2011782117 - pnas.2011782117.sapp.pdf

Supplementary Information

for

Experience

dependent plasticity in an innate social behavior is mediated by

hypothalamic LTP

Stefanos Stagkourakis

, Giada Spigolon

, Grace Liu

, David

J. Anderson

1,3

Corresponding author information:

Stefanos Stagkourakis, David J. Anderson

Email:

stefanos.stagkourakis@caltech.edu

wuwei@caltech.edu

This PDF file includes:

Supplementary Information text

Figures S1 to S5 with legends

able

SI References

www.pnas.org/cgi/doi/10.1073/pnas.2011782117

Supplementary Information Text

Extended

iscussion

The plasticity observed at AHiPM



VMHvl

Esr1

synapses likely has both post

and pre

synaptic

components, as suggested by an increase in the AMPAR/NMDAR ratio

following aggression

training (

Fig. 2J, K), and by the differential responses of VMHvl

Esr1

neurons to trains of pre

synaptic stimuli

(4, 5)

(Fig. 2L, M), respectively.

Surprisingly,

the

form

hypothalamic

LTP

studied

here

does

not

exhibit

“occlusion,”

phenomenon

observed

studies

hippocampal

amygdalar

LTP

(6,

which

following

vivo

behavioral

induction

LTP

the

synaptic

population

interest,

the

magnitude

LTP

that

can

induced

subsequently

vivo

markedly

decreased

(Fig.

4D,

K).

Similarly,

not

observe

the

phenomenon

which

prior

vivo

LTP

can

enhance

the

extent

LTD

that

can

induced

vivo

slices

from

such

animals.

The

reason(s)

for

the

failure

observe

these

phenom

ena

are

not

clear,

and

will

require

further

studies

elucidate.

There

are

number

effects,

however,

which

could

account

for

these

observations.

Firstly,

it is possible that the proportion of synapses modified by the

in vivo

social

experience was smal

l compared to the synapses being sampled in the slice. Another possibility is that

the synapses being assayed in the slice are a different population than the ones modified

in vivo

, or lastly

that new synapses were formed by the

in vivo

experience and they

are the ones primarily contributing to

the LTP and LTD being measured

in vitro

. This last possibility is of particular interest, given that

presented in Fig. 3, an increase in spine density occurs in VMHvl

Esr1

neurons of AGG mice.

The data on LTP pre

sented here, blur the distinction between neural circuits mediating learned

vs. innate behaviors, and reinforce the concept of “learned innate behavior,” in which plasticity within

developmentally hardwired circuits can function to modify the strength of a

n instinctive behavior in

response to social experience

(8, 9)

. An example of the latter in an invertebrate is the post

mating

response in

Drosophila,

a form of memory in which female sexual receptivity is inhibited following mating

(10

12)

This idea notwithstanding, more complex forms of learning, such as classical or operant

conditioning, may utilize circuits that are parallel to those that mediate innate forms of the modified

behavior, as shown in the case of condi

tioned

innate

fear

(13

15)

. In this context, it is worth noting

that mice can learn an instrumental, operant response using successful aggressive encoun

ters as a

reinforcer

(16)

, and that performance

of this instrumental task is facilitated by optogenetic activation of

VMHvl neurons

(17)

. The neural substrates and synaptic mechanisms underlying this operant

conditioning remain to be elucidated, although the nucleus accumbens

based reward system has been

implicated in recent studies

(18)

Aggressiveness can be enhanced not only by repeated successful agonistic enco

unters, as

shown here, but also by prior mating experience

(19)

. Recently, we showed that as little as 30 minutes

of free social interaction with a female was sufficient to transfor

m a socially naive mouse into an AGG

mouse within 24 hrs of the interaction

(8)

. This effect was associated with a change in the neural

representation of ma

. female conspecifics among VMHvl

Esr1

neurons, from partially overlapping to

largely non

overlapping

(8)

. W

hether this change in neural population coding involves synaptic plasticity

within VMHvl, or is inherited from upstream structures, such as the MeA

(9)

, remains to be determined.

In other studies, we have shown that the effect of social isolation stress to promote aggression in non

sexually experienced males is mediated by the neuropeptide Neurokinin B (NkB) and its receptor Nk3R,

acting in the

dorso

medial hypothalamus (DMH)

(20)

. The relationship of this form of experience

dependent plasticity to VMHvl

Esr1

neuronal activity is currently unknown.

Our current findings also provide insights into individual differences in the ability of genetically

identical animals to re

spond to “aggression training”.

Firstly,

show

here

that

several

physiological

parameters

AGG

mice

are

different

from

those

socially

naïve

mice.

These

include

elevated

baseline

VMHvl

Esr1

neuron

activity

(Fig.

J),

increased

spontaneous

excitatory

input

onto

VMHvl

Esr1

neurons

(Fig.

C),

increased

AMPA/NMDA

ratio

at AHiPM



VMHvl

Esr1

synapses

(Fig.

and

altered

synaptic

integration

properties

(Fig.

M).

contrast,

NON

mice

spontaneous

inhibitory

inputs

VMHvl

Esr1

neurons

are

increased,

relative

socially

naïve

mice

(Fig.

S1).

addition,

NON

mice

exhibit

shorter

lasting

LTP

and

longer

lasting

LTD

than

are

observed

AGG

mice

(Fig.

4D,

G).

Whether

increased

LTD

suf

ficient

account

for

the

failure

NON

mice

respond

aggression

training

not

yet

clear.

Another

possibility,

suggested

the

increased

spontaneous

IPSCs,

that

VMHvl

receives

stronger

inhibitory

input

from

GABAergic

neurons

NON

mice.

While

there

are

very

few

GABAergic

neurons

within

VMHvl

itself

(21)

VMHvl

receives

strong

inhibitory

input

from

the

neighboring

tuberal

(TU)

region.

possible

that

the

lack

aggression

NON

mice

reflects

potentiation

these

GABAergic

neurons.

The

synaptic

mechanisms

responsible

for

the

lack

aggression

NON

mice

will

clearly

require

further

investigation.

We also find th

at NON mice have low

levels of circulating T in comparison to AGG mice, and

that

experimental administration of supplemental T can restore the capacity for “aggression learning” in

such animals. While the role of T in promoting male aggressiveness is wel

established

(22

27)

, our

studies provide new insight into the neurophysiological mechanisms that may mediate this effect in the

context of aggression training. Sp

ecifically, we observe that NON animals can only express LTP

in vivo

following administration of exogenous T. Although LTP can be induced optogenetically

ex vivo

in slices

from control NON animals, LTP in slices from T

implanted NON animals exhibited higher

amplitude and

persistence.

Moreover,

in AGG mice

levels of T increased during aggression training. This correlation

suggests either that T acts directly to en

hance LTP at this synapse, which in turn promotes aggression,

or that T acts indirectly, by promoting aggressive behavior which in turn enhances LTP (Fig. S5). Whether

T directly influences synaptic plasticity, and if so the underlying molecular mechanisms

involved, as well

as the mechanistic basis of individual differences in T levels, are interesting topics for future study.

Our experiments have focused on a specific glutamatergic input to VMHvl

Esr1

neurons which have

a causal role in aggression. In addi

tion to our finding

, recent work

reported

that

VMHvl

projecting

Vglut

neurons

the

AHiPM

are

active

during

both

social

investigation

and

attack,

and

that

functional

manipulation

these

neurons

influences

attack

(28,

29)

VMHvl

Esr1

neurons receive inputs from neurons

in over 30 different structures

(30)

, raising the quest

ion of whether other

glutamatergic

inputs to these

cells also display plasticity. Indeed, recently published work has identified synaptic plasticity promoted

by foot

shock stress in a medial amygdala projection that primarily targets the central part of VM

H (VMHc)

(31)

. Although a causal role in promoting aggression was not directly demonstrated for this input, and the

mechanism of potentiation was not established, plasticity at this synapse may regulate stress

induced

aggression

(31)

. The present study demonstrates that AHiPM



VMHvl

Esr1

synapses can undergo

Hebbian LTP, and that potentiation of these synapses occurs during social experience that enhances

offensive aggression. Together, these data suggest that VM

Hvl likely provides a substrate in which

aggression plasticity can occur at multiple synaptic inputs, each of which may play distinct roles in

physiology and/or behavior. Our results also reveal striking effects of aggression training on dendritic

100

spine mo

rphology among VMHvl

Esr1

neurons, although we cannot be certain whether the secondary

101

dendritic branches where we observe this phenomenon receive synaptic input from AHiPM. Other recent

102

studies have identified structural plasticity among VMHvl

derived ax

ons innervating hypothalamic

103

targets in females, which are mediated by changes in sex steroids during estrus

(32)

. The present work,

104

together with these other studies, begins to provide a view of the acute and dynamic changes that can

105

occur through experience and/or hormonal modulation,

in a brain node that controls innate social

106

behaviors.

107

108

109

Extended M

aterials and

ethods

110

Animals.

All mice were housed in ventilated micro

isolator cages in a temperature

controlled

111

environment (median temperature 23 °C), under a reversed 12h dark

light cycle, and had

ad libitum

112

access to food and water. Mouse cages were changed weekly on a fixed day o

n which experiments were

113

not performed.

114

Brain slice electrophysiology.

Acute mouse brain slices were prepared. Slices were cut on a vibratome

115

(Leica VT1000S) to 300

m thickness and continuously perfused with oxygenated aCSF containing (in

116

millimolar): NaC

l (127), KCl (2.0), NaH

(1.2), NaHCO

(26), MgCl

(1.3), CaCl

(2.4), and D

glucose

117

(10). See also

SI Appendix,

Table S1. Whole

cell current

and voltage

clamp recordings were performed

118

with micropipettes filled with intracellular solution containing (in millimolar), K

gluconate (140), KCl (10),

119

HEPES (10), EGTA (10), and Na

ATP (2) or Cesium methanesulfonate (140)

, KCl (10), HEPES (10),

120

EGTA (10), and Na

ATP (2) (pH 7.3 with KOH). Recordings were performed using a Multiclamp 700B

121

amplifier, a DigiData 1440 digitizer, and pClamp 11 software (Molecular Devices). Slow and fast

122

capacitative components were semi

automat

ically compensated. Access resistance was monitored

123

throughout the experiments, and neurons in which the series resistance exceeded 15 M

or changed

124

≥20% were excluded from the statistics. The liquid junction potential was 9.7 mV and not compensated.

125

The r

ecorded current was sampled at 20 kHz. Baseline recordings of EPSCs, IPSCs and optogenetically

126

evoked synaptic currents were performed in normal aCSF conditions and in the absence of GABA and

127

NMDA

receptor

blockers. Spontaneous excitatory currents

were sam

pled at the reversal of Cl

HOLD

128

70 mV), and spontaneous inhibitory currents were sampled at the reversal of fast excitatory

129

neurotransmission (V

HOLD

=0 mV). All recordings were performed at near

physiological temperature

130

(33±1

C). Reagents used in slice

electrophysiology experiments; Neurobiotin

tracer (Vector

131

laboratories) was used in combination with Streptavidin conjugated to Alexa Fluor 647. MATLAB and

132

OriginPro9 were used for electrophysiological data analysis.

CNQX (10

M), D

AP5 (25

M), TTX (50

133

nM), and 4

AP (100 mM) were bath applied to block excitatory transmission and to test if optogenetically

134

evoked responses are monosynaptic

(28)

. All drugs were pre

applied for 5 min in the slice chamber prior

135

to data acquisition.

136

Brain slice Ca

maging.

The spontaneous activity of mouse VMHvl

Esr1

neurons was monitored by

137

imaging fluorescence changes of the jGCaMP7s biosensor, using a CCD camera (Evolve

512,

138

Photometrics), mounted on an Olympus BX51WI

microscope. Recordings were 5 min in duration.

As a

139

subpopulation of VMHvl

Esr1

neurons

likely

expresses T

type Ca

channels, the Ca

transients reported

140

in Fig. 1

reflect

both action

potentials

and subthreshold synaptic potentials. A 60x water

dipping objective

141

was used to focus on VMHvl. Ca

imaging analysis was performed using the MIN1PIPE one

photon

142

based calcium imaging signal extraction pipeline

(30)

, in combination with custom

written MATLAB

143

routines.

144

Cell filling and reconstruction.

Mouse

Esr1

VMHvl neurons were recorded in whole

cell mode with

145

intracellular pipette solution as above, with the addition of 0.2% neurobiotin. After recording, slices were

146

placed in fixative (4% paraformaldehyde/0.16% picric acid), washed in PBS and incubated at 4°C

for 72h

147

in a solution containing s

treptavidin conjugated to Alexa Fluor 647

. After extensive washing, slices were

148

mounted with 2.5% DABCO in glycerol. VMHvl

Esr1

neuron identity of all filled cells was confirmed with

149

colocalization studies of viral

induced

tdTomato expression.

150

Ex vivo

optogenetics.

Photostimulation during slice whole

cell recordings was performed via a 3.4 watt

151

535 nm LED mounted on the microscope fluorescence light source and delivered through the 60X

152

objective’s lens. Photostimulation w

as controlled via the analog outputs of a DigiData 1440A, enabling

153

control over the duration and intensity. The photostimulation diameter through the objective lens was

154

~310

m with illumination intensity typically scaled to 0.35 mW/mm

155

In vivo

optogeneti

cs.

Subjects were coupled via a ferrule patch cord to a ferrule on the head of the

156

mouse using a zirconia split sleeve (Doric Lenses). Ferrules and fiber

optic patch cords were purchased

157

from Thorlabs and Doric Lenses, respectively. The optical fiber was c

onnected to THORLABS fiber

158

coupled LED (

M530F2, 9.6 mW

) via a fiber

optic rotary joint (

FRJ_1x1_FC

FC,

Doric Lenses) to avoid

159

twisting of the cable caused by the animal’s movement. Prior to a testing session, following the coupling

160

of the patch cords with

the optic fiber ferrules,

Esr1

Cre/

animals were given 10 min to acclimate in their

161

home cage in the absence of an intruder. The frequency and duration of photostimulation were controlled

162

using the programmable train generator Pulse Pal

(31)

. Light power was controlled by dialing an analog

163

knob on the LED driver (T

Cube

LED Driver with Trigger Mode, Thorlabs, LEDD1B). Light power was

164

measured from the tip of the ferrule in the patch cord at different laser output settings, using an o

ptical

165

power energy meter and a photodiode power sensor (Thorlabs, PM100D, and S130VC). L

ight power was

166

dialed at 0.5 mW at the fiber tip.

Upon identification of the fiber placement coordinates in brain tissue

167

slides, irradiance (light intensity) was calcu

lated using the brain tissue light transmission calculator based

168

on (

http://www.stanford.edu/group/dlab/cgi

bin/graph/chart.php

) using laser power measured at the tip

169

and the distan

ce from the tip to the target brain region measured by histology. Animals showing no

170

detectable viral expression in the target region and/or ectopic fiber placement were excluded from the

171

analysis.

172

In vivo

electrophysiology.

In vivo

electrophysiology

recordings were performed in freely moving mice,

173

using chronic silicon probe implants. All extracellular recordings were conducted in the left VMHvl, and

174

all mice included in the present study were validated using the following criteria: identification of

the

175

lipophilic dye (DiD) tract targeting VMHvl, phototagging of VMHvl

Esr1

neurons, and photostimulation

176

evoked low

latency attack against a conspecific through optrode mediated VMHvl

Esr1

neuron

177

photoactivation. Recordings were performed using an optrode ba

sed on the A1x32

Poly2

10mm

50s

178

177 NeuroNexus probe and a 100

m optic fiber placed along the probe’s shank terminating 50

m above

179

the probe’s first recording sites. Photostimulation was delivered using fiber

coupled Thorlabs LEDs

180

(M530F2,

9.6 mW

for LTP

/LTD studies, and M617F2

, 13.2 mW

for phototagging), and light power was

181

dialed at 0.33 mW at the optrode’s fiber tip. The probe was implanted 200 μm above the intended

182

recording site, and using the NeuroNexus OH32LP oDrive was lowered over a period of fou

r days to the

183

target coordinates (lowering by 50

m/day).

Only channels that showed photo

responses in the local field

184

potential were used for LFP analysis.

Recordings were performed using the Open Ephys acquisition

185

board with a

sampling rate of 30 kHz, th

e Open Ephys I/O board, and the Open Ephys GUI

(31)

. The

186

LFP

signal was obtained by applying low pass

filtering with a cut

off at 100 Hz on the raw voltage traces.

187

Note that although the exc

itation wavelength for testing LTP

in vivo

was chosen to preferentially activate

188

AHiPM synaptic input to VMHvl neurons via Chronos, due to spectral overlap between the opsins we

189

cannot exclude a contribution of direct ChrimsonR

mediated Esr1 neuron activat

ion to the fEPSPs

190

Immunohistochemistry.

Mice were anesthetized with ketamine (KetaVed, VEDCO) and xylazine

191

(AnaSed, NDC 59399

110

20), then transcardially

perfused with 20 mL of ice

cold fixative. Whole brains

192

were dissected, immersed in ice

cold fixative for 90 min then stored in 0.1M PBS (pH 7.4) containing

193

20% sucrose, 0.02% bacitracin and 0.01% sodium azide for three days, before freezing with dry ice.

194

Coronal sections were cut at a thickness of 14 μm on a cryostat (Microm, Walldorf) and thaw

mounted

195

onto gelatine

coated glass slides. For GFP staining, brain sections were incubated overnight at 4°C using

196

a chicken anti

GFP antibody (Aves Labs, Inc., GFP

1010) at 1:500 dilution. For tdTomato staining brain

197

sections were incubated overnight at 4°C using a rabbit anti

DsRed antibody (Takara, 632392) at 1:500

198

dilution. Primary antibody incubation was followed by Alexa

488

conjugated goat anti

chicken secondar

199

antisera (1:500; Invitrogen), and/or Alexa

568

conjugated donkey anti

rabbit secondary antisera (1:500;

200

Invitrogen). DAPI solution (1mg/mL) was used at 1:10000 dilution.

For further details on reagents, see

201

also

SI Appendix

Table S1.

202

Confocal microscopy

Brain slices were imaged by confocal microscopy (Zeiss, LSM 800). Brain areas

203

were determined according to their anatomy using Paxinos and Franklin Brain Atlas

(8, 33)

204

For

cell

reconstructions,

each

entire

neurobiotin

filled

neuron

was

acquired

63X

(NA

1.4),

μm

step

205

size

using

Zeiss

LSM880

confocal

microscope.

Imaris

9.3

(Bitplane)

was

used

visualize

the

topology

206

the

dendritic

tree

and

the

centrifugal

branch

ordering

method

was

chosen

sort

dendrites,

assigning

207

order

the

root.

order

dendrites

were

then

selected

for

further

imaging

acquisition

perform

spine

208

quantification.

μm

long

dendritic

segments

were

acquired

63X

(NA

1.43),

0.1

μm

step

size

209

and

0.06x0.06

pixel

size

using

Airy

scan

detector

the

LSM880.

Two

segments

were

acquired

for

210

dendrites

longer

that

200

μm.

211

For

spine

quantification,

images

dendritic

segments

were

rendered

Imaris

using

the

Blend

algorithm

212

and

the

Filament

module

was

used

reconstruct

dendrites

and

spines.

Specifically,

the

auto

path

213

method

was

chosen

and

thinnest

spine

diameter

(between

1.5

and

μm),

maximal

distance

from

the

214

dendrite

(between

and

μm)

and

fluorescence

intensity

threshold

were

defined

every

single

dendrite

215

detect

spines.

The

statistics

module

Imaris

was

used

extract

spine

density

values.

Three

six

216

segments

per

neuron

were

quantified

and

values

were

averaged.

217

Tail

tip whole blood sampling.

Whole blood samples of 40

L were collected from the lateral tail

218

vein of

restrained mice

(32)

. Only blood samples acquired within

2 min post

restraining were used for

219

hormone measurements, and the subjects were then returned to their home cage. Briefly, the rodent’s

220

tail was immersed for 30 sec in 40

C water to dilate the tail blood vessels. Immediately after, a 23G

221

needle was used

to puncture the lateral tail vein, and whole blood was collected. Bleeding was stopped

222

via applying gentle pressure to the tail at the level of the puncture with surgical cleaning tissue, and 2%

223

chlorhexidine antiseptic solution was used for tail disinfect

ion at the end of the procedure. Blood samples

224

were refrigerated at 4

C for 30 min and then centrifuged at 4

C at 2000 RCF. Following centrifugation,

225

serum was collected and was frozen at

C for a maximal period of 2 months prior to performing ELISA

226

mea

surements. All blood samples were acquired during the dark phase of the 12h/12h light/dark cycle.

227

For further details on reagents, see also

SI Appendix

Table S1.

228

Testosterone ELISA.

well plates were used in a ready

use kit for testosterone ELISA (R&

229

systems

–

Catalog number KGE010). Linear regression was used to fit the optical densities for the

230

standard curve

the concentration. The standard curve range for corticosterone was 300 to 100000

231

pg/mL. Concentrations were calculated from the optical de

nsity at 450 nm of each sample. Appropriate

232

sample dilutions were carried out to maintain detection in the linear part of the standard curve and

233

typically involved 1 to 10 for mouse serum samples. Data acquired from the performed ELISAs are

234

presented as ab

solute values. Differences between groups were identified by Student’s

test or ANOVA.

235

Viral vectors.

For

ex vivo

imaging studies of VMHvl neurons,

Esr1Cre/

male mice

were injected

236

in VMHvl with 200 nL of AAV9

Syn

FLEX

jGCaMP7s

WPRE (addgene

104491

AAV9) 5.3 × 10

237

genomic copies per mL. For

ex vivo

optogenetic studies,

Esr1Cre/

male mice

were injected in VMHvl

238

with 200 nL of AAV9

FLEX

tdTomato (addgene 28306

AAV9) 4.2 × 10

genomic copies per mL and in

239

AHiPM with 100 nL of AAV5

Syn

Chronos

GFP (addgene 59170

AAV5) 3.7 × 10

genomic copies per

240

mL. For

in vivo

optogenetic and electrophysiology experiments,

Esr1Cre/

male mice

were injected in

241

VMHvl with 100 nL of AAV5

Syn

FLEX

rc[ChrimsonR

tdTomato] (addgene 62723

AAV5) 4.1 × 10

242

genomic co

pies per mL and in AHiPM with 100 nL of AAV5

Syn

Chronos

GFP (addgene 59170

AAV5)

243

3.7 × 10

genomic copies per mL. Control groups were injected in VMHvl with 100 nL of AAV9

FLEX

244

tdTomato (addgene 28306

AAV9) 4.2 × 10

genomic copies per mL and in AHiPM w

ith 100 nL of AAV5

245

CAG

GFP (37825

AAV5) 5.9 × 10

genomic copies per mL.

For further details on reagents, see also

246

Appendix

Table S1.

247

Stereotactic surgery and viral gene transfer.

Adult heterozygous

Esr1Cre/

males were single

housed

248

for at least five

days before undergoing surgical procedures and were operated on at 16

–

20 weeks of

249

age.

Mice were anesthetized using isoflurane (5% induction, 1

–

2% maintenance, in 95% oxygen) and

250

placed in a stereotaxic frame (David Kopf Instruments). Body temperature was

maintained using a

251

heating pad. An incision was made to expose the skull for stereotaxic alignment using the inferior cerebral

252

vein and the Bregma as vertical references. We based the coordinates for the craniotomy and stereotaxic

253

injection of VMHvl on an

anatomical magnetic resonance atlas of the mouse brain (AP:

−

4.68 mm; ML:

254

±0.78 mm; DV:

−

5.80 mm), as previously described

(34)

. Virus suspension was injected using a pulled

255

glass capillary at a slow rate of 8

–

10 nL/min, 100 nl per injection site (Nanojector II, Drummond Scientific;

256

Micro4 controller, Wor

ld Precision Instruments). The glass capillary was withdrawn 10 min after the

257

cessation of injection.

258

Osmotic mini

pumps.

Testosterone was dissolved at 30 mg/ml in sesame oil and was administered for

259

2 weeks at a rate of 0.75 mg/hour via subcutaneous osmot

ic mini

pumps (Alzet, model 1002)

(35)

For

260

further details on reagents, see also

SI Appendix

Table S1.

261

Social behavior assays.

The aggressio

n phenotype of animals defined as aggressive (AGG), or non

262

aggressive (NON) in the present study was based on the expression of aggressive behavior in the five

263

consecutive day resident

intruder test (5cdRI). Animals that did not express any aggressive beha

vior in

264

the 5cdRI were identified as NONs, while all AGGs expressed aggression in a minimum of three out of

265

the five trials, with the majority expressing attack behavior in all five days. As described in Fig. 1, the

266

5cdRI composed of a 15 min social intera

ction test per day in the resident’s home arena, with socially

267

naïve 4

5 month

old residents. Intruders were BALB/c males 2

3 months old and of lower weight/size.

268

Three follow up tests were performed in the 5cdRI experimental design presented in Fig. 1, sp

ecifically,

269

2 weeks, 4 weeks and 12 weeks following

the completion date of the 5cdRI assay. Note that only 15 out

270

of a total of 106 aggressive mice, were used to quantify the effect of aggression training. This was based

271

on the finding that following behav

ioral analysis of the first 15 mice used in the study, the power of the

272

ANOVA test reached

< 0.0001. This suggested that including additional observations would not aid the

273

power of the statistical test. In Fig. 5, following the 5cdRI, on day six a social interaction test was

274

performed

in a novel home

cage

sized arena. In addition to the C57 mal

e, a male with a larger

275

bodyweight/size CD

1 conspecific was introduced. The duration of this experiment was 15 min, following

276

which both animals were returned to their home cage.

277

Statistics.

No statistical methods were used to predetermine sample sizes bu

t our sample sizes are

278

similar to those reported in previous publications

(36)

. Data met the assumptions of the stati

stical tests

279

used and were tested for normality and variance.

Normality was determined by D’Agostino

–

Pearson

280

normality test. All

tests and one

way ANOVAs were performed using GraphPad Prism software

281

(Graphpad Software Inc.). Statistical significance was

set at

0.05.

282

283

284

285

286

287

Fig. S1. Presynaptic plasticity of inhibitory input in VMHvl

Esr1

neurons of non

aggressive male

288

mice.

289

(A) Representative recordings of spontaneous inhibitory post

synaptic currents (sIPSCs) from VMHvl

Esr1

290

neurons, from socially naive, aggressive (AGG) and non

aggressive (NON) mice.

291

(B) Left

–

cumulative frequency distribution plot of sIPSC inter

event interval (IEI) in voltage

clamp

292

recordings collected from VMHvl

Esr1

neurons from socially naive, AGG and N

ON mice (n=11

14 VMHvl

Esr1

293

neuron recording per group, collected from 8

10 mice per group, Kolmogorov

Smirnov test,

< 0.0001

294

between socially naive and AGG mice,

< 0.0001 between socially naive and NON mice). Right

–

295

comparison of sIPSC frequency in vo

ltage

clamp recordings collected from VMHvl

Esr1

neurons from

296

socially naive, AGG and NON mice (n=11

14 VMHvl

Esr1

neuron recording per group, collected from 8

297

mice per group, Kruskal

Wallis one

way ANOVA with uncorrected Dunn’s post hoc test,

= 0.0425

298

between socially naive and AGG mice,

= 0.0480 between socially naive and NON mice).

299

–

cumulative frequency distribution plot of sIPSC amplitude in voltage

clamp recordings

300

collected from VMHvl

Esr1

neurons from socially naive, AGG and NON mice (

n=11

14 VMHvl

Esr1

neuron

301

recording per group, collected from 8

10 mice per group, Kolmogorov

Smirnov test,

P =

0.2780 between

302

socially naive and AGG mice,

< 0.0001 between socially naive and NON mice). Right

–

comparison of

303

sIPSC amplitude in voltage

cla

mp recordings collected from VMHvl

Esr1

neurons from socially naive, AGG

304

and NON mice (n=11

14 VMHvl

Esr1

neuron recording per group, collected from 8

10 mice per group,

305

Kruskal

Wallis one

way ANOVA with uncorrected Dunn’s post hoc test,

= 0.8995 between socially

306

naive and AGG mice,

= 0.0476 between socially naive and NON mice).

307

ns; not significant, *

< 0.05, ****

< 0.0001.

In box plots the median is represented by the center line,

308

the interquartile range is represented by the box edge

s, the bottom whisker extends to minimal value,

309

and the top whisker extends to the maximal value.

310

311

312

313

314

315

Fig. S2. Monosynaptic connectivity between AHiPM and VMHvl

Esr1

neurons.

316

(A) Schematic illustration of the experimental design, transducing AHiPM neuron

s with Chronos and

317

optically evoking postsynaptic responses in VMHvl

Esr1

neurons

ex vivo

318

(

) Quantification of VMHvl

Esr1

neurons with optically

evoked EPSCs (oEPSCs).

319

orange),

320

and in CNQX and AP5 (black); n=5 brain slices, collected from n=5 mice, one

way ANOVA with Dunnett’s

321

post hoc test,

= 0.0002 between baseline and TTX conditions,

= 0.0001 between baseline and

322

TTX+4AP conditions,

= 0.0002 between baseline and

CNQX+AP5 conditions. Shaded region

323

represents the standard error. The vertical scale bar defines current and the horizontal scale bar time.

324

***

< 0.001.

In box plots the median is represented by the center line, the interquartile range is

325

represented by

the box edges, the bottom whisker extends to minimal value, and the top whisker extends

326

to the maximal value.

327

328

Fig. S3. Characterization of LTP

inducing stimulation protocols at the

AHiPM



VMHvl

Esr1

synapse

329

(A) Schematic of the experimental design used to identify the appropriate stimulation protocol for LTP

330

induction

ex vivo

in socially naïve mice.

331

(B) Illustration of the experimental protocols tested to to induce LTP in the AHiPM



VMHvl synapse.

332

ng the optically induced EPSC (oEPSC) prior to, and following application of each of three

333

stimulation protocols (n=8 cells, n=5 socially naïve mice).

334

(D) Alternative quantification/illustration of optically induced EPSC (oEPSC) prior to, and following

335

application of each of three stimulation protocols (n=8 cells, n=5 socially naïve mice

–

similar to panel C).

336

337

338

339

340

341

342

343

344

345

Fig. S4. Optogenetic induction of LTP at AHiPM



VMHvl

Esr1

synapses in socially naïve mice leads

346

to elevated aggression in the first resi

dent

intruder test.

347

(A) Schematic indicative of the experimental design used to induced hypothalamic LTP in the

348

AHiPM



VMHvl synapses, via Chronos

eYFP expression in AHiPM, and ChrimsonR expression in

349

VMHvl

Esr1

neurons.

350

(B) Schematic of the behavior test de

sign used to identify whether induction of LTP in the AHiPM



VMHvl

351

synapses, influences the innate expression of aggression.

352

expressing (LTP) mouse, in the

353

resident

intruder test against

a novel BALBc conspecific.

354

(D) Quantification of attack duration (n=5

6 mice per group, two

tailed unpaired

test,

= 0.0046 between

355

YFP and LTP groups).

356

(E) Quantification of attack latency (n=5

6 mice per group, two

tailed unpaired

test,

= 0.0214 b

etween

357

YFP and LTP groups).

358

(F) Quantification of number of attacks per trial (n=5

6 mice per group, two

tailed unpaired

test,

359

0.0235 between YFP and LTP groups).

360

(G) Quantification of close investigation duration (n=5

6 mice per group, two

tailed unpaired

test,

361

0.7106 between YFP and LTP groups).

362

< 0.05, **

< 0.01.

In box plots the median is represented by the center line, the interquartile range is

363

repre

sented by the box edges, the bottom whisker extends to minimal value, and the top whisker extends

364

to the maximal value.

365

366

367

368

369

370

371

Fig. S5. Schematic summary.

372

(A) The schematic summarizes the findings from AGG mice, and the suggested links between

373

aggression,

serum testosterone and hypothalamic LTP.

374

(B) Similar to panel (A), but summarizing results from experiments in NON mice

this schematic

375

summarizes the identified links among aggression, serum testosterone and hypothalamic LTP. Our

376

results do not distin

guish whether the effect of elevated serum testosterone to increase LTP

in vivo

(Fig.

377

T) is direct, or rather indirect via an effect to increase aggressive behavior, which in turn enhances

378

LTP. However, exogenous administration of T to NON mice (in the

absence of any aggressive

379

experience) enhances LTP amplitude and persistence as tested

ex vivo

(Fig. 6K

O).

380

Materials and Methods

381

Table S1. Reagents and resources.

382

REAGENT or RESOURCE

SOURCE

IDENTIFIER

Antibodies

Rabbit monoclonal anti

DsRed

Takara

632392

Anti

GFP rabbit serum

Invitrogen

6455

Chicken polyclonal anti

GFP

Aves Labs, Inc.

GFP

1010

Donkey anti

Mouse IgG

Alexa Fluor

488

ThermoFisher

21202

Donkey anti

Rabbit IgG

Alexa Fluor

488

ThermoFisher

21206

Donkey anti

Rabbit IgG

Alexa Fluor

568

ThermoFisher

10042

Donkey anti

Rabbit IgG

Alexa Fluor

647

ThermoFisher

31573

Goat anti

Chicken IgY

Alexa Fluor

488

ThermoFisher

11039

Biotinylated Goat Anti

Rabbit IgG Antibody

Vector Laboratories

1000

Donkey anti

Rabbit IgG

Alexa Fluor

568

ThermoFisher

10042

Chemicals, Peptides, and Recombinant Proteins

Picric acid

Sigma

Aldrich

P6744

4% paraformaldehyde (PFA) in PBS

Santa Cruz Biotech.

CAS30525

Streptavidin conjugated to Alexa Fluor 647

ThermoFisher

CS32357

Neurobiotin tracer

VectorLabs

1120

Sodium chloride

Sigma

Aldrich

S9888

Sodium bicarbonate

Sigma

Aldrich

S6297

(+)

Glucose

Sigma

Aldrich

G7528

Sodium phosphate monobasic dihydrate

Sigma

Aldrich

71505

Potassium chloride

Sigma

Aldrich

P9333

Magnesium sulfate heptahydrate

Sigma

Aldrich

63138

Calcium chloride dihydrate

Sigma

Aldrich

C5080

Aminopyridine

Sigma

Aldrich

275875

CNQX disodium salt

TOCRIS

1045

AP5

TOCRIS

0106