Supplementary Information_revised-2 - aat4422_Patriarchi

www.sciencemag.org/

cgi/

content/

full/

science.

aat4422

/DC1

Supplementary

Material

s for

Ultrafast neuronal imaging of dopamine dynamics with designed

genetically encoded sensors

Tommaso Patriarchi, Jounhong Ryan Cho, Katharina Merten, Mark W. Howe,

Aaron Marley, Wei

-Hong Xiong, Robert W. Folk, Gerard Joey Broussard,

Ruqiang Liang, Min Jee Jang, Haining Zhong, Daniel Dombeck, Mark von Zastrow,

Axel Nimmerjahn, Viviana Gradinaru, John T. Williams, Lin Tian*

*Corresponding author. Email:

lintian@ucdav

is.edu

Published

31 May

2018 on

Science

First Release

DOI:

10.1126/science.

aat4422

This PDF file includes:

Materials and Methods

Figs. S1 to S16

Captions for data S1 to S3

References

Other supplementary material for this manuscript includes:

Data S1 to S3

Materials and Methods

Molecular Cloning

All constructs were designed using circular polymerase extension cloning (CPEC)

(

)

restriction cloning, and gBlock gene fragments (Integrated DNA Technologies). Sequences

coding for a hemagglutinin (HA) cleavable secretion motif and a FLAG epitope were placed at

the 5’

end of the construct as in

(

)

. HindIII and NotI cut sites were placed at the 5’

and 3’

end, respectively, for cloning into pEGFP

(Addgene)

to generate

all p

CMV

con

structs

BamHI and HindIII sites were introduced via PCR for final subcloning onto pAAV.

hSynapsin

and pAAV.

CAG

vectors.

To maximize coupling between conformational changes and chromophore fluorescence, we

chose to use a cpGFP module (LSS

cpGFP

DQL)

from GCaMP6

(

)

for insertion into

DRD1

via CPEC

For screening linker variants, we generated a linker library by first creating an

insert DNA carrying a 2 aminoacid

long randomized linker on each side of cpGFP (LSS

cpGFP

DQL). Single colonies were manually picked and grown overnight as described in

(

)

. All

sensor

sequences

generated

in this study are listed in the

Supplementary

Data

file.

Structural modeling and sequence alignments

Sequence alignments were performed using Jalview s

oftware (UK)

(

)

using a

percentage

identity color map

Inactive conformation of the sensor was predicted with rosetta_cm protocol

of rosetta 3 (v

ersion 2015.31)

(

)

. Primary sequence of sensor design was threaded with partial

thread ro

utine onto template PDB structures of inactive

adrenergic receptor

(

)

(ID: 2RH1)

and unbound state of GCaMP3

(

)

(ID: 4IK3). The threaded structures were then hybridized

together with rosetta_cm protocol for membrane protein

(

)

. A total of 374 PDB structures

were generated by rosetta_script routine. Structure with lowe

st total score was considered the

final model.

Cell culture

maging and quantification

HEK293T cells (ATCC #1573) and U2OS cells (ATCC #HTB

96) used in this study we

cultured and transfected as in

(

23, 44

)

. Primary hippocampal neurons were freshly isolated as

previously described

(

)

. Neurons were infected using AAVs (1 x 10

GC/ml) at DIV5, two

weeks prior to imaging. Prior to imaging

cells were washed with HBSS (Life Technologies).

Cell i

maging was performed

using

a 40X oil

based objective on an inverted Zeiss Observer

LSN710 confocal microscope with 488/513 ex/em wavelengths. For testing sensor responses,

neurotransmitters/drugs were directly applied to the bath during the imaging session. For

neurotransmitter tit

rations, a dual buffer gravity

driven perfusion system was used to exchange

buffers between different drug concentrations. Surface labeling was achieved as described

previously

(

23, 44

)

. One

photon emission spectrum for the sensors was d

ertermined

using the

lambda

scan function

of the

confocal

microscope

. Two

photon emission spectr

obtained

with a 40X water

based object

ive on a SliceScope (Scientifica) and w

used to obtain the

normalized two

photon cross

section using a custom

made script on M

ATLAB

For ROIs

selection, masks were generated either on the cell membrane or around the cytosol, depending on

the experiment,

using the threshold function in Fiji. We calculated spatial movies and images of

Δ

F/F

in response to an applied ligand as

퐹

(

푡

)

−

퐹

()*+,,-.+

퐹

()*+,-.+

with

퐹

(

푡

)

the pixel

wise

fluorescence value at each time,

, and mean fluorescence in t

ime points prior to ligand

application,

퐹

()*+,-.+

. To avoid the possibility of infinite pixel values, we added a small offset to

each pixel in

퐹

()*+,-.+

. Based on the

F/F maps, we calculated a corresponding SNR map as

∆

퐹

()*+

-.+

⁄

FF heatmaps were generated using a

custom

ATLAB

script. Surface

expression of sensors was quantified as the ratio of membrane fluorescence over cytoplasm

fluorescence. For titration curves, Kd values were obtained by fitting the data with a Hill

function on Igor Pro

(WaveMetrics)

Dopamine

uncaging

Uncaging experiments on neuronal dendrites were performed in the presence of 100 μM

caged

dopamine (carboxynitroveratryl

DA, CNV

(

)

) in the

HBSS

. Optical recordings

were performed at 153 ms/frame scan rate. Uncaging was achieved by shining 405 nm light

(40% laser power) on a 2 μm wide circular region selected 10 μm away from the dendrite

surface.

Internalization assay with flow

cytometry

24 hours post transfection, cells were re

plated onto 6 well dishes. The following day,

surface levels of receptors were assayed by addition of Alexa

647

conjugated M1 antibody

(Sigma) for 45 minutes, as described previously

(

)

. Fluorescence intensity profiles of cells

populations (>5000 cells) were measured using a FACS

Calibur instrument (BD Biosciences

Each condition was performed in duplicate. Internalization was calculated by measuring the

fraction of surface fluorescence remaining after 30 minutes of 1

M SKF81297 (Tocris) and

divided by the non

treated condition.

Luminescence

based cAMP assay

xperiments were conducted as previously described

(

)

. Briefly, cells were transfected

with the cyclic

permuted lucif

erase pGLOSensor

20F plasmid (Promega) and then treated with

luciferin (GoldBio) in phenol and serum free media for 1 hour in a 24 well dish. Luminescence

values for SKF81297 (Tocris) treated conditions were measured at their peaks and normalized

with refe

rence to 10

M forskolin (Sigma) at its peak.

Total Internal Reflection Fluorescence Microscopy (TIRF

FM) Live Imaging.

Live cell TIRF

FM was conducted with a Nikon Ti

E inverted microscope at 37°C

controlled humidity and CO

controlled chamber as de

scribed previously

(

)

. D1 specific

agonist S

KF81297 (Tocris) was added at 1

M while D1 antagonist SCH23390 (Tocris)

was

added at 10

Preparation and transfection of cortical organotypic slice cultures

Rat hippocampal slice cultures were prepared from P6

–

P7 pups as previously described

(

47, 48

)

. pCMV

dLight

2 (12

g) and pCMV

mCherry (3

g) were mixed with single cell

electroporation solu

tion (160 mM NaCl , 5.4 mM KCl, 12mM MgCl

, 2 mM CaCl

, 5 mM

HEPES, pH 7.4 ) to total volume of 36

l. One

week old hippocampal cultured slices were

transfected with single cell electroporation

(

)

. 5

–

7 days after transfection, single

hippocampal culture

d slice was transferred to the imaging chamber in a custom built two

photon imaging setup. Slices were perfused in 10ml gassed artificial cerebral spinal fluid

(ACSF) containing 4mM Ca

, 4mM Mg

, 0.5

M TTX and 10

M cocaine (DA transporter

blocker) durin

g imaging.

AAV viral production

All dLight

AAV constructs were cloned in the laboratory, and viruses were produced by the UC

Davis Vision Center Vector Design and Packaging Core facility. The viral titers of the viruses

used in this study were: AAV1.

CAG

lex.tdTomato, ~8 x 10

genome copies (GC)

/mL

(University of Pennsylvania); AAV5.

hSyn

apsin

flex

.ChrimsonR.tdTomato, ~4 x 10

/mL

(University of North Carolina); AAV1.

hSynapsin

.NES

jRGECO1a, ~3 x 10

/mL

(University of Pennsylvania); AAV1.

hSynapsin

.dLight

1, ~1 x 10

GC/mL; AAV1.

hSynapsin

.dLight

2, ~2 x 10

GC/mL; AAV9.

hSynapsin

.dLight

2, ~4

GC/mL;

AAV9.

CAG

.dLight

1, ~7

GC/mL; AAV9.

CAG

control_sensor

, ~1 x 10

GC/mL.

Animals

Animal studies were conducted in compli

ance with

the

Guide for the Care and Use of

Laboratory Animals of the National Institutes of Health and approved by the Institutional Animal

Care and Use Committee (IACUC) at the University of California, Davis or the relevant

institutional regulatory body

. Wild type rats were used in this study (0

2 months old) for

neuronal and organotypic slice culture preparation and two

photon imaging in brain slice.

Dopamine receptor

D1 (Drd1)

Cre mice (4

6 months old, Jackson Labs, Strain B6; 129

Tg(Drd1

cre)120Mxu/MM

Jax) were used for two

photon imaging in the dorsal striatum.

Vesicular GABA transporter (VGAT)::IRES

Cre

(

)

(Jackson Labs, Slc32a1tm2(cre)Lowl/J)

and Tyrosine hydroxylase (TH)::IRES

Cre

(

)

(Jackson Labs, B6.129X1

Thtm1(cre)Te/Kieg)

knock

in mice between 3 and 5 m

onths old were used for in vivo imaging using fiber photometry

and for optogenetic manipulations. Wild type

male

mice (2 to 5 months old,

C57BL6/J, Jackson

Laboratories)

were used for two

photon in vivo imaging in the cortex. Sample sizes (number of

mice)

for each experiment are stated in main text. All animals were group housed in standard

plastic cages on a 12

hour light/dark cycle with food and water available

ad libitum

, except for

mice trained on the visuomotor association task (see below). These mice

were house individually

on a reverse 12

hour light/dark cycle.

Anesthesia was performed with 4.5% isoflurane for

induction and 2%

for

maintenance.

Viral injections

Injection procedures were essentially identical to those described in

(

10, 52

)

with a few

exceptions.

For dopamine imaging in brain slice,

male and female Sprague

Dawley rats (p23

7) were

used for intracerebral microinjections. Briefly, anesthetized rats were immobilized in a

Stereotaxic Alignment System (Kopf Instruments). AAV1.

hSyn

apsin1

.dLight

2 (200 nl) was

injected uni

laterally into the dorsal striatum using the following coo

rdinates from bregma (in

mm): 0.48 anteroposterior (AP),

2.19 mediolateral (ML), 4.69 dorsoventral (DV).

For dopamine imaging in the dorsal striatum, mice were injected in the dorsal striatum at

two caudal locations (+0.6 and +0.2 AP and 2.1mm ML to bregm

a) and 3 depths below the

surface (

1.6mm,

1.9mm, and

2.2mm; 100 nL at each depth, total 0.6uL) with either AAV1.

hSynapsin1

.dLight

1 or AAV1.

hSynapsin1

.dLight

2 in combination with

AAV1.

CAG

.flex.tdTomato (diluted 1:100 in PBS).

The flex

tdTomato in o

ur experiments is

expressed in roughly half of the SPNs (Drd1

cre mice) to avoid potential issues, such as

competition of expression when two AAVs are co

expressed in the same neurons.

For dopamine imaging using fiber photometry,

AAV9.

CAG

.dLight

1 or

AAV9.

CAG

control_sensor

were injected into the NAc (+1.3 mm AP,

1.25 mm ML,

4.25

mm DV) and cre

dependent AAV encoding ChrimsonR

tdTomato was injected into the VTA

(

3.3 mm AP,

0.5 mm ML,

4.3 mm DV). For dual color photometry experiments, AAV

encoding

jRGECO1a was 1:1 mixed with dLight

1 virus and injected into the NAc region

(+1.3 mm AP,

5 mm ML,

4.25 mm DV). A 10 μL NanoFil microsyringe (World Precision

Instruments) with a blunt 35

gauge needle was used for viral injection and manipulated by a

crosyringe pump (UMP3, World Precision Instruments) and a controller (Micro4, World

Precision Instruments). 500 nL of AAV was slowly injected into the target coordinates over

10 minutes, and additional 10 minutes were waited to allow diffusion. The needle

was slowly

withdrawn over 10

15 minutes.

For dopamine imaging in the cortex,

anesthetized animals (4% isoflurane for induction;

~1.5% during surgery) were positioned in a computer

assisted stereotactic frame with digital

coordinate readout and atlas target

ing (Angle Two, Leica Biosystems Inc., Buffalo Grove,

IL)

icropipettes were loaded with virus solution and slowly lowered into the brain under a

~32º injection angle until the target depth (~0.2 mm) was reached. Manual pressure was applied

to a 30 mL syr

inge connected to the injection pipette. Virus solution was slowly injected over a

period of 5

–

10 min. Once desired injection volume (200

400 nl) was delivered, the syringe’s

pressure valve was locked and position maintained for approximately 10 min to all

ow virus to

spread and to avoid backflow upon needle retraction. Each mouse received two injections of

AAV9.

hSynapsin1

.dLight

2 (1:50 dilution). Injection coordinates were (AP 3

3.2 mm, ML 1

1.2 mm) and (AP 1.5 mm, ML 1

1.5 mm).

Slice preparation

isoflurane was used to deeply anesthetize rats prior to decapitation. Brains were rapidly

removed and placed in modified Krebs buffer containing (in mM) 126 NaCl, 2.5 KCl, 1.2

MgCl

, 2.4 CaCl

, 1.4 NaH

, 25 NaHCO

, 11 D

glucose, and with 10

M MK

801.

orizontal slices (245

m) were taken using a Leica vibratome and allowed to recover for at

least 30 minutes prior to use in Krebs with MK

801 being continuously bubbled with 95/5%

/CO

. Slicing and recovery was done at elevated temperatures (30

C). Following recovery,

slices were secured in a recording chamber maintained at 34

C and perfused with modified

Krebs at a rate of 3 ml/min.

Surgical procedures for in vivo imaging

Cannula implant for in vivo imaging in dorsal striatum

Two weeks post

njection, a chronic imaging window was implanted over the external

capsule fibers above the striatal injection site as described previously

(

)

Briefly, a 2.75mm

craniotomy was performed on anaesthetized mice, and cortical tissue was slowly aspirated until

the white matter of the external capsule was exposed. A thin layer of Kwik

Sil (WPI) was

applied over the imaging region, and a metal cannula

covered at one end by a glass coverslip was

inserted into the aspiration site down to the fiber surface.

Optical fiber implantation for fiber photometry and optogenetics

Mice were anesthetized with isoflurane and carbogen mixture (4

5% for induction, 1.5

2% for maintenance) and carefully placed to a stereotaxic frame (Kopf Instruments) after

shaving their hair. Head skin was sterilized with chlorohexidine and a midline incision was

made using a sterile scalpel. The skull surface was exposed and cleaned wi

th sterilized cotton

swabs. Bregma and lambda points were identified and leveled to be at the same dorsal

ventral

axis. Small craniotomy holes were made with drill bits (#73 size, Kyocera) over the nucleus

accumbens (NAc) and ventral tegmental area (VTA).

After viral delivery, optical ferrule/fiber

(for VTA/optogenetics, 300 μm diameter, 5

6 mm cut length, NA 0.37, home

made; for

NAc/photometry, 400 μm diameter, 5 mm cut length, NA 0.48, Doric lenses) was mounted to

a stereotaxic cannula holder (Doric lense

s). First, a 400 μm fiber was lowered to the NAc

through a craniotomy hole and stopped at 200 μm above the virus injection target. A layer of

adhesive cement (C&B Metabond, Parkell Inc.) was applied to the anterior portion of skull

surface to strongly hold

the implanted ferrule. Care was taken not to apply Metabond near the

posterior craniotomy sites for VTA. Next, a 300 μm fiber was inserted to the VTA and

stopped at 500 μm above the injection site. Another layer of adhesive cement was applied to

hold the

second fiber. After adhesive cement was fully dried, a thick layer of dental cement

(Lang Dental) was applied to build a head cap.

Head

plate

and cranial window

implantation for in vivo imaging in the cortex

A few weeks after viral injections,

mice were

implanted with a head plate and cranial

window on a custom surgical bed (Thorlabs Inc., Newton, NJ).

Body

temperature was

maintained at 36

–

37 °C with a DC temperature control system and ophthalmic ointment was

used to prevent eyes from drying. Skin was cl

eaned and disinfected with 70% ethanol and

Betadine. A small (~10 mm) incision was performed along the midline. The scalp was pulled

open and periosteum cleaned. First, a portion of the scalp was surgically removed to expose

frontal, parietal, and interpar

ietal skull segments. Then, the metal plate was affixed to the

bone with C&B Metabond Quick Adhesive Cement (Parkell Inc., Edgewood, NY). Next, to

enable chronic two

photon imaging, a custom

made cranial window was implanted, similar to

(

)

. The craniotomy (2.5 mm diameter) was centered around (AP 2 mm and ML 1.5 mm) and

sealed with a custom

three

layered cover glass assembly (each No.1 thickness) with the two

layers closest to the corte

x consisting of two pieces of circular 2.5 mm

diameter cover glass

and the outermost layer consisting of a circular 4 mm

diameter cover glass that rested on the

thinned skull. The dura mater was kept intact.

Behavioral experiments

Sucrose consumption

Mice were handled daily for 5

10 minutes to reduce anxiety associated with

experimenter’s handling. Following recovery from surgery (~7 days) mice were water

restricted to 1.5 mL per day and maintained at 85

90 % of ad libitum weight. All behavioral

experi

ments were carried out in an operant chamber within a sound

attenuating box (Lafayette

Instruments). Behavioral tasks were implemented and controlled by ABET II software

(Lafayette Instruments). TTL pulses were used to synchronize with fiber photometry

rec

ordings. After at least 7 days of water deprivation, mice were introduced to an operant

chamber and allowed to freely explore with a patch cord connected. 50 μL of 5 % sucrose

water was delivered every 60 seconds (total 10 times) so that mice can learn the

position of a

lick port. Lick was detected as a break of infrared beam at the lick port. This habituation

session was repeated two to three times for each mouse until they showed robust lick activity

(>400 licks per session). At the recording day, mice un

derwent the same sessions while

recording dLight1 fluorescence with fiber photometry.

For analysis, mean fluorescence values

were obtained from baseline (

10~

5 s from lick onset) and during consumption (0~5 s).

Unpredictable footshock delivery

Mice were

placed into an operating chamber with a patch cord connected for photometric

recordings. Five electric footshocks (0.6 mA for 1 second) were delivered with variable

intervals (randomly chosen from uniform distribution between 45 and 75 seconds) without

edictive cues. This was performed after all other experiments, since such aversive stimuli

can induce sustained fear and anxiety to the context.

For analysis, mean fluorescence values

were obtained from baseline (

1 s from lick onset) and during footsho

ck (0~2 s).

Cue

reward learning and extinction

In cue

reward learning sessions, CS was turned on for 10 seconds and US was available 7

seconds after CS onset, after variable inter

trial intervals (ITI). In cue

reward extinction sessions,

same CS was turne

d on for 10 seconds without US delivery after variable ITI.

In “expected reward delivery” trials, US was delivered after CS presentation, as in previous

learning sessions.

After sucrose consumption experiments, mice started appetitive Pavlovian

condition

ing, or cue

reward learning,

the same operant chamber. Conditioned stimuli (CS)

consisted of house

light and 70 dB 5kHz tone, and were turned on for 10 seconds with

variable intervals (randomly drawn from uniform distribution between 75 and 105 seconds)

Unconditioned stimulus (US) was 50 μL of 5 % sucrose water, available at the 7th second

after each CS onset. Pump sound from liquid dispenser was audible from 6th to 7th seconds.

US pairings were repeated 20 times per session (therefore ~30 minutes fo

r each session).

Mice underwent total of 12 sessions for cue

reward learning. Chambers and lick port were

sanitized with Accel and 70% ethanol between animals to remove any odor. For all sessions,

lick data and photometry recordings were simultaneously obt

ained. Two to three days after

the last cue

reward learning sessions mice began cue

reward extinction. Identical CS were

given 30 times per session with variable intervals (between 45 and 75 seconds), now without

US delivery. Mice underwent total of 5 exti

nction sessions. To quantify CS

triggered

behavior, number of licks was counted during CS presentation. For photometry data, peak

fluorescence was obtained for CS (0~3 s after CS onset) and US (

2~5 s around US

consumption onset

US consumption onset for

each trial was defined as the lick bout onset

after US is available, where lick bout onset was detected with a threshold of 2 Hz. To examine

fluorescence response shift from US to CS (or vice versa), we calculated CS

US index,

defined as (CS response

–

response) / (CS response + US response).

Unexpected reward delivery and omission

After 12 cue

reward learning sessions, mice underwent reward prediction error

experiments. In “unexpected reward delivery” session, animals were exposed to normal CS

US pairing trials; but in 25% of trials, US was delivered without CS, so that mice can exp

lore

to the lick port without predictive cues and consume reward in an unexpected manner. Peak

fluorescence values after US consumption onset was obtained and compared between

expected and unexpected conditions. In “unexpected reward omission” sessions, US

delivery

was omitted in 4 out of 20 trials (4th, 8th, 13th, and 16th trials) after predictive CS onset. For

analysis, mean fluorescence values were obtained from baseline (

10~0 s from CS onset) and

after CS offset (10.5~11.5 s) and compared.

Visuomotor

learning task

Prior to surgical preparation and behavioral training, mice were handled/tamed on two

consecutive days to reduce stress during experiments. Following recovery from surgery (~7

days) mice were water

restricted to 25 ml kg

per day and mainta

ined at 80

85% of ad libitum

weight. Training was performed in a custom

built setup that included a color LCD monitor

(12.1" LCD Display Kit/500cd/VGA, ICP Deutschland GmbH) on which visual stimuli were

presented. To reduce noise in optical recordings, the

monitor was covered with a color filter

(R342 Rose Pink, Rosco Laboratories Inc.). During training, mice were head

fixed with a

custom

build head holder and placed on a spherical treadmill (Habitrail Mini Exercise Ball,

Animal World Network) facing the LC

D display. An optical encoder (E7P OEM, US Digital)

attached to the treadmill allowed measurement of both speed and direction of ball movement.

Water reward was delivered with a programmable syringe pump (NE

500 OEM Syringe Pump,

New Era Pump Systems, Inc.

). Behavioral signals were acquired through a data acquisition board

(PCI

6221, National Instruments) connected to a breakout box (BNC

2110, National

Instruments) and interfaced to MATLAB using the Data Acquisition Toolbox (Version

R2010bSP2, The MathWorks

Inc.). Behavioral task sequence was controlled by the MATLAB

based software MonkeyLogic (www.monkeylogic.net)

(

56, 57

)

. Custom

written functions were

added to MonkeyLogic to enable analysis

and control of ball rotation parameters. Treadmill

encoder signals, trial marker codes (generated by MonkeyLogic), and imaging data were

acquired in synchrony allowing running parameters, behavioral task events, and image frames to

be linked.

Once the mou

se stopped moving on the ball, a sequence of task events was initiated.

First, a blue square frame was displayed on the monitor and required the animal to continue to

stand still for a period of 10 s (ball rotation/velocity

≤

2 mm/s). If the mouse continue

d to stand

still for the entire stand

still phase, a second stimulus, a filled blue square, was presented for 3 s,

instructing the mouse to initiate a run. If the mouse initiated sustained movement (ball

rotation/velocity >10 mm/s for >1s duration) during

the 3s stimulus phase, a water reward was

delivered. In 20% of pseudo

randomly selected trials the reward was withheld (“Unexpected

reward omissions”; light green traces in Fig. 5,

fig.

and S1

). If no running occurred, the

trials counted as a miss trial. If the animal began to move during the 10 s stand

still phase (ball

rotation/velocity > 2 mm/s), the trial was aborted and no water reward was delivered

(“Spontaneous run trials”; pink traces).

The mouse was able to initiate a new trial after an

intertrial interval (ITI) of 5 s.

During the first two days of training, mice spent ~15

30 min/day in the setup to become

accustomed to head restraint. Mice were then trained daily for 60

90 min during which they

performed ~300

700 trials. Task parameters were adjusted depending on individ

ual animal’s

performance. Initially, the duration of the stand

still phase was set to 3 s, the stimulus phase to

20 s, and the running threshold to 2 mm/s. This increased the chance for the mouse to receive the

task

dependent reward. The stand

still phase

was then progressively extended and the stimulus

phase shortened to establish the association between stimulus onset and running onset. Training

continued until median reaction times (RT) had dropped below 1500 ms, which indicated that

mice had learned to

associate the visual cues with the desired behavior. Task proficiency was

typically reached within 5

7 days. During the initial training phase all successful “Go” trials

were rewarded. Unexpected reward omission trials were introduced during imaging sessio

only.

In vivo optogenetic and pharmacological manipulations

For optogenetic experiments, a patch cord (1 m length, 1.25 mm zirconia ferrule, 300 μm

diameter fiber, Doric lenses) was used to connect to a ferrule on an animal’s head cap and a

swivel com

mutator (fiber

optic rotary joints, Doric lenses) to allow free movement. Another

patch cord was connected from the commutator and to a 635 nm diode

pumped solid

state

laser (Changchun New Industries Optoelectronics Technology). The intensity of laser was

measured with a power meter (PM100D, Thorlabs) and calibrated to be 10 mW at the fiber

tip. An external TTL pulse generator (OTPG4, Doric Lenses) was used to control the laser

output. For activating VTA DA neurons, 5 Hz, 10 Hz, and 20 Hz of 10 ms pulses we

applied for 2 seconds. For activating VTA GABA neurons, 40 Hz pulses were used

(

)

These pulse trains were repeated 40 times per animal with intervals of 30 seconds.

For

pharmacological manipulations, the following drugs were used: DRD1 antagonist (SCH

23390, 0.25 mg/kg) and selective DA reuptake inhibitor (GBR

12909, 10 mg/kg)

(

)

Animals were injected with these drugs or saline (0.9% sodium chloride) through

intraperitoneal delivery 30 minutes before optogenetics/photometry experiments

. Mean or

peak fluorescence values were obtained before (

1 s from onset), during (0.5~2 s) or after

(4~10 s) photoactivation.

Data acquisition and analysis for ex vivo and in vivo imaging

Ex vivo imaging in rat brain slice

Imaging was carried out i

n a custom built 2

photon microscope. Data were acquired and

collected using ScanImage software

(

)

. Slices were scanned at different frequencies,

depending on the experiment, using 920 nm light. Electrical stimulation through a glass mono

polar electrode placed within the slice near the area of imaging was used to evoke dopamine

release. Experiments w

ere carried out with a scan rate of 2 (128x128 pixels) or 15 (32x32 pixels)

Hz. The fluorescence over the entire field (20x20

m) was measured for each frame and plotted

against time. Drugs were applied by superfusion. Line scans were taken at a 500 Hz fra

me rate.

To measure the dopamine response curve, two

photon images were taken every minute. Image

analysis was performed using custom software written in MATLAB.

In vivo imaging in mouse dorsal striatum

Treadmill velocity and acceleration were sampled at

1000Hz by a rotary encoder (E2

5000,

US Digital) attached to the axl

of the treadmill. Two

photon imaging was performed as

described previously

(

)

, using the same collection optics, but without the electric lens. 920 nm

laser light was used for excitation and 1024x512 pixels time series datasets were acquired at 30

Hz. Imaging sessions began after mice were acclimated to head fixation and ran fr

equently on

the treadmill (~1

3 days). Imaging data was collected over 1

2 days from dorsal striatum fields

ranging from 250 to 450

m in diameter. All analyses were carried out using custom software

written in MATLAB. Time series movies were motion corre

cted on the static red channel using

algorithms described previously

(

)

and x and y shifts from the red were used to correct the

green channel. Mean whole field fluorescence was calculated from a large, hand

selected ROI

con

taining all regions of the field with visible cellular structure in the red td

Tomato channel.

Fluorescence time series were converted to

F/F by normalizing signals to an 8th percentile

baseline

(

)

within a sliding window (

15s around each point) to correct for slow drift and

bleaching. Significant positive

going transients were calculated as previously described

(

)

In vivo imaging with fiber photometry

Fiber photometry was performed as in

(

28, 63, 64

)

. For dual color imaging a three LED

system was used (FMC7, Doric Lenses): 490 nm for dLight1, 565 nm for jRGECO1a, and 405

nm to be used as isosbestic wavelength for both indicators.

The following dichroic filters and

excitation/emi

ssion filters were used (all from Semrock): Di02

R405, Di02

R442, FF495

Di03,

FF552

Di02, FF593

Di03, FF01

405/10, FF01

433/24, FF01

475/28, FF02

520/28, FF01

565/24 and FF02

641/75

(

28, 63, 64

)

Acquired photometry data were processed with custom

written codes in MATLAB. Raw data

from each channel were low

pass filtered at 25 Hz (for

single color imaging) or 12 Hz (for dual color imaging) using a 2nd order Butterworth filter

with zero

phase distortion. To calculate

F/F time series, a linear fit was applied to

the

405

signals

nd aligned to the 490 and 565 nm signals. The fitted 405 nm signal was

subtracted from 490 and 565 nm channels, and then divided by the fitted 405 nm signal to

yield

F/F values. Dual color imaging data were down

sampled to 100 Hz. In Pavlovian

conditionin

F/F time

series signal is further normalized using robust Z

score (subtraction

of median and division by median absolute deviation, calculated from the entire session) to

account for potential differences in signal variance across animals and sessions.

Photometry

signals were then extracted around relevant behavioral events (e.g. lick onset, footshock

delivery, CS onset) and averaged. Power spectral density was estimated for averaged

dLight1/

control sensor

fluorescence data upon optogenetic stimulation

of different

frequencies. We used Welch’s method with 2

second window size and 50% overlap to

estimate power from 0.5 to 25 Hz in 0.1 Hz step.

In vivo two

photon imaging in the cortex.

Once mice had reached proficiency on the task, they were imaged daily for up to 9 days.

Imaging was

performed using a resonant scanning two

photon microscope (Sutter Instrument)

equipped with a pulsed femtosecond Ti:Sapphire laser (Chameleon Ultra II, Cohe

rent). dLight

fluorescence was

excited with

910

light, and

detected using a ET525/70M emission filter

(

Chroma Technology Corp.

) and H7422

40 GaAsP photomultiplier tube (Hamamatsu

Photonics)

Average excitation power was 40

130

mW depending on imaging d

epth and dLight

expression levels/duration. Typical recording depth was 80

150

μm

below the pia. Data were

acquired using a Nikon 16x 0.8

NA water immersion objective.

A custom

made blackout curtain

around the microscope’s detector was used to reduce ligh

t contamination by the LCD monitor.

Images (512

512 pixels) were acquired at 30.8 frames/sec. Recording sessions consisted of five

to twelve ~10 min recordings, separated by short imaging breaks (3

5 min). Recordings within a

given session were performed a

t the exact same location to maximize the number of trial

repetitions for analysis. Recordings from different sessions were performed at the same injection

sites but offset either laterally or axially to maximize tissue volume being sampled

Analysis of

cortical in vivo imaging data

All data were analyzed using custom code written in MATLAB (MathWorks). Data from

four

animals were included in the analysis. Mouse #1 underwent 4 recording sessions starting 8

weeks after virus injection. Mouse #2 underwent 9

recording sessions starting 3 weeks after

injection

. M

ouse #3 underwent 2 recording sessions starting 2 weeks after injection

and Mouse

#4 underwent 4 recording sessions starting 6 weeks after injection.

Recording locations

(M1/M2/FrA) are indicated in

g. S1

To confirm the animals’ responsiveness to the “Go” stimulus, we compared proportions

of trials in which runs were triggered by “Go” stimulus presentation (“Hit trials”) and

proportions of trials in which no such response occurred (“Miss trials”)

(

ig. S1

A and C). A

steep increase in “Miss trials” at the end of the session indicated that mice had lost interest in the

water reward. Trials beyond that point (black vertical line in

ig. S1

A) were excluded from

analysis. To confirm that mice had lear

ned the task, we analyzed the animals’ RTs (defined as

the time interval between “Go” stimulus onset and movement onset/ball velocity >10 mm/s) (

ig.

and D). RTs below 1500 ms indicated that mice had learned to associate the visual cues

with the desir

ed behavior.

To reduce random noise in our time

lapse recordings a sliding average filter was applied

(8 frames or 260 ms). This filtering largely retained the temporal dynamics of dLight1

fluorescence signals. Lateral image motion (e.g., due to mouse mov

emen

t) was corrected using a

cross

correlation registration algorithm, with an average image of 50 consecutive frames from

the time

lapse recording serving as the reference image. The same reference image was used to

correct image motion of other recordings from the same location. Noise in

motion

corrected

image data was further reduced using a Kalman filter.

To quantify/classify dLight1 transients, we tiled the field of view (FOV) with equally

sized (~17x17μm) regions of interest (ROIs). To exclude ROIs with little or no dLight1

expression

(e.g., on blood vessels) we first generated a mean fluorescence projection image from

the corresponding time

lapse recording. We then calculated the projection image’s pixel intensity

distribution. ROIs with mean pixel intensity values below the 55

perc

entile of this distribution

were excluded from analysis. To extract fluorescence time traces F(t) from all remaining ROIs,

pixel intensities of each ROI were averaged.

F(t)/F was calculated as (F(t)

–

mean F) / mean F.

ROIs that showed transients with neg

ative amplitudes (predominantly located near blood

vessels) were excluded from further analysis.

Each ROI’s time trace included multiple trials/task repetitions. Based on the animal’s

performance on the trials, time traces were subdivided into “Expected r

eward”, “Unexpected

reward omission”, and “Spontaneous run” trial traces. This was done for all recordings from the

same location using the same ROIs. Extracted trial traces were then sorted by trial type, aligned

to running onset, and averaged (

ig. S1

single ROIs

). To ensure that cued and spontaneous

running bouts were comparable, only traces from spontaneous runs starting > 5 s after stand

still

cue onset and with > 5 mm/s ball rotation/velocity were included. To calculate population

responses, “Expec

ted reward”, “Unexpected reward omission”, and “Spontaneous run” trial

traces from all animals, all sessions, and all significant ROIs were averaged (Fig. 5E

, group

average

To identify ROIs with significant increases in dLight1 fluorescence in response

to the

task, three analysis intervals were defined (

ig. S1

A):

)

Baseline (from

7 to

2 s prior to run

onset),

(II)

reward expectation (from run onset to 3.3 s after run onset), and

(III)

reward interval

(from 4.3 s to 10.8 s after run onset).

F/F tra

ces for each trial were averaged during the three

trial intervals respectively. The distributions of the averaged

F/F traces for particular trial types

were then statistically compared among each other and between intervals. ROIs that showed

significant f

luorescence increases during the reward expectation but not the reward interval, and

for all trial types compared to the baseline interval, were defined as “ROI active during

locomotion” (white squares in Fig. 5 and

fig.

G).

ROIs that showed significant

fluorescence

increases during the reward expectation but not reward interval for cued/triggered but not

spontaneous run trials were classified as “ROI active during reward expectation” (black squares).

Finally, ROIs that showed significant fluorescence in

creases during the reward interval for

rewarded but not un

rewarded or spontaneous run trials were defined as “ROI active during

reward” (red squares). Significance was determined using the Wilcoxon’s rank

sum test

(p<0.05), Bonferroni corrected for multip

le comparisons.

All ROI analysis was performed on motion corrected image data (see above) averaged

across trials of the same type and from the same session. To investigate whether this averaging

introduces artifacts in the DA transients (e.g., due to ster

eotypic running

related image

movements) we inspected our data more closely at cellular and population levels. Close

observation of a 40x45 pixel detail of the motion corrected image data revealed stable and sharp

images even during the most vigorous runni

ng periods (

ig. S1

A).

Next, we plotted the full field

of view (FOV) intensity changes for the average data of each trial type aligned at run onset (

ig.

B).

This revealed that the main characteristics of the single ROI transients are retained in the

ull FOV average, arguing against a pronounced contamination by motion artifacts. Next, to

investigate residual lateral shifts in the average image data for each trial type, we applied our

motion correction algorithm. The detected residual image motion was

<2 pixels for all sessions

and mice (

ig. S1

C and D). To verify that our motion correction algorithm reliably detects sub

and supra

pixel image displacements we artificially introduced random lateral shifts of up to 5

pixels to the same average image dat

a (

ig. S1

E).

The error between the introduced shifts and

the shifts detected by the algorithm across 100 repetitions was 0.089 ± 0.001 pixels (mean ±

SEM). Together, this suggests that DA transients extracted from our 17x17μm ROIs are unlikely

an artifac

t of image motion (

ig. S1

F).

Histology

Histological verification of proper sensor expression was as described previously

(

)

Primary antibodies used were: chicken anti

GFP (1:500; GFP

1020, Aves Labs), chicken anti

TH (1:500; TYH, Aves Labs) and rabbit anti

RFP (1:500; 600

401

379, Rockland A&A).

Hybrid chain reaction comb

ined with immunohistochemistry

assess the expression pattern of ChrimsonR

tdTomato injected into the VTA of

VGAT::IRES

Cre mice, we performed combined in situ hybridization for labeling

vgat

mRNA

and immunohistochemistry (IHC) for detecting DAerg

ic neurons (with TH). To achieve hig

sensitivity

in situ hybridization of

vgat

in tissue slices, we used hybridization chain reaction

(HCR

)

(65,

First, we designed

probes

for targeting

vgat

using a custom

written software

(available at https://github.com/GradinaruLab/HCRprobe). Each probe consists of 20

nt target

sequence, 2

nt spacer, and 18

nt initiator

. We

selected the target sequences that

has

(1)

the GC

content of

60%, (2) no nucleotide repeats more than 3, (3) no

more than 20 hits when blasted,

(4) higher than a

G of

9 kcal/mol to avoid self

dimers. Then we blasted full sequences (target

sequences with spacer and initiator)

and

calculate Smith

Waterman alignment scores between all

possible pair

to exclude probes

forming cross

dimers. The designed probes were synthesized by

Integrated DNA Technologies.

HCR was performed on

thick slices and kept in RNAlater solution

(ThermoFisher

Scientific).

Slices

were permeabilized in PBST (1xPBS with 0.1% Triton X

100) for 1 hour at

room temperature

(RT)

and pre

hybridized in hybridization buffer (2x saline

sodium citrate

(SSC) , 10% ethylene carbonate, 10% dextran sulfate) for another hour at 37

°C

Then the

samples were incubated in pre

warmed hybridization buffer i

ncluding probes (2 nM for each) at

°C

overnight.

After hybridization, w

e performed stringent washes with wash buffer (2xSSC,

10% ethylene carbonate) for 30 min at 37

°C

and washes with 2xSSC for 30 min at

(

twice

for

each)

The amplification step was performed as described in

(

)

overnight.

For IHC labeling of TH on HCR

labeled tissue slices, we used IHC buffer consist

ing

2xSSC, 1% donkey serum, and 0.1% Triton X

100.

incubate

the samples in IHC buffer for

1 hour at

for blocking

and added primary antibody (anti

TH, AB152, Millipore

, 1:200

Primary antibody reaction was performed for overnight at

. Following the wash steps with

2xSSC at

for 30 min twice, the samples were incubated in IHC buffer including secondary

antibodies

(1:200)

for overnight at

. The samples were rinsed with 2xSSC a few times and

mounted on a glass slide with mounting media (Prolong Diamond, ThermoFisher Scientific).

Probe sequences

(5’

>3’)

CACTTCATATCACTCACTaaGACACGGAGGTGGCCACATT

CACTTCATATCACTCACTaaGCGATGCTCAAAGTCGAGAT

CACTTCATATCACTCACTaaCCTGAATGGCATTTGTCACG

CACTTCATATCACTCACTaaACCAGGACTTCTGCGACACG

CACTTCATATCACTCACTaaTTCTTCAGGAAGGCGCAGGG

CACTTCATATCACTCACTaaTTCTCCCAGGCCCAATCACG

CACTTCATATCACTCACTaaGTGTAGCTGAACACGA

TGAT

CACTTCATATCACTCACTaaGCGGCGATGTGTGTCCAGTT

CACTTCATATCACTCACTaaACTTCCTTGGTCTCGTCGGC

CACTTCATATCACTCACTaaCGCGAAGAAGGGCAACGGAT

CACTTCATATCACTCACTaaAGGCGCAAGTGGAAGAGGCT

CCTTGGCCTGGGACTTGTTGaaCCCAATCTCTATCTACCC

ATGTCCATCTGCAGGCCCTGaaCCCAATCTCTATCTACCC

TAGGCCCAGCACGAACATGCaaCCCAATCTCTATCTACCC

CACCGCTGTGGCTATGATGGaaCCCAATCTCTATCTACCC

ACTTGGACACGGCCTTGAGAaaCCCAATCTCTATCTACCC

CGTCGATGTAGAACTTCACCaaCCCAATCTCTATCTACCC

AGGGCAGGAAGATCTGCGACaaCCCAATCTCTATCTACCC

AGAGACCCTTGAGCACGCAGaaCCCAATCTCTATCTACCC

CGGGCAGGTTATCCGTGATGaaCCCAATCTCTATCTACCC

TTCTCCAGCACTTCGACGGCaaCCCAATCTCTATCTACCC

ACAGCAGCTTGCGCCAGAGAaaCCCAATCTCTATCTACCC

Statistical Analysis

All

statistical analyses were performed in M

ATLAB (MathW

orks)

, Igor Pro (WaveMetrics)

or Prism (GraphPad). We used both parametric (t

test,

one

way

ANOVA

, two

way ANOVA

)

and non

parametric (Wilcoxon’s rank

sum) tests.

Pearson’s correlation coefficients and their p

values were calculated to assess how behavior or dLight1 fluorescence evolved across reward

learning or extinction.

All tests were two

tailed. Error bars are standard error of the mean or

standard deviation as

indicated in figure legends and main text. No statistical methods were used

to predetermine sample size.

Fig. S1.

Screening dLight1 variants

(

) Membrane expression and response of original dopamine sensor built by

inserting the original GCaMP6 cpG

FP module (LSS

cpGF

DQL) into DRD1. Peak response (

19.4 ± 0.02 %

∆

F/F, mean ± SEM, n=12 cells). Inset shows membrane intensity profile. Scale bar, 10 μm. (

) Resting fluorescence for

all screened variants that showed expression. P

roportion of respon

se type in horizontal bar

Color

coding indicates

significance value (n=3). (

) Optimizing insertion site of cpGFP in DRD1. (

) Sequence alignment of DRD1 and

adrenergic receptor encompassing TM5

TM6 (replacement sites was indicated as 0 aa in each TM

region) is shown. (

) Representative images showing membrane expression of dLight1 variants. Scale bar, 10 μm. (

) Quantification of

∆

F/F and surface expression from variants shown in

F/F:

3 aa, 70.1 ± 10.8 %;

2 aa,

1.3 ± 6.7 %;

1 aa, 199.4 ±

3.1 %; 0 aa, 223.2 ± 8.2 %; +1 aa, 652.3 ± 9.8 %. Surface expression values quantified in arbitrary units and

normalized to 0 aa control (

3 aa, 33.6 ± 4 %;

2 aa, 30 ± 1.8 %;

1 aa, 88.8 ± 6.4 %; +1 aa, 87 ± 2.7 %). (

)

Quantification of fluorescent signa

l changes and surface expression from variants shown in e.

F/F values:

2 aa,

9.2 ±

2.3 %; 0 aa, 223.2 ± 8.2 %; +2 aa, 205 ± 21 %). Surface expression values quantified in arbitrary units and normalized

to 0 aa control:

2 aa, 36.4 ± 2.9 %; +2 aa, 25.9 ±

0.86 %. All values are shown as mean ± SEM. *p<0.05, **p<0.01,

****p<0.0001 versus 0 aa position control, One

way ANOVA with Dunnett’s post

hoc test.

ig. S2. Characterization of dLight1.3a and dLight1.5.

(

) Top, aminoacid sequence of DRD2 insertion site for the

cpGFP module. Bottom, representative images of dLight1.3a and dLight1.5 expressed on HEK293 cells before and after

addition of DA (10 μM) and corresponding SNR heatmaps. Scale bars, 10 μm. (

) Quanti

fication of fluorescence

response of dLight1.3a (n=12) and dLight1.5 (n=4) to DA titrations on HEK293 cells.

Data were fit with Hill equation

(EC50: dLight1.3a, 2300

nM; dLight1.5, 110

nM). (

) Representative images of dLight1.5 under basal

condi

tions, after addition of quinpirol

(10 μM) and haloperidol (50 μM), and trace showing

quantification of

fluorescence

response during bath application of drugs

(n=15). All data are shown as mean ±

SEM

ig. S3. One photon emission spectrum and two

photon cross

section of

dLight

1 expressed in HEK293T cells.

Emission spectrum of both dLight1.1 and dLight1.2 peaks at 517 nm. Under 2P illumination, emission peaks are driven

at 920 nm.