Article

Intact-Brain Analyses Reveal Distinct Information

Carried by SNc Dopamine Subcircuits

Graphical Abstract

Highlights

d Intact, inclusive approaches to classifying neuronal cell

types

d Differential brain-wide circuit incorporation of SNc dopamine

neuron subpopulations

d Opposite valence encoding of shock by projection target-

defined SNc neurons

d Independently controlled information streams from the SNc

to the DMS and DLS

Lerner et al., 2015, Cell 162, 635–647July 30, 2015 ª2015 Elsevier Inc.http://dx.doi.org/10.1016/j.cell.2015.07.014

Authors

TaliaN. Lerner,CarrieShilyansky, Thomas

J. Davidson, ..., Liqun Luo, Raju Tomer,

Karl Deisseroth

Correspondencedeissero@stanford.edu

In Brief

Exploring the mammalian brain with an

array of intact-brain circuit interrogation

tools—including CLARITY, COLM,

optogenetics, viral tracing, and fiber

photometry—reveals that neurons in the

SNc region present different biophysical

properties, wiring of inputs and outputs,

and activity during behavior, despite

signaling through the same

neurotransmitter.

Article

Intact-Brain Analyses Reveal Distinct InformationCarried by SNc Dopamine SubcircuitsTalia N. Lerner,1,2 Carrie Shilyansky,3 Thomas J. Davidson,1,2 Kathryn E. Evans,4 Kevin T. Beier,3,4,5,6

Kelly A. Zalocusky,1,2,7 Ailey K. Crow,2 Robert C. Malenka,3,5 Liqun Luo,4,6 Raju Tomer,1,2 and Karl Deisseroth1,2,3,6,*1Department of Bioengineering2CNC Program3Department of Psychiatry and Behavioral Sciences4Department of Biology5Nancy Pritzker Laboratory6Howard Hughes Medical Institute7Neuroscience Program

Stanford University, Stanford, CA 94305, USA

*Correspondence: deissero@stanford.edu

http://dx.doi.org/10.1016/j.cell.2015.07.014

SUMMARY

Recent progress in understanding the diversity ofmidbrain dopamine neurons has highlighted theimportance—and the challenges—of defining mam-malian neuronal cell types. Although neurons maybe best categorized using inclusive criteria spanningbiophysical properties, wiring of inputs, wiring ofoutputs, and activity during behavior, linking all ofthese measurements to cell types within the intactbrains of living mammals has been difficult. Here,using an array of intact-brain circuit interrogationtools, including CLARITY, COLM, optogenetics, viraltracing, and fiber photometry, we explore the diver-sity of dopamine neurons within the substantia nigrapars compacta (SNc).We identify two parallel nigros-triatal dopamine neuron subpopulations differing inbiophysical properties, input wiring, output wiringto dorsomedial striatum (DMS) versus dorsolateralstriatum (DLS), and natural activity patterns duringfree behavior. Our results reveal independently oper-ating nigrostriatal information streams, with implica-tions for understanding the logic of dopaminergicfeedback circuits and the diversity of mammalianneuronal cell types.

INTRODUCTION

Dopamine (DA) is a neurotransmitter that is crucial for many bio-

logical processes relevant to health and disease and is thought

to regulate (among other behaviors) voluntary movement, rein-

forcement learning, and motivation (Bromberg-Martin et al.,

2010). Seminal early studies into the information encoded by

midbrain DA neurons suggested that a key function of DA is

to transmit reward prediction error signals (Mirenowicz and

Schultz, 1996; Schultz et al., 1997; Waelti et al., 2001), a hypo-

thesis concordant with temporal difference learning models

(Montague et al., 1996, 2004; Steinberg et al., 2013; Sutton,

1988). However, not all midbrain DA neurons appear to encode

similar information in their activity patterns (Bromberg-Martin

et al., 2010; Lammel et al., 2014; Roeper, 2013). For example,

DA neurons have been observed that differ in their responses

to aversive stimuli, leading to the hypothesis that the DA neu-

rons, which increase their firing in response to these stimuli,

may signal salience rather than value (Brischoux et al., 2009;

Bromberg-Martin et al., 2010; Matsumoto and Hikosaka,

2009), though controversy on this point remains (Cohen et al.,

2012; Fiorillo et al., 2013, 2013; Ungless et al., 2004).

The concept that there are diverse subsets of midbrain DA

neurons transmitting distinct signals naturally leads to the ques-

tion of whether these subsets are functionally incorporated into

different circuits with different roles in the brain. Explorations of

DA neurons in the ventral tegmental area (VTA) have revealed

that these neurons can be divided into distinct categories based

on their projection targets, which include the prefrontal cortex,

nucleus accumbens (NAc) core, NAc medial shell, NAc lateral

shell, and amygdala. When divided by projection target, different

VTA DA neuron classes express varying levels of the DA trans-

porter (DAT), DA D2 autoreceptors, GIRK channels, and HCN

channels mediating the Ih current (Lammel et al., 2008, 2011;

Margolis et al., 2008), all of which could affect the dynamics of

signals represented and transmitted. Furthermore, projection

target-defined subpopulations are located in different subre-

gions of the VTA, their excitatory synapses are differentially

modulated by rewarding and aversive stimuli, and they receive

distinct inputs, which can elicit opposite behaviors when selec-

tively recruited (Lammel et al., 2008, 2011, 2012).

Distinctions among DA neurons have also increasingly been

drawn between DA neurons within the VTA and those within

the substantia nigra pars compacta (SNc). For example, a recent

viral circuit tracing study showed that VTA and SNc DA neurons

receive different proportions of inputs from key brain regions

(Watabe-Uchida et al., 2012). In particular, this study noted

that SNc neurons receive a large proportion of their inputs

from the dorsal striatum. In contrast, studies attempting to func-

tionally characterize direct striatal projections to midbrain DA

neurons using optogenetics and slice recordings that have failed

Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc. 635

DM

S-p

roje

ctin

gD

LS-p

roje

ctin

g

NAc

BNST

GPe CeA

DS

DS

S1

HY

PAG

MRNBNST

DS

GPe

DS

CeAHY LHb

LHb

PAG

MRNNAc

S1

M1/2 S1

Cortex (

Other)

NAc core

NAc med

shell

NAc lat

shell

DMSDLS

FS (IPAC)

GPeSI/V

PBNST

CeA

Hypothala

mus

Thalamus

LHb SC ICSNc

SNrVTA

PAG

Dorsal R

aphe

MRNPPN PB

0

10

20

30

40

% ip

sila

tera

l inp

uts DMS-projecting

DLS-projecting

****

***

CAV-cre AAV

Rabies Virus

RVdG-GFP

EnvA

2 w

eeks

CAG pAG

CAG pATC

A B

E

F

mergeTHTC

mergeTHRVdG-GFP TC

RVdG-GFPDLS

-pro

ject

ing

DM

S-p

roje

ctin

g

DDMS-proj DLS-proj

0

25

50

75

100

% T

H+

star

ter c

ells

THTCRVdG-GFP

SNcVTA

THTCRVdG-GFP

M LV

D

SNcVTA

DLS-projectingDMS-projecting C

Figure 1. Whole-Brain Mapping of Inputs to DMS-Projecting and DLS-Projecting SNc DA Neurons

(A) Viral injection strategy for whole-brain input mapping based on output (TRIO). CAV-cre is injected into the striatum (DMS or DLS), and AAVs expressing cre-

dependent TC and G are injected into the SNc. Two weeks later, RVdG-GFP is injected into the SNc, where it infects TC-expressing cells and spreads one

synapse upstream from G-expressing cells.

(legend continued on next page)

636 Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc.

to identify such connections (Chuhma et al., 2011; Xia et al.,

2011) or found them to be very weak (Bocklisch et al., 2013).

The functional connectivity from striatum to SNc thus remains

an open question.

In addition to differences from VTA neurons, there are hints in

the literature that SNc DA neurons alone could be further divis-

ible into functionally distinct subsets. One pioneering study

demonstrated that DA neurons located more medially within

the SNc express higher levels of K-ATP channels, whichmediate

burst firing (Schiemann et al., 2012). Another study used record-

ings in awake monkeys to demonstrate a correlation between

the depth at which a DA neuron was recorded along the elec-

trode (indicating a more ventromedial location in the midbrain)

and the likelihood that it would decrease rather than increase

its firing in response to an aversive cue (Matsumoto and Hiko-

saka, 2009).

How might such differences among SNc DA neurons relate to

circuit wiring and function? One intriguing hypothesis is that

distinct locations within the SNc give rise to projections targeting

distinct locations within the dorsal striatum, such as dorsomedial

striatum (DMS) and dorsolateral striatum (DLS); distinct SNc

subfields could furthermore be set up to receive, process, and

transmit functionally distinct streams of information from else-

where in the brain. Such circuit organizationwould have powerful

implications. For example, given the highmedial SNc expression

of K-ATP channels that enable bursting, the prediction could be

made that the striatal target field of themedial SNcwould receive

strong novelty-driven bursts of DA during exploratory behavior

that the striatal target field of the lateral SNc could not receive.

However, other lines of research suggest that DAergic projec-

tions to the DMS and DLS carry similar information streams

because restriction of DA signaling to either subregion has

similar effects (Darvas and Palmiter, 2009, 2010). In fact, SNc

DA neurons could in principle even be similar to VTA DA neurons

in terms of the valence of information represented because mice

will self-stimulate for SNc DA neuron activity just as for VTA DA

neuron activity (Ilango et al., 2014; Rossi et al., 2013).

In the end, though this is clearly a question of great potential

significance, it remains unknown whether different SNc DA

neuron populations are fundamentally distinct in their projection

targets in striatum, in the types of information represented in their

electrical activity, or in the sources of their incoming information

across the brain. Here, using an array of intact-brain circuit

interrogation tools, including CLARITY, COLM, optogenetics,

viral tracing, and fiber photometry during behavior, we explore

this issue and find that all three conjectures regarding the

(B) Low-magnification (53) images of DMS- and DLS-projecting SNc ‘‘starter cells

shows TC expression, and blue shows (TH) immunostaining.

(C) Quantification of the percentage of starter cells that were TH+. Error bars are

(D) High-magnification (403) images from the sections shown in (B).

(E) Examples of GFP labeling of ipsilateral inputs to DMS- and DLS-projecting S

nucleus accumbens; DS, dorsal striatum; BNST, bed nucleus of the stria termin

amygdala; LHb, lateral habenula; DR, dorsal raphe; MRN, midbrain reticular nuc

(F) Quantification of ipsilaterally labeled inputs to DMS- and DLS-projecting SNc D

***p < 0.001 and ****p < 0.0001. Other abbreviations: motor cortex (M1/2), somat

limb of the anterior commissure (FS/IPAC), substantia innominata/ventral pallidu

compacta (SNc), substantia nigra pars reticulata (SNr), ventral tegmental area (V

See also Figure S1 and Table S1.

complexity of SNc functional wiring are borne out, with substan-

tial implications for understanding the logic of DAergic signaling

to striatum.

RESULTS

Unbiased Brain-wide Mapping of Inputs to StriatumSubfield-Projecting SNc NeuronsTo directly explore the hypothesis that distinct SNc neuron pop-

ulations receive and deliver separable information streams, we

began with an unbiased approach for globally mapping the

input/output relationships of SNc neurons. We mapped inputs

onto two subpopulations of SNc DA neurons defined by their

outputs to the DMS and DLS. These regions of the dorsal stria-

tum have been functionally distinguished by previous studies

(Castane et al., 2010; Faure et al., 2005; Featherstone and

McDonald, 2004; Yin and Knowlton, 2004; Yin et al., 2004,

2005a, 2009, 2005b), leading us to hypothesize that SNc projec-

tions to these areas might also participate in separable circuits.

We achieved unbiased global visualization of SNc inputs and

outputs using a combination of CLARITY and CLARITY-opti-

mized light-sheet microscopy (COLM) (Chung et al., 2013;

Tomer et al., 2014), along with a variation of rabies-based circuit

mapping (Schwarz et al., 2015); the lattermethod (TRIO) operates

similarly to previously published rabies-based circuit mapping

technologies, utilizing a GFP-expressing rabies virus (RVdG-

GFP) that both lacks the glycoprotein needed to spread trans-

synaptically and is pseudotyped with EnvA (so that it can infect

only neurons expressing TVA, an avian receptor protein normally

absent in mammalian cells [Wickersham et al., 2007]).Glycopro-

tein andTVAcan thenbeexpressedusingcre-dependent vectors

to direct the rabies virus to infect and spread to connected inputs

from a cre-defined subset of neurons. In TRIO, cre is delivered

from the retrograde CAV-cre virus (Hnasko et al., 2006; Soudais

et al., 2001), which transduces axon terminals and thereby de-

fines ‘‘starter cells’’ (fromwhich rabies tracingwill occur) by virtue

of their projection target. We injected CAV-cre into either the

DMS or DLS and then injected AAVs expressing cre-dependent

rabies glycoprotein (G) and TC, a high efficiency version of the

TVA receptor fused to mCherry (Miyamichi et al., 2013), into the

SNc. Two weeks later, we injected RVdG-GFP into the SNc (Fig-

ure 1A). Control experiments revealed that resulting putative

starter cells in the SNc, as defined by neurons that expressed

both RVdG-GFP and TC, were predominantly DAergic (TH+;

96% ± 3% after CAV-cre injection into the DMS, n = 4 mice;

98% ± 2% after CAV-cre injection into the DLS, n = 4 mice;

,’’ fromwhich tracing likely occurred. Green shows RVdG-GFP expression, red

SEM.

Nc DA neurons from various brain regions. S1, somatosensory cortex; NAc,

alis; GPe, globus pallidus external segment; HY, hypothalamus; CeA, central

leus; PAG, periaqueductal gray.

A neurons, shown as a percentage of all ipsilateral inputs. Error bars are SEM.

osensory cortex (S1), fundus of the striatum/interstitial nucleus of the posterior

m (SI/VP), superior colliculus (SC), inferior colliculus (IC), substantia nigra pars

TA), pedunculopontine nucleus (PPN), and parabrachial nucleus (PB).

Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc. 637

Figures 1B–1D). Additional controls demonstrated that the

RVdG-GFPwasproperly EnvA-pseudotypedand that long-range

tracing of inputs was cre dependent (Figures S1A and S1B).

We began our analysis of GFP-labeled inputs to DMS- and

DLS-projecting SNc DA neurons by optically clarifying the

labeled brains (Chung et al., 2013; Tomer et al., 2014) and visu-

alizing using COLM (Tomer et al., 2014). We reasoned that,

although thin sectioning approaches used for analysis of

anatomical tracing have yielded important insights, these are

not ideal for visualization of brain-wide patterns. Thin sectioning

results in particular caveats for the interpretation of negative

results: slices may tear, fragment, or otherwise be damaged or

lost, and strategies to minimize overcounting of somata split

across adjacent sections (such as counting only separated sec-

tions—e.g., every third or sixth) may underestimate cell counts,

particularly in small brain regions. Whole-brain CLARITY/

COLM circumvents these difficulties, allowing an intact global

overview of brain-wide structural patterns.

Using CLARITY/COLM, we noted an interesting GFP expres-

sion pattern in the striatum (Movies S1, S2, S3, and S4). When

inputs to DMS-projecting SNc DA neurons were labeled, we

observed relatively strong labeling in the NAc and DMS (Movies

S1 and S2) in comparison to the DLS, whereas when inputs to

DLS-projecting SNcDA neurons were labeled, we observed rela-

tively strong labeling in the DLS, particularly in the caudal tail of

the striatum (Movies S3 and S4) in comparison to the DMS and

NAc. Moreover, fine details of local neuronal architecture in

GFP-labeled neurons could be observed in higher-resolution vol-

ume renderings (Movie S5).

To quantify these divergent input patterns to SNc subfields de-

pending on their projection target, we followed up with detailed

targeted slicing approaches and quantified the inputs from

each brain region to DMS- and DLS-projecting SNc DA neurons

as a proportion of the total inputs observed (Figures 1E and 1F

and Table S1). Existing atlases used to define major brain re-

gions were further refined to define the boundaries of the DMS

and DLS (Figures S1C and S1D). Because the inputs to SNc

DA neurons are almost entirely ipsilateral (DMS-proj 96.4% ±

0.9%, n = 4; DLS-proj 96.5% ± 0.8%, n = 4, p = 0.91), we focused

on the injected hemisphere (contralateral inputs were examined

separately; Figures S1E and S1F and Table S1), noting first that

brain-wide inputs to SNc DA neurons, regardless of projection

target, broadly matched those observed by Watabe-Uchida

et al. (2012) using RVdG-GFP tracing from the SNc of DAT-cre

mice. However, despite gross similarities between groups, a

two-way ANOVA revealed a significant interaction between

starter cell projection target and input area (p < 0.001). Multiple

comparisons revealed significant differences in inputs from the

DMS (p < 0.001) and the DLS (p < 0.0001; Figure 1F), indicating

a marked preferential reciprocal connectivity of dorsal striatal

subregions to their specific DAergic SNc input subregions.

Injections in D1-tomato bacterial artificial chromosome (BAC)

transgenic mice further revealed that striatal inputs to SNc DA

neurons arise from DA D1 receptor-expressing neurons (Fig-

ure S1G; 92/97 DMS-projecting neurons and 210/214 DLS-

projecting neurons), consistent with the idea that D1 (but not

D2) striatal neurons project directly to the midbrain and with

hypotheses regarding the patch/matrix organization of striatum

638 Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc.

(Crittenden and Graybiel, 2011). These anatomical studies

suggested a fundamental distinction in circuit incorporation for

projection-defined SNc subregions and provided a roadmap

for a more detailed investigation of the circuit.

Organization of SNc DA Neurons Projecting to DistinctRegions of Dorsal StriatumTo quantify the spatial patterns of cre expression following injec-

tions of CAV-cre into the striatum, we injected CAV-cre into

the DMS or DLS of Ai9 tdTomato cre reporter mice (Figure 2A).

Viral spread in the striatum was contained within the targeted

subregion (Figure S2A). In the SNc, two key observations

were made. First, cre expression was largely restricted to DA

neurons, identified by tyrosine hydroxylase (TH) immunostaining,

following injection in both the DMS (344/360 neurons) and the

DLS (722/772 neurons). Although a few scattered TH+ cells

were labeled in the VTA in both groups, adjacent non-DAergic re-

gions did not contain cre-expressing cells (Figure 2B). Second,

the patterns of cre expression following DMS and DLS injection

differed. DMS-projecting neurons were observed in the medial

SNc, whereas DLS-projecting neurons were observed in the

lateral SNc. This differing anatomical localization of DMS- and

DLS-projecting SNc DA neurons supports the idea of parallel

nigrostriatal subcircuits within the SNc.

We next asked whether DMS- and DLS-projecting SNc DA

neurons projected only to the DMS or DLS, or instead collateral-

ized substantially. To visualize the boundaries of SNc DA neuron

axonal arborizations in the striatum, we injected CAV-cre into the

striatum and an adeno-associated virus (AAV) for cre-dependent

expression of membrane-bound GFP (mGFP) and synaptophy-

sin-mRuby (SYP-mRuby) into the SNc (Beier et al., 2015 [in this

issue ofCell]; Figure 2C). With this approach, axons were labeled

in green, and putative presynaptic sites were labeled in red. In

the SNc, both green and red labeling were visible as anticipated

(Figure 2D). The areas of highest fluorescence colocalized with

TH in a pattern concordant with the locations of DMS- and

DLS-projectingSNcDAneuronsobserved inAi9mice (Figure 2B).

In the striatum, axons of DMS-projecting SNc DA neurons

remained within the DMS, and axons of DLS-projecting SNc

DA neurons remained within the DLS (Figures 2E–2G and S2B).

To quantify this differential axon distribution, we acquired a

systematic image series from each mouse corresponding to

the full anterior-posterior span of the striatum (Figure S2B). Pro-

jection fraction was calculated as the axonal coverage area in

one region (DMS or DLS) divided by the total area covered

across the entire striatum (Figure 2H). The entire striatum was

defined to include the NAc, but we observed negligible contribu-

tions of projections to any of the NAc subregions across all con-

ditions (Figure S2C). NoGFP-labeled axonswere observed in the

prefrontal cortex or amygdala (Figure S2D). In dorsal striatum, an

overwhelming fraction of DMS-projecting axons was localized

within the bounds of the DMS (DMS 0.98 ± 0.01, n = 2, versus

DLS 0.02 ± 0.01, n = 2; two-way ANOVA, brain area3 projection

target interaction, p < 0.0001; post hoc Tukey’smultiple compar-

isons, p < 0.0001); similarly, DLS-projecting axons were local-

ized largely within the bounds of the DLS (DMS 0.14 ± 0.02,

n = 2, versus DLS 0.86 ± 0.02, n = 2; post hoc Tukey’s multiple

comparisons, p < 0.0001). These data confirmed that the SNc

SNr

SNc

PBP

IPN

PNVTA

THcre DLS-projecting

SNr

SNc

PBP

IPN

PNVTA

THcre DMS-projecting

DMS DLS0.0

0.5

1.0

Pro

ject

ion

Frac

tion

DMS-projectingDLS-projecting

********

****

****v

ctx

str

ctx

str

v

CAV-cre AAV

hSyn1 pA

mGFPSYP-mRuby T2A

CAV-cre Ai9 tdTomato cre reporter mouseA B

C

344/360 ≈ 96% TH+ 722/772 ≈ 94% TH+

D

G

DLS

-pro

ject

ing

DM

S-p

roje

ctin

g

mergeTHmGFP SYP-mRuby

mergeTHmGFP SYP-Ruby

M L

V

D

F

DLS

-pro

ject

ing

DM

S-p

roje

ctin

g

DLS

-pro

ject

ing

DM

S-p

roje

ctin

g

SYP-mRuby mGFP

DLS

DLSDMS

DMS

SYP-mRuby mGFP

SYP-mRuby mGFP

SYP-mRuby mGFP

E

M L

V

D

Figure 2. Medial-Lateral Distribution of SNc DA Neurons Projecting to Medial and Lateral Subregions of Dorsal Striatum

(A) CAV-cre was injected into the striatum of Ai9 td-Tomato cre reporter mice, labeling SNc neurons projecting to the injected region in red.

(B) DMS- and DLS-projecting SNc neurons are labeled with td-Tomato (red). Blue shows TH+ neurons. Both DMS- and DLS-projecting SNc neurons were found

to be predominantly DAergic. Scale bars are 0.5 mm.

(C) To map the outputs of DMS- and DLS-projecting SNc DA neurons, CAV-cre was injected into the striatum (DMS or DLS) of wild-type mice. Cre was then used

to direct the expression of mGFP-T2A-SYP-mRuby, labeling processes in green and presynaptic sites in red.

(D) DMS- and DLS-projecting SNc DA neurons expressing mGFP and SYP-mRuby. Staining for TH (blue) shows overlap with DA neuron cell bodies. Scale bars

are 0.5 mm.

(E) Example images (53) of patterns of mGFP expression in the striatal projections of DMS- or DLS-projecting DA neurons. Ctx, cortex; Str, striatum; V, ventricle.

(F) Example images (403) of regions of interest (DMS, DLS) in mice expressing mGFP and SYP-mRuby in DMS- and DLS-projecting SNc DA neurons. Scale bars

are 50 mm.

(G) Projection fraction of DMS- and DLS-projecting SNc DA neurons in DMS and DLS. Error bars are SEM. ****p < 0.0001.

See also Figure S2.

projections to the DMS and DLS are largely parallel, defining

separable nigrostriatal subcircuits with the potential to convey

distinct and independent signals to the DMS and DLS.

Distinct Properties of SNc DA Neurons Dependingupon Projection TargetWe next explored whether projection-defined SNc DA neurons

could be distinguished by intrinsic electrophysiological proper-

ties. To mark neurons for recording, we injected red-fluorescent

retrobeads into the DMS or DLS of TH-GFP mice. Retrobead-

labeled SNc neurons were also primarily GFP+ (Figure 3C;

509/515 DMS-projecting neurons and 526/542 DLS-projecting

neurons). Although the specificity of the TH-GFP line has been

questioned (Lammel et al., 2015), we found that GFP expression

is specific for TH+ neurons in the SNc (Figure S3). Retrobead in-

jections remained well localized to the injection site (Figure 3A)

and retrobead-containing neurons within the SNc were readily

identifiable for patching (Figure 3B).

Whole-cell patch-clamp analysis of retrobead+ andGFP+ SNc

neurons revealed similar membrane capacitance, resistance,

Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc. 639

DMS-projectingDLS-projecting

DMS-proj DLS-proj0

20

40

60

80

AP h

eigh

t (m

V)

0.0

0.5

1.0

102030

Freq

uenc

y (H

z)

DMS-projDLS-proj

mEPSCs mIPSCs mEPSCs mIPSCs

-40

-30

-20

-10

0

Ampl

itude

(pA)

DMS-projDLS-proj

DMS-proj DLS-proj0

20

40

60

80

100

Cm

(pF)

DMS-proj DLS-proj0

100

200

300

400

500

Rm

(M!

)

A

B

C

D E F

ML

I h (p

A)DMS-proj DLS-proj

0

200

400

600 **

DMS-proj DLS-proj0

100

200

300

400

500

I leak

(pA)

DLS

-pro

ject

ing

retrobeads TH-GFP merge

509/515 ≈ 99%

526/542 ≈ 97%

100 ms500

pA

mEPSCs

mIPSCs

0.5 s

30 p

A

THretrobeads

THretrobeads

retrobeadsDIC

ctx

str

ctx

str

retrobeads TH-GFP mergeM L

V

D

M LV

D

DM

S-p

roje

ctin

g

DMS-projectingDLS-projectingG H I

DMS-proj DLS-proj0.0

0.5

1.0

1.5

AP h

alf-w

idth

(ms)

*

J

K

1 ms

10 m

V

30 m

V/m

s

10 mV

**** ****

Figure 3. Intrinsic Properties of DMS-Projecting and DLS-Projecting SNc DA Neurons

(A) Retrobeads (red) injected into the striatum of TH-GFP mice. Ctx, cortex; Str, striatum. Scale bars are 0.5 mm.

(B) DIC and red fluorescent images of a patched retrobead-containing neuron. Dotted lines highlight the position of the recording electrode. Scale bars are 10 mm.

(C) Colocalization of retrobead-containing neurons (red) with TH-GFP-labeled neurons (green) in the SNc. Colocalization rates were �99% and �97% in DMS-

and DLS-projecting neurons, respectively.

(D) Membrane capacitance (Cm) of DMS- and DLS-projecting SNc DA neurons. Error bars are SEM.

(E) Membrane resistance (Rm) of DMS- and DLS-projecting SNc DA neurons. Error bars are SEM.

(F) Left, example responses of DMS-projecting SNc neurons (orange) and DLS-projecting SNc neurons (blue) to a hyperpolarizing current injection. Right,

Ih and Ileak current measurements in response to hyperpolarizing current injection from DMS- and DLS-projecting SNc DA neurons. Error bars are SEM.

**p < 0.01.

(G) Average action potential waveforms from DMS- and DLS-projecting SNc DA neurons. Area of light shading is SEM.

(H) Phase plots of average action potential waveforms from DMS- and DLS-projecting SNc DA neurons. Area of light shading is SEM.

(I) Average action potential height from DMS- and DLS-projecting SNc DA neurons. Error bars are SEM.

(J) Average action potential half widths from DMS- and DLS-projecting SNc DA neurons. Error bars are SEM. *p < 0.05.

(legend continued on next page)

640 Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc.

and leak currents of DMS-projecting and DLS-projecting DA

neurons (Figures 3D and 3E; Cm, DMS-proj 69.24 ± 3.589 pF,

n = 21 versus DLS-proj 71.94 ± 4.563 pF, n = 16, p = 0.64; Rm,

DMS-proj 384.7 ± 41.09 MU, n = 21 versus DLS-proj 327.4 ±

30.03 MU, n = 16, p = 0.30; Ileak, DMS-proj 317.1 ± 25.51 pA,

n = 21 versus DLS-proj 345.6 ± 35.81 pA, n = 16, p = 0.51). How-

ever, we identified distinctly larger Ih currents in DLS-projecting

DA neurons (Figure 3F; Ih, DMS-proj 296.2 ± 29.25 pA, n = 21

versus DLS-proj 460.8 ± 49.69 pA, n = 16, p < 0.01). All recorded

neurons displayed significant Ih currents and broad action

potential waveforms consistent with reliable identification of DA

neurons across groups. A small but significant difference was

observed in action potential half-width of DMS- versus DLS-pro-

jecting neurons (Figures 3G–3J; AP heights, DMS-proj 63.34 ±

1.901 mV, n = 21 versus DLS-proj 59.95 ± 2.395 mV, n = 16,

p = 0.27; AP half widths, DMS-proj 1.152 ± 0.06ms, n = 21 versus

DLS-proj 0.9938 ± 0.03 ms, n = 16, p < 0.05).

SNc DA Neurons Are Embedded within a LargelyInhibitory NetworkProjection-defined SNc DA neurons exhibit different intrinsic

properties, project to non-overlapping striatal subregions, and

receive differential inputs, together suggesting fundamentally

different roles in the circuit. However, since the nature of the

signals carried by the differing afferents remained unclear, we

next functionally investigated these afferents using slice electro-

physiology. As a first assessment of global functional afferent

input, we recorded miniature excitatory and inhibitory postsyn-

aptic currents (mEPSCs and mIPSCs) (Figure 3K). Although

no significant differences were observed between DMS- and

DLS-projecting SNc DA neurons, both the amplitudes and

frequencies of mIPSCs were higher than for mEPSCs across

both groups (Figures 3L and 3M; mEPSC amplitude, DMS-proj

�15.27 ± 0.50 pA, n = 15 versus DLS-proj �15.79 ± 0.68 pA,

n = 15; mIPSC amplitude, DMS-proj �29.69 ± 2.23 pA, n = 15

versus DLS-proj �26.71 ± 2.13 pA, n = 15; two-way ANOVA, ef-

fect of mEPSCs versus mIPSCs, p < 0.0001; mEPSC frequency

DMS-proj 0.49 ± 0.16 Hz, n = 15 versus DLS-proj 0.27 ± 0.10 Hz,

n = 15;mIPSC frequency, DMS-proj 7.11± 1.86Hz, n = 15 versus

DLS-proj 15.07 ± 4.29 pA, n = 15; two-way ANOVA, effect of

mEPSCs versus mIPSCs, p < 0.0001). This finding indicates

that SNc DA neurons in general are embedded within a largely

inhibitory network, in agreement with our findings and the find-

ings of Watabe-Uchida et al. (2012) that SNc DA neurons receive

afferents from many areas known to have GABAergic projection

neurons, including the striatum.

Functional Optogenetic Characterization of DorsalStriatal Inputs to DMS- versus DLS-ProjectingSNc DA NeuronsTo extend these findings, we employed targeted optogenetics to

test the functionality of synapses from dorsal striatum to SNc,

(K) Example traces of excitatory (top) and inhibitory (bottom) miniature postsyna

neurons.

(L) Amplitudes of mEPSCs and mIPSCs recorded from DMS-projecting and DLS

(M) Frequencies of mEPSCs and mIPSCs from recorded DMS-projecting and DL

See also Figure S3.

where we had specifically observed differences in inputs onto

projection-defined SNc neurons using TRIO (Figure 1F). First,

we expressed ChR2-eYFP in either DMS or DLS. Then, we in-

jected retrobeads into one of these regions to enable patching

of projection-defined SNc neurons. Blue light pulses stimulated

inputs specifically from the striatal subregion expressing ChR2

(Figure 4A). Examples of injection sites for ChR2 and retrobeads

are shown in Figure 4B, along with the resulting distributions of

red and green fluorescence in SNc (green fibers were observed

both in the SNc and in the underlying SNr, where many direct

pathway striatal neurons project).

Because striatal projection neurons are GABAergic, the gluta-

mate receptor antagonists NBQX (5 mM) and APV (50 mM) were

included in the extracellular solution to isolate inhibitory postsyn-

aptic currents (IPSCs), and a high-chloride internal solution was

used to facilitate event detection. We also included the voltage-

gated sodium channel antagonist TTX (1 mM) and potassium

channel antagonist 4-AP (100 mM) to isolate monosynaptic in-

puts by preventing disynaptic disinhibitory responses through

the GABAergic cells of the SNr (with TTX) while enabling ChR2

to drive neurotransmitter release in the absence of action poten-

tials (with 4-AP) (Petreanu et al., 2009). In all mice, recorded neu-

rons were identified that directly responded to blue light stimula-

tion, with response rates varying by condition. Connections from

DMS to DMS-projecting SNc DA neurons and connections from

DLS to DLS-projecting SNc DA neurons were clearly favored

(Figure 4C; X2 p < 0.0001), mirroring our TRIO results (Figure 1)

and functionally confirming a striking reciprocal connectivity

between dorsal striatal subregions and their DAergic inputs.

DLS Inputs to SNc DA Neurons Are Strongerthan DMS InputsBroad anatomical methods, while useful, cannot resolve key

aspects of function, such as distinguishing very strong from

very weak connections. In contrast, targeted optogenetic exper-

iments can provide information not only about connection

probabilities, but also about input strength. We next found, via

IPSC amplitude quantification for SNc neurons that responded

to stimulation of striatal afferents, that responses detected in

both DMS- and DLS-projecting SNc neurons were much larger

when DLS inputs were stimulated (Figure 4D; ChR2 DMS/

DMS-proj �205.27 ± 146.09 pA, n = 13; ChR2 DMS/DLS-proj

�96.23 ± 69.28 pA, n = 3; ChR2 DLS/DMS-proj �1193.10 ±

324.30 pA, n = 22; ChR2 DLS/DLS-proj �2737.96 ± 369.82 pA,

n = 38; two-way ANOVA, effect of ChR2 injection site, p <

0.01). Additionally, DLS-projecting SNc neurons had larger

responses to DLS stimulation than did DMS-projecting SNc

neurons (Tukey’s multiple comparisons, p < 0.05).

The observed differences in IPSC amplitudes between SNc

neurons receiving inputs from the DMS and DLS were not fully

explained by differences in the numbers of inputs arising from

the two striatal subregions (Figures 1 and 4C), suggesting

ptic currents from DMS-projecting (orange) and DLS-projecting (blue) SNc DA

-projecting SNc DA neurons. Error bars are SEM. ****p < 0.0001.

S-projecting SNc DA neurons. Error bars are SEM. ****p < 0.0001.

Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc. 641

ChR2-eYFPretrobeads

record

o-stim

ChR2 DMS ChR2 DLS0

20

40

60

80

% re

spon

ding

DMS-projectingDLS-projecting

38/49

22/39

3/28

13/32

A B

C D

ChR2 DMS ChR2 DLS

-4000

-3000

-2000

-1000

0

IPSC

am

plitu

de (p

A)

**

*

n.s.

Stri

atum

SN

cChR2 DMS

ChR2 DLS

20 ms500

pA

20 ms50 p

A

TTX4-AP

M LV

D

ChR2retrobeads

ChR2retrobeads

TH

50 ms50 p

A

ChR2 DMS

ChR2 DLS

ChR2 DMS ChR2 DLS

-50

-40

-30

-20

-10

0

qIPS

C a

mpl

itude

(pA)

DMS-projectingDLS-projecting

*

E

Figure 4. Optogenetic Mapping of Dorsal Striatal Inputs to DMS- and DLS-Projecting SNc DA Neurons: Reciprocal Connectivity Bias

and Divergent Strength

(A) ChR2-eYFP was expressed in either the DMS or the DLS. Red retrobeads labeled inputs to either the DMS or the DLS. Retrobead-containing cells in the SNc

were patched, and blue light was flashed to activate ChR2-expressing terminals nearby. Responses were recorded in the presence of TTX (1 mM) and 4-AP

(100 mM) to isolate monosynaptic inputs.

(B) Example images of striatal injection sites in the four experimental groups (top row) and of retrobead and ChR2 expression in the SNc (bottom row). Blue shows

TH immunostaining. Images are of slices used for recordings. Scale bars are 0.5 mm.

(C) Percentage of patched neurons responding to ChR2 stimulation of striatal terminals in the four experimental groups.

(D) IPSC amplitudes of responding neurons from (C). Example traces from each group are shown on the right. The blue bar indicates the time of blue light

stimulation. Error bars are SEM. **p < 0.01 and ****p < 0.0001.

(E) Quantification of the quantal IPSC (qIPSC) amplitude observed in DMS- and DLS-projecting SNc DA neurons in response to stimulation of DMS or DLS ChR2-

expressing terminals. Example traces of evoked asynchronous IPSCs are shown on the right, with the period of qIPSC event collection highlighted in a pop out

box. The blue bar indicates the time of stimulation. Error bars are SEM. *p < 0.05.

additional possible differences in quantal IPSC amplitude, the

number of synapses per input cell, and/or release probability.

To test for differences in quantal amplitude, we replaced calcium

642 Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc.

in the extracellular recording solution with strontium to induce

asynchronous release and found that the average quantal

IPSC amplitude was larger for inputs arising from the DLS versus

the DMS (Figure 4E; ChR2 DMS/DMS-proj �28.86 ± 3.48 pA,

n = 10; ChR2 DMS/DLS-proj �26.77 ± 2.75 pA, n = 8; ChR2

DLS/DMS-proj �36.64 ± 3.77 pA, n = 11; ChR2 DLS/DLS-proj

�40.60 ± 4.94 pA, n = 10, two-way ANOVA, effect of ChR2 injec-

tion site, p < 0.05), providing an additional partial explanation for

the observed differences in IPSC amplitude.

Selective Opposite-Valence Encoding of AversiveStimuli by DMS- versus DLS-Projecting SNc DA NeuronsParallel nigrostriatal SNc DA subcircuits clearly have the

potential to deliver different kinds of information to the DMS

and DLS. To formally test this intriguing possibility, we moni-

tored activity from projection-defined SNc DA neurons during

behavior using fiber photometry, a method for collecting popu-

lation intracellular [Ca2+] fluorescent signals from a genetically

encoded Ca2+ indicator such as GCaMP through a single

chronic fiber optic implant (Gunaydin et al., 2014). We ex-

pressed GCaMP6f (Chen et al., 2013) in DMS- or DLS-projec-

ting SNc DA neurons by injecting CAV-cre into the DMS

or the DLS, followed by injection of a cre-dependent GCaMP6f

construct into the SNc. Through a fiber optic implanted into

SNc at the GCaMP6f injection site (Figure 5A), we delivered

excitation light at 490 nm to stimulate GCaMP6f fluorescence

in a Ca2+-dependent manner and at 405 nm, an excitation

isosbestic wavelength for GCaMP6f that allowed us to perform

ratiometric measurements of GCaMP6f activity, thereby

correcting for bleaching and artifactual signal fluctuations

(Figures S4A–S4D).

We recorded the activity of DMS- or DLS-projecting SNc

DA neurons following either appetitive or aversive stimuli. To

record appetitive signals, we trained mice to lever press for

a sucrose reward retrieved from a reward port. Each press

had a 10% chance of delivering reward, a contingency that

allowed us to record during both rewarded and unrewarded

port entries within the same behavioral session for comparison.

DMS- and DLS-projecting SNc DA neurons reacted similarly

to rewarding outcomes (Figures 5C–5E); a similar peak DF/F

was observed on rewarded port entries for both sets of mice

(DMS-proj 2.164 ± 0.549%, n = 7 versus DLS-proj 1.753 ±

0.247%, n = 9, p = 0.47). Peaks were not observed for non-

rewarded port entries (Figures 5C, 5D, and 5F; DMS-proj

0.043 ± 0.195%, n = 7 versus DLS-proj �0.274 ± 0.083%,

n = 9, p = 0.13).

To record responses to aversive stimuli, we subjected the

same set of GCaMP6f-expressing mice to mild, unpredicted

electrical shocks. Strikingly, DMS- and DLS-projecting SNc

DA neurons responded oppositely to this aversive stimulus.

DMS-projecting neurons showed a marked dip in activity at the

time of shock, whereas DLS-projecting neurons showed an

increase (Figures 5G–5I; peak DF/F i.e., positive or negative

extreme during shock, DMS-proj�2.628 ± 0.513%, n = 8 versus

DLS-proj 1.527 ± 0.358%, n = 9, p < 0.0001). Additionally,

DLS-projecting neurons were returned immediately to baseline,

whereas more persistent changes were observed in DMS-pro-

jecting neurons (Figure 5J; mean DF/F between seconds 1 and

5, DMS-proj 2.403 ± 0.525%, n = 8 versus DLS-proj �0.3146 ±

0.176%, n = 9, p < 0.001). Post hoc histology revealed that

GCaMP6f-expressing axons were located in the DMS or DLS

as expected, that fiber optic implants were placed appropriately

in the SNc relative to GCaMP6f-expressing cell bodies, and that

GCaMP6f-expressing cell bodies were TH+ (Figure S4E). We

also verified that differences in movement between groups

during the reward and shock sessions could not account for

the observed differences in the photometry signal (Figures

S4F–S4K). Together, these results suggest that the projections

from subsets of SNc DA neurons to subregions of the dorsal

striatum are set up so that the DMS and DLS receive fundamen-

tally different signals; the DMS may receive a value signal

(DAergic population firing signals the valence of the outcome),

whereas the DLS may receive a salience signal (DAergic firing

alerts the system to adaptation-relevant events, whether appeti-

tive or aversive).

DISCUSSION

Here, we have developed and applied anatomical and functional

methodologies to provide a deeper understanding of striatoni-

grostriatal circuitry. We began by demonstrating that novel cir-

cuit-tracing and -recording methodologies can be effectively

combined with whole-brain CLARITY/COLM imaging (Chung

et al., 2013; Tomer et al., 2014) to provide a uniquely informative

overview of richly defined cell types and their global connectivity

motifs within the intact experimental subject brain. Further

development of high-throughput processing of these whole-

brain datasets will be needed (Chung and Deisseroth, 2013;

Kim et al., 2013, 2015) to fully capitalize on this opportunity for

the rapid advancement of understanding structure-function

relationships in the brain. For example, looking to the future, it

will be vital to increase experimental-subject group sizes despite

the immensely large datasets acquired for each experimental

subject. Although our current methods successfully captured

the large though complex connectivity differences observed

here, larger group sizes would broaden the potential to identify

biologically and potentially clinically important differences with

smaller effect sizes, including, perhaps, effects of age, gender,

and life experience.

Additionally, we have shown that, despite the utility of

anatomical tracing techniques such as TRIO, following up find-

ings from their use with other experimental modalities examining

the functionality of identified connections is essential. By stimu-

lating inputs from the striatum to SNc DA neurons optogeneti-

cally, we discovered that DLS inputs are an order of magnitude

stronger than DMS inputs, a finding that could not have been

predicted anatomically but is nevertheless likely to be crucial

in thinking about SNc circuit function. Our observations of

unequivocal inputs from the dorsal striatum to the SNc contrast

with certain earlier reports, which failed to find connections

using similar techniques (Chuhma et al., 2011; Xia et al., 2011).

Although it is often difficult to explain negative findings, these

previous studies used younger animals than did this study,

which is consistent with a possible role of age- or experience-

dependent plasticity; moreover, we have shown that striatal

projections to DA neurons may be very difficult to find if the

subregions of the striatum and the SNc are not well matched

(highlighting the value of the CLARITY/COLM/TRIO approach).

For example, if ChR2 expression were predominantly in DMS

Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc. 643

Photoreceiver

DichroicGFP

excitationfilter

Dichroic

Focusing lens

Realtime Modulation and Demodulation

GFP emission filter

405nm LED

490nm LED

405 bandpass filter

MultimodeOptical Fiber

Time (s)-5 0 5 10

ΔF

/F (

%)

-6

-4

-2

0

2

4

6

DMS-projecting

CAV-cre AAV

pA

GCaMP6f

fiber optic implant

A B

G H

Time (s)-5 0 5 10

ΔF

/F (

%)

-2

-1

0

1

2

3

4

DLS-projecting

****

Time (s)-5 0 5 10

ΔF

/F (

%)

-2

0

2

4

6rewardno reward

DMS-projectingC D

Time (s)-5 0 5 10

ΔF

/F (

%)

-2

0

2

4

6rewardno reward

DLS-projecting

E F

I J ***

Figure 5. DMS- and DLS-Projecting SNc DA

Neurons Respond Similarly to Appetitive but

Differentially to Aversive Stimulation

(A) CAV-cre was injected into either the DMS or

DLS. An AAV expressing cre-dependent GCaMP6f

was injected into the SNc. A 400 mm fiber optic

implant was placed in the SNc.

(B) Fiber photometry rig schematic. See Experi-

mental Procedures for amore detailed description.

(C) Example of responses to reward observed in a

mouse expressing GCaMP6f in DMS-projecting

SNc DA neurons. Time 0 is aligned to either re-

warded or non-rewarded port entries in an operant

chamber. Arrow is placed at time 0. Area of light

shading is SEM.

(D) Example of responses to reward observed in a

mouse expressing GCaMP6f in DLS-projecting

SNc DA neurons.

(E) Quantification of the peak DF/F observed in

response to reward in mice expressing GCaMP6f

in DMS-projecting SNc DA neurons or in mice

expressing GCaMP6f in DLS-projecting SNc DA

neurons.

(F) Quantification of the DF/F observed at time 0 in

response to a non-rewarded port entry in mice

expressing GCaMP6f in DMS-projecting SNc

DA neurons or in mice expressing GCaMP6f in

DLS-projecting SNc DA neurons.

(G) Example of responses to shock observed in

a mouse expressing GCaMP6f in DMS-projec-

ting SNc DA neurons. Yellow bar indicates the

time of the shock (0–0.5 s). Area of light shading

is SEM.

(H) Example of responses to shock observed in a

mouse expressing GCaMP6f in DLS-projecting

SNc DA neurons.

(I) Quantification of the peak (extreme, whether

negative or positive) DF/F observed during the

shock in mice expressing GCaMP6f in DMS-pro-

jecting SNc DA neurons or in mice expressing

GCaMP6f in DLS-projecting SNc DA neurons.

****p < 0.0001.

(J) Quantification of the mean DF/F observed 1–5 s

post-shock in mice expressing GCaMP6f in DMS-

projecting SNc DA neurons or in mice expressing

GCaMP6f in DLS-projecting SNc DA neurons.

***p < 0.001. The same mice are shown in (C)–(J).

See also Figure S4.

and laterally located (DLS-projecting) DA neurons were being

patched, we would expect connections to be extremely sparse,

small and difficult to detect.

644 Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc.

Our result regarding the in vivo activity

of DMS- and DLS-projecting SNc DA

neurons is concordant with the hypothe-

sis put forward by Matsumoto and Hiko-

saka (2009) that DA neurons located

more ventrolaterally within the midbrain

increase firing in response to aversive

stimuli. By targeting SNc DA neurons

for observation based on their striatal

projection target, we directly tested and

substantially extended this hypothesis, demonstrating that the

differences in the representation of aversive stimuli among sub-

sets of SNc DA neurons relate to projection target. Additionally,

we observed a sustained activity increase in DMS-projecting

SNc DA neurons following shock. We speculate that this popula-

tion of DA neurons may also encode for a heightened sensitivity

or motivational state following an unfamiliar outcome (unlike the

reward stimulus, the mice had no previous experience of shock

prior to test day) in order to promote future action-outcome

learning.

Regarding anatomy of efferents alone, our finding that the

projections from SNc DA neurons to the DMS and DLS arise

from distinct locations within the SNc is not inconsistent

with several previous findings. For example, in rats that had

the DA neuron toxin 6-OHDA injected into the DLS, degenera-

tion of DA neurons was observed primarily in the lateral SNc

(Faure et al., 2005). Similarly, for mice in which DA neuron

TH expression was restricted to neurons projecting to either

the DLS or the ventromedial striatum (VMS), DLS-projecting

neurons were located more laterally than VMS-projecting neu-

rons (Darvas and Palmiter, 2009; 2010; see also Schiemann

et al., 2012).

Information transmitted along these separable activity

streams will depend on intrinsic properties of the DA neurons

and on their afferents. Indeed, combined with other recent lines

of evidence (Lammel et al., 2008, 2011; Margolis et al., 2008),

we have found that projection-defined subpopulations of DA

neurons display different intrinsic properties. For example,

increasing-amplitude Ih currents are observed progressing

medially to laterally within the midbrain DA system; DA neurons

in the medial paranigral nucleus of the VTA express little to no Ihcurrent (Lammel et al., 2011), whereas DA neurons in the lateral

SNc express the largest Ih currents. Differences in Ih currents are

strongly correlated with action potential rebound delays and can

influence pacemaking activity (Neuhoff et al., 2002), particularly

in the case of calbindin-negative SNc DA neurons (which consti-

tute the majority of the SNc DA neuron population).

Afferents to projection-defined SNc DA neurons also differ; in

particular, we here identify new principles of afferent-projection

complexity in SNc subcircuits (see Beier et al., 2015 for addi-

tional illuminating information regarding VTA subcircuits). A

prominent finding is that the DMS and DLS are preferentially

reciprocally connected with the very same DA neurons that

project back to these areas. This information notably enriches

the ascending spiral model proposed by Haber et al. (2000),

which could not make definitive statements about striatal inputs

onto DAergic versus non-DAergic midbrain cells. We also

observed strong DLS projections to DMS-projecting DA neu-

rons; this finding implies a novel route of lateral to medial infor-

mation flow.

In summary, our identification of two distinct nigrostriatal

DA circuits—differing in inputs, outputs, biophysical properties,

and environmental information representations—both reveals

independently controlled information representations streaming

through SNc and provides a generalizable framework for

brain-wide mapping of diverse populations of neurons defined

by multiple independent types of features. Particularly in the

case of DA neurons, this type of approach may improve our un-

derstanding of the circuit mechanisms underlying normal brain

function, as well as diseases such as depression, addiction,

and schizophrenia.

EXPERIMENTAL PROCEDURES

Animals

All experiments were approved by the Stanford University IACUC committee,

protocol 10747. All mice (Ai9, TH-GFP, D1-tdTomato, and wild-type) were on a

C57BL6/6J background and were 2–4 months old.

Stereotaxic Injections

Viruses and retrobeads were injected and optical implants placed at the

following coordinates, relative to bregma: DMS +0.75 AP, 1.5 ML, �2.8 DV;

DLS +0.25 AP, 2.5 ML, �3.4 DV; medial SNc �3.1, 0.8 ML, �4.7 DV; lateral

SNc �3.1, 1.3 ML, �4.2 DV.

Histology

TH staining was done with 1:500 primary antibody overnight at 4�C and 1:500

secondary antibody coupled to Alexa Fluor 647 for 2–3 hr at room temperature.

GFP staining was done with 1:500 primary antibody coupled directly to either

Alexa Fluor 488 or Alexa Fluor 647 overnight at 4�C. Counterstaining was done

with Neurotrace 435/455 Blue Fluorescent Nissl Stain (1:500) and/or DAPI

(300 nM).

CLARITY

Brains were perfused and incubated with CLARITYmonomer solution contain-

ing 1% acrylamide, 0.0125% bis-acrylamide, and 4% PFA and then polymer-

ized at 37�C for 6–7 hr. Brains were passively cleared in SDS Borate Buffer

(pH 8.5) at 37�C for 4–5 weeks, equilibrated in Focusclear for imaging and

imaged using COLM methods (Tomer et al., 2014).

Slice Electrophysiology

300 mm coronal sections were prepared in an NMDG-based solution at room

temperature. Striatal slices were fixed in 4% PFA and saved for verification

of injection sites. Whole-cell recordings were performed in standard aCSF at

30–32�C. Where indicated, TTX (1 mm), 4-AP (100 mM), NBQX (5 mM), APV

(50 mM), and picrotoxin (50 mM) were added. In some experiments, extracel-

lular Ca2+ was replaced with Sr2+ to induce asynchronous release. K-gluco-

nate internal was used for recording action potentials and Ih currents. EPSCs

were recorded using CeMeSO3 internal and IPSCs were recorded using high

chloride CsCl internal. 5 ms blue light pulses (475 nm, �10 mW/mm2) were

used to stimulate ChR2.

Reward Behavior

Mice were trained to lever press for 20% sucrose reward on an RR10

schedule, earning a maximum of 40 25 ml rewards in a 1 hr session.

Shock Behavior

Mice received 15 mild foot shocks (0.4 mA, 0.5 s) on an RI60 schedule.

Fiber Photometry

A 490 nm LED was sinusoidally modulated at 211 Hz and passed through a

GFP excitation filter. A 405 nm LED was modulated at 531 Hz and passed

through a 405 nm bandpass filter. Both light streams were coupled to a high

NA (0.48), large core (400 mm) optical fiber patch cord, which was mated

to a matching brain implant in each mouse. GCaMP6f fluorescence was

collected by the same fiber, passed through aGFP emission filter, and focused

onto a photoreceiver. Custom software running on a real-time signal processor

controlled the LEDs and independently demodulated the fluorescence bright-

ness due to 405 nm and 490 nm excitation. The timing of behavioral variables

was recorded by the same system. To calculate DF/F, a least-squares linear fit

was applied to the 405 nm signal to align it to the 490 nm signal, producing a

fitted 405 nm signal that was used to normalize the 490 nm as follows: DF/F =

(490 nm signal � fitted 405 nm signal)/fitted 405 nm signal.

Statistics

Unpaired t tests were used for comparisons between two groups (DMS- and

DLS-projecting SNc DA neurons). Two-way ANOVAs were used to assess

how the properties or responses of DMS- and DLS-projecting SNcDA neurons

were affected by other factors (e.g., input area). When a statistically significant

Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc. 645

effect was observed using a two-way ANOVA, post hoc testing with correction

for multiple comparisons was performed using Tukey’s or Sidak’s multiple

comparisons test.

SUPPLEMENTAL INFORMATION

Supplemental Information includes Supplemental Experimental Procedures,

four figures, one table, and fivemovies and can be found with this article online

at http://dx.doi.org/10.1016/j.cell.2015.07.014.

AUTHOR CONTRIBUTIONS

T.N.L. and K.D. designed experiments, with input from L.L. and K.T.B on viral

tracing and with input from R.T. on CLARITY and COLM implementation. L.L.

and K.T.B. provided unpublished TRIO procedures. T.N.L. performed and

analyzed all experiments with contributions from K.T.B. on viral tracing, contri-

butions from R.T. and A.K.C. on whole-brain COLM imaging experiments and

related image analysis, contributions from K.E.E. on image quantification, and

contributions from C.S., T.J.D., and K.A.Z. on fiber photometry methods

development, implementation, and combination with freely moving behavior.

T.N.L. and K.D. wrote the paper with editorial input from all authors. K.D. su-

pervised all aspects of the work.

ACKNOWLEDGMENTS

We thank L. Ye and K. Engberg for advice on CLARITY; C. Ramakrishnan, S.

Pak, A. Shi On Hong, A. Wang, A. Lei, and H. Swanson for assistance with

animal husbandry and reagent production and acquisition; P. Rothwell and

L. Steinberg for advice and sharing of transgenic mouse lines; E. Callaway

for advice on rabies tracing; and the entire K.D. and L.L. labs for invaluable dis-

cussion. We also thank E. Kremer (IGMM) for supplying CAV-cre, M. Kay

(Stanford) for providing the pRC-DJ plasmid used to produce AAVDJ, the

Stanford Viral and Vector Core for packaging AAVDJ, and D. Kim for

supplying pGP-CMV-GCaMP6f (Addgene plasmid #40755), which was re-

cloned in our lab by C. Ramakrishnan. We thank the UNC vector core for sup-

plying the AAV5 and AAV8 vectors used in these studies and the Salk Vector

Core for supplying rabies (amplified in-house by the L.L. lab). T.N.L. was sup-

ported by a Stanford Dean’s Postdoctoral Fellowship and by an NRSA

Postdoctoral Fellowship (1F32MH105053-01). K.A.Z. was supported by an

NRSA Predoctoral Fellowship (1F31MH105151-01). This work was also

supported by the Hughes Collaborative Innovation Award (HCIA) and the

NIMH Silvio Conte Center at Stanford (P50 MH086403 to R.C.M.). K.D. is sup-

ported by the DARPA Neuro-FAST program, NIMH, NIDA, NSF, the Simons

Foundation, the Gatsby Foundation, the Wiegers Family Fund, the Nancy

and James Grosfeld Foundation, the H.L. Snyder Medical Foundation, the

Samuel and Betsy Reeves Fund, the Vincent VC Woo Fund, and the

Albert Yu and Mary Bechman Foundations. All optogenetics (http://www.

optogenetics.org), CLARITY (http://clarityresourcecenter.org), and COLM

(http://clarityresourcecenter.com/COLM.html) reagents and protocols are

distributed and supported freely.

Received: May 11, 2015

Revised: June 26, 2015

Accepted: July 8, 2015

Published: July 30, 2015

REFERENCES

Beier, K.T., Steinberg, E.E., DeLoach, K.E., Xie, S., Miyamichi, K., Schwarz, L.,

Gao, X.J., Kremer, E.J., Malenka, R.C., and Luo, L. (2015). Circuit architecture

of VTA dopamine neurons revealed by systematic input-output mapping. Cell

162, this issue, 622–634.

Bocklisch, C., Pascoli, V., Wong, J.C.Y., House, D.R.C., Yvon, C., de Roo, M.,

Tan, K.R., and Luscher, C. (2013). Cocaine disinhibits dopamine neurons by

potentiation of GABA transmission in the ventral tegmental area. Science

341, 1521–1525.

646 Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc.

Brischoux, F., Chakraborty, S., Brierley, D.I., and Ungless, M.A. (2009). Phasic

excitation of dopamine neurons in ventral VTA by noxious stimuli. Proc. Natl.

Acad. Sci. USA 106, 4894–4899.

Bromberg-Martin, E.S., Matsumoto, M., and Hikosaka, O. (2010). Dopamine in

motivational control: rewarding, aversive, and alerting. Neuron 68, 815–834.

Castane, A., Theobald, D.E.H., and Robbins, T.W. (2010). Selective lesions of

the dorsomedial striatum impair serial spatial reversal learning in rats. Behav.

Brain Res. 210, 74–83.

Chen, T.-W., Wardill, T.J., Sun, Y., Pulver, S.R., Renninger, S.L., Baohan, A.,

Schreiter, E.R., Kerr, R.A., Orger, M.B., Jayaraman, V., et al. (2013). Ultrasen-

sitive fluorescent proteins for imaging neuronal activity. Nature 499, 295–300.

Chuhma, N., Tanaka, K.F., Hen, R., and Rayport, S. (2011). Functional connec-

tome of the striatal medium spiny neuron. J. Neurosci. 31, 1183–1192.

Chung, K., and Deisseroth, K. (2013). CLARITY for mapping the nervous

system. Nat. Methods 10, 508–513.

Chung, K., Wallace, J., Kim, S.-Y., Kalyanasundaram, S., Andalman, A.S.,

Davidson, T.J., Mirzabekov, J.J., Zalocusky, K.A., Mattis, J., Denisin, A.K.,

et al. (2013). Structural and molecular interrogation of intact biological sys-

tems. Nature 497, 332–337.

Cohen, J.Y., Haesler, S., Vong, L., Lowell, B.B., and Uchida, N. (2012). Neuron-

type-specific signals for reward and punishment in the ventral tegmental area.

Nature 482, 85–88.

Crittenden, J.R., and Graybiel, A.M. (2011). Basal Ganglia disorders associ-

atedwith imbalances in the striatal striosome andmatrix compartments. Front.

Neuroanat. 5, 59.

Darvas, M., and Palmiter, R.D. (2009). Restriction of dopamine signaling to the

dorsolateral striatum is sufficient for many cognitive behaviors. Proc. Natl.

Acad. Sci. USA 106, 14664–14669.

Darvas, M., and Palmiter, R.D. (2010). Restricting dopaminergic signaling to

either dorsolateral or medial striatum facilitates cognition. J. Neurosci. 30,

1158–1165.

Faure, A., Haberland, U., Conde, F., and El Massioui, N. (2005). Lesion to the

nigrostriatal dopamine system disrupts stimulus-response habit formation.

J. Neurosci. 25, 2771–2780.

Featherstone, R.E., and McDonald, R.J. (2004). Dorsal striatum and stimulus-

response learning: lesions of the dorsolateral, but not dorsomedial, striatum

impair acquisition of a stimulus-response-based instrumental discrimination

task, while sparing conditioned place preference learning. Neuroscience

124, 23–31.

Fiorillo, C.D. (2013). Two dimensions of value: dopamine neurons represent

reward but not aversiveness. Science 341, 546–549.

Fiorillo, C.D., Yun, S.R., and Song, M.R. (2013). Diversity and homogeneity in

responses of midbrain dopamine neurons. J. Neurosci. 33, 4693–4709.

Gunaydin, L.A., Grosenick, L., Finkelstein, J.C., Kauvar, I.V., Fenno, L.E., Ad-

hikari, A., Lammel, S., Mirzabekov, J.J., Airan, R.D., Zalocusky, K.A., et al.

(2014). Natural neural projection dynamics underlying social behavior. Cell

157, 1535–1551.

Haber, S.N., Fudge, J.L., andMcFarland, N.R. (2000). Striatonigrostriatal path-

ways in primates form an ascending spiral from the shell to the dorsolateral

striatum. J. Neurosci. 20, 2369–2382.

Hnasko, T.S., Perez, F.A., Scouras, A.D., Stoll, E.A., Gale, S.D., Luquet, S.,

Phillips, P.E.M., Kremer, E.J., and Palmiter, R.D. (2006). Cre recombinase-

mediated restoration of nigrostriatal dopamine in dopamine-deficient mice

reverses hypophagia and bradykinesia. Proc. Natl. Acad. Sci. USA 103,

8858–8863.

Ilango, A., Kesner, A.J., Keller, K.L., Stuber, G.D., Bonci, A., and Ikemoto, S.

(2014). Similar roles of substantia nigra and ventral tegmental dopamine

neurons in reward and aversion. J. Neurosci. 34, 817–822.

Kim, S.-Y., Chung, K., and Deisseroth, K. (2013). Light microscopymapping of

connections in the intact brain. Trends Cogn. Sci. 17, 596–599.

Kim, Y., Venkataraju, K.U., Pradhan, K., Mende, C., Taranda, J., Turaga, S.C.,

Arganda-Carreras, I., Ng, L., Hawrylycz, M.J., Rockland, K.S., et al. (2015).

Mapping social behavior-induced brain activation at cellular resolution in the

mouse. Cell Rep. 10, 292–305.

Lammel, S., Hetzel, A., Hackel, O., Jones, I., Liss, B., and Roeper, J. (2008).

Unique properties of mesoprefrontal neurons within a dual mesocorticolimbic

dopamine system. Neuron 57, 760–773.

Lammel, S., Ion, D.I., Roeper, J., andMalenka, R.C. (2011). Projection-specific

modulation of dopamine neuron synapses by aversive and rewarding stimuli.

Neuron 70, 855–862.

Lammel, S., Lim, B.K., Ran, C., Huang, K.W., Betley, M.J., Tye, K.M., Deisser-

oth, K., and Malenka, R.C. (2012). Input-specific control of reward and aver-

sion in the ventral tegmental area. Nature 491, 212–217.

Lammel, S., Lim, B.K., and Malenka, R.C. (2014). Reward and aversion in a

heterogeneous midbrain dopamine system. Neuropharmacology 76 Pt B,

351–359.

Lammel, S., Steinberg, E.E., Foldy, C., Wall, N.R., Beier, K., Luo, L., and Mal-

enka, R.C. (2015). Diversity of transgenic mousemodels for selective targeting

of midbrain dopamine neurons. Neuron 85, 429–438.

Margolis, E.B., Mitchell, J.M., Ishikawa, J., Hjelmstad, G.O., and Fields, H.L.

(2008). Midbrain dopamine neurons: projection target determines action

potential duration and dopamine D(2) receptor inhibition. J. Neurosci. 28,

8908–8913.

Matsumoto, M., and Hikosaka, O. (2009). Two types of dopamine neuron

distinctly convey positive and negative motivational signals. Nature 459,

837–841.

Mirenowicz, J., and Schultz, W. (1996). Preferential activation of midbrain

dopamine neurons by appetitive rather than aversive stimuli. Nature 379,

449–451.

Miyamichi, K., Shlomai-Fuchs, Y., Shu, M., Weissbourd, B.C., Luo, L., and

Mizrahi, A. (2013). Dissecting local circuits: parvalbumin interneurons underlie

broad feedback control of olfactory bulb output. Neuron 80, 1232–1245.

Montague, P.R., Dayan, P., and Sejnowski, T.J. (1996). A framework for

mesencephalic dopamine systems based on predictive Hebbian learning.

J. Neurosci. 16, 1936–1947.

Montague, P.R., Hyman, S.E., and Cohen, J.D. (2004). Computational roles for

dopamine in behavioural control. Nature 431, 760–767.

Neuhoff, H., Neu, A., Liss, B., and Roeper, J. (2002). I(h) channels contribute to

the different functional properties of identified dopaminergic subpopulations in

the midbrain. J. Neurosci. 22, 1290–1302.

Petreanu, L., Mao, T., Sternson, S.M., and Svoboda, K. (2009). The subcellular

organization of neocortical excitatory connections. Nature 457, 1142–1145.

Roeper, J. (2013). Dissecting the diversity of midbrain dopamine neurons.

Trends Neurosci. 36, 336–342.

Rossi, M.A., Sukharnikova, T., Hayrapetyan, V.Y., Yang, L., and Yin, H.H.

(2013). Operant self-stimulation of dopamine neurons in the substantia nigra.

PLoS ONE 8, e65799.

Schiemann, J., Schlaudraff, F., Klose, V., Bingmer, M., Seino, S., Magill, P.J.,

Zaghloul, K.A., Schneider, G., Liss, B., and Roeper, J. (2012). K-ATP channels

in dopamine substantia nigra neurons control bursting and novelty-induced

exploration. Nat. Neurosci. 15, 1272–1280.

Schultz, W., Dayan, P., and Montague, P.R. (1997). A neural substrate of pre-

diction and reward. Science 275, 1593–1599.

Schwarz, L.A., Miyamichi, K., Gao, X.J., Beier, K.T., Weissbourd, B., DeLoach,

K.E., Ren, J., Ibanes, S., Malenka, R.C., Kremer, E.J., and Luo, L. (2015).

Viral-genetic tracing of the input-output organization of a central noradrenaline

circuit. Nature. http://dx.doi.org/10.1038/nature14600.

Soudais, C., Laplace-Builhe, C., Kissa, K., and Kremer, E.J. (2001). Preferential

transduction of neurons by canine adenovirus vectors and their efficient retro-

grade transport in vivo. FASEB J. 15, 2283–2285.

Steinberg, E.E., Keiflin, R., Boivin, J.R., Witten, I.B., Deisseroth, K., and Janak,

P.H. (2013). A causal link between prediction errors, dopamine neurons and

learning. Nat. Neurosci. 16, 966–973.

Sutton, R.S. (1988). Learning to predict by the methods of temporal differ-

ences. Mach. Learn. 3, 9–44.

Tomer, R., Ye, L., Hsueh, B., and Deisseroth, K. (2014). Advanced CLARITY for

rapid and high-resolution imaging of intact tissues. Nat. Protoc. 9, 1682–1697.

Ungless, M.A., Magill, P.J., and Bolam, J.P. (2004). Uniform inhibition of dopa-

mine neurons in the ventral tegmental area by aversive stimuli. Science 303,

2040–2042.

Waelti, P., Dickinson, A., and Schultz, W. (2001). Dopamine responses comply

with basic assumptions of formal learning theory. Nature 412, 43–48.

Watabe-Uchida, M., Zhu, L., Ogawa, S.K., Vamanrao, A., and Uchida, N.

(2012). Whole-brain mapping of direct inputs to midbrain dopamine neurons.

Neuron 74, 858–873.

Wickersham, I.R., Lyon, D.C., Barnard, R.J.O., Mori, T., Finke, S., Conzel-

mann, K.-K., Young, J.A.T., and Callaway, E.M. (2007). Monosynaptic restric-

tion of transsynaptic tracing from single, genetically targeted neurons. Neuron

53, 639–647.

Xia, Y., Driscoll, J.R., Wilbrecht, L., Margolis, E.B., Fields, H.L., and Hjelmstad,

G.O. (2011). Nucleus accumbens medium spiny neurons target non-dopami-

nergic neurons in the ventral tegmental area. J. Neurosci. 31, 7811–7816.

Yin, H.H., and Knowlton, B.J. (2004). Contributions of striatal subregions to

place and response learning. Learn. Mem. 11, 459–463.

Yin, H.H., Knowlton, B.J., and Balleine, B.W. (2004). Lesions of dorsolateral

striatum preserve outcome expectancy but disrupt habit formation in instru-

mental learning. Eur. J. Neurosci. 19, 181–189.

Yin, H.H., Knowlton, B.J., and Balleine, B.W. (2005a). Blockade of NMDA

receptors in the dorsomedial striatum prevents action-outcome learning in

instrumental conditioning. Eur. J. Neurosci. 22, 505–512.

Yin, H.H., Ostlund, S.B., Knowlton, B.J., and Balleine, B.W. (2005b). The role of

the dorsomedial striatum in instrumental conditioning. Eur. J. Neurosci. 22,

513–523.

Yin, H.H., Mulcare, S.P., Hilario, M.R.F., Clouse, E., Holloway, T., Davis, M.I.,

Hansson, A.C., Lovinger, D.M., and Costa, R.M. (2009). Dynamic reorganiza-

tion of striatal circuits during the acquisition and consolidation of a skill. Nat.

Neurosci. 12, 333–341.

Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc. 647

Supplemental Figures

DMSDLS

M1/2 S1

Cortex (

Other) NAc

Dorsal S

triatu

m

FS (IPAC)

GPeSI/V

PBNST

CeA

Hypothala

mus

Thalamus

LHb SC ICSNc

SNrVTA

PAG

Dorsal R

aphe

MRNPPN PB

0

10

20

30

40

% c

ontra

late

ral i

nput

s DLS-projecting***DMS-projecting

ΔRG-GFP only

THRVdG-GFP

No CAV-creA B

THRVdG-GFP RVdG-GFP

DM

S-p

roje

ctin

gD

LS-p

roje

ctin

g

GP

GP

E

F

G

RVdG-GFP D1-tmt merge DMS-projecting 92/97 ≈ 95% DLS-projecting 210/214 ≈ 98%

D1-tmt mergeRVdG-GFP

D

C

DM

S-p

roje

ctin

gD

LS-p

roje

ctin

g M1

M1

LHb

LHb

HY

HY

PAG

PAG

MRNVTASNr

SNc

VTASNr

SNc

MRN

Figure S1. Proper EnvA-Pseudotyping of RVdG-GFP and Cre Dependency of Long-Range Tracing of Inputs, Divisions of the DMS and DLS

Subregions of Dorsal Striatum, Contralateral Inputs to DMS- and DLS-Projecting SNc Dopamine Neurons, and Dopamine D1 Receptor-

Expressing Identity of Striatonigral Inputs Identified by TRIO, Related to Figure 1

(A) Coronal slice showing the SNc of a mouse injected with only the RVdG-GFP virus. TH staining for dopamine neurons is shown in blue. The lack of GFP+ cells

indicates that the RVdG-GFP virus was not able to infect cells in the absence of TC expression.

(B) Coronal slices showing the SNc and striatum of a mouse injected with all the TRIO viruses except for CAV-cre. Left, some GFP+ cells are observed locally in

the SNc, indicating leak of the TC virus. This local background precludes conclusions about local connectivity in TRIO, but does not prohibit conclusions about

long-range connectivity. TH staining for dopamine neurons is shown in blue. Right, no GFP+ neurons are observed in the striatum (or any other structures that

provide long-range inputs to SNc dopamine neurons), indicating that long-range TRIO tracing of inputs here was indeed dependent on CAV-cre injections in the

DMS or DLS.

(C) Example coronal atlas images showing the divisions between DMS (yellow) and DLS (magenta).

(D) Example images showing coronal sections of the dorsal striatum at three distinct locations along the anterior-posterior axis (locations similar to the atlas

images shown in A).

(E) Example images of RVdG-GFP labeling of contralateral inputs to DMS- and DLS-projecting SNc dopamine neurons from various brain regions. M1, motor

cortex. HY, hypothalamus. LHb, lateral habenula. VTA, ventral tegmental area. SNc, substantia nigra pars compacta. SNr, substantia nigra pars reticulata. PAG,

periaqueductal gray. MRN, midbrain reticular nucleus.

(F) Whole-brain quantification of contralaterally labeled inputs to DMS- and DLS-projecting SNc dopamine neurons, expressed as a percentage of all contra-

lateral inputs. Other abbreviations: motor cortex (M1/2), somatosensory cortex (S1), fundus of the striatum/ interstitial nucleus of the posterior limb of the anterior

(legend continued on next page)

Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc. S1

commissure (FS/IPAC), globus pallidus external segment (GPe), substantia innominata/ventral pallidum (SI/VP), bed nucleus of the stria terminalis (BNST), central

amygdala (CeA), superior colliculus (SC), inferior colliculus (IC), pedunculopontine nucleus (PPN), parabrachial nucleus (PB).

(G) Images showing strong overlap of RVdG-GFP labeling in the dorsal striatumwith expression of the D1 dopamine receptor (labeled in a D1-tmt BAC transgenic

mouse). Scale bars are 50 mm.

S2 Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc.

CAV-cre DMS CAV-cre DLS

str

V

D

ctxctx

str

M LV

D

A

BDMS-projecting DLS-projecting

anterior

posterior

M LV

DM L

V

D

C

D

NAc core

NAc med

shell

NAc lat

shell

DMSDLS

0.0000

0.0005

0.0010

0.0015

0.2

0.4

0.6

0.8

1.0

Pro

ject

ion

Frac

tion

DMS-projectingDLS-projecting

DMS-projecting DLS-projecting

PFC

amyg

dala

Figure S2. CAV-cre Injections Contained within Distinct Dorsal Striatal Subregions and Anterior-Posterior Extent and Specificity of Stria-

tonigral Projections to DMS and DLS, Related to Figure 2

(A) CAV-cre was injected into striatum of Ai9 td-Tomato cre reporter mice. Left, CAV-cre injected into DMS. Right, CAV-cre injected into DLS. Dotted white line

highlights the corpus callosum. Ctx, Cortex. Str, Striatum.

(B) Example images of patterns of mGFP expression in the striatal projections of DMS- or DLS-projecting dopamine neurons, showing the full range of slices from

most anterior to most posterior. Green images show the original mGFP expression pattern. Black and white images to the right of each of the green images show

the analyzed image after background subtraction and thresholding.

(C) Projection fraction of DMS- and DLS-projecting SNc dopamine neurons in NAc core, NAc medial shell, NAc lateral shell, DMS and DLS. Error bars are SEM.

Data for the DMS and DLS are the same as shown in Figure 2G. This graph simply displays the full results, which included quantifications of the projection

fractions in NAc subregions.

(D) Histology sections showing a lack of mGFP-labeled axons from DMS- or DLS-projecting dopamine neurons in the PFC and amygdala. No mGFP expression

was observed despite GFP immunostaining to enhance the signal. DAPI counterstaining (light blue) shows the location of the tissue. Scale bars are 0.5 mm.

Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc. S3

GFP+ TH+GFP+ TH-

lateral SNc

95.4%

medial SNc

91.8%

lateral VTA

91.6%

midline VTA

37.3%

B lateral SNc

A

C

VTA

SN

c

TH Ab

TH Ab

TH-GFP

TH-GFP

medial SNc lateral VTA midline VTA

TH A

bTH

-GFP

mer

ge

Figure S3. Specificity of the TH-GFP Mouse Line for TH Immunopositive Neurons in the SNc and Lateral VTA, but Not the Midline VTA,

Related to Figure 3

(A) Low magnification (10x) images showing GFP expression (TH-GFP, green) and TH immunostaining (TH Ab, blue) in the midbrain (SNc and VTA) of TH-GFP

mice. Scale bars are 0.5 mm.

(B) High magnification (40x) images showing TH-GFP expression (TH-GFP, green) and TH immunostaining (TH Ab, blue) in four regions of the midbrain – lateral

SNc, medial SNc, lateral VTA, and midline VTA – which were quantified separately in C. Scale bars are 50 mm.

(C) Quantification of the percentage of GFP+ neurons that were also TH+ in the midbrains of TH-GFP mice (n = 3 mice). Pooled VTA data give a specificity of

71.8%, comparable to the specificity of 69% reported by Lammel et al. (2015).

S4 Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc.

-1 0 1 2 3 4 520

25

30

35

40

45

Peak During Reward

Rew

ards

Ear

ned

R2 = 0.08p = 0.53

-4 -3 -2 -10

5

10

15

20

Peak During Shock

% T

ime

Free

zing

R2 = 0.005p = 0.93

-4 -3 -2 -10.0

0.5

1.0

1.5

2.0

2.5

Peak During Shock

Post

-Sho

ck A

ctiv

ity B

urst

(s)

R2 = 0.18p = 0.57

0 1 2 3 420

25

30

35

40

45

Peak During Reward

Rew

ards

Ear

ned R2 = 0.22

p = 0.21

-1 0 1 2 3 40

5

10

15

20

25

Peak During Shock

% T

ime

Free

zing

R2 = 0.42p = 0.17

-1 0 1 2 3 40.0

0.5

1.0

1.5

2.0

Peak During Shock

Post

-Sho

ck A

ctiv

ity B

urst

(s)

R2 = 0.47p = 0.13

DLS

-pro

ject

ing

DM

S-p

roje

ctin

g

E

F

G

H J

THGCaMP6f

THGCaMP6f

THGCaMP6f

GCaMP6f

GCaMP6f

Striatum SNc SNc

M LV

D

ML

L

V

D

ctx

str

ctx

str

I K

Baseline47

5nm

400n

mA B

D

Stimulation C

filter filter

fit

subtract

400

300

200

100

0

Fluo

resc

ence

Inte

nsity

(A.U

.)

151050Time (s)

50 Hz e-stim

475nm 400nm

490nm Signal405nm Control

20 s2% Δ

F/F

L THGCaMP6f

Figure S4. Fiber Photometry Artifact Correction Using an Isosbestic Excitation Wavelength Control Signal, Expression of GCaMP6f in DMS-and DLS-Projecting Dopamine Neurons, and the Relationship of Recorded Fiber Photometry Signals with Behavior, Related to Figure 5

(A) Cultured neuron expressing GCaMP6f showing increased fluorescence upon 50Hz electrical field stimulation when illuminated with 475 nm but not 400 nm

light.

(legend continued on next page)

Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc. S5

(B) Quantification of the fluorescence time course represented in A.

(C) Schematic of the bleaching and artifact correction workflow. Both the calcium-dependent and –independent signals are filtered. The calcium-independent

(control) signal is then fitted to the calcium-dependent signal and subtracted from it to remove artifacts. A DF/F is further calculated by dividing by the control

signal.

(D) Sample trace from a GCaMP6f-expressing mouse under 405 nm or 490 nm illumination. Red dotted lines indicate reward retrieval times. Note the non-

responsiveness of the 405 nm signal to rewarding outcomes.

(E) Left, Histology showing the expression of GCaMP6f in terminals in the either the DMS or the DLS. Dotted line highlights the corpus callosumdividing the cortex

(ctx) and striatum (striatum). Middle, Histology showing the expression of GCaMP6f in SNc cell bodies in relation to TH staining (blue) and fiber optic placement

(dotted rectangle). Right, higher magnification images of GCaMP6f-expressing cell bodies in the SNc, showing healthy expression and overlap with TH staining

(blue).

(F) The number of rewards earned by individual mice expressing GCaMP6f in DMS-projecting SNc dopamine neurons did not correlate with the photometry

signals observed during reward collection.

(G) The number of rewards earned by individual mice expressing GCaMP6f in DLS-projecting SNc dopamine neurons did not correlate with the photometry

signals observed during reward collection.

(H) The time spent freezing during the shock session by individual mice expressing GCaMP6f in DMS-projecting SNc dopamine neurons did not correlate with the

photometry signals observed during shock delivery.

(I) The time spent freezing during the shock session by individual mice expressing GCaMP6f in DLS-projecting SNc dopamine neurons did not correlate with the

photometry signals observed during shock delivery.

(J) The time of post-shockmotor activity (running or jumping) by individual mice expressingGCaMP6f in DMS-projecting SNc dopamine neurons did not correlate

with the photometry signals observed during shock delivery.

(K) The time of post-shock motor activity (running or jumping) by individual mice expressing GCaMP6f in DLS-projecting SNc dopamine neurons did not correlate

with the photometry signals observed during shock delivery.

S6 Cell 162, 635–647, July 30, 2015 ª2015 Elsevier Inc.

Cell

Supplemental Information

Intact-Brain Analyses Reveal Distinct Information

Carried by SNc Dopamine Subcircuits

Talia N. Lerner, Carrie Shilyansky, Thomas J. Davidson, Kathryn E. Evans, Kevin T.

Beier, Kelly A. Zalocusky, Ailey K. Crow, Robert C. Malenka, Liqun Luo, Raju Tomer,

and Karl Deisseroth

Supplemental  Experimental  Procedures    Animals.  All  animal  experiments  were  approved  by  the  Stanford  University  IACUC  

committee,  protocol  #10747. All  mice  used  in  this  study  were  on  a  C57BL/6J  background.  

In  addition  to  wildtype  mice  (Jackson,  strain  664),  Ai9  tdTomato  cre  reporter  mice  

(Jackson,  strain  7909),  TH-­‐GFP  transgenic  mice  (MMRRC,  stock  #000292-­‐UNC)  and  D1-­‐

tdTomato  transgenic  mice  (Jackson,  strain  16204)  were  used  where  indicated.  Only  

heterozygote  transgenic  mice,  obtained  by  backcrossing  to  C57BL/6J  wildtypes,  were  used  

for  experiments.  All  mice  were  2-­‐4  months  old.  Males  were  used  for  TRIO  and  output  

tracing.  Electrophysiology  experiments  and  behavior  were  performed  in  both  males  and  

females  (counterbalanced  across  conditions)  with  no  significant  effects  of  sex  observed.

 

Injection  and  Implant  Surgeries.  Mice  were  anesthetized  in  an  isoflurane  induction  chamber  

at  3-­‐4%  isoflurane,  and  then  injected  with  buprenorphine  (0.05  mg/kg,  s.q.)  and  carprofen  

(5  mg/kg,  s.q.)  prior  to  the  start  of  surgery.  Anesthesia  was  maintained  using  1-­‐2%  

isoflurane.  Hair  was  trimmed  from  the  scalp  and  the  skin  was  cleaned  with  a  povidone-­‐

iodine  solution  (Betadine)  prior  to  incision.  The  scalp  was  opened  using  a  sterile  scalpel  

and  holes  were  drilled  in  the  skull  at  the  appropriate  stereotaxic  coordinates  (relative  to  

bregma:  +0.75  AP,  1.5  ML,  -­‐2.8  DV  for  DMS,  +0.25  AP,  2.5  ML,  -­‐3.4  DV  for  DLS,  -­‐3.1,  0.8  ML,  -­‐

4.7  DV  for  DMS-­‐projecting  SNc,  and  -­‐3.1,  1.3  ML,  -­‐4.2  DV  for  DLS-­‐projecting  SNc.)  Viruses  

were  infused  at  100-­‐200  nl/min  through  a  33-­‐gauge  injection  needle  using  a  syringe  pump  

(World  Precision  Instruments).  Red  Retrobeads  IX  (Lumafluor,  diluted  1:2  in  sterile  saline,  

80  nl)  were  infused  using  the  same  syringe  pump  system  run  at  80  nl/min.  The  needle  was  

left  in  place  for  5  min  following  the  end  of  the  injection,  then  slowly  retracted  to  avoid  

leakage  up  the  injection  tract.  For  fiber  photometry  experiments,  a  fiber  optic  cannula  was  

then  implanted  over  the  SNc  through  the  same  hole  as  made  for  the  virus  injection.  To  

reduce  autofluorescence  artifacts  and  maximize  light  collection,  cannulae  (special  ordered  

from  Doric  Lenses)  were  fabricated  using  0.48  NA  400μM  BFH48-­‐400  fiber,  non-­‐

fluorescent  epoxy  and  metal  2.5mm  ferrules.  Cannulae  were  fixed  to  the  skull  using  a  thin  

later  of  opaque  Metabond  dental  cement  (Parkell,  S396  Radiopaque  L-­‐powder),  followed  

by  Flow-­‐it  ALC  light  curing  dental  cement  (Pentron  Clinical,  N11VH).  After  surgery,  mice  

were  allowed  to  recover  until  ambulatory  on  a  heated  pad,  then  returned  to  their  

homecage.  

 

Viruses.  Viruses  were  injected  where  indicated  in  the  main  text  at  the  following  volumes  at  

titers:  CAV-­‐cre,  2.5  x  1012  vg/ml,  250  nl;  AAVDJ-­‐hSyn1-­‐mGFP-­‐T2A-­‐mRuby,  2.9  x  1013  vg/ml,  

250  nl;  AAV5-­‐CAG-­‐FLeX-­‐TC,  2.6  x  1012  vg/ml,  325-­‐500  nl;  AAV8-­‐CAG-­‐FLeX-­‐G,  1.3  x  1012  

vg/ml,  325-­‐500  nl;  RVdG-­‐GFP,  5  x  108  pfu/ml,  500  nl;  AAV5-­‐CaMKIIα-­‐ChR2(H134R)-­‐EYFP,  

4  x  1012  vg/ml  250-­‐500  nl;  AAV5-­‐EF1α-­‐DIO-­‐GCaMP6f,  1.9  x  1013  vg/ml,  1000  nl.  

 

Transcardial  Perfusions.  Mice  received  lethal  i.p.  injections  of  Beuthanasia  (Merck),  a  

combination  of  sodium  pentobarbital  and  sodium  phenytoin,  to  induce  a  smooth  and  rapid  

onset  of  unconsciousness  and  death.  Once  unresponsive  to  a  firm  toe  pinch,  an  incision  was  

made  up  the  middle  of  the  body  cavity.  An  injection  needle  was  inserted  into  the  left  

ventricle  of  the  heart,  the  right  atrium  was  punctured  and  solution  (NMDG  cutting  solution  

for  electrophysiology,  or  PBS  followed  by  4%  PFA  for  traditional  histology,  or  PBS  followed  

by  CLARITY  monomer  solution  for  CLARITY)  was  infused  as  the  mouse  was  exsanguinated.  

The  mouse  was  then  decapitated  and  its  brain  was  quickly  removed  for  further  

experimentation.  

 

CLARITY.  Mice  were  perfused  with  CLARITY  monomer  solution  containing  1%  acrylamide,  

0.0125%  bis-­‐acrylamide,  4%  PFA,  and  0.25%  VA-­‐044  diluted  in  PBS  and  left  for  two  

overnights  at  4°C  to  allow  the  monomer  to  fully  penetrate  the  tissue.  Then,  the  samples  

were  degassed  and  moved  to  a  37°C  water  bath  for  6-­‐7  hours  to  polymerize.  After  

polymerization,  excess  hydrogel  was  poured  off  and  the  brains  were  rinsed  with  PBS,  then  

transferred  to  clearing  solution  containing  4%  SDS  and  200mM  boric  acid  in  water,  

adjusted  to  pH  8.5  with  NaOH.  Whole  brains  were  incubated  in  clearing  solution,  shaking,  

at  37°C  for  4-­‐5  weeks  (until  brains  were  clear).  Clearing  solution  was  changed  once  per  

week.  After  clearing,  brains  were  stored  in  PBS  +  0.02%  sodium  azide  at  room  temperature  

until  ready  for  COLM  imaging.  For  imaging,  brains  were  equilibrated  in  Focusclear  

(CelExplorer  Labs).  

 

Tissue  Sectioning  

After  perfusion,  brains  were  fixed  overnight  at  4°C  in  4%  PFA,  then  transferred  to  solution  

of  30%  sucrose  in  PBS,  where  they  were  stored  for  at  least  two  overnights  at  4°C  before  

sectioning.  For  TRIO,  tissue  was  embedded  in  OCT  (Tissue-­‐Tek)  and  sectioned  on  a  cryostat  

at  60  μm  and  mounted  on  slides  for  counterstaining.  All  other  tissue  was  sectioned  on  a  

freezing  microtome  at  40  μm,  stored  in  cryoprotectant  (25%  glycerol  and  30%  ethylene  

glycol  in  PBS,  pH  6.7)  at  4°C,  then  stained  as  free-­‐floating  sections  before  mounting.  

 

Histology  and  Immunostaining.  

TRIO  sections  were  counterstained  with  Neurotrace  435/455  Blue  Fluorescent  Nissl  Stain  

and  DAPI  (Life  Technologies,  Cat.  No.  N-­‐21479,  D1306)  on  slides.  Slides  were  rinsed  with  

PBS  for  5  minutes,  permeabilized  with  0.3%  Triton  X  in  PBS  (PBS-­‐T)  for  20  minutes,  

stained  with  Neurotrace  (1:500  in  PBS)  for  2  hours  at  room  temperature,  stained  with  

DAPI  (300nM)  for  20  minutes,  then  rinsed  with  PBS  for  5  minutes  and  coverslipped  with  

Fluoromount  G  mounting  medium  (Southern  Biotech).  Tyrosine  hydroxlase  (TH)  staining  

was  performed  where  indicated  in  the  text.  Staining  was  performed  on  free  floating  

sections,  which  were  blocked  with  3%  normal  donkey  serum  in  PBS-­‐T  for  1  hour  at  room  

temperature,  then  stained  with  1:500  primary  antibody  (Aves  Labs,  Cat  No.  TYH)  in  

blocking  solution  at  4°C  overnight.  Secondary  staining  was  performed  using  1:500  goat  

anti-­‐chicken  Alexa  Fluor  647  secondary  antibody  (Life  Technologies,  Cat.  No.  A-­‐21449).  

Anti-­‐GFP  staining  was  performed  to  amplify  signals  from  all  sections  containing  mGFP  and  

GCaMP6f  by  blocking  in  3%  normal  donkey  serum  in  PBS-­‐T  for  1  hour  at  room  

temperature,  then  using  1:500  primary  antibody  conjugated  directly  to  either  Alexa  Fluor  

488  or  Alexa  Fluor  647  (Life  Technologies,  Cat.  No.    A-­‐21311,  A-­‐31852)  in  blocking  solution  

at  4°C  overnight.  

 

Traditional  Imaging.    

Axon  collateralization  mapping  sections  and  TRIO  sections  were  imaged  on  a  

epifluorescent  slide  scanner  system  (Leica,  DM6000  B  with  Ariol  SL200  slide  loader)  with  a  

5x  air  objective  using  Neurotrace  or  DAPI  counterstaining  to  focus  the  images.  All  other  

images  were  taken  using  a  confocal  microscope  (Leica,  TCS  SP5)  with  5x  or  10x  air  

immersion  objectives,  or  40x  or  63x  oil  immersion  objectives.  In  sections  where  

fluorescence  intensity  was  quantitatively  compared,  all  acquisition  settings  were  held  

constant.  

 

Axon  Collateralization  Analysis.  

Images  of  mGFP  expression  were  acquired  from  a  systematic  image  series  from  each  

mouse  corresponding  to  the  full  anterior-­‐posterior  span  of  the  striatum  (5-­‐6  sections  per  

brain).  All  images  were  acquired  using  identical  settings  and  were  analyzed  using  ImageJ.  

Regions  of  interest  were  defined  using  the  DAPI  reference  channel.  Images  were  then  

background  subtracted  (rolling  ball  radius  of  50  pixels),  thresholded    and  binarized  (pixel  

values  >30  =  black).  The  projection  fraction  was  calculated  as  (black  area  in  region  of  

interest)  /  (total  black  area  in  all  regions).    

 

COLM  Imaging  and  3D  volume  rendering.  Whole  brain  imaging  was  performed  using  COLM  

(Tomer  et  al.,  2014)  or  second  generation  COLM  (Tomer  et  al.,  in  preparation),  as  described  

in  detail  (Tomer  et  al.,  2014).  The  images  were  acquired  along  the  dorsal-­‐ventral  axis.  

Whole  brain  volume  rendering  movies  were  prepared  using  4x4  fold  downsampled  data  in  

the  x-­‐y  dimensions or  using  full  resolution  data  in  the  case  of  Movie  S5.  Data  stitching  was  

performed  using  an  adapted  Terastitcher  based  pipeline  (Bria  and  Iannello,  2012)  and  

volume  rendering  using  Amira  (FEI)  and  ImageJ  software.  

 

Cell  Counting.  Cell  counting  was  performed  in  a  semi-­‐automated  fashion  using  Imaris  

software  (Bitplane).  For  each  brain  region  counted,  Imaris  spot  detection  parameters  were  

defined  that  most  accurately  counted  GFP-­‐labeled  cell  bodies  as  determined  by  manual  

verification.  Brain  regions  were  defined  with  reference  to  the  Allen  Mouse  Brain  Reference  

Atlas  (http://mouse.brain-­‐map.org/static/atlas),  using  the  Paxinos  and  Franklin  atlas  

(Paxinos  and  Franklin,  2004)  as  a  secondary  resource  (e.g.  for  NAc  core  vs.  shell).  For  

subdivision  of  the  dorsal  striatum  into  dorsomedial  and  dorsolateral  portions,  an  in-­‐house  

atlas  was  created  (Fig.  S1C-­‐D).  Regional  boundaries  were  defined  a  priori  without  biasing  

by  reference  to  labeling  in  our  experiments.  

 

Slice  Electrophysiology.  Three  weeks  after  injections,  mice  were  transcardially  perfused  

with  a  room-­‐temperature  NMDG  slicing  solution  (containing  in  mM:  92  N-­‐methyl-­‐D-­‐

glucamine,  2.5  KCl,  30  NaHCO3,  1.2  NaH2PO4-­‐H2O,  20  HEPES,  25  glucose,  5  sodium  

ascorbate,  2  thiourea,  3  sodium  pyruvate,  adjusted  to  pH  7.4  with  HCl)  and  the  brain  tissue  

was  sliced  in  the  same  NMDG  solution  using  a  vibratome  (VT1200S,  Leica).  Slices  

containing  the  striatum  were  fixed  in  4%  PFA  and  saved  for  verification  of  the  ChR2  and  

retrobead  injection  sites.  300µm  coronal  slices  containing  the  substantia  nigra  were  

allowed  to  recover  for  10  minutes  at  33°C  in  NMDG  solution,  then  another  20  minutes  at  

33°C  in  a  modified  HEPES  artifical  cerebrospinal  fluid  (containing    in  mM:  92  NaCl,  2.5  KCl,  

30  NaHCO3,  1.2  NaH2PO4-­‐H2O,  20  HEPES,  25  glucose,  5  sodium  ascorbate,  2  thiourea,  3  

sodium  pyruvate),  then  another  15  minutes  at  room  temperature  in  the  modified  HEPES  

solution.  Finally,  slices  were  transferred  to  standard  artificial  cerebrospinal  fluid  (aCSF;  

containing  in  mM:  125  NaCl,  2.5  KCl,  2  CaCl2,  1  MgCl2,  26  NaHCO3,  1.25  NaH2PO4-­‐H2O,  11  

glucose)  bubbled  with  95%O2/5%CO2  and  stored  at  room  temperature  until  recording.    

Whole-­‐cell  patch  clamp  recordings  were  performed  in  the  same  aCSF  solution  at  30-­‐32°C.  

Signals  were  amplified  with  a  Multiclamp  700B  amplifier,  acquired  using  a  Digidata  1440A  

digitizer,  sampled  at  10  kHz,  and  filtered  at  2  kHz.  All  data  acquisition  and  analysis  were  

performed  using  pCLAMP  software  (Molecular  Devices).  Neurons  were  visually  identified  

for  patching  using  an  upright  microscope  (Olympus  BX51WI)  equipped  with  IR-­‐DIC  optics,  

filter  sets  for  visualizing  YFP  and  RFP,  and  a  CCD  camera  (RoleraXR,  Q-­‐Imaging).  Resistance  

of  the  patch  pipettes  was  2.5–4MΩ  when  filled  with  intracellular  solution.  To  record  action  

potentials  and  Ih  currents,  we  used  intracellular  solution  containing  the  following  (in  mM):  

150  K-­‐gluconate,  5  NaCl,  0.2  EGTA,  10  HEPES,  1MgCl2,  2  Mg-­‐ATP,  0.3  Na-­‐GTP,  adjusted  to  

pH  7.3  with  KOH.  To  measure  Ileak  and  Ih,  we  held  neurons  at  -­‐40mV  and  stepped  for  500ms  

to  -­‐120mV.  The  average  of  three  sweeps  was  taken,  then  Ileak  was  calculated  as  the  change  

in  current  between  the  baseline  before  the  voltage  step  was  applied  and  the  current  at  ~40  

ms  after  the  voltage  step  was  applied.  Ih  was  calculated  from  the  change  in  current  between  

~40  and  498  ms  after  the  voltage  step  was  applied.  To  record  EPSCs  and  IPSCs,  neurons  

were  voltage-­‐clamped  at  -­‐70mV.  To  record  EPSCs,  we  added  picrotoxin  (50µM)  to  the  

extracellular  solution  and  used  intracellular  solution  containing  the  following  (in  mM):  120  

CsMeSO3,  15  CsCl,  8  NaCl,  0.2  EGTA,  10  HEPES,  2  Mg-­‐ATP,  0.3  Na-­‐GTP,  10  TEA  

(tetraethylammonium),  5  QX-­‐314  (lidocaine  N-­‐ethyl  bromide),  adjusted  to  pH  7.25  with  

CsOH.  To  record  IPSCs,  we  added  NBQX  (5µM)  and  APV  (50µM)  to  the  extracellular  

solution  and  used  intracellular  solution  containing  the  following  (in  mM):  130  CsCl,  1  

EGTA,  10  HEPES,  2  Mg-­‐ATP,  0.3  Na-­‐GTP,  10  TEA  (tetraethylammonium),  5  QX-­‐314  

(lidocaine  N-­‐ethyl  bromide),  adjusted  to  pH  7.25  with  CsOH.  Miniature  EPSCs  and  IPSCs  

were  recorded  in  the  presence  of  TTX  (1µm).  To  isolate  monosynaptic  IPSCs  from  the  DMS  

and  DLS,  we  added  TTX  (1µm)  and  4-­‐AP  (100µM)  to  the  extracellular  solution.  To  stimulate  

ChR2  expressed  in  axon  terminals  from  the  striatum,  5ms  blue  light  pulses  (475nm,  

~10mW/mm2)  were  generated  using  a  Spectra  X  LED  light  engine  (Lumencor)  and  

delivered  to  the  slice  via  a  40x/0.8  water  immersion  objective  focused  onto  the  recorded  

neuron.  Pulses  were  delivered  once  every  30  seconds.  Response  sizes  were  calculated  by  

baseline-­‐subtracting  and  averaging  10-­‐15  traces  together,  then  calculating  the  peak  

amplitude  in  a  50ms  window  after  the  light  pulse.  Neurons  that  did  not  show  a  peak  

amplitude  in  this  window  that  exceeded  5  SD  of  the  baseline  noise  were  counted  as  non-­‐

responders.  To  measure  qIPSC  amplitude,  we  replaced  extracellular  calcium  with  2mM  

strontium  to  induce  asynchronous  release.  We  then  calculated  the  average  amplitude  of  

events  occurring  between  50ms  and  250ms  after  the  blue  light  pulse.  The  amplitude  of  

each  event  was  measured  from  a  local  baseline  just  before  the  event  (from  5-­‐20ms  prior  to  

the  start  of  the  event  to  the  start)  to  avoid  artifacts  arising  from  the  decaying  transient  of  

the  initial  synchronous  release.  

 

Reward  Behavior.  Mice  were  water-­‐restricted  to  ~1mL/day  (maintained  at  >85%  ad  

libitum  weight)  and  trained  to  press  a  lever  for  25µl  of  20%  sucrose  water  on  a  1:1  

continuous  reinforcement  schedule.  Med  Associates  operant  boxes  (ENV-­‐307A-­‐CT)  were  

arranged  with  an  extra-­‐tall  sucrose  port  on  one  wall  (to  allow  for  head  entry  with  the  

implant  attached)  and  active  levers  both  to  the  right  and  to  the  left  of  the  port.  Mice  were  

required  to  retrieve  each  reward  before  the  onset  of  the  next  trial.  Mice  were  trained  on  

this  schedule  to  criterion  performance  of  >60%  available  rewards  earned  in  one  hour,  

which  required  4-­‐8  days.  Mice  were  then  advanced  to  testing  on  a  random  ratio  (RR10)  

schedule  in  which  each  lever  press  had  a  10%  chance  being  rewarded.  This  schedule  was  

chosen  to  allow  for  increased  total  number  of  trials  per  session  and  for  within-­‐session  

comparisons  of  rewarded  and  non-­‐rewarded  trials.  One  mouse  was  not  advanced  to  testing  

due  to  complications  of  surgery  becoming  evident  during  training.  All  sessions  from  start  of  

task  habituation  and  training  were  performed  with  the  mouse  plugged  in  to  a  400  µm  

BFH48-­‐400  patch  cord  using  a  metal  ferrule  sleeve,  regardless  of  whether  signals  were  

being  recorded  in  order  to  habituate  the  mouse  to  tethering.  

 

Shock  Exposure  Behavior.  Mice  were  placed  in  an  operant  chamber  (Med  Associates,  

different  from  the  reward  training  chamber,  but  of  the  same  dimensions)  with  a  shock  

floor,  with  their  fiber  optic  implant  plugged  in  to  a  400  µm  BFH48-­‐400  patch  cord  using  a  

metal  ferrule  sleeve.  Fifteen  mild  shocks  (0.4  mA,  0.5  s)  were  delivered  at  random,  about  

once  per  minute  (RI60  schedule).  Photometry  signal  was  recorded  throughout  session  and  

behavior  was  recorded  by  overhead  camera.  Freezing,  defined  as  the  complete  absence  of  

movement  except  as  required  for  respiration,  was  manually  scored.  Activity  burst,  defined  

as  abrupt  increase  in  velocity  of  movement,  was  manually  timed  until  return  to  baseline  

movements  in  cage.  Behavioral  scoring  was  completed  for  a  subset  of  animals  by  a  single  

experimenter  blind  to  mouse  group  and  photometry  results.

 

Fiber  Photometry.  As  in  Gunaydin  et  al.  (2014),  we  measured  bulk  fluorescence  from  deep  

brain  regions  using  a  single  optical  fiber  for  both  delivery  of  excitation  light  and  collection  

of   emitted   fluorescence.   The   fluorescence   output   of   the   calcium   sensor   is  modulated   by  

varying   the   intensity   of   the   excitation   light,   generating   an   amplitude-­‐modulated  

fluorescence   signal   that   is   demodulated   to   recover   the   original   calcium   sensor   response.  

This  ‘upconversion’  of  the  calcium  signal  avoids  contamination  of  the  signal  by  changes  in  

ambient  light  levels  with  behavior  ,  as  well  as  avoiding  low-­‐frequency  ‘flicker  noise’  in  our  

photodetector.   We   have   extended   this   method   to   the   case   of   multiple   excitation  

wavelengths  delivered  over  the  same  fiber,  each  modulated  at  a  distinct  carrier  frequency,  

to   allow   for   simultaneous  multicolor  measurements.   Two   excitation   LEDs   (490nm   ‘blue’  

and   405nm   ‘violet’,   Thor   Labs  M490F1   and  M405F1)  were   controlled   via   an  RP2.1   real-­‐

time  processor  (Tucker  Davis  Technologies)  running  custom  software.  Blue  excitation  was  

sinusoidally  modulated   at   211Hz   and   passed   through   a   GFP   excitation   filter   (Thor   Labs,  

MF469-­‐35);   violet   excitation   was   modulated   at   531Hz   and   passed   through   a   405nm  

bandpass   filter   (Thor   Labs,   FB405-­‐10).   These   carrier   frequencies   were   chosen   to   avoid  

contamination   from   overhead   lights   (120Hz   and   harmonics)   and   cross-­‐talk   between  

channels  (the  bandwidth  of  GCaMP6f  was  observed  to  be  <15Hz),  while  remaining  within  

the  30-­‐750Hz  bandwidth  of  the  photoreceiver.    Excitation  light  from  each  LED  was  coupled  

into  a  0.39NA,  200um-­‐core  fiber  (chosen  to  avoid  overfilling  the  patchcord  to  the  animal),  

collimated,   then   combined  by   a   425nm   long-­‐pass   dichroic  mirror   (ThorLabs,  DMLP425).  

The  excitation  light  was  coupled  into  a  high-­‐NA  (0.48),  low-­‐autofluorescence  400µm  patch  

cord  (manufactured  by  Doric  Lenses,,  using  BFH48-­‐400  fiber  from  ThorLabs)  using  a  fixed-­‐

focused   high   NA   (0.51)   coupler/collimator   (Thor   Labs,   F240FC-­‐A).   Each   LED  was   set   to  

30µW  at  the  far  end  of  the  patch  cord,  which  was  terminated  with  a  2.5  mm  ferrule.  The  far  

end   of   the   patch   cord   and   the   2.5mm   metal   optical   implant   ferrule   were   cleaned   with  

isopropanol  before  each   recording,   then   securely  attached  via  a  metal  or   zirconia   sleeve.  

The  GCaMP6f  emission  signal  was  collected  through  the  patch  cord  and  collimator,  passed  

through   a   GFP   emission   filter   (Thor   Labs,   MF525-­‐39)   and   focused   onto   a   femtowatt  

photoreceiver  (Newport,  Model  2151)  using  a  lens  (Edmund  Optics,  Cat.  No.  62-­‐561).  The  

photoreceiver   signal   was   sampled   at   6.1   kHz,   and   each   of   the   two   modulated   signals  

generated   by   the   two   LEDs   was   independently   recovered   using   standard   synchronous  

demodulation   techniques   implemented   on   the   RP2.1   real-­‐time   processor:   the   detector  

output  was  routed  to  two  product  detectors,  one  using  the  selected  channel’s  modulation  

signal   as   a   reference,   and   the   other   using   a   90-­‐degree   phase-­‐shifted   copy   of   the   same  

reference.  These  outputs  were  low-­‐pass  filtered  (corner  frequency  of  15Hz),  and  added  in  

quadrature.  This  dual-­‐phase  detection  approach  makes  the  output  insensitive  to  any  phase  

delay  between  the  reference  and  signal.  The  resulting  fluorescence  magnitude  signals  were  

then  decimated  to  382  Hz  for  recording  to  disk  and  further  filtered  using  an  ~2Hz  low-­‐pass  

filter  before  analysis.  Behavioral  variables,  such  as  lever  presses,  reward  port  entry  times  

and   shock   times,  were   fed   into   the   real-­‐time   processor   as   TTL   signals   from   the   operant  

chambers.  Files  were  then  exported  for  analysis  to  MATLAB  (MathWorks)  using  a  custom  

script.  A   least-­‐squares   linear  fit  was  applied  to  the  405nm  control  signal  to  align  it   to  the  

490nm   signal.   The   ΔF/F   time   series   was   then   calculated   for   each   behavioral   session   as  

((490nm  signal  -­‐  fitted  405nm  signal)  /  fitted  405nm  signal).  Peristimulus  time  histograms  

(PSTHs)  were   created   using   the   TTL   timestamps   that   had   been   fed   in   from   the   operant  

chambers.  Peak  reward  ΔF/F    was  calculated  as  the  maximum  ΔF/F    value  in  the  rewarded  

port  entry  PSTH  in  a  one  second  period  around  the  reward  retrieval  (time  0).    No  reward  

ΔF/F  was  calculated  as  the  ΔF/F    at  time  0  in  the  non-­‐rewarded  port  entry  PSTH.  Peak  ΔF/F  

during   shock   was   calculated   as   the   most   extreme   ΔF/F     value   (positive   or   negative)  

observed  during   the   shock  period   (time  0-­‐0.5   seconds).  Mean  ΔF/F     at   1-­‐5   seconds  post  

shock   was   calculated   as   the   mean   ΔF/F     value   at   time   1-­‐5   seconds,   where   the   shock  

occurred  at  time  0-­‐0.5  seconds.  

 

Statistics.  Unpaired  t-­‐tests  were  used  for  comparisons  between  two  groups  (DMS-­‐  and  DLS-­‐

projecting  SNc  dopamine  neurons).  Two-­‐way  ANOVAs  were  used  to  assess  how  the  

properties  or  responses  of  DMS-­‐  and  DLS-­‐projecting  SNc  dopamine  neurons  were  affected  

by  other  factors  (e.g.  input  area).  When  a  statistically  significant  effect  was  observed  using  

a  two-­‐way  ANOVA,  post-­‐hoc  testing  with  correction  for  multiple  comparisons  was  

performed  using  Sidak’s  or  Tukey’s  multiple  comparisons  test.  Statistics  were  performed  

using  Prism  6  (GraphPad)  software.  

   

 

Supplementary  Tables    Table  S1.  Percent  Ipsilateral  and  Contralateral  Inputs  to  DMS-­‐  and  DLS-­‐projecting  SNc  Dopamine  Neurons,  Related  to  Figures  1  and  S1  N=  4  brains  per  condition.  Shown  are  means  ±  SEM.       %  Ipsilateral  Inputs   %  Contralateral  Inputs  Input  Area   DMS-­‐

projecting  DLS-­‐  

projecting  DMS-­‐

projecting  DLS-­‐  

projecting  M1/2   1.66  ±  0.17   1.90  ±  0.20   3.04  ±  0.97   2.33  ±  0.58  S1   0.99  ±  0.07   1.21  ±  0.08   0.91  ±  0.36   0.52  ±  0.22  Cortex  (Other)   0.67  ±  0.08   0.78  ±  0.16   2.27  ±  0.70   1.42  ±  0.13  NAc  core   7.88  ±  2.03   3.49  ±  1.01   1.90  ±  1.39   1.15  ±  0.73  NAc  med  shell   4.23  ±  0.74   1.87  ±  0.62      NAc  lat  shell   2.83  ±  0.33   3.12  ±  0.16      DMS   18.09  ±  3.67   11.72  ±  1.33   0.84  ±  0.39   0.63  ±  0.29  DLS   27.90  ±  2.12   34.78  ±  3.81      FS  (IPAC)   0.93  ±  0.45   2.10  ±  0.38   0.42  ±  0.15   0.07  ±  0.07  GPe   4.47  ±  0.28   6.01  ±  1.20   0.17  ±  0.1   0.66  ±  0.66  SI/VP   2.04  ±  0.12   2.45  ±  0.20   5.33  ±  1.79   2.91  ±  1.70  BNST   1.87  ±  0.43   2.21  ±  0.39   0.34  ±  0.13   0.40  ±  0.31  CeA   1.44  ±  0.35   1.83  ±  0.38   0.24  ±  0.13   0.10  ±  0.07  Hypothalamus   12.11  ±  1.55   11.37  ±  1.50   29.61  ±  0.92   23.49  ±  6.74  Thalamus   2.13  ±  0.17   1.63  ±  0.32   4.89  ±  1.37   3.23  ±  0.72  LHb   0.10  ±  0.03   0.31  ±  0.16   2.79  ±  0.69   1.52  ±  0.18  SC   1.83  ±  0.24   1.50  ±  0.30   4.48  ±  0.56   2.81  ±  0.59  IC   0.28  ±  0.10   0.10  ±  0.10   1.25  ±  0.81   0.63  ±  0.30  SNc   1.80  ±  0.49   1.65  ±  0.57   1.30  ±  0.53   1.96  ±  1.14  SNr   1.61  ±  0.28   1.83  ±  0.41   1.40  ±  0.76   2.86  ±  0.66  VTA   2.93  ±  0.52   2.27  ±  0.56   3.60  ±  2.08   5.82  ±  2.41  PAG   2.64  ±  0.58   2.45  ±  0.64   11.52  ±  2.98   11.74  ±  2.20  Dorsal  Raphe   0.41  ±  0.22   0.62  ±  0.15   0.73  ±  0.43   7.53  ±  1.26  MRN   4.41  ±  1.10   3.38  ±  0.87   11.09  ±  3.12   22.83  ±  5.13  PPN   1.44  ±  0.32   0.66  ±  0.22   0.81  ±  0.51   1.70  ±  0.90  PB   0.72  ±  0.10   0.76  ±  0.33   9.86  ±  4.15   2.75  ±  1.45          

Supplemental  References    Bria,  A.,  and  Iannello,  G.  (2012).  TeraStitcher  -­‐  a  tool  for  fast  automatic  3D-­‐stitching  of  teravoxel-­‐sized  microscopy  images.  BMC  Bioinformatics  13,  316.  

Paxinos,  G.,  and  Franklin,  K.B.J.  (2004).  The  Mouse  Brain  in  Stereotaxic  Coordinates  (San  Diego:  Academic  Press).