Long-range projections coordinate distributed

brain-wide neural activity with a specific

spatiotemporal profile
Alex T. L. Leonga,b,1, Russell W. Chana,b,1, Patrick P. Gaoa,b, Ying-Shing Chanc, Kevin K. Tsiab, Wing-Ho Yungd,
and Ed X. Wua,b,c,e,2
a
Laboratory of Biomedical Imaging and Signal Processing, The University of Hong Kong, Pokfulam, Hong Kong SAR, China; bDepartment of Electrical and
Electronic Engineering, The University of Hong Kong, Pokfulam, Hong Kong SAR, China; cSchool of Biomedical Sciences, Li Ka Shing Faculty of Medicine,
The University of Hong Kong, Pokfulam, Hong Kong SAR, China; dSchool of Biomedical Sciences, Faculty of Medicine, The Chinese University of Hong Kong,
Shatin, Hong Kong SAR, China; and eState Key Laboratory of Pharmaceutical Biotechnology, The University of Hong Kong, Pokfulam, Hong Kong SAR,
China

Edited by Marcus E. Raichle, Washington University, St. Louis, MO, and approved October 25, 2016 (received for review October 4, 2016)

One challenge in contemporary neuroscience is to achieve an inte- (19–21). Although studies demonstrated that rsfMRI connectivity
grated understanding of the large-scale brain-wide interactions, is constrained by well-defined structural networks (3, 17), con-
particularly the spatiotemporal patterns of neural activity that give nectivity itself can reorganize during learning-related tasks (22).
rise to functions and behavior. At present, little is known about the Given these multifaceted findings, it is now imperative to directly
spatiotemporal properties of long-range neuronal networks. We investigate the spatiotemporal properties of large-scale neural
examined brain-wide neural activity patterns elicited by stimulat- interactions. Specifically, how do long-range projection networks
ing ventral posteromedial (VPM) thalamo-cortical excitatory neu- facilitate and govern large-scale neural activity?
rons through combined optogenetic stimulation and functional MRI The cerebral cortex contains long-range axonal projections of
(fMRI). We detected robust optogenetically evoked fMRI activation pyramidal neurons reciprocally connected to the thalamus by
bilaterally in primary visual, somatosensory, and auditory cortices cortico-thalamic and thalamo-cortical projections and within the
at low (1 Hz) but not high frequencies (5–40 Hz). Subsequent elec- cortex via cortico-cortical and interhemispheric projections (23).
trophysiological recordings indicated interactions over long temporal These long-range projections are predominantly glutamatergic
windows across thalamo-cortical, cortico-cortical, and interhemi- and generate most excitatory inputs to their target regions (24).
spheric callosal projections at low frequencies. We further observed Thalamo-cortical circuits, particularly the somatosensory tha-
enhanced visually evoked fMRI activation during and after VPM lamo-cortical circuits in rodents, are excellent models of long-
stimulation in the superior colliculus, indicating that visual pro- range projections often studied for their straightforward circuit
cessing was subcortically modulated by low-frequency activity architecture with monosynaptic excitatory cortical input (25, 26).
originating from VPM. Stimulating posteromedial complex tha- In addition, the thalamus can control cortical states (27–29) and
lamo-cortical excitatory neurons also evoked brain-wide blood- modulate spontaneous cortical activity based on different cortical
oxygenation-level–dependent activation, although with a distinct states (30, 31). These studies suggest that the thalamo-cortico-
spatiotemporal profile. Our results directly demonstrate that low- thalamic networks are integral to facilitating diverse functional
frequency activity governs large-scale, brain-wide connectivity and neural integration across the entire brain.
interactions through long-range excitatory projections to coordinate The neuronal architecture within the brain is an intricate web
the functional integration of remote brain regions. This low-fre- of interconnected excitatory neurons and inhibitory interneurons,
quency phenomenon contributes to the neural basis of long-range
functional connectivity as measured by resting-state fMRI. Significance
fMRI | optogenetic | brain connectivity | low frequency | thalamus What makes the brain tick? A simple yet challenging question
that has captivated our minds for centuries. This sentiment was

T he brain is a highly complex, interconnected structure with par-

allel and hierarchical networks distributed within and between
neural systems (1, 2). This integrative architecture dictates the un-
fittingly reflected in the launch of The BRAIN Initiative 3 years
ago, spurred by the rapid advancement of noninvasive brain
imaging and neuronal mapping technologies that have ad-
derlying principles of how brain-wide neural connectivity supports vanced our understanding of neural networks, which are central
and organizes sensory, behavioral, and cognitive processes (3). Re- to brain functions and behavior. Here, we study the patterns of
cent advances in structural (2, 4) and functional connectivity (5–12) large-scale brain-wide interactions mediated by thalamo-cortical
mapping, as well as neural circuit modulatory tools, such as opto- networks through optogenetics and functional MRI. We found
genetics (13, 14), permit detailed, high-resolution neural examina- that the thalamus can recruit long-range cortical and subcortical
tions at unprecedented scales. In particular, functional connectivity networks and initiate their interactions in a spatiotemporally
mapping using resting-state functional MRI (rsfMRI) produces specific manner. This finding provides a fresh impetus to study
noninvasive visualization of slow and spontaneous hemody- the mysteries of the brain.
namic fluctuations in humans (5–9) and animals (10–12). Coherent
Author contributions: A.T.L.L., R.W.C., and E.X.W. designed research; E.X.W. supervised
low-frequency (<1 Hz) fluctuations across functionally coupled, research; A.T.L.L. and R.W.C. performed research; Y.-S.C., K.K.T., and W.-H.Y. provided
large-scale networks is an intriguing property, although the exact technical assistance; A.T.L.L., R.W.C., P.P.G., and E.X.W. analyzed data; and A.T.L.L. and
underlying neural bases and functional significance remain unclear. E.X.W. wrote the paper.
Previous studies integrating large-scale electrical recordings, volt- The authors declare no conflict of interest.
age-sensitive dye (VSD), and Ca2+-imaging techniques (15–18) This article is a PNAS Direct Submission.
support the hypothesis that slow, oscillating neural activity con- 1
A.T.L.L. and R.W.C. contributed equally to this work.
strains and elicits these hemodynamic fluctuations. Furthermore, 2
To whom correspondence should be addressed. Email: [email protected].
low-frequency activity can temporally synchronize remote brain This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
regions for extended periods to enable information processing 1073/pnas.1616361113/-/DCSupplemental.

E8306–E8315 | PNAS | Published online December 5, 2016 www.pnas.org/cgi/doi/10.1073/pnas.1616361113

 PNAS PLUS
which creates highly dynamic yet stable excitation–inhibition net- to cortical and subcortical regions. We overcame the limitations of
works. Normal operation of this architecture or brain-wide net- electrical stimulation nonspecificity (34, 35) and electrode re-
works requires precise and spatiotemporally structured activity cording spatial coverage by using a multimodal approach. We used
propagation and interaction patterns that support the functional optogenetic functional MRI (fMRI) in combination with electro-
properties of these networks, including downstream activity prop- physiological recordings to interrogate the vibrissae (whisker)
agation characteristics within long-range projections. Intuitively, somatosensory thalamo-cortical system. Integration of fMRI
simple perturbations using local electrical microstimulation will and optogenetics permits large-scale view and detection of brain-
not likely permit a reliable probe of the spatiotemporal activity wide neural activity and enables reversible, millisecond precision
propagation dynamics and their functional output. Previous studies and cell-type–specific modulation of thalamo-cortical neurons.
examining the spatiotemporal response properties within thalamo- We selectively targeted thalamo-cortical excitatory neurons in
cortical circuits were limited by nonspecific electrical stimulation. ventral posteromedial thalamus (VPM), a primary lemniscal
Electrical stimulation cannot target cell-type–specific neurons, somatosensory pathway, and stimulated them at different frequen-
and it excites circuit components in the stimulation region that cies (1–40 Hz). The posteromedial complex of the thalamus (POm),
may not be involved in the particular behavior under physio- a nonlemniscal somatosensory pathway, was also examined. We in-
logical conditions. Indeed, electrical stimulation can disrupt vestigated the spatiotemporal properties of downstream excitatory
both cortico-cortical and cortico-thalamic signal propagation signal propagation beyond the defined thalamo-cortical circuit,
(32) and result in enlarged regions of spatial activation in so- identified the requisite long-range projections, and tested whether
matosensory cortex (33). These studies indicate the inadequacy VPM stimulation modulates other sensory processing in remote
of local electrical stimulation to evoke and characterize down- brain regions.
stream activity propagation and interaction within long-range and
polysynaptic projection networks. Furthermore, current electrode Results
recording technology limits both the recording site density and Brain-Wide fMRI Mapping of Downstream Signal Propagation from VPM.
spatial coverage. Therefore, thalamo-cortical circuits have been We expressed Ca2+/calmodulin-dependent protein kinase II alpha
characterized only between thalamic nuclei and directly inner- (CaMKIIα)-dependent ChR2(H134R) fused with mCherry in VPM
vated cortical regions (34, 35). This characterization yields a lim- thalamo-cortical excitatory neurons in normal adult Sprague-
ited understanding into widespread neural interactions beyond Dawley rats, as the majority of thalamic nuclei, except for the reticular
thalamo-cortico-thalamic networks. Hence, a disconnection exists nucleus, are predominantly populated by CaMKIIα-expressing ex-
at the systems level describing how thalamo-cortical circuits me- citatory neurons (36) (Fig. 1A and Fig. S1A). We confirmed spe-
diate large-scale brain-wide neural activity. cific expression in VPM excitatory neurons through colocalization
In this study, we sought to characterize large-scale brain-wide of mCherry with CaMKIIα staining. We performed optogenetic
neural activity propagation and interaction dynamics and ex- stimulation in lightly anesthetized rats (1% isoflurane). Blue light

NEUROSCIENCE
amine their functional roles. We determined the spatiotemporal pulses at five frequencies (1 Hz: 10% pulse width duty cycle; 5, 10,
response properties of long-range excitatory thalamic projections 20, and 40 Hz: 30% duty cycle; light intensity: 40 mW/mm2) were

Fig. 1. Histological characterization of ChR2::CaMKIIα viral expression in VPM thalamo-cortical excitatory neurons and optogenetic fMRI stimulation setup.
(A) Confocal images of ChR2-mCherry expression in VPM; Lower magnification (Left) and higher magnification (Right). Overlay of images costained for the nuclear
marker DAPI, excitatory marker CaMKIIα, and mCherry revealed colocalization of mCherry and CaMKIIα in the cell body of thalamo-cortical neurons (indicated by white
arrows). (B) Illustration of stimulation site in ChR2 VPM thalamo-cortical neurons during optogenetic fMRI experiments (Left). T2-weighted anatomical MRI image
showing the location of the implanted optical fiber (asterisk: stimulation site) (Right). (C) Optogenetic fMRI stimulation paradigm. Five different frequencies (1, 5, 10, 20,
and 40 Hz) were presented in a pseudorandomized order. All frequencies were presented at 30% duty cycle except for 1 Hz that was presented at 10% duty cycle.

Leong et al. PNAS | Published online December 5, 2016 | E8307

delivered to VPM neurons in an interleaving block paradigm (Fig. 1 not apparent at 1 Hz because the lower duty cycle elicited a weaker
B and C). These frequencies cover natural whisking frequencies BOLD signal in VPM.
(5 Hz and 10 Hz), below (1 Hz) and above (40 Hz). We chose a The above fMRI visualization of large-scale thalamo-cortical
reduced duty cycle of 10% for l Hz of stimulation to avoid exces- downstream signal propagation demonstrates that evoked spatio-
sively long stimulation pulse widths that may not be physiological. temporal dynamics are not restricted to monosynaptic thalamo-
We made the exposed optical fiber cannula (used to deliver blue cortical projections from VPM to somatosensory cortices (Fig. 2
light pulses) opaque using heat-shrinkable sleeves to prevent light and Figs. S1B and S2). We found frequency-specific response
leakage that may cause undesired visual stimulation to animals. properties that transition activation from somatosensory-visual
We detected robust large-scale blood-oxygenation-level– and somatosensory-auditory regions at low frequency (1 Hz) to
dependent (BOLD) fMRI activation within and beyond the so- sensorimotor areas at whisking frequency (5 Hz and 10 Hz). To-
matosensory thalamo-cortical circuit at 1 Hz of ipsilateral VPM gether, these fMRI results indicate that excitatory signals evoked
stimulation (Fig. 2 and Fig. S2). We identified robust and diffuse by optogenetic stimulation of VPM propagate to remote cortical
BOLD responses in a contralateral primary somatosensory barrel regions polysynaptically and robustly at low frequencies.
field (S1BF) and upper lip area (S1ULp), secondary somatosen-
sory cortex (S2), bilateral primary visual (V1) and auditory (A1) Extracellular Recordings Underlying Brain-Wide BOLD Activations.
cortices, and cingulate cortex (Cg), in addition to ipsilateral SIBF, Guided by our fMRI results, we performed local field potential
S1ULp, and S2 (see Fig. 2C and Fig. S2C for the BOLD signal (LFP) recordings using the same stimulation paradigm as shown in
profiles). Despite the long interstimulus duration and low duty cy- Fig. 1C at five sites: ipsilateral VPM, bilateral S1, and bilateral V1
cle, we reliably detected BOLD activation beyond thalamo-cortical (Fig. 3A). Averaged 30-s traces of LFP recordings largely con-
networks, which revealed recruitment of brain-wide and pop- curred with the magnitude and temporal profiles of BOLD acti-
ulation-wide neural activity by low-frequency stimulation. We vation at ipsilateral S1BF and bilateral V1 (Fig. 3B vs. Fig. 2C).
also performed 1 Hz of stimulation at a lower light intensity The temporal evolution of BOLD signals was generally slower due
(20 mW/mm2; Fig. S3A) and observed no apparent change in to the inherent nature of the hemodynamic response to neural
the overall spatial extent of broad cortical BOLD responses activity (37). Similar levels of LFP responses were elicited at all
although BOLD signal amplitudes decreased as expected. At frequencies (1–40 Hz) for the first optogenetic pulse at each re-
5 and 10 Hz, BOLD activation was no longer present in con- cording site. LFP response levels to optogenetic pulses were largely
tralateral somatosensory cortices and bilateral V1 and A1 (Fig. constant at 1 Hz. However, they decreased quickly during con-
2 and Fig. S2). Instead, robust and diffuse BOLD activation was secutive pulses at 5 Hz and 10 Hz and sharply decayed after the
detected in ipsilateral primary motor cortex (M1), in addition to first pulse at 20 Hz and 40 Hz. LFP recordings performed in naive
ipsilateral S1BF, S2, and S1ULp. At 20 Hz and 40 Hz, we detected animals showed no evoked LFPs at these five sites upon simple
only BOLD activation in ipsilateral S1BF, S2, and S1ULp. We VPM light stimulation (Fig. S4). In addition, laser light pulses can
observed no BOLD activation in naive animals under the same generate high photocurrents in neurons (38), which may produce
light stimulation paradigm. The absolute BOLD signal amplitude unphysiological states. Therefore, we examined LFP recordings
within the ipsilateral S1BF at 1 Hz was generally lower than signal during 1 Hz of stimulation using different light intensities with the
amplitudes at 5 Hz and 10 Hz mainly due to the reduced-pulse- default 40 mW/mm2 vs. 20 mW/mm2 and 15 mW/mm2 (Fig. S3B).
width duty cycle of 10% used at 1 Hz. VPM BOLD activation was We observed no apparent changes in LFP characteristics.

Fig. 2. Low-frequency (1 Hz), not high-frequency (5–40 Hz), optogenetic excitation of VPM thalamo-cortical excitatory neurons evokes brain-wide positive
BOLD activations in remote and bilateral sensory cortices. (A) Illustration of Paxinos atlas-based ROI definitions. S1BF: primary somatosensory barrel field;
V1: primary visual cortex; M1: motor cortex; S1ULp: primary somatosensory upper lip area; S2: secondary somatosensory cortex; A1: primary auditory cortex;
SC: superior colliculus; LGN: lateral geniculate nucleus; V2: secondary visual cortex; Cg: cingulate cortex. (B) Averaged BOLD activation maps at 1, 5, 10, 20,
and 40 Hz of optogenetic stimulation (n = 6; P < 0.001; asterisk: stimulation site). Robust and diffuse positive BOLD activations were observed in bilateral S1BF,
V1/V2, A1, S2, and Cg cortices upon 1 Hz of stimulation. When stimulation frequency was increased to 5 and 10 Hz, robust and diffuse positive BOLD ac-
tivations were observed in ipsilateral S1BF, M1, S1ULp, and S2. Further increase in stimulation frequency to 20 and 40 Hz activated only the ipsilateral S1BF,
S1ULp, and S2. (C) BOLD signal profiles extracted from ROIs defined in A (error bar indicates ± SEM).

E8308 | www.pnas.org/cgi/doi/10.1073/pnas.1616361113 Leong et al.

 PNAS PLUS
Fig. 3. LFP recordings in sensory cortices and VPM confirm neuronal activity underlying BOLD activation and indicate adaptation at the thalamic and cortical
levels. (A) Illustration of the location of recording electrodes in bilateral S1BF, bilateral V1, and optrode in VPM. (B) Averaged LFP from a representative animal

NEUROSCIENCE
during VPM optogenetic stimulation (eight blocks; 1 Hz, 10% duty cycle; 5, 10, 20, and 40 Hz, 30% duty cycle). Response levels of evoked LFPs remained fairly
constant at 1 Hz. However, evoked LFPs decreased quickly during the initial consecutive pulses at 5–40 Hz. These LFP recordings were consistent with the observed
BOLD responses. (C) Adaptation characterization at ipsilateral VPM during 5 Hz of optogenetic stimulation. Adaptation was quantified by normalization of the
amplitudes of positive peaks at each of the evoked potentials to the amplitude of the first positive peak, which was then fitted with a monoexponential decay
curve to quantify its decay constant, τ. (D) There was additional adaptation in ipsilateral S1 at 5 Hz (τVPM = 2.10 ± 0.15 s vs. τS1 = 1.50 ± 0.17 s; error bar indicates
SD; paired t test with post hoc Holm–Bonferroni correction; **P < 0.01) and 10 Hz (τVPM = 1.00 ± 0.13 s vs. τS1 = 0.73 ± 0.11 s; *P < 0.05). This finding indicates that
the observed adaptation involved interactions within the thalamo-cortico-thalamic network and was not solely from sustained VPM membrane depolarization.

These LFP recordings showed that this frequency-dependent Long-Range Excitatory Projections Support Signal Propagation at
adaptation occurs first in VPM. We measured the LFP peak Low Frequencies and Interactions with Long Temporal Windows. To
amplitude decay constant, τ, for 5 and 10 Hz stimulation (Fig. 3 C identify downstream signal propagation pathways at 1 Hz, we
and D). The measurements indicated further adaptation in ipsi- analyzed the latency of evoked LFPs by using the first LFP peak
lateral S1, as τS1 was larger than τVPM (2.10 ± 0.15 s vs. 1.50 ± location to represent the initial neuronal population activity (Fig. 5
0.17 s, P < 0.01 for 5 Hz; 1.05 ± 0.13 s vs. 0.73 ± 0.11 s, P < 0.05 A–C). Negative LFP peaks were observed first and used for ipsi-
for 10 Hz). Hence, simple failure of VPM excitatory neurons due lateral VPM and ipsilateral and contralateral S1, whereas positive
to sustained membrane depolarization at high stimulation fre- ones were observed for ipsilateral and contralateral V1. Analysis
quencies does not underlie this adaptation phenomenon. Both revealed a response latency of 3.4 ± 0.3 ms for VPM. As shown in
cortico-thalamic and intracortical interactions can contribute to Fig. 5D, a delay of 9.6 ± 0.6 ms occurred between VPM and ipsi-
these dynamic properties. lateral S1, consistent with the reported 8.5-ms latency value for
Multiunit activity (MUA) recordings in VPM using an optrode monosynaptic VPM to S1 projection (41). Signal then propagated
showed that spikes were optogenetically evoked in thalamo-cor- from ipsilateral S1 to ipsilateral V1 through cortico-cortical cir-
tical neurons at all stimulation frequencies (Fig. 4A). Spike fre- cuit projections in 21.5 ± 1.6 ms, in line with the reported value of
quency analysis demonstrated that optogenetic pulses successfully 25 ms (17), and to contralateral S1 through interhemispheric
evoked action potentials in a stimulus-locked manner, even at high monosynaptic callosal projection in 6.7 ± 0.5 ms, consistent with the
frequencies including 20 Hz and 40 Hz, the range at which non- 6.0 ms estimated from reported axonal conduction velocity (42).
cell-type–specific electrical stimulation fails (39). We observed an Finally, the signal propagated in 6.2 ± 0.9 ms from ipsilateral V1 to
onset latency of 3.3 ± 1.4 ms, consistent with VPM thalamo-cor- contralateral V1 through interhemispheric callosal projection.
tical neuron characteristics (40). Surprisingly, MUA recordings However, we cannot preclude that propagation from contralateral
revealed different spiking patterns that transitioned from thalamic S1 to contralateral V1 occurs through cortico-cortical circuit pro-
burst activity to tonic activity with increasing stimulation fre- jections in 21.0 ± 1.8 ms because it closely matches the ipsilateral S1
quency. As long stimulation pulse durations can confound evoked to ipsilateral V1 delay. Taken together, these analyses indicate di-
thalamic burst activity, we repeated these recordings using a 10-ms rect participation of long-range excitatory projections as demon-
stimulation pulse and observed similar spiking characteristics (Fig. strated through their initial population activity propagation delays.
4B). These findings indicate that the occurrence of evoked burst As shown in Fig. 5A and in contrast to Fig. 5E, the averaged
activity in VPM is determined through interstimulus intervals, not LFP evoked at 1 Hz typically persisted for 200–300 ms, which
stimulus duration. included strong secondary peaks long after the primary peaks,

Leong et al. PNAS | Published online December 5, 2016 | E8309

 served no significant changes in overall latencies and LFP temporal
characteristics (Fig. 5F).

Visual Subcortical BOLD Activation Enhanced by Low-Frequency VPM

Stimulation. To test whether low-frequency activity over long-
range projections influences cortical and subcortical processing of
visual stimuli, we used a dual stimulation paradigm with a con-
tinuous 1-Hz VPM optogenetic stimulation and binocular visual
stimulation (Fig. 6A). Visual stimulation was applied before
(baseline), during (OG-On), and after (OG-Off) continuous
VPM stimulation.
Baseline visual stimulation evoked positive BOLD activation
along the visual pathway, including the lateral geniculate nucleus
(LGN), superior colliculus (SC), and visual cortices (V1 and V2)
(Fig. 6B). This finding was expected and consistent with the
previous studies (43–45). In the presence of continuous opto-
genetic stimulation, visually evoked BOLD responses signifi-
cantly increased in the SC (Fig. 6 C and D; ipsilateral SC, P <
0.01; contralateral SC, P < 0.01; one-way ANOVA). Moreover,
visual-stimulation–evoked BOLD responses remained elevated
in SC after optogenetic stimulation during the ∼2.5-h experi-
mental window (Fig. 6 C and D). These findings indicate that
excitability in the visual pathway is modulated by low-frequency
activity over long-range projections originating from the primary
somatosensory thalamic nucleus, VPM.

fMRI Mapping of Downstream Signal Propagation from POm. We also

used optogenetic fMRI to examine brain-wide BOLD activity
evoked by excitatory neuronal stimulation in the nonlemniscal so-
matosensory thalamus, POm (Fig. 7). We observed activation of
thalamo-cortical and thalamo-tectal projections upon 1 Hz of
stimulation, which evoked robust BOLD activation in ipsilateral
S1BF, bilateral SC, and LGN. At 5 Hz, BOLD activation was de-
tected in ipsilateral S1BF, bilateral V1, LGN, and SC. At 10 Hz,
BOLD activation was observed only in ipsilateral S1BF (Fig. 7 B
and C). Neither ipsilateral S1BF nor SC was activated at 40 Hz of
stimulation. These findings again demonstrate the brain-wide and
population-wide neural activity evoked by low-frequency stimula-
tion of POm. Furthermore, these spatiotemporal response charac-
teristics are distinct from those observed by VPM stimulation,
highlighting the differential roles of VPM and POm in coordinating
Fig. 4. MUA recordings of VPM thalamo-cortical excitatory neurons reveal
brain-wide neural activity. Histological examination of POm ani-
sustained thalamic burst activity at low frequency (1 Hz) and a transition to
tonic activity at high frequencies (20 Hz and 40 Hz). (A) MUA measured during
mals confirmed that local viral expression was restricted to the POm
optogenetic stimulation at five frequencies. Twenty-one-second trace of VPM nucleus with no observable expression in VPM (Fig. S5).
MUA from a representative animal during VPM optogenetic stimulation (1 Hz,
10% duty cycle; 5, 10, 20, and 40 Hz, 30% duty cycle; Top Left to Bottom Left). Discussion
Corresponding 2-s VPM MUA trace magnified at the beginning (Bottom Left) Long-range projections are essential for neural communication
of stimulation and illustration of evoked thalamic burst (multiple spikes) and between remote brain regions. The brain has an intrinsic capa-
tonic activity (single spike) (Bottom Right) for each stimulation frequency, re- bility to recruit various projections within interconnected net-
spectively. Similar levels of MUA responses were evoked at all frequencies by works to serve various functional demands without requiring
the first stimulation pulse. Analysis of optogenetically evoked spikes revealed
individual long-range projections to perform a single function.
an onset latency of 3.3 ± 1.4 ms (mean ± SD), consistent with VPM thalamo-
cortical neuron characteristics. Burst activity was sustained at 1 Hz, decreased
Despite the wealth of positive evidence, it remains a challenge to
at 5 Hz and 10 Hz, and transitioned to tonic activity at 20 Hz and 40 Hz. VPM establish how long-range projection recruitment and modulation
spiking summary for two animals (Top Right to Bottom Right). Spike-rate occur across the broad range of neural targets. Furthermore,
analysis showed that spikes can be successfully generated at high frequencies although long-range projections are prevalent in all mammalian
(20 Hz and 40 Hz). (B) MUA measured using short-pulse stimulation (10 ms). brains, most studies focus on large-scale cortico-cortical inter-
Corresponding 2-s VPM MUA trace magnified at the beginning of stimulation actions with minimal emphasis on modulation through subcortical
for 1 Hz (Left), 5 Hz (Middle), and 10 Hz (Right). These findings demonstrate interactions. Here, we examined both issues using optogenetics and
that similar spiking characteristics and occurrence of evoked burst activity in fMRI to monitor brain-wide neural responses during thalamic
VPM are determined by interstimulus intervals, not by stimulus duration.
stimulation. Our fMRI and subsequent extracellular recording re-
sults show that long-range cortico-cortical and interhemispheric
callosal projections are recruited polysynaptically during low-fre-
particularly in contralateral and ipsilateral V1 and contralateral S1. quency stimulation of VPM excitatory neurons. Similarly, POm
This finding indicates that a long interstimulus duration (>200 ms) stimulation leads to brain-wide neural activity but with a distinct
is required to allow and enable recurrent, long-lasting interactions spatiotemporal profile. Our results indicate that this low-frequency
between and within long-range cortico-cortical and thalamo-cor- phenomenon likely is causally related to reciprocal interactions
tico-thalamic networks. We also performed LFP recordings at 1 Hz over thalamo-cortico-thalamic, cortico-cortical, and interhemispheric
using a shorter pulse duration (10 ms instead of 100 ms). We ob- callosal projections to facilitate activity propagation with extended

E8310 | www.pnas.org/cgi/doi/10.1073/pnas.1616361113 Leong et al.

 PNAS PLUS
NEUROSCIENCE
Fig. 5. LFP recordings in sensory cortices and VPM reveal downstream signal propagation supported by low frequency (1 Hz) over long-range projections and
their interactions with long temporal windows. (A) Averaged evoked LFPs during 1 Hz of VPM optogenetic stimulation (n = 3; 10% duty cycle; error bar
indicates ± SD). (B) Summary of propagation delays from the onset of stimulation (magnified from A). Latency is measured as the time between the onset and
the first evoked LFP peak location, indicated by a cross (x). (C) Summary of latency measured at the five different locations (n = 3; error bar indicates ± SD,
one-way ANOVA with post hoc Bonferroni correction; **P < 0.01; *P < 0.05). (D) Schematic diagram of downstream signal propagation at 1 Hz. Optogenetic
stimulation of VPM thalamo-cortical neurons elicited action potentials that excited ipsilateral S1 through thalamo-cortical projections. Subsequently, low-
frequency activity propagated cortico-cortically from ipsilateral S1 to ipsilateral V1, and interhemispheric callosal projections between ipsilateral and con-
tralateral S1 were recruited. Signal then was propagated from ipsilateral V1 to contralateral V1 via interhemispheric callosal projections. (E) Averaged evoked
LFPs during 5 Hz of VPM optogenetic stimulation (n = 3; 30% duty cycle; error bar indicates ± SD). Comparision with 1 Hz reveals the absence of secondary
peaks in ipsilateral and contralateral V1. This indicates that long-lasting and recurring interactions between and within cortico-cortical and thalamo-cortico-
thalamic networks require long interstimulus durations. (F) Evoked potentials from a representative animal during 1 Hz of VPM optogenetic stimulation using
a short stimulation pulse (1% duty cycle) shows no significant differences in overall latencies and LFP temporal characteristics.

time windows. Furthermore, visual subcortical SC responses to excitatory pyramidal neurons and inhibitory interneurons at thalamo-
visual stimulation are enhanced during and after sustained acti- cortical synapses (46). Most pyramidal neurons have axon collaterals
vation of long-range projections via VPM stimulation, which that project back to and synapse in superficial layers, whereas others
suggests that long-range projections and interactions shape mul- distribute excitation horizontally. These collaterals form a strong,
tisensory integration processes in the brain. Together, these results recurrent excitatory network both within the local microcircuit and
indicate that low-frequency activity mediates brain-wide interac- among long-range thalamo-cortico-thalamic and cortico-cortical
tions through long-range excitatory projections. Such flexible char- networks (47). The strong input amplification elicited through this
acteristics of low-frequency, brain-wide neural activity can positive feedback loop is tightly regulated by various GABAergic
contribute to the neural basis of long-range functional organiza- interneurons (48) interposed within this pyramidal network to target
tion and connectivity as measured by rsfMRI (5–12), extracellular different neuronal subdomains (23). In our study, evoked LFPs by
recording (15, 16), VSD (17), and Ca2+ imaging (18). 1 Hz of VPM stimulation persisted over long periods and exhibited
Surprisingly, we demonstrated that low-frequency stimulation multiple peaks in the ipsilateral VPM, bilateral S1, and V1 (Fig. 5 A
of excitatory neurons in VPM evokes BOLD responses in regions and E). This observation suggests that feedforward and feedback
far beyond somatosensory cortices (Fig. 2). Reciprocal projections mechanisms governing cortico-cortical and thalamo-cortico-thalamic
between the thalamus and recurrent cortical circuits distinguish networks play a causal role to modulate these LFP responses and
the mammalian sensory cortex. The laminar and columnar ar- directly contribute to robust brain-wide BOLD activations detected
chitecture of the sensory cortex offers the most intuitive structural during 1 Hz of stimulation.
design with these specific functional characteristics. Sensory cor- The question remains how and why cortical long-range projec-
tical pyramidal neurons receive thalamic input at specific cortical tions are recruited. We found that low-frequency VPM stimula-
layers and distribute excitatory signals within and outside the cortex, tion elicited thalamic burst activity, whereas high-frequency
whereas inhibitory interneurons are interspersed throughout corti- stimulation elicited tonic activity (Fig. 4A). Thalamo-cortical ex-
cal layers to provide the requisite excitation–inhibition balance. citatory neurons have intrinsic membrane potential properties
Despite exclusive stimulation of excitatory thalamo-cortical that allow them to fire in two modes, namely burst and tonic/
neurons in VPM, downstream signal propagation recruits both S1 single spike (49), which require cortico-thalamic activity (50, 51).

Leong et al. PNAS | Published online December 5, 2016 | E8311

 Thalamic burst activity, originally identified in the sleep/anes-
thetized state, is a proposed mechanism to decouple thalamo-
cortical activity from the peripheral world (52). However, recent
evidence suggests that thalamic bursting plays a fundamental role
in sensory processing (53, 54). Studies of visual (55) and somato-
sensory (30, 53) pathways demonstrate the presence of thalamic
bursting in awake and attentive states, implying that bursting ac-
tivity modulates sensory perception. Intermittent activation of
multiple long-range circuits to integrate major sensory cortices
suggests that multisensory integration is essential to shape sensory
perceptions that could not otherwise be achieved with only one
sense (56, 57). This premise raises an intriguing proposition that
excitatory long-range projections and their dynamic interactions
are recruited during low-frequency thalamic stimulation through
burst activity propagation from VPM to S1 and, possibly, to other
sensory cortices. In this study, we observed that thalamic burst
activity was successfully elicited during VPM stimulation and sus-
tained only when the interstimulus interval was more than 200 ms
(Fig. 4A), when brain-wide BOLD activation was observed. Such
temporal characteristics indicate that the brain-wide integration of
spiking information from multiple thalamic neurons simulta-
neously, possibly through bursting activity, requires extended time
windows. These interactions presumably synchronize remote brain
regions that require large conduction delays (19–21).
Synaptic depression by anesthesia at the stimulation site may
diminish polysynaptic activity propagation at high frequencies
(39). However, our LFP recordings after cessation of light an-
esthesia show that evoked V1 responses did not exhibit apparent
changes, despite increases in fidelity of signal transmission from
VPM to S1 (Fig. S6). A previous study in awake rats showed that
desynchronized activity occurred during whisking, as indicated by
decreased slow oscillations (∼1 Hz) in somatosensory cortex, and
that these effects could be mimicked using high-frequency opto-
genetic stimulation of thalamo-cortical excitatory neurons (28).
These findings support our contention that high-frequency tha-
lamic stimulation may not activate long-range circuits beyond its
projected area, even during the awake state, when high-frequency
signals (>20 Hz) are transmitted reliably to the cortex. Therefore,
the recruitment of long-range projections by low frequency is not
an effect of anesthesia.
In this study, we show that constant recruitment of dynamic net-
work interactions through continuous low-frequency VPM stimula-
tion increased SC BOLD activation to external visual stimulus and
that such SC enhancement persisted after cessation of VPM opto-
genetic stimulation (Fig. 6). This observation indicates that sensory
gain, a fundamental component of sensory information processing
(58, 59), can be modulated in the SC during and after continual
activation of long-range circuits. Low-frequency stimulation co-
incided with VPM burst activity, which is modulated by cortico-
thalamic input (50, 51). Furthermore, thalamic bursting activity can
increase due to enhancement of sensory feedback gain from cortical
to thalamic regions (50). Hence, the enhanced SC activation ob-
served here could result from burst information relayed to sensory
cortices from VPM and facilitated through direct cortico-tectal
Fig. 6. Low-frequency (1 Hz) optogenetic VPM stimulation enhances visually
evoked subcortical SC BOLD activation. (A) Illustration of binocular visual and
projections from V1 and/or S1 (24, 60). Neural plasticity mechanisms
optogenetic fMRI stimulation setup (Top Left). Block design visual stimulation are also likely recruited. These mechanisms can readjust the neu-
paradigm (20 s on 60 s off) presented before (baseline), during (OG-On), and ronal components stimulating homeostatic regulation of neural cir-
after (OG-Off) continuous 1 Hz of optogenetic stimulation (Right). Typically, cuit functions, such as neuronal or network excitability changes.
five baseline visual fMRI scans were acquired before five OG-On scans were Here, we define neural plasticity as the ability of the nervous system
interleaved with five OG-Off scans (Bottom Left). (B) Illustration of atlas-based to adopt a new functional or structural state in response to extrinsic
ROI definitions (Top). SC: superior colliculus; LGN: lateral geniculate nucleus; and intrinsic factors (61). Plasticity occurs at different neuro-
V1/V2: primary/secondary visual cortex. Averaged BOLD activation maps be- developmental stages spontaneously (62) and as driven by sensory
fore, during, and after 1 Hz of optogenetic stimulation (n = 6; P < 0.001; as-
experiences (63). Such sequential combination of spontaneously
terisk: stimulation site) (Bottom). (C) BOLD signal profiles extracted from ROIs
defined in B (error bar indicates ± SEM). BOLD activations in ipsilateral and
contralateral SC were enhanced during optogenetic stimulation, which
remained elevated after cessation of optogenetic stimulation. These findings β-values extracted from ROIs defined in B (n = 6; error bar indicates ± SEM,
indicate that excitability in the visual pathway is modulated by low-frequency one-way ANOVA with post hoc Bonferroni correction; **P < 0.01; n.s, not
activity over long-range projections originating from VPM. (D) Averaged significant).

E8312 | www.pnas.org/cgi/doi/10.1073/pnas.1616361113 Leong et al.

 PNAS PLUS
also performed rsfMRI measurements (Fig. S7). Our preliminary
rsfMRI results show an increase in long-range interhemispheric
connectivity in S1BF, S2, V1, and A1 during and after low-fre-
quency VPM stimulation. Our findings suggest that functional
connectivity can be directly altered by low-frequency stimulation.
Thus, brain-wide functional connectivity is intimately linked to
low-frequency neural activity.
In this study, we also examined the nonlemniscal somatosensory
thalamus, POm (Fig. 7). We observed brain-wide neural activity
upon POm stimulation at low frequencies (1 and 5 Hz). Interest-
ingly, its spatiotemporal response properties are different from
those observed through VPM stimulation. The thalamus is a relay
center for cortical sensory information. Axonal tracing (68), elec-
trophysiology (41, 69), and behavioral (70) studies show parallel,
but functionally distinct, thalamo-cortical projections, including
lemniscal VPM vs. paralemniscal POm. Previous studies also
characterized the differences between VPM and POm thalamo-
cortical projections in S1BF, in which VPM thalamo-cortical pro-
jections define an essential cortical activity pattern, whereas POm
thalamo-cortical projections affect the effectiveness of transmission
(69, 71). Here, our fMRI results reveal that POm stimulation yields
highly different spatiotemporal response characteristics from those
by VPM stimulation. These findings suggest that POm is func-
tionally distinct from VPM and that the paralemniscal pathway also
plays an essential role in multisensory processing.
Optogenetic VPM or POm stimulation might antidromically
activate somatosensory cortex. However, the experimental evi-
dence from our fMRI and electrophysiological recordings strongly
indicates that the observed spatiotemporally distinct brain-wide
responses are likely not caused by antidromic S1 activation. Spe-
cifically, MUA recordings show the latency between optogenetic
pulse onset and initial spikes as 3.3 ± 1.4 ms and 4.3 ± 2.3 ms for

NEUROSCIENCE
VPM and POm stimulation, respectively (Fig. 4A and Fig. S5B).
These latencies are characteristic of orthodromic activation of
Fig. 7. Optogenetic stimulation of the posteromedial complex of the thala-
mus (POm), a nonlemniscal somatosensory pathway, reveals the distinct long-
VPM or POm thalamo-cortical excitatory neurons (40, 72),
range downstream signal propagation characteristics. (A) Illustration of Pax- whereas antidromic activation exhibits a shorter latency with less
inos atlas-based ROI definitions. S1BF: primary somatosensory barrel field; variation (73). The VPM to ipsilateral S1 propagation time of
LGN: lateral geniculate nucleus; SC: superior colliculus; S1ULp: primary so- 9.6 ± 0.6 ms (Fig. 5D) also corresponds well to published values
matosensory upper lip area; S2: secondary somatosensory cortex; M1: motor for orthodromic propagation (41). These timing measurements
cortex; V1: primary visual cortex, V2: secondary visual cortex; A1: primary au- support that our VPM or POm stimulation activates S1 ortho-
ditory cortex. (B) Averaged BOLD activation maps at 1, 5, 10, and 40 Hz of dromically rather than antidromically, at least in a predominant
optogenetic stimulation (n = 6; P < 0.001; asterisk: stimulation site). A 1-Hz manner. In addition, antidromic activation is unlikely because the
stimulation showed positive BOLD activation at visual subcortical regions, LGN
anterograde and monosynaptic nature of our used viral trans-
and SC. Robust positive BOLD activation was observed in ipsilateral S1BF
during 5 Hz and 10 Hz of stimulation, which then expanded to visual sub-
duction expresses the optogenetic construct only in VPM or POm
cortical activations at 5 Hz. Stimulation at 40 Hz did not show any significant and its ascending cortical targets (i.e., AAV5-CaMKIIα-ChR2)
BOLD activation. These findings indicate a specific spatiotemporal response (Figs. S1 and S5), without further transfecting descending cortico-
profile different from that by VPM stimulation, highlighting the differential thalamic projections from somatosensory cortex (34, 74).
roles of POm and VPM in coordinating brain-wide neural activity. (C) BOLD In summary, our results directly demonstrate that low-frequency
signal profiles extracted from ROIs defined in A (error bar indicates ± SEM). activity supports large-scale brain-wide interactions over long-range
excitatory projections, such that low-frequency activity plays a pivotal
role in functional integration of remote brain regions. Such spatio-
generated and experience-dependent neural activity endows the temporal characteristics can contribute to the neural basis of brain
brain with the ability to accommodate changing, dynamic inputs functional connectivity measured by rsfMRI. In addition, this study
throughout the life span. Studies demonstrate that both spontaneous presents the combined use of large-view fMRI, optogenetic neuro-
and experience-dependent neural activity predominates at low fre- modulation, and external stimuli or tasks as a valuable means to dissect
quencies (64–66), which further corroborates our findings that low- brain long-range interactions and their functions over a large scale.
frequency activity is intimately coupled with long-range projections
and their dynamic interactions. Materials and Methods
At present, functional connectivity is defined as the temporal co- Subjects. All animal experiments were approved by the University of
herence of neural activity patterns among spatially segregated brain Hong Kong’s Committee on the Use of Live Animals in Teaching and
regions at low and ultra-low frequencies, which has been most Research (CULATR). Five groups of adult male Sprague-Dawley rats
widely measured by rsfMRI (5–12). However, interpretation of were used: group I (VPM, n = 6; POm, n = 6) for fMRI experiments; group
II (VPM, n = 3 for LFP and n = 2 for MUA) for electrophysiological re-
functional connectivity studies remains challenging because func-
cordings; group III (VPM, n = 6) for dual stimulation (visual and optogenetic)
tional coupling changes dynamically. This dynamic property sug- fMRI experiments; and group IV (VPM, n = 8) for optogenetic and rsfMRI
gests that it is constrained by, but not fully dictated by, anatomical experiments.
connectivity. Furthermore, the neural mechanisms underlying
long-range, functional coupling remain unknown. Few studies Virus Packaging. Recombinant AAV vectors were serotyped with AAV5-coated
have investigated this question directly (21, 67). In this study, we proteins and produced by the Vector Core of the University of North Carolina.

Leong et al. PNAS | Published online December 5, 2016 | E8313

Viral titer in particles per milliliter was 4 × 1012 for AAV5-CaMKIIα-ChR2 rsfMRI analysis was performed according to a procedure reported in our
(H134R)-mCherry. Maps are available online from www.stanford.edu/group/ previous study (10).
dlab/optogenetics.
LFP and MUA Recordings and Analyses. Simultaneous LFP recordings at five
Stereotactic Surgery for Viral Injection. After the craniotomy was made, viral brain regions were performed on animals using the same anesthesia protocol
injections were performed at two depths in VPM [−3.6 mm posterior to (1.0% isoflurane) as in MRI experiments. Craniotomies (1-mm diameter) were
Bregma, +3.0 mm medial-lateral (ML) right hemisphere, −5.8 mm and −6.2 mm made over bilateral S1 (+0.1 mm anterior to Bregma, ±5.0 mm ML), bilateral
from surface of dura] and POm (−3.6 mm posterior to Bregma, +2.1 mm ML V1 (−6.0 mm posterior to Bregma, ±4.0 mm ML), and a 5-mm-diameter
right hemisphere, −5.0 mm and −5.4 mm from surface of dura) with subsequent craniotomy over the stimulation site VPM (−3.6 mm posterior to Bregma,
verification by high-resolution anatomic MRI. Viral constructs were delivered at +3.0 mm ML right hemisphere). Single tungsten microelectrodes (1 MΩ and
a 1.5-μL volume at each depth through a 5-μL syringe and 33-gauge beveled 10-μm tip diameter) were placed in bilateral S1 and V1 approximately be-
needle injected at a rate of 150 nL/min. Animals recovered for 4 wk before fMRI tween layers IV and V. For VPM, a homemade optrode was used. It consisted
experiments were performed. of a single tungsten microelectrode (2 MΩ and 10-μm tip diameter) coupled
to the fiber using protocols as described previously (77). LFP and MUA re-
Animal Preparation for MRI Experiment. Surgery was performed under 2% cordings were acquired using a 32-channel OpenEphys system with a 30-kHz
isoflurane to implant an optical fiber cannula with a 450-μm diameter at the sampling frequency. The optogenetic stimulation paradigm was the same as
injection site with high-resolution anatomic MRI verification. The fiber-tip used in fMRI experiments. Again, these optical fiber cannulas were made
surface was beveled to facilitate the fiber insertion and minimize injury to opaque using heat-shrinkable sleeves to prevent light leakage that causes
brain tissue. Dental cement fixed the fiber on the skull. inadvertent direct visual stimulation.
Additional optogenetic stimulation paradigms were used to compare the
MRI-Synchronized Optogenetic and Visual Stimulation. For optogenetic stim- effects of short- and long-light pulse durations (10/100, 10/60, and 10/30 ms
ulation, blue light was delivered to VPM or POm nucleus using a 473-nm diode- for 1, 5, and 10 Hz, respectively) and different light intensities (40, 20, and
pumped solid-state laser with calibrated power of 8 mW (40 mW/mm2) at the 15 mW/mm2) on the evoked LFP responses. Additional anesthesia protocols
fiber tip. Optical fiber cannulas were made opaque using heat-shrinkable were used to determine the effects of different anesthesia levels (0.8, 0.5,
sleeves to prevent light leakage that may cause undesired visual stimulation. and 0.0% isoflurane) on evoked LFP responses. Recordings under 0.0% iso-
Both VPM and POm are large thalamic nuclei (Figs. S1 and S5). The optical fiber flurane were performed at 5 min and 15 min after cessation of isoflurane.
tip was typically situated in the center of the VPM or POm nucleus through For LFP analysis, raw data were band-pass-filtered (1–300 Hz) and notch-
stereotaxic implantation. The spatial spread from the fiber tip for 473 nm of filtered for 50 Hz using Matlab. Individual LFP and group-averaged LFP
blue light is rather small (200 μm and 350 μm at 50% and 10% of initial light traces were generated. The LFP latency was measured from the onset of
intensity, respectively) and has little or no spreading in the backward direction light pulse to the first peak location. MUA analysis was performed using
(74), confining the optogenetic stimulation within the VPM or POm nucleus. Offline Sorter software (Plexon Inc.) for spike sorting and Neuroexplorer
For both VPM and POm fMRI experiments, five frequencies were used (1 Hz software (Nex Technologies) and Matlab for calculating spike frequencies
with 10% duty cycle; 5, 10, 20, and 40 Hz with 30% duty cycle). Here, we chose and generating averaged traces. For spike identification, spikes were first
a reduced duty cycle of 10% for 1 Hz of stimulation (in contrast to 30% for sorted using an automated shape-based spike-sorting method, in which a
5–40 Hz) to avoid the use of very long stimulation pulse widths (300 ms). Long negative and positive threshold crossing at twice the noise SD was imposed,
ChR2 optogenetic pulse width can cause neuronal output silencing and ad- and then this sorting was verified by visual inspection before spike frequency
versely affect the neural activity efficacy (75). VPM or POm was stimulated analysis. To quantify the optogenetically evoked spike frequency during the
with a block design paradigm that consisted of 20-s-on and 60-s-off periods. A 20-s optogenetic stimulation block, spikes were identified during the optogenetic
total of nine fMRI sessions were performed on each animal. Each session in- stimulation pulse period and verified to be stimulus-locked. That is, only the first
cluded 10 blocks, with each frequency occupying two blocks. Different fre- spike within a predefined 2- to 5-ms window after the onset of each optogenetic
quencies were presented in a pseudorandom manner. For dual optogenetic/ stimulation pulse was counted as a spiking event, and the subsequent spikes
visual stimulation, additional blue light stimulation was presented via a sep- within the stimulation pulse period were discarded. To quantify the spontaneous
arate optical patch cable at 20 mm in front of both eyes, with power calibrated spike frequency before and after the 20-s optogenetic stimulation block, the
at 0.5 mW. Visual stimulation (1 Hz and 10% duty cycle) was presented in a sorted spikes were counted by imposing an interspike interval of 100 ms that was
block design paradigm of 20-s-on and 40-s-off periods. Each fMRI session in- determined based on the temporal characteristics of the thalamic burst activity
cluded four blocks of visual stimulation. Fifteen fMRI sessions were performed predominantly observed in the absence of stimuli. The above procedure ensured
on each animal, 5 before (baseline), 5 during (OG-On), and 5 after (OG-Off) that both thalamic burst (typically, consecutive multiple spikes within the 20- to
optogenetic stimulation. For rsfMRI experiments, 15 fMRI sessions were per- 80-ms window) and tonic activity (single spikes lasting less than 5 ms) were
formed on each animal, 5 baseline, 5 OG-On, and 5 OG-Off. characterized as one single spiking event to measure optogenetically evoked
spike frequency. The procedure was sensitive to detecting the failure of
optogenetic pulses in evoking spikes.
MRI Acquisitions. All MRI experiments were carried out on a 7T MRI scanner. Scout
and anatomical images were first acquired for accurate positioning and reference.
fMRI and rsfMRI measurements were made using a single-shot gradient-echo Histology, Immunohistochemistry, and Confocal Imaging. Upon completion of
echo-planar imaging (GE-EPI) sequence with 32 mm2 × 32 mm2 field of view, 64 × all fMRI and electrophysiological recording experiments, a selected number of
64 matrix, flip angle 56°, and echo time/repetition time (TE/TR) = 20/1,000 ms. animals (n = 3 for VPM; n = 2 for POm) were transcardially perfused, and their
During MRI, anesthesia was maintained with 1.0% isoflurane. Animal heart rate, brains sectioned and prepared. Consecutive sections (120 μm apart) were
respiration rate, oxygen saturation, and rectal temperature were continuously prepared according to a previously published procedure (78). Double or triple
monitored (10, 43, 76). Note that the fiber outside brain was made opaque using immunofluorescence was assessed using a confocal scanning laser microscope.
heat-shrinkable sleeves to avoid artificial visual stimulation.
Data Availability. The data that support the findings of this study are available
fMRI and rsfMRI Data Analysis. fMRI analysis was performed using the pro- from the corresponding author on request.
cedure established in our previous studies (43, 76). In brief, all GE-EPI images
were first coregistered for each animal using SPM8 and registered to a ACKNOWLEDGMENTS. We thank Drs. I. Timofeev, M. M. Poo, X. Han,
custom-made brain template. For individual animals, data from fMRI ses- J. H. Lee, D. H. Sanes, J. F. He, and C. S. Poon for insightful comments; Drs.
A. To, S. J. Fan, K. K. Y. Wong, Y. I. Shih, Q. Li, L. Xu, X. Li, W. Chen, H. Lei, L. C. Ho,
sions were averaged, smoothed, and high-pass-filtered. Then, a general
M. Lyu, Y. Liu, and A. Mok for technical assistance; and Dr. K. Deisseroth for
linear was applied to calculate the response coefficient (β) maps for each providing us with the ChR2 viral construct. This work was supported by the
stimulus. Student’s t test was performed to identify activated voxels using Hong Kong Research Grant Council (HKU17103015 to E.X.W.), State Key
the threshold P < 0.001 and cluster size ≥4. BOLD temporal profiles were Laboratory of Pharmaceutical Biotechnology at the University of Hong
extracted from regions of interest (ROIs) delineated from the rat brain atlas. Kong, and a generous donation from the Lam Woo Foundation.

1. Stern P (2013) The heavily connected brain. Connection, connection, connection... 3. Park H-J, Friston K (2013) Structural and functional brain networks: From connections
Introduction. Science 342(6158):577. to cognition. Science 342(6158):1238411.
2. Oh SW, et al. (2014) A mesoscale connectome of the mouse brain. Nature 508(7495): 4. Kasthuri N, et al. (2015) Saturated reconstruction of a volume of neocortex. Cell
207–214. 162(3):648–661.

E8314 | www.pnas.org/cgi/doi/10.1073/pnas.1616361113 Leong et al.

 PNAS PLUS
5. Fox MD, Raichle ME (2007) Spontaneous fluctuations in brain activity observed with 42. Debanne D, Campanac E, Bialowas A, Carlier E, Alcaraz G (2011) Axon physiology.
functional magnetic resonance imaging. Nat Rev Neurosci 8(9):700–711. Physiol Rev 91(2):555–602.
6. Buckner RL, Krienen FM, Yeo BT (2013) Opportunities and limitations of intrinsic 43. Zhou IY, Cheung MM, Lau C, Chan KC, Wu EX (2012) Balanced steady-state free
functional connectivity MRI. Nat Neurosci 16(7):832–837. precession fMRI with intravascular susceptibility contrast agent. Magn Reson Med
7. Smith SM, et al.; WU-Minn HCP Consortium (2013) Resting-state fMRI in the Human 68(1):65–73.
Connectome Project. Neuroimage 80:144–168. 44. Pawela CP, et al. (2008) Modeling of region-specific fMRI BOLD neurovascular re-
8. Wang D, et al. (2015) Parcellating cortical functional networks in individuals. Nat sponse functions in rat brain reveals residual differences that correlate with the dif-
Neurosci 18(12):1853–1860. ferences in regional evoked potentials. Neuroimage 41(2):525–534.
9. Liang X, Zou Q, He Y, Yang Y (2013) Coupling of functional connectivity and regional 45. Lau C, et al. (2011) BOLD responses in the superior colliculus and lateral geniculate
cerebral blood flow reveals a physiological basis for network hubs of the human nucleus of the rat viewing an apparent motion stimulus. Neuroimage 58(3):878–884.
brain. Proc Natl Acad Sci USA 110(5):1929–1934. 46. Isaacson JS, Scanziani M (2011) How inhibition shapes cortical activity. Neuron 72(2):
10. Zhou IY, et al. (2014) Brain resting-state functional MRI connectivity: Morphological 231–243.
foundation and plasticity. Neuroimage 84:1–10. 47. Douglas RJ, Martin KA (2007) Recurrent neuronal circuits in the neocortex. Curr Biol
11. Lu H, et al. (2016) Low- but not high-frequency LFP correlates with spontaneous BOLD 17(13):R496–R500.
fluctuations in rat whisker barrel cortex. Cereb Cortex 26(2):683–694. 48. Markram H, et al. (2004) Interneurons of the neocortical inhibitory system. Nat Rev
12. Belcher AM, et al. (2016) Functional connectivity hubs and networks in the awake
Neurosci 5(10):793–807.
marmoset brain. Front Integr Nuerosci 10:9.
49. Sherman SM (2001) Tonic and burst firing: Dual modes of thalamocortical relay.
13. Boyden ES, Zhang F, Bamberg E, Nagel G, Deisseroth K (2005) Millisecond-timescale,
Trends Neurosci 24(2):122–126.
genetically targeted optical control of neural activity. Nat Neurosci 8(9):1263–1268.
50. Krahe R, Gabbiani F (2004) Burst firing in sensory systems. Nat Rev Neurosci 5(1):
14. Deisseroth K (2011) Optogenetics. Nat Methods 8(1):26–29.
13–23.
15. Steriade M, Contreras D, Curró Dossi R, Nuñez A (1993) The slow (< 1 Hz) oscillation in
51. Mease RA, Krieger P, Groh A (2014) Cortical control of adaptation and sensory relay
reticular thalamic and thalamocortical neurons: Scenario of sleep rhythm generation
mode in the thalamus. Proc Natl Acad Sci USA 111(18):6798–6803.
in interacting thalamic and neocortical networks. J Neurosci 13(8):3284–3299.
52. Steriade M, McCormick DA, Sejnowski TJ (1993) Thalamocortical oscillations in the
16. Crunelli V, Hughes SW (2010) The slow (<1 Hz) rhythm of non-REM sleep: A dialogue
sleeping and aroused brain. Science 262(5134):679–685.
between three cardinal oscillators. Nat Neurosci 13(1):9–17.
53. Swadlow HA, Gusev AG (2001) The impact of ‘bursting’ thalamic impulses at a neo-
17. Mohajerani MH, et al. (2013) Spontaneous cortical activity alternates between motifs
cortical synapse. Nat Neurosci 4(4):402–408.
defined by regional axonal projections. Nat Neurosci 16(10):1426–1435.
54. David F, et al. (2013) Essential thalamic contribution to slow waves of natural sleep.
18. Stroh A, et al. (2013) Making waves: Initiation and propagation of corticothalamic
Ca2+ waves in vivo. Neuron 77(6):1136–1150. J Neurosci 33(50):19599–19610.
19. Buzsáki G, Logothetis N, Singer W (2013) Scaling brain size, keeping timing: Evolu- 55. Bezdudnaya T, et al. (2006) Thalamic burst mode and inattention in the awake LGNd.
tionary preservation of brain rhythms. Neuron 80(3):751–764. Neuron 49(3):421–432.
20. Siegel M, Donner TH, Engel AK (2012) Spectral fingerprints of large-scale neuronal 56. Wallace MT, Ramachandran R, Stein BE (2004) A revised view of sensory cortical
interactions. Nat Rev Neurosci 13(2):121–134. parcellation. Proc Natl Acad Sci USA 101(7):2167–2172.
21. Wang L, Saalmann YB, Pinsk MA, Arcaro MJ, Kastner S (2012) Electrophysiological 57. Stein BE, Stanford TR, Rowland BA (2014) Development of multisensory integration
low-frequency coherence and cross-frequency coupling contribute to BOLD connec- from the perspective of the individual neuron. Nat Rev Neurosci 15(8):520–535.
tivity. Neuron 76(5):1010–1020. 58. Hillyard SA, Vogel EK, Luck SJ (1998) Sensory gain control (amplification) as a
22. Ekman M, Derrfuss J, Tittgemeyer M, Fiebach CJ (2012) Predicting errors from re- mechanism of selective attention: Electrophysiological and neuroimaging evidence.
configuration patterns in human brain networks. Proc Natl Acad Sci USA 109(41): Philos Trans R Soc Lond B Biol Sci 353(1373):1257–1270.
16714–16719. 59. Schwartz O, Simoncelli EP (2001) Natural signal statistics and sensory gain control. Nat

NEUROSCIENCE
23. Somogyi P, Tamás G, Lujan R, Buhl EH (1998) Salient features of synaptic organisation Neurosci 4(8):819–825.
in the cerebral cortex. Brain Res Brain Res Rev 26(2-3):113–135. 60. Zakiewicz IM, Bjaalie JG, Leergaard TB (2014) Brain-wide map of efferent projections
24. Brown SP, Hestrin S (2009) Intracortical circuits of pyramidal neurons reflect their from rat barrel cortex. Front Neuroinform 8:5.
long-range axonal targets. Nature 457(7233):1133–1136. 61. Ganguly K, Poo MM (2013) Activity-dependent neural plasticity from bench to bed-
25. Viaene AN, Petrof I, Sherman SM (2011) Synaptic properties of thalamic input to the side. Neuron 80(3):729–741.
subgranular layers of primary somatosensory and auditory cortices in the mouse. 62. Feller MB (1999) Spontaneous correlated activity in developing neural circuits. Neuron
J Neurosci 31(36):12738–12747. 22(4):653–656.
26. Viaene AN, Petrof I, Sherman SM (2011) Synaptic properties of thalamic input to 63. Katz LC, Shatz CJ (1996) Synaptic activity and the construction of cortical circuits.
layers 2/3 and 4 of primary somatosensory and auditory cortices. J Neurophysiol Science 274(5290):1133–1138.
105(1):279–292. 64. Tritsch NX, Yi E, Gale JE, Glowatzki E, Bergles DE (2007) The origin of spontaneous
27. MacLean JN, Watson BO, Aaron GB, Yuste R (2005) Internal dynamics determine the activity in the developing auditory system. Nature 450(7166):50–55.
cortical response to thalamic stimulation. Neuron 48(5):811–823. 65. Gao L, et al. (2009) Entrainment of slow oscillations of auditory thalamic neurons by
28. Poulet JF, Fernandez LM, Crochet S, Petersen CC (2012) Thalamic control of cortical repetitive sound stimuli. J Neurosci 29(18):6013–6021.
states. Nat Neurosci 15(3):370–372. 66. Timofeev I (2011) Neuronal plasticity and thalamocortical sleep and waking oscilla-
29. Alonso JM, Swadlow HA (2015) Thalamus controls recurrent cortical dynamics. Nat tions. Prog Brain Res 193:121–144.
Neurosci 18(12):1703–1704. 67. Schölvinck ML, Maier A, Ye FQ, Duyn JH, Leopold DA (2010) Neural basis of global
30. Fanselow EE, Sameshima K, Baccala LA, Nicolelis MA (2001) Thalamic bursting in rats resting-state fMRI activity. Proc Natl Acad Sci USA 107(22):10238–10243.
during different awake behavioral states. Proc Natl Acad Sci USA 98(26):15330–15335. 68. Ohno S, et al. (2012) A morphological analysis of thalamocortical axon fibers of rat
31. McCormick DA, McGinley MJ, Salkoff DB (2015) Brain state dependent activity in the posterior thalamic nuclei: A single neuron tracing study with viral vectors. Cereb
cortex and thalamus. Curr Opin Neurobiol 31:133–140. Cortex 22(12):2840–2857.
32. Logothetis NK, et al. (2010) The effects of electrical microstimulation on cortical signal 69. Viaene AN, Petrof I, Sherman SM (2011) Properties of the thalamic projection from
propagation. Nat Neurosci 13(10):1283–1291.
the posterior medial nucleus to primary and secondary somatosensory cortices in the
33. Millard DC, Whitmire CJ, Gollnick CA, Rozell CJ, Stanley GB (2015) Electrical and op-
mouse. Proc Natl Acad Sci USA 108(44):18156–18161.
tical activation of mesoscale neural circuits with implications for coding. J Neurosci
70. Diamond ME, von Heimendahl M, Knutsen PM, Kleinfeld D, Ahissar E (2008) ‘Where’
35(47):15702–15715.
and ‘what’ in the whisker sensorimotor system. Nat Rev Neurosci 9(8):601–612.
34. Cruikshank SJ, Urabe H, Nurmikko AV, Connors BW (2010) Pathway-specific feed-
71. Sherman SM, Guillery RW (1998) On the actions that one nerve cell can have on
forward circuits between thalamus and neocortex revealed by selective optical
another: Distinguishing “drivers” from “modulators.” Proc Natl Acad Sci USA 95(12):
stimulation of axons. Neuron 65(2):230–245.
7121–7126.
35. Crandall SR, Cruikshank SJ, Connors BW (2015) A corticothalamic switch: Controlling
72. Chung S, Li X, Nelson SB (2002) Short-term depression at thalamocortical synapses
the thalamus with dynamic synapses. Neuron 86(3):768–782.
contributes to rapid adaptation of cortical sensory responses in vivo. Neuron 34(3):
36. Benson DL, Isackson PJ, Hendry SH, Jones EG (1991) Differential gene expression for
glutamic acid decarboxylase and type II calcium-calmodulin-dependent protein kinase 437–446.
in basal ganglia, thalamus, and hypothalamus of the monkey. J Neurosci 11(6): 73. Bezdudnaya T, Keller A (2008) Laterodorsal nucleus of the thalamus: A processor of
1540–1564. somatosensory inputs. J Comp Neurol 507(6):1979–1989.
37. Logothetis NK, Pauls J, Augath M, Trinath T, Oeltermann A (2001) Neurophysiological 74. Yizhar O, Fenno LE, Davidson TJ, Mogri M, Deisseroth K (2011) Optogenetics in neural
investigation of the basis of the fMRI signal. Nature 412(6843):150–157. systems. Neuron 71(1):9–34.
38. Mattis J, et al. (2011) Principles for applying optogenetic tools derived from direct 75. Herman AM, Huang L, Murphey DK, Garcia I, Arenkiel BR (2014) Cell type-specific
comparative analysis of microbial opsins. Nat Methods 9(2):159–172. and time-dependent light exposure contribute to silencing in neurons expressing
39. Castro-Alamancos MA (2002) Properties of primary sensory (lemniscal) synapses in the Channelrhodopsin-2. eLife 3:e01481.
ventrobasal thalamus and the relay of high-frequency sensory inputs. J Neurophysiol 76. Gao PP, Zhang JW, Fan SJ, Sanes DH, Wu EX (2015) Auditory midbrain processing is
87(2):946–953. differentially modulated by auditory and visual cortices: An auditory fMRI study.
40. Landisman CE, Connors BW (2007) VPM and PoM nuclei of the rat somatosensory Neuroimage 123:22–32.
thalamus: Intrinsic neuronal properties and corticothalamic feedback. Cereb Cortex 77. Sparta DR, et al. (2011) Construction of implantable optical fibers for long-term op-
17(12):2853–2865. togenetic manipulation of neural circuits. Nat Protoc 7(1):12–23.
41. Ahissar E, Sosnik R, Haidarliu S (2000) Transformation from temporal to rate coding in 78. Lee JH, et al. (2010) Global and local fMRI signals driven by neurons defined opto-
a somatosensory thalamocortical pathway. Nature 406(6793):302–306. genetically by type and wiring. Nature 465(7299):788–792.

Leong et al. PNAS | Published online December 5, 2016 | E8315