Disruption of NREM sleep and sleep-related spatial memory consolidation in mice lacking adult hippocampal neurogenesis

Abstract

Cellular plasticity at the structural level and sleep at the behavioural level are both essential for memory formation. The link between the two is not well understood. A functional connection between adult neurogenesis and hippocampus-dependent memory consolidation during NREM sleep has been hypothesized but not experimentally shown. Here, we present evidence that during a three-day learning session in the Morris water maze task a genetic knockout model of adult neurogenesis (Cyclin D2^-/-) showed changes in sleep macro- and microstructure. Sleep EEG analyses revealed a lower total sleep time and NREM fraction in Cyclin D2^-/- mice as well as an impairment of sleep specific neuronal oscillations that are associated with memory consolidation. Better performance in the memory task was associated with specific sleep parameters in wild-type, but not in Cyclin D2^-/- mice. In wild-type animals the number of proliferating cells correlated with the amount of NREM sleep. The lack of adult neurogenesis led to changes in sleep architecture and oscillations that represent the dialog between hippocampus and neocortex during sleep. We suggest that adult neurogenesis-as a key event of hippocampal plasticity-might play an important role for sleep-dependent memory consolidation and modulates learning-induced changes of sleep macro- and microstructure.

Figure 1

Experimental design. EEG/EMG implantation took place at day-10, mice were then allowed 9–10 days of recovery including a 48 h period of habituation to the recording setup. This was followed by a 24 h baseline EEG/EMG recording. EEG/EMG signals were continuously recorded for the next 3 days (EEG day 1–3). Each morning at 6 a.m. on day 1–3 the Morris water maze (MWM) was performed. These 3 days are summarized as the learning period. Mice were perfused for histological analysis afterwards.

Figure 2

Cell proliferation. WT red, Cyclin D2^−/− blue. Ki67⁺ cells per DG as a measure for neurogenesis levels (proliferating cells) in both groups (n = 10 per group; two-sided t-test; ***: p < 0.001).

Figure 3

MWM performance. WT red, Cyclin D2^−/− blue. (A) Mean route efficiency in % (100% = direct swimming path) for each day (6 trials per day) in both groups. (B) Discrimination and ranking of different search strategies. Hippocampus-independent strategies were ranked with 0 points and only hippocampus-dependent strategies (direct search, focal search, and direct swimming) were ranked with 1, 2 or 3 points. (C) Proportional appearance in % of different search strategies over all trials during the 3 days of learning the MWM task in both groups. (D) Mean strategy ranking for each day (6 trials per day) in both groups. (n = 10 for each group, 6 trails per day; one-way repeated-measure ANOVA with post-hoc Bonferroni-adjusted two-sided t-test as appropriate; *: p < 0.05; **: p < 0.01; ***: p < 0.001; data presented as mean ± 1 SEM).

Figure 4

Sleep macrostructure. WT red, Cyclin D2^−/− blue. Each panel shows the group comparison per 24 h recording at baseline and during the learning period (mean of days 1–3). (A) Total sleep time (TST) in min. (B) NREM sleep duration in min. (C) REM sleep duration in min. (D) NREM to REM sleep ratio. (E) Arousal index defined as arousals per minute. (n = 10 per group; two-sided unpaired t-test for group comparison and two-sided paired t-test for baseline learning comparison; *: p < 0.05; **: p < 0.01; ***: p < 0.001).

Figure 5

Sleep microstructure and arousal index correlations. WT red, Cyclin D2^−/− blue. Each sub-figure shows the group comparison per 24 h recording at baseline and during the learning period (mean of days 1–3). (A) Slow oscillation count. (B) Sleep spindle count. (C) Sleep spindle density in count per min. (D) Modulation index of phase-amplitude coupling (PAC). (E) Correlation of sleep spindle count with arousal index (r = − 0.612; p = 0.004). (F) Correlation of arousal index NREM sleep duration (r = − 0.573; p = 0.008). (G) Correlation of arousal index with NREM to REM sleep ratio (r = − 0.722; p < 0.001). (n = 10 per group; two-sided unpaired t-test for group comparison and two-sided paired t-test for baseline learning comparison; *: p < 0.05; **: p < 0.01; Pearson’s correlation coefficient with two-sided test).

Figure 6

Correlation of Ki67⁺ cells and macroscopic sleep parameters during learning period in WT animals. (A) Correlation of Ki67⁺ cell number with NREM and REM sleep fraction. (B) Correlation of Ki67⁺ cell number with NREM to REM sleep ratio (r = 0.837; p = 0.003). (n = 10; Pearson’s correlation coefficient with two-sided test).

Figure 7

Correlation of different sleep parameters (mean d1-3) and route efficiency (mean d2-3). WT red, Cyclin D2^−/− blue. (A) Correlation of NREM sleep duration in min with route efficiency in % (WT: r = 0.936; p < 0.001; Cyclin D2^−/−: r = − 0.036; p = 0.922). (B) Correlation of sleep spindles count with route efficiency in % (WT: r = 0.876; p = 0.001; Cyclin D2^−/: r = − 0.071; p = 0.846). The regression line is only shown for the WT animals, as no association was found for the Cyclin D2^−/− mice. (Pearson’s correlation coefficient with two-sided test).