Time-restricted feeding ameliorates non-alcoholic fatty liver disease through modulating hepatic nicotinamide metabolism via gut microbiota remodeling

Five-to-six weeks old male C57BL/6J mice were obtained from the Animal Center of the First Affiliated Hospital of Sun Yat-sen University. The mice were housed in a specific-pathogen-free (SPF) environment under a 12-h light/dark cycle at a temperature of 25 ± 2°C and a humidity of 55–60%. Following a week of acclimatization, the mice were randomly divided into four groups and subjected to different dietary interventions for a duration of 12 weeks (n = 8 per group): (1) chow diet (CD); (2) HFD containing 60% fat (TROPHIC, TP23400); (3) TRF: HFD for the initial 6 weeks, followed by a 16:8 regimen consisting of an 8-h eating period and a 16-h fasting period, with ad libitum access to water; (4) pair feeding (PF): HFD for the initial 6 weeks, with the average daily intake matching that of the TRF group but without time window restriction for another 6 weeks. Throughout the intervention, daily food intake was recorded, and body weight was measured weekly. Fresh fecal samples were collected under sterile environment during the last 3 days of week 12 for subsequent microbiota transplantation. At the end of the 12-week period, the mice were anesthetized with 1% sodium pentobarbital (100 mg/kg) via intraperitoneal injection after a 4-h fast. Blood samples were immediately collected from the retro-orbital plexus into EDTA-K2-coated microtubes. Plasma was obtained by centrifugation at 3000 rpm for 15 min at 4°C. The liver and epididymal adipose tissue was excised and weighed, which was fixed by 4% paraformaldehyde or cryopreserved at −80°C. The liver index was determined by dividing the liver weight by the body weight. Cecal contents were also collected and cryopreserved at −80°C for metagenome sequencing analysis. All animal experiments were approved by the Institutional Animal Care and Use Committees of the First Affiliated Hospital of Sun Yat-sen University and complied with the National Institutes of Health Guidelines on the Care and Use of Laboratory Animals.

The left lateral lobe of the liver and epididymal adipose tissues fixed by 4% paraformaldehyde for 24 h were subsequently embedded in paraffin. Sections with a thickness of 4 µm were stained using the hematoxylin and eosin (H&E) method following standard protocols. Representative images were captured by an optical microscope at 200× magnification. The NAFLD activity score (NAS) was determined based on the steatosis grade (0–3), lobular inflammation (0–3), and hepatocyte ballooning (0–2), which collectively reflected the progression and severity of NAFLD.²⁴ For oil red O staining, the liver tissues were embedded in an optimal cutting temperature compound and frozen sectioned. The oil red O positive areas and the size of epididymal adipocytes were calculated using ImageJ software (National Institutes of Health, USA).

The levels of total triglyceride (TG), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-c), alanine aminotransferase (ALT), alanine aminotransferase (AST), and reduced glutathione (GSH) in plasma or liver were determined by commercial assay kits following the manufacturer’s protocol (Nanjing Jiancheng Bioengineering Institute, A110-1-1, A111-1-1, A113-1-1, C009-2-1, C010-2-1, A006-2-1). Liver malondialdehyde (MDA) levels were also determined using the corresponding assay kits (Solarbio, BC0025). To prepare the liver homogenates, livers were weighed and mixed with 9 times the volume of ice-cold anhydrous ethanol or saline. The mixture was mechanically homogenized at 60 Hz for 1 min and then centrifuged at 2500–3500 rpm for 10 min at 4°C.

The plasma samples (50 µL) stored at −80°C were thawed on ice and then mixed with 300 µL extraction solution (acetonitrile: methanol = 1:4 V/V) containing internal standards in a 2 mL microcentrifuge tube. The mixture was vortexed for 3 min and then centrifuged at 12,000 rpm for 10 min at 4°C. Two hundred microliter supernatant was collected and placed in a −20°C environment for 30 min, followed by centrifugation at 12,000 rpm for 3 min at 4°C. Subsequently, 180 µL supernatant was obtained for further analysis using ultra-performance liquid chromatography (UPLC, ExionLC AD, https://sciex.com.cn/↗) and tandem mass spectrometry (MS/MS, QTRAP® System, https://sciex.com/↗).

The UPLC conditions were as follows: (1) chromatographic column: Waters ACQUITY UPLC HSS T3 C18 (1.8 µm, 2.1 mm *100 mm); (2) mobile phase: 0.1% formic acid in water as phase A and 0.1% formic acid in acetonitrile as phase B; (3) gradient elution: 95:5 V/V at 0 min, 10:90 V/V at 11.0 min, 10:90 V/V at 12.0 min, 95:5 V/V at 12.1 min, 95:5 V/V at 14.0 min; (4) flow rate: 0.4 mL/min; (5) column temperature: 40°C; (6) sample size: 2 µL. The specific parameters of MS/MS were presented as follows: (1) electrospray ionization (ESI): 500°C; (2) voltage: 5500 V (positive), −4500 V (negative); (3) ion source gas: gas I (GSI), gas II (GSII), and curtain gas (CUR) were set at 55, 60, and 25 psi, respectively; (4) collision-activated dissociation (CAD): high. Each ion pair was detected based on optimized de-clustering potential and collision energy.

Metabolite qualitative profiling was performed using the MWDB database, while quantitative analysis was conducted using multiple reaction monitoring (MRM). The orthogonal partial least squares discriminant analysis (OPLS-DA) was carried out using R software (version 4.2.2; https://www.r-project.org↗) with the MetaboAnalyst R package. The differential metabolites were annotated and subsequently mapped to the KEGG Pathway database, which were then subjected to metabolite set enrichment analysis (MSEA) in MetaboAnalyst 5.0.

Fifty microliter plasma samples stored at −80°C were thawed on ice and then combined with 200 µL ice-cold methanol containing internal standard. One hundred milligram liver samples were also thawed and mechanically homogenized with 1 mL ice-cold methanol. The above mixture was further vortexed and centrifuged at 14,000 rpm for 10 min at 4°C. The supernatant was collected and lyophilized by a freeze dryer. The lyophilized powder was dissolved in 50 µL 50% acetonitrile. Subsequently, the clear supernatant was transferred to glass vials with microinserts, and LC-MS analysis was conducted using Agilent 1290 Infinity II UPLC −6470B Triple-quadrupole LC-MS system with a dual AJS ESI source (Agilent Technologies, Inc., USA). The chromatographic settings were as follows: (1) chromatographic column: Thermo Hypersil Gold C18 column (1.9 µm, 2.1 mm *100 mm); (2) mobile phase: 0.1% formic acid in water as phase A and 0.1% formic acid in acetonitrile as phase B; (3) gradient elution: 0–0.5 min, 95:5 V/V, 0.5–3 min, 95:5 -V/V→5:95 V/V, 3–3.5 min, 5:95 V/V→95:5 V/V, equilibrate for 5 min; (4) flow rate: 0.3 mL/min; (5) sample size: 1 µL. MRM in positive electrospray ionization mode (Agilent Jet Stream) was utilized with the following transitions: N-Me-6-PY at m/z 152.9 → 107.9 and m/z 152.9 → 67.0; N-Me-4-PY at m/z 152.9 → 135.9 and m/z 152.9 → 92; 1-methylnicotinamide (1-MNA) at m/z 138 → 78.6 and m/z 138 → 94.6; NAD+ at m/z 664 → 428 and m/z 664 → 136; NAM at m/z 123 → 80 and m/z 123 → 53; nicotinic acid (NA) at m/z 124 → 80 and m/z 124 → 78. The ion source parameters were as follows: (1) gas temperature: 300°C; (2) gas flow: 5 L/min; (3) nebulizer: 45 psi; (4) sheath gas flow: 11 L/min; (5) capillary voltage: 3500 V (positive) and 3500 V (negative); (6) nozzle voltage: 500 V (positive) and 500 V (negative). All metabolites were quantified according to the standard curve running at the same batch of samples.

The microbial DNA was extracted using Magnetic Soil and Stool DNA Kit (TIANGEN BIOTECH, DP712). The DNA degradation degree and potential contamination were monitored by 1% agarose gels. The DNA concentration was measured using Qubit® dsDNA Assay Kit on the Qubit® 2.0 Fluorometer (Life Technologies, CA, USA). One microgram DNA was used as input material for DNA sample preparation. Sequencing libraries were generated by the NEBNext® Ultra™ DNA Library Prep Kit for Illumina (NEB, USA) according to the manufacturer’s instructions. Index codes were subsequently added to assign sequences to each sample. The DNA sample was initially fragmented into approximately 350 bp length by sonication. Subsequently, the DNA fragments underwent end-polishing, A-tailing, and ligation with a full-length adaptor specifically designed for Illumina sequencing. PCR amplification was then conducted to enrich the DNA fragments. PCR products were purified using the AMPure XP system, and the libraries were assessed for size distribution using the Agilent 2100 Bioanalyzer and quantified using real-time PCR. The clustering of the index-coded samples was performed on a cBot Cluster Generation System according to the manufacturer’s instructions. Once cluster generation was complete, the library preparations were sequenced on an Illumina NovaSeq platform, which generated paired end (PE) reads.

The raw data underwent quality control using the Fastp software. The assembly analysis was conducted using the MEGAHIT Assembly software, with the following parameters: –k-min 35, –k-max 95, –k-step 20, –min-contig-len 500. Subsequently, the clean data was aligned to the contigs of each sample using the Bowtie2 software to identify PE reads that were not utilized. The comparison parameters used were: -I 200, -X 400. The unutilized reads from each sample were then combined for a mixed assembly, using the same assembly parameters as mentioned above. To predict Open Reading Frames (ORF), MetaGeneMark was applied to all contigs (≥500 bp). The predicted genes with a length less than 100nt were filtered out. CD-HIT software was used to remove the redundancy, so as to obtain a non-redundant initial gene catalog. By default, identity 95% and coverage 90% were used for clustering, and the longest sequence was selected as the representative sequence. Parameters: -c 0.95, -g 0, -AS 0.9, -g 1, -d 0. Then, Bowtie2 was used to compare the clean data of each sample to the initial gene catalog, and the number of reads that matched each gene was calculated. The comparison parameters used were as follows: –end-to-end, –sensitive, -I 200, -X 400. Genes with supporting reads ≤2 in each sample were filtered out to obtain a gene catalog (Unigenes) for subsequent analysis. Based on the number of reads and gene length compared, the abundance information of each gene in each sample was calculated.

To explore the taxonomy abundance information of each sample at each classification level (phylum, class, order, family, genus, and species), the Unigenes were compared with sequences of bacteria, fungi, archaea, and viruses extracted from the NCBI NR database via DIAMOND software and LCA algorithm of MEGAN software (blastp, evalue ≤1e-5). Relative abundance comparison, principal coordinate analysis (PCoA), cluster analysis, and Metastats multivariate analysis were performed based on the abundance tables at each classification level.

Five-to-six weeks old male C57BL/6J mice fed a HFD for 9 weeks were set as recipients. The recipients were administered with an antibiotic cocktail consisting of 1 mg/mL neomycin (MCE, HY-B0470), 1 mg/mL streptomycin (MCE, HY-B1906) and 1 mg/mL bacitracin (APExBIO, B1670) in their drinking water for 1 week prior to FMT. Mice from HFD and TRF groups were selected as the donors for FMT. The feces of the donors were collected in a sterile environment and stored at −80°C. The feces were then mixed with sterile phosphate-buffered saline (PBS) (100 mg feces/1 ml PBS). The mixture was vortexed for 10 min and centrifuged at 1000 rpm for 5 min at 4°C. The supernatant was administrated to recipients (200 µL per mice) by oral gavage 3 times a week for 4 weeks.

HepG2 cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) (Gibco, C11995500BT) supplemented with 10% fetal bovine serum (FBS) (Gibco 10,091,148). The cells were incubated at 37°C and with 5% carbon dioxide. N-Me-6-PY (MCE, HY-113432) and N-Me-4-PY (MCE, HY-113472) were dissolved in DMEM and stored at −20°C. Different concentrations of N-Me-6-PY, N-Me-4-PY or their vehicle were added to the cells 1 h prior to treatment. The cells were then treated with free fatty acids (FFA) consisting of oleic acid (OA) (Sigma-Aldrich 75,090) and palmitic acid (PA) (Solarbio, P9830) in a ratio of 2:1, at a total concentration of 0.6 mM for 24 h to establish an in vitro NAFLD model.

The cytotoxicity analysis was conducted by the CCK8 method (Beyotime, C0037). After HepG2 cells were seeded in a 96-well plate and cultured for 24 h, different concentrations of N-Me-6-PY or N-Me-4-PY were added and cultured for an additional 24 h. Then, 10 µL CCK8 reagent was added to each well and incubated for 1 h at 37°C. Following the incubation, the absorbance of each well was measured using a microplate reader at a wavelength of 450 nm. The median inhibitory concentration (IC50) was calculated by GraphPad Prism Version 9.0 software.

HepG2 cells seeded on a 24-well plate were fixed with 4% paraformaldehyde for 15 min at room temperature. After being rinsed by PBS, the cells were stained with filtered oil red O solution (Sigma-Aldrich, O0625) for 30 min at 37°C. Subsequently, the cells were destained with 60% isopropyl alcohol for 5 s. Nuclei were stained with hematoxylin (Solarbio, G4070) for 1 min. The oil red O positive areas, indicating lipid accumulation, were evaluated using ImageJ software. Intracellular TG and TC were measured following the manufacturer’s instructions (Nanjing Jiancheng Bioengineering Institute, A110-1-1, A111-1-1). In brief, cells were digested and resuspended in PBS. After adequate sonication, the levels of TG and TC were measured and normalized to the corresponding protein concentrations.

N-Me-PYs, consisting of N-Me-6-PY (150 µg/kg) and N-Me-4-PY (75 µg/kg) dissolved in sterile PBS, were administered via intraperitoneal injection. Hydralazine (HYD, Sigma-Aldrich, H1753), a potent AOX1 inhibitor, was also dissolved in sterile PBS and administered via oral gavage at a dose of 50 mg/kg.²⁵ Five-to-six weeks old male C57BL/6J mice were randomly divided into the following five groups for a 6-week intervention: (1) CD; (2) CD + N-Me-PYs; (3) HFD; (4) HFD + N-Me-PYs; (5) HFD + HYD. Body weight, food intake, plasma, liver, and epididymal adipose tissue samples were collected and processed as described above.

Proteins from cells and livers were extracted by RIPA buffer (Fdbio science, FD009) supplemented with protease inhibitor cocktail (CWBIO, CW2200S) and phosphatase inhibitor cocktails (CWBIO, CW2383S). Protein concentration was determined by BCA assay kit (Fdbio science, FD2001). A total of 40 µg protein was separated on 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) gels and further transferred on 0.2 µm polyvinylidene difluoride (PVDF) membrane. After being blocked by 5% skimmed milk for 1 h, the membrane was incubated with different primary antibodies overnight at 4°C. The primary antibodies against NNMT (15123–1-AP, 1:1000 dilution), AOX1 (19495–1-AP, 1:500 dilution), FASN (10624–2-AP, 1:1000 dilution), ACC1 (67373–1-Ig, 1:1000 dilution), FABP4 (12802–1-AP, 1:1000 dilution), FATP2 (14048–1-AP, 1:500 dilution) and β-actin (66009–1-Ig, 1:5000 dilution) were acquired from Proteintech, CD36 (YT5585, 1:1000 dilution) and SREBP-1c (YT6055, 1:1000 dilution) were obtained from Immunoway. The membrane was then rinsed with tris-buffered saline with 0.1% Tween-20 (TBS-T) and incubated with corresponding secondary antibodies (Proteintech, SA00001–1; Asbio, AS006; 1:5000 dilution) at room temperature for 1 h. The protein bands were visualized using electrochemiluminescence (ECL) reagents (Millipore, WBKLS0500) according to the manufacturer’s instructions. The relative band density was quantified by ImageJ software, with β-actin as an internal standard.

Total RNA was extracted using the AG RNAex Pro Reagent (Accurate Biotechnology, AG21102) following the manufacturer’s instructions. The concentration and purity of RNA were measured using a NanoDrop 2000 spectrophotometer. Subsequently, cDNA was obtained using the Evo M‐MLV RT kit (Accurate Biotechnology, AG11728). The primer sequences utilized in this study were obtained from PrimerBank, synthesized in GENEray Biotechnology, and listed in Table 1. qRT-PCR was conducted in a 10 μL reaction volume consisting of 1 μL cDNA, 0.5 μL specific forward/reverse primers, 5 μL SYBR Green Premix Pro Taq HS qPCR Kit (Accurate Biotechnology, AG11702), and 3 μL RNase-Free water on the LightCycler 480 instrument. The relative expression of different genes was calculated by the 2^−∆∆Ct method and normalized to GAPDH.

All data were presented as mean ± standard error of the mean (SEM). The normal distribution of the data was initially assessed by the Shapiro–Wilk test. Statistical comparisons between the two groups were conducted using the unpaired two-tailed Student’s t-test for normally distributed data, and the Welch’s t-test for non-normally distributed data. Differences among three or more groups were analyzed using one-way analysis of variance (ANOVA) with Fisher’s LSD post hoc test for normally distributed data, and the Welch’s ANOVA for non-normally distributed data. The non-parametric Spearman test was employed to evaluate correlations between quantitative data. Statistical analyses in this study were performed using GraphPad Prism Version 9.0 software, R Studio, and MetaboAnalyst 5.0. p value < 0.05 was considered statistically significant.

HFD was utilized to establish a mouse model of NAFLD, with standard CD serving as the control. After a duration of 6 weeks, mice fed a HFD were randomly divided into three groups – HFD ad libitum, TRF, or PF. The TRF regimen involved unrestricted food intake from 8:00 to 16:00 each day, with only water allowed during the remaining 16 h. The PF group consumed a similar amount of food per day as the TRF group, but without any restrictions on the time window for food intake (Figure 1(a)). During the 6-week intervention, TRF resulted in a 35% reduction in daily food consumption, accompanied by a 20% decrease in body weight (Figures 1(b–c) and S1A). Likewise, a comparable weight loss was observed in PF mice.

We then explored the impacts of different dietary patterns on liver steatosis. As expected, 12-week HFD led to apparent elevations in liver weight and liver index (Figures 1(d–e)), accompanied by the presence of yellow dot-like structures on the liver surface indicating lipid deposition (Figure 1(f)). Histological examination revealed augmented hepatic vacuolation, steatosis, as well as inflammatory cell infiltration (Figure 1(f)). Oil red O staining also confirmed a notable increase in lipid accumulation, and the NAS was markedly elevated (Figures 1(f–h)). Noteworthily, a six-week TRF resulted in an approximate 50% reduction in liver weight, liver index, and lipid deposition area in the liver (Figures 1(d–h)). The extent of steatosis, lobular inflammation, and hepatocyte ballooning, which constitute the three components of NAS, were all mitigated to some extent (Figure S1B), indicating a substantial improvement in NAFLD under the intervention of TRF. To further determine the impact of TRF on lipid accumulation, epididymal adipose tissue size was measured. Similarly, the enlargement of adipocytes induced by HFD was declined following TRF (Figure S1C), indicating that TRF ameliorated lipid overload in the organism. In contrast, although food intake restriction partially improved the aforementioned liver and adipose parameters, the protective effect of PF appeared to be weaker than that of TRF (Figures 1(d–h) and S1B-C). Supporting this, TRF exhibited a more pronounced effect, compared to PF, on modulating lipid homeostasis in both plasma and liver (Figures 1(i–j)). Moreover, while both TRF and PF displayed equivalent improvement on liver enzyme, as indicated by the decline of ALT (Figure 1(k)), TRF was sufficient to restore the hepatic GSH and MDA levels, which were unnoticeable in PF mice (Figure 1(l)). These data suggested a unique anti-oxidative effect of TRF. Taken together, our findings highly implicated that the alleviation of hepatic steatosis through TRF might not be solely attributed to decreased food intake and body weight.

Considering the robust association between NAFLD and organic metabolic dysfunction,^26,27 along with the proven preservation of metabolic health by TRF,²⁷ we conducted comprehensive untargeted metabolomic profiling of mice plasma samples to explore the metabolic alterations during dietary intervention. A total of 972 metabolites were identified in the analysis. OPLS-DA revealed an apparent disparity in the metabolite composition between TRF and HFD mice, which was less pronounced in PF versus HFD (Figure 2(a)). We then applied a combination of criteria including variable importance on projection (VIP) >1 and fold change (FC) >1.2 or <0.83 to identify the differential metabolites among various dietary patterns. Concretely, HFD led to an upregulation of 185 metabolites and the downregulation of 131 metabolites in comparison to CD (Figure 2(b)). However, upon dietary intervention with TRF, the plasma metabolites of HFD mice were significantly changed, with 126 metabolites elevated and 227 declined (Figure 2(c)). Of note, in spite of equal food intake, there remained 300 differential metabolites observed between TRF and PF mice (Figure 2(d)). The MSEA enrichment sets, which depicted the biologically meaningful contexts related to these altered metabolites,²⁸ were shown in Figure S2A-C, respectively. These data further supported that the metabolic modulatory effect of TRF may not merely rely on food intake reduction.

To investigate the particular regulatory effects of TRF on host metabolism, distinct from the effects of food restriction, we proceeded to identify the metabolites that were specifically influenced upon TRF intervention. As shown by the Venn diagrams, a total of 57 metabolites, including 8 potential beneficial (characterized as metabolites that were downregulated by HFD but restored by TRF rather than PF) and 49 potential harmful metabolites (opposite to beneficial metabolites), were searched out (Figures 2(e-f)). The relative abundances of these metabolites were depicted in Figure S2D. Using MSEA, these 57 differential metabolites were primarily involved in nicotinate (NA) and NAM metabolism, glyoxylate and dicarboxylate metabolism, arginine and proline metabolism, and aminoacyl-tRNA biosynthesis pathways (Figure 2(g)). Interestingly, the NA and NAM metabolism pathway ranked the top among them (Figure 2(g)). Consistent with this, the differential metabolites mediated by TRF were predominantly enriched in this pathway as well (Figure S2B). These findings highly announced the alteration of NA and NAM metabolism might be associated with the metabolic protective effect of TRF dietary pattern.

NA and NAM are two forms of vitamin B3 that act as precursors for the coenzymes nicotinamide adenine dinucleotide (NAD+) and nicotinamide adenine dinucleotide phosphate (NADP+), which were essential for maintaining energy and cellular homeostasis.²⁹ On the other hand, NAM can be converted to 1-MNA by nicotinamide N-methyltransferase (NNMT) and subsequently metabolized to N-Me-6-PY and N-Me-4-PY by AOX1, predominantly in the liver³⁰ (Figure 2(h)). Of note, the plasma levels of NA, NAM, N-Me-6-PY, and N-Me-4-PY were prominent increased in HFD mice, whereas NAD+ was almost completely exhausted (Figure 2(i)). Moreover, the conversions of 1-MNA to N-Me-6-PY and N-Me-4-PY in the liver were largely enhanced, as evidenced by the elevation of N-Me-6-PY/1-MNA and N-Me-4-PY/1-MNA ratios (Figures 2(j) and S2E). Supporting this, both the protein and gene expressions of hepatic AOX1, but not NNMT, were upregulated in HFD mice (Figures 2(k) and S2F). These observations demonstrated the augmentation of hepatic NAM catabolism in HFD mice. On the contrary, TRF intervention was capable of restoring the upregulation of hepatic AOX1 (Figures 2(k) and S2F) and the augmentation of hepatic NAM catabolism (Figures 2(j) and S2E) induced by HFD, thereby mitigating the overaccumulation of N-Me-6-PY and N-Me-4-PY in plasma (Figure 2(i)). While a mild ameliorating effect was also observed in PF mice, it seemed less pronounced compared to TRF (Figures 2(i–k) and S2E-F). In addition, both the plasma and the hepatic levels of NAD+ were largely increased upon TRF intervention (Figures 2(i) and S2E), suggesting an improvement of NAD+ metabolism. Intriguingly, the plasma levels of N-Me-6-PY and N-Me-4-PY presented dramatically positive correlations with various parameters that associated with hepatic steatosis and lipid metabolic disturbance, including body weight, oil red O positive area, plasma lipid parameters, ALT, and hepatic MDA, whereas displayed negative association with the hepatic GSH (Figure 2(l)), suggesting the potential pro-steatotic effect of N-Me-6-PY and N-Me-4-PY. Collectively, our findings indicated that TRF might alleviate NAFLD by suppressing the excessive accumulation of NAM metabolic end products, N-Me-6-PY and N-Me-4-PY.

We then utilized a human liver cell line HepG2 to clarify the impact of N-Me-6-PY and N-Me-4-PY on hepatosteatosis. CCK8 assay showed that the IC50 for N-Me-6-PY and N-Me-4-PY in HepG2 cells were 694.2 μM and 489.7 μM, respectively (Figure S3A). As previously reported, N-Me-6-PY is predominant in human and N-Me-4-PY in rodents,³¹ with the physiological plasma concentrations approximate to 1 μM and 0.5 μM, respectively.^31,32 Thus, a gradient dose of N-Me-6-PY and N-Me-4-PY ranging from 2.5-fold to 10-fold of their physiological concentrations was used for in vitro intervention. Following stimulation with FFA, both N-Me-6-PY and N-Me-4-PY dose-dependently increased the burden of intracellular lipid droplets (Figures 3(a–b)) and raised intracellular TC and TG levels (Figure S3B) in hepatocytes. Furthermore, the mRNA expression levels of genes associated with de novo lipogenesis (SREBP-1c, ACC1, FASN) and fatty acid uptake (CD36, FATP2, FABP4) in the cells were upregulated following pretreatment with N-Me-6-PY and N-Me-4-PY (Figure 3(c)), while genes related to fatty acid oxidation (PPARα, CPT1α, ACOX1) and inflammation (TNFα, IL-6, IL-1β) showed no obvious alteration (Figure S3C). These findings were consistent at the protein level as well (Figures 3(d) and S3D).

To further determine the pro-steatotic effect of N-Me-6-PY and N-Me-4-PY in vivo, five-to-six-week-old male C57BL/6J mice fed a CD or HFD were intraperitoneally injected with N-Me-PYs, a mixture of N-Me-6-PY (150 µg/kg) and N-Me-4-PY (75 µg/kg), or its vehicle for 6 weeks (Figure 3(e)). Moreover, HYD, a selective inhibitor of hepatic AOX1,^33,34 was utilized to validate the therapeutic potential of AOX1 inhibition on NAFLD (Figure 3(e)). The administration of N-Me-PYs showed no obvious influence on body weight and liver weight (Figures 3(f) and S3E), yet mildly decreased the food intake of HFD mice (Figure S3F). As expected, supplementation with N-Me-PYs prominently aggravated hepatic lipid deposition in mice fed either CD or HFD (Figure 3(g)). Of note, both plasma and hepatic levels of TG, TC, and LDL-c, as well as plasma ALT were significantly increased in CD mice following N-Me-PYs administration (Figures 3(h–i), S3G). Similarly, plasma TC, LDL-c, and hepatic TG, LDL-c were apparently increased in HFD mice treated with N-Me-PYs, whereas the plasma TG, ALT, and hepatic TC were unaltered (Figures 3(h–i), S3G). In addition, the epididymal adipose tissue weight and adipocyte size were uninfluenced upon N-Me-PYs treatment (Figure S3H-I). These findings demonstrated that N-Me-PYs aggravated hepatic steatosis and contributed to lipid metabolic disturbance. Conversely, HYD effectively decreased body weight and liver weight (Figures 3(f) and S3E), mitigated hepatic lipid deposition and lipid metabolic disorder (Figures 3(g–i)), and improved the condition of epididymal adipose tissue (Figure S3H-I) in HFD mice, suggesting that inhibition of hepatic AOX1 might be a potential therapeutic strategy for NAFLD. Altogether, these results demonstrated that overaccumulation of N-Me-6-PY or N-Me-4-PY promoted de novo lipogenesis and fatty acid uptake capability in hepatocytes, thus aggravating hepatic steatosis.

We further explored the mechanism by which TRF modulated AOX1 expression and reprogramed NAM metabolism. Previous studies have shown that TRF profoundly impacts gene expressions via realigning the circadian clock.^35,36 However, the transcriptome landscape verified that the rhythmicity of liver AOX1 expression was not significantly altered under TRF in mammals,³⁵ and proteome also exhibited relatively stable daily rhythms,³⁷ suggesting the alteration of AOX1 expression by TRF might be independent of rhythm resynchronization. Previous studies have indicated that TRF can remodel the gut microbiota on both composition and metabolic function,^38,39 thereby regulating host metabolic homeostasis. In addition, gut microbiota exhibits a modifiable effect on host gene regulation.^40–42 To this, we hypothesized that gut microbiota might mediate the modulatory effects of TRF on host NAM metabolism, thus ameliorating hepatic steatosis.

Metagenome sequencing was performed on cecal contents obtained from the aforementioned mice. In terms of bacterial diversity, the α-diversity was obviously declined in HFD mice, while both TRF and PF displayed a trend of improvement, although statistically insignificant (Figures 4(a–b)). In addition, the principal coordinates analysis (PCoA) analysis based on Bray-Curtis distance demonstrated obvious differences in β-diversity among groups (Figure 4(c)). We also compared the gut microbiota composition profiles at different taxonomic levels. At the class level, the most predominant classes were Clostridia, Bacteroidia, Erysipelotrichia, Verrucomicrobiae, and Bacilli, and their relative abundance significantly differed among the four groups (Figure 4(d)). At the family level, Lachnospiraceae, Muribaculaceae, Erysipelotrichaceae, Oscillospiraceae, and Akkermansiaceae accounted for the largest proportion. Notably, TRF enriched the abundance of Akkermansiaceae and Rikenellaceae (Figure 4(e)). Previous studies have demonstrated that genera from these two families are less abundant in individuals with obesity, diabetes, and NAFLD,^43,44 which is also associated with the improvement of HFD-induced metabolic disorders.^45,46

Next, we employed Metastats analysis⁴⁷ to identify the bacterial species potentially involved in N-Me-6-PY and N-Me-4-PY production, as well as NAFLD progression. As the overlapping process shown in Figure S4A, a total of 26 bacterial species were searched out. The relative abundances of these species among the four groups were depicted in Figure S4B. Indeed, species such as Lactobacillus iners, Salipaludibacillus neizhouensis, Alteribacillus sp YIM-98480, uncultured Eubacterium sp, and Lachnospiraceae bacterium 14–2, which were downregulated following TRF, displayed apparently positive correlation with hepatic steatosis and lipid metabolic disturbance (Figure 4(f)). Conversely, Bacteroidaceae bacterium, Candidatus Curtissbacteria bacterium GW2011-GWC2-38-9, Alkalihalobacillus alcalophilus, Ruminococcus sp AM28-29LB, and Kosmotoga pacifica showed negative correlations with these indices. These species were depleted upon HFD, whereas they were restored by TRF intervention (Figure 4(f)). Importantly, Ruminococcus sp AM28-29LB and Lactobacillus iners exhibited the strongest correlation with the plasma levels of NAM catabolites, especially N-Me-6-PY, N-Me-4-PY, and NAD+, as well as the hepatic conversion ratios from 1-MNA to N-Me-6-PY and N-Me-4-PY (Figure 4(f)), suggested these bacterial species might potentially influence the hepatic NAM catabolism and the progression of NAFLD. Taken together, our findings implicated that the alteration of gut microbiota might mediate the regulatory effect of TRF on hepatic NAM catabolism, thereby ameliorating hepatic steatosis.

To probe into the causal relationship between TRF-mediated gut microbiota alteration and the improvement of NAFLD, an FMT experiment was conducted on HFD-induced obese mice. After a week of gut microbiota depletion using an antibiotic cocktail, mice were orally administered fecal bacterial suspensions from either HFD or TRF mice for another four consecutive weeks (three times per week) (Figure 5(a)). Consequently, mice transplanted with fecal microbiota from TRF donors exhibited a prominent decrease in body weight gain compared to those receiving FMT from HFD donors (Figure 5(b)). Surprisingly, TRF-dependent FMT led to pronounced reductions in liver weight, liver index, area of lipid deposition, and NAS (Figures 5(c–g) and S5A), along with the size of epididymal adipocytes (Figure S5B). In parallel, plasma lipids, liver enzymes, liver homogenate lipids, and hepatic oxidative stress levels were effectively improved after TRF-dependent FMT as well (Figures 5(h–k)). These findings further confirmed that gut microbiota, at least partially, mediates the protective effect of TRF pattern on hepatic steatosis and lipid metabolic disorder.

Finally, we evaluated whether FMT can transfer the modulatory effect of TRF on NAM catabolism. Interestingly, mice received TRF fecal microbiota showed reductions of N-Me-6-PY and N-Me-4-PY in plasma by 43% and 32%, respectively (Figure 5(l)). Additionally, the protein expression levels of hepatic NNMT and AOX1 were declined by 41% and 38% (Figure 5(m)). Consistently, the gene expression level of hepatic AOX1 was also remarkably decreased (Figure S5C). Aligned with the in vitro findings, the reduction of N-Me-6-PY and N-Me-4-PY contributed to dramatic improvement in de novo lipogenesis and fatty acid uptake in the liver, as evidenced by the downregulation of FASN, ACC1, SREBP-1c, as well as CD36, FATP2, and FABP4 at both protein and gene expression levels (Figures 5(n) and S5D). Therefore, we concluded that gut microbiota mediated the metabolic modulatory effect of TRF, which suppressed the hyperactivation of NAM catabolism, and stabilized hepatic lipid metabolism function, thereby alleviating NAFLD.

The untargeted plasma metabolomic and gut microbial metagenome sequencing data are both available on the Dryad database under doi:10.5061/dryad.83bk3j9zz.