Functional identification of a neurocircuit regulatingblood glucoseThomas H. Meeka, Jarrell T. Nelsona, Miles E. Matsena, Mauricio D. Dorfmana, Stephan J. Guyeneta, Vincent Damiana,Margaret B. Allisonb, Jarrad M. Scarletta, Hong T. Nguyena, Joshua P. Thalera, David P. Olsonb, Martin G. Myers Jr.c,Michael W. Schwartza, and Gregory J. Mortona,1

aDiabetes and Obesity Center of Excellence, Department of Medicine, University of Washington, Seattle, WA 98195; bDepartment of Pediatrics andCommunicable Diseases, University of Michigan, Ann Arbor, MI 48105; and cDepartment of Internal Medicine, University of Michigan, Ann Arbor, MI 48105

Edited by Gerald I. Shulman, Howard Hughes Medical Institute/Yale University, New Haven, CT, and approved February 16, 2016 (received for review October28, 2015)

Previous studies implicate the hypothalamic ventromedial nucleus(VMN) in glycemic control. Here, we report that selective inhibitionof the subset of VMN neurons that express the transcription factorsteroidogenic-factor 1 (VMNSF1 neurons) blocks recovery from insu-lin-induced hypoglycemia whereas, conversely, activation of VMNSF1

neurons causes diabetes-range hyperglycemia. Moreover, this hyper-glycemic response is reproduced by selective activation of VMNSF1

fibers projecting to the anterior bed nucleus of the stria terminalis(aBNST), but not to other brain areas innervated by VMNSF1 neurons.We also report that neurons in the lateral parabrachial nucleus(LPBN), a brain area that is also implicated in the response to hypo-glycemia, make synaptic connections with the specific subset of glu-coregulatory VMNSF1 neurons that project to the aBNST. Theseresults collectively establish a physiological role in glucose homeo-stasis for VMNSF1 neurons and suggest that these neurons are part ofan ascending glucoregulatory LPBN→VMNSF1→aBNST neurocircuit.

glucoregulatory circuit | counter regulation | ventromedial nucleus |bed nucleus of the stria terminalis | hyperglycemia

Because the brain relies exclusively on glucose as a fuel source,brain function is rapidly compromised when circulating glucose

levels drop below the normal range. Consequently, hypoglycemiaelicits a robust, integrated, and redundant set of counterregulatoryresponses (CRRs) that ensure the rapid and efficient recovery ofplasma glucose concentrations into the normal range (1). Com-ponents of the CRR include increased secretion of the hormonesglucagon, epinephrine, and glucocorticoids, inhibition of glucose-induced insulin secretion, increased sympathetic nervous system(SNS) outflow to the liver, and increased food intake (1–3). Owingto this redundancy, recovery from hypoglycemia is difficult to blockin normal humans and animal models, even when adrenal or glu-cagon responses are prevented. Only when multiple responses areblocked is the ability to recover from hypoglycemia significantlycompromised (4). This arrangement is perhaps unsurprising, giventhe threat to survival posed by hypoglycemia.Although glucose sensing can occur at peripheral (e.g., neu-

rons innervating the hepatic portal vein) as well as central sites(3, 5), the brain is the organ responsible both for transducing thisinformation into effective glucose counterregulation and forterminating this response once euglycemia is restored. Of themany brain areas that have been investigated, the hypothalamicventromedial nucleus (VMN) has emerged as potentially beingboth necessary and sufficient to elicit this powerful response.This assertion is based on evidence that, whereas electricalstimulation of the VMN activates the CRR and thereby raisescirculating glucose levels (6), glucose infusion directly into theVMN can suppress the CRR during hypoglycemia (7) andthereby impair recovery of normal blood glucose levels (8).Moreover, two recent papers identified a circuit comprised ofneurons in the lateral parabrachial nucleus (LPBN) that projectto the VMN, activation of which seems to be required for ef-fective glucose counterregulation (9, 10). The relevant VMN

neurons seem to be glutamatergic because genetic disruption ofglutamatergic signaling within a specific subset of VMN neurons[known as steroidogenic factor-1 (SF1) neurons] also attenuatesrecovery from insulin-induced hypoglycemia (11).Together, these observations support a model in which, during

hypoglycemia, inputs from the LPBN and other glucose-responsiveneurocircuits converge upon and activate VMN neurons, and inwhich this activation is required for effective glucose counter-regulation. Whether these VMN neurons or any other specificneuronal subset(s) are truly necessary or sufficient for this impor-tant adaptive response has yet to be established, however. Recenttechnological advances, including the use of optogenetics to selec-tively activate or inactivate neurons in a regionally and temporallyspecific manner, now enable such questions to be addressed (12).A relevant parallel can be drawn to knowledge recently gained

regarding control of feeding behavior by neurons that expressAgouti-related peptide (Agrp), found in the adjacent hypotha-lamic arcuate nucleus. Since their discovery more than a decadeago (13, 14), innumerable papers were published implicating theseneurons as key drivers of fasting-induced hyperphagia (15, 16), butwhether they are in and of themselves necessary or sufficient fornormal food intake was unknown until recently. The past fewyears have shown (among other things) that selective activation ofAgrp neurons is sufficient to potently drive intake whereas in-hibition of these neurons blunts the effect of fasting to stimulatefeeding (17, 18). These findings offer direct, compelling supportfor the hypothesis that Agrp neurons are primary drivers of need-based feeding. Whether activation of VMNSF1 neurons is bothnecessary and sufficient to explain the powerful adaptive responseselicited by hypoglycemia is a question of similar importance. Given

Significance

Hypoglycemia is an important and frequently encountered com-plication of diabetes treatment. Here, we identify a subset ofneurons located in the ventromedial hypothalamic nucleus, ac-tivation of which is both necessary and sufficient to mediateadaptive counterregulatory responses to hypoglycemia thatreturn low blood glucose levels into the normal range. Theseneurons receive ascending input from neurons in the lateralparabrachial nucleus and in turn control blood glucose levels viaprojections to the anterior bed nucleus of the stria terminalis.Together, this work identifies a previously unrecognized func-tional neurocircuit involved in glycemic control.

Author contributions: T.H.M., M.D.D., J.M.S., J.P.T., M.W.S., and G.J.M. designed research;T.H.M., J.T.N., M.E.M., S.J.G., V.D., and H.T.N. performed research; M.B.A., J.M.S., D.P.O.,M.G.M., M.W.S., and G.J.M. contributed new reagents/analytic tools; T.H.M., M.D.D.,M.W.S., and G.J.M. analyzed data; and T.H.M., M.W.S., and G.J.M. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. Email: gjmorton@u.washington.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1521160113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1521160113 PNAS | Published online March 21, 2016 | E2073–E2082

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

020

the many sites that participate in glucose sensing (5, 19, 20) andhow difficult it is to prevent recovery from hypoglycemia (owing toredundancy inherent in the CRR), it seems improbable that theentire response should hinge on activation of a select population ofneurons in a relatively small brain area. To test this hypothesis, wefirst determined whether optogenetic inactivation of VMNSF1

neurons during insulin-induced hypoglycemia blocks increasedglucagon and corticosterone secretion, inhibition of glucose-induced insulin secretion, and recovery of normal blood glucoselevels. We also asked whether, conversely, optogenetic activationof VMNSF1 neurons elicits hyperglycemia by activating the CRRin otherwise normal mice. Last and perhaps most importantly, wesought to identify projections of VMNSF1 neurons to downstreambrain areas involved in glucose counterregulation and to de-termine whether these VMNSF1 neuronal subsets make synapticcontacts with the previously identified upstream neurons locatedin the LPBN. Our data collectively indicate that (i) VMNSF1

neurons are key components of a circuit that extends from theLPBN to the VMN and then to the anterior bed nucleus of the striaterminalis (aBNST) (LPBN→VMN→aBNST), (ii) VMNSF1 neuronsfunction as a primary motor output driving the CRR, and (iii)activation of VMNSF1 neurons is required for effective recoveryfrom insulin-induced hypoglycemia.

ResultsVMNSF1 Neurons Are Required for Intact Responses to Insulin-InducedHypoglycemia.Because of the threat to survival posed by inadequateglucose delivery to the brain, the CRR to hypoglycemia is mediatedby a highly integrated, redundant, and powerful combination ofbehavioral, autonomic, and neuroendocrine responses that areconserved across mammalian species. To investigate the physio-logical role played by VMNSF1 neurons in this response, we soughtto determine whether activation of these neurons is required for theability to recover from hypoglycemia. To this end, we expressed alight-activated inhibitory channel selectively in VMNSF1 neurons.This objective was achieved through bilateral stereotaxic microin-jection of an adeno-associated viral construct (AAV) into the VMNof SF1-Cre-positive (SF1-Cre+) mice (in which Cre recombinase isexpressed under the control of the VMN-specific SF1 promoter)(21). The AAV expresses a modified channelrhodopsin chloride-conducting anion channel fused with the fluorescent reporterEYFP (AAVDJ8-DIO-SwiChRCA-EYFP, referred to hereafter as“SwiChRCA virus”) and is “Cre-inducible,” meaning that viral ex-pression is restricted to cells that also express Cre recombinase.Through an optic fiber implanted immediately dorsal to each VMNinjection site (Fig. 1 A–C), delivery of light (“Laser On”) causeshyperpolarization of VMNSF1 neurons, but not other neurons in theVMN or surrounding brain areas (Fig. 1C).We report that, whereas photoinhibition of VMNSF1 neurons

had no effect on fasting blood glucose levels in the basal staterelative to the “Laser-Off” condition (Fig. 1G), the ability of miceto recover from insulin-induced hypoglycemia was severely im-paired by light-induced silencing of VMNSF1 neurons (Fig. 1D).Thus, activation of a discrete and uniquely identified subpopulationof VMN neurons is required to effectively recover from hypogly-cemia. By comparison, VMNSF1 neuron activation is not requiredfor maintenance of normal blood glucose levels in the basal state(e.g., in the absence of a threat to brain glucose delivery).The striking inability of mice to respond to insulin-induced hy-

poglycemia when VMNSF1 neurons are silenced implies that theiractivation is required for secretion of counterregulatory hormonesin this setting. Consistent with this hypothesis, we found, in aseparate experiment, that the increase of circulating glucagon andcorticosterone levels that normally occurs during hypoglycemia wasblocked when VMNSF1 neurons were silenced (Fig. 1 E and F).These data demonstrate that activation of VMNSF1 neurons isindispensable for an intact CRR to hypoglycemia, as previously

hypothesized (11), and thus constitute direct evidence of aphysiological role for these neurons in glucose homeostasis.

Photoactivation of VMNSF1 Neurons Induces Hyperglycemia. Based onour finding that activation of VMNSF1 neurons is required for anintact response to hypoglycemia, we hypothesized that activation ofthese neurons in otherwise normal mice will trigger the CRRand thereby cause blood glucose levels to increase. To test this

Fig. 1. Photoinhibition of VMNSF1 neurons disrupts recovery from insulin-induced hypoglycemia. (A) Scheme demonstrating bilateral stereotaxic micro-injection of the Cre-dependent inhibitory channelrhodopsin-EYFP (SwiChRCA)virus into the VMN of SF1-Cre+ mice and (B) implantation of the optic fiberdorsal to the injection site. (C) Representative image of EYFP fluorescenceshowing bilateral infection and expression specifically within the VMN of SF1-Cre+ mice. 3V, third cerebral ventricle; VMN, ventromedial hypothalamic nu-cleus. (D) Blood glucose and plasma (E) glucagon and (F) corticosterone levels inSF1-Cre+ mice during photoinhibition of VMNSF1 neurons (Laser On) relative tothe Laser Off during insulin-induced hypoglycemia (n = 7–8 per group). *P <0.05 vs. Laser off. All data are expressed as mean ± SEM. (G) Blood glucose levelsbefore (pre), during (inhib), and 1 h after (post) bilateral photoinhibition ofVMNSF1 neurons in SF1-Cre+ mice in the basal state (n = 8) analyzed by one-wayANOVA with Sidak’s post hoc test. All comparisons are nonsignificant.

E2074 | www.pnas.org/cgi/doi/10.1073/pnas.1521160113 Meek et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

020

hypothesis, we expressed a light-activated excitatory channelselectively in VMNSF1 neurons through unilateral stereotaxicmicroinjection of a Cre-dependent, AAV-expressing channelrho-dopsin-EYFP virus (AAV5-DIO-ChR2-EYFP) into the VMN ofSF1-Cre+ mice, followed by implantation of an optic fiber abovethe injection site (Fig. 2 A and B). As expected, EYFP fluorescencewas restricted to the VMN of SF1-cre+ mice, and, 1 h after pho-tostimulation of these neurons, a robust induction of c-Fos(a marker of neuronal activation) was detected in the VMN andsurrounding hypothalamic areas (Fig. 2C). As predicted, photo-activation of VMNSF1 neurons raised blood glucose levels rapidlyand markedly, with glycemia returning to baseline values within 1 hafter cessation of photostimulation (Fig. 2D). Because this effectwas not observed in Cre-negative littermate controls (VMNControl)that underwent the same viral microinjection and light exposureprocedure (Fig. 2E), we conclude that the observed hyperglycemicresponse was a specific consequence of VMNSF1 neuron activation.To further characterize the effect of VMNSF1 neuronal activationon glucose homeostasis, we performed an intraperitoneal (i.p.)glucose tolerance test (IPGTT) in both the presence and absenceof unilateral photoactivation of VMNSF1 neurons. Glucose toler-ance was markedly impaired by VMNSF1 neuron activation (Fig.2F) (GluAUC; Laser Off, 21,560 ± 904 vs. Laser On, 39,659 ±4,253; P < 0.05), further establishing its powerful diabetogenic ef-fect. To determine whether increased hepatic gluconeogenesiscontributes to this impairment of glucose tolerance, we performeda pyruvate tolerance test (PTT) in a separate cohort of SF1-Cre+

mice in both the presence and absence of VMNSF1 neuron pho-toactivation. Our finding that the rise of blood glucose levels inresponse to a pyruvate challenge was markedly increased duringactivation of VMNSF1 neurons (Fig. 2G) (GluAUC; Laser Off,13,723 ± 897 vs. Laser On, 20,855 ± 903; P < 0.05) implicatesincreased hepatic gluconeogenesis in the hyperglycemia and glu-cose intolerance observed in this setting.The VMN comprises a heterogeneous population of neurons,

many of which express SF1. A large subset of VMN neurons alsoexpress the long form of the leptin receptor (LepRb), and a ma-jority of these neurons also express SF1 (21, 22). To determinewhether activation of LepRb-expressing neurons in the VMN(VMNLepRb) also causes hyperglycemia, we microinjected theCre-dependent channelrhodopsin virus into the VMN of leptinreceptor-IRES-Cre (Lepr-Cre+) mice (Fig. 3 A–C) so as to enablephotoactivation of VMNLepRb neurons. Successful transduction ofthese neurons was confirmed by observing EYFP fluorescence inthe VMN, with EYFP fluorescence also detected in a few cells inadjacent areas, including the dorsomedial hypothalamus (DMH)and arcuate nucleus (ARC) (Fig. 3C). However, little or no c-Foswas colocalized with transduced LepRb neurons outside the VMNafter photostimulation (Fig. 3C). Thus, our photostimulationprotocol seems to have effectively targeted only VMNLepRb neu-rons and not adjacent hypothalamic LepRb subsets.In contrast to the potent diabetogenic effect of VMNSF1 neuron

activation, photostimulation of VMNLepRb neurons did not affectblood glucose levels (Fig. 3D), despite the fact that VMNLepRb

neurons were clearly activated by the photostimulation protocoland that many VMNLepRb neurons coexpress SF1 (21, 22). Theseobservations suggest that a subset of VMNSF1 neurons that do notexpress LepRb are responsible for the hyperglycemic effect ob-served during VMN stimulation. Alternatively, it is possible that

Fig. 2. Photoactivation of VMNSF1 neurons induces hyperglycemia andcauses impaired glucose and pyruvate tolerance in nondiabetic mice. (A)Scheme depicting unilateral stereotaxic microinjection of the Cre-dependentlight-activated channelrhodopsin (ChR2-EYFP) virus into the VMN of SF1-Cre+

mice and (B) implantation of the optic fiber dorsal to the injection site. (C)Representative image indicating unilateral infection and expression of EYFPfluorescence within the VMN of SF1-Cre+ mice and c-Fos within both the VMNand surrounding area 1 h after photostimulation. 3V, third cerebral ventricle;

VMN, ventromedial hypothalamic nucleus. Blood glucose levels before (Pre),during (Stim) and 1 h after (Post) unilateral stimulation of VMNSF1 neurons in(D) SF1-Cre+ mice and (E) VMNControl animals (Cre-negative) (n = 5–8 pergroup). *P < 0.05 vs. Pre. Blood glucose levels during an i.p. (F) glucose- (GTT)and (G) pyruvate-tolerance test (PTT) in both the presence and absence ofunilateral photoactivation of VMNSF1 neurons in SF1-Cre+ mice. *P < 0.05 vs.Laser Off. All data are expressed as mean ± SEM.

Meek et al. PNAS | Published online March 21, 2016 | E2075

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

020

activation of a subset of VMNLepRb neurons that do not expressSF1 exerts an inhibitory effect on those neurons that do, andthereby blocks the activation of responses underlying hyperglycemia.

Neuroendocrine Effects of VMNSF1 Photoactivation. To explain howVMNSF1 neuron photoactivation elicits its robust diabetogeniceffect, we hypothesized a role for both an autonomic mechanism,whereby glucose-induced pancreatic insulin secretion is blocked,and a neuroendocrine component involving increased secretionof counterregulatory hormones. To test the former hypothesis,we measured plasma insulin levels before and during photo-activation of VMNSF1 neurons. Despite the marked hypergly-cemia elicited by activation of VMNSF1 neurons (Fig. 4A),plasma insulin levels remained at their normoglycemic baseline(Laser Off) (Fig. 4B). As expected, neither blood glucose norplasma insulin levels were affected by light stimulation of theVMN of either VMNLepRb or VMNControl mice (Fig. 4B). Thus, thehyperglycemic response to VMNSF1 neuron photoactivation seemsto be mediated, in part, by a potent inhibition of glucose-inducedinsulin secretion, consistent with earlier observations after electricalstimulation of the VMN (23, 24).In addition to inhibitory effects on the pancreatic beta cell, the

hyperglycemic response to photoactivation of VMNSF1 neurons wasassociated with a marked rise of both plasma corticosterone andglucagon levels (Fig. 4 C and D). By comparison, plasma levels of

corticosterone and glucagon did not change in response to photo-activation of the VMN in either VMNLepRb or VMNControl mice(Fig. 4 C and D). Because increased hepatic glucose production(HGP) figures prominently in the metabolic actions of these hor-mones, we measured hepatic expression of the key gluconeogenicgenes phosphoenolpyruvate carboxykinase (Pepck) and glucose-6-phosphatase (G6Pase). Indeed, expression of both genes was elevatedafter light exposure in VMNSF1 mice (Fig. 4 E and F). Interestingly,relative to VMNControl mice, hepatic expression of Pepck was alsoelevated by light exposure of the VMN in VMNLepRb mice, consis-tent with previous evidence of a role for hypothalamic leptin actionin the control of hepatic glucose flux (25, 26), even though this effectoccurred in the absence of changes of glycemia or glucoregulatoryhormones (Fig. 4E). Collectively, these data suggest that the com-bination of reduced insulin secretion and increased secretion ofcorticosterone and glucagon contribute to the hyperglycemic re-sponse to VMNSF1 neuronal stimulation and that increased HGPlikely contributed to this response.

Identification of Downstream Projections of VMNSF1 Neurons RegulatingGlycemia. Among brain areas that receive dense projections fromVMN neurons are the periaqueductal gray (PAG), the bed nucleusof the stria terminalis (BNST), the central nucleus of the amygdala(CeA), and the hypothalamic paraventricular nucleus (PVN). Eachof these brain areas is implicated in autonomic, neuroendocrine,and behavioral regulation and therefore could function within aglucoregulatory circuit involving VMNSF1 neurons (27, 28). To testwhether VMNSF1 neurons are among those VMN neurons thatproject to these four brain areas, we used fluorescently taggedchannelrhodopsin to visualize the projection fields. Specifically,AAV5-DIO-ChR2-EYFP was microinjected into the VMN ofSF1-Cre+ mice, followed by histological imaging to detect theEYFP reporter in axonal projections (Fig. 5 A–E). Labeled pro-jections of VMNSF1 neurons were detected in both ipsilateral andcontralateral PAG, particularly in the dorsal region superior to thecerebral aqueduct. In the BNST, fibers arising from VMNSF1

neurons were detected in the dorsomedial region, along with adense plexus in the anterior BNST and the anterior-medial sub-division (aBNST). Projections of VMNSF1 neurons were also de-tected in both the CeA and PVN; in the latter, innervation wasparticularly robust in medial regions along the border of the thirdventricle (Fig. 5 B–E). Because these findings are consistent withprevious work using standard tract-tracing methods to detect pro-jections from unspecified VMN neurons (27), the distribution ofprojections from the subset expressing SF1 does not seem to differsignificantly from the population of VMN neurons overall.To investigate whether functional connectivity exists between

cell bodies of SF1 neurons in the VMN and any of these fourdownstream innervation sites, we used an optogenetics approachin which a ChR2-expressing virus was unilaterally microinjectedinto the VMN of SF1-cre+ mice, followed by unilateral implan-tation of an optic fiber above each of the four projection fields(aBNST, CeA, PAG, and PVN) in separate cohorts of mice. Aftertransport of ChR2 from cell bodies to axons of transduced neurons,the axons can be depolarized by laser stimulation of terminalprojections. Six weeks after surgery, stimulation of VMNSF1 axonswas achieved using the same 1-h photostimulation protocol de-scribed earlier. We found that hyperglycemia was induced afterphotostimulation of VMNSF1 neuronal projections to the aBNST(VMNSF1→aBNST), but not after stimulation of projections to anyof the other three areas (Fig. 5 F–I). To investigate whether thiseffect was specific to the aBNST, we microinjected the cre-dependent ChR2-EYFP virus into the VMN of a separate cohort ofSF1-Cre+ mice and placed the optic fiber above either the aBNSTor medial BNST (mBNST). Consistent with our earlier observa-tions, we found that unilateral photoactivation of VMNSF1 projec-tions to the aBNST resulted in hyperglycemia whereas activation ofprojections to the mBNST had no effect on blood glucose levels

Fig. 3. Photoactivation of VMNLepRb neurons fails to raise blood glucoselevels. (A) Scheme demonstrating unilateral stereotaxic microinjection of theCre-dependent light-activated channelrhodopsin (ChR2-EYFP) virus into theVMN of Lepr-IRES-Cre mice and (B) implantation of the optic fiber dorsal tothe injection site. (C) Representative image depicting unilateral infection andexpression of EYFP fluorescence within the VMN of Lepr-IRES-Cre mice andc-Fos within both the VMN and surrounding area 1 h after photostimulation.3V, third cerebral ventricle; VMN, ventromedial hypothalamic nucleus.(D) Blood glucose levels before (Pre), during (Stim), and 1 h after (Post) unilateralstimulation of VMNLepRb neurons in Lepr-IRES-Cre mice. All comparisons arenonsignificant (n = 11 per group). All data are expressed as mean ± SEM.

E2076 | www.pnas.org/cgi/doi/10.1073/pnas.1521160113 Meek et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

020

(Fig. S1). Collectively, these findings suggest that the glucoregula-tory subset of SF1 neurons project specifically to the aBNST.Consistent with the interpretation that VMNSF1→aBNST

stimulation recapitulates the glycemic response to VMNSF1 neu-ronal activation, this effect was accompanied by elevated plasmalevels of both corticosterone and glucagon (Fig. 5 N and R) andwith inhibition of glucose-induced insulin secretion, such thatplasma insulin levels did not increase despite the rise of bloodglucose concentrations (Fig. 5 F and J). By comparison, no effecton levels of blood glucose, plasma insulin, or plasma glucagon wasdetected during stimulation of the VMNSF1 neuron fibers supplyingthe PVN although a nonsignificant increase of plasma corticoste-rone levels was observed (Fig. 5 F–R). Collectively, these observa-tions implicate the subset of VMNSF1 neurons that project to theaBNST, but not to the PAG, CeA, PVN, or mBNST, as compo-nents of a functional glucoregulatory circuit.To localize the cell bodies of VMNSF1 neurons that project to

the aBNST, we used a two-pronged strategy. First, we injectedcholera toxin subunit B (CTB), a fluorescently labeled retro-grade neuronal tracer, into the aBNST and examined fluores-cence within the VMN (Fig. 6 A and B). As expected, we founddense CTB expression throughout the VMN, supporting pre-vious evidence that axonal projections from neuronal cell bodiesin the VMN supply the aBNST (27). CTB was also detectedin other mediobasal hypothalamic areas, including the ARC(among other areas) (Fig. S2), consistent with previous evidence(29) that neurons distributed throughout the mediobasal hypo-thalamus (MBH) project to the BNST (Fig. 6 C and D).We next identified cell bodies of neurons in the VMN that

were activated (as assessed by expression of c-Fos) after photo-stimulation of axonal projections of VMNSF1 neurons supplyingthe aBNST (VMNSF1→aBNST) (Fig. 6 E–H). The c-Fos wassomewhat widespread after activation of the aBNST projections,mimicking channelrhodopsin expression, with little c-Fos stain-ing observed in the absence of light treatment (Fig. 6I). Theresultant c-Fos (+) neurons were concentrated in the central (c)and dorsomedial (dm) regions of the VMN (Fig. 6 J–N), areaspreviously implicated in metabolic regulation (30, 31). By com-parison, few activated neurons were present in the ventrolateral(vl) portion of the VMN, an area associated with reproduction

and aggressive behavior (32–35). Together, these findings im-plicate subsets of SF1 neurons located in central and dorsome-dial regions of the VMN in glycemic control (10).

The LPBN Sends Afferents to aBNST-Projecting VMNSF1 Neurons. Inaddition to the VMN, recent evidence suggests that the CRR tohypoglycemia involves a subset of neurons in the LPBN markedby coexpression of leptin receptor and CCK (9, 10). Becausethese LPBN LepRbCCK neurons also project to the VMN, wesought to determine whether they are anatomically coupled toVMNSF1 neurons identified herein as having a physiological rolein glucose counterregulation. To first verify that neurons in theLPBN project to VMN, we injected CTB to the VMN and, 4 dlater, examined the brain for alexa-488 fluorescence label. In-cluded among several sites upstream of the VMN identified bythis approach were the medial preoptic area, amygdala, andBNST (Fig. S3), as well as the LPBN (Fig. 7 A–C).We next sought to determine whether the subset of VMNSF1

neurons that project to the aBNST are synaptically contacted byLPBN neurons. Accomplishing this goal was made possiblethrough the use of a two-virus strategy. The first virus is amodified rabies viral construct, SADΔG-EGFP, engineered tolack the gene encoding rabies glycoprotein G that is required fortranssynaptic spread (36), and packaged in a single envelopeglycoprotein (EnvA). The second virus is a Cre-dependent AAV-expressing TVA construct (that expresses the receptor for theavian sarcoma leucosis virus glycoprotein, EnvA) linked with RG(rabies envelope glycoprotein) using a 2A-cleavage peptide,AAV8-DIO-TVA-RG (TVA-RG). With this strategy, microin-jection of TVA-RG into the VMN of SF1-Cre+ mice selectivelyrenders VMNSF1 neurons (i) uniquely susceptible to infection byEnvA and (ii) capable of retrograde monosynaptic spread due tothe presence of RG (Fig. 7D). By subsequently injecting theSADΔG-EGFP virus (EnvA) into the aBNST of the same mice,infection by the latter virus is restricted initially to VMNSF1

terminals projecting to this region previously transfected by theTVA-RG virus. Subsequently, the SADΔG-EGFP virus is able tocross a single synapse and thereby infect neurons upstream ofaBNST-projecting VMNSF1 neurons.

Fig. 4. Photoactivation of VMNSF1 neurons medi-ates CRR responses. (A) Blood glucose, (B) plasmainsulin, and (C) plasma corticosterone either prior(Pre), or 1 h after Cre-dependent channelrhodopsinstimulation (Stim) and (D) plasma glucagon andhepatic expression of (E) Pepck and (F) G6Pase asmeasured by real-time PCR 1 h after Cre-dependentchannelrhodopsin activation of VMNSF1, VMNControl,and VMNLepRb cell types, respectively (n = 5–8 pergroup). *P < 0.05 vs. VMNControl; #P < 0.05 vs.VMNLepRb. All data are expressed as mean ± SEM.

Meek et al. PNAS | Published online March 21, 2016 | E2077

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

020

Fig. 5. Photoactivation of VMNSF1→aBNST projections selectively promote hyperglycemia in nondiabetic mice. (A) Schematic showing unilateral stereotaxic micro-injection of the Cre-dependent light-activated channelrhodopsin (ChR2-EYFP) virus into the VMN of SF1-Cre+ mice (i.e., VMNSF1 neurons) and implantation of the opticfiber above each of four projection sites in separate cohorts of animals. Image adapted from connectivity.brain-map.org/; experiment ID 182337561. aBNST, anterior bednucleus of the stria terminalis; CeA, central nucleus of the amygdala; PAG, periaqueductal gray; PVN, paraventricular nucleus; VMN, ventromedial nucleus. Fluorescentlylabeled projections of channelrhodopsin in VMNSF1 neurons detected in terminal targets including the (B) aBNST, (C) PVN, (D) CeA, and (E) PAG. (Magnification: 4×.) 3V,third cerebral ventricle; CA, cerebral aqueduct. (F–I) Blood glucose, (J–M) plasma insulin, and (N–Q) plasma corticosterone (CORT) levels before (Pre), during (Stim), and1 h after (Post) photoactivation of VMNSF1 neuron axon projection fields. (aBNST, n = 7; PVN, n = 6; CeA, n = 5; and PAG, n = 6 per group). *P < 0.05 vs. Pre. (R) Plasmaglucagon 1 h after photoactivation of each site. *P < 0.05 vs. CeA and PAG; no significant difference from PVN. All data are expressed as mean ± SEM.

E2078 | www.pnas.org/cgi/doi/10.1073/pnas.1521160113 Meek et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

020

As expected, analysis of the brains of these mice revealed largenumbers of EGFP (+) cells in the VMN, potentially representingnot only those VMNSF1 neurons initially infected with TVA-RG but also local VMN cells that synapse onto these neurons (Fig.7E). We also detected EGFP (+) neurons in the LPBN as well asseveral other brain regions (Fig. S4). Specifically, labeled neuronswithin the PBN were limited to the external, central, and ventralaspects of the LPBN, with none detected in either the superiorLPBN or the medial PBN (Fig. 7F). These observations demon-strate that a subset of LPBN neurons make synaptic connectionswith those VMNSF1 neurons that project to the aBNST.

DiscussionTo cope with the brain’s exclusive reliance on glucose as a fuel, asophisticated system has evolved to detect and respond to thethreat posed by reduced glucose availability. The brain clearlyplays a central role in this regulatory system, and, whereas pre-vious studies have implicated a role for the VMN, the underlyingneurocircuitry remains largely unidentified. Here, we report acrucial role for VMNSF1 neurons in the physiological response tohypoglycemia and identify them as part of an ascending glucor-egulatory LPBN→VMNSF1→aBNST neurocircuit (Fig. S5). Theseobservations offer direct evidence of a previously unrecognizedneurocircuit involved in glycemic control.A key finding is that, whereas inhibition of VMNSF1 neurons

has no glucose-lowering effect under basal conditions, it severelyimpairs recovery from hypoglycemia because it blocks the pow-erful CRRs that normally restore low blood glucose levels into thenormal range. Conversely, activation of VMNSF1 neurons in other-wise normal mice causes diabetes-range hyperglycemia because itactivates CRRs despite the absence of hypoglycemia. Thesefindings offer compelling, direct evidence that VMNSF1 neuronsfunction as “primary motor output” neurons for raising circulatingglucose concentrations in times of need, analogous to the role infood intake played by Agrp neurons situated in the hypothalamicarcuate nucleus. Although relatively inactive in sated mice, Agrpneurons powerfully stimulate feeding when activated in responseto depleted fuel stores (37, 38). These neurons are thereforepoised to drive positive energy balance to replenish depleted fuelstores but otherwise are relatively inactive. Consequently, in-hibition of Agrp neurons has little effect on food intake in satedmice but impairs fasting-induced feeding (17, 18). Thus, inhibitionof Agrp neurons has little impact on food intake under usualcircumstances but blocks need-based feeding. Similarly, because

activation of VMNSF1 neurons causes marked hyperglycemia, itmakes physiological sense for these neurons to remain inactiveunless blood glucose levels are threatened. In the latter setting,however, inhibition of VMNSF1 neurons has devastating conse-quences because it blocks the powerful CRRs that normally re-store blood glucose levels into the normal range (11, 30), asexpected for neurons that function as a primary motor output forraising blood glucose levels when glucose availability is threatened.Details regarding the cellular phenotype of VMNSF1 neurons

involved in this response await further analysis. Recent work hasfocused on the role of intracellular glucose sensors, neuro-transmitters such as gamma-aminobutyric acid (GABA), gluta-mate, serotonin, and neuropeptides [corticotrophin-releasinghormone (CRH) and urocortins] (1, 30). Most VMNSF1 neuronsare glutamatergic (39, 40), and previous evidence points to a rolefor glutamate release from these neurons in the CRR response(11). However, our data suggest that not all VMNSF1 neuronsparticipate in glucose counterregulation because photoactivationof VMNLepRb neurons, many of which express SF1 and are alsoglutamatergic (21, 22), was without effect on either glycemia orsecretion of counterregulatory hormones. To shed additional lighton the subset of VMNSF1 neurons involved in glucose homeo-stasis, we used fluorescently tagged channelrhodopsin to charac-terize the circuit architecture. Consistent with previous studies(27), we found that the heaviest projections of VMNSF1 neuronsterminate in the PAG, CeA, PVN, and aBNST. Our finding thatphotostimulation of VMNSF1 neurons that project to the aBNST,but not to the PAG, PVN, CeA, or mBNTS, mimics the glycemicresponse elicited by VMNSF1 neuronal stimulation offers clear evi-dence that projections of VMNSF1 neurons to the aBNST contributeto observed effects on glucose homeostasis. Moreover, the glucose-raising effect of VMNSF1→aBNST fiber activation was accom-panied by hormonal responses similar to that induced by stim-ulation of VMNSF1 cell bodies, including increases of plasmaglucagon and corticosterone levels and inhibition of glucose-stimulated insulin secretion. Collectively, these observationsimplicate axonal projections to the aBNST in the glycemic re-sponse observed after photostimulation of VMNSF1 cell bodies.Our finding that photostimulation of VMNSF1→aBNST axons

induces activation of VMNSF1 neuron cell bodies predominantlyin the dmVMN and cVMN, and not the vlVMN, is consistentwith established evidence implicating the dmVMN and cVMN inautonomic control of metabolism whereas neurons in the vlVMNparticipate in the control of aggression and reproduction

Fig. 6. Cell bodies of VMNSF1 neurons projectingto the aBNST are located primarily in central anddorsomedial VMN. (A) Diagram showing unilateralstereotaxic microinjection of fluorescently labeledretrograde neuronal tracer cholera toxin subunit B(CTB) into the aBNST of WT mice and (B) expressionof the virus at injection site and associated coronalillustration for orientation. AC, anterior commis-sure; LV, lateral cerebral ventricle. Expression ofthe fluorescently labeled retrograde neuronaltracer CTB throughout the mediobasal hypothala-mus (MBH) at either magnification (C) 4× or (D)10×. (E) Diagram showing unilateral microinjectionof ChR2-EYFP into the VMN of SF1-Cre+ mice fol-lowed by 1 h photoactivation of the aBNST pro-jection field. (F) Representative 10× magnificationof VMNSF1 projections within the aBNST (green-labeled fibers), with (G) c-Fos (red stain) and the(H) merged image after aBNST photoactivation ofVMNSF1projections. (I) Control staining for c-Foswithin the aBNST when light treatment was withheld (10× magnification) showing low immunoactivity. (J) Expression of ChR2 within the VMN of SF1-Cre+

mice (4× magnification). (K) Representative images of VMNSF1-infected neurons (green), (L) c-Fos (red), and the (M) merged image after photoactivation ofVMNSF1 projections in the aBNST. (Magnification: 10×.) (N) Bilateral view of the VMN showing isolated expression of c-Fos within only the left hemisphere. 3V,third ventricle; c, central; dm, dorsomedial; vl, ventral lateral.

Meek et al. PNAS | Published online March 21, 2016 | E2079

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

020

(33–35). These findings collectively support the hypothesis thatVMNSF1 neurons involved in glucose homeostasis (i) are locatedin dmVMN and cVMN, (ii) project to the aBNST, and (iii) aredistinct from those neurons expressing leptin receptors.In addition to the VMN, recent work implicates a subset of

neurons situated in the LPBN in the CRR to hypoglycemia. TheseLPBN neurons express both leptin receptor and CCK, and, likephotoactivation of VMNSF1 neurons, pharmacogenetic activationof LPBN LepRbCCK neurons raises blood glucose levels in asso-ciation with increased secretion of glucagon and corticosterone.Conversely, inhibition of LPBN LepRbCCK neurons blunts theglycemic response to glucoprivation (9, 10), an effect resemblingthe consequences of optogenetic silencing of VMNSF1 neuronsduring insulin-induced hypoglycemia in the current studies. Becausethese LPBN LepRbCCK neurons project to the dmVMN and cVMNand because the induction of CRRs after activation of PBNLepRbCCK neurons is attenuated by pharmacogenetic inhibitionof VMNSF1 neurons (10), we hypothesized that the formerneurons provide ascending, stimulatory input to the subset ofVMNSF1 neurons that project to the aBNST.In support of this hypothesis, we found that aBNST-projec-

ting VMNSF1 neurons are synaptically connected to neurons inthe LPBN, among other regions. These data offer direct evidencein support of an LPBN→VMNSF1→aBNST neurocircuit involved

in glycemic regulation, and we anticipate that future studies willdemonstrate that LepRbCCK neurons are among those LPBNneurons involved in this circuit. Interestingly, several areas addi-tional to the LPBN send afferents to the subset of VMNSF1

neurons that project to the aBNST. How this information isprocessed and the extent to which these projections contribute toglucose homeostasis are questions that await additional study.The liver plays a key role as the primary source of circulating

glucose mobilized by CRRs to hypoglycemia. Activation of thehypothalamic-pituitary-adrenal (HPA) axis and increased glucagonsecretion (41) each raise blood glucose levels by increasing HGP viaincreases of both glycogenolysis and gluconeogenesis. The latter canbe assessed indirectly by the expression of the hepatic gluconeo-genic genes G6Pase and Pepck, and we found that hyperglycemiainduced by photoactivation of VMNSF1 neurons is associated withincreased hepatic expression of both genes, suggesting that in-creased hepatic gluconeogenesis may have contributed to hyper-glycemia elicited by photoactivation of VMNSF1 neurons. Consistentwith this hypothesis, we found that the glycemic excursion in re-sponse to a pyruvate challenge was also markedly increased byphotoactivation of VMNSF1 neurons.Additional evidence of a role of the VMN in these responses

(42, 43) includes the finding that electrical stimulation of themediobasal hypothalamus elicits hyperglycemia (6), driven in

Fig. 7. aBNST→VMNSF1 afferents originate withinthe PBN. (A) Schematic of CTB distribution afterVMN injection. LPBN, lateral parabrachial nucleus;LSN, lateral septal nucleus; mAMG, medial amyg-dala; MPO, medial preoptic nucleus; PP, peri-peduncular nucleus. (B) Microphotograph of CTB atthe injection site within the VMN. (C) Representa-tive image showing the presence of CTB in the lat-eral PBN 4 d after VMN injection. (Magnification:Left, 4×; Right, 10×.) Additional microphotographsof CTB spread are included in Fig. S3. (D) Diagramshowing procedure and timeline of the modifiedrabies approach (both virus and mode of action).(E) Back propagated modified rabies virus within theVMN from aBNST projecting terminals at 4× (Left)and 10× (Right) magnification. (F) Representativeimage of modified rabies virus in VMNSF1 afferentneurons found within the LPBN at 4× (Left) and 10×(Right) magnification.

E2080 | www.pnas.org/cgi/doi/10.1073/pnas.1521160113 Meek et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

020

part by increased glucagon secretion (44). Similarly, glucagonrelease is triggered by administration of 2-deoxyglucose (2-DG)locally into the VMN to induce neuroglucopenia in this brainarea (45). Conversely, the glucagon response to systemic hypo-glycemia is blocked by intra-VMN administration of glucose (7)and in rats with bilateral VMH lesions (46). Combined with ourfindings that photoinhibition of VMNSF1 neurons blocks theplasma glucagon and corticosterone response to hypoglycemia,whereas photoactivation of these neurons has the opposite ef-fect, we conclude that a neuronal network involving VMNSF1

neurons is critically involved in the CRR to hypoglycemia.In addition to increased counterregulatory hormone secretion,

plasma insulin levels were unchanged during photoactivation ofVMNSF1 neurons, despite an associated, dramatic increase ofblood glucose levels. Although this failure to raise plasma insulinlevels may seem unimpressive, the observed increase of bloodglucose constitutes a powerful stimulus to insulin secretion, andincreased SNS outflow to the pancreas is among very few mech-anisms capable of blocking this response in normal animals. Froma teleological perspective, it makes intuitive sense that, whenglucose availability is threatened, responses that increase glucoseproduction would be activated (e.g., elevated glucagon and corti-costerone levels) in concert with responses that reduce glucoseclearance (e.g., inhibition of insulin secretion), and much of theliterature indicates that pancreatic beta cells are subject to in-hibitory control via the sympathetic nervous system (SNS) (47, 48).For example, bilateral VMN lesioning induces hyperinsulinemia(49, 50) whereas electrical stimulation of the VMN suppressesglucose-induced insulin secretion (51). Further, the VMN isknown to regulate autonomic outflow to the pancreas (52, 53), andactivation of islet sympathetic nerves during glycopenic stress (54)inhibits insulin secretion via a mechanism involving activation ofα2-adrenoreceptors on the beta cell (47, 48). However, the neu-rocircuitry that underlies inhibitory control of insulin secretion bythe brain remains unknown.In conclusion, we report that optogenetic silencing of VMNSF1

neurons profoundly impairs recovery from insulin-induced hy-poglycemia by blocking powerful neuroendocrine and autonomiccomponents of the CRR. These findings offer unambiguous ev-idence that activation of VMNSF1 neurons is required foreffective glucose counterregulation. At the same time, tonicinhibition of these neurons seems to be permissive for mainte-nance of glucose homeostasis under usual conditions becausephotoactivation of VMNSF1 neurons rapidly induces diabetes-range hyperglycemia with impaired glucose tolerance, owing to acombination of increased CRR hormone secretion and in-hibition of glucose-induced insulin secretion. A subset ofVMNSF1 neurons that project to the aBNST and are anatomi-cally linked to upstream neurons in the LPBN are implicated inthis glucoregulatory neurocircuit. An improved understanding ofthe functional organization of this neurocircuit may help toidentify future strategies for prevention of both insulin-inducedhypoglycemia and hypoglycemic unawareness, two of the mostcommon and costly complications of diabetes treatment (55).

Experimental ProceduresAnimals. All procedures were performed in accordance with NIH guidelines forthe care and use of animals and were approved by the Animal Care Committeeat theUniversity ofWashington. All studied animalswere individually housed ina temperature-controlled room with a 12:12-h light:dark cycle under specific-pathogen free (SPF) conditions and provided with ad libitum access to waterand chow unless otherwise stated (PMI Nutrition). SF1-cre mice (approvedmouse gene name, Nr5a1) have been generated and described previously (21,22) and were bred in our colony (56) whereas Lepr-IRES-Cre mice were pur-chased from The Jackson Laboratory (stock no. 008320).

Viral Generation, Injections, and Fiber Placement. The viral vectors AAV5-EF1α-DIO-hChR2(H134R)-EYFP and AAVDJ8-EF1α-DIO-SwiChRCA-TS-EYFP-WPRE usedin these studies have been described previously (57–59). All viruses were

packaged at the Gene Therapy Center at the University of North Carolina,except the SwiChR2 virus, which was packaged into a DJ8 AAV vector by theUniversity of Washington Diabetes Research Center Viral Vector and Trans-genic Mouse Core. The Cre-inducible EF1α-DIO-SwiChRCA-TS-EYFP-WPRE con-struct was kindly provided by Karl Deisseroth (Stanford University, Stanford,CA) (57). Details regarding stereotaxic delivery of viruses to specific brain areasis provided in SI Experimental Procedures.

Optogenetic in Vivo Photostimulation and Photoinhibition. Optogenetic stud-ies were supported by the Nutrition Obesity Research Center (NORC) EnergyBalance and Glucose Metabolism Core at the University of Washington. Fordelivery of light pulses with millisecond precision, the output beam from adiode laser (473 nm, DPSS continuous wave laser system; Laserglow) wascontrolled using an AMPI Master-9 stimulator (Laserglow) through a singlefiber port (17, 60). The light was then split using multimode optical fibers witha 200-nm diameter core, N.A. 0.22 (Thor Laboratories), and passed through afiber optic rotary joint (Thor Laboratories). A terminal fiber attached to therotary joint was fixed to a 230-μm-bore ceramic ferrule and a mating sleevethat allowed delivery of light to the brain through coupling with a 200-nmferrule-capped fiber implanted within the mouse brain.

Unless otherwise indicated, for all in vivo photostimulation experiments,5-ms pulses, 40 pulses per 1 s, were used for 1 h,which is slightly higher than thetraditional stimulation patterns used with electrical activation of the VMN (6,61). Photoinhibition experiments used a constant beam of light for 1–2 h asindicated for each experiment. Irradiance at target regions was estimated at21 mW/mm2 for photostimulation and 10 mW/mm2 for photoinhibition ex-periments based upon the previously described relationship of light scat-tering in the brain of mammals with light power exiting the fiber tipcorresponding to 8 mW and 4 mW for activation and inhibition, respectively(web.stanford.edu/group/dlab/cgi-bin/graph/chart.php). The nature of thenumerical aperture (N.A. 0.22), a measurement dictating the range of angleslight is transmitted along, allowed a greater depth of light penetration intotissue while resulting in a narrow illumination cone. Based on these calcu-lations and the placement of the optic fiber (0.4 mm dorsal to the VMN), theintensity of light reaching the lateral ventral portion of the VMN likelyapproached theminimum threshold to result in regular burst firing (1 mW/mm2)(12). However, in regions that receive an asymmetrical distribution of VMNterminals along a transverse plane, the stimulation of fibers adjacent ratherthan directly below the fiber would have been limited.

Viral Track-Tracing Studies. To investigate the origin of BNST or VMN afferentprojections, 150 nL of 1 mg/mL cholera toxin b subunit conjugated with Alexadye 488 (CTB-488; Life Technologies) in PBS was injected into the region ofinterest at 50 nL/min. Animals were perfused 4 d after injections to harvestbrains for histological analysis.

For retrograde rabies-tracing studies, a modified rabies viral construct,AAV-DIO-TVA-RG (TVA-RG), serotype 8, was injected unilaterally into theVMN (200 nL) of 7- to 8-wk-old mice (titer 1.1 × 1012 genome copies permilliliter). Twenty-one days later, SADΔG-EGFP (EnvA) rabies was unilaterallyinjected into the aBNST (100 nL), followed by a 10-min wait. One week afterSADΔG-EGFP (EnvA) rabies injection, mice were perfused, and brains weremounted for sectioning as described below.

Metabolic Studies and Tissue Processing. Intraperitoneal glucose tolerancetests (GTTs) (2 g/kg body weight) or pyruvate tolerance tests (PTTs) (2 g/kgbody weight) were conducted in 6-h fasted animals. Tail vein blood wascollected for measurement of blood glucose levels at the times indicated andas described in the SI Experimental Procedures.

Immunohistochemistry and RT-PCR. Details for immunostaining and geneexpression estimation can be found in SI Experimental Procedures and aspreviously described (62, 63).

Statistical Analyses. All results are presented as means ± SEM. Statistical analyseswere performed using PASW Statistics (version 18; IBM SPSS). P values for pair-wise comparisons were calculated by two-tailed Student’s t test. Data for com-parisons across more than two groups were analyzed using a one-way ANOVAwith post hoc comparisons, where appropriate. Time course comparisons be-tween groups were analyzed using a two-way repeated-measures ANOVA withmain effects of light (laser on/off) and time (6–8 instances). All post hoc com-parisons were determined using Sidak’s correction for multiple comparisons. Inall instances, probability values of <0.05 were considered significant.

ACKNOWLEDGMENTS. We acknowledge the technical expertise provided byAlexis Cubelo, Annelise Paige Matsuo, Justin Ngoc Lam, Loan Nguyen, and

Meek et al. PNAS | Published online March 21, 2016 | E2081

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

020

Jonathan D. Fischer (University of Washington). This work was supported byNational Institutes of Health Grants DK089056 (to G.J.M.), DK083042,DK090320, and DK101997 (to M.W.S.), and DK098853 (to M.G.M.), by NIHK01 Career Development Award DK097859 (to T.H.M.), by the Nutrition

Obesity Research Center (NORC Grant DK035816), by the Diabetes ResearchCenter (DRC Grant DK17047), and by Diabetes and Metabolism TrainingGrants F32 DK097859 and T32 DK0007247 (University of Washington) andthe Michigan Diabetes Research Center (DRC Grant DK020572).

1. Beall C, Ashford ML, McCrimmon RJ (2012) The physiology and pathophysiology ofthe neural control of the counterregulatory response. Am J Physiol Regul Integr CompPhysiol 302(2):R215–R223.

2. Brodows RG, Pi-Sunyer FX, Campbell RG (1975) Sympathetic control of hepatic gly-cogenolysis during glucopenia in man. Metabolism 24(5):617–624.

3. Ritter RC, Slusser PG, Stone S (1981) Glucoreceptors controlling feeding and bloodglucose: Location in the hindbrain. Science 213(4506):451–452.

4. Cryer PE (2004) Diabetes Mellitus: A Fundamental and Clinical Text, eds LeRoith D,Olefsky JM, Taylor SI (Lippincott Williams & Wilkins, Philadelphia), 3rd Ed.

5. Saberi M, Bohland M, Donovan CM (2008) The locus for hypoglycemic detection shiftswith the rate of fall in glycemia: The role of portal-superior mesenteric vein glucosesensing. Diabetes 57(5):1380–1386.

6. Frohman LA, Bernardis LL (1971) Effect of hypothalamic stimulation on plasma glu-cose, insulin, and glucagon levels. Am J Physiol 221(6):1596–1603.

7. Borg MA, Sherwin RS, Borg WP, Tamborlane WV, Shulman GI (1997) Local ventro-medial hypothalamus glucose perfusion blocks counterregulation during systemichypoglycemia in awake rats. J Clin Invest 99(2):361–365.

8. Routh VH (2010) Glucose sensing neurons in the ventromedial hypothalamus. Sensors(Basel) 10(10):9002–9025.

9. Flak JN, et al. (2014) Leptin-inhibited PBN neurons enhance responses to hypoglyce-mia in negative energy balance. Nat Neurosci 17(12):1744–1750.

10. Garfield AS, et al. (2014) A parabrachial-hypothalamic cholecystokinin neurocircuitcontrols counterregulatory responses to hypoglycemia. Cell Metab 20(6):1030–1037.

11. Tong Q, et al. (2007) Synaptic glutamate release by ventromedial hypothalamicneurons is part of the neurocircuitry that prevents hypoglycemia. Cell Metab 5(5):383–393.

12. Aravanis AM, et al. (2007) An optical neural interface: In vivo control of rodent motorcortex with integrated fiberoptic and optogenetic technology. J Neural Eng 4(3):S143–S156.

13. Ollmann MM, et al. (1997) Antagonism of central melanocortin receptors in vitro andin vivo by agouti-related protein. Science 278(5335):135–138.

14. Shutter JR, et al. (1997) Hypothalamic expression of ART, a novel gene related toagouti, is up-regulated in obese and diabetic mutant mice. Genes Dev 11(5):593–602.

15. Schwartz MW, Woods SC, Porte D, Jr, Seeley RJ, Baskin DG (2000) Central nervoussystem control of food intake. Nature 404(6778):661–671.

16. Xu Y, Elmquist JK, Fukuda M (2011) Central nervous control of energy and glucosebalance: Focus on the central melanocortin system. Ann N Y Acad Sci 1243:1–14.

17. Aponte Y, Atasoy D, Sternson SM (2011) AGRP neurons are sufficient to orchestratefeeding behavior rapidly and without training. Nat Neurosci 14(3):351–355.

18. Krashes MJ, et al. (2011) Rapid, reversible activation of AgRP neurons drives feedingbehavior in mice. J Clin Invest 121(4):1424–1428.

19. Ritter S, Li AJ, Wang Q, Dinh TT (2011) The value of looking backward: The essentialrole of the hindbrain in counterregulatory responses to glucose deficit. Endocrinology152(11):4019–4032.

20. Verberne AJ, Sabetghadam A, Korim WS (2014) Neural pathways that control theglucose counterregulatory response. Front Neurosci 8:38.

21. Dhillon H, et al. (2006) Leptin directly activates SF1 neurons in the VMH, and thisaction by leptin is required for normal body-weight homeostasis. Neuron 49(2):191–203.

22. Bingham NC, Anderson KK, Reuter AL, Stallings NR, Parker KL (2008) Selective loss ofleptin receptors in the ventromedial hypothalamic nucleus results in increased adi-posity and a metabolic syndrome. Endocrinology 149(5):2138–2148.

23. King BM (2006) The rise, fall, and resurrection of the ventromedial hypothalamus inthe regulation of feeding behavior and body weight. Physiol Behav 87(2):221–244.

24. Shimazu T, Ishikawa K (1981) Modulation by the hypothalamus of glucagon and insulinsecretion in rabbits: Studies with electrical and chemical stimulations. Endocrinology108(2):605–611.

25. Gutiérrez-Juárez R, Obici S, Rossetti L (2004) Melanocortin-independent effects ofleptin on hepatic glucose fluxes. J Biol Chem 279(48):49704–49715.

26. Liu L, et al. (1998) Intracerebroventricular leptin regulates hepatic but not peripheralglucose fluxes. J Biol Chem 273(47):31160–31167.

27. Canteras NS, Simerly RB, Swanson LW (1994) Organization of projections from theventromedial nucleus of the hypothalamus: A Phaseolus vulgaris-leucoagglutininstudy in the rat. J Comp Neurol 348(1):41–79.

28. Saper CB, Swanson LW, Cowan WM (1976) The efferent connections of the ventro-medial nucleus of the hypothalamus of the rat. J Comp Neurol 169(4):409–442.

29. Weller KL, Smith DA (1982) Afferent connections to the bed nucleus of the striaterminalis. Brain Res 232(2):255–270.

30. Chan O, Sherwin R (2013) Influence of VMH fuel sensing on hypoglycemic responses.Trends Endocrinol Metab 24(12):616–624.

31. Choi YH, Fujikawa T, Lee J, Reuter A, Kim KW (2013) Revisiting the ventral medialnucleus of the hypothalamus: The roles of SF-1 neurons in energy homeostasis. FrontNeurosci 7:71.

32. Falkner AL, Dollar P, Perona P, Anderson DJ, Lin D (2014) Decoding ventromedialhypothalamic neural activity during male mouse aggression. J Neurosci 34(17):5971–5984.

33. Lee H, et al. (2014) Scalable control of mounting and attack by Esr1+ neurons in theventromedial hypothalamus. Nature 509(7502):627–632.

34. Lin D, et al. (2011) Functional identification of an aggression locus in the mouse hy-pothalamus. Nature 470(7333):221–226.

35. Silva BA, et al. (2013) Independent hypothalamic circuits for social and predator fear.Nat Neurosci 16(12):1731–1733.

36. Wickersham IR, et al. (2007) Monosynaptic restriction of transsynaptic tracing fromsingle, genetically targeted neurons. Neuron 53(5):639–647.

37. Morton GJ, Meek TH, Schwartz MW (2014) Neurobiology of food intake in health anddisease. Nat Rev Neurosci 15(6):367–378.

38. Sternson SM (2013) Hypothalamic survival circuits: Blueprints for purposive behaviors.Neuron 77(5):810–824.

39. Ovesjö ML, Gamstedt M, Collin M, Meister B (2001) GABAergic nature of hypotha-lamic leptin target neurones in the ventromedial arcuate nucleus. J Neuroendocrinol13(6):505–516.

40. Ziegler DR, Cullinan WE, Herman JP (2002) Distribution of vesicular glutamatetransporter mRNA in rat hypothalamus. J Comp Neurol 448(3):217–229.

41. Cryer PE (1981) Glucose counterregulation in man. Diabetes 30(3):261–264.42. Begg DP, Woods SC (2013) Interactions between the central nervous system and

pancreatic islet secretions: A historical perspective. Adv Physiol Educ 37(1):53–60.43. Woods SC, Porte D, Jr (1974) Neural control of the endocrine pancreas. Physiol Rev

54(3):596–619.44. Nonogaki K, Iguchi A (1997) Role of central neural mechanisms in the regulation of

hepatic glucose metabolism. Life Sci 60(11):797–807.45. Borg WP, Sherwin RS, During MJ, Borg MA, Shulman GI (1995) Local ventromedial

hypothalamus glucopenia triggers counterregulatory hormone release. Diabetes44(2):180–184.

46. Borg WP, et al. (1994) Ventromedial hypothalamic lesions in rats suppress counter-regulatory responses to hypoglycemia. J Clin Invest 93(4):1677–1682.

47. Ahrén B (2000) Autonomic regulation of islet hormone secretion: Implications forhealth and disease. Diabetologia 43(4):393–410.

48. Nonogaki K (2000) New insights into sympathetic regulation of glucose and fat me-tabolism. Diabetologia 43(5):533–549.

49. Frohman LA, Bernardis LL (1968) Growth hormone and insulin levels in weanling ratswith ventromedial hypothalamic lesions. Endocrinology 82(6):1125–1132.

50. Rohner-Jeanrenaud F, Jeanrenaud B (1980) Consequences of ventromedial hypotha-lamic lesions upon insulin and glucagon secretion by subsequently isolated perfusedpancreases in the rat. J Clin Invest 65(4):902–910.

51. Dubuc PU, Leshin LS, Willis PL (1982) Glucose and endocrine responses to hypotha-lamic electrical stimulation in rats. Am J Physiol 242(3):R220–R226.

52. Osundiji MA, Evans ML (2013) Brain control of insulin and glucagon secretion.Endocrinol Metab Clin North Am 42(1):1–14.

53. Buijs RM, Chun SJ, Niijima A, Romijn HJ, Nagai K (2001) Parasympathetic and sym-pathetic control of the pancreas: A role for the suprachiasmatic nucleus and otherhypothalamic centers that are involved in the regulation of food intake. J CompNeurol 431(4):405–423.

54. Havel PJ, Taborsky GJ, Jr (1989) The contribution of the autonomic nervous system tochanges of glucagon and insulin secretion during hypoglycemic stress. Endocr Rev10(3):332–350.

55. Frier BM (2014) Hypoglycaemia in diabetes mellitus: Epidemiology and clinical im-plications. Nat Rev Endocrinol 10(12):711–722.

56. Meek TH, et al. (2013) Leptin action in the ventromedial hypothalamic nucleus issufficient, but not necessary, to normalize diabetic hyperglycemia. Endocrinology154(9):3067–3076.

57. Berndt A, Lee SY, Ramakrishnan C, Deisseroth K (2014) Structure-guided trans-formation of channelrhodopsin into a light-activated chloride channel. Science344(6182):420–424.

58. Boyden ES, Zhang F, Bamberg E, Nagel G, Deisseroth K (2005) Millisecond-timescale,genetically targeted optical control of neural activity. Nat Neurosci 8(9):1263–1268.

59. Zhang F, et al. (2011) The microbial opsin family of optogenetic tools. Cell 147(7):1446–1457.

60. Betley JN, Cao ZF, Ritola KD, Sternson SM (2013) Parallel, redundant circuit organi-zation for homeostatic control of feeding behavior. Cell 155(6):1337–1350.

61. Frohman LA, Bernardis LL, Stachura ME (1974) Factors modifying plasma insulin andglucose responses to ventromedial hypothalamic stimulation. Metabolism 23(11):1047–1056.

62. Morton GJ, et al. (2005) Leptin action in the forebrain regulates the hindbrain re-sponse to satiety signals. J Clin Invest 115(3):703–710.

63. Blevins JE, Schwartz MW, Baskin DG (2004) Evidence that paraventricular nucleusoxytocin neurons link hypothalamic leptin action to caudal brain stem nuclei con-trolling meal size. Am J Physiol Regul Integr Comp Physiol 287(1):R87–R96.

64. Morton GJ, et al. (2005) Leptin regulates insulin sensitivity via phosphatidylinositol-3-OH kinase signaling in mediobasal hypothalamic neurons. Cell Metab 2(6):411–420.

65. Franklin KBJ, Paxinos G (1997) The Mouse Brain in Stereotaxic Coordinates. (Aca-demic, San Diego).

66. German JP, et al. (2011) Leptin activates a novel CNS mechanism for insulin-independentnormalization of severe diabetic hyperglycemia. Endocrinology 152(2):394–404.

67. German JP, et al. (2010) Leptin deficiency causes insulin resistance induced byuncontrolled diabetes. Diabetes 59(7):1626–1634.

E2082 | www.pnas.org/cgi/doi/10.1073/pnas.1521160113 Meek et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

12, 2

020