Adult-born neurons maintain hippocampal cholinergic inputs and support working memory during aging

Abstract

Adult neurogenesis is reduced during aging and impaired in disorders of stress, memory, and cognition though its normal function remains unclear. Moreover, a systems level understanding of how a small number of young hippocampal neurons could dramatically influence brain function is lacking. We examined whether adult neurogenesis sustains hippocampal connections cumulatively across the life span. Long-term suppression of neurogenesis as occurs during stress and aging resulted in an accelerated decline in hippocampal acetylcholine signaling and a slow and progressing emergence of profound working memory deficits. These deficits were accompanied by compensatory reorganization of cholinergic dentate gyrus inputs with increased cholinergic innervation to the ventral hippocampus and recruitment of ventrally projecting neurons by the dorsal projection. While increased cholinergic innervation was dysfunctional and corresponded to overall decreases in cholinergic levels and signaling, it could be recruited to correct the resulting memory dysfunction even in old animals. Our study demonstrates that hippocampal neurogenesis supports memory by maintaining the septohippocampal cholinergic circuit across the lifespan. It also provides a systems level explanation for the progressive nature of memory deterioration during normal and pathological aging and indicates that the brain connectome is malleable by experience.

Figure 1. Increased cholinergic hilar inputs in mice living with diminished neurogenesis.

(A) 2 month-old mice were exposed to focal hippocampal X-irradiation or VGCV. CAV2-GFP was injected into the dorsal hilus 5 months after X-irradiation or VGCV treatment. CAV2-GFP was also injected into a group of X-irradiated mice after two months. (B) GFP+ cells in the MS-NDB and the LC in NG+ and NG− mice (n=12 per group). (C) NG− mice without neurogenesis for 5 months showed significantly more cells projecting to the dorsal hilus than NG+ mice from the MS-NDB both ipsilateral (I) (t(22)=2.238, p= 0.0357) and contralateral (C; t(22)=3.316, p= 0.0033) to the injection site. NG− mice without neurogenesis for 2 months (n=5) and NG+ mice (n=5) showed similar connectivity from MS-NDB and LC to the dorsal hilus. NG−^TK mice with suppressed neurogenesis for 5 months showed significantly more cells projecting to the dorsal hilus than NG+^TK mice from the MS-NDB ipsilateral to the injection site (t(7)=2.927, p= 0.0221). (D) Cartoon schematic for CAV2-STOPLIGHT injections into hilus of ChAT-CRE mice. (E) Coronal sections from ChAT-Cre mice injected with CAV2-STOPLIGHT reveal exclusively GFP+ cells in the MS-NDB, and exclusively dsRed+ cells in the entorhinal cortex (EC) and the DG granule cell layer. (F, G) In the MS-NDB of NG+ and NG− animals from B and C, 80-90% of GFP labeled cells overlap with a primary marker of cholinergic cells, acetylcholineserase in 5 and 2 month groups. I= Ipsilateral, C= Contralateral. Bars represent mean ± SEM.

Figure 2. A working memory deficit emerges in mice after a prolonged reduction of adult neurogenesis.

(A) Behavior of NG− mice was assessed at 2, 4, 5 and 12 months without neurogenesis and compared to age matched NG+ controls. (B-E) Spontaneous alternation pattern (SAP) in a 4-arm spontaneous alternation task. (B,C) NG− mice after 2 and 4 months without neurogenesis had similar SAP scores of about 60% as NG+ mice (2 months n=15 per group, 4 months NG+ n=12 NG− n=13). (D) However, after 5 months without neurogenesis a deficit emerged in NG− mice where they showed SAP scores around 50% (NG+ n=10, NG− n=14;t(22)=3.835, p= 0.0009). (E) After 12 months without neurogenesis SAP scores declined to chance levels of about 44% (dotted line) NG− mice (NG+ n=7, NG− n=5; t(10)=4.06, p= 0.0023). (F) In the open field, NG− mice (n=11) without neurogenesis for 5 months and NG+ mice (n=8) showed no differences in total distance travelled in 30 min (cm) or the percentage of time spent in the center. (G) In the elevated plus maze NG− mice (n=8) without neurogenesis for 5 months and NG+ mice (n=11) showed similar open arm duration (s) and similar ratios of open to closed arm entries. (H) In a beam walking task, NG− mice (n=6) without neurogenesis for 5 months and NG+ mice (n=6) showed a similar number of total footslips and traversal latency (s). Bars represent mean ± SEM.

Figure 3. Hippocampal acetylcholine release declines in mice after a prolonged reduction of adult neurogenesis.

(A) NG− mice without neurogenesis for 12 months and NG+ mice were administered an acetlycholinesterase inhibitor physostigmine and spontaneous alternation pattern was assessed in a 4-arm spontaneous alternation task. Performance was compared across all groups (NG+ C n= 7, NG− C n=5, NG+ P n=7, NG− P n=5, main effect of group F₃₂₆ = 8.683, p= 0.0007). NG− C mice performed below all groups and at chance. Bars represent mean ± SEM, Tukey’s post hoc * p<0.05, ** p<0.01. (B) Mice were treated with X-irradiation or Sham treatment and bilaterally implanted with cannulae. In one group of mice microdialysis measurements were taken from the left dorsal hilus after 4 months without neurogenesis. In a separate group of mice microdialysis measurements were taken from the right dorsal hilus at 5 months and the left dorsal hilus at 7 months without neurogenesis. Microdialysis measurements were taken at baseline (B1-B4), while mice were in the 4-arm spontaneous alternation maze (M1-M2) and after being removed from the maze (A1-A2). (C) After 4 months without neurogenesis NG+ and NG− mice showed similar patterns of acetylcholine release that increased when animals were in the maze (NG+ n= 7, NG− n=8; main effect of time F_7,91 = 8.504, p< 0.0001). (D) After 5 months without neurogenesis NG− mice showed significantly lower acetylcholine release (n=6 per group; main effect of group F_1,10 = 5.109, p= 0.0473), main effect of time F_5,50 = 10.87, p< 0.0001, interaction effect F_5,50 = 4.198, p= 0.0029, (E) Reduced acetylcholine in NG− mice remained after 7 months with reduced neurogenesis (NG+ n= 5, NG− n=6; main effect of group F_1,9 = 7.31, p= 0.0242, main effect of time F_7,63 = 13.44, p< 0.0001, interaction effect F_7,63 = 6.904, p< 0.0001). (C) Mice without neurogenesis for 4 months demonstrated an increase in acetylcholine above baseline during the spontaneous alternation task. NG− mice showed a slightly attenuated increase in acetylcholine while performing spontaneous alternation compared to NG+ controls (main effect of time F_7,91 = 8.921, p< 0.0001, main effect of group F_7,91 =1.874, p=0.1942, time x group interaction F_7,91 = 1.786, p= 0.0996).b (D) Mice without neurogenesis for 5 months demonstrated an increase in acetylcholine above baseline during the spontaneous alternation task (main effect of time F_5,40 = 3.695, p= 0.0076) as did animals after 7 months without neurogenesis (main effect of time F_7,63 = 10,68, p< 0.0001) (E). Bars and points represent mean ± SEM, Bonferroni post hoc *p<0.05, ** p<0.01.

Figure 4. Hippocampal acetylcholine release kinetics are altered in mice after reduction of adult neurogenesis.

(A) NG− mice without neurogenesis for 2 months and NG+ mice were administered a muscarinic acetylcholine receptor antagonist scopolamine and spontaneous alternation pattern was assessed in a 4-arm spontaneous alternation task. Performance was compared across all groups (NG+ C n= 15, NG− C n=15, NG+ S n=5 and NG− S n=6, main effect of group F_3,37 = 4.977, p= 0.0053). The performance of NG− S mice was reduced compared to all groups to near chance levels (dotted line). (B) NG+ and NG− mice were injected with AAV9-hSyn-GRAB-ACh3.0 into the dorsal hilus and implanted with a fiber optic cannula. Fiber photometery was performed during spontaneous alternation testing at 3 and 6 months after neurogenesis ablation. Average signal analyzed across time (C) and in 1-minute bins. (D) revealed a robust increase in ACh signaling after mice entered the maze (broken line) but no differences between NG+ and NG− animals after 3 months of living without neurogenesis (NG+ n=8, NG− n=8; main effect of time F_9,99=8.844, p<0.0001; main effect of group F_1,11=0.008527, p=0.9281; time x group interaction F_9,99=0.4304, p=0.9158; 2-way RM ANOVA and Bonferroni post hoc). (E) Task-evoked rise time of baseline acetylcholine was also not different between groups (NG+ n=8, NG− n=8, t(14)=0.1491, p=0.8836). Both the slope and magnitude of rise in response to the maze was significantly lower in NG− mice (slope: NG+ n=8, NG− n=8, t(14)=3.434, p=0.0040; magnitude: NG+ n=8, NG− n=8, t(14)=2.357, p=0.0335; *p<0.05, ** p<0.01). (F,G) After 6 months without neurogenesis, NG− mice exhibited an attenuated task-evoked increase in acetylcholine signaling compared to NG+ animals (NG+ n=9, NG− n=9; main effect of time F_{3.505, 45.57}=9.909, p<0.0001, main effect of group F_1,13=40.18, p<0.0001, time x group interaction F_9,117=7.776, p<0.0001; 2-way RM ANOVA and Bonferroni post-hoc, *p<0.05, ** p<0.01). (H) Task-evoked rise time did not differ between NG− and NG+ mice at 6 months (NG+ n=9, NG− n=9, t(16)=1.429, p=0.1723). Both the slope and magnitude of rise was significantly lower in NG− compared to NG+ mice after 6 months (slope: NG+ n=9, NG− n=8, t(15)=6.031, p<0.0001; magnitude: NG+ n=9, NG− n=8, t(15)=3.841, p=0.0016; **p<0.01, **** p<0.0001). Bars represent mean ± SEM.

Figure 5. Reorganization of cholinergic septotemporal projection after a prolonged reduction in adult neurogenesis.

(A) In NG+ and NG− mice without neurogenesis for 5 months, CAV-GFP was injected into the dorsal hilus and CAV-Cherry was injected into the ventral hilus. (B,C) In NG+ mice (n=5) 10% of GFP+ cells in the MS-NDB showed Cherry labeling compared to 40% in NG− mice (n=4; t(7)=5.985, p= 0.0006). (D) NG− mice (n=5) show a greater number of cells projecting from the MS-NDB to the ventral hilus compared to NG+ mice (n=4, t(7)=4.754, p= 0.0021). (E) The increase in the number of cell bodies in NG− animals is in the medial NDB (t(7)=3.331, p= 0.0126) and lateral MS (t(7)=11.14, p<0.0001). Bars represent mean ± SEM. * p<0.05, ** p<0.0. Mapping hilar input reorganization in NG− mice. (F) Septohippocampal circuit was microdissected from NG− mice, immunolabeled, and subjected to iDISCO clearing. (G) 3-Dimensional reconstruction of confocal microphotographs through a unilateral septohippocampal projection. The brains were immunolabeled for mCherry, which was injected into the ventral DG and GFP, which was injected into the dorsal DG. Note the three neuronal populations 1) solely innervating the dorsal DG (green), 2) solely innervating the ventral DG (red), and 3) innervating both dorsal and ventral DG (yellow). An mCherry+, GFP+, and double+ axon was traced and all three were overlayed onto the reconstruction. (H) Axonal tracings from a MS-NDB neurons projecting to the dorsal DG (Green), ventral DG (Red), and both (Yellow). Filled circles represent cell bodies of origin. Open white circles represent axonal branching points in the fornix as exhibited in panels D-K. Dotted red line indicates region of discontinuity. (I) An mCherry+GFP+ axon forms branches in the fornix that traverse the molecular layer to innervate the dorsal hilus (arrows). (J) Boxed region magnified. (K) The same mCherry+GFP+ axon forms branches in the ventral DG (arrows). (L) Boxed region magnified. (M) A GFP+ axon in the fornix forms branches that traverse the granule cell layer to innervate the dorsal hilus. (N) Boxed region magnified. (O) An mCherry+ axon in the fornix forms branches that traverse the granule cell layer to innervate the ventral hilus. (P) Boxed region magnified. Note Abbreviations: GCL, granule cell layer; Fx, fornix; H, hilus.(Q) Structural and functional reorganization of the septohippocampal circuit in NG− mice. NG+ mice show acetylcholine release that supports working memory and cholinergic afferent organization within the septohippocampal projection. NG− mice without neurogenesis for 2 months show an emerging deficit in acetylcholine release in the hippocampus but maintain cholinergic afferent organization within the septohippocampal projection. NG− mice without neurogenesis for 5 months show significant reductions in hippocampal acetylcholine release and rewiring of cholinergic septohippocampal inputs with septal neurons that normally project to the ventral hilus innervating the dorsal hilus and increased innervation of the ventral hilus.

Figure 6. Reorganization of cholinergic septotemporal projection after a prolonged reduction in adult neurogenesis.

(A) NG+ and NG− mice were injected with an HSV encoding a cre-dependent form of the activated DREADD hM3Dq and mCherry. Spontaneous alternation was tested with and without CNO and brains were analyzed for hilar mCherry fluorescence and as well as localization and morphology of mCherry+ axons. (B-E) Representative confocal micrograph labeled for markers of Astrocytes (GFAP and S100b) and cholinergic axons (mCherry). Note the proximity of axon segments with ample varicosities surrounding astrocytes. (F) A cutout from (E) demonstrates cholinergic axonal varicosities within 400-800nM from an astrocyte process. (G) Total hilar mCherry fluorescence was greater in NG− (n=6) animals compared to NG+ (n=5) controls (t(9)=2.493, p=0.0171. (H) NG− Mice with DG projecting hM3Dq-expressing cholinergic neurons performed better at the spontaneous alternation task when they received CNO injection prior to testing compared to vehicle control (t(5)=2.134, p=0.043; n=6,6).