Short-Chain Fatty Acids Protect Against High-Fat Diet–Induced Obesity via a PPARγ-Dependent Switch From Lipogenesis to Fat Oxidation

The shift in Western and developing countries from a traditional high-fiber, low-fat, low-calorie diet toward a low-fiber, high-fat, high-calorie diet is accompanied by a growing prevalence of obesity and insulin resistance (1,2). Dietary fiber supplementation, on the other hand, has been shown to reduce body weight, insulin resistance, and dyslipidemia (3–5). The main products of intestinal fermentation of dietary fibers are short-chain fatty acids (SCFAs), of which acetate, propionate, and butyrate are the most abundant (6). Although these three SCFAs are rapidly assimilated into host carbohydrates and lipids—providing ∼10% of our daily energy requirements (7)—there are clear differences in the way each is metabolized: propionate is primarily a precursor for gluconeogenesis, whereas acetate and butyrate are rather incorporated into fatty acids and cholesterol (8). Besides serving as an energy source, SCFAs also regulate metabolism by inhibition of histone deacetylases and chain-length–dependent activation of the endogenous G-protein–coupled receptor (GPR) 41 and 43 (9–11). The longer butyrate is more selective for GPR41, the shorter acetate is more selective for GPR43, and propionate binds to both receptors (12,13). The SCFAs propionate and butyrate were recently shown to increase intestinal gluconeogenesis (IGN), resulting in beneficial effects on glucose and energy homeostasis (14). The mechanism, however, was different for the different SCFAs. Butyrate acted through a cAMP-dependent mechanism, whereas propionate, itself a substrate of IGN, activated IGN gene expression via a gut-brain neural circuit involving GPR41.

Despite the differences between the metabolism and GPR affinities of the three SCFAs, they ameliorate high-fat diet (HFD)–induced obesity and insulin resistance to a similar extent when given as a dietary supplement (15–17). This suggests that there is a common, GPR-independent molecular mechanism for the beneficial effects of acetate, propionate, and butyrate. A GPR-independent mechanism is in line with recent findings that GPR41-deficient mice are still sensitive to the effects of SCFAs on body weight and insulin sensitivity (17). It is likely that a common mechanism may involve AMPK because acetate and butyrate have been shown to increase energy expenditure by activating AMPK in liver and muscle tissue (15,18), whereas propionate activates AMPK in colon cancer cells and reduces lipid synthesis in isolated rat hepatocytes (19,20). The mechanism by which SCFAs activate AMPK is, however, unknown.

In this study we identify a unifying, AMPK-dependent mechanism by which the three SCFAs mediate the beneficial effects of dietary fiber on the metabolic syndrome. We reveal a cascade of events that starts with downregulation of peroxisome proliferator–activated receptor-γ (PPARγ) activity and in which liver and adipose PPARγ play distinct, complementary roles.

Male C57Bl/6J mice (Charles River, L’Arbresle, France), 2 months of age, were housed in a light- and temperature-controlled facility (lights on from 6:30 a.m. to 6:30 p.m., 21°C) with free access to water and food. The experimental groups were fed a semisynthetic HFD (D1245; Research Diet Services, Wijk bij Duurstede, the Netherlands) in which 45% of calories were from palm oil fat. For the SCFA diets, sodium acetate (S2889; Sigma-Aldrich), sodium propionate (P1880; Sigma-Aldrich), or sodium butyrate (303410; Sigma-Aldrich) was incorporated into the diet at 5% (w/w). A normal-fat control group received a chow diet (RMH-B; Hope Farms, Woerden, the Netherlands). Mice in which exons 1 and 2 of the PPARγ gene were loxP-flanked (PPARγ f/f) were provided by Ronald M. Evans (Salk Institute) and have been described previously (21). PPARγ lox/lox mice were crossed with C57Bl/6J transgenic mice expressing Cre recombinase under the control of the albumin promoter, which is expressed in liver (L-KO), or the aP2 promoter, which targets adipose tissue (A-KO). Experimental procedures were approved by the University of Groningen Ethics Committee for Animal Experiments.

In vivo lipogenesis was determined by incorporation of [1-¹³C]acetate into palmitate by providing 2% (w/v) [1-¹³C]acetate in drinking water for 24 h, as described previously (22). Fatty acid β-oxidation capacity was determined in fresh liver and adipose homogenates according to Hirschey et al. (23). Briefly, tissue was homogenized in sucrose/Tris/EDTA buffer, incubated for 30 min in the reaction mixture (pH 8.0) containing [1-¹⁴C]palmitic acid, and trapped [¹⁴C]CO₂ was measured.

Intraperitoneal glucose tolerance was tested after an intraperitoneal injection of glucose (2 g/kg body weight) after an overnight fast. Intraperitoneal insulin tolerance was tested after an intraperitoneal injection of insulin (NovoRapid; 0.75 units/kg body weight) after a 4-h fast. Hyperinsulinemic-euglycemic clamp studies were performed as previously described (24).

The mice were fasted from 6:00 to 10:00 a.m. Blood glucose concentrations were measured using a EuroFlash meter (Lifescan Benelux, Beerse, Belgium). Mice were subsequently killed by cardiac puncture under isoflurane anesthesia. Liver and epididymal fat pads were weighed, snap-frozen in liquid nitrogen, and stored at −80°C. Part of the fat tissue was fixed in 4% paraformaldehyde in PBS and embedded in paraffin. For adipocyte histology, paraffin sections (3 μm) were stained with hematoxylin and eosin and analyzed at original magnification ×20. Blood was centrifuged at 4,000g for 10 min at 4°C, and plasma was stored at −20°C. Plasma nonesterified fatty acid (NEFA) concentrations were determined using a commercially available kit (Roche Diagnostics, Mannheim, Germany). Plasma leptin and insulin levels were determined using ELISA (ALPCO Diagnostics, Salem, NH). Hepatic triglyceride content was determined using a commercially available kit (Roche) after lipid extraction (25).

Hepatic malonyl-CoA and adenosine concentrations were determined by high-performance liquid chromatography according to Demoz et al. (26) and Miller et al. (27), respectively. Carnitine palmitoyltransferase I (CPT-1) activity was determined in liver homogenates at different dilutions (20, 10, 5, 2, and 1 ng/μL) according to van Vlies et al. (28) with 50 μmol/L palmitoyl-CoA and 2 mmol/L l-carnitine as substrates. The reaction was quenched at 0, 2, 5, and 10 min.

Oxygen consumption, energy expenditure, respiratory exchange ratio (RER), food intake, and activity patterns were measured simultaneously for each mouse using a Comprehensive Laboratory Animal Monitoring System (TSE Systems GmbH, Bad Homburg, Germany). The energy balance was determined by measuring the energy content of diet and that of dried, homogenized feces using a bomb calorimeter (CBB 330; standard benzoic acid 6,320 cal/g, BCS-CRM no. 90N).

Mitochondria were isolated from fresh liver tissue according to Mildaziene et al. (29). The rates of oxygen consumption in isolated liver mitochondria were measured at 37°C using a two-channel high-resolution Oxygraph-2k (Oroboros, Innsbruck, Austria) in mitochondrial respiration medium (30) with palmitoyl-CoA as substrate. Maximal ADP-stimulated oxygen consumption (i.e., state 3) was achieved by adding 4.8 units/mL hexokinase, 12.5 mmol/L glucose, and 1 mmol/L ATP. The resting state (i.e., state 4) oxygen consumption rate was determined after blocking ADP phosphorylation with 1.25 μmol/L carboxyatractyloside. The respiratory control ratio was calculated by dividing oxygen consumption rate in state 3 by oxygen consumption rate in state 4.

HepG2, 3T3-L1, and C2C12 cells, purchased from American Type Culture Collection (Manassas, VA), were maintained at 37°C in 5% CO₂ in DMEM with 10% FBS and 1% penicillin/streptomycin. For HepG2 experiments, cells were plated in DMEM with 10% FBS and 1% penicillin/streptomycin in six-well plates and incubated with 0.1, 1, or 3 mmol/L sodium acetate, sodium propionate, or sodium butyrate for 24 h with or without 100 nmol/L rosiglitazone (Sigma-Aldrich) or 10 μmol/L GW9662 (Sigma-Aldrich) as indicated. For 3T3-L1 experiments, cells were differentiated in six-well plates according to Zebisch et al. (31) and incubated with 0.1, 1, or 3 mmol/L sodium acetate, sodium propionate, or sodium butyrate for 24 h, with or without 100 nmol/L rosiglitazone or 10 μmol/L GW9662, as indicated. For C2C12 experiments, cells were differentiated with 2% horse serum in six-well plates according to Fujita et al. (32) and incubated with 3 mmol/L sodium acetate, sodium propionate, or sodium butyrate for 24 h.

RNA was extracted from liver and adipose tissue using Tri reagent (Sigma-Aldrich, St. Louis, MO) and converted into cDNA by a reverse transcription procedure using Moloney murine leukemia virus and random primers according to the manufacturer’s protocol (Sigma-Aldrich). For quantitative PCR, cDNA was amplified using the appropriate primers and probes. The sequence of the other primers can be found in Supplementary Table 1. mRNA levels were calculated relative to 36b4 expression and normalized for expression levels of mice fed an HFD.

For immunoblot analysis, whole-cell lysate was prepared in lysis buffer, and the protein concentrations were determined using the BCA Protein Assay kit (Pierce). Individual samples were mixed with loading buffer, heated for 5 min at 96°C, and subjected to SDS-PAGE. Antibodies and their sources were as follows: AMPK (no. 2532; Cell Signaling), phosphorylated AMPK (pAMPK Thr172, no. 2531; Cell Signaling), acetyl-CoA carboxylase (ACC, no. 45174; Abcam), phosphorylated ACC (pACC S79, no. 31931; Abcam), uncoupling protein 2 (UCP2, no. 6525; Santa Cruz Biotechnology), PPARγ (no. 2435; Cell Signaling), and fatty acid synthase (no. 3180; Cell Signaling). As loading control, β-actin (no. 2066; Sigma-Aldrich) was used for liver and adipose tissue, and TOM20 (no. 11415; Santa Cruz Biotechnology) was used for isolated liver mitochondria. Finally, horseradish peroxidase–conjugated anti-rabbit from donkey (Amersham Pharmacia Bioscience) or horseradish peroxidase–conjugated anti-goat from donkey (Dako, Glostrup, Denmark) and SuperSignal West Pico Chemiluminescent Substrate System (Pierce) were used. The immunoblots were analyzed by densitometry using Image Laboratory software (Bio-Rad).

Data are presented as mean values ± SEM. Statistical analysis was assessed by one-way ANOVA using the Tukey test for post hoc analysis. Statistical significance was reached at a P value of <0.05.

To examine the effects of SCFAs on the development of obesity and insulin resistance, wild-type (WT) C57Bl/6J mice were fed an HFD with acetate, propionate, or butyrate (5% w/w) for 12 weeks. This provided the animals with a physiological amount of SCFA because the contribution of SCFA to the total energy intake has been estimated at ∼10% (7). The substantial raise in body weight that was observed in controls, fed an HFD only, was attenuated by all three SCFAs to a similar extent (Fig. 1A). This coincided with reductions in white adipose tissue (WAT) mass, adipose cell size, and plasma leptin concentrations (Supplementary Fig. 1A–C). The lower body weight in the SCFA groups was not due to alterations in food intake or physical activity because neither of these was significantly affected by SCFA supplementation (Supplementary Fig. 1D and E). We did, however, observe enhanced energy expenditure in the SCFA-treated mice (Fig. 1B and Supplementary Fig. 1F) as well as a shift from carbohydrate to fatty acid oxidation, as indicated by lower RER (Fig. 1C). This suggests that SCFAs reduce HFD-induced body weight gain by enhancing energy expenditure through increased lipid oxidation.

SCFA-fed and control-fed mice had similar fasting blood glucose levels and glucose tolerance (Supplementary Fig. 1G–I) but much lower fasting insulin levels (Fig. 1D). Together with enhanced disposal of glucose upon insulin injection (Supplementary Fig. 1J), this points to higher insulin sensitivity in these mice. We therefore performed hyperinsulinemic-euglycemic clamp studies under matched insulin exposure. The glucose infusion rate required for maintaining euglycemia (a measure of whole-body insulin sensitivity) in SCFA-fed mice was ∼1.5-fold higher than that in control mice (Fig. 1E). Although the hepatic glucose production rate during the clamp was similar in all groups (Ra in Fig. 1F), the degree to which insulin stimulated the rate of glucose uptake by peripheral tissues (primarily muscle and adipose tissue) was much higher in SCFA-fed mice (Rd in Fig. 1F). This indicates improved peripheral insulin sensitivity in these mice. Collectively, our observations demonstrate that supplementation of any of the three SCFAs enhances the insulin sensitivity of HFD-fed mice to a similar extent.

SCFA supplementation completely prevented HFD-induced obesity and insulin resistance, as indicated by similar body weight, WAT mass, plasma leptin and insulin levels, and peripheral insulin sensitivity compared with mice fed the normal-fat chow diet (Supplementary Fig. 2A–G). The HFD did not significantly affect the basal glucose level, which was similar in mice fed chow and the HFD (Supplementary Fig. 2D).

Finally, we wondered whether SCFAs could also be used to treat existing obesity and insulin resistance. To this end, we first fed mice an HFD for 12 weeks to induce obesity and then supplemented the HFD with SCFAs for 6 weeks. Indeed, the mice supplemented with SCFAs showed significantly lower body weight and WAT mass after this treatment than controls as well as a shift toward fatty acid oxidation and enhanced insulin sensitivity, without affecting the food intake (Supplementary Fig. 3A–I). After 18 weeks, the HFD control mice had the same basal glucose level as the HFD and chow controls at 12 weeks (compared with Supplementary Figs. 3E and 2D).

Control mice fed the HFD had high plasma concentrations of NEFAs and high liver concentrations of triglycerides (hepatic steatosis). These aspects of the metabolic syndrome were reduced when SCFAs were supplemented from the beginning (Supplementary Fig. 4A and B) or after the mice were already obese (Supplementary Fig. 3H and I). This prompted us to study hepatic fatty acid synthesis and oxidation. SCFA-fed mice had lower transcript levels of genes involved in hepatic lipogenesis and a lower concentration of hepatic fatty acid synthase protein (Supplementary Fig. 4C and D). In agreement, these mice had a twofold reduction in in vivo hepatic lipid synthesis (Fig. 2A). The capacity for hepatic lipid oxidation in SCFA-fed mice was twofold higher than that of controls (Fig. 2B). Clearly, hepatic lipid metabolism was shifted toward a more oxidative state.

We wondered whether the metabolic effects of SCFAs might be mediated through activation of AMPK, which is known to shift metabolism from lipid synthesis to oxidation (33). Indeed, we observed increased phosphorylation of AMPK and its downstream target ACC (Fig. 2C), without affecting the total hepatic AMPK and ACC levels (Supplementary Fig. 4E). Phosphorylation inactivates ACC and should lead to lower concentrations of its product malonyl-CoA, an endogenous inhibitor of CPT-1, the first enzyme in the β-oxidation of fatty acids (34). Consistently, SCFA treatment reduced hepatic malonyl-CoA concentrations (Supplementary Fig. 4F) and increased the enzyme capacity (V_max) of CPT-1 in diluted liver homogenates (Fig. 2D). This implies that CPT-1 is stimulated by SCFAs via a dual mechanism: by raising its V_max and by reducing the concentration of its inhibitor.

SCFA-fed mice had decreased hepatic ATP concentrations and increased AMP-to-ATP ratios (Supplementary Fig. 4G and H). The latter is a sensitive reflection of the energetic state of the cell and a direct activator of AMPK (35). Reduced ATP concentrations can be a result of increased mitochondrial proton leakage, leading to mitochondrial uncoupling and subsequently reduced ATP synthesis (36). Therefore, we examined oxygen consumption by isolated liver mitochondria using palmitoyl-CoA as a respiratory substrate, both in the presence of ADP (state 3) and in the presence of an inhibitor of ATP production (state 4). We observed lower respiratory control ratios (the rate of state 3 divided by state 4 respiration) in liver mitochondria from SCFA-fed mice. This could be attributed to an increased state 4 respiration rate (Fig. 2E), demonstrating that there is intrinsic uncoupling of mitochondrial oxidative phosphorylation in the livers of these mice. In line with this, SCFA feeding led to increased expression of UCP2 (Fig. 2F), suggesting that proton leak via UCP2 may be responsible for the observed uncoupling (36).

Next, we studied how SCFAs activate the UCP2-AMPK-ACC pathway. Possible candidates were PPARα and γ, which are known regulators of UCP2 expression, fatty acid oxidation, and whole-body lipid metabolism (37–40). SCFAs did not significantly affect the expression of PPARα, whereas expression of target genes involved in fatty acid oxidation was decreased rather than increased (Supplementary Fig. 5A–C). This suggests that PPARα is not responsible for the enhanced fatty acid oxidation capacity of SCFA-treated mice. In contrast, SCFAs did reduce expression of PPARγ—and its target genes Cd36, Lpl, Fabp4, and Pltp—in liver and adipose tissue, but not in muscle (Supplementary Fig. 5D–G). A reduced expression of PPARγ expression or activity stimulates UCP2 expression and fatty acid oxidation and reduces lipogenesis and hepatic triglyceride levels (41,42), suggesting that PPARγ may well be the mediating factor between SCFAs and the UCP2-AMPK-ACC pathway. PGC-1α mRNA expression in liver, adipose, and muscle tissue and UCP-1 mRNA expression in brown adipose tissue did not change after SCFA feeding (Supplementary Fig. 5H and I), suggesting that neither PGC-1α nor UCP-1 play a role in the SCFA-induced effect.

To find out whether PPARγ may be causally involved in the induction by SCFAs of the UCP2-AMPK-ACC signaling pathway, we first investigated this in vitro in liver cells (HepG2), differentiated adipose cells (3T3-L1), and muscle cells (C2C12). Treating cells for 24 h with 3 mmol/L of any of the SCFAs reduced mRNA and protein levels of PPARγ and its target genes in HepG2 and 3T3-L1 cells but not in C2C12 cells (Fig. 3A–C). SCFAs also enhanced the activity of the UCP2-AMPK-ACC signaling pathway in HepG2 and 3T3L1 cells but not in C2C12 cells (Fig. 3D–F), in line with our in vivo results. The induction of the UCP2-AMPK-ACC pathway by SCFAs was abolished when the partial repression of PPARγ expression and activity was compensated by the PPARγ agonist rosiglitazone (Supplementary Fig. 6A). However, complete inhibition of the activity of PPARγ by the PPARγ antagonist GW9662 did not affect the SCFA-induced reduction in PPARγ expression but did abolish the accompanying increase in the activity of the UCP2-AMPK-ACC pathway (Supplementary Fig. 6B). Apparently, activation or inhibition of the activity of PPARγ abolished the SCFA-induced increase of the activity of the UCP2-pAMPK-pACC pathway.

The SCFA concentrations of 3 mmol/L used here represent typical intestinal concentrations rather than portal blood concentrations (43), although even portal SCFA concentrations may exceed 2 mmol/L postprandially after a fiber-rich diet is digested (44–46). Therefore, we also incubated HepG2 and differentiated 3T3-L1 cells at different concentrations of SCFAs. In both HepG2 and 3T3-L1 cells, there was only a significant effect on PPARγ expression after 24 h exposure to 3 mmol/L SCFAs, whereas 1 and 0.1 mmol/L showed no reduction compared with the control (Supplementary Fig. 6C and D).

Altogether, these results indicate that the activation of the UCP2-pAMPK-pACC pathway by SCFAs in HepG2 and 3T3-L1 cells was due to the observed reduction of PPARγ expression and activity, whereas PPARα had no role in these SCFA-induced effects.

To distinguish between the role of PPARγ in the liver and adipose tissue, L-KO mice or A-KO of PPARγ mice were fed an HFD, with or without SCFA supplementation. The SCFA-induced effects in these KO mice were compared with those of the WT above to analyze whether they were mediated by liver or adipose PPARγ.

Like the WT mice, the L-KO mice also showed a reduction of body weight, WAT mass, and insulin levels in response to SCFA supplementation, demonstrating that these effects were not mediated by hepatic PPARγ (Fig. 4A and B and Supplementary Fig. 7A–C). This is in line with our previous observation that SCFAs increased peripheral but not hepatic insulin sensitivity (Fig. 1F). Whereas in WT mice the RER values and plasma NEFA concentrations were reduced by SCFAs, they were increased by SCFAs in L-KO mice (Supplementary Fig. 7D and E). In addition, the SCFA-induced increase in hepatic lipid oxidation capacity and the concomitant reduction in hepatic triglycerides that was observed in WT mice were abolished in PPARγ L-KO mice (Fig. 4C and D). Finally, we were no longer able to detect any differences in hepatic protein levels of UCP2, pAMPK, and pACC between the SCFA-fed and non-SCFA-fed PPARγ L-KO mice (Fig. 4E), nor did we see any effects on the target genes of PPARγ (Supplementary Fig. 7F). These results suggest that the beneficial effect of SCFAs on liver lipid metabolism was directly mediated by hepatic PPARγ.

Our results with the PPARγ A-KO mice were almost opposite to those with the L-KO mice. Disruption of adipose PPARγ abolished the effects of SCFAs on body weight gain, WAT mass, and insulin levels (Fig. 5A and B and Supplementary Fig. 8A–C). In contrast to our observations in the PPARγ L-KO mice, SCFAs had no effect on RER and plasma NEFA concentrations in the PPARγ A-KO mice (Supplementary Fig. 8D and E), whereas the strong SCFA-induced reduction in liver triglycerides was preserved in these mice (Fig. 5C). Finally, no increase occurred in adipose lipid oxidation capacity or in adipose protein levels of UCP2, pAMPK, and pACC when SCFA-fed and non-SCFA-fed PPARγ A-KO mice were compared (Fig. 5D and E), nor did we see any effects of SCFAs on the target genes of PPARγ (Supplementary Fig. 8F). Further reflecting the differences between the L-KO and the A-KO mice was that the adipose tissue of the L-KO mice preserved the SCFA-induced effects on lipid oxidation capacity, PPARγ protein levels, transcript levels of the PPARγ target genes and the concomitant increase of UCP2, pAMPK, and pACC protein levels (Fig. 4D and E and Supplementary Fig. 7G), whereas the A-KO mice preserved these characteristics in liver tissue (Fig. 5D and E and Supplementary Fig. 8G).

Taken together, our results indicated that SCFA-induced protection against HFD-induced obesity and insulin resistance is impaired by adipose PPARγ deficiency, whereas hepatic PPARγ deficiency impairs the SCFA-induced reduction in hepatic steatosis.

In this study, we demonstrate that the beneficial metabolic effects of SCFAs—protection against HFD-induced obesity and improved insulin sensitivity—are mediated by downregulation of PPARγ. Our results are in line with other studies in which similar physiological effects of SCFAs have been found (15,17). However, by identifying PPARγ as a central regulator of the systemic response, we are the first to integrate the role of all three major SCFAs (acetate, propionate, and butyrate) into a single mechanism.

The combined results of our in vivo and in vitro experiments suggest an SCFA-induced cascade in which downregulation of PPARγ activates an UCP2-AMPK-ACC network. This cascade shifts metabolism in adipose and liver tissue from lipogenesis to fatty acid oxidation. SCFAs activate the supply of fatty acid substrates via CPT1 and the demand for their oxidation products via uncoupling of the respiratory chain. This is a clear example of what has been described as multisite modulation; that is, the phenomenon that metabolic fluxes can only be altered by simultaneous modulation of multiple enzymes in a pathway (47).

The observed mitochondrial uncoupling may be partly explained by the increased expression of UCP2. UCP2 belongs to a family of mitochondrial UCPs comprising UCP1, 2, and 3, all three with proton leak activity (48,49). Whereas UCP1 has clear-cut proton-conducting activity, UCP2 catalyzes the proton-driven exchange of phosphate for small dicarboxylic acids across the mitochondrial inner membrane (50). This allows UCP2 to dissipate the proton gradient across the mitochondrial inner membrane in a substrate-dependent manner.

SCFAs activate the expression of leptin via the SCFA receptor GPR43 (51). Yet, we found reduced serum levels of the satiety factor leptin in SCFA-treated animals. This can be explained from the overall decrease in WAT mass. The same observation was made previously in mice that overexpress GPR43 (52). Despite their lower leptin levels, the food intake of SCFA-treated animals was not increased. This may be due to other satiety factors, which we did not investigate further.

In addition to our finding that PPARγ is a central regulator of the beneficial effects of SCFAs, we also demonstrated that liver and adipose tissue contribute independently to this mechanism: the beneficial effects on whole-body fat accumulation, hepatic steatosis, hypertriglyceridemia, and insulin sensitivity appear to be mediated differently in these two tissues. Whereas an SCFA-induced reduction of PPARγ in the liver reduced hepatic triglyceride concentrations, the same reduction in adipose tissue reduced body weight and improved insulin sensitivity. Cultured adipose and liver cells respond to SCFAs in the same way, by reducing PPARγ and activating the UCP2-AMPK-ACC network, which suggests that the organs respond directly to the SCFAs that have been taken up into the body. We recently showed that the fluxes of SCFA uptake into the body—but not the SCFA concentrations in the cecum—are correlated with the physiological effects of dietary fiber (53). This corroborates the idea that SCFAs need to be taken up and do not exert their full effect on organ physiology through intestinal receptors. In addition, De Vadder et al. (14) recently showed that the intestinal gluconeogenesis pathway is also an important mediator of SCFA-induced reduction of body weight and insulin resistance as PPARγ is in our study. They showed—by capsaicin-induced periportal nervous deafferentation—that propionate exerts its effect via a gut-brain communication axis. Intestinal knockout of the gluconeogenic gene G6Pc abolished the benefits of fiber and SCFAs on body weight and glucose homeostasis, as did our A-KO of PPARγ. How this intestinal pathway relates to the PPARγ pathway in liver and adipose is currently unclear. Similarly, whether and how the PPARγ-dependent pathway interacts with GPR41 and/or GPR43 signaling remains to be established.

That we established PPARγ as a central mediator of the beneficial effects of SCFAs has possible consequences for the treatment of metabolic disorders. A key finding in the PPARγ field is that adipose mass increases almost proportionally to PPARγ activity, whereas inhibition or activation of PPARγ sensitizes the body for insulin (54). With regard to treatment, the most suitable molecules are most likely those that uncouple these different functions of PPARγ such that the receptor sensitizes the body for insulin, without any undesired adipogenesis. Such selective PPARγ modulators have been proposed by others (55), and our results suggest that SCFAs act as highly effective endogenous selective PPARγ modulators. The few human studies that have been conducted showed that intravenous administration of acetate or propionate reduced plasma free fatty acids in humans (56,57), whereas 12 weeks of dietary vinegar (acetate) supplementation in obese subjects led to lower body weight, body fat mass, and serum triglycerides levels than in the placebo-controlled group (16). This makes the inexpensive SCFAs very attractive compounds to prevent and reverse HFD-induced obesity and insulin resistance.

Funding. This work was funded by the Netherlands Genomics Initiative via the Netherlands Consortium for Systems Biology (91108004) from Netherlands Organisation for Scientific Research (NWO)/ZonMw. A.B. and K.v.E. currently receive an unrestricted research grant from Top Institute Food and Nutrition. J.W.J. is supported by a European Research Council grant (IRG-277169), by the Human Frontier Science Program (CDA00013/2011-C), the NWO (VIDI grant 016.126.338), the Dutch Digestive Foundation (grant WO 11-67), and the Dutch Diabetes Foundation (grant 2012.00.1537). M.H.O. and B.M.B. hold a Rosalind Franklin Fellowship from the University of Groningen.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. G.d.B. conceived the study, designed and performed experiments, analyzed data, and wrote the manuscript. A.B., A.G., K.v.E., R.H., T.H.v.D., and M.H.O. performed experiments, analyzed data, and provided input into the manuscript. J.W.J., A.K.G., D.-J.R., and B.M.B. conceived the study, interpreted data, and edited the manuscript. B.M.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.