early-life stimulation of neuronal proliferation may lead to reduced neurogenesis later in life.
Conclusion-The current data suggests that apoE4 domain interaction negatively impacts cognition not only at older ages but even early in life before any classical AD pathologies are likely to begin to appear. The increase in BDNF and TrkB receptor mRNA expression (which was not sustained later in life) as well as the increase in doublecortin-positive cells in Arg-61 mice may be an attempt by the system to repair itself. These compensatory attempts are consistent with the recent discovery that the apoE4 allele appear to be "beneficial" in earlier age and to confer risk of cognitive decline later in life and, therefore, as an example of antagonistic pleiotropy (13,67,102). We further demonstrated here that these compensatory attempts are obviously not adequate, as this is accompanied with an increase of cleaved caspase-3 expression in younger age (an effect sustained till the older age) and reduction of doublecortin expression later in life at old age. These data, including the apparent greater effect on female animals, are consistent with apoE4-dependent cognitive phenotype (69). Taken together, these data further strengthen the idea that domain interaction may be the main pathologic feature of apoE4 (14,15). It further suggests that the cognitive pathologies associated with apoE4 may be independent of old age or classical AD pathologies including neurofibrillary tangles and A␤-plaques. Similar observations have been made in humans in which healthy adults carrying the apoE4 gene show signs of cognitive deficits. Thus, apoE4 domain interaction may be a viable therapeutic and/or prophylactic target for Alzheimer disease and cognitive impairment especially early in life in apoE4 subjects.