Augmenting neurogenesis rescues memory impairments in Alzheimer's disease by restoring the memory-storing neurons

Abstract

Hippocampal neurogenesis is impaired in Alzheimer's disease (AD) patients and familial Alzheimer's disease (FAD) mouse models. However, it is unknown whether new neurons play a causative role in memory deficits. Here, we show that immature neurons were actively recruited into the engram following a hippocampus-dependent task. However, their recruitment is severely deficient in FAD. Recruited immature neurons exhibited compromised spine density and altered transcript profile. Targeted augmentation of neurogenesis in FAD mice restored the number of new neurons in the engram, the dendritic spine density, and the transcription signature of both immature and mature neurons, ultimately leading to the rescue of memory. Chemogenetic inactivation of immature neurons following enhanced neurogenesis in AD, reversed mouse performance, and diminished memory. Notably, AD-linked App, ApoE, and Adam10 were of the top differentially expressed genes in the engram. Collectively, these observations suggest that defective neurogenesis contributes to memory failure in AD.

Graphical abstract

Figure 1.

Rescue of hippocampus-dependent memory following augmentation of neurogenesis in FAD mice. (A) Breeding scheme. Abbreviations: corn oil (Corn, C)– or tamoxifen (TAM, T)-treated NestinCreER^T2; Bax^fl/fl (NB) and NestinCreER^T2; Bax^fl/fl; 5XFAD (NBF). (B) Experimental paradigm. (C) The number of DCX-expressing neuroblasts and immature neurons in the DG of corn oil (C)– or tamoxifen (T)-treated NestinCreER^T2; Bax^fl/fl (NB) and NestinCreER^T2; Bax^fl/fl; 5XFAD (NBF) quantified by unbiased stereology. N = 5 for C-NB, T-NB, and C-NBF; N = 6 for T-NBF; Kruskal–Wallis test with Dunn’s post hoc test (**P = 0.009, ****P = 0.0001). (D) The extent of survival of new neurons (BrdU⁺NeuN⁺) following treatment with either tamoxifen or corn in the DG of corn oil (C)– or tamoxifen (T)-treated NestinCreER^T2; Bax^fl/fl (NB) and NestinCreER^T2; Bax^fl/fl; 5XFAD (NBF), as quantified by unbiased stereology. N = 5; *P < 0.05, ****P < 0.0001. (E) The number of new neurons (DCX⁺NeuN⁺) in the DG of corn oil (C)– or tamoxifen (T)-treated NestinCreER^T2; Bax^fl/fl (NB) and NestinCreER^T2; Bax^fl/fl; 5XFAD (NBF), as quantified by unbiased stereology. *P < 0.05, ****P < 0.0001. (F) Confocal images of new neurons (DCX⁺NeuN⁺) in brain sections of tamoxifen (T)-treated NestinCreER^T2; Bax^fl/fl; 5XFAD (NBF). Scale bar = 50 μm; insert = 20 μm. (G) A scheme of the NOL test. (H and I) Performance of corn oil (C)– or tamoxifen (T)-treated NestinCreER^T2; Bax^fl/fl (NB) and NestinCreER^T2; Bax^fl/fl; 5XFAD (NBF) in the NOL test. Results indicate the percentage of time spent at the novel location. One-way ANOVA with Fisher’s LSD post hoc test F(7, 104) = 6.032; P < 0.0001 and *P < 0.05, **P < 0.005 (H) and discrimination index. Kruskal–Wallis test with Dunn’s post hoc test; ***P = 0.009, **P = 0.005 (I). (J) Experimental design of the CFC test. (K) Percentage of freezing in the CFC test (***P < 0.001, *P = 0.012).

Figure S1.

Validation measures of experimental models. (A) Bax expression level in neural progenitor cells cultured from the hippocampi of corn oil (C)– or tamoxifen (T)-treated NestinCreER^T2Bax^fl/fl (NB) or NestinCreER^T2; Bax^fl/fl; 5XFAD (NBF) mice, as tested by PCR (Student’s t test, *P < 0.05). (B) The number of DCX⁺ and BrdU⁺ proliferating neuroblasts in the brains of corn oil (C)– or tamoxifen (T)-treated NestinCreER^T2Bax^fl/fl (NB) or NestinCreER^T2; Bax^fl/fl; 5XFAD (NBF) mice 5 wk after BrdU injection. The number of proliferating neuroblasts was similar among most groups. A significant difference was observed between C-NB and C-NBF and T-NB and C-NBF (Kruskal–Wallis test with Dunn’s post hoc test; *P < 0.05, ***P = 0.0008). (C) 4.5-mo-old corn oil (C)– or tamoxifen (T)-treated NestinCreER^T2Bax^fl/fl (NB) or NestinCreER^T2; Bax^fl/fl; 5XFAD (NBF) mice (N = 12–16) were subjected to the light–dark behavioral test. The time spent in the light or dark was comparable between the groups. Comparable anxiety levels were observed in wild type and FAD mice (two-way ANOVA with Fisher’s LSD post hoc test F(3, 94) = 0.009558, P = 0.9987). (D) Quantification of the number of surviving new neurons (BrdU⁺NeuN⁺) in mice that were not injected with viral engram cocktail (Kruskal–Wallis test with Dunn’s post hoc test; *P < 0.05). (E–I) Specificity of eYFP and Egr-1 expression in mouse brains following CFC. Quantification of the number of eYFP and Egr-1 expressing cells in brain sections of mice that were either maintained in standard housing (home cage) or underwent CFC. eYFP (E; Kruskal–Wallis test with Dunn’s post hoc test; *P < 0.05, ***P = 0.0001), Egr-1 (F: *P < 0.05, **P < 0.005), DCX⁺Egr-1⁺ (G: **P < 0.01, ****P < 0.0001), eYFP⁺Egr-1⁺ (H: *P = 0.0157, **P < 0.01, ****P < 0.0001), DCX⁺eYFP⁺Egr-1⁺ (I: *P < 0.05, **P < 0.01, ***P = 0.0002). (J–P) A comparison of the neurons in the engram in Corn oil- and tamoxifen-treated NestinCreER^T2; Bax^fl/fl mice including total number of eYFP⁺ (J), NeuN⁺eYFP⁺ (K), NeuN⁺BrdU⁺eYFP⁺ (L), NeuN⁺DCX⁺eYFP⁺ (M), % NeuN⁺DCX⁺eYFP⁺/total eYFP⁺ (N), % DCX⁺eYFP⁺Egr-1⁺/total DCX⁺ (O), % DCX⁺eYFP⁺Egr-1⁺/total eYFP⁺Egr-1⁺ (P; P = ns, **P = 0.001).

Figure 2.

Augmenting neurogenesis rescues the recruitment of immature neurons into the memory circuit. (A) Experimental design aiming to identify immature neurons incorporating in the contextual memory circuit in the DG of corn oil (C)–treated NestinCreER^T2; Bax^fl/fl (NB) and corn oil (C)– or tamoxifen (T)-treated NestinCreER^T2; Bax^fl/fl; 5XFAD (NBF) mice. (B) Confocal images demonstrating the recruitment of a granule neuron into the memory circuit during memory acquisition (eYFP) and retrieval (c-fos). eYFP expression in granule neurons infected with the viral engram kit AAV9-cFos-tTA and AAV9-TRE-ChR2-eYFP in the DG of NestinCreER^T2; Bax^fl/fl mice was upregulated following CFC training (shock). c-fos was upregulated following the test. Brain sections were immunostained with antibodies raised against c-fos and show co-localization of c-fos and eYFP. Scale bar = 20 μm. (C and E–G) Quantification of cells incorporated in memory circuit during acquisition. Total number of activated cells (C; eYFP⁺; Kruskal–Wallis test with Dunn’s post hoc test, P = ns); total number of activated neurons (E; NeuN⁺eYFP⁺; P = ns); number of new neurons (F; NeuN⁺DCX⁺eYFP⁺; **P < 0.01). The number of new neurons born 6 wk prior to CFC (G; NeuN⁺BrdU⁺eYFP⁺; **P = 0.0011) in the DG of corn oil (C)–treated NestinCreER^T2; Bax^fl/fl (NB) and corn oil (C)– or tamoxifen (T)-treated NestinCreER^T2; Bax^fl/fl; 5XFAD (NBF). (D) Co-localization of enhanced eYFP and endogenous c-fos in granule neurons in the DG of tamoxifen (T)-treated NestinCreER^T2; Bax^fl/fl; 5XFAD (NBF) mouse. Scale bar = 75 μm. (H) Percentage of new neurons recruited into the engram to the total eYFP⁺ cells (*P = 0.0284 and **P = 0.0019). (I and J) Co-localization of eYFP and DCX⁺ (I), and eYFP⁺, DCX⁺, and NeuN⁺ (J) in immature neurons in the DG of tamoxifen (T)-treated NestinCreER^T2; Bax^fl/fl; 5XFAD (NBF) mouse. Scale bar = 20 μm.

Figure 3.

Enhanced number of immature neurons reactivated in the retrieval of contextual memory in FAD mice following augmented neurogenesis. (A–G) Quantification of the number of cells recruited into the contextual memory circuit in corn oil (C)– or tamoxifen (T)-treated NestinCreER^T2; Bax^fl/fl; 5XFAD (NBF) compared to corn oil–treated NestinCreER^T2; Bax^fl/fl (C-NB) mice following memory retrieval on test day of CFC using unbiased stereology. (A) Total number of cells (Egr-1⁺). (B) New neurons recruited during memory retrieval (DCX⁺Egr-1⁺; Kruskal–Wallis test with Dunn’s post hoc test; *P =0.0109, **P = 0.0058). (C) Total number of cells incorporated in both memory acquisition and reactivated in memory retrieval (eYFP⁺Egr-1⁺; *P = 0.0162, **P = 0.0037). (D) New neurons incorporated in both memory acquisition and reactivated in memory retrieval (DCX⁺eYFP⁺Egr-1⁺; **P = 0.0015). (E) New neurons incorporated in memory retrieval but not during acquisition (DCX⁺Egr-1⁺eYFP⁻). (F) Mature granule neurons incorporated in both memory acquisition and reactivated in memory retrieval (DCX⁻Egr-1⁺eYFP⁺; P = ns). (G) The portion of new and mature granule neurons in the contextual memory engram in the three experimental groups. Pie graphs were scaled based on the total size of the engram in each experimental group. (H and I) Confocal images of eYFP⁺ and Egr-1⁺ cells in the DG of mice injected with engram kit cocktail (H), scale bar = 50 μm; and a representative new neuron (DCX⁺) co-expressing eYFP⁺ and Egr-1⁺ (I), scale bar = 20 μm.

Figure S2.

Representative confocal images of whole DG taken from brain sections of the three experimental groups. Immunohistochemistry of brain sections followed by confocal imaging captured engram cocktail–induced eYFP⁺ cells, Egr-1⁺ cells, immature neurons (DCX⁺), and their co-expression in the DG of C-NB, C-NBF, and T-NBF mice. Scale bar = 950 μm.

Figure 4.

Rescue of synaptic density in FAD mice following augmentation of neurogenesis. (A) Confocal images showing dendritic spines in eYFP⁺Egr-1⁺DCX⁺ cells in brain sections of C-NB, C-NBF, and T-NBF mice. Scale bar = 25 μm. (B–D) The density of total (B; F(2, 6) = 43.02), mushroom (C; F(2, 6) = 12.46), and thin (D; F(2, 6) = 8.332) spines was quantified in 10 μm dendrite segments of eYFP⁺Egr-1⁺DCX⁺ engram cells in brain sections of C-NB, C-NBF, and T-NBF mice using Neurolucida 360. N = 3/group and N = 25/10 μm dendrite/animal. One-way ANOVA with Tukey’s multiple comparisons post hoc test, *P < 0.05, **P < 0.005, and ***P < 0.0005. (E–G) Total spine density (E; F(2, 15) = 28.87), thin spine density (F; F(2, 15) = 21.88), and mushroom spine density (G; F(2, 15) = 19.76) as a function of distance from the cell body. Two-way ANOVA with Tukey’s multiple comparisons post hoc test, *P < 0.05.

Figure S3.

Rescue of dendritic tree of engram cells following enhanced neurogenesis in FAD mice. (A) Confocal images of DCX-expressing cells in the DG of the corn oil (C)–treated NestinCreER^T2; Bax^fl/fl (NB), corn oil (C)– or tamoxifen (TAM)-treated NestinCreER^T2; Bax^fl/fl; 5XFAD (NBF). Scale bar = 100 μm; insert = 50 μm. (B) Confocal images of tertiary dendrites of neurons incorporated in the contextual memory circuit. Scale bar = 20 μm. (C) Quantification of tertiary dendrites in engram cells, as observed in the outer molecular layer of the DG (*P < 0.05). (D–I) Validation of actuator and injection effect on freezing behavior. (D) Experimental paradigm. (E) Treatment of wild type mice with either vehicle (H₂O), 1 mg/kg CNO, or 5 mg/kg CNO results in comparable freezing behavior, suggesting lack of effect of CNO on the behavioral phenotype (N = 10; one-way ANOVA with Fisher’s LSD post hoc test F(2, 27) = 0.1289, P = 0.8796, ns). (F) Gardner-Altman plot of actuator. (G) Gardner-Altman plot of injection site. (H) Two group plot. (I) Cummings estimation plot. For F–I, colored points represent raw data, black points represent mean difference between condition compared to reference group (indicated by the gray horizontal line), and black vertical lines with distribution represent the bias-corrected and accelerated 95% confidence interval after performing bootstrap resampling 5,000 times. For H and I, vertical black lines in the top row next to colored points represent standard deviation; gaps between vertical black lines represent the mean (Ho et al., 2019).

Figure 5.

Augmentation of neurogenesis modulates synaptic plasticity of mature granule neurons participating in the engram in the DG. (A) Confocal images showing dendritic spines of mature neurons (eYFP⁺Egr1⁺NeuN⁺) in the DG of C-NB, C-NBF, and T-NBF. (B–E) Total spine density (B; F(2, 6) = 24.31), mushroom (C; F(2, 6) = 15.07), stubby (D; F(2, 6) = 3.228), and thin (E; F(2, 6) = 3.713) spines were quantified in 10 μm dendrite segments of eYFP⁺Egr1⁺NeuN⁺ engram cells in DG using Neurolucida 360. N = 3/group and N = 50/10 μm dendrite/animal were analyzed. (F and G) Total spine density (F; F(2, 75) = 90.06) and mushroom spine density (G; F(2, 75) = 22.04) as a function of distance from the cell body. Two-way ANOVA with Tukey’s multiple comparisons post hoc test, *P < 0.05, **P < 0.005, and ***P < 0.0005. (H) Percentage of each type of spines/10 μm. F(2, 18) = 22.59. (I–L) Comparison of total (I; F(1, 8) = 10.24), thin (J; F(1, 8) = 0.009), mushroom (K; F(1, 8) = 0.0068), and stubby (L; F(1, 8) = 27.82) spine density in mature and immature engram neurons in the DG of C-NBF versus T-NBF. Two-way ANOVA with Bonferroni’s multiple comparisons post hoc test, *P < 0.05, **P < 0.005, and ***P < 0.0005.

Figure 6.

New neurons are required for the formation of CFC memory in FAD. (A) Experimental design aiming at determining level of mouse freezing following actuator-induced inactivation of new neurons infected with RV-HM4-eGFP. (B–D) Confocal images of RV-HM4-eGFP–infected new neurons and c-fos⁺ cells in brain sections of tamoxifen-treated NestinCreER^T2; Bax^fl/fl; 5XFAD (T-NBF; merged image [B], RV-HM4-GFP [C], and c-fos [D]). Scale bar = 100 μm. (E) Representative image of HM4-GFP⁺ neuron in the granular cell layer of tamoxifen-treated NestinCreER^T2; Bax^fl/fl; 5XFAD (T-NBF). Scale bar = 5 μm. (F) Tamoxifen-treated NestinCreER^T2; Bax^fl/fl; 5XFAD (T-NBF) injected with RV-HM4-eGFP or RV-eGFP followed with actuator (+) or not (−) exhibit compromised memory compared to vehicle-treated (two-way ANOVA with Fisher’s LSD post hoc test for virus type: F(1, 34) = 0.8621, P = 0.3597; and for actuator: F(1, 34) = 12.99, **P = 0.0021, **P = 0.0029, respectively). (G–I) RV-HM4-eGFP-injected C-NB (G), wild type C57Bl6 (H), and T-NB (I) treated with CNO or water during CFC show similar behavior. Student’s two-tailed t test, P = 0.310 (G); Student’s two-tailed t test, P = 0.829 (H); Student’s two-tailed t test, P = 0.906 (I).

Figure 7.

In situ sequencing of immature and mature engram neurons. (A) A scheme of the in situ sequencing workflow. (B–E) Representative images of individual gene expression readout following sequencing and imaging. Slc17a6 (B), Tpbg (C), Unc5c (D), and Vipr2 (E). Scale conversion for individual images = 0.32 μm/pixel. (F–H) Cell segmentation. An image of DAPI-stained nuclei is shown with the estimated cell border boundary overlaid. (I–L) Representative example of DAPI (I), NeuN (J), eYFP⁺NeuN⁺Egr-1⁺DAPI⁺ neuron (L, white arrowhead)–stained cells that underwent in situ sequencing. (M) t-SNE plots of the in situ sequencing data. Points represent 50 μm²–binned areas covering the entire section. Areas in the DG for C-NB, C-NBF, and T-NBF are shown in blue, orange, and green, respectively, above other areas (gray). Scale bar = 1,000 μm (B–F); 225 μm (G); 55 μm (H); 175 μm (I–K); 75 μm (L).

Figure S4.

Quantification and distribution of neurons in the dentate gyrus based on in situ sequencing data. (A) Quantification of the number of eYFP⁺ new, mature and combined neurons in the DG of the three experimental groups, C-NB, C-NBF, and T-NBF, based on in situ sequencing data. t test analysis. (B) Plots showing the distribution of all eYFP⁺, eYFP⁻ excitatory, inhibitory, immature, and mature neurons in the DG.

Figure 8.

Altered engram profile in FAD is rescued following augmentation of neurogenesis. (A–C) Heatmap analysis of genes differentially expressed in eYFP⁺ compared to eYFP⁻ total neurons (A), mature neurons (B), and new neurons (C) in the DG of C-NB, C-NBF, and T-NBF mice. Plotted values are z-scored % expression. Gray scale plots the mean expression over all data sets for each gene. Only genes with P < 0.1 (nominal P value) for at least one of the eYFP⁺ versus eYFP⁻ comparisons were retained. eYFP⁻ columns are indicated by gray box below heatmap; eYFP⁺ columns indicated by yellow-green box. C-NB: blue; C-NBF: orange; T-NBF: green. (D) Intergroup directionality of the genes with the lowest P values for percent expression [(eYFP⁺ %) − (eYFP⁻ %)] in new neurons. (E) The genes with the 10 lowest P values in newborn neurons for each group, where the bar indicates the difference in percent expression [(eYFP⁺ %) − (eYFP⁻ %)].

Figure 9.

Transcript profile of the engram in FAD and following augmentation of neurogenesis. (A–C) Representative scatter plots showing eYFP⁺ neurons located within the DG and hilus for C-NB (A), C-NBF (B), and T-NBF (C) mice. Red points indicate eYFP⁺ neurons, and gray points indicate all other cells. (D–F) 2D volcano plots for each comparison. For 2D volcanos: Log₂FC vs. −log₁₀(P) for C-NB/C-NBF (D), T-NBF/C-NBF (E), and T-NBF/C-NB (F). (G–I) 3D volcano plots for each comparison: C-NB/C-NBF (G), T-NBF/C-NBF (H), and T-NBF/C-NB (I). X axis: log₂FC; y axis: −log(P value); z axis: log(cell count) = log[(# positive cells in group 1) + (# positive cells in group 2)]. (J and K) ANOSIM. Principal component analysis plots of fractional gene expression of all DG neurons (J) and of eYFP⁺ neurons (K) in the three experimental groups (for statistical analysis, see Fig. S4). (L–N) Heatmaps representing the log₂FC of the percentage of eYFP⁺ neurons expressing each gene for C-NB/C-NBF log₂FC, and T-NBF/C-NBF log₂FC, and T-NBF/C-NB log₂FC, ranked by the log₂FC for C-NB/C-NBF (L), T-NBF/C-NBF (M), and T-NBF/C-NB (N).

Figure S5.

Excitatory and inhibitory neurons in the granule cell layer of the DG. (A) Excitatory and inhibitory percentage of mature neurons for each DG (left) and mean excitatory and inhibitory neurons (middle) and percent of total mature neurons (right) for each group. (B) Mean number of neurons for each group by excitatory/inhibitory, immature/mature, and eYFP^+/−. (C) ANOSIM of fractional gene expression for each cell type between groups. (D–L) 3D volcano plots for eYFP⁺ total, mature, and immature neurons for each comparison. X axis: log₂FC; y axis: negative log(P); z-axis: log(cell count).

Figure 10.

Alteration of transcription profile of new and mature engram neurons in the DG of FAD following enhanced neurogenesis. (A–C) Representative scatter plots of eYFP⁺ new and mature engram neurons in the DG of C-NB (A), C-NBF (B), and T-NBF (C) groups. Blue: eYFP⁺ mature neurons; orange: eYFP⁺ new neurons; gray: other cells. (D–L) Mean cell count of the five genes with the lowest P values for each comparison and cell type. Blue: C-NB; orange: T-NBF; green: C-NBF. (M–O) Consistency of log₂FC direction for the union of the 20 genes with the lowest P values across the C-NB/C-NBF and T-NBF/C-NBF comparisons for all eYFP⁺ neurons (M), eYFP⁺ mature neurons (N), and eYFP⁺ new neurons (O). Blue: genes with the same log₂FC direction; orange: genes with opposite log₂FC directions; dark blue/orange: T-NBF/C-NBF comparison; light blue/orange: C-NB/C-NBF comparison. (P) Percentage of same and opposite log₂FC directions of genes shown in M–O. Significance: ***P < 0.001.