Angiopoietin-like 4 governs diurnal lipoprotein lipase activity in brown adipose tissue

Objective Brown adipose tissue (BAT) burns fatty acids (FAs) to produce heat, and shows diurnal oscillation in glucose and triglyceride (TG)-derived FA-uptake, peaking around wakening. Here we aimed to gain insight in the diurnal regulation of metabolic BAT activity. Methods RNA-sequencing, chromatin immunoprecipitation (ChIP)-sequencing, and lipidomics analyses were performed on BAT samples of wild type C57BL/6J mice collected at 3-hour intervals throughout the day. Knockout and overexpression models were used to study causal relationships in diurnal lipid handling by BAT. Results We identified pronounced enrichment of oscillating genes involved in extracellular lipolysis in BAT, accompanied by oscillations of FA and monoacylglycerol content. This coincided with peak lipoprotein lipase (Lpl) expression, and was predicted to be driven by peroxisome proliferator-activated receptor gamma (PPARγ) activity. ChIP-sequencing for PPARγ confirmed oscillation in binding of PPARγ to Lpl. Of the known LPL-modulators, angiopoietin-like 4 (Angptl4) showed the largest diurnal amplitude opposite to Lpl, and both Angptl4 knockout and overexpression attenuated oscillations of LPL activity and TG-derived FA-uptake by BAT. Conclusions Our findings highlight involvement of PPARγ and a crucial role of ANGPTL4 in mediating the diurnal oscillation of TG-derived FA-uptake by BAT, and imply that time of day is essential when targeting LPL activity in BAT to improve metabolic health.


INTRODUCTION
White adipose tissue (WAT) is the most abundant type of adipose tissue that humans and other mammals have, which primarily stores energy from food. In contrast, the main function of brown adipose tissue (BAT) is to convert energy into heat, a process known as nonshivering thermogenesis. These two types are intertwined and highly related to one another as for example white adipocytes are capable of transdifferentiating into so-called beige adipocytes with brown-like characteristics. Brown and beige adipocytes produce heat via expression of uncoupling protein 1 (UCP1), which dissipates the mitochondrial proton motive force generated by fatty acid (FA) oxidation thereby releasing energy as heat [1]. Cold exposure is the main physiological activator of non-shivering thermogenesis and triggers the breakdown of intracellularly stored triglycerides (TGs) into FAs. To replenish intracellular lipid stores, BAT takes up glucose and FAs from circulating TG-rich lipoproteins (TRLs) [2]. Consequently, the uptake of nutrients by BAT can be used as a proxy for the presence and metabolic activity of this tissue [3e6]. Importantly, because BAT has the capability of clearing large amounts of glucose and TGderived FAs from the circulation, many favorable metabolic effects have been attributed to its thermogenic activity in mice, including the protection from atherosclerosis development [7]. Likewise, the presence of active BAT in humans has been associated with cardiometabolic health [8], highlighting its potential as therapeutic target.
We and others have shown that nutrient uptake by BAT is characterized by strong diurnal oscillations, with peak uptake of glucose and TGderived FAs around wakening [9e15]. This diurnal pattern is most likely related to the daily need for non-shivering thermogenesis to maintain body temperature and/or facilitate the rise in body temperature prior to wakening. We hypothesized that insight in the mechanism driving oscillations of BAT activity may lead to the identification or optimization of strategies to promote thermogenesis, and may provide an explanation for the increased incidence of cardiometabolic diseases among shift workers [16]. Thus far, studies in rodents have revealed that, for example, the core-clock protein REV-ERBA suppresses Ucp1 expression during the wakeful (active) phase [17]. Interestingly, cold exposure appears to add an extra dimension to the diurnal regulation of BAT as it suppresses Nr1d1 expression (encoding REV-ERBA) to relieve the inhibition on Ucp1 expression [17]. Similarly, Adlanmerini et al. [18] recently revealed that cold exposure introduces oscillation of genes involved in de novo lipogenesis with peak expression near the end of the feeding (active) phase of mice, which probably serves to provide BAT with extra fuel to maintain body temperature during the resting (inactive) phase and/or for combustion towards the start of the subsequent wakeful phase.
Here we aimed to gain further insight into the diurnal regulation of metabolic BAT activity. To this end we used an unbiased RNA sequencing and lipidomics approach in BAT of mice, and identified angiopoietin-like (Angptl)4 as critical mediator in the diurnal regulation of lipoprotein lipase (LPL) activity, thereby governing the diurnal regulation of TG-derived FA-uptake by BAT.

RESULTS & DISCUSSION
2.1. The transcriptome of murine brown adipose tissue consists of four clusters with distinct oscillating expression phases RNA-sequencing was performed on interscapular BAT (iBAT) samples collected at 3-hour intervals throughout a 24-hour period from chowfed male C57BL/6J mice exposed to mild cold (22 C, which is approximately 8 C below the thermoneutral zone [19]). All mice were entrained to a 12h:12h light:dark cycle, and therefore time is denoted as Zeitgeber Time (ZT) where ZT0 indicates the onset of the light (inactive) phase. Oscillation of transcripts was assessed by JTK cycle analysis, which provided the phase, amplitude, and significance values for each transcript and showed that in total, 5,486 genes (40.5% of all expressed genes; Figure 1A) were oscillating (P < 0.05) (full details are provided in supplementary file: 'RNA-sequencing'). Within this selection of oscillating genes, hierarchical clustering of standardized residuals identified four gene clusters with distinct expression phases ( Figure 1A). Additional horizontal clustering showed that samples collected at the same or sequential time points clustered together, highlighting their close similarities (Suppl Fig S1). On the four identified gene clusters, we carried out gene ontology and transcription factor enrichment analyses. Cluster 1 (consisting of 1,314 genes) showed peak expression around ZT18 ( Figure 1B) and was enriched in genes involved in posttranscriptional RNA processing (top hit GO:0006396; P ¼ 7.77 $ 10 À29 ) ( Figure 1B). These pathways are under direct control of the cellular clock machinery and contribute to the circadian regulation of gene expression in many tissues [20]. Cluster 2 (consisting of 1,499 genes) peaked around the onset of the light phase at ZT0 ( Figure 1B), and was enriched in genes involved in metabolic and biosynthetic processes, with the top hit being cilium assembly (GO:0060271; P ¼ 4.56 $ 10 À7 ) ( Figure 1B). This is of interest because primary cilia have a pivotal role in signaling pathways, including the transduction of signals that promote adipogenesis. In addition, many G-protein-coupled receptors (GPR) have been found to be selectively targeted by cilia. Whether this is true for the GPRs implicated in BAT functioning such as the b-adrenergic receptors [21,22], GPR120 [23] and GPR3 [24] is unknown, but represents an interesting topic for further studies. Transcription factor enrichment analysis identified general transcription factor II-I (GTF2I) as a potential mediator of the gene expression in clusters 1 and 2 ( Figure 1B). GTF2I is involved in various aspects of general cell physiology and is known to interact with the circadian locomotor output cycles kaput (CLOCK) [25]. Within cluster 2, also oscillating genes involved in de novo lipogenesis were identified (Suppl Fig S2) in line with recent findings [18], including ATP citrate lyase (Acly) and acetyl-CoA carboxylase (Acaca). Glucokinase (Gck), another gene involved in de novo lipogenesis, displayed oscillation with a slightly earlier peak expression at ZT18 corresponding to cluster 1.
Despite the sub-thermoneutral housing temperature of 22 C, no oscillations were identified for FA synthase (Fasn), stearoyl-CoA desaturase-1 (Scd1), and elongation of very long chain FAs protein (Elovl)3 (Suppl Fig S2). This indicates that the dependency on de novo lipogenesis may be determined by the degree of cold exposure, as severe cold stress induces oscillations of more de novo lipogenic genes [18] than we observed at mild cold stress. Cluster 3 (consisting of 1,660 genes) had a peak expression around ZT8 ( Figure 1B), and was enriched in genes involved in among others cellular organization, adhesion, and localization, with extracellular matrix organization as top hit (GO:0030198; P ¼ 7.92 $ 10 À7 ; Figure 1B). As yet, little is known about the interaction between circadian clocks and the cellular microenvironment in BAT or other tissues, which thus warrants further exploration. Lastly, cluster 4 (consisting of 1013 genes) was identified with peak expression around ZT14 ( Figure 1B), coinciding with the previously reported peak in metabolic BAT activity as defined by TG-derived FAuptake [13]. Correspondingly, this cluster showed enrichment in catabolic processes, with the top hit being lipid catabolic process (GO:0016042; P ¼ 2.31$10 À6 ) ( Figure 1B). Transcription factor enrichment analysis revealed peroxisome proliferator-activated receptor gamma (PPARg) as the top hit for both cluster 3 and 4. PPARg senses the energy status of the cell to regulate lipid uptake and storage, and in fact has often been described as an important circadian transcription factor in adipose tissue [26,27]. Its expression is regulated by CCAAT/enhancer-binding protein alpha (CEBPA), which we identified as first hit following PPARg in cluster 4 ( Figure 1B), and its activity is enhanced upon interaction with the circadian protein nocturnin (NOCT) [26,28]. In our dataset, Cebpa expression by itself was found oscillating within cluster 3, while expression of Pparg itself and Noct were not oscillating (data not shown), likely because PPARg is broadly regulated at the posttranscriptional level [26]. Strikingly, transcription factor enrichment analysis on the top 100 genes with the largest absolute oscillation amplitude among all oscillating genes also identified PPARg as the top hit (Suppl Fig S3). CEBPA, and CCAAT/enhancer-binding protein beta (CEBPB), both regulators of PPARg, also showed up in the top 10 of this transcription factor enrichment analysis, as well as sterol regulatory element binding transcription factor 1 (SREBF1) (also known as SREBP1), which is required for de novo lipogenesis [29] and is regulated by PPARg [30]. Collectively, these data predict a central role for PPARg in driving transcriptional oscillation within BAT, likely mediated by post-transcriptional regulation.
To obtain further insight in the processes that take place during the time that BAT is metabolically most active (i.e., around ZT12), relative amplitudes of individual genes were compared within clusters 3 and 4 (Suppl Table S1). We identified nuclear receptor subfamily 1 group D member 1 (Nr1d1) and D-box binding protein (Dbp), both involved in the core clock machinery, as the genes with largest relative amplitude in cluster 3 and 4, respectively. The genes that follow in cluster 3 are involved in a broad variety of cellular processes including cytokine binding and amino acid metabolism, while the three genes that follow in cluster 4 are all involved in the cellular core clock machinery.
Although of interest, these data do not explain the diurnal metabolic activity in BAT, and we therefore proceeded with comparing absolute amplitudes of individual genes (Suppl Table S1). NADH-ubiquinone oxidoreductase chain 1 (mt-nd1), a subunit of NADH dehydrogenase critical for the electron transport chain, was identified as the gene with the highest diurnal amplitude within the third cluster, with estimated peak expression near ZT8. The four genes that follow within this cluster encode for mitochondrial complexes and Ucp1, suggesting that large diurnal changes in mitochondrial dynamics are required for the flexible regulation of FA combustion [31e33]. Interestingly, we identified a large absolute diurnal amplitude for Lpl, encoding for the protein responsible for liberating FAs from TRLs [2], in cluster 4 with estimated peak expression at ZT12. Other genes with large amplitudes in clusters 3 and 4 were identified as being involved in lipid storage (i.e. glycerol-3-phosphate acyltransferase 4 (Gpat4), diacylglycerol O-acyltransferase 2 (Dgat2), cell death inducing DFFA like effector C (Cidec), and Perilipin 4 (Plin4)) and intracellular lipolysis (i.e. acyl-CoA synthetase long chain family member 1 (Acsl1), patatin like phospholipase domain containing 2 (Pnpla2), and carboxylesterase 1 D (Ces1d)).
To substantiate the idea that PPARg is involved in the regulation of diurnal gene expression in clusters 3 and 4, we performed chromatin immunoprecipitation (ChIP)-sequencing for PPARg on pooled iBAT samples. Out of 66,066 peaks (i.e. PPARg binding sites), 8,153 demonstrated diurnal oscillation as determined by JTK-cycle analysis (full details are provided in supplementary file: 'ChIP-sequencing').
Peaks were annotated to genes associated with the nearest transcriptional start site, and these data were used to identify the genes with oscillating PPARg binding and oscillating gene expression ( Figure 2A; Suppl Fig S4). Strikingly, within cluster 4, genes with oscillating PPARg binding showed gene ontology enrichment in (lipid) catabolic processes ( Figure 2B), while genes with non-oscillating PPARg binding did not show such enrichment (Suppl Fig S5). These data suggest a role of PPARg in regulating transcription of genes involved in intracellular and extracellular lipolysis when BAT is metabolically most active. This notion is further strengthened by strong oscillations in PPARg binding to sites annotated to abovementioned genes with largest diurnal amplitude (Suppl Table S1) that are involved in lipid storage and intracellular lipolysis (i.e. Acsl1, Ces1d, Dgat2, Gpat4, Plin4 and Pnpla2), with estimated peaks in PPARg binding on average 2 h prior to estimated peak in gene expression (Suppl Fig S6). Notably, six oscillating binding sites that were annotated to Lpl could be identified with estimated peak in PPARg binding at ZT6-7.5 (Suppl Fig S7). From these results we interpret that intracellular lipolysis at the end of the light (inactive) phase serves to supply BAT with fuel for thermogenesis. This seems to be followed by storage of FA taken up from the circulation after hydrolysis of TRLs by LPL and subsequent lipogenesis around the onset of the dark (active) phase, likely in order to replenish Original Article   CoA dehydratase 4 (Hacd4), and trans-2,3-enoyl-CoA reductase (Tecr) showed peak expression throughout the light phase (Suppl Fig S9). Cluster 2 (peak ZT21) contained the remaining oscillating DGs, characterized by primarily short or medium acyl chains (<C18:x). In addition, this cluster contained 67 of the 74 oscillating TGs ( Figure 3B), probably as the result of food intake, de novo lipogenesis, and metabolically inactive BAT. Cluster 3 (peak ZT16) contained 11 of the 13 oscillating FAs and eight of the nine oscillating monoacylglycerols (MGs; Figure 3B). These data are in concordance with our previous observations in young mice [34], and consistent with a peak in lipolytic activity at the onset of the dark (active) phase as indicated by the RNA-sequencing data and previous functional data on TG-derived FA-uptake by BAT [13], as both intracellular and extracellular lipolysis yield MGs and FAs. Of note, five of the identified FAs within cluster 3 are poly-unsaturated FAs (PUFAs), which are known ligands for PPARg [35] and UCP1 [36] and therefore likely contribute to thermogenic activation of the tissue followed by the uptake and storage of TG-derived FA from the circulation.
2.3. The lipoprotein lipase pathway follows diurnal oscillations with a peak at the onset of the active phase We observed peak expression of genes involved in lipid catabolic processes around ZT12 (onset of the dark phase) and identified Lpl as the gene with the largest amplitude within this cluster, highlighting its relevance in the diurnal oscillation of TG-derived FA-uptake by BAT [13]. To delineate which components of this TRL-processing pathway are oscillating, we next characterized the diurnal expression of genes involved in the LPL-mediated lipolytic processing of TRLs, as well as cellular uptake and transport of TG-derived FAs. Hereto, we visualized the diurnal oscillations of a manual selection of genes related to LPLmediated TRL processing, including the LPL-regulators Angptl3, Angptl4, and Angptl8, apolipoprotein (Apo)c1, Apoc2, Apoc3, Apoa5 [37], and related genes including glycosylphosphatidylinositol anchored high density lipoprotein binding protein 1 (Gpihbp1), FA transporter cluster of differentiation 36 (Cd36), and phospholipid transfer protein (Pltp). Strikingly, besides the strong oscillation of Lpl expression ( Figure 4A), expression of the LPL-inhibitors [38,39] Angptl4 and Angptl8 was found to oscillate with high amplitude, while expression of Angptl3 was low and not oscillating (Figure 4BeD). In particular, the expression of Angptl4 showed a 5-fold difference between highest (ZT3) and lowest (ZT15) measured levels opposing the oscillation of Lpl expression ( Figure 4B). It was previously reported that in WAT, oscillations of Angptl4 expression follow the pattern in BAT albeit with a 5-hour delay [40], illustrating tissue-specific regulation and a possible prioritization of fuel utilization by BAT prior to wakening [40]. ANGPTL8 forms a complex with hepatic-derived ANGPTL3 to strengthen its inhibition of LPL, as well as with ANGPTL4 to counteract its inhibitory effect on LPL [41], therefore the opposing oscillation of Angptl8 may amplify ANGPTL4 activity at its peak and further inhibit it at its nadir. Importantly, ANGPTLs modulate LPL on protein but not mRNA level, instead oscillations of Lpl gene expression may be driven by PPARg as described above. Additionally, Lpl and Angptl4 may be driven by daily fluctuations in insulin [1,42]. In contrast to ANGPTLs, apolipoproteins known to modulate LPL activity were either not oscillating (e.g. Apoc1; Figure 4E) or not detected in BAT (e.g. Apoc2, Apoc3, and Apoa5; data not shown). Similarly, gene expression of Gpihbp1, which protects LPL from catalysis by ANGPTLs, and the FA Original Article transporter Cd36 showed no oscillation (Figure 4FeG). Interestingly, we identified oscillation in the expression of Pltp ( Figure 4H). This protein contributes to the transfer of phospholipids from TRLs to highdensity lipoprotein (HDL) following liberation during LPL-mediated lipolysis, thereby facilitating the transport of TRL-remnant-derived cholesterol to the liver, as part of reverse cholesterol transport [43]. A schematic overview of LPL-mediated lipolysis and the autocrine regulation hereof by BAT via these genes and corresponding proteins is depicted in Figure 4I. Taken together, of the known LPL modulators expressed in BAT, Angptl4 showed the largest diurnal amplitude and opposite to Lpl. Accordingly, we hypothesized that ANGPTL4 is the main regulator of the diurnal variation in LPL activity within BAT.

Angiopoietin-like 4 modulation flattens oscillation of lipoprotein lipase activity and triglyceride-derived fatty acid uptake by brown adipose tissue
To delineate the role of Angptl4 in the diurnal regulation of LPL activity in BAT, we utilized whole-body Angptl4 knockout (KO) and transgenic overexpression (Tg) mice, which we compared to wild type (WT) mice at ZT0 (corresponding to the nadir in Lpl expression) and ZT12 (corresponding to the peak in Lpl expression). Knockout ( Figure 5A) or overexpression ( Figure 5B) of Angptl4 was confirmed by quantitative polymerase chain reaction (qPCR). Diurnal gene expression of Lpl was unaltered in Angptl4 KO mice ( Figure 5A), which is in line with the observation that ANGPTL4 inhibits LPL primarily at the protein rather than mRNA level [42]. On the other hand, gene expression of Lpl was increased in Angptl4 Tg mice at ZT12 ( Figure 5B), which might be a compensatory mechanism for reduced TG-derived FA-uptake due to highly suppressed LPL activity. Protein abundance of LPL was 3-fold higher at ZT12 compared to ZT0 in WT mice, in line with what was previously reported [13] (Figure 5CeD). In Angptl4 KO mice and Angptl4 Tg mice, LPL protein levels were not oscillating and equal to the abundance in WT mice at ZT12 and ZT0, respectively (Figure 5CeD), indicating that LPL is fully suppressed at ZT0 and fully activated at ZT12 in WT mice. The TG-hydrolase activity of iBAT-derived extracellular LPL was assessed in vitro ( Figure 5E,G) and matched the abundance of LPL (Figure 5CeD). To determine the functional consequences on the diurnal oscillation of TG-derived FA-uptake by BAT, Angptl4 KO and Angptl4 Tg mice were injected with TRL-like particles containing glycerol tri[ 3 H] oleate. In line with LPL abundance and activity, Angptl4 KO mice showed constitutively increased [ 3 H]oleate uptake by BAT ( Figure 5F), and Angptl4 Tg mice constitutively reduced [ 3 H]oleate uptake by BAT ( Figure 5H) compared to WT mice. Although generally regulated in the same direction, the abundance and TG-hydrolase activity of LPL in vitro at ZT0 did not translate one-to-one to TG-derived FA-uptake by BAT, indicating that other oscillating factors, such as ANGPTL8, may further finetune TG-derived FA-uptake in vivo.

CONCLUSIONS & PERSPECTIVE
We aimed to gain further insight into the diurnal regulation of metabolic BAT activity; the main findings of the transcriptomics and lipidomics data and our interpretations are summarized as a hypothetical model in Figure 6. Briefly, metabolic BAT activity peaks around the onset of the dark (active) phase (i.e. ZT12) [9e15]. At this time, FA for combustion are supplied through lipolysis of intracellular lipid stores and by LPL-mediated lipolysis of circulating TRLs, both possibly under transcriptional control of PPARg. This is followed by (de novo) lipogenesis and FA elongation, likely to replenish intracellular lipid stores and efficient combustion during the dark phase. The relative contribution of the various processes is probably context dependent, as cold exposure adds an extra dimension to the diurnal regulation of BAT activity by, for example, increasing the amplitude and expression of genes involved in de novo lipogenesis during the second half of the dark phase [18], when nutrient availability is expression, which encodes a protein that uncouples ATP production from mitochondrial oxidative phosphorylation (i.e. thermogenesis), and thereby likely brown adipose tissue (BAT) thermogenic activity is lowest. Genes involved in fatty acid (FA) elongation peak throughout the light phase, coinciding with increased abundance of long-acyl chain diacylglycerols (DGs). In the second half of the light phase, expression of mitochondrial complexes, as well as Ucp1 peak, suggesting peak thermogenic activity within the tissue.
Genes involved in intracellular lipolysis peak shortly before the highest abundance of FAs and monoacylglycerols (MGs), possibly to supply mitochondria with FAs for b-oxidation and to allosterically activate UCP1 [1]. To replenish intracellular lipid stores, the peak in triglyceride (TG)-rich lipoprotein-derived FA-uptake follows at the onset of the dark (active) phase (i.e. ZT12) [13], driven by lipoprotein lipase (LPL)-mediated hydrolysis, and likely contributing to the FA and MG abundance at the onset of the dark phase. This is followed by peak expression of lipogenic genes probably to store these FAs as TGs in intracellular lipid droplets. Throughout the dark phase there is ample glucose availability from food intake, allowing for the conversion of glucose to FA during peak expression of de novo lipogenic genes. This may be an additional pathway to supply BAT with lipids to replenish intracellular lipid stores, resulting in highest TG abundance at the end of the dark phase.
highest due to feeding. A limitation of the current study is that the experiment was performed only in male mice that were housed at room temperature.
The question remains what the main driving force is of the diurnal metabolic activity of BAT. Previous experiments indicated that oscillating BAT activity is modulated by glucocorticoids [15], but is independent of glucocorticoid receptor expression in BAT, suggestive of an indirect mechanism. Sympathetic denervation of BAT resulted in attenuated oscillations of TG-derived FA-uptake by BAT, but those experiments should be interpreted with caution given the complete abolishment of metabolic activity [13,44]. Based on the current study we suggest that stimulation of intracellular lipolysis [1] around wakening, possibly as a result of increased sympathetic activity, promotes thermogenesis and activation of PPARg [45] to regulate expression of genes involved in uptake, storage [46], and intracellular lipolysis of lipids. These data may explain why circadian disruption by prolonged daily light exposure [44] or by flattened corticosterone oscillation [15] attenuates TG-derived FAuptake by BAT and promotes adiposity in mice.
The current study primarily focused on BAT because it shows a strong diurnal rhythm in TG-derived FA-uptake as opposed to e.g. WAT [13]. However, we anticipate that a comparable approach in other tissues may provide valuable insights in the mechanisms driving their diurnal oscillations and contribute to the development of novel pharmacological strategies. Here, we identified oscillations of ANGPTL4 and LPL as important mediators in the diurnal regulation of metabolic BAT activity. ANGPTL4 is considered a therapeutic target for reducing cardiometabolic disease, as in humans, loss-of-function gene variants are associated with reduced circulating TG levels and lower cardiovascular disease risk, which could be mimicked in mice with the use of anti-ANGPTL4 monoclonal antibodies [47e53]. However, our data suggest that the ability of ANGPTL4 to modulate LPL activity might be dependent on time of day. ANGTPL4 inhibition at the onset of the resting phase may therefore be clinically most relevant for stimulating TG-derived FA-uptake by BAT.

Animals
All mice were housed under standard conditions with a 12h:12h light:dark schedule at 22 C with ad libitum access to a chow diet (Rat and Mouse No.3 Breeding, SDS, Horley, United Kingdom) and water. All mouse experiments were conducted in accordance with the Institute for Laboratory Animal Research Guide for the Care and Use of Laboratory Animals and were approved by the National Committee for Animal experiments. In a first experiment, non-fasted male C57BL/6J mice (Charles River Laboratories, Wilmington, MA, USA; 10 weeks old) were killed by CO 2 inhalation at eight time points over a 24-hour period (corresponding to ZT0, ZT3, ZT6, ZT9, ZT12, ZT15, ZT18 and ZT21; n ¼ 4 per time point). iBAT was collected and snap-frozen to assess oscillating gene expression by RNA-sequencing, oscillating lipid levels by lipidomics, and oscillating chromatin binding of PPARg by ChIP-sequencing (see below). In a second and third experiment, male whole-body Angptl4 KO and Angptl4 Tg mice, respectively, were compared with C57BL/6J mice (both on a C57BL/6J background; in-house breeding), which were obtained as described previously [54,55]. Angptl4 Tg mice were compared to littermates and Angptl4 KO mice to mice from another inhouse breeding. Angptl4 KO mice do not fully express the Angptl4 gene, resulting in a non-functional ANGPTL4 protein [54,55], whereas Angptl4 Tg mice overexpress Angptl4 under its own promotor [56].
Angptl4 KO (9e12 weeks old; n ¼ 7e9 per group per time point), Angptl4 Tg (9e12 weeks old; n ¼ 8 per group per time point) and their WT controls were killed at the onset of the light phase (corresponding to ZT0) and at the onset of the dark phase (corresponding to ZT12) to assess oscillating organ uptake of TG-derived FA (see below).

RNA sequencing
Total RNA of iBAT (approx. 10e20 mg; experiment 1) was isolated using the nucleospin kit (MachereyeNagel, Düren, Germany; 740955.50) after homogenization by a FastPrep-24Ô 5G bead beating grinder and lysis system Heatmaps of standardized residuals were generated using the 'heatmap.2' function using the complete linkage method and the complement of the Pearson distance (gplots version 3.0.4). Functional enrichment by gene ontology was performed using GOrilla [62], and transcription factor enrichment analyses were performed using ChIP-X enrichment analysis version 3 (ChEA3) [63].

ChIP-sequencing
ChIP was performed as previously described [64]. Briefly, iBAT samples pooled from eight mice (approx. 45 mg per mouse combined; experiment 1) were homogenized in lysis buffer (

Lipidomics
Lipidomics was performed as previously described [34,68], with minor adjustments. Briefly, the following amounts of internal standards   4.9. Statistical analyses P < 0.05 was considered statistically significant. In experiment 2 and 3, data were tested for normality by an Anderson-Darling test. Comparisons between groups were made by two-way ANOVA with a posthoc Tukey test when data were normally distributed. In case of nonnormal distributions, comparisons were made by a KruskalleWallis test. Statistical analyses were performed with GraphPad Prism software, version 8.4.2 (GraphPad, La Jolla, California) and R (http://www. r-project.org/; version 4.0.2). Data are presented as means AE SEM.

DATA AVAILABILITY
The RNA-sequencing and ChIP-sequencing datasets discussed in the current study have been deposited in NCBI's Gene Expression Omnibus (GEO) [70] and are accessible through GEO Series accession numbers GSE182045 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi? acc¼GSE182045) and GSE197261 (https://www.ncbi.nlm.nih.gov/ geo/query/acc.cgi?acc¼GSE197261). The lipidomics dataset discussed in the current study has been deposited in MetaboLights [71] and is accessible through study identifier MTBLS4082 (https://www. ebi.ac.uk/metabolights/MTBLS4082). The remaining datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.