Microglia modulate stable wakefulness via the thalamic reticular nucleus in mice

Abstract

Microglia are important for brain homeostasis and immunity, but their role in regulating vigilance remains unclear. We employed genetic, physiological, and metabolomic methods to examine microglial involvement in the regulation of wakefulness and sleep. Microglial depletion decreased stable nighttime wakefulness in mice by increasing transitions between wakefulness and non-rapid eye movement (NREM) sleep. Metabolomic analysis revealed that the sleep-wake behavior closely correlated with diurnal variation of the brain ceramide, which disappeared in microglia-depleted mice. Ceramide preferentially influenced microglia in the thalamic reticular nucleus (TRN), and local depletion of TRN microglia produced similar impaired wakefulness. Chemogenetic manipulations of anterior TRN neurons showed that they regulated transitions between wakefulness and NREM sleep. Their firing capacity was suppressed by both microglial depletion and added ceramide. In microglia-depleted mice, activating anterior TRN neurons or inhibiting ceramide production both restored stable wakefulness. These findings demonstrate that microglia can modulate stable wakefulness through anterior TRN neurons via ceramide signaling.

Fig. 1. Microglia are essential for stable wakefulness at night.

a Microglia were noninvasively depleted by injecting DT (i.p.) into CX3CR1^Cre-ERT2/+:R26^iDTR/+ transgenic mice. EEG/EMG signals were recorded before and after injection for 6 h during the middle of day and night periods. Three groups of mice were included: CX3CR1^Cre-ERT2/+:R26^iDTR/− mice with tamoxifen (TAM) and DT injection (MG^DTR− DT), CX3CR1^Cre-ERT2/+:R26^iDTR/+ mice with TAM and vehicle injection (MG^DTR+ Veh.), CX3CR1^Cre-ERT2/+:R26^iDTR/+ mice with TAM and DT injection (MG^DTR+ DT). b Four weeks after TAM treatment, DT injection sharply decreased CX3CR1-expressing cells in the brain (p = 1.1 × 10⁻¹⁰), with no effect on kidney and spleen of CX3CR1^Cre-ERT2/+:R26^iDTR/+ mice. Control, 3 mice; n = 14 images from brain, n = 20 images from kidney, n = 22 images from spleen; DT, 2 mice; n = 10 images from brain, n = 16 images from kidney, n = 18 images from spleen, Each dot represents cell density obtained from one image. Scale bar, 20 µm. c DT depleted ~80% of microglia (cells co-labeled with GFP and Iba1) across the whole brain (day 0 vs day 4, p < 0.01 for all examined regions; day 0 vs day 6, p < 0.01 for all examined regions). Representative images show cortical microglia. Microglial density was normalized to day 0 values. CTX cortex, VLPO ventrolateral preoptic nucleus, TRN thalamic reticular nucleus, LH lateral hypothalamus, VLPAG ventrolateral periaqueductal gray, PZ parafacial zone. At least four images were collected for each brain region of one mouse, each dot represents cell density obtained from one mouse. day 0, n = 5 mice; day 4, n = 4 mice; day 6, n = 7 mice. Scale bar, 50 µm. d, e A hypnogram measured from one mouse during the day (top, left) and at night (top, right) before and after microglial depletion (W wakefulness, N NREM sleep, R REM sleep). Sleep architecture (bottom in d: total duration for each state; e: bout duration for each state) was compared before (day 0) and after (day 4) DT or vehicle injection for each animal, and the difference (day 4–day 0) was further compared between three groups of mice. Each dot represents the difference between day 0 and day 4 in one mouse. For ∆ % total duration at night in d; W: p = 0.025 for MG^DTR− DT vs MG^DTR+ DT; p = 0.005 for MG^DTR+ Veh. vs MG^DTR+ DT; N: p = 0.008 for MG^DTR− DT vs MG^DTR+ DT; p = 0.002 for MG^DTR+ Veh. vs MG^DTR+ DT. For ∆ bout duration in e; W: p = 0.01 for MG^DTR− DT vs MG^DTR+ DT; p = 0.003 for MG^DTR+ Veh. vs MG^DTR+ DT. MG^DTR+ DT, n = 8 mice; MG^DTR+ Veh., n = 5 mice. MG^DTR− DT, n = 5 mice. f Microglial depletion shortened wakefulness stability (bottom) but not quick arousals (top) at night. g Microglial depletion enhanced the number of wakefulness to NREM sleep (WN) and NREM sleep-to-wakefulness (NW) transitions at night, the transition from NREM sleep to REM sleep (NR) and REM sleep-to-wake were still intact (RW). WN: p = 0.005 for MG^DTR− DT vs MG^DTR+ DT; p = 0.008 for MG^DTR+ Veh. vs MG^DTR+ DT; NW: p = 1.9 × 10⁻⁴ for MG^DTR− DT vs MG^DTR+ DT, p = 6.2 × 10⁻⁴ for MG^DTR+ Veh. vs MG^DTR+ DT. MG^DTR+ DT, n = 8 mice; MG^DTR+ Veh., n = 5 mice. MG^DTR− DT, n = 5 mice. *p < 0.05; two-sided unpaired t-test for b, c; one-way ANOVA with Fisher’s post-hoc test for d, e, and g; two-sided Kolmogorov–Smirnov test for f. Data are reported as mean ± SEM. See also Supplementary Figs. 1–3.

Fig. 2. Circadian variation of ceramide correlated with sleep-wake behavior.

a Subcortical brain regions were collected during the day when mice were asleep (Sleep_day, n = 6 mice) and at night when mice were awake (Wake_night, n = 6 mice) for metabolomic analysis. b A heatmap of lipid species showing significant Sleep_day/Wake_night expression differences. The p value for each lipid was obtained by compare the difference between Sleep_day and Wake_night. LCFAs, long chain saturated fatty acids, p = 0.019; LC-MUFAs long chain monounsaturated fatty acids, p = 0.006; LC-PUFAs long chain polyunsaturated fatty acids, p = 0.016; DA dicarboxylic fatty acids, p = 0.01; LCACs long chain saturated acyl carnitines, p = 0.005; MUACs monounsaturated acyl carnitines, p = 0.004; PUACs polyunsaturated acyl carnitines, p = 0.022; Carnitine, p = 0.036; ACs acylcholines, p = 0.048; HFAs dihydroxy fatty acids, p = 0.008; EC endocannabinoids, p = 0.04; PLP phospholipids, p = 0.018; PC phosphatidylcholines, p = 0.01; PE phosphatidylethanolamine, p = 0.019; PS phosphatidylserine, p = 0.024; PG phosphatidylglycerol, p = 0.019; LP lysophospholipid, p = 0.026; PL plasmalogen, p = 0.034; LyPL lysoplasmalogen, p = 0.044; MG monoacylglycerol, p = 0.021; DAG diacylglycerol, p = 0.031; CERs ceramides, p = 0.04; HexCERs hexosylceramides, p = 0.018; dhSMs dihydrosphingomyelins, p = 0.043; SMs sphingomyelins, p = 0.021; SPHs sphingosines, p = 0.003; Sterol, p = 0.042. c A pathway analysis indicated that lipid-related pathways were most affected by Sleep_day/Wake_night. d A volcano plot of measured sphingolipids; differentially represented metabolites with >1.3-fold change (Sleep_day vs Wake_night) and p values <0.10 are indicated in color; n.s. nonsignificant. e Sleep deprivation promoted subcortical ceramide accumulation. Subcortical region was sampled at ZT12 (timepoint of light off). p = 0.035 for Ctl vs SD12. Ctl control mice, n = 8; SD6, mice with 6 h of sleep deprivation, n = 8; SD12, mice with 12 h of sleep deprivation, n = 8. f Tight correlation between brain ceramide level and wakefulness. Wakefulness was defined with video recording for 1 h before sampling (ZT12). Pearson’s correlation, p value was calculated using two-sided hypothesis test. Eight mice. *p < 0.05; two-sided unpaired t-test for b, d, and e. Data are reported as mean ± SEM. See also Supplementary Fig. 4.

Fig. 3. Ceramide-mediated microglial modulation of stable wakefulness.

a Experimental procedure for measurement of subcortical ceramide in CX3CR1^Cre-ERT2/+:R26^iDTR/+ transgenic mice. b Sleep_day/Wake_night ceramide levels in microglia depleted (right; Sleep_day, n = 10 mice, Wake_night, n = 10 mice) and control mice (left; daytime, n = 8 mice; night-time, n = 7 mice). p = 3.6 × 10⁻⁵ for Sleep_day vs Wake_night in control mice. c An increase in nocturnal ceramide levels induced by carmofur administration facilitated state transitions between wakefulness and NREM sleep and decreased stable wakefulness in wild-type mice. Car. carmofur, Veh. vehicle. EEG/EMG signals were recorded for 4 h after administration at night (n = 18 pairs of recordings from 6 mice). Each line represents change in transition number between recording sessions with vehicle/carmofur injection collected from one mouse. Veh. vs Car.: p = 5.1 × 10⁻⁴ for WN; p = 3.7 × 10⁻⁴ for NW. d A decrease in ceramide by GW-4869 administration rescued stable wakefulness at night in microglia-depleted mice. GW-4869 or vehicle was injected three times before microglial depletion. Night-time EEG/EMG signals were recorded for 6 h after administration on day 0 (before depletion) and on day 4 (after depletion). Each dot represents the difference in % total duration for each brain state between day 0 and day 4 in one mouse. Veh. vs GW: p = 0.032 for W; p = 0.026 for N. GW, GW-4869 (n = 5 mice); Veh., vehicle (n = 5 mice). *p < 0.05; two-sided unpaired t-test for b, and left panel in d, two-sided paired t-test for left panel in c, two-sided Kolmogorov–Smirnov test for the right panels in c and d. Data are reported as mean ± SEM. See also Supplementary Fig. 4.

Fig. 4. High sensitivity of TRN microglia to brain ceramide.

a Diurnal differences in the total lengths of microglial processes in sleep/wakefulness-related brain regions (Sleep_day, 3 mice; Wake_night, 3 mice). A sagittal brain cartoon showing the regions examined, with red dots representing regions with significant Sleep_day/Wake_night differences in microglial morphology. The white dashed line on the representative microglia indicates the measured processes. Scale bar, 10 µm. The p value for each brain region was obtained by compare the difference of microglial process length between Sleep_day and Wake_night, each dot represents log₂ of ratio (Sleep_day/mean value of Wake_night). MnPO, median preoptic area, p = 0.015 for Sleep_day (n = 65) vs Wake_night (n = 49); LPO lateral preoptic area, p = 5.4 × 10⁻⁴ for Sleep_day (n = 84) vs Wake_night (n = 92); VLPO ventrolateral preoptic area (Sleep_day, n = 53; Wake_night, n = 55); TRN thalamic reticular nucleus, p = 1.5 × 10⁻⁶ for Sleep_day (n = 64) vs Wake_night (n = 70); SCN suprachiasmatic nucleus, p = 0.006 for Sleep_day (n = 48) vs Wake_night (n = 69); LH lateral hypothalamus (Sleep_day, n = 59; Wake_night, n = 70); DM dorsomedial hypothalamus, p = 0.039 for Sleep_day (n = 83) vs Wake_night (n = 88); VLPAG ventrolateral periaqueductal grey matter (Sleep_day, n = 64; Wake_night, n = 59); LDTg laterodorsal tegmental nucleus (Sleep_day, n = 89; Wake_night, n = 53); PZ parafacial zone (Sleep_day, n = 16; Wake_night, n = 25); LPGi lateral paragigantocellular nucleus (Sleep_day, n = 37; Wake_night, n = 48). b Exogenous application of C2-ceramide specifically altered Iba1 fluorescence and microglial morphology by a retraction of processes in the TRN region (Veh. Vehicle, 6 mice; MnPO, n = 38 cells; LPO, n = 25 cells; TRN, n = 85 cells; DM, n = 32 cells; SCN, n = 45 cells. C2-Cer, C2-ceramide, 4 mice; MnPO, n = 25 cells; LPO, n = 55 cells; TRN, n = 57 cells; DM, n = 31 cells; SCN, n = 51 cells). p = 2.5 × 10⁻⁵ in TRN region for Veh. vs C2-Cer. Each dot represents one microglia. Brain tissue was collected 3 h after i.c.v. injection. Representative images with Iba1 staining were taken from TRN-containing brain sections, with the TRN outlined with white dashed lines. The dashed square within the cartoon inset indicates the brain regions from which images were taken. Scale bar, 200 µm. c Total process lengths of TRN microglia in wild-type (WT) and Acer3 knockout (Acer3^−/−) mice (WT, n = 30 cells from 2 mice; Acer3^−/−, n = 31 cells from 3 mice; p = 8.1 × 10⁻¹⁰ for WT vs Acer3^−/−). Each dot represents one microglia. Scale bar, 10 µm. d Correlation between brain ceramide level and microglial morphology in naïve mice. Subcortical ceramide level and total process length of TRN microglia were analyzed in the same mouse. Total process length of at least 20 TRN microglia was measured for each mouse, and the averaged value was used for further analysis. Pearson’s correlation, p value was calculated using two-sided hypothesis test. n = 8 mice. e TRN microglia retracted processes following sleep deprivation. Representative image with Iba1 staining shown TRN microglia in control and sleep-deprived mice. Scale bar, 20 µm. Ctl control mice, 86 cells from 4 mice; SD6, mice with 6 h of sleep deprivation, 91 cells from 4 mice; SD12, mice with 12 h of sleep deprivation, 81 cells from 4 mice. Each dot represents one microglia. f Local injection of clodronate liposomes into the anterior thalamic reticular nucleus (aTRN) effectively depleted microglia (top) and facilitated NREM sleep specifically at night (bottom). Scale bar, 100 µm. Clodronate vs Control at night: p = 0.003 for W; p = 0.003 for N. Clodronate liposomes, n = 6 mice; control liposomes, n = 6 mice. Each open circle represents the difference, before and after local administration, in one mouse, and then these differences were compared between clodronate-treated and control mice. g aTRN neuronal firing rate (n = 14 mice with 248 neurons been collected) during daytime and night-time. Each line represents change in mean firing rate of aTRN neurons between day and night in one mouse. p = 0.012 for day vs night. h Compared to control mice (11 neurons from 3 mice), action potentials induced by higher current injections were reduced in aTRN neurons from microglia-depleted mice (17 neurons from two mice; p < 0.05 for current injection high than 120 pA). i Application of C2-ceramide (10 µM) reduced aTRN neuronal excitability (9 neurons from 3 mice; p < 0.05 for current injection high than 200 pA). *p < 0.05; two-sided unpaired t-test for a–c, e, f, and h. Two-sided paired t-test for g and i. Data are reported as mean ± SEM. See also Supplementary Fig. 5.

Fig. 5. aTRN neuronal activity controls transitions between wakefulness and NREM sleep.

a Simultaneous recordings of aTRN neuronal activity and sleep architecture via tetrode/EEG/EMG implants. The TRN is represented with parvalbumin (PV) staining (red signal, tetrode tip location). b Left, scatter plot of aTRN neuron firing rates (FR) during wakefulness (y-axis) against during NREM sleep (x-axis). Right, distribution of firing preferences between wakefulness and NREM sleep (191 neurons from 14 mice). c, d Left side, activity of aTRN neurons 20 s before and after transition onsets from wakefulness to NREM sleep (c, W → N), or d from NREM sleep to wakefulness (N → W, 428 cells from 23 mice). The black trace indicates average normalized neuronal activity. The raster plots represent spike events from five representative aTRN neurons. Right side, lag time distributions relative to transition onset (the red-dashed line), and an example of logistic fitting is shown at the top. e, f Chemogenetic inhibition (e via DIO-KORD), or activation (f via DIO-hM3Dq), of aTRN neuron controls state transitions between wakefulness and NREM sleep. Top, experimental procedure (left) and validation (right) of virus expression in the aTRN of GAD2^cre mice (scale bars: left = 1 mm; right = 50 µm). Bottom, differences in state transition numbers between wakefulness and NREM sleep in mice with chemogenetic inhibition (e, DIO-KORD, p = 0.002 for W → N and p = 0.010 for N → W, n = 21 pairs of recordings from 7 mice; DIO-mCherry, n = 15 pairs of recordings from 5 mice) or with chemogenetic activation of aTRN neurons (f, DIO-hM3Dq, p = 0.004 for W → N and p = 0.002 for N → W, n = 13 pairs of recordings from 4 mice; DIO-mCherry, n = 15 pairs of recordings from 5 mice). Each line represents the change in transition number between recording sessions with vehicle or CNO/SalB injection collected from one mouse. *p < 0.05; two-sided paired Wilcoxon signed rank test for e and f. Data are reported as mean ± SEM. See also Supplementary Figs. 6–8.

Fig. 6. Attenuation of aTRN neuronal excitability is responsible for changes in wakefulness stability following microglial depletion.

a aTRN neuronal activity and sleep architecture were recorded at night, before and after microglial depletion. b Averaged firing rates (FR) of aTRN neurons before and after DT or vehicle administration (n = 6 DT-injected mice, n = 6 vehicle-injected mice). Each line represents change in mean firing rate of aTRN neuron from one mouse. W wakeful, N NREM sleep. DT 0 vs DT 4, p = 0.007 for wake state. c, e Differences in neuronal firing rates between wakefulness and NREM sleep during transitions from wakefulness to NREM sleep (c DT 0, n = 63 neurons from 6 mice; DT 4, n = 81 neurons from 6 mice; p = 0.028 for DT 0 vs DT 4) and from NREM sleep to wakefulness (e DT 0, n = 69 neurons from 6 mice; DT 4, n = 81 neuron from 6 mice s; p = 0.017 for DT 0 vs DT 4) before/after microglial depletion. Spike events occurring 5–20 s before/after transition onsets were included. d, f aTRN activity during transitions from wakefulness to NREM sleep (d) and from NREM sleep to wakefulness (f) before/after microglial depletion. Neuronal activities within ±20 s from transition onsets were normalized to their minimum firing rates. Right panels, averaged aTRN neuronal activities during transitions. g Experimental procedure for chemogenetic activation of aTRN in microglia-depleted mice. Top, validation of hM3Dq expression. Scale bar = 1 mm. h, i The number of state transition (h left: WN wakefulness to NREM sleep; right: NW NREM sleep to wakefulness; GFP, DT 0 vs DT 4: p = 0.015 for WN, p = 0.020 for NW) and % total duration of wake and NREM sleep (i left, total duration of wake; right, total duration of NREM sleep; GFP, DT 0 vs DT 4: p = 0.005 for wake, p = 0.007 for NREM sleep) were compared before/after microglial depletion in GFP-expressing (n = 7) or hM3Dq-expressing (n = 7) mice. Each line represents change in transition number from one mouse. j Left panel: in control mice expressing GFP, microglial depletion significantly shortened the bout durations of stable wakefulness (n = 7 mice). Right panel: chemogenetic activation of the aTRN abolished the effect of microglial depletion on wakefulness stability (n = 7 mice). k A proposed model for ceramide-mediated microglia-neuron interactions in wakefulness/NREM sleep regulation. Microglia regulate the diurnal variation in ceramide levels, which determines the neuronal activity of the aTRN and, in turn, controls transition states between wakefulness and NREM sleep. Green cells: aTRN microglia; blue cells: aTRN neurons; red molecules: ceramide; black raster plot, action potentials produced by aTRN neurons. *p < 0.05; two-sided unpaired t-tests for c and e; two-sided paired t-tests for b, h, and i. two-sided Kolmogorov–Smirnov test for j. Data are reported as mean ± SEM. See also Supplementary Fig. 9.