Neuroinflammatory suppression of astrocytic BMAL1 defines a selective reactive astrocyte program

Abstract

Astrocyte reactivity is a hallmark of neuroinflammatory diseases. Astrocytes contain functional circadian clocks that drive daily homeostatic rhythms, making them potential regulators of the transition from homeostatic to reactive states. However, it remains unclear whether neuroinflammatory conditions alter endogenous astrocyte clocks and whether these clock alterations contribute to pathway changes associated with reactivity. Here, we combined in vivo and in vitro lipopolysaccharide (LPS)-based neuroinflammation models with circadian profiling, chromatin analysis, and bulk RNA-sequencing of purified microglia and astrocytes. In vivo and in vitro, neuroinflammation reduced the amplitude of Bmal1 and Per1 rhythms and changed clock gene expression patterns across circadian time, while reactive markers gained rhythmic expression. In vitro, these clock changes occurred in astrocytes exposed to the inflammatory glial environment, but not after direct LPS treatment, indicating that paracrine glial signaling is the main driver. RNA-sequencing of microglia from the same neuroinflammatory glial cultures identified Il1b and Tnf as strong candidate mediators, and combined IL-1β/TNFα treatment was sufficient to suppress astrocyte clock gene expression and induce reactive marker expression. This clock suppression was accompanied by a modest reduction in BMAL1 protein but a stronger reduction in average BMAL1 occupancy at multiple clock target loci, including Per1 and Nr1d1. To define BMAL1-dependent pathways in astrocytes, we compared gene expression in wild-type and Bmal1-deficient astrocytes, and then compared these changes with those observed under neuroinflammatory conditions. Bmal1 loss alone did not reproduce the full inflammatory response. Instead, it identified a selective subset of the reactive astrocyte transcriptome, mainly involving suppression of cell-cycle and chromosome-segregation pathways with limited innate immune activation. Together, these findings identify astrocytic BMAL1 as a regulator of a selective pathway subset within reactive astrocytes in response to neuroinflammatory conditions.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12974-026-03953-6.

Introduction

Astrocytes are the most abundant glial cells in the brain and maintain neuronal homeostasis through metabolic support, neurotransmitter uptake, and synapse assembly and maintenance [1, 2]. During brain injury or inflammation, this homeostatic state is disrupted and astrocytes undergo reactive changes. Reactive astrocytes are common across neurological and neuroinflammatory diseases, including Alzheimer’s disease, multiple sclerosis, and sepsis-associated encephalopathy [3, 4]. One major route to this transition is early microglial activation, which releases cytokines including interleukin-1α (IL-1α), tumor necrosis factor α (TNFα), and complement component 1q (C1q) that can drive astrocytes into a neuroinflammatory reactive state [5–7]. Reactive astrocytes are characterized by upregulation of pan-reactive markers, including Gfap, Lcn2, and Serpina3n, and by morphological hypertrophy [2, 8]. Their transcriptomes also change in a stimulus-dependent manner, and at least half of the altered genes can be injury-specific [8–10]. Depending on the insult and disease state, reactive astrocytes can amplify neuroinflammatory signaling, contribute to neuronal damage, form glial scars, or support tissue protection [11–15]. However, the cell-intrinsic mechanisms that shift astrocytes from homeostatic to reactive states and shape distinct reactive outputs remain incompletely understood.

Circadian rhythms are generated by a cell-autonomous transcriptional-translational feedback loop (TTFL), the circadian clock. In this loop, the BMAL1/CLOCK heterodimer drives transcription of thousands of target genes, including its own repressors, Period (Per1/2/3) and Cryptochrome (Cry1/2), by binding E-box cis-regulatory element in their promoters [16, 17]. PER/CRY complexes suppress BMAL1/CLOCK activity, and their subsequent degradation relieves this repression to initiate the next ~ 24-hour cycle [18, 19]. A secondary interlocked loop, in which RORs activate and REV-ERBα/β, also known as NR1D1/NR1D2, repress Bmal1 transcription, reinforces the core TTFL and stabilizes oscillation amplitude [20, 21]. This molecular clock is present in most cell types, including brain cells, and drives daily rhythms in cell physiology through rhythmic expression of clock-controlled genes [22–24]. Through clock-controlled transcription, BMAL1 also regulates cell-type-specific programs, including immune responses, metabolism, and stress responses [25, 26].

In peripheral immune cells, inflammatory signaling and the circadian clock interact bidirectionally, establishing a clock-immunity axis [25]. In macrophages, nuclear factor kappa B (NF-κB) activation suppresses Bmal1 expression in part through miR-155-dependent post-transcriptional regulation [27]. At the chromatin level, the NF-κB subunit RELA directly binds the BMAL1 transactivation domain and suppresses BMAL1/CLOCK-dependent transcription at E-box sites by competing for the coactivator CREB-binding protein (CBP)/p300 [28, 29]. BMAL1 also restricts NF-κB-driven inflammatory gene expression by counteracting CLOCK-dependent NF-κB transactivation [30]. These findings support bidirectional clock-immune cross-talk and raise the possibility that similar mechanisms operate in astrocytes. A previous study showed that TNFα can directly modulates the molecular clock of astrocytes and alter clock gene oscillations [31]. Astrocyte-specific deletion of Bmal1 induces a cell-autonomous reactive phenotype, including increased Gfap and Serpina3n, and inflammatory-response genes [32, 33]. However, whether neuroinflammatory conditions alter astrocyte clock gene expression and BMAL1 transcriptional activity, and whether these alterations contribute to pathway changes associated with astrocyte reactivity, remains unknown.

Here, we combined in vivo and in vitro LPS-based neuroinflammation models with circadian profiling, chromatin analysis, and genome-wide transcriptomics to examine clock control in reactive astrocytes. We show that neuroinflammatory conditions suppress astrocyte clock gene expression, reduce average BMAL1 occupancy at multiple clock target loci, including Per1 and Nr1d1, and induce rhythmic reactive marker expression. In vitro, this clock suppression was driven by paracrine glial signaling rather than direct LPS exposure. We further used Bmal1-deficient astrocytes to identify BMAL1-dependent pathways and compare them with the astrocyte response to neuroinflammation. This analysis separates BMAL1-linked pathways from the broader inflammatory response in reactive astrocytes. Together, these findings identify BMAL1-dependent clock regulation as a link between neuroinflammatory signaling and selective reactive astrocyte pathways.

Materials and methods

Mouse

Bmal1 ^+/− mice (JAX stock #009100) were purchased from The Jackson Laboratory [34]. Bmal1^−/− (BKO) mice were generated by crossing Bmal1^+/− mice. Wild-type (WT) and BKO mice were maintained on a C57BL/6J background under a 12-h light/12-h dark cycle with ad libitum access to food and water. Both male and female adult mice aged 8–15 weeks were used for the in vivo circadian experiments. Data from male and female mice were pooled for the main analysis. All experimental procedures were approved by the Animal Research Ethics Subcommittee at City University of Hong Kong and the Department of Health, the Government of the Hong Kong Special Administrative Region.

Systemic inflammation (SI) in vivo

To induce systemic inflammation, 8- to 15-week-old mice were intraperitoneally injected with Lipopolysaccharide (LPS, 5 mg/kg body weight, L7895, Sigma-Aldrich ) dissolved in endotoxin-free Dulbecco’s Phosphate-Buffered Saline (DPBS, #A1285601, Gibco) [35]. Control mice received an equivalent volume of vehicle. At 3, 7, and 14 days post-injection, mice were euthanized; brains were either processed immediately for astrocyte purification followed by RNA extraction, or fixed for immunohistochemistry.

Primary glial culture

Primary glia were isolated from the cortices of postnatal day 1–2 (P1–P2) WT or BKO pups. After decapitation and cranium removal, brains were placed in cold dissection medium [1× HBSS with 2.5 mM HEPES, 5.4 g/L glucose, 100 U/mL penicillin, and 100 µg/mL streptomycin (all from Gibco)]. Cortices were dissected, meninges were removed, and tissues were digested in 0.25% trypsin for 15 min at 37 °C. Digestion was stopped with culture medium [DMEM/F-12 containing 10% fetal bovine serum (Biosera), 1× GlutaMAX, 16 mM HEPES, 110 U/mL penicillin, and 110 µg/mL streptomycin (all from Gibco)]. After centrifugation at 400 × g for 5 min, the cell pellet was resuspended in culture medium, triturated sequentially with P1000, P200 pipettes, and a 21G needle, then filtered through a 70 μm cell strainer (Falcon). Cells were plated on poly-D-lysine (PDL)-coated T75 or T25 flasks and cultured at 37 °C in a humidified 5% CO₂ incubator. Medium was changed every 3–4 days. Mixed glial cultures reached confluency at 6–8 days in vitro (DIV) and were used for subculture or experiments within passage 3.

Glial culture synchronization and treatment

To induce neuroinflammation in vitro, 1 × 10⁶ glial cells were plated per 6 cm dish for 24 h. Cultures were then synchronized with 0.1 mM dexamethasone (D4902, Sigma) for 2 h, after which the medium was replaced with fresh medium containing either 0.5 ng/µL LPS or vehicle (endotoxin-free DPBS). To assess time-dependent responses, samples were harvested every 4 h on day 3 after LPS treatment.

For cytokine treatment, 1 × 10⁶ purified astrocytes were plated per 6 cm dish. After 24 h, cells were treated for 48 h with recombinant cytokines [50 ng/mL IL-1β (#211-11B, Peprotech) and 10 ng/mL TNF-α (#315–01 A, Peprotech)] or vehicle (PBS with 0.1% BSA). Total RNA was then extracted for subsequent analysis.

Purification of astrocytes

Astrocytes were purified by magnetic-activated cell sorting (MACS, Miltenyi) according to the manufacturer’s instructions and previously described [36].

Purification from primary glial culture

Cells were trypsinized with 0.05% trypsin for 7 min at 37 °C. After detrypsinization with culture medium, the cell suspension was centrifuged at 400 × g for 5 min. The cell pellet from a T25 flask was resuspended in 80 µL of MACS buffer [0.5% (w/v) bovine serum albumin (BSA; Sigma-Aldrich) and 2 mM EDTA (Affymetrix) in PBS]. 10 µL of FcR Blocking Reagent (#130-097-678, Miltenyi) was added, and the mixture was incubated at 4 °C for 10 min with rotation. Then 10 µL of Anti-ACSA-2 Microbeads (Miltenyi) were added, followed by incubation at 4 °C for 15 min with rotation. Cells were washed, resuspended in 500 µL of MACS buffer, and applied to a pre-wetted MS column (#130-042-201, Miltenyi). After washing, astrocytes were eluted with 1 mL of MACS buffer.

Purification from adult mouse cortex

Brains were minced and digested in 0.025% trypsin for 15 min at 37 °C. The digested tissue was gently triturated, filtered through a 70 μm cell strainer, and centrifuged at 400 × g for 5 min. The pellet was resuspended in red blood cell lysis buffer (155 mM NH₄Cl, 12 mM NaHCO₃, 0.1 mM EDTA) and incubated for 5 min at room temperature (RT). After washing with MACS buffer, myelin was removed by incubating the cells with 60 µL myelin removal beads (#130-096-731, Miltenyi) per brain at 4 °C for 20 min, followed by passage through a pre-wetted LS column (#130-042-401, Miltenyi). The flow-through (myelin-depleted fraction) was collected, and cells were then subjected to astrocyte-specific purification: they were resuspended in 80 µL MACS buffer, mixed with 10 µL FcR Blocking Reagent, and incubated at 4 °C for 10 min with rotation. Then 10 µL Anti-ACSA-2 Microbeads was added, and the mixture was incubated at 4 °C for 15 min with rotation. After washing and resuspension in 500 µL MACS buffer, the cell suspension was applied to a pre-wetted MS column. After washing, astrocytes were eluted with 1 mL of MACS buffer.

RNA extraction and Quantitative Real-time PCR (qRT-PCR)

Total RNA was extracted using TRIzol reagent (Invitrogen) or a column-based method (TaKaRa MiniBEST Universal RNA Extraction Kit) according to the manufacturer’s instructions. RNA concentration and purity were assessed using a NanoDrop spectrophotometer. For reverse transcription, 500 ng of total RNA was converted to cDNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). qRT-PCR was performed using SYBR Premix Ex Taq (Tli RNase H Plus) (TaKaRa) on a QuantStudio 3 Real-Time PCR System (Applied Biosystems). Each 20 µL reaction contained 2µL of diluted cDNA. The cycling conditions were as follows: 95 °C for 30 s, followed by 35 cycles of 95 °C for 5 s and 60 °C for 30 s. Gene expression was normalized to the ribosomal protein gene Rps13 and calculated using the 2^−ΔΔCT method. Primer sequences used are listed in Table 1.

Table 1.

List of the primers used in this study

Gene	Forward sequence	Reverse sequence
Rps13	TCCCTCCCAGATAGGTGTAATCC	TCCTTTCTGTTCCTCTCAAGGT
Rpl30	GCTGGAGTCGATCAACTCTAGG	CCAATTTCGCTTTGCCTTGTC
Bmal1	TGACCCTCATGGAAGGTTAGAA	GGACATTGCATTGCATGTTGG
Per1	CAGCTGGGCCGGTTTTG	CACTTTATGGCGACCCAACA
Cry1	CTGGCGTGGAAGTCATCGT	CTGTCCGCCATTGAGTTCTATG
Itgam	ATGGACGCTGATGGCAATACC	TCCCCATTCACGTCTCCCA
Iba1	CTTTTGGACTGCTGAAGGC	GTTTCTCCAGCATTCGCTTC
Gfap	AGAAAGGTTGAATCGCTGGA	CGGCGATAGTCGTTAGCTTC
Lcn2	GGACCAGGGCTGTCGCTACT	GGTGGCCACTTGCACATTGT
Serpina3n	CTCTCAGGTGGTCCACAAGG	GAGGGTACAGTTTCGCAGACA

Flow Cytometry (FCM)

Cultured cells were trypsinized, washed twice with DPBS (Gibco), and fixed in 4% paraformaldehyde (PFA, VWR) for 15 min at RT. Cells were then permeabilized with ice-cold 90% methanol in 1×PBS for 10 min. After washing with 1×PBS containing 1% BSA, cells were incubated with primary antibodies diluted in 1% BSA for 1.5 h at RT. Cells were then washed and incubated with fluorescently conjugated secondary antibodies in 1% BSA for 1 h under the darkness. After two final washes with 1×PBS, cells were resuspended in 1 mL 1×PBS and analyzed on a CytoFLEX Flow Cytometer (Beckman Coulter). Data were analyzed using FlowJo V10 software [37]. Antibodies used are listed in Table 2.

Table 2.

List of antibodies used in this study

Antibody (Host)	Catalog number	Manufacturer	Dilution/Amount	Application
Primary Antibodies
GFAP (Mouse)	#835,301	Biolegend	1:1000	FCM, IHC
CD11b (Rat)	#101,202	Biolegend	1:100	FCM
BMAL1 (Rabbit)	#A302-616 A	Bethyl	1:1000/5 µg	FCM, ChIP
H3K27ac (Rabbit)	#ab4729,	abcam	5 µg	ChIP
RNAPII (Mouse)	#2629	Cell Signaling	5 µg	ChIP
Secondary Antibodies
Anti-Mouse IgG, AF®488-conjugated	#715-545-150	Jackson ImmunoResearch	1:400	IHC
Anti-Mouse IgG, AF®647-conjugated	#715-605-150	Jackson ImmunoResearch	1:400	FCM
Anti-Rat IgG,  AF®594-conjugated	#712-585-150	Jackson ImmunoResearch	1:400	IHC
Anti-Rat IgG, AF®647-conjugated	#A21247	Invitrogen	1:400	FCM
IgG (Rabbit)	#ab6715	abcam	5 µg	ChIP
IgG (Mouse)	#ab18413	abcam	5 µg	ChIP

Immunostaining

Mice were anesthetized and transcardially perfused with 40 mL ice-cold 1×PBS followed by 40 mL 4% (w/v) PFA in 1×PBS. Brains were harvested, post-fixed in 4% PFA for 40 h at 4 °C, and then dehydrated in 30% (w/v) sucrose (ChemCruz) in 1×PBS at 4 °C until the tissue sank. Brains were embedded in OCT compound (Epredia) and frozen. Coronal Sect.  (20 μm thickness) were cut using a cryostat and mounted on adhesion slides (SuperFrost Plus, Thermo). Then, sections were blocked in PGBA blocking buffer [0.1% (w/v) gelatin, 1% (w/v) BSA, 0.002% (w/v) sodium azide, 10% (v/v) normal goat serum, and 0.5% (v/v) Triton X-100 in 0.1 M phosphate buffer (PB)] for 1 h at RT. Sections were then incubated with primary antibodies diluted in the same blocking buffer overnight at 4 °C. The next day, sections were washed three times with 0.1 M PB (10 min each) and then incubated with species-appropriate fluorescent secondary antibodies (all from Jackson ImmunoResearch, 1:400) in blocking buffer for 1 h at RT in the dark. After incubation, sections were washed three times with 0.1 M PB, and nuclei were counterstained with DAPI contained in the mounting medium (Vectashield, H-1200). Antibodies used are listed in Table 2.

Imaging

Fluorescent images were acquired using an A1R HD25 Confocal Microscope (Nikon) with a 20× objective and NIS-Elements software (Nikon). All images for comparison were captured with identical exposure and laser settings. Fluorescence intensity was quantified on randomly selected fields from at least 3 independent samples per condition. Mean fluorescence intensity (intensity per area) for GFAP was measured using Fiji/ImageJ [38]. To quantify GFAP-positive astrocyte density in the brain, brain sections from mice treated with PBS or LPS at 3, 7, and 14 days post-injection were analyzed. For each animal, counting frames (0.685 mm per side, area 0.47 mm²) were randomly placed within the cortex. GFAP-immunoreactive cells with a clearly identifiable DAPI-labeled nucleus residing within the 20-µm-thick optical section were manually counted. Cell counting was performed under single-blind conditions.

Chromatin immunoprecipitation (ChIP)-Quantitative PCR (qPCR)

ChIP

ChIP-qPCR was performed as previously described using non-synchronized control and NI mixed glial cultures to compare average BMAL1 occupancy under each condition [39]. Mixed glial cultures grown in 15 cm dishes were trypsinized, pelleted at 400 g for 5 min, washed with 1×PBS, and resuspended in culture medium. Cells were counted using a Countess 3 automated cell counter (Invitrogen), and 0.8 × 10⁷ cells in 3 mL medium were fixed with 1% TEM-grade PFA (Tousimis) for 10 min with rotation. Fixation was quenched by adding 0.25 M glycine (Affymetrix) for 5 min with rotation. After centrifugation at 800 g for 5 min at 4 °C and two washes with 1×PBS, cell pellets were resuspended in 300 µL cell lysis buffer [0.5% NP-40, 5 mM PIPES pH 8, 85 mM KCl] containing protease inhibitor cocktail and 1 mM DTT. Following centrifugation at 1,000 g for 5 min at 4 °C, the supernatant was removed and the nuclear pellet was resuspended in 200 µL nuclear lysis buffer [1% SDS, 10 mM EDTA, 50 mM Tris-HCl pH 8] with protease inhibitor cocktail and DTT. Chromatin was sonicated using a Bioruptor (Diagenode) on high setting with 30 s on/off cycles for 4 cycles. After centrifugation at 12,000 g for 10 min at 4 °C, the supernatant containing sheared chromatin was transferred to a new 1.5 ml tube on ice. Sonication efficiency was verified by mixing 20 µL of chromatin with 80 µl ChIP dilution buffer [0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl pH 8, 167 mM NaCl] and 1 µl proteinase K (#P8107, New England BioLabs) at 62 °C for 2 h with shaking at 1,500 rpm, followed by heat reversal at 95 °C for 10 min, DNA purification using a PCR purification kit (#28104, QIAGEN), and electrophoresis on a 2% agarose gel.

For immunoprecipitation, chromatin was diluted with ChIP dilution buffer containing protease inhibitor cocktail to achieve 5 µg chromatin per reaction in 500 µL. A 5 µL aliquot of each reaction was saved as input. Antibodies (Table 2) were added, and 20 µL of Dynabeads™ Protein G (pre-washed three times with ChIP dilution buffer) were added to each sample, followed by overnight rotation at 4 °C. Beads were then washed sequentially with low salt wash buffer [0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl pH 8, 300 mM NaCl], high salt wash buffer [500 mM NaCl], LiCl wash buffer [1 mM EDTA, 10 mM Tris-HCl pH 8, 1% NP-40, 250 mM LiCl], and finally TE buffer (10 mM Tris-HCl pH 8, 1 mM EDTA). Crosslinking was reversed by incubating beads in 100 µL TE buffer containing 1% SDS, 150 mM NaCl, and 1:100 proteinase K at 62 °C for 2 h with shaking at 1,500 rpm. DNA was then boiled at 95 °C for 10 min and purified using a PCR purification kit, eluting in 50 µL TE buffer.

qPCR

Purified DNA was diluted 1:5 with TE buffer, and 2 µL of diluted DNA was used as template per qPCR reaction. Enrichment at multiple clock target E-box regions was quantified by ChIP-qPCR. Primer sequences for all ChIP-qPCR assays are listed in Table 3. All samples, including input DNA, were run in technical triplicate in each experiment. Enrichment was calculated using the percent input method.

Table 3.

List of the primers for ChIP-qPCR

Gene	Forward sequence	Reverse sequence
Per1 E-box	ATCCTGATCGCATTGGCTGACTGA	TCTCTTCCTGGCATCTGATTGGCT
Nr1d1 E-box	GCTGCTGGAAAAGTGTGTCA	ATGGAGAAATGAGGCACCAG
Per2 E-box	GCGTCTCCACCAATTGATGA	CCCTCCGCTGTCACATAGT
Dbp E-box	ACACCCGCATCCGATAGC	CCACTTCGGGCCAATGAG

Statistical analysis

Data are presented as mean ± SEM, as indicated in the figure legends. For pairwise comparisons, except for RNA-sequencing and circadian rhythmicity analyses, an unpaired two-tailed Student’s t-test was used. Circadian rhythmicity of all examined genes was analyzed using MetaCycle [40, 41]. To quantify the magnitude of circadian parameter changes between groups, amplitude change percentage and estimated phase change percentage were calculated using the following formulas: amplitude change (%) = (amplitude difference / |mean amplitude of control group|) × 100; estimated phase change (%) = (circular phase difference / 24 h) × 100. Estimated phase differences were calculated on a 24-h circular scale. Statistical significance of rhythmic parameter differences between control and LPS-treated groups was assessed by bootstrap resampling with 1,000 iterations [42]. All statistical analyses were performed using GraphPad Prism 9 (San Diego, CA), unless otherwise indicated. P < 0.05 was considered statistically significant.

RNA sequencing and data analysis

RNA-sequencing data processing

Raw FastQ files were processed using the Rsubread package (v2.18.0) [43]. Sequence alignment was performed against the mouse reference genome (GRCm39/mm39). Gene quantification was conducted with the featureCounts function using GENCODE vM30 annotations [44]. The raw count matrix was integrated with metadata and stored as a SummarizedExperiment object (v1.36.0) for downstream analysis in R (v4.4.3) [45].

Differential Expression Gene (DEG) analysis

DEA was performed using the DESeq2 (v1.46.0) [46]. For both astrocytes and microglia, genes with fewer than 10 counts in at least 3 samples were pre-filtered. Principal component analysis (PCA) was used for quality control via the variance stabilizing transformation (vst). Differentially expressed genes (DEGs) were identified using the Wald test with apeglm (v1.28.0) for log₂ fold change shrinkage [47]. Significance thresholds: shrunken |log2FC| ≥ 1.0, BH-adjusted P < 0.01, baseMean ≥ 50 for microglia.

Gene Set Enrichment Analysis (GSEA)

Functional grouping and semantic similarity clustering were performed using the simplifyEnrichment package (v2.16.0) and visualized using ComplexHeatmap (v2.22.0) [48, 49]. GSEA was performed with clusterProfiler (v4.12.0) using Gene Ontology (GO) Biological Process terms [50]. Genes were ranked by the DESeq2 Wald statistic. Significance was defined as adjusted P < 0.01 and |normalized enrichment score (NES)| ≥ 2.0 were retained for downstream analysis. These significant terms were functional grouped by semantic similarity using simplifyEnrichment (v2.16.0) and visualized with ComplexHeatmap (v2.22.0).

Results

Systemic inflammation reduces the amplitude of astrocyte clock gene rhythms in the brain

To test whether the reactive astrocyte state transition is coupled to alterations in astrocyte circadian clocks, we induced systemic inflammation (SI) in mice by intraperitoneal LPS injection. Astrocyte reactivity was confirmed at 3 days post-treatment (Supplementary Fig. 1), and brains were collected across the circadian cycle for astrocyte purification and transcript quantification (Fig. 1A). We examined the expression of Bmal1, a positive-loop gene, and Per1, a negative-loop gene, because changes in either limb indicate altered clock gene rhythms. SI astrocytes showed reduced amplitudes in both Bmal1 and Per1 rhythms than vehicle-treated controls. Estimated phase of these rhythms also differed between groups (Fig. 1B). In the same cells, pan-reactive markers Gfap, Lcn2, and Serpina3n were induced across most circadian time points (Fig. 1C). Reactive markers also displayed a statistically rhythmic expression pattern under SI. Although statistical analysis classified these markers as rhythmic in controls, their expression was near-baseline and therefore difficult to interpret biologically in the resting state. Systemic inflammation thus reduces the amplitude of astrocyte clock gene rhythms and induces rhythmic reactive-state gene expression across the circadian cycle.

Fig. 1.

Systemic inflammation reduces the amplitude of astrocyte clock gene rhythms in the brain. (A) Experimental design for systemic inflammation (SI). Mice received intraperitoneal (i.p.) injection of LPS or vehicle (PBS; Ctrl), remained under a light/dark cycle (LD) for 2 days, and were then transferred to constant darkness (DD) for 1 day. Cortices were collected every 4 h across circadian time (CT), followed by astrocyte purification and qRT-PCR analysis. (B) Relative mRNA levels of core clock genes (Bmal1, Per1) and control gene (Rpl30), and (C) pan-reactive astrocyte markers (Gfap, Lcn2, Serpina3n), measured by qRT-PCR in purified astrocytes from Ctrl and SI mice across CT. (B, C) Lower panels: Rhythmicity was analyzed using MetaCycle, and estimated phase and amplitude differences were assessed by bootstrap resampling with 1,000 iterations

Neuroinflammation suppresses astrocyte clock gene expression through paracrine signaling in glial cultures

To understand how neuroinflammation reduces astrocyte clock gene expression, we used primary mixed glial cultures treated with LPS to model neuroinflammatory conditions in vitro. These cultures were mainly astrocyte-enriched, with 95.7% astrocytes and 3.66% microglia (Fig. 2A). LPS treatment in glial cultures increased activated microglial markers Itgam and Iba1 and reactive astrocyte markers Lcn2 and Serpina3n, confirming induction of neuroinflammation (NI) in the culture (Fig. 2B).

Fig. 2.

Neuroinflammation suppresses astrocyte clock gene expression through paracrine signaling in glial cultures. (A) Flow cytometry analysis of mixed glial cultures. Astrocytes were identified by GFAP staining and microglia by CD11b staining. (B) Relative mRNA levels of activated microglial marker genes (Itgam, Iba1), reactive astrocyte marker genes (Lcn2, Serpina3n), and control gene (Rpl30) measured by qRT-PCR in non-synchronized glial cultures after LPS treatment to induce neuroinflammatory (NI) conditions. Samples were collected at 24, 48, 72, and 96 h after treatment. Untreated cultures served as controls (0 h). (C) Experimental workflow for testing direct LPS versus paracrine glial signaling in astrocyte clock suppression. In (i), astrocytes were purified from mixed glial cultures and treated directly with PBS (control; Ctrl) or LPS for 2 days. In (ii), mixed glial cultures were treated with PBS or LPS (neuroinflammation; NI) for 2 days, followed by astrocyte purification and qRT-PCR. (D) Relative mRNA levels of core clock genes (Bmal1, Per1), pan-reactive astrocyte marker gene (Lcn2), and control gene (Rpl30) measured by qRT-PCR in purified astrocytes after direct LPS treatment (left) or after purification from Ctrl and NI glial cultures (right). Each dot represents one independent experiment. (E) Experimental workflow for circadian gene expression profiling in vitro. Mixed glial cultures were synchronized with dexamethasone (dexa) for 2 h, then treated with LPS in fresh medium to induce NI conditions or with PBS as control. Cells were harvested every 4 h across circadian time (CT) for qRT-PCR analysis. (F) Relative mRNA levels of core clock genes (Bmal1, Per1) and the control gene (Rpl30), and (G) pan- reactive astrocyte marker genes (Lcn2, Serpina3n), measured by qRT-PCR in glial cultures under Ctrl and NI conditions across a 24-h cycle. (F and G) Lower panels: Rhythmicity was analyzed using MetaCycle, and estimated phase and amplitude differences were assessed by bootstrap resampling with 1,000 iterations. Data are shown as mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001 (Student’s two-tailed t-test)

We first measured average gene expression after LPS treatment in non-synchronized cultures. In this experiment, cellular clocks were not phase-aligned. Therefore, this experiment measured the average expression of clock genes and reactive markers across the mixed-phase culture after LPS treatment. Clock genes Bmal1, Per1, and Cry1 were reduced after LPS treatment, whereas reactive astrocyte markers Lcn2 and Serpina3n were induced (Supplementary Fig. 2). Thus, LPS-induced neuroinflammation decreases average clock gene expression while inducing reactive marker expression.

We next asked whether astrocyte clock suppression is caused by direct LPS exposure or by inflammatory signals from the glial environment. We compared clock gene expression between purified astrocytes treated directly with LPS and astrocytes purified from neuroinflammatory glial cultures exposed to the same LPS dose (Fig. 2C). Direct LPS caused statistically non-significant reductions of 7.83% in Bmal1 (P = 0.216) and 11.8% in Per1 (P = 0.221). In contrast, astrocytes purified from NI glial cultures showed 39.43% and 29.76% reductions in Bmal1 (P = 0.00119) and Per1 (P = 0.00255), respectively (Fig. 2D). These results indicate that the glial environment, rather than direct LPS exposure alone, is the main driver of astrocyte clock suppression in this model.

We then examined clock gene rhythms after circadian synchronization. Glial cultures were synchronized with dexamethasone and harvested every 4 h across one circadian cycle (Fig. 2E). Under neuroinflammatory conditions, Bmal1 and Per1 showed reduced amplitudes compared with controls. Their estimated phase also differed between groups (Fig. 2F). Reactive markers Lcn2 and Serpina3n were strongly induced and showed circadian rhythmic expression under NI conditions (Fig. 2G). Together, these data show that in vitro neuroinflammation suppresses astrocyte clock gene expression through paracrine inflammatory signaling. This response parallels the astrocyte clock gene changes observed in mice with systemic inflammation.

Microglia-derived IL-1β and TNFα are sufficient to suppress astrocyte clock gene expression

The paracrine dependence of astrocyte clock suppression pointed to microglia-secreted factors as candidate mediators. To identify candidates, we performed bulk RNA-seq on microglia purified from control and neuroinflammatory glial cultures (Fig. 3A). NI microglia showed 2,263 DEGs relative to control microglia, including 1,271 upregulated and 992 downregulated genes (shrunken |log2FC| ≥ 1.0, BH-adjusted P < 0.01, baseMean ≥ 50; Fig. 3B, Supplementary Table 1). Upregulated DEGs included multiple cytokines and chemokines, including Il1b and Tnf. Gene set enrichment analysis (GSEA) of Gene Ontology (GO) Biological Process (BP) terms identified 367 significant pathways in NI microglia (adjusted P < 0.01, |NES| ≥ 2.0). Based on grouped analysis, these pathways were divided into nine functional categories: interferon signaling, antigen presentation, and adaptive immune activation (Group 1); leukocyte activation, chemotaxis, and cytotoxicity (Group 2); interleukin and TNF signaling and production (Group 3); cytokine/chemokine signaling and inflammasome activation (Group 4); leukocyte effector function (Group 5); innate immune and host defense response (Group 6); cell cycle and chromosome segregation (Group 7); reactive oxygen/nitrogen species metabolism (Group 8); and other pathways (Group 9) (Fig. 3C, Supplementary Table 2).

Fig. 3.

Microglia-derived IL-1β and TNFα are sufficient to suppress astrocyte clock gene expression. (A) Experimental workflow for bulk RNA sequencing (RNA-seq). Primary microglia were purified from control (Ctrl) and LPS-treated neuroinflammatory (NI) mixed glial cultures and subjected to RNA-seq. (B) Heatmaps of the top upregulated and downregulated differentially expressed genes (DEGs) in NI microglia relative to Ctrl microglia. Heatmap color indicates relative expression level across samples, and the left side color scale indicates shrunken log2 fold change for each gene. (C) Gene set enrichment analysis (GSEA) of Gene Ontology (GO) Biological Process (BP) terms based on transcriptomic profiles of NI versus Ctrl microglia. Bars indicate representative-term normalized enrichment score (NES), and grayscale indicates −log10(adjusted P value). Only positive-NES functional groups are shown. Functional groups were defined as follows: Group 1, interferon signaling, antigen presentation, and adaptive immune activation; Group 2, leukocyte activation, chemotaxis, and cytotoxicity; Group 3, interleukin (IL) and tumor necrosis factor (TNF) signaling and production; Group 4, cytokine/chemokine signaling and inflammasome activation; Group 5, leukocyte effector function; Group 6, innate immune and host defense response; Group 8, reactive oxygen/nitrogen species metabolism; and Group 9, others. (D) Top 10 enriched GO BP terms within functional Group 3 (IL and TNF signaling and production) identified from the NI versus Ctrl microglial RNA-seq dataset. Bar length indicates NES, and grayscale indicates −log10(adjusted P value). (E) Relative mRNA levels of core clock genes (Bmal1, Per1, Cry1), pan-reactive astrocyte marker gene (Lcn2), and control gene (Rpl30) measured by qRT-PCR in purified primary astrocytes under control (Ctrl) or IL-1β + TNFα-treated (Cyto) conditions. Expression levels were normalized to a control gene. Each dot represents one independent experiment. Data are shown as mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001 (Student’s two-tailed t-test)

Groups 1 and 2 were positively enriched in NI microglia. These groups represented broad IFN-response and immune activation programs, including antigen presentation, leukocyte activation, chemotaxis, and cytotoxicity. However, they did not point to a focused secreted cytokine axis for functional testing. IFN ligand genes were also not identified as strongly upregulated microglial DEGs (Supplementary Table 1). In contrast, among the upregulated DEGs, the cytokine genes Il1b and Tnf were linked to enriched IL-1/TNF production and signaling pathways in Group 3 (Fig. 3, C and D, Supplementary Tables 1 and 2). Therefore, we prioritized IL-1β and TNFα as testable microglia-derived candidate mediators of astrocyte clock suppression.

We tested this directly by treating purified astrocytes with recombinant IL-1β and TNFα for 48 h. Bmal1 was reduced by approximately 40%, and Per1 and Cry1 were also significantly downregulated (Fig. 3E), closely matching the suppression observed in astrocytes purified from neuroinflammatory glial cultures (Fig. 2D). Lcn2 was induced in the same cells, confirming a reactive-like response (Fig. 3E). IL-1β and TNFα together are therefore sufficient to suppress astrocyte clock gene expression.

Neuroinflammation modestly reduces BMAL1 protein but more strongly reduces average BMAL1 occupancy at multiple clock target loci

Reduced astrocyte clock gene expression under neuroinflammatory conditions could reflect reduced BMAL1 protein abundance or reduced BMAL1 occupancy at clock target loci. To examine this, we first quantified BMAL1 protein in GFAP+ astrocytes from non-synchronized glial cultures by flow cytometry. Approximately 90% of astrocytes remained BMAL1-positive under both control and NI conditions. The BMAL1^hi fraction, defined as GFAP+ cells with BMAL1 fluorescence above the mode of the control distribution, decreased from 45.27 ± 1.75% to 36.90 ± 3.09% under NI conditions (mean reduction, 8.37 ± 4%; P < 0.05; Fig. 4A). Thus, BMAL1 protein was modestly reduced and shifted to lower levels across the astrocyte population under NI conditions.

Fig. 4.

Neuroinflammation modestly reduces BMAL1 protein but more strongly reduces average BMAL1 occupancy at multiple clock target loci. (A) Flow cytometry analysis of BMAL1 protein in GFAP+ astrocytes from control (Ctrl astro) and neuroinflammatory (NI astro) glial cultures. BMAL1hi astrocytes were defined as GFAP+ cells with BMAL1 fluorescence intensity above the mode of the Ctrl distribution. (B and C) ChIP-qPCR analysis of BMAL1, H3K27ac, and RNA polymerase II (RNA Pol II) enrichment at the Per1 (B) and Nr1d1 (C) promoter E-box regions in Ctrl and NI glial cultures. Rabbit IgG (Rb IgG) and mouse IgG (Mo IgG) were used as negative controls. Each dot represents one independent experiment. Data are shown as mean ± SEM. **P < 0.01, ***P < 0.001, ****P < 0.0001 (Student’s two-tailed t-test)

We next measured BMAL1 occupancy by ChIP-qPCR in non-synchronized glial cultures. This experiment was designed to compare average BMAL1 occupancy under control and neuroinflammatory conditions, not maximal or phase-specific occupancy. ChIP-qPCR showed a more than 3-fold decrease in BMAL1 occupancy at the Per1 promoter E-box region under NI conditions (Fig. 4B). BMAL1 occupancy at an additional canonical target locus, Nr1d1, was also reduced under NI conditions (Fig. 4C). H3K27ac and RNA Pol II enrichment at the Per1 and Nr1d1 promoters were also reduced (Fig. 4, B and C), supporting reduced transcriptional activity at these target sites. Additional ChIP-qPCR analysis showed reduced BMAL1 occupancy and RNA Pol II enrichment at the Per2 and Dbp promoter E-box regions under NI conditions (Supplementary Fig. 3). Together, these results show that neuroinflammation weakens BMAL1-dependent clock regulation, affecting BMAL1 occupancy at clock target loci more strongly than BMAL1 protein abundance.

BMAL1 deletion increases selected reactive astrocyte markers

We next used Bmal1-/- (BKO) astrocytes to define BMAL1-dependent changes and to determine whether BMAL1 deletion is sufficient to induce reactive marker expression without exogenous inflammatory stimulation. For this, we measured reactive marker expression in astrocytes purified from wild-type (WT) and BKO glial cultures under basal conditions. Bmal1 and Per1 were reduced in BKO astrocytes, confirming loss of BMAL1-dependent clock gene expression. In the same cells, Gfap and Serpina3n were increased, whereas Rpl30 was unchanged (Fig. 5A). These results show that Bmal1 loss alone is sufficient to increase selected reactive astrocyte markers under basal conditions. We then compared BKO astrocytes with astrocytes under neuroinflammatory conditions to identify BMAL1-dependent pathways that are also altered during neuroinflammation.

Fig. 5.

BMAL1 deletion increases selected reactive astrocyte markers and partially overlaps with neuroinflammatory astrocytes at the pathway level. (A) Relative mRNA levels of core clock genes (Bmal1, Per1), pan-reactive astrocyte marker genes (Gfap, Serpina3n), and the control gene (Rpl30) measured by qRT-PCR in purified astrocytes from wild-type (WT) and Bmal1-/- (BKO) glial cultures. Each dot represents one mouse. Data are shown as mean ± SEM. ***P < 0.001 (Student’s two-tailed t-test). (B) Experimental workflow for RNA-seq analysis of four groups: (i) neuroinflammatory glial cultures induced by LPS treatment (NI glia), (ii) purified astrocytes from NI glial cultures (NI astro), (iii) purified astrocytes from WT and BKO glial cultures (BKO astro), and (iv) purified astrocytes treated with IL-1β + TNFα (Cyto astro). (C) Summary table showing the number of enriched Gene Ontology (GO) Biological Process (BP) terms identified by GSEA and assigned to the same nine functional groups defined in Figure 3C, across the four contrasts shown in (B). (D) Number of shared positive-normalized enrichment score (NES; black) and negative-NES (gray) GO BP terms between indicated contrasts. (E) Heatmaps of leading-edge genes from Group 7 (cell cycle and chromosome segregation) and Groups 1 and 6 (Group 1+6; interferon signaling/adaptive immune activation and innate immune/host defense response) across BKO, NI, and Cyto astrocytes. Color indicates scaled relative expression. Lower panel, quantification of unique and shared leading-edge (LE) genes and the percentage of shared genes changing in the same direction between groups

BMAL1 deletion partially overlaps with neuroinflammatory astrocytes at the pathway level

To define BMAL1-dependent pathways that overlap with astrocyte changes under neuroinflammatory conditions, we performed bulk RNA-seq under four conditions: NI glial cultures (NI glia), astrocytes purified from NI glial cultures (NI astrocytes), astrocytes purified from BKO glial cultures (BKO astrocytes), and astrocytes treated with IL-1β + TNFα (Cyto astrocytes) (Fig. 5B). GSEA of GO BP terms (adjusted P < 0.01, |NES| ≥ 2.0) identified 365, 287, 73, and 398 significant pathways in NI glia, NI astrocytes, BKO astrocytes, and Cyto astrocytes, respectively. These pathways were grouped into the same nine functional categories applied to the microglial analysis (Fig. 5C, Supplementary Table 3). NI and Cyto astrocytes showed broad enrichment of immune and inflammatory programs spanning Groups 1–6, with Cyto astrocytes showing the broadest profile. NI astrocytes also showed strong Group 7 enrichment, which represents cell-cycle and chromosome-segregation pathways. In contrast, BKO astrocytes showed a narrower profile, dominated by Group 7, with additional enrichment in Group 9 pathways and limited representation of immune and inflammatory programs (Fig. 5C). Thus, Bmal1 loss alone did not reproduce the broad inflammatory profile of NI astrocytes or Cyto astrocytes. Instead, it induced a more restricted transcriptional state dominated by cell-cycle and chromosome-segregation pathway changes.

Pairwise pathway overlap supported this pattern (Fig. 5D, Supplementary Table 4). NI glia and NI astrocytes shared 106 concordant GO BP terms, nearly all with positive NES, indicating that inflammatory pathway programs induced in neuroinflammatory glial cultures were maintained after astrocyte purification. The largest overlap was between NI astrocytes and Cyto astrocytes, which shared 166 concordant terms. These were dominated by positive NES immune pathways, together with a subset of negative NES Group 7 terms and one Group 9 term, acetyl-CoA metabolic process. This indicates that IL-1β + TNFα reproduces the major pathway-level features of the NI astrocyte state. Overlap with BKO astrocytes was smaller. NI astrocytes and BKO astrocytes shared 34 concordant terms, and Cyto astrocytes and BKO astrocytes shared 27. These overlapping pathways were dominated by negative NES Group 7 terms, with limited positive NES overlap in immune categories. These results indicate that Bmal1 loss overlaps with the NI astrocyte program mainly through suppression of cell-cycle and chromosome-segregation pathways, not through broad activation of inflammatory pathways.

Leading-edge gene analysis then defined the gene programs underlying these shared overlaps (Fig. 5E). For Group 7, BKO astrocytes, NI astrocytes, and Cyto astrocytes contributed 222, 465, and 216 unique leading-edge genes, respectively. Among them, 120 genes were shared between BKO and NI astrocytes, and 60 were shared across all three conditions. In these shared gene sets, 72.7% of BKO-NI shared genes and 98.3% of genes shared across all three conditions changed in the same direction. These shared genes included core regulators of mitotic progression and chromosome segregation, including Mybl1, Ska3, Fbxo5, Ska1, Smc5, Rad54b, Cep192, Espl1, Kntc1, Sgo1, Top2a, Cdc23, Fancd2, Aurka, Ttk, and Spag5, all downregulated in both BKO and NI astrocytes (Fig. 5E). This pattern identifies suppression of the mitotic program as the dominant shared transcriptional feature of Bmal1 loss and neuroinflammatory astrocyte activation.

The immune module was weaker in BKO astrocytes. For Groups 1 and 6, BKO astrocytes contributed only 52 leading-edge genes compared with 324 in NI astrocytes and 568 in Cyto astrocytes. Only 17 genes were shared between BKO astrocytes and NI astrocytes and 16 across all three conditions. Directional concordance was also lower than in Group 7. BKO astrocytes showed induction of only a restricted subset of interferon-stimulated genes, including Gbp3, Gbp2, Bst2, Stat1, Gbp6, Ifit1, Ifnar1, Oas1a, Oas1b, Oas1g, and Xaf1, and the magnitude of induction was lower than in NI astrocytes or Cyto astrocytes (Fig. 5E). Thus, Bmal1 loss alone engages only part of the innate immune program and does not reproduce the broader inflammatory activation induced by NI or cytokine treatment. Together, these leading-edge analyses show that the strongest overlap between Bmal1 loss and neuroinflammatory astrocyte activation is in suppression of cell-cycle and chromosome-segregation pathways, whereas the broader immune program requires additional inflammatory signals.

Together, these data support a model in which neuroinflammation weakens BMAL1-dependent clock regulation in astrocytes, while BMAL1 loss alone is sufficient to induce a selective reactive-like astrocyte state. This state does not reproduce the broad inflammatory activation observed in astrocytes under neuroinflammatory conditions or cytokine-treated astrocytes. Instead, it is marked by reactive marker induction, strong suppression of cell-cycle and chromosome-segregation pathways, and limited innate immune activation. Thus, BMAL1 regulates a selective pathway axis within reactive astrocytes, rather than acting as a single driver of the full inflammatory response.

Discussion

This study shows that neuroinflammation suppresses astrocyte clock gene expression through paracrine inflammatory signaling. Although BMAL1 protein was modestly reduced in astrocytes, average BMAL1 occupancy at multiple clock target loci was more strongly reduced. BMAL1 deletion alone did not reproduce the full inflammatory astrocyte response. Instead, it separated the selective BMAL1-linked pathways in reactive astrocytes. Together, these findings identify astrocytic BMAL1 as a regulator of selected reactive astrocyte pathways under neuroinflammatory conditions.

Neuroinflammation weakens astrocyte clock gene rhythms while reactive genes gain rhythmic expression

Our in vivo and synchronized in vitro data show that neuroinflammation weakens astrocyte clock gene rhythms. The clock genes Bmal1 and Per1 showed reduced amplitudes and altered expression patterns across circadian time. Consistent with reduced clock regulation, average BMAL1 occupancy, H3K27ac, and RNA Pol II were reduced at the Per1 and Nr1d1 promoters under neuroinflammatory conditions. These data support weaker BMAL1-dependent transcriptional regulation at clock target loci. At the same time, reactive markers gained rhythmic expression under inflammatory conditions. Thus, neuroinflammation weakens BMAL1-dependent core clock regulation while allowing rhythmic expression of reactive genes.

The in vivo systemic inflammation (SI) model has a limitation. Intraperitoneal LPS can alter activity, feeding, body temperature, and stress hormone signaling. These systemic physiological effects can affect circadian clocks. Therefore, the in vivo astrocyte clock phenotype likely reflects both systemic inflammation-associated brain inflammatory responses and systemic physiological effects. We used this model because, in many disease settings, systemic inflammation can induce brain inflammatory responses, including astrocyte reactivity. The in vitro neuroinflammation model removes these systemic physiological effects and allows us to test inflammatory signaling within the glial environment more directly. Together, these two models allow us to distinguish systemic physiological effects from local glial inflammatory signaling and identify glial paracrine mediators as a direct mechanism of astrocyte clock suppression.

IL-1β and TNFα are sufficient candidate mediators of astrocyte clock suppression

Our data indicate that astrocyte clock suppression under neuroinflammatory conditions is driven by inflammatory signals from the glial environment rather than by direct LPS exposure. Among the induced microglial factors, IL-1β and TNFα were sufficient to reproduce the magnitude of clock suppression observed in neuroinflammatory (NI) astrocytes. However, they did not reproduce the full downstream astrocyte response. This indicates that clock suppression can be induced by this cytokine pair, whereas broader reactive pathway changes require additional inflammatory signals. Thus, IL-1β and TNFα represent a sufficient cytokine axis for astrocyte clock suppression, but not the full glial inflammatory response. Consistent with this interpretation, broader immune activation pathways were enriched in NI microglia, suggesting that additional microglial inflammatory signals may also contribute to astrocyte clock suppression and other aspects of astrocyte reactivity.

BMAL1 occupancy is reduced more strongly than BMAL1 protein abundance

BMAL1 protein was modestly reduced under neuroinflammatory conditions, whereas average BMAL1 occupancy at multiple clock target loci showed a larger decrease. Thus, neuroinflammation not only lowers BMAL1 abundance but also reduces BMAL1 occupancy at clock target loci. The parallel loss of H3K27ac and RNA Pol II at the Per1 and Nr1d1 promoters supports reduced transcriptional activity at these target sites. Together, these data suggest that neuroinflammation weakens BMAL1-dependent clock regulation mainly at the level of chromatin occupancy.

One possible explanation for the modest BMAL1 protein reduction but stronger loss of BMAL1 occupancy is that inflammatory signaling interferes with BMAL1/CLOCK transcriptional complexes. NF-κB can interact with BMAL1 and interfere with BMAL1/CLOCK-dependent transcription at E-box sites, in part by affecting CBP/p300 coactivator availability. Neuroinflammation may also reduce chromatin accessibility or histone acetylation at BMAL1 target loci, making E-box-containing regions less permissive for BMAL1 binding. In addition, BMAL1 post-translational modifications, such as phosphorylation, acetylation, SUMOylation, or ubiquitylation, could alter BMAL1 stability, DNA binding, or interaction with transcriptional cofactors [27–30]. These mechanisms remain to be directly tested in reactive astrocytes under neuroinflammatory conditions.

The ChIP-qPCR experiment was performed in non-synchronized, astrocyte-enriched glial cultures. We used this design to compare average BMAL1 occupancy under control and neuroinflammatory conditions without selecting one or a few circadian time points. Therefore, the data represent average occupancy across mixed-phase cultures, not peak or phase-specific occupancy. A separate limitation is that astrocytes were not purified from the cultures before ChIP-qPCR. Therefore, the occupancy data are not astrocyte-specific. However, the cultures were strongly astrocyte-enriched, and reduced BMAL1 occupancy was accompanied by reduced H3K27ac and RNA Pol II enrichment at the Per1 and Nr1d1 promoter E-box regions. Thus, these data support reduced average BMAL1-dependent regulation in this astrocyte-dominant glial system.

BMAL1 deletion separates a selective reactive-like astrocyte state from full inflammatory activation

Reactive astrocytes are diverse, and their transcriptional and functional outputs vary with the inducing stimulus and disease state [8]. Microglia-derived IL-1α, TNFα, and C1q can drive a neuroinflammatory reactive state [6, 7], whereas other injuries and disease settings induce overlapping but distinct astrocyte responses [12, 51]. Multi-omic studies further show that reactive astrocytes can carry both damaging and protective signatures at the same time [13, 14]. Our data place BMAL1 in this framework as a regulator of a selective reactive-like state rather than the full inflammatory state. BKO astrocytes upregulated Gfap and Serpina3n, but their strongest transcriptomic overlap with neuroinflammatory astrocytes was suppression of cell-cycle and chromosome-segregation pathways rather than broad induction of inflammatory pathways. This indicates that BMAL1 does not contribute equally to all components of the reactive astrocyte response, but contributes more strongly to specific pathway components.

Although BKO astrocytes overlapped only partially with NI and Cyto astrocytes, this partial overlap is informative. BKO astrocytes reflect BMAL1-dependent changes in the absence of exogenous inflammatory stimulation, whereas NI and Cyto astrocytes reflect extrinsic inflammatory stimulation. Their comparison therefore separates a BMAL1-dependent reactive component from the broader inflammatory response induced by neuroinflammation or cytokine treatment. Previous work showed that astrocyte-specific Bmal1 deletion induces astrocyte activation and inflammatory gene expression [32]. Our data extend that finding by separating the reactive response into components. The dominant shared output between BKO astrocytes and NI astrocytes was suppression of cell-cycle and chromosome-segregation genes, including Aurka, Top2a, Ttk, and Ska3, whereas innate immune activation in BKO astrocytes was limited and incomplete. Thus, reactive marker induction, cell-cycle suppression, and full inflammatory activation are separable outputs. BMAL1 loss is sufficient to induce the first two reactive-like features, whereas full inflammatory activation requires additional extrinsic signals and mechanisms.

The BKO model also has a limitation. Because this model is constitutive and global, developmental effects cannot be excluded. Therefore, BKO astrocytes should not be considered a direct model of acute NI-induced BMAL1 reduction. Instead, we used BKO astrocytes to identify BMAL1-dependent pathways from gene-expression changes and then compared these changes with those in astrocytes under neuroinflammatory conditions. A transient astrocyte-specific Bmal1 knockdown or inducible deletion model will be needed to define the direct effects of acute BMAL1 reduction under neuroinflammatory conditions.

Together, these data support a model in which neuroinflammation weakens BMAL1-dependent clock regulation in astrocytes, while BMAL1 loss alone is sufficient to induce a restricted reactive-like astrocyte state. This state does not reproduce the broad inflammatory activation observed in astrocytes under neuroinflammatory conditions or after cytokine treatment. Thus, astrocytic BMAL1 helps explain why reactive astrocyte responses can diverge into distinct pathway outputs under inflammatory conditions.

Authors’ contributions

X.M. performed the experimental and bulk RNA-seq analyses and prepared Figs. 1, 2, 3 and 5. M.H.C. conducted experiments and contributed to all figures. X.Z. supported data analysis and organization, performed ChIP-qPCR experiments, and prepared Figure 4. J.Y.K. conceptualized the study, designed the experiments, supervised all experimental work and data analysis, and wrote the manuscript. All authors reviewed the manuscript.

Funding

We acknowledge funding from the following. Research Grants Council of the Hong Kong Special Administrative Region (CityU 11103418 to J.Y.K.). Tung Biomedical Sciences Centre, City University of Hong Kong (9609323 to J.Y.K.). City University of Hong Kong (7006041 to J.Y.K.).

Data availability

The RNA-seq data generated in this study have been deposited in the NCBI Gene Expression Omnibus (GEO) under accession number GSE325658, which will be publicly released upon publication. All bioinformatics workflows and source codes, including implementations for UpSetR (v1.4.0) and ggplot2 (v4.0.2), are available at: https://zenodo.org/records/19872779. All other data are available in the main text or supplementary materials.

Declarations

Competing interests

The authors declare no competing interests.