Optogenetic activation of cholinergic neurons in thePPT or LDT induces REM sleepChrista J. Van Dorta,b,c,1, Daniel P. Zachsa,b,c, Jonathan D. Kennya,b,c, Shu Zhengb, Rebecca R. Goldblumb,c,d,Noah A. Gelwana,b,c, Daniel M. Ramosb,c, Michael A. Nolanb,c,d, Karen Wangb,c, Feng-Ju Wengb,e, Yingxi Linb,e,Matthew A. Wilsonb,c, and Emery N. Browna,b,d,f,1

aDepartment of Anesthesia, Critical Care, and Pain Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114; and bDepartmentof Brain and Cognitive Sciences, cPicower Institute for Learning and Memory, eMcGovern Institute for Brain Research, fHarvard-MIT Division of HealthSciences and Technology, and dInstitute for Medical Engineering and Science, Massachusetts Institute of Technology, Cambridge, MA 02139

Contributed by Emery N. Brown, December 3, 2014 (sent for review September 19, 2014; reviewed by Helen A. Baghdoyan and H. Craig Heller)

Rapid eye movement (REM) sleep is an important component ofthe natural sleep/wake cycle, yet the mechanisms that regulateREM sleep remain incompletely understood. Cholinergic neuronsin the mesopontine tegmentum have been implicated in REM sleepregulation, but lesions of this area have had varying effects on REMsleep. Therefore, this study aimed to clarify the role of cholinergicneurons in the pedunculopontine tegmentum (PPT) and laterodor-sal tegmentum (LDT) in REM sleep generation. Selective optoge-netic activation of cholinergic neurons in the PPT or LDT duringnon-REM (NREM) sleep increased the number of REM sleep episodesand did not change REM sleep episode duration. Activation ofcholinergic neurons in the PPT or LDT during NREM sleep wassufficient to induce REM sleep.

rapid eye movement sleep | acetylcholine | optogenetics |mesopontine tegmentum | mouse

Rapid eye movement (REM) sleep is tightly regulated, yet themechanisms that control REM sleep remain incompletely

understood. Early pharmacological and unit recording studiessuggested that ACh was important for REM sleep regulation (1,2). For example, injection of cholinergic drugs into the dorsalmesopontine tegmentum reliably induced a state very similar tonatural REM sleep in cats (3–6). Unit recordings from thecholinergic areas of the mesopontine tegmentum revealed cellsthat were active during wakefulness and REM sleep, as well asneurons active only during REM sleep (7–13). Electrical stimu-lation of the laterodorsal tegmentum (LDT) in cats increased thepercentage of time spent in REM sleep (14), and activation ofthe pedunculopontine tegmentum (PPT) in rats induced wake-fulness and REM sleep (15). If cholinergic PPT and LDT neu-rons are necessary for REM sleep to occur, as the early studiessuggest, then lesioning the PPT or LDT should decrease REMsleep. In cats, lesions of the PPT and LDT do disrupt REM sleep(16, 17), but lesions in rodents have had little effect on REMsleep or increased REM sleep (18–22). Additionally, c-fos studieshave found very few cholinergic cells activated under high-REMsleep conditions. When c-fos–positive cholinergic neurons in thePPT and LDT are found to correlate with the percentage of REMsleep, they still account for only a few of the total cholinergic cellsin the area (23). Juxtacellular recordings of identified cholinergicneurons in the LDT found these cells had wake and REM activefiring profiles, with the majority firing the highest during REMsleep (13). These discrepancies have led to alternative theories ofREM sleep regulation, where cholinergic neurons do not playa key role (18, 19, 23, 24 and reviewed in 25, 26).The PPT and LDT are made up of heterogeneous populations

of cells, including distinct populations of cholinergic, GABAergic,and glutamatergic neurons (27–29). Many GABAergic neuronsare active during REM sleep, as indicated by c-fos (23), and bothGABAergic and glutamatergic neurons have been found withmaximal firing rates during REM sleep in the LDT and medialPPT (13). To distinguish the differential roles of each cell type in

REM sleep regulation, a method that can modulate specific celltypes in the behaving animal is needed. Optogenetics now pro-vides this ability to target specific subpopulations of neuronsand control them with millisecond temporal resolution (30).Therefore, we aimed to determine the role of cholinergicneurons in the PPT and LDT in REM sleep regulation usingoptogenetics.

ResultsChannelrhodopsin Expression Was Selective to Cholinergic Neurons inthe PPT and LDT and Functional in Vitro. Mice expressing chan-nelrhodopsin (ChR2) conjugated to YFP under the cholineacetyltransferase (ChAT) promoter and WT littermates wereused (six PPT ChAT-ChR2+, five PPT ChAT-ChR2−, five LDTChAT-ChR2+, and six LDT ChAT-ChR2− mice per group,and one ChAT-ChR2+ patch-clamp mouse) (31). Immunohisto-chemistry for ChAT confirmed that ChR2 was expressed selec-tively in cholinergic neurons in the PPT and LDT. Quantificationrevealed that 96.1% (2,636 of 2,742) and 94.0% (1,992 of 2,119)of ChAT-positive neurons were also positive for ChR2-YFP inthe PPT and LDT, respectively. Colocation of ChAT and ChR2-YFP in this range is consistent with previous reports demon-strating selective expression in the cortex (100%), striatum(100%), globus pallidus (100%), and medial habenula (98.2%)for the same mouse strain (31). No ChR2-YFP–only neuronswere found (Fig. 1). Fig. 2A demonstrates that LDT neurons

Significance

Rapid eye movement (REM) sleep is a critical component ofrestful sleep, yet the mechanisms that control REM sleep areincompletely understood. Brainstem cholinergic neurons havebeen implicated in REM sleep regulation, but heterogeneouscell types in the area have made it difficult to determine thespecific role of each population, leading to a debate about theimportance of cholinergic neurons. Therefore, we selectivelyactivated brainstem cholinergic neurons to determine theirrole in REM sleep regulation. We found that activation ofcholinergic neurons during non-REM sleep increased thenumber of REM sleep episodes but not REM sleep duration.Our data demonstrate that brainstem cholinergic neurons areimportant modulators of REM sleep and clarify their role inREM sleep initiation.

Author contributions: C.J.V., M.A.W., and E.N.B. designed research; C.J.V., D.P.Z., S.Z.,R.R.G., N.A.G., D.M.R., M.A.N., K.W., F.-J.W., and Y.L. performed research; C.J.V., D.P.Z.,J.D.K., and S.Z. analyzed data; C.J.V. and E.N.B. wrote the paper; and F.-J.W. and Y.L.performed patch-clamp studies.

Reviewers included: H.A.B., University of Tennessee; and H.C.H., Stanford University.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1To whom correspondence may be addressed. Email: vandortc@mit.edu or enb@neurostat.mit.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1423136112/-/DCSupplemental.

584–589 | PNAS | January 13, 2015 | vol. 112 | no. 2 www.pnas.org/cgi/doi/10.1073/pnas.1423136112

(one shown, n = 2 tested) had reliable rapid-onset action potentialsfollowing 5-ms light pulses. Fig. 2B shows the ability of a cell tofollow the light pulses at 5 Hz over 2 s. The cell never missed thefirst light pulse, but the probability of a spike occurring de-creased with progressive light pulses. The likelihood of a spikeoccurring varied as a function of the starting membrane poten-tial. If the cell started near −55 mV, it was more likely to fire andfollow the light pulses. If the cell started closer to −60 mV, itmissed more light pulse-induced action potentials. The latencyfrom start of the light pulse to the beginning of the actionpotential ranged from 4.6 ms for the first light pulse to 20 msfor the 10th light pulse in the series.

Activation of Cholinergic Neurons in the PPT and LDT Increases theProbability of REM Sleep. Mice expressing ChR2 in cholinergicneurons (ChR2+) and their WT littermates (ChR2−) wereimplanted with bilateral fiber optics for the PPT or LDT, as wellas EEG and electromyogram (EMG) electrodes. REM sleep wasclassified by high levels of theta (5–9 Hz) in the EEG and nomuscle tone. EEG and EMG traces demonstrate that opto-genetic activation of cholinergic neurons in the PPT during non-REM (NREM) sleep induced REM sleep (Fig. 3). The prob-ability of REM sleep over time increased between ChR2+ mice(n = 11) and ChR2− mice (n = 11) for all stimulations for boththe PPT and LDT (Fig. 4). REM sleep probability was signifi-cantly higher [nonoverlapping confidence intervals (CIs)] for 30–60 s beyond the stimulation for ChR2+ PPT and LDT 60-s and80-s stimulations compared with ChR2− 60-s and 80-s stim-ulations. REM sleep probability for the PPT 180-s stimulation

reached its peak of 0.30 (95% CI: 0.26–0.35) at 2 min and 25 s,and then started to decline. REM sleep probability for the LDT180-s stimulation significantly increased between 2 and 3 min(nonoverlapping CIs) compared with the ChR2− 180-s stimula-tion. Group data from 22 mice and an average of 173 stimulationsper condition (range: 99–226 stimulations per condition; Table 1)show that optogenetic activation of the PPT or LDT increasedREM sleep (Fig. 5 A and D). PPT 180-s stimulation increasedREM sleep more than the LDT 180-s stimulation for ChR2+

mice (difference of means = 15.72 s, 99% CI: 3.90–27.51 s).NREM sleep decreased between ChR2+ mice and ChR2− micefor all PPT stimulations and the 60-s LDT stimulation (Fig. 5 Band E). Wakefulness did not change for either PPT or LDTstimulation, except for a small increase in wakefulness for the 60-sPPT stimulation (Fig. 5 C and F). Inferences of the differencesbetween groups were calculated using 99% CIs (details are pro-vided in Table 2 and SI Materials and Methods, Data Analysis).

Increase in REM Sleep Is Due to More REM Sleep Episodes. The in-crease in REM sleep occurred by increasing the number of REMsleep episodes (percentage of stimulations that induced REMsleep; Fig. 6 A and D) but not the duration of REM sleep epi-sodes (Fig. 6 B and E). The induced REM sleep was electro-physiologically similar to natural REM sleep. Power spectralanalysis of the EEG during induced REM sleep was not signifi-cantly different from the power spectra during natural REM sleepin the same ChR2+ mice (Fig. 6 C and F), as indicated by over-lapping 95% CIs.

Fiber Optics Were Localized to the PPT and LDT. Fig. 7 summarizesthe results of the histological analyses demonstrating that the tipsof the fiber optics were localized just above the PPT or LDTbilaterally. Placing the tips of the fibers at the top of the nucleiensured as complete as possible activation of the entire nucleibased on Yizhar et al.’s (32) calculations of light dissipation inthe brain. Stained sections were compared with the mouse brainatlas (33) to identify their final location (Fig. 7A). The averagePPT stereotaxic coordinates were 4.80 mm posterior to bregma,1.14 mm lateral to the midline, and −3.15 mm ventral. The av-erage LDT stereotaxic coordinates were 5.03 mm posterior tobregma, 0.59 mm lateral to the midline, and −3.03 mm ventral.

B

AChAT ChR2-YFP Merge

ChAT ChR2-YFP Merge

PP

TLD

T

100 µm

100 µm

20 µm

20 µm

Fig. 1. ChAT and ChR2-YFP were colocalized in the PPT (A; bregma, −4.84mm)and LDT (B; bregma, −5.02 mm). Confocal images of coronal brain sectionsstained for ChAT and ChR2-YFP in the PPT (A, Top) and LDT (B, Top) areshown. (Magnfication: 25×.) Confocal images show robust colocalization ofChAT to the cell bodies and ChR2-YFP to the cell membrane in the PPT(A, Bottom) and LDT (B, Bottom). (Magnification: 40×.)

200ms

10pA

Time (s)

Tria

l #

1

3

5

7

9

11

0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4

Firin

g P

roba

bilit

y

0

0.5

10.75

0.25

20m

V

A

B

Fig. 2. Blue light-induced reliable action potentials in cholinergic LDTneurons expressing ChR2-YFP. (A) Current-clamp (Top) and voltage-clamp(Bottom) recordings of a ChR2-YFP–expressing cholinergic LDT neuron dur-ing photostimulation (5-ms blue light pulse at 5 Hz for 2 s, blue bars). (B)Twelve 10-pulse trials for one neuron, showing the ability of the neuron tofollow the light pulse (Top) and the probability of an action potential foreach light pulse (Bottom).

Van Dort et al. PNAS | January 13, 2015 | vol. 112 | no. 2 | 585

NEU

ROSC

IENCE

Wake and REM Sleep Stimulations. Eighty-second (5 ms at 5 Hz)stimulations during wakefulness caused a small increase in theprobability of wakefulness for the 4 min after the beginning ofthe stimulation [0.92 (95% CI: 0.91–0.93, n = 167) for ChAT-ChR2+ mice compared with 0.90 (95% CI: 0.90–0.91, n = 75) forChAT-ChR2− mice; difference of the means = 1.6 (95% CI: 0.9–2.4)]. Eighty-second (5 ms at 5 Hz) stimulations during REMsleep did not change the duration of REM sleep between ChAT-ChR2+ mice 73.2-s (95% CI: 62.9–83.8, n = 87), ChAT-ChR2−

mice 71.5-s (95% CI: 57.3–86.7, n = 51), and baseline days withno stimulation for 82.8 s in both ChAT-ChR2+ and ChAT-ChR2− mice (95% CI: 77–88.5, n = 333). Lack of a significantdifference was calculated using the difference of the means and98.3% (Bonferroni correction for three comparisons) CIs ofthe difference.

DiscussionPPT and LDT Cholinergic Neurons Are Important for REM SleepInitiation but Not REM Sleep Maintenance. Activation of choliner-gic neurons in the PPT or LDT during NREM sleep increasedthe probability of REM sleep (Fig. 4) and the number of REMsleep episodes (Fig. 5) but not the duration of REM sleep epi-sodes (Fig. 6). REM sleep probability for the PPT 180-s stimu-lation reached its peak of 0.30 (95% CI: 0.26–0.35) at 2 m and25 s, and then started to decline (Fig. 4C). As has been suggestedbefore (34), there are likely different mechanisms controllingREM sleep initiation vs. REM sleep maintenance. Activation ofcholinergic cells in the basal forebrain during REM sleep wasable to increase the duration of REM sleep episodes (35),whereas stimulation of the PPT during REM sleep in this studywas not able to prolong REM sleep episodes. The present studysuggests that cholinergic PPT and LDT neurons are importantmodulators of REM sleep initiation but not REM sleep main-tenance. PPT 180-s stimulation increased REM sleep more thanthe LDT 180-s stimulation for ChR2+ mice (difference of means15.72 s, 99% CI: 3.90–27.51 s). This difference suggests that

EE

G

0 1 2 3 4 5Time (min)

EM

G5 Hz Stimulation

WAKE

NREM

REM

EE

G

Freq

uenc

y (H

z)

0

10

20

500 µV

A

C

Pow

er (d

B)

50

60

70B

Fig. 3. Representative example of optogenetic induction of REM sleep inChAT-ChR2+ mice showing the unprocessed EEG and EMG traces. (A) Hyp-nogram of wake, NREM sleep, and REM sleep shows the changing sleepstates across the stimulation. The black bar indicates the timing of opticalstimulation (5-ms pulse at 5-Hz for 60 s). Sleep-wake states were determinedby analyzing the power spectra of the EEG (B) and the raw EEG and EMG (C).

0

0.1

0.2

0.3

0.4

0

0.1

0.2

0.3

0.4

Time (min)0 1 2 3 4

0

0.1

0.2

0.3

0.4

Time (min)0 1 2 3 4

Pro

babi

lity

of R

EM

sle

ep

ChR2+ChR2-

PPT LDT

Pro

babi

lity

of R

EM

sle

epP

roba

bilit

y of

RE

M s

leep

A

B

C

D

E

F

Fig. 4. Probability of REM sleep as a function of time (mean and 95% CI).Probability of REM sleep in ChR2+ (blue) and ChR2− (red) mice for PPTstimulation across 60-s (A), 80-s (B), and 180-s (C) stimulations (stimulationtime marked by black bars). Probability of REM sleep for LDT stimulationacross 60-s (D), 80-s (E), and 180-s (F) stimulations (stimulation time markedby black bars).

Table 1. Number of stimulations per condition

Stimulation length, s ChR2+ ChR2−

PPT 60 150 15180 168 146

180 171 99LDT 60 196 181

80 226 219180 163 183

0

10

20

30

40

50

0

30

60

90

120

150

180

0

5

10

15

20

25

30

D ChR2+ LDTChR2- LDT

E

60 80 180Stimulation Length (s)

F

RE

M (s

)M

ean

and

99%

CI A

NR

EM

(s)

Mea

n an

d 99

% C

I B

60 80 180

Wak

e (s

)M

ean

and

99%

CI

Stimulation Length (s)

C

ChR2+ PPTChR2- PPT

*

*

**

**

*

*

*

*

Fig. 5. Group data demonstrate that optogenetic activation of choliner-gic neurons in the PPT and LDT increased REM sleep. Activation of the PPT(A) increased REM sleep during the stimulation window for 60-s, 80-s, and180-s stimulations, whereas activation of the LDT (D) increased REM sleepfor the 60-s and 80-s stimulations. PPT activation decreased NREM sleep forall stimulations (B), whereas LDT activation only decreased NREM sleep forthe 60-s stimulation (E ). (C and F) Wakefulness during the stimulationwindow did not change, except for a small increase for the 60-s PPTstimulation. *Significant differences between groups with 99% confidence(details of the numbers used to generate the inference are provided inTable 2).

586 | www.pnas.org/cgi/doi/10.1073/pnas.1423136112 Van Dort et al.

although both areas contain cholinergic neurons, the PPT maybe better situated in the REM sleep circuitry to control REMsleep initiation.

Implications for the Role of PPT and LDT Cholinergic Neurons in REMSleep Modulation. The PPT and LDT contain multiple sub-populations of REM-on neurons, which have made it difficult todistinguish the differential roles of each subtype. C-fos studiesshow both cholinergic and GABAergic REM-on neurons in rats(23), and juxtacellular recordings have found both GABAergicand glutamatergic REM-on neurons (13). Interestingly, thisjuxtacellular study did not find any cholinergic REM-on neurons.Instead, the identified cholinergic neurons were active duringboth wake and REM sleep, with the majority firing the highestduring REM sleep (13). The cholinergic agonist carbachol in-duced prolonged REM sleep by inhibiting presumably cholin-ergic REM-on neurons and exciting presumably noncholinergicREM-on neurons in the cat pons (10). The present study dem-onstrated that selective optogenetic activation of cholinergicneurons in the PPT and LDT induced REM sleep in mice.Monoamines are thought to inhibit REM sleep, and cholinergic

neurons in the LDT are inhibited by serotonin in both rat (36) andguinea pig (37) brain slices. The selective 5-hydroxytryptamine 1A(5-HT1A) receptor agonist 8-hydroxy-2-(di-n-propylamino)tetralin(8-OH-DPAT) selectively inhibits presumably cholinergic PPTREM-on neurons but not wake-on/REM-on neurons in cats (8).Therefore, it was surprising that Grace et al. (20) found that localPPT delivery of 8-OH-DPAT increased REM sleep in rats. Intheory, this manipulation should inhibit cholinergic REM-oncells, which would have been expected to decrease REM sleep.In light of the multiple cell types that are now known to exhibita REM-on firing pattern in the LDT and PPT (13), it is possiblethat serotonin inhibited GABAergic or glutamatergic REM-onneurons. In addition, it is possible that serotonin inhibition ofcholinergic PPT occurs through other receptor subtypes or thatselectivity for the 5-HT1A receptor by 8-OH-DPAT is lost athigh concentrations. Also, 5HT1A serotonin receptor mRNAhas only been found in GABAergic but not cholinergic neurons

in the PPT of the mouse (38). Taken together, these data sug-gest that it is possible Grace et al. (20) inhibited GABAergic orglutamatergic REM-on neurons to get an increase in REM sleep,whereas we selectively activated cholinergic REM-on neurons toget an increase in REM sleep. Interestingly, both manipulationsresulted in a change in the number of REM sleep episodes and didnot change REM sleep episode duration, suggesting that the PPTis involved in REM sleep initiation but not REM sleep mainte-nance. Additional studies are needed to tease apart the differen-tial roles of cholinergic, GABAergic, and glutamatergic REM-oncells in the PPT and LDT. There are about 50% fewer cholinergiccells compared with GABAergic cells in the PPT and LDT (29), inaddition to intermingled glutamatergic cells. The fact that acti-vation of a small number of cholinergic cells can elicit a strongincrease in REM sleep adds to the strength of the evidence thatthose cholinergic neurons are important for REM sleep initiation.Basal forebrain cholinergic neurons also contribute to the

regulation of sleep and wakefulness. Carbachol injection into thebasal forebrain of cats increased wakefulness (39). A recentstudy optogenetically activated cholinergic neurons in the basalforebrain during wake, NREM sleep, and REM sleep (35). Ac-tivation of basal forebrain cholinergic neurons during NREMsleep induced transitions to wake and REM sleep. However,these states were shorter in duration than natural wake and REMsleep episodes, and the stimulation did not influence whetherwake or REM sleep was induced. Stimulation of the cholinergicbasal forebrain during wakefulness in Han et al. (35) decreasedNREM sleep. In the present study, stimulation of cholinergicneurons in the PPT during wakefulness caused a small increasein the probability of wakefulness. Optogenetic stimulation of

0

30

60

90

120

150

A

60 80 180

% S

timul

atio

ns th

at

indu

ced

RE

MM

ean

and

99%

CI

RE

M E

piso

de D

urat

ion

(s)

Mea

n an

d 99

% C

I

Stimulation Length (s)

B

D

60 80 180Stimulation Length (s)

E

C F

Frequency (Hz)

Pow

er (d

B)

0 5 10 15 20 25-30

-26

-22

-18

-14

-10ChR2+ Induced REMChR2+ Natural REM

Frequency (Hz)0 5 10 15 20 25

ChR2+ LDTChR2- LDT

**

0

10

20

30

40

50

60 ChR2+ PPTChR2- PPT

*

**

Fig. 6. Optogenetic activation of cholinergic neurons in the PPT and LDTincreased the number of REM sleep episodes but not REM sleep episodeduration. The percentage of stimulations that induced REM sleep pernumber of REM sleep episodes increased for the PPT (A) and LDT (D). Theduration of REM sleep episodes that started within the stimulation windowdid not change for the PPT (B) or LDT (E). The power spectra were not dif-ferent for PPT-induced (C) or LDT-induced (F) REM sleep compared withnoninduced REM sleep in the same ChR2+ mice. *Significant differencesbetween groups with 99% confidence (details of the numbers used togenerate the inference are provided in Table 2).

Table 2. Difference of ChR2+ and ChR2− means and 99% CIs

Stimulationlength, s

Differenceof means

LowerCI

UpperCI Significance

PPTREM 60 4.2 1.5 7.2 Increase

80 8.0 3.0 13.0 Increase180 24.5 11.6 37.0 Increase

NREM 60 −8.4 −12.1 −4.8 Decrease80 −11.2 −16.5 −5.7 Decrease

180 −30.9 −43.8 −16.2 DecreaseWake 60 4.2 1.6 7.2 Increase

80 3.2 0.0 6.3 No change180 6.4 −2.2 13.9 No change

Stimulationsthat inducedREM, %

60 0.2 0.1 0.3 Increase80 0.2 0.1 0.3 Increase

180 0.4 0.2 0.5 IncreaseREM episode

duration60 8.2 −49.2 54.2 No change80 21.5 −8.1 50.5 No change

180 −17.7 −65.0 29.9 No changeLDT

REM 60 3.9 1.1 6.9 Increase80 4.8 1.1 8.4 Increase

180 5.1 −4.7 15.6 No changeNREM 60 −4.1 −7.6 −0.6 Decrease

80 −2.6 −6.9 1.7 No change180 −3.1 −14.9 8.7 No change

Wake 60 0.2 −2.3 2.8 No change80 −2.2 −5.1 0.5 No change

180 −1.9 −10.3 6.4 No changeStimulations

that inducedREM, %

60 0.1 0.0 0.2 Increase80 0.1 0.0 0.2 Increase

180 0.1 −0.1 0.2 No changeREM episode

duration60 12.3 −33.4 52.0 No change80 4.1 −32.3 39.2 No change

180 1.2 −26.3 28.8 No change

Van Dort et al. PNAS | January 13, 2015 | vol. 112 | no. 2 | 587

NEU

ROSC

IENCE

cholinergic basal forebrain neurons during REM sleep increasedthe duration of REM sleep episodes, whereas stimulation of PPTcholinergic neurons during REM sleep in the present study didnot prolong REM sleep. Basal forebrain carbachol injectiondecreased the amount of pontine carbachol injection-inducedREM sleep-like state (39), suggesting that the basal forebrainarousal promoting system interacts with the pontine REMpromoting system. Taken together, these findings suggest thatthere may be a forebrain component to REM sleep mainte-nance. Han el al. (35) concluded that the basal forebraincholinergic neurons are responsible for terminating NREMsleep. This interpretation is complementary to our interpretationthat brainstem cholinergic neurons are important for initiatingREM sleep.

Implications for the Role of PPT and LDT Cholinergic Neurons inWakefulness. Given the wakefulness and REM sleep firingprofile of cholinergic PPT and LDT neurons (7–13), we wouldhave expected a larger effect on wakefulness. Instead, we onlyfound two conditions that slightly increased wakefulness: the60-s PPT stimulations during NREM sleep and the 80-s PPTstimulations during wakefulness. The PPT and LDT are wellpositioned to activate the cortex during both wakefulness andREM sleep. However, when activated during NREM sleep, thebrainstem cholinergic neurons preferentially shift an animal toREM sleep vs. wakefulness.

Possible Explanations for the Delay in REM Sleep Onset AfterStimulation. The probability of REM sleep builds over thetime course of the stimulation. A few possibilities exist for whythe REM sleep transition is not an immediate switch with shortlatency: (i) cholinergic tone must build up to a certain leveland meet a network threshold for the transition to occur; (ii)incomplete activation of the nucleus due to limited spread ofthe light out of the fiber optic or nonoptimal placement of thefibers; (iii) only one cholinergic brainstem center was activatedat a time due to the small size of the mouse brain and limited

space to target both the PPT and LDT simultaneously; and (iv)cholinergic cells in the PPT and LDT are modulators of REMsleep and exert their effect via projections to other REM-onareas, such as the sublaterodorsal nucleus. Slice recordingssupport cholinergic modulation of the sublaterodorsal nucleus,where carbachol activated spinally projecting sublaterodorsalnucleus neurons (40). Animals with fibers that were optimallypositioned over the PPT and had the highest fiber trans-mittance had the strongest REM induction effect.

Limitations. The transgenic mice used in the present study havebeen found to express extra copies of the vesicular ACh transportergene and have increased cholinergic tone (41). Behaviorally,these mice have prolonged motor endurance and impairedattention and memory. These mice were reported to be moreactive at night than WT mice but had no difference in activitylevel during the day. In our hands, we recorded sleep during theday and found that the ChR2+ mice had 4% REM sleep com-pared with 3.7% REM sleep in the ChR2− mice, and this dif-ference was not statistically different. All of our experimentswere performed during the day, when locomotor activity andpercentage of REM sleep were the same between experimentalgroups; therefore, we think our results are representative of thetrue effect. The transgenic mice used in the present studyexpressed ChR2 in cholinergic neurons throughout the brain.Hence, it is possible that activation of cholinergic fibers of pas-sage from other brain regions to the PPT or LDT contributed toREM sleep induction. Future studies that selectively inhibitcholinergic neurons in the PPT and LDT of nonhypercholinergicmice are needed to determine if cholinergic neurons are neces-sary for REM sleep generation.

ConclusionsThe present findings demonstrate that activation of cholinergicneurons in the PPT or LDT during NREM sleep is sufficient toincrease REM sleep in mice. The induced REM sleep state closelyresembles natural REM sleep. Selectively increasing cholinergic

C

A

PPT

-4.72mm

-4.84mm

LDT

-4.96mm1mm

-5.20mm

B

LDT

LDT

D-5.02mm

PPT

PPT

Fig. 7. Tips of the fiber optics were localized to the top of the PPT or LDT. (A) Location of each fiber optic tip above the PPT (●) and LDT (■) is shown oncoronal sections modified from the mouse brain atlas (33). Blue indicates ChR2+ mice, and red indicates ChR2− mice. (B) Vertical lines on the sagittal sectionshow the span of the coronal sections. Coronal DAPI-stained sections show representative tracts over the PPT (C) and LDT (D) left by 1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI; pink)–coated fiber optics.

588 | www.pnas.org/cgi/doi/10.1073/pnas.1423136112 Van Dort et al.

tone in the PPT or LDT increases the number of REM sleepepisodes but not the duration of REM sleep episodes. There-fore, cholinergic neurons in the PPT and LDT remain potentmodulators of REM sleep initiation. This modulation of REMsleep expression may occur via activation of other REM-onneuron populations, such as the pontine reticular formation andsublaterodorsal nucleus.

Materials and MethodsAdult male mice (n = 23) expressing ChR2 under the ChAT promoter (stockno. 014546; The Jackson Laboratory) (31) and their WT littermates wereimplanted with EEG and EMG electrodes and bilateral fiber optics in thePPT or LDT. All experiments were approved by the Massachusetts Instituteof Technology Committee on Animal Care. The mice were placed in a re-cording chamber while EEGs and EMGs were recorded for 6–8 h. Bluelight from a laser was used to stimulate the cholinergic PPT or LDT neu-rons optogenetically during NREM sleep. Stimulations were 60 s, 80 s, or180 s long, separated by at least 1 min, and each experiment consisted of

∼25 stimulations. EEGs and EMGs were used to score sleep manually.Variables calculated included the amount of time spent in NREM sleep,REM sleep, and wake within the stimulation; REM sleep episode duration;and percentage of REM sleep induced. The probability of REM sleep overthe time course of the stimulation was also plotted. Fiber placement andspecific expression of ChR2 in cholinergic neurons were confirmed byhistology. Patch-clamp of a ChAT-ChR2+ mouse LDT slice was performedto confirm light-induced action potentials. Statistical inferences weredetermined by calculating 95% CIs for each group. The decision rule wasto reject the null hypothesis if zero is not in the 95% CI of the differencebetween the groups. Details of the experiment procedures are provided inthe SI Materials and Methods.

ACKNOWLEDGMENTS. This study was supported by NIH Grant DP1-OD003646 (to E.N.B.), NIH Grant TR01-GM104948 (to E.N.B. and M.A.W.),and NIH Grant T32-HL07901 (to C.J.V.); a Massachusetts General Hospital Ex-ecutive Committee on Research Fellowship (to C.J.V.); and the Department ofAnesthesia, Critical Care, and Pain Medicine, Massachusetts General Hospital.

1. Hobson JA, McCarley RW, Wyzinski PW (1975) Sleep cycle oscillation: Reciprocal dis-charge by two brainstem neuronal groups. Science 189(4196):55–58.

2. McCarley RW (2007) Neurobiology of REM and NREM sleep. Sleep Med 8(4):302–330.3. Amatruda TT, 3rd, Black DA, McKenna TM, McCarley RW, Hobson JA (1975) Sleep

cycle control and cholinergic mechanisms: Differential effects of carbachol injectionsat pontine brain stem sites. Brain Res 98(3):501–515.

4. Baghdoyan HA, Rodrigo-Angulo ML, McCarley RW, Hobson JA (1987) A neuroana-tomical gradient in the pontine tegmentum for the cholinoceptive induction of de-synchronized sleep signs. Brain Res 414(2):245–261.

5. George R, Haslett WL, Jenden DJ (1964) A cholinergic mechanism in the brainstemreticular formation: Induction of paradoxical sleep. Int J Neuropharmacol 3:541–552.

6. Vanni-Mercier G, Sakai K, Lin JS, Jouvet M (1989) Mapping of cholinoceptive brain-stem structures responsible for the generation of paradoxical sleep in the cat. ArchItal Biol 127(3):133–164.

7. el Mansari M, Sakai K, Jouvet M (1989) Unitary characteristics of presumptive cho-linergic tegmental neurons during the sleep-waking cycle in freely moving cats. ExpBrain Res 76(3):519–529.

8. Thakkar MM, Strecker RE, McCarley RW (1998) Behavioral state control through dif-ferential serotonergic inhibition in the mesopontine cholinergic nuclei: A simulta-neous unit recording and microdialysis study. J Neurosci 18(14):5490–5497.

9. Kayama Y, Ohta M, Jodo E (1992) Firing of ‘possibly’ cholinergic neurons in the ratlaterodorsal tegmental nucleus during sleep and wakefulness. Brain Res 569(2):210–220.

10. Sakai K, Koyama Y (1996) Are there cholinergic and non-cholinergic paradoxicalsleep-on neurones in the pons? Neuroreport 7(15-17):2449–2453.

11. Grant SJ, Highfield DA (1991) Extracellular characteristics of putative cholinergicneurons in the rat laterodorsal tegmental nucleus. Brain Res 559(1):64–74.

12. Steriade M, Datta S, Paré D, Oakson G, Curró Dossi RC (1990) Neuronal activities inbrain-stem cholinergic nuclei related to tonic activation processes in thalamocorticalsystems. J Neurosci 10(8):2541–2559.

13. Boucetta S, Cissé Y, Mainville L, Morales M, Jones BE (2014) Discharge profiles acrossthe sleep-waking cycle of identified cholinergic, GABAergic, and glutamatergicneurons in the pontomesencephalic tegmentum of the rat. J Neurosci 34(13):4708–4727.

14. Thakkar M, Portas C, McCarley RW (1996) Chronic low-amplitude electrical stimula-tion of the laterodorsal tegmental nucleus of freely moving cats increases REM sleep.Brain Res 723(1-2):223–227.

15. Datta S, Siwek DF (1997) Excitation of the brain stem pedunculopontine tegmentumcholinergic cells induces wakefulness and REM sleep. J Neurophysiol 77(6):2975–2988.

16. Webster HH, Jones BE (1988) Neurotoxic lesions of the dorsolateral pontomesence-phalic tegmentum-cholinergic cell area in the cat. II. Effects upon sleep-waking states.Brain Res 458(2):285–302.

17. Shouse MN, Siegel JM (1992) Pontine regulation of REM sleep components in cats:Integrity of the pedunculopontine tegmentum (PPT) is important for phasic eventsbut unnecessary for atonia during REM sleep. Brain Res 571(1):50–63.

18. Lu J, Sherman D, Devor M, Saper CB (2006) A putative flip-flop switch for control ofREM sleep. Nature 441(7093):589–594.

19. Boissard R, et al. (2002) The rat ponto-medullary network responsible for paradoxicalsleep onset and maintenance: A combined microinjection and functional neuroana-tomical study. Eur J Neurosci 16(10):1959–1973.

20. Grace KP, Liu H, Horner RL (2012) 5-HT1A receptor-responsive pedunculopontinetegmental neurons suppress REM sleep and respiratory motor activity. J Neurosci32(5):1622–1633.

21. Inglis WL, Semba K (1997) Discriminable excitotoxic effects of ibotenic acid, AMPA, NMDAand quinolinic acid in the rat laterodorsal tegmental nucleus. Brain Res 755(1):17–27.

22. Petrovic J, Ciric J, Lazic K, Kalauzi A, Saponjic J (2013) Lesion of the pedunculopontinetegmental nucleus in rat augments cortical activation and disturbs sleep/wake statetransitions structure. Exp Neurol 247:562–571.

23. Maloney KJ, Mainville L, Jones BE (1999) Differential c-Fos expression in cholinergic,monoaminergic, and GABAergic cell groups of the pontomesencephalic tegmentumafter paradoxical sleep deprivation and recovery. J Neurosci 19(8):3057–3072.

24. Verret L, Léger L, Fort P, Luppi PH (2005) Cholinergic and noncholinergic brainstemneurons expressing Fos after paradoxical (REM) sleep deprivation and recovery. Eur JNeurosci 21(9):2488–2504.

25. Fuller PM, Saper CB, Lu J (2007) The pontine REM switch: past and present. J Physiol584(Pt 3):735–741.

26. Luppi PH, et al. (2006) Paradoxical (REM) sleep genesis: the switch from an aminergic-cholinergic to a GABAergic-glutamatergic hypothesis. J Physiol Paris 100(5-6):271–283.

27. Wang HL, Morales M (2009) Pedunculopontine and laterodorsal tegmental nucleicontain distinct populations of cholinergic, glutamatergic and GABAergic neurons inthe rat. Eur J Neurosci 29(2):340–358.

28. Boucetta S, Jones BE (2009) Activity profiles of cholinergic and intermingledGABAergic and putative glutamatergic neurons in the pontomesencephalic teg-mentum of urethane-anesthetized rats. J Neurosci 29(14):4664–4674.

29. Ford B, Holmes CJ, Mainville L, Jones BE (1995) GABAergic neurons in the rat ponto-mesencephalic tegmentum: Codistribution with cholinergic and other tegmental neu-rons projecting to the posterior lateral hypothalamus. J Comp Neurol 363(2):177–196.

30. Zhang F, Aravanis AM, Adamantidis A, de Lecea L, Deisseroth K (2007) Circuit-breakers: Optical technologies for probing neural signals and systems. Nat Rev Neu-rosci 8(8):577–581.

31. Zhao S, et al. (2011) Cell type–specific channelrhodopsin-2 transgenic mice for opto-genetic dissection of neural circuitry function. Nat Methods 8(9):745–752.

32. Yizhar O, Fenno LE, Davidson TJ, Mogri M, Deisseroth K (2011) Optogenetics in neuralsystems. Neuron 71(1):9–34.

33. Paxinos G, Franklin K (2007) The Mouse Brain in Stereotaxic Coordinates (Academic,San Diego), 3rd Ed.

34. Luppi PH, Clément O, Fort P (2013) Paradoxical (REM) sleep genesis by the brainstem isunder hypothalamic control. Curr Opin Neurobiol 23(5):786–792.

35. Han Y, et al. (2014) Selective activation of cholinergic basal forebrain neurons inducesimmediate sleep-wake transitions. Curr Biol 24(6):693–698.

36. Luebke JI, et al. (1992) Serotonin hyperpolarizes cholinergic low-threshold burstneurons in the rat laterodorsal tegmental nucleus in vitro. Proc Natl Acad Sci USA89(2):743–747.

37. Leonard CS, Llinás R (1994) Serotonergic and cholinergic inhibition of mesopontinecholinergic neurons controlling REM sleep: An in vitro electrophysiological study.Neuroscience 59(2):309–330.

38. Bonnavion P, Bernard JF, Hamon M, Adrien J, Fabre V (2010) Heterogeneous distri-bution of the serotonin 5-HT(1A) receptor mRNA in chemically identified neurons ofthe mouse rostral brainstem: Implications for the role of serotonin in the regulationof wakefulness and REM sleep. J Comp Neurol 518(14):2744–2770.

39. Baghdoyan HA, Spotts JL, Snyder SG (1993) Simultaneous pontine and basal forebrainmicroinjections of carbachol suppress REM sleep. J Neurosci 13(1):229–242.

40. Weng FJ, et al. (2014) Carbachol excites sublaterodorsal nucleus neurons projecting tothe spinal cord. J Physiol 592(Pt 7):1601–1617.

41. Kolisnyk B, et al. (2013) ChAT-ChR2-EYFP mice have enhanced motor endurance butshow deficits in attention and several additional cognitive domains. J Neurosci 33(25):10427–10438.

Van Dort et al. PNAS | January 13, 2015 | vol. 112 | no. 2 | 589

NEU

ROSC

IENCE

Supporting InformationVan Dort et al. 10.1073/pnas.1423136112SI Materials and MethodsAnimals. Twenty-three adult male transgenic mice expressingChR2 conjugated to YFP under the ChAT promoter (stock no.014546; The Jackson Laboratory) and WT littermates were used(six PPT ChAT-ChR2+, five PPT ChAT-ChR2−, five LDTChAT-ChR2+, and six LDT ChAT-ChR2− mice per group, andone ChAT-ChR2+ patch-clamp mouse). Line 6 of these ChAT-ChR2 mice was characterized for colocalization of ChAT andChR2-YFP in the cortex (100%), striatum (100%), globus pal-lidus (100%), and medial habenula (98.2%) by Zhao et al. (1).Mice were bred in-house and genotyped for the YFP-containingtransgene by sending ear tissue to Transnetyx. All mice werekept on a 12:12-h light/dark cycle (lights on at 7:00 AM, lights offat 7:00 PM), with ad libitum access to food and water. All animalprocedures were reviewed and approved by the MassachusettsInstitute of Technology Committee on Animal Care.

Surgery. Mice were anesthetized with 3% (vol/vol) isofluraneanesthesia in oxygen and placed in a stereotaxic frame (DavidKopf Instruments). An incision was made in the skin, and cra-niotomies were made above the target regions for PPT bregma(−4.72 mm anterior posterior and ±1.25 mm lateral to themidline) and for LDT bregma (−5.02 mm anterior posterior and±0.5 mm lateral to the midline). Bilateral fiber optics (200-μminner diameter, 0.22 N.A.; Doric Lenses) coated with 1′-dio-ctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchloratewere lowered into the brain 3.25 mm for the PPT and 3.15 mmfor the LDT so that the tip of the fiber was at the top of the PPTor LDT. Craniotomies were also made for EEG electrodes(0.005-inch stainless steel; A-M Systems) over the prefrontalcortex and somatosensory cortex, as well as for six anchor screwsand a ground screw (0.7-mm diameter, 2 mm long; AntrinMiniature Specialties). EMG electrodes (0.002-inch seven-stranded stainless steel; A-M Systems) were placed in the nuchalmuscle. EEG and EMG electrodes were connected to an eight-channel electrode interface board (Neuralynx). Anchor screws,electrodes, an electrode interface board, and fibers were adheredto the skull with dental acrylic. Animals were singly housed aftersurgery and had a minimum of 1 wk to recover before experi-ments began. Each experiment was separated by at least 3 d.

Optogenetic Stimulation and Sleep Recordings. Animals were con-ditioned to handling and the recording chamber after surgery.Experiments started at least 1 wk after surgery to allow for re-covery, and began at the beginning of the light cycle between 7:00and 9:00 AM. The mice behaved freely while EEGs (filtered 0.5–500 Hz) and EMGs (filtered 10–500 Hz) were recorded for 6–8 husing a 64-channel Neuralynx Digital recording system ata sampling rate of 254.344 Hz (Neuralynx). A patchcord (DoricLenses) was connected to the implanted fiber optic with a ce-ramic sleeve to join the 1.25-mm metal ferrules on each end ofthe fibers. Blue light was generated with a 200-mW MBL-H473-nm laser with a PSU-H-FDA power supply (Opto Engine).Stimulation parameters were controlled using the analog outputof a Multichannel Systems stimulus generator STG4000 (ALAScientific Instruments), and the start and end of each pulse weremarked by a transistor–transistor logic (TTL) pulse into theNeuralynx recording system. The laser power output was mea-sured using a PM100D power meter (Thor Labs). The tip of thepatch cord measured 30 mW with continuous light. Due to lightloss at the connections, the fibers transmitted about 50% of theinitial continuous light, so the power at the tip of the fiber was

∼15 mW. According to Yizhar et al. (2), you need 1–5 mW/mm2

to activate ChR2. We wanted to activate as much of the PPT andLDT as possible, so we targeted the middle of each nucleus in theanterior-to-posterior plane, which spans ∼0.7 mm, and placed thefiber at the top of each nucleus, which has a depth of ∼0.8 mm.Blue light attenuates in brain tissue at a measured rate and can becalculated using the tool at www.optogenetics.org/calc. With ourparameters of 473 mm of light; 200-μm diameter, 0.22 N.A. fiberoptics; and 15 mW of power output, we have 477 mW/mm2 im-mediately out of the tip and 9.5 mW/mm2 at 0.8 mm below the tipof the fiber, so we should be able to activate the entire depth ofthe nuclei as well as the anterior-to-posterior span of the nuclei.To mimic the natural firing rate of these neurons during REM

sleep, the stimulation frequency was based on published firingrates (3, 4) for the PPT and LDT across sleep and wakefulness inthe cat, where presumably cholinergic REM-on neurons fired at∼5 Hz during REM sleep. In our pilot multiunit in vivo recordingstudies, light pulses shorter than 5 ms did not induce reliableaction potentials. Therefore, the final values used for thesestudies were 5-ms pulses at 5 Hz. The duration of stimulationsused in these experiments was based on pilot behavioral studies,which found that 60-s stimulations increased the percentage ofREM sleep but not the duration of REM sleep. We chose 60-sstimulations to start with because based upon the literature,a 60-s stimulation should not cause damage to the brain due toheat (2) and it was a physiologically relevant length of time forREM sleep to occur. To determine if continued activation ofcholinergic neurons could prolong the REM sleep episode du-ration, we next tried 180-s stimulations. These stimulations in-creased the percentage of REM sleep even more than the 60-sstimulations but still did not increase the REM sleep episodeduration. The peak REM sleep probability in the 180-s stim-ulations occurred at 80 s, so next we tested if 80 s of lightstimulation was sufficient to generate the same amount of REMsleep as the 180-s stimulations. Experiments consisted of allow-ing the animals to adjust to the recording chamber for ∼2 h.Then, once the animal naturally went into NREM sleep for atleast 60 s, the light pulses were initiated. Each animal had ∼25stimulations per 6- to 8-h session, where each stimulation wasseparated by at least 1 min. Each stimulation was consideredindependent because it was separated in time. Stimulations frommultiple animals were compiled to yield the group data. Ex-periment days were separated by at least 3 d. The experimentsreported in this study used four optogenetic stimulation con-ditions in random order: baseline sleep (no stimulation) and60-s, 80-s, or 180-s stimulations (5-ms pulses at 5 Hz). EEG,EMG, and videotapes were recorded and used to score sleep.

Wake and REM Stimulations. In a subset of animals, we alsostimulated during wake and REM sleep with the same param-eters (5-ms pulses at 5 Hz for the middle-duration 80 s) ona separate experiment day from the NREM sleep stimulations.Wake stimulations began when the animal had been awake for atleast 60 s. REM sleep stimulations began when the animal hadbeen in REM sleep for 5–10 s. For wake stimulations, theprobability of wakefulness was calculated for the 4 min after thebeginning of the stimulation between ChAT-ChR2+ PPT miceand ChAT-ChR2− PPT mice. For REM sleep stimulations,REM sleep duration was compared between baseline dayswhere no stimulation occurred, ChAT-ChR2+ PPT mice, andChAT-ChR2− PPT mice.

Van Dort et al. www.pnas.org/cgi/content/short/1423136112 1 of 3

Patch-Clamp Recordings: Acute Slice Preparation. Mice were anes-thetized with isoflurane and decapitated. The brains were re-moved and immediately immersed in carbogenated (95% O2, 5%CO2) ice-cold cutting solution containing 105 mM N-methyl-D-glucamine, 2.5 mM KCl, 1.24 mM NaH2PO4, 10 mM MgCl2,0.5 mM CaCl2, 26 mM NaHCO3, 15 mM glucose, and 1 mM Na-ascorbate at an osmolarity of 300 mOsm. The pH was adjusted to7.3 with HCl. Brains were then rapidly blocked, and 350-μmcoronal slices were cut in the same solution with a vibratingblade microtome (VT1200; Leica). Slices containing the PPTand LDT were transferred to an incubation chamber filled withcarbogenated warm (32 °C) cutting solution for 10 min and then50% cutting solution/50% artificial cerebral spinal fluid (ACSF)(32 °C) for 20 min for recovery. Slices were then stored ina holding chamber filled with carbogenated room temperature(∼23 °C) ACSF containing 119 mM NaCl, 2.5 mM KCl, 1.24 mMNaH2PO4, 1.3 mM MgCl2, 2.5 mM CaCl2, 26 mM NaHCO3, and10 mM glucose at an osmolarity of 300 mOsm (pH 7.3) for atleast 1 h before being used for patch experiments.

In Vitro Electrophysiology.After recovery, slices were transferred toa recording chamber perfused with carbogenated room tempera-ture (∼23 °C) ACSF at a flow rate of 2 mL/min. A borosilicateglass pipette (tip resistance of 3–6 MΩ) was filled with internalsolutions containing 130 mM K-gluconate, 10 mM KCl, 1 mMMgCl2, 10 mM Hepes, 0.2 mM EGTA, 4 mM Mg-ATP, and 0.5mM Na-GTP at an osmolarity of 290 mOsm (pH 7.25). Photo-currents were recorded in voltage-clamp mode with membranepotential held at −70 mV, and photoevoked spikes were recordedin current-clamp mode with a holding potential at around −55mV. Cholinergic ChR2-YFP–expressing cells were identified byYFP expression, and cells were photostimulated by 470-nm, 5-ms light pulses at 5 Hz for 2 s. Data were acquired with anAxon Multiclamp 700B amplifier and a Digidata 1440 digitizer(Molecular Devices). Data were analyzed using Axon Clampfit.Recordings with access resistance greater than 25 MΩ or withchanges in access resistance greater than 15% were discarded.

Histology.At the end of all experiments, the optical fiber positionswere verified by postmortem histological analyses. Animals wereperfused transcardially with PBS, followed by 10% (vol/vol) neutralbuffered formalin. The brains were postfixed in 10% neutralbuffered formalin overnight. Brains were sliced at 50 μm usinga Leica VT1000 S vibratome (Leica Microsystems). Confirmationof specific expression in the targeted neuronal population wasdone with immunohistochemical staining for ChAT (goat anti-ChAT, 1:200 dilution, catalog no. AB144P; Millipore), a marker ofcholinergic neurons (Fig. 1), and ChR2-YFP as marked by YFPcolocalization (rabbit anti-GFP Alexa Fluor 488 conjugate, 1:400dilution, catalog no. A21311; Invitrogen). The secondary antibodywas donkey anti-goat conjugated to Alexa 568 (1:200 dilution,catalog no. A21432; Invitrogen). High-resolution images weretaken with a Zeiss 710 laser scanning confocal fluorescent mi-croscope (Keck Imaging Facility, Whitehead Institute, Massa-chusetts Institute of Technology). The expression of ChR2-YFP incholinergic neurons in the ChAT-ChR2-YFP line 6 strain hasbeen well characterized in many other brain regions and is con-sistent between animals (1). To document expression of ChR2-YFP in cholinergic neurons of the PPT and LDT of this mousestrain, we counted multiple 50-μm sections spanning the entirePPT and LDT of three animals using the cell counter plug-in ofFiji/ImageJ (Image Processing and Analysis in Java) (5). Toidentify fiber placement, images were taken with a Zeiss AxioImager M2 fluorescent microscope. Confirmation of optical fiberplacement in the correct brain region was done by comparingimages with the mouse brain atlas (6).

Data Analysis.Sleep scoring. Wakefulness, NREM sleep, and REM sleep werescored visually in 2-s epochs using Spike2 (CED), followingstandard mouse sleep scoring parameters (7), by the same ex-perienced nonblinded scorer. Briefly, wake was classified by ac-tivated EEG and muscle tone. NREM sleep was classified byhigh delta power (0.5–4 Hz) in the EEG and decreased muscletone. REM sleep was classified by high theta (5–9 Hz) in theEEG and no muscle tone. A videotape was used to help classifyany epochs that were not clear from the EEG and EMG.Making inferences by constructing 99% bootstrap CIs. To assess thesignificance of the differences between groups, we used a boot-strap procedure to construct CI data. The bootstrap is especiallyuseful in situations like ours, in which empirical distributions canbe easily constructed and used to draw random (bootstrap)samples. The bootstrap procedure is carried out by drawing re-peated random samples from the empirical distribution, esti-mating the quantity of interest from each sample, and thusconstructing its bootstrap distribution, which is the Monte Carloestimate of the probability density of the quantity of interest.Statistical inferences can be made about this quantity of interestbecause the bootstrap distribution is an estimate of the truedistribution of the quantity of interest.For example, if the quantity of interest is the difference in REM

sleep time between the ChAT-ChR2+ and ChAT-ChR2− mice,the bootstrap distribution provides an estimate of the probabilitydensity of this difference. The 99% CI for the true value of thisdifference can be constructed by taking the 0.5th and the 99.5thpercentiles of the bootstrap distribution. We report 99% CIs toconduct our inferences because, unlike P values, the CIs estimatethe quantity of interest and its associated uncertainty. In con-trast, the P value only provides an assessment of how likely theobserved quantity of interest is to have occurred by chance if theassumed null hypothesis is true. As such, the P value answersa binary question. It is well known that a 95% CI can be used toconduct a test of the null hypothesis against a two-sided alter-native at the 5% level of significance. The decision rule is toreject the null hypothesis if zero is not in the 95% CI of thedifference or with the Bonferroni correction 99% (1 to 0.05/5)based on comparing five variables: NREM sleep, REM sleep,and wake time within the stimulation; REM sleep episode du-ration; and percentage of REM sleep induced. Furthermore, ifzero lies outside the CI, the magnitude of the difference betweenthe two groups can be appreciated by assessing the distancebetween zero and the bounds of the CIs; the greater the distance,the greater is the effect.Table 2 shows the difference of themeans, CI of the differences,

and resulting significance for all data with statistical inferences.Bootstrap procedure.

i) Draw n samples with replacement from XChAT+ (1), . . .,XChAT+(n).

ii) Draw m samples with replacement from XChAT− (1), . . .,XChAT−(m).

iii) Compute the mean of n values (x) and the mean of mvalues (y).

iv) Save x in the bootstrap vector BChAT+, and save y in thebootstrap vector BChAT−.

v) Compute BChAT+ − BChAT− = dChAT+ − ChAT−.vi) Repeat steps 1 through 5 for a total of 10,000 times.vii) Sort in ascending order the bootstrap vectors BChAT+,

BChAT−, and dChAT+ − ChAT−.viii) Index the 50th and 9,950th values of each group to get the

lower 99% confidence bound and upper 99% confidencebound, respectively.

Spectrogram. The spectrogram of the EEG was calculated using themtspecgramc function of Chronux, an open-source MATLAB

Van Dort et al. www.pnas.org/cgi/content/short/1423136112 2 of 3

(MathWorks) software package for analyzing neural data (chronux.org) (8), using a moving 2.0-s window and a 0.1-s step size. A time-frequency spectrum was generated with a half-bandwidth of 1 Hz,a frequency band from 0 to 25 Hz, and three tapers. The spec-trum values were converted to decibels.Probability of REM sleep. The probability of REM sleep was cal-culated by marking each 2-s state epoch with a 1 for REM sleep ora 0 for NREM sleep and wake. Two hundred forty seconds (120epochs) starting at the beginning of the stimulation were ana-lyzed. This binary vector was arrayed in an n by 120 matrix, where120 was the number of 2-s epochs and n was the total number ofstimulations. The binary data from each epoch (column) werepooled from all n stimulations to create a binary time series. Theprobability of REM sleep was calculated by finding the mean andlower and upper 95% confidence bounds using a hidden Markov

model state-space smoothing algorithm (9). This method has theadvantage of using the entire dataset to calculate CIs so thatthere are not multiple comparisons. The probability of REMsleep is significantly different if the 95% CIs do not overlap.Power spectra. EEG values were normalized using the standardscore to compute z values, which are a dimensionless quantity.z= ðx− xÞ=SD, where x is an EEG value, x is the mean of thepopulation, and SD is the standard deviation of the pop-ulation. The normalized EEG was used to compute the powerspectra densities for natural ChR2+ mouse REM sleep andstimulation-induced ChR2+ REM mouse sleep using theChronux toolbox in MATLAB (8). The multitaper methodwas used with 20 tapers and a bandwidth of 5 Hz. The Jackknifemethod in Chronux was used to construct 95% CIs around thepower spectra.

1. Zhao S, et al. (2011) Cell type–specific channelrhodopsin-2 transgenic mice for opto-genetic dissection of neural circuitry function. Nat Methods 8(9):745–752.

2. Yizhar O, Fenno LE, Davidson TJ, Mogri M, Deisseroth K (2011) Optogenetics in neuralsystems. Neuron 71(1):9–34.

3. el Mansari M, Sakai K, Jouvet M (1989) Unitary characteristics of presumptive cholin-ergic tegmental neurons during the sleep-waking cycle in freely moving cats. Exp BrainRes 76(3):519–529.

4. Thakkar MM, Strecker RE, McCarley RW (1998) Behavioral state control through dif-ferential serotonergic inhibition in the mesopontine cholinergic nuclei: A simultaneousunit recording and microdialysis study. J Neurosci 18(14):5490–5497.

5. Schneider CA, Rasband WS, Eliceiri KW (2012) NIH Image to ImageJ: 25 years of imageanalysis. Nat Methods 9(7):671–675.

6. Paxinos G, Franklin K (2007) The Mouse Brain in Stereotaxic Coordinates (Academic,San Diego), 3rd Ed.

7. Coleman CG, Lydic R, Baghdoyan HA (2004) M2 muscarinic receptors in pontine re-ticular formation of C57BL/6J mouse contribute to rapid eye movement sleep gener-ation. Neuroscience 126(4):821–830.

8. Bokil H, Andrews P, Kulkarni JE, Mehta S, Mitra PP (2010) Chronux: A platform foranalyzing neural signals. J Neurosci Methods 192(1):146–151.

9. Smith AC, et al. (2004) Dynamic analysis of learning in behavioral experiments.J Neurosci 24(2):447–461.

Van Dort et al. www.pnas.org/cgi/content/short/1423136112 3 of 3