Depletion of adult neurogenesis exacerbates cognitive deficits in Alzheimer’s disease by compromising hippocampal inhibition

Animal care procedures were conducted according to the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Our colony is maintained via group housing (≤5 mice per cage) in a high barrier facility under a 14:10 light:dark cycle with free access to food and water. FAD-linked APPswe/PS1ΔE9 mice [34] and Nestin-δ-HSV-TK mice [35] were generated as previously described. To maintain consistency of pathology progression and avoid gender-related disparate observations, mice used in this study were all females. Mouse euthanasia was performed using isofluorane and cervical dislocation.

Valganciclovir powder was mixed into mouse chow (.09%, Valcyte, valganciclovir hydrochloride) and given ad libitum for an average dose of ~90 mg/kg/day (Custom Animal Diets, LLC (Easton, PA)). Animals were fed either Valganciclovir- containing chow or standard chow (“vehicle”) right after weaning for 2 months. At the end of this period mice were subjected to behavioral tests, i.e., contextual fear conditioning and pattern separation sequentially. A separate group of mice was subjected to pattern separation followed by a single probe test, as described in Fig. 5. Following sacrifice, brains were analyzed as described below.

Performed as previously described [36]. Conditioning was conducted in two distinct contexts: context A with the shock, and context C without the shock. Both test cages (17.8 × 17.8 × 30.5 cm) were encased by isolation cubicles. Context A had two plexiglass walls, two metal walls and a stainless steel grid floor (Coulbourn Instruments). The light and fan were turned on. A mild lemon blossom scent was used, and 70% ethanol for cleaning. Context C had the light and fan turned off, the chamber door ajar, a mild anise scent, and Clorox disinfecting wipes for cleaning. The shape of the chamber was altered with a plastic circular insert, and a plastic flooring was placed on top of the stainless steel grid floor with cage bedding added. Motion was recorded by a digital video camera mounted above the test cage. On day 0 mice were tested only in context A, then for 3 consecutive days in both A and C context with an hour separation. In test cage A the mice received a single 2 s foot shock (0.75 mA at the 185th second). FreezeFrame and FreezeView software (Actimetrics) were used for recording and analyzing freezing behavior. Percentage of freezing during the first 180 s in each context for each day was computed. Discrimination ratios were calculated using the formula:

Performed as previously described [33]. Conditioning was conducted in two similar contexts: the shock context A, and the similar context B. Both cages had two clear Plexiglas walls, two grey metal walls and a stainless steel grid floor (Coulbourn Instruments). In cage A the house light and fan were turned on. A mild lemon blossom scent was used. Cages were cleaned with 70% ethanol prior to mouse placement. Test cage B differed from test cage A in that the metal walls had black and white inserts, the house light and fan were turned off and the chamber door was left ajar. A mild peppermint scent was used, and Clorox disinfecting wipes. On day 0 mice were exposed only to the training context A. For the next nine consecutive days mice were exposed to A and B in a randomized order. Discrimination ratios were calculated using the formula:

Two doses of 5′-bromo-2′-deoxyuridine (BrdU; Sigma) were administered four hours apart intraperitoneally (100 mg/kg), in physiological saline. Animals were sacrificed four weeks later.

For immunohistochemi​cal staining, all mice were anesthetized using overdose of isofluorane and transcardially perfused with ice-cold PBS. Removed brains were halved on the sagittal plane, and half placed into 4% paraformaldehyde and half saved for western blot.

50 μm sagittal sections cut using a sliding freezing microtome (Leica Biosystems, Buffalo Grove, IL) were stored in cryoprotectant (glycerol, ethylene glycol, 1X PBS) at −20 °C. The following antibodies were used: rat anti-BrdU (1:400; Accurate Chemical & Scientific Corp., Westbury, NY), mouse anti-nestin (1:100; Millipore Corporation, Billerica, MA), goat anti-doublecortin (DCX; 1:400; Santa Cruz Biotechnology, Santa Cruz, CA), mouse anti-NeuN (1:400, Millipore, Temecula, CA), rabbit anti-Egr-1 (1:250, Santa Cruz Biotechnology, Santa Cruz, CA), goat anti-green fluorescent protein (GFP, 1:1000, Abcam, Inc. Cambridge, MA) and mouse anti-green fluorescent protein (GFP, 1:200; Santa Cruz Biotechnology, Santa Cruz, CA). Secondary antibodies from Jackson ImmunoResearch Laboratories (West Grove, PA): biotinylated species-specific anti-IgG (all used at 1:250), Cy3-conjugated Donkey anti-Rabbit (1:250), Cy3-conjugated Donkey anti-Rat (1:250), Alexa Fluor 647 (AF647)-conjugated Donkey anti-Mouse and anti-Rat (1:250), Cy2-conjugated Streptavidin (1:250), DAPI Nucleic Acid Stain (1:2500, Life Technologies, Grand Island, NY).

Cell counts were performed using design-based stereology (StereoInvestigator version 8, MBF Bioscience). For the analysis, every sixth section of brain tissue was quantified by applying the Nv × VRef method. Sections were traced using a Zeiss AX10 microscope (Carl Zeiss Ltd., Hertfordshire, England) in low magnification (5×) and counting was performed at high magnification (63×), counting frame = 100 μm × 100 μm, grid size 100 μm × 100 μm, and all sections were counted using 12.5-μm top and bottom guard zones.

Entorhinal cortex underwent extraction for isolating the water soluble Aβ fraction for dot-blot analysis as previously described [37]. Briefly, entorhinal cortex was homogenized in 1X PBS containing protease inhibitor cocktail (1:100, Sigma-Aldrich #P8340) followed by ultracentrifugation at 100,000 g for 1 h at 4 °C. The water-soluble fraction was quantified via the BCA method and 25 μg total protein was blotted onto prewet nitrocellulose membrane in the dot-blot apparatus (Bio-Rad Bio-Dot Microfiltration Apparatus). The membrane was then washed once in TBS and was blocked in 5% nonfat milk in 1X TBS + 0.01% Tween 20 (TBST) for 2 h at room temperature and incubated overnight at 4 °C in polyclonal rabbit anti-amyloid oligomer (A11; 1:5000; EMD Millipore #AB9234). The membrane was then washed three times in TBST and then incubated for 1 h at room temperature in IRDye 800CW donkey anti-rabbit IgG (1:20,000; Li-Cor #925–32213). The membrane was imaged using an Odyssey Fc (800 channel, 30 s acquisition) and protein expression levels were quantified using Image Studio Lite (version 5.2.5; Li-Cor).

Protein extraction of brain tissue was performed as previously [30]. Antibodies used for Western blot included mouse anti-actin (1:5000, Millipore, Temecula, CA), rabbit anti-amyloid precursor protein (APP, Abcam, Cambridge, MA), mouse anti-phospho-PHF-tau (AT8, 1:500, ThermoFisher Scientific) and mouse anti-tau (tau-5, 1:1000, Millipore, Temecula, CA). The following secondary antibodies were used, mouse anti-HRP (1:10,000, Thermo Scientific, Rockford, Il) and rabbit anti-HRP (1:15,000, Promega, Madison, WI). Protein expression was measured in ImageJ and was normalized to actin.

Stereological quantification was analyzed using two-tail or one-tail (Fig. 5), unpaired t-test, or Welch’s unequal variance t-test where appropriate. Contextual fear conditioning, pattern separation, and discrimination ratios were analyzed using repeated measures, two-way ANOVA with Holm-Sidak multiple comparison testing. Western blot data was analyzed using two-tail, unpaired t-test, or Welch’s unequal variance t-test where appropriate. All statistical analysis was done in GraphPad Prism (Version 7.01; GraphPad Software Inc., La Jolla, CA, USA). All data shown represent mean ± S.E.M. and a probability of less than 0.05 was considered statistically significant.

To address the hypothesis that decreased adult neurogenesis will exacerbate the cognitive deficits associated with Alzheimer’s disease, we temporally ablated neural progenitor cells from the brains of FAD-linked APPswe/PS1ΔE9 transgenic mice. For this purpose, we bred the APPswe/PS1ΔE9 mice with the Nestin-δ-HSV-TK transgenic mouse line, [35]. The Nestin-δ-HSV-TK transgenic line contains a modified version of the herpes simplex virus thymidine kinase (TK), as well as an enhanced green fluorescent protein (GFP), driven by the nestin promoter and its second intron regulatory element (Fig. 1a). Administration of valganciclovir in mouse chow, which is specifically phosphorylated by Nestin-δ-HSV-TK, kills dividing nestin expressing cells in these mice by acting as a toxic thymidine analog. The number of GFP-expressing nestin positive cells in the subgranular layer of the hippocampus of Nestin-δ-HSV-TK treated with valganciclovir is reduced compared to vehicle-treated Nestin-δ-HSV-TK mice (Fig. 1b-f). Likewise, a dramatic decrease in GFP+ cells was observed in brain sections of APPswe/PS1ΔE9;Nestin-δ-HSV-TK treated with valganciclovir, compared to vehicle-treated APPswe/PS1ΔE9;Nestin-δ-HSV-TK mice (Fig. 1g-n). To determine the nature of neurogenic deficits in valganciclovir-treated Nestin-δ-HSV-TK and APPswe/PS1ΔE9;Nestin-δ-HSV-TK, we quantified the number of GFP+ neural progenitor cells (NPCs) by unbiased stereology following a two-month valganciclovir treatment. No change in the number of GFP + DCX- (nestin expressing NPCs; Fig. 1d) or GFP + DCX+ (neuroblasts; Fig. 1e) was observed between valganciclovir- and vehicle-treated Nestin-δ-HSV-TK mice. However, a significant reduction in the number of GFP-DCX+ (immature neurons) was observed in Nestin-δ-HSV-TK mice treated with valganciclovir (Fig. 1f; two-tailed, unpaired t-test; t ₇  = 5.589, P = 0.0008). Taken together, this may suggest that in the tested conditions, reduced neurogenesis in the valganciclovir treated Nestin-δ-HSV-TK mice is manifested at the immature neuron stage. Interestingly, APPswe/PS1ΔE9;Nestin-δ-HSV-TK mice treated with valganciclovir displayed a significant decrease at earlier stages. Specifically, a significant reduction in total number of nestin expressing NPCs (GFP + DCX-, Fig. 1i; two-tailed, unpaired t-test; t ₉  = 3.253, P = 0.0099), neuroblasts (GFP + DCX+, Fig. 1j, two-tailed, unpaired t-test; t ₉  = 3.305, P = 0.0092) and immature neurons (GFP-DCX+, Fig. 1k, two-tailed, unpaired t-test; t ₉  = 2.932, P = 0.0167) in the subgranular layer following treatment (Valganciclovir N = 5, Vehicle N = 6). This may suggest that in the APPswe/PS1ΔE9;Nestin-δ-HSV-TK mice, treatment with valganciclovir affects earlier neurogenic populations compared to the Nestin-δ-HSV-TK mice and is manifested by a significant reduction in the number of NPCs, neuroblasts and immature neurons. This may further support previous reports suggesting impairment of NPCs in FAD [30, 38]. To assess the impact on the number of newly born neurons we quantified the number of BrdU + NeuN+ cells in the granular cell layer of the DG of the APPswe/PS1ΔE9;Nestin-δ-HSV-TK mice. Unbiased stereology demonstrated a trending decrease in the total number of BrdU+ (Fig. 1l, two-tailed, Welch’s unequal variance t-test; t _4.034  = 2.204, P = 0.0917), BrdU + NeuN+ new neurons (Fig. 1m, two-tailed, Welch’s unequal variance t-test; t _4.062  = 2.077, P = 0.1054), as well as in the number of new glia BrdU + NeuN- cells (Fig. 1n, two-tailed, Welch’s unequal variance t-test; t _4.137  = 2.457, P = 0.0679) following valganciclovir treatment (Valganciclovir N = 4, Vehicle N = 5). Combined, this data demonstrates a successful ablation of adult neurogenesis in APPswe/PS1ΔE9;Nestin-δ-HSV-TK animals, resulting in less new neurons and glia in the granular cell layer of the DG. In addition, these results may imply that NPCs and neuroblasts are more vulnerable in the brains of APPswe/PS1ΔE9;Nestin-δ-HSV-TK mice compared to the Nestin-δ-HSV-TK mice.

To determine cognitive performance of FAD mice following depletion of neurogenesis, mice were tested in the contextual fear conditioning and the DG-specific pattern separation tasks as previously described [36, 39]. In the contextual fear-conditioning task (Fig. 2a), age matched wild type (Nontransgenic, N = 11, Fig. 2b) and APPswe/PS1ΔE9 (Fig. 2d, n =14) mice that were fed with vehicle chow, were able to successfully distinguish between the two contexts, based on percentage freezing, by the end of the second day. Discrimination ratio shows no difference between vehicle- and valganciclovir-fed wild type (Fig. 2c) or APPswe/PS1ΔE9 mice (Fig. 2d), suggesting no side effect of the valganciclovir on animals’ behavior. Similarly, Nestin-δ-HSV-TK animals fed with either vehicle- or valganciclovir-containing chow were also able to successfully learn the task (vehicle-treated Nestin-δ-HSV-TK N = 12, valganciclovir- treated Nestin-δ-HSV-TK N = 8, Fig. 2f-h respectively), suggesting that ablation of adult neurogenesis alone is insufficient for the disruption of this behavioral task. Vehicle treated APPswe/PS1ΔE9;Nestin-δ-HSV-TK animals exhibited some difficulty learning the task on the first day, but successfully distinguished between the two contexts on day 2 and 3 (N = 7, Fig. 2i). Importantly, valganciclovir-fed APPswe/PS1ΔE9;Nestin-δ-HSV-TK mice were unable to distinguish between context A or C on any of the days of the test (N = 9, Fig. 2j). It is important to note that this deficit was only observed with the combination of ablated neurogenesis and the FAD mouse background. Nevertheless, discrimination ratios revealed no significant difference in the animal’s ability to discriminate between contexts between days (Fig. 2k), which may suggest that the effect is very mild. However, valganciclovir-fed Nestin-δ-HSV-TK mice had a significant Context x Days interaction (Fig. 2l, repeated measures two-way ANOVA, Context x Days: F_2,36 = 3.56 P = 0.0388). Furthermore, on Days 2 and 3 valganciclovir-fed Nestin-δ-HSV-TK exhibited significantly increased abilities to discriminate between contexts A and C compared to valganciclovir-fed APPswe/PS1ΔE9;Nestin-δ-HSV-TK and APPswe/PS1ΔE9 mice (Fig. 2l; repeated measures two-way ANOVA, * P < 0.05, ^# P < 0.05, ^## P < 0.01) with the genotype having a significant influence on discrimination between contexts A and C (repeated measures two-way ANOVA, Genotype: F_2,22 = 7.00 P = 0.0044). To examine whether freezing behavior is the result of anxiety rather than learning, mice were subjected to the Light/Dark box anxiety task. However we found no differences in anxiety level between any of the groups (data not shown), suggesting that these observed deficits are not a result of increased anxiety in the APPswe/PS1ΔE9;Nestin-δ-HSV-TK animals.

We next asked whether depletion of neurogenesis in the FAD mice would induce difficulty learning in a known DG-specific task, pattern separation. The challenge in this task is the high similarity between the two contexts, and typically the learning process is longer compared to the contextual fear conditioning (Fig. 3a).

Vehicle chow-fed age matched wild type (Nontransgenic, N = 11, Fig. 3b) or APPswe/PS1ΔE9 animals (N = 13, Fig. 3d) were able to successfully distinguish between the two contexts, based on percentage freezing, by the end of the pattern separation task (repeated measures two-way ANOVA, ** P < 0.01, ***P < 0.001). Discrimination ratios of vehicle and valganciclovir-treated wild type (Fig. 3c) and APPswe/PS1ΔE9 (Fig. 3e) show no difference between the treatment groups, excluding the possibility of side effect of valganciclovir on mouse behavior. Unexpectedly, Nestin-δ-HSV-TK animals fed with either vehicle or valganciclovir containing chow were able to successfully learn the task (Vehicle-treated N = 12, valganciclovir-treated N = 7, Fig. 3f-h; repeated measures two-way ANOVA, * P < 0.05, **P < 0.05, ***P < 0.001, ****P < 0.0001), suggesting that either ablation of adult neurogenesis using the Nestin-δ-HSV-TK transgenic line is insufficient for the disruption of this behavioral task, or that our specific contexts were not sufficiently challenging, masking the deficit. Interestingly, vehicle-fed APPswe/PS1ΔE9;Nestin-δ-HSV-TK were able to discriminate between the tasks albeit not all days were statistically significant (N = 7, Fig. 3i; repeated measures two-way ANOVA, * P < 0.05, **P < 0.05, ***P < 0.001). This may suggest a genetic background effect in comparison to vehicle-fed APPswe/PS1ΔE9 mice. Importantly, ablation of neurogenesis in the APPswe/PS1ΔE9;Nestin-δ-HSV-TK mice completely disrupted the performance of mice in the first three days of the task (N = 8, Fig. 3j; repeated measures two-way ANOVA, * P < 0.05). Discrimination ratios within genotype revealed no significant differences in the animal’s ability to discriminate between contexts between days (Fig. 3k), suggesting a mild behavioral effect of neurogenesis in this task. Additionally, when comparing discrimination ratios between valganciclovir-fed Nestin-δ-HSV-TK, APPswe/PS1ΔE9 and APPswe/PS1ΔE9;Nestin-δ-HSV-TK mice there was no significance within days but there was an overall days effect among the groups (Fig. 3l; repeated measures two-way ANOVA, Days: F_5,100 = 10.50 P < 0.0001).

Next, we asked whether, in addition to learning and memory, hippocampal neurogenesis would affect AD neuropathology. Western blot analysis of hippocampal protein lysate of vehicle- and valganciclovir-treated APPswe/PS1ΔE9;Nestin-δ-HSV-TK animals showed that the levels of phosphorylated tau, as detected by AT8 antibodies, demonstrated a trending increase following ablation of adult neurogenesis in the valganciclovir-treated APPswe/PS1ΔE9;Nestin-δ-HSV-TK (Fig. 4a,b). Total tau expression levels did not change (tau-5, Fig. 4a,c), while the ratio of phosphorylated tau to total tau was significantly increased (Fig. 4d; two-tailed, unpaired t-test; t ₄  = 4.368, P = 0.0120). This observation suggests that ablation of adult neurogenesis induces upregulation of tau hyperphosphorylation in the hippocampus. Immunohistochemical analysis of hippocampal sections stained for phosphorylated tau (p-tau) and doublecortin (DCX) clearly show AT8 immunoreactivity in DCX+ neuroblasts and new neurons in valganciclovir-treated APPswe/PS1ΔE9;Nestin-δ-HSV-TK, but not in vehicle-treated APPswe/PS1ΔE9;Nestin-δ-HSV-TK mice (Fig. 4e). This may suggest that depletion of neurogenesis compromises the regulation of tau phosphorylation in newly differentiated neurons. Given the critical role of tau in neuronal maturation, this may suggest that the reduced level of neurogenesis may lead to the incorporation of aberrant new neurons into the hippocampus. To make sure that the observed alterations in tau phosphorylation are not a side effect of the valganciclovir treatment, we examined p-tau expression in vehicle- and valganciclovir- treated wild type mice and observed no change in p-tau expression (Additional file 1: Figure S1). This suggests that induction of hyperphosphorylated tau observed in the valganciclovir-treated APPswe/PS1ΔE9;Nestin-δ-HSV-TK was due to the depletion of neurogenesis.

To address the effect of neurogenesis on the amyloidogenic pathway we examined levels of full length APP in hippocampal protein lysates of vehicle- and valganciclovir-treated APPswe/PS1ΔE9;Nestin-δ-HSV-TK animals. We observed that APP levels do not change after ablation of neurogenesis (N = 3, Fig. 4a). In addition, the level of oligomeric Aβ was comparable in water-soluble protein extracts prepared from the entorhinal cortices of these mice (Fig. 4f). These results suggest that the effect of depletion of neurogenesis on tau may not be modulated by amyloidosis.

To assess the impact of loss of neurogenesis on the activity of the hippocampal circuitry we examined expression of the immediate early gene Egr-1 (zif268), as an indicator of neuronal activation following learning. For this purpose, Nestin-δ-HSV-TK mice fed with vehicle or valganciclovir were subject to pattern separation, as above. To assess the neuronal population activated by the long-term memory of context A, ten days after task completion, a single probe trial of only the shock context was performed, and mice were immediately sacrificed and examined for Egr-1 expression (Fig. 5a). Using unbiased stereology the number of Egr-1 expressing cells was quantified in the granular cell layer of the DG, the CA1 region, and the CA3 region of these mice (N = 4). Ablation of adult neurogenesis caused a trending increase in the number of Egr-1 expressing cells in the granular cell layer of the DG (Fig. 5b), with a significant increase in the CA3 region (Fig. 5c,e,f; one-tailed, unpaired t-test; t ₄  = 2.302, P = 0.0414), and the CA1 region (Fig. 5d,g,h; one-tailed, unpaired t-test; t ₆  = 2.039, P = 0.0438). These results suggest that eliminating the presence of new neurons relieves inhibitory circuitry in the DG-CA3-CA1. Alternatively, this may be the result of decreased excitatory synapses onto interneurons in the hilus.