Insulin Preferentially Regulates the Activity of Parasympathetic Preganglionic Neurons over Sympathetic Preganglionic Neurons

Uisu Hyun; Yoon Young Kweon; Jong-Woo Sohn

doi:10.3803/EnM.2023.1725

Articles

Page Path: HOME > Endocrinol Metab > Volume 38(5); 2023 > Article

Original Article Diabetes, obesity and metabolism Insulin Preferentially Regulates the Activity of Parasympathetic Preganglionic Neurons over Sympathetic Preganglionic Neurons Keypoint The study explored insulin's effects on neurons in the brainstem and spinal cord using mouse models. Results confirmed that insulin inhibits parasympathetic cholinergic neurons in the dorsal motor nucleus of the vagus nerve and GABAergic neurons in the brainstem. However, it did not affect sympathetic neurons in the spinal cord. The findings suggest that insulin's inhibition of parasympathetic neurons may not play a critical role in regulating glucose homeostasis and gastrointestinal motility.: Uisu Hyun, Yoon Young Kweon, Jong-Woo Sohn; Endocrinology and Metabolism 2023;38(5):545-556.
DOI: https://doi.org/10.3803/EnM.2023.1725
Published online: September 26, 2023

Department of Biological Sciences, Korea Advanced Institute of Science and Technology, Daejeon, Korea

Corresponding author: Jong-Woo Sohn Department of Biological Sciences, Korea Advanced Institute of Science and Technology, 291 Daehak-ro, Yuseong-gu, Daejeon 34141, Korea Tel.: +82-42-350-2631, Fax: +82-42-350-2610, E-mail: jwsohn@kaist.ac.kr

• Received: May 2, 2023 • Revised: July 4, 2023 • Accepted: August 1, 2023

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

1,382 Views
78 Download

Full Article

Download PDF

ABSTRACT
GRAPHICAL ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
Supplementary Material
Article information
References

ABSTRACT

Background
Insulin is a peptide hormone that regulates post-prandial physiology, and it is well known that insulin controls homeostasis at least in part via the central nervous system. In particular, insulin alters the activity of neurons within the autonomic nervous system. However, currently available data are mostly from unidentified brainstem neurons of the dorsal motor nucleus of the vagus nerve (DMV).
Methods
In this study, we used several genetically engineered mouse models to label distinct populations of neurons within the brainstem and the spinal cord for whole-cell patch clamp recordings and to assess several in vivo metabolic functions.
Results
We first confirmed that insulin directly inhibited cholinergic (parasympathetic preganglionic) neurons in the DMV. We also found inhibitory effects of insulin on both the excitatory and inhibitory postsynaptic currents recorded in DMV cholinergic neurons. In addition, GABAergic neurons of the DMV and nucleus tractus solitarius were inhibited by insulin. However, insulin had no effects on the cholinergic sympathetic preganglionic neurons of the spinal cord. Finally, we obtained results suggesting that the insulin-induced inhibition of parasympathetic preganglionic neurons may not play a critical role in the regulation of glucose homeostasis and gastrointestinal motility.
Conclusion
Our results demonstrate that insulin inhibits parasympathetic neuronal circuitry in the brainstem, while not affecting sympathetic neuronal activity in the spinal cord.
Keywords: Autonomic nervous system; Medulla oblongata; Spinal cord; Cholinergic neurons; Electrophysiology; Autonomic function

GRAPHICAL ABSTRACT

INTRODUCTION

It is well known that insulin, a peptide hormone secreted from the pancreatic β-cells, regulates blood glucose levels and energy balance [1]. Multiple organs express insulin receptors (InsRs), which mediate the metabolic effects of insulin. For instance, InsRs expressed by the liver and the skeletal muscle are responsible for the glucose-lowering effects of insulin [2]. Conversely, neuron-specific deletions of InsRs led to body weight gain [3]. A more recent study demonstrated that InsRs expressed by steroidogenic factor-1 (SF-1) neurons in the ventromedial nucleus of the hypothalamus (VMH) mediated diet-induced obesity as well as diet-induced insulin resistance [4]. In addition, InsRs expressed by pro-opiomelanocortin (POMC) neurons and agouti-related peptide (AgRP) neurons within the arcuate nucleus of the hypothalamus (ARH) were shown to be required for the maintenance of normal glucose balance [5,6]. Insulin was also demonstrated to lower glucose production and decrease food intake through the dorsal vagal complex (DVC) of the brainstem [7,8]. Therefore, the available evidence suggests that insulin stimulates InsRs within the central nervous system, as well as peripheral InsRs, to exert metabolic effects.

Previous studies have suggested that the acute effects of insulin on specific neuronal populations may explain the in vivo metabolic effects. For example, insulin-induced inhibition of ARH POMC and AgRP neurons was suggested to be responsible for the maintenance of glycemia by InsRs expressed by these neurons [5]. In addition, insulin-induced inhibition of VMH SF-1 neurons was suggested to underlie the promotion of diet-induced obesity and insulin resistance by the InsRs expressed by these neurons [4]. However, it is less clear how insulin works in the brainstem to control blood glucose levels and food intake. While previous studies have reported that insulin inhibited neurons within the dorsal motor nucleus of the vagus nerve (DMV) [9], the recordings were obtained from unidentified neurons of the DMV. Since DMV contains other types of neurons, as well as parasympathetic preganglionic neurons [10-12], it is necessary to specifically label parasympathetic preganglionic neurons for recordings. Moreover, no data are currently available regarding the effects of insulin on the sympathetic preganglionic neurons of the spinal cord.

In this study, we took advantage of genetically engineered mouse models to label parasympathetic and sympathetic preganglionic neurons and performed whole-cell patch clamp experiments to address the acute effects of insulin on the brainstem and spinal cord autonomic circuitry. We also assessed the in vivo metabolic functions of InsRs expressed by the autonomic preganglionic neurons.

Animals: All experimental procedures involving the use of animals were approved by the Institutional Animal Care and Use Committee of Korea Advanced Institute of Science and Technology (KAIST; Protocol No. KA2021-126). Animals were housed in a standard 12-hour light-dark cycle (light on at 7:00 AM and light off at 7:00 PM) with ad libitum access to water and standard mouse chow (Envigo 2018C Teklad Global 18% protein rodent diet, Inotiv, West Lafayette, IN, USA). All experiments were conducted using 5- to 16-week-old male mice. The mice used in this study were maintained in a specific pathogen-free facility of KAIST Laboratory Animal Resource Center. ChAT-IRES-Cre mice (#006410, Jackson Laboratory, Bar Harbor, ME, USA), pairedlike homeobox 2b (Phox2b)-Cre mice (#016223, Jackson Laboratory), or vesicular GABA transporter (Vgat)-IRES-Cre (#016962, Jackson Laboratory) were crossed with tdTomato reporter mice (#007908, Jackson Laboratory) to label selective neuronal populations. All Cre and tdTomato mouse lines were used only in the heterozygous state. Insulin receptor flox (InsR^f/f) mice (#006955, Jackson Laboratory) were used for some experiments.
Electrophysiology: Whole-cell patch clamp recordings from tdTomato-expressing neurons were maintained in the acute brainstem or thoracic spinal cord slice preparations, as previously described [13,14]. Briefly, 6- to 15-week-old male mice were deeply anesthetized by isoflurane inhalation and transcardially perfused with an ice-cold modified artificial cerebrospinal fluid (ACSF) (described below), in which an equimolar concentration of sucrose was substituted for NaCl. Then, the brainstem or the thoracic spinal cord was rapidly prepared and submerged in ice-cold modified ACSF. Acute coronal sections (250 μm) of the brainstem or the spinal cord were obtained using a vibratome (VT1200S, Leica, Wetzlar, Germany). The sections were then allowed to recover at 34°C for 1 hour in ACSF (126 mM NaCl, 2.8 mM KCl, 1.2 mM MgCl₂, 2.5 mM CaCl₂, 1.25 mM NaH₂PO₄, 26 mM NaHCO₃, and 5 mM glucose).; Afterward, slices were transferred to a recording chamber with a continuous flow (approximately 2 mL/min) of oxygenated ACSF (32°C to 34°C). Epifluorescence was briefly used to target fluorescent cells; at which time the light source was switched to infrared differential interference contrast imaging to obtain whole-cell patch clamp recordings (Olympus, Tokyo, Japan; BX51 WI equipped with a fixed stage and an optiMOS scientific CMOS camera). Electrophysiological signals were recorded using an Axopatch 700B amplifier (Molecular Devices, San Jose, CA, USA), low pass filtered at 2 to 5 kHz, and analyzed offline with pCLAMP programs (Clampfit 10.4, Molecular Devices) and MiniAnalysis software (Synaptosoft/Bluecell, Anyang, Korea).; The recording electrodes had a resistance of 2.5 to 5 MΩ when filled with K⁺- or Cs⁺-based pipette solutions (described below). For current-clamp recordings, the pipette solution contained 120 mM K-gluconate, 10 mM KCl, 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 1 mM CaCl₂, 1 mM MgCl₂, 5 mM ethylene glycol-bis(β-aminoethyl ether)-N,N,N´, N´-tetraacetic acid (EGTA), 2 mM Mg-ATP, and 0.03 mM Alexa fluor 488 dye. The baseline membrane potential was determined as a stable voltage maintained for at least 3 minutes. We assessed the input resistance by measuring voltage deflection at the end of responses to small hyperpolarizing rectangular current pulse steps (500 ms of 0 to –50 pA). For voltage-clamp experiments, K⁺ was replaced by equimolar concentrations of Cs⁺. We recorded excitatory postsynaptic current (EPSC) and inhibitory postsynaptic current (IPSC) at holding potentials of –60 and –10 mV, respectively. We analyzed current traces only when the access resistance was maintained below 30 MΩ throughout the recording. An acute effect was required to be associated temporally with drug application, and the response had to be stable within a few minutes. A neuron was considered hyperpolarized only if a change in membrane potential was at least 2 mV in amplitude. Membrane potential values were not adjusted to account for junction potential (–8 mV).; Insulin (insulin aspart, NovoRapid, Novo Nordisk, Bagsværd, Denmark) was applied at least for 5 minutes to bath solutions at a concentration of 50 nM. We used 500 nM tetrodotoxin (TTX; #T-550, Alomone lab, Jerusalem, Israel) to block voltage-gated sodium channels. Furthermore, 1 mM kynurenic acid (#K3375, Sigma-Aldrich, St. Louis, MO, USA) and 50 μM picrotoxin (#P1675, Sigma-Aldrich) were used to block ionotropic glutamate receptors and GABAA receptors, respectively. Stock solutions were prepared according to the manufacturer’s specifications.
Glucose tolerance test and insulin tolerance test: Five- to 7-week-old male mice were used for the glucose tolerance test (GTT) and insulin tolerance test. For intraperitoneal glucose tolerance test (IPGTT) and oral glucose tolerance test (OGTT), single-housed InsR^f/f mice and choline acetyltransferase (ChAT)^cre/+::InsR^f/f mice were provided with clean bedding and were fasted for 18 hours. Glucose (#G8270, Sigma-Aldrich) was dissolved in saline and then filtered to make glucose solutions. Glucose (1.25 g/kg) was administered at 10:00 AM (Zeitgeber time [ZT]=3), which was considered time 0. Blood glucose levels were measured from tail veins using a glucometer (AGM-4100, Allmedicus, Anyang, Korea) at 0 minute, 15 minutes, 30 minutes, 1 hour, 2 hours, and 4 hours after glucose injections. For intraperitoneal insulin tolerance test (IPITT), ad libitum fed mice were injected intraperitoneal with insulin (1 U/kg, insulin aspart, NovoRapid) at 10:00 AM (ZT=3), which was considered time 0. Blood glucose levels were measured at 0 minute, 15 minutes, 30 minutes, 1 hour, 2 hours, and 4 hours after insulin injections.
Cannulation: Seven- to 11-week-old male InsR^f/f mice and ChAT^cre/+::InsR^f/f mice were anesthetized and had their heads fixed in a stereotaxic frame (Stoelting, Wood Dale, IL, USA). Lidocaine (0.5 wt/vol) was applied for analgesia, and an incision was made with a sterile scalpel. The skull was drilled and a guide cannula (C315G/SPC, 26 gauge, Plastics One, Boerne, TX, USA) was placed in the lateral ventricle (anterior-posterior –0.34 mm; medial-lateral ±1.00 mm; dorsal-ventral –2.20 mm from the bregma). The guide cannula was held in place with acrylic dental cement and three anchoring skull screws. After the cement hardened, a dummy cannula (C315DC/SPC, Plastics One) was inserted at the outer tip of the guide cannula. The mice were then allowed to recover for 3 to 7 days. After recovery, the dummy cannula was removed and the internal cannula (C315I/SPC, 33 gauge, Plastics One) was inserted. The internal cannula was connected to a Hamilton syringe via a polyethylene tube. Angiotensin II (ATII; 50 μg/mL, 2 μL, #A9525, Sigma-Aldrich) was injected via the internal cannula, which caused drinking within 10 minutes, to confirm on-target placement of cannula. After the ATII test, the internal cannula was removed and the dummy cannula was inserted. Mice were handled for 5 to 7 days and were allowed to habituate to single-cage housing at least 1 week before the day of experiment.
Assessment of gastrointestinal motility: Seven- to 11-week-old male wild-type mice were fasted for 18 hours before oral gavage with charcoal solutions (100 μL) at 10:00 AM (ZT=3). Charcoal solutions containing 10% charcoal (#C9157, Sigma-Aldrich) and 5% arabic gum (#G9752, Sigma-Aldrich) were dissolved in saline. For intraperitoneal injection studies, Dulbecco’s phosphate-buffered saline (DPBS) or insulin (0.5 U/kg) was injected intraperitoneal 30 minutes before oral gavage. For intracerebroventricular injection studies, 2 μL of DPBS or insulin (4 mU) was injected intracerebroventricular 30 minutes before oral gavage using a Hamilton syringe and the internal cannula at a speed of 5 μL/min. The internal cannula was removed 10 seconds after the drug injection was completed. Mice were sacrificed 50 minutes after oral gavage to collect the gastrointestinal (GI) tract from the stomach to small intestine. We measured the length between the distal end of the pylorus and the most distal end of charcoal in the GI tract, after which we normalized the measurements to the length between the distal end of the pylorus and the distal end of the ileum to determine GI motility.; Gastric emptying was measured using 7- to 11-week-old male InsR^f/f mice and ChAT^cre/+::InsR^f/f mice with the acetaminophen assay, as described previously [15]. Food was removed at the starting time, and acetaminophen (90 mg/kg, #BP371, Sigma-Aldrich) was given immediately by oral gavage. Acetaminophen was given at 10:00 PM (ZT=15) in the dark cycle and at 10:00 AM (ZT=3) in the light cycle. The mice were fed ad libitum for the dark cycle assay, and were fasted for 18 hours for the light cycle assay. Blood was collected from tail veins 15 minutes and 1 hour after oral gavage to analyze the concentration of acetaminophen using the paracetamol (Acetaminophen) 3 Reagent Assay Kit (#K8002, Cambridge Life Sciences, Ely, UK) according to the manufacturer’s protocol.
Data analysis: All statistical analyses were performed using GraphPad Prism version 7 (GraphPad Software, San Diego, CA, USA). Statistical data are expressed as mean±standard error of the mean, where n represents the number of cells. The significance of differences between groups was evaluated using Fisher’s exact test, the Wilcoxon matched-pairs signed rank test, Mann-Whitney test, two-way analysis of variance with the Bonferroni multiple comparisons test, with a confidence level of P<0.05, P<0.01, or P<0.001.

Insulin directly inhibits parasympathetic preganglionic neurons: The parasympathetic preganglionic neurons within the DMV and the sympathetic preganglionic neurons within the intermediolateral column (IML) are cholinergic neurons that express ChAT. Therefore, we used ChAT-IRES-Cre mice [16] crossed with tdTomato reporter mice [17] (ChAT^cre/+::tdTomato mice) to label parasympathetic and sympathetic preganglionic neurons.; We first targeted tdTomato-expressing neurons within the DMV for whole-cell patch clamp recordings and tested the acute effects of insulin (Fig. 1A, B). We found that insulin (50 nM) hyperpolarized nine of 21 (43%) DMV cholinergic neurons by –12.7±2.0 mV (n=9) in current-clamp modes (Fig. 1C, I, J). We applied small hyperpolarizing current steps before and during insulin treatments (arrows in Fig. 1C) to calculate input resistance (Fig. 1D). We found that insulin decreased input resistance by 35.7% (from 0.91±0.07 to 0.59±0.07 GΩ, n=9) with a reversal potential of –82.2±3.4 mV (n=9), which suggested the opening of a putative potassium conductance (Fig. 1E). In the course of patch clamp recordings, we took pictures of cells at a low magnification to examine the location in the rostrocaudal and mediolateral extent of the DMV with respect to their responses to insulin (Fig. 1F). We noted that the hyperpolarizing responses were located mostly in caudal sections of the DMV. We repeated the same series of experiments in the presence of TTX (500 nM) and a cocktail of synaptic blockers (SBs, 1 mM kynurenic acid and 50 μM picrotoxin), and observed that insulin still hyperpolarized six of 11 (55%) DMV cholinergic neurons by –11.3±2.1 mV (Fig. 1G, I, J). Consistent with results in Fig. 1F, most hyperpolarizing responses were observed in the caudal DMV (Supplemental Fig. S1, upper panels). We also confirmed insulin-induced decreases in the input resistance by 55.2% (from 0.36±0.08 to 0.16±0.03 GΩ, n=6) with a reversal potential of –87.1±8.0 mV (n=6) in the presence of TTX and SBs. Subsequently, we deleted InsRs selectively in the cholinergic neurons by crossing the ChAT^cre/+::tdTomato mice with the InsR^f/f mice [18] (ChAT^cre/+::InsR^f/f::tdTomato mice). The hyperpolarizing effect of insulin on DMV cholinergic neurons was no longer observed (0.1±0.3 mV, n=13) in the brainstem slices from the ChAT^cre/+::InsR^f/f::tdTomato mice (Fig. 1H-J, Supplemental Fig. S1, lower panels). Together, these results suggest that insulin stimulates InsRs to directly hyperpolarize the membrane potential of DMV neurons independently of synaptic activity.
Insulin suppresses synaptic input into parasympathetic preganglionic neurons: We also tested the acute effects of insulin on synaptic currents. We recorded EPSCs and IPSCs on DMV cholinergic neurons in voltage-clamp modes. We observed a significant reduction of spontaneous EPSCs (sEPSCs) in both frequency (by 27.4%, from 10.7±1.6 to 7.3±1.2 Hz, P=0.0002) and mean amplitudes (by 18.8%, from 19.6±1.8 to 15.3±1.1 pA, P<0.0001) by insulin treatment (Fig. 2A-C, Supplemental Fig. S2). Interestingly, insulin also significantly decreased spontaneous IPSCs (sIPSCs) in both frequency (by 17.2%, from 5.4±1.4 to 3.7±1.0 Hz, P=0.0129) and mean amplitudes (by 10.6%, from 34.0±2.6 to 29.4±2.2 pA, P=0.0110) (Fig. 2D-F, Supplemental Fig. S2). These results indicate that insulin downregulates glutamatergic and GABAergic neurotransmission onto DMV cholinergic neurons.; We also recorded miniature EPSCs and IPSCs (mEPSCs and mIPSCs) in the presence of 500 nM TTX and tested the acute effects of insulin. Similar to the results from sEPSC recordings, insulin treatment significantly reduced the frequency (by 25.9%, from 5.3±0.7 to 3.5±0.6 Hz, P=0.0103) and mean amplitude (by 17.4%, from 15.7±0.9 to 12.9±0.7 pA, P=0.0001) of mEPSCs (Fig. 2G-I, Supplemental Fig. S2). However, insulin treatment did not affect the mean amplitude of mIPSCs (18.4±1.2 pA in TTX only vs. 18.2±1.1 pA in TTX⁺insulin, P>0.05) while the mIPSC frequency was significantly reduced by 37.8% (from 2.5±0.7 to 1.2±0.4 Hz, P=0.0015) (Fig. 2J-L, Supplemental Fig. S2). Therefore, we predict that insulin not only inhibits the activity of parasympathetic preganglionic neurons, but also suppresses local neurocircuitry within the DVC whether it is excitatory or inhibitory on the parasympathetic preganglionic neurons.
Insulin inhibits GABAergic neurons in the DVC: Given the suppression of sEPSCs by insulin, we hypothesized that insulin inhibits glutamatergic neurons projecting to DMV cholinergic neurons. The nucleus tractus solitarius (NTS), which is located dorsal to the DMV, contains Phox2b-positive (or Phox2b) neurons that innervate the DMV [16]. In addition, Phox2b neurons in the NTS are exclusively glutamatergic [19], and DMV neurons receive glutamatergic inputs from the neurons of the NTS [20]. Therefore, we assumed that Phox2b neurons may also send insulin-sensitive glutamatergic input to DMV cholinergic neurons. We labeled Phox2b neurons in the NTS using the Phox2b-Cre mice [21] bred with tdTomato reporter mice (Phox2b^cre/+::tdTomato mice) to test the acute effects of insulin on these neurons. However, we found only one of 24 cells (4%) that was hyperpolarized by –6.9 mV in response to insulin, while the other 23 cells remained non-responsive (from –50.8±1.5 to –51.0±1.5 mV, n=23) (Fig. 3A-C). These results suggest that NTS Phox2b neurons are not likely to be the source of the insulin-sensitive sEPSCs recorded in DMV cholinergic neurons.; It has been previously shown that the NTS also sends GABAergic synaptic input to the DMV [20]. As we observed insulininduced inhibition of sIPSCs as well, we tested the possibility that GABAergic neurons within the NTS may be inhibited by insulin. To selectively label GABAergic neurons, we crossed the Vgat-IRES-Cre recombinase expressing mice [22] with the tdTomato reporter mice (Vgat^cre/+::tdTomato mice) and tested the acute effects of insulin on NTS GABAergic neurons. We found that insulin hyperpolarized three of 16 cells (19%) by –11.7±3.0 mV (n=3), which was accompanied by a 47.0% decrease in input resistance (from 0.79±0.10 to 0.42±0.15 GΩ, n=3) with a reversal potential of –72.0±1.7 mV (n=3) (Fig. 3D-F). Subsequently, we targeted GABAergic neurons within the DMV and found that insulin hyperpolarized three of 15 cells (20%) by –16.0±4.1 mV (n=3), which was accompanied by a 36.0% decrease in input resistance (from 0.86±0.12 to 0.55±0.07 GΩ, n=3) with a reversal potential of –89.8±6.0 mV (n=3) (Fig. 3G-I). Together, these results support the idea that GABAergic neurons within the DVC may serve as the source of insulin-sensitive sIPSCs recorded in DMV cholinergic neurons, and that insulin-sensitive glutamatergic input may originate from neurons outside of the DVC (Fig. 3J).
Insulin does not alter the activity of sympathetic preganglionic neurons: Subsequently, we targeted the cholinergic sympathetic preganglionic neurons within the IML for current-clamp recordings (Fig. 4A, B). However, we noted that insulin treatments did not change the membrane potential of the sympathetic preganglionic neurons (Fig. 4C). The average membrane potential was –52.6±2.2 and –52.8±2.2 mV before and after insulin treatment, respectively (n=18) (Fig. 4D). Input resistance also remained unchanged by insulin treatment (from 0.68±0.08 to 0.66±0.08 GΩ, n=18) (Fig. 4E). These data suggest that sympathetic preganglionic neurons may not be necessary for the metabolic effects of insulin.
Insulin receptors expressed by cholinergic neurons are not critical for glucose homeostasis and gastrointestinal motility: Our patch clamp results demonstrated that insulin inhibited parasympathetic preganglionic neurons, but did not affect sympathetic preganglionic neurons. In addition, it has been previously reported that injections of insulin into the DVC lowered glucose production and decreased food intake [7,8]. Given these results, we tested the role of InsRs expressed by these neurons in the regulation of autonomic function. Firstly, we performed a GTT using the InsR^f/f mice and the ChAT^cre/+::InsR^f/f mice (Supplemental Fig. S3). Glucose (1.25 g/kg) was administered either intraperitoneally (IPGTT) or by oral gavage (OGTT) to 5-week-old male InsR^f/f mice and ChAT^cre/+::InsR^f/f mice. However, glucose tolerance did not show any difference between the genotypes (Supplemental Fig. S3A, B). We also injected insulin (1 U/kg) IPITT, but there was no difference in insulin sensitivity between the genotypes (Supplemental Fig. S3C). These results suggest that the InsRs expressed by parasympathetic preganglionic neurons may not play a critical role in the regulation of glucose balance.; Subsequently, we tested the possibility that the inhibitory effects of insulin on the parasympathetic preganglionic neurons may result in altered GI motility. We injected (intraperitoneal) overnight-fasted wild-type mice with DPBS or insulin (0.5 U/kg). Thirty minutes after the injections, mice were given charcoal solutions (100 μL) by oral gavage and were sacrificed after 50 minutes to measure the distance of charcoal movement in the GI tract (Supplemental Fig. S4A). The intraperitoneal injections of insulin did not lead to any significant differences in GI motility compared to DPBS injections (Supplemental Fig. S4B). We also performed intracerebroventricular injections to deliver insulin directly to the cerebrospinal fluid, but there was no significant difference in GI motility between DPBS- and insulin-injected groups (Supplemental Fig. S4C). These results suggest that exogenous insulin does not affect GI motility.; We also measured gastric emptying by the acetaminophen assay, as described previously [15]. Since acetaminophen is poorly absorbed in the stomach, it can be a good indicator of gastric emptying. We gave 5-week-old male InsR^f/f mice and ChAT^cre/+:: InsR^f/f mice acetaminophen (90 mg/kg) by oral gavage and measured serum concentrations of acetaminophen at 15 minutes and 1 hour after administration. Consistent with the charcoal movement experiments described above, there was no significant difference in serum concentrations of acetaminophen between the genotypes according to whether the assay was done in the dark cycle or in the light cycle after 18-hour fasting (Supplemental Fig. S4D). These results suggest that InsRs expressed by autonomic preganglionic neurons do not play a critical role in the regulation of gastric emptying.

DISCUSSION

In this study, we demonstrated that insulin directly and indirectly inhibited parasympathetic preganglionic neurons. Notably, we found that insulin did not affect the membrane potential of sympathetic preganglionic neurons. Therefore, our data suggest that insulin may affect autonomic function preferentially via the parasympathetic nervous system, while the contribution of the sympathetic nervous system is likely to be minimal.

It is well known that insulin affects the autonomic nervous system to regulate energy balance and glucose homeostasis [23]. However, it was not clear whether the sympathetic nervous system may also contribute to the autonomic effects of insulin. In this study, we recorded chemically identified sympathetic preganglionic neurons of the IML and found no acute effects of insulin on those neurons (Fig. 4). However, we observed that insulin stimulates InsRs, leading to direct inhibition of the cholinergic neurons within the DMV (Fig. 1). A previous study also showed that insulin inhibited stomach-projecting DMV neurons by hyperpolarizing the membrane potential and decreasing action potential frequency [9]. These observations are consistent with insulin-induced inhibition of hypothalamic neurons [4-6], suggesting that insulin directly inhibits the activity of multiple types of neurons.

We note that the acute effects of insulin on synaptic currents are not consistent between our current findings and those reported previously. It was previously reported that insulin decreased sEPSC and mEPSC frequency, but not the mean amplitude, while the sIPSC frequency remained unchanged by insulin [9]. Based on these findings, the authors proposed that insulin acts at receptors on presynaptic glutamatergic terminals onto DMV neurons. By contrast, we found that insulin inhibited both the frequency and mean amplitude of sEPSCs and mEPSCs (Fig. 2). Our findings suggest that insulin not only acts on presynaptic glutamatergic terminals, but also on the somata of presynaptic glutamatergic neurons and the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors expressed by DMV cholinergic neurons. In addition, we found insulin-induced suppression of sIPSCs and the frequency of mIPSCs (Fig. 2), suggesting that insulin acts both on presynaptic GABAergic terminals and somata to decrease inhibitory input. Therefore, our results provide evidence that insulin broadly inhibits the neural circuitry involving the parasympathetic preganglionic neurons of the DMV. This hypothesis is supported by our observation that insulin inhibited GABAergic neurons in the NTS and DMV (Fig. 3). Since only 4% of NTS Phox2b neurons were inhibited by insulin, these neurons may not be the source of insulin-sensitive glutamatergic input. Instead, glutamatergic innervation of the DVC from arcuate POMC neurons [24] and VMH SF-1 neurons [25] may be sensitive to insulin, as both neuron types are inhibited by insulin [4,6,26,27]. This hypothesis needs to be clarified in future studies. We note that the authors of the previous study injected gastric musculature with the attenuated Bartha strain of pseudorabies virus (PRV614) and recorded labeled neurons within the DMV [9], and that PRV614 may also label neurons that polysynaptically innervate the target organ [28]. Therefore, it is likely that non-preganglionic neurons were also targeted for recording, which may explain the discrepancy between the studies.

We found in this study that not all parasympathetic preganglionic neurons were inhibited by insulin; instead, insulin inhibited nine of 12 (43%) and six of 11 (55%) of recorded neurons in the absence and presence of TTX and SBs, respectively (Fig. 1). In addition, insulin inhibited about 20% of GABAergic neurons within the NTS and the DMV (Fig. 3), and previous studies demonstrated that insulin inhibited 25% of arcuate POMC neurons and about 23% of VMH SF-1 neurons [26,27]. The heterogeneous responses to insulin may represent the functional heterogeneity of parasympathetic preganglionic neurons. Indeed, it has been previously shown that stimulation of the rostral and caudal DMV resulted in gastric contraction and relaxation, respectively [29]. Although it is unclear which physiological functions are carried out by insulin-responsive parasympathetic preganglionic neurons, it may not be the regulation of GI motility, given our results in Supplemental Fig. S4. Future studies are warranted to identify specific physiological functions of insulinresponsive parasympathetic preganglionic neurons. We also need to delineate whether insulin-responsive arcuate POMC neurons, VMH SF-1 neurons, and DVC GABAergic neurons are upstream of parasympathetic preganglionic neurons in the DMV. This information would be helpful for understanding what insulin does in the brainstem vagal circuitry.

We unexpectedly found that deleting InsRs in cholinergic neurons did not alter glucose homeostasis (Supplemental Fig. S3), which appears to contradict results from a previous study, which showed that insulin signaling in the brainstem parasympathetic nervous system regulated glucose metabolism [7]. However, Filippi et al. [8] injected exogenous insulin directly into the brainstem, while we observed the effects of endogenous insulin. It is thus possible that the roles of exogenous and endogenous insulin may be different, or that the inhibition of parasympathetic preganglionic neurons by endogenous insulin is not as strong as that by exogenous insulin (i.e., enough to produce a discernable phenotype).

In conclusion, we demonstrated that insulin preferentially inhibited the brainstem parasympathetic neurocircuitry over the spinal cord sympathetic neurons to inhibit the parasympathetic preganglionic neurons of the DMV directly and indirectly. These findings may serve as neural correlates to regulate homeostasis via the insulin-dependent regulation of autonomic activity. Further studies are required to directly address the in vivo role of InsRs expressed by the parasympathetic preganglionic neurons.

Supplementary Material

Supplemental Fig. S1.

Summary of insulin effects on parasympathetic preganglionic neurons. Drawings of four rostrocaudal levels of the mouse brainstem summarize location of acute effects of insulin on parasympathetic preganglionic neurons from ChAT^cre/+::tdTomato mice in the presence of tetrodotoxin (TTX) and synaptic blockers (SBs; upper panels) and ChAT^cre/+::InsR^f/f mice in control artificial cerebrospinal fluid (ACSF) solutions (lower panels). Blue dots indicate hyperpolarized cells while black dots indicate cells with no effects. 4V, fourth ventricle; 12N, nucleus of the hypoglossal nerve; NTS, nucleus tractus solitaries; DMV, dorsal motor nucleus of the vagus nerve; AP, area postrema; Cu, nucleus cuneatus; Gr, nucleus gracilis; ChAT, choline acetyltransferase; InsR^f/f, insulin receptor flox.

enm-2023-1725-Supplemental-Fig-S1.pdf

Supplemental Fig. S2.

Summary of insulin effects on postsynaptic currents onto parasympathetic preganglionic neurons. Drawings of four rostrocaudal levels of the mouse brainstem summarize location of acute effects of insulin on spontaneous excitatory postsynaptic current (sEPSC), spontaneous inhibitory postsynaptic current (sIPSC), miniature EPSC (mEPSC), and miniature IPSC (mIPSC) (from top to bottom) recorded from parasympathetic preganglionic neurons. Orange and blue dots indicate increased and decreased frequency (left to dashed line) or amplitude (right to dashed line), respectively. Black dots indicate cells with no effects.

enm-2023-1725-Supplemental-Fig-S2.pdf

Supplemental Fig. S3.

Insulin receptors expressed by the cholinergic neurons are dispensable for glucose tolerance and insulin sensitivity. (A) Graphs demonstrate results of intraperitoneal glucose tolerance tests (IPGTTs; intraperitoneal glucose, 1.25 g/kg) from insulin receptor flox (InsR^f/f) mice (n=8, black dots and lines) and ChAT^cre/+::InsR^f/f mice (n=12, blue dots and lines). (B) Graphs demonstrate results of oral glucose tolerance test (OGTT; oral glucose, 1.25 g/kg) from InsR^f/f mice (n=3, black dots and lines) and ChAT^cre/+::InsR^f/f mice (n=8, blue dots and lines). (C) Graphs demonstrate results of intraperitoneal insulin tolerance test (IPITT; intraperitoneal insulin, 1 U/kg) from InsR^f/f mice (n=8, black dots and lines) and ChAT^cre/+::InsR^f/f mice (n=12, blue dots and lines). No statistical significance was observed by two-way analysis of variance (ANOVA) with Bonferroni’s multiple comparisons test. ChAT, choline acetyltransferase.

enm-2023-1725-Supplemental-Fig-S3.pdf

Supplemental Fig. S4.

Insulin and insulin receptors expressed by the cholinergic neurons do not regulate gastrointestinal (GI) motility. (A) Image demonstrates GI tract with charcoal (black) in the lumen. X represents the length between the distal end of pylorus and the most distal end of charcoal traveled within the lumen. Y represents the length between the distal end of pylorus and the distal end of ileum. (B) Graph demonstrates effects of intraperitoneal (i.p.) insulin (0.5 U/kg, blue rectangles, n=10) vs. intraperitoneal Dulbecco՚s phosphate-buffered saline (DPBS; black circles, n=10) on charcoal movement (X/Y). No statistical significance was observed by Mann-Whitney test. (C) Graph demonstrates effects of intracerebroventricular (i.c.v.) insulin (4 mU, blue rectangles, n=9) vs. intraperitoneal DPBS (black circles, n=8) on charcoal movement (X/Y). No statistical significance was observed by Mann-Whitney test. (D) Graph demonstrates results of the acetaminophen assay in the insulin receptor flox (InsR^f/f) mice (black circles, n=8 in the dark cycle, n=6 in the light cycle) and ChAT^cre/+::InsR^f/f mice (blue rectangles, n=15 in the dark cycle, n=14 in the light cycle). Mice were fed ad libitum in the dark cycle experiments, and were fasted for 18 hours prior to the light cycle experiments. No statistical significance was observed by two-way analysis of variance (ANOVA) with Bonferroni’s multiple comparisons test. ChAT, choline acetyltransferase.

enm-2023-1725-Supplemental-Fig-S4.pdf

Article information

CONFLICTS OF INTEREST

No potential conflict of interest relevant to this article was reported.

AUTHOR CONTRIBUTIONS

Conception or design: U.H., J.W.S. Acquisition, analysis, or interpretation of data: U.H., Y.Y.K., J.W.S. Drafting the work or revising: U.H., J.W.S. Final approval of the manuscript: U.H., J.W.S.

Acknowledgements

This work was supported by the National Research Foundation of Korea (NRF-2022R1A2C3005613 to Jong-Woo Sohn) funded by the Korean Ministry of Science and ICT.

Fig. 1.

Insulin directly inhibits parasympathetic preganglionic neurons. (A) Image demonstrating the location of the brainstem section (left) and a parasympathetic preganglionic neuron of the dorsal motor nucleus of the vagus nerve (DMV; right) targeted for whole-cell patch clamp recordings, obtained from ChAT^cre/+::tdTomato mice. (B) Bright-field, tetramethylrhodamine (TRITC), fluorescein isothiocyanate (FITC), and merged images (from left to right) of the targeted neuron (arrows) in the DMV of ChAT^cre/+::tdTomato mice. Scale bar=50 μm. (C) Membrane potential hyperpolarizes in response to insulin (50 nM) treatment. The dashed line indicates the baseline membrane potential. (D) Voltage responses to hyperpolarizing current steps from –50 to 0 pA, applied as indicated by arrows in (C). (E) The voltage-current relationship demonstrates insulin-induced decreases in input resistance. (F) Drawings of four rostrocaudal levels of the mouse brainstem summarize the location of the acute effects of insulin on parasympathetic preganglionic neurons. Blue dots indicate hyperpolarized cells, while black dots indicate cells with no effects. (G) The membrane potential hyperpolarized in response to insulin treatment in the presence of tetrodotoxin (TTX, 500 nM) and synaptic blockers (SBs, 1 mM kynurenic acid+50 μM picrotoxin). The dashed line indicates baseline membrane potential. (H) The membrane potential did not change after insulin treatment parasympathetic preganglionic neurons from ChAT^cre/+::InsR^f/f::tdTomato mice. The dashed line indicates baseline membrane potential. (I) Plots summarizing insulin-induced changes in membrane potential. (J) Histogram summarizing the percentage of neurons showing insulin responses. ACSF, artificial cerebrospinal fluid; 4V, fourth ventricle; 12N, nucleus of the hypoglossal nerve; NTS, nucleus tractus solitaries; AP, area postrema; CC, central canal; Cu, nucleus cuneatus; Gr, nucleus gracilis; ChAT, choline acetyltransferase; InsR^f/f, insulin receptor flox. ^aP<0.01, Fisher’s exact test for the occurrence of hyperpolarization and no effects.

Fig. 2.

Insulin reduces excitatory and inhibitory postsynaptic currents onto parasympathetic preganglionic neurons. (A) Traces demonstrating spontaneous excitatory postsynaptic currents (sEPSCs) before (upper) and during (lower) insulin (50 nM) treatment. Holding potential=–60 mV. (B, C) Insulin reduced the frequency (B) and mean amplitude (C) of sEPSCs. (D) Traces demonstrating spontaneous inhibitory postsynaptic currents (sIPSCs) before (upper) and during (lower) insulin treatment. Holding potential=–10 mV. (E, F) Insulin reduced the frequency (E) and mean amplitude (F) of sIPSCs. (G) Traces demonstrating miniature EPSCs (mEPSCs) before (upper) and during (lower) insulin treatment. Holding potential=–60 mV. (H, I) Insulin reduced the frequency (H) and mean amplitude (I) of mEPSCs. (J) Traces demonstrating miniature IPSCs (mIPSCs) before (upper) and during (lower) insulin treatment. Holding potential=–10 mV. (K, L) Insulin reduced the frequency (K), but not the mean amplitude (L) of mIPSCs. ACSF, artificial cerebrospinal fluid; TTX, tetrodotoxin; ns, not significant (P>0.05). ^aP<0.05; ^bP<0.01; ^cP<0.001 by the Wilcoxon matched-pairs signed rank test.

Fig. 3.

Insulin hyperpolarizes GABAergic neurons of the nucleus tractus solitaries (NTS) and dorsal motor nucleus of the vagus nerve (DMV). (A) Bright-field, tetramethylrhodamine (TRITC), fluorescein isothiocyanate (FITC), and merged images (from left to right) of the targeted neuron (arrows) in the NTS of Phox2bcre/+::tdTomato mice. Scale bar=50 μm. (B) The membrane potential hyperpolarized in response to insulin (50 nM) treatment. The dashed line indicates the baseline membrane potential. (C) Plots summarizing insulin-induced changes in membrane potential. (D) Bright-field, TRITC, FITC, and merged images (from left to right) of the targeted neuron (arrows) in the NTS of Vgat^cre/+::tdTomato mice. Scale bar=50 μm. (E) The membrane potential hyperpolarized in response to insulin treatment. The dashed line indicates the baseline membrane potential. (F) Plots summarize insulin-induced changes in the membrane potential. (G) Bright-field, TRITC, FITC, and merged images (from left to right) of the targeted neuron (arrows) in the DMV of Vgat^cre/+::tdTomato mice. Scale bar=50 μm. (H) The membrane potential hyperpolarized in response to insulin treatment. The dashed line indicates the baseline membrane potential. (I) Plots summarizing insulin-induced changes in the membrane potential. (J) Illustration summarizing modulation of parasympathetic preganglionic neurons and their upstream neurons by insulin and insulin receptors. Phox2b, paired-like homeobox 2b; Vgat, vesicular GABA transporter; ACSF, artificial cerebrospinal fluid; AP, area postrema; ChAT, choline acetyltransferase; InsR, insulin receptor; GABA, γ-aminobutyric acid.

Fig. 4.

Insulin does not affect the activity of sympathetic preganglionic neurons. (A) Image demonstrating the location of the spinal cord section (left) and the sympathetic preganglionic neuron of intermediolateral column (IML, right) targeted for whole-cell patch clamp recordings, obtained from ChAT^cre/+::tdTomato mice. (B) Bright-field, tetramethylrhodamine (TRITC), fluorescein isothiocyanate (FITC), and merged images (from left to right) of targeted neuron (arrows) in the IML of ChAT^cre/+::tdTomato mice. Scale bar=50 μm. (C) The membrane potential did not change in response to insulin (50 nM) treatment. The dashed line indicates the baseline membrane potential. (D, E) Plots summarizing insulin-induced changes in the membrane potential (D) and input resistance (E). Not significant (ns) (P>0.05) by Wilcoxon matched-pairs signed rank test. ChAT, choline acetyltransferase; ACSF, artificial cerebrospinal fluid.

References

1. Chen C, Cohrs CM, Stertmann J, Bozsak R, Speier S. Human beta cell mass and function in diabetes: recent advances in knowledge and technologies to understand disease pathogenesis. Mol Metab 2017;6:943–57.Article PubMed PMC
2. Sonksen P, Sonksen J. Insulin: understanding its action in health and disease. Br J Anaesth 2000;85:69–79.Article PubMed
3. Brüning JC, Gautam D, Burks DJ, Gillette J, Schubert M, Orban PC, et al. Role of brain insulin receptor in control of body weight and reproduction. Science 2000;289:2122–5.Article PubMed
4. Klockener T, Hess S, Belgardt BF, Paeger L, Verhagen LA, Husch A, et al. High-fat feeding promotes obesity via insulin receptor/PI3K-dependent inhibition of SF-1 VMH neurons. Nat Neurosci 2011;14:911–8.Article PubMed PMC PDF
5. Konner AC, Janoschek R, Plum L, Jordan SD, Rother E, Ma X, et al. Insulin action in AgRP-expressing neurons is required for suppression of hepatic glucose production. Cell Metab 2007;5:438–49.Article PubMed
6. Hill JW, Elias CF, Fukuda M, Williams KW, Berglund ED, Holland WL, et al. Direct insulin and leptin action on proopiomelanocortin neurons is required for normal glucose homeostasis and fertility. Cell Metab 2010;11:286–97.Article PubMed PMC
7. Filippi BM, Yang CS, Tang C, Lam TK. Insulin activates Erk1/2 signaling in the dorsal vagal complex to inhibit glucose production. Cell Metab 2012;16:500–10.Article PubMed
8. Filippi BM, Bassiri A, Abraham MA, Duca FA, Yue JT, Lam TK. Insulin signals through the dorsal vagal complex to regulate energy balance. Diabetes 2014;63:892–9.Article PubMed PDF
9. Blake CB, Smith BN. Insulin reduces excitation in gastricrelated neurons of the dorsal motor nucleus of the vagus. Am J Physiol Regul Integr Comp Physiol 2012;303:R807–14.Article PubMed PMC
10. Grabauskas G, Moises HC. Gastrointestinal-projecting neurones in the dorsal motor nucleus of the vagus exhibit direct and viscerotopically organized sensitivity to orexin. J Physiol 2003;549(Pt 1):37–56.Article PubMed PMC PDF
11. Cork SC, Richards JE, Holt MK, Gribble FM, Reimann F, Trapp S. Distribution and characterisation of glucagon-like peptide-1 receptor expressing cells in the mouse brain. Mol Metab 2015;4:718–31.Article PubMed PMC
12. Zheng Z, Lewis MW, Travagli RA. In vitro analysis of the effects of cholecystokinin on rat brain stem motoneurons. Am J Physiol Gastrointest Liver Physiol 2005;288:G1066–73.Article PubMed PMC
13. Sohn JW, Harris LE, Berglund ED, Liu T, Vong L, Lowell BB, et al. Melanocortin 4 receptors reciprocally regulate sympathetic and parasympathetic preganglionic neurons. Cell 2013;152:612–9.Article PubMed PMC
14. Ju SH, Yun H, Oh Y, Choi Y, Sohn JW. Melanocortin-4 receptors activate sympathetic preganglionic neurons and elevate blood pressure via TRPV1. Cell Rep 2022;41:111579.Article PubMed
15. Chambers AP, Smith EP, Begg DP, Grayson BE, Sisley S, Greer T, et al. Regulation of gastric emptying rate and its role in nutrient-induced GLP-1 secretion in rats after vertical sleeve gastrectomy. Am J Physiol Endocrinol Metab 2014;306:E424–32.Article PubMed PMC
16. Rossi J, Balthasar N, Olson D, Scott M, Berglund E, Lee CE, et al. Melanocortin-4 receptors expressed by cholinergic neurons regulate energy balance and glucose homeostasis. Cell Metab 2011;13:195–204.Article PubMed PMC
17. Madisen L, Zwingman TA, Sunkin SM, Oh SW, Zariwala HA, Gu H, et al. A robust and high-throughput Cre reporting and characterization system for the whole mouse brain. Nat Neurosci 2010;13:133–40.Article PubMed PDF
18. Bruning JC, Michael MD, Winnay JN, Hayashi T, Horsch D, Accili D, et al. A muscle-specific insulin receptor knockout exhibits features of the metabolic syndrome of NIDDM without altering glucose tolerance. Mol Cell 1998;2:559–69.Article PubMed
19. Kang BJ, Chang DA, Mackay DD, West GH, Moreira TS, Takakura AC, et al. Central nervous system distribution of the transcription factor Phox2b in the adult rat. J Comp Neurol 2007;503:627–41.Article PubMed
20. Davis SF, Derbenev AV, Williams KW, Glatzer NR, Smith BN. Excitatory and inhibitory local circuit input to the rat dorsal motor nucleus of the vagus originating from the nucleus tractus solitarius. Brain Res 2004;1017:208–17.Article PubMed PMC
21. Scott MM, Williams KW, Rossi J, Lee CE, Elmquist JK. Leptin receptor expression in hindbrain Glp-1 neurons regulates food intake and energy balance in mice. J Clin Invest 2011;121:2413–21.Article PubMed PMC
22. Vong L, Ye C, Yang Z, Choi B, Chua S Jr, Lowell BB. Leptin action on GABAergic neurons prevents obesity and reduces inhibitory tone to POMC neurons. Neuron 2011;71:142–54.Article PubMed PMC
23. Hyun U, Sohn JW. Autonomic control of energy balance and glucose homeostasis. Exp Mol Med 2022;54:370–6.Article PubMed PMC PDF
24. Kwon E, Joung HY, Liu SM, Chua SC Jr, Schwartz GJ, Jo YH. Optogenetic stimulation of the liver-projecting melanocortinergic pathway promotes hepatic glucose production. Nat Commun 2020;11:6295.Article PubMed PMC PDF
25. Yang F, Liu Y, Chen S, Dai Z, Yang D, Gao D, et al. A GABAergic neural circuit in the ventromedial hypothalamus mediates chronic stress-induced bone loss. J Clin Invest 2020;130:6539–54.Article PubMed PMC
26. Sohn JW, Oh Y, Kim KW, Lee S, Williams KW, Elmquist JK. Leptin and insulin engage specific PI3K subunits in hypothalamic SF1 neurons. Mol Metab 2016;5:669–79.Article PubMed PMC
27. Williams KW, Margatho LO, Lee CE, Choi M, Lee S, Scott MM, et al. Segregation of acute leptin and insulin effects in distinct populations of arcuate proopiomelanocortin neurons. J Neurosci 2010;30:2472–9.Article PubMed PMC
28. Card JP, Enquist LW. Transneuronal circuit analysis with pseudorabies viruses. Curr Protoc Neurosci 2014;68:1.5.1–1.5.39.Article PubMed PMC PDF
29. Zhou SY, Lu YX, Yao H, Owyang C. Spatial organization of neurons in the dorsal motor nucleus of the vagus synapsing with intragastric cholinergic and nitric oxide/VIP neurons in the rat. Am J Physiol Gastrointest Liver Physiol 2008;294:G1201–9.Article PubMed