Adult neurogenesis improves spatial information encoding in the mouse hippocampus

Abstract

Adult neurogenesis is a unique form of neuronal plasticity in which newly generated neurons are integrated into the adult dentate gyrus in a process that is modulated by environmental stimuli. Adult-born neurons can contribute to spatial memory, but it is unknown whether they alter neural representations of space in the hippocampus. Using in vivo two-photon calcium imaging, we find that male and female mice previously housed in an enriched environment, which triggers an increase in neurogenesis, have increased spatial information encoding in the dentate gyrus. Ablating adult neurogenesis blocks the effect of enrichment and lowers spatial information, as does the chemogenetic silencing of adult-born neurons. Both ablating neurogenesis and silencing adult-born neurons decreases the calcium activity of dentate gyrus neurons, resulting in a decreased amplitude of place-specific responses. These findings are in contrast with previous studies that suggested a predominantly inhibitory action for adult-born neurons. We propose that adult neurogenesis improves representations of space by increasing the gain of dentate gyrus neurons and thereby improving their ability to tune to spatial features. This mechanism may mediate the beneficial effects of environmental enrichment on spatial learning and memory.

Fig. 1. Environmental enrichment increases spatial information encoding in the DG but this effect is blocked in mice with ablated adult neurogenesis.

a Experimental timeline, including irradiation, surgery, housing environment, and in vivo imaging. b Example of calcium imaging field of view. Cells that were active during a recording are shaded in color. c Immunofluorescence labelling of DCX positive neurons in the DG of non-irradiated (top) and irradiated (bottom) mice. Arrows denote neurogenic subgranular layer in the upper and lower leaves of the DG where DCX-expressing cells can be found. d Number of DCX-expressing neurons in imaged mice from regular cage (RC), enriched environment (EE) groups and corresponding irradiated groups (Irr+RC, Irr+EE). (n_RC = 4 mice, n_Irr+RC = 4 mice, n_EE = 4 mice, n_Irr+EE = 3 mice, average of three 40 µm slices per mouse, Welch ANOVA test with Dunnett’s T3 multiple comparisons, RC vs Irr+RC **p = 0.008, RC vs EE **p = 0.009, Irr+RC vs Irr+EE ns p > 0.999, EE vs Irr+EE **p = 0.004). e Example calcium traces (top) and respective position of animal on the treadmill (bottom). f Accuracy in decoding position of mouse on treadmill from calcium traces (RC vs Irr+RC: *p = 0.006, EE vs Irr+EE: ****p = 2.15 × 10⁻⁵, **RC vs EE: p = 0.03, ANOVA, Holm-Sidak correction for multiple comparisons, n = 4 mice, 42 neurons subsampled per mouse. Dotted line is chance performance level (5%). Error bars represent +/- SEM in all plots. All scale bars = 50 µm.

Fig. 2. Ablating adult neurogenesis decreases spatial information content at the single cell level.

a–d Population Fisher information determined with noise correlations and after random shuffling to disrupt noise correlations (RC vs RC w/o NC: p = 0.057 n.s, n = 5 mice per group, EE vs EE w/o NC: p = 0.15 n.s, n = 5 mice per group, Irr+RC vs Irr+RC w/o NC: p = 0.20 n.s., n = 4 mice per group, Irr+EE vs Irr+EE w/o NC: p = 0.014, n = 4 mice per group, Mann-Whitney U test). e Single-cell spatial information content determined using Fisher information (FI) (RC vs Irr+RC: ****p < 1/100000, n_RC = 5 mice, 277 neurons, n_Irr+RC = 4 mice, 253 neurons, EE vs Irr+EE: ****p < 1/100000, n_EE = 5 mice, 201 neurons, n_Irr+EE = 4 mice, 412 neurons, statistical analysis: bootstrap (two-sided) and Bonferroni correction for multiple comparisons (see methods). Error bars are mean ± SEM. f Distribution of spatial information content across the imaged neurons. g Distance between two positions that DG single-cells are able, on average, to discriminate correctly 70% of the time (RC vs Irr+RC: **p = 0.0038 n_RC = 5 mice, n_Irr+RC = 4 mice, EE vs Irr+EE: **p = 0.0012, n_EE = 5 mice, n_Irr+EE = 4 mice, statistical analysis: ANOVA, Holm-Sidak correction for multiple comparisons). Error bars represent +/- SEM in all plots.

Fig. 3. Ablating adult neurogenesis reduces tuning specificity and activity.

a Example calcium traces of cells with low (yellow) and high (blue) tuning indices. b Tuning vectors from cells in A plotted in polar coordinates. c Tuning indices of cells in non-irradiated and irradiated groups (RC vs Irr+RC: ****p < 1/100000, n_RC = 5 mice, 277 neurons, n_Irr+RC = 4 mice, 253 neurons, EE vs Irr+EE: ****p < 1/100000, n_EE = 5 mice, 201 neurons, n_Irr+EE = 4 mice, 412 neurons). d Activity measured as integrated calcium traces normalized to distance travelled (RC vs EE: ****p < 1/100000, n_RC = 5 mice, 277 neurons, n_EE = 5 mice, 201 neurons, RC vs Irr+RC: **p < 0.00228, n_RC = 5 mice, 277 neurons, n_Irr+RC = 4 mice, 253 neurons, EE vs Irr+EE: ****p < 1/100000, n_EE = 5 mice, 201 neurons, n_Irr+EE = 4 mice, 412 neurons). e Schematic of tuning curve properties of individual cells fitted with Von Mises function. f Cross-validated goodness of fit (R²) of tuning curves (RC vs EE: *p = 0.01617, n_RC = 5 mice, 277 neurons, n_EE = 5 mice, 201 neurons, RC vs Irr+RC: ****p < 1/100000, n_RC = 5 mice, 277 neurons, n_Irr+RC = 4, 253 neurons, EE vs Irr+EE: ****p < 1/100000, n_EE = 5 mice, 201 neurons, n_Irr+EE = 4 mice, 412). g Peak width of tuning curves of well-fitted cells. (RC vs Irr+RC: p = 0.516, n_RC = 5 mice, 173 neurons, n_Irr+RC = 4 mice, 70 neurons, EE vs Irr+EE: p = 1.159, n_EE = 4 mice, 113 neurons, n_Irr+EE = 4 mice, 85 neurons). h Peak amplitude of well-fitted cells (RC vs Irr+RC: p < 1/100000, n_RC = 5 mice, 173 neurons, n_Irr+RC = 4 mice, 70 neurons, EE vs Irr+EE: p = 0.00237, n_EE = 4 mice, 113 neurons, n_Irr+EE = 4 mice, 85 neurons). Statistical analyses: Bootstrap (two-sided) and Bonferroni correction for multiple comparisons. Error bars represent mean ± SEM in all plots.

Fig. 4. Acute chemogenetic silencing of ABNs decreases spatial information content in the DG.

a Experimental timeline. b Immunofluorescence images of HA-tag positive neurons (red) and DAPI labelled nuclei (blue). c Schematic of contextual fear conditioning (CFC) task and images of context A and context B. d Percentage of time spent freezing in shocked and novel contexts (hM4Di- Ctx A vs Ctx B: **p = 0.0015, n_{Ctx A} = 5 mice, n_{Ctx B} = 5 mice, hM4Di+ Ctx A vs Ctx B, *p = 0.0484, n_{Ctx A} = 5 mice, n_{Ctx B} = 5 mice). e Discrimination index of freezing between contexts (hM4Di- (control) vs hM4Di + : *p = 0.0159, n_hM4Di- = 5 mice, n_hM4Di+ = 5 mice). Statistical analyses c) and d): Two-sided Mann-Whitney test, no multiple comparisons correction. f, g Fisher information (Baseline vs CNO: **p = 0.00345, Baseline vs CNO (hM4Di- control): p = 0.08291,). h, i Activity (Baseline vs CNO: ****p = 0.00005, Baseline vs CNO (hM4Di- control): *p = 0.01964). j, k Tuning index (Baseline vs CNO: *p = 0.0173, Baseline vs CNO (hM4Di- control): p = 0.13226). l, m Goodness-of-fit (R²) of the Von Mises function to the tuning curves of all cells (Baseline vs CNO: p = 0.44973, Baseline vs CNO (hM4Di- control): p = 0.47035). Sample size f), h), j) and l): n_Baseline = 9 mice, 471 neurons, n_CNO = 9 mice, 461 neurons. Sample size g), i), k) and m): n_Baseline = 5 mice, 228 neurons, n_CNO = 5 mice, 217 neurons. n, o) Goodness-of-fit of well-fitted (R² > 0.5) cells (Baseline vs CNO: **p = 0.00776, Baseline vs CNO (hM4Di- control): p = 0.47035). p, q Peak width of well-fitted cells (Baseline vs CNO: p = 0.49976, Baseline vs CNO (hM4Di- control): p = 0.17618). r, s Peak amplitude of well-fitted cells (Baseline vs CNO: **p = 0.00636, Baseline vs CNO (hM4Di- control): p = 0.3244). Sample size n), p) and r): n_Baseline = 9 mice, 168 neurons, n_CNO = 9 mice, 161 neurons. Sample size o), q) and s): n_Baseline = 5 mice, 45 neurons, n_CNO = 5 mice, 60 neurons. Statistical analyses f) thru s): Two-sided bootstrap. Error bars represent mean ± SEM in all plots. All scale bars = 50 µm.