Daily rhythms in brain-infecting viruses' entry and disrupted body clocks in hosts

The nervous system is crucial for both neurotropic viral infection and the synchronization of the central clock with peripheral circadian clocks^1,2,39. Although the circadian clock is known to regulate certain viral infections^46–51, whether this regulatory role extends to all neurotropic viruses remains unclear. Given that viral receptors are critical for viral infection^12,14,15, the promoters of well-characterized receptors for neurotropic viruses were examined (Supplementary Table S1). E-BOX or RORE motifs, which are key binding motifs for circadian transcription factors³³, were present in the promoters of all the receptors (Supplementary Fig. S1a). BMAL1 is a core transcription factor that partners with the CLOCK protein to activate the expression of genes, and its genome-wide occupancy rhythms peak at CT6–CT8^33,34. To investigate the direct regulation of neurotropic viral receptors by the circadian clock, cerebral organoids (Supplementary Fig. S1b) were synchronized with hydrocortisone⁵². The results demonstrated that treatment with 100 nM hydrocortisone effectively reset the circadian oscillations of the organoids, with BMAL1 mRNA levels peaking around CT7 within the CT1–CT12 interval (Supplementary Fig. S1c). Accordingly, cerebral organoids were sampled at CT7 for a CUT&Tag assay using an anti-BMAL1 antibody, which revealed a noticeable enrichment in the transcription start site (TSS) region (Fig. 1a, b), with ~63% of BMAL1 peaks located in the promoter regions (Fig. 1c). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of genes associated with these promoter-associated peaks revealed significant enrichment for “Circadian Rhythm,” confirming the reliability of our CUT&Tag-seq data. Furthermore, pathways related to viral infection, including “Herpes Simplex Virus 1 (HSV-1) infection” and “Viral carcinogenesis”, were also significantly enriched (Fig. 1d). Gene Ontology (GO) analysis of peak-associated genes demonstrated enrichment for terms associated with RNA processing, nuclear components, and ligase activity (Fig. 1e). To specifically determine whether BMAL1 binds to the promoters of neurotropic virus receptors, we performed a detailed analysis of the identified peaks. First, we analyzed classical BMAL1 target genes (e.g., REV-ERBα and PER1) and observed pronounced BMAL1 binding peaks at these loci, further confirming that CT7 represents an appropriate circadian time point for investigating BMAL1–promoter interactions. We subsequently detected significant BMAL1 binding signals in the promoter regions of 23 receptors, including PDGFRA, ITGB1, and AXL. These receptors correspond to 32 different viruses with neurotropic or neuroinvasive potential, such as HCMV, Ebola virus (EBOV), and Chikungunya virus (Fig. 1f), indicating a potential regulatory function of the circadian clock on neurotropic viral receptors.

Additionally, we analyzed RNA-seq data from mouse dorsal hippocampus (GSE222744↗⁵³) and salivary gland tissues (GSE234789↗ (not published)), as well as Calu-3 and U2OS cells (GSE176393↗⁵⁴ and GSE248721↗⁵⁵), from the NCBI Gene Expression Omnibus (GEO) database. The results revealed that the expression of most neurotropic virus receptors exhibited significant oscillations at different Zeitgeber times (Supplementary Fig. S1d). Furthermore, the expression of these receptors was either upregulated or downregulated following pharmacological activation of REV-ERBα with SR9009 or upon Rev-erbα knockout (KO) (Supplementary Fig. S1e), with the expression of the RABV receptor p75NTR showing the most substantial changes. To confirm the pervasive nature of this circadian regulation, we utilized three distinct experimental models: the human neuroblastoma cell line SH-SY5Y, human cerebral organoids, and C57BL/6 mice. Synchronization was achieved through a 100 nM dexamethasone pulse^56,57, a 100 nM hydrocortisone pulse⁵², and a 12-h light/dark cycle⁵⁸, respectively. In all three models, the mRNA levels of the investigated neurotropic virus receptors exhibited significant rhythmic expression (Fig. 1g–i). Using a Euclidean distance algorithm, we clustered the expression rhythms of these receptors into two distinct groups. One group, which included the HCMV receptors ITGB1 and PDGFRA, oscillated in phase with the expression profile of BMAL1. The other group, which included the RABV receptor p75NTR, oscillated in phase with REV-ERBα (Supplementary Fig. S1f–h). We subsequently quantified the mRNA levels of these receptors in SH-SY5Y cells treated with DMSO or the REV-ERBα agonist SR9009, as well as in brain tissues from wild-type (WT) or Rev-erbα-KO mice. Under our experimental conditions, no significant adverse effects of SR9009 on SH-SY5Y cell viability were detected (Supplementary Fig. S1i). We found that pharmacological activation of REV-ERBα with SR9009 in SH-SY5Y cells or the brains of Rev-erbα-KO mice significantly altered the expression of most receptors. The most pronounced effect was again observed for the RABV receptor p75NTR (Fig. 1j, k). Taken together, these results demonstrate that the expression of neurotropic viral receptors is robustly regulated by the circadian clock in a conserved manner across different models and species.

To investigate whether circadian clock regulation affects host susceptibility to viruses with neurotropic or neuroinvasive potential, we selected a panel of representative viruses for viral entry assays, including neurotropic viruses (RABV, vesicular stomatitis virus (VSV), Zika virus (ZIKV), Coxsackievirus B3 (CVB3), and HSV-1) and viruses with neuroinvasive potential (EBOV, dengue virus (DENV), and HCMV). As hypothesized, host cell susceptibility to these viruses exhibited a rhythmic pattern that mirrored the oscillatory expression of their respective receptors (Fig. 2a). We subsequently treated SH-SY5Y and U-251 cells with SR9009 to investigate the effect of REV-ERBα activation on viral entry into host cells; under our experimental conditions, SR9009 did not significantly decrease U-251 cell viability (Supplementary Fig. S2a). SR9009 treatment significantly increased RABV entry but inhibited the entry of the other viruses (Fig. 2b); these findings for ZIKV, DENV, and HSV-1 were consistent with those of previous reports^46,49. For a more in-depth mechanistic study, we focused on the RNA virus RABV and the DNA virus HCMV, as their respective receptors exhibit expression patterns that align with the rhythms of the core clock components REV-ERBα and BMAL1. We used shRNA to knock down these clock genes in SH-SY5Y cells and the human glioblastoma multiforme cell line U-251 (Supplementary Fig. S2b). We subsequently infected the knockdown (KD) SH-SY5Y cells with the RABV CVS-11 strain and U-251 cells with the HCMV AD169 strain. Notably, REV-ERBα KD had the most potent effect, strongly inhibiting RABV infection while increasing HCMV infection (Fig. 2c). Within the circadian TTFL, RORA typically functions as a transcriptional activator, whereas REV-ERBα serves as a transcriptional repressor. Both proteins recognize and bind to identical or overlapping DNA motifs^33,34. We observed that they exerted critical yet opposing effects on the outcome of HCMV infection in U-251 cells (Fig. 2c). In contrast, during the RABV infection of SH-SY5Y cells, only REV-ERBα knockdown resulted in a pronounced inhibitory effect (Fig. 2c). Subsequent experiments demonstrated that REV-ERBα overexpression or pharmacological activation with SR9009 markedly increased RABV infection in SH-SY5Y cells, whereas RORA overexpression or stimulation with SR1078 elicited only a modest inhibitory effect (Supplementary Fig. S2c).

For transcriptional activators within the TTFL, rhythmic binding to promoter regions of target genes enables the transmission of their oscillatory patterns to downstream gene expression. Interestingly, as shown in Fig. 1g–i, the rhythmic oscillation patterns of certain neurotropic viral receptors consistently aligned with those of REV-ERBα across different species. This raised the intriguing question of how REV-ERBα, which functions as a transcriptional repressor within the TTFL, could propagate its oscillatory rhythm to these neurotropic viral receptors. Consequently, our study focused primarily on elucidating the molecular mechanisms through which REV-ERBα governs the circadian regulation of neurotropic viral receptor rhythmic expression.

Furthermore, we observed that synchronizing U-251 cells with dexamethasone also induced rhythmic gene expression (Supplementary Fig. S2d). When synchronized SH-SY5Y and U-251 cells were infected with RABV or HCMV at different circadian times, we observed that the overall infection outcome was rhythmic (Fig. 2d). Deletion of BMAL1 is known to abolish cellular circadian rhythms^59,60. Indeed, BMAL1 KO in SH-SY5Y and U-251 cells disrupted the rhythmic expression of REV-ERBα (Supplementary Fig. S2e) and, consequently, abolished the rhythmic susceptibility to RABV and HCMV, respectively (Fig. 2e).

Furthermore, SR9009 promoted RABV infection in SH-SY5Y cells and inhibited HCMV infection in U-251 cells in a dose- and time-dependent manner (Fig. 2f). We also noted that SR9009 treatment increased the number of RABV particles in infected SH-SY5Y cells but reduced the number of HCMV particles in infected U-251 cells (Fig. 2g). Conversely, RABV infection was markedly reduced in REV-ERBα-KO SH-SY5Y cells, whereas HCMV infection was significantly increased in REV-ERBα-KO U-251 cells (Fig. 2h). To determine whether the effect of REV-ERBα on viral infection depends on its transcriptional regulatory function, we engineered a mutant REV-ERBα lacking its DNA-binding domain (REV-ERBα-dDBD). Overexpression of full-length REV-ERBα (REV-ERBα-FL) in SH-SY5Y cells increased RABV infection, but this effect was absent in cells expressing the REV-ERBα-dDBD mutant. In contrast, overexpression of REV-ERBα-FL in U-251 cells suppressed HCMV infection, an effect that was also abrogated in cells expressing a mutant lacking the DNA-binding domain (Fig. 2i). Taken together, these data demonstrate that host cell susceptibility to neurotropic viruses is regulated by the circadian clock and that clock components play key divergent roles in RNA and DNA virus infection.

RABV relies predominantly on the nervous system for its transmission and replication, in contrast to HCMV, which is transmitted primarily via bodily fluids and disseminates to other organs following lytic infection in epithelial cells^61,62. Furthermore, HCMV is human specific and does not naturally infect mice⁶², making RABV an ideal model for investigating the influence of the circadian clock on viral receptor expression and infection mechanisms both in vitro and in vivo. First, analysis of the GEO database revealed that p75NTR is the sole RABV receptor under circadian clock regulation (Fig. 3a; Supplementary Fig. S3a, b). Additionally, in REV-ERBα-KO SH-SY5Y cells, p75NTR mRNA levels were markedly reduced, accompanied by significant suppression of RABV entry (Fig. 3b, d); conversely, treatment with SR9009 or overexpression of REV-ERBα-FL significantly increased p75NTR mRNA levels and RABV entry, whereas the REV-ERBα-dDBD mutant failed to produce these effects (Fig. 3c, e; Supplementary Fig. S3c, d). The knockdown of p75NTR subsequently attenuated the enhancing effect of SR9009 on RABV infection and reduced the variability in cellular susceptibility to RABV across circadian cycles (Fig. 3f, g).

Next, we used a dual-luciferase reporter system to determine whether REV-ERBα influences the promoter activity of p75NTR. A significant increase in p75NTR promoter activity was observed with REV-ERBα-FL, whereas the REV-ERBα-dDBD mutant had no effect (Fig. 3h). Moreover, when both WT and REV-ERBα-KO HEK293 cells were transfected with the p75NTR promoter-luciferase reporter and then treated with DMSO or SR9009, the knockout of REV-ERBα abolished the potentiating effect of SR9009 on promoter activity (Fig. 3i). Collectively, these results indicate that REV-ERBα increases p75NTR promoter activity, thereby upregulating p75NTR expression and facilitating RABV infection.

Given that REV-ERBα is a transcriptional repressor that regulates gene expression by suppressing promoter activity⁶³, we hypothesized that its effect on the p75NTR promoter was not direct but was mediated by another transcription factor. PROMO^64,65 was used to predict transcription factors that bind to the p75NTR promoter. Given the nonrhythmic expression patterns of GRM2 and NCAM1 (Supplementary Fig. S3e), the predicted transcription factors linked to the promoters of GRM2 and NCAM1 were used as a filtering criterion. Ultimately, members of the E2F family, as well as AhR and SRF, were identified as potential regulators of p75NTR (Fig. 3j). The mRNA levels of these transcription factors were subsequently examined in SH-SY5Y cells treated with SR9009 or in cells with REV-ERBα KO. Notably, the expression of E2F5, E2F7, and E2F8 was significantly downregulated following SR9009 treatment and correspondingly upregulated in REV-ERBα-KO SH-SY5Y cells (Fig. 3k). Furthermore, analysis of the GEO dataset GSE176393↗⁵⁴ indicated that treatment of Calu-3 cells with SR9009 resulted in a pronounced decrease in E2F8 mRNA levels (Supplementary Fig. S3f).

To elucidate the regulatory effect of E2F5, E2F7, and E2F8 on p75NTR, these transcription factors were overexpressed in SH-SY5Y cells, with only E2F8 overexpression leading to a significant decrease in the p75NTR mRNA level (Fig. 3l). Conversely, knockdown of E2F8 in SH-SY5Y cells resulted in a significant increase in p75NTR mRNA levels (Supplementary Fig. S3g). In addition, RNA-seq data from the GSE71574↗⁶⁶ dataset revealed that E2F8 KO led to an increase in the mRNA level of p75NTR (Supplementary Fig. S3h). These findings support the hypothesis that E2F8 functions as a repressor of p75NTR transcription and that REV-ERBα increases p75NTR promoter activity by suppressing E2F8. To further validate this model, SH-SY5Y cell lines overexpressing E2F5, E2F7, and E2F8 were infected with CVS-11. Only E2F8 overexpression significantly inhibited RABV infection (Fig. 3m), whereas E2F8 knockdown increased RABV infection (Supplementary Fig. S3i). Additionally, when p75NTR expression was reduced in an SH-SY5Y cell line stably overexpressing E2F8, the inhibitory effect of E2F8 on RABV infection was abolished, indicating that E2F8 suppresses RABV infection primarily through modulation of p75NTR (Fig. 3n). Furthermore, HEK293 cells were cotransfected with a p75NTR promoter-luciferase reporter and REV-ERBα or E2F8. The results revealed that overexpression of E2F8 negated the REV-ERBα–mediated increase in promoter activity (Fig. 3o). In contrast, SR9009 treatment failed to promote RABV infection in an SH-SY5Y cell line in which E2F8 was knocked down (Fig. 3p).

To explore whether the DNA binding domains of E2F8 are essential for the regulation of p75NTR expression and RABV infection, we constructed E2F8 deletion mutants lacking DNA binding domain 1 (E2F8-D1) or DNA binding domain 2 (E2F8-D2). Our results demonstrated that both domains were critical for E2F8-mediated suppression of p75NTR expression and RABV infection (Fig. 3q), suggesting that E2F8 functions by directly binding to the p75NTR promoter. In support of this notion, analysis of ChIP-seq data from the GSE275631↗⁶⁷ dataset in the GEO database revealed that in the p75NTR promoter (Fig. 3r), E2F8 peaks were located exclusively within 0–1 kb of the TSS (100%, purple). This complete proximal enrichment, with no peaks in distal regions, indicated that E2F8 bound directly to the core promoter elements of p75NTR, likely exerting strong and direct transcriptional control. Furthermore, CUT&Tag-qPCR experiments confirmed that E2F8 directly bound to the p75NTR promoter, thereby inhibiting its expression (Fig. 3s). These results demonstrate that E2F8 suppresses the promoter activity of p75NTR by binding to its promoter, whereas REV-ERBα increases the mRNA level of p75NTR by inhibiting E2F8 expression.

The E2F8 protein functions as a negative regulator of the cell cycle, impeding G1 phase progression and stress responses through the modulation of gene transcription⁶⁸. KEGG analysis of the peaks from the BMAL1 CUT&Tag-seq data revealed that the “cell cycle” pathway was enriched (Fig. 1d), prompting us to hypothesize that E2F8 is a clock-controlled gene (CCG). To investigate its relationship with the circadian clock, we analyzed an RNA-seq dataset from GSE234789↗, which revealed significant oscillations in E2f8 mRNA levels across various Zeitgeber times in mouse submandibular gland tissues (Supplementary Fig. S4a). To confirm that E2F8 is a CCG, we evaluated its mRNA and protein levels via qRT‒PCR and western blotting assay at various circadian time points in synchronized SH-SY5Y cells or at different Zeitgeber times in brain tissues from mice entrained to a 12:12 light/dark cycle (Fig. 4a, b; Supplementary Fig. S4b, c). Ablation of BMAL1 or CRY1/2 results in arrhythmicity in cells or animals^60,69,70. We examined the rhythmic pattern of E2F8 expression in BMAL1-KO SH-SY5Y cells and Cry1/2-DKO MEFs and found that when the cells were rendered arrhythmic, the oscillatory expression of E2F8 was also abolished (Supplementary Fig. S4d, e). These findings demonstrated the dependency of E2F8 rhythmicity on the core circadian clock machinery. However, BMAL1 CUT&Tag-seq data revealed no binding peak in the E2F8 promoter region (data not shown). Our previous experiments demonstrated that the clock factor REV-ERBα suppresses E2F8 expression, leading us to hypothesize that REV-ERBα drives its rhythmic expression. Indeed, the reduction in E2F8 protein levels induced by REV-ERBα depended on its DNA-binding domain (Fig. 4c). To determine whether this regulation is direct, we analyzed the public ChIP-seq dataset GSM8123058↗ and performed CUT&Tag-qPCR experiments, which confirmed that REV-ERBα bound to the promoter region of E2F8 (Fig. 4d; Supplementary Fig. S4f).

Given that the cell cycle is in a state of dynamic equilibrium, we hypothesize the presence of a negative feedback regulatory mechanism between E2F8 and REV-ERBα, which may indirectly influence the circadian clock system. Surprisingly, SH-SY5Y cells with overexpressed or knockeddown E2F8 showed significant alterations in the mRNA levels of several circadian clock control factors, including BMAL1, REV-ERBα, CLOCK, RORB, PER2, CRY1 and DBP (Fig. 4e, f). Notably, only the expression of REV-ERBα and PER2 was downregulated following E2F8 overexpression. To elucidate this mechanism, we overexpressed E2F8-FL and mutants lacking DNA-binding domains in SH-SY5Y cells and found that the downregulation of REV-ERBα and PER2 expression required intact E2F8 DNA-binding domains, suggesting direct transcriptional repression (Fig. 4g). To explore whether E2F8 directly binds to the promoter regions of REV-ERBα and PER2, we mined publicly available ChIP-seq datasets for E2F8 (GSE275631↗) and analyzed the genomic distribution of E2F8 peaks relative to the TSS of the REV-ERBα and PER2 promoters using ChIPseeker and custom R scripts. We found that the REV-ERBα promoter had a broader distribution: the largest proportion (~55%, green) resided 5–10 kb from the TSS (distal, likely enhancer regions), ~30% (cyan) was within 0–1 kb (proximal promoter), and ~15% (purple) was within 1–3 kb. In contrast, the PER2 promoter was more proximally biased: ~45% (purple) was within 3–5 kb, ~30% (cyan) was within 1–3 kb, and ~15% (green) was within 0–1 kb, with minimal peaks farther away (Fig. 4h). Collectively, these data demonstrated the direct occupancy of E2F8 at the promoter regions of REV-ERBα and PER2, with a proximal predominance for PER2 and the inclusion of distal enhancers for REV-ERBα. Finally, using CUT&Tag-qPCR, we demonstrated that E2F8 bound directly to the promoter regions of REV-ERBα and PER2 (Fig. 4i). These results suggest that E2F8 is a potential regulator of the circadian clock.

We further hypothesized that E2F8 influences the host circadian clock, thereby affecting the expression of neurotropic virus receptors. RNA-seq analysis of the GEO dataset GSE52157↗⁷¹ revealed that E2F8 deficiency in the mouse spleen significantly altered the expression of receptors for various neurotropic viruses (Supplementary Fig. S4g). The increase in these receptors due to E2F8 overexpression was linked to its effect on the circadian clock, as this increase was abolished following BMAL1 knockout (Fig. 4j, k). Conversely, the decrease in the level of receptor genes due to E2F8 overexpression resulted from its direct binding to their promoters, which persisted despite BMAL1 knockout (Fig. 4j, k). Finally, we demonstrated that E2F8-mediated regulation of receptor expression functionally affected neurotropic virus entry (Fig. 4l). In conclusion, our findings reveal an unconventional role for the cell cycle regulator E2F8 as a key component in a reciprocal feedback loop with the core circadian clock, thereby influencing host susceptibility to neurotropic viruses through both direct and indirect transcriptional mechanisms.

The mutual interaction between viruses and hosts is a dynamic process⁷². Previous research has shown that HSV-1 infection modulates BMAL1 promoter activity⁴⁶, suggesting that viral infections may disrupt host circadian rhythms. However, behavioral evidence from animal models and the underlying molecular mechanism remain scarce. To investigate in detail the impact of neurotropic virus infection on the host circadian clock, we chose RABV as our model because its infection can affect the host’s nervous system⁷³. We examined the rhythmic patterns of integrated behavioral and metabolic functions in mock- and RABV-infected mice. Our analysis revealed that food intake did not differ significantly between RABV-infected and mock-infected mice prior to the onset of encephalitis and paralysis in the RABV group (Supplementary Fig. S5). However, the RABV-infected mice displayed arrhythmic patterns in wheel running activity (Fig. 5a), oxygen consumption (Fig. 5b), and the respiratory exchange ratio (Fig. 5c), indicating that RABV disrupts the host’s circadian clock.

Previous research has shown that BMAL1 is the only clock gene whose disruption of homeostasis results in circadian clock dysfunction⁷⁴. To elucidate the molecular mechanisms underlying RABV-induced circadian rhythm disruption, we assessed changes in BMAL1 protein levels across multiple organs — the brain, spleen, kidney, lung, and liver — at various stages of viral infection. Our findings demonstrated that Bmal1 protein levels in the brain, spleen, and lung significantly decreased from the early to late stages of RABV infection. In contrast, BMAL1 levels in the kidney remained unchanged during the early stages, whereas those in the liver were unaffected throughout the infection (Fig. 5d–h). Notably, among these organs, RABV was detected exclusively in the brain, leading us to hypothesize that RABV disrupts the central clock, consequently impairing peripheral clocks.

Further investigation revealed a significant reduction in BMAL1 protein levels in SH-SY5Y and mouse neuroblastoma N2A cells cultured in vitro following RABV infection (Fig. 5i). Given that viral infections typically activate the host’s innate immune response, we explored whether this response could account for BMAL1 downregulation. Treatment of SH-SY5Y cells with varying concentrations of interferon-beta, interferon-alpha, and interferon-gamma did not alter BMAL1 levels (Fig. 5j), suggesting that viral components may directly modulate the host’s circadian clock.

To identify the specific viral factors involved, we overexpressed the G, M, P, and N proteins of RABV CVS-11 in SH-SY5Y cells. Owing to the large molecular weight of the L protein, we were unable to construct an overexpression plasmid. Our results indicated that only the RABV G protein significantly reduced BMAL1 protein expression (Fig. 5k). Moreover, G proteins from various RABV strains consistently decreased BMAL1 protein levels (Fig. 5l). The primary mechanisms leading to reduced BMAL1 protein levels were K48-linked polyubiquitination, which targets BMAL1 for proteasomal degradation^75,76, and transcriptional repression of the Bmal1 gene by REV-ERBα³⁴. We found that RABV infection did not alter the levels of total ubiquitinated BMAL1 or K48-linked ubiquitinated BMAL1 (Fig. 5m). However, we discovered that RABV infection suppressed BMAL1 promoter activity and that this repression was dependent on the presence of REV-ERBα binding sites. Specifically, a BMAL1 luciferase reporter mutant lacking these binding sites showed no response to RABV infection (Fig. 5n). These findings indicated that RABV infection disrupted the central circadian clock, leading to systemic perturbation of the host circadian system. The observed dysregulation of BMAL1 expression may be indirectly mediated by the effect of the viral G protein on REV-ERBα.

To investigate the relationship between RV-G and REV-ERBα, we utilized the protein interaction network provided in the REV-ERBα record from the NCBI Gene database, which catalogs known interaction partners on the basis of curated experimental data. Previously, we mapped the protein‒protein interaction (PPI) profiles of G proteins from the SAD B19 and Nishigahara strains with host proteins in SH-SY5Y cells using affinity tag purification-mass spectrometry (AP-MS)⁷⁷. However, no overlap was observed between these G protein PPI profiles and those of REV-ERBα (data not shown). After analyzing the proteins that interacted with REV-ERBα, we speculated that the low basal expression levels of certain proteins in neuronal cells might hinder the capture of the existing interactions. Given this, we conducted AP-MS analysis on HEK293 cell lysates following co-immunoprecipitation (co-IP) with 3× Flag-tagged G proteins from the RABV strains Nishigahara and SAD B19. These analyses revealed 42 high-confidence interactions for Nishigahara G and 46 for SAD B19 G. By integrating these data with the 21 known REV-ERBα-interacting host proteins from the NCBI database, we constructed an interactome map (Fig. 6b; Supplementary Tables S2–S4). Notably, the E3 ubiquitin ligase HUWE1 emerged as the sole protein common to all three interactomes (Fig. 6a, b). Subsequent reciprocal assays confirmed the interaction between RV-G and HUWE1 (Fig. 6c). This interaction was conserved across G proteins from different RABV strains (Fig. 6d), suggesting its consistency across viral variants. The RV-G protein comprises three functional domains: the fusion domain (FD), central domain (CD), and pleckstrin homology domain (PHD)⁷⁸. To determine which domain mediates the interaction with HUWE1, we generated RV-G mutants lacking each domain individually. Deletion of the CD domain abolished the RV-G–HUWE1 interaction (Fig. 6e).

REV-ERBα is known to undergo lithium-stimulated degradation, a process mediated by the HUWE1/PAM complex through K48 ubiquitination at serines 55 and 59⁷⁹. Consistent with prior reports, we demonstrated that the proteasome inhibitor MG132 extended the half-life of REV-ERBα. Similarly, overexpression of RV-G prolonged the half-life of REV-ERBα (Fig. 6f). To further explore this mechanism, we engineered a REV-ERBα mutant mimicking phosphorylation at serines 55 and 59 (REV-ERBα-55/59 SD). Treatment with 20 mM LiCl reduced the protein levels of WT REV-ERBα but not those of the REV-ERBα-55/59 SD mutant. Moreover, RV-G inhibited the lithium-induced degradation of REV-ERBα (Fig. 6g). However, RV-G mutants incapable of interacting with HUWE1 failed to prevent LiCl-induced REV-ERBα degradation (Fig. 6h).

On the basis of these findings, we hypothesized that RV-G competes with REV-ERBα for binding to HUWE1. Co-IP experiments revealed that LiCl treatment increased the interaction between REV-ERBα and HUWE1, an effect that was suppressed by RV-G (Fig. 6i). Furthermore, WT RV-G inhibited both total ubiquitination and K48-specific ubiquitination of REV-ERBα, an effect contingent upon its ability to bind HUWE1 (Fig. 6j, k). Finally, we sought to determine whether the mechanism by which RABV disrupts the circadian clock is universal. We found that the G proteins of VSV and EBOV did not significantly affect the half-life of REV-ERBα or its Li⁺-induced degradation (Supplementary Fig. S6a, b), indicating that virus–host interactions with the circadian clock are complex and diverse. Collectively, our results demonstrate that RV-G competes with REV-ERBα for binding to HUWE1, thereby attenuating K48 ubiquitination at serines 55 and 59 of REV-ERBα. This competition stabilizes REV-ERBα, leading to its accumulation, which in turn suppresses BMAL1 expression and disrupts the host’s circadian rhythm.

Despite sharing the same fundamental molecular clockwork, the circadian system in monolayer-cultured cells differs significantly from that in living animals^80,81. To examine the influence of the circadian clock on neurotropic virus infections in vivo, we selected RABV as a representative neurotropic virus and utilized C57BL/6 mice as the experimental model. A standard 12 h light and 12 h dark cycle (12:12 L/D) was used to establish a “time of day” research model. Additionally, 6-h light-phase advances were performed every 2 days⁸² to induce chronic circadian disruption, whereas CRISPR-Cas9 technology was used to generate a “Rev-erbα knockout” model (Fig. 7a). Mice were infected with the CVS-11 strain of RABV via intranasal administration, and their brains were subsequently dissected into the olfactory bulb, cerebrum, and cerebellum to assess viral infection. Daily fluctuations in RABV receptor and clock factor expression were quantified in brain tissue harvested every 6 h from mice entrained to a 12:12 L/D cycle. The rhythmic expression of the clock genes Bmal1 and Rev-erbα in the olfactory bulb, cerebrum, and cerebellum confirmed the efficacy of our “Time of Day” model (Supplementary Fig. S7a–c). Consistent with our cellular-level findings, only the mRNA expression of p75NTR exhibited a daily rhythm, peaking at ZT12 (Fig. 7b; Supplementary Fig. S7d, e). A defining feature of circadian rhythms is their persistence in the absence of environmental cues. The sustained rhythmic expression of Bmal1 and Rev-erbα mRNA in the three brain regions under constant darkness (CD) confirmed this characteristic (Supplementary Fig. S7f–h), with the p75NTR rhythm similarly persisting across these regions (Supplementary Fig. S7i–k).

In the chronic jet lag (CJL) model, mice maintained under a 12:12 L/D cycle served as controls. By monitoring rhythmic behavior and metabolic activity, we confirmed that the “repeated 6 h light phase advances every two days to induce chronic circadian disruption” regimen effectively disrupted the circadian clock in CJL mice (Supplementary Fig. S8a, b). Moreover, the mRNA levels of the clock genes Bmal1 and Rev-erbα lost their rhythmicity in the three brain regions of CJL mice (Supplementary Fig. S8c–e). We subsequently confirmed that the rhythmicity of p75NTR expression was also abolished and that the mRNA and protein levels of p75NTR in the olfactory bulb, cerebellum, and cerebrum of CJL and 12:12 L/D mice displayed dynamic changes across the Zeitgeber time axis, with significant differences observed at most Zeitgeber time points (Fig. 7c; Supplementary Fig. S8f, g). In Rev-erbα-KO mice, despite the critical role of REV-ERBα in circadian rhythms, no substantial changes were observed in the period or amplitude of rhythmic behavior following Rev-erbα knockout (Supplementary Fig. S9), which aligns with the findings of previous studies⁸³. Consistent with our cellular-level results, the mRNA levels of the RABV receptor p75NTR were significantly lower in Rev-erbα-KO mice than in WT mice (Fig. 7d).

We first conducted infection experiments using the “Time of Day” model. Compared with those infected at ZT0 (the onset of the inactive/light phase), mice infected at ZT12 (the onset of the active/dark phase) exhibited more rapid weight loss (Fig. 7e) and reached the disease endpoint (score of 5) earlier (Supplementary Fig. S10a). Upon reaching the clinical endpoint, the mice were euthanized (Fig. 7e). Subsequent analysis of RABV replication kinetics in the mouse brain revealed that the virus initially replicated in the cerebrum approximately 4–6 days post-infection (d.p.i.) before it spread to the olfactory bulb and cerebellum at later stages (8–10 d.p.i.). Notably, at all the measured time points and across all the brain regions, the viral load and replication rate were significantly greater in mice infected at ZT12 compared with those infected at ZT0 (Supplementary Fig. S10b). This finding suggests that viral amplification is more pronounced at ZT12. As anticipated, by the 10th d.p.i., a greater abundance of RABV M protein was detected at ZT12 (Fig. 7h), accompanied by an increased number of viral particles in the brain, compared with those detected at ZT0 (Fig. 7k). Following infection with the rabies challenge strain, the mice developed pathogenic encephalitis. To evaluate the extent of encephalitis induced by RABV infection, CD45-positive cells were identified via immunohistochemical staining, and the levels of interferon β and γ, proinflammatory cytokines, inflammasome activation, and chemokines were quantified by qPCR. CD45-positive cells were more abundant in the brains of mice infected at ZT12 than in those of mice infected at ZT0 (Fig. 7n). Additionally, the mRNA levels of all the immune-related genes assessed were elevated in ZT12 mice compared with ZT0 mice (Supplementary Fig. S10c), indicating that the inflammatory immune response was more robust in the brains of ZT12 mice. Given that circadian rhythms modulate host immune responses, we examined whether antiviral immunity in the mouse brain differs between ZT0 and ZT12 under basal conditions, potentially explaining the greater susceptibility to RABV at ZT12. Analysis of IFNβ secretion and interferon-stimulated gene (ISG) expression revealed no significant differences between ZT0 and ZT12 in the resting state (Supplementary Fig. S10d, e). Collectively, these results suggest that RABV infection is modulated by the time of day, which correlates with the regulation of the RABV receptor p75NTR by the biological clock.

Three weeks after initiating the CJL light-dark regimen, both 12:12 L/D and CJL mice were inoculated with RABV. Compared with their 12:12 L/D counterparts, CJL mice exhibited weight loss and initial clinical symptoms earlier, with accelerated weight loss, faster disease progression, and higher mortality rates (Fig. 7f; Supplementary Fig. S11a). Consistent with the observed disease progression, CJL mice displayed elevated viral loads in the cerebrum, cerebellum, and olfactory bulb (Fig. 7i, l), along with more intense inflammation and increased mRNA levels of immune-related factors (Fig. 7o; Supplementary Fig. S11b–d). Research has indicated that CJL can impair immune function and exacerbate vulnerability to pathogens^84–86; for instance, CJL alters the mouse lung transcriptome, downregulating the expression of immune regulatory genes while upregulating the expression of transcripts that promote viral replication and immune evasion⁸⁷. Nonetheless, reports elucidating the effect of CJL on antiviral immune responses within the host central nervous system are lacking. To investigate whether CJL affects antiviral immune responses in the mouse brain, we measured IFN-β secretion and ISGs expression at different Zeitgeber times (ZT0, ZT4, ZT8, ZT12, ZT16, ZT20, and ZT24) in CJL and 12:12 L/D mice. We observed that, under basal conditions, compared with those in 12:12 L/D mice, IFN-β secretion and expression of ISGs in the brains of CJL mice were persistently modestly reduced (Supplementary Fig. S11e, f). These data suggest that the increased susceptibility of CJL mice to RABV results from a combined effect of impaired antiviral immune responses in the brain and attenuated oscillatory expression of the RABV receptor p75NTR, which remains elevated at most Zeitgeber times relative to mice under regular light-dark cycles. Given that IFN-β secretion and expression of ISGs in CJL mice were only modestly lower than those in 12:12 L/D mice, we propose that the differences in p75NTR rhythmicity are likely the dominant contributing factor.

Finally, in RABV challenge experiments involving WT and Rev-erbα-KO mice, we found that the suppression of p75NTR expression in Rev-erbα-KO mice impeded RABV infection, prolonged the incubation period, and slowed disease progression (Fig. 7g; Supplementary Fig. S12a). Rev-erbα-KO mice exhibited reduced viral loads in the cerebrum, cerebellum, and olfactory bulb (Fig. 7j, m), less severe inflammation and lower mRNA levels of immune-related factors (Fig. 7p; Supplementary Fig. S12b–d). In summary, our findings demonstrate that the timing of RABV infection critically influences its outcome. Disruption of the circadian clock exacerbates disease progression following RABV infection, whereas loss of the clock factor REV-ERBα results in partial resistance to RABV, thereby extending host survival (Supplementary Fig. S13).

C57BL/6 N WT mice (8–12 weeks old, mixed sexes) were purchased from Wuhan Institute of Biological Products Co., Ltd. Rev-erbα heterozygous mice were procured from Cyagen Inc. (Suzhou, China) for breeding to generate Rev-erbα knockout mice and WT mice. All of the mice were fed rodent chow ad libitum. All the studies were approved by the Institutional Animal Care and Use Committee of Medical Research Institute, Wuhan University.

Induced pluripotent stem cells (iPSCs, purchased from SAIOS Biotech, Inc.) were cultured in a Matrigel-coated T25 flask with Essential 8 flex medium (Gibco, A2858501). On Day 0, the iPSCs were counted and seeded at 1 × 10⁶ cells/well in a six-well plate in Essential 8 Flex Medium supplemented with 10 μM ROCK inhibitor Y-27632 (MCE, HY-10071). On Day 1, embryonic bodies were formed, and the medium was replaced with E6 medium (50 mL of E8 basal + 0.1 μM LDN + 10 μM SB431542 + 2 μM XAV + 20 μg/mL insulin + 1% P/S). The medium was changed every two days until the iPSCs were induced into ectodermal cells (approximately 10 days), after which the medium was replaced with NI medium (50 mL of DMEM/F12 + 50 of neurobasal medium + 1% B-27 + 0.5% N-2 + 1× GlutaMAX + 1% P/S), and the cells were cultured for 60 days. The entire culture process was carried out under gentle shaking on a shaker (80 rpm).

Anti-PER2 (ABclonal, A5107), anti-p75NTR (ABclonal, A19088), anti-E2F8 (ABclonal, A1135), anti-BMAL1 (Cell Signaling Technology, D2L7G), anti-REV-ERBα (Cell Signaling Technology, E1Y6D), anti-HCMV-PP65 (gift from Dr. Minhua Luo), anti-RABV-M (gift from Dr. LingZhao), anti-Flag (Sigma‒Aldrich, F3165), anti-HA (Abcam, ab9110), anti-V5 (Proteintech, 14,440–1-AP), and anti-GAPDH (ABclonal, AC033) antibodies were used. SR9009 (MedChemExpress, HY-16989), Lipofectamine™ 2000 Transfection Reagent (Invitrogen 1,679,991), MG132 (Sigma‒Aldrich, C2211), CHX (MedChemExpress, HY-12320), and LiCl (MedChemExpress, HY-Y0649) were obtained from the indicated suppliers.

REV-ERBα-Flag (gene ID: 9572), E2F8-Flag (ge ne ID: 79733), E2F7-Flag (gene ID: 36270), E2F5-Flag (gene ID: 1875), and related mutants were inserted into the pQCXIP retroviral vector (deposited by Martin Dorf). p75NTR-luc, BMAL1-luc, and related mutants were inserted into the PGL4 vector (deposited by Martin Dorf). The plasmids described above were constructed for this study. HUWE1-Flag, HUWE1-HA, VSV-G-Flag, Ebola virus-G, CVS 11-P-Flag, CVS 11-M-Flag, CVS 11-N-Flag, CVS 11-G-Flag, ERA-G-Flag, B19-G-Flag, Nishigahara-G-Flag, DRV-Mexico-G-Flag, Ub-WT-HA, and Ub-K48-HA plasmids were preserved in the laboratory.

HCMV was a gift from Dr. Minhua Luo. Coxsackievirus B3 was a gift from Dr. Yin Wu. HSV-1 was a gift from Dr. Junjie Zhang. RABV, Zika virus, and DENV were preserved in the laboratory. The pseudoviruses were constructed through a lentiviral packaging system, as previously described⁷⁷. Cells for various infection assays were seeded in 24-well plates. When the cell density reached 90% in both the experimental and control groups, RABV or HCMV was added at a multiplicity of infection (MOI) of 0.1 or 5, respectively. Following a 48 h incubation period post-infection, the cells were harvested to assess the viral load via western blot analysis. In the viral entry assay, cells were cultivated in 24-well plates to 90% confluency. Viruses were added at an MOI of 10 and incubated at 4 °C for 1 h to allow sufficient attachment to the cell surface. Nonadherent viruses were removed by washing with PBS, followed by a 1 h incubation at 37 °C. Cells were then collected, and viral genome abundance was detected by qPCR. For all the RABV infection experiments involving animals in this study, the animals were anesthetized with isoflurane. The RABV strain CVS-11, with a titer of 10³ TCID₅₀ in 25 μL, was administered intranasally to mice as described by Rosseels et al.¹⁰⁰. At the required time points after viral infection, the mice were euthanized, and mouse brain tissue was collected for qPCR, western blotting assay, immunofluorescence staining, and immunohistochemistry analysis of virus abundance and brain tissue inflammation.

For daily cycle experiments, WT mice were maintained under constant environmental conditions and entrained to a 12 h light/12 h dark cycle (light period from 8:00 a.m. to 8:00 p.m.) for at least 2 weeks before the experiments. The animals were euthanized at 8:00, 14:00, 20:00, and 2:00 the next day and at 24:00, corresponding to ZT0, ZT6, ZT12, ZT18, and ZT24, respectively. For experiments conducted under constant darkness, the animals were kept in constant darkness for 3 days prior to sampling at CT0, CT6, CT12, CT18, and CT24. After euthanasia, the brain tissues were collected for further analysis of the rhythmic changes in circadian clock factors and viral receptors. REV-ERBα-KO mice, as well as their WT littermates, were subjected to a regular light-dark cycle (12 h light:12 h dark) for more than two weeks. At ZT12 (20:00), which has been reported as the time point of maximal REV-ERBα activity at promoter sites¹⁰¹, intranasal viral infection was initiated.

To establish a chronic jet lag model in mice, we progressively advanced the light-dark cycle of the mouse housing environment by 6 h every 2 days, while the control group of mice remained under a regular light-dark cycle of 12 h each. This altered lighting schedule was maintained for a minimum of three weeks. Concurrent with the final light-phase advancement, both the CJL model and control groups of mice were infected with CVS-11.

A Comprehensive Lab Animal Monitoring System (CLAMS; Columbus Instruments) was utilized to assess the wheel-running activity, metabolism, and feeding behavior of the mice. All monitored mice were individually housed in cages with ad libitum access to water and food. For experiments involving the effect of virus on mouse circadian rhythms, mice were maintained under a regular 12 h light and 12 h dark cycle, and the experimental endpoint was determined by the loss of locomotor activity following virus-induced encephalitis. In the behavioral monitoring experiments of the CJL model mice, the environmental lighting conditions were consistent with those mentioned above, whereas the WT mice were housed under standard regular lighting conditions.

The experiments were conducted according to the instructions provided by Vazyme (TD904). In brief, cerebral organoids were digested into single cells using Cell Recovery Solution for Organoids (Yeasen, 41421ES60), and approximately 2.5 × 10⁵ cells were collected. The dissociated cells were cross-linked with Enhancer buffer, combined with activated ConA beads and incubated overnight with anti-BMAL1 antibodies. Following incubation, the supernatant was removed, and the samples were incubated with a secondary antibody (Vazyme, Ab207) for 1 h at room temperature with rotation. The supernatant was subsequently discarded, and the ConA beads were incubated with pA/G-Tnp Pro and TTBL, each for 1 h at both room temperature and 37 °C. After TTBL incubation, 2 mL of 10% SDS was added to each sample and incubated at 55 °C for 10 min. The fragmented DNA in the supernatant was enriched using activated DNA Extract Beads Pro, and a next-generation sequencing library was constructed according to the manufacturer’s protocol (Vazyme, TD202). Sequencing was performed on the Illumina NovaSeq platform by Frasergen Bioinformatics Co., Ltd.

The precipitated chromatin DNA samples were obtained with a Hyperactive Universal CUT&Tag Assay Kit (TD904, Vazyme), in which the fragmented DNA was released from the DNA extract beads using CUT&Tag stop buffer (TD904-C1), followed by qPCR to quantify the abundance of the target gene. The samples were analyzed on a Bio-Rad CFX96 real-time PCR system using Taq Pro Universal SYBR qPCR Master Mix (Vazyme). The enrichment fold change of the REV-ERBα and E2F8 binding sites relative to that of DNA spike-in was determined using threshold cycle values and the 2^−ΔΔCt method.

We acquired the GSE176393↗, GSE248721↗, GSE222744↗, GSE71574↗, GSE52157↗, GSE234789↗, GSM7503111↗ and GSM8123058↗ datasets, which contain RNA-seq data or ChIP-seq data (GSM7503111↗ and GSM8123058↗) from diverse cell types and murine tissues that are pertinent from the public repository of the GEO website. The narrowPeak format files from the GSM7503111↗ dataset were processed using the ChIPSeeker package, and the bigWig format file of the GSM8123058↗ dataset was processed using the rtracklayer and GenomicRanges packages; the results were visualized with the ggplot2 package. A gene was defined as a differentially expressed gene (DEG) when there was a statistically significant difference in the transcription level of a gene in different test environments. The DEGs were identified from the long-expression values using the LIMMA package with Benjamini‒Hochberg correction to control the rate of false discovery and DESEq2 in multiple testing options. Cutoff criteria (P value < 0.05 and |logFC| ≥ 1.0) were applied to detect significant DEGs from all the datasets. All big data analyses were conducted using the R programming language (v4.4.2).

All analyses were performed using R software (v4.5.2). For rhythmic gene fitting curves that exhibited approximate symmetry, autoregressive spectral estimation (ARS) was applied to assess rhythmic significance. Analyses were conducted in R using the MetaCycle package, which implements ARS alongside other rhythmicity detection algorithms. The ARS procedure was run via the meta2d() function, which specified the range of expected periods (minper and maxper) and a default period (ARSdefaultPer) within this range.

For rhythmic gene fitting curves that were asymmetric or based on sparse sampling points, we applied cosine fitting combined with joint significance testing of harmonic components (CFJHC) to evaluate rhythmicity. Specifically, we used the base R functions for linear modeling and the MASS package for multivariate normal sampling. The observed values were fitted against cosine and sine terms corresponding to the specified period (e.g., 24 h). The model provided estimates of amplitude (defined as the square root of the sum of squared cosine and sine coefficients), peak-to-trough difference (twice the amplitude), and phase (calculated as the arctangent of the ratio between sine and cosine coefficients). To account for uncertainty in amplitude, we generated repeated parameter samples from the joint distribution of the model coefficients using the mvrnorm() function in the MASS package and calculated the amplitude for each sample. To determine rhythmicity, amplitudes were compared with a predefined threshold derived from nonrhythmic genes. A one-sample t-test was performed using the base R’s t.test() function to evaluate whether the mean amplitude exceeded this threshold, and genes showing significant results were classified as rhythmic.

The CUT&Tag-seq data that support the findings of this study have been deposited in the NCBI Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/↗) with the accession number GSE310611↗. All the relevant data are available from the authors upon reasonable request.