Vagal Sensory Neuron-Derived FGF3 Controls Insulin Secretion

Summary:

Vagal nerve stimulation (VNS) has emerged as a promising modality for treating a wide range of chronic conditions including metabolic disorders. However, the cellular and molecular pathways driving these clinical benefits remain largely obscure. Here, we demonstrated that Fgf3 mRNA is upregulated in the mouse vagal ganglia under acute metabolic stress. Systemic and vagal sensory overexpression of Fgf3 enhanced glucose-stimulated insulin secretion (GSIS), improved glucose excursion, and increased energy expenditure and physical activity. Fgf3-elicited insulinotropic and glucose-lowering responses were recapitulated when overexpression of Fgf3 was restricted to the pancreas-projecting vagal sensory neurons. Genetic ablation of Fgf3 in pancreatic vagal afferents exacerbated high-fat diet-induced glucose intolerance and blunted GSIS. Finally, electrostimulation of the vagal afferents enhanced GSIS and glucose clearance independently of efferent outputs. Collectively, we demonstrate a direct role for the vagal afferent signaling in GSIS and identify Fgf3 as a vagal sensory-derived metabolic factor that controls pancreatic β-cell activity.

eTOC Blurb

Adaptive β-cell response is a common feature of the prediabetic insulin-resistant state in humans and rodents. In this issue of Developmental Cell, Tahiri et al. identify Fgf3 as a vagal sensory-derived metabolic factor that controls insulin secretion and systemic glucose homeostasis and demonstrate a direct role for the vagal afferent signaling in pancreatic β-cell activity.

Graphical Abstract

Introduction:

Energy homeostasis in mammalian systems is regulated by a myriad of humoral factors and neuronal networks that facilitate the perception, integration and processing of peripherally-derived metabolic cues by the central nervous system (CNS) ¹. The vagus nerve is a key neuroanatomical substrate that mediates intercommunication between the CNS and peripheral metabolic organs. It plays important modulatory roles in energy balance and constitutes an attractive target for neuromodulation-based therapies to treat metabolic disorders ^2,3. The vagus nerve is composed of sensory (afferent) and motor (efferent) fibers with the number of afferent neurons exceeding the efferent neurons in a ratio of 4:1 ^4–6. The vagal afferents convey metabolic information via the vagal jugular-nodose ganglia (JNG) to the nucleus tractus solitarius (NTS) in the brainstem where information is processed and transmitted to the periphery through the efferent neurons of the dorsal motor nucleus of the vagus (DMV) ⁷. In addition to their sensing capabilities, the vagal afferents exert efferent functions as they express and release active neurotransmitters and neuropeptides, including glutamate, substance P (SP), calcitonin gene-related protein (CGRP) and cocaine- and amphetamine-regulated transcript (CART) ^8–15. Although the notion that sensory neurons carry out efferent functions in the periphery is relatively understudied, it is relevance is underscored in various physiological and pathophysiological contexts, including in urodele limb regeneration ^16,17, epidermal repair/hair follicle growth ^18,19, maintenance of gut homeostasis ^20,21, neurogenic inflammation ^22,23, energy balance ^24,25 and cancer formation ^26,27.

There is mounting evidence to support the role of sensory neurons in pancreatic β-cell function ^28,29 and dysfunction ^30,31. The vagus nerve projects into the pancreas ^32,33; vagal afferents can be retrogradely traced from the pancreas in rodents ^29,34,35; and vagal terminal endings can be mapped in the vicinity of islet β cells in mice and humans ^36–38. The vagal afferent system relays to the CNS the status of the islet-cell activity via sensing of locally released endocrine hormones, including glucagon ³⁹, insulin ⁴⁰ and serotonin ³⁸. Vagotomy impairs compensatory pancreatic β-cell expansion in response to physiological and pathophysiological insulin resistance ^41–43. Selective vagotomy and vagus nerve stimulation (VNS) paradigms revealed differential roles for efferent and afferent neurons in insulin secretion ^44–47. Nevertheless, the inherent lack of specificity associated with vagotomy and VNS paradigms precludes differentiating the unique roles of the afferent, efferent and afferent-mediated efferent signals of the vagus nerve ⁴⁸. Recent genetically guided anatomical mapping studies reported the existence of molecularly and functionally distinct vagal efferent ⁴⁹ and afferent ⁵⁰ neurons projecting in metabolic tissues. However, the identity and role of the molecular signals mediating the downstream biological effects of these neuronal subsets are currently unknown, particularly those pertaining to the control of islet β-cell activity. The identification of such neuromodulators will improve our understanding of pancreatic β-cell neurobiology and unravel druggable targets for individuals with impaired insulin secretion.

Here, we used RNA sequencing (RNA-Seq) to interrogate the molecular profile of the mouse vagal afferents. We identified the fibroblast growth factor 3 (Fgf3) mRNA transcript as highly elevated in the vagal jugular-nodose ganglia under acute metabolic stress. Systemic and vagal ganglia overexpression of Fgf3 in mice improved glucose clearance, insulin secretion, energy expenditure and physical activity. The β-cell-centered effects were recapitulated when Fgf3 overexpression was solely restricted to the pancreas-projecting vagal sensory neurons. Mice lacking Fgf3 specifically in vagal sensory neurons innervating the pancreas showed impaired glucose tolerance and defective insulin secretion under high-fat diet conditions. Ex vivo assays revealed that Fgf3 enhances glucose-stimulated insulin secretion (GSIS) directly from isolated pancreatic islets in an FGF receptor (FGFR)-dependent manner. Vagal afferent stimulation improved glucose clearance and insulin release independently of DMV preganglionic efferents. Together, we provide evidence for the regulation of insulin secretion by the vagal afferent signaling and reveal a neuroendocrine role for the vagal sensory-derived Fgf3 molecules in the control of pancreatic β-cell activity and glucose homeostasis.

Results:

Transcriptome profiling of the vagal sensory ganglia under acute metabolic stress:

To identify genes encoding secretory proteins in vagal sensory neurons potentially implicated in adaptive β-cell function during metabolic stress, we performed an unbiased RNA-seq analysis of JNG derived derived from vehicle- and S961-treated mice. Briefly, six-week-old C57BL/6J male mice were continuously infused with the insulin receptor antagonist S961 (10 nmoles/week) or vehicle (PBS) for two weeks before vagal ganglia were harvested for RNA sequencing (Figure 1A). As previously described ⁵¹, S961-treated mice maintained normal body mass but developed progressive hyperglycemia, increased β-cell mass, glucose intolerance, systemic insulin resistance and increased insulin secretion (Figure S1A–F). RNA-Seq yielded 361 million high-quality reads from sequencing three biological replicates of PBS- and S961-treated mice (Table S1). To evaluate the transcriptome profiles, we conducted a principal component analysis (PCA) and found that PBS and S961 genomic datasets were clustered into two independent groups (Figure 1B), suggesting that nodose ganglia derived from PBS- and S961-treated mice exhibit distinct transcriptome signatures. To identify genes that are differentially expressed between control and experimental groups, we quantified the expression levels of all mouse genes and compared them using the DESeq2 method. We demonstrated that 90 genes were differentially expressed (FDR < 0.05), including 66 up-regulated and 24 downregulated genes in response to S961 (Figure 1C and Table S2). Furthermore, we found that the differences in gene expression between PBS and S961 conditions are consistent among the three biological replicates (Figure 1D). To determine the differential biological pathways in PBS- and S961-treated groups, we performed Gene Ontology (GO) enrichment analysis and Gene Set Enrichment Analysis (GSEA). We found noticeable differences in GO biological pathways (FDR < 0.05, Figure 1E and Table S3) between groups. Among pathways, we observed differential expression in glucose-mediated signaling pathways and regulation of extracellular signal-regulated kinases ERK1 and ERK2 and p38 mitogen-activated protein kinase (p38MAPK) cascades. Moreover, several pathways related to the regulation of endocrine hormone and peptide secretion and regulation of hormone levels were upregulated in the S961 model (Figure 1E). Finally, we observed that GO terms related to the regulation of insulin secretion were evident in S961-infused mice (Figure 1E) consistent with the enhanced insulin levels as an adaptive response to acute insulin resistance (Figure S1F). The differential regulation of the endocrine hormone secretion pathway was also reflected by GSEA analyses (Figure 1F and Table S4). Interestingly, differential gene expression analyses revealed the enrichment (Log₂ (S961/PBS) ≈ 3.6) of Fgf3 mRNA transcript in the jugular-nodose ganglia derived from S961-treated mice. Using quantitative RT-PCR, we confirmed the elevation of Fgf3 mRNA expression levels in JNG harvested from S961- vs. PBS-treated mice (Figure 1G). However, in vagal ganglia derived from mice fed with a high-fat diet (HFD) for twenty-four weeks (Figure S1G–J), the mRNA levels of Fgf3 were comparable to those harvested from mice fed with a chow diet (Figure 1H). Together, these findings suggest that acute insulin resistance induces molecular changes and alterations in signaling pathways potentially relevant to compensatory insulin secretion and highlight Fgf3 as a putative vagal sensory-derived endocrine hormone that regulates pancreatic β-cell activity.

Figure 1: Transcriptome profiling of the vagal sensory ganglia under acute metabolic stress.

A. Schematic of the experimental workflow for the transcriptome profiling of the vagal sensory ganglia. B. PCA plot of gene expression in vagal ganglia harvested from PBS- and S961-treated mice. C. Volcano plots of genes expressed in vagal ganglia derived from PBS- and S961-treated mice. The differentially expressed genes (FDR < 0.05) were highlighted in red for upregulated genes and in blue for downregulated genes. D. Heatmap depicting differential gene relative expression in vagal ganglia isolated from PBS- and S961-treated mice. E. Gene ontology (GO) enrichment analysis in vagal ganglia derived from PBS- and S961-treated mice. FDR < 0.05 was used as the cutoff criteria. F. Gene set enrichment analysis (GSEA) plots of gene clusters that are enriched in vagal ganglia harvested from S961 versus PBS conditions. G. Relative Fgf3 mRNA expression levels in vagal ganglia isolated from PBS- versus S961-treated mice. H. Relative Fgf3 mRNA expression levels in vagal ganglia harvested from mice fed with a chow diet (CD) or a high-fat diet (HFD). Data represent mean ± SEM. *p ≤ 0.05, (n=3) in S961 studies and (n=3–5) in HFD studies.

Systemic and vagal sensory-derived Fgf3 enhance insulin secretion:

To investigate the role of Fgf3 in glucose homeostasis, we administered adeno-associated viruses encoding GFP (AAV9-GFP) or FGF3 (AAV9-FGF3) proteins via tail vein injection (10¹¹ genome copies/mouse) in four-week-old C57BL/6J male mice. Mice were then metabolically characterized for fifteen weeks after AAV transduction (Figure 2A). The overexpression of Fgf3 mRNA transcript was confirmed by RT-PCR in livers harvested from AAV9-FGF3 mice (Figure 2B). Weekly monitoring of body weight did not reveal differences between AAV9-FGF3 and AAV9-GFP mice (Figure 2C). Random-fed blood glucose levels tended to be lower in mice overexpressing FGF3 (Figure 2D). Glucose clearance was enhanced in AAV9-FGF3 vs. AAV9-GFP mice (Figure 2E) while circulating insulin levels and whole-body insulin sensitivity remained similar between groups (Figures 2F and 2G). Interestingly, glucose-induced insulin release was largely improved in AAV9-FGF3 mice (Figure 2H). Islet β-cell proliferation, area and mass were unaltered between control and experimental groups (Figure 2I–K). To further characterize the metabolic profile of AAV9-GFP and AAV9-FGF3 mice, we used the comprehensive laboratory analysis monitoring system (CLAMS) system. Indirect calorimetry assays demonstrated that AAV9-FGF3 mice have a higher respiratory exchange ratio (RER) indicative of predominant usage of carbohydrates as the main metabolic fuel (Figure 2L). Next, we examined whether vagal sensory-restricted overexpression of FGF3 molecules generates similar metabolic benefits. To this end, we injected AAV9-GFP or AAV9-FGF3 (150 nL, 10¹³ genome copies/mL) in the left JNG of eight-week-old C57BL/6J male mice (Figure 2M) as previously described ^52,53. RT-PCR assays confirmed that Fgf3 mRNA levels are higher in the left versus right vagal ganglia (right ganglia 0.4 ± 0.2 versus left ganglia 2.4 ± 0.7, p < 0.05, n = 5). Fasting blood glucose (Figure 2N) and body mass (Figure 2O) were similar between groups. However, mice intraganglionically injected with AAV9-FGF3 had enhanced glucose excursion in comparison to age-matched control AAV9-GFP mice (Figure 2P). Further, AAV9-FGF3 mice had a markedly improved glucose-induced insulin release (Figure 2Q) but similar insulin sensitivity (Figure 2R) as compared to control AAV9-GFP mice. Interestingly, AAV9-FGF3 mice exhibited higher oxygen consumption (Figure 2S) and CO₂ release (Figure 2T), and normal RER (Figure 2U). Finally, AAV9-FGF3 mice exhibited higher energy expenditure (Figure 2V) and physical activity (Figure 2W). Taken together, these experiments suggest that systemic and vagal sensory overexpression of Fgf3 increased β-cell activity, enhanced glucose excursion and improved overall energy homeostasis.

Figure 2: Systemic and vagal sensory-derived Fgf3 enhance insulin secretion.

A. Schematic of the experimental workflow for the study of the effect of systemic Fgf3 in glucose homeostasis. B. Expression of FGF3 in the liver as assessed by RT-PCR. C. Body mass. D. Random-fed blood glucose levels. E. Glucose tolerance test. F. Insulin tolerance test. G. Random-fed blood insulin levels. H. Glucose-stimulated insulin secretion. I. Percentage of Ki67⁺ insulin⁺ cells. J. Islet β-cell area. K. Islet β-cell mass. L. Respiratory exchange ratio (RER). M. Schematic of the experimental workflow for the study of the effect of vagal sensory-restricted overexpression of Fgf3 in glucose homeostasis. N. Fasting blood glucose levels. O. Body mass. P. Glucose tolerance test. Q. Glucose-stimulated insulin secretion. R. Insulin tolerance test. S. Oxygen consumption (ml/kg/hour). T. Carbon dioxide release (ml/kg/hour). U. Respiratory exchange ratio (RER). V. Energy expenditure (kcal/hour). W. Locomotor activity (counts). Shaded bars refer to daytime; solid bars refer to nighttime. Data represent mean ± SEM. *p ≤ 0.05, **p ≤ 0.001 (n=4–5 per group).

Pancreas-projecting vagal sensory Fgf3 modulates insulin secretion:

To determine whether the impact of vagal sensory-derived Fgf3 on pancreatic β-cell activity arises from those directly projecting into the pancreas, we used retrogradely trafficked adeno-associated viruses (AAVrg) that allow for gene delivery in projection neurons ⁵⁴. To validate the efficacy of the AAVrg serotype to transduce the vagal afferents innervating the pancreas, we injected AAVrg particles expressing GFP (3.10¹⁰ genome copies) via the intraductal route in the pancreas of six-week-old C57BL/6J male mice. Four weeks post-transduction, JNG were collected and immunostained for the pan-neuronal marker PGP9.5. In agreement with recent studies ^34,38, we demonstrated that 5–10% of PGP9.5-labelled vagal afferents express GFP, indicating that nearly ~ up to 1 out of 10 vagal afferents projects directly in the pancreas. Interestingly, these proportions were observed in the left (LNG, Figures S2A–C and S2G) and right (RNG, Figures S2D–F and S2G) nodose ganglia, thus ruling out left-right asymmetry ^53,55 in the origin of the vagal sensory innervation of the pancreas. In parallel studies, we examined pancreatic vagal afferent projections in the brain. Consistent with the pseudounipolar nature of the vagal afferents, we identified GFP signals in the vagus nerve terminal endings in the brain stem regions near the area postrema (AP) and the nucleus tractus solitarius (NTS) (Figure S3A–F). However, GFP signals were localized in axonal tract endings (Figure S3D’–F’), not in the soma, thus excluding potential trans-synaptic transmission ⁵⁶. Unsurprisingly, there were no GFP⁺ cells in the dorsal motor of the vagus (DMV) (Figure S3G–L) nor in deeper structures such as the paraventricular nucleus of the hypothalamus (PVN) (Figure S3M–O). To examine the metabolic effects of Fgf3 overexpression in pancreas-innervating vagal sensory neurons, we injected AAVrg particles expressing GFP alone (AAVrg-GFP) or FGF3 and GFP (AAVrg-FGF3) via the intraductal route into the pancreas (3.10¹⁰ genome copies) of six-week-old C57BL/6J male mice and characterized the overall metabolic health of the resultant mice (Figure 3A). Mice injected with AAVrg expressing FGF3 and GFP displayed fluorescence signals in the vagal ganglia (Figure S4A) but not in islet endocrine cells nor in pancreatic peri-islet space (Figure S4B–F). In line with the observations from the anterograde-transduction experiments, AAVrg-FGF3 mice exhibited increased glucose clearance (Figure 3B), enhanced GSIS (Figure 3C) and no change in insulin sensitivity (Figure 3D) as compared to age-matched AAVrg-GFP. Interestingly, AAVrg-FGF3 mice had normal RER (Figure 3E), unaltered energy expenditure (Figure 3F) and unaffected physical activity (Figure 3G), hence suggesting that the Fgf3-induced glucose-lowering effects are secondary to enhanced pancreatic β-cell activity. To demonstrate whether FGF3 acts directly on pancreatic islets, we incubated mouse pancreatic islets ex-vivo for twenty-four hours in presence of 1 nM FGF3 with or without pan-FGFR inhibitors (10 nM LY2874455 or 10 nM TAS120) and measured basal (3.9 mM glucose) and stimulated (11.1 mM glucose) insulin secretions. Using insulin immunoassays, we demonstrated that chronic incubation with recombinant FGF3 did not affect basal insulin release. As expected, insulin secretion was increased in control (vehicle) settings under stimulatory conditions. Remarkably, islets treated with FGF3 displayed superior GSIS compared to control islets under similar glucose conditions. The insulinotropic action of FGF3 was largely reduced when islets were co-incubated with pan-FGFR inhibitors LY2874455 or TAS120 (Figure S5). Together, these data indicate that FGF3 acts directly on pancreatic islets to enhance GSIS through FGFR signaling. Next, we sought to evaluate the physiological relevance of Fgf3 derived from pancreas-projecting vagal sensory neurons in β-cell activity in normal and insulin-resistant states. To this end, six-week-old mice carrying homozygous floxed Fgf3 alleles (Fgf3^fl/fl) were injected via the pancreas intraductal route with AAVrg (3.10¹⁰ genome copies) expressing GFP or Cre recombinase (Figure 3H). Two weeks after transduction, control (AAVrg-GFP) and experimental (AAVrg-Cre) groups were fed a regular chow diet or a high-fat diet and characterized for aspects of glucose and insulin homeostasis. We did not observe significant metabolic differences between AAVrg-GFP and AAVrg-Cre under regular chow diet conditions: glucose clearance (Figure 3I) and glucose-induced insulin release (Figure 3J) were not distinguishable. However, in high-fat diet conditions, mice lacking Fgf3 in pancreas-projecting afferents exhibited impaired glucose tolerance (Figure 3K). Insulin sensitivity (Figure 3L) and body weight (Figure 3M) did not differ between AAVrg-GFP and AAVrg-Cre mice. GSIS was blunted in AAVrg-Cre relative to control AAVrg-GFP mice (Figure 3N). Collectively, these studies suggest that pancreas-specific vagal sensory Fgf3 is necessary to regulate insulin secretion and glucose clearance during the adaptive metabolic response to insulin resistance.

Figure 3: Pancreas-projecting vagal sensory Fgf3 modulates insulin secretion.

A. Schematic of the experimental strategy for the overexpression of Fgf3 in pancreas-selective vagal afferents. B. Glucose tolerance test. C. Glucose-stimulated insulin secretion. D. Insulin tolerance test. E. Respiratory exchange ratio (RER). F. Energy expenditure (kcal/hour). G. Locomotor activity (counts). H. Schematic of the experimental strategy for the ablation of Fgf3 in pancreas-selective vagal afferents. I. Glucose tolerance test in mice fed with chow diet. J. Glucose-stimulated insulin secretion in mice fed with chow diet. K. Glucose tolerance test in mice fed with high-fat diet. L. Insulin tolerance test in mice fed with high-fat diet. M. Body weight in mice fed with high-fat diet. N. Glucose-stimulated insulin secretion in mice fed with high-fat diet. Shaded bars refer to daytime; solid bars refer to nighttime. Data represent mean ± SEM. *p ≤ 0.05, **p ≤ 0.01, (n=4–5 per group).

Vagal afferent signaling regulates insulin secretion:

Numerous studies reported the role of the vagal efferent neurons in the regulation of blood glucose and β-cell activity ^57–61. However, the role of afferent signaling in these processes remains elusive. We set out to investigate whether selective electrical stimulation of the vagus nerve afferents can modulate glucose-induced insulin release and glucose clearance. Briefly, six-to-eight-week-old C57BL/6J male mice were subjected to bilateral subdiaphragmatic vagotomy ⁶². Twenty-four hours post-surgery, vagotomized mice were fasted for 16 hours and subjected to afferent vagal nerve stimulation (aVNS) on the proximal tip of the sectioned vagal trunk (5V, 50Hz, 100 ms, 15 min), before they were injected with glucose (2g/kg body weight; ip). Control mice (Sham) received the same procedures but using a non-electrical wood “toothpick” instead of the electrodes (Figure 4A). Electrostimulation of the proximal tip of the sectioned right vagus nerve improved glucose clearance (Figure 4B) and enhanced insulin release (Figure 4C). To rule out possible confounding metabolic effects of supradiaphragmatic organs (e.g., lung; ⁶³) innervated by non-severed afferents and preganglionic efferents, we conducted bilateral cervical vagotomy. Six-to-eight-week-old C57BL/6J male mice were subjected to bilateral cervical neurectomy sectioning the two vagal trunks and thereby all potential direct efferent signals derived from the DMV preganglionic neurons ⁶⁴. To circumvent the commonly encountered respiratory failure after bilateral cervical vagotomy, we performed a tracheotomy before neurectomy, which allowed for protracting survival for 24 hours ⁶⁵. We then performed a specific afferent vagal stimulation by stimulating the proximal tip of the right cervical vagal trunk which can only trigger an afferent signal toward the brain (Figure 4A). Using a low-pass frequency filter at 25Hz, we demonstrated that electrocardiogram (ECG) recording is normal (Figure 4D) with a transient decrease in heart rate during aVNS from an average of 441 beats per minutes (bpm) before stimulation to an average of 337 bpm during stimulation (Figure 4E), thus validating the afferent vagal nerve stimulation (aVNS) paradigm. Fasting glycemia and insulinemia were comparable between Sham and aVNS groups prior to glucose administration. Remarkably, aVNS—shortly before glucose administration—rapidly and substantially lowered glucose compared to Sham mice (Figure 4F). The glucose-lowering effects of aVNS were recapitulated (Glycemia [mg/dl] at 15 min after glucose injection for Sham: 447.7 ± 13.2 vs. 50Hz: 270.7 ± 16.0 vs. 1Hz: 278.6 ± 24.9. n=3–4, p<0.01, One-way ANOVA vs. Sham) at a slower stimulation frequency (1Hz). Consistent with the ameliorated glucose clearance, aVNS increased circulating insulin levels compared to Sham treatment (Figure 4G). Together, these results demonstrate that specific afferent vagal stimulation enhances glucose-induced insulin release and improves glucose tolerance independently of the efferent outputs originating from the DMV.

Figure 4: Vagal afferent signaling regulates insulin secretion.

Six-to-eight-week-old C57BL/6J male mice underwent subdiaphragmatic vagotomy. Twenty-four hours post-surgery, mice were fasted for 16 hours and subjected to afferent vagal nerve stimulation (aVNS: 5 V, 50 Hz, 100 milliseconds pulse width) or control treatment (Sham) for 15 min before they were injected with glucose (2g/kg body weight). Blood was collected at 0, 5, 15, 30 and 60 minutes after glucose administration to evaluate glucose excursion and assess insulin levels. A. Experimental schematic showing stimulation of the proximal tip of the right sectioned vagal trunk. B. Glucose tolerance test. C. Glucose-stimulated insulin secretion. Data represent mean ± SEM. *p ≤ 0.05, (n=3 per group). Six-to-eight-week-old C57BL/6J male mice underwent cervical vagotomy, fasted for 16 hours and subjected to afferent vagal nerve stimulation (aVNS: 5 V, 50 Hz, 100 milliseconds pulse width) or control treatment (Sham) for 15 min before they were injected with glucose (2g/kg body weight). D. Electrocardiogram (ECG) recording. E. ECG recorded before, during and after aVNS. F. Glucose tolerance test. G. Glucose-stimulated insulin secretion. Data represent mean ± SEM. *p ≤ 0.05, (n=4 per group).

Discussion:

The vagus nerve regulates several facets of energy homeostasis, including food intake ⁶⁶, hepatic glucose production ⁶⁷, gastric function ⁶⁸ and pancreatic secretions ⁴⁷. It is conventionally accepted that the metabolic actions mediated by the vagus nerve are carried out by the preganglionic efferent neurons emanating from the DMV ⁴⁹. Nevertheless, recent studies highlighted the presence of neurotransmitters and neuropeptides in the nodose ganglia, thus suggesting overlooked efferent-like functions mediated by the vagal afferents ^10,69. Our studies unraveled Fgf3 as a vagal-sensory derived factor that regulates islet β-cell function, substantiating the molecular basis of potential feedback mechanisms moderating the vagal-sensory-islet intercommunication ^38,40. These findings highlight an insulinotropic role of the vagal afferents and set forth a starting point for identifying additional vagal sensory-derived signaling molecules that control islet β-cell function. Mapping the islet-specific sensory nerve subsets within the vagal ganglia will uncover an untapped wealth of neurochemical information that can tie islet cell-type specific secretions to selective subsets of vagal sensory neurons ⁷.

A salient observation in our studies is the dual opposing actions of vagal sensory-derived Fgf3 on insulin secretion and energy expenditure. Insulin and energy expenditure/physical activity are two antagonistic metabolic regulatory mechanisms; insulin is an anabolic hormone while physical exercise and other forms of energy expenditure serve catabolic functions ⁷⁰. A plausible explanation is that there are at least two Fgf3-expressing vagal sensory subtypes in the jugular-nodose ganglia: a subset of pancreas-projecting vagal afferents that modulate glucose-induced insulin release via potentially efferent-like mechanisms and another afferent subtype that regulates energy expenditure and physical activity through transmission in the brainstem. In support of this hypothesis, Fgf3-elicited insulinotropic properties were preserved when Fgf3 was solely overexpressed in pancreas-selective vagal afferents while the central effects on energy expenditure were largely reduced. Moreover, the stipulation that Fgf3 acts centrally to regulate energy expenditure is consistent with the mode of action of the previously metabolic hormone-like Fgfs ^71–73. Furthermore, our ex vivo assays demonstrated that Fgf3 enhances GSIS in pancreatic islets in an FGFR-dependent manner, thus suggesting a potential direct peripheral action of the vagal sensory-derived Fgf3 on pancreatic β cells in the form of an axon reflex.

An intriguing biological feature unveiled by the electrical stimulation studies is the notion that vagal afferent signaling can enhance insulin secretion in the absence of an intact vago-vagal reflex. Traditionally, the vagal control of insulin secretion is believed to hinge on the activity of a bidirectional neural system where the ascending vagal afferents detect local variations in islet hormone levels ^38–40 and transfer these signals to the NTS/DMV where signaling molecules are instructed to be released the periphery through descending vagal efferents to regulate islet secretions ^57–61. How does the ascending vagal afferent signaling toward the brain communicate insulinotropic impulses to pancreatic β cells when the DMV efferent pathways are surgically severed? Anatomically speaking, one could envisage a vago-sympathetic reflex where ascending vagal signals are interpreted in the brainstem and conveyed to the peripheral tissues via the descending spinal efferent neurons ⁷⁴. It is also plausible that vagal afferent signaling stimulates the release of CNS-derived neuroendocrine molecules such as pituitary hormones ⁷⁵ that act systemically to regulate peripheral functions. Further experimentation is warranted to delineate the contribution(s) of these alternative pathways in the control of islet β-cell activity via vagal ascending signaling.

In summary, we identify a unique role for vagal afferent signaling in the modulation of islet β-cell function and unravel Fgf3 as a vagal-sensory derived metabolic factor that regulates insulin secretion and glucose homeostasis. These findings provide insights into the autonomic control of pancreatic β-cell function and lay ground for further investigations to identify elements of the vagal sensory signaling that can be leveraged to normalize β-cell activity in individuals with compromised insulin secretion.

Limitations of the study:

We do not know which FGFR mediates the insulinotropic action of Fgf3. Fgf3 binds to Fgfr1b and Fgfr2b ⁷⁶, which are known to be expressed in islet β cells in the adult mouse pancreas ⁷⁷ and genome-wide association studies (GWAS) have linked genetic variants of Fgfr1 and Fgfr2 to the development of type 2 diabetes ^78,79. Further studies are warranted to identify the Fgfr(s) and the specific downstream signaling pathways implicated in the Fgf3-mediated regulation of GSIS. One caveat of this study is the absence of dual viral strategies to target the pancreas projecting vagal afferents independently of other potentially confounding neuronal pathways (e.g., celiac ganglion). Nevertheless, the combinatory use of retrograde and anterograde transduction approaches and gain- and loss-of-function models indicate that vagal sensory-derived Fgf3 molecules contribute to the modulation of insulin secretion. Our studies do not exclude a vago-vagal reflex wherein vagal sensory-derived FGF3 molecules are released in the hindbrain and signal through the NTS-DMV circuits to convey insulinotropic impulses through the pancreas-projecting DMV preganglionic nerves. Although the electrostimulation experiments unequivocally suggest that insulin secretion can be regulated by the vagal afferents independently of the vago-vagal reflex, the physiological contribution of the efferent pathways, the vago-spinal reflex and the CNS-derived neuroendocrine hormones has not been addressed. Additional studies are needed to tease apart the proportional significance of these pathways in regulating islet β-cell activity.

STAR Methods

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact: Abdelfattah El Ouaamari (aelouaam@NYMC.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

The RNA-seq data (raw and processed files) have been deposited in NCBI GEO under the accession number GSE239596. This paper does not report any original code. The scripts used in this study are available at GitHub repository (https://github.com/Jerry-Zhao/NG2023). Microscopy images reported in this paper will be made available by the lead contact upon request. Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request.

Experimental model and study participant details

Animals:

All studies were performed on C57BL/6J male mice (Strain # 000664, The Jackson Laboratory) unless indicated otherwise. The Fgf3 mutant (B6;129-Fgf3tm1.2Sms/J; Strain # 023900) mice were obtained from the Jackson Laboratory. Mice were housed in pathogen-free facilities and maintained on a 12-hour light/dark cycle in the Animal Care Facility at the Child Health Institute of New Jersey and New York Medical College. All studies and protocols were approved by the Rutgers University and New York Medical College Animal Care and Use Committees and were in accordance with the National Institutes of Health guidelines. Blood glucose was monitored using an automated glucose monitor and plasma insulin was detected by ELISA. Animals were fasted overnight for 16 hours for glucose tolerance tests and glucose-stimulated insulin secretion assays and 6 hours for insulin tolerance tests.

S961 and high-fat diet studies:

C57BL/6J male mice were anesthetized ketamine (100 mg/kg BW) and xylazine (10 mg/kg BW) and infused, using osmotic pumps (ALZET) implanted subcutaneously, with PBS alone or PBS with the insulin receptor antagonist S961 at the dose of 10 nmoles/week for two weeks. For high-fat diet experiments, mice were challenged with a diet containing 60% kcal fat (catalog# D12492, Research Diet) for twenty-four weeks.

Method details

RNA sequencing:

RNA samples were quantified and assessed for quality using Agilent BioAnalyzer (Agilent Technologies, Palo Alto, CA, USA). The RIN values were 7.95 +/− 0.86 [6.6 to 9.1] and a median concentration of 48.5 ng/μL but with a very wide range [1.8 to 266.6] ng/μL. Illumina sequencing libraries were prepared using the Clontech/Takara Pico-Input Strand-Specific Total RNA-Seq for Illumina (Takara, San Jose, CA, USA) and barcodes for sequencing together as a pool. To compensate for the wide range of concentrations the same amount of RNA was used for each sample, roughly 6 ng of total RNA. The libraries were assessed for quality using a TapeStation (Agilent Technologies, Palo Alto, CA, USA) as well as the Kapa library quantification kit (KAPA Biosystems, Wilmington, MA, USA) The libraries had a median size of 407 +/− 13 bp. Samples were pooled in equimolar amounts and sequenced on an Illumina HiSeq 2000 in two separate lanes at 101 bp with 8×8bp barcodes. FASTQ files were produced using Illumina bcl2fastq program with a read length of 100bp.

RNA sequencing data analysis:

Image analysis and base calling were performed by the HiSeq Control Software (HCS). Raw sequence data (.bcl files) was derived by Illumina HiSeq 2000 and was converted into the FASTQ files and de-multiplexed using Illumina bcl2fastq 2.20 software. The RNA-seq data analyses were performed as previously described ^80,81. Briefly, the FASTQ files were aligned to mouse mm10 genome using STAR 2.7.7a ^80,82 with the following parameters: ‘–runThreadN 40 –outFilterMultimapNmax 1 –outFilterMismatchNmax 3 – outFilterScoreMinOverLread 0.25 –outFilterMatchNminOverLread 0.25’. The total number of mapped reads and the number of reads aligned to the exons of each gene were obtained with a Perl script. Subsequently, the table of raw read counts underwent normalization, and the top 1000 variable genes were selected for the principal component analysis using DESeq2 ⁸³. To compare gene expression levels between the groups, the DESeq2 and the Wald significance tests were utilized. The criteria for differentially expressed genes (DEGs) was false discovery rate (FDR) < 0.05. To carry out the functional enrichment analysis, we used the g:Profiler (version_ e109_eg56_p17_1d3191d) ⁸⁴ and Gene Set Enrichment Analysis (GSEA) software (version_4.3.2) ⁸⁵. Briefly, the list of DEGs were used as input data for g:Profiler to identify enriched Gene Ontology (GO), Reactome, and KEGG pathway terms. To determine the statistical significance, the Benjamini-Hochberg FDR method with a threshold of 0.05 was employed. The gene ratio, the intersection size divided by the query size, in conjunction with the intersection size and adjusted P-value, were used as inputs to generate the dot plot to visualize the results of GO enrichment analysis. The Counts Per Million (CPM) values were obtained from DESeq2 and were used to perform the GSEA analysis. The GSEA analysis was derived by using the msigdb.v2022.1.Mm.symbols.gmt gene sets database and the Mouse-Gene_Symbol_Remapping_MsigDB.v2022.1.Mm.Chip platform. GSEA was executed with the following parameters: ‘1000 permutations’, ‘collapse’ option, and ‘gene_set’ as the permutation type.

Intravenous tail-vein injection:

Mice were warmed under a heat lamp for 5–10 min for tail vein vasodilation and placed into a restrainer (Mouse Tail Illuminator Restrainer, Braintree Scientific). The skin at the injection site was wiped with a gauze sponge moistened with 70% ethanol. Mice were injected (28G needle), each, in the lateral tail vein with a total volume of 100 μL of phosphate-buffered saline (PBS) containing 10¹¹ genome copies of AAVs expressing GFP or FGF3. Mice were removed from the restrainer after injection and allowed to recover in their home cage.

Nodose ganglia injection:

Mice were anesthetized with ketamine (100 mg/kg BW) and xylazine (10 mg/kg BW) and received a subcutaneous injection of buprenorphine (0.05 mg/kg BW). Complete anesthesia was assessed by pedal reflex. A longitudinal incision of 1 cm was made along the ventral neck and the left jugular-nodose ganglion was exposed. 150 nL of PBS solution containing AAVs expressing GFP or FGF3 (titer ~10¹³ genome copies/ml) and 0.05% Fast Green FCF Dye (Sigma) with a pulled glass pipette (Sutter Instruments) using a Nanoinject III Injector (Drummond). Successful injections were validated by the presence of Fast Green FCF throughout the ganglion. The incision site was closed using simple interrupted stitches with Polyamide Monofilament (B. Braun). Metabolic phenotyping was conducted four weeks after injection.

Pancreas intraductal injection:

Mice were anesthetized with ketamine (100 mg/kg BW) and xylazine (10 mg/kg BW) and received a subcutaneous injection of buprenorphine (0.05 mg/kg BW). Complete anesthesia was assessed by pedal reflex. A midline incision of 2 cm was made along the abdomen, the duodenum was exposed and the common bile duct was clamped caudal to the liver side. The syringe (28G1/2 needle) was entered into the pancreatic duct from the duodenum through the ampulla of Vater and advanced in a retrograde manner toward the gallbladder. A volume of 100 μL of AAV retrograde particles expressing GFP, FGF3 or Cre recombinase (3.10¹⁰ genome copies) was slowly administered over one minute. After clamp removal, the wound on the ampulla was sealed with 3M Vetbond Tissue Adhesive (3M Animal Care) and the abdominal muscle layer was closed using simple interrupted stitches with Polyamide Monofilament (B. Braun), and the overlying skin was sutured using 9 mm autoclips (Braintree Scientific).

Islet isolation and glucose-stimulated insulin secretion:

Pancreatic islets were isolated from 12-week-old C57BL/6J male mice using intraductal collagenase technique ⁵¹. Islets were handpicked and cultured overnight in RPMI 1640 medium containing 5.6 mM glucose and 10% fetal bovine serum (FBS). Ten size-matched islets were incubated for 24 hours in RPMI 1640 medium containing 5.6 mM glucose with or without 1 nM recombinant human FGF3 protein, 10 μg/ml heparan sulfate, 10 nM Futibatinib (TAS120) or 10 nM LY2874455 as indicated. For glucose-stimulated insulin secretion assays, islets—in the presence (or absence) of the same compounds as above—were incubated in Krebs-Ringer bicarbonate (KRB) solution containing 3.9 mM glucose for 60 min, then transferred to KRB solution containing 11.1 mM glucose for 60 min. Insulin concentration in the buffer and insulin content in islets were measured using an insulin ELISA kit. Basal and stimulated insulin secretions were normalized to total islet insulin content.

Afferent vagal nerve stimulation:

C57BL/6J male mice were anesthetized (isoflurane 1–2%) and subjected to bilateral cervical vagotomy as previously reported ⁶⁴. Prior to neurectomy, mice underwent tracheotomy to bypass the severe respiratory insufficiency induced by the bilateral cervical vagotomy ⁶⁵. Mice were fasted for 16 hours and glycemia and insulinemia were measured (t=0) before electrical stimulation. We then performed a specific afferent vagal stimulation (aVNS) by stimulating (5V, 50Hz, 100 ms, 15 min) the proximal tip of the right vagal trunk which can only propagate afferent signal toward the brain stem. Control mice (Sham) received the same procedures but using a non-electrical wood “toothpick” instead of the stimulation electrodes ⁶². Some experiments were performed at a slow stimulation frequency (1Hz). Immediately after aVNS (or Sham) treatment, mice received a glucose challenge (2g/kg body weight; ip) and tail nick blood collection was performed to measure glucose and insulin levels at the indicated time points (5, 15, 30, 60, and 120 minutes) post glucose administration.

Electrocardiogram (ECG) recording:

Mouse ECG lead II was recorded with ECG100c module attached to STM100c MP150 Biopac research system (MP150, Biopac Systems, Goleta, CA) as we previously reported ⁸⁶. Unipolar stainless-steel electrodes (Biopac Systems, Goleta, CA) were inserted in the right and left forelimbs as well as in the right hindlimbs (as a ground electrode). The data acquisition was done using Acqknowledge 3.9 software (Biopac Systems, Goleta, CA). The acquisition rate was set to 2000 samples/second with a frequency-band filter of 25–1000 Hz. The high-pass frequency was not altered and was set to the acquisition rate default. Afferent vagus nerve stimulation was performed using platinum-iridium bipolar electrodes (Microprobes, Gaithersburg, MD) to deliver an alternating (1.5-second on-and-off cycle) current (5V, 1 or 50Hz, 100 milliseconds pulse width) for 15 minutes.

Real-Time Quantitative Reverse Transcription Polymerase Chain Reaction (qRT-PCR):

cDNA synthesis was carried out by using High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, Cat# 4368814) with total RNA samples. The cycling parameters for qRT-PCR amplification reactions were: AmpliTaq activation at 95°C for 10min, denaturation at 95°C for 15s, and annealing/extension at 60°C for 1min (40 cycles). The Ct value of FGF3 was normalized by the Ct value of internal reference (β-actin) and the relative expression values were calculated per the 2−ΔΔCt method. Primers for FGF3: 5′-TCCACAAACTCACACTCTGC-3′ (forward) and 5′-GAACAGCGCCTATAGCATCC-3′ (reverse) and for β-actin: 5’-CCAGTTGGTAACAATGCCATG-3’ (forward) and 5’-GGCTGTATTCCCCTCCATCG-3’ (reverse).

Pancreas immunostaining and islet morphometric studies:

Pancreas sections were analyzed by immunostaining using anti-Ki67 (BD) and anti-insulin (Abcam) antibodies and nuclear DAPI staining. Ki67+ β cells were examined by immunofluorescence microscopy (BX53, Olympus). Insulin+ cells showing nuclear DAPI staining were considered as β cells. Insulin+ cells displaying nuclear colocalization of DAPI and Ki67 staining were counted as Ki67+ β cells. The islet β-cell area was calculated by dividing the area of insulin-positive areas over the total pancreas area. The islet β-cell mass was determined by multiplying the percentage of islet β-cell area by the pancreas weight. Cell counting and area estimation was performed using ImageJ software ⁸⁷.

Quantification and statistical analysis

All data are presented as mean ± SEM. Data were analyzed using Student’s ‘t’ test or two-way ANOVA and a “p” value of less than 0.05 was considered statistically significant.

Supplementary Material

2

Table S1: Statistics of sequencing and alignment results of the RNA-Seq data (related to Figure 1).

3

Table S2: List of differentially expressed genes (DEGs) (related to Figure 1).

4

Table S3: Gene Ontology (GO) enrichment analysis (related to Figure 1).

5

Table S4: Gene set enrichment analysis (GSEA) (related to Figure 1).

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Purified Mouse Anti-Ki-67	BD Pharmingen	Cat# 556003
Guinea Pig Anti-insulin	Bio-Rad	Cat# 5330–0104G
Rabbit Anti-PGP9.5	Abcam	Cat# ab108986
Chicken Anti-GFP	enQuire Bioreagents	Cat# CGFP-45A
Goat Anti-Choline Acetyltransferase	Sigma-Aldrich	Cat# AB144P
Alexa Fluor® 594 AffiniPure Donkey Anti-Guinea Pig IgG (H+L)	Jackson ImmunoResearch	Cat# 706–585-148
Alexa Fluor® 488 AffiniPure Donkey Anti-Mouse IgG (H+L)	Jackson ImmunoResearch	Cat# 715–545-150
Alexa Fluor® 594 AffiniPure Donkey Anti-Rabbit IgG (H+L)	Jackson ImmunoResearch	Cat# 711–585-152
Donkey anti-Chicken IgY (H+L) Alexa Fluor™ 488	Thermo Fisher Scientific	Cat# A78948
Donkey anti-Goat IgG (H+L) Alexa Fluor™ 555	Thermo Fisher Scientific	Cat# A-21432
Bacterial and virus strains
AAV9-EF1a-6xMyc-GFP	Vigene Biosciences	N/A
AAV9-EF1a-mFGF3–6xMyc-IRES-GFP	Vigene Biosciences	N/A
AAV2/retro-CAG-GFP-WPRE	Boston Children’s Hospital Viral Core	N/A
AAV2/retro-CAG-Cre-WPRE	Boston Children’s Hospital Viral Core	N/A
AAV2/retro-miniCAG-6xmyc-GFP	Boston Children’s Hospital Viral Core	N/A
AAV2/retro-miniCAG-mFGF3-IRES-GFP	Boston Children’s Hospital Viral Core	N/A
Biological samples
Dextrose 50% IV Injection Solution	Henry Schein	Cat# 2486614
Humulin R Insulin 100u/mL	Henry Schein	Cat# 1238578
Rodent Diet With 60 kcal% Fat (Formula D12492)	Research Diets	Cat# D12492
Chemicals, peptides, and recombinant proteins
S961 peptide (gift from Dr. Lauge Schäffer)	Novo Nordisk	Cat# NNC0069–0961
Recombinant human FGF3 protein	R&D	Cat# 1206-F3
Heparan sulfate	Selleckchem	Cat# S5992
Futibatinib (TAS120)	MedChemExpress	Cat# HY-100818
LY2874455	Selleckchem	Cat# S7057
Critical commercial assays
Bayer Contour Blood Glucose Monitoring System	Amazon	Cat# B000REKQ80
Ultra Sensitive Mouse Insulin ELISA Kit	Morinaga	Cat# M1108
iTaq™ Universal SYBR® Green Supermix	Bio-Rad	Cat# 1725124
Deposited data
Raw and analyzed data	This paper	GEO: GSE239596
Computational code and scripts	This paper	github.com/Jerry-Zhao/NG2023
Experimental models: Organisms/strains
C57BL/6J mice	Jackson Laboratory	Stock# 000664
B6;129-Fgf3tm1.2Sms/J mice	Jackson Laboratory	Stock# 023900
Oligonucleotides
FGF3 Primer: 5′-TCCACAAACTCACACTCTGC-3′ (forward)	Integrated DNA Technologies	N/A
FGF3 Primer: 5′-GAACAGCGCCTATAGCATCC-3′ (reverse)	Integrated DNA Technologies	N/A
β-actin Primer: 5’-CCAGTTGGTAACAATGCCATG-3’ (forward)	Integrated DNA Technologies	N/A
β-actin Primer: 5’-GGCTGTATTCCCCTCCATCG-3’ (reverse)	Integrated DNA Technologies	N/A
Recombinant DNA
pA-6xMyc-GFP (gift from Dr. Thomas Schimmang)	University of Valladolid Spain	N/A
pA-6xMyc-mFGF3 (gift from Dr. Thomas Schimmang)	University of Valladolid Spain	N/A
Software and algorithms
STAR	Dobin et al. 82	https://github.com/alexdobin/STAR
R	N/A	https://www.r-project.org/
DESeq2	Love et al. 83	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
g:Profiler	Kolberg et al. 84	https://biit.cs.ut.ee/gprofiler/gost
GSEA	Subramanian et al. 85	https://www.gsea-msigdb.org/gsea/index.jsp
pheatmap	N/A	https://cran.r-project.org/web/packages/pheatmap/index.html
ImageJ	Schneider et al. 87	https://imagej.nih.gov/ij/

Highlights.

• Vagal sensory-derived Fgf3 is elevated in acute insulin resistance

• Systemic and vagal sensory-derived Fgf3 promotes insulin secretion

• Pancreas-projecting vagal sensory Fgf3 controls insulin release

• Electrical stimulation of vagal afferents enhances insulin secretion