Transitions in chromatin conformation shaped by fatty acids and the circadian clock underlie hepatic transcriptional reorganization in obese mice

Male C57BL/6 J mice were housed under 12 h light:12 h dark (LD12:12) with food and water available ad libitum. 7–8 weeks old mice were randomly distributed in two groups, and were fed for 12 weeks with either a chow diet (CD group, 18% calories from fat; 2018S Tekland, ENVIGO) or a high-fat diet (HF group, 53% calories from fat; TD.160547 Tekland, ENVIGO). Mice were sacrificed either at ZT6 or ZT18 (n = 6 mice per time point and diet), and livers were snap frozen and kept at -80ºC for further processing. Animal care was in accordance with the Internal Committee for the Care and Use of Laboratory Animals (CICUAL) at the Instituto de Investigaciones Biomédicas, protocol number ID230 (UNAM, Mexico).

At 10 and 11 weeks of feeding paradigm, mice were subjected to a glucose tolerance and insulin tolerance tests (GTT / ITT). For the GTT, mice were fasted for 12 h, and injected intraperitoneally (IP) with D-glucose (G7021, SIGMA) at 2 mg/kg. For the ITT, insulin (HI0210, Eli Lilly) at 0.6 U/kg was IP injected 4 h post fasting. Circulating glucose was measured from tail using a glucometer (ACCU-CHEK, ROCHE) at 0, 15, 30, 60, 90 and 120 min after IP injection. Both procedures were performed at ZT1 and ZT13, with n = 6 mice per group.

Frozen livers were embedded in OCT medium (Tissue-Tek, 4583) and cut into 10-μm sections using a Leica cryostat. Tissue sections were fixed in slides with 4% paraformaldehyde for 2 min and immediately washed twice with deionized water. Oil Red O (SIGMA, O1391) working solution (3.75 mg/ml in 60% isopropanol) was applied on slides for 5 min at room temperature (RT) and washed with abundant deionized water to eliminate colorant excess. The images were documented with Axiocam EEc 5 s camera coupled to a ZEISS Primovert microscope, using a 40X magnification.

60 mg of tissue were homogenized (Benchmark Scientific, D1000 homogenizer) for 30 s with 0.75 ml of Trizol^™ (15,596,018, Invitrogen). Homogenates were incubated for 5 min at room temperature (RT), then 0.15 ml of chloroform was added, shaked and incubated at RT for 3 min followed by centrifugation for 15 min at 13,000 rpm at 4 °C. The aqueous phase was recollected, and 0.4 ml of isopropanol was added. After a 10 min incubation at RT, RNA was precipitated by centrifugation for 20 min at 13,000 rpm and 4 °C. RNA was washed twice with 0.75 ml of 75% ethanol and resuspended in 50 µl of molecular biology grade water. For cDNA synthesis, 1 μg of RNA was retrotranscribed with SuperScript^™ III RT (Invitrogen, 18,080,093) according to manufacturer’s instructions, using 100 ng of random hexamers (Invitrogen, N8080127) per sample.

RT-qPCR reactions were performed with 10 ng of cDNA using the iTaq Universal SYBR^® Green Mix (1,725,121, BioRad) in a final volume of 10 µl. A CFX96 Touch Real-Time PCR Detection System (BioRad) was used with the following program: 30 s at 95 °C, followed by 40 cycles of 5 s at 95 °C, 30 s at 60 °C. Single-product amplification was verified by an integrated post-run melting curve analysis. Values were normalized to the housekeeping genes Gapdh and B2m. The geometric mean was used to calculate Ct values of the housekeeping genes and expression values for the genes of interest were determined using ΔCT methodology. Primers are listed in Supplementary Table S1.

Baits were defined in regions with regulatory potential for each gene as follows: Dbp bait was located in the intron 2 of Dbp gene, containing two E-boxes essential for rhythmic expression of the gene as described in [10], Ppara and Pparg2 baits were located in their promoter regions, and Srebp1c bait was designed within intron 1, which shows strong regulatory potential with high H3K27ac levels in mouse liver. Two livers per dietary condition and time point were randomly selected for 4C experiments. 4C templates were prepared from 0.5 to 0.7 g of liver tissues homogenized in PBS 1X pH 7.4, filtered in 70 µm cell strainer and diluted to 30 ml with PBS. Cells were pelleted and crosslinked in 25 ml of DMEM with 1% of formaldehyde at room temperature under rotation for 15 min and the reaction was immediately quenched with glycine at 0.125 M final concentration for 5 min at 4 °C. Nuclear isolation was performed in NI buffer (10 mM Tris HCl, 10 mM NaCl, 0.2% NP-40) for 1 h at 4 °C. For first restriction, selected six-cutter enzymes were EcoRI for Pparγ2 bait, and HindIII for Dbp, Pparα and Srebp1c baits. Ten million nuclei per sample were resuspended in 500 µl of 1X NEBbuffer 2.1 (New England BioLabs) containing 0.3% SDS and incubated 10 min at 60 °C and 50 min at 37 °C. Triton X100 was added at 1.8% v/v, incubated for 1 h at 37 °C, and 400 U of restriction enzyme were added and incubated at 37 °C overnight with agitation. First ligation was performed with 100 U of T4 DNA ligase (ThermoScientific Cat. No. EL0011) with 5% Polyethylene glycol in 7 ml final volume, for 6 h at 16 °C. DNA crosslinking was reversed by adding 30 µl of 10 mg/ml proteinase K (EMD Millipore Cat. No. 124568) and incubating for 6 h at 65 °C, followed by addition of 30 µl of 10 mg/ml RNAse A (Sigma Cat. No.R5000) and incubation for 45 min at 37 °C. DNA was purified with phenol/chloroform extraction and sodium acetate/ethanol precipitation.

The purified DNA had a second restriction with 200 U of Csp6I (New England BioLabs) in 500 µl reaction volume. Finally, a second round of ligation with 200 U of T4 DNA ligase with 5% PEG in a 14 ml final volume was performed. Circular DNA was resuspended in molecular grade H2O. Efficiency of each reaction was evaluated by loading DNA in 0.8% agarose gel.

Libraries were prepared with a two-step PCR procedure. For each 4C template, eight reactions of 100–200 ng were amplified by PCR using high fidelity Hot Start Phusion II DNA polymerase (ThermoFisher, F549) and long-template specific-bait primers containing the NexteraXT overhang sequences (Supplementary Table), for 2 min at 98 °C, (15 s at 98 °C, 1 min at 48 °C, 3 min at 72 °C) × 30 cycles and a final extension of 7 min at 72 °C. PCRs were pooled to serve as template for a subsequent PCR using NexteraXT index kit. PCR products were size selected with 2% agarose gel to 200 bp-2 kb, purified with gel purification kit (Qiagen) and fragment size distribution analyzed with bioanalyzer tapestation (Agilent). S1

Libraries were sequenced with Illumina HiSeq platform, in Dbp and Pparγ2 cases using single end 100-bp read length, meanwhile Pparα and Srebp1c libraries were sequenced with paired-end 150-bp read length. Two biological samples per condition were analyzed.

Raw data quality was evaluated with FastQC. Adapters and primer sequences were trimmed using Trim Galore, and files were aligned to mm9 mouse genome with bowtie2. Close cis interactions were identified using the FourCSeq pipeline in R [23]. Briefly, Bam files were aligned to a in silico fragmented genome with the corresponding restriction enzymes (HindIII/Csp6I or EcoRI/Csp6I) and normalized counts (RPKM) for every restriction fragment were obtained. The first two fragments upstream and downstream from each bait were not considered for analysis. Count values were transformed with a variance stabilizing transformation and z-scores were calculated, considering as significant interactions those with a z-score ≥ 2.0 and an FDR ≤ 0.05 in both replicates. In the particular case of the Srebp1c bait, a z-score ≥ 1.5 and an FDR ≤ 0.05 were considered as threshold. For analysis of differential interactions between conditions, we calculated an average z-score ratio, setting as threshold a ratio ≥ 1.5, while keeping an FDR ≤ 0.05.

RPKM bedgraph files were used as input for the 4Cker program[24], with n = 2 and k = 20, using the transAnalysis function. Coordinates defining the interactions for each bait were visualized using RCircos package[25]. These analyses were performed in R.

Previously processed HiC data from mouse liver was obtained from GSE104129↗ [14]. These HiC matrix files were further processed with the Galaxy HiCExplorer version [26]. HiC.matrix files were converted to.cool files with hicConvertFormat tool and plotted with hicPlotTADs tool. Matrices were visualized together with the TAD and subTAD coordinates in BED files downloaded from GSE104129↗, and with the obtained tracks with 4C-seq signal and significant interactions. The method for calling TADs and SubTADs is described in Ref. [14], and is based on the directionality index method as detailed in Ref. [27]. A tolerance distance of 200 kbps was applied to determine overlap.

30–50 mg of liver tissues were lysed with cold RIPA buffer (1 mM DTT, 10 mM Tris–HCl, 30% glycerol, 1 mM EDTA, 1% Triton X-100, 1 mM Na3VO4, 1 mM PMSF, 15 mM sodium azide and 1:25 v/v cOmplete ROCHE). Homogenates were centrifuged at 19,000 × g for 30 min at 4 °C, and supernatants were recovered. Protein content was determined with Bradford method. 30 µg of protein per sample were separated in SDS PAGE, using a 7% acrylamide gel to allow for detection of differences in BMAL1 phosphorylated states, and transferred to PVDF membranes. Membranes were blocked with 5% w/v non-fat milk in 0.05% PBST for one hour and hybridized with the corresponding primary antibodies overnight at 4 °C. Membranes were washed three times with PBST and incubated with the corresponding secondary antibody for 1 h at RT, followed by three washes with PBST. For specific band visualization, the Immobilon Western Chemiluminiscent HRP Substrate (Millipore, WBKLS0100) was used, and signals were recorded in a Gel Logic 1500 Imaging System (KODAK). Protein bands were quantified by densitometric analysis using ImageJ software. Relative densitometric units were calculated by dividing the densitometric band value of protein of interest by the GAPDH protein for each sample, and then normalized to CD ZT6 condition unless otherwise specified in the figure caption. Primary antibodies used in this study were: PPARY (2443) and REVERBA (13,418), both at 1:1000 dilution, from Cell Signaling; CEBPA (365,318) and CEBPB (7962), 1:500, from Santa Cruz Biotechnology; BMAL1 (93,806), 1:2000, from Abcam; PER2 (Per21A), 1:1000, from Alpha Diagnostics; CRY1 (A302-614A), 1:2000, from Bethyl laboratories; GAPDH (GT-239), 1:20,000, from GeneTex; TUBULIN (T5168), 1:8000, from Sigma. Secondary antibodies were from Sigma, anti-rabbit IgG (A0545, 1:10,000 for PPARY, 1:80,000 for CRY1, REVERBA and PER2, 1:150,000 for BMAL1) and anti-mouse (A09044, 1:20,000 for CEBPA and CEBPB, 1:80,000 for TUBULIN).

Enhancer candidates were considered as regions colocalizing with the interactions detected by 4C-seq +–5 kb with enrichment in genomic features associated with active enhancers, as H3K27ac, chromatin accessibility measured with ATAC and FAIRE-seq (see data availability section). Also, as a plus feature, some of these regions colocalized with annotated regulatory regions by ENCODE [28]. Genomic coordinates of candidates are indicated in Supplementary Table S3.

CistromeDB Toolkit [29] was used to search for TFs defined to specific coordinates in the mm10 version of the mouse genome. ChIP-seq data were filtered for liver and adipose tissue TFs in defined intervals according to genomic coordinates of enhancer sequences. Over-represented TFs were detected as a measure of overlapped peaks in the regions of interest.

The enhancer sequences were compared to the human genome (hg38) using NCBI nucleotide blast tool [30], searching for very similar regions. Percentage of conservation and genomic coordinates in human were obtained.

Enhancer sequences were cloned in pGL4.23 plasmid (Promega) upstream of the minimal promoter using KpnI (ThermoFisher Scientific, ER0501) and HindIII (New England Biolabs, R0104S) restriction enzymes. Briefly, 10 µg of vector were digested with 30 U of each enzyme in 70 µl of reaction mix in NEBuffer TM 2.1 1X (New England Biolabs, B27202S) for 6 h at 37 °C. Reaction was inactivated at 80 °C for 10 min. Digested vector was recovered by purification on a 0.7% agarose gel using QIAquick Gel Extraction kit (QIAGEN, 28,704). Enhancer sequences were amplified from mouse genomic DNA using specific primers flanked by the restriction enzymes recognition sequences (Supplementary Table 1). Each insert was amplified with the high fidelity Phusion Hot Start II DNA Polymerase (ThermoFisher Scientific, F549S). DNA was recovered by purification on a 1.0% agarose gel using QIAquick Gel Extraction kit (QIAGEN, 28,704). Inserts were digested with 10 U of each enzyme in 50 µl of reaction mix containing NEBuffer TM 2.1 1X (NEB, B27202S) for 4 h at 37 °C and inactivation at 80 °C for 10 min. DNA was cleaned with MinElute Reaction Cleanup Kit (QIAGEN, 28,204) and eluted in molecular grade water. Ligations were performed using 50 ng of digested vector in a 1:10 vector:insert molar proportion, with 5U of T4 DNA ligase (ThermoFisher Scientific, EL001) in 10 µl of volume reaction at 16 °C overnight and inactivation of ligase at 65 °C for 10 min. Constructions were cloned in TOP10 E coli, and DNA was purified with E.Z.N.A Plasmid Maxi Kit (Omega BIO-TEK, D6922). Successful cloning was verified by Sanger sequencing.

AML12 cell line (ATCC, CRL-2254) was cultured in DMEM-F12 medium (Gibco, 12,400–029) supplemented with 10% fetal bovine serum (FBS), ITS 0.5X (5 µg/ml insulin, 2.75 µg/ml transferrin, 2.5 ng/ml selenium; Gibco, 41,400,045), Penicillin/Streptomycin 1X (Biowest, MS00P01014) and 40 ng/ml of dexamethasone (Sigma, D4902). For treatment with a mixture of palmitate/oleate, 12,500 cells were seeded in 24 well plates in the complete medium with 5% FBS. After 48 h, cells were treated with 0.25 mM of palmitate (Sigma, P0500)/oleate (Sigma, O1008) mixture in a 1:2 proportion for 48 h prior to transfection. Cells treated with 1% of free fatty acid albumin (FFA-BSA) were used as control.

Palmitate/Oleate fatty acids were solubilized using alkaline NaOH 0.1 N at 80 °C to generate the corresponding salt and complexing with FFA-BSA in a 6:1 proportion FA/FFA-BSA for 1 h at 37 °C in 0.9% NaCl solution with constant agitation at 1000 rpm. ORO staining of cells was performed following our previously described protocols [31].

AML12 cells seeded in 24-well plates were transfected with 200 ng of enhancers fused to the luciferase vector, 200 ng of LacZ reporter, 100 ng myc–CLOCK–pCDNA3, 200 ng myc–BMAL1–pCDNA3, which have been described previously[32]. pCDNA3 vector was used to adjust DNA quantity. DNA was transfected using BioT transfection reagent (Bioland Scientific LCC, B01-00) in a proportion 1.5:1 BioT:DNA, except for Fig. 5F, where Lipofectemine 2000 (Thermo Fisher) was used to a 2:1 ratio. The total amount of applied DNA per well was adjusted by adding pcDNA3 vector. Cells were allowed to recover for 24 h more and luciferase assay was performed. Lipid treatment was maintained during transfection and until the end of the protocol.

Cells were washed with 700 µl of PBS 1X and lysed in 100 µl of lysis buffer (25 mM Tris–phosphate pH 7.8, 2 mM EDTA, 1 mM DTT, 10% glycerol, 1% TritonX-100) for 20 min at room temperature with vigorous orbital agitation, followed by subsequent incubation at -80 °C for 40 min. Supernatants were recovered for luciferase assay. 20 µl of lysate were added to a 96-well luciferase plate, and 100 µl of luciferase reaction buffer (20 mM Tris–phosphate pH 7.8, 1.07 mM MgCl₂, 2.7 mM MgSO₄, 0.1 mM EDTA, 33.3 mM DTT, 470 µM luciferin, 530 µM ATP, 270 µM Coenzyme-A) were subsequently added to the lysates. Luminescence was measured using a Synergy H1 (BioTek) multi-mode reader, with 1 s of integration time.

For β-galactosidase assay, 40 µl of lysate were assessed in a 96 well plate containing 120 µl of tampon Z (60 mM NaHPO₄, 40 mM NaH₂PO₄-H₂O, 10 mM KCl, 1 mM MgSO₄, 0.7 mg/ml ONPG and 3.50 µl/ml β-mercaptoethanol). The plate was incubated 37 °C until the yellow color appeared, and the reaction was stopped with 50 µl of 1 M Na₂CO₃. β-galactosidase activity was measured at 450 nm using a Synergy H1 (BioTek) multi-mode reader. Transfection efficiency was normalized by dividing luciferase signal by the β-galactosidase signal.

Plasmids expressing BMAL1 and CLOCK have been previously described[32] and were a kind gift from Paolo Sassone-Corsi. Plasmids expressing REVERBβ CEBPβ and A-CEBPβ were purchased from Addgene, Cat. No. 22745 gifted by Bruce Spiegelman and described in [33], Cat. No. 49198 gifted by Jed Friedman and Cat. No. 33363 gifted by Charles Vinson respectively.

ChIP experiments were performed as previously described [34]. Briefly, 200 mg of liver tissue were homogenized in PBS. Dual crosslinking was performed in a final volume of 1 ml using 2 mM of DSG (Disuccinimidyl glutarate, ProteoChem, CAS: 79,642–50-5) for 10 min at RT on a rotary shaker. DSG was washed out and 1% formaldehyde (Sigma-Aldrich, F8775) in PBS was added an incubated for 15 min at RT with rotataion. Crosslinking was topped with 0.125 M glycine for 5 min on ice. Two washes with ice-cold PBS were performed, and nuclei were isolated in 600 μL of ice-cold nuclei preparation buffer (NPB: 10 mM HEPES, 10 mM KCl,1.5 mM MgCl2, 250 mM sucrose, 0.1% IGEPAL CA-630) at 4 °C for 5 min in rotation. Nuclei were collected by centrifugation at 1,500 g for 12 min at 4 °C and, and resuspended in 600 μL of cold nuclear lysis buffer (10 mM Tris pH 8, 1 mM EDTA, 0.5mMEGTA, 0.3% SDS, 1 × cOmplete™ Protease Inhibitor Cocktail, Roche) for 30 min on ice. 300 μL of lysates were sonicated using a Bioruptor Pico Sonicator (Diagenode) for 15 cycles (30 s ON/30 s OFF). Chromatin fragments (100–500 bp) were evaluated on agarose gels using 10 μL of sonicated chromatin for DNA purification using the phenol method. 600 μL of ice-cold ChIP-dilution buffer (1% Triton X-100, 2 mM EDTA, 20 mM Tris pH 8, 150 mM NaCl, 1 mM PMSF, 1 × cOmplete™ Protease Inhibitor Cocktail, Roche) was added to the chromatin, and 10% volume was recovered as the Input. Immunoprecipitations were set overnight at 4 °C, by adding 20 μL of magnetic beads (Cat. No. #16–662, Sigma-Aldrich) and the antibodies: 8 μl Rev-Erbα (E1Y6D) Rabbit mAb (Cell Signaling Cat. No. #13,418) or 3 μl C/EBP beta (H-7) mAb (Santa Cruz Cat. No. sc-7962), in 900 μL final volume. Immunoprecipitations with 3 μL of IgG (Sigma-Aldrich, Cat. No. 18765) were done simultaneously. Sequential washes of the magnetic beads were performed for 10 min at 4 °C, as follows: Wash buffer 1 (20 mM Tris pH 8, 0.1% SDS, 1% Triton X-100, 150 mM NaCl, 2 mM EDTA),Wash buffer 2 (20mMTris pH 8, 0.1% SDS, 1% Triton X-100, 500 mM NaCl, 2 mM EDTA), Wash buffer 3 (10 mM Tris pH 8, 250 mM LiCl, 1% IGEPAL CA-630, 1% sodium deoxycolate) and TE buffer (10 mM Tris pH 8, 1 mM EDTA). To isolate immunoprecipitated DNA, 400 μL of fresh elution buffer (10 mM Tris pH 8, 0.5% SDS, 300 mM NaCl, 5 mM EDTA, 0.05 mg/mL proteinase K) was added to the magnetic beads followed by incubation overnight at 65 °C. A treatment with RNase A at 0.1 mg/ml for 30 min at 37 °C was performed. The DNA was purified from the IPs and Inputs by the phenol-based method, and DNA was precipitated with ethanol in the presence of 20 μl glycogen (10,901,393,001, Roche) at − 80 °C overnight, and resuspended in 40 μl of MQ water. qRT-PCR reactions were set using primers described in Supplementary Table S1.

Normalized ICED HiC matrices corresponding to ZT10 in the WT and liver-specific Rev-erbα KO conditions were obtained from Kim et al., 2018 [14] (GSE104129↗). The HiCExplorer [35] function hicConvertFormat was used to convert the matrices to cool format. Subsequently, the Log₂ ratio between the WT and KO contact matrices was calculated and plots were generated using the hicCompareMatrices function. Increased interactions were defined as described in REF. [14].

Gro-seq data were obtained from GSM1437746↗ which were analyzed as described [9]. Briefly, uniquely mapped reads were extended to 150 bp in the 5’ to 3’ direction, transcription tag counts were normalized by Reads Per Kilobase of transcript per Million (RPKM) and Bigwig files were generated using HOMER v4.3 [36] and visualized in the WashU epigenome browser [37] in the strand where the gene is transcribed. Gene body transcription level was calculated and plotted by counting reads beginning at the TSS. Microarray expression data form livers from WT and RevErbα liver-specific KO mice were obtained from GSE59460↗ and analyzed using GEO2R, which uses limma to perform differential expression analysis and to calculate p value [38].

Analyses were performed using GraphPad Prism V8. Data are presented as the mean +–standard error, using one-way or two-way analysis of variance (ANOVA) followed by Tukey´s posttest, unless otherwise indicated. When two-way ANOVA was used, the independence between the factors (e.g., time and diet, or plasmid and treatment) was tested and reported using the “interaction” term. Differences between groups were considered as statistically significant with a P < 0.05.

We implemented a mouse model of diet-induced obesity by feeding mice with either a high fat diet (60% Kcal from fat, HF group) or a chow diet (CD group) for 12 weeks (See methods section). Significantly increased body weight was observed in the HF group compared to controls (Figure S1A, P < 0.001, Two-way ANOVA followed by Tukey posttest). As expected, obese mice showed impaired glucose tolerance (Figure S1B, P < 0.001, One-way ANOVA followed by Tukey posttest) and displayed poor ability to lower circulating glucose levels upon insulin injection (Figure S1C; P < 0.001, One-way ANOVA followed by Tukey posttest), evidencing abnormal high glucose levels and systemic insulin resistance associated to HF-diet feeding. This effect was more pronounced at the onset of the rest phase (ZT1; ZT is Zeitgeber time, where light is on at ZT0 and off at ZT12), coincident with previous reports showing that insulin sensitivity follows a circadian rhythm [39]. Because hepatic lipid deposition is a major feature of insulin resistance in the liver, we investigated hepatic steatosis at the macroscopic level, showing that HF-fed mice developed hepatomegaly and a yellow-orange liver coloration (Figure S1D). Subsequent ORO staining confirmed hepatic lipid accumulation in HF-fed mice (Figure S1D). Then, we analyzed expression of circadian genes at two representative time points, ZT6 (day-time) and ZT18 (night-time). As expected, the expression of Bmal1, Reverbα, Cry1, and Per2 genes showed significant fluctuations between ZT6 and ZT18 (Figure S1E; P < 0.01, Two-way ANOVA with Bonferroni post-test). At the protein level, REVERBα and PER2 also fluctuated between ZT6 and ZT18. Although these two time points do not capture CRY1 variations in the livers of CD-fed mice, CRY1 was significantly increased at ZT18 in the livers of obese mice due to its advanced phase under these conditions [20, 34, 40, 41]. A similar pattern was observed for BMAL1 phosphorylation (Figure S1F). Together, these data demonstrate a functional circadian machinery in the liver of these cohort of mice and indicate that our mice fed a HF diet developed a metabolic phenotype with obesity, insulin resistance and hepatic steatosis, commonly associated to type 2 diabetes.

CLOCK:BMAL1 are pioneering transcription factors for rhythmic genes, and their disfunction (i.e., through genetic or pharmacologic interventions) has been associated with metabolic disorders, including fatty liver disease [42]. Bmal1 deficiency reduces hepatic insulin response, opposing AKT phosphorylation and decreasing de novo lipogenesis [43]. During high-fat feeding, hepatic Bmal1 deficiency results in marked insulin resistance and liver steatosis [44]. Because BMAL1 participates in delineating chromatin loops for rhythmic transcription, it is possible that rhythmic interactions are altered in the liver of obese mice promoting, at least to some extent, hepatic metabolic derangements. Hereby, we applied Chromatin Conformation Capture followed by sequencing (4C-seq) to define long range chromatin interactions in livers from control diet and HF-fed mice. The 4C-seq technique was designed to detect interactions between a selected locus, also known as “bait” or “viewpoint”, and other genomic regions, thus providing highly specific and spatially resolved contact profiles around the bait [45–47]. As bait, we first selected the Dbp gene, based on our previous studies describing that this locus engages in rhythmic chromatin interactions during the circadian cycle which depend on BMAL1 in mouse embryonic fibroblasts [10]. Additionally, we sought to explore the long-range interactions from viewpoints located within genes participating in the metabolic phenotype activated by diet, including Pparγ, Pparα and Srebp1c. Specifically, the reorganization in circadian gene expression in mouse fatty liver is largely directed by the coordinated action between the transcription factors PPARγ, PPARα and SREBP, through the sequential activation of their targets along the day [21, 40]. Hereby these genes constitute important objectives to investigate, as their regulation at the chromatin level in response to lipids might underlie a large part of the transcriptional reprogramming.

In mouse liver, two main phases define alternate chromatin recruitment of circadian transcriptional activators and repressors, centered around the circadian times ZT6 and ZT18 respectively [7, 52]. Hence, rhythmic positioning of clock proteins underlies chromatin transitions at the regulatory elements of clock-controlled genes [53]. Hereby, these two time points were selected to profile chromatin interactions. Differences in contact frequencies between distinct conditions were calculated using DESeq2 methodology [23, 54], calling differential contacts when Padj. < 0.05 (Wald test). As expected, for each bait we found architectural variations depending on the ZT [10, 14, 15] illustrated for example by increased interactions at ZT6 vs ZT18 for the Dbp and Pparγ2 baits (Fig. 1A–D). Interestingly, for all selected baits, rhythmic chromatin interactions were extensively reshaped in HF fed mice when compared with the control mice. This reorganization was also apparent for the interactions in trans (Figure S2A–D), some of which were previously described to be rhythmic for the Dbp promoter in mouse embryonic fibroblasts [10].

Hereby, we found a specific rhythmic interactome for each nutritional condition, with a diet-associated reorganization of long-range interactions whose extent appeared markedly gene-specific [6, 22]. We then turned our focus to understand the function of differential chromatin landscapes depending on time-of-day and diet.

Extensive transcriptional alterations are triggered by high-fat diet, including reprogramming of rhythmic transcription. Hereby, we sought to study the network of long-range interactions involving regulatory regions of the genes Pparγ, Pparα and Srebp1c, largely responsible for such transcriptional responses [21, 40]. In this context, we reasoned that the remodeling of genomic regulatory elements governing the transcriptional activation of these key genes during high-fat feeding might initiate the transcriptional responses to high fat feeding.

Pparγ gene encodes two isoforms, Pparγ1 and Pparγ2, using different promoters and alternative splicing [62]. Under high fat feeding, PPARγ2 protein is overexpressed in hepatocytes promoting the adipogenic program to store fatty acids in lipid droplets [62–64] (Fig. 2E). This is accompanied by Pparγ2 overexpression at the mRNA and pre-mRNA levels, evidencing transcriptional control (Fig. 2F). Consistently, Pparγ2 promoter engaged in multiple long-range interactions in the liver of high-fat diet fed mice, both at ZT6 and ZT18 (Figs. 2G, 2H). Most of these (6 out of 8) were conserved between both ZTs, which appeared within gene bodies including Tsen2 (TRNA Splicing Endonuclease Subunit 2), Timp4 (Tissue inhibitor of metalloproteinases 4) and Gm17482. We explored whether the diet-induced chromatin contacts influenced transcription of these genes, and found that overall, their expression remained unchanged independently of ZT or diet (Fig. 2I).

Collectively, these findings sustain the notion that a high-fat diet promotes precise reconfigurations of chromatin contacts within the regulatory elements essential for metabolic adaptations to dietary changes. Notably, specific contacts exhibit dependency on the time of day, and appear concurrent with fluctuations in the transcriptional activity of these pivotal genes, highlighting the interplay between chromatin architecture and the temporal regulation of gene expression in response to a high-fat diet.

In addition, we observed specific distal interactions with the Pparγ bait which appeared in response to high fat feeding and were situated within a genomic region exhibiting a notable increase in the H3K27ac mark in the livers of obese mice. These identified regulatory elements were designated as Pparg_RE1 and Pparg_RE2 (Fig. 4B). A parallel pattern emerged for the genomic locus housing the Pparα gene, revealing two contacts enriched with H3K27ac. These included a distal upstream region labeled as Ppara_RE1 and a contact within the Pparα intron 1 designated as Ppara_I2 (Fig. 4C). Lastly, a similar distinctive interaction was selected upstream of the Srebp1c gene, denoted as Srebp1_RE (Fig. 4D). This pattern of specific distal interactions, coupled with the distinctive H3K27ac enrichment, underscores potential regulatory roles for these elements in the context of high-fat feeding-induced genomic responses.

To validate the potential responsiveness of the selected genomic regions to specific transcriptional regulators in the context of high-fat feeding, we conducted an unbiased analysis utilizing the Cistrome Database toolkit [29]. This analysis involved a comprehensive search for potential transcriptional regulators enriched in these regions, leveraging experimental ChIP-seq data from mouse liver and adipose tissues [68, 69]. Our analyses revealed that these regulatory elements exhibit the capacity to recruit a diverse array of transcription factors associated with the clock machinery, such as BMAL1, CLOCK, REVERBα, as well as hepatic and lipid-responsive transcription factors including CEBPβ, HNF4A (hepatocyte nuclear factor 4 alpha), GR (glucocorticoid receptor), RXRA/G (retinoid X receptor), PPARγ2, and PPARα. Additionally, our results implicated epigenetic regulators, such as KTM2B (lysine methyltransferase 2B), ARID1A (AT-rich interactive domain-containing protein 1A), SMARCA4 (SWI/SNF related, matrix associated, actin dependent regulator of chromatin, subfamily A, member 4), or HDAC3. Notably, specific for the Ppara_RE locus, the recruitment of proteins involved in supporting genome architecture, including CTCF (CCCTC-Binding Factor), RAD21 (RAD21 cohesin complex component), MED1, or BRD4 (bromodomain-containing protein 4), was evident (Figure S4A–D). When visualizing ChIP-seq data from mouse liver for the circadian components, it was evident that the selected loci could recruit the clock machinery (Fig. 4E–G). Interestingly, the contacts Fgf21_RE, Pparg_RE2, Ppara_I2 and SREBP1_RE were bound by both transcriptional activators of the CEBP family, and the co-repressor complex formed by NCoR and HDAC3 (Fig. 4E–G). Notably, NCoR and HDAC3 are recruited to specific loci by REVERBα to regulate rhythmic hepatic lipid metabolism [70, 71], while CEBP proteins are well-known drivers of adipogenesis and have been extensively implicated in the control of hepatic lipid homeostasis [72, 73]. Together, these data suggest that under high fat feeding, new long-range interactions appear in concordance with transcriptional responses, which might be mediated by components of the circadian machinery and specific lipid-responsive transcription factors.

These data unveil a role for CEBPβ in delineating the three-dimensional chromatin landscape in the liver of obese mice, sustaining a transcriptional program oriented to adjust to dietary conditions in collaboration with the clock machinery.

Below is the link to the electronic supplementary material. Supplementary file1 Figure S1. Physiological characterization of mice fed chow diet or high fat diet. A) Weekly weight from C57BL/6J mice fed a normocaloric (CD) or hypercaloric (HF) diet. (n= 6 per group). Two-way ANOVA and post test of Tukey * p≤0.05, ** p≤0.01, *** p≤0.001. B, C) Glucose (B, GTT) and insuline (C, ITT) tolerance tests at ZT1 and ZT13 after 10 weeks of feeeding with CD or HF diet, and their respective area under the curve (AUC) (n= 6 mice per group). One-way ANOVA and post test of Tukey * p≤0.05, ** p≤0.01, *** p≤0.001, **** p≤0.0001. AUC (area under the curve) is plotted as arbitrary units, and CD data was set to 1. D) Representative images of macroscopy visualization of livers (left), and their lipid staining with oil-red-o (ORO, right) from control (CD) and obese (HF) mice. E) RT-qPCR of from mouse liver fed either a control (CD) or a high fat (HF) diet, at two circadian times (ZT6 and ZT18) for the indicated transcripts of the clock machinery. (means ± s.e.m. of 5-6 biological replicates). F) Protein expression assessed by western blot from the indicated clock proteins in mouse liver. Quantification data is shown on the right for the three biological replicates. Data is presented as mean ± s.e.m, For BMAL1, the ratio between phosphorylated BMAL (upper signal) to total BMAL is plotted. For REVERBα and CRY1, CD at ZT6 was set to 1. For PER2, CD at ZT18 was set to 1. For histograms, Two-way ANOVA was applied for statistical analyses. When interaction was positive, Bonferroni was used as post-test. *p˂ 0.05, **p˂0.01; ***p˂0.001; ns, non-significative. Figure S2. Interchromosomal contacts engaged by the baits in the mouse liver. A-D) Circos plots representing the genome-wide view of detected inter-chromosomal interactions engaged by regulatory elements for the genes Dbp (A), Pparγ2 (B), Pparα (C), and Srebp1c (D) in livers form mice fed a chow (CD) or high-fat (HF) diets, at two circadian times, ZT6 (day) and ZT18 (night). Figure S3. Chromatin loops induced by high fat feeding do not necessarily modify expression of all their contained genes. A) RT-qPCR of from mouse liver fed either a control (CD) or a high fat (HF) diet, at two circadian times (ZT6 and ZT18) for the indicated transcripts. (means ± s.e.m. of 5-6 biological replicates). Two-way ANOVA was applied for statistical analyses. n.s.: Non significant. Figure S4. Hepatic chromatin loops induced by high fat feeding exhibit a distinctive cistrome. A-D) Cistrome toolkit analyses showing enrichment of binding sites for the transcription factors shown in the x axes, for the indicated regulatory elements for the genes Dbp (A), Pparγ (B), Pparα (C) and Srebp1c (C), uncovered in mouse liver after high-fat feeding. Figure S5. Activity of enhancers detected after high-fat feeding is boosted by CLOCK:BMAL1. A-D) Luciferase-based enhancer activity assays with the identified enhancer regions in mouse liver after high fat-feeding. AML12 cells were transiently transfected with plasmids expressing CLOCK/BMAL (C/B) in the presence or absence of lipids (P/O). Light units were normalized to an internal LacZ control, and the relative light units (RLU) from basal expression was set to 1 (means ± s.e.m. of 4 replicates). A non-regulatory region (NR) is shown as a negative control, while the previously described reporter plasmid Dbp-Luc was used as a control of the experimental set-up. E) ChIP experiments followed by qPCR were performed in livers from CD and HF diets fed mice at ZT10, corresponding to the time when REVERBα is highly recruited to chromatin, using the indicated antibodies. n=3 biological and 2 technical replicates. (means ± s.e.m.; * p≤0.05, ** p≤0.01, *** p≤0.001; Two-way ANOVA with Bonferroni´s post-test) (PDF 8657 KB) Supplementary file2 (PDF 129 KB) Supplementary file3 Sequencing reads per sample and mapping results. (PDF 77 KB) Supplementary file4 Genomic coordinates of contacts (enhancers) studied from the 4C-seq data. (PDF 61 KB)