REV-ERBα mediates complement expression and diurnal regulation of microglial synaptic phagocytosis

While analyzing our previously published transcriptomic dataset from global Bmal1 knockout (BMKO) hippocampal tissue, we observed a striking upregulation of several complement transcripts, in particular C4b and C3 (Griffin et al., 2019), two genes which are critical for synaptic phagocytosis (Figure 1A). Other complement-related transcripts including C1qc, C1qb, C1qa, and C1ra were also increased in BMKO hippocampus (Figure 1A). Analysis of second published microarray dataset from our group which was derived from cerebral cortex samples from brain-specific Bmal1 KO mice (Nestin-Cre;Bmal1^f/f) mice (Lananna et al., 2018) also revealed specific upregulation of C4b, but less so for other complement factors (Figure 1A). C4b was similarly increased in cerebral cortex samples from 4mo tamoxifen-inducible global BMAL1 KO mice (CAG-Cre^ERT2;Bmal1^f/f) in which Bmal1 was deleted at 2mo, demonstrating that this is not a developmental phenomenon (Figure 1—figure supplement 1).

To determine the cell type(s) in which BMAL1 deletion induces complement gene expression, we examined cerebral cortex tissue from pan-neuronal-(CamK2a-iCre;Bmal1^fl/fl) (Izumo et al., 2014), astrocyte- (Aldh1l1-Cre^ERT2;Bmal1^fl/fl) (Lananna et al., 2018), and microglia-specific (Cx3cr1-Cre^ERT2;Bmal1^fl/fl) (Parkhurst et al., 2013) BMAL1 knockout mice. Both astrocyte- and microglia-specific Bmal1 KO mice were treated with tamoxifen at 2mo and harvested 2 months later. Notably, C4b mRNA was strongly induced in the neuron-specific Bmal1 KO mice, while C3 was not (Figure 1B). C4b was also induced to a lesser degree in astrocyte-specific Bmal1 KO mice, although C3 was not (Figure 1C). Under basal conditions, neither C4b nor C3 was induced in microglia specific Bmal1 KO mice (Figure 1D). Deletion of Bmal1 in primary Bmal1^fl/fl neuron cultures via infection with an AAV8-Cre viral vector (versus AAV8-eGFP control) suppressed the BMAL1 transcriptional target Nr1d1 (which encodes REV-ERBα) by 85% and induced C4b expression but caused no increase in C3 (Figure 1E). Bmal1 deletion also increased expression of Fabp7 (Figure 1E), a known target of REV-ERBα-mediated transcriptional repression (Schnell et al., 2014). As REV-ERBα mRNA (Nr1d1) is also suppressed by ~80% in the brain following brain-specific Bmal1 deletion (Figure 1—figure supplement 1), our data suggested that the upregulation of C4b gene expression observed with loss of BMAL1 could be mediated by transcriptional de-repression as consequence of downstream REV-ERBα loss. Accordingly, global deletion of REV-ERBα caused striking increases in C4b, C3, and other complement transcripts in the hippocampus as assessed by microarray analysis (Figure 1A) and confirmed in separate samples by qPCR (Figure 1F). Increased C3 protein expression was observed in activated astrocytes (Figure 1G, Figure 1—figure supplement 2) in the hippocampus of 5mo REV-ERBα KO (RKO) mice. There was less overlap of C3 expression and the microglial marker Iba1 (Figure 1H), although microglial C3 expression did increase in RKO brain (Figure 1—figure supplement 2). The finding that Bmal1 deletion in neurons (and to a less degree astrocytes) induces C4b mRNA but that C3 only increases in global BMKO and RKO mice suggests that REV-ERBα directly represses C4b expression in neurons and astrocytes, but the induction of C3 in both BMKO and RKO brain is likely secondary to a multicellular inflammatory glial activation which occurs over time.

We previously demonstrated that global REV-ERBα deletion induced microglial activation in vivo (Griffin et al., 2019). Given those results and the observation of increased C4b and C3 expression in RKO brains, we examined the possibility that these changes would enhance synaptic phagocytosis in 4–6mo RKO mice. We primarily focused on the mossy fiber synapses in the CA2/3 region of the hippocampus, as these large synapses (thorny excrescences) can easily be stained and imaged using standard confocal microscopy. Triple-labeling of tissue sections was performed with antibodies against synaptophysin (a marker of presynaptic neuronal terminals – Figure 2A), CD68 (a microglial lysosome marker – Figure 2B) and Iba1 (to define microglial cell bodies and processes – Figure 2C). CD68 was used to ensure that the colocalized synaptic material was actually within the microglia phagosome. 3D reconstructions were made and total volumes of engulfed synaptic material were calculated (Figure 2D, Figure 2—figure supplement 1). In the RKO microglia, we observed synaptic material in the microglial process and cell body (Figure 2Eii, Eiii), whereas WT microglia only had engulfed synaptic material in the cell body (Figure 2Ei). In all, we noted a 13.6-fold increase in engulfed synaptic material in the hippocampus of RKO mice compared to their WT littermates (Figure 2F).

To corroborate these results, we also performed large area scanning electron microscopy (SEM) experiments of the CA3 region of 4–6mo WT and RKO mice. Using this method, we visually confirmed an increased number of presynaptic terminals within or in contact with microglia in the hippocampus of RKO mice compared to WT (Figure 2G and H, 2H(i-v), 2I). We also noted downregulation in the expression of the Sirpa gene in our RKO mice which codes for the protein SIRPα (Figure 1A). SIRPα was recently described as a surface receptor on microglia that serves as a receptor for a ‘do-not-eat-me’ signal (Lehrman et al., 2018). Notably, Sirpa was not downregulated in Nestin-Cre-Bmal1 KO mice (Figure 1A), which have normal microglial BMAL1 and REV-ERBα expression, suggesting a possible cell-autonomous effect of REV-ERBα on microglial Sirpa expression. Taken together, our results suggest that REV-ERBα deletion induces synaptic engulfment.

Following our observations of increased synaptic phagocytosis in the RKO mice, we investigated the status of the synapses in the CA3 region of the hippocampus. Synapses were double labeled by using the presynaptic marker synaptophysin and the postsynaptic marker homer1, followed by confocal imaging. In RKO mice, we observed a significant decrease in the synaptic volume in the CA3 region of the hippocampus by synaptophysin staining (Figure 3A), homer1 staining (Figure 3B), and their colocalization (Figure 3C) compared to their WT littermates. To further confirm these results, we counted synapses in large area SEM images from the stratum lucidum of WT or RKO mice. Again, we noted a decrease in the number of synapses in the RKO mouse CA3 by SEM, as compared to their WT littermates (Figure 3D).

To ensure that our observations were not due purely to changes in synaptophysin protein expression in the presynaptic terminal, synapses were also stained with a second presynaptic marker, synaptoporin, and quantified. Synaptoporin was used because it is enriched in the mossy fiber synapses of the hippocampus (Singec et al., 2002). Again, we observed a similar decrease in synaptic volume in CA3 of RKO mice compared to their WT littermates using synaptoporin staining (Figure 3E). Interestingly, we did not observe significant differences in synaptic volumes between WT and RKO mice in the CA1 region of the hippocampus (Figure 2—figure supplement 1), suggesting that some terminals may be more susceptible to loss than others. To determine whether the loss in synapses was due to a loss of neuronal cell bodies, we quantified the volume and width of neuronal nuclei of the dentate gyrus, which project to CA3, via NeuN staining. We found no significant difference between the neuronal nuclear width or volumes of the dentate gyrus between the WT and RKO mice (Figure 3—figure supplement 2). Moreover, we did not observe any change in 3-nitrotyrosine staining in the dentate gyrus, suggesting there is no increase in neuronal nitrosative stress in RKO mice (Figure 3—figure supplement 3). Taken together, our data suggest that deletion of REV-ERBα results in robust synaptic loss in the CA3 region without obvious neuronal loss.

In mouse cerebral cortex and ventral midbrain, REV-ERBα displays daily oscillation in mRNA level with its peak around zeitgeber time (ZT) eight and its trough around ZT20 (Chung et al., 2014). We have also previously described oscillations in microglial activation, with low activation at ZT5 and high activation at ZT17 (Griffin et al., 2019). Consequently, we investigated microglial synaptic engulfment at these times of day. WT (C57BL/6) and RKO mice were sacrificed ZT17 (11PM), and hippocampal sections were triple labeled with Iba1, CD68, and synaptophysin, and compared to those sacrificed at ZT5 (11AM, shown in Figure 2). Microglia from the CA3 region of the hippocampus in WT mice sacrificed at ZT17 showed significantly more engulfed presynaptic protein than those at ZT5 (Figure 4A–C). When compared to the WT mice, RKO mice harvested at ZT5 and 17 showed a persistently increased level of synaptic engulfment in microglia, but with no time-of-day variation (Figure 4C), suggesting that daily oscillations in REV-ERBα may mediate rhythms in microglial synaptic phagocytosis. To further establish these findings, we used large area SEM to count the number of presynaptic terminals in contact with or within microglia at ZT17, and compared to mice harvested at ZT5 (shown in Figure 2G–H). In that experiment, we also noted a higher number of presynaptic terminals in contact with and within the microglia at ZT17 (Figure 4D–F). As a final confirmation of this finding, we injected Cx3cr1^GFP mice, which express GFP in microglia (Jung et al., 2000), with an AAV-Camk2a-mCherry viral vector in the retrospenial cortex, to express mCherry in neurons. 4 weeks later, we collected brain samples at ZT5 and 17, and calculated the volume of mCherry+ material in individual GFP+ microglia. We observed that microglia at 11PM (ZT17) showed a qualitative increase in mCherry engulfment as compared to those at 11AM (ZT5) (Figure 4—figure supplement 1). Overall, our data establish a REV-ERBα-dependent rhythm for the engulfment of neuronal materials by microglia in the brain.

Nr1d1^+/–mice on C57bl/6 background were obtained from The Jackson Laboratory (Bar Harbor, ME) and bred at at Washington University. Heterozygous mice were bred together to generate Nr1d1^+/+ (wt) and Nr1d1^-/– (referred to as REV-ERBα KO or RKO) littermates which were used for experiments. For all experiments, a mix of male and female mice was used. No obvious difference between sexes was noted, though experiments were not powered to detect sex differences. WT mice were C57Bl/6J mice from Jackson Labs. Constitutive Bmal1^-/- (BMKO), CAG-Cre^ERT2, Aldh1L1-Cre^ERT2, Cx3cr1-Cre^ERT2, and Bmal1^fl/fl mice were obtained from the Jackson Laboratory, and bred so that mice were heterozygous for Cre and homozygous for the floxed Bmal1 allele. In these experiments, Cre-;Bmal1^f/f littermates were used as controls. All inducible knockout lines were were treated with tamoxifen (Sigma) dissolved in corn oil via oral gavage, 2.5 mg/day for 5 days, at 2mo. Camk2a-iCre+;Bmal1^fl/fl mice were bred at University of Texas- Southwestern Medical Center. Cx3cr1^GFP mice were obtained from The Jackson Laboratory. Mice were housed on a 12/12 light/dark cycle and fed ad libitum. All procedures performed on the mice were approved by the Washington University IACUC.

The following primary antibodies were used (with dilution): Gfap (Rabbit polyclonal, 1:2500, Dako/Agilent Cat# Z0334), C3 (Rat monoclonal, 1:500, Novus Biologicals Cat# NB200-540), Iba1 (Goat polyclonal, 1:500, Abcam Cat# ab5076), Synaptophysin (Mouse monoclonal, 1:100, Abcam Cat# ab8049), CD68 (Rat monoclonal, 1:250, BioRad, Cat# MCA1957), Homer1 (Rabbit polyclonal, 1:500, Synaptic Systems Cat# 160 003), Synaptoporin (Rabbit polyclonal, 1:1000, Synaptic Systems Cat# 102 002), NeuN (Mouse monoclonal, 1:1000, EMD/Millipore Cat# 14-5698-82). Specificity of all antibodies was confirmed via staining without primary antibody, and in the case of the anti-C3 antibody, was tested in C3 KO mouse tissue.

Mice were anesthetized with intraperitoneal (i.p.) injection of pentobarbital (150 mg/kg), followed by pump perfusion for three mins with ice cold Dulbecco’s modified Phosphate Buffered Saline (DPBS) containing 3 g/l heparin sulfate. One hemisphere was drop fixed in 4% paraformaldehyde (PFA) for 24 hr at 4°C, then cryoprotected with 30% sucrose in PBS also at 4°C for at least 48 hr. Brains were embedded in OCT and frozen in acetone with dry ice. Twelve μm serial coronal sections were cut on a cryostat and mounted directly onto the glass slides for synaptic pruning analysis or synaptic quantification. Sections were washed three times in Tris buffered saline (TBS), blocked for 45 min in TBS containing 20% goat (or donkey) serum, 2% Mouse-on-mouse (M.O.M) blocking reagent and 0.4% Triton X-100 (Sigma-Aldrich, St. Louis, MO). Sections were then incubated in TBS containing 10% goat (or donkey) serum, 8% M.O.M protein concentrate and 0.4% Triton X-100 with primary antibody overnight at 4°C. Sections were then washed three times and incubated for 4 hr at room temperature with 1:1000 fluorescent secondary antibody in a solution of TBS containing 10% goat (or donkey) serum, 8% M.O.M protein concentrate and 0.4% Triton X-100. For other non-synaptic staining, 50 μm serial coronal sections were cut on a freezing sliding microtome and stored in cryoprotectant (30% ethylene glycol, 15% sucrose, 15% phosphate buffer in ddH₂O). Sections were washed three times in Tris buffered saline (TBS), blocked for 30 min in TBS containing 3% goat serum and 0.25% Triton X-100 (Sigma-Aldrich, St. Louis, MO) then incubated in TBS containing 1% goat serum and 0.25% Triton X-100 with primary antibody overnight at 4°C. Sections were then washed three times and incubated for 1 hr at room temperature with 1:1000 fluorescent secondary antibody and mounted on slides. Confocal images were taken on the Nikon Elements software on the Nikon A1Rsi scanning confocal microscope. Z-stacks were taken at a step size of 0.5–1 µm from dark to dark through the tissue.

Previously-published transcriptomic datasets were analyzed for expression of complement transcripts. Both datasets were from Agilent 4 × 44 k mouse microarrays. One dataset was derived from 5- to 6-month-old Bmal1 and REV-ERBα KO mice with WT littermate controls, and used hippocampal tissue. This dataset was previously published (Griffin et al., 2019) and is freely available on the Array Express Server: https://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-7590↗. The second dataset was derived from Nestin-Cre;Bmal1^f/f mice (and Cre- controls) aged to 12 months, and used cerebral cortex tissue. This dataset was also previously published (Lananna et al., 2018) and is also available on Array Express Server: https://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-7151↗. For this dataset, the samples harvested at 6pm are shown.

Flash-frozen brain tissue was homogenized with a mechanical handheld homogenizer for 20 seconds in RNA kit lysis buffer (PureLink RNA Mini Kit, Life Technologies, Carlsbad, CA) plus 1% β-mercaptoethanol. RNA was then purified using the kit protocol. Cells well collected and lysed in Trizol (Life Technologies). The aqueous layer was collected following chloroform extraction (added at 1:5 then spun at 13,000xg for 15 min) with RNA isolation protocol. RNA concentrations were then measured using the Nanodrop spectrophotometer and cDNA was made using a high-capacity RNA-cDNA reverse transcription kit (Applied Biosystems/LifeTechnologies) with 1 μg RNA used per 20 μL reaction. Real-time quantitative PCR was performed with Taqman primers and PCR Master Mix buffer (Life Technologies) on the ABI StepOnePlus 12 k Real-Time PCR thermocyclers. β-actin (Actb) mRNA levels were used for normalization during analysis. The following primers were used (all from Life Technologies, assay number is listed): Actb: Mm02619580_g1, Nr1d1: Mm00520708_m1, C4b: Mm00437893_g1, C3: Mm01232779_m1, Fabp7: Mm00437838_m1.

Animals were perfused with a fixative mix consisting of 2.5% glutaraldehyde + 2% paraformaldehyde (fresh, EM grade) in 0.15 M cacodylate buffer with 2 mM CaCl2 (final concentrations). Brains were extracted, blocked, and 100 µm coronal sections were made using a Leica 1200S vibratome. Tissues were then washed 3 × 10 min in cold cacodylate buffer containing 2 mM calcium chloride, and then incubated in a solution of 1% OsO₄ containing 3% potassium ferrocyanide in 0.3M cacodylate buffer with 4 mM calcium chloride for 1 hr in the dark. Following incubation, tissues were incubated for 20 min in a 1% thiocarbohydrazide (TCH) solution, rinsed 3 × 10 min in ddH2O at room temperature and thereafter placed in 2% osmium tetroxide (NOT osmium ferrocyanide) in ddH20 for 30 min, at room temperature. Tissues were washed 3 × 10 min at room temperature in ddH2O then placed in 1% uranyl acetate (aqueous) and incubated at 4° overnight. Tissues were incubated in a lead aspartate solution at 60°C oven for 30 min, washed 5 × 3 min in ddH2O, and returned to the lead aspartate solution at 60°C for 30 min. Tissues were washed 3 × 10 min in room temperature ddH₂O and dehydrated in 50%, 70%, 90%, 100%, 100% acetone (anhydrous), 10 min each. Samples were embedded in Durcupan ACM resin and polymerized 60°C oven for 48 hr. Seventy nm slices were made using a Leica UC7 ultramicrotome, and sections were picked up on a Si wafer (Ted Pella, Redding, CA). Images were acquired on a Zeiss Merlin FE-SEM using a solid state backscatter detector (8kV, 900 pA) at 7 nm resolution with 5 µs pixel dwell times and 4x line averaging. Large area scans of ~150 µm x 150 µm field of view were acquired and stitched using Atlas 5.0 (Fibics, Ottawa, Canada).

Stereotactic viral vector injections were performed on Cx3cr1^GFP mice. Briefly, mice were anesthetized with inhaled isoflurane, the head was shaved and the skull exposed and cleaned with iodine, and the head supported in ear bars. A hole was drilled through the skull. The coordinates used for the retrospenial cortex were 0.3 mm mediolateral (M/L), −2 mm anterior-posterior(A/P) and −1 mm dorso-ventral(D/V) with bregma as a reference. The needle was placed in the target location and then allowed to rest for 2 min before infusion. The AAV2-Camk2a-mCherry virus was obtained from the UNC viral vector core and infused at a rate of 0.2 µL/min. In total, the mice were injected with 2 μL in the retrospenial cortex. After the infusion, we waited 5 min for the virus to diffuse in the parenchyma and then the needle was slowly removed. The skin over the skull was then stitched up and antibiotic was applied to the area. The mice were then allowed to recover in an empty cage on a heating pad. After the surgery the mice were checked twice daily, 4 hr apart to ensure survival for 3 days. The mice were then allowed to age for 1 month before being sacrificed for sectioning and imaging.

Imaris visualization and analysis software (Version 9, Bitplane, South Windsor, CT, USA) was used at the Washington University Center for Cellular Imaging. For all analyses, Z-stacks were saved in the. nd2 file format and loaded into the software. 3-D surfaces had a surface detail ranging from 0.1 to 0.3 μm. To quantify the volume of the synapses, we generated 3-D surfaces from each of the synaptic markers (Synaptophysin, Synaptoporin and Homer1). We used the Batch colocalization function to colocalize the Synaptophysin and Homer1 volumes. The total volume for each Z-stack was summed. For the synaptic engulfment analyses, we first generated volumes from synaptophysin, CD68 and Iba1 staining. We then colocalized the synaptophysin volume with the CD68 volume. The resulting volume was then colocalized with the Iba1 volume to produce a final microglial synaptic engulfment volume. All values were normalized to the average values for the control group in each experiment.

Neuronal cultures were made from age E18 Bmal1^f/f mouse pups. Cortices plus hippocampi were dissected and stripped of meninges in ice-cold DMEM (Life Technologies) and then incubated in 0.05% Trypsin-EDTA at 37°C for 15 min. Tissue was gently triturated in 37°C DMEM plus 10% FBS (Gibco). Triturated cells were transferred to a second tube to remove debris, then diluted in Neurobasal (Life Technologies) plus B27 (Life Technologies) prior to plating on a PDL-coated plates. Cells were treated with 1 uM cytosine arabinoside for 24 hr, after which media was replaced. On DIV 4, Bmal1^f/f neuron-enriched cultures were treated with either AAV8-CMV-GFP or AAV8-CMV-Cre for 4 days. Media was then again changed, and cells were harvested for in Trizol reagent at DIV 10.

Statistical analyses were performed using GraphPad Prism 8. We performed t-tests with Welch’s correction for multiple comparisons.