Glucose Controls Glucagon Secretion by Regulating Fatty Acid Oxidation in Pancreatic α-Cells

Circulating glucose levels are under strict control by the release of glucagon and insulin from pancreatic islets. Glucagon secretion is increased when glucose levels are low to stimulate glucose production from the liver. Circulating glucagon levels are also elevated in individuals with obesity and people with diabetes, and this is thought to contribute to the development of hyperglycemia (1–4). Treating hyperglucagonemia with glucagon receptor antagonists lowers HbA_1c in patients with diabetes (5), but increases the risk of hepatic steatosis (6) and possibly α-cell hyperplasia (7). However, our limited knowledge of α-cell function has made therapeutic intervention difficult.

Glucose is considered the major intrinsic regulator of glucagon secretion at low glucose (<5 mmol/L). However, the mechanism remains widely debated (8–10). The current hypotheses include, but are not limited to, regulation of ATP-sensitive K⁺ (K_ATP) channels (11), store-operated Ca²⁺ channels (12), and reductions in cAMP (13). Common for all is the assumption that at low glucose concentrations, K_ATP channels are mostly closed and that inhibition of glucagon release by glucose relies on an increase in ATP derived from glucose oxidation. This creates a paradox, where ATP needs to be high at both low and high glucose concentrations. While previous findings suggest that glucose increases intracellular ATP in α-cells, more recent findings suggest that α-cells rely on fatty acid (FA) oxidation (FAO) for ATP production at low glucose (14) to maintain glucagon secretion. Although glucose and nonesterified FAs (NEFAs) circulate in concentrations of 5–8 mmol/L and 0.07–0.56 mmol/L, respectively (15–17), the mechanisms that control glucose-regulated glucagon secretion are often investigated with glucose as the only substrate (11,13,14,18–27). Here we hypothesize that the effect of glucose on glucagon secretion and α-cell metabolism depends on the presence of NEFA.

All work with human material was approved by the National Committee for the Ethical Approval of Research, Denmark, and the ethics committee at Lund University. Human islets were obtained from the ADI Islet Core Laboratory at the University of Alberta (Edmonton, Alberta, Canada), or Nordic Network for islet transplantation via Human Tissue Laboratory, EXODIAB/LUDC, Lund University (Lund, Sweden). None of the donors had diabetes (Table 1). Islets were cultured at 37°C, 5% CO₂ in RPMI 1640 (61870-010; 10% FBS, 1% penicillin/streptomycin).

All experiments were approved by the Danish Animal Inspectorate. Mice were housed on a 12–12-h light-dark cycle at 22°C with ad libitum access to standard chow and were 12–20 weeks old. Mice with overexpression of pyruvate dehydrogenase kinase 4 (Pdk4) were generated as previously described (28). Briefly, mouse Pdk4 cDNA and rabbit β-globin polyadenylation sequence, preceded by a loxP flanked neomycin cassette, was inserted via PhiC31 recombinase-mediated cassette exchange downstream of the chicken β-actin promoter position at the Gt(ROSA26)Sor locus, in mouse C57BL/6N JM8F6 embryonic stem cells. Recombinant embryonic stem cells were then injected into blastocysts, and the resulting chimeras were bred with C57BL/6J mice to generate the line of conditional Pdk4 overexpression mice. In these mice, Pdk4 transcription can be activated in a Cre recombinase-dependent manner. The mice carrying the Pdk4 transgene were crossed with mice carrying iCRE under the prepro-glucagon (PPG) promoter (29). Mice with conditional knockout of Cpt1a were generated as previously described (30). α-Cell–specific GCamp3 mice were generated as previously described (20). C57Bl6/J mice, referred to as wild-type (WT) mice, were purchased from Janvier Labs (Le Genest-Saint-Isle, France). For in vivo and ex vivo studies using the Pdk4α knock-in (KI) and αCpt1a mice, results from both sexes were pooled, unless otherwise stated.

Mice were euthanized by cervical dislocation, and islets were isolated by Liberase injection (Roche, 05401020001) into the common bile duct, as previously described (14,31), handpicked into 11 mmol/L glucose RPMI 1640 (61870-010; 10% FBS, 1% penicillin/streptomycin), and incubated at 37°C in 5% CO₂.

Pancreata from control and Pdk4αKI mice were fixed in 10% neutral buffered formalin (Sigma-Aldrich, HT501128), cryo-embedded, and cut into 5-µm sections. Sections were costained for glucagon (1:500; Sigma-Aldrich, G2654) and Pdk4 (dilution 1:50; Protein Tech, 12949-1-Ab) overnight at 4°C. Isolated islets were stained for glucagon, insulin (1:100; Cell Signaling, L6B10) and green fluorescent protein (GFP) for Perceval detection (1:100; Abcam, ab13970). FACS-sorted α-cells were seeded overnight before staining with glucagon (Abcam, ab108426). Secondary antibodies were added before counterstaining with DAPI. Imaging was performed on a Leica SP5-X. Intensity of Pdk4 staining was analyzed in ImageJ software (https://imagej.net/software/fiji/). Sections were blinded and assigned to the respective genotype following analysis to prevent bias.

To facilitate reporter expression in α-cells, adenoviruses carrying H2B-GFP or Perceval High Range (HR) under glucagon promoter were created. A GUTR2 construct (32) was used as a template to amplify GCG-H2B-mCherry cassette. The amplified 3.5-kb fragment was cloned into pGEM, sequenced, and subsequently subcloned into shuttle pDUAL2-V5His-bGH (Vector Biolabs, Malvern, PA). EcoRI restriction site was then introduced downstream of the reporter. PercevalHR was amplified from FUGW-PercevalHR obtained from Addgene (no. 49083) and cloned into pGEM. H2B-mCherry was cut from the GCG promoter cassette in pDUAL2-V5His-bGH using NotI and EcoRI and substituted with PercevalHR or H2B-GFP. Sequences of all constructs were verified. Adenoviruses were produced by Vector Biolabs.

Time-lapse imaging was performed on whole islets using a Leica SP5-X with a 40× objective. For viral induction and dye loading, islets were incubated and parameters set as stated in Table 2. Viral infection with PercevalHR (33) or GFP was used to specifically identify changes in α-cells. Islets were (unless otherwise stated) preincubated for 30 min in 1 mmol/L glucose Krebs Ringer Buffer (KRB; 140 mmol/L NaCl, 4.6 mmol/L KCl, 2.6 mmol/L CaCl₂, 1.2 mmol/L MgCl₂, 1 mmol/L NaH₂PO₄, 25 mmol/L NaHCO₃, 10 mmol/L HEPES, pH7.4, and 6.6 mg/mL FA-free BSA) (6). Islets were perifused with KRB containing a physiological relevant mix and concentration of FAs (0.36 mmol/L NEFA) (21% palmitate, 45.5% oleic acid, and 22.8% linoleic acid bound to FA-free BSA in a 3.6:1 molar ratio) (16), FA-free BSA only, or no BSA at 0.2 mL/min at 35°C. All solutions contained d-glucose and/or carbonyl cyanide-4- (trifluoromethoxy)phenylhydrazone as indicated. Images were taken every 5 min for ATP-Red and every 2 min for all other experiments. Fiji/ImageJ was used for analysis where fluorescence was normalized to the initial condition (F/F₀), and the last 6 min of each condition was averaged for statistical analysis. Calcium imaging was performed as previously described (20), and islets were perfused with KRB, with or without 0.36 mmol/L NEFA and d-glucose, as indicated, at a rate of 0.2 mL/min at 35°C. Images were acquired in 1.28-Hz intervals. GCaMP calcium traces of individual islet cells were analyzed using MATLAB R2018b. Peaks were identified by using built-in MATLAB functions msbackadj(), trapz(), and findpeaks(), a minimal peak height of two times the average intensity of all cell traces, a minimum peak distance of two and width of one, with a peak prominence of at least 1 SD of the individual cell trace.

Islets were isolated from B6J mice and cultured overnight in the incubator (800 islets in 2.5 mL, 7 mmol/L glucose RPMI 1640 [61870-010], 10% FBS, and 1% penicillin/streptomycin). The next day, islets were dispersed using 50% TrypLE for 5 min in a 37°C water bath. Then, 7 mmol/L glucose RPMI was added before centrifugation and dispersion in KRB + 3 mmol/L glucose + 5% serum. Dispersed cells were subjected to single-cell sorting based on their size and autofluorescence. The sort was gated for exclusion of duplets and other cell aggregates based on side scatter (SSC) and forward scatter (FSC). α-Cells were distinguished from non–α-cells based on size (FSC parameter) (34) and autofluorescence (FITC parameter) (35) and sorted into 7 mmol/L glucose RPMI.

Cells were centrifuged and resuspended in 7 mmol/L glucose RPMI, and 10,000 cells were incubated at 37°C in 5% CO₂ in 100 µL medium to recover for 1 h. The medium was then replaced with KRB containing 1 mmol/L glucose and 0.36 mmol/L NEFA for 30 min before being replaced with KRB containing 1 mmol/L or 5 mmol/L [U-¹³C]glucose and 0.36 mmol/L NEFA for 2 h. Cells were gently centrifuged, and the supernatant was removed. Then, 300 µL 90% high-performance liquid chromatography-grade methanol was added, and cell pellets were frozen at −80°C. After three cycles of snap-freezing in liquid nitrogen and vortexing, cells were incubated on ice to precipitate proteins. Samples were then centrifuged at 12,000 rpm at 4°C for 15 min, and the supernatant was transferred to liquid chromatography–mass spectrometry and evaporated to completeness under a stream of N₂. Ketovaline (not detected in samples in a pilot experiment) was added as an internal standard together with sample extract. Dried extracts were derivatizatized with 3-nitrophenylhydrazine essentially according to Hodek et al. (36) and thereafter analyzed by using an ultrahigh performance liquid chromatography system (Agilent 1290 Infinity II) connected to a Bruker timsTOF Pro mass spectrometer (Bruker, Bremen, Germany) operated in negative ionization mode. Extraction and derivatization blanks and calibration curves were derivatized and analyzed as samples. Data processing was performed with Bruker Compass DataAnalysis 5.2 software and TASQ 2021b (Bruker). Calculation of [¹³C] labeling, including correction for natural isotopes, was done according to Lindén et al. (37).

Electrical activity of α-cells within intact mouse islets was measured using a perforated-patch technique. During the experiments, the islets were perfused with an extracellular solution containing (mmol/L) 140 NaCl, 3.6 KCl, 1.3 CaCl₂, 0.5 MgSO₄, 10 HEPES, 0.5 NaH₂PO₄, and 5 NaHCO₃ at pH 7.4, with NaOH and glucose/NEFA as indicated. The temperature of the solution was controlled between 32 and 34°C. The solution within the pipette contained (in mmol/L) 76 K₂SO₄, 10 KCl, 10 NaCl, 1 MgCl₂, and 5 HEPES (pH 7.35 using KOH). The pore-forming antibiotic amphotericin B (240 μg/mL) was included in the intracellular buffer to achieve membrane perforation. α-Cells were distinguished by their functional fingerprinting (38). Electrophysiological measurements were performed using an EPC-10 patch-clamp amplifier (HEKA Electronics, Ludwigshafen/Rhein, Germany) and PatchMaster 2 × 91 software.

Glucagon secretion was measured from groups of 10 islets/replicate. Following isolation, islets were incubated for 1 h in 11 mmol/L glucose RPMI, washed once with KRB containing 0 mmol/L glucose, and preincubated in 2 mmol/L glucose KRB with 0.36 mmol/L NEFA, unless otherwise indicated, for 1 h at 37°C in 5% CO₂. Islets were then incubated sequentially at 1 mmol/L and 5 mmol/L glucose with 0.36 mmol/L NEFA, or as indicated, for 1 h. The supernatant was collected, and islets were harvested in acid ethanol and sonicated. Glucagon and insulin were determined using U-PLEX insulin and glucagon kits (Meso Scale Diagnostics; K15145C-3 or K15303K-4), according to the manufacturer’s instructions.

FAO was measured from 50 islets from WT or αCPT1a knock-out mice (αCPT1aKO) and preincubated in 1 mmol/L glucose with 0.36 mmol/L NEFA, using radiolabeled palmitic acid [9,10-³H(N)] (Perkin Elmer), as previously described (14).

Body weight, blood glucose, and blood ketones were measured between 8 a.m. and 10 a.m. using a Contour NEXT glucose meter (Bayer) and FreeStyle Precision ketone meter (Abbott).

Glucose tolerance tests were performed after a 6-h fast (6 a.m.–12 p.m.). Mice were injected with d-glucose (2 g/kg; Sigma-Aldrich) in sterile PBS, and blood glucose was measured at 0, 30, 60, 90, and 120 min using a Contour NEXT (Bayer). Blood was collected at 0 and 30 min on ice into aprotinin (Sigma-Aldrich, A11S3), centrifuged at 2600g at 4°C for 10 min, and plasma was collected and stored at −80°C. Plasma glucagon levels were measured using the R-PLEX glucagon assay (Meso Scale Diagnostics, F201K-3, L45SA-2-2).

All data are presented as mean ± SEM. All statistics were performed using GraphPad Prism 9 software (GraphPad Software, San Diego, CA). Values outside of 2 SDs were considered outliers and excluded. Specific statistical tests and post hoc analyses are stated in each figure legend.

Data and materials from this study will be available upon reasonable request.

The effect of single FAs on glucagon secretion has previously been investigated (39,40). Here we explored how NEFA, in a physiologically relevant mix and concentration, would affect glucose-regulated glucagon secretion. Islets store FAs as triglycerides (41), which can potentially be used for FAO at low glucose. We therefore preincubated WT islets at 5 mmol/L glucose, 1 mmol/L glucose, or 1 mmol/L glucose supplemented with 0.36 mmol/L NEFA. In islets preincubated in 5 mmol/L glucose, an increase in glucose from 1 to 5 mmol/L reduced glucagon secretion (Fig. 1A). When islets from the same mice were preincubated in 1 mmol/L glucose, glucagon secretion was both lower at 1 mmol/L glucose and not reduced in response to 5 mmol/L (Fig. 1A). However, adding 0.36 mmol/L NEFA to the 1 mmol/L glucose preincubation restored glucose-induced inhibition of glucagon (Fig. 1A), without changes in glucagon content (Fig. 1B). These findings were not due to changes in insulin secretion (Supplementary Fig. 1A and B). The effect of NEFA on glucagon secretion at 1 mmol/L glucose was dose dependent and higher at 0.36 mmol/L than at 0 mmol/L NEFA, following a preincubation with the indicated amount of NEFA at 1 mmol/L glucose (Fig. 1C). No differences in secretion were observed at 5 mmol/L glucose, and glucose was only repressive in the presence of 0.36 mmol/L NEFA (Fig. 1C and D). The presence of NEFA also affected glucose-stimulated insulin secretion in the presence of 0.07 mmol/L NEFA (Supplementary Fig. 1C and D). Islets also maintained the expected response to 10 mmol/L glucose in the presence of 0.36 mmol/L NEFA, lowering glucagon and stimulating insulin secretion (Supplementary Fig. 1E–H). The glucose-induced reduction in glucagon secretion in the presence of NEFA was correlated with reduced FAO measured in isolated mouse islets, preincubated in 1 mmol/L glucose with 0.36 mmol/L NEFA, when glucose was increased from 1 mmol/L to 5 mmol/L (Fig. 1E). Collectively, this suggests that α-cells rely on the presence of NEFA to maintain the inhibitory effect of glucose on glucagon secretion.

We explored how addition of NEFA affected intracellular ATP by expressing the ATP sensor PercevalHR (33) specifically in α-cells (Fig. 2A and B). Exposing mouse or human islets to 0.36 mmol/L NEFA in the presence of 1 mmol/L glucose increased intracellular ATP in α-cells (Fig. 2C–F). When glucose levels were then increased to 5 mmol/L in the presence of 0.36 mmol/L NEFA, intracellular ATP in α-cells was reduced (Fig. 2C–F). The observed reduction in ATP was not an artifact due to changes in intracellular pH (Supplementary Fig. 2A–D). Similar data were obtained using ATP-Red 1 dye (42) in α-cells with nuclear GFP expression (Supplementary Fig. 2F–H). In these experiments, the non-GFP cells tended to increase fluorescence intensity (Supplementary Fig. 2G and H), as is expected for non–α-cells (24). The effect of glucose on ATP in α-cells required the presence of NEFA as preincubation of islets at 1 mmol/L glucose removed the reduction in intracellular ATP at both 5 mmol/L glucose and 20 mmol/L glucose (Fig. 2G–J). The response to glucose was variable in these cells, which could suggest that the degree of intracellular FA depletion may differ between individual α-cells. When WT islets were preincubated in 3 mmol/L glucose, which would spare FA availability, intracellular ATP also decreased in response to 20 mmol/L glucose (Fig. 2K and L). Notably, these ATP measurements were performed in a buffer containing BSA (6.6 mg/mL), which is required to bind NEFA. Previous ATP measurements that show increased intracellular ATP in response to increased glucose were performed without the addition of BSA (11,14,19,21,43). Indeed, when the ATP measurements were made in the absence of BSA in our experimental system, increasing from 1 to 20 mmol/L glucose increased intracellular ATP (Supplementary Fig. 2I). This suggested that α-cell function may be affected by the presence of BSA. To test this, we preincubated islets in 5 mmol/L glucose, with and without BSA, in the absence of 0.36 mmol/L NEFA. In the presence of BSA, glucagon secretion responded as expected (Supplementary Fig. 2J). However, complete removal of BSA impaired secretion at both 1 mmol/L and 5 mmol/L glucose (Supplementary Fig. 2J). These differences were present without changes in glucagon content (Supplementary Fig. 2K) or insulin secretion and content (Supplementary Fig. 2L–M). Together, these data suggest that under physiological conditions, α-cells use FAs for ATP production and that elevations of extracellular glucose lead to reductions in intracellular ATP.

Normal fuel homeostasis requires reciprocal regulation of glucose and FAO. In the fed state, FAO is inhibited by the increased availability of glucose through the glucose-FA cycle (44–46). The pathway requires pyruvate, derived from glycolysis, to enter the tricarboxylic acid cycle (TCA). Once pyruvate is converted to acetyl-CoA by pyruvate dehydrogenase (PDH), it is condensed with oxaloacetate by citrate synthase to form citrate. Citrate is either used in the TCA cycle or exported to the cytosol, where it acts as substrate for malonyl-CoA synthesis. Production of malonyl-CoA near the mitochondrial membrane leads to inhibition of carnitine palmitoyl 1a (CPT1a) and thereby inhibits long-chain FAO (46). Therefore, to understand how glucose lowers intracellular ATP in α-cells, we used stable isotope glucose tracer metabolite profiling. To obtain a pure fraction, α-cells were FACS sorted using flavin-adenine dinucleotide autofluorescence (35). Unlike previously, here sorting was done in 3 mmol/L glucose KRB with 5% FBS. Under these conditions, flavin-adenine dinucleotide fluorescence is high in other islet cells, and α-cells can be sorted as the least fluorescent fraction (Fig. 3A). This led to an α-cell fraction with 89% purity, confirmed by glucagon staining. To understand how glucose is metabolized in α-cells, the sorted α-cell fraction was preincubated in 1 mmol/L glucose with 0.36 mmol/L NEFA and subsequently transferred to 1 or 5 mmol/L [U-¹³C]glucose with 0.36 mmol/L NEFA for 2 h (Fig. 3A).

Analysis by mass spectrometry revealed that the content of all measured metabolites, except for citrate, increased or trended toward an increase in 5 mmol/L compared with 1 mmol/L glucose (Fig. 3B). The percentage of ¹³C enrichment was only increased in α-ketoglutarate (α- KG) and trended toward an increase in citrate (P = 0.101), while the enrichment for succinate decreased (Fig. 3C). The similar citrate content and small increase in ¹³C enrichment could suggest that citrate is being used to generate other metabolites not measured here. The increased percentage of ¹³C enrichment in α-KG could suggest that citrate may be converted to α-KG. Alternatively, cataplerosis of citrate could explain why there is no increase in the total content of the metabolite (Fig. 3D). The combination of increased α-KG and decreased succinate enrichment suggests that the increases in total concentrations of fumarate, malate, and succinate content may be derived from anaplerotic metabolism. On the basis of the findings in this experimental setup, it is likely that a proportion of citrate is used in the glucose FA cycle or for FA synthesis (Fig. 3D).

For glucose to contribute to the glucose FA cycle, pyruvate derived from glucose must enter the TCA cycle as oxaloacetate through pyruvate carboxylase or as acetyl-CoA through PDH. The two metabolites can then be condensed by citrate synthase to form citrate. The activity of PDH is determined partly through inhibitory phosphorylation by pyruvate dehydrogenase kinases (PDK) (Fig. 4A). Mouse and human α-cells both express high levels of Pdk4 mRNA compared with other islet cell types (47,48). PDK4 expression is regulated by changes in substrate availability, and consequently, expression is increased in both skeletal muscle (49) and islets (50) during fasting. To determine whether the glucose-FA cycle is important for the effect of glucose on intracellular ATP and glucagon secretion, we generated a mouse model in which Pdk4 was specifically overexpressed in α-cells (Pdk4αKI). In these mice, the higher PDH phosphorylation would lead to lower influx of pyruvate into the TCA cycle through PDH. PDK4 staining intensity was observed exclusively in α-cells and was twofold higher in Pdk4αKI mice than in controls (Fig. 4B and C). In these mice, the glucose-induced reduction in intracellular ATP (Fig. 4D and E) and glucagon secretion (Fig. 4F) was lost, without changes in glucagon content (Fig. 4G). In addition, the loss of glucose-induced suppression of glucagon secretion in Pdk4αKI mice led to an increase in insulin secretion at 5 mmol/L (Supplementary Fig. 3A and B), supporting observations that glucagon can potentiate insulin secretion at low glucose concentrations (51). This suggests that PDK4 plays a key role in the regulation of glucagon secretion and that entry of pyruvate as acetyl-CoA into the TCA cycle is important for the reduction of glucagon secretion and intracellular ATP in α-cells. These findings were confirmed by treatment of islets with the PDK inhibitor dichloroacetate, which removed the effect of 5 mmol/L glucose on glucagon secretion (Fig. 4H).

Cpt1aKO in α-cells leads to decreased long-chain FAO and reduced glucagon secretion at low glucose (14), identical to the effects observed when islets are incubated with low substrate levels (Fig. 1A). We therefore investigated the effect of glucose on ATP levels in islets from mice with Cpt1aKO specifically in α-cells (αCPT1aKO). Intracellular ATP did not change in α-cells from these mice when exposed to 0.36 mmol/L NEFA or 5 mmol/L glucose, but instead continuously decreased (Fig. 4J and K), suggesting that α-cells rely on FAO to maintain ATP, as previously observed (14). Some studies have suggested that inhibition of mitochondrial FAO can activate peroxisomal FAO (52–54). We therefore measured whole-islet FAO in control and αCPT1aKO mice at 1 and 5 mmol/L glucose. FAO in whole islets was lower in αCPT1aKO mice at 1 mmol/L glucose and was unaffected by 5 mmol/L glucose (Fig. 4K). These findings indicate that the glucose-FA cycle is active in α-cells and that glucose inhibits glucagon secretion by lowering FAO and intracellular ATP.

To test whether PDK4 expression in α-cells contributes to the regulation of glucagon secretion in vivo, we measured blood glucose and plasma glucagon in Pdk4αKI mice (Fig. 5). Glucose tolerance and plasma glucagon levels were unaffected in male and female Pdk4αKI mice (Fig. 5A–I). However, unlike controls, glucagon levels failed to reduce in female Pdk4αKI mice after glucose administration (Fig. 5I). This was also the case when sexes were combined to increase statistical power (Fig. 5C). No changes in body weight, blood glucose, or ketones were observed between genotypes (Fig. 5J–L). This suggests that, while PDK4 expression in α-cells is an important regulator of glucose-regulated glucagon secretion, it is not enough to drive the development of hyperglycemia or hyperketonemia in vivo.

Pancreatic α-cells express K_ATP channels (48,55). It was previously suggested that closure of the channel leads to reduced action potential height (via voltage gated Na⁺ channels), reduced Ca²⁺ influx, and inhibition of glucagon secretion, a mechanism that would require increased ATP in response to glucose. The observation that glucose decreased intracellular ATP in the presence of NEFA therefore prompted us to investigate the electrical activity in α-cells under these experimental conditions. The addition of 0.36 mmol/L NEFA at 1 mmol/L glucose had little effect on plasma membrane potential (Fig. 6A–C). Unlike in control recordings without NEFA and BSA (Fig. 6A), adding 5 mmol/L glucose in the presence of 0.36 mmol/L NEFA repolarized the membrane (Fig. 6B and C) and reduced the firing frequency of action potentials (Fig. 6D). Subsequent addition of the K_ATP channel blocker tolbutamide (100 µmol/L) caused a rapid and strong depolarization (Fig. 6B and C). This could indicate that the reduction in ATP caused by increasing glucose levels leads to opening of K_ATP channels and suggests that membrane potential in α-cells is more negative than previously reported (11). Electrical activity (11) and FAs (40) have both been suggested to regulate glucagon secretion through changes in cytosolic calcium. To investigate whether the addition of 0.36 mmol/L NEFA affected this parameter, we used mice expressing the cytosolic calcium sensor GCaMP3 specifically in α-cells (20). The addition of 0.36 mmol/L NEFA did not affect oscillation frequency in α-cells at 1 mmol/L. Increasing glucose to 5 mmol/L lowered oscillation frequency (Fig. 6E and F) as expected, in line with previous observations recorded in the absence of NEFA (20). This suggests that at low glucose, the membrane potential is depolarized and that increasing glucose repolarizes the plasma membrane to inhibit electrical activity, calcium entry, and glucagon secretion.

Here we propose that the regulation of ATP production in α-cells is highly dependent on enzymes that promote FAO, such as PDK4 and CPT1a. We find that inhibition of pyruvate entry into the TCA cycle as acetyl-CoA, or FA transport into the mitochondria, disconnects changes in glucose levels from changes in ATP production and glucagon secretion. Based on the observations made here, we suggest that glucose regulates glucagon secretion, not by increasing intracellular ATP, but by inhibiting FAO to lower intracellular ATP.

Our findings suggest that in α-cells, FAO is subject to suppression from glucose, as suggested by the glucose-FA cycle. Despite this, α-cells do oxidize glucose to some extent (56,57), at least in the absence of other substrates and BSA, where increases in extracellular glucose results in ATP production (11,19,21,24,58). However, we show here that in the absence of BSA, α-cells do not secrete much glucagon and do not respond to glucose. While it is not clear why BSA is important for α-cell function, albumin has previously been shown to impact both intracellular lipid, pH, and redox homeostasis (59,60). Depleting endogenously released FAs by adding FA-free BSA to dispersed β-cells has been shown to prevent signaling that stimulates Ca²⁺ oscillations and insulin secretion (33). However, here we observed intact regulation of glucagon secretion with FA-free BSA when whole islets are preincubated in 5 mmol/L glucose, suggesting that this is not occurring in the current experimental paradigm. Whether this discrepancy is due to differences in the experimental paradigm or the two cell types is unclear.

Previous hypotheses of how glucose regulates glucagon secretion suggest that increased ATP from glucose oxidation leads to membrane depolarization in α-cells (11,22). However, as with the previous measurements of ATP, these experiments were performed with glucose as the only substrate. The findings we present here show that glucose repolarizes the plasma membrane in α-cells when applied in the presence of NEFA, consistent with the observed reduction in ATP under the same experimental conditions. The finding that this effect was reversed by tolbutamide suggests it reflects activation of K_ATP channels. It has previously been suggested that a reduction in intracellular ATP could lead to reduced Na⁺/K⁺ pump activity (14); however, this would lead to a depolarization of the membrane, unlike what we see here. It is, therefore, more likely that activation of K_ATP channels drives the change in membrane potential in α-cells in response to increased glucose levels. In addition, the current observation that ATP is reduced may also be aligned with the proposed reduction in intracellular cAMP in α-cells (13). While, the apparent K_m of adenylate cyclase for ATP is ∼0.6 mmol/L (61), for some isoforms, the K_m is substantially altered when bound to G_αs (62) supporting the possibility that lowering of intracellular ATP could contribute to the lowering cAMP. The lower cAMP could also be caused by increases in intracellular FAs as a consequence of the lower FAO at higher glucose, as adenylate cyclase in other tissues has been suggested to be inhibited by increases in intracellular FA levels (59).

Concentrations of 5 mmol/L glucose and 0.36 mmol/L NEFA mimic plasma levels of macronutrients after an overnight fast; however, after a glucose challenge, postprandial NEFA levels may be as low as 0.07 mmol/L, with plasma glucose concentrations of 8 mmol/L in human subjects (17). Under conditions with 0.07 mmol/L NEFA, we observed lower glucagon secretion at low glucose and no effect of increasing glucose concentrations. This suggests that the lowering of plasma levels of glucagon in response to a glucose tolerance test may also be driven by changes in FA availability. This is supported by the ex vivo experiments presented here. However, overexpression of PDK4 in α-cells results in a rather mild phenotype. That PDK4 overexpression in α-cells alone is not enough to drive the development of hyperglycemia or hyperketonemia in vivo is not surprising. Other models of impaired glucagon secretion also have relatively mild phenotypes (7,14,21,63). In the case of this model, this may reflect that paracrine factors also contribute to the inhibition glucagon secretion. Despite this, our data indicate that changing PDH activity in α-cells can affect circulating glucagon levels. Thus, α-cells may rely on sensing circulating levels of FA as well as glucose. However, it should be considered that other substrates, such as amino acids, could also contribute and thereby regulate glucagon secretion in α-cells (64).

In conclusion, we propose a framework for α-cell metabolism and glucose-regulated glucagon secretion, reciprocal to that observed in β-cells (65), in which the metabolic phenotype of α-cells enables a specialized glucose response, which lowers intracellular ATP and leads to reduced glucagon secretion through activation of K_ATP channels and repolarization of the plasma membrane. This model of α-cell metabolism and glucagon secretion suggests that α-cells can act as sensors of changes in both circulating glucose and NEFA concentrations.

Acknowledgments. The authors thank Dorthe Nielsen (University of Copenhagen) for technical assistance during data collection, Professor Leanne Hodson and Dr. Katherine Pinnick (University of Oxford) for scientific advice, and Professor Seung Kim (Stanford University) for providing the GUTR2 construct used for live cell imaging experiments. Imaging experiments were performed at the Centre for Advanced Bioimaging (CAB) at the University of Copenhagen.

Funding. A.H. is supported by a fellowship from Svenska Sällskapet for Medicinsk Forskning (SSMF). Financial support for the transgenic core at the Wellcome Centre for Human Genetics was provided by the Wellcome Trust Core Award Grant Number 203141/Z/16/Z. P.C. is supported by long term structural funding - Methusalem funding by the Flemish government, the Fund for Scientific Research-Flanders (FWO-Vlaanderen), European Research Council Advanced Research Grant EU-(ERC743074), and a Novo Nordisk Foundation (Denmark) NNF Laureate Research Grant. T.M. is supported by the Novo Nordisk Foundation (grant no. NNF18CC0034900) L.E. is supported by the Swedish Research grant (SRA-Exodiab and project grant), the Swedish Foundation for Strategic Research (IRC-LUDC), and The Swedish Diabetes Foundation. P.R. is supported by the Swedish Research Council, the Helmsley Trust, and the Medical Research Council (MRC), J.G.K. is supported by a Novo Nordisk Fonden Excellence Emerging Investigator Grant- Endocrinology & Metabolism (no. 0054300) and an Independent Research Fund Denmark Sapere Aude Fellowship (no. 0169-00067B).

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

Author Contributions. S.L.A., A.F., M.V.C., H.D., L.A.-M., A.H., G.K.-P., T.M., and J.G.K. contributed to the investigation. S.L.A., A.F., M.V.C, H.D., L.A.-M., A.H., G.K.-P., P.C., B.D., T.M., L.E., P.R., and J.G.K. reviewed and edited the manuscript. S.L.A. and J.G.K. conceptualized the study. S.L.A. and J.G.K. conceived the experimental design. S.L.A. and J.G.K. wrote the original draft. M.V.C., P.C., B.D., L.E., and P.R. contributed materials. J.G.K. 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.

Prior Presentation. Parts of this study were presented as an abstract at the 56th Annual Meeting of the European Association for the Study of Diabetes, virtual meeting, 21–25 September 2020, and at the 81st Scientific Sessions of the American Diabetes Association, virtual meeting, 25–29 June 2021.