ADGRL1 binds with glucose

We used a 30 KDa glucose–biotin–PAA conjugate (ESM Fig. 1) and MACS technique to enrich glucose-bound hypothalamic neurons obtained from C57BL/6J mice. The enriched neurons were then lysed and subjected to proteomics analyses (Fig. 1a) to identify the most enriched receptors. We observed that ADGRL1 (encoded by ADGRL1, also known as LPHN1) was at the top of the list of potential glucose-binding proteins (ESM Tables 1 and 2).

Fig. 1

Glucose binds with ADGRL1. (a) Flow chart displaying the steps used in this study to identify a hypothalamic glucoreceptor. (b) Colorimetric assay showing dose-dependent binding of biotinylated glucose with human ADGRL1 stably expressed in CHO cells, n=4. (c, d) Glucose–ADGRL1 dose-dependent binding response obtained using SPRi, in which different concentrations of ADGRL1-expressing CHO cells were injected over immobilised biotinylated glucose (c), or different concentrations of soluble biotinylated glucose were injected over immobilised cells (d), n=3. M, mol/l of CHO cells stably expressing ADGRL1. nM, concentration of biotinylated glucose in nmol/l. KD was 3.44×10−14 mol/l with the approach described in (c) and 4.86×10−9 mol/l with the method described in (d), both demonstrating that glucose binds ADGRL1. Ctrl, control

To verify glucose–ADGRL1 binding, we produced a stable CHO-K1 cell line expressing human ADGRL1 (Uniprot O94910) and incubated the cells with different concentrations of glucose–biotin–PAA or its control conjugate (without glucose). We performed qRT-PCR to validate stable expression of ADGRL1 in CHO-K1 cells (\(^}_}}\) = 10,485, ADGRL1-expressing cells vs 1.0, control cells, n=3, fold relative to control). We then used biotin–streptavidin chemistry to detect the interaction between the conjugates and cells. The cells incubated with the glucose–biotin–PAA displayed a dose-dependent increase in absorbance, which was absent in the control group (Fig. 1b), thereby unequivocally demonstrating the binding of glucose with human ADGRL1. After validating glucose–ADGRL1 binding, we measured KD to quantify binding affinity between the human ADGRL1-expressing cells and the biotinylated glucose or its control conjugate. When the conjugates were used as ligand (immobilised on a biochip) and the cells as analyte (injected over the conjugates), their interaction yielded Kon 3.5×108 [mol/l]−1 s−1, Koff 1.24×10−5 s−1, and KD 3.44×10−14 mol/l (Fig. 1c). In this approach, the cells are presented in a suspension form, hence they offer more binding surface area and may interact with several glucose molecules at the same time, thereby overestimating the binding affinity. To address this limitation, we also immobilised the cells (5×105 cells/ml) on a biochip and then injected different concentrations of soluble glucose conjugates over them. This alternative approach yielded Kon 9.33×104 [mol/l]−1 s−1, Koff 4.54×10−4 s−1, and KD 4.86×10−9 mol/l (Fig. 1d), which may represent a more realistic measurement because only a portion of the cells was available to bind glucose, simulating a natural in vivo interaction. The cells without ADGRL1 yielded a negligible response with the glucose conjugate (ESM Fig. 2a). For the immobilisation procedure, we used a contact-spotting approach (described in the Methods) that demonstrated more optimal conditions and more consistent results than the microfluidic approach (ESM Fig. 2b, c). The use of the cells enabled us to study glucose–ADGRL1 interactions in situ, which presented the receptor in a form closest to its native state.

To account for the possibility that the glucose polymer, used in our initial studies to identify ADGRL1 as a glucose receptor, may not reflect the physiological properties of regular glucose, we further validated glucose–ADGRL1 binding using different concentrations of regular glucose (non-conjugated) immobilised onto 96 well plates and then adding recombinant human ADGRL1 protein solution into the wells to assess glucose–ADGRL1 interaction. Gel electrophoresis and colorimetry confirmed that glucose binds with recombinant ADGRL1 (ESM Fig. 3a, b).

We also determined the potential signal transduction pathway of the glucose–ADGRL1 interaction by measuring second messengers such as cAMP or Ca2+ in the stable CHO cells expressing ADGRL1 or its control in the presence or absence of glucose. We observed that glucose dose-dependently decreased cAMP (ESM Fig. 3c) in an assay used to detect coupling with Gαi protein. There was no change in cAMP or Ca2+ in assays used to detect couplings with Gs or Gq (ESM Fig. 3d, e).

Global or hypothalamic ADGRL1 deficiency causes obesity in mice

To determine the physiological role of ADGRL1 in energy balance, we produced global Adgrl1 KO mice and measured their body weight and food intake. The Adgrl1 KO mice exhibited obesity and hyperphagia was observed when they were about 12 weeks of age (Fig. 2a,b). We validated the knockout of Adgrl1 using RT-qPCR (Fig. 2c). Because the hypothalamus is a major hub controlling body weight, feeding and glucose homeostasis [23, 27,28,29] we next determined tissue-specific distribution of Adgrl1 mRNA in the hypothalamus and defined the role of hypothalamic ADGRL1 in energy homeostasis in male and female mice as described below.

Fig. 2

Adgrl1-deficient mice develop obesity. (a, b) Male global Adgrl1 KO mice have higher body weight (a) and exhibit hyperphagia (b) (at 12 weeks of age). (c) Adgrl1 gene KO was validated by qRT-PCR in the hypothalamus. Adgrl1 KO mice and their littermate WT mice were euthanised at 29 weeks of age for qRT-PCR. (d, e) Adgrl1 knockdown in the VMH (Adgrl1VMH KD) was induced in 8-week-old male and female mice. (d) Body weight in male Adgrl1VMH KD mice and littermate controls was measured at different times; n=15 Ctrl and n=25 Adgrl1VMH KD mice. (e) Body weight in female Adgrl1VMH KD mice and their littermate controls; ovariectomy in corresponding groups was performed 1 week after inducing the Adgrl1 deficiency; n=10 Ctrl and n=8 Adgrl1VMH KD mice. Data are shown as means ± SEM. Two-tailed Student’s unpaired t test or repeated measures ANOVA (1- or 2-way) followed by Bonferroni multiple comparison test: *p<0.05, ***p<0.001 vs Ctrl, over the indicated time points. Ctrl, control; Ovx, ovariectomy/ovariectomised mice

We used an RNA in situ hybridisation procedure to assess the distribution of Adgrl1 in the mouse hypothalamus. We report that Adgrl1 is highly expressed in the VMH (ESM Fig. 4a, b) relative to other hypothalamic regions such as the paraventricular (PVH) and arcuate nucleus (ARC) of the hypothalamus. Adgrl1 is colocalised with Sf1, which encodes steroidogenic factor 1 (SF1), a VMH marker (ESM Fig. 4c). Further, we observed that Adgrl1 was mainly expressed in neurons as determined by the neuronal marker neuronal nuclear protein (NeuN; also known as Rbfox3; ESM Fig. 4d). We analysed Adgrl1–NeuN co-localisation using CellProfiler analysis software. About 94% of Adgrl1-expressing cells in the VMH showed NeuN staining (three sections/mouse brain and four mice/group were analysed).

To knock down Adgrl1 specifically in the VMH (Adgrl1VMH), we generated Adgrl1loxP/loxP mice (ESM Fig. 5a, b) and injected AAV-Cre or its control vector into the VMH of these mice using a stereotaxic surgical procedure (ESM Fig. 5c). The ADGRL1-deficient mice had higher body weight about 9 weeks after inducing ADGRL1 deficiency and gained ~47% more weight compared with their littermate controls by the 24th week following the Adgrl1 knockdown (Fig. 2d). Given the involvement of the VMH in sex-dependent regulation of energy balance [30, 31], we determined whether ADGRL1 deficiency in the VMH produced obesity in female mice. We found that ADGRL1 deficiency alone did not cause obesity in female mice (Fig. 2e). To answer whether oestrogen is responsible for protecting the ADGRL1-deficient mice from obesity, we performed ovariectomy in these mice and their littermate controls. Indeed, the ovariectomised ADGRL1-deficient mice had higher body weight relative to their control ovariectomised mice (Fig. 2e), thereby unmasking the contribution of ADGRL1 in body weight regulation. After completion of the study, we validated the lack of ADGRL1 selectively in the VMH (ESM Fig. 5d) using RNA in situ hybridisation.

Adgrl1VMH-knockdown mice exhibited higher food intake (ESM Fig. 6a) on the 21st week after inducing the ADGRL1 deficiency, at which time the mice were already obese. We then used proton magnetic resonance spectroscopy (1H-MRS) to measure their body composition. Adgrl1VMH-deficient mice had higher fat and lean mass relative to their littermate controls (ESM Fig. 6b, c). Moreover, physical activity was reduced (ESM Fig. 6d) in the ADGRL1-deficient mice compared with their littermate controls. To determine whether the decrease in physical activity contributed to the development of obesity or was a consequence of obesity, we measured physical activity in Adgrl1VMH-deficient mice on the 6th week after inducing Adgrl1VMH deficiency, during which time the mice displayed normal body weight (ESM Fig. 6e). The non-obese Adgrl1VMH-deficient mice were also less active than their littermate controls (ESM Fig. 6f). These findings suggest that the ADGRL1 deficiency is likely responsible for reducing physical activity in mice before they develop obesity. Moreover, the obese and weight-matched ADGRL1-deficient mice had normal energy expenditure when adjusted by total body mass using the MMPC ANCOVA analysis, but lower respiratory exchange ratio compared with their littermate controls (ESM Fig. 7a–d), suggesting elevated lipid oxidation in the ADGRL1-deficient mice.

ADGRL1 in the VMH is involved in feeding responses to glucose or fasting

Glucose influences food intake [2,3,4,5]. According to the glucostatic theory and previous publications [2,3,4,5, 7, 8, 21], food intake is inversely proportional to blood glucose levels under normal circumstances. To determine the role of ADGRL1 in regulating glucose-mediated changes in food intake, we measured feeding responses to glucose administration or overnight fasting in non-obese Adgrl1VMH-deficient mice 3 weeks after inducing ADGRL1 deficiency using AAV-Cre as described above. Based on previous studies using this viral vector approach, 3 weeks are sufficient for mice to recover from stereotaxic surgery and for knockdown of desired genes [32]. The ADGRL1-deficient mice had impaired feeding responses compared with their littermate controls. The control mice did reduce their food intake, as anticipated, following oral or intra-VMH glucose administration, but lack of ADGRL1 in the VMH suppressed these responses (Fig. 3a,b). Conversely, we observed a temporary hyperphagia, as expected, in the control mice when they were refed ad libitum food following an overnight fast (Fig. 3c). However, this response was attenuated in ADGRL1-deficient mice (Fig. 3c). To clarify whether the levels of ADGRL1 are regulated by fasting, we used qRT-PCR to measure Adgrl1 gene expression in the VMH of mice that were fasted overnight. We observed that the fasting downregulated the Adgrl1 expression (Fig. 3d), further suggesting the role of ADGRL1 in responding to changes in energy balance. These results demonstrate the contribution of ADGRL1 in the VMH to mediating feeding responses to glucose and fasting in mice.

Fig. 3

Impaired feeding responses in male Adgrl1VMH KD (knockdown of Adgrl1 in the VMH) mice and effect of fasting on Adgrl1VMH expression. (a, b) Feeding responses to oral (a) or intra-VMH (b) glucose administration (500 mg glucose in 300 µl water, oral gavage, or 5 mmol/l glucose in 2 µl PBS, intra-VMH, administered at 16:00 and 17:00 hours for both) in the 3rd week following Adgrl1VMH deficiency in 11-week-old Adgrl1VMH KD mice and their littermate controls; n=6 or 8 Ctrl and Adgrl1VMH KD mice. Two-tailed Student’s t test: **p<0.01 vs corresponding group without glucose. (c) Feeding response following overnight (18:00 to 09:00 hours) fasting in the 3rd week after Adgrl1VMH deficiency in 11-week-old Adgrl1VMH KD mice and their littermate controls; n=8 Ctrl and Adgrl1VMH KD mice. (d) Overnight (18:00 to 09:00 hours) fasting reduces Adgrl1 mRNA levels (measured by qRT-PCR) in the VMH of 8-week-old C57BL/6 male mice. Two-tailed Student’s unpaired t test or repeated measures two-way ANOVA followed by Bonferroni multiple comparison tests: **p<0.01 vs all other groups at corresponding times. †p<0.05, ††p<0.01 vs Adgrl1VMH KD (non-fasted) at the corresponding time. Ctrl, control; Glu, glucose

ADGRL1 in the VMH regulates insulin secretion and insulin sensitivity

VMH lesions cause hypersecretion of insulin [33, 34]. Because ADGRL1 is highly expressed in the VMH, we measured fasting plasma insulin levels and glucose-stimulated insulin secretion in Adgrl1VMH-deficient mice to determine the involvement of VMH ADGRL1 in regulating insulin secretion. Compared with the control group, the ADGRL1-deficient mice had increased fasting plasma insulin levels at baseline as well as after glucose administration (Fig. 4a), thereby amplifying glucose-stimulated insulin secretion. As expected from previous studies [35, 36], the control mice showed a progressive increase in plasma insulin levels under fasting and glucose-stimulated conditions with age, which was augmented in Adgrl1VMH-deficient mice. The first analyses of fasting plasma insulin levels and glucose-stimulated insulin secretion were performed 3 weeks after inducing the ADGRL1 deficiency. At this time, the ADGRL1-deficient mice and their control group had similar body weights and insulin sensitivity (ADGRL1-deficient mice vs control group: body weight, 26.5 ±1 vs 25.1 ±0.6 g; AUC of blood glucose levels obtained during insulin tolerance test, 966 ± 83.34 vs 884.2 ± 71.48 mmol/l × min, n=11 and 14, respectively), indicating that the fasting hyperinsulinaemia and glucose-stimulated hypersecretion of insulin were consequences of ADGRL1 deficiency and not secondary to obesity or insulin resistance. Given the contribution of the vagus nerve to insulin secretion following hypothalamic lesions [33], we denervated the pancreatic vagal nerve to determine whether it mediates hyperinsulinaemia in the ADGRL1-deficient mice. Within three days of pancreatic vagotomy, the insulin hypersecretion was reversed (Fig. 4b) These findings imply that ADGRL1 in the VMH may be responsible for keeping a check on insulin secretion via the vagus nerve to maintain physiological levels of circulating insulin at baseline and following meals.

Fig. 4

Hyperinsulinaemia in male Adgrl1VMH KD (knockdown of Adgrl1 in the VMH) mice. (a) Fasting (08:00 to 14:00 hours) plasma insulin levels (0 min) and glucose-stimulated insulin secretion (20 min). The weeks on the x-axis indicate the time after inducing Adgrl1VMH deficiency; n=11 Ctrl and n=14 Adgrl1VMH KD (3 weeks following Adgrl1VMH deficiency), n=13 Ctrl and Adgrl1VMH KD (7 weeks following Adgrl1VMH deficiency), n=16 Ctrl and Adgrl1VMH KD (18 weeks following Adgrl1VMH deficiency). Repeated measures two-way ANOVA followed by Bonferroni multiple comparison tests: **p<0.01, ***p<0.001 vs Ctrl at 0 min (for the corresponding time post Adgrl1 deficiency). †p<0.05, ††p<0.01, †††p<0.001 vs all other groups at corresponding times. (b) Pancreatic vagotomy reverses hyperinsulinaemia in Adgrl1VMH KD mice within 3 days of the vagotomy. Vagotomy was performed 4 weeks after inducing the Adgrl1 deficiency. Two-way ANOVA followed by Bonferroni multiple comparison test: ***p<0.001 vs all other groups. (c, d) Impaired insulin sensitivity in Adgrl1VMH KD mice. Insulin tolerance tests (raw data and presented as % of baseline blood glucose level) on the 15th week (c) and 24th week (d) after inducing Adgrl1 deficiency; n=12 Ctrl and n=10 Adgrl1VMH KD mice in (c), n=16 Ctrl and Adgrl1VMH KD mice in (d). Two-way repeated measures ANOVA followed by Bonferroni multiple comparison tests: ***p<0.001. Ctrl, control; Pan. Vgx, pancreatic vagotomy

Next, we measured insulin sensitivity and glucose tolerance in Adgrl1VMH-deficient mice to further determine whether these mice show insulin resistance followed by impairments in glucose regulation. We observed that the Adgrl1VMH-deficient mice had reduced insulin sensitivity, as measured by insulin tolerance tests, compared with their littermate controls at different times throughout the study (Fig. 4c,d). The Adgrl1VMH-deficient mice exhibited impaired glucose tolerance on the 17th and 24th week after inducing the ADGRL1 deficiency (Fig. 5a,b). The impaired glucose tolerance was likely due to the chronic insulin resistance observed in the Adgrl1VMH-deficient mice (Fig. 4c,d). The ADGRL1-deficient mice exhibited fasting hyperglycaemia (Fig. 5c) on the 24th week following ADGRL1 knockdown. Similarly, the ovariectomised female Adgrl1VMH-deficient mice also manifested hyperinsulinaemia, hyperglycaemia and impaired glucose tolerance (Fig. 5d–f). As expected, ovariectomy itself induced metabolic abnormalities; however, these were exacerbated in the ADGRL1-deficient mice, thereby revealing the effects of ADGRL1 on energy and glucose homeostasis in female mice.

Fig. 5

Impaired glucose homeostasis in Adgrl1VMH KD (knockdown of Adgrl1 in the VMH) male and ovariectomised female mice. (a, b) Oral glucose tolerance test in the 17th week (a; n=12 Ctrl and n=20 Adgrl1VMH KD mice) and 24th week (b; n=16 Ctrl and n=12 Adgrl1VMH KD male mice) after inducing Adgrl1VMH deficiency. Bar graphs in (a) and (b) represent the corresponding AUC. (c) Fasting (08:00 to 14:00 hours) blood glucose levels on the 24th week following Adgrl1 deficiency, n=17 Ctrl and n=28 Adgrl1VMH KD male mice. (d, e, f) Ovariectomised Adgrl1VMH knockdown (Adgrl1VMH KD) female mice have higher plasma insulin (d), fasting hyperglycaemia (e) and impaired glucose tolerance (f) compared with their littermate controls. Ovariectomy or sham surgery was performed in 9-week-old mice 1 week after inducing Adgrl1VMH deficiency. Plasma insulin was measured 3 weeks after inducing the Adgrl1 deficiency. Fasting blood glucose levels and oral glucose tolerance were measured 18 weeks after inducing the Adgrl1 deficiency. Data are shown as means ± SEM. Two-way repeated measures ANOVA followed by Bonferroni multiple comparison tests or two-tailed Student’s unpaired t test: *p<0.05, **p<0.01, ***p<0.001 vs Ctrl or sham, ††p<0.01, †††p<0.001 vs all other groups. Ctrl, control; Ovx, ovariectomised mice

To further validate insulin resistance and impaired glucose homeostasis, we performed hyperinsulinaemic–euglycaemic clamps in awake Adgrl1VMH-deficient mice and their littermate controls (Fig. 6a). The ADGRL1-deficient mice needed less exogenous glucose to maintain their clamped blood glucose levels compared with the control mice (Fig. 6b), indicating insulin resistance in the ADGRL1-deficient mice. Moreover, the ADGRL1 deficiency led to impaired baseline glucose production and desensitised insulin-mediated suppression of hepatic glucose production (Fig. 6c–e). The Adgrl1VMH-deficient mice had defective glucose turnover, glycolysis and glycogen synthesis under the hyperinsulinaemic conditions (Fig. 6f–h), indicating impaired insulin sensitivity. The clamp procedure further revealed the sites of insulin resistance. The ADGRL1-deficient mice had decreased insulin-mediated glucose uptake in white adipose tissue (epididymal) but not gastrocnemius skeletal muscle (Fig. 6i,j). We repeated the hyperinsulinaemic–euglycaemic clamp study in the non-obese Adgrl1VMH-deficient mice (weights shown in ESM Fig. 6e), 6 weeks after inducing ADGRL1 deficiency in the VMH, when their weights were similar to that of the control littermates. The weight-matched Adgrl1VMH-deficient mice also manifested insulin resistance (ESM Fig. 8a–f) except in gastrocnemius skeletal muscle (ESM Fig. 8g), thereby supporting the findings obtained from the obese Adgrl1VMH-deficient mice. Together, these results demonstrate that VMH ADGRL1 is indispensable for maintaining physiological plasma insulin levels and insulin sensitivity to consequently influence glucose homeostasis.

Fig. 6

Hyperinsulinaemic–euglycaemic clamps in obese male Adgrl1VMH KD (knockdown of Adgrl1 in the VMH) mice. (a) Clamped blood glucose levels; (b) glucose infusion rate; (c) baseline and (d) clamp hepatic glucose production; (e) hepatic insulin sensitivity; (f) whole-body glucose turnover; (g) whole-body glycolysis; (h) whole-body glycogen synthesis; (i) glucose uptake in adipose tissue and (j) skeletal muscle during hyperinsulinaemic–euglycaemic clamps in the 22nd week following Adgrl1VMH deficiency in 30-week-old male Adgrl1VMH KD mice and their littermate controls; n=7 Ctrl and Adgrl1VMH KD. Data are shown as mean ± SEM. Two-tailed Student’s t test: *p<0.05, **p<0.01, ***p<0.001. Ctrl, Control; HGP, hepatic glucose production

ADGRL1 contributes to glucose sensing

We performed mouse brain slice electrophysiology to determine the role of VMH ADGRL1 in glucose sensing according to the criteria established in previous publications [19, 37]. ADGRL1-expressing and knocked-down neurons were identified in the VMH using a fluorescent microscope as described in the Methods. ADGRL1 neurons exhibited a heterogenous response to glucose and were classified as glucose-excited (GE), glucose-inhibited (GI), high-glucose-excited (HGE), and high-glucose-inhibited (HGI) neurons (Fig. 7), in line with previous publications [19, 37]. We did not observe any GI neurons (0 out of 34 neurons) in the VMH in the ADGRL1-deficient mice, while 18% of recorded neurons (6 out of 34 neurons) were GI in the control group (Fig. 7a,g). Similarly, we did not observe HGE neurons in the ADGRL1-deficient mice, compared with 18% HGE neurons (6 out of 34 neurons) observed in their control littermates (Fig. 7b,h). Lack of ADGRL1 reduced the proportion of glucose-sensing neurons mainly by reducing the number of GI and HGE neurons, thereby increasing the number of glucose non-responsive neurons (Fig. 7c–h). These data demonstrate a role for ADGRL1 in mediating the effects of glucose on neuronal excitability. In addition, the frequency of spontaneous excitatory postsynaptic currents was decreased with 0.2 and 2.5 mmol/l glucose, without affecting the amplitude, in ADGRL1-deficient neurons (Fig. 7i,j). This finding suggests that ADGRL1 plays a role in the ability of glucose to alter glutamate release at VMH synapses, but not on postsynaptic glutamate receptor responses. Altogether, these data implicate ADGRL1 as a crucial component in determining cellular excitability responses to changes in glucose.

Fig. 7

Electrophysiological characterisation of glucose-responsive neurons in the presence (control) and knockdown (Adgrl1VMH KD) of Adgrl1 in the VMH 6 weeks after inducing Adgrl1 deficiency in 14-week-old non-obese male mice. Representative traces are shown for the categories that were different between the groups in their response to glucose. Six out of 34 recorded neurons showed an increase in action potential firing in response to change in glucose concentration from 2.5 to 0.2 mmol/l (GI neurons); only two neurons in the Adgrl1VMH KD group showed such a trend and this was non-significant (a). Six out of 34 recorded neurons showed an increase in action potential firing in response to change in glucose concentration from 2.5 to 10 mmol/l (HGE neurons); only three neurons in the Adgrl1VMH KD group showed such a trend and this was non-significant (b). (c, d) The number of GE neurons (c) and HGI neurons (d) was similar between the groups. (e, f) The remaining recorded neurons were not responsive to low (e) or high (f) glucose. (g, h) Summary of the proportion of changes in glucose-sensing neurons. (i, j) The frequency of spontaneous excitatory postsynaptic currents was reduced in the Adgrl1VMH KD mice at 2.5 and 0.2 mmol/l glucose (i), but there was no change in the mean amplitude of sEPSCs at different glucose concentrations (j). Data are presented from individual recorded neurons. Two or three neurons per brain slice were recorded and two or three slices per mouse (five mice per group) were included in this study. Two-tailed Student’s paired t test or χ2 test: *p<0.05, **p<0.01, ***p<0.001. NR, not responsive; sEPSC, spontaneous excitatory postsynaptic current