Impact of chemogenetic activation of dorsal vagal complex astrocytes in mice on adaptive glucoregulatory responses

The dorsal vagal complex (DVC) regulates diverse aspects of physiology including food intake and blood glucose homeostasis. Astrocytes play an active role in regulating DVC function and, by extension, physiological parameters. DVC astrocytes in ex vivo slices respond to low tissue glucose. The response of neurons to low glucose is conditional on intact astrocyte signalling in slice preparations, suggesting astrocytes are primary sensors of glucose deprivation (glucoprivation). Based on these published findings we hypothesised that in vivo DVC astrocyte manipulation with chemogenetics would be sufficient to alter physiological responses that control blood glucose. We found that 2‐h after systemic 2‐DG‐induced glucoprivation there were no observable changes in morphology of glial fibrillary acidic protein (GFAP)‐immunoreactive DVC cells, specifically those in the nucleus of the solitary tract (NTS). Chemogenetic activation of DVC astrocytes was sufficient to suppress nocturnal food intake by reducing both meal size and meal number and this manipulation also suppressed 2‐DG‐induced glucoprivic food intake. Chemogenetic activation of DVC astrocytes did not increase basal blood glucose nor protect against insulin‐induced hypoglycaemia. In male mice, chemogenetic DVC astrocyte activation did not alter glucose tolerance. In female mice, the initial glucose excursion was reduced in a glucose tolerance test, suggesting enhanced glucose absorption. Based on our data and published work, we propose that DVC astrocytes may play an indispensable homeostatic role, that is, are necessary to maintain the function of glucoregulatory neuronal circuitry, but alone their bulk activation is not sufficient to result in adaptive glucoregulatory responses. It is possible that there are state‐dependent effects and/or DVC astrocyte subsets that have this specialised role, but this was unresolvable using the experimental approaches employed here.

conditional on intact astrocyte signalling in slice preparations, suggesting astrocytes are primary sensors of glucose deprivation (glucoprivation). Based on these published findings we hypothesised that in vivo DVC astrocyte manipulation with chemogenetics would be sufficient to alter physiological responses that control blood glucose. We found that 2-h after systemic 2-DG-induced glucoprivation there were no observable changes in morphology of glial fibrillary acidic protein (GFAP)-immunoreactive DVC cells, specifically those in the nucleus of the solitary tract (NTS). Chemogenetic activation of DVC astrocytes was sufficient to suppress nocturnal food intake by reducing both meal size and meal number and this manipulation also suppressed 2-DG-induced glucoprivic food intake. Chemogenetic activation of DVC astrocytes did not increase basal blood glucose nor protect against insulin-induced hypoglycaemia. In male mice, chemogenetic DVC astrocyte activation did not alter glucose tolerance. In female mice, the initial glucose excursion was reduced in a glucose tolerance test, suggesting enhanced glucose absorption. Based on our data and published work, we propose that DVC astrocytes may play an indispensable homeostatic role, that is, are necessary to maintain the function of glucoregulatory neuronal circuitry, but alone their bulk activation is not sufficient to result in adaptive glucoregulatory responses. It is possible that there are state-dependent effects and/or DVC astrocyte subsets that have this specialised role, but this was unresolvable using the experimental approaches employed here.
K E Y W O R D S astrocyte, brainstem, feeding, glucoprivic, hypoglycaemia

| INTRODUCTION
The dorsal vagal complex (DVC) is a brain centre involved in regulating many facets of homeostasis including food intake, digestion, cardiovascular reflexes, respiratory reflexes and glucose homeostasis. [1][2][3][4][5] Situated in the brainstem, the DVC consists of three nuclei: the area postrema (AP; a circumventricular organ), the nucleus of the solitary tract (NTS; a neuronal hub integrating input from the vagus nerves), and the dorsal motor nucleus of the vagus (DMX; containing the cell bodies of preganglionic parasympathetic neurons that comprise the efferent branch of the vagus nerves). Of these constituent nuclei the NTS is the primary sensory integrator of signals from the periphery conveyed by the vagus nerves.
Neurons in the brainstem, particularly those in the NTS and ventrolateral medulla (VLM), are proposed to be intrinsic glucose sensors, sensing deviations in blood glucose (both hypo-and hyperglycaemia) and mediating physiological responses, including glucoprivic feeding and hormone secretion, to restore blood glucose. [5][6][7][8] Local tissue glucoprivation in the NTS and/or VLM induced by injection of 2-deoxyglucose (2-DG; an inhibitor of glycolysis) into the brain is sufficient to induce counter-regulatory food intake in rats, suggesting that cells in these sites can directly sense changes in glucose. 9 The potential importance of direct brainstem glucose sensing as part of counter-regulatory feeding responses is further supported by the observation that subdiaphragmatic vagotomy does not abolish glucoprivic stimulus-induced food intake in rats. 10 Within the brainstem, catecholaminergic neurons, identified by their expression of tyrosine hydroxylase (TH) and/or dopamine betahydroxylase (DBH), are a key component of the circuitry underlying glucoprivic responses since their ablation eliminates glucoprivic feeding in rats 11 and their chemogenetic inhibition attenuates glucoprivic feeding in mice. 12 In addition, GABAergic NTS neurons have glucoregulatory capacity since their optogenetic or chemogenetic activation increases blood glucose and glucagon in mice, suggesting the ability of these neurons to drive homeostatic endocrine responses to hypoglycaemia. 13,14 Although NTS TH neurons are demonstrably required to generate glucoprivic feeding, 11,12 it is not clear whether they sense glucose levels cell-autonomously or instead are downstream of other glucose-sensing cells. Indeed, astrocytes in the NTS have been proposed as the primary cell type underlying low-glucose detection. 15 In support of this hypothesis, NTS astrocytes in ex vivo brain slices from both rats and mice show increases in intracellular calcium ([Ca 2+ ] i ) in response to low glucose conditions or bath application of 2-DG. [16][17][18] It appears that NTS astrocytes then relay this signal to neighbouring neurons (including NTS TH neurons) by modulating extracellular purine levels. 17,18 NTS astrocyte activity and purinergic signalling is required for 2-DG-induced increases in blood glucose in anaesthetised rats, providing functional evidence for astrocyte detection of glucoprivation. 19 Glucose transporter 2 (GLUT2) is proposed to play a glucose-sensing role in astrocytes 20,21 as re-expression of GLUT2 in astrocytes of GLUT2 À/À mice is sufficient to restore NTS sensitivity to glucoprivation. 20 In addition, the astrocyte Ca 2+ response to bath application of low glucose in brain slices is abolished with pharmacological GLUT2 blockade. 21 Taken together, these data suggest NTS astrocytes can detect low glucose, modulate extracellular purine levels (possibly via direct release, i.e., gliotransmission) to excite neighbouring neurons, and drive physiological responses to restore blood glucose. 15 Thus, evidence from ex vivo brain slice experiments suggest astrocytes in the NTS may be the primary sensors of local glucopriva- The rostrocaudal "bins" used to quantify cell counts were as follows (all values-mm from Bregma): rostral 6.96-7.2, postremal 7.32-7.64, caudal 7.76-8. Sections were imaged at 10x magnification on an upright fluorescence microscope (Leica AF6000) and compared to a reference brain atlas 22 to determine the distance from Bregma. For colocalization analysis of FOS and GFAP-immunoreactivity, two postremal-level NTS sections were imaged at 20x magnification on a confocal microscope (Leica DMi8) and mean cell counts from these sections were calculated for each mouse (n = 8 mice per group, 5 male, 3 female).

| Viral vector injection surgery
Viral vectors were delivered to the DVC as described previously. 23,24 In brief, mice were anaesthetised with ketamine (70 mg/kg) and medetomidine (0.5 mg/kg) administered as a cocktail i.p. The skin on the back of the head and neck was shaved and the mouse was placed in a stereotaxic frame. Under aseptic conditions, the skin was incised, and the muscles parted to expose the atlanto-occipital membrane. This membrane was then incised, and a Hamilton needle inserted into the left side of the brain (from obex r/c 0 mm, m/l 0.2 mm, d/v 0.5 mm) at a 25 angle. Then, 180 nL of viral vector was injected at a rate of 100 nL/min and the needle was kept in place for 2 min after the injection. This process was then repeated on the contralateral (right) side and the wound was closed with sutures in the muscle and skin. The mouse was subcutaneously injected with carprofen (5 mg/kg) and atipamezole (1 mg/kg) and transferred to a heated (24-26 C) recovery cage. The following morning mice received a second injection of carprofen (5 mg/kg). Mice were given >3 weeks to allow for recovery and viral expression before being used in experiments. Viral vectors used were AAV9/2-hGFAP-hM3Dq-mCherry (ETH Zurich Viral Vector Facility, Zurich, Switzerland; titre = 5 x 10 8 viral genomes/mL), AAV5/2-hG-FAP-mCherry (ViGene Biosciences; titre = 3.89 x 10 13 ) or AAV5/2-h-Syn-DIO-eGFP (ETH Zurich Viral Vector Facility) and were diluted 1:2 in sterile saline prior to injection when one vector was injected. When two vectors were coinjected they were mixed 1:1:1 with sterile saline.

| Food intake, water intake and activity monitoring
Mice were individually housed in Promethion metabolic monitoring cages (Sable Systems Europe) which contained three sensors: a mass monitor attached to the food hopper, a mass monitor attached to the water bottle, and an XY-beam array surrounding the cage. These permitted the accurate quantification of food intake, water intake and activity using the Promethion system for data acquisition and a macro for data analysis. On the day of the experiment mice were injected i.p. with saline or clozapine-N-oxide (CNO; 1 mg/kg, Tocris/Bio-Techne) immediately prior to lights off. Data were then acquired for 23 h. Drug order was not randomised due to the potential duration of CNOmediated effects and for this reason the experimenter was not blinded.
Saline was given first followed at least 72 h later by CNO. Following this, the same mice were used for glucose tolerance testing (see below).

| Glucoprivic feeding
The protocol for assessing glucoprivic feeding was modified from Lewis et al. 25 Mice were individually housed and habituated to T A B L E 1 Results of three-way analysis of variance (ANOVA) of the effects, and interactions of sex, drug treatment and rostrocaudal (RC) position or time on FOS-immunoreactive (FOS-IR) cell number, GFAP-immunoreactive cell number, food intake, glucoprivic feeding, glucose tolerance test (GTT) or insulin tolerance test (ITT). Statistically significant effects (p < .05) are shown in bold font for clarity. For glucoprivic feeding only saline/2-deoxyglucose (2-DG) and clozapine-N-oxide (CNO)/2DG groups were analysed to examine the main effect of CNO.   the opposite drug (saline or CNO) given in the first injection. The initial drug allocation was randomised by coin toss and the investigator was blinded to this allocation until after the experiment concluded.

| Perfusion and histology
All viral-vector injected mice were perfused as described above ("Sys-

| Statistical analysis
All values were collated in Prism 9 (GraphPad) for tests of statistical  These experiments were performed in both male and female mice and the data stratified by sex are shown in Figure S1. In general, effects were consistent across sexes with the exception of FOSimmunoreactivity in the rostral NTS, which was increased in males injected with 2-DG but not females ( Figure S1B,C).

| Chemogenetic activation of DVC astrocytes suppressed food intake through reduced meal number and size
We sought to characterise hypophagia induced by chemogenetic DVC astrocyte modulation using hM3Dq 24,30 in mice of both sexes by combining this methodology with non-invasive monitoring of food intake, water intake and activity. Mice received bilateral injections of an adeno-associated viral (AAV) vector containing the hM3Dq receptor fused to a fluorescent mCherry reporter under the control of the gfaABC 1 D GFAP promotor fragment 31 (Figure 2A). This widely used approach drives expression of hM3Dq in astrocytes at the injection site. [32][33][34][35][36][37] In these mice, mCherry immunoreactivity was consistently observed in the DVC while the degree of extra-DVC transduction varied between mice ( Figure 2B,C). At the target injection site (postremal NTS) mCherry-immunoreactivity was observed associated with 85.72 ± 5.4% of GFAP-immunoreactive cells (n = 3 mice, 1 section per mouse) in a diffuse pattern consistent with a membrane-bound protein ( Figures 2D and S2A). 38,39 Of note, GFAP-immunoreactivity was not observable in all DVC astrocytes and some areas had sparse labelling. We found that the mCherry-immunoreactivity was largely overlapping with S100B, a marker of the astrocyte cytoplasm which more robustly labelled astrocytes throughout the DVC ( Figure S2B). In the same area, mCherry-immunoreactivity was absent from NeuN- Water intake was also reduced on CNO injection days as compared with saline ( Figures S3A,C). Similarly, activity measured by the cumulative number of beam break events was reduced by injection with CNO relative to saline (Figures S3B,D). As water intake and activity (food seeking) are linked to feeding, from our experiment it is unclear whether these effects are secondary to reduced food intake or driven by distinct mechanisms.
In parallel, we tested mice injected with AAV-GFAP-mCherry and therefore lacking the hM3Dq receptor. In this group, treatment with CNO had no effect on food intake, meal patterning or water intake but did induce a slight, statistically significant, decrease in total activity ( Figure S4).
F I G U R E 1 Legend on next page.

| Chemogenetic activation of DVC astrocytes suppressed glucoprivic feeding
We reasoned that if they are indeed primary detectors of low glucose, chemogenetic activation of DVC astrocytes may enhance the subsequent homeostatic physiological responses. If so, this could occur by enhancing glucoprivic feeding, elevating basal blood glucose, promoting absorption of glucose from the blood, and/or ameliorating insulininduced hypoglycaemia.
To test the effect of DVC astrocyte activation on glucoprivic feeding, DVC::GFAP hM3Dq mice were injected i.p. with saline or CNO (1 mg/kg) during the light phase. Then, 30 min later mice were injected i.p. with either saline or 2-DG (0.4 g/kg) and food intake was measured from this time point ( Figure 3A). Since the observed effects were consistent across sexes, data were pooled (data are shown separated by sex in Figure S5). In line with other studies, 25 in the absence of chemogenetic activation of DVC astrocytes, i.p. 2-DG treatment increased food intake at 2, 4 and 8 h compared to injection with saline ( Figure 3B). In contrast to our hypothesis, injection with CNO prior to 2-DG attenuated glucoprivic feeding to levels equivalent to injection with saline alone ( Figure 3B). This indicates that glucoprivic feeding is suppressed by chemogenetic activation of DVC astrocytes.
At 24 h after injection, compared to saline treated controls, there was no longer a statistically significant effect of 2-DG treatment on food intake measurements in DVC::GFAP hM3Dq mice ( Figure 3B).
However, CNO-treatment in DVC::GFAP hM3Dq mice resulted in a statistically significant reduction in 24 h food intake compared to saline treatment regardless of whether 2-DG was also administered ( Figure 3B). This is consistent with our feeding data ( Figure 2G,K) where the rate of food intake is reduced for $12 h after CNO injection before returning to normal. In these studies and our prior studies of fast-induced refeeding 24 we did not see a compensatory increase in food intake following the offset of CNO's presumed effect window.
A control group of DVC::GFAP mCherry mice showed that CNO alone did not suppress glucoprivic feeding ( Figure S6). We did, however, find a statistically significant increase in food intake in CNO injected DVC::GFAP mCherry mice compared with saline alone 8 and 24 h after injection ( Figure S6). Since we have not observed this previously ( Figure S4) it is possible this finding is spurious. 24 Additionally, the direction of this finding is opposite to our observation of food intake suppression in DVC::GFAP hM3Dq mice therefore CNO alone does not account for our findings.

| Chemogenetic activation of DVC astrocytes
was not sufficient to alter basal blood glucose levels was not sufficient to initiate mechanisms to increase blood glucose.  Figure 4B). This effect appeared to be specific to the initial Due to the divergent effect of CNO between sexes ( Figure 4B,F) data for all blood glucose measurements were stratified by sex rather than pooled as for other data.

| Chemogenetic activation of DVC astrocytes was not sufficient to alter glucose homeostasis responses in the insulin tolerance test
Finally, we tested whether chemogenetic activation of DVC astrocytes altered the response to subsequent insulin-induced hypoglycaemia in the insulin tolerance test. We hypothesised that increasing DVC astrocyte activity would be protective against this stimulus by reducing the severity of hypoglycaemia and/or improving blood glucose recovery by stimulating hepatic glucose production. 13 In contrast to our hypothesis, at the time points assessed no difference was observed in the response to insulin-induced hypoglycaemia between saline or CNO injection in female or male DVC::GFAP hM3Dq mice, neither in the initial reduction in blood glucose nor subsequent recovery ( Figure 4D,H).
Taken together, the results from the glucose homeostasis assessments suggest that activation of DVC astrocytes was not sufficient to modulate basal blood glucose nor was it protective against subsequent insulin-induced hypoglycaemia. There was, however, a statistically significant reduction in the peak glucose excursion in female mice during the glucose tolerance test.
In parallel, we performed all experiments evaluating glucose homeostasis parameters in control DVC::GFAP mCherry mice and observed no statistically significant effects of CNO on basal blood glucose, glucose tolerance or insulin tolerance ( Figure S7).

| Chemogenetic activation of DVC astrocytes did not selectively activate catecholaminergic NTS neurons
We previously observed that chemogenetic activation of DVC astrocytes induced FOS-immunoreactivity in the DVC and lateral parabrachial nucleus. 24 In the current study we sought to identify whether NTS catecholaminergic neurons, identified by their expression of TH (NTS TH neurons), were among those recruited, perhaps being selectively recruited over non-catecholaminergic NTS neurons. This neuronal population was specifically selected for study as NTS TH neurons are involved in both appetite suppression and glucoprivic feeding in mice. 12 GFAP mCherry mice. In addition, we observed many FOSimmunoreactive cells that did not express TH ( Figure 5E). This indicates that although there is some recruitment of NTS TH neurons by chemogenetic DVC astrocyte activation this represents a minority of NTS TH neurons. Furthermore, since far more non-TH neurons are activated, there appears to be no preferential recruitment of NTS TH neurons by this manipulation.

| DISCUSSION
In this study we set out to determine the ability of astrocytes in the NTS and wider DVC to sense acute systemic 2-DG-induced glucoprivation, and whether their chemogenetic activation was sufficient to modulate glucose homeostasis. We found that NTS GFAP-expressing astrocytes did not increase their FOS-immunoreactivity following acute systemic 2-DG treatment nor did they alter their primary process morphology. Furthermore, the number of GFAP-expressing cells in the NTS was equivalent between groups. In addition, we verified that chemogenetic activation of DVC astrocytes cells suppressed nocturnal food intake in mice of both sexes, and attenuated glucoprivic feeding following systemic 2-DG administration. We found no effect visible, other studies report rapid reorganisation of astrocyte morphology by food intake or hormone administration in the order of 1-2 h, albeit using different methodology (electron microscopy). 45,46 Additionally, we only examined the morphology of the primary processes (visualised by GFAP immunoreactivity) which does not label the fine processes that make up much of the astrocyte cell volume. 47 As such it is certainly possible that the morphology of these finer processes may change, altering the proximity of the astrocyte to neuronal synapses and modulating transmission as a result.
In order to unify our findings with brain slice imaging, in vivo recording of astrocyte Ca 2+ with fibre photometry or imaging is a promising future direction. 48 Due to the location of the DVC below the craniovertebral junction there are significant technical challenges associated with doing this in awake freely moving mice, although recently fibre photometry recordings have been published for both astrocytes in the striatum and neurons in the NTS. 49,50 As such, it is possible that combining these approaches may reveal the real-time responses of astrocytes to deviations in energy status in intact animals which were not evident using the approach employed here.

| Chemogenetic activation of DVC astrocytes suppressed food intake even under glucoprivic conditions
While chemogenetic activation of DVC astrocytes suppresses food intake under conditions of physiological hunger (12 h overnight fast), 24 fast-induced food intake and glucoprivic feeding are functionally separable at the circuit level. 51

| Uncovering neural circuits downstream of chemogenetic stimulation of DVC astrocytes
Using immunohistochemistry, we examined whether TH neurons were preferentially recruited following chemogenetic activation of DVC astrocytes. Instead, we found that the most FOSimmunoreactivity was in TH-negative cells. The percentage of NTS TH neurons that were activated increased in more rostral sections relative to caudal. At these levels, there are NTS TH neurons that both stimulate and inhibit feeding 63 and so it is possible that more feeding-suppressive than feeding-stimulating neurons are recruited. Communication between astrocytes and neighbouring neurons induced by this manipulation may be non-random which raises the question of which neuronal populations are expressing FOS. From our functional data we can speculate that PPG, ChAT and GABA neurons are not recruited since the published effects of chemogenetic manipulation of these cells on blood glucose and glucose tolerance 13,40,41 were not recapitulated in our studies, although further investigation is required to confirm this assertion.
It is also, however, possible that one or more of these neuronal populations are recruited non-selectively and their actions occlude one another or are diluted by bulk CNO-induced NTS neuronal activation in our experiment.
Building on our published work in male mice, 24 following chemogenetic stimulation of DVC astrocytes, we found food intake was reduced in both male and female animals associated with both reduced meal size and frequency. We previously observed FOS-immunoreactivity in the lateral parabrachial nucleus after chemogenetic stimulation of DVC astrocytes. 24 Together, these findings suggest that the activated neurons are likely glutamatergic since these neurons account for all appetite suppressing NTS subtypes 64 and NTS projections to the lateral parabrachial nucleus have been observed to be glutamatergic. 42,65

| CONCLUSION
We initially posited that DVC astrocytes could be primary glucoprivation sensors, whose activation alone is sufficient to stimulate counterregulatory circuitry and downstream glucose homeostasis responses, but this was not supported by our experimental observations using bulk chemogenetic activation of astrocytes across the mouse DVC.
Superficially, this appears to contrast with the published data indicating that DVC astrocytes can detect low glucose in ex vivo rat brain slices [16][17][18] and that pharmacologically inhibiting their activity prevents 2-DG-induced compensatory hyperglycaemia in anaesthetised rats. 19 This may potentially be accounted for by a divergence of necessity (as suggested by published studies looking at pharmacological inhibition) and sufficiency (as suggested by our studies herein looking at chemogenetic activation). However, the picture is likely far more complex with regional and physiological state-dependent differences in astrocyte function influencing their roles in brainstem circuity. As stated above, the results presented here do not rule out the possibility that subsets of DVC astrocytes are specialised for sensing glucoprivation by virtue of their location and/or unique functional capabilities, and that these preferentially communicate with the appropriate neurons to drive responses that restore glucose homeostasis. This is supported by data from the rat brain slice experiments showing that only a proportion ($40%) of NTS astrocytes respond to glucoprivation. 18 Within the brainstem, neuronal circuits that regulate fast-induced food intake and glucoprivic feeding are functionally separable at the circuit level. 11,51 Tools for selective in vivo manipulation of functionally distinct astrocyte subsets that would facilitate a similar type of experimental circuit dissection are not yet available.
The absence of overt morphological or immunoreactivity differences in DVC astrocytes following in vivo glucoprivation may suggest that DVC astrocytes provide metabolic and/or functional support to counter-regulatory neural circuits that is indispensable to their function. Of the currently available tools, future work measuring the regional, state-dependent Ca 2+ dynamics of DVC astrocytes in vivo during glucose fluctuations may help reconcile our findings with those from ex vivo brain slice imaging.