Microglial BMAL1 deficiency disrupts hippocampal-dependent memory in aged mice
The circadian clock declines with aging and is associated with deficits in learning and memory [6, 23, 24]. We have previously shown that macrophage circadian gene expression and immune responses decline in aging mice [25]. Aging microglia also exhibit a diminished capacity for normal function that is associated with impaired cognition [26–28]. To investigate the interaction between microglial circadian function and aging, we generated myeloid-specific Bmal1 conditional knock out mice (CD11bcre;Bmal1lox/lox; cKO) and confirmed reduced BMAL1 protein expression in peritoneal macrophages (Additional file 1: Fig. S1 A; p = 0.0178). We then subjected young and aged WT and cKO mice to a series of cognitive tasks. The novel object recognition task, a non-spatial recall test that measures discrimination between a familiar and novel object, revealed impaired memory in aged Bmal1 cKO compared to littermate WT mice (Fig. 1A). Young Bmal1 cKO mice performed similarly to WT in their examination of the novel object (Fig. 1A). Spatial learning and memory, which is encoded primarily by the hippocampus, was tested using the Barnes maze task. Primary escape latency over the 5-day training period decreased by a larger degree for young WT compared to aged WT and Bmal1 cKO mice (Fig. 1B). Aged Bmal1 cKO showed a significant increase in escape latency compared to WT mice (Fig. 1C; p = 0.0118), indicating a deficit in spatial memory. Young Bmal1 cKO performed better on the Barnes maze task compared to WT mice (Fig. 1C; p = 0.0319) consistent with a recent study showing that microglial loss of Bmal1 improved cognition in young mice [29]. In the open field test, which assesses motor function, exploratory activity, and anxiety, there were no differences in motor function but there was increased anxiety in aged Bmal1 cKO compared to WT mice (Fig. 1D-E; p = 0.0301). There were no differences in motor activity, exploration or anxiety between young Bmal1 cKO and WT mice (Fig. 1D-E). These results indicate that microglial BMAL1 deficiency accelerates age-associated deficits in spatial and non-spatial learning and memory.
Microglial Bmal1 Deficiency Suppresses Long-term Potentiation In The Ca1 Hippocampal Region Of Aged Mice
Next, we asked if neural correlates of learning and memory in aged mice were altered in Bmal1 cKO mice. First, we evaluated short-term synaptic plasticity by measuring paired-pulse ratio (PPR) in the CA1 region of the hippocampus. PPR measures the probability of activity-dependent pre-synaptic vesicular release following an action potential [30]. The ratio of the amplitude of the second response to that of the first to stimulation is inversely related to the release probability. PPR was significantly reduced at 10 ms (Fig. 2A; p = 0.0339) and increased at 50 ms (p = 0.0049), 100 ms (p = 0.0595), 200 ms (p = 0.0467), and 500 ms (p = 0.0404) in aged Bmal1 cKO compared to WT mice (Fig. 2A), consistent with perturbation of pre-synaptic short-term plasticity. The input/output (I/O) curve reflecting the efficacy of pre- and post-synaptic neurotransmitter release was significantly increased at higher stimulation strengths (50 to 65 µA) in aged Bmal1 cKO mice, consistent with disrupted basal synaptic transmission (Fig. 2B; p = 0.0231).
We next examined long-lasting, activity-dependent changes in synaptic efficacy from the CA3 to CA1 Schaffer collateral pathway in the hippocampus to assess post-synaptic plasticity. We found significant impairment in long-term potentiation (LTP) in hippocampal slices from aged Bmal1 cKO compared to WT mice across 70 minutes of recording (Fig. 2C; main effect of genotype for 70 minutes post-induction: F1,18= 20.77, p = 0.0002). This impairment was apparent immediately after LTP induction (Fig. 2C; p < 0.0001). Thereafter, LTP in aged Bmal1 cKO mice slowly decayed to ~ 35% potentiation, whereas WT mice stabilized LTP at ~ 80% with respect to baseline.
Calcium/calmodulin-dependent protein kinase II (CaMKII) is critical for the induction of LTP, and inhibition of CAMKII prevents LTP [31–34]. To investigate the candidate mechanisms underlying the reduced LTP in aged Bmal1 cKO mice, we tested whether microglial Bmal1 deficiency might alter hippocampal levels of CAMKII or phosphorylated CAMKII. Quantitative immunoblotting showed increased total CAMKII expression and CAMKII phosphorylation, however, there was no difference in the ratio of phosphorylated CAMKII/total CAMKII between aged Bmal1 cKO and WT mice (Fig. 2D-E; p = 0.6433). The AMPA receptor subunit, GluA1, mediates LTP maintenance by enhancing AMPA receptor conductance and anchoring at the synapse [35]. To reconcile the increase in I/O ratio with decreased LTP induction and amplitude, we assessed levels of GluA1 expression. Aged Bmal1 cKO mice exhibited decreased GluA1 signal in the CA1 hippocampal region compared to WT mice (Fig. 2F-H; p = 0.0040). These data suggest that microglial BMAL1 deficiency alters hippocampal plasticity in aged mice by a mechanism involving reduced insertion of AMPA receptors in dendritic spines in aged mice.
Bmal1-deficient Microglia Increase Synaptic Density In The Ca1 Hippocampal Region Of Aged Mice
Microglia actively maintain neural circuitry and plasticity through selective pruning of synapses [36–38]. Given the cognitive deficits and altered synaptic plasticity observed in aged Bmal1 cKO mice, we measured expression levels of presynaptic and postsynaptic proteins, SNAP25 and PSD95, respectively, in the CA1 hippocampal region. Aged Bmal1 cKO mice showed higher immunoreactivity for both PSD95 (p = 0.0487) and SNAP25 (p = 0.0439) compared to aged WT mice (Fig. 3A-B). No significant differences were observed between young Bmal1 cKO and WT mice (Additional file 1: Fig. S1 B-C). These results were confirmed using quantitative immunoblotting for PSD95 (p = 0.0010) and SNAP25 (p = 0.0441) (Fig. 3C-D). Assessment of synaptic morphology and density by Golgi staining confirmed increased dendritic spine density in aged Bmal1 cKO compared to WT mice (Fig. 3E-F; p = 0.0245). Aged Bmal1 cKO mice had a higher density of filopodia-like (p = 0.00497), stubby (p = 0.0185) and mushroom spines (p = 0.01498) compared to WT mice (Fig. 3G), indicating a persistence of immature synapses in aged Bmal1 cKO mice. These data suggest a significant impairment in pruning of synapses resulting from microglial BMAL1 deficiency in aged mice.
Microglial Bmal1 Deficiency Decreases C1q Expression And Synaptic Pruning In Aged Ca1 Hippocampus
To investigate mechanisms that may underlie defective synaptic pruning in aged Bmal1 cKO mice, we measured expression of C1q, an opsonin that is deposited on synaptic terminals and promotes synapse removal through interaction with the microglial C3 receptor, CR3 [37, 39, 40]. C1q is primarily expressed by microglia and increases with age, although neuronal expression has also been reported [39, 41, 42]. In the CA1 hippocampal region of aged Bmal1 cKO mice, C1q colocalized with PSD95 but its abundance was significantly reduced (Fig. 4A-B; p = 0.0010) in contrast to PSD95 levels (Fig. 4A-B; p = 0.0012) which were significantly elevated (Fig. 3A-B). CD68, a microglial lysosomal protein, was observed in close proximity to PSD95/C1q and was also decreased in aged Bmal1 cKO (Fig. 4A-B; p = 0.0427), suggesting reduced microglial production of both C1q and CD68. There were no differences in C1q, PSD95 or CD68 levels between young Bmal1 cKO and WT mice (Additional file 2: Fig. S2 A-B).
C1q promotes the formation of C3 convertase through activation of complement proteins, C4 and C2. C3 convertase cleaves C3 to produce C3a, a pro-inflammatory mediator, and C3b, an opsonin that initiates phagocytosis via the CR3 receptor on microglia [43]. We observed a reciprocal increase in C3 immunoreactivity in the CA1 hippocampal region of aged Bmal1 cKO mice compared to WT (Fig. 4C-E; p = 0.0208), suggesting that C3 may accumulate in response to reduced levels of C3 convertase.
Next, we assessed microglial engulfment of synapses by measuring colocalization of PSD95 and CD68 within IBA1 + microglia. Superresolution microscopy was used to quantify PSD95 within CD68 + structures in IBA + microglia in order to identify engulfed synapses. Aged Bmal1 cKO microglia demonstrated a significant decrease in internalized PSD95 compared to WT (Fig. 5A-C; p = 0.0469; Additional file 3: Video S1), consistent with decreased synapse engulfment. Taken together, these data indicate that microglial loss of BMAL1 decreases C1q deposition on synapses and synaptic engulfment.
Microglia deficient in BMAL1 exhibit microgliosis and reduced expression of lysosomal proteins in the CA1 hippocampal region
To further characterize Bmal1-deficient microglia, we measured IBA1 and CD68 expression in the CA1 hippocampal region. A significant increase was observed in IBA1 immunoreactivity (p = 0.0028) and number of microglia (p = 0.0126) in aged Bmal1 cKO compared to WT mice (Fig. 6A-C). Quantitative immunoblot analysis confirmed the increase in IBA1 expression in aged Bmal1 cKO mice (Fig. 6D-E; p = 0.0067). Consistent with our previous observation shown in Fig. 4B, CD68 was significantly decreased in aged Bmal1 cKO mice (Fig. 6A-C; p = 0.0494). The proportion of IBA1 + cells immunoreactive for CD68 was also significantly reduced, indicating an overall decrease in CD68 expression (Fig. 6A-C; p = 0.0197). Morphological analysis of microglial complexity revealed decreases in branch length (p = 0.0489), branch number (p = 0.0385) and number of junctions (p = 0.0398) in aged Bmal1 cKO compared to WT mice (Fig. 6F-G). No differences between young WT and Bmal1 cKO mice in either IBA1 or CD68 expression, proportion of IBA1 + cells immunoreactive for CD68, number of microglia, or morphological parameters were observed (Additional file 4: Fig. S3 A-E).
CD68 is a member of the lysosomal/endosomal-associated transmembrane glycoprotein (LAMP) family whose expression is upregulated with inflammatory stimuli such as bacterial lipopolysaccharide (LPS). Microglial CD68 expression and lysosomal dysfunction increases with aging and is associated with neurodegeneration [44–47]. We investigated additional lysosomal proteins including LAMP1, whose expression was similarly reduced in the CA1 hippocampal region of aged Bmal1 cKO mice (Additional file 5: Fig. S4 A-B; p = 0.0335), as was p62 (Additional file 5: Fig. S4 C-D; p = 0.0483) which is required for the formation of the omegasome, a precursor structure of the autophagolysosome [48, 49]. These data suggest that although microglia in aged Bmal1 cKO mice are proliferating and are morphologically activated, their reduced levels of CD68, LAMP1 and p62 suggest a deficit in lysosomal function.
Deficiency Of Bmal1 In Microglia Alters The Omegasome And Viral Response Genes
To identify biological pathways affected by microglial loss of BMAL1, we performed RNA sequencing on CD11b + microglia isolated from young and aged, WT and Bmal1 cKO mice. Principal component analysis (PCA) showed a large difference in transcriptional profiles driven by age (PC1) with variability among the aged mice (PC2) (Fig. 7A). No distinction between genotypes in either aged group was observed at the full transcriptome level. At the gene level, differential expression analysis (DEA) revealed 150 differentially expressed genes (DEGs) between Bmal1 cKO and WT aged mice. 4/19 (21%) down-regulated and 1/144 (< 1%) upregulated genes overlapped between young and aged Bmal1 cKO mice (Fig. 7B). Of those, 13/150 (9%) were down-regulated and 137/150 (91%) were up-regulated (Fig. 7C). 18 DEGs were identified between young Bmal1 cKO and WT mice, with 10/18 (56%) up-regulated and 8/18 (44%) down-regulated genes in Bmal1 cKO (Additional file 6: Fig. S5). Gene ontology (GO) analysis showed significant enrichment only in downregulated genes for biological functions that include innate immunity, autophagy and viral response (Additional file 7: Table S1). Interestingly, GO enrichment for cellular component was significant for the omegasome pathway, in line with our observations of decreased p62 protein levels in aged Bmal1 cKO microglia.
Microglial Deficiency Of Bmal1 Disrupts The Sleep-wake Cycle
Sleep is a major regulator of microglial function, and neuron-microglia interactions change during wake, NREM and REM states, reflecting changes in synapse remodeling [50]. Indeed, the sleep-wake cycle drives changes in synaptic ultrastructure and the daily dynamics of synaptic protein phosphorylation relevant to spine dynamics [51, 52]. The circadian clock plays a substantial role in sleep architecture, and global loss of BMAL1 disrupts sleep-wake patterns [53, 54]. Therefore, we carried out continuous 24-hour electroencephalogram/electromyogram (EEG/EMG) recordings to determine baseline sleep-wake patterns in aged Bmal1 and WT mice over 3 days. A trend towards increased wakefulness was observed in the Bmal1 cKO mice, (Fig. 8A-C). During the non-active light phase/subjective night, aged Bmal1 cKO mice spent less time in REM sleep (Fig. 8C; p = 0.0108) compared to WT mice. Similarly, during the active dark phase/subjective day, aged Bmal1 cKO mice spent more time awake (Fig. 8A; p = 0.0366) and less time in REM sleep (Fig. 8C; p = 0.0253).
During both the light and dark phases, aged Bmal1 cKO mice exhibited an overall trend towards prolonged sleep-wake cycling. For example, during the non-active light phase/subjective night, the average duration of individual wake (Fig. 8D; p = 0.0051), NREM (Fig. 8E; p = 0.0055), and REM (Fig. 8F; p = 0.0034) bouts was increased, while the total number of wake (Fig. 8G; p = 0.0068), NREM (Fig. 8H; p = 0.0061), and REM (Fig. 8I; p = 0.0009) bouts, and transitions between all sleep-wake states (Fig. 8J; p = 0.0032) were decreased in the aged Bmal1 cKO mice compared to WT. Similarly, during the active dark phase/subjective day, the average wake bout duration was increased (Fig. 8D; p = 0.0154) and the total number of wake bouts (Fig. 8G; p = 0.0064), NREM bouts (Fig. 8H; p = 0.0063), REM bouts (Fig. 8I; p = 0.0101), and state transitions (Fig. 8J; p = 0.0038) was decreased in the aged Bmal1 cKO mice. Additionally, the average REM bout latency (time from NREM onset to REM onset) was increased in the aged Bmal1 cKO mice during both the light (Fig. 8K; p = 0.0062) and dark (Fig. 8K; p = 0.0252) phases compared to WT.
Next, we performed an EEG spectral power analysis by averaging the relative values of delta 1 (0.5-2 Hz), delta 2 (2.5–4.5 Hz), theta (5–9 Hz), alpha (6–10 Hz) and beta (15.5–20 Hz) activity during wake, NREM, and REM sleep during the non-active light phase/subjective night. Aged Bmal1 cKO mice exhibited a relative shift in spectral power characteristics during the wake period, with delta 1 power significantly decreased (p = 0.0084), and theta power increased (p = 0.0144), compared to aged WT mice (Additional file 8: Fig. S6 A). No differences were observed during NREM and REM periods (Additional file 8: Fig. S6 B-C). In humans, increased delta power and decreased theta power during wake has been correlated with subjective feelings of sleepiness [55, 56], and in mice the wake-promoting drug modafinil decreases delta power and increases theta power during wakefulness [57, 58]. Taken together, the observed decreased delta and increased theta power, which are signatures of lower sleep pressure, suggests increased wakefulness in the aged Bmal1 cKO mice. Interestingly, the delta 2 EEG band, which is also associated with homeostatic sleep pressure [59] was unaffected in the Bmal1 cKO mice (Additional file 8: Fig. S6).
These results indicate increased wakefulness during both light and dark periods in aged Bmal1 cKO mice and suggest a role of circadian-tuned microglia in regulating overall neuronal excitability during the sleep-wake cycle.
Microglial Deficiency Of Bmal1 Cko Alters Response To Sleep Deprivation
Core clock genes influence compensatory mechanisms induced by sleep deprivation [53, 60]. Given that the loss of Bmal1 in microglia in aged mice was sufficient to disrupt the sleep-wake cycle and change sleep architecture, we explored the effects of sleep deprivation in these mice. We subjected aged Bmal1 cKO and WT mice to a 4-hour sleep deprivation protocol during the non-active light phase/subjective night and then assessed sleep recovery. During the light phase recovery period, a rebound effect was observed in both the WT and Bmal1 cKO mice, with both groups spending less time awake (Fig. 9A and D; WT: p = 0.0004; Bmal1 cKO: p < 0.0001) and more time in NREM sleep (Fig. 9B and E; WT: p = 0.0010; Bmal1 cKO: p = 0.0004) compared to their baseline levels from the same timeframe. An increase in the time spent in REM sleep was observed in aged Bmal1 cKO mice (Fig. 9C and F; p = 0.0445) compared to WT mice. During the dark phase recovery period, WT mice continued to spend less time awake (Fig. 9A and D; p < 0.0001) and more time in NREM sleep (Fig. 9B and E; p = 0.0001) compared to baseline, whereas the amount of time Bmal1 cKO mice spent in wake, NREM, and REM sleep did not differ from baseline (Fig. 9A and F).
The cummulative amount of time spent in wake, NREM, and REM sleep did not differ between groups during the light phase recovery period (Fig. 9G-H). However, during the dark phase recovery period, the cummulative time spent by WT mice in wake was decreased (Fig. 9G-H; p < 0.0001), and increased in NREM (Fig. 9G-H; p < 0.0001). Unlike WT mice, Bmal1 cKO mice exhibited no compensatory changes to sleep deprivation in wake and NREM states (Fig. 9G-H). While the cumulative time spent by WT mice in REM did not differ from baseline after sleep deprivation in both the light and dark phase recovery periods, the cumulative time spent in REM was significantly increased from baseline in Bmal1 cKO mice during the light phase (Fig. 9I; p = 0.0007) and dark phase (Fig. 9I; p < 0.0001) recovery periods.
Taken together, these results suggest that the homeostatic response to sleep deprivation is markedly blunted in aged Bmal1 cKO mice.