Graphical abstract

   graphic file with name fx1.jpg
   [43]Open in a new tab

Highlights

     * •
       Hepatic metabolism is affected differently by different weight loss
       interventions
     * •
       Intermittent fasting results in increased lipogenesis and affects
       multiple pathways
     * •
       Bariatric surgery uniquely affects one-carbon metabolic pathways
       and hepatic redox
     * •
       Transmethylation cycle may be a promising target for treatment of
       metabolic disease
     __________________________________________________________________

   Surgery; Omics

Introduction

   The worldwide prevalence of obesity is steadily increasing. In 2015, an
   estimated 12% of adults and 5% of children worldwide had obesity.
   Excess body weight accounted for about 4 million deaths and 120 million
   disability-adjusted life-years, most of these because of cardiovascular
   disease and diabetes.[44]^1 Weight loss can prevent many of the adverse
   consequences of obesity, and can be achieved through either dietary and
   lifestyle interventions, pharmacotherapy, or bariatric surgery.[45]^2
   Some of these interventions have been purported to have benefits beyond
   those conferred by weight loss alone. For example, with regards to
   mortality and cardiovascular adverse events, the benefits of bariatric
   surgery were evident even with minimal weight loss, and greater degrees
   of weight loss were required to achieve the same effects through
   dietary intervention alone.[46]^3^,[47]^4 Likewise, an intermittent
   fasting (IF) regimen resulted in improved insulin sensitivity when
   compared to a similar degree of weight loss achieved through continuous
   energy restriction diets, both in humans and in animal
   models,[48]^5^,[49]^6^,[50]^7 in addition to reported benefits for
   unrelated morbidities such as autoimmune disease, inflammatory
   disorders, and cancer.[51]^8^,[52]^9^,[53]^10^,[54]^11 The mechanisms
   underlying these beneficial effects remain largely unknown. Numerous
   unique metabolic alterations have been described with both bariatric
   surgery and IF. These include changes in serum amino acid and bile acid
   concentrations,[55]^12^,[56]^13 in the composition of the gut
   microbiome,[57]^14^,[58]^15 and in intestinal lipid handling.[59]^16
   The clinical significance of some of these alterations remains
   unclear.[60]^17^,[61]^18 We and others have previously demonstrated a
   specific and weight loss-independent effect of surgery on hepatic
   insulin sensitivity,[62]^19^,[63]^20^,[64]^21 suggesting that changes
   in liver metabolism may explain some of the positive effects of
   surgery. Some effect of bariatric surgery on hepatic one-carbon
   metabolism, although not well characterized, has been identified in
   several human and animal studies.[65]^22^,[66]^23^,[67]^24^,[68]^25

   As the central organ for fatty acid (FA) metabolism,[69]^26 the liver
   is also one of the primary organs affected by obesity. Excessive weight
   is often associated with an alteration of hepatic FA metabolism in
   which the liver’s capacity to handle FAs and other metabolic energy
   substrates is overwhelmed,[70]^27 leading to an accumulation of lipids
   within hepatocytes, a condition known as non-alcoholic fatty liver
   (NAFL) or hepatic steatosis.[71]^28^,[72]^29 In the present work, we
   seek to describe the changes in hepatic metabolic metabolism brought
   about by weight loss, and whether different interventions differ in
   this regard. We analyzed the metabolome of the hepatic blood supply and
   the metabolome and transcriptome of the liver itself following sleeve
   gastrectomy (SG), a common bariatric surgery, and IF with caloric
   restriction (IF-CR), in rats fed a high-fat, high-sucrose diet (HFHSD).
   Both interventions, when compared to ad libitum (AL)-fed animals, led
   to a similar loss of weight but resulted in distinct and in some cases
   contradictory effects on several metabolic pathways, when compared to
   an ad libitum-fed control group. The effect of bariatric surgery is
   apparent chiefly within pathways related to methionine and methyl group
   metabolism. In contrast, IF-CR has a profound effect on fatty acid
   metabolism and synthesis, which remain relatively unaffected by SG, as
   compared to ad-libitum fed rats.

Results

   At the conclusion of 12 weeks of HFHSD, animals attained an average
   weight of 456.3 g (+/− 14.4 g) in the SG group, 464.7 g (+/− 26.6 g) in
   the IF-CR group, and 492.8 g (+/−39.2 g) in the AL group
   ([73]Figure 1B). Morning non-fasting serum glucose levels were
   unaltered between the experimental groups ([74]Figure 1C). Following
   oral glucose loading, the IF-CR group showed improved glucose tolerance
   when compared to the SG group ([75]Figure 1D), as manifested in a
   decreased glucose area under the curve.

Figure 1.

   [76]Figure 1
   [77]Open in a new tab

   Overview and liver histology

   (A) Experimental design. Animals were fed a high-fat, high-sucrose diet
   (HFHSD) for 7 weeks, and then underwent either sleeve gastrectomy (SG),
   sham surgery followed by intermittent fasting and caloric restriction
   (IF-CR), or no intervention (AL). At day 80, animals were sacrificed
   following a 10-h fast, and blood and tissue collected for analysis.

   (B) Body weight of animals in the 3 experimental groups (mean ± SD).

   (C) Morning non-fasting serum glucose measurements carried out in the
   three experimental groups.

   (D) Oral glucose tolerance tests carried out at week 10 in the SG and
   IF-CR groups (mean ± SD). Left: Glucose measurements at five time
   points following oral gavage. Right: area under the curve (AUC; mean ±
   SD).

   (E) Top: hematoxylin and eosin-stained liver sections, 10X
   magnification. Bottom: periodic acid-Schiff-stained liver sections, 10X
   magnification.

   (F) Glycogen content in liver homogenates (mean ± SD).

   ∗p < 0.05; ∗∗p < 0.01 (ANOVA with Tukey post-hoc test).

   Livers were examined histologically, which revealed a relative
   abundance of glycogen in the IF-CR group ([78]Figure 1E). Other than
   this finding, which was corroborated by a biochemical glycogen assay
   ([79]Figure 1F), there were no significant differences in liver
   histology, which appeared normal.

Metabolomic analysis of serum entering the liver

   In analysis of the metabolic composition of the hepatic input, 87 of
   197 metabolites were identified as altered between the experimental
   groups ([80]Figures 2A–2C). PLS-DA showed separation of the IF-CR group
   from the SG and AL groups along the first component, and of the SG
   group from the two other groups primarily along the second component
   ([81]Figure 2D). Pathway enrichment analysis showed component one to be
   enriched for fatty acid beta oxidation and biosynthesis, linolenic and
   linoleic acid metabolism, and ketone body metabolism ([82]Figure 2E).
   Component two was enriched for glycine and serine metabolism (with
   relevant metabolites including methionine, betaine, creatine, glycine,
   guanidinoacetic acid, serine, glutamate, threonine, and
   alpha-ketoglutarate), glutathione metabolism, carnitine synthesis, and
   methionine metabolism ([83]Figure 2F).

Figure 2.

   [84]Figure 2
   [85]Open in a new tab

   Metabolomic analysis of blood entering the liver

   (A–C) Volcano plots showing the relative changes in metabolite
   concentrations in the hepatic input (80% portal blood, 20% aortic
   blood) between the experimental groups (SG: sleeve gastrectomy; IF-CR:
   intermittent fasting caloric restriction; AL: ad libitum; FDR = false
   discovery rate-adjusted p value of between-group t-tests; PUFA:
   poly-unsaturated long-chain free fatty acids; SFA/MUFA:
   saturated/mono-unsaturated long-chain free fatty acids; SCFA/MCFA:
   short-/medium-chain free fatty acids).

   (D) Partial least squares-discriminant analysis (PLS-DA) scores plot of
   metabolite abundance demonstrating separation of the IF-CR group from
   the AL group along the first component and of the SG group from the AL
   group along the second component.

   (E and F) Enrichment analysis by small molecule pathway database
   (SBPDB) pathways of metabolites in components 1 (E) and 2 (F) of
   PLS-DA. Metabolites were defined as belonging to component one, two, or
   neither, based on their VIP score for that component being above 1.0
   and greater than the VIP score for the other component.

   (G) Heatmap of normalized concentrations of 87 metabolites identified
   as significantly altered between the experimental groups by ANOVA with
   FDR of <0.05.

   Within the differentially regulated metabolites in the hepatic input,
   several patterns of between-group differences were seen ([86]Figure 2G,
   [87]Table S1). One group of compounds was similarly affected by both SG
   and IF-CR. Levels of serum kynurenine, a tryptophan breakdown product,
   of serum creatine, and of glutathione and its precursors
   γ-glutamyl-cysteine and cystathionine, were all reduced in both weight
   loss intervention groups. In contrast, levels of the micronutrients
   thiamine and ascorbic acid (vitamins B1 and C), and of the amino acids
   methionine, threonine, and serine, were all lower in the AL-fed group.

   Another group of metabolites was upregulated in the SG group and
   downregulated in the IF-CR group. Among these were numerous metabolites
   related to one-carbon metabolism and methyl transfer reactions,
   including betaine, dimethylglycine, sarcosine, and glycine. Other
   metabolites showing a similar pattern were the pyrimidine nucleotides
   (cytidine and thymidine) and the polyunsaturated ω-3 long-chain fatty
   acids (LCFAs).

   Metabolites downregulated exclusively in the IF-CR group include the
   ketone bodies acetoacetate (AcAc) and beta-hydroxybutyrate (beta-HB)
   and certain saturated and mono-unsaturated LCFAs. Notable compounds
   upregulated with IF-CR include the gluconeogenic precursors alanine and
   glycerate, proline, the branched-chain amino acids valine, leucine, and
   isoleucine, and gut microbiota-derived metabolites, including short-
   and medium-chain free fatty acids (butyric, caproic, and caprylic
   acids), and the tryptophan breakdown product indole 3-acetate (IAA),
   which functions as a central mediator in gut microbiota-host
   interactions.[88]^30^,[89]^31

   Looking separately at each component of the hepatic input, i.e., aortic
   and portal blood, we find that the two can be readily distinguished by
   their metabolic composition ([90]Figures 3A–3C). PLS-DA with separate
   analysis of each blood pool from the three experimental groups showed
   clear separation of the IF-CR group from the two other groups, as well
   as clear separation of aortic and portal blood. However, there was
   considerable overlap between aortic samples from the SG and AL groups,
   as well as between portal samples from these two groups
   ([91]Figure 3D). Correlation analysis using data from all three
   experimental groups revealed several metabolites whose concentrations
   were either higher or lower in the portal versus aortic blood. Notably,
   ratios of metabolite concentrations between aortic and portal blood
   were consistent across the experimental groups, and no statistically
   significant differences were seen in these ratios for all metabolites
   examined, despite numerous between-group differences in absolute
   metabolite concentrations, as detailed above. Portal blood from all
   groups was characterized by higher concentrations of asparagine,
   arginine, histidine, alanine, glycine, and proline, and by lower
   concentrations of glucose, glutamine, and creatinine, when compared to
   aortic blood. Levels of choline and, predictably, bile salts, were also
   higher in the portal blood.

Figure 3.

   [92]Figure 3
   [93]Open in a new tab

   Metabolomic comparison of aortic and portal blood

   (A–C) Volcano plots showing the relative concentrations of metabolites
   in aortic vs. portal blood in each experimental gorup (SG: sleeve
   gastrectomy; IF-CR: intermittent fasting caloric restriction; AL: ad
   libitum; FDR = false discovery rate-adjusted p value of between-group
   t-tests).

   (D) Partial least squares-discriminant analysis (PLS-DA) scores plot of
   metabolite abundance demonstrating separation of the IF-CR group from
   the AL group along the first component and of aortic from portal
   samples along the second component.

   (E and F) Enrichment analysis by small molecule pathway database
   (SBPDB) pathways of metabolites in components 1 (E) and 2 (F) of
   PLS-DA. Metabolites were defined as belonging to component one, two, or
   neither, based on their VIP score for that component being above 1.0
   and greater than the VIP score for the other component.

   (G) Heatmap of the log2(portal to aortic ratio) of 58 metabolites
   identified as significantly correlated with aortic or portal source
   across experimental groups (FDR <0.1).

Metabolomic analysis of liver

   In the liver itself, significant changes were seen in the
   concentrations of 51 of 241 metabolites across the experimental groups
   ([94]Figures 4A–4C). PLS-DA showed separation of the IF-CR group from
   the SG and AL groups along the first component, and separation of the
   SG group from the two other groups along the second component
   ([95]Figure 4D). Pathway enrichment showed component one to be enriched
   for glutamate metabolism, fatty acid and plasmalogen biosynthesis,
   purine metabolism, and carnitine synthesis. Component two was enriched
   for glycine and serine metabolism as well as glycolysis/gluconeogenesis
   and pentose phosphate pathway-related metabolites and ammonia recycling
   ([96]Figures 4E and 4F).

Figure 4.

   [97]Figure 4
   [98]Open in a new tab

   Metabolomic analysis of liver

   (A–C) Volcano plots showing the relative changes in metabolite
   concentrations in the liver between the experimental groups (SG: sleeve
   gastrectomy; IF-CR: intermittent fasting caloric restriction; AL: ad
   libitum; FDR = false discovery rate-adjusted p value of between-group
   t-tests; LCFA: long-chain free fatty acids; SCFA/MCFA:
   short-/medium-chain free fatty acids).

   (D) Partial least squares-discriminant analysis (PLS-DA) scores plot of
   metabolite abundance demonstrating separation of the IF-CR group from
   the AL group along the first component and of the SG group from the AL
   group along the second component.

   (E and F) Enrichment analysis by small molecule pathway database
   (SBPDB) pathways of metabolites in components 1 (E) and 2 (F) of
   PLS-DA. Metabolites were defined as belonging to component one, two, or
   neither, based on their VIP score for that component being above 1.0
   and greater than the VIP score for the other component.

   (G) Heatmap of normalized concentrations of 51 metabolites identified
   as significantly altered between the experimental groups by ANOVA with
   FDR of <0.05.

   As in the serum, several metabolites showed opposite trends in the SG
   and IF-CR groups when compared to the AL group ([99]Figure 4G,
   [100]Table S2). Compounds whose concentrations were increased in the SG
   groups and decreased in the IF-CR group included one-carbon
   metabolism-related metabolites and the biogenic amines serotonin and
   histamine. Showing an opposite pattern were glutathione and the
   phosphorylated sugars glucose and fructose 6-phosphate. Other
   metabolites were exclusively affected in the IF-CR group. Consistent
   with their plasma concentrations, alanine, proline, and isoleucine, and
   the short-chain free fatty acids (butyric and hexanoic acid), were all
   increased in the IF-CR group. In contrast, the LCFAs, the majority of
   which showed decreased serum concentrations in the IF-CR group, as
   mentioned above, were seen to be increased intrahepatically in this
   group in comparison to the two others.

Transcriptomic analysis of liver

   2645 differentially expressed genes were identified, of which 1222 were
   upregulated and 819 were downregulated in the IF-CR group, while 636
   were upregulated and 902 were downregulated in the SG group compared to
   the AL group ([101]Figures 5A–5C). Genes upregulated in the IF-CR group
   were found to be enriched for citrate cycle, pyruvate metabolism,
   glycolysis/gluconeogenesis, pentose phosphate pathway, glycine, serine,
   and threonine metabolism, and fatty acid biosynthesis, whereas
   downregulated pathways included fatty acid degradation and linoleic
   acid metabolism ([102]Figure 5D). For the SG group, prominent
   downregulated pathways were those of glycine, serine, and threonine
   metabolism, cysteine and methionine metabolism, citrate cycle, and
   pyruvate metabolism, whereas fatty acid degradation, linoleic acid
   metabolism, and tryptophan and nicotinamide metabolism were upregulated
   ([103]Figure 5E). PCA showed distinct separation of the IF-CR group
   from the two other groups along the first principal component, and
   lesser separation of the SG group from both other groups along the
   second principal component ([104]Figure 5F). A joint pathway analysis
   incorporating both significantly altered metabolites and genes
   ([105]Figure 5G) showed glycolysis/gluconeogenesis, citric acid cycle,
   and pyruvate metabolism to be significantly altered in the IF-CR
   compared to the AL group, and glycine, serine, and threonine metabolism
   and cysteine and methionine metabolism as significantly altered in the
   SG compared to the AL group.

Figure 5.

   [106]Figure 5
   [107]Open in a new tab

   Transcriptomic analysis of liver

   (A–C) Volcano plots showing the relative changes in transcript
   abundance in the liver between the experimental groups (SG: sleeve
   gastrectomy; IF-CR: intermittent fasting caloric restriction; AL: ad
   libitum; FDR = false discovery rate-adjusted p value of between-group
   t-tests). Red denotes differentially expressed transcripts.

   (D and E) Enrichment analysis by Kyoto Encyclopedia of Genes and
   Genomes (KEGG) pathways of genes identified as upregulated (top) or
   downregulated (bottom) in the IF-CR (D) and SG (E) groups; genes were
   defined as up- or down-regulated if identified by ANOVA (FDR <0.05) and
   post-hoc analysis as significantly altered against at least one of the
   other experimental groups, and with either a negative or positive
   fold-change compared to both other groups.

   (F) Principal component analysis (PCA) scores plot of transcript
   abundance demonstrating separation of the IF-CR group from the AL group
   along the first component and of the SG group from the AL group along
   the second component.

   (G) Joint pathway enrichment analysis by KEGG pathways of genes and
   metabolites identified as significantly altered by ANOVA (FDR <0.05)
   and post-hoc analysis between the IF-CR and AL groups (left) and
   between the SG and AL groups (right).

One-carbon metabolism: Decreased utilization of one-carbon donors following
SG

   Transfer of methyl groups is centered around the transmethylation
   cycle, which involves conversion of methionine to S-adenosylmethionine
   (SAM), the universal methyl donor, and then to homocysteine. Methionine
   is then regenerated in a reaction utilizing either betaine or
   tetrahydrofolate as the source of a one-carbon unit. As described
   above, the most prominent effects of SG were seen in metabolites and
   transcripts related to one-carbon metabolic pathways. Levels of the
   major one-carbon donors, including betaine, methionine, serine,
   glycine, sarcosine, and dimethylglycine (DMG), were increased in both
   the serum and intrahepatically. With IF-CR, decreased levels of
   betaine, DMG, sarcosine, and glycine, which are intermediates in the
   sequential breakdown of choline to glycine, were seen in the serum
   ([108]Figure 6A). Levels of methyl-utilizing enzymes involved in the
   transfer of one-carbon groups for regeneration of SAM, including
   betaine homocysteine S-methyltransferase (BHMT), dimethylglycine
   dehydrogenase (DMGDH), mitochondrial serine hydroxymethyltransferase
   (SHMT2), and glycine decarboxylase (GLDC), were all decreased in the SG
   group compared to the AL and IF-CR group ([109]Figure 6B), suggesting
   that reduced consumption of one-carbon groups through the
   transmethylation cycle underlies the observed increase in potential
   methyl donors.

Figure 6.

   [110]Figure 6
   [111]Open in a new tab

   One-carbon metabolism

   (A and B) Heatmap of concentrations (area under the curve [AUC]
   normalized to total ion count [TIC]) of metabolites related to
   one-carbon transfer in the serum (hepatic input, A) and liver (B) of
   the SG and IF-CR groups, expressed as log[2] of the fold-change versus
   the AL group. (DMG: dimethylglycine; Sarc: sarcosine; Met: methionine;
   GAA: guanidinoacetate; pEtN: phosphoethanolamine; Hcy: homocysteine;
   CDP-EtN: cytidine diphosphate-ethanolamine; CDP-Cho: cytidine
   diphosphate-choline; GPC: glycerophosphocholine; SAM:
   S-adenosylmethionine; SAH: S-adenosylhomocysteine; pCho:
   phosphorylcholine; Cys: cystathionine; GGC: gamma-glutamylcysteine;
   GSH: reduced glutathione).

   (C) Ratios in liver of methylated phosphatidylcholine (PC) to
   unmethylated phosphatidylethanolamine (PE), methylated creatine to
   unmethylated GAA, and SAM to SAH, in the three experimental groups
   (concentrations in nmol/mg for PC and PE, and AUC normalized to TIC for
   creatine, GAA, SAM, and SAH, expressed as standardized z-scores).

   (D) Liver concentrations of reduced glutathione (GSH, AUC normalized to
   TIC) and ratios of GSH to GSSG (expressed as standardized z-scores).

   (E) Ratios of hepatic pyruvate to lactate (Pyr/Lac, left), pyruvate to
   malate (Pyr/Mal, middle), and acetoacetate to beta-hydroxybutyrate
   (AcAc/beta-HB, right), reflecting cytosolic NAD+/NADH, cytosolic
   NADP+/NADPH, and mitochondrial NAD+/NADH ratios, respectively (AUC
   normalized to TIC of each compound was divided by that of the other,
   and expressed as standardized z-scores).

   (F) Heatmaps of relative transcript abundance of significantly altered
   genes (ANOVA FDR <0.05) related to methyl (CH[3]) group transfer, PC
   and PE synthesis, GSH synthesis, and PC and bile acid (BA) export.
   Expressed as log[2] of the fold-change versus the AL group.

   (G) Plot summarizing changes in one-carbon metabolic pathways. In red,
   metabolites or transcripts upregulated in the SG group. In blue,
   metabolites or transcripts down regulated in the SG group.

   ∗p < 0.05 ANOVA with Tukey post-hoc test.

   ∗∗p < 0.01 ANOVA with Tukey post-hoc test.

   Levels of SAM and SAM/S-adenosylhomocysteine (SAH) ratios were not
   significantly altered between the experimental groups ([112]Figure 6C).
   However, changes in the ratios of methylated products to their
   unmethylated precursors could account for the apparent changes in
   methyl donor utilization. Conversion of the phospholipid
   phosphoethanolamine (PE) to phosphatidylcholine (PC), and the
   methylation of guanidinoacetic acid (GAA) to form creatine, are the two
   major methyl-consuming reactions in the liver and the body as a
   whole.[113]^32^,[114]^33 The ratio of creatine to GAA was reduced
   post-SG when compared to both other groups. In contrast, the
   intrahepatic PC/PE ratio was increased in the IF-CR group
   ([115]Figure 6C). Levels of choline kinase (CHKA), which catalyzes the
   rate-limiting step in the synthesis of PC from choline (Kennedy
   pathway), were reduced post-SG ([116]Figure 6F), suggesting that the
   apparent reduced consumption of methyl groups is not because of
   increased synthesis of PC via this pathway. Levels of ABCB4/MDR3, which
   is responsible for secretion of PC into bile,[117]^34 were decreased in
   both the SG and IF-CR groups when compared to the AL group.

Redox balance: Decreased hepatic glutathione concentrations following SG

   Homocysteine generated in the transmethylation cycle serves as a
   precursor for the synthesis of glutathione, the major hepatic
   antioxidant. Consistent with decreased transmethylation activity, we
   found decreased levels of glutathione in the post-SG livers, as well as
   a decreased ratio of reduced to oxidized glutathione (GSH/GSSG;
   [118]Figure 6D). This finding contrasts with an increased expression of
   subunits of the rate-limiting enzyme in GSH biosynthesis,
   glutamate-cysteine ligase (GCLC and GCLM; [119]Figure 6F), suggesting
   upstream substrate availability as the cause of decreased GSH post-SG.
   Additional evidence for a more oxidized metabolic state post-surgery
   was seen with an increased ratio of AcAc to beta-HB, which provides an
   estimate of the mitochondrial ratio of oxidized to reduced nicotinamide
   adenine dinucleotide (NAD+/NADH),[120]^35 as well as an increase in the
   ratios of pyruvate to malate and lactate, which provide an estimate of
   the cytosolic ratios of oxidized to reduced nicotinamide adenine
   dinucleotide phosphate (NADP+/NADPH) and NAD+/NADH, respectively
   ([121]Figure 6E). This effect was particularly prominent in comparing
   the SG to the AL group, and less so with respect to the IF-CR group.
   Increased glutathione oxidation is linked to attenuated endoplasmic
   reticulum (ER) stress,[122]^36 and we found decreased expression of
   unfolded protein response genes (HSPA5/BIP, EIF2AK3/PERK), which are
   markers of ER stress, in the SG group ([123]Figure 6F). The changes in
   one-carbon and redox pathways following SG are summarized in
   [124]Figure 6G.

Fatty acid metabolism: Lipogenic pathways increased by IF-CR

   SG had little effect on lipid metabolism, with the exception of an
   increase in the transcript levels of key enzymes of fatty acid
   beta-oxidation, including carnitine palmitoyl transferase 1 (CPT1) and
   acyl-CoA oxidase 1 (ACOX1), which catalyze the rate-limiting steps in
   mitochondrial and peroxisomal β-oxidation,
   respectively.[125]^37^,[126]^38 Both enzymes were upregulated by
   approximately 2-fold in the SG group when compared to the AL-fed
   animals. In contrast, these two enzymes were downregulated in the IF-CR
   group when compared to the AL group, in addition to numerous other
   changes relating to lipid metabolism. We witnessed in the IF-CR group a
   decreased concentration of long-chain FAs (LCFAs) in blood entering the
   liver but an increased concentration of intrahepatic LCFAs
   ([127]Figure 7A) when compared to both the SG and AL groups. An almost
   3-fold increase in serum triglycerides (TGs) was seen in the IF-CR
   group in comparison to the SG group ([128]Figure 7B), as well as a
   decrease in serum ketone bodies (AcAc and beta-HB, [129]Figure 7C).
   These changes were paralleled by a dramatic increase in the expression
   of enzymes involved in de novo lipogenesis (DNL) and in TG mobilization
   within the liver in the IF-CR group. In particular, the expression of
   rate-limiting enzymes of DNL, acetyl-CoA carboxylase (ACACA), fatty
   acid synthetase (FASN), acetyl-CoA citrate lyase (ACLY), and
   stearoyl-CoA desaturase (SCD),[130]^39 was increased between 8- and
   50-fold in the IF-CR group compared to the AL group ([131]Figure 7D).

Figure 7.

   [132]Figure 7
   [133]Open in a new tab

   Fatty acid metabolism

   (A) Heatmap of concentrations of saturated and mono-unsaturated
   long-chain free fatty acids (SFA/MUFA), poly-unsaturated long-chain
   free fatty acids (PUFA), and short- and medium-chain free fatty acids
   (SCFA/MCFA) in the serum (hepatic input) and liver of the SG and IF-CR
   groups, expressed as log[2] of the fold-change versus the AL group. (∗
   - ANOVA FDR <0.05 in serum or liver).

   (B) Serum lipid profile showing increased triglyceride (TG)
   concentrations in the IF-CR group, as well as decreased high-density
   lipoprotein (HDL) cholesterol levels in the SG group and no change in
   low-density lipoprotein (LDL) cholesterol levels.

   (C) Serum concentrations (area under the curve [AUC] normalized to
   total ion count [TIC]) of the major ketone bodies, acetoacetate (AcAc)
   and beta-hydroxybutyrate (beta-HB), showing a decrease in the IF-CR
   group.

   (E) Heatmaps of relative transcript abundance of significantly altered
   genes (ANOVA FDR <0.05) in pathways of fatty acid (FA) oxidation and
   synthesis, glycolysis and gluconeogenesis, pentose phosphate pathway
   (PPP), and tricarboxylic acid (TCA) cycle, expressed as log[2] of the
   fold-change versus the AL group, and showing upregulation of FA
   synthesis in the IF-CR group.

   (F) Liver concentrations of glycolytic/gluconeogenic intermediates (AUC
   normalized to TIC, expressed as standardized z-scores) (G6P: glucose
   6-phosphate; F6P: fructose 6-phosphate; GA3P: glyceraldehyde
   3-phosphate; DHAP: dihydroxyacetone phosphate; PEP: phosphoenol
   pyruvate; Pyr: pyruvate).

   (G) plot summarizing changes in fatty acid metabolism pathways,
   glycolysis and gluconeogenesis, pentose phosphate pathway, and TCA
   cycle. In red, metabolites or transcripts upregulated in the IF-CR
   group. In blue, metabolites or transcripts down regulated in the IF-CR
   group. In purple, long-chain FAs – downregulated in serum and
   upregulated in the liver of the IF-CR group.

   ∗p < 0.05 ANOVA with Tukey post-hoc test.

   ∗∗p < 0.01 ANOVA with Tukey post-hoc test.

   In parallel with an increase in fatty acid synthetic pathways, we
   observed in the IF-CR group an increased expression of glycolytic,
   pentose phosphate pathway, and citrate cycle enzymes ([134]Figure 5E),
   and increased levels of glycolytic intermediates in the liver
   ([135]Figure 7F). The changes related to fatty acid metabolism
   following IF-CR are summarized in [136]Figure 7G.

Discussion

   In this study, we sought to compare the changes in liver metabolism
   following weight loss induced by either SG or IF-CR in a rat model,
   using combined metabolomic and transcriptomic approaches. We chose an
   experimental design in which animals were fed a moderately high-fat
   high-sucrose diet for a relatively short period. This design was meant
   to examine the effects of SG and IF-CR but separated from the metabolic
   disturbances wrought by frank liver steatosis or the development of
   impaired glucose tolerance seen with prolonged high-fat feeding. Thus,
   both interventions resulted in significant, though modest, weight loss
   when compared to AL-fed animals.

   Importantly, each of the interventions had a distinct effect on hepatic
   metabolism. The metabolic changes following SG concentrated primarily
   on one-carbon metabolic pathways. The levels of most major methyl
   donors in the blood entering the liver were increased post-SG. The
   levels of methyl acceptors were either markedly increased, as in the
   case of GAA, the precursor to creatine, or unchanged, e.g., in the case
   of PE and PC. Contrasting with the increase in these substrates,
   quantities of transcripts encoding for key enzymes within the
   transmethylation cycle were decreased. Levels of glutathione, generated
   downstream of the transmethylation cycle, were also reduced.

   The increased levels of both methyl donors and methyl acceptors suggest
   altered regulation at the entry point of the transmethylation cycle.
   Decreased glutathione and decreased allocation of methyl groups to
   creatine synthesis are features commonly observed in methionine
   restricted diets.[137]^40^,[138]^41^,[139]^42^,[140]^43 Despite an
   abundance of methionine and other methyl donors, these features were
   seen in our study post-SG. Methionine adenosyltransferase 1A (MAT1A)
   catalyzes the entry of methionine into the transmethylation cycle.
   Levels of MAT1A mRNA were not altered post-SG; nevertheless, the
   metabolic profile established post-SG is similar to that described with
   specific downregulation of MAT1A.[141]^44 The activity of this enzyme
   is modulated post-transcriptionally by reactive oxygen species,
   glutathione, and nitric oxide
   (NO),[142]^45^,[143]^46^,[144]^47^,[145]^48 through oxidation or
   nitrosylation of a specific cysteine residue. Levels of NO are
   increased following bariatric surgery,[146]^49^,[147]^50 an effect
   perhaps mediated by increased GLP-1-signalling.[148]^51^,[149]^52 Thus,
   changes in either NO metabolism or hepatic redox status post-SG may
   explain the change in transmethylation activity.

   One-carbon metabolism is central to many cellular processes, including
   phospholipid metabolism, creatine synthesis, nucleotide synthesis,
   maintenance of DNA, RNA, and protein methylation, and, as noted above,
   redox metabolism. An intriguing option, to be probed in future studies,
   is that the epigenetic changes observed following bariatric
   surgery[150]^53 are driven at least partly by decreased
   transmethylation. However, changes in hepatic redox metabolism may have
   a more direct effect. Although oxidative stress is generally thought of
   as a deleterious phenomenon, the early stages of the development of
   liver steatosis are characterized by increased antioxidant capacity and
   GSH synthesis,[151]^54^,[152]^55 and several studies have pointed to a
   beneficial effect of reactive oxygen species (ROS) generation and
   glutathione depletion early in the development of
   NAFL[153]^36^,[154]^44^,[155]^56^,[156]^57^,[157]^58 or in the context
   of a methionine- and choline-deficient diet.[158]^58 Indeed, an
   anti-obesity effect of MAT1A downregulation was found to be mediated by
   a decrease in glutathione production.[159]^44 Intriguingly, as noted
   above, decreased glutathione may itself lead to a decrease in MAT1A
   activity, resulting in a feedback mechanism which stabilizes this new
   metabolic state. A decrease in antioxidant capacity may underlie both
   the beneficial effect in the majority of patients, and the puzzling
   worsening of NAFL features in a subset (approximately 10%) of patients
   undergoing bariatric surgery,[160]^59 as well the increased
   susceptibility to acetaminophen toxicity which has been reported in
   several human studies.[161]^60^,[162]^61^,[163]^62

   Beyond effects on the liver itself, alterations in hepatic one-carbon
   metabolism may have profound systemic consequences. The liver is the
   major consumer of methyl groups in the body. Therefore, alterations in
   hepatic methyl group utilization may affect the availability of
   one-carbon groups in other organs. Dietary intake of methionine
   influences immune responses through an effect on T cell activation,
   mediated by SAM availability.[164]^63^,[165]^64 Thus, an increase in
   one-carbon donors following surgery may in part mediate both beneficial
   and deleterious effects on the immune system seen following this
   intervention.[166]^65 For example, a negative effect on the clinical
   course of multiple sclerosis is described following bariatric
   surgery,[167]^66 whereas methionine restriction was found to have a
   positive effect on this disease in pre-clinical models.[168]^63
   Likewise, SG was found to result in altered anti-cancer immunity, with
   a dampened anti-tumor immune response compared to weight-matched
   controls, but improved response to immune checkpoint blockade.[169]^11

   Levels of glycine have been consistently observed to be lower in
   individuals with obesity and the metabolic syndrome, and to increase
   following bariatric surgery.[170]^67 Our findings, showing higher
   glycine levels in the portal compared to the aortic blood pool, and
   decreased expression of glycine catabolic enzymes in the liver, suggest
   that decreased hepatic glycine consumption, reflecting a decreased
   demand for methyl donors, underlies the increased circulating glycine
   levels following bariatric surgery. Other explanations for this
   phenomenon have also been proposed.[171]^68

   Although high-fat diets and the development of obesity have been shown
   to result in depletion of methyl donors,[172]^69 the changes observed
   in one-carbon metabolism post-SG cannot be attributed to either loss of
   weight or decreased consumption of fat alone, as they were in direct
   opposition to the changes seen in the IF-CR group, thus indicating a
   unique effect of the surgical intervention itself. Our transcriptomic
   data points to a direct link between changes in liver metabolism
   following surgery and at least some of the observed alterations in
   plasma levels of one-carbon metabolites. However, the metabolic
   composition of the portal blood also reflects changes in the metabolism
   of the organs drained by this blood vessel, including visceral fat, the
   stomach, intestine, colon, spleen, and pancreas. Intestinal amino acid
   catabolism in particular is an important determinant of amino acid
   availability to extraintestinal tissues.[173]^70^,[174]^71 Glutamine is
   the major amino acid consumed by enterocytes, as reflected in our data
   by the relatively low levels of glutamine in the portal blood. However,
   the intestinal mucosa also utilizes other dietary amino acids,
   including methionine, serine, and glycine, derived either directly from
   the diet or from arterial blood supply. The gastrointestinal tract is
   the direct target of the bariatric surgical intervention, and
   intestinal metabolic remodeling may therefore underlie the changes in
   amino acid concentrations, with altered liver metabolism as a
   downstream effect. Indeed, changes in intestinal methionine and
   threonine uptake have been observed with high-fat feeding,[175]^72 and
   marked changes in intestinal handling of dietary glucose and amino
   acids[176]^73^,[177]^74 are described following various types of
   bariatric surgery, due in part to accelerated gastric emptying.

   IF-CR resulted in changes primarily related to fatty acid metabolism.
   Despite lower levels of circulating free LCFAs, their intrahepatic
   concentrations were either the same or higher than those seen in the
   two other groups, likely reflecting increased hepatic DNL. The
   expression of enzymes involved in DNL was indeed dramatically increased
   compared to SG and AL fed animals, a finding also reported in a recent
   study of the intermittent fasting response in mice.[178]^7 Similarly,
   an increase in lipogenic enzymes was observed in white adipose tissue
   in rats following cycles of stravation and refeeding.[179]^75 Elevated
   circulating TGs in the IF-CR probably indicate increased hepatic export
   of fatty acids to peripheral tissues. The apparent increase in
   lipogenesis is a somewhat surprising consequence of IF. However, it can
   be seen as a logical outcome of a once-daily intermittent feeding
   schedule, where a short period of caloric intake is followed by a
   prolonged fast. An increased capacity for energy storage, in the form
   of both fat and glycogen, may be developed as an adpatation to this
   feeding program, enabling utilization of the calories consumed during a
   short period throughout the day. This adaptation to fasting may also
   underlie the relative decrease in ketone bodies seen in the IF-CR
   group, as the 10-h period without food may not have been sufficient to
   induce ketogenesis in animals accustomed to daily fasting (in contrast
   to animals habituated to constant access to food). We do note that our
   conclusion regarding increased lipogenetic flux is limited by an
   imperfect correlation between mRNA abundance and metabolic flux, as
   enzyme activities also depend on post-translational modifications and,
   critically, substrate availability[180]^76; therefore, definite
   conclusions regarding the effect of IF-CR on hepatic DNL would require
   more direct measurement of lipogenesis at different fasting intervals,
   and quantification with stable isotope techniques. Nevertheless,
   expression of key enzymes of lipogenesis was increased up to 50-fold in
   the IF-CR group, strongly suggesting increased activity of this pathway
   as a result of the intervention.

   DNL hinges on a supply of acetyl-CoA and reducing power in the form of
   NADPH, derived from the pentose phosphate pathway, the citric acid
   cycle, and serine metabolism.[181]^77 Indeed, marked upregulation of
   these pathways at the transcriptomic level was observed. Likewise, TG
   export relies on the liver’s ability to synthesize sufficient PC,
   either directly from dietary choline, or via the methylation of PE,
   with SAM serving as the methyl donor. Indeed, the ratio of PC to PE was
   increased in the IF-CR group compared to both other experimental
   groups. Increased demand for PC may account for the upregulation of
   enzymes related to methyl group transfer and regeneration of SAM, and
   the relatively depleted levels of methyl group donors, primarily
   betaine, seen following IF-CR, in direct opposition to the effect of SG
   on these pathways.

   In addition to the effects on fatty acid metabolism, IF-CR also showed
   an effect on glucose metabolism, with marked upregulation of genes
   involved in glycolysis and gluconeogenesis, the pentose phosphate
   pathway, and the tricarboxylic acid cycle. Most prominent were an
   upregulation of glucose-6-phosphate dehydrogenase and of malic enzyme
   1, both key enzymes in the generation of cytosolic NADPH, presumably in
   support of lipogenesis. Although generation of reducing equivalents may
   indeed underlie the observed increase in glucose-metabolizing enzymes,
   we are aware that many of these also serve gluconeogenesis, which is
   likely to be predominant after prolonged fasting.

   To conclude, our study highlights the disparate metabolic effects of
   different weight-loss interventions. Overall, although IF-CR resulted
   in changes across multiple central metabolic pathways, the effects of
   SG concentrated on one-carbon metabolic pathways. We highlight a unique
   effect of SG on hepatic one-carbon metabolism and redox balance, which
   has not been hitherto described, and, intriguingly, find that the
   metabolic profile established in the liver following SG is similar to
   that recently described with inhibition of MAT1A,[182]^44 pointing to a
   potential therapeutic target which may mimic the beneficial effect of
   surgery on the liver, and through the liver on whole body metabolism.
   Moreover, the increase in one-carbon donors following surgery may in
   part mediate the effects on the immune system described following this
   intervention.

Limitations of the study

   The main strength of our study is in the combined transcriptomic and
   metabolomic analysis, and in the simultaneous examination of the portal
   and aortic blood pools. However, absent stable isotope tracing studies,
   we are unable to definitively conclude whether observed alterations in
   metabolite levels are because of changes in intake or absorption,
   endogenous synthesis, or consumption. Also, as noted above, mRNA
   abundances are an imperfect surrogate for metabolic flux. However, the
   combined data provides strong circumstantial evidence in many cases and
   allows us to draw conclusions which may be validated by future,
   directed tracing studies. For example, serum levels of glycine were
   highest in the SG group, and significantly higher in the portal
   compared to aortic blood in the AL and IF-CR groups, coinciding with
   decreased expression of glycine catabolic enzymes in the liver post-SG.
   We can thus conclude that the increased levels of glycine seen post-SG
   are likely because of decreased hepatic consumption of this metabolite.

   The comparison of two weight loss interventions with an AL-fed control
   group allowed us to separate the effects of each intervention from the
   effects of weight loss in general. The groups were compared following
   an identical period of fasting, reducing the confounding effects of
   different feeding times on the observed results. A possible limitation
   of our study is that sham surgery was not performed in the AL group.
   The decision not to perform a sham intervention in this group was
   related to our primary objective of identifying metabolic alterations
   following equivalent weight loss in SG and IF-CR compared with no
   intervention at all, including the transient weight-loss induced by the
   sham surgery. The changes identified could theoretically be attributed
   to the surgery itself. However, we note that there were few changes in
   gene expression and metabolite abundance common to the IF-CR and SG
   groups which could therefore be attributed solely to the effects of
   laparotomy or anesthesia performed over a month earlier. A further
   limitation is that we are unable to establish a mechanistic link
   between the interventions and the observed metabolic and transcriptomic
   changes, or between the observed changes and the clinical benefit
   associated with these interventions.

STAR★Methods

Key resources table

   REAGENT or RESOURCE SOURCE IDENTIFIER
   Chemicals, peptides, and recombinant proteins
     __________________________________________________________________

   Isoflurane, USP 100% (Terrell) Piramal Critical Care, Inc. NDC code:
   66794-017-10
   Meloxicam 5 mg/ml (Loxicom) Norbrook ANADA code: #200-491
   D-(+)-Glucose Sigma-Aldrich/Merck Cat#16325
   UltraPure DNase/RNase-Free Water Bio-Lab Ltd. Cat#1249301
   Povidone iodine 10% (Polydine) Dr. Fischer, Israel License
   #131.96.23410
   Pentobarbital sodium 200 mg/ml (Pental) CTS Chemical Industries, Ltd.
   077-29-91952-00
   Methanol LC-MS J.T. Baker Cat#9830
   Acetonitrile LC-MS J.T. Baker Cat#9829
   Isotope-labeled internal standards (glucose, glutamine, pyruvate,
   glutamate, lactate) Cambridge Isotope Laboratories Cat#CLM-1396-5,
   CLM-1822-H-PK, CLM-2440-1, CLM-3949-0.25, CLM-1579-0.5
   Ethanol J.T. Baker Cat#8025
   Isopropyl alcohol Gadot group, Israel Cat#830111370
   Mayer's hematoxylin solution Sigma-Aldrich/Merck Cat#MHS32-1L
   Eosin solution Sigma-Aldrich/Merck Cat#HT110116-500ML
   Ammonium carbonate Fluka Cat#74415-250G
     __________________________________________________________________

   Critical commercial assays
     __________________________________________________________________

   TRI Reagent Sigma-Aldrich/Merck Cat#93289
   Pierce BCA protein assay kit Thermo scientific Cat#23227
   Phosphatidylcholine assay kit Abcam Cat#ab83377
   Phosphatidylethanolamine assay kit Abcam Cat#ab241005
   Glycogen assay kit Abcam Cat#ab65620
   PAS staining kit Merck Cat#101646
   Alcian blue solution Merck Cat#101647
   Cobas C111 Triglycerides Roche diagnostics Cat#04657594190
   Cobas C111 LDL-C Gen3 Roche diagnostics Cat#07005806190
   Cobas C111 HDL-C Gen4 Roche diagnostics Cat#07528604190
   Cobas C111 Cholesterin Roche diagnostics Cat#04718917190
   RNA ScreenTape kit Agilent Technologies Cat#5067-5576
   D1000 ScreenTape kit Agilent Technologies Cat#5067-5582
   Qubit® RNA HS Assay kit Invitrogen Cat#[183]Q32852
   Qubit® DNA HS Assay kit Invitrogen Cat#32854
   KAPA Stranded mRNA-Seq Kit Kapa Biosystems Cat#KK8421
   NextSeq 500/550 High Output v2 kit (75 cycles) cartridge Illumina
   Cat#FC-404-1005
     __________________________________________________________________

   Deposited data
     __________________________________________________________________

   RNA sequencing and metabolomics data Mendeley Data
   [184]https://doi.org/10.17632/fy7zbmxfjk.1
     __________________________________________________________________

   Experimental models: Organisms/strains
     __________________________________________________________________

   Hsd:Sprague Dawley SD male rats Envigo Israel Hsd:Sprague Dawley SD
     __________________________________________________________________

   Software and algorithms
     __________________________________________________________________

   Thermo Xcalibur Thermo Fisher Scientific Version 4.1
   TraceFinder Thermo Fisher Scientific Version 4.1
   Compound Discoverer Thermo Fisher Scientific Version 3.0
   FastQC Babraham Bioinformatics Version 0.11.9
   Cutadapt NBIS Version 3.1
   TopHat Johns Hopkins University Version 2.1.1
   DESeq2 Bioconductor Version 3.12
   Matlab MathWorks Version R2020a
   MetaboAnalyst GenomeCanada Version 4.1.3
   R R Project Version 4.0.1
     __________________________________________________________________

   Other
     __________________________________________________________________

   45% kcal fat diet Envigo/Inotiv TD.08811
   Endo GIA surgical stapler Covidien Cat#EGIA60AMT
   Accu-Chek Performa glucometer Roche Diabetes Care Ltd. Cat#06987788039
   Accu-Chek Performa test strips Roche Diabetes Care Ltd. Cat#06454011018
   Coated Vicryl absorable suture Ethicon Cat#W9501
   Scalp vein set, 23G X 3/4" Anhui Hongyu Wuzhou, Ltd. Cat#OMG-W-01-055B
   MiniCollect lithium heparin collection tube Greiner Bio-One Cat#450535
   CK14 soft tissue homogenizing tubes Bertin Corp. Cat#P000933-LYSK0-A
   Precellys tubes Bertin Technologies Cat#P000945-LYSK0-A
   [185]Open in a new tab

Resource availability

Lead contact

   Further information and requests for resources and reagents should be
   directed to and will be fulfilled by the lead contact, Arnon Haran
   ([186]arnon@hadassah.org.il).

Materials availability

   This study did not generate new unique reagents.

Experimental model and study participant details

Rats

   All experiments were approved by the institutional animal care and use
   committee. Five-week-old male Hsd:Sprague-Dawley rats purchased from
   Envigo (Ness-Ziona, Israel) were used for the studies. Animals were
   housed in pairs and were maintained on a 12-hour light, 12-hour dark
   cycle in a temperature-controlled room (22 ± 2°C).

Experimental model details

   The animals were fed a HFHSD (Envigo TD.08811) for a total of 12 weeks,
   during which they were divided into three experimental groups (see
   [187]Figure 1A). One group (8 rats) underwent SG following seven weeks
   of high-fat, high-sucrose feeding, a second group (6 rats) underwent
   sham operation at the same time and then underwent an IF-CR diet with
   the goal of reaching the same end-experimental weight as the SG-treated
   rats, and a third group (6 rats) was allowed to consume food freely and
   did not undergo any intervention (ad libitum [AL]). During caloric
   restriction, food was provided once daily in the evening. The amount of
   food was initially restricted to 10 grams/animal/day to and was then
   gradually increased up to 20 grams/animal/day.

   Oral glucose tolerance tests were carried out in the SG and IF-CR
   groups at week 10. Oral gavage of 2 g/kg glucose was administered
   following an overnight fast, and serum glucose was determined using an
   Accu-Chek Performa glucometer (Roche Diagnostics Ltd.) at intervals of
   15-30 minutes up to 120 minutes post-gavage. Random, non-fasting
   measurements of serum glucose were also carried out in all experimental
   groups once to twice weekly.

Details of surgical interventions

   Rats were fasted for the 12 hours before and for the first 24 hours
   after surgical interventions and were then maintained on a liquid diet
   for a further 72 hours, following which the preoperative diet was
   returned. Surgery was performed under inhalational anesthesia
   (isoflurane) and post-operative analgesia (meloxicam) was administered
   daily for the 5 days following surgery.

   SG was performed with the use of a Covidien Endo GIA surgical stapler.
   A midline incision was made, and the stapler was positioned to achieve
   a removal of 70-80% of the stomach, following previously described
   protocols.[188]^78 For sham operations, a midline incision was followed
   by manual manipulation of the stomach only.

Tissue collection

   At week 12, five weeks after surgery in the interventional groups, and
   following a 10-hour overnight fast, blood samples from the portal vein
   and the aorta were obtained, as well as tissue samples from the liver.
   For the AL and SG groups, food was removed from the cages at 10 hours
   prior to sacrifice; for the IF-CR group, once-daily feeding was
   provided at 10 hours prior to sacrifice, and any remaining food was
   then removed. Anesthesia was achieved with intraperitoneal
   phenobarbital. All samples were immediately frozen in liquid nitrogen
   and then stored at -80°C until their analysis.

Method details

   All animals in the IF-CR and AL groups were included in the final
   analysis. In the SG groups, two animals died due to surgical
   complications and one animal died prior to obtaining blood and tissue
   samples due to anesthetic complications. Therefore, data from only five
   animals in this group were included in the final analysis.

Sample preparation for mass spectrometry

   Metabolomic analysis of the hepatic input and metabolomic and
   transcriptomic analysis of the liver itself were performed in the three
   experimental groups. Portal and aortic blood samples were analyzed
   separately, and the hepatic input was then defined as their combination
   in a 4:1 ratio.[189]^79

   Metabolite extraction was performed as follows: frozen rat plasma and
   liver samples were thawed on ice. 10 μL of each plasma sample were
   diluted in 190 μL of precooled (-20° C) methanol:acetonitrile 3:1
   extraction solution and vortexed for 10 minutes at 4°C to extract
   metabolites. The extraction solution was prepared with isotope-labeled
   internal standards (20 μM ^13C[6]-glucose, 4 μM ^13C[5]-glutamine, 2 μM
   ^13C[3]-pyruvate, 10 μM ^13C[5]-glutamate, 10 μM ^13C[3]-alanine, and
   10 μM ^13C[3]-lactate; Cambridge Isotope Laboratories). Vortexed plasma
   samples were centrifuged at 4°C for 15 minutes at 16,000g to pellet
   proteins. The supernatant was then transferred into high-performance
   liquid chromatography (HPLC) glass vials containing a glass insert and
   stored at -80°C until liquid chromatography-mass spectrometry (LC-MS)
   analysis. Frozen liver samples weighing 30 mg each were suspended in
   CK14 soft tissue homogenizing tubes (Bertin Corp., Rockville, MD, USA)
   containing 1,000 μL precooled (-20° C) methanol:acetonitrile:water
   5:3:2 extraction solution and 1.4 mm ceramic beads. The extraction
   solution was prepared with the same isotope-labeled internal standards
   as specified above for plasma samples. Samples were homogenized using a
   Precellys 24 tissue homogenizer (Bertin Technologies,
   Montigny-le-Bretonneux, France) cooled to 4° C (3 X 20 seconds at
   6000 rpm, with a 30 second gap between each of the three cycles).
   Homogenized extracts were centrifuged in the Precellys tubes at
   18,000 g for 15 min at 4°C, and supernatants were collected in
   microcentrifuge tubes and centrifuged again at 18,000 g for 10 min at
   4°C. The supernatants were then transferred to glass HPLC vials and
   kept at -80 °C prior to LC-MS analysis.

Analysis of metabolites by mass spectrometry

   LC-MS analysis was conducted as described previously[190]^80 with minor
   adaptations. Briefly, a Thermo Ultimate 3000 HPLC system coupled to a
   Q-Exactive Orbitrap Mass Spectrometer (Thermo Fisher Scientific) was
   used with a resolution of 35,000 at 200 mass/charge ratio (m/z),
   electrospray ionization, and polarity switching mode to enable both
   positive and negative ions across a mass range of 67 to 1000 m/z. HPLC
   setup consisted of a ZIC-pHILIC column (SeQuant; 150 mm × 2.1 mm, 5 μm;
   Merck) with a ZIC-pHILIC guard column (SeQuant; 20 mm ×2.1 mm). 5 μl of
   rat plasma or liver extract were injected, and the compounds were
   separated with mobile phase gradient of 15 min, starting at 20% aqueous
   (20 mM ammonium carbonate adjusted to pH 9.2 with 0.1% of 25% ammonium
   hydroxide) and 80% organic (acetonitrile) and terminated with 20%
   acetonitrile. Flow rate and column temperature were maintained at
   0.2 ml/min and 45°C, respectively, for a total run time of 27 min. All
   metabolites were detected using mass accuracy below 5 ppm. Thermo
   Xcalibur was used for data acquisition. Quality control standards
   containing only extraction mix and processed as described above, as
   well as a quality control mix of all samples, were used to assess for
   technical variability. Using the software TraceFinder 4.1 (Thermo
   Fisher Scientific), raw data were extracted and matched to a reference
   library built by running commercial standards for all detected
   metabolites, based on expected retention time and mass. For data
   normalization, raw data files were processed with Compound Discoverer
   3.0 to obtain total measurable ion peak intensities for each sample.
   Automatic identification was followed by manual curation to ensure
   accuracy. Peaks were quantified using the area under the curve and
   normalized to the total ion count (TIC) measured for each sample. Fatty
   acids measured were non-esterified, free fatty acids.

RNA sequencing analysis

   RNA was isolated from liver tissue using a TRI reagent-based protocol.
   cDNA libraries were prepared using 1000 ng of RNA per sample and
   following poly(A) enrichment, using the KAPA Stranded mRNA-Seq kit
   (Kapa Biosystems). RNA and DNA quantity and quality were assessed using
   an Agilent TapeStation system (Agilent Technologies, Santa Clara,
   California) and a Qubit fluorometer (Invitrogen, Carlsbad, California),
   respectively. RNA integrity numbers were >8.2 for all samples.
   Libraries were normalized and pooled together. Pooled samples were then
   loaded on the Illumina NextSeq 500 sequencing system, with 75 cycles
   and single-read sequencing conditions. Raw reads were inspected for
   quality issues with FastQC. Following that, the reads were
   quality-trimmed with Cutadapt[191]^81, using a quality threshold of 32
   for both ends. The processed reads were aligned to the rat
   transcriptome and genome with TopHat,[192]^82 allowing for five
   mismatches. Quantification was done with HTseq-count.[193]^83

Analysis of phospholipids, glycogen, cholesterol, and triglycerides

   Phosphatidylcholine (PC), phosphatidylethanolamine (PE), and glycogen
   content in liver tissue were measured using Abcam
   colorimetric/fluorimetric assays (ab83377, ab241005, and ab65620,
   respectively), following the manufacturer’s instructions. Serum
   triglycerides (TG), low-density lipoprotein (LDL) cholesterol, and
   high-density lipoprotein (HDL) cholesterol, were determined using a
   Cobas c111 analyzer (Roche Diagnostics Ltd.).

Quantification and statistical analysis

   All metabolomic statistical analyses were carried out using
   MetaboAnalyst.[194]^84 One-way analysis of variance (ANOVA) was used to
   recognize individual metabolites or genes whose concentrations or
   expression, respectively, were significantly altered between the
   experimental groups. Correction for multiple comparisons was carried
   out using the Benjamini-Hochberg procedure[195]^85 with a false
   discovery rate (FDR) or adjusted p-value cutoff of 0.05. ANOVA post-hoc
   analyses were performed using Fisher’s least significant differences
   (LSD) procedure or, where noted, Tukey’s honestly significant
   difference (HSD) test.

   Differential gene expression analysis was carried out using
   DESeq2[196]^86 and Matlab. Low expression genes were filtered out prior
   to analysis based on a threshold of average expression of less than 1
   copy per million across samples. The significance threshold was set as
   an adjusted p-value of <0.05, following correction for multiple
   comparisons carried out using the Benjamini-Hochberg procedure.[197]^85

   Multivariate dimensionality reduction using partial least
   squares-discriminant analysis (PLS-DA) was used to identify groups of
   metabolites discriminating between the experimental groups. Metabolites
   were defined as belonging to components one, two, or neither, based on
   their variable importance in projection (VIP) score for that component
   being above 1.0 (an accepted threshold[198]^87) and greater than the
   VIP score for the other component. Pathway enrichment of each component
   was then carried out using the small molecule pathway database (SBPDB).

   For differentially expressed genes, enrichment analysis by Kyoto
   Encyclopedia of Genes and Genomes (KEGG) pathway was performed
   separately for upregulated and downregulated genes in each of the
   intervention groups (i.e., SG and IF-CR); these were defined as genes
   identified by ANOVA and post-hoc analysis as significantly altered
   against at least one of the other experimental groups, and with either
   a negative or positive fold-change compared to both other.

Acknowledgments