Hypothalamic Irak4 is a genetically controlled regulator of hypoglycemia-induced glucagon secretion

Abstract

Objectives

Glucagon secretion to stimulate hepatic glucose production is the first line of defense against hypoglycemia. This response is triggered by so far incompletely characterized central hypoglycemia-sensing mechanisms, which control autonomous nervous activity and hormone secretion. The objective of this study was to identify novel hypothalamic genes controlling insulin-induced glucagon secretion.

Methods

To obtain new information on the mechanisms of hypothalamic hypoglycemia sensing, we combined genetic and transcriptomic analysis of glucagon response to insulin-induced hypoglycemia in a panel of BXD recombinant inbred mice.

Results

We identified two QTLs on chromosome 8 and chromosome 15. We further investigated the role of Irak4 and Cpne8, both located in the QTL on chromosome 15, in C57BL/6J and DBA/2J mice, the BXD mouse parental strains. We found that the poor glucagon response of DBA/2J mice was associated with higher hypothalamic expression of Irak4, which encodes a kinase acting downstream of the interleukin-1 receptor (Il-1R), and of Il-ß when compared with C57BL/6J mice. We showed that intracerebroventricular administration of an Il-1R antagonist in DBA/2J mice restored insulin-induced glucagon secretion; this was associated with increased c-fos expression in the arcuate and paraventricular nuclei of the hypothalamus and with higher activation of both branches of the autonomous nervous system. Whole body inactivation of Cpne8, which encodes a Ca⁺⁺-dependent regulator of membrane trafficking and exocytosis, however, had no impact on insulin-induced glucagon secretion.

Conclusions

Collectively, our data identify Irak4 as a genetically controlled regulator of hypoglycemia-activated hypothalamic neurons and glucagon secretion.

Highlights

• •A genetic screening in mice for hypoglycemia-induced glucagon secretion yielded two QTLs.
• •Hypothalamic Irak4 was identified as a candidate regulator of glucagon secretion.
• •High arcuate Irak4 expression in DBA/2J mice correlated with low glucagon secretion.
• •Inhibition of central Il-1β/Irak4 signaling in DAB/2J mice restored glucagon secretion.

1. Introduction

The brain mostly relies on glucose as a source of metabolic energy. Hence, homeostatic glucoregulatory mechanisms are needed to maintain blood glucose at the minimum level of ∼5 mM to ensure sufficient glucose availability to the brain [1]. Hypoglycemia-sensing cells located in the central nervous system and in the periphery, e.g., in the hepatoportal vein and in the carotid bodies, initiate a counterregulatory response (CRR) when the blood glucose level falls below the euglycemic level [[2], [3], [4]]. CRR involves the activation of the autonomous nervous system and the hypothalamo-pituitary-adrenal axis by the brain, leading to the secretion of glucagon by pancreatic alpha cells and of catecholamines and glucocorticoids by the adrenal glands [4]. This promotes hepatic neoglucogenesis, stimulates lipolysis, and inhibits insulin secretion as well as glucose uptake by muscle and fat. This coordinated response restores euglycemia to preserve sufficient glucose provision to the brain. While the CRR prevents hypoglycemia in healthy individuals, this response becomes progressively blunted in type 1 and insulin-treated type 2 diabetic patients, leading to recurrent hypoglycemic episodes of increasing severity, a condition known as hypoglycemia-associated autonomic failure (HAAF), which represents a major limitation in the insulin-based therapy of diabetes [5].

In the brain, glucose sensing neurons have been identified in the hypothalamus and in the brainstem [6,7], and they are commonly classified in two categories depending on their activation by a rise (glucose excited, GE) or a fall in blood glucose concentration (glucose inhibited, GI) [[7], [8], [9]]. In the brainstem, glucose-sensing neurons are found in the dorsal vagal complex (DVC) composed of the area postrema (AP), the nucleus of the solitary tract (NTS) and the dorsal motor nucleus of the vagus (DMNX) [6]. In the hypothalamus, glucose-sensing neurons are distributed in the arcuate (ARH), the paraventricular (PVN), the lateral (LH), the dorsomedial (DMH), and the ventromedial hypothalamic nuclei (VMN) [6]. Among these nuclei, the VMN has been extensively studied for its implication in glucose sensing and in the control of glucagon secretion [[10], [11], [12], [13]]. Glucose sensing by GE neurons has been proposed to depend on a Glut/glucokinase/K_ATP channel signaling pathway similar to that controlling glucose-stimulated insulin secretion in pancreatic β cells [[14], [15], [16], [17], [18]]. However, glucose sensing by GE neurons of the VMN is not suppressed when the glucokinase gene is inactivated [19] and the Na⁺/glucose co-transporters SGLT1 and SGLT3 are required for glucose sensing by specific populations of GE neurons [20]. Activation of GI neurons firing by hypoglycemia requires activation of AMP-activated protein kinase [[21], [22], [23]] and the regulated activity of CFTR [24], anoctamine 4 [25], two-pore-domain K⁺ channels [26], or the Na⁺/K⁺ATPase [27,28]. Thus, the mechanisms of neuronal glucose sensing are diverse and are still incompletely understood.

Unbiased identification of genes involved in the CRR and its deregulation is of significant physiological and pathophysiological interest. Such identification is possible by screening genetic reference populations for quantitative trait loci controlling glucagon secretion. The BXD mice consist of a large panel of recombinant inbred lines derived from the cross of C57BL/6J and DBA/2J mice [29]. In a previous study, we screened a panel of 36 BXD mouse lines to identify QTLs controlling neuroglucopenia (2-deoxy-d-glucose, 2DG)-induced glucagon secretion. This led to the identification of a QTL on chromosome 7 and of Fgf15 produced by DMH neurons, as a negative a regulator of glucagon secretion but a positive inducer of hepatic glucose production through direct activation of sympathetic nervous activity [30,31].

Here, we performed a new screening aimed to discover hypothalamic genes controlling glucagon secretion in response to insulin-induced hypoglycemia, as this may be more relevant to the condition of insulin-induced HAAF. This led to the identification of two QTLs, one on chromosome 8 and the second one on chromosome 15. We then searched for candidate genes in these QTL based on the assumption that they were expressed in the hypothalamus and that their level of expression correlated with the glucagon trait. Irak4, encoding a protein kinase acting downstream of the Il-1R and Toll-like receptor (TlR) signaling pathway [32], was found to be the best candidate on chromosome 15. Physiological studies showed that high Irak4 as well as Il-1ß expression in the hypothalamus of DBA/2J mice were responsible for the low hypoglycemia-induced glucagon response observed in these mice.

2. Research design and methods

2.1. Mice

BXD mice (see list of the strains used in Supplemental Table 1) were purchased from the Jackson Laboratory (Bar Harbor, Maine, USA). C57BL/6J and DBA/2J mice were purchased from Charles River Laboratories (Saint Germain Nuelles, France). Mice were housed on a 12-h light/dark cycle and fed a standard rodent chow diet (Diet 3436, Provimi Kliba AG, Kaiseraugst, Switzerland). Experiments were performed with 8- to 12-week-old male mice. All animal experimentations were approved by the Veterinary Office of Canton de Vaud under license agreement VD 3363.

2.2. Biochemical measurements

Blood was collected from the submandibullary (glucagon) or tail veins (glycemia). Glycemia was measured using a glucometer (Ascensia Breeze 2, Bayer Healthcare, Lerverkusen, Germany). ELISA was used to quantify plasma glucagon (cat. number:10-1271-01, Mercodia, Uppsala, Sweden). Pancreatic glucagon content was determined as described previously [31].

2.3. Phenotyping of insulin-induced glucagon secretion in BXD mice and their parental strains

Insulin-induced glucagon secretion was assessed in 5 mice for each of the 36 BXD lines and in the parental C57BL/6J and DBA/2J strains. Mice were handled daily for two weeks before the experiments. Before the first experiment, mice were fasted for 6 h, and glycemia was measured at 2:00 p.m. The mice then received an i.p. injection of NaCl 0.9%. Blood was collected 1 h later for basal glucagon quantification. Mice were then allowed to recover for two weeks with daily handling. The same protocol was used with i.p. injection of insulin (Actrapid, Novo Nordisk Pharma, 0.8 U/kg).

2.4. QTL and eQTL mapping

QTL and eQTL mappings were performed using the R package R/qtl as previously described [31,33]. QTL interval mapping was calculated using the expected-maximization algorithm, a 5% genotyping error rate, and pseudomarkers were generated every cM. QTL location was obtained with 1.5 LOD score (equivalent to 6.915 likelihood ratio statistics (LRS)) support intervals as suggested [34]. Significant QTLs were determined for each trait using 5% false discovery rate threshold estimated from 1000 permutations.

2.5. RNA-Seq analysis

RNA-Seq from pools of 3–6 hypothalami of 12-week-old BXD mice had been previously generated [31]. Read counts were normalized in transcripts per million (TPM), and Pearson's correlations were calculated between gene expression levels and physiological traits. RNA-Seq data are accessible via GEO under accession number GSE87586.

2.6. I.c.v. cannulation

Surgeries were performed under ketamine/xylazine anesthesia. Cannulas were placed in the lateral ventricle (−0.7 mm from the Bregma; −1.3 mm from the midline; −2.0 mm from the surface of the skull) [35]. The animals were allowed to recover for one week before experiment with daily handling and body weight monitoring.

2.7. I.c.v. injections

At 1:00 p.m. on the test day, mice received either an i.c.v. injection of saline or Anakinra (50 μg; Kineret®, recombinant human IL-1ra, Swedish Orphan Biovitrum AB) in a total volume of 5 μL. One hour later, mice were injected i.p. with saline or 0.8 U/kg of insulin. Blood was collected 1 h later for glucagon quantification, and the brains were fixed 2 h later for immunofluorescence detection of c-fos.

2.8. Physiological measurements

Insulin-induced hypoglycemia tests (Actrapid, 0.8 U/kg) were performed in 6-h food-deprived mice injected i.c.v. with NaCl 0.9% or Anakinra 1 h before i.p. injection of NaCl 0.9% or 0.8 U/kg insulin. Glycemia was measured before i.c.v. and i.p. injections and 1 h after i.p. injections. Blood was collected for glucagon measurement 1 h after the i.p. injections. Hyperinsulinemic-hypoglycemic clamps were performed as previously described [36]. Briefly, 6 h-fasted C57BL/6J and DBA/2J mice received at 2:00 p.m. on the test day either i.c.v. NaCl 0.9% or Anakinra (50 μg) at the start of the clamp procedure. After 90 min and after at least 30 min of stable glucose infusion rates and glycemia, blood was sampled to quantify plasma glucagon.

2.9. Immunofluorescence microscopy

C-fos immunodetection was performed in C57BL/6J that received i.p. injection of saline or insulin 0.8 U/kg, and in DBA/2J mice that received an i.c.v. injection of saline or Anakinra 1 h before an i.p. injection of insulin 0.8U/kg. Two hours later, the mice were fixed by cardiac perfusion of 4% cold paraformaldehyde (PFA) in sodium phosphate buffer (0.1 M, pH 7.4). Brains were then dissected and kept for 2 h in PFA at 4 °C, incubated overnight in a 20% sucrose solution at 4 °C, and frozen at −80 °C. Serial 25-μm-thick hypothalamic cryosections were prepared and incubated first for 1 h in 0.1M phosphate buffer pH 7.4 containing 4% normal goat serum and 0.3% Triton X-100 and then for 24 h at room temperature with rabbit monoclonal antibodies against c-fos (cat. number: 2250, 1/1000, Cell Signaling, Danvers, USA) and for 2 h at room temperature with Alexa Fluor 488- or 568-conjugated goat anti-rabbit IgG antibodies (cat. number: 11008, 1/500, Life Technologies, Carlsbad, USA). Nuclei were counterstained with DAPI (cat. number: D9542, Sigma Aldrich, St. Louis, USA), and the slides were mounted in Mowiol (cat. number: 81381, Sigma Aldrich, St. Louis, USA).

Images were acquired with a Zeiss Axio Imager D1 microscope interfaced with Axiovision software equipped with ApoTome.2 and a Camera Axiocam 702 mono (Zeiss, Oberkochen, Germany). The number of c-fos-positive cells in the ARH and PVN was normalized to the respective surface of each nuclei using Image J software.

2.10. In situ hybridization

For in situ hybridization detection of Irak4 based on Atto 550 fluorescence, brains of C57BL/6J and DBA/2J mice were dissected as described above for c-fos immunodetection. Twenty-five micrometer hypothalamic cryosections were then prepared, and in situ hybridization was performed using Advanced Cell Diagnostics probes (cat. number: 444451) and RNAscope Fluorescent Multiplex Detection Reagents (Advanced Cell Diagnostics, Newark, USA) following the manufacturer's instructions. Irak4 mRNA spots were quantified using ImageJ software on 2 hemisections between the Bregma −0.82 and −0.94 mm for the PVN and −1.70 to 1.82 mm for the ARH on 4 to 5 animals per group. The number of spots was then normalized to the respective surface of each nucleus.

2.11. Autonomous nervous system activity recording

Unipolar parasympathetic and sympathetic activities were recorded along the carotid artery as previously described [30,31]. Recordings were performed for 1.5 h under isoflurane anesthesia (30 min during basal condition after i.c.v. NaCl 0.9%/Anakinra and before i.p. insulin 0.8 U/kg, 1 h during insulin-induced hypoglycemia) using the LabChart 8 software (AD Instrument, Oxford, UK). Data were digitized with PowerLab 16/35 (AD Instrument, Oxford, UK). Signals were amplified 10⁵ times and filtered using a 100/1000 Hz band pass filter. Firing rate analysis was performed using the LabChart 8 software.

2.12. Statistical analysis

Data are expressed as mean ± SEM. Statistical analysis was performed using GraphPad Prism® 9.1.0, either by a mixed-effects analysis followed by a Sidak's post hoc test, a repeated-measures two-way ANOVA followed by a Sidak's post hoc test, or by an unpaired two-tailed Student's t-test. P-values of <0.05 were considered to be significant.

3. Results

3.1. Identification of QTLs for insulin-induced glucagon secretion

To search for the loci controlling hypoglycemia-induced glucagon secretion, we performed a genetic screening of a panel of 36 BXD mouse lines and their parental C57BL/6J and DBA/2J strains. In the first experiment, five mice from each line were fasted for 6 h and then injected intraperitoneally (i.p.) with a saline solution. Blood was collected 1 h later for plasma preparation and glucagon measurement. The experiment was repeated two weeks later, but mice received i.p. injections of 0.8 U/kg insulin instead of saline to induce hypoglycemia. Plasma glucagon levels after saline injections were in the same low range for all the mouse lines tested but varied markedly between lines upon insulin-induced hypoglycemia (Figure 1A). The glucagon response was not correlated with the pancreas glucagon content of the BXD lines (Figure 1B); it was, however, correlated with the level of insulin-induced hypoglycemia (Figure 1C) and varied up to 23-fold when compared with BXD 70 (7.2 pM) and BXD 98 (169.2 pM) (Figure 1 D). The plasma glucagon data of Figure 1D were then used for QTL mapping and led to the identification of two genome-wide significant QTLs on chromosome 8 (LRS = 19.23) and on the distal part of chromosome 15 (LRS = 26.27) (Figure 1E). As the glucagon response showed a strong correlation with the hypoglycemia levels, the identified QTLs may also control insulin sensitivity. We thus performed a QTL analysis for insulin-induced hypoglycemia. This led to the identification of a single QTL on chromosome 8 (Figure 1F), located in the same genomic interval as the glucagon QTL. This region may thus influence both insulin sensitivity and hypoglycemia-induced glucagon secretion. This result also indicated that the QTL on chromosome 15 was related to glucagon secretion rather than to insulin sensitivity.

Figure 1.

Identification of QTLs controlling hypoglycemia-induced glucagon secretion. A: Basal and hypoglycemia-induced glucagon secretion in the 36 BXD lines and the parental C57BL/6J and DBA/2J strains. Each circle represents the mean glucagonemia for each strain. B: Scatterplot of the correlation between plasma glucagon levels and total pancreas glucagon content for all the mouse lines tested. C: Scatterplot of the correlation between plasma glucagon levels and glycemia for all the mouse lines tested. D: Plasma glucagon levels 1 h after i.p. insulin injection in the indicated BXD lines and the parental strains. E: Whole genome significant QTLs on chromosome 8 and 15. The red line indicates the whole genome significant threshold (p ≤ 0.05). The blue line indicates the whole genome suggestive threshold (p ≤ 0.63). F: Whole genome significant QTL on chromosome 8 for glycemia 1 h after i.p. insulin injection. The red line indicates the whole genome significant threshold (p ≤ 0.05). The blue line indicates the whole genome suggestive threshold (p ≤ 0.63).

The QTL on chromosome 8 spans ∼6.38 Mb between markers rs13479624 and rs33450716 with a peak LRS on rs13479628 (Figure 2A). This QTL contains 155 genes and explains 41.4% of the variance of the trait. To identify candidate genes in the QTL controlling insulin-induced glucagon secretion, we postulated that they must be expressed in the hypothalamus and that their expression level correlated to the insulin-induced glucagon secretion trait. We then analyzed the transcriptomic data from the hypothalamus of naïve BXD and their parental strains that we previously generated [31]. Of the 155 genes present in this locus, the level of expression of 3 of them was significantly correlated to the trait, named 261005L07Rik, Agpat5, and LOC547150 (figure 2B).

Figure 2.

Candidate genes in chromosome 8 QTL. A: Localization of the QTL on chromosome 8 between markers rs13479624 and rs33450716 with a peak LRS = 19.23 on marker rs13479628. The QTL spans ∼6.38 Mb and contains 155 genes. B: Table showing chromosome 8 QTL genes with hypothalamic expression correlated to insulin-induced glucagon secretion ranked by p-value. r: Pearson's correlation coefficient. TPM: mean hypothalamic expression among BXD strains in transcripts per million.

The QTL on the chromosome 15 spans ∼4.94 Mb between pseudomarker c15.loc87 and marker rs13482723 with a peak LRS on rs3685284 (Figure 3A). This QTL contains 42 genes and explains 51.8% of the variance of the trait. Forty-two genes are present in this locus, and the expression level of 8 of them was significantly correlated with the glucagon trait (Figure 3B). The most strongly correlated genes were Irak4 (r = −0.584; p = 1.0 × 10⁻⁴) and Tmem117 (r = −0.553; p = 3.0 × 10⁻⁴) (Figure 3B and C). The correlation of the other 6 genes with insulin-induced glucagon secretion showed at least 10-fold less significant p-values. We found that a cis eQTL between markers rs3685284 and rs45781537 with peak LRS = 40.6 on rs13482723 controls the expression levels of Irak4 (Figure 3D) and that the presence of the DBA/2J allele at rs13482723 in BXD mice led to higher hypothalamic Irak4 expression than the presence of the C57BL/6J allele (Figure 3E). Thus, the same genomic interval on chromosome 15 controls both hypothalamic Irak4 expression and insulin-induced glucagon secretion.

Figure 3.

Candidate genes in chromosome 15 QTL. A: Localization of the QTL on chromosome 15 between markers c15.loc87 and rs13482723 with a peak LRS = 26.27 on marker rs3685284. The QTL spans ∼4.94 Mb and contains 42 genes. B: Table showing the chromosome 15 QTL genes with hypothalamic expression correlated to insulin-induced glucagon secretion ranked by p-value. r: Pearson's correlation coefficient. TPM: mean hypothalamic expression among BXD strains in transcripts per million. C: Scatterplot of the correlation between plasma glucagon and hypothalamic Irak4 expression in BXD mice and their parental strains. D: eQTL mapping identifies a cis eQTL on chromosome 15 controlling Irak4 expression between markers rs3685284 and rs45781537 with peak LRS = 40.6 at rs13482723. E: Effect of DBA/2J and C57BL/6J alleles at rs13482723 on hypothalamic Irak4 expression. Orange: BXD strains with DBA/2J allele; Cyan: BXD strains with C57BL/6J allele; Gray: heterozygous mice; Red: DBA/2J mice; Blue: C57BL/6J mice.

Because Irak4 showed the strongest, negative correlation with the glucagon secretion trait, we further examined its role in the response to hypoglycemia. The physiological roles of Tmem117 and Agapt5 in insulin-induced glucagon will be published separately. Here, we will also report on the role of Cpne8 encoding a calcium- and lipid-binding protein present in the QTL of chromosome 15 [37].

3.2. Hypothalamic Irak4 and insulin-induced glucagon secretion

The Irak4 (interleukin-1 receptor-associated kinase 4) gene encodes a 459-amino-acid protein kinase. It acts downstream of Il-1R and TlRs to activate an inflammatory response via the NF-κB and MAPK signaling pathways [32,38]. In DBA/2J mice, the Irak4 gene sequence shows two missense variants as compared to the C57BL/6J gene, namely I11V and N183R [39,40]. These variants are located outside of the functional domains of Irak4 and are not predicted to modify its function (https://www.uniprot.org/uniprot/Q8R4K2) [41]. Analysis of the level of expression of genes that pertain to the Il-1β and TlR signaling pathways (Table 1) showed that Irak4 was expressed at a higher level (log-fold change (LogFC) = 0.689) in the hypothalamus of DBA/2J mice as compared to that of C57BL/6J mice. We also found higher expression in DBA/2J of Il-1ß (LogFC = 2.524) and of Tlr6 (LogFC = 1.399), a receptor that recognizes pathogen-associated molecular patterns.

Table 1.

Differential expression of genes belonging to the Il-1R and TlR signaling pathways in the hypothalamus of DBA/2J vs. C57BL/6J mice.

Pathway	Gene list	Differential gene expression in the hypothalamus of DBA/2J vs. C57BL/6J mice (LogFC)
Il-1r signaling	Il-1a	−0.282
	Il-1b	2.524
	Il-1r1	−0.566
	Il-1r2	−0.368
	Il-1rap	−0.228
Tlr signaling	Tlr1	−0.382
	Tlr2	−0.169
	Tlr3	−0.116
	Tlr4	0.046
	Tlr5	−0.810
	Tlr6	1.399
	Tlr7	0.075
	Tlr9	−0.848
	Tlr12	−0.760
	Tlr13	0.405
Common signaling	Myd88	−0.527
	Irak4	0.689

To identify hypothalamic sites activated by insulin-induced hypoglycemia, where Irak4 expression levels would direct the differential glucagon response of C57Bl/6J and DBA/2J mice, we analyzed c-fos expression 2 h after i.p. saline or insulin injections in C57BL/6J mice. Figure 4A–D shows that hypoglycemia strongly induced c-fos expression in the ARH and the PVN. We thus analyzed the expression level of Irak4 by in situ hybridization in these hypothalamic nuclei. We found that Irak4 was expressed at a 2.5-fold higher level in the ARH, but not the PVN of DBA/2J mice as compared to C57BL/6J mice (Figure 4E–J).

Figure 4.

Hypoglycemia-induced neuronal activation and Irak4 expression in the hypothalamus of C57BL/6J and DBA/2J mice. A-B: Representative micrographs of c-fos-positive cells in the ARH 2 h after i.p. NaCl (A) or insulin injection (B) in C57BL/6J mice. Scale bar = 100 μm. C-D: Representative micrographs of c-fos-positive cells in the PVN 2 h after i.p. NaCl (C) or insulin injection (D) in C57BL/6J mice. Scale bar = 100 μm. E-F: Representative micrographs of Irak4 expression in the ARH of C57BL/6J mice. Scale bar: 50 μm. G-H: Representative micrographs of Irak4 expression in the ARH of C57BL/6J mice. Scale bar: 50 μm. I: Number of Irak4 mRNA spots in the ARH of the hypothalamus of C57BL/6J as compared to that in DBA/2J mice. (n = 5 and 4 mice; total number of spots counted: 845 and 1403 for C57BL/6J and. DBA/2J mice, respectively). J: Number of Irak4 mRNA spots in the PVN of C57BL/6J as compared to that in BA/2J mice. (n = 5 and 4 mice; total number of spots counted: 433 and 338 for C57BL/6J and. DBA/2J mice, respectively). Data are expressed as means ± SEM. ∗∗p < 0.005. Student's t-test (I).

DBA/2J mice secrete markedly less glucagon upon insulin-induced hypoglycemia as compared to C57BL/6J mice (11.9 ± 0.84 pM vs. 42.6 ± 10.99 pM for DBA/2J vs. C57BL/6J mice, respectively) (Figure 1B) despite reaching similar hypoglycemic levels (4.3 ± 0.28 mM vs 4.8 ± 0.20 mM for DBA/2J and C57BL/6J mice, respectively). Thus, the higher expression of Irak4, especially in the ARH, and of Il-1ß in the hypothalamus of DBA/2J mice suggested that the Il-1β/Irak4 signaling pathway could regulate insulin-induced glucagon secretion. To test this hypothesis, we evaluated the effect of blocking hypothalamic Il-β signaling on glucagon secretion. C57BL/6J and DBA/2J mice were injected i.c.v. with saline or the Il-1R antagonist Anakinra followed 1 h later by an i.p. injection of saline or insulin, and blood was collected after 60 min for plasma glucagon measurements. In C57BL/6J mice, i.c.v. injection of Anakinra had no effect on glycemia after i.p. saline (Figure 5A) or i.p. insulin injections (Figure 5B), and glucagon secretion increased to the same extent in mice previously injected with saline or Anakinra (Figure 5C). In DBA/2J mice, i.c.v. Anakinra had no effect on glycemia after i.p. saline (Figure 5D) but led to a deeper hypoglycemia after insulin injection (Figure 5E) and a markedly increased insulin-induced glucagon secretion (Figure 5F). Thus, increased insulin-induced glucagon secretion following the inhibition of Il-1β signaling in DBA/2J mice could result from increased insulin sensitivity, thereby causing more severe hypoglycemia and/or to higher sensitivity to hypoglycemia of the glucose-sensing system controlling glucagon secretion.

Figure 5.

Inhibition of Il-1R signaling restores hypoglycemia-induced glucagon secretion in DBA/2J mice. A: Glycemia in C57BL/6J mice injected i.c.v. with NaCl 0.9% or Anakinra at −60 min and i.p. with NaCl 0.9% at 0 min (n = 6–8). B: Glycemia in C567BL/6J mice injected i.c.v. with NaCl 0.9% or Anakinra at −60 min and i.p. with Anakinra at 0 min (n = 10–11). C: Plasma glucagon 60 min after i.p. injection of NaCl 0.9% or insulin in C57BL/6J mice that received i.c.v. injection of NaCl 0.9% or Anakinra at −60 min (n = 6–11). D: Glycemia in DBA/2J mice injected i.c.v. with NaCl 0.9% or Anakinra at −60 min and i.p. with NaCl 0.9% at 0 min (n = 7–9). E: Glycemia in DBA/2J mice injected i.c.v. with NaCl 0.9% or Anakinra at −60 min and i.p. with Anakinra at 0 min (n = 7). F: Plasma glucagon 60 min after i.p. injection of NaCl 0.9% or insulin in DBA/2J mice that received i.c.v. injection of NaCl 0.9% or Anakinra at −60 min (n = 7–9). G-I: Hyperinsulinemic-hypoglycemic clamps in DBA/2J mice that received i.c.v. NaCl or Anakinra at 0 min. I: Plasma glucagon measured at 90 min (n = 8–10). Data are means ± SEM. ∗p < 0.05; ∗∗p < 0.005. Mixed effects analysis followed by Sidak's multiple comparisons test (E and F). Student's t-test (I).

To determine whether hypoglycemia sensing in DBA/2J mice was affected by Il-1β signaling, we performed hyperinsulinemic-hypoglycemic clamps in mice injected i.c.v. with saline or Anakinra. Insulin was infused at a constant rate and glucose at a variable rate to reach ∼2.5 mM (Figure 5G). Anakinra administration did not affect the glucose infusion rate (Figure 5H) but significantly increased glucagon secretion (Figure 5I). Thus, in DBA/2J mice, Il-1β signaling negatively controlled hypoglycemia sensing and glucagon secretion.

3.3. Il-1β antagonism increases hypoglycemia-activated autonomic nervous activity in DBA/2J mice

To determine whether ARH and PVN were also sensitive to hypoglycemia in DBA/2J mice and whether Anakinra would increase this sensitivity, we injected mice with i.c.v. saline or i.c.v. Anakinra and 1 h later with i.p. insulin and prepared their brains 2 h later for c-fos immunofluorescence microscopy detection. One hour after insulin injection, the two groups of mice reached the same glycemic level (3.48 ± 0.32 mM vs. 3.91 ± 0.37 mM after saline and Anakinra administration, respectively). In the ARH (Figure 6A–E) and PVN (Figure 6F–J), we found that Anakinra administration significantly increased insulin-induced neuronal activation as compared to saline treatment.

Figure 6.

Inhibition of Il-1R signaling restores hypoglycemia-induced neuronal activation in DBA/2J mice. A-D: DBA/2J mice received i.c.v. NaCl or Anakinra. One hour later, they received i.p. insulin and were killed 2 h later. Representative micrographs of the ARH showing c-fos staining (A,C) or DAPI staining (B,D). Scale bar = 50 μm. E: Quantification of c-fos-positive cells in the ARH. F–I: The same DBA/2J mice as shown in A-D. Representative micrographs of the PVN showing c-fos staining (F,H) and DAPI staining (G,I). Scale bar = 50 μm. J: Quantification of c-fos positive cells in the PVN. Data are means ± SEM. ∗p < 0.05; ∗∗p < 0.005. Student's t-test (E and J).

The ARH and PVN nuclei are connected to pre-autonomic neurons [[42], [43], [44], [45]] that control both the parasympathetic (PNS) and sympathetic (SNS) nerves, which when activated by hypoglycemia trigger glucagon secretion [46]. Thus, we next recorded PNS and SNS activities in DBA/2J mice that received either i.c.v. saline or Anakinra before i.p. insulin. Upon insulin-induced hypoglycemia, we observed no increase in PNS activity in DBA/2J mice pre-injected with saline. This activity was, however, markedly higher following i.c.v. Anakinra injection (Figure 7AandB). Similarly, while insulin did not induce SNS activity in mice pre-injected with saline, a marked activation was observed in mice pre-injected with Anakinra (Figure 7C and D).

Figure 7.

Inhibition of Il-1R signaling restored hypoglycemia-induced autonomous nervous activity in DBA/2J mice. A-B: Parasympathetic nerve firing in the basal state or following i.p. insulin injection in DBA/2J mice that received i.c.v. injections of NaCl 0.9% or Anakinra 60 min before starting the recording. A: quantification of the firing activity and B: representative trace (n = 5 to 9). C-D: Sympathetic nerve firing in the basal state or following i.p. insulin injection in DBA/2J mice that received i.c.v. injections of NaCl 0.9% or Anakinra 60 min before starting the recording. C: quantification of the firing activity and D: representative trace (n = 5). Data are expressed as means ± SEM. ∗p < 0.05; ∗∗p < 0.005. Repeated-measures two-way ANOVA followed by Sidak's multiple comparisons test (A and C).

3.4. Copine 8 and insulin-induced glucagon secretion

As shown in Figure 3B, Tmem117 was the second-best correlated gene with the glucagon trait; its characterization and role in controlling glucagon secretion will be the subject of another publication. The next genes in the list were Pdzrn4, a PDZ domain and ring domain containing protein that may be involved in the control of cell proliferation; Prickle1, a nuclear hormone receptor and possible regulator of the Wnt/beta-catenin signaling pathway; Gxylt1, encoding a glucoside xylosetransferase; Zcrb1, a zinc finger containing RNA binding protein involved in splicing; Cpne8, a Ca⁺⁺-dependent and CII domain-containing protein, and Twf1, an actin-binding protein.

Copine8 is involved in Ca⁺⁺-dependent regulation of membrane trafficking and exocytosis [47] and, thus, synaptic vesicle release upon membrane depolarization. In addition, we found that the expression of the Cpne8 mRNA was enriched in the VMN (Fig. S1A), and its hypothalamic expression was under the control of a cis eQTL located in the QTL of chromosome 15, between markers rs13482702 and rs45781537 with peak LRS = 22.9 (Fig. S1B). Thus, to explore the potential role of Cpne8 in insulin-induced glucagon secretion, we generated mice with whole body Cpne8 inactivation (Fig. S1C). Recombination of the gene was confirmed by PCR analysis (Fig. S1D), and quantitative RT-PCR demonstrated a complete loss of Cpne8 expression in the hypothalamus of Cpne8^−/− mice as compared to that in Cpne8^+/+ mice (Fig. S1E). Saline- and insulin-induced glucagon secretion were assessed as described above. We found that Cpne8 inactivation had no impact on insulin-induced hypoglycemia (Fig. S1F) and insulin-induced glucagon secretion (Figure 1G).

4. Discussion

The present genetic and genomic screenings identified two QTLs involved in the control of hypoglycemia-induced glucagon secretion. In the QTL of chromosome 8, three genes showed expression levels strongly correlated with the glucagon secretion trait, named Agpat5 and two genes encoding proteins of unknown function. In the QTL of chromosome 15, 8 genes showed significant correlations with the glucagon trait. Here, we focused on the role of hypothalamic Irak4 and Cpne8. We found that in DBA/2J mice, which display a very low glucagon response to insulin-induced hypoglycemia, the level of hypothalamic Irak4 and of Il-1ß were significantly higher than that in the hypothalamus of C57BL/6J mice. We found that blocking hypothalamic Il-1R signaling in DBA/2J mice markedly increased hypoglycemia-induced glucagon secretion, c-fos expression in the ARH and PVN, and parasympathetic and sympathetic nerve activities. Collectively, our data suggest that hypothalamic Irak4 expression is genetically determined and that elevated Irak4 expression limits hypoglycemia-induced glucagon secretion.

In a previous genetic screening for genes controlling glucagon secretion induced by 2DG-induced neuroglucopenia, a procedure often used as a substitute for insulin-induced hypoglycemia, we identified a QTL on chromosome 7 and Fgf15 as a candidate gene. We found that Fgf15 expression defines a subpopulation of neurons in the DMH which, when activated suppress PNS activity and glucagon secretion but stimulate SNS activity to increase hepatic glucose production [30,31]. Our present screening for insulin-induced glucagon secretion identified two different QTLs on chromosomes 8 and 15. Thus, the central mechanisms of neuroglucopenia- and hypoglycemia-induced glucagon secretion recruit different neuronal circuits. This is not too surprising as 2DG suppresses glycolysis and ATP production, which leads to a rapid CRR and a marked hyperglycemia, whereas insulin can affect neuronal activity by binding to neuronal insulin receptors and/or as a result of the slowly developing hypoglycemia. Thus, our genetic data confirm that 2DG recruits non-overlapping neuronal circuits, both of which, however, converge on pre-autonomic regions to control PNS and SNS activity. This further supports the notion that hypoglycemia detection by the brain consists of a distributed glucose-sensing system, which is precisely integrated to control the counterregulatory response.

Irak4 encodes a protein kinase acting downstream of the Il-1R and TlRs [32,38,48]. Upon ligand binding, the Il-1R heterodimerizes with the Il-1R accessory protein (IL-Rap), and TlRs homodimerize or heterodimerize with other TlRs. These dimers associate via their cytoplasmic Toll/interleukin-1 receptor domains to the adaptor protein Myd88, which recruits Irak4 to induce the synthesis of pro-inflammatory cytokines such as Il-1β, Il-6 and TNFα through the NF-κB and MAPK signaling pathways [49,50]. The role of Irak4 in the response to TlR ligands and innate immunity is well characterized [38]. Its role in glucoregulation or specifically in the CRR is less well established.

Irak4 is expressed at a higher level in the hypothalamus of DBA/2J mice as compared to that in C57BL/6J mice, and in situ hybridization experiments showed that Irak4 is widely distributed in the hypothalamus of both strains with a markedly higher expression in the ARH, but not in the PVN, of DBA2/J mice, suggesting that the ARH is where Irak4 has a major role in controlling the differential glucagon response to hypoglycemia. The expression of Il-1ß was also found to be higher in the hypothalamus of DBA/2J mice. As our genetic screening identified Irak4, not Il-1ß as a genetic determinant of glucagon secretion, this suggests that Irak4 is at the center of an autoregulated signaling loop that controls Il-1ß expression, Il-1R signaling and, eventually glucagon secretion. To determine whether the high expression of Irak4 and Il-1R signaling was the cause of the lower glucagon secretion in DBA/2J mice, we blocked Il-1R signaling by i.c.v. injection of Anakinra before insulin-induced hypoglycemia. This increased glucagon secretion in DBA/2J mice but not in C57BL/6J mice. In DBA/2J mice, this was associated with an increased expression of c-fos in the ARH and the PVN, and with enhanced PNS and SNS nervous activity upon insulin-induced hypoglycemia. It is well known that ARH and PVN neurons can regulate PNS activity through their projections to the DVC [51] and AgRP neurons of the ARH, 40% of which are GI neurons [52] that innervate neurons of the PVN, which regulate SNS outflow to the liver, the pancreas, and the adrenal medulla [[53], [54], [55]].

In hyperinsulinemic-hypoglycemic clamps we also found that pretreatment with Anakinra increased glucagon secretion at the end of the hypoglycemic period. This experiment confirmed that Il-1R signaling modulates hypoglycemia-induced glucagon secretion in DBA/2J mice. This was an important point to verify as Anakinra pre-treatment of DBA/2J mice also led to deeper insulin-induced hypoglycemia than in saline pre-injected mice, suggesting that hypothalamic Il-1R signaling may also control insulin sensitivity. This aspect of Il-1β action will require further investigations.

Il-1β is expressed widely in the brain, neurons, astrocytes, and microglial cells [56,57]. Its expression can be induced by hypoglycemia, which induces an inflammatory reaction in the brain [[57], [58], [59]]. Previous studies have shown that i.p. injections of Il-1β induce hypoglycemia, associated with augmented insulin secretion [60,61], increased glucose utilization, and reduced hepatic glucose production [62]. Intraperitoneal injection of Il-1β also stimulates Il-1β mRNA expression in the hypothalamus, and the associated hypoglycemia is markedly reduced by i.c.v. injection of an Il-1R antagonist [63], indicating that the central Il-1β signaling pathway negatively impacts the CRR. The role of hypothalamic Il-1β in suppressing the counterregulatory response is further supported by a recent observation that hypoglycemia-activated microglial cells release cytokines, including Il-1β, in close proximity to NPY/AgRP neurons leading to impaired CRR [57]. In this context, our study shows that hypothalamic Il-1β signaling is a physiological regulator of glucagon secretion and the CRR.

Il-1β-induced hypoglycemia, in contrast to that induced by insulin, does not stimulate food consumption [64], a response that involves the activation of AgRP neurons [65]. A possible explanation for a lack of feeding stimulation is that Il-1β increases glucose utilization by neurons and astrocytes [66,67]. Thus, even though hypoglycemia develops, the metabolic effect of Il-1β may prevent cellular energy depletion and the consequent activation of AMP-dependent protein kinase, which is required for the activation of GI neurons [23,68]. Therefore, prolonged inflammatory conditions, such as those induced by multiple hypoglycemic episodes, which increase central Il-1β production may contribute to HAAF.

In summary, our genetic and transcriptomic screening identified novel candidate genes controlling the physiological response of the brain to hypoglycemia. We showed that the hypothalamic Il-1R/Irak4 signaling pathway, through a genetic control of Irak4 expression, is differentially expressed depending on the mouse genetic background, and an increased activity of this pathway negatively impacts insulin-induced glucagon secretion. This likely depends on the known effect of Il-1β to increase neuronal and astrocyte glucose metabolism, thereby preventing the normal activation of GI neurons by hypoglycemia. In DBA/2J mice, blocking Il-1R signaling increased the response of ARH and PVN neurons to hypoglycemia, the activation of both branches of the autonomic nervous system, and the secretion of glucagon. Collectively, our results provide new genetic and cellular information about the complexity of central hypoglycemia sensing and the counterregulatory response. They will also pave the way for a better understanding of how insulin induces hypoglycemia associated autonomic failure in diabetic patients.

Conflict of interest

The authors report no conflict of interest. JCA is an employee of Novo Nordisk.

Author contributions

BT and AP conceived the project and designed the experiments. AP, XB, JCA, DT, and SC performed the experiments. AP, MJ, and ARSA analyzed the genetic and transcriptomic data. AP, JCA, and BT analyzed the data and wrote the paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article: