Peripheral insulin administration enhances the electrical activity of oxytocin and vasopressin neurones in vivo

Oxytocin neurones are involved in the regulation of energy balance through diverse central and peripheral actions and, in rats, they are potently activated by gavage of sweet substances. Here, we test the hypothesis that this activation is mediated by the central actions of insulin. We show that, in urethane‐anaesthetised rats, oxytocin cells in the supraoptic nucleus show prolonged activation after i.v. injections of insulin, and that this response is greater in fasted rats than in non‐fasted rats. Vasopressin cells are also activated, although less consistently. We also show that this activation of oxytocin cells is independent of changes in plasma glucose concentration, and is completely blocked by central (i.c.v.) administration of an insulin receptor antagonist. Finally, we replicate the previously published finding that oxytocin cells are activated by gavage of sweetened condensed milk, and show that this response too is completely blocked by central administration of an insulin receptor antagonist. We conclude that the response of oxytocin cells to gavage of sweetened condensed milk is mediated by the central actions of insulin.


| INTRODUC TI ON
Insulin is widely known for its role in glucose homeostasis on peripheral tissues, although its central effects are not yet fully elucidated. Once secreted into the circulation, insulin is transported into the brain by a saturable transport mechanism. 1,2 Both exogenous insulin administration and glucose-stimulated insulin secretion result in a progressive increase of insulin in the cerebrospinal fluid (CSF) in several species, including humans. [3][4][5][6] Accordingly, insulin concentrations in the CSF correlate with levels in plasma, although they are approximately 15-fold lower than plasma concentrations in fasted rats. 3 In the brain, regions sensitive to insulin include the hypothalamus, 7,8 which contains insulin-responsive neurones in several nuclei. [9][10][11] Amongst these, the insulin receptor (InsR) is abundantly expressed in the supraoptic nucleus (SON), [12][13][14] which exclusively contains magnocellular oxytocin and vasopressin cells, and i.p. administration of insulin induces the expression of Fos protein in parvo-and magnocellular oxytocin cells of the paraventricular nucleus in rats. 15 Explants of the hypothalamo-neurohypophysial system, including the SON and its projections to the posterior pituitary, release oxytocin and vasopressin in response to direct application of insulin, 16 and central administration of insulin increases peripheral secretion of oxytocin in mice by a direct action on oxytocin cells. 17 In addition to their classical roles in reproduction, 18,19 stress 20 and water balance, 21 oxytocin and vasopressin have roles in energy homeostasis. 22 Both central and peripheral oxytocin administration exert anorexigenic effects, increase energy expenditure and induce lipolysis. [23][24][25][26] Peripheral administration of both oxytocin and vasopressin can induce the release of insulin from the pancreas [27][28][29][30] and systemically administered oxytocin in humans (administered intranasally) has been reported to curb the meal-related increase in plasma glucose, 31 as well as to improve β-cell responsivity and glucose tolerance in healthy men. 32 Studies using well-validated radioimmunoassays in extracted plasma samples 33 indicate that patients with metabolic syndrome exhibit higher circulating oxytocin concentrations than normal individuals, 34 and patients with diabetes have higher concentrations of vasopressin, as well as copeptin (which is co-secreted with vasopressin). 35,36 In the present study, we examine whether peripheral (i.v.) administration of insulin affects the electrical activity of oxytocin and vasopressin cells in the SON of urethane-anaesthetised rats. The effect of different feeding states, and consequently different blood glucose concentrations, on these responses was also investigated.
We investigated the role of brain InsR in these responses by blocking these receptors using an InsR antagonist. Finally, we tested whether the previously reported enhanced electrical activity of oxytocin cells in response to sweet food gavage 37 was mediated by endogenous insulin release acting on brain InsRs.

| Animals
We used adult male Sprague-Dawley rats weighing 300-350 g. The rats had ad lib. access to food and water and were maintained under a 12:12 hour light/dark cycle (lights on 7.00 am) at a room temperature of 20-21°C. In most experiments, we used fasted rats to reduce the variability of blood glucose and gastric signals (induced by prior food consumption) that could affect neural activity and so, in these experiments, the food was removed overnight (~15 hours). All procedures were conducted on rats under deep terminal anaesthesia in accordance with the UK Home Office Animals Scientific Procedures Act 1986 and a project licence approved by the Ethical Committee of the University of Edinburgh.  38 was given into the third ventricle using a 31-gauge needle inserted through the median eminence; 1 nmol (4.8 µg) of S961 was injected at 1 µL min -1 . We chose a dose expected to be sufficient to block insulin receptors throughout the brain when given i.c.v., although lower than that needed to antagonise insulin actions if given peripherally. The affinity of S961 for both isoforms of the insulin receptor is close to that of insulin itself. 38 Previous studies have reported that bilateral injections of 100 ng of S961 into the arcuate nucleus block the effects of insulin microinjected into the arcuate nucleus on lumbar sympathetic nerve activity in late pregnant rats. 39 Studies using the closely related antagonist S661, which has properties indistinguishable from those of S961, indicated that peripheral doses of 30 nmol kg -1 or more are needed to block the effects of i.v. administration of 30 mmol kg -1 insulin on blood glucose levels. 38 As detailed below, the i.c.v. application of S961 in our hands had no significant effect on plasma glucose concentrations.

| Sweet condensed milk gavage
In fasted rats, a gavage tube was inserted orally into the stomach to deliver a total volume of 5 mL of sweetened condensed milk (SCM; Nestle, Vevey, Switzerland) diluted 50% v/v in distilled water (40.8 kJ, 1.68 g sugar, 0.24 g fat) at 0.16 mL min -1 .

| In vivo electrophysiology
Rats were briefly anaesthetised with isoflurane inhalation anaesthesia, and then urethane (ethyl carbamate 25% solution) was in- Design, Cambridge, UK) connected to a PC running spike2, version 7.20 (Cambridge Electronic Design). Most recordings were made from single neurones; in some experiments, the spike activity of two cells was recorded simultaneously; in these cases, the spikes were discriminated and analysed offline using the waveform function of spike2. Recordings were made between 12.00 pm and 5.00 pm (lights on 7.00 am to 7.00 pm). Rats were tested only once with insulin.
Supraoptic nucleus neurones were antidromically identified through stimulation of the pituitary stalk by matched biphasic pulses (1 ms, <1 mA peak to peak), which produce an antidromic spike at a constant latency (~10 ms) ( Figure 1A). Oxytocin cells were discriminated from continuous-firing vasopressin cells ( Figure 1B given at 20 μg kg -1 , comprising a transient excitation of oxytocin cells, and no effect or short inhibition of vasopressin cells ( Figure 1E, F). 40,41 CCK was given at the end of the experiments to identify continuously-firing cells.

| Effect of i.v. insulin
The spontaneous spiking activity of SON neurones was recorded for 20 minutes (basal activity) and for at least 60 minutes after i.v.
insulin. Blood samples (50 μL) were taken to measure glucose immediately before administration of insulin or vehicle, as well as 15, 30, 60, 90 and 120 minutes later.

| Effect of restoring circulating glucose content in insulin-responsive neurones
The basal activity of SON neurones was recorded for 20 minutes, and for another 30 minutes after i.v. insulin. Then, glucose was given i.v. and the spike activity recorded for further 30 minutes.
Blood glucose concentrations were measured before insulin, 30 minutes later (ie, before i.v. glucose) and 5 and 20 minutes after the first glucose injection. Only rats exhibiting in the last sample a blood glucose concentration within 15% of the value in the basal sample were used.

| Blockade of central InsRs
The basal spike activity of SON neurones in fasted rats was recorded for 20 minutes. Then, S961 was given i.c.v. and spike activity recorded for 15 minutes. After this, insulin was given i.v. and the spike activity recorded for another 30 minutes. Blood glucose concentrations were measured using an Accu-Chek Aviva meter (Roche Diagnostics GmbH, Mannheim, Germany) immediately before S961 injection, 15 minutes later (ie, before i.v. insulin) and 30 minutes after i.v. insulin.

| Effect of central InsR blockade on SCMstimulated activity of oxytocin cells
The basal spike activity of SON neurones was recorded for 20 minutes. Then, rats were injected i.c.v. with either vehicle or S961 and activity recorded for 10 minutes. After this, SCM was gavaged (over 30 minutes) and spike activity recorded for 1 hour.
Blood samples (300 μL) were taken immediately before the i.c.v. injection, 10 minutes later (ie, before SCM gavage) and at 30 and 60 minutes after the start of gavage. Blood glucose concentrations were measured immediately after sampling; then, samples were centrifuged in EDTA-coated tubes, and plasma collected and stored at −80°C for insulin measurements using a rat/mouse insulin ELISA kit (catalogue no. EZRMI-13K; EMD Millipore, Burlington, MA, USA). When plotted this way, a negative exponential distribution (the distribution characteristic of random events) becomes a constant 'hazard' proportional to the average firing rate. Deviation from this then become interpretable as periods of decreased or increased excitability. Consensus hazard functions were calculated from the means of hazard functions.

| Statistical analysis
Data were analysed using Prism, version 6 (GraphPad Software Inc., San Diego, CA, USA). Responses to insulin were analysed by comparing the mean firing rate in the 60-minute after insulin with the (basal) firing rate over the 20-minute control period. The changes were compared using a two-tailed Wilcoxon signed-rank test. The activity of phasic cells was analysed in spike2; detection of a burst of activity was defined by spike activity lasting at least 5 seconds and containing >20 spikes followed by >5 seconds of spike silence between bursts. The mean burst duration, interburst interval and activity quotient (percentage of active time over the total time) over the 20-minute basal and 60 minutes after insulin were compared using Wilcoxon matched-pairs signed-rank test.
The effect of glucose on insulin-responsive cells was analysed by comparing the mean change in firing rate (spikes s -1 in 10-minute bins) before and after glucose (ie, 0-30 minutes vs 30-60 minutes) using Wilcoxon matched-pairs signed-rank test.
The effect of blockade of central InsRs was analysed by testing whether the mean change in firing rate in the 15-minute after S961 injection was significantly different from 0 (ie, from the basal rate) using a two-tailed Wilcoxon signed-rank test. Then, the mean change in firing rate over 30-minute after insulin was compared with the firing rate in the 15 minutes after S961 using a two-tailed Wilcoxon signed-rank test. One-way ANOVA followed by a post-hoc Bonferroni test was used to compare glucose profiles.
The mean change in firing rate over 60 minutes and the glucose profiles between fasted and non-fasted rats were compared using two-tailed Mann-Whitney test and two-way ANOVA followed by post-hoc Bonferroni multiple comparison tests, respectively. We also compared the change in firing rate to determine whether different treatments affect the responses of SON neurones to insulin using two-way ANOVA followed by a post-hoc Bonferroni test.
The effect of prior blockade of central InsRs on SCM-induced activity was analysed using a two-tailed Mann-Whitney test comparing the mean change in firing rate over 60 minutes between i.c.v.
control-and S961-treated rats. The change in firing rate (in 10-minute bins), blood glucose concentrations and plasma insulin content between the two groups were compared using two-way ANOVA, followed by a post-hoc Bonferroni test.
All data are reported as the mean ± SEM. P < 0.05 was considered statistically significant, unless otherwise stated. Recordings were made from 10 oxytocin cells in 10 fasted rats and from 10 cells in nine non-fasted rats (including one double recording). In non-fasted rats, the mean ± SEM (range) basal firing rate of 2.5 ± 0.4 (0.7-4.1) spikes s -1 increased by 0.9 ± 0.3 (0.1-2.5) spikes s -1 (averaged over the 60 minutes after i.v. insulin; P = 0.002, Wilcoxon signed-rank test) (Figure 2A,B). In fasted rats, oxytocin cells responded more strongly ( Figure 2C

| Vasopressin cells
In six fasted rats, recordings were made from 10 vasopressin cells signed-rank test) ( Figure 2D). In the eight phasic cells, insulin increased the burst duration (from 73 ± 17 seconds to 328 ± 137 seconds). In these cells, the interburst period was reduced (from 64 ± 26 to 61 ± 18 seconds); the activity quotient was increased from 0.6 ± 0.1 to 0.7 ± 0.1, and the intraburst frequency was increased from 6.6 ± 0.8 to 7.2 ± 0.7 spikes s -1 .
Eight of 10 vasopressin cells in fasted rats and nine of sixteen vasopressin cells in non-fasted rats increased their activity by more than 10% after i.v. insulin, and the mean response of all vasopressin cells tested was greater in fasted rats than in non-fasted rats, although this did not reach statistical significance (Mann-Whitney U test, P = 0.63).

| Hazard functions
In oxytocin cells, the hazard functions conformed to the profile previously reported as typical of oxytocin cells, reflecting a prolonged post-spike refractoriness of 30-50 ms followed by a stable plateau of excitability. 42 Insulin did not affect the duration of the post-spike refractoriness but elevated the plateau level of excitability ( Figure 2E).
In vasopressin cells, the hazard functions also conformed to the profile previously reported as typical of vasopressin cells, reflecting a post-spike refractoriness of 20-50 ms followed by a period of hyperexcitability (reflecting a depolarising afterpotential) before reaching a stable plateau of excitability. 42 Insulin did not affect the duration of the post-spike refractoriness or the plateau level of excitability but enhanced the post-spike hyperexcitability ( Figure 2F).

| Blockade of central InsRs before i.v. insulin
To test whether the activation of SON neurones by insulin involves  F I G U R E 3 Effect of i.v. glucose infusion in insulin-responsive neurones in fasted rats. A, Blood glucose concentrations were lowered after i.v. insulin, but not i.v. vehicle, in fasted (F) and non-fasted (NF) rats (*P < 0.05, Two-way ANOVA followed by a Bonferroni post-hoc test). B, Blood glucose concentrations after i.v. insulin and after i.v. insulin, and 5% glucose solution injections (arrows: as required) of all 10 rats where neuronal activity was recorded. After 30-minutes of i.v. insulin, the glucose concentration was significantly lower compared to all other blood samples (one-way ANOVA for repeated measures; ***P < 0.001, Bonferroni post-hoc test) with no significant differences between other samples. B, Blood glucose concentrations after i.v. insulin, and 5% glucose solution injections (arrows: 400 µL, *300 µL, *100 µL; * if required) of all animals (n = 10) where neural activity was recorded. C, D, After insulin, no significant differences in firing rate of (C) oxytocin and (D) vasopressin cells in glucose-treated rats were detected compared to non-glucose-treated fasted rats. Data are the mean ± SEM

| Effect of blockade of central InsRs on oxytocin spike activity induced by SCM gavage
Gavage of food rich in sugars, but not fat, results in a rise of blood glucose and insulin plasma concentration and a progressive increase in the electrical activity of oxytocin cells. 37 Here, we tested whether this involves brain InsRs.
Both vehicle-and S961-injected rats exhibited a significant increase in both blood glucose concentration and plasma insulin concentration following SCM gavage ( Figure 5A) with no significant differences between groups (glucose: two-way ANOVA for repeated measures: interaction, In vehicle-injected rats, as expected, 37   to the posterior pituitary, also release large amounts of oxytocin within the brain from their dendrites. This dendritic release is likely to have important effects at relatively local sites, including the amygdala and the ventromedial nucleus of the hypothalamus where abundant oxytocin receptors are expressed but which contain only sparse oxytocin fibres. 23,45 In addition, it has recently become apparent that many magnocellular neurones have extensive axonal projections to diverse brain regions, including notably to the nucleus accumbens. 46 In the present study, systemic administration of insulin increased the electrical activity of both oxytocin and vasopressin SON cells, consistent with previous reports in humans and rats that insulin increases secretion of oxytocin and vasopressin. [47][48][49] As originally conceived in the design of the present experiments, the dose and route of insulin administration followed the conventional design of insulin tolerance tests 50 to produce an acute maintained hypoglycaemia. This bolus injection raises peripheral insulin concentrations above the normal physiological range, which are then rapidly cleared. The evolution of oxytocin cell activity after insulin injections thus mirrored neither the changes in plasma glucose, nor the expected changes in peripheral insulin concentration. Insulin crosses the bloodbrain barrier by an active transport mechanism that is saturated: at least 50% of maximal transport capacity is reached at euglycemic levels of plasma insulin; thus, supraphysiological levels of insulin in the plasma have little additional effect on insulin penetration into the brain beyond that seen at high physiological levels. 1,51 Thus, the expected evolution of CNS insulin following i.v. bolus injection is a progressive rise when peripheral levels are elevated above normal levels, possibly explaining the progressive rise in oxytocin cell activity.

| D ISCUSS I ON
Brain InsRs play an important role in the control of energy balance as shown by selective genetically-induced decreased expression of brain InsRs which is linked to a peripheral metabolic alterations, including increased food intake, fat and body weight, as well as increased glucose and insulin resistance in rodents. 52,53 Moreover, injection of the InsR antagonist S961 into the ventromedial nucleus increases blood glucose concentration in rats. 54 In the present study, central adminis- In non-fasted rats, which exhibited a more pronounced hyperglycaemia than fasted rats, the responses of oxytocin cells were less prominent than in fasted rats. This may reflect InsR desensitisation in oxytocin cells, similarly to that shown in skeletal muscle in vivo 56 and fibroblasts in vitro, 57 where acute exposition to high glucose concentration reduced insulin-stimulated glucose uptake and impaired InsR intracellular signalling, respectively. Alternatively, because, in fasted animals, blood glucose concentrations fell following insulin administration to concentrations lower than immediately after anaesthesia, this might stimulate the hypothalamic-pituitary-adrenal as occurs in the insulin tolerance test, 55 potentiating the release of oxytocin (and vasopressin).
A recent study 17 raised a question about the capacity of SON neurones to respond to insulin administration because insulin given i.c.v. induced an increase in Fos expression after 90-minutes in 13% of the PVN, but not SON, oxytocin cells compared to control mice.
Nevertheless, SON neurones appear to be intrinsically sensitive to insulin and glucose because they express InsR 12-14 and the enzyme glucokinase, 58 a marker for glucose sensing. Moreover, vasopressin and oxytocin are released from SON explants in the presence of medium containing glucose and insulin. 16 Although Fos protein has been widely used as a marker for neuronal activation, its lack of expression does not necessarily exclude changes in neural activity as observed in some conditions, and increased spike activity is not invariably linked to Fos expression. 59,60 It appears that insulin might not induce the expected rapid expression of Fos (ie, 60-90 minutes) because Griffond et al 15 reported that, at 1 hour after insulin i.p.
(20 mg kg -1 ), there was little expression of Fos in PVN oxytocin cells.
A limitation of the present study is that it involved urethane-anesthetised rats. Urethane has long been the anaesthetic of choice for SON electrophysiological recordings because it provides a deep long-lasting stable anaesthesia compatible with transpharyngeal surgery without affecting the physiological responses of SON neurones. 40 However, urethane raises blood glucose concentrations 61,62 by increasing sympathetic tone 63 and consequently increasing gluconeogenesis. Thus, blood glucose concentrations in both non-fasted and fasted anaesthetised rats were higher than in conscious Sprague-Dawley rats. 64 However, they were lower in fasted rats than in non-fasted rats, and changed in the expected manner in response to i.v. insulin.

ACK N OWLED G EM ENTS
This work was supported by the BBSRC (BB/S000224/1).

CO N FLI C T O F I NTE R E S T S
The authors declare that they have no conflicts of interest.

AUTH O R CO NTR I B UTI O N S
The study was designed by GL and performed by LP. LP and GL analysed the data and wrote the paper together. GL had full access to all the data and analyses, and takes responsibility for the integrity of the data and the accuracy of the analyses.

DATA AVA I L A B I L I T Y
The datasets generated during and/or analysed during the present study are available from the corresponding author upon reasonable request.