Alcohol and Liver Clock Disruption Increase Small Droplet Macrosteatosis, Alter Lipid Metabolism and Clock Gene mRNA Rhythms, and Remodel the Triglyceride Lipidome in Mouse Liver

Eight to ten week old male liver-specific Bmal1 knockout mice (LKO mice: Alb^Cre+/−; Bmal1^Fl/Fl) and control littermates (Fl/Fl mice: Alb^Cre−/−; Bmal1^Fl/Fl) on a C57BL/6J background were weight-matched into pairs and then assigned into a control or alcohol (ethanol) diet feeding group as described in our previous study (Udoh et al., 2017). Mice were pair-fed iso-caloric (1 kcal/ml) liquid diets (control diet: F1259SP and ethanol diet: F1258SP) from Bio-Serv (Frenchtown, NJ; Filiano et al., 2013). Mice in the alcohol-fed group were acclimated to the diet by gradually increasing the concentration of alcohol in the diet from 0 to 3% (w/v) over a 1 week period and maintained at this level for 4 weeks (Udoh et al., 2017). This concentration of alcohol (3%, w/v) in the diet is equivalent to 22% total daily caloric intake. This alcohol-feeding protocol is widely used and well accepted as a model of the earliest stage of ALD – hepatic steatosis (Lieber et al., 1989; Brandon-Warner et al., 2012). Mice were single-housed in standard cages with bedding and kept in a temperature-, humidity-, and light-controlled room with a 12-h light:12-h dark (12-h L:D) schedule. The amount of liquid diet consumed by each mouse was measured every 24 h with diets replaced daily between zeitgeber time (ZT) 10–12 (ZT 0 = beginning of lights on/inactive period and ZT 12 = beginning of lights off/active period). At the end of the feeding study, mice were euthanized by decapitation every 4 h at ZT 3, 7, 11, 15, 19, and 23. During the dark period (ZT 12–24), an infrared viewer was used for tissue collection at ZT 15, 19, and 23 with photic input to the suprachiasmatic nucleus (SCN) prevented by removing the eyes prior to light exposure (Filiano et al., 2013). Blood was collected in EDTA-treated tubes and placed on ice before centrifugation to obtain plasma. Livers were quickly divided into portions and placed into liquid nitrogen, RNAlater™ stabilization solution (Invitrogen, Carlsbad, CA), or 10% buffered formalin. All procedures were approved by the UAB IACUC and compliant with the NIH Guide for the Care and Use of Laboratory Animals (8th ed., National Academy of Sciences, 2011).

Total RNA was isolated from liver using TRI-Reagent® (Sigma-Aldrich, St. Louis, MO). Isolated RNA was DNase treated with DNA-free™ DNase Treatment and Removal Reagents (Thermo Fisher Scientific, Waltham, MA) and the DNase-treated RNA (260/280 ratio > 1.8) was converted to cDNA using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, Waltham, MA). Measurements of mRNA levels were done by Real Time PCR (RT-PCR) using commercially available gene-specific TaqMan™ primers from Thermo Fisher Scientific (Waltham, MA) with either an Applied Biosystems 7900HT Real-time PCR or QuantStudio6 Flex Real-time PCR system (Thermo Fisher Scientific, Waltham, MA). Relative levels of mRNA for each gene were determined using the ΔΔCt method (Livak and Schmittgen, 2001) with values normalized to either glyceraldehyde-3-phosphate dehydrogenase (Gapdh) or peptidyl-prolyl isomerase A (Ppia). Data are presented as fold-change set to the control diet-Fl/Fl mice group trough (Udoh et al., 2017). A list of gene primers is provided in the supplemental data section (Supplementary Table 13).

Plasma TG was determined using reagents from Pointe Scientific (Canton, MI) and reported as mg TG/dL. For hepatic TG content, lipids were extracted by methods described previously (Filiano et al., 2013), TG was quantified using L-Type Triglyceride M colorimetric assay reagents (FUJIFILM Wako Diagnostics, Mountain View, CA) and reported as μg TG/mg liver.

Liver was collected every 4 h at ZT 3, 7, 11, 15, 19, and 23 and a small piece of liver was dissected from the large outer right lobe of all mice, placed into 10% buffered formalin, and embedded in paraffin before being sectioned at a 5 μm thickness. Liver was mounted onto slides and stained with hematoxylin and eosin (H and E) according to accepted guidelines (Kleiner et al., 2005). Slides from each experimental group and for all time-points were coded with a unique number designation for unbiased examination and scoring by pathologists. Steatosis (percentage of hepatocytes containing lipid droplets), lobular inflammation (number of inflammatory foci), ballooning, and fibrosis were scored by pathologists blinded to the experimental design using the NAFLD activity scoring (NAS) protocol (Brunt et al., 1999). While the NAS protocol is not intended for ALD, we applied this system to assign a histopathology score to cases in this experimental animal study. Steatosis was scored as 0, <5%; 1, 5–33%; 2, >34–65%; and 3, >66% of hepatocytes containing lipid droplets. Lobular inflammation was scored as: 0, no foci; 1, <2 foci; 2, 2–4 foci, and 3, >4 foci. Ballooned hepatocytes were rare and fibrosis was undetectable; thus, histopathology scores reported in Table 1 largely reflect steatosis and inflammation. Pathologists also recorded the percentage of hepatocytes containing large and small droplet macrosteatosis.

Livers were homogenized in 0.25 M sucrose buffer, pH 7.4, supplemented with protease and phosphatase inhibitor cocktails (Sigma-Aldrich, St. Louis, MO) and equal amounts of protein were separated on 4–20% Criterion™ TGX™ Precast Gels (Bio-Rad Laboratories, Hercules, CA) from all time-points of tissue collection (Udoh et al., 2015, 2017). Proteins were transferred from gels to nitrocellulose membranes and membranes were incubated at room temperature for 1 h in LI-COR® Blocking Buffer (LI-COR® Biosciences, Lincoln, NE). Membranes were incubated with primary antibodies overnight at 4°C: total ACC, 1:1,000 (cat. no. 3676, Cell Signaling Technologies, Danvers, MA); phospho-ACC-Ser79, 1:1,000 (cat. no. 11818, Cell Signaling Technology, Danvers, MA); FASN, 1:1,000 (cat no. 3180S, Cell Signaling Technology, Danvers, MA); and MCD, 1:1,000 (cat. no. PA5-22081, Thermo Fisher Scientific, Waltham, MA). An anti-β-actin antibody (cat. no. A5441, Sigma-Aldrich, St. Louis, MO) was used as the loading control (1:5,000). Membranes were washed and incubated in secondary antibodies from LI-COR® Biosciences (Lincoln, NE) for 1 h at room temperature. The secondary antibody used for total and phospho-ACC was IRDye 680RD (rabbit, cat. no. 926-68071), for MCD and FASN was 800CW (rabbit, cat. no. 926-32213), and for β-actin was 800CW (mouse, cat. no. 926-32210) or IRDye 680 RD (mouse, cat. no. 926-68070). Fluorescence was detected and densitometry of protein bands was quantified using the LI-COR Odyssey® CLx imaging system (Udoh et al., 2015, 2017). Data are presented as the average of protein densitometry across all time-points normalized to β-Actin for each treatment group as done in (Hatori et al., 2012). The antibodies used for total and phospho-ACC detect both ACC1 (265 kDa) and ACC2 (280 kDa) isoforms. Due to the close proximity of isoform bands on images, values represent the combined densitometry for total ACC1 and 2 and phospho-ACC1 and 2, respectively.

Lipidomics analysis was performed on liver lipid extracts collected at ZT 3 and ZT 15 from control and alcohol-fed Fl/Fl and LKO mice. Lipids were extracted from tissue using a modified Bligh and Dyer method (Bligh and Dyer, 1959). Briefly, 50–100 mg of frozen liver tissue was homogenized in 0.9 ml of dichloromethane and 2.0 ml of methanol (Honeywell Burdick and Jackson, Muskegon, MI). After homogenization, 10 μl of EquiSPLASH LIPIDOMIX Quantitative Mass Spec internal standard (cat. no. 330731, Avanti Polar Lipids, Alabaster, AL) was added and samples were incubated at room temperature for 30 min. One milliliter of liquid chromatography-mass spectrometry grade water (Honeywell Burdick and Jackson, Muskegon, MI) and 0.9 ml of dichloromethane were added, samples were centrifuged at 1200 rpm for 10 min at 4°C to separate phases, and the bottom hydrophobic organic phase containing lipids was collected and placed into a fresh glass test tube. The remaining upper hydrophilic phase was re-extracted with dichloromethane and the two organic phases were combined and dried under a stream of nitrogen. Lipids were re-solubilized in 300 μl injection solvent (dichloromethane:methanol, 50:50, with 5 mM ammonium acetate) and MS/MS^ALL mass spectrometric analysis and lipid species identification was performed as previously described in (Gajenthra Kumar et al., 2018) using a SCIEX Triple-TOF 6600+ (SCIEX, Framingham, MA) in the positive ion mode. Lipid species based on precursor fragment ion pairs were determined using the LipidView comprehensive target list (SCIEX, Framingham, MA) and the detected lipid species present in the TG pool (Supplementary Table 8) were compiled into a spreadsheet for statistical analyses. From these results, the percent TG FA saturation and percent TG FA composition were determined for the hepatic TG pool. The percent TG FA saturation was determined by dividing the sum of saturated fatty acids (SFA), monounsaturated fatty acids (MUFA), diunsaturated fatty acids (DUFA), or polyunsaturated fatty acids (PUFA) by the sum of all TG FA in the samples and multiplying by 100. Similarly, the percent TG FA composition was determined by dividing the sum of the most abundant individual TG FA species (12:0, 12:1, 14:0, 14:1, 16:0, 16:1, 18:1, 18:2, 20:0, 20:1, 20:2, 20:4, 22:1, and 22:6) by the sum of all TG FA in the samples and multiplying by 100.

Two or three-factor ANOVA, where indicated, were used to determine statistical significance of main effects of genotype, diet, time, and/or interactions with Tukey’s HSD test for post hoc analyses. To measure variation over time of day, cosinor analysis was performed to determine whether experimental measures fit a 24 h rhythm. For this, data were fitted to a cosine wave equation, f(t) = mesor +amplitude * cos[(2πt/T) + acrophase], in SPSS (IBM) using a non-linear regression module (Filiano et al., 2013; Udoh et al., 2015, 2017). The mesor (midline estimating statistic of rhythm) = mean of the rhythm; amplitude = 1/2 the distance between the peak and trough; t = time-point (ZT 3, 7, 11, 15, 19, or 23); T = the period (fixed to 24 h); and acrophase = ZT time of the peak of the rhythm, i.e., cosine maximum. Rhythmicity was determined using a linear regression model f(t) = M + cos (2πt/T) + sin (2πt/T), and data were considered rhythmic if the p of the R² was ≤0.05. Student’s t-test was used to compare parameter estimates among the experimental treatment groups when the overall R² was significant. Data that significantly fit a cosine function (rhythmic) are represented in graphs by solid lines, whereas non-significant data (arrhythmic) are represented by dashed lines. Sample sizes are included in figure legends with statistical significance set at p ≤ 0.05. Liver and plasma TG measurements were analyzed using non-parametric algorithm JTK_CYCLE in R (Hughes et al., 2010), with the period set to 24 h to test for rhythmicity in which statistical significance was set at p ≤ 0.05. Significant interactions of diet and genotype for the histopathology score was determined by ordinal contingency analysis comparing each treatment group, collapsed across ZTs. Steatosis grade, inflammation, and hepatocyte ballooning were assessed for significant differences in Fl/Fl vs. LKO control and alcohol-fed mice using Fisher’s Exact Test. For macrosteatosis, Fl/Fl control and LKO mice were compared across circadian time within diet groups. Hepatocytes containing small and large lipid droplets (macrosteatosis) were counted and quantified as a percentage of total hepatocytes in each image. Large droplet macrosteatosis percentages in both control diet groups were too low to warrant analysis. Distributions for large and small droplet macrosteatosis were positively skewed, which is often the case for count and percentage data. Positive skew for the small droplet macrosteatosis percentages within the alcohol group was corrected by a square root transformation. As a result, small droplet macrosteatosis data for alcohol-fed Fl/Fl and LKO mice were analyzed via cosinor analysis after square root transformation. For genotype comparisons of small droplet macrosteatosis percentages within the control diet group, transformation did not correct the positive skew. Thus, data were stratified by genotype and a negative binomial regression (Dupont, 2002) with linearized cosine functions as fixed factors [(Cos(2πt/24)) and Sin(cos(2πt/24))] was performed in SPSS (IBM) followed by planned post hoc comparisons between groups at each ZT. Likewise, large droplet macrosteatosis data were positively skewed and were analyzed similarly via negative binomial regression. For lipidomics analysis, we performed three-factor ANOVA to determine main effects of genotype, diet, time, and/or interactions with Tukey’s HSD for post hoc pairwise comparisons between each group. Two-factor ANOVA was performed for TG FA, where indicated, in cases that showed significant two-way interaction by three-factor ANOVA, but did not significantly show three-way interactions (Supplementary Tables 9–12).

Consistent with our previous findings (Filiano et al., 2013), chronic alcohol consumption significantly altered rhythmicity of clock gene mRNA levels in the liver (Cosinor analysis; Figure 1 and Supplementary Table 1). Specifically, alcohol decreased the midline estimating statistic of rhythm (mesor) and amplitude of Bmal1 in liver of Fl/Fl mice (Figure 1A). A low amplitude Bmal1 rhythm persists in livers of LKO mice that is most likely due to non-hepatocyte cell types in liver that do not express Cre recombinase (Figure 1B). Compared to control-fed mice, alcohol abolished rhythmic mRNA levels of the other positive regulator, Clock, in livers of Fl/Fl mice (Figure 1C). Diet had no effect on arrhythmic mRNA levels of Clock in livers from LKO mice (Figure 1D).

Nr1d1/REV-ERBα forms an important secondary feedback loop that regulates the clock via Bmal1 transcriptional repression (Preitner et al., 2002). Rhythmic mRNA levels of Nr1d1/REV-ERBα were evident in livers from all four groups (Supplementary Table 1). In Fl/Fl mice, alcohol induced a significant ~3 h phase advance compared to control-fed mice (Figure 1E). Liver clock disruption significantly phase delayed and reduced the amplitude of Nr1d1/REV-ERBα rhythms in livers from LKO mice fed either diet compared to Fl/Fl mice (Figures 1E,F). LKO livers showed a ~5-fold reduction in Nr1d1 amplitude compared to Fl/Fl mice (Figure 1F; Supplementary Table 1).

We next examined rhythmic transcription of Per2 and Cry1, negative regulators of the clock (Kume et al., 1999; Zheng et al., 1999). Although alcohol had a minimal effect on diurnal rhythms in mRNA levels of Per2 and Cry1 in livers of Fl/Fl mice (Figures 1G,I), the combination of liver clock disruption and alcohol resulted in arrhythmic Per2 mRNA levels (Figure 1H). Per2 mRNA levels remained rhythmic in livers of control-fed LKO mice, albeit with a significant phase delay in the peak of Per2 compared to control-fed Fl/Fl mice (Figure 1H). Cry1 mRNA levels were arrhythmic in livers of LKO mice (Figure 1J).

Other factors that regulate the clock, or are themselves regulated by the clock, may also be affected by alcohol and/or liver clock disruption. For example, casein kinase 1 delta and epsilon (Csnk1d and Csnk1e) are two kinases that regulate clock timing through phosphorylation of core clock proteins (Vielhaber et al., 2000; Etchegaray et al., 2009; Lee et al., 2009). Measurements of mRNA levels for these genes were arrhythmic and unaffected by alcohol (Figures 1K–N). However, several important clock-controlled genes showed sensitivity to alcohol or the combination of alcohol and liver clock disruption. Both nocturnin (Noct; an mRNA deadenylase; Kojima et al., 2012) and Dbp (a D-box binding b-ZIP transcription factor) showed rhythmic mRNA levels in livers of control-fed mice; however, amplitude was significantly reduced in LKO mice compared to Fl/Fl mice (Figures 1O–R). Importantly, alcohol altered the diurnal rhythms in mRNA levels of Noct and Dbp in livers from mice of either genotype (Figures 1O–R). Although Dbp in livers of alcohol-fed Fl/Fl mice was still rhythmic, the amplitude was decreased by more than half (Figure 1Q). While control-fed LKO mice exhibited low amplitude Dbp and Noct rhythms, these rhythms were lost when LKO mice were fed alcohol (Figures 1P,R). Another D-box binding b-ZIP transcription factor, Nfil3/E4BP4 (interleukin 3 regulated/E4 promoter-binding protein 4; Mitsui et al., 2001; Ripperger and Schibler, 2006) displayed a significant diurnal rhythm in livers of control-fed Fl/Fl mice that was anti-phase to Nr1d1/REV-ERBα and Dbp (Figure 1S). Alcohol, liver clock disruption, or both, abolished Nfil3/E4BP4 rhythmicity (Figures 1S,T). Together, these results extend our previous findings and show that genetic disruption of the liver clock exacerbates the circadian rhythm-impairing effects of alcohol.

In order to assess whether alcohol-induced effects on overall clock mRNA levels depend on genotype (independent of time), we performed two-factor ANOVA on clock and clock-controlled genes (Supplementary Table 2). Alcohol reduced overall levels of Bmal1, Clock, Cry1, Nfil3/E4BP4, and Dbp, regardless of genotype (main effect of diet). In addition, livers from LKO mice generally had higher mRNA levels of Clock, Per2, Cry1, Csnk1d, and Csnk1e, as well as lower expression of Bmal1, Dbp, and Nr1d1/REV-ERBα, compared to livers from Fl/Fl mice (main effect of genotype). Although no Genotype X Diet interactions reached significance, there was a trend for Bmal1, Dbp, and Nfil3/E4BP4. Specifically, alcohol-fed Fl/Fl mice had lower mRNA levels of Bmal1 and Dbp, whereas LKO mice on either diet had minimal levels of mRNA for these genes (Figures 1B,R). Higher Nfil3/E4BP4 mRNA levels in livers of control-fed LKO mice tended to be reduced when mice were fed alcohol (Figures 1S,T).

TG was significantly elevated in livers of alcohol-fed Fl/Fl and LKO mice (Figures 2B,D) compared to control diet counterparts (Figures 2A,C). Two-factor ANOVA showed significant interaction of Genotype X Diet for liver TG (Figure 2I). Plasma TG were also increased by alcohol in both genotypes (Figures 2F,H) compared to control diet mice (Figures 2E,G) with the highest overall levels of plasma TG observed in alcohol-fed LKO mice (Figure 2H). Plasma TG were rhythmic in control-fed Fl/Fl and LKO mice (Fl/Fl, p = 0.049; LKO, p = 0.0005; JTK cycle). Similar to liver TG, two-factor ANOVA for plasma TG shows a significant interaction of Genotype X Diet (Figure 2I). Taken together, these data illustrate differential impacts of liver clock disruption on TG metabolism in alcohol-fed mice.

A significantly greater percentage of livers from alcohol-fed LKO mice had higher overall histopathology scores (one or more) compared to the other three groups (Table 1). Steatosis was analyzed by comparing livers scored at grade 0 to those scored at grades 1 + 2 in control-fed and alcohol-fed mice. There was a statistically significant difference in steatosis grade in livers of alcohol-fed mice with increased numbers of livers exhibiting grade 1 + 2 compared to grade 0, and there was no effect in control-fed mice (Table 1). We observed no significant differences in lobular inflammation between groups. Ballooned hepatocytes were only seen in livers of LKO mice that were collected during the inactive period (ZT 0–12) of the day (Table 1). Mildly reactive Kupffer cells were also rare and only detected in livers of two alcohol-fed LKO mice.

We also examined livers for temporal differences in both large and small droplet macrosteatosis (Figure 3). Representative images are presented for livers collected at ZT 15 (Figures 3I–L). Both large (Figures 3A–D) and small droplet (Figures 3E–H) macrosteatosis were present; however, small droplet macrosteatosis was more prevalent with the highest levels seen in livers of alcohol-fed mice (Figures 3F,H). Low levels of large and small droplet macrosteatosis were observed in livers of some control-fed Fl/Fl mice (Figures 3A,E), with higher levels in livers of control-fed LKO mice (Figures 3C,G). In alcohol-fed mice, large droplet macrosteatosis displayed a significant 24 h rhythm (Figure 3B; χ²(5) = 12.8, p = 0.005) in livers of Fl/Fl mice but not in livers of LKO mice (Figure 3D; χ²(5) = 1.23, p = 0.805). Specifically, large droplet macrosteatosis in livers of alcohol-fed Fl/Fl mice were significantly less likely to be observed at ZT 3, 7, and 19 compared to ZT 11 (Figure 3B; χ²(5) = 12.7, p = 0.03). In LKO mice fed a control diet, livers had a decreased likelihood of large droplet macrosteatosis at ZT 3, 7, 15, 19, and 23 compared to ZT 11 (Figure 3C; χ²(5) = 53.6, p = 2.5E-10).

We also examined temporal dynamics of small droplet macrosteatosis. Cosinor analysis revealed that small droplet macrosteatosis exhibited a 24 h rhythm in livers of alcohol-fed Fl/Fl mice (Figure 3F; R² = 0.29; p = 0.006; peak phase = ZT 14.44) but not in livers of alcohol-fed LKO mice (Figure 3H; R² = 0.15; p = 0.11). In contrast, small droplet macrosteatosis was rhythmic in livers of control-fed LKO mice (Figure 3G; χ²(5) = 48.1, p = 3.3E-9), but not in livers of control-fed Fl/Fl mice (Figure 3E; χ²(5) = 8.0, p = 0.16). Livers of control-fed LKO mice had a decreased likelihood of small droplet macrosteatosis specifically at ZT 3, 15, 19, and 23 compared to ZT 7 and ZT 11 (Figure 3G; χ²(5) = 25.3, p < 0.05). Together, these results show that alcohol-induced macrosteatosis varies over the 24h day in a liver-clock dependent manner.

Sterol regulatory element binding transcription factor 1/protein 1C (Srebf1/SREBP-1C) mRNA levels were rhythmic, peaking in the inactive period of the day in livers of control-fed Fl/Fl mice with alcohol inducing arrhythmicity (Figure 4A). Srebf1/SREBP-1C was arrhythmic in livers of LKO mice (Figure 4B). MLX interacting protein like/carbohydrate-responsive element binding protein (Mlxipl/ChREBP) was arrhythmic in all groups (Figures 4C,D). Nuclear receptor subfamily 1 group H member 3/Liver X receptor α (Nr1h3/LXRα) was also arrhythmic in all groups (Figures 4E,F), whereas Nr1h2/LXRβ was rhythmic in livers of control-fed Fl/Fl mice, but arrhythmic in livers of alcohol-fed Fl/Fl (Figure 4G) and LKO mice (Figure 4H). Peroxisome proliferator-activated receptor alpha (PPARα), a transcription factor that activates expression of many FAO and FA transport genes, was surprisingly arrhythmic in livers of control-fed Fl/Fl mice, but rhythmic in livers of alcohol-fed Fl/Fl mice (Figure 4I). Ppara was also arrhythmic in livers of LKO mice (Figure 4J). Results of cosinor analyses for lipid metabolism transcription factors are included in Supplementary Table 3.

Two-factor ANOVA was performed to determine significant main effects of alcohol and genotype on mRNA levels of lipid metabolism transcription factors, regardless of time (Supplementary Table 4). Independent of genotype, alcohol-fed mice generally displayed reduced mRNA levels of Srebf1/SREBP-1C, Nr1h3/LXRα, Nr1h2/LXRβ, and Ppara. Further, genetic disruption of the liver clock resulted in increased mRNA levels of Srebf1/SREBP-1C and Nr1h2/LXRβ, while decreasing mRNA levels of Mlxipl/ChREBP and Ppara (main effect of genotype). We also observed a trend toward a significant interaction of Genotype X Diet for Mlxipl/ChREBP.

Acetyl CoA carboxylase alpha/1 (Acaca/ACC1), an enzyme that catalyzes the formation of malonyl-CoA (substrate for de novo FA synthesis) by the carboxylation of acetyl-CoA (Thampy and Wakil, 1988), was rhythmic in livers of control-fed Fl/Fl mice and arrhythmic in livers of alcohol-fed Fl/Fl mice and LKO mice (Figures 5A,B). Acacb/ACC2 also catalyzes the generation of malonyl-CoA, which acts as an allosteric inhibitor of FAO by decreasing FA uptake into mitochondria by carnitine palmitoyltransferse-1A (CPT1A) activity (Abu-Elheiga et al., 2000). Acacb/ACC2 mRNA levels were rhythmic in livers of control and alcohol-fed Fl/Fl mice (Figure 5C) with alcohol significantly decreasing the mesor. While Acacb/ACC2 was rhythmic in livers of control-fed LKO mice (Figure 5D), the mesor was decreased and the peak was phase delayed in livers of control-fed LKO mice compared to Fl/Fl mice. Acacb/ACC2 was arrhythmic in livers of alcohol-fed LKO mice (Figure 5D). Malonyl CoA decarboxylase (Mlycd/MCD), an enzyme that catalyzes malonyl-CoA degradation to acetyl-CoA, was rhythmic in livers of control and alcohol-fed Fl/Fl mice (Figure 5E); however, alcohol feeding significantly decreased the mesor and induced a phase delay. Mlycd/MCD was rhythmic in livers of control-fed LKO mice (Figure 5F) with increased mesor and delayed peak mRNA levels compared to livers of control-fed Fl/Fl mice. Mlycd/MCD was arrhythmic in livers of alcohol-fed LKO mice (Figure 5F). Cpt1a was rhythmic in livers of Fl/Fl mice (Figure 5G) and alcohol-fed LKO mice (Figure 5H) with peak mRNA levels phase advanced in livers of alcohol-fed LKO mice compared to alcohol-fed Fl/Fl mice.

Fatty acid synthase (Fasn), a multi-functional enzyme that uses acetyl-CoA and malonyl-CoA to eventually produce palmitic acid (16:0), was rhythmic in livers of control and alcohol-fed Fl/Fl mice with peak mRNA levels occurring in the early active period (Figure 5I). Fasn was arrhythmic in livers of LKO mice (Figure 5J). FA elongation and biosynthesis of MUFA and PUFA is catalyzed through actions of stearoyl CoA desaturase 1 (Scd1), fatty acid desaturase 1 (Fads1), Fads2, and fatty acid elongase 5 (Elovl5) and Elovl6 (Zhang et al., 2016a). Scd1, the rate-limiting step in the synthesis of MUFA, was arrhythmic in all groups (Figures 5K,L). Elovl5 was rhythmic in livers of Fl/Fl mice (Figure 5M), but arrhythmic in livers of LKO mice (Figure 5N). Interestingly, diurnal rhythms in Elovl6 mRNA levels were anti-phase to Elovl5 mRNA levels when comparing control-fed Fl/Fl mice (Figures 5M,O) and was arrhythmic in livers of alcohol-fed Fl/Fl mice (Figure 5O) and livers of LKO mice (Figure 5P). Fads1 mRNA levels were arrhythmic in all groups (Figures 5Q,R), whereas Fads2 mRNA levels were rhythmic only in livers of control-fed Fl/Fl mice (Figures 5S,T). Results of cosinor analyses for these FA metabolism genes are provided in Supplementary Table 5.

Two-factor ANOVA (Supplementary Table 4) revealed that the overall mRNA levels of Acaca, Acacb, Mlycd, Elovl6, Fads1, and Fads2 had significant main effects of genotype, independent of diet. In LKO mice, we observed overall increased mRNA levels of Acaca, Mlycd, and decreased Acacb, Elovl6, Fads1, Fads2, and Scd1. Independent of genotype, alcohol decreased overall mRNA levels of Acacb and Mlycd and increased mRNA levels of Elovl6. A significant interaction of Genotype X Diet was detected for Acacb where alcohol decreased Acacb in Fl/Fl, but not LKO mice. In contrast, alcohol increased Fasn mRNA levels in livers of LKO, but not Fl/Fl mice. In addition, a trend toward a significant interaction was observed for Acaca and Mlycd. Specifically, in LKO mice, mRNA levels of Mlycd tended to be decreased and Acaca mRNA levels increased by alcohol feeding.

We also examined protein abundance and phosphorylation for a small set of FA metabolism enzymes, ACC, MCD, and FASN, to complement gene expression analyses. As no time-of-day differences were detected, results were analyzed by pooling all time points for each treatment group, as done in previous studies (Hatori et al., 2012). While the antibodies used for total and phospho-ACC detect both ACC1 (265 kDa) and ACC2 (280 kDa) isoforms, our data largely showed the presence of the lower molecular weight lipogenic ACC1 isoform in liver (Figure 6), which is in agreement with other studies in rodents (Bianchi et al., 1990). Total ACC was unchanged in all treatment groups (Figure 6A), whereas p-ACC levels were significantly lower in livers of alcohol-fed LKO mice compared to levels measured in control-fed LKO mice (Figure 6B). Two-factor ANOVA showed a significant main effect of diet for p-ACC and a trend toward a significant Genotype X Diet interaction (p = 0.07). MCD was significantly lower in livers of LKO mice compared to Fl/Fl mice (Figure 6C) with a significant effect of genotype. Similarly, FASN was lower in livers of alcohol-fed Fl/Fl mice and LKO mice compared to control-fed Fl/Fl mice (Figure 6D). Two-factor ANOVA showed a significant genotype effect for FASN.

Previous studies report diurnal rhythms in mRNA levels of TG metabolism genes (Figure 7A) in the liver (Adamovich et al., 2014); however, the impact alcohol and liver clock disruption have on these day/night differences is unknown. Glycerol-3-phosphate acyltransferase 1 (Gpat1), the first enzyme of TG synthesis that catalyzes acylation of glycerol-3-phosphate (G3P) to lysophosphatidic acid (LPA; Coleman and Mashek, 2011), was rhythmic in livers of control-fed Fl/Fl and LKO mice, whereas mRNA levels were arrhythmic in livers of alcohol-fed mice (Figures 7B,C). The next step in TG synthesis involves 1-acyl-sn-glycerol-3-phosphate acyltransferase 1 and 2 (AGPAT1, 2) that transfer an additional FA to LPA producing phosphatidate (PA). Agpat1 was rhythmic in livers of both control-fed Fl/Fl and LKO mice (Figures 7D,E) with peak Agpat1 mRNA levels phase delayed in livers of LKO mice. Agpat1 was arrhythmic in livers of alcohol-fed Fl/Fl mice, but rhythmic in livers of alcohol-fed LKO mice (Figures 7D,E). Agpat2 mRNA levels were rhythmic in livers of Fl/Fl mice (Figure 7F), as well as livers of control-fed LKO mice, but not livers of alcohol-fed LKO mice (Figure 7G). The peak of Agpat2 was phase advanced ~3.5 h in livers of control-fed LKO compared to control-fed Fl/Fl mice. Lipins catalyze diglyceride (DG) formation through phosphatidate phosphatase-1 activity (Reue and Wang, 2019). Lpin1 displayed a high amplitude rhythm in livers of control-fed Fl/Fl mice that was significantly lower in livers of alcohol-fed Fl/Fl mice (Figure 7H) and control-fed LKO mice (Figure 7I). Lpin1 was arrhythmic in livers of alcohol-fed LKO mice (Figure 7I). Lpin2 mRNA levels were also rhythmic in livers of control-fed Fl/Fl mice, but arrhythmic in livers of alcohol-fed Fl/Fl mice (Figure 7J) and LKO mice (Figure 7K). The last step of TG synthesis is the conversion of DG to TG by diacylglycerol O-acyltransferase (DGAT) enzymes. Dgat2 mRNA levels were rhythmic in livers of control and alcohol-fed Fl/Fl mice (Figure 7L) and control-fed LKO mice (Figure 7M), but arrhythmic in livers of alcohol-fed LKO mice (Figure 7M).

We also measured 24 h rhythms in mRNA levels for patatin like phospholipase domain containing 2/adipose triglyceride lipase (Pnpla2/ATGL) and Pnpla3, which are lipases that mediate the first step in TG breakdown (Figure 7A; Trepo et al., 2016; Schreiber et al., 2019). Pnpla2/ATGL mRNA levels were rhythmic in livers of both control-fed and alcohol-fed Fl/Fl mice; however, alcohol feeding significantly decreased the mesor and amplitude of Pnpla2/ATGL (Figure 7N). Lower mRNA levels of Pnpla2/ATGL were also seen in livers of control-fed LKO mice with alcohol inducing arrhythmicity (Figure 7O). Pnpla3 exhibited a high amplitude rhythm in livers of control-fed Fl/Fl mice that was significantly decreased by alcohol feeding (Figure 7P). Pnpla3 was arrhythmic in livers of LKO mice (Figurer 7Q). Cosinor analyses results for TG metabolism genes are included in Supplementary Table 6.

Assessment of mRNA levels of TG metabolism genes was next performed by two-factor ANOVA to determine whether the observed effects of alcohol were dependent on genotype, regardless of time of day (Supplementary Table 4). Alcohol decreased mRNA levels of Agpat1, Lpin1, Lpin2, and Pnpla2/ATGL and Pnpla3. Significant main effects of genotype were found for Agpat1 and Agpat2, which were generally increased in livers of LKO mice. In contrast, Pnpla3 mRNA levels were significantly decreased in livers of LKO mice. We also observed a significant interaction of Genotype X Diet for Pnpla3, due to the dramatic alcohol-induced decrease in Pnpla3 mRNA levels observed in Fl/Fl mice.

Lipid droplet turnover is regulated by the coordinated actions of multiple proteins located on the lipid droplet surface (Figure 8K). First, we measured 24 h rhythms in mRNA levels of two members of the perilipin (PLIN) family, Plin2 and Plin5, which recruit lipases to the lipid droplet surface initiating lipolysis (Itabe et al., 2017). Plin2 mRNA levels were rhythmic in livers of Fl/Fl mice (Figure 8A) and arrhythmic in LKO mice (Figure 8B). In contrast, Plin5 was arrhythmic in livers of Fl/Fl mice (Figure 8C), but rhythmic in livers of LKO mice (Figure 8D). Abhydrolase domain containing 5/comparative gene identification-58 (Abdh5/CGI-58), postulated to activate Pnpla2/ATGL and start TG hydrolysis (Lass et al., 2006), was arrhythmic in all groups (Figures 8E,F). Lipase E/hormone sensitive type (Lipe/HSL), which catalyzes DG hydrolysis, was rhythmic in livers of control-fed Fl/Fl mice, but arrhythmic in livers of alcohol-fed Fl/Fl mice (Figure 8G) and livers of LKO mice (Figure 8H). The final step in lipolysis is catalyzed by monoacylglycerol lipase (Mgll), which works in concert with Lipe/HSL to hydrolyze DG to FA and glycerol (Fredrikson et al., 1986). Mgll was rhythmic in livers of control and alcohol-fed Fl/Fl mice (Figure 8I), but arrhythmic in livers of LKO mice (Figure 8J). Results of cosinor analyses for lipid droplet-associated genes are included in Supplementary Table 7.

We examined genotype and alcohol-induced effects on mRNA levels of lipid droplet components by two-factor ANOVA, independent of time (Supplementary Table 4). Livers of LKO mice generally exhibited increased mRNA levels of Plin5 and Abdh5/CGI-58 and decreased Lipe/HSL and Mgll (main effect of genotype). Alcohol decreased mRNA levels of Plin2, Plin5, Lipe/HSL, and Abdh5/CGI-58, independent of genotype (main effect of diet). Plin2 mRNA levels were lower in livers of alcohol-fed LKO mice, with a significant Genotype X Diet interaction. A trend toward a significant Genotype X Diet interaction was found for Plin5 and Abdh5/CGI-58, where increased mRNA levels of these genes in livers of LKO mice tended to be decreased by alcohol.

Finally, to gain greater insight into the effects alcohol and liver clock disruption have on lipid metabolism we conducted a lipidomics analysis on livers collected at ZT 3 and ZT 15. A complete list of the TG species detected in livers is provided in Supplementary Table 8 and these results were used to calculate the percent TG FA saturation and percent TG FA composition in the hepatic TG pool. First, we analyzed for effects on FA saturation. Hepatic TG FA were grouped accordingly to the saturation state (SFA, MUFA, DUFA, or PUFA) and data were analyzed by three-factor ANOVA (Supplementary Table 9). Alcohol significantly decreased the percentage of TG SFA in livers of Fl/Fl and LKO mice compared to control diet mice (Figure 9A) with significant main effects of genotype and diet. In contrast, TG MUFA are increased in livers of alcohol-fed mice (vs. control-fed mice) and at ZT 15 (vs. ZT 3; Figure 9B) with significant main effects of diet and time. Alcohol also increased TG DUFA levels (compared to control-fed mice) and livers from LKO mice had higher TG DUFA than livers from Fl/Fl mice (main effects of genotype and diet; Figure 9C). TG PUFA made up the smallest percentage of FA detected in the hepatic TG pool. Alcohol decreased the percentage of TG PUFA (main effect of diet; Figure 9D).

Next, we analyzed lipidomics results to look for effects on the individual FA species in the hepatic TG pool (Figure 10A). Results for three-factor ANOVA are provided in Supplementary Table 9. The significant interaction of Genotype X Diet X Time was observed for several long chain FA with 20 or more carbons. Specifically, pair-wise comparisons (Supplementary Table 10) show that the combination of alcohol and liver clock disruption significantly increased liver content of arachidic acid (20:0; Figure 10J), eicosadienoic acid (20:2; Figure 10L), and docosenoic acid (22:1; Figure 10N), especially at ZT 15. In contrast, alcohol and liver clock disruption significantly decreased arachidonic acid (20:4), specifically at ZT 3 (Figure 10M).

Significant main effects of genotype, diet, or time were also found for several FA species. Independent of genotype and time (main effect of diet; Supplementary Table 9), alcohol significantly increased lauric acid (12:0; Figure 10B), lauroleic acid (12:1; Figure 10C), and the most abundant FA, oleic acid (18:1; Figure 10H). The PUFA docosahexanoic acid (22:6) also tended to be increased in livers of alcohol-fed mice compared to control-fed mice (Figure 10O). Interestingly, alcohol induced an overall decrease in palmitic acid (16:0), the enzymatic product of FASN and a long chain FA precursor, compared to mice fed the control diet (Figure 10F). Genetic disruption of the liver clock alone (main effect of genotype; Supplementary Table 9) induced an overall increase in lauric acid (12:0; Figure 10B), lauroleic acid (12:1; Figure 10C) and linoleic acid (18:2; Figure 10I). We also observed a general decrease in the levels of palmitic acid (16:0; Figure 10F) in livers of LKO mice compared to Fl/Fl mice (main effect of genotype; Supplementary Table 9). A time-of-day difference was observed for the overall level of myristoleic acid (14:1; Figure 10E), which was higher at ZT 3 regardless of genotype or diet (main effect of time; Supplementary Table 9).

Where significant two-factor interactions (Genotype X Diet or Diet X Time) and no three-factor interaction was present, we conducted two-factor ANOVA with pairwise comparisons between groups. Specifically, corresponding data were collapsed together across time for a Genotype X Diet interaction or across genotype for a Diet X Time interaction. No Genotype X Time interactions were present. These additional analyses revealed that independent of time there was a significant interaction of Genotype X Diet (and) for myristic acid (14:0;) and myristoleic acid (14:1). Alcohol significantly increased overall levels of myristic acid (14:0) in livers of Fl/Fl mice (); however, pair-wise comparisons did not reach significance (). Pair-wise comparison revealed increased levels of myristoleic acid (14:1) in livers of control-fed LKO mice compared to control-fed Fl/Fl mice and alcohol-fed LKO mice (). Time-of-day dependent effects of alcohol feeding (significant interaction of Diet X Time;) were observed for linoleic acid (18:2) and eicosenoic acid (20:1), regardless of genotype. Specifically, at ZT 3 alcohol-fed mice have increased linoleic acid levels, whereas alcohol feeding has no significant effect at ZT 15 (). Conversely, higher levels of eicosenoic acid are present in livers of alcohol-fed mice compared to control-fed mice at ZT 15, but not ZT 3 (). Collectively, these exciting new lipidomics findings indicate that chronic alcohol consumption and liver clock disruption differentially affect FA saturation and composition of the hepatic TG pool. Supplementary Tables 9 11 Supplementary Figure 1A Supplementary Figure 1B Supplementary Figure 1A Supplementary Figure 1C Supplementary Figures 1B,C Supplementary Table 8 Supplementary Figures 2A,C Supplementary Figures 2B,C

PB was employed by the Avanti Polar Lipids, Inc (USA).

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.