Changes in neuronal activity across the mouse ventromedial nucleus of the hypothalamus in response to low glucose: Evaluation using an extracellular multi‐electrode array approach

Abstract The hypothalamic ventromedial nucleus (VMN) is involved in maintaining systemic glucose homeostasis. Neurophysiological studies in rodent brain slices have identified populations of VMN glucose‐sensing neurones: glucose‐excited (GE) neurones, cells which increased their firing rate in response to increases in glucose concentration, and glucose‐inhibited (GI) neurones, which show a reduced firing frequency in response to increasing glucose concentrations. To date, most slice electrophysiological studies characterising VMN glucose‐sensing neurones in rodents have utilised the patch clamp technique. Multi‐electrode arrays (MEAs) are a state‐of‐the‐art electrophysiological tool enabling the electrical activity of many cells to be recorded across multiple electrode sites (channels) simultaneously. We used a perforated MEA (pMEA) system to evaluate electrical activity changes across the dorsal‐ventral extent of the mouse VMN region in response to alterations in glucose concentration. Because intrinsic (ie, direct postsynaptic sensing) and extrinsic (ie, presynaptically modulated) glucosensation were not discriminated, we use the terminology ‘GE/presynaptically excited by an increase (PER)’ and ‘GI/presynaptically excited by a decrease (PED)’ in the present study to describe responsiveness to changes in extracellular glucose across the mouse VMN. We observed that 15%‐60% of channels were GE/PER, whereas 2%‐7% were GI/PED channels. Within the dorsomedial portion of the VMN (DM‐VMN), significantly more channels were GE/PER compared to the ventrolateral portion of the VMN (VL‐VMN). However, GE/PER channels within the VL‐VMN showed a significantly higher basal firing rate in 2.5 mmol l‐1 glucose than DM‐VMN GE/PER channels. No significant difference in the distribution of GI/PED channels was observed between the VMN subregions. The results of the present study demonstrate the utility of the pMEA approach for evaluating glucose responsivity across the mouse VMN. pMEA studies could be used to refine our understanding of other neuroendocrine systems by examining population level changes in electrical activity across brain nuclei, thus providing key functional neuroanatomical information to complement and inform the design of single‐cell neurophysiological studies.


| INTRODUC TI ON
The brain is critical for maintaining systemic glucose homeostasis, in part by direct sensing of changes in glucose levels. Neurophysiological studies in rodent brain slices have identified glucose-sensing neurones that change their firing frequency in response to alterations in local glucose levels. Glucose-sensing neurones are broadly subdivided into two groups: glucose-excited (GE) and glucose-inhibited(GI).GEneuronesshowenhancedfiringactivityinresponseto higher ambient glucose concentrations, whereas the reverse is true for GI neurones. The hypothalamus is a key brain region mediat- The VMN is one of the best studied hypothalamic nuclei with respecttoglucose-sensingneurones,withbothGEandGIneurones observed. 1,5 VMN glucose-sensing neurones play an important role in detecting and reacting to glucose deficit, regulating both the counter regulatory response to hypoglycaemia and glucoprivic feeding. 6 GEandGIneuronesaredefinedasbeingdirectly/intrinsically responsive to alterations in extracellular glucose levels. 7 Indeed,in the presence of the voltage-gated sodium channel blocker, tetrodotoxin(TTX),directpostsynapticchanges(ie,intherestingmembranepotential)areseeninbothGEandGIneuronesinvariousbrain areas. 8,9 Three subtypes of non-intrinsic, glucose-sensing VMN neurones have also been described, which are modulated by presynaptic glutamatergic inputs. These extrinsically glucose-sensing neurones include PED neurones, which are presynaptically excited by a decrease(PED)inextracellularglucoselevels(from2.5to0.1mmolL -1 glucose),andPERandPIRneurones,whichareeitherpresynaptically excited(PER)orinhibited(PIR)byanincreaseinextracellularglucose levels (from 2.5 to 5-10 mmol L -1 glucose), respectively. 10 Unlikeotherhypothalamicnucleiwheretheneuropeptide/neurotransmitter phenotype of glucose-sensing neurones has already been identified, 1  because only a single cell can be studied at a given time, and given that intrinsic GE and GI neurones are estimated to together comprise only approximately 20% of the VMN neuronal population, 11 thismakesthepatchclamptechniquebothtime-consumingandof relatively low yield. Furthermore, because few glucose-sensing neurones can be recorded per brain slice, the investigator, who, in this circumstance, is ultimately searching for specialised glucose-sensing neurones, is potentially introducing unintentional anatomical selec-tionbiasintheirrecordingsbytargetingpartsoftheVMNknownto be enriched in these cells.

Multi-electrodearrays(MEAs)compriseanelectrophysiological
tool enabling the electrical activity of large neuronal groups to be simultaneously recorded across multiple electrode sites (channels).
Thispowerfulneurophysiologicalextracellularrecordingtechnique is used both in vivo and ex vivo. 12 Anadvantageofthismethodcompared to others is the high spatial and temporal resolution that it offers, enabling the functional investigation and mapping of complex and heterogeneous nuclei, such as the VMN. 13,14 Each MEA electrode site can record the electrical activity from a population of neurones:referredtoasmulti-unitactivity(MUA).Useofconventional spikesortingmethodsallowstheactivityofsinglecells(single-unit activity[SUA])tobediscriminatedfrom theMUA. 15 Wehavepreviously employed perforated MEA (pMEA) technology for ex vivo hypothalamic recordings from mouse brain slices 16 because of the improved slice perfusion rate, long-term recording stability and high signal-to-noise ratio. 17 Inthepresentstudy,wehaveusedanequiv-alentpMEAextracellularrecordingapproachinexvivoadultmouse brain slices aiming to objectively evaluate changes in neuronal activity across the dorsal-ventral extent of the medial portion of the VMN in response to changes in extracellular glucose concentration. To our knowledge,thisisthefirststudyutilisingtheMEAsystemtoeval-uateglucoseresponsivityacrossthemouseVMNneuralnetwork.

| In vitro multi-electrode recordings
MEA recordings from acute brain slices were performed as described previously. 16,19 After 1 hour of rest, a VMN-containing mouse brain slice was placed recording side down, onto a 60pMEA100/30iR-Ti-gr perforated multi-electrode array (pMEA; Multi Channel Systems, MCS GmbH, Hanover, Germany). These arrays are comprised of 60 electrodes in a 6 × 10 layout (one is a reference electrode), with each electrode 100 µm apart, and with a diameter of 30 µm, covering a total area of approximately 707 μm 2 . Thus, anatomically, a single array can cover the entirety of the mouse VMN. 18 Contrastillumination(bothunderneathand above the brain slice) was used to ensure accurate placement of the

| Glucose responsiveness
To investigate the glucose-sensing capability of VMN neural networks to changing glucose concentrations, recordings were com-

| Anatomical locations
Offline overlay of pMEA electrode sites and respective recorded sliceimages,withtheaidofkeylandmarks(shapeofthehippocampus, third ventricle, median eminence and location of the optic tracts) andreferencetotheMouseBrainAtlas, 18  to be studied and, importantly, to be distinguished from non-VMN regions, with the latter being excluded from VMN-related analyses ( Table 2).
In summary, although we found evidence of glucose-sensing

| More glucose-excited (GE/PER) channels were found in the dorsomedial compared to the ventrolateral subdivision of the VMN
Within the VMN, the DM-VMN had significantly more GE/PER channels than the VL-VMN in response to a lowering of the extracellular glucose concentration from 2.5 to 0.1 mmol L -1 for 40 minutes (Figures6Aand4andTable3).Therewasnosignificantdifference in the distribution of GI/PED channels between these VMN subregions ( Figure 4 and Table 3).

| Glucose-excited (GE/PER) channels in the ventrolateral portion of the VMN displayed a higher spontaneous basal firing frequency
ThebaselinefiringfrequencyofGE/PERchannels(ie,in2.5mmolL -1 glucose;MUAanalysis)wassignificantlyhigherintheVL-VMNcompared to the DM-VMN region VMN in response to a lowering of the extracellular glucose concentration from 2.5 to 0.1 mmol L -1 for 40 minutes (Figure 6B-D). From these data, a representative heat map was generated in which the firing activity across the VMN was plotted relative to anatomical location: this analysis indicated that channels with the highest average basal firing frequency were located in the VL-VMN proximal to the central region (Figure 7).

| Analysis of the single-unit data supported the observations of the multi-unit analysis
TheSUAsupportedtheMUAdata,with60%ofsingle-unitsacross the VMN being GE/PER (Table 4) and < 7% being GI/PED in response to a lowering of the extracellular glucose concentration from 2.5 to 0.1 mmol L -1 for 40 minutes, with the latter being a higher percentage than was seen for the GI/PED MUA analysis.
There were insufficient single-units per slice within the VMN sub-regions to provide meaningful analysis of the within VMN distribution.

| D ISCUSS I ON
In the present study, we report that, using the pMEA recording technique in ex vivo brain slices, both multi-unit and single-unit    neurones reported in the rat VL-VMN. 23 Assuch,thereappearstobe variabilityinthereportedfrequencyofGIneuronesintheVMN.It is possible that, using our extracellular recording method, the preva- Nevertheless,inthepresentstudy,thepercentageofGI/PEDchannels across the VMN was significantly lower than that of GE/PER channels.

GIneuronesintheVMHthatspecificallyexpressneuronalnitric
oxide synthase have been shown to be important for both neuronal glucosensing and regulation of the counter-regulatory response to hypoglycaemia in vivo. 46 Furthermore, VMN GE and GI neurones have been shown to express SF-1. 47 AsubpopulationofVMNSF-1 neurones has been shown to co-localise with pituitary adenylate cyclase-activatingpeptide(PACAP) 48 andrecentworkhasidentified apopulationofintrinsicallyGIVMNPACAP-expressingneurones. 9 Althoughcomprisingarelativelysmallneuronalpopulation,theydisplay a wide distribution across the VMN and are considered to play a role in systemic glucose regulation because chemogenetic stimulation of these neurones in vivo resulted in inhibition of insulin secretion, which led to a reduced glucose tolerance. 9 Interestingly,GI VMNPACAP-expressingneurones(15%ofwhichwerefoundtobe neuronal nitric oxide synthase-positive) send projections both within the VMH and to other brain areas, including the paraventricular nucleus, lateral hypothalamus, aBNST, paraventricular nucleus of the thalamusandperiaqueductalgrey, 9 in line with the findings reported byMeeket al. 36 Insummary,inthepresentstudy,wehaveutilisedthepMEArecordingtechniquetoprovideanunbiasedassessmentofthepercentage and distribution of glucose-sensing neurones across the mouse VMN (ex vivo). Although the pMEA method cannot provide finer, more detailed information regarding the neurophysiological propertiesofcells,suchascanbeobtainedbythepatchclamptechnique, it is still a very useful neurophysiological tool for examining both population and single-cell level responses simultaneously across brainnuclei,thusproviding,usingahigh-throughputapproach,key functional neuroanatomical information that could complement and inform the design of future single-cell studies. In practical terms,

MEArecordingsarearguablymoreuser-friendlyandtechnicallyless
demanding to perform than the gold standard patch clamp method and, at the same time, offer a high spatial and temporal resolution. 14 Finally, the in vitro MEA method permits cell-non-invasive, longterm, stable recordings to be performed, and, in combination with optogenetic, chemogenetic, pharmacological and/or calcium imaging approaches, this method can be used to refine our understanding ofotherneuroendocrinenetworks.

ACK N OWLED G EM ENTS
We thank Dr Jonathan Brown (University of Exeter) and Dr

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

DATA AVA I L A B I L I T Y
The data that support the findings of this study are available from thecorrespondingauthoruponreasonablerequest.