Glucose-dependent insulinotropic polypeptide regulates body weight and food intake via GABAergic neurons in mice

The development of single-molecule co-agonists for the glucagon-like peptide-1 (GLP-1) receptor (GLP-1R) and glucose-dependent insulinotropic polypeptide (GIP) receptor (GIPR) is considered a breakthrough in the treatment of obesity and type 2 diabetes. But although GIPR–GLP-1R co-agonism decreases body weight with superior efficacy relative to GLP-1R agonism alone in preclinical1–3 and clinical studies4,5, the role of GIP in regulating energy metabolism remains enigmatic. Increasing evidence suggests that long-acting GIPR agonists act in the brain to decrease body weight through the inhibition of food intake3,6–8; however, the mechanisms and neuronal populations through which GIP affects metabolism remain to be identified. Here, we report that long-acting GIPR agonists and GIPR–GLP-1R co-agonists decrease body weight and food intake via inhibitory GABAergic neurons. We show that acyl-GIP decreases body weight and food intake in male diet-induced obese wild-type mice, but not in mice with deletion of Gipr in Vgat(also known as Slc32a1)-expressing GABAergic neurons (Vgat-Gipr knockout). Whereas the GIPR–GLP-1R co-agonist MAR709 leads, in male diet-induced obese wild-type mice, to greater weight loss and further inhibition of food intake relative to a pharmacokinetically matched acyl-GLP-1 control, this superiority over GLP-1 vanishes in Vgat-Gipr knockout mice. Our data demonstrate that long-acting GIPR agonists crucially depend on GIPR signaling in inhibitory GABAergic neurons to decrease body weight and food intake.

The development of single-molecule co-agonists for the glucagon-like peptide-1 (GLP-1) receptor (GLP-1R) and glucose-dependent insulinotropic polypeptide (GIP) receptor (GIPR) is considered a breakthrough in the treatment of obesity and type 2 diabetes.But although GIPR-GLP-1R co-agonism decreases body weight with superior efficacy relative to GLP-1R agonism alone in preclinical [1][2][3] and clinical studies 4,5 , the role of GIP in regulating energy metabolism remains enigmatic.Increasing evidence suggests that long-acting GIPR agonists act in the brain to decrease body weight through the inhibition of food intake 3,[6][7][8] ; however, the mechanisms and neuronal populations through which GIP affects metabolism remain to be identified.Here, we report that long-acting GIPR agonists and GIPR-GLP-1R co-agonists decrease body weight and food intake via inhibitory GABAergic neurons.We show that acyl-GIP decreases body weight and food intake in male diet-induced obese wild-type mice, but not in mice with deletion of Gipr in Vgat(also known as Slc32a1)-expressing GABAergic neurons (Vgat-Gipr knockout).Whereas the GIPR-GLP-1R co-agonist MAR709 leads, in male diet-induced obese wild-type mice, to greater weight loss and further inhibition of food intake relative to a p ha rmacokinetically matched acyl-GLP-1 control, this superiority over GLP-1 vanishes in Vgat-Gipr knockout mice.Our data demonstrate that long-acting GIPR agonists crucially depend on GIPR signaling in inhibitory GABAergic neurons to decrease body weight and food intake.
We found that chow-fed male Vgat-Gipr KO mice show normal body weight but improved glucose metabolism.Male Vgat-Gipr KO mice also mimic the phenotype of the global 24 and neuronal 3 Gipr KO mice when fed a regular chow diet.Chow-fed Vgat-Gipr KO mice show no overt differences in body weight, body composition or food intake relative to wild-type controls (Fig. 2a-d).Absorption and utilization of nutrients is also not different between chow-fed Vgat-Gipr KO mice and wild-type controls (Fig. 2e-g), although fatty acid oxidation is decreased in the Vgat-Gipr KO mice (Fig. 2h).Vgat-Gipr KO mice show no differences in locomotor activity (Fig. 2i), energy expenditure (Fig. 2j) or fasting levels of blood glucose (Fig. 2k), but insulin levels are decreased (Fig. 2l).Chow-fed Vgat-Gipr KO mice further show improved glucose tolerance (Fig. 2m) without alterations in insulin sensitivity (Fig. 2n) or plasma levels of triglycerides and cholesterol (Fig. 2o,p).Similar to the chow-fed male Vgat-Gipr KO mice, female Vgat-Gipr KO mice show no difference in body weight relative to wild-type controls, even when fed a HFD (Extended Data Fig. 3a).Female Vgat-Gipr KO mice show no difference in body composition or food intake relative to wild-type controls, but locomotor activity is slightly enhanced, without changes in fatty acid oxidation, nutrient utilization and energy expenditure (Extended Data Fig. 3b-h).Female Vgat-Gipr KO mice further show decreased blood glucose with normal fasting levels of plasma insulin, but improved glucose tolerance without changes in insulin sensitivity, HbA1c and plasma levels of triglycerides, cholesterol and free fatty acids (Extended Data Fig. 3i-p).
We found that Vgat-Gipr KO mice are resistant to central GIP effects on cFos neuronal activation and food intake.Based on recent data indicating that GIP acts on hindbrain GIPR neurons to regulate food intake 6 , we next assessed whether validated long-acting (fatty acid acylated) GLP-1R and GIPR agonists 2,3,8 (Extended Data Fig. 4a) differ in their ability to induce neuronal activation in this area.After single subcutaneous administration, we find fluorescently labeled acyl-GLP-1 Cy5 and acyl-GIP Cy5 to substantially accumulate in the area postrema (Fig. 3a); however, although we find acyl-GLP-1 Cy5 to induce cFos neuronal activity in the area postrema and the nucleus tractus solitarius (NTS), acyl-GIP Cy5 induced cFos activity only in the area postrema, but not in the NTS (Fig. 3a-c).Notably, while acyl-GIP Cy5 in wild-type mice solidly induced cFos activity in the area postrema, this effect was largely blunted in the Vgat-Gipr KO mice, despite unchanged accumulation of acyl-GIP Cy5 in the area postrema (Fig. 3d,e).Preserved accumulation of acyl-GIP Cy5 in the area postrema of the Vgat-Gipr KO mice is not unexpected, given that Vgat is only expressed in ~18% of Gipr cells in mice 3,8 , but not in mice with Nes-cre-mediated neuronal loss of Gipr 3 .Accumulating evidence indicates that GIPR agonism is also a vital constituent to GIPR-GLP-1R co-agonism.The GIPR-GLP-1R co-agonist MAR709 leads relative to a pharmacokinetically matched acyl-GLP-1 to greater weight loss and further inhibition of food intake, but this superiority vanishes in mice with neuronal loss of Gipr 3 .And while the GIPR-GLP-1R co-agonist tirzepatide promotes insulin secretion in isolated human islets primarily via the GIP receptor 10 , long-acting GIPR agonists attenuate the emetic effect of GLP-1R agonism in experimental animals 11,12 and hence likely contribute to greater tolerability of GIPR-GLP-1R co-agonism relative to GLP-1R agonism at higher doses.Although the mechanisms and neuronal populations through which GIP affects body weight and food intake have yet to be identified, these data collectively indicate that GIPR agonism is a vital constituent to the metabolic efficacy and tolerability of GIPR-GLP-1R co-agonism.
Several studies have recently assessed the expression profile of Gipr in the brain using single-cell RNA-sequencing (scRNA-seq) analysis, revealing that Gipr is expressed in a variety of different cells types within the hypothalamus and the hindbrain, including neurons, mesenchymal cells, mural cells and oligodendrocytes 7,[13][14][15] .In both of these brain areas, expression of Gipr is found in cells or neurons that express Slc32a1 (also known as Vgat), a marker indicative of inhibitory GABAergic neurons 6,14,16,17 .GABAergic Gipr neurons seem crucial for the anti-emetic effect of GIPR agonism 11,13 , but their role in energy metabolism remains unknown.Emphasizing their potential role in energy balance regulation, Vgat-expressing GABA neurons are implicated in the control of eating behavior [18][19][20] , and while optogenetic stimulation of VGAT neurons in the lateral hypothalamus promotes food intake, genetic ablation of these neurons has the opposite effect 18 .Although being expressed in only 14-18% of all Gipr cells in the hypothalamus 15 and the hindbrain 14 , Vgat it is found in around 32% of hypothalamic Gipr neurons 15 and, depending on the study, in up to 55% of hindbrain Gipr neurons 14,16,17,21 (Extended Data Table 1).Based on the expression of Gipr in GABAergic neurons 7,11,14,15 , and the demonstration that selective activation of hypothalamic and hindbrain GIPR neurons decreases food intake in mice 6,7 , we here assessed whether the metabolic effects of GIP and GIPR-GLP-1R co-agonism depend on GIPR signaling in inhibitory GABAergic neurons.

Letter
https://doi.org/10.1038/s42255-023-00931-7 the hindbrain, including for example oligodendrocytes and endothelial cells 14 (Extended Data Table 1).Nonetheless, consistent with previous scRNA-seq data showing that Vgat is particularly enriched in Gipr neurons within the area postrema 11,[13][14][15]17,21 (Extended Data Table 1), these data hence demonstrate that the acyl-GIP Cy5 -induced cFos activation in the area postrema is almost exclusively attributed to Vgat-expressing inhibitory GABAergic neurons. In wid-type and Vgat-Gipr KO mice, we also found that acyl-GIP Cy5 , after a single subcutaneous administration, accumulates in the hypothalamic median eminence, along with increased cFos activation in the arcuate nucleus and paraventricular nucleus in wild-type mice, but not in Vgat-Gipr KO mice (Extended Data Fig. 5a-d).No changes were observed in cFos activation in the ventromedial and dorsomedial hypothalamus (Extended Data Fig. 5b,e,f) or the lateral parabrachial nucleus (Extended Data Fig. 5gh).We next assessed whether acyl-GIP differentially affects food intake in HFD-fed diet-induced obese (DIO) wild-type and Vgat-Gipr KO mice.While single subcutaneous bolus administration of acyl-GIP (100 nmol kg −1 ) acutely decreased food intake in male and female HFD-fed wild-type mice, administration of acyl-GIP failed to affect food intake in male and female Vgat-Gipr KO mice (Fig. 3f-i).These data hence indicate  Cumul.food intake (g) that acyl-GIP essentially requires GIPR signaling in Vgat-expressing inhibitory GABAergic neurons to acutely decrease food intake.We found that GIP and GIPR-GLP-1R co-agonism decreases body weight and food intake in mice via GIPR signaling in inhibitory GABAergic neurons.We assessed whether the body weight-lowering effect of acyl-GIP and GIPR-GLP-1R co-agonism depend on GIPR signaling in GABAergic neurons.In DIO wild-type mice, daily subcutaneous treatment with acyl-GIP (100 nmol kg −1 ) for 26 days significantly decreased body weight and food intake relative to vehicle controls (Fig. 4a-c and Extended Data Fig. 6a,b).The decrease in body weight in mice treated with acyl-GIP was accompanied by a decrease in fat but not lean tissue mass (Fig. 4d,e), without overt changes in blood glucose, insulin, glucose tolerance, insulin sensitivity or plasma levels of triglycerides (Fig. 4f-j).Consistent with the ability of acyl-GIP to decrease body weight and food intake in DIO wild-type mice (Fig. 4a-c), treatment with the GIPR-GLP-1R co-agonist MAR709 led to greater weight loss and further decrease in food intake and fat mass relative to mice treated with acyl-GLP-1 (Fig. 4a-d).In DIO wild-type mice, MAR709 and acyl-GLP-1 equally decreased fasting levels of blood glucose and insulin (Fig. 4f-g), but with superiority of MAR709 over acyl-GLP-1 to improve glucose tolerance (Fig. 4h).Insulin sensitivity, as estimated by HOMA-IR, was equally improved by MAR709 and acyl-GLP-1 (Fig. 4i), without effects of either treatment on plasma levels of triglycerides (Fig. 4j).
In contrast to DIO wild-type mice, treatment with acyl-GIP in Vgat-Gipr KO mice failed to affect body weight, food intake or body composition relative to vehicle controls (Fig. 4k-o).Treatment of Vgat-Gipr KO mice with acyl-GIP had no effect on blood glucose despite a slight decrease in fasting insulin levels (Fig. 4p,q).Vgat-Gipr KO mice treated with acyl-GIP show no difference in glucose control relative to vehicle controls, but display improved insulin sensitivity, as estimated by HOMA-IR (Fig. 4r,s), without changes in plasma triglycerides (Fig. 4t).Consistent with the demonstration that acyl-GIP loses its ability to decrease body weight and food intake in Vgat-Gipr KO mice (Fig. 4k-m), we found that weight loss induced by MAR709 was indistinguishable from that under acyl-GLP-1 treatment in Vgat-Gipr KO mice (Fig. 4k,l).The observed superiority of MAR709 over acyl-GLP-1 to further inhibit food intake in wild-type mice (Fig. 4c) likewise vanished in the Vgat-Gipr KO mice (Fig. 4m).MAR709 and acyl-GLP-1 further equally decreased fat mass, as well as fasting levels of blood glucose and insulin in Vgat-Gipr KO mice (Fig. 4o-q), with nearly identical improvements in glucose tolerance and insulin sensitivity (Fig. 4e,s) and unchanged plasma levels of triglycerides (Fig. 4t).In summary, these data show that GIPR signaling in inhibitory GABAergic neurons is essential for the ability of acyl-GIP to decrease body weight and food intake and also, for the superior metabolic effects of the GIPR-GLP-1R co-agonist MAR709 over a matched acyl-GLP-1 control.Our data are consistent with our previous report demonstrating that acyl-GIP and MAR709 act in the brain to regulate body weight and food intake via central nervous system GIPR signaling 3 .Furthermore, our data corroborate that GIPR signaling is a genuine contributor to the metabolic efficacy of GIPR-GLP-1R co-agonism, in that it drives greater weight loss and further inhibition of food intake relative to GLP-1R agonism alone.Notably, our data are further in line with the recent demonstration that a long-acting GIPR agonist decreases body weight in healthy humans 25 .
We here show that long-acting GIPR agonists and the GIPR-GLP-1R co-agonist MAR709 decrease body weight and food intake in DIO mice via GIPR signaling in inhibitory GABAergic neurons.Limitations to our study include the lack of publicly available antibodies to reliably detect GIPR, which would be of value to further delineate the neuronal mechanisms by which GIP regulates energy metabolism in GABAergic neurons.Such antibodies would have also been useful to demonstrate lack of co-staining between VGAT and GIPR in Vgat-Gipr KO mice and to further characterize the Vgat/Gipr coexpressing neurons.Published scRNA repositories 14,16,17,21 , however, show that ~81-93% of the Vgat/Gipr-expressing cells in the hypothalamus and the dorsal vagal complex (DVC) are neurons (Extended Data Fig. 6c) and that Vgat/Gipr coexpressing neurons in the DVC belong to five out of seven GABAergic neuron clusters with ~37% and ~19% localizing to clusters GABA5 and GABA4, respectively (Extended Data Fig. 6d-g).
Further limitations of our studies include that we cannot exclude the potential for variable long-acting GIPR agonists to differ in central GIPR signaling based on differences in pharmacokinetics and pharmacodynamics.Differences in pharmacokinetics might also explain why we see that acyl-GIP Cy5 restricted cFos activation in the area postrema, whereas other studies using different GIPR agonists found cFos activation in the NTS 6,11 .Although unlikely, the possibility that the pharmacokinetics of the acyl-GIP that we used slightly differs from acyl-GIP Cy5 may be a similar consideration.Another limitation is that the HFD-fed wild-type and Vgat-Gipr KO mice had a strong tendency to shred food, which prevented deeper analysis of eating behavior, including assessment of cumulative food intake and meal patterns.Notably, the demonstrated low (~14-18%) abundance of Vgat in Gipr cells using scRNA-seq analysis may be an underestimate due to limitations in complementary DNA library preparations or due to the general difficulty to adequately detect low-abundant transcripts.Finally, clarification is needed on whether acyl-GIP and GIPR-GLP-1R co-agonists decrease body weight and food intake exclusively via hindbrain GABAergic neurons or whether also GABAergic GIPR neurons in other brain regions are required for the effects on body weight and food intake shown in this study.

Animals and housing conditions
Experiments were performed in accordance with the Animal Protection Law of the European Union after permission by the Government of Upper Bavaria.Mice were fed ad libitum with either chow diet (1314, Altromin) or HFD (D12331, Research Diets) and were kept at 22 ± 2 °C with constant humidity (45-65%) and a 12-h light-dark cycle.C57BL/6J Vgat-ires-cre knock-in mice were purchased from The Jackson Laboratory (028862).Gipr flx/flx mice 22,23 were crossed with Vgat-cre +/− mice to obtain Vgat-cre +/− Gipr flx+/flx+ (Vgat-Gipr KO).Vgat-cre +/− Gipr flx-/flx-were used as WT controls.For metabolic phenotyping and assessment of drug effects, age-matched mice were grouped based on genotype and double-housed at room temperature (22 ± 2 °C).For the examination of drug effects, mice were fed a HFD approximately 20 weeks before the start of the studies.Body composition was analyzed using a magnetic resonance whole-body composition analyzer (EchoMRI).For pharmacological studies, mice were treated for the indicated time duration with either acyl-GIP (100 nmol kg −1 ) or 10 nmol kg −1 of acyl-GLP-1 or MAR709 (Extended Data Fig. 1a).

Drug development
Commercially available maleimide-functionalized fluorophores were conjugated to the free cysteine residue of the appropriate and food intake (c) of 38-week-old male C57BL/6J WT mice treated daily with vehicle (Vhcl), acyl-GIP (100 nmol kg −1 ) or 10 nmol kg −1 of either acyl-GLP-1 or MAR709 (n = 7-8 each group).d-j, Fat and lean tissue mass in 38-week-old male C57BL/6J WT mice (d,e), fasting levels of blood glucose and insulin in 38-week-old male C57BL/6J WT mice (f,g), glucose tolerance in 38-week-old male C57BL/6J WT mice (h), HOMA-IR in 38-week-old male WT mice (i) and plasma levels of triglycerides in 35-week-old male C57BL/6J WT mice (j) (n = 7-8 each group).k-m, Body weight development (k), placebo-corrected weight loss after 26 days of treatment (l) and food intake (m) of 35-week-old male C57BL/6J Vgat-Gipr KO mice treated daily with vehicle, acyl-GIP (100 nmol kg −1 ) or 10 nmol kg −1 of either acyl-GLP-1 or MAR709 (n = 7-8 each group).n-t, Fat and lean tissue mass in 35-week-old male C57BL/6J KO mice (n,o), fasting levels of blood glucose and insulin in 35-week-old male WT mice (p,q), glucose tolerance in 35-week-old male C57BL/6J WT mice (r), HOMA-IR in 35-week-old male WT mice (s) and plasma levels of triglycerides in 35-week-old male C57BL/6J WT mice (t) (n = 7-8 mice each group).Data in a,h,k,r were analyzed using repeated measures two-way ANOVA and with Bonferroni post hoc comparison for individual time points.Data in c,m were analyzed using Fishers LSD test.Data in b,d-g,i,j,l,n-q,s,t were analyzed using ordinary one-way ANOVA.Data in g,i were analyzed using a Student's two-tailed t-test.Food intake in c,m was assessed per cage in doublehoused mice.Cages with mice shredding food were excluded from the analysis.Data area mean ± s.e.m.; NS, not significant; *P < 0.05; **P < 0.01 and ***P < 0.001.Individual P values are shown in the Source Data file, unless P < 0.0001.

Fasting glucose, insulin, ipGTT and ipITT
Fasting levels of blood glucose and insulin were measured in 6-h fasted mice.Insulin was measured by Ultra-Sensitive Mouse Insulin ELISA kit (90080, Crystal Chem) according to the manufacturer's instructions.For assessment of glucose tolerance, glucose was administered intraperitoneally at a dose of 1.75 g kg −1 .For assessment of insulin tolerance, insulin (Humalog, Eli Lilly and Co.) was injected intraperitoneally at a dose of 0.75 UI kg −1 (HFD-fed mice) or 0.5 UI kg −1 (chow-fed mice).

Bomb calorimetry
Assimilated energy and assimilation efficiency were assessed using the C200 Oxygen Bomb Calorimeter (IKA).Feces were collected over 7 days, dried at 65 °C until weight was consistent, before measuring the food/fecal energy content.

Indirect calorimetry
Energy expenditure, food intake, RER, FA oxidation and locomotor activity were assessed for 3-4 consecutive days in single-house mice using a climate-controlled indirect calorimetric system (TSE Phenomaster, TSE Systems).Mice were given a 24-h acclimatization phase before the start of the measurements.Data for energy expenditure were analyzed using ANCOVA with body weight as a covariate as previously suggested 26,27 .FA oxidation (kcal h −1 ) was calculated by the formula, energy expenditure (kcal h −1 ) × (1 − RER)/0.3(ref.28).

Immunofluorescence
For assessment of cFos, mice were habituated to a daily injection by subcutaneous saline administration for three constitutive days.On day 4, mice were treated subcutaneous with a single dose of acyl-GIP Cy5 or acyl-GLP-1 Cy5 (150 nmol kg −1 ).The mice were exposed to CO 2 overdose 90 min after drug exposure and were briefly perfused with ice-cold TBS following by buffered 4% formaldehyde.Twenty-four hours after fixation, brains were coronally cryosectioned and 35-μm-thick slices were immunolabelled with the monoclonal rabbit anti-cFos antibody (MA5-15055, Invitrogen, 1:400 dilution) and the anti-rabbit Alexa546 secondary antibody (A10040, Invitrogen, 1:2,000 dilution).According to the Allen Mouse Brain Atlas, the DVC containing the area postrema and NTS was captured by ×20 objective in z-stack mode using a Leica SP8 confocal microscope.In each region, the number of cFos-positive cells was automatically counted in using Fiji/ImageJ software.For cFos quantification, whole brain slices were scanned in z-stack mode with AxioScan 7 digital slide scanner (Zeiss, ZEN Blue v.3.5, ×20 objective) and imaged using LAS X (v.3.5.7.23225,Leica Microsystems).The Allen Brain Atlas was imputed and aligned to the whole slide images using Fiji with the ABBA plugin and the number of cFos-positive cells in each identified region was measured by using QuPath v.0.4.4 software 30 .

Immunohistochemistry for α-and β-cell volume and islet size
Pancreata were fixed in 10% formalin (HT501128, Sigma-Aldrich) for 24 h at room temperature and processed for paraffin embedding (Tissue Tec VIP.6, Sakura Europe).Paraffinized pancreata were exhaustively cross-sectioned into 3-4 parallel, equidistant slices per case.Maintaining their orientation, the tissue slices were vertically embedded in paraffin.After co-staining for insulin (monoclonal rabbit anti-insulin, 3014, Cell Signaling 1:800 dilution; AlexaFluor750-conjugated goat anti-rabbit, A21039, Invitrogen 1:100 dilution) and for glucagon (polyclonal guinea pig anti-glucagon, M182, Takara 1:3,000 dilution; goat anti-guinea pig AF555, A21435, Invitrogen 1:200 dilution) nuclei were labeled with Hoechst33342 (H1399, Thermo Fisher, 7.5 μg ml −1 ).The stained tissue sections were scanned with an AxioScan 7 digital slide scanner (Zeiss, ZEN Blue v.3.5)equipped with a ×20 magnification objective.Quantification of insulin-or glucagon-expressing cells was performed on the entire tissue sections using the image analysis software Visiopharm.The insulin-or glucagon-expressing cells were classified automatically using the fluorescence intensity of each hormone.The β-cell volume (mg) was calculated by multiplying the detected relative insulin-positive cell area by total pancreatic weight.The α-cell volume (mg) was similarly calculated based on the detected glucagon-positive cell area.The area of the pancreatic islet was calculated based on the insulin and glucagon-positive area.

Serum analysis
Blood was collected by cardiac puncture during organ withdrawal, stored on ice and centrifuged (2,500g for 10 min at 4 °C) for serum separation and collection.The levels of leptin (KMC2281, Leptin mouse ELISA, Invitrogen), GIP (EZRMGIP-55K, rat/mouse GIP ELISA, Merck Millipore) and GLP-1 (1508, Mouse GLP-1 ELISA, Crystal Chem) were measured according to the manufacturer's instructions.

Statistical analysis
Statistical analyses were performed using GraphPad Prism v.9 and SPSS v.28.0.1.1.Analysis of energy expenditure were performed using ANCOVA with body weight as a covariate as previously suggested 26,27 .For each analysis, the statistical tests and sample sizes are indicated in the figure legends.P < 0.05 was considered statistically significant.All data represent mean ± s.e.m.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
The data used for the statistical analysis are available in source data files, along with the GraphPad Prism-derived statistical reports as appropriate, which contain the mean difference between the treatment groups, 95% confidence intervals, the significance summary and exact P values (unless P < 0.0001).The used scRNA-seq datasets are available via the Gene Expression Omnibus under accession codes GSE160938, GSE166649 and GSE168737.The HypoMap is available in an interactive CELLxGENE viewer (https://www.mrl.ims.cam.ac.uk).The corresponding Seurat object is deposited at University of Cambridge's Apollo Repository (https://doi.org/10.17863/CAM.87955).Other used databases include the Allen Mouse Atlas.All raw images are provided in the source data files, with the exception of the histology pictures for Extended Data Figs.1m-s and 2d-f, which were too large for public repositories; due to the large file size of these pictures, they are only available upon request.Source data are provided with this paper.and Gipr in each cell, with the color corresponding to log-normalized UMI counts scaled to the maximum of each gene (f).The dot plot of the top 5 genes most associated with each GABAergic neuron population as determined by Wilcoxonrank-sum test (g).Data in panel A were analyzed using ordinary 1-way ANOVA, data in panel B were analyzed using ordinary 1-way ANOVA and using Holm-Sidak multiple comparison test.Date represent means ± SEM; asterisks indicate * p < 0.05; ** p < 0.01 and *** p < 0.001.Individual p values are shown in the Data Source file, unless p < 0.0001.Data shown in panels C, D, E, F, G were generated based on publicly available snRNA Seq datasets of hypothalamus (Steuernagel et al. 15 ) or DVC (Lugwig et al. 17 ).