Selective FFA2 Agonism Appears to Act via Intestinal PYY to Reduce Transit and Food Intake but Does Not Improve Glucose Tolerance in Mouse Models

Currently, the most effective antiobesity treatment is bariatric surgery, which leads to rapid, long-lasting diabetes remission independent of weight loss (1). The increased mortality and expense burden associated with obesity and its comorbidities, including type 2 diabetes, highlight the need for development of knifeless therapeutic strategies. Host-microbiome chemosensation is emerging as an important system in the development and pathogenesis of metabolic syndrome and associated diseases (2–4). Indeed, the colonic microbial composition is altered after Roux-en-Y gastric bypass and is coupled with improved glucose tolerance, reduced adiposity, and weight loss in humans (5) and mice (6).

Short-chain fatty acids (SCFAs) are generated by bacterial fermentation of colonic dietary fiber, reaching high concentrations (70–130 mmol/L) under physiological conditions in humans (7). These endogenous ligands activate two free fatty acid (FFA) receptors, specifically, FFA2 (GPR43) and FFA3 (GPR41). Acetate and propionate exhibit a higher potency for FFA2 than FFA3, while butyrate and other SCFAs have a lower potency for both (8). FFA2 couples to G_q and G_i/o proteins (9) and is expressed in rodent and human intestine, specifically in enteroendocrine L cells (10–12) that are thought to corelease the incretin and gastric emptying inhibitor glucagon-like peptide 1 (GLP-1) (13) and the satiety-inducing peptide YY (PYY). FFA2 knockout leads to impaired glucose tolerance and GLP-1 secretion (14) and hyperphagia, obesity, and insulin resistance (15); therefore, FFA2-mediated GLP-1 secretion may be of therapeutic benefit. Propionate stimulates GLP-1 secretion from murine colonic cultures, and this response is impaired in cultures from FFA2^−/− mice (14). Moreover, apically perfused intrasmall intestinal or colonic administration of propionate to rats increases the release of and elevates plasma GLP-1 and PYY (16,17), while in hyperinsulinemic humans, intracolonic acetate increases circulating GLP-1 and PYY (18), confirming this SCFA’s ability to release both L-cell products. The higher postprandial levels of these gut hormones observed post–Roux-en-Y gastric bypass are associated with improved glucose levels and satiety, indicating that upregulating these peptides may be beneficial in obese patients (19).

PYY is released postprandially in direct response to nutrient ingestion (16). Within the gastrointestinal (GI) tract, PYY has a spectrum of activities, reducing gastric emptying and activating the ileal and colonic brakes in humans and rodents (20,21). PYY also inhibits mucosal anion and fluid secretion along the length of the intestine, particularly in the distal bowel, where L-cell frequency (22) and electrolyte absorptive capacity are greatest in mammals (23). In addition, peripheral PYY(3-36) acts to increase GLP-1 secretion and improve glucose tolerance (24) and is a well-known satiety factor. Indeed, the development of PYY-Y2 antiobesity agonists remains promising (25,26).

GLP-1 inhibits gastric emptying and food intake and is the principal incretin responsible for regulating postprandially elevated insulin secretion (27). GLP-1 mimetics are proven antidiabetic agents (28), associated with improved glucose tolerance and reduced incidence of hypoglycemia, in addition to significant weight reduction (29). Identifying novel therapeutic targets that further promote GLP-1 release is a current focus of drug-development programs. FFA2-induced GLP-1 and PYY release has received much interest in recent years, and FFA2 has been proposed as a promising target for the treatment of obesity and type 2 diabetes (30,31). Nevertheless, recent research shows that FFA2 acts in a cell-autonomous manner in β-cells to inhibit insulin secretion (32), raising doubts about its validity as a diabetes target.

Here, we synthesized a potent FFA2 agonist (Compound 1 [Cpd 1]; (2S,5R)-5-(2-chlorophenyl)-1-1(2′-methoxy-[1,1′-biphenyl]-4-carbonyl)pyrrolidine-2-carboxylic acid; patent no. WO 2011/076732 A1) (Supplementary Fig. 1A) to explore the functional effects of selective agonism. Although FFA2 agonists have been used to examine GLP-1 release in vitro (33), this is the first reported study of selective FFA2 effects in native tissue and more importantly in vivo. Cpd 1 displayed good pharmacokinetic properties and in vitro potency at FFA2 (Supplementary Figs. 1B and 2). There was no activity at FFA1 or FFA3 at concentrations up to 10 μmol/L (Supplementary Fig. 2C). Furthermore, it showed no significant enzymatic activity or radioligand displacement when tested at 10 μmol/L in a broad selectivity panel against 124 known drug targets and was not cytotoxic using our in-house assay. Thus, we deem Cpd 1 to be suitable for use as an in vivo tool agonist for FFA2. Our primary aims were to determine the PYY and GLP-1 mechanisms underpinning L-cell FFA2 regulation of colonic mucosal function and on intestinal and colonic transit, food intake, and glucose tolerance in lean and diet-induced obesity (DIO) mice, using the FFA2 selective agonist Cpd 1 and FFA2^−/− mice as tools.

All experiments were performed in accordance with the Animals (Scientific Procedures) Act 1986 and had approval from internal ethics committees for animal experimentation. FFA2^−/− mice were generated on a 129/SvEv background as previously described (14). For all in vivo studies, the vehicle was 5:5:90 Tween 80:PEG400:saline at 5 mL/kg i.p.

The preparation of colonic cells from wild-type (WT) mice and secretion studies were performed as previously described (34). Cpd 1 dissolved in DMSO was applied to the culture medium and incubated for 3 h at 37°C. Secreted and cellular GLP-1 was extracted as previously described (34) and active GLP-1 quantified by ELISA.

Mice were fasted for 14 h overnight before receiving either dipeptidyl peptidase IV inhibitor (DPPIVi) or vehicle and 30 min later were injected with Cpd 1 or vehicle. After 15 min, mice received a liquid mixed-nutrient meal challenge (Ensure Plus). Ten minutes later, terminal plasma samples (via blood collected in aprotinin-coated tubes) were analyzed using a Meso Scale Discovery Active GLP-1 Assay kit. Whole blood samples for insulin measurements were taken via tail bleed into heparinized capillary tubes and assayed by ELISA.

Descending colon mucosa from WT or FFA2^−/− male mice were bathed in Krebs-Henseleit buffer and voltage clamped at 0 mV in Ussing chambers as previously described (35). Vectorial ion transport was measured as short-circuit current (I_sc) (μA ⋅ cm⁻²), and all peptide and antagonist additions were basolateral. Addition of Cpd 1 to the apical or basolateral reservoirs occurred 10 min after vasoactive intestinal peptide (30 nmol/L), and internal controls, i.e., PYY (10 nmol/L) or exendin 4 (100 nmol/L), were obtained after Cpd 1. Endogenous PYY or GLP-1 mediation in FFA2 activity was determined using optimized pretreatments, i.e., the Y1 antagonist (BIBO3304, 300 nmol/L) with or without Y2 antagonist (BIIE0246, 1 μmol/L) with or without GLP-1 antagonist [exendin(9-39), 1 μmol/L (36)]. In glucose sensitivity studies, colonic mucosa was bathed with Krebs-Henseleit buffer containing glucose (11.1 mmol/L) on the basolateral side, but mannitol (11.1 mmol/L) replaced glucose on the apical side, and apical FFA2 responses were recorded. As a control, blockade of the sodium/glucose cotransporter 1 was achieved with apical phloridzin (50 μmol/L).

WT or FFA2^−/− male mice acclimatized to handling for 3 days prior to experimentation were fasted overnight prior to testing. No drugs, Cpd 1, or vehicle was administered 10 min prior to bead insertion, and colonic transit was measured using the propulsion assay described previously (36).

WT or FFA2^−/− mice acclimatized to handling for 3 days prior to experimentation were fasted for 16 h prior to testing. CD-1 mice were fasted for 4 h and then administered Cpd 1 or vehicle 10 min prior to testing. A charcoal meal (10% plant charcoal in 5% gum acacia [21]) was given by intragastric gavage, and 30 min later the small intestine was isolated from the pyloric to ileocecal junctions. Upper-GI transit (UGIT), encompassing gastric emptying and small intestinal motility, was determined as previously described (36).

C57Bl6, FFA2^−/−, or WT male mice (6–8 weeks old) received a high-fat diet (HFD) (45% or 60% kcal as fat) for 21–23 weeks as indicated. For the phenotyping study, body weight was monitored weekly and body composition assessed by EchoMRI. For the chronic dosing study, mice were housed singly on a reverse-phase light/dark cycle and received Cpd 1 or vehicle twice daily for 21 days; an oral glucose tolerance test (OGTT) was performed on day 22.

Lean mice were habituated to a 4-h daily presentation of wet mash (wet powdered chow diet) over 2 weeks. On the study day, standard chow was removed and mice were dosed with Cpd 1 or vehicle 1 h before presentation of the wet-mash meal. After 4 h, mash was replaced with standard chow.

Mice were fasted overnight for 15 h. For compound studies, animals were administered Cpd 1 or vehicle 1 h prior to challenge, consisting of 10 mL/kg p.o. liquid mixed-nutrient meal challenge and 1.5 g/kg or 2 g/kg glucose p.o. as indicated. Tail snip blood samples were analyzed for glucose using either a handheld monitor or standard enzyme assay and insulin by ELISA.

All data are presented as mean ± 1 SEM. Analyses were performed using GraphPad Prism, version 5.01, or InVivoStat, by Student t test, one-way ANOVA, ANCOVA, or two-way ANOVA with Dunnett or Bonferroni multiple-comparison or planned comparison post hoc tests, as indicated.

Addition of Cpd 1 to primary colonic cultures increased GLP-1 secretion in a concentration-dependent manner, reaching significance at 10 μmol/L, where there was a doubling of secretion compared with negative controls (Fig. 1A). In comparison, the positive control (10 mmol/L acetate) resulted in an approximately threefold increase over negative control (vehicle).

In vivo Cpd 1 had no effect on plasma GLP-1 levels after nutrient challenge in the absence of a DPPIVi but significantly increased GLP-1 when given in the presence of a DPPIVi (Fig. 1B). In the absence of the DPPIVi, Cpd 1 caused a significant decrease in plasma insulin levels measured at a single time point 10 min after meal challenge; however, no change was observed in the presence of the DPPIVi (Fig. 1C).

Apical Cpd 1 evoked responses in WT colon mucosa with an EC₅₀ of 8.9 nmol/L (range: 3.2–60.7) (Fig. 2A) and were similar to basolateral WT responses, while the comparative responses to apical Cpd 1 were reduced 82% in FFA2^−/− mucosa (Fig. 2B). Basal I_sc and resistance levels were not significantly different in WT and FFA2^−/− tissue (Fig. 2C), although there was a slightly higher resistance in FFA2^−/− colon mucosa (P = 0.06 [Student t test]). In Fig. 2D, the Y1 and Y2 antagonists (BIBO3304 and BIIE0246, respectively) raised basal I_sc revealing endogenous PYY tone within WT colon (described previously [21]). Notably, the monophasic I_sc responses to Cpd 1 were BIBO3304 (and BIIE0246) sensitive, but exendin(9-39) insensitive (Fig. 2E), revealing a predominantly PYY-Y1–mediated mechanism of action. Exogenous PYY and exendin 4 responses were abolished by their respective antagonists (see [35]). Apical Cpd 1 responses were glucose sensitive, as replacement with mannitol apically resulted in significantly reduced, but not abolished, responses (Fig. 2F). As previously observed (35), the sodium/glucose cotransporter 1 inhibitor phloridzin (used as a control) was only effective when apical glucose was present.

Confirming that FFA2 receptors are involved in slowing GI motility, under basal conditions FFA2^−/− mice exhibited a significantly faster colonic transit (i.e., reduced transit time) compared with that of WT mice (Fig. 3A). In WT mice, Cpd 1 significantly increased the time until bead expulsion compared with vehicle and to an extent also in FFA2^−/− mice (where a significant difference at P < 0.05 was only observed using the Student t test). The efficacy of Cpd 1 on FFA2^−/− colonic transit was, however, significantly reduced compared with that of WT mice. Furthermore, basal UGIT in FFA2^−/− mice was significantly faster than in WT mice, and Cpd 1 significantly decreased UGIT in WT mice (Fig. 3B).

FFA2^−/− mice with established DIO showed no change in body weight compared with WT mice (Fig. 4A); however, they exhibited increased body fat (Fig. 4B). Blood glucose levels were increased in FFA2^−/− DIO mice compared with WT after administration of glucose or a mixed-nutrient meal challenge (Fig. 4C and E); however, no change in blood insulin levels was seen (Fig. 4D and F).

To investigate whether FFA2 activation affects food intake, we tested the effect of Cpd 1 on meal feeding in lean WT mice. At 10 mg/kg, Cpd 1 had no effect on food intake (Fig. 5A). However, at 30 mg/kg Cpd 1 resulted in a significant decrease in food consumed within 1 h after meal presentation and, between 4 and 24 h, at the same dose. All mice appeared well with no signs of ill health or malaise.

To investigate whether the effects on food intake in lean mice translated into obese animals, the effects of Cpd 1 were tested in DIO mice. At 10 mg/kg, there was no effect of Cpd 1 over the first 2 weeks of administration. Therefore, the dose was increased to 20 mg/kg on day 14 and subsequently 30 mg/kg on day 17 (Fig. 5B). At 20 mg/kg, there was a small but nonsignificant reduction in daily food intake. However, at 30 mg/kg the food intake was reduced by ∼40% on the first day, and this remained significantly lower than vehicle for the following 2 days. Corresponding to this reduction in food intake, a body weight loss was also seen (Fig. 5C) significantly after the 30 mg/kg dose (day 19 and 20). Throughout the study, mice appeared healthy, with no signs of ill health or malaise.

In DIO mice, an acute dose of Cpd 1 significantly increased blood glucose between 30 and 60 min in response to a meal tolerance test (MTT) (Fig. 6A). No change in the premeal glucose levels at time 0 was observed. By 90 min, the blood glucose had returned to a level equivalent to vehicle. In the same mice, insulin levels were significantly lower after Cpd 1 at 10 min (Fig. 6B). There was no difference at 60 min, when the vehicle group’s values had fallen back to a level similar to the level in those given agonist.

After chronic dosing of Cpd 1 to DIO mice, an OGTT evoked a response on plasma glucose (Fig. 6C) and insulin (Fig. 6D) similar to that observed after an acute MTT. There was no effect of Cpd 1 alone on basal glucose but a significant elevation of blood glucose 30 min after meal challenge. A small but significant reduction in the basal insulin level was observed but only prior to glucose challenge. Postchallenge, there was a non–statistically significant lower level of insulin in compound-treated animals.

Our studies demonstrate that selective FFA2 agonism induces PYY-dependent rather than GLP-1 mucosal mechanisms, slowing of GI transit, and activation of anorexigenic pathways without improving glucose tolerance in vivo. Since the majority of L-cell vesicles contain one or the other peptide (37), independent release is possible but has not been a common observation to date.

SCFAs stimulate GLP-1 secretion in murine colonic cultures, while FFA2 knockout lowers GLP-1 concentrations in vivo (14). Propionate stimulates GLP-1 and PYY secretion via FFA2 from rodent and human colonic cells, while SCFA-induced peptide secretion is impaired in FFA2^−/− mice (17,38) and selective orthosteric FFA2 agonists stimulate the release of GLP-1 from enteroendocrine STC-1 cells (33). Moreover, intracolonic administration of propionate to rodents elevates portal and jugular PYY and GLP-1 (17). In our study, Cpd 1 only stimulated GLP-1 secretion in ex vivo colonic cells at high concentrations and with a lower magnitude than acetate. In vivo Cpd 1 increased plasma GLP-1 but only in the presence of a DPPIVi, administered to enhance the measurable level of GLP-1. Nevertheless, it is possible that Cpd 1 is stimulating nondetectable GLP-1 release, which may still have relevant biological actions. After systemic elevation of GLP-1, corresponding increase in insulin levels would be expected; however, Cpd 1 inhibited insulin secretion after challenge. This suggests that the direct inhibitory effect of FFA2 agonism on insulin secretion, which may involve FFA2 receptors on pancreatic β-cells (32), is more potent than the incretin effects mediated by GLP-1. Acetate has a direct effect on pancreatic FFA2 and FFA3 inhibiting insulin release; however, knockout of both receptors was needed to abolish this response (32). It is likely that GLP-1 responses observed by others with SCFAs are due to effects on other receptors, including FFA3 (14,32). Our data suggest that selective FFA2 agonism alone is sufficient to inhibit insulin release. We attempted to measure PYY secretion; however, we were unable to reliably measure concentrations from cells or ex vivo samples.

Within murine L cells, a recent study shows apparent segregation of vesicular GLP-1 and PYY immunofluorescence (37). Release profiles for these two hormones also differ postprandially, and our data implicate differential mucosal PYY and GLP-1 release, as FFA2 agonism is predominantly PYY mediated. In further support, intraluminal oleic acid and butyrate infused into rat colon induced the release of PYY, but not GLP-1, although no effect on either peptide has also been reported (39–41). Moreover, in humans intraduodenal glucose infusion leads to a fourfold increase in GLP-1 but little change in plasma PYY (42).

It is not known whether FFA2 is targeted to apical and basolateral membranes of L cells and therefore whether they primarily detect luminal or plasma SCFA. However, both apical and basolateral additions of lipid-soluble Cpd 1 inhibited colonic I_sc levels, with apical Cpd 1 displaying a low nmol/L potency. This corresponds well with data generated in recombinant cells expressing mouse FFA2, where Cpd 1 shows an EC₅₀ of 81 nmol/L (Supplementary Fig. 2). Apical mucosal responses to Cpd 1 were significantly reduced but not abolished in FFA2^−/− tissue. Residual activity may indicate involvement of other mechanisms at 100 nmol/L; however, Cpd 1 shows no activity at FFA1 or FFA3 up to 10 μmol/L (Supplementary Fig. 2C). Apical acetate or proprionate (5 mmol/L) evoked biphasic responses in WT colon, the secondary reductions in I_sc being absent from FFA2^−/− colon (Supplementary Fig. 3). In contrast, mucosal Cpd 1 responses were monophasic and PYY, not GLP-1, mediated. Epithelial Y receptors couple to G_i, resulting in sustained inhibition of transepithelial Cl⁻ secretion (43). Use of competitive Y1 and Y2 antagonists revealed that FFA2 activation causes a predominant Y1-mediated response, which is consistent with PYY's capacity to mediate 90% of endogenous Y1 tone in mouse colon (21). In contrast, exendin(9-39) did not alter Cpd 1 responses, confirming that GLP-1 receptors are not involved. Notably, Cpd 1 responses were inhibited when apical glucose was removed; thus, FFA2 agonism exhibited glucose sensitivity as observed for other L cell–specific G-protein–coupled receptors (35,44).

FFA2^−/− mice had significantly faster distal colon motility and UGIT compared with WT mice, revealing a role for FFA2 in both colonic and ileal brakes. Cpd 1 slowed colonic transit and UGIT in WT mice, and although Cpd 1 did slow colonic bead expulsion in FFA2^−/− mice, this was residual compared with WT transit. The PYY-mediated mucosal results with Cpd 1, combined with previous data, suggest that FFA2 agonism leads to slowing of transit via PYY release. Indeed PYY^−/− mice have significantly faster UGIT and colonic transit than WT mice (21,45) and PYY activates ileal and colonic brakes under normal and stressed conditions (21,46).

Here, DIO resulted in a higher proportion of fat mass and impaired glucose tolerance in FFA2^−/− mice, although obese FFA2^−/− and WT mice did not have different body weights. In contrast, FFA2 deficiency exacerbated obesity in mice and SCFA-mediated FFA2 activation suppressed adipose insulin signaling, leading to inhibition of fat accumulation in adipose tissue (15). We found that FFA2 agonism using Cpd 1 in DIO mice reduced free-feeding HFD food intake by ∼40% on the first day of an effective dose (30 mg/kg) and this remained significantly lower than vehicle for the next 2 days; correspondingly, a significant body weight loss was also seen at this dose. Similarly, Cpd 1 significantly reduced food intake in lean mice habituated to receive a daily palatable meal. Although the pharmacokinetics of Cpd 1 are relatively short, the effects on food intake appear to remain after the compound is likely to be cleared; no rebound feeding effect was observed, and after an acute dose a significant difference in food consumed was evident 4–24 h after administration, which may be due to downstream effects of FFA2 activation.

Our data suggest that Cpd 1 most likely exerts its anorexigenic effects by accelerating satiety through FFA2- and PYY-dependent mechanisms. In all feeding studies, no signs of ill health, malaise, or sickness were observed, and in the acute study it was noted that animals displayed characteristic satiety behaviors. Furthermore, the pattern of anorectic effects was similar to sibutramine, a well-characterized anorectic agent in both acute and chronic models. Circulating PYY(3-36) induces satiety, at least partially, via Y2 receptors in the arcuate nucleus (47). Targeted propionate delivery to the colon by oral ingestion increased plasma GLP-1 and PYY in a manner similar to that of a 1,000-calorie meal and prevented weight gain in overweight humans (38); thus, the induction of satiety via activation of PYY mechanisms after Cpd 1 is likely after intraperitoneal administration to mice.

The mucosal PYY-mediated activity of Cpd 1 together with the reduced colonic motility in vivo observed in FFA2^−/− mice is congruent with the drug's anorectic action being most likely due to an increase in satiety rather than an adverse effect of the compound. However, we have been unable to confirm that this effect is solely due to actions at FFA2 using FFA2^−/− mice, as the background 129SvEv strain is not amenable to meal habituation. It is therefore possible that at least a small proportion of in vivo activity could be due to pharmacological activity at another unidentified receptor.

In DIO FFA2^−/− mice, no changes in insulin secretion were observed in response to the meal or glucose challenge, despite the significant glucose intolerance. This may be reflective of the severity of disease in this model, as although obese, these mice only display mild glucose intolerance and not hyperinsulinemia. Agonism of FFA2 in DIO mice dramatically suppressed insulin levels in response to a meal challenge, and although this may be an artifact response due to the timing of glucose and insulin measurements, blood glucose was elevated throughout the duration of the study, indicating that insulin release was suppressed. In agreement, recent findings suggest that FFA2 receptors in pancreatic β-cells have a role in suppressing insulin release under diabetic conditions (32). After chronic dosing of Cpd 1 to DIO mice over 21 days, there was a significant reduction in the basal insulin levels prior to glucose challenge. This did not translate into an elevated basal glucose level, although a significant elevation of blood glucose 30 min after challenge was seen. Moreover, no significant reduction in insulin release was observed, despite an evident trend. Taken together, these data indicate that selective FFA2 agonists have no beneficial effect on glucose homeostasis during obesity.

In conclusion, we describe the first reported use of a selective FFA2 agonist on ex vivo and in vivo responses. We have shown that proven FFA2 agonism has the potential to activate FFA2 receptors on L cells and cause responses that are mediated by PYY release, therefore preferentially producing the beneficial effects of this gut hormone. These include in vivo activation of colonic and ileal brakes and suppression of food intake, leading to a reduction of body weight in DIO mice. However, in contrast to SCFAs, we were unable to detect any responses supporting GLP-1 secretion with Cpd 1, and Cpd 1 did not improve glucose tolerance in lean or obese mice; in fact, it suppressed insulin levels in vivo. FFA2 agonists may have promising utility to treat obesity and consequently reduce the incidence of type 2 diabetes comorbidity. However, therapies targeting FFA2 should be treated with caution given their effects on suppressing insulin secretion.

Acknowledgments. The authors acknowledge RenaSci Ltd for performing the chronic dosing mouse DIO study with Cpd 1.

Funding. The studies were funded initially by a Biotechnology and Biological Sciences Research Council Centre for Integrative Biomedicine pre-Collaborative Awards in Science and Engineering award.

Duality of Interest. The studies were subsequently funded directly by Takeda Cambridge Ltd. S.S., G.C., H.H., K.R., R.N., R.D., M.B., and J.G. are paid employees of Takeda Cambridge Ltd. H.C. consults for Novo Nordisk A/S, Denmark. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. S.F. and S.S. cowrote the manuscript and researched and interpreted data. G.C., H.H., K.R., and R.N. researched data. R.D., M.B., and J.G. contributed to the discussion and reviewed the manuscript. H.C. designed the mucosal and transit studies, interpreted data, contributed to the discussion, and reviewed and edited the manuscript. S.S. 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.