All mouse lines were maintained in Laboratory Animal Research Unit, City University of Hong Kong. mPer2::Luc (RRID: IMSR_JAX:006852, B6.129S6-Per2tm1Jt/J) (40); CX3CR-1^GFP (RRID: IMSR_JAX:005582, B6.129P2(Cg)-Cx3cr1^tm1Litt/J) (41); NG2^DsRed (RRID:I MSR_JAX:008241, STOCK Tg(Cspg4-DsRed.T1)1Akik/J) (42); Aldh1l1-EGFP/Rpl10a (RRID: IMSR_JAX:030248, B6; FVB-Tg(Aldh1l1-EGFP/Rpl10a) JD133Htz/J) (43) and Bmal knockout mice (RRID: IMSR_JAX:009100, B6.129-Arntl^tm1Bra/J) (44) were purchased from Jackson Laboratories. All mice were maintained in a 12:12h light-dark cycle at 20-24°C with 50%–70% humidity. All mouse experiments were performed in accordance with the protocol approved by the Institutional Animal Research Ethics Sub-Committee of City University of Hong Kong and Department of Health, The Government of The Hong Kong Special Administrative Region.

Primary microglia were cultured from cortex of the postnatal day 0-2 (P0–2) pups. After decapitation, craniotomy and meninges removal in dissection medium [1x HBSS (Gibco) + 2.5 mM HEPES (Gibco) + 5.4 g/L glucose (Sigma-Aldrich) + 100 units/mL penicillin/100 μg/mL streptomycin (Gibco)], dissected brains from postnatal day 0-2 (P0-P2) C57BL/6J mice were digested. The tissue was triturated after trypsin neutralization, filtered through a 70 μm cell strainer (Falcon), and seeded in 0.1 mg/mL Poly-D-Lysine (PDL)-coated flasks. Cultures were maintained at 37°C with 5% CO&₂, with medium changed every 3–4 days. Mixed glial cells were ready for subculture or experiments after fully confluence for experimentation at 6–8 days in vitro.

Mixed glia cultures were digested with trypsin. After trypsinization determined by culture medium [1x HBSS (Gibco) + 2.5mM HEPES (Gibco) + 5.4g/L glucose (Sigma-Aldrich) + 100 units/mL penicillin + 100 μg/mL streptomycin (Gibco)], microglia were purified by cluster of differentiation molecule 11B (CD11b) beads (#130-097-142, Miltenyi) and followed the magnetic-activated cell sorting (MACS) protocol according to Miltenyi’s instructions. After elution and centrifugation, microglia were resuspended with culture medium and plated in a dish pre-coated by Poly-D-Lysine.

Primary microglia were cultured from the cortex of P0–2 pups of the mPer2::Luc mouse line, which expresses a PER2::LUCIFERASE reporter in various tissues for around 1 week until full growth, then purified and cultured for 3 days. To synchronize all the cells in the same plate, microglia were treated with 100 μM Dexamethasone (Cat# D4902, Sigma) in Lumicycle Medium [1% DMEM (Gibco) + 0.348% Glucose (Sigma-Aldrich) + 10 mM HEPES (Gibco) + 10% Horse serum (Gibco) + 350 mg/L Bicarbonate solution (Gibco) + 140 U/mL Penicillin-Streptomycin (Gibco)] for 2 h. After synchronization, culture media was changed by Lumicycle medium with 100 μM D-Luciferin (Cat# E1602, Promega) as the substrate of luciferase. The culture plate was sealed with a glass coverslip by vacuum grease and kept in the LumiCycle machine (Actimetrics) placed in the cell incubator at 37°C with 5% CO₂. Bioluminescence (counts/sec) produced by luciferase from each plate was recorded and the oscillation curve was calculated by the software Lumicycle Analysis (45).

8- to 12-week-old adult mice were perfused with 4% (wt/vol) paraformaldehyde (PFA) in 1x phosphate-buffered saline (PBS) at circadian time (CT) 06 and their brains were dehydrated in 30% (wt/vol) sucrose (ChemCruz). Fixed coronal sections with 20μm thickness in proper position were attached to adhesion slides and blocked with PGBA blocking buffer containing 0.1% gelatin (Sigma-Aldrich), 1% BSA (Sigma-Aldrich), 0.002% sodium azide (Sigma-Aldrich), 10% normal goat serum (Jackson Immuno Research), and 0.5% Triton X-100 (Sigma-Aldrich) in 0.1 M phosphate buffer [PB, 71.7 mM K₂HPO₄ (Sigma-Aldrich) + 28.3 mM KH₂PO₄ (Sigma-Aldrich)] for 1 hour under room temperature. Then, the sections were incubated with primary antibodies in blocking buffer overnight at 4°C. After washing for three times with 0.1M PB, each time 10 minutes, brain sections were incubated with secondary antibodies in blocking buffer for 1 hour under room temperature without light. Again, three times of wash with PB were performed to remove secondary antibodies. Finally, samples were mounted with mounting medium containing 4’,6-diamidino-2-phenylindole (DAPI; Vetashield, H-1200).

The following antibodies were used in this study for IHC: IBA1 (#019-19741, Wako), 1:500; CD11b (#101202, Biolegend), 1:100; S100β (ab52642, Abcam), 1:100; GFAP (#835301, Biolegend), 1:1000; NG2 (#AB5320, Merck), 1:200; Ki67 (#14-5698-80, Invitrogen), 1:500; Alexa Fluor 488 donkey anti-mouse, Alexa Fluor 594-donkey anti-mouse, Alexa Fluor 647-donkey anti-mouse, Alexa Fluor 488-donkey anti-rabbit, Alexa Fluor 594-donkey anti-rabbit, Alexa Fluor 594-donkey anti-rat, Alexa Fluor 647-donkey anti-rat antibody, 1:400 (all from Jackson Immuno Research).

Fluorescent images were obtained using A1R HD25 Confocal Microscope (Nikon) by NIS-Elements imaging software (Nikon) (46). Images were taken with 20x objective lens and magnification of 1024*1024. Images with Z stacks had 5 stacks and the step between each stack was 2 μm. When making Z stacks with max intensity projection, 3 stacks with the highest and best signal intensities were used for the morphology analysis or cell count in the injection site.

Lipopolysaccharides (LPS, L7895, Sigma) was dissolved in endotoxin-free DPBS (#A1285601, Gibco) to make the 1 mg/mL stock for the injection. To induce the neuroinflammatory reaction in mouse brain, 5 mg/kg LPS was injected intraperitoneally (i.p.) to 8- to 12-week-old adult mice. To maintain the consistent circadian times across animals, LPS injection was performed at CT04. After specific days of injection, mice were sacrificed and their brains were harvested for the following experiments.

Brains were harvested at CT04 from 8- to 12-week-old adult mice, and dissected cortices were digested in trypsin. The pellet was triturated slowly and gently in 4 mL culture medium, and then transfer cell suspension to 70 μm cell strainer and washed by another 1 mL culture medium. The cells were pelleted and resuspended in 2 mL brain red blood cell lysis buffer [155 mM NH₄Cl (Sigma-Aldrich) + 12 mM NaHCO₃ (Sigma-Aldrich) + 0.1 mM EDTA (Sigma-Aldrich)] and incubated for 5 minutes at room temperature. Cells were washed once with MACS buffer [0.5% w/v Bovine Serum Albumin (BSA; Sigma-Aldrich) + 2 mM Ethylenediaminetetraacetic Acid (EDTA; Affymetrix) in phosphate buffered saline [PBS; 137 mM NaCl (Affymetrix) + 2.7 mM KCl (Sigma-Aldrich) + 8 mM Na₂HPO₄ (Sigma-Aldrich) + 1.5 mM KH₂PO₄ (Sigma-Aldrich)] to remove red blood cell lysis buffer thoroughly. Then, single cells were resuspended in MACS buffer.

To purify glia cells and remove neurons and oligodendrocytes from the mixed single cells, 60 μL myelin removal beads (#130-096-731, Miltenyi) were added to the cell suspension and purified by MACS according to Miltenyi’s instructions. To purify microglia from the mix glia suspends, 10 μL CD11b beads (#130-097-142, Miltenyi) were added to the myelin-removed cell suspension purified by MACS according to Miltenyi’s instructions. Countess II Automated Cell Counter (AMQAX1000, Invitrogen) was used to determine cell number with 0.4% of Trypan blue staining 0.4% (2295044, Invitrogen) for live cell concentration.

For circadian gene expression analyses, brains were harvested every 4 hours over 28 hours from 8- to 12-week-old adult mice, providing seven time points that span over a complete circadian cycle under both control and post-LPS conditions. Purified microglia were dissolved in TRIzol (15596026, Invitrogen) and then kept in -80°C for the RNA extraction. Total RNA was extracted following manufacturer’s instruction by Invitrogen. Nanodrop spectrophotometry was used to assess RNA purity and concentration and 400 ng of total RNA was used for the following steps. Genomic DNA removal and RNA reverse transcription were performed by PrimeScript RT Reagent Kit with gDNA Eraser (Perfect Real Time) (TaKaRa). qRT-PCR with cDNA products were performed with SYBR Premix Ex Taq (Tli RNase H Plus) (TaKaRa) in Applied Biosystems QuantStudioTM 3 Real-Time Polymerase Chain Reaction System. Primers used for the qRT-PCR are listed in Table 1.

A total of 1x10⁷ cells purified from cortices harvested at CT04 within 2 to 3 adult mice (8- to 12-week-old adult mice) were lysed by 1 mL lysate buffer [1% Igepal (CA-630, Sigma-Aldrich), 20 mM Tris-HCl (pH 8.0) (Affymetrix), 137 mM NaCl (Affymetrix), 2 mM EDTA (Affymetrix), 200 mM Sodium Orthovanadate (Sigma-Aldrich), 5 mM NaF (Sigma-Aldrich) with freshly added cOmplete^™ protease inhibitor (Roche)] with gently rock at 4°C for 30 minutes. After incubation, cell suspension was collected into a clean tube by centrifugation. Protein concentration was measured by using Braford assay (Bio-rad). Cytokine screening was performed by using Mouse Cytokine Antibody Array, Panel A (ARY006, R&D). Bio-Rad ChemiDoc (Bio-rad) was used to take the images. Image quantification was performed with ImageJ measurement.

To transplant microglia into the corpus callosum (CC) of recipient mouse brain, 8- to 12-week-old adult mice were narcotized with 200 μl Ketamine (final concentration: 1%; #HK-37715, Alfasan)/Xylazine (final concentration: 0.1%; #HK-56179, Alfasan) in filtered saline (0.9% NaCl in water). Digital Mouse Stereotaxic Instruments (Stoelting) was used to fix the mice head and the skull was exposed by vertical cut on the head skin. After verifying skull flatness by moving the injector with glass needle filled with mineral oil (M8410-100ML, Sigma-Aldrich), the injection coordinates relative to the bregma in mm were used to reach the CC, AP: -1.70, ML: ± 1, DV: 0.25, and 1x10⁵ microglia in total volume of 1 μL were injected. with Nanoject III Programmable Microinjector and Wiretrol™ I glass micropipette (Drummond) at a rate of 13 nl/s in each side. Pipettes remained post-injection for 5 minutes to prevent reflux. The wound on the scalp was sealed with Vetbond surgical glue (3M), followed by postoperative monitoring until recovery (47). Microglia for injection were purified at CT04, corresponding to the circadian activation phase, and transplanted into the CC at CT10, based on the time required for microglia isolation and preparation. Brains were collected three days post-transplantation, allowing the transplanted microglia to remain in the CC for approximately three full circadian cycles.

To evaluate whether our sample size was sufficient to detect group differences, we calculated statistical power based on the observed effect size. Assuming a two-tailed Student’s t-test, α = 0.05, and a large expected effect size (Cohen’s d = 1.5), the statistical power was approximately 65% for comparisons with n = at least 3 per condition. This effect size was based on consistent and robust differences observed in microglial branch number, marker intensity, and OPC recruitment. Although the sample size is modest, statistically significant differences were observed across multiple independent experiments. The combination of reproducible findings and large effect size supports the biological relevance and reliability of our conclusions. A Cohen’s d of 1.5 is considered a very large effect, and under such conditions, statistical significance can be reasonably detected with limited replication. While the power is below the conventional 80% threshold, we interpret these results as meaningful, and future studies with larger cohorts will further validate and extend these findings.

Data were presented as mean ± standard deviation (SD). Prior to applying significance analysis, for data sets with N > 20, the assumption of normality for continuous variables was evaluated by quantile-quantile (Q-Q) plots. Departures from the straight diagonal reference line were interpreted as evidence of non-normality. When the number of biological replicates was fewer than seven (N < 7), normality testing was not conducted, as standard normality tests are not reliable with small sample sizes. Instead, all individual data points are presented in each graph to allow a transparent assessment of data distribution. Difference comparisons between two different conditions were carried out using an unpaired two-tailed Student’s t-test. All images were created and analyzed by NIS-Elements BR 5.21.00 and quantified using ImageJ. All statistical analyses were performed using GraphPad Prism 9 (San Diego, CA). In all cases, results were considered statistically significant at P < 0.05.

Statistical analysis of circadian rhythmicity of all examined genes was determined by CircaCompare (48) or MetaCycle (49).

Most cell types, including microglia, contain endogenous circadian clocks. However, the robustness of rhythmicity can vary across cell types and may be reinforced by interactions with neighboring cells (50, 51). To determine whether microglial clocks are self-sustained without external inputs, we cultured cortical microglia from mPer2::Luciferase knock-in mice (mPer2::Luc) (52) and monitored real-time bioluminescence for six consecutive days ( Figure 1A ). The luciferase recordings revealed sustained oscillatory rhythms with a period of 25–26 hours ( Figures 1B, C ). This confirms that microglia contain endogenous molecular clocks capable of generating self-sustained circadian rhythms independently of external inputs.

Since microglia closely interact with neurons and glia (oligodendrocytes and astrocytes) in the brain, we next examined their in vivo circadian rhythms by monitoring activation states across the day ( Figure 2A ). Immunostaining for IBA1, a pan-microglial marker, revealed no significant change in microglial number across circadian time (CT) points. However, the number of branches per IBA1&⁺ cell oscillated, peaking at CT18 and troughing at CT06 ( Figures 2B, C ; Table 2 )—reduced branching at CT06 indicating a more activated microglial state. In addition, the intensity of IBA1 and CD11b—both associated with microglial activation—also exhibited rhythmic oscillations, peaking around CT10 ( Figures 2B, D ; Table 2 ). These results indicate that microglial activation states oscillate under physiological conditions, with a higher activity around CT06 and a lower activity around CT18, correlating with circadian rhythmicity.

Microglia transition into activated states in response to a variety of pathological conditions (8, 10). To compare activation states under physiological versus pathological conditions, we induced neuroinflammation through intraperitoneal (i.p.) LPS injection ( Figure 3A ). Immunostaining revealed a progressive increase in IBA1 and CD11b intensity from Day 1 to Day 3 post-injection, with peak expression observed on Day 3 ( Supplementary Figure S1 ). At CT06—a time point when microglia are normally most active—neuroinflammatory brains showed microglia with fewer branches and markedly increased IBA1 and CD11b intensity compared to PBS-injected controls ( Figures 3B, C ). These findings indicate that while microglial activation exhibits circadian oscillations under physiological conditions, neuroinflammation leads to a markedly enhanced activation state. This implies a functional shift from homeostatic surveillance to immune response.

Since microglial activation is elevated under neuroinflammatory conditions, we next examined whether this state is associated with changes in microglial clocks. Microglia were purified every 4 hours across a full circadian cycle from both control and neuroinflammatory mice, and gene expression was analyzed ( Figure 4A ). In control microglia, Iba1 and Itgam exhibited circadian oscillations, consistent with IBA1 and CD11b immunostaining results ( Figure 4B ; Table 3A ). In contrast, neuroinflammatory microglia showed overall induced expression of these genes, with Iba1 displaying a significant phase shift and Itgam losing rhythmicity ( Table 3A ). These results suggest that neuroinflammation disrupts clock-regulated expression patterns of microglial activation genes.

To investigate the potential mechanism, we analyzed core clock components. Bmal1 expression was phase-shifted in neuroinflammatory microglia, and its downstream target Per1 lost rhythmicity ( Figure 4C ; Table 3B ). These findings indicate that the intrinsic microglial clocks are disrupted during neuroinflammation and may contribute to persistently enhanced microglial activation marker expression.

Under neuroinflammatory conditions, microglia secrete various cytokines and chemokines to promote pro-inflammatory reactions (1, 53, 54). To investigate whether microglial clocks modulate cytokine expression, we analyzed protein levels of secreted cytokines and chemokines in purified microglia. Under neuroinflammatory conditions, several factors, including CC chemokine ligands CCL3 and CCL5, were upregulated ( Supplementary Figure S2A ). Gene expression analysis revealed that Ccl5 was rhythmic under control conditions but lost rhythmicity during neuroinflammation. In contrast, Ccl3 remained arrhythmic in both conditions, with overall expression induced during neuroinflammation ( Figure 4D ; Table 2C ). To further assess whether microglial clocks regulate cytokine expression more broadly, we examined additional genes, including Ccl12 and Il-1β, and found that their expression was non-rhythmic even under physiological conditions ( Supplementary Figures S2B, C ). These findings suggest that microglial clocks selectively regulate a subset of immune-related genes, and that this regulation is disrupted during neuroinflammation.

Disrupted BMAL1 downstream in neuroinflammatory microglia led us to hypothesize that clock dysfunction contributes to the loss of microglial physiological functions. To investigate the role of microglial clocks under physiological conditions, we used two transgenic mouse lines: CX₃CR-1^GFP knock-in mice, which express enhanced green fluorescent proteins (EGFP) in microglia and other myeloid cells ( Supplementary Figure S3 ) (41), and NG2^DsRed mice, which express red fluorescent protein under the neural/glial antigen 2 (NG2) promoter to label NG2⁺ OPCs (42).

Under physiological conditions, microglia support oligodendrocyte lineage cells by secreting trophic factors such as insulin-like grow factor 1 (IGF-1) and transforming growth factor β (TGF-β), which promote OPC recruitment, proliferation, and differentiation (55–58). These OPCs replace senescent or damaged oligodendrocytes and contribute to spontaneous myelin turnover in the brain (59). To initiate repair, OPCs first migrate to affected areas and then differentiate into mature oligodendrocytes that re-form myelin around axons (60–62). To test whether microglial clocks contribute to OPC recruitment, we crossed CX₃CR-1^GFP mice with conventional Bmal1 knockout (BKO) mice to generate CX₃CR-1^GFP: BKO line. Purified GFP-expressing microglia from CX₃CR-1^GFP (mWT) and CX₃CR-1^GFP: BKO (mBKO) mice were transplanted into the corpus callosum (CC) of NG2^DsRed mice ( Figure 5A ). The CC provides an ideal site for OPC analysis due to its high density of parenchymal OPCs and aligned myelinated axons that serve as migration tracks. Three days after transplantation, significantly more NG2&⁺ cells were detected at mWT injection sites compared to mBKO ( Figures 5B, D ). In the mBKO-injected CC, some red fluorescence overlapped with nuclear counterstain, indicating non-specific signals, as NG2 is a cytoplasmic marker. These results suggest that functional microglial clocks are required for effective OPC recruitment.

To determine whether this increase was due to OPC migration rather than proliferation, we co-stained NG2&⁺ cells with the proliferation marker Ki67. Most NG2&⁺ cells were Ki67–negative at both the injection site (#1) and surrounding areas (#2) ( Supplementary Figure S4 ), indicating that microglial clocks primarily influence recruitment rather than local OPC proliferation.

To test whether this effect is specific to microglia, we transplanted purified GFP-expressing astrocytes from Aldh1l1-EGFP mice ( Supplementary Figure S5 ) into the CC of NG2^DsRed mice. In contrast to mWT microglia, astrocyte transplantation did not increase NG2&⁺ cell numbers compared to mBKO-injected sites ( Figures 5C, D ). These results indicate a microglia-specific, clock-dependent mechanism underlying OPC recruitment under physiological conditions.

To test whether neuroinflammatory conditions impair the clock-dependent function of microglia in supporting OPC recruitment, we injected purified microglia from control and neuroinflammatory CX₃CR-1^GFP mice into the CC of NG2^DsRed mice ( Figure 6A ). Three days after transplantation, significantly fewer NG2&⁺ cells were observed at sited injected with neuroinflammatory microglia compared to control microglia ( Figures 6B, C ). This result suggests that, under neuroinflammatory conditions, microglia shift their functional priority from physiological roles, such as OPC recruitment, toward immune responses.

In summary, our study demonstrates that functional microglial clocks are essential for supporting OPC recruitment under physiological conditions. During neuroinflammation, however, Bmal1 expression in microglia undergoes a phase shift, and its target gene Per1 – as well as microglial activation markers Iba1 and Itgam – loses rhythmicity while their expression levels increase. This disruption in circadian clock output is accompanied by sustained cytokine production and loss of rhythmic regulation, leading to a persistently activated microglia state in neuroinflammation. These findings highlight the importance of intact microglial clocks in maintaining physiological roles and suggest that preserving circadian clock function may help mitigate excessive neuroinflammatory responses while supporting beneficial microglial activity.