AVP-induced counter-regulatory glucagon is diminished in type 1 diabetes

Angela Kim, Jakob G. Knudsen, Joseph C. Madara, Anna Benrick, Thomas Hill, Lina Abdul Kadir, Joely A. Kellard, Lisa Mellander, Caroline Miranda, Haopeng Lin, Timothy James, Kinga Suba, Aliya F. Spigelman, Yanling Wu, Patrick E. MacDonald, Ingrid Wernstedt Asterholm, Tore Magnussen, Mikkel Christensen, Tina Vilsboll, Victoria Salem, Filip K. Knop, Patrik Rorsman, Bradford B. Lowell, Linford J.B. Briant


Introduction
A tight regulation of blood glucose is critical for normal brain function. For this reason, hypoglycaemia evokes a multitude of responses to counter deleterious declines in blood glucose and avert brain failure and ultimately death (Cryer, 2007). A critical component to this "counter-regulatory" response is the elevation of plasma glucagon -a hormone that is considered the first line of defence against hypoglycaemia (Cryer, 1994).
Glucagon is secreted from alpha-cells of the pancreatic islets, and counters hypoglycaemia by potently stimulating hepatic glucose production. The importance of glucagon for glucose homeostasis is demonstrated in both type 1 and type 2 diabetes mellitus, wherein hyperglycaemia results from a combination of complete/partial loss of insulin secretion and over-secretion of glucagon (Unger & Cherrington, 2012). Despite the centrality of glucagon to diabetes aetiology and its importance in countering hypoglycaemia, there remains considerable uncertainty regarding how its secretion is regulated.
The mechanism(s) by which glucagon secretion is controlled is a contentious topic (Gylfe, 2013;Lai et al., 2018). Based on observations in isolated (ex vivo) islets, islets have been shown to respond to hypoglycaemia (and adjust glucagon secretion accordingly) via intrinsic (Rorsman et al., 2014;Basco et al., 2018;Yu et al., 2019) and paracrine mechanisms (Briant et al., 2018;Vergari et al., 2019). While it is indisputable that the islet is a critical component of the body's glucostat (Rodriguez-Diaz et al., 2018) and has the ability to intrinsically modulate glucagon output, it is clear that such an 'islet-centric' viewpoint is overly simplistic (Schwartz et al., 2013). Indeed, many studies have clearly demonstrated that brain-directed interventions can profoundly alter islet alpha-cell function, with glucose-sensing neurons in the hypothalamus being key mediators (Garfield et al., 2014;Lamy et al., 2014;Meek et al., pg. 5 2016). This ability of the brain to modulate glucagon secretion is commonly attributed to autonomic innervation of the pancreas (Marty et al., 2005;Lamy et al., 2014;Thorens, 2014).
However, counter-regulatory glucagon secretion is not only restored in pancreas transplant patients but also insensitive to adrenergic blockade (Diem et al., 1990), suggesting that other (non-autonomic) central mechanisms may also contribute to counter-regulatory glucagon secretion in vivo.
Arginine-vasopressin (AVP) is a hormone synthesised in the hypothalamus (see review by Bourque (2008)). AVP neurones are divided into two classes: parvocellular AVP neurons (which either project to the median eminence to stimulate ACTH and glucocorticoid release or project centrally to other brain regions) and magnocellular AVP neurons (which are the main contributors to circulating levels of AVP). The parvocellular neurones reside solely in the paraventricular nucleus of the hypothalamus (PVH), whereas magnocellular neurones are found in both the PVH and supraoptic nucleus (SON). Stimulation of the magnocellular AVP neurons causes release of AVP into the posterior pituitary, where it enters the peripheral circulation.
Animals that have had glucagon signalling blocked with an antibody against the glucagon receptor exhibit profound hyperglucagonaemia (Sloop et al., 2004). In these mice, transcripts for the vasopressin 1b receptor (Avpr1b) are highly upregulated in alpha-cells . Even under normal conditions, Avpr1b is one of the most enriched transcripts in alphacells (Lawlor et al., 2017;van der Meulen et al., 2017). This raises the possibility that AVP may be an important regulator of glucagon secretion under physiological conditions and that this regulatory pathway is utilised in conditions of interrupted glucagon signalling to bolster glucagon output. Here, we have investigated the regulation of glucagon by circulating AVP.

pg. 6
We show that glucagon secretion in response to hypoglycaemia is due to AVP and we identify a novel role for glucagon; maintaining plasma glucose in the dehydrated state.
Finally, we explore the link between copeptin (co-secreted with AVP but with a longer circulating half-life) and glucagon in humans and provide evidence that this pituitary-alphacell axis is impaired in type 1 diabetes.

Glucagon secretion in response to an insulin tolerance test is driven by AVP
Hypoglycaemia induced by an insulin tolerance test (ITT) robustly stimulates glucagon secretion in vivo (Fig. 1a, b). However, reducing the glucose concentration from 8 to 4 mM (mimicking an ITT) does not stimulate glucagon secretion from isolated (ex vivo) islets (Fig.   1c). Therefore, mechanisms extrinsic to the islet must control glucagon secretion in vivo, at least during an ITT. We hypothesised that this extrinsic stimulus is AVP. First, we investigated whether AVP neuron activity is increased in response to an ITT in vivo using fibre photometry . Exogenous insulin evoked an increase in AVP neuron activity , whereas saline vehicle treatment was ineffective . Finally, we confirmed that plasma AVP is elevated following an ITT by measuring its surrogate marker, copeptin (the C-terminal fragment of pre-pro-AVP; Fig. 1i).
To investigate the kinetics and glucose-dependency of this response, we simultaneously recorded AVP neuron activity and plasma glucose (by continuous glucose monitoring; Fig.   2a, b). This revealed that AVP neuron activity initiates at 4.9 ±0.4 mM glucose (Fig. 2c) at a delay of ~10 mins (Fig. 2d). Once the threshold was reached, AVP neurons displayed random bursts of activity. The amount of AVP neuronal activation did not correlate with the extent of hypoglycemia. Neither frequency nor peak amplitude correlated with the blood glucose change. Insulin-induced hypoglycemia was accompanied by a significant decrease in the body temperature (Fig. 2e), as previously reported (Freinkel et al., 1972;Buchanan et al., 1991).
To determine whether AVP contributes to glucagon secretion during an insulin-induced hypoglycaemia, we injected wild-type mice with the V1bR antagonist SSR149415 pg. 8 (nelivaptan;Serradeil-Le Gal et al. (2002)) prior to the ITT ( Fig. 3a-b). This reduced insulininduced glucagon secretion by 70%. Similarly, in Avpr1b knockout mice (Avpr1b -/-; Lolait et al. (2007)) glucagon secretion was decreased by 65% compared to wild-type littermates (Avpr1b -/-; Fig. 3c-d). There was not a change in the depth of hypoglycaemia induced following V1bR antagonism (Fig. 3b) or in Avpr1b -/mice (Fig. 3d), despite the drastic reduction in plasma glucagon. Thus, it appears that the supraphysiological concentrations of insulin likely dominate the hyperglycaemic action of endogenous glucagon. This 'glucagonostatic' effect of Avpr1b antagonism or KO was also observed during deeper hypoglycaemia, achieved with a higher dose of insulin .
AVP has also been reported to increase circulating corticosterone (Antoni, 1993) and adrenaline (Grazzini et al., 1996), but in our experiments neither adrenaline ( Supplementary   Fig. 1a-c) nor corticosterone (Supplementary Fig. 1d and e) increased glucagon output or alpha-cell activity in isolated mouse or human islets.
We next investigated the role of AVP in controlling glucagon secretion evoked by 2-deoxy-D-glucose (2DG). 2DG is a non-metabolisable glucose molecule that evokes a state of perceived glucose deficit (mimicking hypoglycaemia) that triggers a robust counterregulatory stimulation of glucagon secretion (Marty et al., 2005). We hypothesised that AVP contributes to this response. We monitored AVP neuron activity in response to 2DG by in vivo fibre photometry (as per Fig. 1d), and correlated this to changes in plasma glucose during this metabolic challenge ( Supplementary Fig. 2). Injection (i.p.) of 2DG increased blood glucose (Supplementary Fig. 2a) and triggered a concomitant elevation of GCaMP6s signal in AVP neurons . The hyperglycaemic response to 2DG was suppressed by pre-treatment with either the V1bR antagonist SSR149415 or the glucagon receptor antagonist LY2409021 (Supplementary Fig. 2e; Kazda et al. (2016)). Finally, the elevation in plasma glucagon by 2DG injection was reduced following pretreatment with the V1bR antagonist SSR149415 (Supplementary Fig. 2f). We conclude that AVP contributes to the hyperglycaemic response to 2DG by stimulating glucagon release.

AVP evokes hyperglycaemia and hyperglucagonaemia
Next, we investigated the mechanisms by which AVP stimulates glucagon during hypoglycaemia. We first investigated the metabolic effects of AVP in vivo ( Fig. 4a-d). We expressed the modified human M3 muscarinic receptor hM3Dq (see Alexander et al. (2009)) in AVP neurons by bilaterally injecting a Cre-dependent virus containing hM3Dq (AAV-DIO-hM3Dq-mCherry) into the supraoptic nucleus (SON) of mice bearing Avp promoterdriven Cre recombinase (Avp ires-Cre+ mice; Fig. 4a). Expression of hM3Dq was limited to the SON ( Supplementary Fig. 3), thus allowing targeted activation of magnocellular AVP neurons (which determine circulating AVP). Patch-clamp recordings from brain slices of the SON prepared from these mice confirmed that bath application of clozapine-N-oxide (CNO; 5-10 µM)a specific, pharmacologically inert agonist for hM3Dq -induced membrane depolarisation and increased the firing rate in hM3Dq-expressing AVP neurons (Supplementary Fig. 3a and b). Injection of CNO (3 mg/kg i.p.) in vivo increased blood glucose (Fig. 4b) and copeptin (Fig. 4c). To establish the contribution of glucagon to this hyperglycaemic response, we pre-treated mice with the glucagon receptor antagonist LY2409021. This completely abolished the hyperglycaemic action of CNO (Fig. 4b).
Similarly, to understand the contribution of vasopressin 1b receptor (V1bR) signalling, we pre-treated mice with the V1bR antagonist SSR149415. This also abolished the hyperglycaemic effect of CNO (Fig. 4b). Measurements of plasma glucagon during CNO treatment revealed it was elevated by ~40% (Fig. 4d). We note that this response was modest, pg. 10 which could reflect the variability in hM3Dq expression or the fact that the elevation in blood glucose exerts a glucagonostatic effect. CNO did not change food intake ( Supplementary Fig.   3f) and did not have an off-target effect on blood glucose in Avp ires-Cre+ mice expressing a passive protein (mCherry) under AAV transfection in the SON (Supplementary Fig. 3g).
Exogenous AVP also caused an increase in glucose and glucagon relative to saline injection in wild-type mice . In summary, CNO stimulates AVP release, which in turn elevates plasma glucose through stimulation of glucagon release.
To determine whether the hyperglycaemic and hyperglucagonaemic actions of AVP are due to AVP directly stimulating glucagon secretion from alpha-cells, we assessed the response of islets to AVP. First, we assessed this response in vivo; we isolated islets from Gcg Cre+ mice crossed with a Cre-dependent GCaMP3 reporter mouse (from hereon, Gcg-GCaMP3 mice) and implanted them in the anterior chamber of the eye (Fig. 4e-h; see Salem et al. (2019)).
This allows the cytoplasmic Ca 2+ concentration ([Ca 2+ ]i) in individual alpha-cells to be imaged in vivo. Administration (i.v.) of AVP resulted in a biphasic elevation of [Ca 2+ ]i consisting of an intial spike followed by rapid oscillatory activity ( Fig. 4f-g).
To understand whether AVP stimulates glucagon secretion directly from islets, we made a detailed assessment of the response of isolated (ex vivo) mouse islets and the in situ perfused pancreas to AVP. First, we characterised the expression of vasopressin receptors in mouse islets. V1bR mRNA (Avpr1b) was expressed in whole mouse islets, whereas vasopressin receptor subtypes 1a and 2 mRNA were not; instead, they were detected in various extrapancreatic tissue, consistent with their distinct roles in the regulation of blood pressure and diuresis (Bourque, 2008), respectively ( Supplementary Fig. 4a). To determine whether Avpr1b expression was enriched in alpha-cells, mice bearing a proglucagon promoter-driven pg. 11 Cre-recombinase (Gcg Cre+ mice) were crossed with mice expressing a Cre-driven fluorescent reporter (RFP). qPCR of the fluorescence-activated cell sorted RFP + and RFPfractions revealed that Avpr1b is upregulated in mouse alpha-cells with ~43-fold enrichment (Supplementary Fig. 4c and d).
AVP dose-dependently stimulated glucagon secretion from isolated mouse islets ( Supplementary Fig. 5). In dynamic measurements using in situ perfused mouse pancreas, AVP was found to produce a biphasic increase in glucagon secretion (Supplementary Fig. 5d). In mouse islets, the AVP-induced increase in glucagon was prevented by SSR149415 ( Supplementary Fig. 5c). AVP did not affect insulin secretion . To understand the intracellular mechanisms by which AVP stimulates glucagon secretion, we performed perforated patch-clamp recordings of electrical activity and Ca 2+ imaging in alphacells in isolated islets from Gcg Cre+ mice crossed with a Cre-dependent GCaMP3 reporter mouse (from hereon, Gcg-GCaMP3 mice). AVP increased action potential firing (Supplementary Fig. 5i,j) and Ca 2+ oscillations ( Supplementary Fig. 5k, l) in a dosedependent manner. The Ca 2+ response was abolished following application of the V1bR antagonist SSR149415 ( Supplementary Fig. 5m, n). AVP-induced Ca 2+ activity was dependent on Gq-protein activation, because it was blocked with the Gq-inhibitor YM254890 (Takasaki et al. (2004); Supplementary Fig. 6a, b) and it increased intracellular diacylglycerol ( Supplementary Fig. 6c, d). In line with Gq activation, electrical activity evoked by AVP exhibited membrane potential oscillations (as revealed by power-spectrum analysis; Supplementary Fig. 6e). Conducting mathematical modelling of the intracellular signalling pathway following Gq receptor activation demonstrated that the canonical Gq pathway can evoke Ca 2+ oscillations, further supporting the involvement of VlbR, a Gq pg. 12 protein-coupled receptor, in stimulating glucagon secretion from alpha cells ( Supplementary   Fig. 6f, g).

Hypoglycaemia evokes AVP secretion via activation of A1/C1 neurons
Many physiological stressors activate hindbrain catecholamine neurons, which release noradrenaline (A1) or adrenaline (C1) and reside in the ventrolateral portion of the medulla oblongata (VLM). Activation of C1 neurons (by targeted glucoprivation or chemogenetic manipulation) is known to elevate blood glucose (Ritter et al., 2000;Zhao et al., 2017;Li et al., 2018) and plasma glucagon (Andrew et al., 2007). Furthermore, C1 cell lesions severely attenuate the release of AVP in response to hydralazine-evoked hypotension (Madden et al., 2006), indicating that this hindbrain site may be a key regulator of AVP neuron activity during physiological stress.
Furthermore, these EPSCs could be blocked with TTX, but reinstated with addition of 4-AP pg. 13 To explore the consequences of A1/C1 activation in vivo, we injected AAV-DIO-hM3Dq-mCherry bilaterally into the VLM of Th Cre+ mice (Fig. 5f, Supplementary Fig. 7b, c).
Activation of A1/C1 neurons with CNO evoked a ~4.5 mM increase in plasma glucose (Fig. 5g,h). Pre-treatment with the glucagon receptor antagonist LY2409021 inhibited this response (Fig. 5g, h). In line with this, plasma glucagon was increased following CNO application (Fig. 5i). The hyperglycaemic response was also dependent on functional V1bRs, because it was abolished following pre-treatment with the V1bR antagonist SSR149415 (Fig.   5g, h). CNO had no effect on blood glucose in Th Cre+ mice expressing mCherry in A1/C1 neurons ( Supplementary Fig. 7c). Together, these data indicate that A1/C1 neuron activation evokes AVP and glucagon secretion. We therefore hypothesised that hypoglycaemia-induced AVP release is due to projections from A1/C1 neurons.
In line with this hypothesis, c-Fos expression (a marker of neuronal activity) was increased in A1/C1 neurons following an insulin bolus (Supplementary Fig. 8a and b).
To determine the contribution of A1/C1 neurons to AVP neuron activity during an ITT, we inhibited A1/C1 neurons whilst monitoring AVP neuron activity. To this end, we expressed an inhibitory receptor (the modified human muscarinic M4 receptor hM4Di; Armbruster et al. (2007)) in A1/C1 neurons by injecting AAV-fDIO-hM4Di-mCherry into the VLM, whilst co-injecting AAV-DIO-GCaMP6s into the SON of Dbh flp+ x Avp ires-Cre+ mice (Fig. 5j). We then measured AVP neuron population [Ca 2+ ]i activity (with in vivo fibre photometry) and plasma glucagon following inhibition of A1/C1 neurons with CNO ( Fig. 5j and pg. 14 Supplementary Fig. 8c). AVP neuron population activity during an ITT was reduced by A1/C1 silencing compared to vehicle injection ( Fig. 5k-m). Furthermore, glucagon secretion was reduced following CNO silencing of A1/C1 neurons (Fig. 5n). Together, these data suggest that AVP-dependent glucagon secretion during an ITT is mediated by A1/C1 neurons.

AVP-induced glucagon secretion maintains glucose homeostasis during dehydration
Collectively, these data point to AVP being an important regulator of glucagon secretion but they do not explain why a receptor for a hormone commonly known for its anti-diuretic action is expressed in alpha-cells. We speculated that glucagon may play an important homeostatic role during dehydration -a physiological state where circulating AVP is elevated.
To explore this, mice were water restricted for 24 hours, but given unrestricted access to food. This resulted in a reduction in food intake (Fig. 6a)a critically important behavioural response to water restriction known as dehydration anorexia (Watts & Boyle, 2010). Despite the 30% reduction in food intake, plasma glucose was maintained (Fig. 6b). We hypothesised that this was due to AVP-induced glucagon secretion. In support of this hypothesis, glucagon was indeed elevated in these animals following 24 hour water restriction (Fig. 6c). In contrast, when the same mice were given ad lib access to water (which would not elevate plasma AVP) but had their food consumption restricted to that consumed in the dehydration trial, blood glucose fell and plasma glucagon was not elevated (Fig. 6b, c). The elevation in plasma glucagon was due to AVP, because pre-treatment with the V1bR antagonist SSR149415 (or vehicle) during water deprivation, blunted this increase (Fig. 6d, e).

pg. 15
Therefore, in addition to serving an antidiuretic function, AVP stimulates glucagon secretion to aid the maintenance of plasma glucose in the dehydrated state (in spite of the reduction in feeding).

Insulin-induced AVP secretion may underlie counter-regulatory glucagon secretion in human, and is diminished in subjects with Type 1 Diabetes
To understand whether this physiological pathway contributes to counter-regulatory glucagon secretion in human, samples were analysed from an ongoing clinical trial (NCT03954873). In two "saline arms" of the trial, healthy volunteers were given a hypoglycaemic clamp during one visit and a euglycaemic clamp during another visit in a randomized order ( To understand whether circulating AVP could be contributing to counter-regulatory glucagon secretion in human subjects, islets isolated from human donors were studied. In human islets from 8 donors, AVPR1B was the most abundant of the vasopressin receptor family (Fig. 7e).
These data are supported both by a recent meta-analysis of single-cell RNA-seq data from human donors (Mawla & Huising, 2019), and bulk sequencing of human islet fractions (Nica et al., 2013). In human islets from 9 donors, AVP resulted in an increase in glucagon secretion (Fig. 7f). Finally, AVP increased Ca 2+ (Fluo-4) activity in islets isolated from 5 human donors (Fig. 7g-h).

pg. 16
In T1D, deficiency of the secretory response of glucagon to hypoglycaemia is an early (< 2 years of onset) acquired abnormality of counter-regulation and leads to severe hypoglycaemia (Cryer, 2002). We therefore measured copeptin and glucagon during hypoglycaemic clamps in subjects with T1D and healthy controls from two previously published studies conducted on male participants (Christensen et al., 2011;Christensen et al., 2015). The BMI and age of these subjects did not differ (Supplementary Table 1). At 60 minutes, the achieved glucose clamp was similar between the subjects ( However, copeptin was significantly elevated in controls but not subjects with T1D (Fig. 8b).
Insulin-induced hypoglycaemia failed to increase circulating glucagon in all subjects with T1D ( Fig. 8d). Finally, a correlation of copeptin and glucagon demonstrated that control subjects either had a greater change in copeptin in response to hypoglycaemia, or a similar change in copeptin but this resulted in a larger increase in plasma glucagon (Fig. 8d).

Discussion
In this study, we report that AVP is an important regulator of glucagon secretion in vivo. This axis between the pituitary and the alpha-cell appears to be important for maintaining plasma glucose during dehydration as well as mediating the classical counter-regulatory response to hypoglycaemia. It would therefore appear that AVP regulates multiple physiological parameters that facilitate survival during stress, including water retention, blood pressure and plasma glucose.

Is AVP mediating glucagon secretion during counter-regulation?
The ability of exogenous AVP and AVP analogues to potently stimulate glucagon secretion ex vivo and in situ has been known for some time (Dunning et al., 1984). However, the physiological role of AVP in vivo in regulating glucagon secretion has remained enigmatic because previous studies have employed in vitro or in situ methods (Dunning et al., 1984;Gao et al., 1990;Gao et al., 1992;Li et al., 1992;Yibchok-Anun et al., 2000). Furthermore, these studies have used supraphysiological concentrations of AVP (Gao et al., 1990), immortalised islet cell lines (Yibchok-Anun et al., 2000) and/or synthetic AVP analogues (Dunning et al., 1984).
Importantly, our data demonstrate that glucagon is controlled by systemic concentrations of AVP in vivo under various physiological and non-physiological challenges. Therefore, we suggest that the increase in Avpr1b expression in response to interrupted glucagon signalling  is an exploitation of an important, established regulatory pathway from the hypothalamus to the islet alpha-cell (although we recognise that the phenotype in these mice is driven by hyperaminoacidaemia).

pg. 18
We did not observe an effect of AVP on insulin secretion from isolated mouse islets, whether at low or high glucose. In contrast, earlier studies have demonstrated that AVP stimulates insulin secretion (Dunning et al., 1984;Gao et al., 1990;Gao et al., 1992;Li et al., 1992).
The physiological interpretation of these earlier studies is difficult because the V1bR is the only receptor from the vasopressin receptor family expressed in mouse islets, and its expression is restricted to alpha-cells (DiGruccio et al., 2016;Taveau et al., 2017), something we also observe. An explanation for the reported increase in insulin could lie in the recently elucidated role of glucagon as a paracrine regulator of insulin (Svendsen et al., 2018). In particular, it could be that AVP stimulates glucagon secretion so potently (especially at the supraphysiological concentrations used in these aforementioned studies; > 10 nM) that it binds to the GLP-1 receptor on beta-cells, thereby increasing insulin secretion.

A1/C1 neurons mediate the AVP response to insulin-induced hypoglycaemia
Catecholaminergic neurons in the VLM are a key component of the central counterregulatory circuit and the ability of these neurons to evoke hyperglycaemia (Ritter et al., 2000;Zhao et al., 2017;Li et al., 2018) and hyperglucagonaemia (Andrew et al., 2007) is well-established. Recent studies have clearly demonstrated that spinally-projecting C1 neurons evoke hyperglycaemia by stimulating the adrenal medulla (Zhao et al., 2017;Li et al., 2018), suggesting that the response is likely to be mediated by corticosterone and/or adrenaline. We also show that activation of A1/C1 neurons evokes hyperglycaemia, but by promoting glucagon release. The important distinction here is that the elevation of plasma glucagon cannot be explained by signalling from adrenal factors, because neither corticosterone nor adrenaline stimulated glucagon release or Ca 2+ activity in isolated mouse and human islets when applied at circulating concentrations. Neurons in the VLM have a well-documented ability to increase plasma AVP (Ross et al., 1984;Madden et al., 2006).

pg. 19
Our data strongly support the notion that the hyperglycaemic and hyperglucagonaemic effect of activating A1/C1 neurons is, at least in part, mediated by stimulation of AVP release: we show that A1/C1 neurons send functional projections to the SON, A1/C1 function is required for insulin-induced AVP neuron activation and that AVP is an important stimuli of glucagon secretion. However, we recognise that other circuits must be involved in the activation of AVP neurons. For example, the PVN is richly supplied with axons from the BNST (Sawchenko & Swanson, 1983) and hypothalamic VMN neurons -key drivers of the glucose counter-regulatory response -project to the BNST (Meek et al., 2016). Therefore a VHN-BNST circuit may also be important for driving AVP neuron activity in response to hypoglycaemia, explaining why A1/C1 inhibition only partially prevented the activation of AVP neurons. In addition, AVP can stimulate glycogen breakdown from the liver (Hems & Whitton, 1973). Therefore it is likely that part of the increase in plasma glucose following A1/C1 neuron stimulation is mediated by the action of AVP on the liver. While there may be direct effects of AVP on hepatic glucose production, the findings presented here demonstrate that inhibition of either AVP signalling or glucagon secretion during stimulation of A1/C1 neurons abolishes the increase in plasma glucose.
The signal driving AVP neuron activation is unlikely to be due to a direct action of insulin on AVP or A1/C1 neurons, because 2-DG caused a similar activation. The widely accepted view is that the activation of A1/C1 neurons (and consequently AVP neurons) during hypoglycaemia depends on peripheral glucose sensing at multiple sites, including the hepatic portal system (Marty et al., 2007;Verberne et al., 2014). The exact location of the glucose sensing is still controversial. Here, we show that activity of AVP neurons is increased at ~4.9 mM glucose; while it is unknown whether this threshold is sufficient to directly activate A1/C1 neurons, vagal afferents in the hepatic portal veinwhich convey vagal sensory pg. 20 information to the A1/C1 neurons via projections to the nucleus of the solitary tract (Verberne et al., 2014) -are responsive to alterations in glucose at this threshold (Niijima, 1982). We therefore suggest that activation of A1/C1 neurons and the subsequent increase in AVP release is critical for counter-regulatory glucagon release during the early stages of hypoglycaemia (4-5 mM glucose), when intrinsic islet mechanisms are yet to be recruited.

Relevance of findings to diabetes and diabetes complications
The present findings may have clinical relevance. Insulin-induced hypoglycaemia is a major safety concern with diabetic patients (Cryer, 2002). Patients with type 1 diabetes acquire early abnormalities in their counter-regulatory response, putting them at increased risk of hypoglycaemia (McCrimmon & Sherwin, 2010). We found that hypoglycaemia fails to stimulate glucagon secretion in our cohort of T1D patients. This is a known, early acquired abnormality of counter-regulation in type 1 diabetes (Siafarikas et al., 2012), which may in part be explained by the major structural and functional changes that occur to the islets in T1D. However, we also found that insulin-induced copeptin secretion was reduced, with some subjects exhibiting no elevation in copeptin. Therefore, part of this defective glucagon secretion may be due to an insufficient insulin-induced copeptin response. The cause for this change is unknown, but we speculate that recurrent hypoglycaemia in patients with T1D may result in changes in glucose and/or insulin sensitivity in the A1/C1 region. Regardless of the brain region involved in this attenuated copeptin response, monitoring of copeptin may prove an important tool for stratification of hypoglycaemia risk in patients with type 1 diabetes.

Data availability statement
The authors declare that all data supporting the findings of this study are available within the article and its Supplementary Information or from the lead author on reasonable request.

Declaration of interests
The authors declare no competing interests.
pg. 23  c) This analysis was conducted for each trial. Grouped analysis as depicted in b) of the glucose concentration at which the GCaMP6 signal crosses >2 SD from baseline, crosses > 3 SD from baseline and first exhibits a peak, following an insulin tolerance test. n=2 mice, 6 trials. One-way ANOVA, p<0.05=*. d) Grouped analysis as depicted in b) of the delay (from insulin injection) taken for the GCaMP6 signal to cross >2 SD from baseline, cross > 3 SD from baseline and first exhibits a peak. n=2 mice, 6 trials. One-way ANOVA, p<0.05=*. e) Body temperature following insulin injection (dashed line). Mean ± SEM of six trials in n=2 mice. Dotted line indicates (average) time GCaMP6 activity first exhibits a peak, as per d). c) Same as a) but for a larger dose of insulin (0.75 U/kg). d) Same as b) but for a larger dose of insulin (0.75 U/kg). e) Plasma glucagon following an ITT (injection at 0 min) in Avpr1b -/mice and littermate controls (Avpr1b +/+ ). Two-way RM ANOVA with Sidak's multiple comparisons test. Avpr1b -/vs. Avpr1b +/+ ; p<0.001= † † † (30 minutes). 0 vs. 30 minutes; p<0.001 (Avpr1b +/+ ); p=0.097 (Avpr1b -/-). Time, p<0.0001; Genotype, p=0.008; Interaction, p=0.009. n=8-9 mice. Note that even after removal of the two high plasma glucagon samples at 30 minutes in Avpr1b +/+ mice, the glucagon is still higher than in Avpr1b -/mice (p=0.0028). Interaction, p<0.0001. N=6 mice. c) As in (b), but plasma copeptin 30 minutes following saline (-) or CNO (+). N=12 mice. Mann Whitney t-test (p=0.0025). N=18 mice. Removal of < 1 pg/ml measurements (N=3 in saline treatment) yields p=0.007.   subjected to a 24 hour water restriction trial, wherein food consumption during was monitored. During the second trial (food restriction trial), the same mice were given unrestricted access to water, but had the same quantity of food available consumed in the first (water restriction) trial. One-way RM ANOVA; Dunnett's post-hoc, p<0.05=*. N=8 wild-type mice.     pg. 36 i) AVP (10 µg/kg, i.p.) was injected into wild-type mice and blood glucose was measured with glucose test strips. Two-way RM ANOVA with Tukey's (within treatments) and Sidak's (between treatments) multiple comparison. AVP injection caused an increase in plasma glucose at 15 min compared to 0 min (p=0.023, *).

pg. 25
Vehicle did not increase plasma glucose (all p>0.2). At 15 mins, blood glucose for AVP treatment was significantly different to saline treatment (p=0.038, †). j) Same cohort as in (e), but plasma glucagon measurements. Two-way RM ANOVA; p<0.05=*.

Supplementary Figure 4: Expression of the vasopressin 1b receptor in mouse
a) mRNA expression of Avpr family in mouse heart, kidney, islets and adrenal glands.
Samples from N=3 wild-type mice, each run in triplicate. Calculated with the Pfaffl method, using Actb as the reference gene. b) mRNA expression in sorted islet cells. Fractions were sorted from mice with a fluorescent reporter (RFP) in alpha-cells (Gcg Cre+ -RFP mice) into an alpha-cell fraction (RFP(+)) and non-alpha-cell fraction (RFP(-)). Data from 4 sorts from N=4 Gcg Cre+ -RFP mice. Ratio paired t-test; p<0.05=*, p<0.01=**. c) Same data as in c). mRNA expression in alpha-cells (RFP(+) fraction) represented as fold of RFP(-) expression. Scale = log10. All data represented as mean ± SEM.

Animals
All animals were kept in a specific pathogen-free (SPF) facility under a 12:12 hour light:dark cycle at 22 °C, with unrestricted access to standard rodent chow and water. C57BL/6J mice used in this study are referred to as wild-type mice. To generate alpha-cell specific expression of the genetically-encoded Ca 2+ sensor GCaMP3, mice carrying Cre recombinase under the control of the proglucagon promoter (Gcg Cre+ mice) were crossed with mice with a floxed green calmodulin (GCaMP3) Ca 2+ indicator in the ROSA26 locus (The Jackson Laboratory).
These mice are referred to as Gcg-GCaMP3 mice. To generate mice expressing RFP in alphacells, Gcg Cre+ were crossed with mice containing a floxed tandem-dimer red fluorescent pg. 43 protein (tdFRP) in the ROSA26 locus (Gcg-RFP mice). Both of these mouse models were kept on a C57BL/6J background. Other transgenic mouse strains usednamely, Avp ires-Cre+ (Pei et al., 2014), Th Cre+ (The Jackson Laboratory), Dbh flp+ (MMRCC) and Avp GFP (MMRCC) -were heterozygous for the transgene and maintained on a mixed background.

Isolation of mouse islets
Mice of both sex and 11-16 weeks of age were killed by cervical dislocation (UK Schedule 1 procedure). Pancreatic islets were isolated by liberase digestion followed by manual picking.

Patch-clamp electrophysiology in islets
Mouse islets were used for patch-clamp electrophysiological recordings. These recordings (in intact islets) were performed at 33-34 ⁰C using an EPC-10 patch-clamp amplifier and PatchMaster software (HEKA Electronics, Lambrecht/Pfalz, Germany). Unless otherwise stated, recordings were made in 3 mM glucose, to mimic hypoglycaemic conditions in mice.
Currents were filtered at 2.9 kHz and digitized at > 10 kHz. A new islet was used for each recording. Membrane potential ( ) recordings were conducted using the perforated patchclamp technique, as previously described (Briant et al., 2018  transients were then automatically detected using the built in peak-find algorithm; the amplitude of peaks to be detected was dependent on the SNR but was typically > 20% of the maximal signal intensity. Following this, frequency of Ca 2+ transients could be determined.

GCaMP3 imaging in mouse islets
For plotting Ca 2+ data, the data was imported into MATLAB.

DAG measurements in mouse islets
The effects of AVP on the intracellular diacylglycerol concentration (DAG) in pancreatic islet cells was studied using a recombinant circularly permutated probe, Upward DAG was excited at 488 nm and fluorescence emission collected at 530 nm. The pinhole diameter was kept constant, and frames of 1388x1040 pixels were taken every 3 sec. The mean intensity for each islet was determined by manually drawing an ROI around the islet in ImageJ. Data analysis and representation was performed with MATLAB. All data was processed using a moving average filter function (smooth) with a span of 20 mins, minimum subtracted and then normalised to maximum signal intensity in the time-series. AUC was calculated using the trapz function and then divided by the length of the condition.

Pancreatic islet isolation, transplantation and in vivo imaging of islets implanted into the anterior chamber of the eye (ACE).
Pancreatic islets from Gcg-GCaMP3 mice were isolated and cultured as described above. For transplantation, 10-20 islets were aspirated with a 27-gauge blunt eye cannula (BeaverVisitec, UK) connected to a 100 µl Hamilton syringe (Hamilton) via 0.4-mm polyethylene tubing (Portex Limited). Prior to surgery, mice (C57BL6/J) were anaesthetised with 2-4% isoflurane (Zoetis) and placed in a stereotactic frame. The cornea was incised near the junction with the sclera, then the blunt cannula (pre-loaded with islets) was inserted into the ACE and islets were expelled (average injection volume 20 μl for 10 islets). Carprofen

pg. 47
(Bayer, UK) and eye ointment were administered post-surgery. A minimum of four weeks was allowed for full implantation before imaging. Imaging sessions were performed with the mouse held in a stereotactic frame and the eye gently retracted, with the animal maintained under 2-4% isoflurane anaesthesia. All imaging experiments were conducted using a spinning disk confocal microscope (Nikon Eclipse Ti, Crest spinning disk, 20x water dipping 1.0 NA objective). The signal from GCaMP3 (ex. 488 nm, em. 525±25 nm) was monitored at 3 Hz for up to 20 min. After a baseline recording, mice received a bolus of AVP (10 µg/kg) i.v.
(tail vein). Data were imported into ImageJ for initial movement correction (conducted with the StackReg and TurboReg plugins) and ROI selection. Analysis was then conducted in MATLAB.

Hormone secretion measurements from mouse and human islets
Islets, from human donors or isolated from wild-type mice, were incubated for 1 h in RPMI or DMEM supplemented with 7.5 mM glucose in a cell culture incubator. Size-matched batches of 15-20 islets were pre-incubated in 0.2 ml KRB with 2 mg/ml BSA (S6003, Sigma-Aldrich) and 3 mM glucose for 1 hour in a water-bath at 37 ⁰C. Following this islets were statically subjected to 0.2 ml KRB with 2 mg/ml BSA with the condition (e.g. 10 pM AVP) for 1 hour. After each incubation, the supernatant was removed and kept, and 0.1 ml of acid:etoh (1:15) was added to the islets. Both of these were then stored at -80 ⁰C. Each condition was repeated in at least triplicates.
Glucagon and insulin measurements in supernatants and content measurements were performed using a dual mouse insulin/glucagon assay system (Meso Scale Discovery, MD, U.S.A.) according to the protocol provided.

Hormone secretion measurements in the perfused mouse pancreas
Dynamic measurements of glucagon were performed using the in situ perfused mouse pancreas. Briefly, the aorta was cannulated by ligating above the coeliac artery and below the superior mesenteric artery, and the pancreas was perfused with KRB at a rate of ~0.45ml/min using an Ismatec Reglo Digital MS2/12 peristaltic pump. The KRB solution was maintained at 37 ⁰C with a Warner Instruments temperature control unit TC-32 4B in conjunction with a tube heater (Warner Instruments P/N 64-0102) and a Harvard Apparatus heated rodent operating table. The effluent was collected by cannulating the portal vein and using a Teledyne ISCO Foxy R1 fraction collector. The pancreas was first perfused for 10 min with 3 mM glucose before commencing the experiment to establish the basal rate of secretion.
Glucagon measurements in collected effluent were performed using RIA.

Flow cytometry of islet cells (FACS), RNA extraction, cDNA synthesis and quantitative PCR
The expression of the AVPR gene family was analysed in tissues from 12-week old C57BL6/J mice (3 mice) and pancreatic islets from human donors (2 samples, each comprised of pooled islet cDNA from 7 and 8 donors, respectively). Total RNA was isolated using a combination of TRIzol and PureLink RNA Mini Kit (Ambion, Thermofisher Scientific) with incorporated DNase treatment.
Pancreatic islets from Gcg-RFP mice were isolated and then dissociated into single cells by trypsin digestion and mechanical dissociation. Single cells were passed through a MoFlo Legacy (Beckman Coulter). Cells were purified by combining several narrow gates. Forward and side scatter were used to isolate small cells and to exclude cell debris. Cells were then gated on pulse width to exclude doublets or triplets. RFP + cells were excited with a 488 nm pg. 49 laser and the fluorescent signal was detected through a 580/30 bandpass filter (i.e. in the range 565-595 nm). RFP-negative cells were collected in parallel. The levels of gene expression in the RFP + and in the RFP -FAC-sorted fractions were determined using real-time quantitative PCR (qPCR). RNA from FACS-sorted islet cells was isolated using RNeasy Micro Kit (Qiagen). cDNA was synthesized using the High Capacity RNA-to-cDNA kit (Applied Biosystems, Thermofisher Scientific). Real time qPCR was performed using SYBR Green detection and gene specific QuantiTect Primer Assays (Qiagen) on a 7900HT Applied Biosystems analyser. All reactions were run in triplicates. Relative expression was calculated using ΔCt method Actb as a reference gene.

Fibre photometry experiments and analysis of photometry data.
In vivo fibre photometry was conducted as previously described (Mandelblat-Cerf et al. (2017)). A fibre optic cable (1-m long, metal ferrule, 400 µm diameter; Doric Lenses) was attached to the implanted optic cannula with zirconia sleeves (Doric Lenses). Laser light (473 nm) was focused on the opposite end of the fibre optic cable to titrate the light intensity entering the brain to 0.1-0.2 mW. Emitted light was passed through a dichroic mirror (Di02-R488-25x36, Semrock) and GFP emission filter (FF03-525/50-25, Semrock), before being focused onto a sensitive photodetector (Newport part #2151). The GCaMP6 signal was passed through a low-pass filter (50 Hz), and digitized at 1 KHz using a National Instruments data acquisition card and MATLAB software.

pg. 51
All experiments were conducted in the home-cage in freely moving mice. Animals prepared for in vivo fibre photometry experiments (outlined above), were subjected to an ITT or 2DG injection after overnight fasting. Prior to insulin or 2DG injection, a period of GCaMP6s activity was recorded (3 min) to establish baseline activity. Insulin (i.p. 2 U/kg), 2DG (i.p. 500mg/kg), or saline vehicle was then administered, and GCaMP6 activity recorded for a further 40 min. In some experiments, mice were pre-treated i.p. with CNO (1mg/kg) or saline, 30 minutes prior to insulin or 2DG. The recorded data was exported and then imported into MATLAB for analysis. Fluorescent traces were down-sampled to 1 Hz and the signal was normalised to the baseline (F0 mean activity during baseline activity), with 100% signal being defined as the maximum signal in the entire trace (excluding the injection artefact).
Following the ITT, the signal was binned (1 min) and a mean for each bin calculated. These binned signals were compared to baseline signal using a one-way RM ANOVA.

Surgery for continuous glucose monitoring
Animals that have undergone fibre photometry surgeries (3 weeks prior) were anesthetized and maintained with isoflurane. Once mice were fully anesthetized, the ventral abdomen and underside of the neck were shaved and disinfected. Animals were placed on their backs on a heated surgical surface. For transmitter implantation, a ventral midline abdomen incision was made and the abdominal wall was incised. The transmitter was placed in the abdominal cavity with the lead exiting cranially and the sensor and connector board exteriorized. The incision was sutured incorporating the suture rib into the closure. For glucose probe implantation, a midline neck incision was performed and the left common carotid artery was isolated. The vessel was then perforated and the sensor of the glucose probe (HD-XG, Data Sciences International) was advanced into the artery towards the heart, within a final placement in the aortic arch. Once in place, the catheter was secured by tying the suture pg. 52 around the catheter and vessel, and overlying opening in tissue was closed. Mice were kept warm on a heating pad and monitored closely until fully recovered from anaesthesia.

Simultaneous AVP fiber photometry and continuous glucose monitoring
All experiments were conducted in the home-cage in freely moving mice. Animals prepared for in vivo fibre photometry and continuous glucose monitoring (outlined above), were subjected to an ITT after overnight fasting. After establishing > 3 min of baseline activity, insulin (i.p. 1 or 1.5 U/kg) or saline vehicle was administered. GCaMP6s activity, blood glucose, and body temperature were recorded throughout 2 h of experiment. Each recording was separated by at least 48h. GCaMP6s recording was performed as described above. Blood glucose and body temperature were acquired using Dataquest A.R.T. 4.36 system and analysed using MATLAB. Calibration of HD-XG device was performed as per manufacturer's manual.

In vivo measurements of plasma glucose, glucagon and copeptin
Samples for blood glucose and plasma glucagon measurements were taken from mice in response to different metabolic challenges (described in detail below). Both sexes were used for these experiments. Blood glucose was measured with an Accu-Chek Aviva (Roche Diagnostic, UK) and OneTouch Ultra (LifeScan, UK). Plasma copeptin was measured using an ELISA (MyBioSource, USA and Neo Scientific, USA).

Insulin tolerance test
Mice were restrained and a tail vein sample of blood was used to measure fed plasma glucose. A further sample was extracted into EDTA coated tubes for glucagon measurements.
Aprotinin (1:5, 4 TIU/ml; Sigma-Aldrich, UK) was added to all blood samples. These blood pg. 53 samples were kept on ice until the end of the experiment. Mice were first administered with any necessary pre-treatment and then individually caged. Pre-treatments included SSR149415 (30 mg/kg in PBS with 5% DMSO and 5% Cremophor EL), LY2409021 (5 mg/kg in PBS with 5% DMSO), CNO (1-3 mg/kg in PBS with 5% DMSO) or the appropriate vehicle. After a 30 minute period, mice were restrained again, and blood was taken via a tail vein or submandibular bleed. This was used for blood glucose measurements, and also for glucagon.
Insulin (0.75, 1 or 1.5 U/kg) was then administered i.p., and the mice were re-caged. At regular time intervals after the insulin injection, mice were restrained and a blood sample extracted. Blood glucose was measured, and blood was taken for glucagon measurements. At the end of the experiment, blood samples were centrifuged at 2700 rpm for 10 min at 4 °C to obtain plasma. The plasma was then removed and stored at -80 °C. Plasma glucagon measurements were conducted using the 10-µl glucagon assay system (Mercodia, Upsala, Sweden), according to the manufacturer's protocol.

Fasting experiments
Wild-type mice were used for fasting experiments. Mice (8-9 weeks of age) were restrained, and blood was taken for blood glucose and glucagon measurements (as above). The mice were then home caged for the period of the fast. Food was removed at 08:30, and samples were taken at 14:30 and 16:00 (7.5 hour fast). SSR149415 (30 mg/kg) or vehicle was administered i.p. at 14:30 (90 minutes prior to termination of the fast). Blood glucagon was measured as indicated above. All animals had free access to water during the fast.

Water restriction (dehydration) experiments
Wild-type mice (8-9 weeks of age) were used for water restriction experiments. Mice were single housed one week prior to experimental manipulation. For trial 1 (water restriction),

pg. 54
blood was taken at 16:00 on day 1 and 2 for blood glucose and glucagon measurements (as above). Mice were water restricted for 24h between day 1 and 2, but had ad lib access to food. Amount of food consumed during this period was measured at the end of the trial and used for trial 2. For trial 2 (food restriction), blood was taken at 16:00 on day 1 and 2 for blood glucose and glucagon measurements. Mice had ad lib access to water between day 1 and 2, but were given the same amount of food consumed in trial 1. The two trials were separated by 2 days without any manipulation.

Glucoprivic response to 2-Deoxy-D-glucose
Wild-type mice were used for 2-Deoxy-D-glucose (2DG) experiments. The mice were single housed one week prior to experimental manipulation. On the experimental day, food was removed 4 hours prior to the experiment. 2DG (500 mg/kg) or saline vehicle was then administered i.p., and blood samples taken at regular intervals for blood glucose and plasma glucagon measurements. In some cohorts, the V1bR antagonist SSR149415 (30 mg/kg in PBS with 5% DMSO and 5% Tween 80), glucagon receptor antagonist LY240901 (5 mg/kg in PBS with 5% DMSO and 5% Tween 80) or appropriate vehicle was administered i.p. 30 minutes prior to administration of 2DG. Plasma glucagon was measured as described above.

Brain slice electrophysiology
To prepare brain slices for electrophysiological recordings, brains were removed from anesthetized mice (4-8 weeks old) and immediately placed in ice-cold cutting solution consisting of (in mM): 72 sucrose, 83 NaCl, 2.  (20-24 °C) for at least 60 min prior to recording. A single slice was placed in the recording chamber where it was continuously super-fused at a rate of 3-4 ml per min with oxygenated aCSF. Neurons were visualized with an upright microscope equipped with infrared-differential interference contrast and fluorescence optics. Borosilicate glass microelectrodes (5-7 MΩ) were filled with internal solution. All recordings were made using Multiclamp 700B amplifier, and data was filtered at 2 kHz and digitized at 10 kHz. All analysis was conducted off-line in MATLAB.
Brain slices were prepared (as above) from these mice. The SON was located by using the bifurcation of the anterior and middle cerebral arteries on the ventral surface of the brain as a ChR2-positive A1/C1 fibres, an LED light source (473 nm) was used. The blue light was focused on to the back aperture of the microscope objective, producing a wide-field exposure around the recorded cell of 1 mW. The light power at the specimen was measured using an optical power meter PM100D (ThorLabs). The light output is controlled by a programmable pulse stimulator, Master-8 (AMPI Co. Israel) and the pClamp 10.2 software (AXON Instruments).

Activation of hM3Dq with CNO in AVP neurons
The modified human M3 muscarinic receptor hM3Dq (Alexander et al., 2009)

Clamping studies in human participants
Clamping studies were conducted at Gentofte Hospital, University of Copenhagen.

Comparison of copeptin in healthy subjects undergoing a hypoglycaemic and euglycaemic clamp
Samples from the "saline arm" from 10 male subjects enrolled in an ongoing, unpublished clinical trial (https://clinicaltrials.gov/ct2/show/NCT03954873) were used to compare copeptin secretion during euglycaemia and hypoglycaemia.

pg. 58
For the study, two cannulae were inserted bilaterally into the cubital veins for infusions and blood sampling, respectively. For the euglycaemic study, participants were monitored during fasting glucose levels. For the hypoglycaemic clamp, an intravenous insulin (Actrapid; Novo Nordisk, Bagsvaerd, Denmark) infusion was initiated at time 0 min to lower plasma glucose.
Plasma glucose was measured bedside every 5 min and kept >2.2 mM. Arterialised venous blood was drawn at regular time intervals prior to and during insulin infusion.

Comparison of copeptin in subjects with T1DM and healthy controls
Samples from 20 male (N=10 control and N=10 T1DM patients) from the "saline arms" of two previously published studies (Christensen et al., 2011;Christensen et al., 2015) were used to compare copeptin secretion during a hypoglycaemic clamp between T1DM and control subjects. The samples from healthy individuals (Controls) were from Christensen et al. (2011). These 10 healthy male subjects were of age 23 ± 1 years, BMI 23 ± 0.5 kg/m 2 and HbA1c 5.5 ± 0.1%. The T1DM patient samples were from (Christensen et al., 2015). These patients were; C-peptide negative, age 26 ± 1 years, BMI 24 ± 0.5 kg/m 2 , HbA1c 7.3 ± 0.2%, positive islet cell and/or GAD-65 antibodies, treated with multiple doses of insulin (N = 9) or insulin pump (N = 1), without late diabetes complications, without hypoglycemia unawareness, and without residual β-cell function (i.e., C-peptide negative after a 5-g arginine stimulation test). For the study, a hypoglycaemic clamp was conducted as outlined above.

Measurement of copeptin, glucagon and AVP in human plasma
Copeptin in human plasma was analysed using the KRYPTOR compact PLUS (Brahms Instruments, Thermo Fisher, DE). Glucagon was measured using human glucagon ELISA from Mercodia. Plasma AVP was measured using a human AVP ELISA kit (Cusabio, China).

pg. 59
Statistical tests of data

Mathematical model of Gq-mediated Ca 2+ oscillations
The mathematical model of the processes governing Ca 2+ signalling in a single cell following activation of a metabotropic receptor (i.e. V1bR) is based on the formalism by Li and Rinzel (1994) and Lemon et al. (2003).