An Intrinsic Disorder Region Controlling Condensation of a Circadian Clock Component and Rhythmic Transcription in the Liver

SUMMARY

Circadian gene transcription is fundamental to metabolic physiology. Here we report that the nuclear receptor REV-ERBα, a repressive component of the molecular clock, forms circadian condensates in the nuclei of mouse liver. These condensates are dictated by an intrinsically disordered region (IDR) located in the protein’s hinge region which specifically concentrates nuclear receptor corepressor 1 (NCOR1) at the genome. IDR deletion diminishes the recruitment of NCOR1 and disrupts rhythmic gene transcription in vivo. REV-ERBα condensates are located at high-order transcriptional repressive hubs in the liver genome that are highly correlated with circadian gene repression. Deletion of the IDR disrupts transcriptional repressive hubs and diminishes silencing of target genes by REV-ERBα. This work demonstrates physiological circadian protein condensates containing REV-ERBα whose IDR is required for hub formation and the control of rhythmic gene expression.

Graphical Abstract

eTOC Blurb

Zhu et al report that a core component of the molecular clock, REV-ERBα, forms circadian condensates controlling liver rhythmic gene repression.

INTRODUCTION

The molecular components and genetic mechanisms of circadian clocks have been extensively explored during the past few decades, leading to the well-established transcription-translation-feedback-loops (TTFLs) consisting of core clock factors that regulate the oscillation of behavior and physiology with a 24-hour period^1,2. Transcriptional control of mammalian clock genes lies at the center of the cellautonomous circadian clock and is highly orchestrated. Core molecular oscillators govern a large number of clock-controlled genes (CCGs) in a tissue-specific manner with the cooperative roles of tissue-specific transcription factors (TFs), variable chromatin accessibility and high-order chromatin topology^3–6. The main positive arm of TTFLs comprises transcription factors Bmal1, Clock and its homologue Npas2, which activates clock-controlled genes (CCGs) as well as core clock TFs that feedback to their expression to complete the TTFLs⁷. A major negative feedback pathway is mediated by nuclear receptors REV-ERBα and β, which are rhythmically expressed with a huge circadian amplitude that is anti-phase to repression targets Bmal1 and Npas2⁸. In general, REV-ERBα is of primary importance, with β serving a redundant function⁹.

In liver, REV-ERBs couple circadian rhythms and physiological metabolism². In addition to their role in the clock, they potently repress numerous other genes critical for the metabolic function of the liver in a diurnal manner¹⁰. In mice, REV-ERBα gene expression and protein levels peak around ZT10 (5PM) and drop to undetectable levels by ZT22 (5 AM)^8,11. Many transcripts repressed by REV-ERBα in liver have been shown to be located near genomic binding sites to which REV-ERBα recruits nuclear receptor corepressor 1 (NCOR1) in complex with histone deacetylase 3 (HDAC3), resulting in reduced histone acetylation and enhancer RNA (eRNA) expression^3,12. REV-ERBα binding has been shown to alter 3D genome architecture in a diurnal manner^5,6, but how its binding to cis-regulatory elements controls the specificity and magnitude of target gene repression remains mysterious.

Recent advances in the understanding of gene transcription have suggested a role of TF binding and function at hubs where, via looping and other mechanisms, multiple regions of chromatin come together^13–15. One way that TFs or coactivators congregate at hubs is mediated by intrinsically disordered regions (IDRs)¹⁶. Proteins containing IDRs or multivalent proteins tend to form dynamic higher-order assemblies/condensates, which have been linked to important biological functions¹⁷. The list of intracellular assemblies/condensates driven by liquid-liquid phase condensation is rapidly evolving, but there is little evidence of their function in the physiological context. Much of the evidence for TF condensates and hubs have been based on the study of transcriptional activation, although one study performed in Ciona embryos showed that the transcription repressor Hes forms liquid droplets for gene silencing¹⁸. Moreover, the vast majority of studies have been performed in cell-free systems and cells in culture^19–21, and thus little is known about the physiological importance of condensates and related phenomena in living animals. Here we demonstrate the physiological rhythmic spatiotemporal organization of a mammalian circadian clock component REV-ERBα, in mouse liver. We identify an IDR that is required both for liquid-liquid phase separation of REV-ERBα and recruitment of NCOR1 corepressor in cultured cells as well as the robust formation of REV-ERBα-containing chromatin hubs which are responsible for the synchronized physiological repression of clock-controlled genes in intact, living livers.

RESULTS

REV-ERBα forms dynamic physiological circadian condensates in vivo

In mouse liver, REV-ERBα mRNA and protein exhibit a physiological circadian expression pattern with peak around ZT10 (5PM) and nadir at ZT22 (5AM)²². Immunohistochemistry of frozen sections using an antibody validated in livers depleted of REV-ERBα (Figure S1A) revealed that endogenous REV-ERBα is concentrated in discrete puncta at the times of day when the protein is expressed (ZT7 and ZT10) (Figures 1A and S1B). Intriguingly, while most REV-ERBα puncta displayed homogenous and round shape (puncta P1,P2,P3 in Figure 1B and puncta P7,P8 in Figure S1C), some puncta showed an annulus or donut-like structure (puncta P4, P5, P6, P9 in Figures 1B and S1C) when imaged using high-resolution confocal microscopy in Airyscanning mode to reveal fine detailed structures²³(Figures 1B and S1C). Consistent with mouse liver staining, endogenous REV-ERBα also exhibits punctate patterns in the nucleus of human U2OS cells when stained with two independent REV-ERBα antibodies (Figures S1D and S1E).

Figure 1. REV-ERBα forms dynamic physiological circadian condensates in vivo.

(A) IHC staining of mouse liver frozen section at different circadian time points (ZT1, 4, 7, 10, 13, 16, 19, 22), the rectangle box at ZT7 and ZT10 represents magnified individual puncta; (B) High resolution imaging of a mouse liver frozen slide at ZT10 using Airyscanning mode. Puncta 1,2,3,4 (annotated as P1, P2, P3, P4 in the dashed rectangle boxes) are shown zoomed in at the bottom; (C) Real-time tracking of EYFP-REV-ERBα puncta dynamics in the live U2OS cells, see also Movie S1; (D-E) FRAP experiments examining REV-ERBα puncta movement, see also Movie S2; (F-G) REV-ERBα puncta are treated by 10% 1,6- hexanediol or 10% 2,5-hexanediol for ~1 min before image acquisition (1 A&B scale bar=2 μm; 1 C-G scale bar=10 μm). See also Figures S1.

The punctate pattern of endogenous nuclear REV-ERBα was reminiscent of protein condensates that have been suggested to play a role in gene transcription^19–21. Protein condensation is favored by liquid-liquid phase separation that may be mediated by intrinsically disordered regions (IDR)²⁴. REV-ERBα has two major IDRs, one positioned at the N terminus (nIDR: 1–123 aa) and one in the hinge region (hIDR: 218–412 aa) (Figure S1F). To elucidate the biophysical properties of REV-ERBα IDRs, we expressed fluorescently-tagged proteins in U2OS cells. Nuclear droplets were observed with each of 3 different fluorescent tags, suggesting that the formation of REV-ERBα droplets is independent of different tagging (Figures S1G and S1H). Droplet formation by Enhanced Yellow Fluorescent Protein (EYFP)-REV-ERBα was concentration-dependent and occurred even at expression levels lower than that of the endogenous REV- ERBα (Figures S1I and S1J).

Nuclear condensates driven by multivalent IDR-IDR interactions exhibit the biophysical properties of rapid movement and fusion²⁵. In live cells, EYFP-REV-ERBα droplets coalesced into single larger droplets upon contact one with another, indicating that the droplets possess liquid-like properties (Figure 1C; Movie S1). Fluorescence Recovery After Photobleaching (FRAP) analysis revealed that the intensity and rounded structure of REV-ERBα puncta were recovered rapidly after bleaching (T_1/2=~18s), demonstrating dynamic exchange with the surrounding nucleoplasm (Figure 1D and 1E; Movie S2). Moreover, the condensation of EYFP-REV-ERBα was disrupted by 1,6-Hexanediol (Figure 1F), an agent that disrupts biomolecular condensates^15,19,26. In contrast, 2,5-Hexanediol, an aliphatic alcohol that has been shown lower dissolution capabilities on protein condensates²⁷, has less disruptive effects on REV-ERBα puncta (Figure 1G), consistent with the notion that REV-ERBα puncta formation might be driven by transient intermolecular interactions between IDRs^28,29. These results collectively show that REV-ERBα forms dynamic nuclear condensates in vivo as well as in cultured cells.

We next determined the contribution of individual IDR regions to the transcriptional and biophysical properties of REV-ERBα. hIDR was required for REV-ERBα-dependent repression of a Bmal1-promoter luciferase reporter in U2OS cells, whereas the potency of a REV-ERBα mutant lacking nIDR (ΔnIDR) was similar to that of REV-ERBα wildtype (WT) (Figure S2A). Consistent with the reporter assay, the ΔhIDR mutant lost the repression effects on Npas2 gene in the liver while the ΔnIDR mutant still had similar effects as the WT (Figure S2B). A major difference between hIDR and nIDR was that, using a LacI-LacO genomic assay to visualize condensates (Figure S2C)^21,26,30, hIDR puncta exhibited a hollow core encompassed by an outer shell whereas nIDR formed homogenous puncta in both U2OS and AML12 cells (Figures 2A, S2D and S2E). As control, the structured DNA-binding domain (DBD) did not form condensates in this assay (Figures S2D and S2E). Moreover, a mutant lacking the hIDR (ΔhIDR) lost the ability to form droplets (Figure S2F).

Figure 2. IDR perturbation disrupts REV-ERBα condensates-mediated liver gene repression.

(A) Live imaging of EYFP-tagged REV-ERBα DBD (125–202 aa), hIDR (218–412 aa) in the U2OS cells; (B-C) Live imaging of co-expressed EYFP-REV-ERBα or EYFP-hIDR together with mCherry-HSPA8 in the U2OS cells, droplet within dashed box is magnified at the bottom right; (D) Gene expression change of REV-ERBα targeted genes in scramble and HSPA8 depleted liver by shRNA knockdown (n=4~5); (E) Workflow of AAV8 virus packaging and delivery strategy for REV-ERBα WT or ΔhIDR together with Cre to mouse liver; (F) Gene expression of REV-ERBα targeted genes at ZT10 in Ctrl or DKO liver injected by viruses labeled in the x axis, each individual dot representing one biological replicate is stacked by the geom_dotplot from the ggplot2 package (n=3~4); (G) H3K27ac ChIP-seq signal at the promoter regions (−1 to 1 kb) of REV-ERBα target genes, genomic tracking of REV-ERBα canonical targets Npas2 is shown on the right (dashed box highlighted the promoter region of Npas2). (scale bar=10 μm, error bar represents SEM, students’ t-test, * p< 0.05, ** p<0.01, *** p<0.001). See also Figures S2, S3, S4.

The annulus/donut-like condensate architecture was shared between hIDR and full-length REV-ERBα protein (Figure 2B), and resembled the donut-like (anisosome) structure of TDP-43, which utilizes HSP70 family members as its core scaffold (18). Among HSP70 family members, HSPA5, HSPA8 and HSPA9 are highly and constitutively expressed in the mouse liver during a 24-hr circadian period (Figures S3A and S3B). Of these, only HSPA8 (also called HSC70) is nuclear as well as cytoplasmic (19), while HSPA5 and HSPA9 are localized in endoplasmic reticulum and mitochondria, respectively. Live imaging of mCherry-tagged HSPA8 and EYFP-tagged REV-ERBα or hIDR in U2OS cells showed that HSPA8 is located in the center of the full-length REV-ERBα (Figures 2B, S3C and S3D) or hIDR annulus/donut-shaped condensates (Figure 2C) but not in the nIDR puncta or DBD truncated protein (Figure S3E and S3F).

Functionally, ectopic expression of HSPA8 augmented REV-ERBα repression activity in a Bmal1-promoter luciferase reporter assay in U2OS cells (Figure S3G). Furthermore, an AAV8-shRNA that efficiently knocked down HSPA8 mRNA and protein in the liver (Figures S3H–S3J) led to the de-repression of canonical REV-ERBα target genes, including Npas2, Ndrg1, Bmal1, Cry1 and Nfil3 (Figures 2D, S3K). Therefore, hIDR was critical for transcriptional repression by REV-ERBα and sufficient for the annulus/donut-like structure formation, HSPA8 recruitment.

We next investigated the physiological role of REV-ERBα hIDR in the circadian control of transcription in mouse liver. REV-ERBα wildtype (WT) and ΔhIDR mutant were expressed in hepatocytes of adult mice which were also depleted of endogenous REV-ERBα and β (DKO) as previously described^8,22, (Figures 2E, S4A and S4B). As expected⁸, a large number of rhythmic genes including core clock genes Npas2, Bmal1, Cry1, Nfil3, and other rhythmic genes such as Loxl4 and Nrdg1 were derepressed in the REV-ERB DKO at the time of day when physiological REV-ERB expression is normally highest (ZT10; 5 PM) (Figures S4C and S4D). Re-expression of REV-ERBα in the DKO rescued a large percentage of de-repressed genes involved in circadian rhythm, lipid and glutathione metabolism (135/508, Figures S4E and S4F). However, the hIDR mutant was less effective than WT at repressing these genes (Figure 2F), indicating that the additional roles of hIDR for fully efficient repression of these genes. Of note, though all of these REV-ERBα target genes showed a statistically significant difference between WT and IDR mutant groups, the magnitude of the IDR-dependence of repression by REV-ERB varied considerably, suggesting gene-specificity of the importance of the IDR relative to other mechanisms, including the direct structured interaction between the REV-ERBα LBD and NCOR1^31,32. Consistent with the lesser rescue of repression by the hIDR mutant, the levels of H3K27ac surrounding the promoter of the REV-ERBα repressed genes were elevated relative to WT rescue (Figure 2G). These results suggest that the hIDR required for forming REV-ERBα condensates is critical for its physiological role in regulating the repression of genes involved in the core clock and metabolic pathways in the liver.

IDR-mediated REV-ERBα and NCOR1 partitioning regulate liver circadian gene transcription

Transcriptional repression mediated by REV-ERBα relies on the recruitment of NCOR1¹². Unlike REV-ERBα, NCOR1 is constitutively expressed in the liver over the course of the day (Figure S5A). Intriguingly, NCOR1 contains extended IDRs that make up over 50% of this large protein (Figure 3A), and constitutively formed puncta in the liver nucleus during a 24-hr cycle by the frozen section staining of an endogenous NCOR1 antibody validated before³³(Figures 3B, S5B and S5C), which is in agreement with its constant expression. Interestingly, overexpressed mCherry-tagged NCOR1 protein alone also formed droplets in the live U2OS cells (Figure 3C), consistent with the endogenous NCOR1 puncta identified in mouse liver. In U2OS cells, NCOR1 co-localized with REV-ERBα in the outer layer of annulus/donut-like condensates (Figure 3D). In contrast, the well-established MED1 IDR condensates¹⁹ were not partitioned from REV-ERBα droplets (Figure S5D), suggesting the specificity for the composition of the repressive complexes.

Figure 3. IDR-mediated REV-ERBα and NCOR1 partitioning controls liver circadian gene transcription.

(A) IDR distribution of NCOR1 protein predicted by DisPro tool; (B) Frozen section staining of mouse liver harvested at ZT22 and ZT10 by an endogenous NCOR1 antibody; (C) Live imaging of NCOR1-mCherry in the U2OS cells; (D) Live imaging of co-expressed EYFP-REV-ERBα and NCOR1-mCherry in the U2OS cells; (E) Heatmap of NCOR1 binding signal from ZT10 and ZT22 liver at NCOR1 ZT10-enriched binding regions (n=2520); (F) Scatterplot showing correlation between NCOR1 and REV-ERBα binding signal at genomic region defined from E; (G) Genome browser view of REV-ERBα and NCOR1 binding at Loxl4 gene at ZT22 and ZT10; (H) Cartoon model of LacO array genomic visualization in the U2OS cells; (I) Co-staining of EYFP antibody and NCOR1 antibody in the U2OS cells transfected by EYFP-DBD-LacI-NLS, EYFP-hIDR-LacI-NLS, EYFP-REV-ERBα-LacI-NLS or EYFP-REV-ERBα-ΔhIDR-LacI-NLS; (J-K) Scatterplot of NCOR1 binding signal in WT Vs. ΔhIDR, blue dots represent increased binding in ΔhIDR (FC>=1.3, n=3), green dot represents reduced binding in ΔhIDR (FC <=0.77, n=2783), which is shown by heatmap on the bottom (K); (L-M) Scatterplot of H3K27ac intensity in WT Vs. ΔhIDR, blue dots represent reduced binding in ΔhIDR (FC <=0.77, n=46), green dots represent increased binding in ΔhIDR (FC>=1.3, n=522),which is shown by heatmap on the bottom (M). (scale bar=10 μm). See also Figures S5, S6.

To further explore the co-dependence between NCOR1 and REV-ERBα in liver, we performed NCOR1 and REV-ERBα ChIP-seq of livers in 3XHA knock-in REV-ERBα (REV-ERBα^3HA/3HA) mice at ZT22 and ZT10 (Figure S6A). In line with the circadian expression of the 3HA-REV-ERBα that is indistinguishable from wild type³⁴, circadian REV-ERBα has almost no binding signal at ZT22 while massively binding to the genome at ZT10 (Figure S6B). NCOR1 ChIP-seq revealed a subset of REV-ERBα binding sites at which NCOR1 bound rhythmically (Figures 3E and S6C) and, as expected, these sites were also bound by REV-ERBα (Figures 3F, 3G, and S6D). Together these data show that REV-ERBα recruits NCOR1 to its condensates and its genomic locations in a circadian manner.

We next interrogated the role of the hIDR in contributing to the concentration of REV-ERBαNCOR1 on the genome. By LacI-LacO genomic visualization assay in U2OS (Figure 3H), the REV-ERBα DBD did not concentrate endogenous NCOR1 at the LacO locus, whereas hIDR and REV-ERBα wild type (WT) but not ΔhIDR could efficiently concentrate NCOR1 (Figure 3I). To determine the function of the IDR in vivo, we performed NCOR1 and H3K27ac ChIP-seq on DKO livers in which WT or hIDR mutant REV-ERBα was re-expressed. NCOR1 binding was greater at a large number of REV-ERBα sites in WT-expressing livers than in those expressing the hIDR mutant (Figures 3J and 3K). For example, NCOR1 binding was markedly reduced while H3K27ac was increased at the REV-ERBα binding site in the Loxl4 gene in the DKO, and both were restored by WT REV-ERBα but much less by the hIDR mutant (Figure S6E). NCOR1 complexes contain HDAC3³⁵ and, consistent with its enzyme activity, H3K27ac was higher at REV-ERBα binding sites in livers expressing the hIDR mutant than from those rescued with WT REV-ERBα (Figures 3L and 3M). Collectively, these data demonstrate that REV-ERBα concentrated NCOR1 through hIDR at REV-ERBα bound loci to repress rhythmic gene expression.

Circadian REV-ERBα condensation drives rhythmic liver gene expression

Having found that REV-ERBα droplets selectively concentrate NCOR1 through hIDR-mediated condensation, we next considered whether the REV-ERBα hIDR has a physiological role in a circadian context. To test this, we used the AAV8 liver gene delivery system with the mouse REV-ERBα core promoter sequence (Nr1d1p) to drive circadian expression of cargo in liver (Figure 4A). Using this system, REV-ERBα mRNA and protein both exhibited rhythmic expression patterns, with ZT10>ZT22 as expected after re-expression from AV8-Nr1d1p in DKO livers (Figures S7A and S7B). Moreover, this led to efficient repression of REV-ERBα canonical target genes Npas2, Bmal1, Cry1 and Nfil3 in a circadian manner (Figure S7C), demonstrating that this circadian transgene delivery system could mimic REV-ERBα circadian patterns in liver. We then compared the functions of wild type and hIDR mutant REV-ERBα using this circadian transgene expression system. As expected, core clock genes regulated by REV-ERBα displayed robust amplitude between ZT22 and ZT10 and were depressed in the DKO group at ZT22. With REV-ERBα re-expression driven by Nr1d1p promoter, these genes were highly repressed again at ZT10 but showed no repression (Figures 4B, 4D and 4E) or a lesser degree of repression (Figure 4C) at ZT22. In contrast, the REV-ERBα hIDR mutant failed to repress these genes both at ZT10 or ZT22. In addition to core clock genes, other REV-ERBα rhythmic target genes also showed a similar pattern by IDR mutation (Figure S7D). Collectively, these data show that hIDR was required for REV-ERBα control of circadian gene repression in vivo.

Figure 4. REV-ERBα hIDR mediates liver rhythmic gene expression.

(A) Workflow of circadian REV-ERBα expression cargo design and delivery into a mouse liver; (B-E) Cyclic expression pattern of REV-ERBα canonical core clock components Bmal1, Npas2, Nfil3, Cry1 in different mouse groups injected with the AAV virus labeled in the top axis. (Error bar represents SEM, students’ t-test, * p< 0.05, ** p<0.01, *** p<0.001). See also Figure S7.

REV-ERBα disordered region enhances repressive hubs formation for circadian gene regulation

Nuclear condensates are linked to chromatin architecture through multivalent interactions between molecules^15,36,37. In Drosophila, one remarkable chromatin architecture change is the rhymimc repositioning of clock genes from periphery to the center of nucleus regulated by PERIOD (PER) condensates³⁸, and therefore we hypothesized that REV-ERBα condensates could govern hepatic circadian gene expression by a related mechanism. However, visualization of the genomic loci for one canonical REV-ERBα target gene, Npas2, by DNA-FISH showed no obvious change in their nucleus position between ZT22 and ZT10, despite the dramatic difference in their expression levels at these two time points (Figure S8A), suggesting there might be other mechanisms to regulate liver circadian expression. Considering the low resolution and input of conventional DNA FISH for mapping chromatin architecture, we developed a high-resolution method to map 3D chromatin architecture in vivo utilizing Tissue In-situ ChIA-PET with improved ChIP-seq using Tn5 transposase to map the occupancy of REV-ERBα condensates on the mouse genome in both 2-dimensional and 3-dimensional space (Figure S8B). To validate the method, we also mapped the general 3D chromatin architecture of mouse liver using CTCF antibody with In-situ ChIA-PET, which served as a positive control^39,40. Analysis of CTCF In-situ ChIA-PET using a modified ChIA-PIPE tool (Figure S8C)⁴¹ demonstrated the robustness and reproducibility of the 3D chromatin architecture map compared to the gold standard of In-situ HiC (Figure S8D).

Utilizing REV-ERBα^3HA/3HA mice, we performed REV-ERBα In-situ ChIA-PET and Tn5-based ChIP-seq at the peak (5PM/ZT10) and trough (5AM/ZT22) of REV-ERBα expression in mouse liver. At ZT10 REV-ERBα bound at known multiple sites⁵ at the locus for Npas2, a canonical REV-ERBα target gene, but no binding signal was noted at ZT22 when REV-ERBα was not expressed (Figure 5A). These REV-ERBα binding peaks were strongly interconnected with one another and with the transcriptional start site (TSS) as marked by H3K4me3 and H3K27ac, which also showed rhythmic modification pattern correlating transcription of the Npas2 (Figure 5A). In addition to Npas2, gene expression was highly correlated with H3K4me3 and H3K27ac at the majority of promoters of REV-ERBα targeted genes (Figures S9A–S9C).

Figure 5. REV-ERBα disordered region enhances repressive hubs formation for circadian gene regulation.

(A) Example of ChIA-PET and ChIP-seq in REV-ERBα^3HA/3HA liver hepatocytes at ZT22 and ZT10 at Npas2 locus; (B-C) Definition and analysis pipeline of REV-ERBRα associated Super Anchor (RESA); (D) Rank-order plot of all RESA hubs, the top 187 are marked in red while the rest (2829) are in grey; (E) Percentage of derepressed genes in DKO ranked by RESA binding score; (F) Correlation between repression fold change and normalized RESA binding score; (G) Workflow of AAV8-CRISPRa system; (H) Gene expression of Ndrg1 and its nearby genes (Wisp1, St3gal1, Sla) from REV-ERBα^3HA/3HA mice injected by gLacZ or gNdrg1 at ZT10, dSaCas9 expression are also checked in the two groups; (I) ChIP-qPCR of HA-REV-ERBα binding at Ndrg1 A1 anchor in REV-ERBα^3HA/3HA mice injected by gLacZ or gNdrg1, B6-WT mice were used as control for anti-HA antibody here. See Figure S9G for the definition of A1 anchor. (J) Npas2 genomic track view of REV-ERBα WT and ΔhIDR ChIA-PET in DKO liver; (K) Boxplot of RESA interconnection score of top RESAs in REV-ERBα WT and ΔhIDR. (Error bar represents SEM, students’ t-test, * p< 0.05, ** p<0.01, *** p<0.001). See also Figures S8, S9.

In total, 57,867 high-confidence REV-ERBα ChIA-PET loops (PET >=3) were identified, of which 46,801 were characterized by strong REV-ERBα binding at one or both anchors, and 27,326 were linked to one or multiple TSS regions directly or transitively (Figure 5B), consolidating the understanding that REV-ERBα mainly functions in the circadian gene transcription. We focused on these promoter-associated REV-ERBα loops for further analyses. For each active TSS annotated by H3K4me3 signal, all REV-ERBα ChIA-PET loop anchors linking it directly or transitively through intermediate anchor nodes were defined as a REV-ERBα-associated Super Anchors (RESA) hub for this gene (Figure 5C; STAR Methods). RESA hubs covered the majority (76%) of de-repressed genes in the DKO group (Figure S9D), suggesting that the RESA was a highly predictive marker for REV-ERBα-mediated repression. A rank-order plot revealed that 187 genes (red dots) had a particularly high RESA binding score (Figure 5D), and the majority of top RESA scoring genes were derepressed in the DKO whereas the low RESA genes were less likely to be REV-ERBα targets (Figure 5E). This strong relation (p = 2.82e⁻¹¹) between RESA binding score and gene repression was further illustrated by a scatterplot linking RESA score to magnitude of gene upregulation in the DKO livers (Figure 5F). Together these data demonstrate the RESA score, which is a reflection of REV-ERBα binding intensity at hubs of physical interactions between cis-binding sites and the gene promoter, to be an excellent predictor of REV-ERBα-mediated transcription repression.

To interrogate a causal link between the RESA and transcriptional repression in vivo, we developed a one-vector AAV8-CRISPRa system to deliver nuclease dead SaCas9 (dSaCas9) fused to the VP64 transcriptional activator along with gRNAs to modulate endogenous gene transcription in hepatocytes by targeting RESA (Figures 5G and S9E). dCas9 has been widely used to modulate endogenous gene activity by competing with transcription factors through targeting their binding sites^42,43. The Ndrg1 RESA contains 4 major REV-ERBα-occupied anchors that interact with the TSS (Figure S9G) and a highly efficient gRNA targeting anchor 1 (Figure S9F, gNdrg1 #1), which is located in the Wisp1 gene body ~60 kb from the Ndrg1 promoter, strongly activated endogenous liver Ndrg1 expression at ZT10, but not nearby genes Wisp1, St3gal1 and Sla (Figure 5H). Binding of REV-ERBα was reduced at the A1 anchor, consistent with the gNdrg1 sequence specifically targeting at this site (Figure 5I). We also targeted RESAs of canonical REV-ERBα target gene Npas2 by AAV8-CRISPRa, which activated Npas2 but not its nearby target genes (Figures S9H and S9I). Collectively, these data reveal that RESAs are necessary for REV-ERBα-mediated rhythmic gene repression.

To further assess the relationships between the ability of IDR to form condensates, hub formation as revealed by RESA, and gene repression, we performed In-situ ChIA-PET on DKO livers re-expressing WT or hIDR mutant REV-ERBα. As illustrated by the Npas2 gene, the REV-ERBα hIDR mutant failed to restore interactions between REV-ERBα binding sites mediated by WT REV-ERBα, although both WT and hIDR mutant bound at Npas2 loci similarly (Figure 5J). Indeed, RESA interaction scores were markedly reduced in the hIDR mutant relative to WT livers (Figure 5K), further demonstrating the links between IDRs that mediate physiological condensate formation, REV-ERBα hub formation, and repression of rhythmic gene expression by REV-ERBα.

DISCUSSION

Through multiple orthogonal and advanced approaches, our results reveal a new mechanism for mammalian circadian gene regulation. Specially, our studies reveal that a core component of mammalian circadian clock, REV-ERBα, forms dynamic circadian condensates in vivo via an intrinsic disordered region that contributes to physiological repression of rhythmic hepatic gene transcription. In contrast to the PER/CRY complex, a second negative regulator of molecular clock which relies on the translocation between cytoplasm and nucleus¹, REV-ERBα is constitutively localized in the nucleus to mediate its repression roles. Nevertheless, given the extensive distribution of IDR in all mammalian clock core components (Figure S10A), the formation of clock protein condensates mediated by IDR could be a general mechanism for circadian gene regulation. Indeed, preliminary high-resolution imaging of other clock oscillators further supports the idea that the condensation of clock proteins could be a widespread phenomenon in mammalian physiology (Figures S10B and S10C). Although our work does not establish the precise the molecular details of REV-ERBα condensate formation, the differential disruption of the condensates by 1,6-Hexanediol versus 2,5-Hexanediol suggests that intermolecular interactions such as labile cross-β interactions might be important driving forces^27–29.

Transcriptional condensates have been suggested to play roles in chromatin hub formation¹⁵. Our work identified the formation of mammalian circadian clock condensates, which is consistent with several studies showing that clock protein condensation is essential for circadian rhythmicity in Cyanobacteria^44,45, Neurospora^46,47 and Drosophila³⁸. These findings suggest that circadian protein condensation is an evolutionarily conserved mechanism for controlling rhythmic biological processes in different organisms. However, it remains a mystery how the circadian condensates can interact with DNA or chromosome to control the cyclic expression of genes. By developing a high-resolution method to map 3D chromatin architecture in vivo we uncovered that the IDR region is essential for repressive transcriptional hub formation in the liver. Through looping, REV-ERBα condensates leverage distant REV-ERBα binding sites to efficiently control the robustness of circadian gene repression. Moreover, REV-ERBα ChIA-PET assays showed that REV-ERBα could couple the rhythmic expression profiling of several genes either located on the same chromosome or on different chromosome, resulting in the synchronized repression of its targeted genes in a more efficient and specific manner. This phenomenon of exploiting the high-order chromatin organization in mammalian cells might explain the relative paucity of discrete foci while still controlling a large number of target genes in the same phase. It would be of interest to investigate whether other mammalian clock components, such as PER, which has been shown to form only a few puncta in Drosophila melanogaster³⁸ and our mouse study (Figure S10B), can also occupy transcriptional hubs and control a large number of cyclic genes simultaneously.

Transcriptionally active condensates were originally discovered in vitro and found to mediate gene activation by facilitating concentration of transcriptional activators to chromatin hubs²⁰. Our studies demonstrate a repressive role for circadian REV-ERBα condensates in recruiting NCOR1 to RESAs and repress physiologically rhythmic gene expression in the mammalian liver. NCOR1 can directly interact with the LBD of REV-ERBα via structured interactions involving CoRNR motifs^48,49, the live cell and mouse liver data presented here demonstrate a second and mutually non-exclusive mechanism by which REV-ERBα may concentrate NCOR1 on chromatin to mediate transcriptional repression. In contrast to heterochromatin or polycomb repressive complex-mediated gene silencing, which persists for relatively long times⁵⁰, rhythmic transcription by REV-ERBα occurs over a 24 h-cycle and thus requires a more dynamic and fast on-off switch, which is provided by the huge amplitude of the rhythm of REV-ERBα protein expression in liver²². More generally, our findings reveal that multivalent IDR-IDR interactions of REV-ERBα/NCOR1 binding at multiple sites form repressive hubs which link to the promoters of gene targets that are physiologically repressed at times of day when REV-ERBα is expressed.

Multiphase condensates have been discovered in different membraneless organelles, including TDP-43 anisosomes and stress granules, both of which share HSP70 chaperones as their scaffold^51,52. Intriguingly, REV-ERBα similarly recruits HSPA8, a chaperone in the HSP70 family, for liquid-phase condensation, and HSPA8 plays a critical role in facilitating REV-ERBα condensation in cultured cells. The loss of HSPA8 in mouse liver causes derepression of rhythmic genes targeted by REV-ERBα hubs, suggesting a functional role for HSPA8 incorporation into REV-ERBα condensates in mouse liver. The hinge region IDR of REV-ERBα is critical for recruiting HSPA8 to induce the donut-like structures, and further mutagenesis of HSPA8 and REV-ERBα will be required to understand how HSPA8 contributes REV-ERBα condensates formation in animal studies. A critical role of HSP70 is to maintain TDP43 and FUS as liquid-like droplets, preventing their amyloid-like aggregation in the context of neurodegenerative diseases^53,54, thus raising the possibility that REV-ERBa-HSP70 condensates could play additional roles in pathological conditions involving proteostasis stress. Moreover, as other core clock components also contain IDRs, our work opens the possibility that circadian condensation may play critical roles in physiological and pathological contexts.

Limitations of the study

Due to current technological limitations, we were unable to visualize the dynamics of endogenous REV-ERBα puncta in live mice. However, our fixed liver staining and live imaging in cells have shown that REV-ERBα exists as circadian puncta in liver nuclei and exhibits liquid-like condensates by elucidating their biophysical properties. More advanced tools need to be developed in this field to study physiological condensates in vivo. Although we have successfully performed DNA FISH in mouse liver tissue, we cannot obtain good co-staining of chromatin and protein due to the very different methodologies for DNA FISH and protein condensate staining. Thus, new methods such as dCas9-based DNA staining need to be developed for in vivo co-staining. However, our REV-ERBα ChIA-PET data in liver strongly shows that REV-ERBα puncta in vivo are associated with repressive chromatin loops, and REV-ERBα WT and hIDR mutant ChIA-PET data further implies that the formation of REV-ERBα puncta is crucial for repressive chromatin looping. Finally, although our finding that the REV-ERB IDR is required for condensate formation, recruitment of NCOR1 corepressor, robust formation of REV-ERBα-containing chromatin hubs, and physiological repression of REV-ERB target genes in mouse liver, the relationship between condensate formation and transcriptional regulation in vivo must be considered strongly correlative rather than causative based on the present data.

STAR*METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Dr. Mitchell A. Lazar (lazar@pennmedicine.upenn.edu)

Materials availability

All plasmids generated in this study are available upon request. M.A.L. obtained REV-ERBβ floxed mice under a limited material transfer agreement with the Centre Europeen de Recherche en Biologie et Medecine.

Data and code availability

• All RNA-seq, ChIP-seq, ChIA-PET data and processed files are available at GEO: GSExxx. All blots and microscopy images have been deposited at Mendeley Data and are publicly available as of the date of publication. Accession numbers and DOI are listed in the key resources table.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
REV-ERBα	Novus Biologicals	NBP-84931
REV-ERBα	Thermo Fisher	MA5–20771
REV-ERBα	Abnova	H00009572-M16
NCOR1	Thermo Fisher	PA1–844A
NCOR1	Everest	EB09500
EYFP	GeneTex	GTX26673
Pierce™ Anti-HA Magnetic Beads	Thermo Fisher	88837
CTCF	Abcam	ab70303
H3K4me3	Millipore	04–745
H3K27ac	Active Motif	39133
NCOR1	In-house	N/A
HA-Tag	CST	3724
REV-ERBα	Novus Biologicals	NBP-84931
HSC70	Abcam	ab51052
β-Actin (HRP conjugate)	CST	5125
Vinculin (HRP conjugate)	CST	18799
Bacterial and virus strains
One Shot™ Stbl3™	Thermo Fisher	C737303
NEB® 5-alpha	NEB	C2987
AAV8-TBG-Cre	Penn Vector Core	N/A
AAV8-TBG-GFP	Penn Vector Core	N/A
Other AAV8 virus strains	In-house	N/A
Chemicals, peptides, and recombinant proteins
DMEM	Thermo Fisher	11995065
DMEM/F-12	Thermo Fisher	11330032
ITS	Thermo Fisher	41400045
1,6-Hexanediol	Sigma-Aldrich	88571–100ML-F
Fluoromount-G	SouthernBiotech	0100–01
2,5-Hexanediol	Sigma-Aldrich	H11904
Rubber Cement	Elmer’s	B000EFQ2I0
Normal-chow diet	LabDiet	5010
NEBuilder HiFi Assembly	NEB	E5520s
CloneAmp™ HiFi PCR Premix	Takara	639298
TGX™ Precast Protein Gel	Bio-Rad	4561034; 4561094
0.45-um PVDF membrane	Thermo Fisher	88518
chemiluminescent substrates	Thermo Fisher	32106; 34095
8-well chamber slide	Thermo Fisher	154534
mounting medium	SouthernBiotech	0100–01
FuGENE® HD	Promega	E2311
Lipofectamine® 3000	Thermo Fisher	L3000008
μ-Slide 8 Well chamber	ibidi	80826
Janelia HaloTag® Ligands A549	Promega	GA1110
TRIzol Reagent	Ambion	5457
SYBR Green PCR Master Mix	Thermo Fisher	4334973
LipoD293™	SignaGen® Laboratories	SL100668
Amicon® Ultra Centrifugal Filters	Millipore	UFC510024
EGS	Thermo Fisher	21565
O.C.T. Compound	Fisher Scientific	23–730-571
SPRIselect	Beckman coulter	B23318
AluI	NEB	R0137L
Dynabeads™ M-280 Streptavidin	Thermo Fisher	11206D
IGEPAL CA-630	Sigma	I8896
DynaBeads Protein G	Thermo Fisher	10004D
Critical commercial assays
Gel DNA Recovery kit	Zymo Research	D4007
RNeasy Mini Kit	Qiagen	74106
QIAquick PCR Purification Kit	Qiagen	28104
QIAprep Spin Miniprep Kit	Qiagen	27104
Dual-Luciferase® Reporter Assay	Promega	E1910
High Capacity cDNA Reverse Transcription Kit	Thermo Fisher	4368814
Illumina Ribo-Zero Plus rRNA Depletion Kit	Illumina	20040526
TruSeq Stranded Total RNA Library Prep kit	Illumina	20020599
Nextera XT DNA Library Preparation Kit	Illumina	FC-131–1024
Tagment DNA Enzyme and Buffer Kit	Illumina	20034197
Deposited data
RNA-seq data	This study	GEO: GSE202605
ChIP-seq data	This study	GEO: GSEXXX
ChIA-PET data	This study	GEO: GSEXXX
RNA-seq data	Guan et al.8	GEO: GSE143528
Unprocessed blots and images	This study	https://doi.org/10.17632/z359cvncmj.2
Experimental models: Cell lines
HEK 293FT	Thermo Fisher	R70007
U2OS 2–6-3	Janicki et al.30	PMID: 15006351
AML12	ATCC	CRL-2254
AAVpro® 293T	Takara	632273
Experimental models: Organisms/strains
REV-ERBαf/fREV-ERBβf/f	Guan et al.8	N/A
REV-ERBα3HA/3HA	Adlanmerini et al.34	N/A
Oligonucleotides
Oligonucleotides used in this study	This paper	See table S1
Recombinant DNA
pMY-EYFP-REV-ERBα	This paper	N/A
pMY-EYFP-DBD-LacI-NLS	This paper	N/A
pMY-EYFP-nIDR-LacI-NLS	This paper	N/A
pMY-EYFP-hIDR-LacI-NLS	This paper	N/A
pMY-mCherry-REV-ERBα	This paper	N/A
pMY-HaoTag-REV-ERBα	This paper	N/A
pMY-EYFP-REV-ERBα-LacI-NLS	This paper	N/A
pMY-EYFP-REV-ERBα-ΔhIDR-LacI-NLS	This paper	N/A
pMY-3HA-NCOR1-mCherry	This paper	N/A
pMY-mCherry-HSPA8	This paper	N/A
pMY-mCherry-mPer1	This paper	N/A
pMY-mCherry-Nfil3	This paper	N/A
pWZL-hygro-Flag HA TRAP220 wt	addgene	17433
pMY-mCherry-MED1IDR-LacI-NLS	This paper	N/A
pAAV-TBG-PIeGFP.WPRE	addgene	105535
pAAV-TBG-PI-3HA-REV-ERBα-ΔhlDR-NLS.WPRE	This paper	N/A
pAAV-Nr1d1p-PI-3HA-REV-ERBα	This paper	N/A
pAAV-Nr1d1p-PI-3HA-REV-ERBα-ΔhIDR-NLS	This paper	N/A
pAAV-U6-shScramble-TBG-3F-tdTomato	This paper	N/A
pAAV2/8	Addgene	112864
pAdDeltaF6	Addgene	112867
pX603-AAV-CMV∷NLS-dSaCas9-NLS-3xHA-bGHpA	Addgene	61594
pAAV-U6-shHpsa8-TBG-3F-tdTomato	This paper	N/A
pAAV-U6-gLaz-TBG-dSaCas9-VP64	This paper	N/A
pAAV-U6-gNdrg1-TBG-dSaCas9-VP64	This paper	N/A
pAAV-U6-gNpas2-TBG-dSaCas9-VP64	This paper	N/A
Software and algorithms
Samtools	Li et al.53	Version 1.6 http://samtools.sourceforge.net
CellProfiler	CellProfiler	Version 4.1.3 https://cellprofiler.org/
ImageJ (FIJI)	Open Source	https://imagej.net/software/fiji
SRA tool kit	NCBI	Version 2.8.2 https://github.com/ncbi/sra-tools
Bedtools	Quinlan and Hall54	Version 2.29.2 http://bedtools.readthedocs.io/en/latest/
STAR aligner	Dobin et al.55	Version2.7.3 https://github.com/alexdobin/STAR
DEseq2	Love et al.56	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
Bowtie	Langmead et al.57	Version 1.0.0 http://bowtie.cbcb.umd.edu.
MobiDB	Piovesan et al.58	https://mobidb.bio.unipd.it/
ChIA-PIPE	Lee et al.41	https://github.com/TheJacksonLaboratory/ChIA-PIPE
cutadapt	Martin	https://cutadapt.readthedocs.io/en/stable/
BWA mem	Li and Durbin59	https://github.com/lh3/bwa
featureCounts	Liao et al.60	https://rnnh.github.io/bioinfo-notebook/docs/featureCounts.html
picard	N/A	https://broadinstitute.github.io/picard/
MACS2	Zhang et al.61	https://hbctraining.github.io/Intro-to-ChIPseq/lessons/05_peak_calling_macs.html
karyoploteR	Gel et al.62	karyoploteR 1.25.0 http://bioconductor.org/packages/release/bioc/html/karyoploteR.html
deepTools	Ramírez et al.63	https://deeptools.readthedocs.io/en/develop/

EXPERIMENTAL MODEL AND SUBJECT DETAILS

U2OS, HEK 293FT, AAVpro^® 293T cells

U2OS, HEK 293FT, AAVpro^® 293T (Takara, #632273) cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM, #11995065, ThermoFisher Scientific) which containing 10%FBS, 1%P/S. All these cells were grown at 37 °C with 5% CO₂ in a humidified sterile incubator.

AML12 cells

AML12 cells were cultured in DMEM/F12 (ThermoFisher Scientific, #11330032) following by adding 10%FBS, 1%P/S, ITS (1X, ThermoFisher Scientific, #41400045), 40 ng/ml DEX. AML12 cells were cultured at 37 °C with 5% CO₂ in a humidified sterile incubator.

Animal studies

All mice in this study were housed under 12:12 day/night cycle (lights on at 7:00 am [ZT10] and off at 7:00 pm [ZT12]) with ad libitum to standard diet (LabDiet, #5010) and water. All mice in this study were performed at the age of 2–4 month and male gender is used in the study.

METHOD DETAILS

Recombinant DNA construction

All constructs generated in this paper are followed by the NEBuilder HiFi DNA Assembly Cloning Kit protocol (NEB, #E5520S). Briefly, the mouse cDNA library or available plasmids from addgenes (See resource table) were used as templates for PCR amplification (Takara, # 639298), these fragments together with linearized backbone were purified by Gel DNA Recovery kit (Zymo, #D4007) and assembled by the HiFi DNA Assembly master mix.

Western blot

Denatured protein samples extracted from cultures cells or mouse livers were run on a 10% or 4–20% TGX^™ Precast Protein Gel (Bio-Rad, #4561034 or #4561094). Protein was wet transferred to a 0.45-um PVDF membrane (ThermoFisher Scientific, #88518). After transfer, the membrane was blocked with 5% non-fat milk in TBST for 1 hr at RT, followed by primary antibody inculcation in 5% BSA at 4 °C overnight. After 3 times wash with TBST, the membrane was incubated with secondary antibody for 1hr at RT and then developed with chemiluminescent substrates (ThermoFisher Scientific, #32106 or #34095) and acquired images in the darkroom.

Immunofluorescence

U2OS cells seeded on the 8-well chamber slide (Thermo Scientific, #154534) were fixed by 4%PFA at RT for 10 min or cold acetone for 10 min. Then the cells were blocked by 5%BSA or 10% donkey serum in PBST for 1hr at RT. Primary Ab was added and incubated at 4 °C overnight. After 3 times wash, secondary antibody labeled with Alexa Fluor^™ 488 or 594 was incubated at 4 °C for 1hr. DAPI staining was performed before the slides were mounted by coverslip and mounting medium (SouthernBiotech, #0100–01)

Microscopy imaging

U2OS and AML12 cells were transfected with different fluorescent proteins by FuGENE^® HD (Promega, #E2311) and Lipofectamine^® 3000 (ThermoFisher Scientific, #L3000008) according to the kit protocols. Live imaging was collected under LSM710 confocal (Zeiss) after 48–72 hrs in μ-Slide 8 Well chamber (ibidi, #80826). HaloTag-fused proteins were stained by Janelia Fluor^® HaloTag^® Ligands A549 (Promega, #GA1110) for 15 mins before imaging collection. For condensates disrupting experiment, the original culture medium for U2OS was replaced by 10% 1,6-Hexanediol containing one, immediately after replacement, cells were imaged. FRAP experiments were performed according to previously described protocols¹⁹, briefly a random region containing one droplet in the nucleus was selected and bleached with high power laser, after bleaching, imaging was collected in 10 seconds interval within total 10 mins. Fluorescence intensity were quantified by ZEISS ZEN software.

RNA extraction and quantitative PCR

Total RNA was extracted by Trizol reagent (Ambion, # 5457) from frozen liver tissue and purified by RNeasy Mini Kit (Qiagen, #74106). Fresh total RNA was reversed into cDNA by High Capacity cDNA Reverse Transcription Kit according to its protocol (ThermoFisher Scientific, #4368814). Quantitative PCR were performed by gene primers (table S1) and SYBR Green PCR Master Mix (ThermoFisher Scientific #4334973) under a QuantStudio 6 Flex instrument (Applied Biosystems). Data analysis was performed by the standard curve method, gene expression was normalized to the mRNA level of the housekeeping gene Rpl19.

Adeno-associated virus 8 (AAV8) packaging and purification

REV-ERBα wild type (WT) and mutant AAV8 virus were packaged according to previously described protocol⁶⁴. Basically, two helper plasmids pAdDF6 (addgene #112867), pAAV2/8 (addgene, #112864) together with REV-ERBα WT or mutant AAV constructs were transfected into AAVpro^® 293T cells by LipoD293^™ following by the protocol (SignaGen^® Laboratories, #SL100668), 24 hrs later, medium was replaced by fresh one and 72 hrs later cells and supernatant medium were collected together following by chloroform-sodium chloride purification. AAV virus was further concentrated and purified by Amicon^® Ultra Centrifugal Filters (Millipore, UFC510024) and titer was measured by qPCR Using AAV2/8 ITR primers.

AAV8-CRISPRa liver delivery system

In order to achieve higher gene modulation efficiency in vivo, we optimized one vector CRISPRa within AAV8 backbone (AAV8-CRISPRa): briefly, we adopted nuclease dead SaCas9 (dSaCas9) to increase transgene size packaging capacity⁶⁵. TBG promoter was used to achieve the hepatocyte type specificity and the smaller synthetic polyA (sPA) was positioned at the end of transgene ORF⁶⁶. Human U6 promoter was used to drive expression of gRNA whose sequence is designed by CCTop tool⁶⁷. AAV8-CRISPRa virus was packaged, purified and titer-measured for further tail-vein injection experiment.

AAV8-shRNA knockdown in mouse liver

shRNA oligoes against mouse HSPA8 gene or scramble were synthesized and cloned into the AAV8 shRNA backbone developed in this paper. Briefly the AAV8 shRNA backbone was adapted from pAAV.TBG-PI-eGFP.WPRE.bGH plasmid (addgene # 105535) by inserting a human U6 promoter followed by AvrII/KpnI double cloning sites and replacing the GFP sequence with the 3XFlag-tdTomato cassette. The forward and reverse oligoes were mixed with T4 PNK enzyme and annealed according to the following program (37 °C, 30 min; 95 °C 5 min and then ramp down to 25 °C at 5 °C/min). Then diluted annealed oligos and ligated with linearized AAV8 shRNA backbone by AvrII/KpnI (NEB, #R0174S/R3142S). The sequence confirmed plasmid was used for AAV8 virus production according to the AAV8 packaging and purification procedure described earlier.

Frozen section preparation and IHC staining

A mouse liver harvested at different circadian time points was chopped into small pieces (~0.5×0.5 cm) and immersed in 30% sucrose/PBS at 4 °C until the liver sunk to the bottom of a test tube. Then tissue chunks were immersed by 1:1 ratio of 30% sucrose and O.C.T. (Fisher Scientific, #23–730-571) for 15 min. Finally, the tissue was completely embedded in O.C.T compound and frozen at −80 °C prior to cryostat sectioning.Frozen section sliced by cryostat was first fixed by cold acetone for 10 min at −20 °C, and permeabilized by 1% TritonX-100 at RT for 10 min. Then blocked with 10% donkey serum at RT for 1hr and incubated with primary antibody at 4 °C overnight or RT for 3 hr. Secondary antibody labeled with Alexa Fluor^™ 488 or 594 was added to the slide and incubated at RT for 1hr. Nuclear staining was performed by DAPI dye before mounting medium was added.

Mouse liver DNA FISH

DNA FISH for mouse liver slides was performed according to the empire genomics kit protocol. Briefly, the frozen liver slides were fixed by fixative buffer (3:1 methanol: acetic acid) for 5 min and permeabilizated by 0.4 mg/ml pepsin for 10 min, and by 1% TritonX-100 subsequently. Liver slides dehydrated through 70%, 85%, 100% ethanol, each for 2 min. Hybridization mix containing a DNA probe was added and denatured at 73 °C for 2 min before hybridization at 37 °C for at least 16 hr. After hybridization, the slides were washed and mounted before image collection.

Luciferase reporter assay

Dual-Luciferase Reporter Assay was performed to measure REV-ERBα-mediated transcriptional activity according to the kit protocol (Promega, #E1910): Briefly, Bmal1 promoter-driven firefly luciferase and CMV-driven renilla luciferase constructs were transfected into AML12 or 293FT cells with other plasmids indicated in the figure legend, 48 hrs later, cells were lysed and luciferase activity was measured by a plate reader. Bmal1-driven luciferase activity was normalized with corresponded renilla activity to remove transfection and cell number bias between groups. The luciferase activity was further normalized to control vector to reflect the activation fold change over control.

RNA-seq library preparation

1 μg total RNA extracted from each biological replicated liver was used for library construction: briefly ribosome RNA was first depleted by RiboZero Magnetic rRNA removal kit (Illumina, #20040526), then remained RNA was converted to DNA library for sequencing according to TruSeq Stranded Total RNA Library Prep kit protocol (Illumina, #20020599).

RNA-seq data processing

The pair-end fastq reads sequencing from RNA-seq library were mapped to transcripts derived from mm10 version genome by STAR aligner⁵⁵. Low mapping quality reads (MAPQ<30) were removed by samtools⁵³. Filtered bam files were converted to bigwig files for genome browser visualization. Total reads number of each gene were quantified with GRCm38.p6 annotation by featureCounts from filtered bam files⁶⁰ and differential expressed genes were identified by DEseq2⁵⁶.

Tn5-based ChIP-seq

Liver nuclei were isolated and fixed from liver tissue by different AAV infection or at two circadian time points (ZT22 and ZT10) according to previous described protocol with minor modification⁶⁸: Briefly, 15 ml cold swelling buffer (10 mM HEPES, 2 mM MgCl2, 3 mM CaCl3, 1x protease inhibitor cocktail) was used to dounce 0.1–0.2 g liver tissues by piston A about 10 times. The liver clumps after spinning down at 400 g in 4 °C for 10 min and washed once by DPBS. The tissue was fixed in 10 ml 1% formaldehyde in DPBS for 20 min at room temperature. 0.125 M glycine were added to quench the crosslinking reaction for 10 mins at room temperature. 15 ml cold swelling buffer was added after centrifugation to remove supernatant. Piston B was used to dounce the cells 10 times to get single nuclei. The pellets were filtered by 100 μm cell strainer and then resuspended in 10 ml cold swelling buffer containing 10% glycerol. 10 ml cold lysis buffer (swelling buffer + 10 % glycerol + 1% Igepal) was then added drop by drop and incubated on ice for 5 mins. Following incubation, an additional 30 ml cold lysis buffer was added to the sample. The supernatant was removed after a 10 minute centrifugation at 600g, 4 °C. The remaining pellets were washed once by 25 ml cold lysis buffer and 10 ml DPBS, then resuspended in 10 ml 1.5 mM EGS (ThermoFisher Scientific, #21565) in 37 °C for 45 min and quenched by 0.125 M glycine for 10 mins at room temperature. The dual-crosslinked pellets were kept in −80 °C for ChIP-seq and In-situ ChIA-PET experiment. The dual-crosslinked liver nuclei generated above were used to perform anti-HA (ThermoFisher Scientific, #88837), H3K4me3 (Millipore, #04–745), H3K27ac (Active Motif, #39133), NCoR1 (In-house) ChIP experiments. Tn5 (Illumina, #20034197) was used to fragment Immunoprecipied DNA and add adaptors simultaneously. Adapted DNA was amplified by PCR within 10–13 cycles dependents the initial DNA amount for tagmentation (Illumina, #FC-131–1024). Size selection of PCR products were performed by SPRIselect (Beckman coulter, #B23318) and subjected to pair-end sequencing on Illumina Hiseq or Novaseq platform.

ChIP-seq data analysis

The ChIP-seq data was processed by standard ChIP-seq pipeline: briefly, adaptor sequences were removed by cutadapt⁶⁹ before BWA mem⁷⁰ mediated alignment on mm10 version genome. Read duplicates and low MAPQ (<=30) were removed by picard⁷¹ and samtools⁵³, respectively. MACS2⁶¹ was used to call peaks from filtered bam files. BigWig files converted from bam files were used for genome browser visualization with a customized R script based on karyoploteR package⁶². Peak intensity was calculated by multiBigwigSummary module from deepTools⁶³.

Tissue In-situ ChIA-PET

In-situ ChIA-PET for liver tissue was optimized according to previously described protocol⁷²: briefly, 30 ul AluI enzyme (NEB, #R0137L) were used to digested genome of dual-crosslinked liver nuclei at 37 °C O/N. Exposed ends of genomic DNA were dA-tailed and ligated with dT-overhang biotinylated bridge linker (IDT, sequence is listed in the table SX) at 16 °C O/N. The ligated chromatin was sheared by sonication and pull-down by antiHA-conjugated magnetic beads or CTCF antibody (abcam, #ab70303). Immunoprecipied DNA was subjected to Tn5 (Illumina, #20034197) tagmentation, M280 Dynabeads (ThermoFisher Scientific, #11206D) streptavidin pull-down and PCR amplification (Illumina, #FC-131–1024). PCR products were subjected to size-selection and pair-end sequencing on Illumina Hiseq or Novaseq platform.

In-situ ChIA-PET data processing and visualization

ChIA-PET data was processed by optimized ChIA-PIPE tool⁴¹. Briefly, ChIA-PET paired-end reads were trimmed by bridge linker and split into 3 types of tags: none-linker containing tags, one-end tags and pair-end tags. Only Tags longer than 18 bp were aligned to mm10 genome by BWA. PET clusters were generated by merging and collapsing non-duplicated pair-end tags if both of their ends were overlapped with distance less than 1 kb. Each PET cluster has pair anchors containing chromosome name, start and end at each side. Total reads count for each PET cluster was defined as PET count to reflect the interaction frequency between these two anchor regions. Based on the distance of paired tags of each PET cluster, they were classified into interchromosomal PETs, intra-chromosomal PETs (two tags distance >=8 kb) and self-ligated PETs (two tags distance <8 kb). Intra-chromosomal PETs (PET count >=3) were reported and further analyzed. All 3 category tags were used to generate bigwig files for genome binding profiling. ChIA-PET data was visualized by customized R script.

REV-ERBα-associated Super Anchor (RESA) definition and analysis

REV-ERBα ChIA-PET data was processed to generate Intra-chromosomal PETs (PET count >=3) for downstream analysis. These REV-ERBα PET loops were further filtered by at least one of both anchors occupied by REV-ERBα binding peaks called from REV-ERBα ChIP-seq. These loops were overlapped with activated TSS sites annotated by H3K4me3 ChIP-seq at ZT22 and ZT10. Since loops linking to a promoter represent the most transcriptional function-relevant in our study, we focus on these promoter-associated REV-ERBα loops for further analyses. For each activated TSS supported by REV-ERBα ChIA-PET loops, we count the total loop anchors linking to promoter directly or recursively through middle anchor nodes. We define all anchors linking to the tss region of a specific gene as a REV-ERBα-associated Super Anchor (RESA). We quantify each RESA binding intensity by summing all REV-ERBα ChIP binding signal at these interconnected loop anchors, RESA interconnection score was summed by these interconnected loop PET count.

QUANTIFICATION AND STATISTICAL ANALYSIS

Variation in all panels is indicated by the standard error of the mean (SEM). Two-tailed unpaired Student’s t-tests were used to test the significance of differences between two groups. Statistical significance is displayed as NS (not significant, or p>=0.05), * (p<0.05), ** (p<0.01), *** (p<0.001) unless specified otherwise.

Supplementary Material

1

Movie S1. Motion of EYFP-REV-ERBα droplets. Related to Figure 1.

2

Movie S2. FRAP movie of EYFP-REV-ERBα. Related to Figure 1.

Highlights.

• Clock component REV-ERBα physiologically forms circadian condensates in vivo

• An intrinsic disorder region (IDR) of REV-ERBα drives circadian condensates

• The REV-ERBα IDR controls hub formation and rhythmic transcription repression in liver

• REV-ERBα condensates selectively partition corepressor NCOR1 for circadian repression