A Parabrachial-Hypothalamic Cholecystokinin Neurocircuit Controls Counterregulatory Responses to Hypoglycemia

1Due to the serious pathophysiological consequences of low blood glucose, normoglycemia is a tightly defended state regulated by a conserved and coordinated network of peripheral and central systems. Sensory information from peripheral glucosensors is integrated into the wider glucoregulatory/sensory circuitry, defined principally by the brainstem and hypothalamus, which in turn engages the sympathetic nervous system (SNS) and hypothalamic-pituitary-adrenal axis to stimulate glucose production and inhibit glucose uptake. Yet, it is unclear how these disparate structures converge to orchestrate sensorimotor responses to dysglycemia.

2The lateral parabrachial nucleus (LPBN) forms part of the preautonomic circuitry that subserves physiological responses to numerous viscerosensory modalities, including nociception, thermostasis, malaise, and energy homeostasis. As an assimilatory interoceptive relay for ascending sensory information, the LPBN is therefore neuroanatomically positioned to respond to various aspects of homeostatic dysregulation. Indeed, the LPBN has been implicated in responses to hypoglycemia, but detailed characterization has hitherto been lacking.

3The functional diversity of the LPBN is underscored by a complex anatomical substructure and neurochemical composition. To determine regions of hypoglycemic responsiveness, glucoprivation was pharmacologically induced through administration of the glucose anti-metabolite 2-deoxyglucose (2DG) or insulin (INS) and cFOS-immunoreactivty (IR), a molecular correlate of neuronal activation, assessed across the rostral-caudal extent of the LPBN. Both stimuli increased LPBN cFOS-IR, although the distribution and number of activated neurons was greater upon 2DG administration (Figure 1A–1D; Figure S1 available online). Of particular note was the common induction of cFOS-IR within the superior LPBN (sLPBN) (Figures 1A–1C and 1E). The neuropeptide CCK is highly expressed within the sLPBN, visualized here using a CCK-ires-Cre::R26-loxSTOPlox-L10-GFP line (Figures 1F and 1G). 2DG and INS treatment significantly increased cFOS-IR within CCKLPBN neurons (Figures 1H–1K). This identifies CCKLPBN neurons as one neurochemically defined subpopulation of LPBN cells responsive to states of glucoprivation, in addition to other non-CCK-expressing cells. To determine the proximal glucoprivic stimulus to which CCKLPBN neurons are responsive, we assessed their capacity to sense shifts in extracellular glucose concentration. Whole-cell recordings from transgenically labeled CCKLPBN neurons demonstrated that a downward step from 5 mM to 0.5 mM resulted in reversible membrane depolarization in 5/14 cells and in spontaneously firing cells an increase in action potential firing rate (Figures 1L–1N; responding cells defined by a poststimulus response greater than 2×SD ± mean of recorded baseline), establishing these cells as a population of glucose-inhibited neurons.

4To probe the relevance of CCKLPBN neurons to glucoregulation, a chemogenetic interrogation of their physiological function using Designer Receptors Activated by Designer Drugs (DREADD) was undertaken. Cre-dependent viral transduction of CCKLPBN neurons with stimulatory hM3Dq-mCherry facilitated robust activation in response to the designer drug clozapine-N-oxide (CNO), as determined by ex vivo slice electrophysiology (Figures 2A and 2B) and in vivo cFOS-IR (Figures S2A and S2B). In freely behaving animals, CCK-ires-Cre::hM3Dq-mCherryLPBN neuron activation prompted an increase in blood glucose concentration that reached a maximal elevation above baseline by 60 min (Figure 2C) (CNO administration had no effect on blood glucose concentration in the absence of DREADD-receptor expression; Figure S2C). Coincident with this rise in blood glucose, CNO administration significantly elevated serum levels of the CR hormones glucagon and corticosterone (Figures 2D and 2E) but had no significant effect on serum INS levels, although there was a trend toward a decrease (Figure S2D). Furthermore, CCKLPBN neuron activation also stimulated sympathoexcitatory drive to the adrenal glands, as indicated by increased serum epinephrine levels (Figure 2F), indicating that these cells couple to the SNS-circuitry requisite for autonomic control of endogenous glucose production. Consistent with these findings, mRNA levels of the glucogenic gene glucose-6-phosphatase (G6pc) increased 2-fold in livers of CNO-treated CCK-ires-Cre::hM3Dq-mCherryLPBN mice, compared to saline controls (Figure 2G). We next investigated the necessity of CCK neurotransmission for CCK-ires-Cre::hM3Dq-mCherryLPBN-induced hyperglycemia. Systemic pretreatment of CCK-ires-Cre::hM3Dq-mCherryLPBN mice with the CCK-receptor antagonist proglumide abrogated the previously observed elevation in blood glucose (Figure 2H), revealing that LPBN-derived CCK is the functionally relevant neurotransmitter in this physiological context. These data demonstrate that the CCKLPBN neurons, and specifically CCK neuropeptide, are sufficient to drive CR-like responses. Interestingly, despite the involvement of both the LPBN and CCK in appetite control, CCKLPBN neuron activation had no impact on feeding behavior (Figures 2I, 2J, and S2E).

5This ability of CCKLPBN neurons to elevate blood glucose concentration, together with their responsiveness to glucoprivic stimuli, alluded to an involvement in the CR-responses to hypoglycemia, wherein their activation facilitates the re-establishment of normoglycemia. To address the physiological relevance of CCKLPBN neurons to counterregulation, we investigated the consequence of their chemogenetic inhibition within the context of glucoprivation. Ex vivo electrophysiology demonstrated that in the presence of CNO, CCK-ires-Cre::hM4Di-mCherryLPBN neurons exhibited membrane hyperpolarization and a decrease in action potential firing (Figures 3A and 3B), establishing a capacity to effectively silence CCKLPBN neurons. In freely behaving normoglycemic (ad-libitum-fed) CCK-ires-Cre::hM4Di-mCherryLPBN mice, CNO administration had no effect on blood glucose concentration (Figure 3C), revealing that these cells are not requisite to the regulation of baseline glycemia. However, CCKLPBN neuron silencing prior to the induction of acute glucoprivation resulted in a significantly diminished CR-response. Specifically, while 2DG administration prompted the expected CR elevation in blood glucose, prior CCK-ires-Cre::hM4Di-mCherryLPBN neuron inhibition markedly attenuated the hyperglycemic response (Figure 3D). Similarly, under INS-induced hypoglycemia, CCKLPBN neuron silencing engendered an exaggerated hypoglycemic response that impaired re-establishment of normoglycemia (Figure 3E). Thus, CCKLPBN neurons form part of the neurocircuitry underlying centrally regulated glycemia and are both necessary and sufficient for CR-responses to glucoprivation. CCKLPBN neuron silencing did not affect feeding behavior (Figures S3A–S3C).

6To probe the neurocircuit underlying CCKLPBN neuron-mediated glucoregulation, their specific efferent targets were assessed using genetically encoded neuronal tract-tracing. Unilateral viral transduction of CCKLPBN neurons with Cre-dependent synaptophysin-mCherry revealed a strictly ascending projection profile, with terminals concentrated predominantly within the ipsilateral hypothalamus (Figures 4A–4C and S4A–S4F). The densest site of CCKLPBN neuron innervation was the ventromedial nucleus of the hypothalamus (VMH), and in particular the dorsomedial compartment (dmVMH) (Figure 4C). This robust innervation, together with the well-established glucoregulatory capacity and preautonomic function of the VMH, supported its role as a site of functional outflow for CR CCKLPBN neurons.

7SF1 is required for the terminal differentiation of the VMH and is expressed predominantly in the dmVMH of adult mice. Conditional genetic manipulations have demonstrated that SF1 neurons are required for CR-response to hypoglycemia. Thus, we investigated whether SF1VMH neurons form part of a glucoregulatory CCKLPBN→VMH microcircuit. Concordant with the lateralized nature of CCKLPBN neuron projections, unilateral CCK-ires-Cre::hM3Dq-mCherryLPBN neuron activation induced an increase in cFOS-IR within transgenically labeled ipsilateral SF1VMH neurons, as compared to the contralateral side, indicating that SF1VMH neurons lie downstream of CCKLPBN neurons (Figure 4D). Consistent with this observation and the excitatory effect of CCK on the VMH, electrophysiological recordings from synaptically isolated SF1VMH neurons at low (0.5 mM) and normal (5 mM) glucose conditions revealed that 55% (5/9) were stimulated by exogenous CCK, with responding cells increasing their firing rate 2.5-fold (Figures 4E, 4F, and S4G); importantly, pretreatment of slices with a CCKB-receptor antagonist (CI988) blocked responses to CCK in all cells tested (n = 12). Non-SF1VMH neurons did not exhibit a response to CCK (0/12 cells; Figures S4H and S4I). In light of evidence that LPBN-derived CCK is the sole source of CCK terminals within the VMH, these data suggest SF1VMH neuron responsiveness to CCK lies in their innervation by CCKLPBN neurons. Furthermore, the robustly glutamatergic nature of the LPBN led us to consider the potential contribution of fast neurotransmission from CCKLPBN neurons. Interestingly, we did not observe VMH projections from glutamatergic vGLUT2LPBN neurons (Figure S4J), and vGLUT2-ires-Cre::hM3Dq-mCherryLPBN stimulation failed to elevate blood glucose levels (Figure S4K). These data suggest that VMH-projecting CCKLPBN neurons are not glutamatergic and support our observation that CCK receptor blockade is sufficient to abrogate the hyperglycemic response to CNO in CCK-ires-Cre::hM3Dq-mCherryLPBN mice (Figure 2H).

8Taking advantage of the spatial exclusivity of CCK and SF1 expression in the LPBN and VMH, respectively, we next generated compound CCK-ires-Cre::SF1-Cre mice to facilitate neurochemically explicit manipulation of this CCKLPBN→SF1VMH microcircuit. Specifically, we sought to functionally occlude CCK-ires-Cre::hM3Dq-mCherryLPBN-induced CR-responses through the simultaneous silencing of downstream SF1-Cre::hM4Di-mCherryVMH neurons (Figure 4G). Concomitant CNO-induced activation and inhibition of CCKLPBN and SF1VMH neurons, respectively, resulted in abrogation of CCK-ires-Cre::hM3Dq-mCherryLPBN-mediated elevation in blood glucose (Figure 4H). Silencing of SF1VMH neurons alone had no effect on blood glucose concentration, indicating a causal interaction between these two populations, and not simply a counteracting additive effect (Figure 4I). Furthermore, like CCKLPBN neurons, silencing of SF1VMH cells within the context of 2DG-glucoprivation induced an attenuated CR-response (Figure S4L). Control experiments performed to rule out any contribution from CCK neurons within the dorsomedial nucleus of the hypothalamus, compact part (cDMH), which neighbors the SF1VMH domain, revealed that CCKcDMH neurons had neither any effect on blood glucose levels nor were engaged by glucoregulatory CCKLPBN neurons (Figures S4M–S4P). These data suggest that the CR function of CCKLPBN neurons is predicated upon their engagement of downstream SF1VMH neurons. Interestingly, systemic antagonism of CCKB-receptors, the predominant central receptor isoform (and enriched in the dmVMH), was sufficient to significantly attenuate the hyperglycemic response to 2DG in wild-type mice (Figure 4J), highlighting the salience of CCKB-receptor signaling to CR-responses, although not definitively identifying specific receptor-expressing neuronal populations.

9Here we demonstrate that CCKLPBN neurons are a population of glucose-sensitive neurons and a requisite component of the neurocircuitry underlying CR-responses to hypoglycemia. Furthermore, the necessity for SF1VMH neurons as downstream effectors of CCKLPBN-regulated glycemia befits their established role in CR. In sum, the present study defines the physiological function of a discrete CCKLPBN→SF1VMH microcircuit that drives hepatic glucose production and mediates CR responses to hypoglycemia. Furthermore, these data identify a critical function for CCK in the physiological response to glucoprivation. These observations have salience to both homeostatic function in health and dysregulation in disease, in particular diabetes.

10CCK-ires-Cre (Jackson Laboratories), SF1-Cre, vGLUT2-ires-Cre, and R26-loxSTOPlox-L10-GFP mice were generated and maintained as previously described. Animal care and experimental procedures were performed with approval by the Beth Israel Deaconess Medical Center Institutional Animal Care and Use Committee or were performed in accordance with the UK Animals (Scientific Procedures) Act 1986.

11The DREADD viruses used have been described previously: AAV8-hSyn-DIO-hM3Dq-mCherry and AAV8-hSyn-DIO-hM4Di-mCherry (University North Carolina Vector Core). A Cre-dependent expression cassette for hEF1alpha-DIO-synaptophysin-mCherry-WPRE was generated (MIT Viral Gene Transfer Core) and the construct packaged in AAV serotype-8 at a titer of 1.3 × 1013 vg/ml (Virovek, Inc). Nucleus specific delivery of viruses was achieved through stereotaxic delivery based upon coordinates defined by the Mouse Brain Atlas.

12On test days, mice were transferred to new cages and food removed at 9:00 am. Mice remained fasted until 1:00 pm, when the experiment would commence. Blood glucose concentration was determined by tail bleed using a OneTouch Ultra glucometer and test strips (LifeScan, Johnson and Johnson Company). Basal blood glucose concentration was determined prior to injection of any substances and at 15, 30, 60, 120, and 180 min postadministration. Pretreatments were given 30 min prior to pharmacological stimulus. Experiments were within study design (unless otherwise stated), with mice assessed following saline and 1 mg/kg CNO. Animals undergoing loss-of-function experiments involving glucoprivic stimuli (INS or 2DG) were counterbalanced and given 3 weeks recovery time between treatments to ensure normal counterregulatory responses were intact.

13Brains were sectioned on a freezing microtome at 30 μm. Dual immunofluorescence histochemistry for cFOS-IR, mCherry-IR, and GFP-IR was conducted as previously described. Primary antibodies (1/1,000) were as follows: rabbit anti-cFOS (EMD Millipore), rabbit anti-RFP (Clontech Laboratories, Inc.), or chicken anti-GFP (EMD Millipore).

14For assessment of serum chemistry and hepatic gene expression animals were decapitated, and fresh trunk blood and liver samples were collected. Blood was collected in tubes containing 250 KIU/ml aprotinin (EMD Millipore), allowed to coagulate at room temperature for 30 min, centrifuged at 1,500 × g for 10 min, and the resulting serum transferred to a new tube. Samples were flash frozen in liquid nitrogen and kept at −80°C until analyzed. Liver extracts were flash frozen in liquid nitrogen and kept at −80°C until RNA was extracted.

15To assess the effects of CNO on CCKLPBN neurons, 5- to 7-week-old CCK-ires-Cre mice were injected with either AAV8-DIO-hM3Dq-mCherry or AAV8-DIO-hM4Di-mCherry into the LPBN 3 weeks before recording. Slices were maintained in aCSF containing 10 mM glucose. After acquisition of stable, whole-cell recordings for 2 to 5 min, aCSF solution containing CNO (5 μM) was perfused into the brain slice preparation (for hM4Di-mCherry analysis, cells were injected with 5 pA of current). To assess the effect of CCK on VMH neurons (SF1+ or SF− neurons), 6- to 8-weeks-old SF1-Cre::R26-loxSTOPlox-L10-GFP mice were used. CCK-8S (100 nM; Tocris Biosciences) and CI988 (500 nM; Tocris Biosciences) were applied to bath solution containing either 5 mM or 0.5 mM glucose through perfusion. After acquisition of stable, whole-cell recording for 2–5 min, acSF solution containing 100 nM of CCK-8S was perfused for 3 to 5 min. Synaptic blockers (1 mM kynurenate and 100 μM picrotoxin) were added in aCSF to synaptically isolate VMH neurons. The effect of glucose concentration on CCKLPBN neuron activity was assessed using a 5 mM→0.5 mM→5 mM step protocol, with osmolarity adjusted with sucrose.

16Statistical analyses were performed using Prism 6 (Graphpad Software, Inc). Data were analyzed using t test, one-way ANOVA, or two-way ANOVA with post hoc comparisons, where appropriate. Data are presented as mean ± SEM, and statistical significance was set at p < 0.05

17A.S.G. conceived experiments with input from L.K.H., M.G.M., M.L.E., B.B.L., C.M.P., and J.F. A.S.G. performed experiments with help from L.K.H., B.P.S., L.K.B., and J.C.M. Synaptophysin-mCherry construct was made by R.L.N. A.S.G. interpreted data with help from L.K.H., M.L.E., B.B.L., and M.G.M. A.S.G. wrote the manuscript with input from L.K.H., M.L.E., B.B.L., and M.G.M.