FFA2-, but not FFA3-agonists inhibit GSIS of human pseudoislets: a comparative study with mouse islets and rat INS-1E cells

The expression of short chain fatty acid receptors FFA2 and FFA3 in pancreatic islets raised interest in using them as drug targets for treating hyperglycemia in humans. This study aims to examine the efficacy of synthetic FFA2- and FFA3-ligands to modulate glucose-stimulated insulin secretion (GSIS) in human pseudoislets which display intact glucose responsiveness. The FFA2-agonists 4-CMTB and TUG-1375 inhibited GSIS, an effect reversed by the FFA2-antagonist CATPB. GSIS itself was not augmented by CATPB. The FFA3-agonists FHQC and 1-MCPC did not affect GSIS in human pseudoislets. For further drug evaluation we used mouse islets. The CATPB-sensitive inhibitory effect of 100 µM 4-CMTB on GSIS was recapitulated. The inhibition was partially sensitive to the Gi/o-protein inhibitor pertussis toxin. A previously described FFA2-dependent increase of GSIS was observed with lower concentrations of 4-CMTB (10 and 30 µM). The stimulatory effect of 4-CMTB on secretion was prevented by the Gq-protein inhibitor FR900359. As in human pseudoislets, in mouse islets relative mRNA levels were FFAR2 > FFAR3 and FFA3-agonists did not affect GSIS. The FFA3-agonists, however, inhibited GSIS in a pertussis toxin-sensitive manner in INS-1E cells and this correlated with relative mRNA levels of Ffar3 > > Ffar2. Thus, in humans, when FFA2-activation impedes GSIS, FFA2-antagonism may reduce glycemia.


Results
In human pseudoislets FFA2-agonists inhibit GSIS. Adequate glucose responsiveness of insulinsecreting beta-cells is a prerequisite for functional testing. Previously, we described that reaggregation of isolated human islet cells into pseudoislets resulted in markedly improved GSIS 42 . The comparison of GSIS of isolated islets from human organ donors and of pseudoislets prepared thereof confirmed a better responsiveness of pseudoislets compared to islets (Fig. 1a, Table 2, Supplementary Fig. S1). In the pseudoislet preparations, insulin secretion at 12 mM glucose was ninefold higher than at 2.8 mM glucose (9.60 ± 0.93 (n = 41) and 1.09 ± 0.14 (n = 43) % of insulin content, respectively). The responsiveness was still variable, but did not correlate to the amount of stored insulin (Table 2). Interestingly, GSIS correlated positively to glucagon mRNA levels (Fig. 1b). In addition to an improved regulation of insulin secretion, glucagon secretion of pseudoislets was inhibited when raising glucose from 2.8 to 12 mM (Fig. 1c). Inhibition of glucagon secretion by raising glucose was not significant in isolated islets ( Supplementary Fig. S1).
In accordance with the FFA2-effects on GSIS in pseudoislets, the relative mRNA levels of FFAR2 were always higher than those of FFAR3, especially in donor #8 that showed a pronounced acetate-and FFA2-agonist mediated inhibition of GSIS (Fig. 1i, Table 2). Of note, the mRNA levels of FFAR1, FFAR2, FFAR3 and FFAR4 as well as of INS (insulin), GCG (Glucagon) and SST (somatostatin) were comparable between pseudoislets and isolated islets of the same human donors (Fig. 1i, Table 2). These results suggest that FFA2-and FFA3-agonists are unsuitable for the treatment of insufficient insulin secretion in humans. FFA2-antagonists, in contrast, may augment GSIS, but only under conditions of FFA2-dependent inhibition of insulin secretion. Effects of FFA2 and FFA3 ligands and SCFAs on GSIS in human pseudoislets. Human pseudoislets were prepared and incubated with test substances as indicated in each experiment and described under methods. (a,d-h) Insulin and (c) glucagon secretion are calculated as % of content and (b,i) relative mRNA levels are expressed as 2 −ΔCt (RPS13 was used as housekeeping gene). In insert in b the highest value is excluded. Results are presented as mean ± SEM of 4 replicates/conditions/donor of (a-c) n = 11, (d) n = 4, (e) n = 1, (f) n = 3, (g) n = 9, (h) n = 4, (i) n = 9 donors (see Supplementary Table S1). Significance **p < 0.01, ***p < 0.001 vs 2.8 mM glucose; # p < 0.05, ## p < 0.01, ### p < 0.001 vs 12 mM glucose, § p < 0.05 vs 12 mM glucose + 1 or 10 µM 4-CMTB, respectively, one-way ANOVA, followed by Tukey´s test. & p < 0.05, unpaired Student's t-test. www.nature.com/scientificreports/ In mouse islets FFA2-agonist 4-CMTB exerted a dual concentration-dependent effect on GSIS through distinct pathways. In view of the unresponsiveness of the human preparations to FFA3-agonists and the absence of stimulatory effects of FFA2-agonists on GSIS, we evaluated the ligands in established rodent cell models, i.e. isolated mouse islets and INS-1E cells. This is possible since the FFA2-and FFA3-agonists have similar affinities and selectivity to mouse as to human receptors (Table 1). In contrast to the consistent inhibitory effect of 4-CMTB in human pseudoislets, in mouse islets the FFA2-agonist 4-CMTB displayed a concentrationdependent dual effect on GSIS (Fig. 2a). At low concentrations, 10 and 30 µM, 4-CMTB stimulated GSIS, while addition of 100 µM inhibited secretion. To examine whether the dual effect of 4-CMTB is due the activation of different G-protein regulated signalling pathways, the effects of 4-CMTB on GSIS was analysed in the presence specific inhibition of G q/11 -proteins with FR900359 43 and of G i/o -proteins with pertussis toxin 44 . Preincubation of mouse islets with the G q -inhibitor FR900359 counteracted the stimulation of insulin secretion induced by 10 µM 4-CMTB (Fig. 2b). Of note, FR900359 abolished muscarinic acetylcholine receptor M3-dependent augmentation of GSIS by carbachol. In contrast, the inhibition of GSIS by 100 µM 4-CMTB was still significant in the presence of FR900359, but partly reversed by pertussis toxin (PTx) pretreatment (Fig. 2c). As control, adrenaline, a potent physiological inhibitor of GSIS, was used, which activates alpha2-adrenergic receptors linked to a PTx-sensitive G-protein 45 . Pertussis toxin treatment was efficient, since adrenaline-mediated inhibition of GSIS was no longer significant in PTx-treated mouse islets (Fig. 2c). The orthosteric FFA2-agonist, TUG-1375 which mimics the interaction of SCFAs with the receptor, did not influence GSIS at 1, 10 and 100 µM (Fig. 2d). Neither SCFAs nor FFA3-agonists affected GSIS (Fig. 2e,f). These results confirm that in mouse islets FFA2-agonist can activate a stimulatory but also an inhibitory pathway. As in human islets, in mouse islets relative mRNA levels of Ffar2 were higher than of Ffar3 (Fig. 2g). The order of relative mRNA abundance was Ffar2 > Ffar3 > Ffar1 > > Ffar4. The Ffar1-4 mRNA levels of isolated islets of different mouse strains (C3HeB/FeJ and C57BL/6N) were comparable and did not change significantly when the mice were held under germ-free conditions (Fig. 2g). Using FACS-sorted GFP-labelled insulin-producing cells of C57BL/6N RIP-Cre mT/mG mice, Ffar2 mRNA levels were discernible in the beta-and non-beta-cell fractions, while Ffar3 mRNA was enriched in beta-cells (Fig. 2h).
These results suggest that the regulation of insulin secretion is comparable between human pseudoislets and mouse islets in regard to the inhibitory effect of high concentrations of FFA2-agonists and unresponsiveness to FFA3-agonists. A stimulatory effect of 4-CMTB on GSIS is only observed in mouse islets.

In rat INS-1E cells SCFAs and FFA3 agonists inhibit GSIS.
Since FFA3-agonists had no effect on GSIS in human pseudoislets and mouse islets, we tested FHQC and 1-MCPC on GSIS in INS-E cells, a rat Table 2. Individual characteristics of human islets and the respective pseudoislet preparation. Data are expressed as mean ± SEM. PI, pseudoislet; SCFAs, short chain fatty acids; GSIS, glucose-stimulated insulin secretion (secretion at 2.8 mM was set to 1); GIGS glucose-inhibited glucagon secretion (secretion at 12 mM was set to 1). www.nature.com/scientificreports/ insulin-secreting cell line. At the highest concentration tested FHQC and 1-MCPC significantly reduced GSIS (Fig. 3a). A comparable inhibitory effect on GSIS was exerted by 1 mM propionate and 1 mM butyrate but not by acetate (Fig. 3b). In INS-1E cells, the relative mRNA levels of Ffar3 were remarkably high, those of Ffar2 were low (Fig. 3c). This expression pattern could explain the inhibitory effect of propionate and butyrate on GSIS in INS-1E cells, since the potency of FFA3 activation declines from butyrate = propionate > > acetate 1,36 . Since no specific FFA3-antagonist was available, the expression of FFA3 was reduced by transfecting INS-1E cells with siRNA against FFA3 (Fig. 3d). The efficient downregulation of FFA3 (by 85%) abrogated the inhibitory effect of FHQC and propionate on secretion without affecting GSIS (Fig. 3e). Next, we examined whether the FFA3-dependent inhibition of GSIS was transmitted via a PTx sensitive G i/o protein as suggested previously 18 . After PTx pretreatment of the cells, both agonists, FHQC and 1-MCPC, were unable to reduce GSIS (Fig. 3f). PTx also abrogated the inhibitory effect of adrenaline. In comparison to the pronounced inhibitory effect of adrenaline on secretion, the FFA3-agonist attenuated secretion by 40%.

Discussion
This study revealed FFA2-antagonist by preventing FFA2-mediated inhibition of GSIS as putative therapeutic targets for the treatment of hyperglycaemia. The results further suggest that FFA3 is not functionally expressed in human islets since FFA3 agonists did not affect GSIS and mRNA levels of FFAR3 were very low, i.e. at the detection limit.
In human pseudoislets as well as in mouse islets and in INS-1E cells, the FFA2-agonist 4-CMTB invariably inhibited GSIS at 100 µM. In contrast to mouse islets, where 10 µM 4-CMTB augmented GSIS, in human pseudoislets, 4-CMTB at lower concentrations inhibited GSIS or had no effect. The concentration-dependent dual effect elicited by 4-CMTB in mouse islets, was observed in a Min6 cell preparation with another FFA2-agonist and might be, therefore, mouse specific 46 . Another, structural related FFA2-agonist stimulated GSIS in human islets at 1 µM concentration 22 . Using a higher number of donor preparations (n=11, Fig. 1) compared with that ones used by McNelis (n=3) 22 , we observed highly variable effects. Therefore, this stimulatory effect might be donor specific.
The inhibition of GSIS by 4-CMTB was efficiently reversed by the FFA2-antagonist CATPB confirming that the effect of 4-CMTB was transmitted through FFA2. Furthermore, in INS-1E cells siRNA against Ffar3 did not (a,f,h) Insulin secretion is expressed as % of content or (b,e,g) relative to secretion at 12 mM glucose. (c,d) mRNA levels are expressed as (2 −ΔCt ). RPS13 was used as housekeeping gene. Results are presented as mean ± SEM of n = 3-4 independent experiments. Significance ***p < 0.001 vs the respective control at 2.8 mM glucose; # p < 0.05, ## p < 0.01, ### p < 0.001 vs the respective stimulation at 12 mM glucose, one-way ANOVA followed by Tukey's test; n.s not significant. www.nature.com/scientificreports/ affect the inhibition by 4-CMTB. Since the antagonist had no effect on GSIS, a favourable effect on insulin secretion is expected only under FFA2-stimulation. This conclusion endorses the proposal made by Stefan Offermann that FFA2-antagonists may be useful for the treatment of hyperglycaemic episodes by improving insulin secretion but our results suggest utility limitations as the antagonist only counteracted FFA2-mediated inhibition of GSIS 18 . Physiological stimuli of FFA2 and FFA3 so far identified are SCFAs, acetate, propionate and butyrate. Using static incubations, we observed a heterogeneous response of human pseudoislets to SCFAs. A significant inhibition of GSIS comparable to the 4-CMTB-induced inhibition was observed in one out of 9 preparations. This inhibition by acetate and low concentration of 4-CMTB correlated with 10-times higher relative mRNA levels of FFA2 compared to the other donors (2 −ΔCt of 0.133 vs < 0.02, respectively; Table 2). Due to restricted information available from organ donors we are unable to speculate about the reason of this heterogeneous expression of FFA2 and response to SCFAs. Acetate, the main SCFA in blood derives mainly from the gut microbiome or from alcohol consumption. That the gut microbiome does not impact on FFA2 and FFA3 expression in islets is corroborated by the finding that FFA2 and FFA3 mRNA levels were not affected by the maintenance of mice under germ-free conditions. Alcohol consumption increases blood acetate levels above 1 mM and associates with decreased fasting and 2 h-postload insulin levels compared to non-drinker 47,48 . In contrast to acetate, plasma concentrations of propionate and butyrate remain at low micromolar range, concentrations which do not activate the receptors.
Our results confirm previous observations that both stimulatory and inhibitory effects can be triggered by SCFAs, and suggest that this heterogeneity is not just a result of different experimental settings 49 . In three previous studies using static incubations of human islets, acetate either inhibited 18 or did not alter GSIS 19,22 . In other studies using perifused human islets, acetate and propionate potentiated GSIS 20,21 . We found in static incubations of pseudoislets stimulation, inhibition or no effect of SCFAs on GSIS, even though glucose stimulated insulin secretion to the same extent and the FFAR2 and FFAR3 mRNA levels were similar in responsive and unresponsive pseudoislets.
Another explanation for the heterogeneous effect of SCFAs and FFA2 agonists on GSIS in human pseudoislets could be a heterogeneous composition of beta:alpha:delta-cells. Somatostatin (from delta-cells) is a potent inhibitor not only of insulin but also of glucagon secretion in humans and rodents 30 . On average, less than 10% of the endocrine islet cells are delta-cells. A recent publication convincingly demonstrated that FFA2-agonists stimulate somatostatin secretion in mice 28 . FFA2 may be expressed in delta-cells, as suggested by the expression pattern of FACS-separated mouse beta-cells and non-beta-cells (Fig. 2h). A variable number of delta-cells may affect the efficiency of the paracrine inhibition of GSIS by somatostatin. In the pseudoislet preparation of donor #8 FFA2-agonists and acetate potently inhibit GSIS and the mRNA levels of somatostatin and FFA2 were higher compared to pseudoislets of the other donors. However, somatostatin acts via PTx-sensitive pathways. In mouse islets, the inhibitory effect of high concentrations of 4-CMTB was only partially reversed by PTx. This indicates both, an activation of a receptor which links to PTx-sensitive G i/o proteins but also the involvement of a PTxinsensitive pathway as has been observed previously 19 .
We decided to use pseudoislets, since they display a much better GSIS than the isolated islets from the same human donor 42 . The mRNA levels of insulin, somatostatin and of the fatty acid receptors FFAR1-4, were comparable between islets and pseudoislets, suggesting that expression of hormones and receptors is not significantly different between islets and pseudoislets ( Table 2) 42 . Interestingly, glucagon mRNA levels were higher in pseudoislets compared to islets yielding in a positive correlation between glucagon mRNA levels and insulin secretion. A higher glucose responsiveness has been found in glucagon rich, dorsal islets compared to glucagon poor, ventral islets isolated from the same rat suggesting that islet glucagon contributes to higher glucose responsiveness 50 . Further studies will help to understand the mechanism underlying the increased production of glucagon in pseudoislets.
The specificity of FFA2-agonists, the FFA2-antagonist [37][38][39]51,52 and FFA3 agonists, FHQC and 1-MCPC 53,54 have been evaluated in expression systems and, at the concentrations used, they activate the respective murine and human receptors (Table 1). However, the FFA3-agonists had no significant effect on GSIS in human pseudoislets and mouse islets. Only in INS-1E cells a significant inhibitory effect on GSIS was observed and correlated with higher mRNA levels of Ffar3 compared to Ffar2. Thus, the absence, or very low expression of FFA2 in INS-1E cells could uncover a FFA3-dependent inhibition of GSIS. The assumption that the relative ratio of FFA2/FFA3 expression in beta-cells determines stimulation or inhibition may explain the results obtained in transgenic (tg) mice overexpressing FFA3 55 . In the FFA3 tg mice, i.e. 40-fold higher expression of FFA3 over FFA2 in beta-cells, plasma glucose excursions during glucose stimulation (oGTT) were increased (due to FFA3-mediated inhibition of GSIS), whereas the absence of FFA3 slightly attenuated the elevation of plasma glucose during oGTT (due to FFA2-mediated stimulation of GSIS). Accordingly, the deletion of FFA2 was not sufficient to overcome the inhibitory effect of acetate on GSIS in isolated mouse islets 18 . Only a deletion of both FFA2 and FFA3 overcame the inhibitory effect of acetate on insulin secretion. The results in our sorted mouse islet cells suggest that FFA3 is enriched in the beta-cell fraction, while FFA2 is expressed on beta-and non-beta-cells. The reason for the very different Ffar2 mRNA levels (more than 2 orders of magnitude) of mouse islets and rat INS-1E cells remains elusive, but expression of FFA2 in non-beta-cell population could contribute to the high islet mRNA levels.
In conclusion, although SCFAs have direct effects on insulin secretion in human islets, these effects are highly heterogeneous among individuals. While SCFAs indirectly augment GSIS by increasing incretin secretion, they may inhibit insulin secretion and subdue the incretin effects via direct effects within the islets 56 . Further experimental evidence is needed to determine whether the beneficial metabolic effects of fibre ingestion include SCFA-effects on beta-cell's differentiation, survival and protection against stress factors 57,58 . This study suggests that FFA2, but not FFA3, is functionally expressed in human islets and that FFA2-antagonists may exert beneficial effects on hyperglycaemic episodes by counteracting FFA2-dependent inhibition of GSIS.  Supplementary Table S1. The islets were cultured overnight in CMRL1066, containing 5 mM glucose, 10% (v/v) FCS (Serva, Heidelberg, Germany), 10 mM HEPES, 2 mM l-glutamine, and 1% penicillin/streptomycin at 37 °C in a 5% CO 2 -humidified atmosphere. The pseudoislets were prepared after dissociation of islets into single cells with 0.25% Trypsin-EDTA in PBS at 37 °C for 5 min as already described in detail 42 . In brief, 2000 cells were reaggregated in hanging drops of 20 µl medium on the top of a petri dish. After 3d of culture, the reaggregated pseudoislets were harvested and placed into 24-well plates with one pseudoislet in 0.5 ml medium/well and cultured for further 2d. Semi-quantitative analysis of cellular mRNA. For cellular mRNA detection and quantification, islets, pseudoislets, sorted islet cells and INS-1E cells were lysed and the Nucleospin RNA isolation kit (Macherey Nagel, Düren, Germany) was used to isolate RNA. Following an evaluation of RNA integrity (Agilent Technologies, Santa Clara, CA, USA), cDNA of 0.1 µg RNA was synthesised using the Transcriptor first strand cDNA synthesis kit (Roche Diagnostics, Rotkreuz, Switzerland). Semi-quantitative PCR was performed with the Light-Cycler 480 system (Roche Diagnostics) using the primers (Invitrogen, Carlsbad, CA, USA) listed in Supplementary Table S2.  Table 1. FR900359 was prepared as previously described, used at a final concentration of 1 µM and added to the cell preparations 1 h before the incubation 43 . A radioimmunoassay (Millipore, Burlington, MA, USA) or a sensitive ELISA (Mercodia, Uppsala, Sweden) was used to measure insulin and glucagon in the supernatant and in the islets/cells following extraction with acid ethanol (80%(v/v) ethanol).