Identification of peripheral neural circuits that regulate heart rate using optogenetic and viral vector strategies

ARTICLE

Identi

cation of peripheral neural circuits that

regulate heart rate using optogenetic and viral

vector strategies

Pradeep S. Rajendran

, Rosemary C. Challis

, Charless C. Fowlkes

, Peter Hanna

, John D. Tompkins

Maria C. Jordan

, Sarah Hiyari

, Beth A. Gabris-Weber

, Alon Greenbaum

, Ken Y. Chan

2,6

Benjamin E. Deverman

2,6

, Heike Münzberg

, Jeffrey L. Ardell

, Guy Salama

, Viviana Gradinaru

Kalyanam Shivkumar

Heart rate is under the precise control of the autonomic nervous system. However, the wiring

of peripheral neural circuits that regulate heart rate is poorly understood. Here, we develop a

clearing-imaging-analysis pipeline to visualize innervation of intact hearts in 3D and

employed a multi-technique approach to map parasympathetic and sympathetic neural cir-

cuits that control heart rate in mice. We identify cholinergic neurons and noradrenergic

neurons in an intrinsic cardiac ganglion and the stellate ganglia, respectively, that project to

the sinoatrial node. We also report that the heart rate response to optogenetic versus

electrical stimulation of the vagus nerve displays different temporal characteristics and that

vagal afferents enhance parasympathetic and reduce sympathetic tone to the heart via

central mechanisms. Our

ndings provide new insights into neural regulation of heart rate,

and our methodology to study cardiac circuits can be readily used to interrogate neural

control of other visceral organs.

https://doi.org/10.1038/s41467-019-09770-1

OPEN

Cardiac Arrhythmia Center and Neurocardiology Research Program of Excellence, David Geffen School of Medicine, University of California - Los Ange

les

(UCLA), Los Angeles, CA 90095, USA.

Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, CA 91125, USA.

Department of Computer Science, University of California - Irvine, Irvine, CA 92697, USA.

Department of Cell Biology, University of Pittsburgh, Pittsburgh,

PA 15261, USA.

Neurobiology of Nutrition and Metabolism Department, Louisiana State University, Baton Rouge, LA 70808, USA.

Present address: Stanley

Center for Psychiatric Research, Broad Institute, Massachusetts Institute of Technology, Cambridge, MA 02142, USA. Correspondence and requests f

materials should be addressed to V.G. (email:

viviana@caltech.edu

) or to K.S. (email:

kshivkumar@mednet.ucla.edu

)

NATURE COMMUNICATIONS

| (2019) 10:1944 | https://doi.org/10.1038/s41467-019-09770-1 | www.nature.com/naturecommunications

1

1234567890():,;

S

ituation-dependent changes in heart rate are essential for

survival and are under the precise control of the autonomic

nervous system (ANS)

. Heart rate reduction during sleep

and elevation during exercise

result from changes in para-

sympathetic and sympathetic tone. In fact, heart rate variability

has been utilized extensively as an index of ANS function

–

Although it is well known that the parasympathetic and sympa-

thetic nervous systems innervate the sinoatrial (SA) node

and

regulate heart rate

–

, the wiring of these neural circuits in the

periphery is not well characterized. Anatomical and functional

maps of these fundamental cardiac circuits are needed to

understand physiology, characterize remodeling in disease (e.g.,

sick sinus syndrome

), and develop novel therapeutics. However,

these efforts have been hindered by a shortage of tools that target

the peripheral nervous system (PNS) with speci

fi

city and

precision.

Cardiac circuit anatomy has traditionally been studied using thin

sections

and whole-mount preparations

. However, these

methods do not preserve the structure of intact circuits and only

provide 2D information. In contrast, tissue clearing methods render

tissues optically transparent while preserving their molecular and

cellular architecture and can be combined with a variety of labeling

strategies to enable 3D visualization of intact circuits

.Totrace

cardiac circuits, dyes and proteins have historically been used.

However, achieving cell type-speci

fi

city and/or sparse labeling

needed for singe-cell tracing and delineating circuit connectivity is

dif

fi

cult or not possible with these methods

. Additionally, many

peripheral neuronal populations such as intrinsic cardiac ganglia are

challenging to access surgically for tracer delivery. Adeno-associated

viruses (AAVs) can address these limitations since they can be used

to express genes of interest (e.g.,

fl

uorescent proteins) in de

fi

ned cell

populations in wild-type and transgenic animals

–

. In addition,

intersectional strategies can be used to titrate gene expression to

achieve sparse labeling

. AAVs can also be delivered systemically

to target dif

fi

cult-to-reach populations

–

Functional mapping of cardiac circuits has relied on electrical

or pharmacological manipulation of the ANS with simultaneous

physiological measurements

. However, each of these methods

has disadvantages. Electrical techniques lack spatial precision and

speci

fi

city. Autonomic nerves such as the vagus contain motor

and sensory

fi

bers

, and electrical stimulation typically activates

both

fi

ber types

as well as surrounding tissues

. Pharma-

cological techniques exhibit improved selectivity but lack

temporal resolution. In contrast, optogenetics, which uses

light-sensitive ion channels (e.g., channelrhodopsin-2 (ChR2),

halorhodopsin), enables precise spatiotemporal control of de

fi

ned

cell populations.

Here, we develop a clearing-imaging-analysis pipeline to

visualize innervation of whole hearts in 3D and employ a multi-

technique approach, which includes AAV-based sparse labeling

and tracing, retrograde neuronal tracing with cholera toxin sub-

unit B (CTB), and optogenetics with simultaneous physiological

measurements, to map peripheral parasympathetic and sympa-

thetic neural circuits that regulate heart rate in mice.

Results

Tissue clearing and computational pipeline to assess cardiac

innervation

. To characterize global innervation of the mouse

heart in 3D, we developed a clearing-imaging-analysis pipeline

(Fig.

a). We stained whole hearts with an antibody against the

pan-neuronal marker protein gene product 9.5 (PGP9.5) and

rendered them optically transparent using an immunolabeling-

enabled three-Dimensional Imaging of Solvent-Cleared Organs

(iDISCO) protocol (Fig.

. We used confocal microscopy to

image large tissue volumes (Fig.

c and Supplementary Movie 1)

and both confocal and lightsheet microscopy to image entire

hearts (Supplementary Movie 2)

. We observed cardiac

ganglia surrounding the pulmonary veins and a dense network of

nerve

fi

bers coursing through the atrial and ventricular

myocardium (Fig.

c). In contrast to whole-mount stained hearts,

innervation was seen throughout the entire thickness of the

myocardium in iDISCO-cleared hearts, with large-diameter nerve

fi

ber bundles located near the epicardium and smaller

fi

ber

bundles in the mid-myocardium and endocardium (Fig.

Supplementary Fig. 1, and Supplementary Movie 3). To analyze

these data, we created a semiautomated computational pipeline to

detect nerve

fi

bers over large tiled volumes and to measure

microanatomical features of

fi

bers such as diameter and orien-

tation (Fig.

d and e). Large-diameter nerve

fi

ber bundles typi-

cally entered near the base of the dorsal heart. These bundles

coursed perpendicular to the atrioventricular (AV) groove and

branched into smaller

fi

ber bundles as they progressed towards

the apex. These data from healthy hearts will be important for

future characterization of neural remodeling in cardiovascular

diseases such as myocardial infarction (MI) in which innervation

patterns are disrupted and nerve sprouting occurs

While the iDISCO protocol was speci

fi

cally developed for

immunostaining applications, other clearing techniques such as

the PAssive Clarity Technique (PACT) are better suited for

visualizing endogenous

fl

uorescence in large tissue volumes or

whole organs

. We demonstrate that PACT preserves the

fl

uorescence of virally labeled cholinergic and endogenously labeled

noradrenergic neurons and nerve

fi

bers in the heart (Supplementary

Fig. 2 and Supplementary Movies 4

–

6). Therefore, the application

should dictate the choice of clearing technique.

AAV-based labeling and tracing of cholinergic neurons on the

heart

. After visualizing global cardiac innervation, we assessed

whether we could identify a subset of cholinergic neurons that form

an anatomical circuit with the SA node to potentially regulate heart

rate. We used an AAV-based system to trace

fi

bers, presumably

from cholinergic neurons in intrinsic cardiac ganglia. Multicolor

labeling strategies allow individual cells to be distinguished from

one another (Fig.

a) and sparse labeling reduces the fraction of

labeled cells to allow individual

fi

bers to be visualized (Fig.

.To

demonstrate this, we systemically co-administered Cre-dependent

vectors expressing

fl

uorescent proteins (XFPs) from the

tetracycline-responsive element (TRE)-containing promoter at a

high dose and the tetracycline transactivator (tTA) from the human

synapsin I promoter (hSyn1) at a lower dose in ChAT-IRES-Cre

transgenic mice (Fig.

. Compared to dense multicolor labeling

(Fig.

a), sparse multicolor labeling resulted in a labeling density in

intrinsic cardiac ganglia that was lower and that would more easily

allow for tracing (Fig.

. To trace cholinergic

fi

ber, we utilized

sparse single-color labeling with tdTomato. Three weeks after viral

delivery, hearts were collected and stained with an antibody for

hyperpolarization-activated cyclic nucleotide-gated potassium

channel 4 (HCN4). HCN4 staining along with anatomical land-

marks were used to identify the SA node, the AV node, and the

conduction system

. We observed cholinergic

fi

bers, presumably

from cardiac ganglia, coursing along the SA node, AV node, and

ventricles (Fig.

c), identifying cholinergic neurons that are poten-

tially involved in chronotropic, dromotropic, and ventricular con-

trol, respectively.

Optogenetic stimulation of cholinergic neurons in the inferior

pulmonary vein-ganglionated plexus

. Next, to functionally assess

whether cholinergic neurons in the inferior pulmonary vein-

ganglionated plexus (IPV-GP) regulate heart rate and AV con-

duction, we used an optogenetic approach. We expressed ChR2 in

ARTICLE

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09770-1

2

NATURE COMMUNICATIONS

| (2019) 10:1944 | https://doi.org/10.1038/s41467-019-09770-1 | www.nature.com/naturecommunications

cholinergic neurons by crossing transgenic ChAT-IRES-Cre mice

with reporter mice containing a Cre-dependent ChR2-enhanced

yellow

fl

uorescent protein allele (ChR2-eYFP; offspring from this

cross are subsequently referred to as ChAT-ChR2-eYFP mice).

ChR2-eYFP expression in intrinsic cardiac neurons was con

fi

rmed

by staining hearts for GFP (to amplify ChR2-eYFP detection) and

PGP9.5 (Fig.

a). All GFP

+

neurons were PGP9.5

+

(100.0 ± 0.0%)

and the majority of PGP9.5

+

neurons were GFP

+

(96.4 ± 1.2%)

(Fig.

b). GFP staining was also present in PGP9.5

+

nerve

fi

bers in

the atria and ventricles (Fig.

a). These data are consistent with

previous studies reporting that the majority of intrinsic cardiac

neurons are cholinergic

and that the ventricles as well as atria

receive cholinergic innervation

–

After verifying ChR2 expression, we next assessed whether

selective stimulation of cholinergic neurons in the IPV-GP

modulated heart rate and AV nodal conduction using optoge-

netics in ex vivo Langendorff-perfused hearts. A blue laser-

coupled optical

fi

ber was positioned for focal illumination of the

IPV-GP while cardiac electrical activity was simultaneously

recorded (Fig.

c, d). Optogenetic stimulation resulted in a

decrease in heart rate that was dependent on light pulse power,

frequency, and pulse width (Fig.

–

g and Supplementary Table 1)

but did not change the AV interval (35.7 ± 1.2 ms before

stimulation versus 35.9 ± 1.0 ms during stimulation) (Fig.

h).

The lack of change in AV nodal conduction suggests that

fi

bers

from this ganglion may pass through the AV node without

synapsing. In addition, stimulation prolonged the P wave

duration (9.3 ± 1.0 ms before stimulation versus 12.3 ± 2.4 ms

during stimulation) and caused P wave fractionation (Fig.

i, j). P

wave fractionation has been reported in humans following

administration of adenosine

, which mimics the effects of

acetylcholine released from cholinergic nerve terminals

During stimulations at higher frequencies, we occasionally

observed ectopic atrial rhythms (

=

3/6 mice) and even asystole

(

=

1/6 mice) (Fig.

d), demonstrating the profound effect of the

IPV-GP on the SA node and atrial function. The response to

stimulation was abolished by administration of the muscarinic

receptor antagonist atropine (

−

33.5 ± 11.0% with stimulation

before atropine versus

−

0.3 ± 0.2% with stimulation after

atropine) (Fig.

k and Supplementary Table 1), indicating that

the bradycardic response was indeed mediated by selective

stimulation of cholinergic neurons.

Since ChR2 is expressed in both preganglionic cholinergic

inputs to and postganglionic cholinergic neurons in the IPV-GP

Imaging using

confocal or lightsheet

microscopy

Automated nerve fiber

detection and 3D visualization

Derivation of microanatomical

data on nerve fibers

Perfusion and

fixation

Staining

Clearing

After

Before

iDISCO

PGP9.5

Ventral

Dorsal

Cardiac ganglion

Ventricle

Nerve diameter

Diameter (

Nerve orientation

Orientation

Fig. 1

Tissue clearing and computational pipeline to assess cardiac innervation.

Schematic of the clearing-imaging-analysis pipeline for assessing cardiac

innervation.

A whole heart (top) was rendered transparent (bottom) using the iDISCO protocol.

3D confocal projections of the ventral (1500

μ

mz-

stack) and dorsal side of a cleared heart (1200

μ

m z-stack) with PGP9.5 staining (gray). Inset 1 shows a maximum intensity projection (MIP) confocal

image of a cardiac ganglion. Inset 2 shows 785

μ

m-thick 3D projections of the entire left ventricular wall.

A semiautomated computational pipeline was

used to detect nerve

bers in the dorsal heart image from (

) and derive their diameter (

) and orientation (

). Insets show higher magni

cation images.

Scale bars are 2 mm (

), 1 mm (

(top),

(left),

(left)), and 100

μ

(bottom),

(right),

(right))

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09770-1

ARTICLE

NATURE COMMUNICATIONS

| (2019) 10:1944 | https://doi.org/10.1038/s41467-019-09770-1 | www.nature.com/naturecommunications

3

in ChAT-ChR2-eYFP mice, we assessed whether we could

stimulate only postganglionic cholinergic neurons in the IPV-

GP using optogenetics and still modulate heart rate. We

fi

rst

evaluated whether we could preferentially deliver transgenes to

peripheral cholinergic neurons in intrinsic cardiac ganglia rather

than central cholinergic neurons in the medulla using systemic

AAVs. We used AAV-PHP.S, a capsid variant that more

fi

ciently transduces the PNS and many visceral organs including

the heart, as compared to AAV9

. We packaged a Cre-

dependent genome that expresses eYFP from the ubiquitous

CAG promoter into AAV-PHP.S and systemically administered

the virus to ChAT-IRES-Cre transgenic mice. Three weeks later,

we evaluated eYFP expression in the medulla, vagus nerve, and

cardiac ganglia with GFP staining. Central cholinergic neurons in

the dorsal motor nucleus of the vagus nerve (DMV) (1.5 ± 0.9%

expressed GFP) and

fi

bers in the vagus nerve were weakly

transduced (Supplementary Fig. 3a

–

c). In contrast, we observed

robust transduction of peripheral cholinergic neurons in cardiac

ganglia (91.0 ± 1.5%) (Supplementary Fig. 3d and e), likely due to

the strong tropism AAV-PHP.S displays for the PNS over the

CNS. Further, GFP expression was highly speci

fi

c for ChAT

+

neurons in cardiac ganglia (100.0 ± 0.0%) (Supplementary Fig. 3e).

For functional studies, we packaged a Cre-dependent genome that

expresses ChR2-eYFP from the ubiquitous CAG promoter in

AAV-PHP.S, systemically administered the virus to ChAT-IRES-

Cre transgenic mice, and evaluated expression 5 weeks later. In

Langendorff-perfused hearts, we were able to optogenetically

stimulate postganglionic cholinergic neurons in the IPV-GP and

decrease heart rate (Supplementary Fig. 4). Taken together, our

anatomical and functional data establish an IPV-GP-SA node

Cardiac ganglion

Dense labeling in ChAT-IRES-Cre mouse

ssAAV-PHP.S:CAG-DIO-XFPs (3 × 10

vg total)

Sparse labeling in ChAT-IRES-Cre mouse

ssAAV-PHP.S:TRE-DIO-XFPs (3 × 10

vg total)

ssAAV-PHP.S:ihSyn1-DIO-tTA (1 × 10

vg)

Cardiac ganglion

SVC

tdTom

HCN4

Sparse labeling in ChAT-IRES-Cre mouse

ssAAV-PHP.S:TRE-DIO-tdTomato (1 × 10

vg)

ssAAV-PHP.S:ihSyn1-DIO-tTA (1 × 10

vg)

Dorsal atrium

SA node

IPV-GP

AV node

Tract-tracing

tdTom

HCN4

SA node

IPV-GP

AV node

Dorsal heart

One-component expression system for

cell type-specific dense labeling

Two-component expression system for

cell type-specific sparse labeling

Variable dose

tTA

High dose

Cre transgenic

mice

High dose

Cre transgenic

mice

SA node

IVC

SVC

PVs

AV node

IPV-GP

Fig. 2

AAV-based labeling and tracing of cholinergic neurons on the heart.

Schematic of the one-component expression system for cell type-speci

dense multicolor labeling (top). To densely label cholinergic neurons in cardiac ganglia, ChAT-IRES-Cre mice were systemically co-administered 3

Cre-

dependent vectors expressing either mRuby2, mNeonGreen, or mTurquoise2 from the ubiquitous CAG promoter (ssAAV-PHP.S:CAG-DIO-XFPs) (1 × 10

vector genomes (vg) each; 3 × 10

vg total). A MIP image of a densely labeled cardiac ganglion (bottom left). Inset shows a higher magni

cation image

(bottom right).

Schematic of the two-component expression system for cell type-speci

c, sparse multicolor labeling (top). Expression of XFPs is

dependent on cotransduction of an inducer in Cre-expressing cells. The dose of the inducer vector can be titrated to control extent of XFP labeling. To

sparsely label cholinergic neurons in cardiac ganglia, ChAT-IRES-Cre mice were systemically co-administered 3 Cre-dependent vectors expressing

XFPs

from the TRE-containing promoter (ssAAV-PHP.S:TRE-DIO-XFPs) (1 × 10

vg each; 3 × 10

vg total) and a Cre-dependent inducer vector expressing the

tTA from the ihSyn1 promoter (ssAAV-PHP.S:ihSyn1-tTA) (1 × 10

vg). A MIP image of a sparsely labeled cardiac ganglion (bottom left). Inset shows a

higher magni

cation image (bottom right).

To trace cholinergic

bers, presumably from cardiac ganglia, sparse labeling was performed by systemically

co-administering ssAAV-PHP.S:TRE-DIO-tdTomato at a high dose (1 × 10

vg) and ssAAV-PHP.S:ihSyn1-DIO-tTA at a lower dose (1 × 10

vg). Cartoon of

the dorsal heart depicting the orientation of images (left). A MIP image of the dorsal atrium with native tdTomato

uorescence (red) and HCN4 staining

(green) (middle). Fibers were traced with neuTube and overlaid on a grayscale MIP image (right). Orange

bers coursed along the right atrium (RA)

including the sinoatrial (SA) and atrioventricular (AV) nodes and blue

bers along the ventricles. Scale bars are 50

μ

) and 200

μ

). All images

were acquired on tissue collected 3 weeks after intravenous injection. IPV-GP inferior pulmonary vein-ganglionated plexus, IVC inferior vena cava

, LA left

atrium, LV left ventricle, PV pulmonary vein, RV right ventricle, SVC superior vena cava

ARTICLE

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09770-1

4

NATURE COMMUNICATIONS

| (2019) 10:1944 | https://doi.org/10.1038/s41467-019-09770-1 | www.nature.com/naturecommunications

circuit involved in heart rate regulation (Fig.

l). Furthermore, the

engineered AAV, AAV-PHP.S

, can be a powerful tool to dissect

out the roles of peripheral versus central circuits on organ

control.

Optogenetic versus electrical stimulation of the vagus nerve

Electrical vagus nerve stimulation (VNS) has been used in

numerous preclinical and clinical studies for the treatment of

cardiovascular diseases

and other conditions (e.g., rheumatoid

arthritis

). However, the relative contributions of vagal efferent

and afferent

fi

bers on cardiac function are not well understood

because conventional techniques do not allow for

fi

ber type-

speci

fi

c stimulation. To address this limitation, we examined

whether we could selectively stimulate efferent

fi

bers in the vagus

nerve using optogenetics. We also assessed whether there was a

difference between optogenetic and electrical VNS on heart rate

(Fig.

a), as a previous study showed that electrical stimulation of

motor nerves results in a non-orderly, non-physiological

recruitment of

fi

bers, with larger

fi

bers activated

fi

rst

. The

P wave duration (ms)

Pre-

stimulation

Stimulation

Pre-

stimulation

Stimulation

AV interval (ms)

IPV-GP

Langendorff preparation

EP catheter

Optical

fiber

ECG

Pre-atropine

stimulation

Post-atropine

stimulation

–80

–60

–40

–20

10 Hz

20 Hz

1 s

100 ms

Cardiac ganglion

Ventricle

PGP9.5

GFP

ChAT-ChR2-eYFP

GFP

PGP9.5

Dorsal heart

GFP

PGP9.5

SA node

IVC

SVC

PVs

IPV-GP

Dorsal heart

10 Hz

1 s

10 ms

100 150 200 250

–80

–60

–40

–20

Power (mW)

Heart rate (%)

–80

–60

–40

–20

Heart rate (%)

–80

–60

–40

–20

Heart rate (%)

Frequency (Hz)

0510

Pulse width (ms)

GFP+ & PGP9.5+/

GFP+

GFP+ & PGP9.5+/

PGP9.5+

100

Cardiac ganglion

neurons (%)

QRS

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09770-1

ARTICLE

NATURE COMMUNICATIONS

| (2019) 10:1944 | https://doi.org/10.1038/s41467-019-09770-1 | www.nature.com/naturecommunications

5

vagus and other autonomic nerves contain both motor and sen-

sory

fi

bers that vary in diameter and myelination

and non-

electrical techniques such as optogenetics are needed to study

their physiological role.

To con

fi

rm that ChR2-eYFP expression was limited to vagal

efferents in ChAT-ChR2-eYFP mice, we stained for GFP and

PGP9.5 in the nodose/jugular ganglion complex, which contains

the cell bodies of vagal sensory neurons, and the cervical vagus

Fig. 3

Ex vivo optogenetic stimulation of cholinergic neurons in the IPV-GP.

A 3D projection (1200

μ

m z-stack) of the dorsal side of a heart from a ChAT-

ChR2-eYFP mouse whole-mount stained with PGP9.5 (red) and GFP (green). Insets 1 and 2 show MIP images of a cardiac ganglion and the ventricle,

respectively. Blue dashed boxes indicate location of higher magni

cation images in blue boxes.

Percentage of cardiac ganglion neurons expressing GFP

and PGP9.5 over those expressing GFP or PGP9.5.

Langendorff-perfused hearts were used for optogenetic stimulation of cholinergic neurons in a cardiac

ganglion. A blue laser-coupled optical

ber was positioned for focal illumination of the IPV-GP (circle). A surface electrocardiogram (ECG) was recorded

with bath electrodes and intracardiac electrograms with an electrophysiology (EP) catheter.

Representative ECGs during stimulation (blue shading) at

10 Hz, 10 ms, and 221 mW (top) and 20 Hz, 10 ms, and 221 mW (bottom). Insets show the ECGs before and during stimulation.

–

Dose response

curves summarizing the effects of altering light pulse power (

), frequency (

), and pulse width (

) on heart rate.

Summary of the AV interval before and

during stimulation at 10 Hz and 10 ms (

=

0.1656,

=

0.8765).

Representative ECG during stimulation (blue shading) at 10 Hz, 10 ms, and 221 mW.

Insets show a single beat before and during stimulation, with gray boxes showing a higher magni

cation of the P wave.

The P wave duration before versus

during stimulation (

=

2.920, *

=

0.0330).

Summary of the heart rate response to stimulation before versus after atropine administration (

=

2.993,

=

0.0402).

Cartoon of the dorsal heart depicting the IPV-GP-SA node circuit.

=

6 mice (

–

) and 5 mice (

); mean ± s.e.m.; paired, two-tailed

-test. Scale bars are 1 mm (

(left),

) and 100

μ

(right))

BVNx cranial RVNS

Frequency (Hz)

Intact

BVNx rostral

Time to peak heart

rate response (s)

Electrical RVNS

abc

GFP

PGP9.5

GFP

PGP9.5

Nodose/jugular complex

Vagus nerve

Trachea

In vivo optical versus electrical

vagus nerve stimulation

1 s

300

400

500

600

700

Heart rate (bpm)

20 Hz intact RVNS

Electrical

Optical

ChAT-ChR2-eYFP

–80

–60

–40

–20

Frequency (Hz)

Heart rate (%)

Intact RVNS

Frequency (Hz)

RVNx caudal RVNS

–80

–60

–40

–20

Frequency (Hz)

BVNx caudal RVNS

RVNx cranial RVNS

–20

–15

–10

–5

–80

–60

–40

–20

Heart rate (%)

–20

–15

–10

–5

Frequency (Hz)

Fig. 4

In vivo optogenetic versus electrical stimulation of the vagus nerve.

Cartoon depicting optogenetic and electrical vagus nerve stimulation strategy

in ChAT-ChR2-eYFP mice. The right vagus nerve was surgically exposed in anesthetized mice and either light or electricity was used for stimulation.

MIP

images of the right nodose/jugular ganglion complex and vagus nerve whole-mount stained with PGP9.5 (red) and GFP (green).

Representative heart

rate responses during optogenetic versus electrical right vagus nerve stimulation (RVNS; magenta shading) at identical frequencies (20 Hz) and pul

widths (10 ms) in the intact state. The light pulse power for optogenetic stimulation was 57 mW and the current for electrical stimulation was 5

μ

Frequency response curves summarizing the effects of optogenetic versus electrical RVNS on heart rate in the intact state.

Frequency response curves

summarizing the effects of optogenetic versus electrical RVNS of the caudal end on heart rate following right vagotomy (RVNx) (

) and bilateral vagotomy

(BVNx) (

Frequency response curves summarizing the effects of optogenetic versus electrical RVNS of the cranial end on heart rate following RVNx

(

=

3.576, *

=

0.0232 at 10 Hz;

=

5.229, **

=

0.0064 at 20 Hz) (

) and BVNx (

=

8.588, **

=

0.0010 at 20 Hz) (

). In

–

insets show a

schematic of the stimulation protocol.

The time to peak heart rate response during electrical RVNS in the intact state versus of the cranial end following

BVNx (

=

6.335, **

=

0.0032).

=

8 mice (

), three mice (

), and

ve mice (

); mean ± s.e.m.; paired, two-tailed

-test. Scale bar is 200

μ

)

ARTICLE

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09770-1

6

NATURE COMMUNICATIONS

| (2019) 10:1944 | https://doi.org/10.1038/s41467-019-09770-1 | www.nature.com/naturecommunications

nerve (Fig.

b). eYFP was not detected in PGP9.5

+

cell bodies in

the nodose/jugular ganglion complex and was only present in a

subset of PGP9.5

+

vagal

fi

bers (

=

5 mice). After verifying

expression, we next performed functional studies in anesthetized

mice in which we positioned a laser-coupled optical

fi

ber for focal

illumination above and a hook electrode underneath the right

vagus nerve. In this context, optogenetic VNS activates only

efferent

fi

bers (GFP

+

), whereas electrical VNS presumably

activates subsets of efferent and afferent

fi

bers (PGP9.5

+

)in

the vagus nerve (Fig.

b). With both vagi intact, optogenetic and

electrical right VNS resulted in a similar decrease in heart rate

(Fig.

c and d); however, the slopes of the responses were

dramatically different (

=

6/8 mice) (Fig.

c), likely due to

differential

fi

ber recruitment

. The heart rate response to

optogenetic versus electrical stimulation of the caudal end of

the right vagus nerve was also similar following either right or

bilateral vagotomy (Fig.

e, f and Supplementary Table 2). In

contrast, optogenetic stimulation of the cranial end of the right

vagus nerve following either right or bilateral vagotomy did not

affect heart rate, whereas electrical stimulation surprisingly

resulted in a decrease in heart rate at 10 Hz (

−

6.5 ± 1.8% versus

−

0.1 ± 0.1% following right vagotomy for electrical versus

optogenetic stimulation) and 20 Hz (

−

9.9 ± 1.8% versus

−

0.1 ±

0.1% following right vagotomy and

−

3.2 ± 0.4% versus

−

0.2 ±

0.1% following bilateral vagotomy for electrical versus optoge-

netic stimulation) (Fig.

g, h and Supplementary Table 2). The

decrease in heart rate to electrical stimulation of the cranial end of

the right vagus nerve following right vagotomy was also greater

than that following bilateral vagotomy at 20 Hz (

=

3.123,

=

0.0354, paired, two-tailed

-test) (Fig.

g, h, red line and

Supplementary Table 2), suggesting that the response following

ipsilateral vagotomy was in part due to a vagal afferent-mediated

increase in parasympathetic efferent out

fl

ow through the intact

contralateral vagus nerve. Furthermore, the decrease in heart rate

to electrical stimulation of the cranial end of the right vagus nerve

following bilateral vagotomy (Fig.

h, red line and Supplementary

Table 2) indicates that vagal-afferents mediate a decrease in

sympathetic efferent out

fl

ow, since both vagi were transected. In

addition, there was an increased latency to peak heart rate

response with electrical stimulation of the cranial end of the right

vagus nerve following bilateral vagotomy compared to that of the

intact right vagus nerve (8.7 ± 0.4 ms versus 3.9 ± 0.5 ms) (Fig.

i),

which further supports that vagal afferents cause withdrawal of

sympathetic tone

. Taken together, our data suggest that

optogenetic stimulation selectively activates vagal efferents in

ChAT-ChR2-eYFP mice and that vagal afferents act centrally to

(1) increase parasympathetic efferent out

fl

ow and (2) decrease

sympathetic efferent out

fl

ow to the heart.

Location and optogenetic stimulation of cardiac-projecting

noradrenergic neurons in the stellate ganglia

. The sympathetic

nervous system, along with the parasympathetic nervous system,

precisely regulates heart rate in normal physiology. To anato-

mically and functionally dissect noradrenergic neurons that form

a circuit with the SA node, we used a retrograde neuronal tracer

and an optogenetic approach. Noradrenergic nerve

fi

bers densely

innervate the heart as shown by staining for tyrosine hydroxylase

(TH), the rate-limiting enzyme in norepinephrine synthesis

(Fig.

a). To identify the location of cardiac-projecting sympa-

thetic neurons, we injected the retrograde neuronal tracer CTB

conjugated to Alexa Fluor 488 into the heart (Fig.

b). The

ab c

0.5

1.0

1.5

2.0

0.5

1.0

x

(mm)

y

(mm)

RSG

0.5

1.0

1.5

2.0

2.5

0.5

1.0

1.5

x

(mm)

y

(mm)

LSG

0.5

1.0

1.5

×10

–3

CTB+ cells/

CTB-488

Ansa

T2G

MCG

Left paravertebral chain ganglia

Heart

T2G

CTB-488

Paravertebral

chain

Dorsal heart

Fig. 5

Cardiac-projecting neurons in stellate ganglia (SG) are clustered craniomedially.

A3Dprojection(1200

μ

m z-stack) of the dorsal side of a C57BL/6J

mouse heart whole-mount stained with TH (green).

Cartoon depicting cholera toxin subunit B (CTB)-Alexa Fluor 488 (CTB-488) injections into the heart to

retrogradely trace neurons in the paravertebral chain ganglia that project to the heart.

A MIP image of the left paravertebral chain from the middle cervical

ganglion (MCG) to the second thoracic ganglion (T2G) showing the location of CTB-488

+

neurons that project to the heart.

Summary heat map of the right

stellate ganglion (RSG) and the left stellate ganglion (LSG).

=

5mice(

). Scale bars are 1 mm (

). CTB injections were performed in C57BL/6J mice

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09770-1

ARTICLE

NATURE COMMUNICATIONS

| (2019) 10:1944 | https://doi.org/10.1038/s41467-019-09770-1 | www.nature.com/naturecommunications

7

majority of CTB

+

neurons were located in the stellate ganglia of

the paravertebral chain, with fewer labeled neurons in the middle

cervical and second thoracic (T2) ganglia (Fig.

c and Supple-

mentary Movie 7). On average, 236 ± 39 neurons were labeled in

the right and 261 ± 34 neurons labeled in the left stellate ganglion

(

=

5 mice). Heat maps of the right stellate ganglion (RSG) and

left stellate ganglion show that cardiac-projecting sympathetic

neurons are clustered in the craniomedial aspect (Fig.

d) and

suggest that these ganglia may have a viscerotopic organization.

Next, we assessed whether we could selectively stimulate

noradrenergic neurons in the paravertebral chain using optoge-

netics and modulate heart rate (Fig.

a). We expressed ChR2 in

noradrenergic neurons by crossing transgenic TH-IRES-Cre mice

with reporter mice containing a Cre-dependent ChR2-eYFP allele

(offspring from this cross are subsequently referred to as TH-

ChR2-eYFP mice). ChR2-eYFP expression in stellate ganglion

neurons was con

fi

rmed by staining for GFP and TH (Fig.

b). The

majority of GFP

+

neurons were TH

+

(98.7 ± 0.4%) and all TH

+

neurons were GFP

+

(100.0 ± 0.0%) (Fig.

c). After verifying

expression, we performed functional studies in open-chest

anesthetized mice in which we positioned a laser-coupled optical

fi

ber for focal illumination above the craniomedial RSG or right

T2 ganglion (Fig.

a). Optogenetic stimulation of the craniomedial

RSG resulted in a frequency- and pulse width-dependent increase

in heart rate (Fig.

–

f and Supplementary Table 3). Although a

small number of cardiac-projecting neurons were located in the

T2 ganglion (Fig.

c), there was no heart rate response to

stimulation of this ganglion and/or preganglionic sympathetic

fi

bers coursing through this region (0.1 ± 0.1% with right T2

ganglion stimulation versus 9.5 ± 1.8% with RSG stimulation)

(Fig.

g and Supplementary Table 3). The response to craniome-

dial RSG stimulation was abolished by administration of the

β

adrenergic receptor antagonist propranolol (5.6 ± 1.4% with

stimulation before propranolol versus 0.1 ± 0.1% with stimulation

after propranolol) (Fig.

h and Supplementary Table 3), indicating

that the tachycardic response was indeed mediated by selective

GFP+ &TH+/

GFP+

TH+ & GFP+/

TH+

100

Stellate ganglion

neurons (%)

Pre-propranolol

RSGS

Post-propranolol

RSGS

–5

def

TH-ChR2-eYFP

400

450

500

550

600

Heart rate (bpm)

RSGS

1 s

Heart

T2G

In vivo optical stimulation of

paravertebral ganglia

T2G

GFP

Right paravertebral chain ganglia

Ventral heart

RSG

SA node

SVC

RSGS

RT2GS

–5

Frequency (Hz)

Heart rate (%)

0510

Pulse width (ms)

Fig. 6

In vivo optogenetic stimulation of noradrenergic neurons in the RSG.

Cartoon depicting optogenetic SG and T2G stimulation strategy in TH-ChR2-

eYFP mice. The right paravertebral chain was surgically exposed in anesthetized mice and light was used for stimulation.

A MIP image of the right

paravertebral chain whole-mounted stained with TH (red) and GFP (green) and showing the SG and the T2G. Inset shows single-plane images of the SG.

Blue dashed boxes indicate location of higher magni

cation images in blue boxes.

Percentage of stellate ganglion neurons expressing GFP and TH over

those expressing GFP or TH.

Representative heart rate response during 10 Hz, 10 ms, and 126 mW craniomedial RSG stimulation (RSGS).

Dose

response curves summarizing the effects of altering craniomedial RSGS frequency (

) and pulse width (

) on heart rate.

Summary of the heart rate

response to craniomedial RSGS versus right T2G stimulation (RT2GS) (

=

5.435, **

=

0.0016).

Summary of the heart rate response to craniomedial

RSGS before versus after propranolol administration (

=

3.951, *

=

0.0289).

Cartoon of the ventral heart depicting the craniomedial RSG-SA node

circuit

=

6 mice (

), seven mice (

ve mice (

), and four mice (

); mean ± s.e.m.; paired, two-tailed

-test. Scale bars are 200

μ

(left) and 50

μ

(

(right)

ARTICLE

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09770-1

8

NATURE COMMUNICATIONS

| (2019) 10:1944 | https://doi.org/10.1038/s41467-019-09770-1 | www.nature.com/naturecommunications

stimulation of noradrenergic neurons. Taken together, our

anatomical and functional data establish a craniomedial RSG-SA

node circuit involved in heart rate regulation (Fig.

i).

Discussion

We developed a clearing-imaging-analysis pipeline to visualize

innervation of whole hearts and employed a multi-technique

approach to dissect fundamental parasympathetic and sympa-

thetic neural circuits involved in heart rate regulation. We report

several novel

fi

ndings: (1) cholinergic neurons in the IPV-GP and

noradrenergic neurons in the craniomedial RSG project to the SA

node and modulate its function; (2) the evoked cardiac response

to optogenetic versus electrical stimulation of the vagus nerve

displays different temporal characteristics; and (3) vagal afferents

enhance parasympathetic and reduce sympathetic efferent out-

fl

ow to the heart via central mechanisms.

Despite advances in tissue clearing

and imaging

techniques

, high-resolution, 3D datasets of global cardiac

innervation do not exist. We show, for the

fi

rst time, innervation

of entire cleared mouse hearts with cardiac ganglia located

around the pulmonary veins and a dense network of nerve

fi

bers

throughout the myocardium. To analyze these data, we developed

a semiautomated computational pipeline to detect nerve

fi

bers

and to measure microanatomical features such as diameter and

orientation. These analytical tools and resulting measurements

are needed to build a reference atlas of cardiac innervation and

for quantitative descriptions of innervation in healthy versus

diseased states such as MI. Following MI, innervation around the

infarct scar and of remote regions of the heart is altered

and this neural remodeling can modulate the arrhythmia sub-

strate

. Understanding changes in innervation post-MI can

provide new insights into arrhythmia mechanisms. Furthermore,

our clearing-imaging-analysis pipeline can be readily applied to

assess innervation of other visceral organs including endogen-

ously and virally labeled tissues.

It is well known that the parasympathetic and sympathetic

nervous systems are critical for heart rate regulation. The SA node

and conduction system are densely innervated

, and stimulation

of the vagus nerve

, stellate ganglia

, and noradrenergic

fi

bers

modulates heart rate. However, the precise wiring of

the underlying neural circuits has not been delineated. We used a

novel sparse AAV labeling system

and an optogenetic approach

to anatomically and functionally characterize cholinergic neurons

that regulate heart rate. Although we were unable to directly

visualize synapses, we identi

fi

ed cholinergic

fi

bers, presumably

from cardiac ganglia, that coursed along the SA node, the AV

node, and the ventricles. Selective optogenetic stimulation of

cholinergic neurons in the IPV-GP modulated heart rate, con-

sistent with a recent study that stimulated cholinergic

fi

bers in the

right atrium also using optogenetics

. Previous studies showed

that electrical stimulation of pulmonary vein ganglia results in

biphasic heart rate responses (initial bradycardia followed by

tachycardia)

. However, since electrical stimulation is non-

speci

fi

c, it is dif

fi

cult to interpret whether the biphasic response

was due to activation of a mixed population of neurons contained

in cardiac ganglia (i.e., parasympathetic, sympathetic, and sen-

sory)

and/or pass through

fi

bers. Our

fi

ndings demonstrate an

IPV-GP-SA node circuit and highlight the importance of using

techniques such as optogenetics, which confer cell type-speci

fi

city, to dissect cardiac neural circuity. Furthermore, electrical and

optogenetic techniques (using traditional transgenic and AAV-

based approaches for ChR2 delivery) stimulate both central

preganglionic inputs to and postganglionic neurons in cardiac

ganglia. Therefore, we used a novel engineered AAV, AAV-PHP.

, that has a strong tropism for the PNS to preferentially deliver

ChR2 to postganglionic cholinergic neurons on the heart rather

than preganglionic cholinergic neurons in the medulla. Future

studies of peripheral neural circuits should use AAV-PHP.S and

other engineered AAVs

to dissect the role of central versus

peripheral neuronal populations on organ function.

To map noradrenergic neurons that regulate heart rate, we

used a retrograde neuronal tracer and optogenetic approach. A

previous study in canines using horseradish peroxidase showed

that sympathetic postganglionic neurons that innervate the heart

are primarily located in the middle cervical ganglia of the para-

vertebral chain

. However, we report, using CTB and con

fi

with optogenetic stimulation, that the stellate ganglia have a

viscerotopic organization with cardiac-projecting neurons clus-

tered in the craniomedial aspect, consistent with a study in cats

In addition to the heart, the stellate ganglia project to many other

thoracic structures, including the sweat glands in the forepaw

the lung and trachea

, the esophagus

, and brown fat

. Char-

acterizing stellate ganglia target innervation and cell-type speci-

fi

cation is an area of ongoing investigation

that is of interest

from a developmental, physiological, and therapeutic perspective.

Current understanding of the role of vagal efferent and afferent

fi

bers on cardiac function is largely based on studies using elec-

trical stimulation, which is non-speci

fi

c and results in non-

orderly, non-physiological recruitment of

fi

bers

. Electrical sti-

mulation of the vagus nerve typically activates large-diameter

myelinated A

fi

bers, followed by medium-diameter myelinated B

fi

bers and then small-diameter unmyelinated C

fi

bers

. In vivo

we report that optogenetic stimulation of motor

fi

bers in the

vagus nerve results in a heart rate response that has a slower onset

than electrical stimulation of motor and sensory

fi

bers, likely due

to differential

fi

ber recruitment

. Our

fi

ndings suggest that non-

electrical techniques such as optogenetics are needed to char-

acterize neural control of cardiac physiology. Although there was

robust expression of ChR2-eYFP as con

fi

rmed by immunohis-

tochemistry, we cannot rule out that incomplete expression or

optical capture affected our results. We also show that activation

of vagal afferents decreases heart rate by enhancing para-

sympathetic efferent out

fl

ow and reducing sympathetic efferent

out

fl

ow centrally, consistent with a prior study showing that

global activation of vagal afferents in Vglut2-ChR2 mice and

selective activation of Npy2r-ChR2 vagal afferents causes a pro-

found bradycardia

. In support of our functional data, anato-

mical tracing studies have previously shown that cardiac vagal

afferent neurons in the nodose/jugular ganglion complex project

to neurons in the nucleus tractus solitarii of the medulla

. These

neurons then project to the nucleus ambiguus and the DMV in

the medulla to modulate parasympathetic efferent out

fl

and

to the paraventricular nucleus of the hypothalamus to modulate

sympathetic efferent out

fl

. Future studies aimed at identify-

ing cardiac-speci

fi

c vagal efferent and afferent

fi

bers, similar to

those performed in the lungs

and the gastrointestinal system

are needed to better understand vagal control of cardiac phy-

siology and to design next-generation VNS therapies.

Overall, our data highlight the complexity of cardiac neural

circuitry and demonstrate that a multi-technique approach is

needed to delineate circuit wiring. Understanding the neural

control of organ function in greater detail is critical as neuro-

modulation therapies are emerging as promising approaches to

treat a wide range of diseases. Tools such as optogenetics and

AAVs are already providing new scienti

fi

c insights into the

structure and function of peripheral neural circuits

. A com-

bination of these approaches will help disentangle neural control

of autonomic physiology and enable a new era of targeted neu-

romodulation approaches.

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09770-1

ARTICLE

NATURE COMMUNICATIONS

| (2019) 10:1944 | https://doi.org/10.1038/s41467-019-09770-1 | www.nature.com/naturecommunications

9