NMDA Receptors in POMC Neurons Connect Exercise With Insulin Sensitivity

A sedentary lifestyle combined with a high-calorie diet increases the risk of obesity and associated metabolic disorders such as type 2 diabetes, cardiovascular disease, and stroke (1). Health care providers emphasize dietary adjustments and regular exercise to combat these conditions (2). Exercise enhances insulin sensitivity, promoting muscle glucose uptake and reducing hepatic glucose production (3,4). While effects of exercise on muscle and liver are well studied, emerging evidence underscores the brain, particularly the hypothalamus, as a crucial mediator of metabolic benefits of exercise (5–8).

Within the hypothalamus, the arcuate (ARC) nucleus anorexigenic proopiomelanocortin (POMC) neurons and orexigenic neuropeptide Y (NPY)/agouti-related peptide (AgRP) neurons are crucial for maintaining energy balance and blood glucose homeostasis (9–12). These neurons respond dynamically to various metabolic cues and challenges (13–15). After high-intensity interval exercise (HIIE), ARC NPY/AgRP neurons are transiently inhibited while ARC POMC neurons are activated and maintain this activity for days (16–18), potentially leading to reduced appetite and improved insulin sensitivity (3,16,19). However, whether there is a direct connection between exercise-induced changes in hypothalamic neuron activity and altered metabolism remains unknown.

Activation of ARC POMC neurons postexercise resembles the activation of ARC NPY/AgRP neurons in response to food deprivation (20). The fasting-induced activation of ARC NPY/AgRP neurons is dependent on N-methyl-d-aspartate receptors (NMDARs) (20). While NR1, a compulsory subunit of NMDARs, has been shown to control energy balance in ARC AgRP neurons, prenatal deletion of NR1 in ARC POMC neurons failed to alter energy homeostasis (20). Importantly, whether NMDAR activity is crucial for the exercise-induced activation of hypothalamic POMC neurons or for maintaining proper glucose metabolism through the activation of ARC POMC neurons postexercise remains unclear.

The aim of the current study is to test the hypothesis that NMDARs are necessary to activate ARC POMC neurons following exercise and improve glucose metabolism. Exercise-induced effects on intrinsic membrane properties and synaptic responses of ARC POMC neurons are assessed. We also examine the requirement of NMDARs in ARC POMC neurons to regulate basal glucose homeostasis as well as insulin sensitivity following exercise.

Male pathogen-free mice were housed under standard laboratory conditions (12-h day/night cycle, lights on at zeitgeber time 0 [ZT0], or 6:00 a.m.) and in a temperature-controlled environment. All experiments were performed in accordance with the guidelines established by the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by The University of Texas at Austin Institutional Animal Care and Use Committee. Mice were provided a Harlan Teklad 18% protein chow diet or high-fat/high-sucrose (D12331i; Research Diets) diet and water ad libitum unless otherwise noted. For NR1 deletion in adult POMC neurons, POMC-CreER^T2 mice (21) were crossed with NR1^loxp/loxp mice (20), generating POMC-CreER^T2::NR1^loxp/loxp mice. Then we crossed with td-tomato reporter mice to identify POMC neurons.

Tamoxifen (15 mg/mL; Sigma-Aldrich) dissolved in corn oil (Sigma-Aldrich) was administered according to body weight (150 mg/kg i.p.) in both 8-week-old male POMC-CreER^T2::NR1^loxp/loxp::td-tomato and control mice. Mice were allowed 2 weeks of minimal disturbance following tamoxifen injection; afterward they were cleared for experiments.

Body weight was measured weekly up to 22 weeks of age and body composition was measured with nuclear magnetic resonance (Bruker minispec; Bruker) at 24 weeks of age. For the high-fat diet (HFD) cohort, chow diets were replaced by HFD (D12331i; Research Diets) at 8 weeks of age. The HFD cohort was also injected with tamoxifen at 8 weeks of age.

Age- and body weight–matched chow-fed male mice were divided into sedentary and exercise groups. Mice that were at least 25 g were selected for this protocol. Motorized treadmills (Exer-6; Columbus Instruments, Columbus, OH) were used for exercise. Mice were coaxed to stay on the treadmill (sedentary group) or continue running on the treadmill with use of an electric stimulus (0.25 mA × 163 V and 1 Hz) generated by a shock grid present at the treadmill base and tapping of their tails with a brush.

Mice were familiarized to the treadmills for 2 days before the exercise bout: familiarization day 1, 5 min at the speed of 0 m/min followed by 5 min at 8 m/min and then 5 min at 10 m/min; familiarization day 2, 5 min at 0 m/min followed by 5 min at 10 m/min and then 5 min at 12 m/min.

Mice rested on the treadmill for 5 min prior to performing the 1 h of HIIE, consisting of 3 × 20 min intervals (5 min at the speed of 12 m/min, 10 min at the speed of 17 m/min, 5 min at the speed of 22 m/min), without rest between intervals (16). Mice rested a minimum of 1 week between any experiments involving this exercise protocol.

Mice were placed on the treadmills with the shock grid for 1 h, while the speed was 0 m/min. This accounted for the stress that can be experienced by being coaxed to stay on the treadmill (22).

The fasting period began at ZT2. After a 4-h fast, 10- to 14-week-old male mice received injections of 1.5 g/kg d-glucose i.p. Blood glucose was measured from tail blood with a glucometer at serial time points as indicated in the figures.

The fasting period began at ZT2. After a 4-h fast to empty the stomach, 10- to 14-week-old male mice received injections of insulin (1.2 units/kg i.p.). Blood glucose was measured from tail blood as previously described (23).

Mice were subjected to the familiarization of treadmills 2 days prior to the exercise bout, as described in Exercise Groups and Sedentary Group. The fasting period began at ZT2. For the 3 h after HIIE ITT, mice performed HIIE in the first hour of a 4-h fast. After the fast, mice received injections of insulin (0.5 units/kg i.p.) (Fig. 3A). Mice performed the same exercise bout for the 24 and 48 h after HIIE ITT; however, they had ad libitum access to chow for 20 and 44 h after a bout of HIIE, before beginning the 4-h fasting period, respectively. The mice rested a minimum of 1 week from exercise between each ITT.

The fasting period began at ZT2. After a 6-h fast to empty the stomach, 10- to 14-week-old male mice received injections of 2 mg/kg pyruvate i.p. Blood glucose was measured from tail blood as previously described (23).

Brain slices were prepared as previously described (16). Briefly, male mice were deeply anesthetized with intraperitoneal injection of 7% chloral hydrate and transcardially perfused with a modified ice-cold artificial cerebrospinal fluid (ACSF) (described below). Mice were then decapitated, and the brain was removed and submerged in ice-cold, carbogen-saturated (95% O₂ and 5% CO₂) ACSF (126 mmol/L NaCl, 2.8 mmol/L KCl, 1.2 mmol/L MgCl₂, 2.5 mmol/L CaCl₂, 1.25 mmol/L NaH₂PO₄, 26 mmol/L NaHCO₃, and 5 mmol/L glucose).

Coronal sections (250 µm) were cut with a Leica VT1000 S vibratome following incubation in oxygenated ACSF at room temperature for at least 1 h before recording. The slices were bathed in oxygenated ACSF (32–34°C) at a flow rate of ∼2 mL/min.

The pipette solution for whole-cell recording was modified to include an intracellular dye (Alexa Fluor 350 hydrazide dye) for whole-cell recording: 120 mmol/L k-gluconate, 10 mmol/L KCl, 10 mmol/L HEPES, 5 mmol/L EGTA, 1 mmol/L CaCl₂, 1 mmol/L MgCl₂, and 2 mmol/L Mg ATP [0.03 mmol/L Alexa Fluor 350 hydrazide dye, pH 7.3]). Epifluorescence was briefly used to target fluorescent cells, at which time the light source was switched to infrared differential interference contrast imaging for obtaining the whole-cell recording (Zeiss Axioskop 2 FS Plus equipped with a fixed stage and a QuantEM:512SC electron-multiplying charge-coupled device camera). Electrophysiological signals were recorded with an Axopatch 700B amplifier (Molecular Devices), low-pass filtered at 2–5 kHz, and analyzed offline on a PC with pCLAMP programs (Molecular Devices). Membrane potential and firing rate were measured with whole-cell current clamp recordings from POMC in brain slices. Recording electrodes had resistances of 2.5–5.0 MΩ when filled with the k-gluconate internal solution.

After tamoxifen induction, we validated NR1 deletion within adult ARC POMC neurons using an approach previously described (20,24). Briefly, electrical stimulation in acute brain slices confirmed loss of NMDA-induced currents. Evoked excitatory postsynaptic currents (EPSCs) were recorded in the presence of picrotoxin (to block GABA receptor–mediated inhibitory postsynaptic currents), and the NMDAR component was subsequently isolated with use of cyanquixaline to block AMPA receptors (AMPARs) (20). NR1 subunit deletion in POMC neurons caused loss of evoked NMDAR-mediated EPSCs (Supplementary Fig. 1). As anticipated from previous work (20,24), NMDA-evoked currents were absent in neurons lacking NR1 subunits.

Results are reported as mean ± SEM unless otherwise indicated. Significance was set at P < 0.05 for all statistical measures. All data were evaluated with a two-tailed Student t test, an unpaired t test, or ANOVA where appropriate. All graphs were carried out with GraphPad Prism 10 software or CorelDRAW X8 (64 bit).

Our initial focus was on NR1 subunits within adult POMC neurons and their role in regulating body weight. Using a tamoxifen-inducible POMC-CreER^T2::NR1^loxp/loxp mouse model, we deleted NR1 in POMC neurons of adult mice to assess its impact on body weight. There was no significant difference in the body weight of POMC-CreER^T2::NR1^loxp/loxp mice fed an ad libitum chow diet in comparisons with littermate controls ([t(16)] = P > 0.05) (Fig. 1A). Similarly, POMC-CreER^T2::NR1^loxp/loxp mice did not show a difference in body weight on HFD ([t(16)] = P > 0.05) (Fig. 1D). Body composition was also measured for both dietary groups, and no significant difference was found in either lean or fat mass after 24 weeks of age in comparisons with littermate controls (Fig. 1B, C, E, and F). These findings suggest that NMDARs in ARC POMC neurons do not play a critical role in energy balance, aligning with prior observations of prenatal NR1 deficiency (20).

Next, we investigated the role of NR1 in the regulation of glucose metabolism by POMC neurons independent of changes in energy balance. GTT, ITT, and PTT were conducted on male mice fed standard chow or HFD. NR1 deletion in adult POMC neurons significantly impaired glucose tolerance and pyruvate tolerance and increased insulin resistance (Fig. 2). This was evident in both dietary groups. Chow-fed mice displayed a marked difference in glucose tolerance [GTT, chow area under the curve (AUC): t(11) = 16.29, P < 0.05 (Fig. 2A)], while the HFD-fed mice exhibited a trend toward impaired glucose handling [GTT, HFD AUC: t(11) = 1.97, P = 0.07] (Fig. 2D). ITTs showed a significant increase in insulin resistance in both chow-fed mice [ITT, chow AUC: t(11) = 14.84, P < 0.05 (Fig. 2B)] and HFD-fed mice [ITT, HFD AUC: t(9) = 2.367, P < 0.05 (Fig. 2E)]. Moreover, when pyruvate was provided as a fuel source (PTT), both chow-fed mice [PTT, chow AUC: t(11) = 10.27, P < 0.05 (Fig. 2C)], and mice fed HFD [PTT, HFD AUC: t(9) = 3.163, P < 0.05 (Fig. 2F)] displayed increased serum glucose levels, indicating increased glucose production. These data highlight a specific role for NR1 subunits in POMC neurons to regulate glucose metabolism independent of body weight.

We further probed the role of NR1 subunits in POMC neurons to modulate insulin sensitivity postexercise. Mice fed a standard chow diet underwent a single bout of HIIE followed by an ITT (Fig. 3A). Exercise significantly enhanced insulin sensitivity compared with a sedentary state (Fig. 3B and C), evidenced by a greater glucose disposal rate during the ITT, which resulted in reduced AUC for blood glucose (Fig. 3F). To better evaluate insulin sensitivity, we focused on the early response to insulin administration. The 15-min interval was chosen to minimize the influence of counterregulatory responses triggered by hypoglycemia (25,26). Here, mice that underwent HIIE demonstrated a pronounced insulin-mediated reduction in blood glucose levels compared with a sedentary state (Fig. 3E). Notably, deletion of NR1 in adult POMC-expressing cells abrogated the HIIE-induced enhancement in insulin sensitivity observed postexercise (Fig. 3E and F). This suggests a crucial role for NR1 in the exercise-mediated improvement of insulin responsiveness. Notably, the same mice underwent an ITT following HIIE at 14 months of age. The aged wild-type (WT) mice exhibited improved insulin sensitivity; however, this response remained blunted in mice deficient for NR1 in POMC neurons (Fig. 3D, G, and H), suggesting that this mechanism functions independent of age.

Considering the sustained impact of exercise on insulin sensitivity, which can last 2–3 days postexercise (27,28), we further assessed how the absence of NR1 in POMC neurons affects this prolonged response. At both 24 and 48 h postexercise, mice lacking NR1 in POMC neurons showed a significant reduction in insulin sensitivity in comparisons with control littermates. This reduction in sensitivity was transient, with a notable reversal observed 48 h after exercise, supporting a temporal modulation of insulin sensitivity by NR1 in response to exercise [ITT AUC, WT 3 h after HIIE t(36) = 7.61, P < 0.05, WT 24 h after HIIE t(36) = 3.60, P < 0.05, and WT 48 h after HIIE t(36) = 0.50, P > 0.05; ITT, knockout (KO) 3 h after HIIE t(36) = 3.89, P < 0.05, KO 24 h after HIIE t(36) = 3.17, P < 0.05, and KO 48 h after HIIE t(36) = 2.49, P > 0.05 (Fig. 3F); ITT AUC, WT sedentary vs. WT 3 h after HIIE t(14) = 3.63, P < 0.05, WT 3 h after HIIE vs. KO sedentary KO 3 h after HIIE t(14) = 1.27, P < 0.05, WT 3 h after HIIE vs. KO 3 h after HIIE t(14) = 2.30, P > 0.05, WT sedentary vs. KO 3 h after HIIE t(14) = 1.21, P > 0.05, WT sedentary vs. KO sedentary after HIIE t(14) = 0.33, P > 0.05, and KO sedentary vs. KO 3 h after HIIE t(14) = 0.90, P > 0.05 (Fig. 3H)].

To determine whether loss of NR1 in POMC neurons abrogates exercise-induced activation of ARC POMC neurons, we performed whole-cell patch clamp electrophysiological recordings on adult ARC POMC neurons from POMC-CreER^T2::NR1^loxp/loxp::td-tomato mice and control mice (POMC-hrGFP [POMC–humanized renilla green fluorescent protein]). Whole-cell patch clamp recordings were made in 68 POMC neurons throughout the rostral-caudal axis of the ARC from seven POMC-hrGFP mice that were subjected to either HIIE (n = 4 mice) or a sedentary state (n = 3 mice) (Fig. 4). For WT sedentary mice there was an average resting membrane potential of ARC POMC neurons −45 ± 1.0 mV (refer to Fig. 4 for data presentation) and overshooting action potentials (n = 32). For exercised mice there was an average resting membrane potential of ARC POMC neurons −42 ± 0.7 mV and overshooting action potentials (n = 36) [WT HIIE t(66) = 2.63, P < 0.05 (Fig. 4F)]. Similar to findings of previous reports (16), the exercise-induced depolarization was associated with an increase in spontaneous EPSCs (sEPSCs) [WT HIIE t(63) = 2.40, P < 0.05 (Fig. 4H)] but no significant difference in sEPSC amplitude or action potential frequency.

For evaluation of the role of NR1 in the activation of ARC POMC neurons after an acute bout of exercise, whole-cell patch clamp recordings were made in 37 POMC neurons throughout the rostral-caudal axis of the ARC nucleus from six POMC-CreER^T2::NR1^loxp/loxp::td-tomato mice that were subjected to either HIIE (n = 3 mice) or a sedentary state (n = 3 mice) (Fig. 5). For sedentary mice lacking NR1 there was average resting membrane potential of ARC POMC neurons −42 ± 1.3 mV (refer to Fig. 5 for data presentation) and overshooting action potentials (n = 17). For exercised mice without NR1 there was average resting membrane potential of ARC POMC neurons −42 ± 1.0 mV and overshooting action potentials (n = 20). In comparisons with values of sedentary POMC-CreER^T2::NR1^loxp/loxp mice, POMC neurons from exercised mice exhibited no significant differences in resting membrane potential. Similarly, sEPSC frequency and amplitude as well as action potential frequency were not significantly different between sedentary and exercised POMC CreER^T2::NR1^loxp/loxp::td-tomato mice. Together these data suggest that NR1 subunits are required for the activation of ARC POMC neurons after an acute bout of exercise.

Selective deletion of NR1 subunits in POMC neurons negatively impacts glucose and insulin sensitivity, independent of body weight and age. Moreover, the temporal exercise-induced improvements in insulin sensitivity were abrogated in mice lacking NR1 subunits in POMC neurons. This observation aligns with the blunted activation of POMC neurons following exercise in mice deficient for NR1 subunits in POMC neurons. These findings collectively support that increased POMC neuronal activity following exercise plays a pivotal role in enhancing peripheral insulin sensitivity.

Increased melanocortin signaling consistently results in improved energy balance and glucose homeostasis (29,30). Acute bouts of exercise can result in significant improvements in peripheral insulin sensitivity and hepatic glucose production (3,4). In agreement with these previous reports, we found that exercise increased insulin sensitivity in mice for up to 2 days after a single bout of exercise, an effect that appears to be independent of age (Fig. 3). Notably, mice deficient for NR1 subunits in POMC neurons also exhibited an abrogated exercise-induced insulin sensitivity, both in absolute magnitude and temporal duration. While it is surprising that NMDARs deficiency in POMC neurons disrupts ARC POMC neuron activation and postexercise glucose metabolism improvements, an analogous observation has been made in ARC NPY/AgRP neurons following food deprivation. Deficiency of NMDARs in ARC NPY/AgRP neurons was sufficient to abrogate fasting-induced activation of ARC NPY/AgRP neurons and resulting hyperphagia (20). Additional intracellular signals and channels (e.g., small-conductance calcium-activated potassium channels) have been suggested in the fasting-induced activation of ARC NPY/AgRP neurons and might be downstream of calcium signaling via NMDARs (31). Determination of whether these mechanisms may be conserved in ARC POMC neurons after exercise remains an interesting area of future investigation.

ARC POMC neurons are highly heterogeneous, exhibiting diverse physiological and functional characteristics (11,32,33). We previously demonstrated that exercise-induced activation of ARC POMC neurons is particularly robust in those expressing leptin receptors (LepRs), suggesting that leptin, an adipose tissue–derived hormone, may contribute to these changes in POMC cell activity (16). However, a direct link between leptin action and the modulation of POMC neuron function in response to exercise remains to be elucidated. Importantly, POMC neurons expressing LepRs encompass various subpopulations of POMC neurons (33,34). Thus, the ARC LepR POMC population may be selecting for POMC neurons that are sensitive to the effects of exercise independent of any exercise-dependent effects via LepR signaling.

Another intriguing subpopulation of POMC neurons is those expressing glucagon-like peptide 1 receptors (GLP-1Rs). GLP-1R–expressing POMC neurons have been shown to express LepRs (35,36); however, there may be a principal GLP-1R–expressing POMC cell population that does not express LepR (36). We previously showed that GLP-1R signaling in ARC POMC neurons is required for the effects of long-acting GLP-1R agonists (GLP-1Rags) on energy balance and blood glucose levels (37). Interestingly, combination of GLP-1Rags with exercise results in enhanced activation of ARC POMC neurons in comparison with either alone (37). In addition to peripherally administered GLP-1Rags, increasing the activity of GLP-1 neurons in the brain results in improved blood glucose control (38–40). Moreover, these GLP-1 neurons in the hindbrain are glutamatergic and project to the ARC nucleus where POMC neurons that express GLP-1Rs reside (40). This is potentially clinically relevant as reports suggest additive benefits of energy metabolism, glucose balance, and insulin sensitivity when exercise and GLP-1R agonism are combined (41,42). The need for better understanding of the interaction between exercise and GLP-1R signaling on the plasticity of metabolically relevant neuronal circuits warrants further investigation.

Our investigation reveals a novel aspect of neural adaptation to exercise, demonstrating significant excitement of ARC POMC neurons through elevation of EPSC frequency. Traditionally, this increase in EPSC frequency is linked to enhanced presynaptic release of glutamate (43,44). However, recruitment of AMPARs topographically to the postsynaptic domain alone may increase the frequency of excitatory events independent of synaptic strength (45,46). NMDARs in POMC neurons may initiate intracellular events, potentially recruiting AMPARs in a postsynaptic manner. Alternatively, exercise could enhance presynaptic glutamatergic neuron activity on ARC POMC neurons. Understanding the sources of excitatory inputs to ARC POMC neurons is crucial for unraveling the neural circuits involved in the modulation of glucose metabolism by exercise. Future studies with use of circuit tracing techniques, optogenetic manipulation, and pharmacological interventions are poised to determine the presynaptic origins of the excitatory inputs and their functional significance in the context of blood glucose control.

While the current study connects NR1 subunits in POMC neurons to exercise and metabolism, these findings should be interpreted with certain considerations in mind. First, POMC neurons are expressed not only in the ARC but also in the hindbrain’s nucleus tractus solitarius (NTS) (47,48). Previous work suggests that activation of NTS POMC neurons acutely regulates energy balance, while activation of ARC POMC neurons regulates chronic changes in energy balance (30,49). This highlights a complex and region-specific role of POMC neurons in energy homeostasis. However, it is crucial to consider cellular heterogeneity within these populations, as subpopulations of ARC POMC neurons may also contribute to acute energy balance regulation (36). The exact mechanisms by which these neuron subpopulations interact and respond to exercise-induced stimuli remain an area for future research. Additionally, the specific contribution of NTS POMC neurons to exercise-induced metabolic improvements remains unclear; more targeted studies are required to elucidate their role.

Building off previous work (16), the current study focused on the metabolic outcomes following HIIE. The metabolic impact of various physical activities, each with unique intensities and durations (22), presents fertile ground for expanding our understanding of how various forms of exercise influence POMC neuron-mediated metabolic regulation. Exploring a wider range of exercise modalities, including endurance training, resistance training, and moderate-intensity continuous exercise, could provide a more comprehensive understanding of the relationship between physical activity and metabolic health.

Finally, with the focus of the current study on lean male mice, the influence of sex and physiological status on exercise-induced metabolic and cellular responses should be considered. Investigating the effects of exercise in a diverse set of animal models, including females and those with different metabolic statuses, particularly in the context of obesity or metabolic disorders (50), would provide a more complete understanding of the broader applicability of the findings. Thus, while our study provides valuable insights into the role of NR1 subunits in POMC neurons in mediating exercise-induced metabolic benefits, it also underscores the need for further research to explore these additional biological implications.

In summary, deletion of NR1 subunits in POMC neurons impaired glucose metabolism and attenuated the increase in insulin sensitivity that occurred after an acute bout of exercise. Additionally, electrophysiological recordings suggest that exercise-induced plasticity in POMC neurons was abrogated by selective deletion of NR1 subunits. However, specific deletion of NR1 subunits in POMC neurons did not affect body weight, regardless of diet. Together, these data highlight the role of NMDARs in POMC neurons as essential for glucose metabolism and exercise-induced improvements in insulin sensitivity.

Acknowledgments. During the preparation of this work the authors used ChatGPT for the purpose of editing. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

Funding. This work was supported by grants to E.H. (National Research Foundation of Korea [NRF 2021R1A6A3A14044733]) and K.W.W. (National Institute of Diabetes and Digestive and Kidney Diseases [R01 DK119169 and DK119130-5830]).

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

Author Contributions. B.P. designed and performed all experiments, analyzed data, and wrote the manuscript. E.H. performed electrophysiological experiments, analyzed data, and reviewed the manuscript. J.A. and K.G. performed electrophysiological experiments and analyzed data. L.L., B.W., and A.K. assisted in performing experiments. J.K.E. supervised development of the mouse models and reviewed the manuscript. K.W.W. conceived and designed the study, supervised development of the mouse models, designed experiments, and edited the manuscript. K.W.W. 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 in abstract form at the 84th Scientific Sessions of the American Diabetes Association, Orlando, FL, 21–24 June 2024. A video presentation can be found in the online version of the article at https://doi.org/10.2337/dbi24-0002.