The isolated perfused rat and mouse small intestinal studies were conducted with permission from the Danish Animal Experiments Inspectorate (2018-15-0201-01397) and our local Institutional Animal Care and Use Committee (Department of Experimental Medicine, Protocol nr. P18-555 and P-20-067). Experiments were performed in accordance with the guidelines of the Danish legislation governing animal experimentation (1987) and the National Institutes of Health (publication number 85-23) and the European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes (Council of Europe No. 123, Strasbourg 1985).

COS-7 cells were transiently transfected using the calcium-phosphate precipitation method (15) with either human MC4R or rat MC4R. Since MC4R primarily is a Gαs-coupled receptor (16) intra-cellular cAMP production was used as marker of receptor activation. Procedures are described in details elsewhere (17). In brief, the cells were seeded in white 96-well plates at a density of 3.5 ∗ 10⁴ per well one day after the transfection. The following day, the cells were washed twice with HEPES-buffered saline (HBS) and incubated with HBS and 1 mM 3-isobutyl-1-methylxanthine (IBMX) for 30 min at 37°C (95% O₂, 5% CO₂). In order to determine agonistic activities, alpha-MSH and NDP-alpha-MSH were added in increasing concentrations and incubation continued for 30 min at 37°C. In order to determine the antagonistic properties of HS014, the cells were pre-incubated for 10 min with increasing concentrations of HS014 followed by addition of fixed concentrations of alpha-MSH or NDP-alpha-MSH, corresponding to 60–80% max activity, and incubated for an additional 20 min (18). After incubation, cells were thoroughly washed and then lysed. Determination of cAMP was done with a HitHunterTM cAMP XS assay (DiscoverX, Herlev, Denmark) according to manufacturer’s instructions.

Male Wistar rats (~250 g) and male C57BL/6JRj mice (~28 g, 11–12 weeks) were purchased from the Janvier Labs (Le Genest-Saint-Isle, France). Mice and rats were housed with ad libitum access to standard chow and water following a 12:12 h light:dark cycle and were acclimatized for a least a week before experiments. On the day of experiment, rats and mice were transferred to our perfusion facility and anesthetized with a subcutaneous injection of hypnorm/midazolam (rats) (0.3 ml/100 g body weight, per ml: 0.08 mg fentanyl, 2.5 mg fluanisone, 0.45 mg Methyl Parahydroxybenzoate, 0.05 mg Propyl Parahydroxybenzoate, Midazolam: 1.25 mg, Matrix Pharmaceuticals, Hellerup, Denmark) or with an intraperitoneal injection of Ketamine/Xylazine (0.1 ml/20 g) (90 mg/kg Ketamine, Ketaminol Vet, MSD Animal Health; 10 mg/kg Xylazine, Rompun Vet., Bayer Animal Health) respectively, to block the activity of pain sensing nerves and to induce surgical anesthesia. For distal perfusions, the rat or mouse was placed on a heated operating table (37°C), the abdominal cavity was opened by a mid-line incision and the large intestine was excised after tying off the supplying vasculature. Furthermore, the upper half of the small intestine was removed. In the rat, the lower half of the small intestine (~50 cm) was retained and perfused in situ after insertion of a catheter into the upper mesenteric artery, while perfusion effluent was collected through a catheter inserted in the portal vein. The mouse intestine (10–12 cm of the distal part) was perfused by a slightly different pathways (since the upper mesenteric artery is too small to cannulate), perfusing retrogradely through a catheter inserted in the abdominal aorta. To avoid perfusion of other abdominal organs, the spleen and stomach was removed and the kidneys were excluded from the circulation by tying off the renal arteries. For the perfusion of the proximal small mouse intestine the procedure was the same, however the distal part of the intestine was removed and the most proximal part (~11–12 cm) of the small intestine was left untouched. The length of the retained intestine varied little between experiments (coefficient of variation <15%). The perfusion buffer consisted of a Krebs–Ringer bicarbonate buffer supplemented with 0.1% (w/v) BSA (fraction V), 5% (w/v) dextran T-70 [to balance oncotic pressure (Pharmacosmos, Holbaek, Denmark)], and 3.5 mmol/L glucose, and 5 mmol/L pyruvate, fumarate, and glutamate as well as 10 µmol/l 3-isobutyl-1-methylxanthine (IBMX, cat. no. I5879, Sigma Aldrich, Brøndby, Denmark) and 2 ml/L (rat) or 5 ml/L (mouse) Vamin (a mixture of essential and nonessential amino acids; Fresenius Kabi, Bad Homburg, Germany). pH was adjusted to 7.4–7.5. A vascular flow rate of 7.5 ml/min was used for rat perfusions and 2.5 ml/min for mouse perfusions. Luminal stimulations was done at a flow rate of 2.5 ml/min in rats and 0.035 ml/min in mice. Prior to perfusions, the perfusion buffer was gassed with 95% O₂ and 5% CO₂. Perfusions were carried out using a UP100 Universal Perfusion System from Hugo Sachs (Harvard Apparatus, March Hugstetten, Germany), which includes heating to 37°C.

Nle(4),D-Phe(7)]α-melanocyte-stimulating hormone (NDP-α-MSH) was from Phoenix Pharmaceuticals, Inc. (Cat. no. 043-06, Burlingame, CA, USA). HS014 was from Tocris (Cat. no. 1831, Abingdon, UK). Human α-MSH, D-glucose and bombesin were from Sigma Aldrich (Cat. no. M4135, G8270 and B4272, Brøndby, Denmark).

All RNA sequencing data were analyzed using R (version 4.0.3). Datasets included in this analysis are listed in the Table 1 below.

Bulk RNA sequencing datasets were converted into DESeq2 format (1.30.0; Bioconductor). Transcript lengths were obtained from biomaRt (2.42.1; Bioconductor) and imported into the DESeq2 dataset. Fragments per kilobase million (FPKM) values were exported using DESeq2’s in-built function (fpkm). Bulk RNA-seq data presented in heat maps are log2 (average FPKM across samples).

Single-cell datasets were converted into Seurat format (4.0.2; Bioconductor), filtering for only cells with counts for >200 genes, and only genes with counts for >3 cells. Data were normalized to FPKM manually using gene transcript lengths and total reads per sample. For data from Billing et al. (20), all cells were first clustered to obtain five populations as originally published, and data filtered to the two L-cell populations (Insl5 and Nts). Single-cell RNA-seq data presented in heat maps are log2 (average FPKM across cells).

cDNA was analyzed from previously described mouse villus and crypt L-cell samples in the duodenum (19), and mouse ileal L-cell samples (8), including corresponding negative samples. Subsequent RT-qPCR was carried out as previously described (22), with three samples per region, and three technical replicates per sample. The following probes obtained from Life Technologies (Paisley, UK) were used: Gpr119: Mm00731497_s1; Mc4r: Mm00457483_s1. Relative expression of each gene of interest was calculated by comparison to expression of the housekeeper β-actin. The cycle threshold (CT) per gene was calculated as the mean over technical replicates, and the difference (ΔCT) calculated by subtracting each gene CT from β-actin CT. Relative gene expression is represented in figures as 2^ΔCT.

Total GLP-1 concentrations in venous effluents were quantified using an in-house RIA (code: 89390) (23). The assay employs an antibody that exclusively reacts with the amidated C-terminal domain of GLP-1 and thus measures intact GLP-1 (7–36 amide) and the primary metabolite (9–36 amide) with equal potency and it also measures cleaved sequences potentially resulting from mid-site cleavage [e.g. cleaved by neprilysin (24)]. Samples were assayed non-extracted and concentrations were calculated by a four-parameter interpolation to a standard curve containing synthetic GLP-1 7–36 amide (H-6795-GMP, Bachem, Bubendorf, Switzerland) ranging from 5 to 320 pmol/L. I¹²⁵-labeled GLP-1 7–36 amide (a gift from Novo Nordisk A/S, Bagssværd, Denmark) was used as tracer. The standards were prepared in perfusion buffer and run in parallel with the samples. The experimental detection limit was 1 pmol/L and the coefficient of variation was <6% at 20 pmol/L, allowing detection of secretion rates down to 2.5 fmol/min (mouse) and 7.5 fmol/min (rat).

GLP-1 total output (fmol/min) was calculated by multiplying perfusion flow (mouse: 2.5 ml/min and rat: 7.5 ml/min) with the GLP-1 concentration in the perfusion effluents (presented in fmol/L). Hormone outputs are presented as means ± SEM. To test for statistical significance of responses, total GLP-1 outputs were calculated by summing up the outputs during the entire period of stimulus administration (e.g. 15 min of NDP-α-MSH administration); these outputs were compared to the total GLP-1 outputs from the preceding basal period which was of the same duration in case of the rat experiments. In the mouse experiments, the length of the initial baseline period was 10 min whereas stimulation periods were 15 min. In this case, outputs are expressed as average outputs within respective periods (fmol/min). Baseline outputs were calculated based on a similar number of samples from the baseline leading up to first period of test compound stimulation constituting the first baseline and the second baseline being derived from the eight samples immediately before stimulus administration and last seven samples during subsequent baseline period (immediately before BBS stimulation). Baseline outputs were calculated in this manner to control for potential baseline drift over the course of the experiment. EC₅₀ and IC₅₀ values for MC4R activation/inhibition in the in vitro studies were calculated from four-parameter logarithmic fitting. Statistical calculations were performed using GraphPad Prism 7 software (La Jolla, CA, USA), employing one-way ANOVA for repeated measurements followed by Tukey multiple comparison test or Student t-test as indicated in the figure legends. P <0.05 was considered significant. Prior to testing, a D’Agostino–Pearson omnibus normality test was performed to confirm Gaussian distribution of the data. Graphs were constructed in GraphPad Prism 7 (La Jolla, CA, USA) and figures were prepared in an Adobe Illustrator (Adobe Systems Incorporated, San Jose, CA, USA).

In order to use appropriate concentrations of MC4R agonists and antagonist in the experiments on isolated perfused small intestine, we first determined the potencies of alpha-MSH (an endogenous MC4R agonist) and NDP-alpha-MSH (a MC4R super-agonist) in cAMP measurement experiments on rat and human MC4R transfected into COS-7 cells. We moreover determined the inhibitory properties of the MC4R antagonist, HS014 on the two MC4R’s. The rat MC4R was activated by alpha-MSH and NDP-alpha-MSH with EC50 values of 1.5 × 10⁻⁸ M and 9.1 × 10⁻¹¹ M (LogEC50 −7.8 ± 0.06 and −10.04 ± 0.5, n = 3). NDP-alpha-MSH had an E_max that was approximately 120% of that of alpha-MSH ( Figure 1A , n = 3). HS014 [often described as an MC4R antagonist (25, 26)] was in our setup a partial agonist of rat MC4R with an EC₅₀ of 1.5 × 10⁻⁸ M (LogEC50 −7.8 ± 0.54 here, n = 3) and E_max of around 40% of alpha-MSH E_max. Activation of human MC4R showed similar relationships but with potencies that were slightly lower for all three compounds (alpha-MSH, 8.4 × 10⁻⁸ log EC50 7.1 ± 0.1, NDP-alpha-MSH, 2.8 × 10⁻¹⁰ log EC50 −9.6 ± 0.2, and HS014, 3.5 × 10⁻⁸ log EC50 7.5 ± 0.8 ( Figure 1D , n = 3). The antagonistic property of HS014 on the rat MC4R was tested using submaximal concentrations of NDP-alpha-MSH or alpha-MSH corresponding to 60–80% of E_max. Here, an increasing concentration of HS014 revealed inhibition of agonist-induced activity with IC₅₀ values of 2.6 × 10⁻⁸ M and 6.1 × 10⁻⁹ M (log IC50 7.6 ± 0.3 and −8.2 ± 0.2) ( Figures 1B, C , n = 3). Again, HS014 had comparable effects on human MC4R activity ( Figures 1E, F , n = 3).

To investigate whether any effects of systemic MC4R agonist administration on plasma GLP-1 concentrations in mice (11, 12) and humans (14) may result from direct MC4R-activation at the level of the L-cell, we investigated MC4R expression in mouse and human L-cells by reanalyzing raw data from both bulk RNA sequencing (mouse and human) and single cell RNA sequencing (mouse) (8, 19–21) as well as by qPCR (mouse). In mice, MC4R was expressed at higher levels than the other melanocortin receptor isoforms and, as previously reported, was enriched in L-cells relative to the surrounding epithelium ( Figure 2A , left panel), which is predominantly constituted from absorptive enterocytes. Expression levels were relatively low and, especially in the large intestine, lower than the expression of other L-cell-expressed GPCRs (GPR119 and bile acid receptor GPBAR1) with established GLP-1 secretory potential. qPCR of mouse duodenum and ileum further demonstrated enrichment of Mc4r in duodenal, but not ileal, L-cells, and the low expression relative to Gpr119 ( Figure 2B ). By contrast, in humans, MC4R was barely detectable in the intestinal epithelium, and instead MC1R was detectable and enriched in small intestinal L-cells ( Figure 1A , right panel). Although MC1R expression levels were comparable to GPR119 and GPBAR1 in acutely isolated human L-cells, in ileal organoid cultures expression levels of MC1R were much lower.

As the relatively low Mc4r expression, compared with other established pro-secretory GPCRs such as Gpr119, puts into question that GLP-1 responses to systemic MC4R agonists observed in vivo derive from direct actions on the L-cell, other stimulatory intestinal pathways could drive GLP-1 secretion secondary to intestinal MC4R activation. To assess this possibility, we performed studies on isolated and perfused mouse and rat small intestine.

In isolated perfused mouse small intestine—upper half—total GLP-1 secretion during the first baseline period (1–10 min) was on average 25 ± 4.0 fmol/min. Luminal NDP-α-MSH infusion (1 µmol/L) had no effects on GLP-1 output (average secretion: 23 ± 4.8 fmol/min, P = 0.98, n = 6, Figures 3A, B ). Similarly, vascular NDP-α-MSH administration (1 µmol/L) had no effect on GLP-1 secretion (average outputs; preceding baseline = 50 ± 4.8 fmol/min, during NDP-α-MSH administration = 45.9 ± 4.3 fmol/min, P = 0.90, n = 6, Figures 3A, B ). Vascular administration of bombesin (positive control) at the end of experiments robustly increased GLP-1 output ( Figures 3C, D ).

Similar secretory dynamics were found when isolating and perfusing the distal part of the mouse small intestine. In this case, average outputs during initial baseline was 17.4 ± 2.4 fmol/min, whereas average output during luminal administration with NDP-α-MSH (1 µmol/L) was 15 ± 2.2 fmol/min, P = 0.75). Consistent with data from the perfused proximal part of the small intestine, vascular administration of NDP-α-MSH at the same dose also did not stimulate GLP-1 output (average output during stimulation: 26 ± 2.0 fmol/min vs. average output during preceding baseline: 24 ± 2.1 fmol/min. P = 0.37). Vascular administration of bombesin (positive control) at the end of experiment, again, robustly increased GLP-1 output ( Figure 3C ).

In isolated perfused rat distal small intestine, total GLP-1 secretion was equally unaffected by luminal and vascular NDP-α-MSH instillation (1 µmol/L; 100-fold higher than the concentration needed to fully activate rat MC4R shown in the in vitro studies). In this case, total GLP-1 output was initially (1–15 min) 1.1 ± 0.2 pmol and 1.1 ± 0.1 pmol during subsequent luminal NDP-α-MSH administration (P >0.99, n = 6, Figures 3E, F ). Similarly, vascular (intra-arterial) administration of NDP-α-MSH had no effects on GLP-1 secretion (total output during 15 min of stimulation = 1.9 ± 0.2, total output during immediate preceding baseline period (15 min) = 1.9 ± 0.3 pmol, P = 0.98, Figures 3E, F ). Baseline secretion drifted slightly upwards during the course of the experiment (P <0.01) in a manner that was unrelated to the administration of test compounds.

Luminal glucose is a powerful stimulus for GLP-1 secretion. To investigate if the GLP-1 response to glucose is partially driven by MC4R activity, we stimulated the perfused rat small intestine twice with luminal glucose and blocked MC4R activity (by intra-arterial HS014 infusion, 30 nmol/L; >4 fold higher than the IC₅₀ for alpha-MSH-stimulated MC4R activation) during the second glucose stimulation ( Figures 3G, H ). In a separate line of control experiments we applied the same protocol but without HS014 infusion. In both experiments, the first glucose stimulation resulted in a robust and immediate increase in GLP-1 secretion which was similar in effect size (total GLP-1 outputs during respective baselines and luminal glucose administration (in both cases 15 min in total) were: control experiment: Baseline = 1.8 ± 0.2 pmol, luminal glucose = 4.8 ± 0.34 pmol, P <0.001, and in the glucose + HS014 experiment: Baseline = 2.0 ± 0.5 pmol, luminal glucose = 5.6 ± 1.0 pmol, P = 0.01, Figures 3G, H , n = 6). The second glucose response in the control experiment tended to have lower initial peak values compared to the first glucose response, but total secretion during glucose administrations was not different: first response = 4.8 ± 0.3 fmol, second response = 4.7 ± 0.7 fmol, P >0.99, Figure 3H ). Administration of HS014 together with glucose had no effects on the glucose response. Glucose again robustly increased GLP-1 output and the total output with HS014 was similar to the total output during the first glucose stimulation (total GLP-1 outputs (15 min) were: baseline = 3.2 ± 0.6 pmol, glucose + HS014 = 5.4 ± 1.1 pmol, P <0.05 between baseline and glucose response, P = 0.98 between glucose responses, n = 6, Figure 3I ). Thus, the total secretory outputs during the second glucose administration were similar between the control experiment and the experiment with co-administration of HS014 (P = 0.62, n = 6, Figure 3J ).