GABAergic RIP-Cre Neuronsin the Arcuate Nucleus SelectivelyRegulate Energy ExpenditureDong Kong,1,6 Qingchun Tong,1,3,6 Chianping Ye,1 Shuichi Koda,1,4 Patrick M. Fuller,2 Michael J. Krashes,1 Linh Vong,1

Russell S. Ray,5 David P. Olson,1 and Bradford B. Lowell1,*1Division of Endocrinology, Department of Medicine2Department of Neurology

Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, MA 02115, USA3Brown Foundation Institute of Molecular Medicine, University of Texas Health Science Center at Houston, Houston, TX 77030, USA4Asubio Pharma, Kobe 650-0047, Japan5Department of Genetics, Harvard Medical School, Boston, MA 02115, USA6These authors contributed equally to this work

*Correspondence: blowell@bidmc.harvard.edu

http://dx.doi.org/10.1016/j.cell.2012.09.020

SUMMARY

Neural regulation of energy expenditure is incom-pletely understood. By genetically disruptingGABAergic transmission in a cell-specific fashion,and by combining this with selective pharmacoge-netic activation and optogenetic mapping tech-niques, we have uncovered an arcuate-based circuitthat selectively drives energy expenditure. Specifi-cally, mice lacking synaptic GABA release fromRIP-Cre neurons have reduced energy expenditure,become obese and are extremely sensitive to high-fat diet-induced obesity, the latter due to defectivediet-induced thermogenesis. Leptin’s ability tostimulate thermogenesis, but not to reduce feeding,is markedly attenuated. Acute, selective activationof arcuate GABAergic RIP-Cre neurons, whichmonosynaptically innervate PVH neurons projectingto the NTS, rapidly stimulates brown fat andincreases energy expenditure but does not affectfeeding. Importantly, this response is dependentupon GABA release from RIP-Cre neurons. Thus,GABAergic RIP-Cre neurons in the arcuate selec-tively drive energy expenditure, contribute to leptin’sstimulatory effect on thermogenesis, and protectagainst diet-induced obesity.

INTRODUCTION

Neural circuits operating within and extending beyond the

hypothalamus regulate food intake and energy expenditure.

Proopiomelanocortin (POMC)-expressing neurons and adja-

cent agouti-related peptide (AgRP)-expressing neurons, both

located in the arcuate nucleus (ARC) of the hypothalamus, are

key components of this circuitry. The ARC also contains

many other neurons that express neither POMC nor AgRP

(i.e., non-POMC, non-AgRP neurons). Due to their location in

the ARC, these could also regulate energy balance. A subset

of these ‘‘non-POMC, non-AgRP’’ neurons has clear neuroen-

docrine functions, such as those expressing kisspeptin, growth

hormone-releasing hormone, or dopamine. ‘‘Non-POMC, non-

AgRP’’ neurons whose function is something other than neuro-

endocrine also likely exist. These neurons, like POMC and AgRP

neurons, could be additional, important regulators of energy

balance.

In Rip-Cre transgenic mice (Postic et al., 1999), the rat

insulin-2 (Ins2) promoter drives cre expression in both pancre-

atic b cells and the brain (Song et al., 2010; Wicksteed et al.,

2010). Neurons expressing cre activity, hereafter referred to

as ‘‘RIP-Cre neurons,’’ are distributed in many hypothalamic

sites, including the ARC. Within the ARC, RIP-Cre neurons

are intermingled with, but are distinct from, POMC and AgRP

neurons (Choudhury et al., 2005). Deletion of various loxed

alleles in Rip-Cre mice, often with the initial intent of manipu-

lating gene expression in pancreatic b cells, has marked effects

on energy balance (Chakravarthy et al., 2007; Choudhury et al.,

2005; Covey et al., 2006; Kubota et al., 2004; Lin et al., 2004;

Mori et al., 2009). It is generally believed that this is secondary

to altered function of RIP-Cre neurons as opposed to pancre-

atic b cells. Notably, disruption of leptin signaling in Rip-Cre

transgenic mice causes obesity (Covey et al., 2006). Because

RIP-Cre neurons are found in many locations in the brain,

the specific subgroup of RIP-Cre neurons responsible for

regulating energy balance is unknown. Similarly, the relevant

neurotransmitter(s) and downstream circuitry are likewise

unknown.

Synaptic transmission via glutamate and GABA are critical

components of the hypothalamic circuitry regulating energy

balance (Cowley et al., 2001; Liu et al., 2012; Pinto et al., 2004;

Tong et al., 2008; van den Pol, 2003; Vong et al., 2011; Wu

et al., 2009; Yang et al., 2011). Consequently, it is of interest to

Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc. 645

Figure 1. Generation of Mice Lacking VGAT

in RIP-Cre Neurons

(A) Immunodetection of GFP in the brain of Rip-

Cre, lox-GFP mice. 3V, third ventricle.

(B and C) In situ hybridization for VgatmRNA in the

brain of (B) control (Vgatflox/flox) and (C) Rip-Cre,

Vgatflox/flox littermates. Arrows indicate the regions

with notable reduction of Vgat mRNA signal in

Rip-Cre, Vgatflox/flox mice.

(D and E) Quantitative PCR results of Vgat mRNA

in (D) mediobasal hypothalamus and (E) isolated

pancreatic islets of 2-month-old littermates

(mean ± SEM; n = 4). N.D., nondetectable. *p <

0.05, unpaired t tests.

Scale bars in (A)–(C) represent 100 mm. See also

Table S1.

establish if synaptic release of glutamate or GABA, specifically

by RIP-Cre neurons, regulates energy balance. For the purpose

of addressing such questions, we previously generated mice

with loxed alleles of the vesicular glutamate transporter 2

(VGLUT2, required for synaptic release of glutamate from hypo-

thalamic neurons) and the vesicular GABA transporter (VGAT,

required for synaptic release of GABA from GABAergic neurons)

(Tong et al., 2007, 2008). In the present study, we have deleted

Vglut2 and Vgat specifically from RIP-Cre neurons. This

has uncovered a critical role for synaptic release of GABA,

but not glutamate, from RIP-Cre neurons, in selectively stimu-

lating energy expenditure. Using pharmacogenetic techniques

(Alexander et al., 2009; Krashes et al., 2011), we go on to

establish that RIP-Cre neurons located specifically in the ARC

drive this effect. Finally, using channelrhodopsin-assisted circuit

mapping (Atasoy et al., 2008; Petreanu et al., 2007), we identify

the downstream circuitry selectively engaged by ARC RIP-Cre

neurons.

RESULTS

Neuronal Expression of Cre Recombinase in Rip-Cre

Transgenic MiceRip-Cre transgenicmicewere crossedwith cre-dependent green

fluorescent protein (GFP) reporter mice (Novak et al., 2000) to

generate Rip-Cre, lox-GFP mice, and immunohistochemistry

for GFP was performed. GFP-positive neurons (i.e., RIP-Cre

neurons)wereprimarily found in the hypothalamusandoccasion-

ally in the cortex and striatum (Figure 1A; see Table S1 available

online). Within the hypothalamus, GFP-positive neurons were

observed in the ARC, the ventromedial hypothalamus (VMH),

the medial tuberal nucleus (MTu), the suprachiasmatic nucleus

(SCN), and the dorsomedial hypothalamus (DMH). These obser-

vations are consistent with previous reports by Choudhury et al.

646 Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc.

(2005), Lin et al. (2004), Song et al.

(2010), and Wicksteed et al. (2010).

Generation of Mice Lacking VGATin RIP-Cre NeuronsRip-Cre transgenic mice were crossed

with Vgatflox/flox mice. The resulting Rip-

Cre, Vgatflox/+ mice were then crossed with Vgatflox/flox mice

to obtain Rip-Cre, Vgatflox/flox study subjects and their control

littermates (Vgatflox/flox mice and Rip-Cre, Vgatflox/+ mice). Rip-

Cre, Vgatflox/+ littermates were included as controls to rule

out nonspecific effects of the Rip-Cre transgene (Lee et al.,

2006). In situ hybridization studies for Vgat mRNA were then

performed to assess for GABAergic neurons and also for

sites where the Rip-Cre transgene disrupts Vgat expression.

In Vgatflox/flox control mice, Vgat mRNA signal was detected

in sites known to contain GABAergic neurons. With regard to

sites shown in Figure 1B, these include, within the hypo-

thalamus, the ARC, DMH and MTu; and beyond the hypothal-

amus, the central amygdala (CeA) and the reticular nucleus of

thalamus (RT). In Rip-Cre, Vgatflox/flox mice (Figure 1C), Vgat

signal was substantially reduced in sites containing RIP-Cre

neurons, such as the ARC, DMH, and MTu. In sites where

RIP-Cre neurons are not found, such as the CeA and RT,

Vgat signal, as expected, was unchanged. Finally, we mea-

sured Vgat mRNA levels in the mediobasal hypothalamus,

a region that includes the ARC, DMH, and MTu, which are

three sites suggested by the aforementioned analysis to

contain GABAergic RIP-Cre neurons. This quantitative analysis

confirmed that Vgat mRNA, but not that of a control transcript

(Ucp2 mRNA), was substantially reduced in Rip-Cre, Vgatflox/flox

mice (Figure 1D).

Vgat mRNA was undetectable in both control and Rip-Cre,

Vgatflox/flox pancreatic islets (Figure 1E), whereas, as noted

above, using the identical assay, Vgat mRNA was readily de-

tected in the hypothalamus (Figure 1D). Confirming the intact

nature of pancreatic islet RNA, Ucp2, a gene expressed in islets

(Zhang et al., 2001), was readily detected in the same samples

(Figure 1E). Thus, mouse islets express little or no Vgat mRNA.

Consequently, deletion of the Vgat gene in pancreatic b cells

should produce no effects.

Figure 2. Energy Balance in Mice Lacking VGAT in Rip-Cre Neurons

(A–D) (A) Body weight, (B) body fat mass (3 months old), (C) daily food intake

(2months old), and (D) locomotor activity (2 months old) of ad libitum chow-fed

male littermates (n = 8–10). Black bars in (D) indicate dark cycles.

(E and F) Oxygen consumption of 2-month-old male littermates expressed (E)

per body weight and (F) per animal (n = 8).

(G) H&E of iBAT from 2-month-old littermates.

(H and I) (H) Temperature (n = 12) and (I)Ucp1mRNA expression (n = 4) in iBAT.

Data are presented asmean ±SEM. *p < 0.05 and ***p < 0.001, unpaired t tests

compared with Vgatflox/flox group. See also Figures S1 and S2.

Energy Balance in Mice Lacking VGAT in RIP-CreNeuronsWhen fed a standard chow diet, Rip-Cre, Vgatflox/flox mice have

modestly increased body weight (Figure 2A) and, at 3 months

of age, markedly increased fat stores (Figure 2B). Lean body

mass, on the other hand, was unchanged (data not shown).

The possible causes of positive energy balance were then as-

sessed in 2-month-old animals. Food intake (Figure 2C) and

locomotor activity (Figure 2D) were found to be unchanged.

Obesity, in the face of normal food intake, strongly implicates

reduced energy expenditure in Rip-Cre, Vgatflox/flox mice. This

was confirmed by direct assessments of energy expenditure.

Oxygen consumption was markedly reduced in Rip-Cre,

Vgatflox/flox mice, and this was apparent when data were ex-

pressed per body weight (Figure 2E) or per animal (Figure 2F).

Thus, obesity in Rip-Cre, Vgatflox/flox mice is due entirely to a

selective reduction in energy expenditure.

We next analyzed brown adipose tissue (BAT), a well-estab-

lished mediator of thermogenesis (Cannon and Nedergaard,

2004). Interscapular BAT (iBAT) of Rip-Cre, Vgatflox/flox mice

was markedly enlarged and pale in comparison with that from

control littermates. As shown in Figure 2G, BAT from Rip-Cre,

Vgatflox/floxmice contained larger cells with unilocular triglyceride

deposits, similar to that observed in animals with defective

sympathetic activation of BAT (Bachman et al., 2002). As as-

sessed by biotelemetry probes implanted subcutaneously in

the interscapula fossa beneath iBAT (Enriori et al., 2011), iBAT

temperature, an index of BAT thermogenesis, was reduced in

Rip-Cre, Vgatflox/flox mice (Figure 2H). In contrast, subcutaneous

temperature of a flank site devoid of BAT was similar in the

two groups (Figure S1A). Finally, expression of Ucp1 mRNA,

which encodes the BAT-specific thermogenic molecule, uncou-

pling protein 1, was significantly lower in Rip-Cre, Vgatflox/flox

mice (Figure 2I). These results indicate that GABA release from

RIP-Cre neurons regulates BAT thermogenic function and

suggest that decreased BAT activity is, at least in part, respon-

sible for reduced energy expenditure inRip-Cre,Vgatflox/floxmice.

Given that Rip-Cre transgenic mice express cre in some SCN

neurons (Figure S1B) and that virtually all SCN neurons are

GABAergic (Figure S1C), Rip-Cre, Vgatflox/flox mice could have

altered circadian regulation, which could in turn affect energy

balance (Bass and Takahashi, 2010). To address this possibility,

body temperature (Tb), a reliable indicator of circadian clock

activity (Fuller et al., 2008), was measured using implanted

biotelemetry. As shown in Figures S1D and S1E, Rip-Cre,

Vgatflox/flox mice displayed diurnal and circadian Tb patterns

(phasing, amplitude, or period) that were comparable to controls.

Thus, the circadian clock does not appear to be altered or other-

wise dysfunctional in Rip-Cre, Vgatflox/flox mice and unlikely con-

tributes to the metabolic phenotypes of the Rip-Cre, Vgatflox/flox

mice. Likewise, serum T4 and corticosterone levels, two other

potential regulators of energy balance, were found to be

unchanged in Rip-Cre, Vgatflox/flox mice (Figures S1F and S1G).

Energy Balance in Mice Lacking VGLUT2in RIP-Cre NeuronsSome RIP-Cre neurons are glutamatergic, such as those in

the VMH. To assess if glutamatergic RIP-Cre neurons regulate

energy homeostasis, mice lacking VGLUT2 in RIP-Cre neurons

(Rip-Cre, Vglut2flox/flox mice) were generated as was done for

Rip-Cre, Vgatflox/flox mice. In situ hybridization analyses revealed

that, as expected, Vglut2 mRNA was dramatically reduced in

the VMH of Rip-Cre, Vglut2flox/flox mice (Figures S2A and S2B).

Of note, Rip-Cre, Vglut2flox/flox mice had normal body weight,

oxygen consumption, and food intake (Figures S2C–S2E),

indicating that glutamate release from RIP-Cre neurons is not

required for regulation of energy balance.

Diet-Induced Obesity in Mice Lacking VGATin RIP-Cre NeuronsIncreased energy expenditure following the ingestion of highly

palatable, calorically dense diets, a phenomenon often referred

Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc. 647

Figure 3. Diet-Induced Obesity in Mice

Lacking VGAT in RIP-Cre Neurons

(A and B) Body weight on HFD (A) and daily food

intake (B) averaged over the first 2 weeks on HFD

(n = 8–10).

(C) Oxygen consumption expressed per body

weight during the transition from chow to HFD

(n = 8). CD, averaged oxygen consumption over

3 days on chow; HD1, HD2 and HD3, oxygen

consumption during day 1, 2, and 3, respectively,

on HFD. The percent increase in oxygen con-

sumption on HFD above that on chow diet is

indicated above each bar.

(D) Ucp1 mRNA level in iBAT of HFD-treated

littermates (n = 6–8).

Data are presented as mean ± SEM. *p < 0.05

and **p < 0.01, unpaired t tests compared with

Vgatflox/flox group.

to as diet-induced thermogenesis, plays an important role in

resisting diet-induced obesity (Bachman et al., 2002). To deter-

mine if GABAergic RIP-Cre neuron-driven energy expenditure

is involved in this adaptive response, Rip-Cre, Vgatflox/flox mice

and control animals were fed a high-fat, high-sucrose diet

(HFD) from between 6 and 26 weeks of age. As shown in

Figure 3A, compared to control mice, Rip-Cre, Vgatflox/flox mice

developed massive obesity. Of interest, this diet-induced

obesity was not caused by increased food intake (Figure 3B),

strongly implicating impaired diet-induced thermogenesis. To

directly assess this, oxygen consumption was measured during

the transition from chow to HFD (3 days on chow followed by

3 days on HFD) in 7-week-old mice (note, body weight is normal

at this early age). As shown in Figure 3C, oxygen consumption of

Rip-Cre,Vgatflox/flox mice, on both chow and HFD, was markedly

lower than that observed in control animals. Importantly,

whereas HFD increased energy expenditure by 7%–12% in

control mice, this response was markedly blunted in Rip-Cre,

Vgatflox/flox mice (increased by only 3%–5% in response to

HFD). In addition, HFD-treated Rip-Cre,Vgatflox/flox mice

(6 months old and 20 weeks on HFD) exhibited reduced

Ucp1 mRNA expression in BAT (Figure 3D). In summary, Rip-

Cre, Vgatflox/flox mice are extremely sensitive to diet-induced

obesity, and this is due entirely to a defect in diet-induced

thermogenesis.

Effects of Leptin Treatment in Mice Lacking VGATin RIP-Cre NeuronsGenetic deletion of leptin receptors (LEPRs) from RIP-Cre

neurons causes marked obesity without affecting food intake

(Covey et al., 2006), suggesting a defect in energy expenditure.

Given that Rip-Cre,Vgatflox/flox mice also have defective energy

648 Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc.

expenditure, this raises the possibility

that GABA release from RIP-Cre neurons

mediates leptin’s effects on energy

expenditure. To test this, changes in

body weight and food intake were moni-

tored following injection of saline or leptin.

In control mice, treatment with leptin

reduced body weight (Figures 4A and 4B) and food intake

(Figures 4C and 4D). Of note, whereas the ability of leptin to

reduce food intake in Rip-Cre,Vgatflox/flox mice was completely

intact (Figure 4D), its ability to reduce body weight was markedly

attenuated (Figure 4B). Attenuation of body weight loss in

the face of intact inhibition of food intake indicates that leptin’s

ability to increase energy expenditure is impaired in Rip-Cre,

Vgatflox/flox mice. To directly assess leptin action on energy

expenditure, in particular, in stimulating BAT activity, iBAT

temperature and Ucp1 mRNA were measured. In control mice,

leptin but not saline dramatically and rapidly increased the

temperature of iBAT (as previously described by Enriori et al.,

2011) (Figures 4E and 4F), but not the temperature of a subcuta-

neous flank site devoid of BAT (Figures S3A and S3B). Leptin

also markedly increased Ucp1 mRNA levels (Figure 4G). Of

note, these stimulatory effects of leptin on iBAT temperature

and Ucp1 mRNA levels were attenuated in Rip-Cre,Vgatflox/flox

animals (Figures 4E–4G). Thus, GABA release from RIP-Cre

neurons is required for leptin to fully stimulate energy expendi-

ture, but not for leptin to inhibit feeding.

To determine which subset of RIP-Cre neurons in the hypo-

thalamus expresses LEPRs, leptin-induced phosphorylation of

STAT3 (Tyr705, pSTAT3), a marker for LEPR activity (Munzberg

et al., 2004), was assessed in Rip-Cre, lox-GFP mice. The

neurons double positive for pSTAT3 and RIP-Cre activity were

mainly observed in the ARC and the VMH (Figures 4H, 4I, and

S3C–S3N). The DMH, which contained both RIP-Cre neurons

and pSTAT3-positive neurons, exhibited negligible colocali-

zation (Figures 4I and S3I–S3K). Because VMH neurons are

glutamatergic and not GABAergic (Vong et al., 2011), all leptin-

responsive, GABAergic RIP-Cre neurons appear to be located

in the ARC. Collectively, given the aforementioned observations,

Figure 4. Response to Leptin in Mice Lack-

ing VGAT in RIP-Cre Neurons

(A–F) The effects of saline or leptin on (A and B)

body weight, (C and D) daily food intake, and

(E and F) iBAT temperature in 2-month-old male

littermates (n = 8–12). *p < 0.05, paired t tests

compared with animals of the same genotype

before leptin injection (i.e., time point 0); #p < 0.05,

unpaired t tests compared with control animals

at given time point.

(G) Ucp1 mRNA level in iBAT 4 hr after saline (Sal)

or leptin (Lep) injection (n = 6). *p < 0.05 and

**p < 0.01, unpaired t tests compared with saline-

injected animals of the same genotype; #p < 0.05,

unpaired t tests compared with Vgatflox/flox animals

of the same treatment.

(H) Double immunohistochemistry for GFP (green)

and leptin-induced phosphorylation of STAT3

(Tyr105, pSTAT3, magenta) in the ARC of Rip-Cre,

lox-GFP mice. Arrows indicate the neurons with

coexpression of GFP and pSTAT3.

(I) Quantification of the neurons that expressed

GFP, pSTAT3, or both in the hypothalamic nuclei

(n = 3 mice).

Data are presented as mean ± SEM. See also

Figures S3 and S5.

it is likely that GABAergic RIP-Cre neurons in the ARC mediate

leptin’s stimulatory effect on energy expenditure.

Pharmacogenetic Activation of RIP-Cre Neuronsin the ARCTo directly test the ability of ARC RIP-Cre neurons to drive

energy expenditure, we used the pharmacogenetic approach

referred to as designer receptors exclusively activated by

designer drugs (DREADD). The stimulatory DREADD, hM3Dq,

is activated by the otherwise inert, brain-penetrable compound,

clozapine-N-oxide (CNO) (Alexander et al., 2009). The cre-

Cell 151, 645–657,

dependent adeno-associated virus, AAV-

Flex-hM3Dq-mCherry (Krashes et al.,

2011), was stereotaxically injected into

the ARC of 3- to 4-week-old Rip-Cre

transgenic mice (Figure 5A), and studies

were performed 2–3 weeks after injec-

tion. Brain slice electrophysiology studies

confirmed that CNO depolarizes and

increases the firing rate of hM3Dq-ex-

pressing RIP-Cre neurons (Figure 5B),

but not control non-hM3Dq-expressing

RIP-Cre neurons (Figure S4A). hM3Dq

virus was then bilaterally injected into the

ARC of 5- to 6-week-old Vgatflox/flox mice,

Rip-Cre mice, or Rip-Cre, Vgatflox/flox

mice. Studies were performed 3 weeks

after injection. The mCherry fusion tag

was exclusively detected in the ARC of

Rip-Cre mice and Rip-Cre, Vgatflox/flox

mice and was absent in the ARC of

Vgatflox/flox mice (because these mice

lack cre activity that enables hM3Dq expression) (Figure 5C).

When CNO was injected in vivo, c-fos immunoreactivity was

markedly increased in the ARC of Rip-Cre mice and Rip-Cre,

Vgatflox/flox mice, but not in the ARC of Vgatflox/flox mice (Fig-

ure 5D). Thus, in vivo treatment with CNO activates hM3Dq-

expressing RIP-Cre neurons in the ARC.

To assess effects on energy expenditure, virus-injected

animals were housed individually in metabolic cages, and

oxygen consumption was monitored following injection with

saline or CNO. After an acclimation period, each mouse was

injected with saline on the first day followed by CNO on the

October 26, 2012 ª2012 Elsevier Inc. 649

Figure 5. Pharmacogenetic Activation of

ARC RIP-Cre Neurons

(A) Diagram of AAV-Flex-hM3Dq-mCherry (left)

and schematic indication of the stereotaxic injec-

tion into the ARC of Rip-Cre transgenic mice

(right).

(B) Representative whole-cell, current-clamp

recording from an ARC RIP-Cre neuron marked by

mCherry fluorescence from a Rip-Cre, lox-GFP

mouse injected with AAV-Flex-hM3Dq-mCherry

virus.

(C and D) Immunohistochemistry for (C) mCherry

and (D) CNO-induced c-Fos (DAB, black stain) in

the ARC of virus-injected Vgatflox/flox (top), Rip-Cre

(middle), and Rip-Cre, Vgatflox/flox (bottom) male

mice.

(E–G) Oxygen consumption over a 24 hr period

(left) and during the first 4 hr (right) following i.p.

injection of saline or CNO (n = 8). *p < 0.05, **p <

0.01, ***p < 0.001, paired t tests compared to

saline groups.

(H–J) iBAT temperature (Temp) over 4 hr following

saline or CNO injection (n = 6–8). *p < 0.01, paired

t test compared to animals of the same genotype

before CNO injection; #p < 0.01, paired t tests

compared to saline-injected animals at given time

point.

(K–M) Ucp1 mRNA in iBAT 6 hr after saline or

CNO injection (n = 4–6). *p < 0.05, unpaired t tests

compared with saline-injected animals of the

same genotype.

Data are presented as mean ± SEM. See also

Figure S4.

second day. Selective activation of ARC RIP-Cre neurons with

CNO rapidly increased oxygen consumption (Figure 5F), and

this effect lasted for approximately 9 hr. Importantly, CNO

had no effect on oxygen consumption in Rip-Cre, Vgatflox/flox

mice (Figure 5G), which are notable for their inability to release

GABA. Similarly, CNO had no effect on oxygen consumption

in control mice (i.e., Vgatflox/flox mice), which do not express

hM3Dq (Figure 5E). Remarkably, and consistent with our earlier

findings suggesting that GABAergic RIP-Cre neurons selec-

tively control energy expenditure, food intake was unaltered

650 Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc.

by CNO treatment (Figure S4B). We

next assessed BAT activity during

stimulation of ARC RIP-Cre neurons. In

Rip-Cre mice, CNO significantly in-

creased iBAT temperature (Figure 5I),

but not subcutaneous flank tempera-

ture (Figure S4C), and also markedly

increased Ucp1 mRNA (Figure 5L).

CNO did not stimulate iBAT temperature

or Ucp1 mRNA in Vgatflox/flox mice (which

do not express hM3Dq) or in Rip-Cre,

Vgatflox/flox mice (which are unable to

release GABA) (Figures 5H, 5J, 5K, and

5M). These results demonstrate that

synaptic GABA release from ARC RIP-

Cre neurons selectively stimulates BAT activity and energy

expenditure.

Electrophysiologic Effects of Leptin on ARCRIP-Cre NeuronsWhole-cell current-clamp recordings were performed as previ-

ously described by Dhillon et al. (2006) on arcuate RIP-Cre

neurons visualized by expression of GFP in Rip-Cre, lox-GFP

mice. In data not shown, ARC RIP-Cre neurons exhibited

heterogeneous responses to leptin: 30% of neurons (6 of 20)

Figure 6. Projection of ARC RIP-Cre

Neurons

(A) Diagram of AAV-Flex-ChR2(H134R)-mCherry

(left) and schematic indication of the stereotaxic

injection into the ARC of Rip-Cre transgenic mice

(right).

(B) Representative voltage tracing showing light-

driven spikes in a current-clamped arcuate neuron

marked by mCherry fluorescence. Blue tick marks

represent 0.5 ms light flashes at 0.5 Hz.

(C–E) Immunohistochemistry for mCherry in the

hypothalamus of virus-injected Rip-Cre transgenic

mice. mCherry-expressing RIP-Cre neurons in the

ARC are shown in (C) and in a zoomed view in (D).

(E) mCherry-expressing RIP-Cre neuron fibers in

the PVH. ME, median eminence; fx, fornix; opt,

optic tract.

(F) (i) Light-evoked IPSCs in a PVH neuron before

(left) and after (right) the addition of 20 mM bicu-

culline in response to clusters of light pulses.

Blue tick marks represent 0.5 ms light flash at

0.5 Hz. (ii) Zoomed in view of response to a single

pulse of light.

See also Figure S6 and Table S2.

were excited (depolarized membrane potential and increased

firing rate), 35% (7 of 20) were inhibited (hyperpolarized

membrane potential and decreased firing rate), and 35% (7 of

20) were not affected by leptin. As discussed in the following

section, the paraventricular hypothalamus (PVH) is the likely

downstream site that mediates the thermogenic effects of

GABAergic ARC RIP-Cre neurons. To assess the effects of leptin

on PVH-projecting ARC RIP-Cre neurons, retrograde red fluo-

rescent beads were stereotaxically injected into the PVH of

Rip-Cre, lox-GFP mice (Figures S5A and S5B). Retrogradely

transported beads were observed in sites known to innervate

the PVH, including the ARC, DMH, SCN, nucleus of the solitary

tract (NTS), and posterodorsal medial amygdala (MEpd) (data

not shown). In the ARC, 62% (194 of 312) of RIP-Cre neurons

contained red beads (Figure S5C). Recordings were then per-

formed on PVH-projecting (GFP+/beads+) and in non-PVH-pro-

jecting (GFP+/beads�) ARC RIP-Cre neurons. Ionotropic gluta-

mate (kynurenate) and GABA (PTX) receptor blockers were

added to minimize indirect effects of leptin. As shown in Fig-

ure S5D, 9 of 15 PVH-projecting ARC RIP-Cre neurons were

directly excited by leptin, 6 of 15 were unaffected by leptin,

Cell 151, 645–657,

and 0 of 15 were inhibited by leptin. In

contrast, 0 of 12 non-PVH-projecting

ARC RIP-Cre neurons were excited by

leptin, 9 of 12 were unaffected by leptin,

and 3 of 12 were directly inhibited by lep-

tin (Figure S5E). Thus, leptin excites the

majority of PVH-projecting ARC RIP-Cre

neurons.

Downstream NeurocircuitryEngaged by ARC RIP-Cre NeuronsWenext used channelrhodopsin2 (ChR2)-

assisted circuit mapping (Atasoy et al.,

2008; Petreanu et al., 2007) to identify proximal downstream

neurons that could mediate stimulation of energy expenditure

by GABAergic ARC RIP-Cre neurons. A virus that conditionally

expresses ChR2-mCherry fusion protein (AAV-Flex-ChR2

(H134R)-mCherry) (Zhang et al., 2007) in the presence of cre re-

combinase was unilaterally injected into the ARC of 3- to 4-

week-oldRip-Cremice (Figure 6A).Micewere studied 2–3weeks

after injection. Expressed ChR2 was functional as evidenced by

light-evoked action potentials in RIP-Cre neurons (Figures 6B

and S6A–S6C). mCherry-positive dendrites and soma were

abundantly and exclusively detected in the ARC (Figures 6C

and 6D). mCherry-expressing axons were primarily observed in

the PVH (Figure 6E). By comparison, much less abundant, scat-

tered mCherry-expressing axons were seen in the DMH, the

bed nuclei of the stria terminalis (BST), and the medial preoptic

area (MPO) (Figure 6E; Table S2). Thus, the PVH is the dominant

target of ARC RIP-Cre neurons. This is of interest because it

has been suggested that release of GABA in the PVH stimulates

BAT thermogenesis (Madden and Morrison, 2009).

To determine if GABAergic RIP-Cre neurons are functionally,

synaptically connected to PVH neurons, inhibitory postsynaptic

October 26, 2012 ª2012 Elsevier Inc. 651

Figure 7. Downstream Neurocircuitry

Engaged by ARC RIP-Cre Neurons

(A–D) (A) Diagram illustrating ChR2/retrobead

double-injection experiment in Rip-Cre trans-

genic mice. AAV-ChR2, AAV-Flex-ChR2(H134R)-

mCherry; Green Beads, fluorescent retrograde

green beads. (B) Immunohistochemistry against

mCherry (red) and native fluorescence of retro-

grade green beads (green) in the PVH. (C and D)

Representative tracings of light-driven IPSCs

recorded in (C) a PVH neuron with green beads

(19 of 23) and in (D) an adjacent PVH neuron

without green beads (13 of 14).

(E–H) (E) Diagram illustrating ChR2/retrobead

double-injection experiment in Agrp-ires-Cre

mice. (F) Immunostaining against mCherry (red)

and native fluorescence of retrograde green beads

(green) in the PVH. (G and H) Representative

tracings of light-driven IPSCs recorded in (G)

a PVH neuron with green beads (16 of 16), and in

(H) an adjacent PVH neuron without green beads

(2 of 9).

(I and J) (I) Diagram illustrating retrograde tracing

assay with green beads injected into the RPa of

Vgat-ires-Cre, lox-tdTomato mice. (J) Native fluo-

rescence of green beads (green) and immunore-

activity of tdTomato (red) in the NTS. Arrows

indicate the neurons labeled with both green

beads (i.e., the neurons projecting to the RPa)

and tdTomato (i.e., GABAergic neurons); two such

neurons are zoomed in the dashed squares.

(K–M) (K) Diagram illustrating AAV-Flex-

ChR2(H134R)-mCherry virus injected into the NTS

of Vgat-ires-Cre mice. (L) Representative voltage

tracing showing light-driven spikes in a current-

clamped NTS neuron marked by mCherry fluo-

rescence. (M) Representative tracing of light-

driven IPSCs recorded in RPa neurons (five of

eight).

Blue tick marks represent 0.5 ms light flashes of

0.5 Hz.

See also Figure S7.

currents (IPSCs) were assessed in PVH neurons following

illumination of ChR2-expressing RIP-Cre terminals. Light-driven

IPSCs were reliably evoked in a small subset of randomly

selected PVH neurons that were surrounded by mCherry fluo-

rescent terminals (2 out of 14), and these were completely

blocked by bicuculline, a GABAA receptor antagonist (Fig-

ure 6F). The latency between onset of light and onset of IPSC

in these two neurons was 1.2 and 2.3 ms. The low frequency

of responders (i.e., 2 out of 14) likely relates to complexity

within the PVH, which is composed of numerous subsets of

652 Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc.

functionally distinct neurons (Biag et al.,

2012; Simmons and Swanson, 2009).

As explained below, we have used site

of projection to enrich for PVH neurons

likely to control energy expenditure

and, therefore, likely to receive mono-

synaptic input from GABAergic RIP-Cre

neurons.

Given that the NTS receives abundant projections from

the PVH (Geerling et al., 2010), contains neurons in polysyn-

aptic contact with BAT (Bamshad et al., 1999; Cano

et al., 2003; Oldfield et al., 2002), is known to regulate sym-

pathetic outflow (Spyer, 1994), and inhibit BAT function (Cao

et al., 2010), we hypothesized that PVH neurons projecting

to the NTS drive RIP-Cre neuron-mediated energy expenditure.

To test if NTS-projecting PVH neurons receive monosynaptic

input from GABAergic ARC RIP-Cre neurons, dual-injection

studies were performed as illustrated in Figure 7A, with

AAV-Flex-ChR2(H134R)-mCherry virus injected into the ARC

and retrograde green fluorescent beads injected into the NTS

of Rip-Cre mice. Histologic studies confirmed injection of

beads into the NTS and, unavoidably, the nearby dorsal motor

nucleus of the vagus nerve (DMV) (Figure S7A). Correct target-

ing of the NTS is further confirmed by the presence of retro-

gradely transported beads in sites known to project to the

NTS, including the PVH, lateral hypothalamus, CeA, and dorsal

raphe (Geerling and Loewy, 2006; Saper et al., 1976; Saw-

chenko and Swanson, 1982) (Figure S7B). As expected, PVH

neurons labeled by NTS-injected beads are not neuroendocrine

cells as evidenced by their lack of colabeling following periph-

eral administration of the retrograde tracer, Fluoro-Gold, which

is taken up by median eminence- and posterior pituitary-pro-

jecting neurons (Luther et al., 2002) (Figure S7C).

In agreement with previous results from Biag et al. (2012),

retrogradely transported beads were found in the ventral zone

of the medial parvicellular part of the PVH (PVHmpv) (Figure 7B).

Notably, ChR2-mCherry-expressing ARC RIP-Cre fibers were

found commingling with bead-positive neurons in the PVHmpv

(Figure 7B). Light-evoked IPSCs were then assessed in bead+

or adjacent bead� PVH neurons. Light-evoked IPSCs were

observed in most (19 of 23) bead+ PVH neurons, i.e., neurons

that project to the NTS (Figure 7C). The latency between onset

of light and onset of IPSC was 3.4 ± 1.5 ms (mean ± SD, 19

neurons). In contrast, light failed to evoke IPSCs in nearly all

(13 of 14) bead� PVH neurons (Figure 7D). Thus, GABAergic

ARC RIP-Cre neurons are selectively connected to NTS-projec-

ting PVH neurons.

To test the specificity of NTS-projecting PVH neurons with

respect to afferent input, we next determined if AgRP neurons,

like RIP-Cre neurons, provide monosynaptic input. AgRP neu-

rons are a useful comparator because, like RIP-Cre neurons,

they release GABA, originate in the arcuate, and project to the

PVH, but in striking contrast with RIP-Cre neurons, AgRP

neurons inhibit energy expenditure (Krashes et al., 2011), and

in addition, they also stimulate food intake (Aponte et al.,

2011; Krashes et al., 2011). Connectivity between AgRP neurons

and NTS-projecting PVH neurons was determined using similar

methods to those described above except that Agrp-ires-Cre

mice (Tong et al., 2008) were used to enable ChR2 expression

(Figure 7E). Expressed ChR2 was functional as evidenced by

light-evoked action potentials in AgRP neurons (Figure S7D).

Consistent with previous reports (Cone, 2005), ChR2-mCherry-

expressing AgRP fibers were found to heavily innervate the

PVH and to commingle with bead-positive neurons in the

PVHmpv (Figure 7F). However, unlike the situation with RIP-

Cre neuron ChR2-assisted circuit mapping, light-evoked IPSCs

were absent in all (16 of 16) bead+ PVH neurons, i.e., neurons

that project to the NTS (Figure 7G). In contrast, light-evoked

IPSCs were observed in two out of nine bead� PVH neurons

(Figure 7H), with a latency between onset of light and IPSC in

these two neurons of 3.5 and 5.7 ms. Afferent control of

NTS-projecting PVH neurons by RIP-Cre but not AgRP neurons

is consistent with the opposite and different functions of

these two arcuate neurons and demonstrates the specificity

of the ARC RIP-Cre neuron / NTS-projecting PVH neuron

connection.

The neural pathway by which NTS neurons regulate sym-

pathetic outflow to BAT has yet to be established. To address

this, we tested for connections between the NTS and the

raphe pallidus (RPa), a physiologically important hindbrain site

where numerous BAT sympathetic preautonomic neurons are

located (Morrison and Nakamura, 2011). Retrograde beads

were stereotaxically injected into the RPa of Vgat-ires-Cre,

lox-tdTomato reporter mice (to allow for identification of

GABAergic neurons; Vong et al., 2011) (Figure 7I). Consistent

with previous studies by Hermann et al. (1997) and Yoshida

et al. (2009), RPa-projecting neurons were found to be located

in the MPO, DMH, PAG, DR, and NTS (Figures S7E–S7G). Of

note, 68 of the 143 RPa-projecting NTS neurons were

GABAergic, as evidenced by colocalization of beads with

tdTomato (Figure 7J). In contrast, few or no GABAergic RPa-

projecting neurons were detected in other sites (MPO, DMH,

PAG, and DR) (Figure S7H). Thus, the NTS sends dense pro-

jections to the RPa, and notably, GABAergic input to the RPa

comes predominantly from the NTS. ChR2-assisted circuit

mapping was then used to confirm a GABAergic NTS / RPa

synaptic connection. AAV-Flex-ChR2(H134R)-mCherry virus

was injected into the NTS of Vgat-ires-Cre mice (Figure 7K),

and the activity of ChR2 within NTS GABAergic neurons was

functionally verified (Figure 7L). Light-evoked IPSCs were then

assessed in randomly selected RPa neurons. Of note, light-

evoked IPSCs were detected in five out of eight RPa neurons

(Figure 7M), with a latency between onset of light and IPSC of

3.1 ± 1.6 ms (mean ± SD, five neurons), and these were

completely blocked by bicuculline (data not shown).

DISCUSSION

RIP-Cre neurons regulate energy balance (Chakravarthy et al.,

2007; Choudhury et al., 2005; Covey et al., 2006; Kubota et al.,

2004; Lin et al., 2004; Mori et al., 2009); however, because

RIP-Cre neurons are located inmany sites, the neurocircuit basis

for this has been unknown. To address this, we undertook

a multistep approach. By deleting VGAT and VGLUT2 from

RIP-Cre neurons, we established that release of GABA, but not

glutamate, from RIP-Cre neurons regulates energy balance.

Remarkably, this regulation was specific for energy expenditure;

food intake was entirely unaffected. Building on this, we hypo-

thesized that GABAergic RIP-Cre neurons located specifically

in the ARC mediate this effect. Using a pharmacogenetic

approach to test this, we selectively stimulated arcuate RIP-

Cre neurons and found that this markedly increased energy

expenditure; again, without any effect on food intake. Notably,

pharmacogenetic stimulation of energy expenditure was entirely

dependent upon release of GABA. These studies establish that

synaptic release of GABA from arcuate RIP-Cre neurons selec-

tively drives energy expenditure.

We then used ChR2-assisted circuit mapping to anatomically

and functionally identify downstream neurons receiving syn-

aptic GABAergic input from arcuate RIP-Cre neurons. We

observed that arcuate RIP-Cre neurons project heavily and

predominately to the PVH. This is notable because it has been

proposed that GABAergic input to the PVH stimulates sympa-

thetic outflow and BAT-mediated energy expenditure (Madden

Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc. 653

and Morrison, 2009). Importantly, PVH neurons, specifically

those that project to the NTS in the brainstem, a site known to

regulate energy expenditure (Cao et al., 2010), receive mono-

synaptic GABAergic input from arcuate RIP-Cre neurons. In

summary, these studies have uncovered an ARC-initiated neuro-

circuit that selectively drives energy expenditure.

Selective Regulation of Energy ExpenditureThe ability of arcuate RIP-Cre neurons to regulate energy expen-

diture, without affecting food intake, is noteworthy. It is seen

following genetic deletion of Vgat (Figures 2 and 3) and Lepr

(Covey et al., 2006) in RIP-Cre neurons and, importantly, also

following acute pharmacogenetic stimulation of RIP-Cre

neurons (Figure 5). The selectivity of arcuate RIP-Cre neurons

for energy expenditure is remarkable because other arcuate

neurons, for example AgRP and POMC neurons, coordinately

regulate both food intake and energy expenditure. This unique

feature of arcuate RIP-Cre neurons is important because it

provides a scheme for experimentally approaching forebrain

control of energy expenditure. Specifically, neurons receiving

GABAergic output from arcuate RIP-Cre neurons are likely to

play important roles in regulating energy expenditure, but not

food intake. In the present study, as will be discussed below,

we have used this approach to uncover an efferent circuit that

likely drives energy expenditure (arcuate RIP-Cre GABAergic

neurons / PVH neurons / NTS neurons).

Paraventricular Nucleus and Regulationof Energy ExpenditureLocal application of GABAA receptor antagonist to the PVH

decreases sympathetic outflow and BAT-mediated energy

expenditure, suggesting that GABAergic input to the PVH drives

energy expenditure (Madden and Morrison, 2009). The neurons

providing this GABAergic input, however, have been unknown.

The following five points strongly support the view that

arcuate RIP-Cre neurons are a major source of this BAT-, energy

expenditure-stimulating GABAergic input. First, Rip-Cre,

Vgatflox/flox mice have reduced energy expenditure, altered BAT

morphology, reduced BAT temperature, and reduced Ucp1

gene expression—all suggestive of low BAT activity (Figure 2).

Second, Rip-Cre, Vgatflox/flox mice have reduced diet-induced

thermogenesis (Figure 3). Third, Rip-Cre, Vgatflox/flox mice

have an impaired catabolic and BAT thermogenic response

to leptin treatment (Figure 4). Fourth, GABA release by the

arcuate subpopulation of RIP-Cre neurons stimulates BAT

thermogenesis and energy expenditure (Figure 5). Fifth, arcuate

RIP-Cre neurons project primarily and heavily to the PVH

(Figures 6E and 7B).

We used ChR2-assisted circuit mapping to test the aforemen-

tioned assertion that RIP-Cre neurons provide important

synaptic GABAergic input to PVH neurons. Connectivity,

however, was observed for only a small subset (�15%) of

randomly selected PVH neurons. This low rate of connectivity

is related to complexity of the PVH, which contains numerous

cell types each with unique function (Biag et al., 2012; Simmons

and Swanson, 2009). These include (1) neuroendocrine neurons

that secrete oxytocin, vasopressin, or thyrotropin- or cortico-

tropin-releasing hormone; (2) nonneuroendocrine neurons that

654 Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc.

regulate feeding behavior via presently ill-defined pathways

(Gold et al., 1977; Leibowitz, 1978); and finally, (3) nonneuroen-

docrine neurons that regulate either sympathetic or parasympa-

thetic outflow (Geerling et al., 2010; Swanson and Sawchenko,

1983). Because sympathetic outflow is the primary driver of

brown adipose function and energy expenditure (Bachman

et al., 2002; Bartness et al., 2010), the energy expenditure-

promoting action of RIP-Cre neurons is likely mediated by

aminor subset of nonneuroendocrine neurons, specifically those

that control sympathetic tone.

Role of NTS-Projecting PVH NeuronsFor reasons mentioned below, neurons in the NTS, and there-

fore the subset of PVH neurons projecting to the NTS, are strong

candidates for mediating RIP-Cre neuron-driven energy expen-

diture. NTS neurons (1) receive dense input from the PVH; (2)

are known to control sympathetic outflow (Andresen and Kunze,

1994; Guyenet, 2006); (3) are labeled by the transneuronal

retrograde tract tracer, pseudorabies virus, following its injec-

tion into BAT (Bamshad et al., 1999; Cano et al., 2003; Oldfield

et al., 2002); and (4) have recently been shown to regulate

BAT function (Cao et al., 2010). To test if NTS-projecting PVH

neurons are indeed a major target of GABAergic RIP-Cre

neurons, we injected retrogradely transported fluorescent

beads into the NTS (and, due to its close proximity, the nearby

DMV) and performed ChR2-assisted circuit mapping. Of great

interest, most NTS-projecting neurons receive synaptic

GABAergic input from RIP-Cre neurons as assessed by light-

evoked IPSCs in bead-positive PVH neurons (Figure 7C). In

striking contrast, most other PVN neurons, i.e., those not projec-

ting to the NTS, do not receive GABAergic input from RIP-Cre

neurons (Figure 7D). Thus, NTS-projecting PVH neurons are

the major target of energy expenditure-regulating arcuate

GABAergic RIP-Cre neurons.

AgRP neurons, unlike RIP-Cre neurons, are not synaptically

connected to NTS-projecting PVH neurons (Figure 7G). This

distinction is remarkable because, like RIP-Cre neurons, AgRP

neurons are located in the arcuate, are GABAergic, and send

very dense projections to the PVH. However, in sharp contrast

with RIP-Cre neurons, AgRP neurons have qualitatively different

functions—they inhibit, as opposed to stimulate, energy expen-

diture (Krashes et al., 2011; Tong et al., 2008), and in addition,

they also stimulate food intake (Aponte et al., 2011; Krashes

et al., 2011). These different functions must be explained by

differences in efferent circuitry. Consistent with this, the present

study clearly demonstrates that such differences can robustly

be resolved at the level of the PVH; RIP-Cre neurons are synap-

tically connected with NTS-projecting PVH neurons, whereas

AgRP neurons are not. These results suggest a means for de-

convoluting complex circuitry that traverses the PVH. By using

PVH-projecting neurons of defined function as the entry point,

in this case AgRP versus RIP-Cre neurons, and by performing

ChR2-assisted circuit mapping to identify PVH neurons that

are synaptically downstream, in conjunction with retrograde

tracers that subdivide PVH neurons based upon differential

sites of projection, it is possible to ascribe function and to deter-

mine wiring diagrams of circuits that span three synaptically

coupled sites.

Circuitry Linking NTS Neurons with BATThe RPa is a sympathetic preautonomic area that when acti-

vated potently stimulates BAT activity (Morrison and Naka-

mura, 2011). Indeed, the RPa / sympathetic preganglionic /

postganglionic neuron / BAT circuit likely constitutes the final

common pathway by which the brain controls BAT activity.

With this in mind, our observation that most GABAergic input

to RPa neurons comes from the NTS is of interest and sug-

gests the following pathway: arcuate RIP-Cre GABAergic

neurons / PVH neurons / NTS GABAergic neurons /

RPa neurons / / / BAT activity. The inclusion of a second

GABAergic neuron in this circuit, namely the NTS GABAergic

neuron, allows for stimulation (disinhibition) of BAT activity by

RIP-Cre GABAergic neurons. Future studies will be required

to critically test the functionality, with respect to control of

BAT activity, of the aforementioned pathway. In total, the

circuitry uncovered in this study provides a framework for

understanding homeostatic regulation of BAT activity and

energy expenditure.

EXPERIMENTAL PROCEDURES

Mice

All animal care and experimental procedures were approved by the Beth Israel

Deaconess Medical Center Institutional Animal Care and Use Committee.

Rip-Cre transgenic mice (Postic et al., 1999) were obtained from The Jackson

Laboratory (#003573). lox-Vgat mice were generated previously (Tong et al.,

2008). Chow (Teklad F6 Rodent Diet 8664) or HFDs (Research Diets;

D12331) were used.

Metabolic Studies

Food intake, body weight, fat mass, and locomotor activity was measured as

described by Dhillon et al. (2006). Oxygen consumption was measured with

indirect calorimetry (Columbus Instruments). iBAT temperature was measured

as described by Enriori et al. (2011) with remote biotelemetry (IPTT-300; Bio

Medic Data Systems).

Leptin Treatment Studies

Leptin’s effects on body weight and food intake were measured as reported

by Banno et al. (2010). Leptin’s effects on iBAT temperature were assessed

as reported by Enriori et al. (2011) in animals implanted with biotelemetry

probes (as above). Leptin-induced STAT3 phosphorylation was assessed

in Rip-Cre, lox-GFP mice using methods previously described (Vong et al.,

2011).

Pharmacogenetic Studies

AAV8-Flex-hM3Dq-mCherry virus (Krashes et al., 2011) was stereotaxically

injected into the arcuate of 5- to 6-week-old mice (for oxygen consumption

and iBAT temperature studies) or 3- to 4-week-old Rip-Cre, lox-GFP mice

(for electrophysiological studies). Two to three weeks following viral injection,

electrophysiological responses of mCherry-expressing neurons to 5 mM CNO

or thermogenic responses of animals to 0.3 mg/kg CNO (i.p.) were tested. See

also Extended Experimental Procedures.

ChR2-Assisted Circuitry Mapping

AAV8-Flex-ChR2(H134R)-mCherry virus was stereotaxically injected into sites

of interest of 3- to 4-week-old mice (Atasoy et al., 2008; Zhang et al., 2007).

Two to three weeks after viral injection, light-stimulated firing responses or

IPSCs were tested by whole-cell recordings. For retrograde bead-related

studies, red or green beads (Lumafluor) were stereotaxically injected 5 days

prior to electrophysiological studies. See also Extended Experimental

Procedures.

Statistics

Statistics were performed with GraphPad Prism software.

SUPPLEMENTAL INFORMATION

Supplemental Information includes Extended Experimental Procedures, seven

figures, and two tables and can be foundwith this article online at http://dx.doi.

org/10.1016/j.cell.2012.09.020

ACKNOWLEDGMENTS

The authors acknowledge J. Lu, J.B. Ding, and members of the B.B.L. lab for

helpful discussion; Z. Yang, J. Yu, B. Choi, X. Hu, S. Ma, and D. Cusher for

technical support; C.B. Saper for insightful suggestions and reading

the manuscript; K. Deisseroth for AAV-Flex-ChR2(H134R)-mCherry; and

B.L. Roth for AAV-FLEX-hM3Dq-mCherry. This work was supported by

Grants R01 DK089044, R01 DK071051, R01 DK075632, R37 DK053477,

and BNORC & BADERC Transgenic Core Grants P30DK046200 and

P30DK057521 (to B.B.L.), P&F from BADERC & BNORC Grants

P30DK057521 and P30DK0460200 (to D.K.), R01 DK092605 and AHA

SDG3280017 (to Q.T.), and NS0736313 (to P.M.F.).

Received: March 30, 2012

Revised: June 26, 2012

Accepted: September 6, 2012

Published: October 25, 2012

REFERENCES

Alexander, G.M., Rogan, S.C., Abbas, A.I., Armbruster, B.N., Pei, Y., Allen,

J.A., Nonneman, R.J., Hartmann, J., Moy, S.S., Nicolelis, M.A., et al. (2009).

Remote control of neuronal activity in transgenic mice expressing evolved G

protein-coupled receptors. Neuron 63, 27–39.

Andresen, M.C., and Kunze, D.L. (1994). Nucleus tractus solitarius—gateway

to neural circulatory control. Annu. Rev. Physiol. 56, 93–116.

Aponte, Y., Atasoy, D., and Sternson, S.M. (2011). AGRP neurons are sufficient

to orchestrate feeding behavior rapidly and without training. Nat. Neurosci. 14,

351–355.

Atasoy, D., Aponte, Y., Su, H.H., and Sternson, S.M. (2008). A FLEX switch

targets Channelrhodopsin-2 to multiple cell types for imaging and long-range

circuit mapping. J. Neurosci. 28, 7025–7030.

Bachman, E.S., Dhillon, H., Zhang, C.Y., Cinti, S., Bianco, A.C., Kobilka, B.K.,

and Lowell, B.B. (2002). betaAR signaling required for diet-induced thermo-

genesis and obesity resistance. Science 297, 843–845.

Bamshad, M., Song, C.K., and Bartness, T.J. (1999). CNS origins of the

sympathetic nervous system outflow to brown adipose tissue. Am. J. Physiol.

276, R1569–R1578.

Banno, R., Zimmer, D., De Jonghe, B.C., Atienza, M., Rak, K., Yang, W., and

Bence, K.K. (2010). PTP1B and SHP2 in POMC neurons reciprocally regulate

energy balance in mice. J. Clin. Invest. 120, 720–734.

Bartness, T.J., Vaughan,C.H., andSong,C.K. (2010). Sympathetic andsensory

innervation of brown adipose tissue. Int. J. Obes. (Lond.) 34(Suppl 1), S36–S42.

Bass, J., and Takahashi, J.S. (2010). Circadian integration of metabolism and

energetics. Science 330, 1349–1354.

Biag, J., Huang, Y., Gou, L., Hintiryan, H., Askarinam, A., Hahn, J.D., Toga,

A.W., and Dong, H.W. (2012). Cyto- and chemoarchitecture of the hypotha-

lamic paraventricular nucleus in the C57BL/6J male mouse: a study of immu-

nostaining and multiple fluorescent tract tracing. J. Comp. Neurol. 520, 6–33.

Cannon, B., and Nedergaard, J. (2004). Brown adipose tissue: function and

physiological significance. Physiol. Rev. 84, 277–359.

Cano, G., Passerin, A.M., Schiltz, J.C., Card, J.P., Morrison, S.F., and Sved,

A.F. (2003). Anatomical substrates for the central control of sympathetic

outflow to interscapular adipose tissue during cold exposure. J. Comp. Neurol.

460, 303–326.

Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc. 655

Cao, W.H., Madden, C.J., and Morrison, S.F. (2010). Inhibition of brown

adipose tissue thermogenesis by neurons in the ventrolateral medulla and in

the nucleus tractus solitarius. Am. J. Physiol. Regul. Integr. Comp. Physiol.

299, R277–R290.

Chakravarthy, M.V., Zhu, Y., Lopez, M., Yin, L., Wozniak, D.F., Coleman, T.,

Hu, Z., Wolfgang, M., Vidal-Puig, A., Lane, M.D., and Semenkovich, C.F.

(2007). Brain fatty acid synthase activates PPARalpha to maintain energy

homeostasis. J. Clin. Invest. 117, 2539–2552.

Choudhury, A.I., Heffron, H., Smith, M.A., Al-Qassab, H., Xu, A.W., Selman, C.,

Simmgen, M., Clements, M., Claret, M., Maccoll, G., et al. (2005). The role of

insulin receptor substrate 2 in hypothalamic and beta cell function. J. Clin.

Invest. 115, 940–950.

Cone, R.D. (2005). Anatomy and regulation of the central melanocortin system.

Nat. Neurosci. 8, 571–578.

Covey, S.D., Wideman, R.D., McDonald, C., Unniappan, S., Huynh, F., Asadi,

A., Speck, M., Webber, T., Chua, S.C., and Kieffer, T.J. (2006). The pancreatic

beta cell is a key site for mediating the effects of leptin on glucose homeo-

stasis. Cell Metab. 4, 291–302.

Cowley, M.A., Smart, J.L., Rubinstein, M., Cerdan, M.G., Diano, S., Horvath,

T.L., Cone, R.D., and Low, M.J. (2001). Leptin activates anorexigenic POMC

neurons through a neural network in the arcuate nucleus. Nature 411, 480–484.

Dhillon, H., Zigman, J.M., Ye, C., Lee, C.E., McGovern, R.A., Tang, V., Kenny,

C.D., Christiansen, L.M., White, R.D., Edelstein, E.A., et al. (2006). Leptin

directly activates SF1 neurons in the VMH, and this action by leptin is required

for normal body-weight homeostasis. Neuron 49, 191–203.

Enriori, P.J., Sinnayah, P., Simonds, S.E., Garcia Rudaz, C., and Cowley, M.A.

(2011). Leptin action in the dorsomedial hypothalamus increases sympathetic

tone to brown adipose tissue in spite of systemic leptin resistance. J. Neurosci.

31, 12189–12197.

Fuller, P.M., Jones, T.A., Jones, S.M., and Fuller, C.A. (2002). Neurovestibular

modulation of circadian and homeostatic regulation: vestibulohypothalamic

connection? Proc. Natl. Acad. Sci. USA 99, 15723–15728.

Fuller, P.M., Lu, J., and Saper, C.B. (2008). Differential rescue of light- and

food-entrainable circadian rhythms. Science 320, 1074–1077.

Geerling, J.C., and Loewy, A.D. (2006). Aldosterone-sensitive neurons in the

nucleus of the solitary tract: bidirectional connections with the central nucleus

of the amygdala. J. Comp. Neurol. 497, 646–657.

Geerling, J.C., Shin, J.W., Chimenti, P.C., and Loewy, A.D. (2010). Paraventric-

ular hypothalamic nucleus: axonal projections to the brainstem. J. Comp. Neu-

rol. 518, 1460–1499.

Gold, R.M., Jones, A.P., and Sawchenko, P.E. (1977). Paraventricular area:

critical focus of a longitudinal neurocircuitry mediating food intake. Physiol.

Behav. 18, 1111–1119.

Guyenet, P.G. (2006). The sympathetic control of blood pressure. Nat. Rev.

Neurosci. 7, 335–346.

Haubensak, W., Kunwar, P.S., Cai, H., Ciocchi, S., Wall, N.R., Ponnusamy, R.,

Biag, J., Dong, H.W., Deisseroth, K., Callaway, E.M., et al. (2010). Genetic

dissection of an amygdala microcircuit that gates conditioned fear. Nature

468, 270–276.

Hermann, D.M., Luppi, P.H., Peyron, C., Hinckel, P., and Jouvet, M. (1997).

Afferent projections to the rat nuclei raphe magnus, raphe pallidus and reticu-

laris gigantocellularis pars alpha demonstrated by iontophoretic application of

choleratoxin (subunit b). J. Chem. Neuroanat. 13, 1–21.

Higley, M.J., Gittis, A.H., Oldenburg, I.A., Balthasar, N., Seal, R.P., Edwards,

R.H., Lowell, B.B., Kreitzer, A.C., and Sabatini, B.L. (2011). Cholinergic inter-

neurons mediate fast VGluT3-dependent glutamatergic transmission in the

striatum. PLoS One 6, e19155.

Kong, D., Vong, L., Parton, L.E., Ye, C., Tong, Q., Hu, X., Choi, B., Bruning,

J.C., and Lowell, B.B. (2010). Glucose stimulation of hypothalamic MCH

neurons involves K(ATP) channels, is modulated by UCP2, and regulates

peripheral glucose homeostasis. Cell Metab. 12, 545–552.

Krashes, M.J., Koda, S., Ye, C., Rogan, S.C., Adams, A.C., Cusher, D.S., Mar-

atos-Flier, E., Roth, B.L., and Lowell, B.B. (2011). Rapid, reversible activation

656 Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc.

of AgRP neurons drives feeding behavior in mice. J. Clin. Invest. 121, 1424–

1428.

Kubota, N., Terauchi, Y., Tobe, K., Yano,W., Suzuki, R., Ueki, K., Takamoto, I.,

Satoh, H., Maki, T., Kubota, T., et al. (2004). Insulin receptor substrate 2 plays

a crucial role in beta cells and the hypothalamus. J. Clin. Invest. 114, 917–927.

Lee, J.Y., Ristow, M., Lin, X., White, M.F., Magnuson, M.A., and Hennighau-

sen, L. (2006). RIP-Cre revisited, evidence for impairments of pancreatic

beta-cell function. J. Biol. Chem. 281, 2649–2653.

Leibowitz, S.F. (1978). Paraventricular nucleus: a primary site mediating

adrenergic stimulation of feeding and drinking. Pharmacol. Biochem. Behav.

8, 163–175.

Lin, X., Taguchi, A., Park, S., Kushner, J.A., Li, F., Li, Y., andWhite, M.F. (2004).

Dysregulation of insulin receptor substrate 2 in beta cells and brain causes

obesity and diabetes. J. Clin. Invest. 114, 908–916.

Liu, T., Kong, D., Shah, B.P., Ye, C., Koda, S., Saunders, A., Ding, J.B., Yang,

Z., Sabatini, B.L., and Lowell, B.B. (2012). Fasting activation of AgRP neurons

requires NMDA receptors and involves spinogenesis and increased excitatory

tone. Neuron 73, 511–522.

Luther, J.A., Daftary, S.S., Boudaba, C., Gould, G.C., Halmos, K.C., and

Tasker, J.G. (2002). Neurosecretory and non-neurosecretory parvocellular

neurones of the hypothalamic paraventricular nucleus express distinct

electrophysiological properties. J. Neuroendocrinol. 14, 929–932.

Madden, C.J. (2012). Glucoprivation in the ventrolateral medulla decreases

brown adipose tissue sympathetic nerve activity by decreasing the activity

of neurons in raphe pallidus. Am. J. Physiol. Regul. Integr. Comp. Physiol.

302, R224–R232.

Madden, C.J., and Morrison, S.F. (2009). Neurons in the paraventricular

nucleus of the hypothalamus inhibit sympathetic outflow to brown adipose

tissue. Am. J. Physiol. Regul. Integr. Comp. Physiol. 296, R831–R843.

Merchenthaler, I. (1991). Neurons with access to the general circulation in the

central nervous system of the rat: a retrograde tracing study with fluoro-gold.

Neuroscience 44, 655–662.

Mori, H., Inoki, K., Munzberg, H., Opland, D., Faouzi, M., Villanueva, E.C., Ike-

noue, T., Kwiatkowski, D., MacDougald, O.A., Myers, M.G., Jr., and Guan, K.L.

(2009). Critical role for hypothalamic mTOR activity in energy balance. Cell

Metab. 9, 362–374.

Morrison, S.F., and Nakamura, K. (2011). Central neural pathways for thermo-

regulation. Front. Biosci. 16, 74–104.

Munzberg, H., Flier, J.S., and Bjørbaek, C. (2004). Region-specific leptin resis-

tance within the hypothalamus of diet-induced obese mice. Endocrinology

145, 4880–4889.

Novak, A., Guo, C., Yang, W., Nagy, A., and Lobe, C.G. (2000). Z/EG, a double

reporter mouse line that expresses enhanced green fluorescent protein upon

Cre-mediated excision. Genesis 28, 147–155.

Oldfield, B.J., Giles, M.E., Watson, A., Anderson, C., Colvill, L.M., and

McKinley, M.J. (2002). The neurochemical characterisation of hypothalamic

pathways projecting polysynaptically to brown adipose tissue in the rat.

Neuroscience 110, 515–526.

Petreanu, L., Huber, D., Sobczyk, A., and Svoboda, K. (2007). Channelrhodop-

sin-2-assisted circuit mapping of long-range callosal projections. Nat. Neuro-

sci. 10, 663–668.

Pinto, S., Roseberry, A.G., Liu, H., Diano, S., Shanabrough, M., Cai, X.,

Friedman, J.M., and Horvath, T.L. (2004). Rapid rewiring of arcuate nucleus

feeding circuits by leptin. Science 304, 110–115.

Postic, C., Shiota, M., Niswender, K.D., Jetton, T.L., Chen, Y., Moates, J.M.,

Shelton, K.D., Lindner, J., Cherrington, A.D., and Magnuson, M.A. (1999).

Dual roles for glucokinase in glucose homeostasis as determined by liver

and pancreatic beta cell-specific gene knock-outs using Cre recombinase.

J. Biol. Chem. 274, 305–315.

Saper, C.B., Loewy, A.D., Swanson, L.W., and Cowan, W.M. (1976). Direct

hypothalamo-autonomic connections. Brain Res. 117, 305–312.

Sawchenko, P.E., and Swanson, L.W. (1982). Immunohistochemical iden-

tification of neurons in the paraventricular nucleus of the hypothalamus that

project to the medulla or to the spinal cord in the rat. J. Comp. Neurol. 205,

260–272.

Simmons, D.M., and Swanson, L.W. (2009). Comparison of the spatial distri-

bution of seven types of neuroendocrine neurons in the rat paraventricular

nucleus: toward a global 3D model. J. Comp. Neurol. 516, 423–441.

Song, J., Xu, Y., Hu, X., Choi, B., and Tong, Q. (2010). Brain expression of Cre

recombinase driven by pancreas-specific promoters. Genesis 48, 628–634.

Spyer, K.M. (1994). Annual review prize lecture. Central nervous mechanisms

contributing to cardiovascular control. J. Physiol. 474, 1–19.

Swanson, L.W., and Sawchenko, P.E. (1983). Hypothalamic integration: orga-

nization of the paraventricular and supraoptic nuclei. Annu. Rev. Neurosci. 6,

269–324.

Tong, Q., Ye, C., McCrimmon, R.J., Dhillon, H., Choi, B., Kramer, M.D., Yu, J.,

Yang, Z., Christiansen, L.M., Lee, C.E., et al. (2007). Synaptic glutamate

release by ventromedial hypothalamic neurons is part of the neurocircuitry

that prevents hypoglycemia. Cell Metab. 5, 383–393.

Tong, Q., Ye, C.P., Jones, J.E., Elmquist, J.K., and Lowell, B.B. (2008).

Synaptic release of GABA by AgRP neurons is required for normal regulation

of energy balance. Nat. Neurosci. 11, 998–1000.

van den Pol, A.N. (2003). Weighing the role of hypothalamic feeding neuro-

transmitters. Neuron 40, 1059–1061.

Vong, L., Ye, C., Yang, Z., Choi, B., Chua, S., Jr., and Lowell, B.B. (2011).

Leptin action on GABAergic neurons prevents obesity and reduces inhibitory

tone to POMC neurons. Neuron 71, 142–154.

Wicksteed, B., Brissova, M., Yan, W., Opland, D.M., Plank, J.L., Reinert, R.B.,

Dickson, L.M., Tamarina, N.A., Philipson, L.H., Shostak, A., et al. (2010).

Conditional gene targeting in mouse pancreatic ß-Cells: analysis of ectopic

Cre transgene expression in the brain. Diabetes 59, 3090–3098.

Wu, Q., Boyle, M.P., and Palmiter, R.D. (2009). Loss of GABAergic signaling

by AgRP neurons to the parabrachial nucleus leads to starvation. Cell 137,

1225–1234.

Yang, Y., Atasoy, D., Su, H.H., and Sternson, S.M. (2011). Hunger states

switch a flip-flop memory circuit via a synaptic AMPK-dependent positive

feedback loop. Cell 146, 992–1003.

Yoshida, K., Li, X., Cano, G., Lazarus, M., and Saper, C.B. (2009). Parallel

preoptic pathways for thermoregulation. J. Neurosci. 29, 11954–11964.

Zhang, C.Y., Baffy, G., Perret, P., Krauss, S., Peroni, O., Grujic, D., Hagen, T.,

Vidal-Puig, A.J., Boss, O., Kim, Y.B., et al. (2001). Uncoupling protein-2

negatively regulates insulin secretion and is a major link between obesity,

beta cell dysfunction, and type 2 diabetes. Cell 105, 745–755.

Zhang, F., Wang, L.P., Brauner, M., Liewald, J.F., Kay, K., Watzke, N., Wood,

P.G., Bamberg, E., Nagel, G., Gottschalk, A., and Deisseroth, K. (2007).

Multimodal fast optical interrogation of neural circuitry. Nature 446, 633–639.

Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc. 657

Supplemental Information

EXTENDED EXPERIMENTAL PROCEDURES

AnimalsAll animal care and experimental procedures were approved by the Beth Israel Deaconess Medical Center Institutional Animal Care

and Use Committee. Mice were housed at 22�C–24�C with a 12 hr light/12 hr dark cycle with standard mouse chow (Teklad F6

Rodent Diet 8664; 4.05 kcal/g, 3.3 kcal/gmetabolizable energy, 12.5% kcal from fat; Harlan Teklad, Madison,WI) andwater provided

ad libitum. All diets were provided as pellets. Mice were euthanized by CO2 narcosis.

Generation of Rip-Cre, Vgatflox/flox, and Rip-Cre, Vglut2flox/flox MiceMice with loxp-flanked Vglut2 alleles (Slc17a6) (Vglut2flox/flox) or with loxp-flanked Vgat alleles (Slc32a1) (Vgatflox/flox) have been

described previously (Tong et al., 2007, 2008). Rip-Cre transgenic mice were obtained from the Jackson Laboratory (Strain:

B6.Cg-Tg(Ins2-cre)25Mgn/J; stock number: 003573). Rip-Cre,Vgatflox/+ mice were generated by mating Rip-Cre mice with

Vgatflox/flox mice (129, C57Bl/6 mixed background). Study groups (Vgatflox/flox, Rip-Cre, Vgatflox/+ and Rip-Cre, Vgatflox/flox mice)

were generated by mating Rip-Cre, Vgatflox/+ mice with Vgatflox/flox mice. The same breeding scheme was used to generate study

subjects Rip-Cre, Vglut2flox/flox mice except that Vglut2flox/flox mice were used.

High-Fat Diet StudiesGroups of male controls (Vgatflox/flox mice and Rip-Cre, Vgatflox/+ mice) and Rip-Cre, Vgatflox/flox littermates were switched to high fat,

high sucrose diet (HFD, Research Diets, D12331, 58% kcal from fat, 26% from sucrose, 5.557kcal/gm) at 6 weeks of age and main-

tained on HFD feeding for additional 20 weeks. Body weight was monitored weekly.

Circadian Rhythm AssaysAdult male littermates at 2 months of age were implanted i.p. with biotelemetry transmitters to record body temperature (Tb) and

locomotor activity as previously described (Fuller et al., 2002, 2008). In brief, all animals were anesthetized, a midline celiotomy

was performed, and a sterilized transmitter (TA11TA-F10; Data Sciences International) was inserted into the peritoneal cavity using

aseptic techniques. All incisions were sutured and treated with topical antibiotics. Following surgery, the animals were individually

housed in standard plastic mouse cages with ad libitum food and water. Each cage was placed on top of a telemetry receiver inter-

faced to a microcomputer data acquisition system. Tb values were recorded at 5 min intervals, and activity data were collected in

5 min bins. The cages were housed inside isolation chambers, which provided ventilation, identical LED-based lighting, an ambient

temperature of 24� ± 1�C, and visual isolation. Placement of the mice cages into the chambers was randomwith respect to genotyp-

ing distribution. Tb and activity data were analyzed and plotted using ClockLab (Coulbourn Instruments).

Energy Expenditure and Food IntakeEnergy expenditure was measured as oxygen consumption rate using indirect calorimetry. Individually housed mice maintained on

chow diet at 7–8 weeks of age were placed in chambers of Comprehensive Lab Animal Monitoring System (CLAMS, Columbus

Instruments, Columbus, OH) with ad libitum food andwater. Mice were acclimatized in the chambers for 48 hr prior to data collection.

Diet-induced thermogenesis was assessed as previously described (Dhillon et al., 2006). In brief, oxygen consumption of chow-fed

animals (2-month old male littermates) was measured for 3 days on chow, followed by 3 days on HFD. The averaged oxygen

consumption over the 3 days on chow and that of the individual days on HFD are presented.

Food intake of individually housed mice was measured for 7 days. For food intake assay on HFD-treated animals, food intake was

measured during the first 2 weeks on HFD.

As for acute studies in mice injected with AAV-Flex-hM3Dq-mCherry virus, mice were studied 3 weeks after viral injection. Each

individually-housed male mouse was i.p. injected with saline on the first day and CNO on the following day at 8:00–8:30 am and

immediately placed back into the metabolic chambers for data collection. CNO was administered at 0.30 mg/kg of body weight.

Saline was delivered at the same volume to maintain consistency in the studies. The mice had free access to food and water during

this study.

iBAT and Subcutaneous Flank TemperatureBAT temperature wasmeasured as previously described (Enriori et al., 2011; Madden, 2012) with remote biotelemetry (IPTT-300, Bio

Medic Data Systems). In brief, temperature probes were implanted beneath the iBAT pad between the scapula under anesthesia.

Animals were allowed 1 week recovery and acclimation was performed with saline injection for up to 5 days. For subcutaneous flank

temperature measurements, probes were implanted subcutaneously on the lower back (a site devoid of iBAT). Recovery and

acclimation was as described above.

Beam-Break AssayLocomotor activity was assessed using an OptoM3 apparatus from Columbus Instruments (Columbus, Ohio). Briefly, individually-

housed male mice were placed in their home cages, and ambulatory counts along the x axis were recorded every min for 1 week

using CI BUS software from Columbus Instruments (Columbus, Ohio).

Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc. S1

In Situ Hybridization StudiesA VgatmRNA riboprobe corresponding to the exon 2 was used for in situ hybridization. The primers used to amplify Vgat probe were:

Forward, 50 ttggccccgggaaagagg 30, and Reverse, 50 agcagtatcgcagccccaaag 30. The 35S-labeled cRNA probes were prepared and

the in situ hybridizations were performed as previously described (Kong et al., 2010). The in situ hybridization study of the expression

of Vglut2 was described previously (Tong et al., 2007).

Quantitative PCR AssayQuantitative PCRwas performed as previously reported (Kong et al., 2010). In brief, RNAwas extracted frommouse tissues using the

STAT-60 Reagent (Isotex Diagnostics Inc.). cDNA was generated by reverse transcriptase (Clontech Inc.). The assay probes were

purchased from Applied Biosystems, with IDs for Vgat, Ucp1, and Ucp2 as Mm01208836_g1, Mm01244861_m1, and

Mm01274108_g1, respectively. Relative expression of mRNA was determined using standard curves based on the same cDNA

samples, and samples were adjusted for total RNA content by 18S ribosomal RNA quantitative PCR (Applied Biosystems, Foster

City, CA). Quantitative PCR was performed on an Mx3000p instrument (Stratagene).

ImmunohistochemistryImmunohistochemistry was performed as described (Kong et al., 2010). In brief, brain sections were pretreated with 0.3% hydro-

gen peroxide in PBS (pH 7.4) for 30 min at room temperature and then were incubated in 3% normal donkey serum (Jackson

ImmunoResearch Laboratories, Inc., West Grove, PA) with 0.25% Triton X-100 in PBS (PBT-azide) for 2 hr. Slides were then incu-

bated overnight at room temperature in a rabbit anti-GFP primary antiserum (Molecular Probes, Cat # a-6455; 1:20,000 in PBT-azide).

After washing in PBS, sections were incubated in biotinylated donkey anti-rabbit IgG (Jackson ImmunoResearch Laboratories;

1:1,000) for 2 hr at room temperature, followed by incubation for 1 hr in a solution of avidin-biotin complex (Vectastain Elite ABC

Kit, Vector Laboratories, Burlingame, CA; 1:500) diluted in PBS. The sections were then washed in PBS and incubated in a solution

of 0.04% diaminobenzidine tetrahydrochloride (DAB; Sigma) and 0.01%hydrogen peroxide in PBS. For fluorescent immunolabeling,

sections treated with anti-GFP antiserum were incubated instead in fluorescein-conjugated donkey anti-rabbit IgG (Jackson

ImmunoResearch Laboratory 1:1000) for 2 hr at room temperature. Sections were mounted onto SuperFrost slides and then visual-

ized with Zeiss Confocal Microscope. For anti-mCherry and anti-tdTomato stainings, primary antibody (anti-DsRed, Invitrogen,

1:2500) and secondary antibody (Alex594-anti-rabbit IgG, Invitrogen, 1:200) were used instead.

For leptin-induced pSTAT3 staining, 7- to 8-week-old Rip-Cre, lox-GFPmalemice were fasted overnight, injected intraperitoneally

with 4 mg/kg recombinant mouse leptin (A.F. Parlow, National Hormone and Peptide Program). 45 min later, the animals were

perfused and the brains were fixed for sectioning at 25 mm. pSTAT3 IHC was performed as described (Munzberg et al., 2004;

Vong et al., 2011) with anti-pSTAT3 antibody. Results were imaged by confocal microscopy (Zeiss LSM 510). Cell counts were per-

formed on one of the five brain series. Sections were first organized systematically in a rostral-to-caudal manner. All brain sections in

the series that contained positive cells in each nucleus were analyzed. Cell counts were obtained from both sides of the brain in each

section, and all GFP- and pSTAT3-positive cells in each hypothalamic region were counted irrespective of staining intensity.

c-fos AssayAnimals were handled for 12 consecutive days before the assay to reduce stress response, and then injected with either saline of

CNO (0.3 mg/kg; i.p.). 150 min later, the animals were sacrificed with 7% chloral hydrate diluted in saline (350 mg/kg) for histological

assay.

Leptin Effects on Body Weight, Food Intake, and iBAT TemperatureA protocol similar to that previously described was used for this study (Banno et al., 2010). Briefly, male Vgatflox/flox and Rip-Cre,

Vgatflox/flox littermates (6–7 weeks old) were individually housed and acclimated by handling for a week. These mice were then i.p.

administered with saline (0.9% NaCl) or leptin (A.F. Parlow, NHPP, 2 mg/kg/injection, 8:00 am and 8:00 pm) every 12 hr for

3 days. Body weight and food intake was measured at baseline, and on each day with saline injections or leptin injections. Body

weight and food intake were averaged during the 3 days prior to injections to obtain the baseline values used for calculating percent

changes. Leptin’s effects on iBAT temperature was measured as previously described (Enriori et al., 2011) and 6 mg/kg leptin or

saline was i.p. injected to the animals implanted with remote biotelemetry (IPTT-300, BMDS).

Stereotaxic InjectionsStereotaxic injections were performed as described (Krashes et al., 2011). Mice were anesthetized with 7% chloral hydrate diluted in

saline (350mg/kg) and placed into a stereotaxic apparatus (KOPFModel 963). After exposing the skull via small incision, a small hole

was drilled for injection. A pulled-glass pipette with 20–40 mm tip diameter was inserted into the brain and virus was injected by an air

pressure system. A micromanipulator (Grass Technologies, Model S48 Stimulator) was used to control injection speed at 25 nl/min

and the pipette was withdrawn 5 min after injection. AAV-Flex-hM3Dq-mCherry, serotype 8 (titer 1.23 1013 genomes copies per ml)

or AAV-Flex-ChR2(H134R)-mCherry, serotype 8 (titer 6 3 1012 genomes copies per ml) was injected bilaterally into the arcuate

nucleus of 5–6 weeks (for DREADD-related energy expenditure and iBAT temperature studies) or 3–4 weeks (for ChR2/DREADD-

related electrophysiological studies) old mice (100 nl, coordinates, bregma: AP:�1.40 mm, DV:�5.80 mm, L: ±0.25 mm). For

S2 Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc.

postoperative care, mice were injected intraperitoneally with meloxicam (0.5 mg/kg). For the PVH-related injections, we used coor-

dinate as bregma: AP: �0.8 mm, DV:�5.0 mm, L: ±0.25 mm. All stereotaxic injection sites were verified under electrophysiological

microscopy (for electrophysiology-related studies) or by immunohistochemistry (for anatomy and in vivo studies).

For retrograde beads-related tracing studies, red or green beads (Lumafluor Inc.) were stereotaxically injected to the PVH (30 nl),

the NTS (30 nl), and the RPa (25 nl) 5 days prior to assays. Fluoro-Gold (i.p. 15 mg/kg) was injected to label neuroendocrine cells as

previously described (Luther et al., 2002; Merchenthaler, 1991).

Electrophysiology—Effects of DREADD Activation and LeptinThe protocols of slice preparation and whole cell recording have previously been described (Dhillon et al., 2006). In brief, 5–7 weeks

old mice were anesthetized with isoflurane before decapitation and removal of the entire brain. Brains were immediately submerged

in ice-cold, carbogen-saturated (95% O2, 5% CO2) high sucrose solution (238 mM sucrose, 26 mM NaHCO3, 2.5 mM KCl, 1.0 mM

NaH2PO4, 5.0 mM MgCl2, 10.0 mM CaCl2, 11 mM glucose). Then, 200 mM thick coronal sections were cut with a Leica VT1000S

Vibratome and incubated in oxygenated recording aCSF (126 mM NaCl, 21.4 mM NaHCO3, 2.5 mM KCl, 1.2 mM NaH2PO4,

1.2 mM MgCl2, 2.4 mM CaCl2, 10 mM glucose) at room temperature (21–24�C) for at least 1 hr before recording. The hypothalamic

slice was put in a recording chamber and bathed in oxygenated aCSF at a flow rate of approximately 2 ml/min.

Whole-cell recordings were then made on identified neurons under IR-DIC optics by using a MultiClamp 700B Amplifier (Axon

Instrument) with the software pClamp 10.2 (Axon Instrument). Recording electrodes had resistances of 3.5–5 MU when filled with

the K-gluconate internal solution, which contained (in mM) K-gluconate 128, HEPES 10, EGTA 1, KCl 10, MgCl2 1, CaCl2 0.3, Mg-

ATP 3, and Na-GTP 0.3 (pH 7.35 with KOH).

When the effects of CNO were assessed on RIP-Cre neurons, 5- to 7-week-old Rip-Cre, lox-GFP mice injected with AAV-Flex-

hM3Dq-mCherry into the arcuate nucleus 2–3 weeks before were used. Clozapine-N-oxide (CNO) was applied to bath solution

through perfusion as previously reported (Krashes et al., 2011). After acquisition of stable whole-cell recordings for 2–5 min,

aCSF solution containing 5 mM CNO was perfused into the brain slice preparation.

When the effects of leptin on electrophysiological responses of ARC RIP-Cre neurons were tested, 5- to 7-week-old Rip-Cre, lox-

GFP mice were used. Effects of leptin were assessed as previously described (Dhillon et al., 2006). Since it is difficult to wash out

leptin, brain slices were changed and the system was flushed between each leptin addition electrophysiology study. For assays

in Figure S5, synaptic blockers (1mMKynurenate and 100 mMPTX) were included in the aCSF to synaptically isolate RIP-Cre neurons

and minimize indirect effects of leptin.

ChR2 and CRACMViral Channelrhodopsin-2 expression and activation were similarly performed as previously described (Atasoy et al., 2008; Hauben-

sak et al., 2010; Higley et al., 2011; Zhang et al., 2007). Two to three weeks after viral injection, mice were euthanized for slice prep-

aration. ChR2-expressing neurons were identified by mCherry fluorescence emission using a custom filter set (excitation, 560AF55;

emission, 645AF75; dichroic, 595DP longpass; Omega), and then visually targeted with infrared differential interference optics. For

whole-cell current-clamp recordings, glass electrodes (3.5–5 MU) were filled with internal solution containing (in mM): 128 K gluco-

nate, 10 KCl, 10 HEPES, 1 EGTA, 1MgCl2, 0.3 CaCl2, 5 Na2ATP, 0.3 NaGTP, adjusted to pH 7.3 with KOH. In some RIP-Cre cells that

were detected with frequent spontaneous action potentials, we applied a current (�5 to �30 pA) to hyperpolarize the cell for obser-

vation of light-stimulated action potentials.

Photostimulation-evoked IPSCs in the PVH and RPa neurons were recorded in the whole cell voltage clampmode, with membrane

potential clamped at �60 mV. The intracellular solution contained the following (in mM): 140 CsCl, 1 BAPTA, 10 HEPES, 5 MgCl2,

5 Mg-ATP, 0.3 Na2GTP, and 10 lidocaine N-ethyl bromide (QX-314), pH 7.35 and 290 mOsm. Photostimulation was performed

with 1 mM kynurenic acid to block excitatory postsynaptic currents. All recordings were made using an Axopatch 200B amplifier,

and data were filtered at 2 kHz and digitized at 10 kHz. To photostimulate Channelrhodopsin2-positive fibers, a laser or LED light

source (473 nm; Opto Engine LLC; Thorlabs) was used. The blue light was focused on to the back aperture of the microscope objec-

tive, producing a wide-field exposure around the recorded cell of 10–20 mW/mm2. The light power at the specimen was measured

using an optical power meter PM100D (ThorLabs). The light output is controlled by a programmable pulse stimulator, Master-8 (AMPI

Co. ISRAEL) and the pClamp 10.2 software (AXON Instruments); 20 mM bicuculline was applied to the bath when light-responsive

neurons were recorded. Evoked IPSCs with short latency (%6 ms) upon light stimulation were considered as light-driven IPSCs.

As discussed by others, such currents are most likely monosynaptic (Petreanu et al., 2007).

Double-injection experiments were performed similarly as described (Haubensak et al., 2010). 30 nl retrograde green beads (Green

Retrobeads, Lumafluro Inc.) was stereotaxically injected into the NTS 5 days prior to the recordings. Beads-positive and adjacent

beads-negative neurons were patched for photostimulation recordings.

SUPPLEMENTAL REFERENCES

Atasoy, D., Aponte, Y., Su, H.H., and Sternson, S.M. (2008). A FLEX switch targets Channelrhodopsin-2 to multiple cell types for imaging and long-range circuit

mapping. J. Neurosci. 28, 7025–7030.

Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc. S3

Banno, R., Zimmer, D., De Jonghe, B.C., Atienza, M., Rak, K., Yang, W., and Bence, K.K. (2010). PTP1B and SHP2 in POMC neurons reciprocally regulate energy

balance in mice. J. Clin. Invest. 120, 720–734.

Dhillon, H., Zigman, J.M., Ye, C., Lee, C.E., McGovern, R.A., Tang, V., Kenny, C.D., Christiansen, L.M., White, R.D., Edelstein, E.A., et al. (2006). Leptin directly

activates SF1 neurons in the VMH, and this action by leptin is required for normal body-weight homeostasis. Neuron 49, 191–203.

Enriori, P.J., Sinnayah, P., Simonds, S.E., Garcia Rudaz, C., and Cowley, M.A. (2011). Leptin action in the dorsomedial hypothalamus increases sympathetic tone

to brown adipose tissue in spite of systemic leptin resistance. J. Neurosci. 31, 12189–12197.

Fuller, P.M., Jones, T.A., Jones, S.M., and Fuller, C.A. (2002). Neurovestibular modulation of circadian and homeostatic regulation: vestibulohypothalamic

connection? Proc. Natl. Acad. Sci. USA 99, 15723–15728.

Fuller, P.M., Lu, J., and Saper, C.B. (2008). Differential rescue of light- and food-entrainable circadian rhythms. Science 320, 1074–1077.

Haubensak, W., Kunwar, P.S., Cai, H., Ciocchi, S., Wall, N.R., Ponnusamy, R., Biag, J., Dong, H.W., Deisseroth, K., Callaway, E.M., et al. (2010). Genetic dissec-

tion of an amygdala microcircuit that gates conditioned fear. Nature 468, 270–276.

Higley, M.J., Gittis, A.H., Oldenburg, I.A., Balthasar, N., Seal, R.P., Edwards, R.H., Lowell, B.B., Kreitzer, A.C., and Sabatini, B.L. (2011). Cholinergic interneurons

mediate fast VGluT3-dependent glutamatergic transmission in the striatum. PLoS One 6, e19155.

Kong, D., Vong, L., Parton, L.E., Ye, C., Tong, Q., Hu, X., Choi, B., Bruning, J.C., and Lowell, B.B. (2010). Glucose stimulation of hypothalamic MCH neurons

involves K(ATP) channels, is modulated by UCP2, and regulates peripheral glucose homeostasis. Cell Metab. 12, 545–552.

Krashes, M.J., Koda, S., Ye, C., Rogan, S.C., Adams, A.C., Cusher, D.S., Maratos-Flier, E., Roth, B.L., and Lowell, B.B. (2011). Rapid, reversible activation of

AgRP neurons drives feeding behavior in mice. J. Clin. Invest. 121, 1424–1428.

Luther, J.A., Daftary, S.S., Boudaba, C., Gould, G.C., Halmos, K.C., and Tasker, J.G. (2002). Neurosecretory and non-neurosecretory parvocellular neurones of

the hypothalamic paraventricular nucleus express distinct electrophysiological properties. J. Neuroendocrinol. 14, 929–932.

Madden, C.J. (2012). Glucoprivation in the ventrolateral medulla decreases brown adipose tissue sympathetic nerve activity by decreasing the activity of neurons

in raphe pallidus. Am. J. Physiol. Regul. Integr. Comp. Physiol. 302, R224–R232.

Merchenthaler, I. (1991). Neuronswith access to the general circulation in the central nervous system of the rat: a retrograde tracing studywith fluoro-gold. Neuro-

science 44, 655–662.

Munzberg, H., Flier, J.S., and Bjørbaek, C. (2004). Region-specific leptin resistance within the hypothalamus of diet-induced obese mice. Endocrinology 145,

4880–4889.

Petreanu, L., Huber, D., Sobczyk, A., and Svoboda, K. (2007). Channelrhodopsin-2-assisted circuit mapping of long-range callosal projections. Nat. Neurosci. 10,

663–668.

Tong, Q., Ye, C., McCrimmon, R.J., Dhillon, H., Choi, B., Kramer, M.D., Yu, J., Yang, Z., Christiansen, L.M., Lee, C.E., et al. (2007). Synaptic glutamate release by

ventromedial hypothalamic neurons is part of the neurocircuitry that prevents hypoglycemia. Cell Metab. 5, 383–393.

Tong, Q., Ye, C.P., Jones, J.E., Elmquist, J.K., and Lowell, B.B. (2008). Synaptic release of GABA by AgRP neurons is required for normal regulation of energy

balance. Nat. Neurosci. 11, 998–1000.

Vong, L., Ye, C., Yang, Z., Choi, B., Chua, S., Jr., and Lowell, B.B. (2011). Leptin action on GABAergic neurons prevents obesity and reduces inhibitory tone to

POMC neurons. Neuron 71, 142–154.

Zhang, F., Wang, L.P., Brauner, M., Liewald, J.F., Kay, K., Watzke, N., Wood, P.G., Bamberg, E., Nagel, G., Gottschalk, A., and Deisseroth, K. (2007). Multimodal

fast optical interrogation of neural circuitry. Nature 446, 633–639.

S4 Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc.

Figure S1. Further Assessment of Energy Balance in Mice Lacking VGAT in RIP-Cre Neurons, Related to Figure 2

(A) Subcutaneous flank temperature measured with subcutaneously implanted temperature probes (mean ± SEM, n = 8).

(B) Immunohistochemistry for GFP (DAB, brown stain) in the hypothalamus (Bregma �0.48 mm) of Rip-Cre, lox-GFP transgenic mice.

(C) In situ hybridization for VgatmRNA (35S-labbeld cRNA probe, silver grains) in control (Vgatflox/flox) andRip-Cre, Vgatflox/flox littermates. Arrows indicate the SCN

with notable reduction of Vgat mRNA signal in Rip-Cre, Vgatflox/flox mice.

(D and E) Circadian rhythms of body temperature (Tb) over 30 days recorded in 2-month old control (Vgatflox/flox, n = 3) and Rip-Cre, Vgatflox/flox (n = 5) male

littermates using i.p. biotelemetry implants. The animals were housed under 12:12 light-dark cycles (LD) for 15 days, followed by exposure to constant darkness

(DD); Two representative actograms from each genotype are shown. Under LD conditions, the diurnal rhythm of Tb in Rip-Cre, Vgatflox/flox mice exhibited stable

entrainment with a phase angle that was comparable to that of the wild-types. Similarly, under DD conditions, Rip-Cre, Vgatflox/flox mice showed a persisting

consolidated ‘‘free-running’’ circadian Tb rhythm with a period close to or slightly shorter than 24 hr, indicating that deletion of GABA release from SCN RIP-Cre

neurons does not disturb circadian rhythms.

(F) Serum T4 level and (G) serum corticosterone level (mean ± SEM, n = 8).

Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc. S5

Figure S2. Energy Balance in Mice Lacking VGLUT2 in RIP-Cre Neurons, Related to Figure 2

(A and B) In situ hybridization for Vglut2 mRNA (35S-labbeld cRNA probe, silver grains) in the brain of (A) control (Vglut2flox/flox) and (B) Rip-Cre, Vglut2flox/flox

littermates.

(C) Body weight, (D) oxygen consumption, and (E) daily food intake of ad libitum chow-fed male mice at 3 months of age (mean ± SEM; n = 8–10).

S6 Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc.

Figure S3. Leptin-Induced Energy Expenditure and Leptin-Responsive RIP-Cre Neurons in the Hypothalamus, Related to Figure 4

(A and B) Subcutaneous flank temperature of control (Vgatflox/flox) and Rip-Cre, Vgatflox/flox mice, upon (A) saline or (B) leptin (6 mg/Kg, i.p.) stimulation (mean ±

SEM, n = 8–10).

(C–N) Double immunohistochemistry for GFP (green), leptin-induced phosphorylated STAT3 (Tyr105, pSTAT3) (purple) in the ARC (C–E), VMH (F–H), DMH (I–K),

and SCN (L–N) ofRip-Cre, lox-GFPmice. Neuronswith co-expressedGFP and pSTAT3 (indicated by arrows) weremainly observed in the ARC and VMH,with few

in the DMH and none in the SCN. See also Figure 4I for quantification.

Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc. S7

Figure S4. Pharmacogenetic Activation of Arcuate RIP-Cre Neurons, Related to Figure 5

(A) A representative voltage tracing recorded in an ARCRIP-Cre neuron that was negative for mCherry fluorescence. The brain slice was prepared from aRip-Cre,

lox-GFP mouse that was injected with AAV-Flex-hM3Dq-mCherry virus into the ARC. 5 mM CNO was applied to the bath as indicated.

(B) Food intake over the first 4 hr following saline or CNO injection in AAV-Flex-hM3Dq-mCherry virus-injected Vgatflox/flox,(left), Rip-Cre (middle), and Rip-Cre,

Vgatflox/flox (right) mice (mean ± SEM, n = 8).

(C) Subcutaneous flank temperature following saline or CNO injection in AAV-Flex-hM3Dq-mCherry virus-injected Rip-Cre mice (mean ± SEM, n = 6).

S8 Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc.

Figure S5. Leptin Action on ARC RIP-Cre Neurons, Related to Figure 4

(A) Diagram illustrating experimental design in Rip-Cre, lox-GFP mice.

(B) Fluorescent image indicating the injection site of retrograde red beads.

(C) Immunohistochemistry against GFP (green) and native fluorescence of retrograde red beads (red) in the ARC of beads-injected Rip-Cre, lox-GFP animals.

Arrows indicate RIP-Cre neurons that were labeled with both GFP and red beads (i.e., PVH-projecting RIP-Cre neurons). An adjacent RIP-Cre neuron that was

labeled with GFP only but not with red beads is also shown (i.e., a RIP-Cre neuron that does not project to the PVH). There were also many ARC neurons labeled

with red beads only but not with GFP (i.e., PVH-projecting, non-RIP-Cre neurons) (data not shown).

(D and E) Whole-cell current-clamp recordings were performed in ARC neurons labeled with (D) both GFP and red beads or with (E) GFP only. Synaptic blockers

(1 mM Kynurenate and 100 mM PTX) were included in aCSF to minimize leptin’s indirect effects. (D) Representative tracing of PVH-projecting ARC RIP-Cre

neurons: (D i) that were activated by leptin (membrane potential: from�60.7 ± 2.1 mV to�54.3 ± 1.7mV (n = 9, p < 0.001), firing rate: from 0.12 ± 0.09 Hz to 0.54 ±

0.23Hz (n = 9, p < 0.05); (D ii) that did not respond to leptin (membrane potential: from �59.4 ± 2.8 mV to �59.9 ± 2.8 mV [n = 6], firing rate: from 0.40 ± 0.28 Hz

to 0.42 ± 0.30 Hz [n = 6]). (E) Representative tracing of non-PVH-projecting ARC RIP-Cre neurons: (E i) that did not respond to leptin (membrane potential:

from �59.5 ± 2.0 mV to �60.1 ± 1.8 mV (n = 9), firing rate: from 0.9 ± 0.3 Hz to 1.0 ± 0.4 Hz (n = 9); (E ii) that were inhibited by leptin (membrane potential:

from �60.2 ± 1.1 mV to �66.7 ± 0.9 mV (n = 3, p < 0.001), firing rate: from of 0.72 ± 0.58Hz to 0.06 ± 0.03 Hz [n = 3]). (mean ± SEM, unpaired t test).

Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc. S9

Figure S6. Optogenetic Stimulation of ARC RIP-Cre Neurons, Related to Figure 6

(A–C) Voltage traces showing light-driven spikes in a current-clamped ARC RIP-Cre neuron with mCherry expression in acute brain slice. AAV-Flex-

ChR2(H134R)-mCherry virus was stereotaxically injected into the ARCofRip-Cre transgenicmice at 3–4weeks of age and patch-clamp analyseswere performed

3 weeks later. Light pulses (473 nm) were applied at the frequencies indicated above the traces. Light-evoked spikes at 4Hz (B) and at 10Hz (C) are shown in

a time-expanded view. These results demonstrate that ARC RIP-Cre neurons spike faithfully following ChR2-based photostimulation at a wide range of

frequencies. 0.5 Hz photostimulation was employed in Figures 6 and 7.

S10 Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc.

Figure S7. Retrograde Tracing Studies in the NTS and RPa, Related to Figure 7

(A) Diagram and fluorescent image indicating the injection site of retrograde green beads into the NTS/DMV and (B) retrogradely labeled beads-positive neurons.

(C) PVH neurons labeled with i.p. injected Fluoro-Gold and with NTS-injected retrograde green beads. No colocalization of Fluoro-Gold and green beads was

observed, indicating that the PVH neurons that project to the NTS are not neuroendocrine cells.

(D) Representative voltage tracing showing light-driven spikes in a mCherry+ AgRP neuron from AAV-Flex-ChR2(H134R)-mCherry virus-injected Agrp-ires-Cre

mice.

(E) Diagram and fluorescent image indicating the injection site of retrograde green beads into the RPa and (F and G) retrogradely labeled bead-positive neurons.

(H) Immunoreactivity of tdTomato (red) and native fluorescence of green beads (green) of Vgat-ires-Cre, lox-tdTomato mice with green beads injected into the

RPa. In summary, 1/51 bead-positive neurons in the MPO, 0/74 in the DMH, 3/46 in the PAG, and 0/44 in the DR, are tdTomato immunoreactive.

Cell 151, 645–657, October 26, 2012 ª2012 Elsevier Inc. S11