Persistent and ramping neural activity in frontal cortex anticipates specific movements
1–6
. Preparatory activity is distributed across multiple brain regions
7,8
, but it is unclear which brain areas are involved and how this activity is mediated
by multi-regional interactions. The cerebellum is a brain structure thought to be
primarily involved in short timescale control of movement
9–12
, but roles in cognitive processes have also been proposed
13–16
. In humans, cerebellar damage can cause defects in planning and working memory
13
. We report that persistent representation of information in frontal cortex during
motor planning is dependent on the cerebellum. Mice performed a sensory discrimination
task in which they used short-term memory to plan a future directional movement. A
transient perturbation in the medial deep cerebellar nucleus (fastigial nucleus) disrupted
correct subsequent responses, without hampering movement execution. Preparatory activity
was observed in both frontal cortex and cerebellar nuclei, seconds before movement
onset. Silencing frontal cortex activity abolished preparatory activity in the cerebellar
nuclei. Fastigial activity was necessary for maintaining cortical preparatory activity.
Fastigial output selectively targeted the behaviorally relevant part of frontal cortex
through the thalamus, closing a cortico-cerebellar loop. Our results support the view
that persistent neural dynamics during motor planning is maintained by multi-regional
neural circuits
17
, and that cerebellar computations extend beyond online motor control
13–15,18
.
Head-fixed mice discriminated object locations (anterior or posterior) presented during
a sample epoch (1.3 s) using their whiskers (Fig. 1a). Mice held their responses in
short-term memory and planned for the upcoming movement during a delay epoch (1.3
s). After the delay epoch, an auditory ‘go’ cue (0.1s) signaled the response epoch
and mice reported pole location by skilled directional licking (left, right) to obtain
a water reward.
During the delay epoch, preparatory activity in anterior lateral motor cortex (ALM;
Fig 1b) is critical for correct subsequent movements
19–22
. ALM neurons exhibit persistent and ramping activity that predicts specific future
movements, long before movement onset.
ALM projects to the cerebellum via the basal pontine nucleus
23,24
. We tested cerebellar involvement in the delayed response task by lesioning the fastigial
or dentate nucleus (Fig 1c, Methods). We separately examined behavioral choice (the
direction of the first lick after the ‘go’ cue) and movement execution (licking duration
and frequency). Fastigial lesions impaired initiation of contralesional licking (Fig
1d) and slowed initiation of contralesional licking following the ‘go’ cue (Fig 1e).
After movement initiation licking was unaffected (Fig 1e, Extended Data Fig 1b–c).
Dentate lesions (~1.5 mm away) did not change behavioral choice and movement execution
(Fig 1d).
We tested fastigial involvement during specific behavioral epochs by disrupting activity
using channelrhodopsin-2 (ChR2) activation (photo-activation, Fig 1f, Methods). Fastigial
photo-activation during the early sample epoch had no effect on behavioral choice
(p>0.05 for either trial type, bootstrap, Methods). Photo-activation during the delay
and response epochs both resulted in incorrect choices (Fig 1g–h). Incorrect choices
were induced even when the photostimuli ceased 800 ms before the movement. Photo-activation
during the delay epoch biased future movements to either the contralateral or ipsilateral
direction in individual mice (Fig 1h; Extended Data Fig 1e). Photo-activation immediately
before movement onset (response epoch) produced a contralateral bias (Fig 1h), opposite
from the directional bias induced by fastigial lesions. Distinct patterns of bias
induced by delay versus response epoch photo-activation suggest temporally-specific
roles of the fastigial nucleus in driving movement planning versus initiation.
Fastigial photo-activation did not block or perturb the frequency and duration of
licking movements after initiation of licking bouts (Fig 1i). Fastigial photo-activation
during the delay epoch did not evoke early licks (Extended Data Fig 1f). These data
show that rhythmic licking is independent of fastigial activity, but they do not rule
out fastigial involvement in controlling the direction and timing of licking bouts
via its descending input to oscillators in the medulla that drive licking
25
. Dentate photo-activation produced little effect on behavioral choice (Fig 1g, p=0.06,
bootstrap; Extended Data Fig 1e), consistent with the lesion results.
We tested for necessity by silencing fastigial activity. We photostimulated ChR2 expressing
Purkinje cells, which inhibited CN activity (photo-inhibition, Fig 1j; Extended Data
Fig 2, Methods). Photo-inhibition during the early sample epoch did not affect behavioral
choice (Fig 1k, p>0.05 for either trial type, bootstrap, Methods). Photo-inhibition
during the late sample and delay epochs resulted in incorrect behavioral choices,
with a larger effect induced in the delay epoch (Fig 1k, Extended Data Fig 2g–h).
Photo-inhibition during the delay epoch biased future movements to either the contralateral
or ipsilateral direction in individual mice (Extended Data Fig 2g–h). Photo-inhibition
during movement initiation (response epoch) produced an ipsilateral bias (Fig 1k),
consistent with fastigial lesions, and opposite from the bias induced by fastigial
photo-activation. Photo-inhibition minimally affected movement execution (Fig 1l).
No other movements were consistently evoked during photo-inhibition (Methods, Extended
Data Figs 2e–f, 2i).
Thus, fastigial activity, starting from the late sample epoch and throughout the delay
epoch, is required for correct choice.
We compared the activity in ALM and CN by recording from single-units (ALM, n=1194
units; CN, n=564; Extended Data Fig 3; Methods). Most ALM pyramidal neurons distinguished
“lick left” and “lick right” trials in spike counts (980/1194, p<0.05, two-tailed
t-test, Methods). A similar proportion of CN neurons showed trial-type selectivity
(416/564). Individual ALM and CN neurons exhibited diverse activity patterns, including
selective persistent activity and ramping activity during the delay epoch (Fig 2a–b).
Preparatory activity (selectivity before the response epoch) was more prevalent in
ALM (ALM, 59% of neurons, CN, 35%; p<0.001, chi-square test). Peri-movement activity
(selectivity during response epoch) was equally prevalent in both regions (ALM, 67%,
CN, 66%). In both regions, a similar proportion (<10%) of neurons modulated their
spike rates during individual licks (Extended Data Fig 4).
Selectivity for “lick left” and “lick right” trials emerged gradually in both ALM
and CN population averages (Fig 2c–d), starting during the sample epoch and reaching
a maximum at the response epoch. On error trials, mice licked in the opposite direction
to the instruction provided by object location. A small group of CN neurons signaled
object location. However, most CN neurons predicted the upcoming licking direction
(67%, Extended Data Fig 4), similar to ALM
19,20
. Thus CN delay epoch selectivity is related to future movement, a hallmark of preparatory
activity that instructs subsequent movement.
Preparatory activity was present in all three cerebellar nuclei (Extended Data Fig
3–4). In the fastigial nucleus, distinct patterns of selectivity were observed during
the early versus late delay epoch, where equal numbers of neurons preferred either
contralateral or ipsilateral movement during the early delay, and a population-level
preference for contralateral movement developed right before movement initiation (Extended
Data Fig 4g)
23
, consistent with the directional biases induced by fastigial lesion and optogenetic
manipulations during movement initiation (Fig 1). These results show wide-spread activity
related to motor planning and movement initiation in the CN.
The cerebellum is thought to process information over timescales of hundreds of milliseconds
26–28
. Cerebellar preparatory activity lasting over seconds could be inherited from ALM.
We recorded from the CN while bilaterally photo-inhibiting ALM during the delay epoch
(Fig 3a; Methods). Activity of many CN neurons was rapidly altered by ALM photo-inhibition
(185/389, p<0.01, two-tailed t-test; latency, 10.7 ± 4.0 ms, mean ± s.e.m.), including
neurons with selective persistent and ramping activity (Extended Data Fig 5a). Photo-inhibiting
ALM caused a greater proportion of CN neurons to increase their activity, resulting
in net disinhibition of CN (Fig 3b, 123/185 increase, 62/185 decrease, p<0.001, binomial
test). CN selectivity for “lick left” and “lick right” trials was abolished by photo-inhibiting
ALM (Fig 3c, Extended Data Fig 5b–d).
Photo-inhibiting either side of ALM reduced CN preparatory activity (Extended Data
Fig 6a–d). Photo-inhibiting motor cortex regions posterior to ALM produced less effect
on CN preparatory activity (Extended Data Fig 6e). Thus ALM appears to provide widespread
input to the cerebellum to drive CN preparatory activity during motor planning.
We tested whether ALM preparatory activity depends on CN activity by recording from
ALM while photo-inhibiting the CN during the delay epoch (Fig 3d; Methods). Activity
of many ALM neurons was rapidly altered by CN photo-inhibition (197/454, p<0.01, two-tailed
t-test, Extended Data Fig 5e; latency, 7.1 ± 3.8 ms). However, the average spike rate
was nearly unchanged (control: 4.15 spikes/s; photo-inhibition: 4.39 spikes/s; p=0.58,
two-tailed t-test). Individual ALM neurons increased (108/197) or decreased (89/197)
their responses during CN photo-inhibition (Fig 3e). ALM trial-type selectivity was
abolished by CN photo-inhibition (Fig 3f), similar to partial inactivation of the
thalamus
17
. Thus, CN activity was necessary for maintaining cortical selectivity for planned
movement and preparatory activity is maintained in a cortico-cerebellar loop.
Preparatory activity requires reciprocal excitation between ALM and thalamus
17
. We mapped connectivity between CN, thalamus and ALM using retrograde and anterograde
tracers in ALM and CN (Methods, Fig 4a, Extended Data Fig 7–8). Fastigial projections
mostly overlapped with parts of the thalamus that project to ALM (i.e., ventral medial
nucleus, VM, and parts of anterior-lateral nucleus, VAL), whereas dentate projections
targeted an adjacent region that was more dorsal and lateral (i.e., primarily VAL)
(Fig 4b–c).
We probed the influence of CN on ALM thalamo-cortical loop during motor planning.
We perturbed fastigial or dentate activity using ChR2 activation during the early
delay epoch while monitoring ALM preparatory activity (Fig 4d). Fastigial photo-activation
induced rapid changes in ALM activity (latency, 4.5 ± 2.6 ms; Extended Data Fig 9a),
consistent with a disynaptic pathway through the thalamus. Selective persistent and
ramping activity was abolished in most ALM neurons (Fig 4e, top). In a minority of
neurons, photo-activation created new preparatory activity (Fig 4e, bottom). Transient
photo-activation of the fastigial nucleus altered ALM preparatory activity long (>1
s) after cessation of the photostimulus (Fig 4f; Extended Data Fig 9b). Dentate photo-activation
also induced ALM response changes (Fig 4g; latency, 8.3 ± 5.0 ms). Importantly, activity
quickly recovered after dentate photo-activation (Fig 4g).
We disambiguated the effect of fastigial and dentate photo-activations on ALM preparatory
activity. We analyzed population dynamics in ‘activity space’, where each dimension
corresponds to activity of one neuron
6,20,29
. We estimated a ‘coding direction’ (
cd
) in activity space along which preparatory activity maximally discriminated upcoming
licking directions (Fig 4h, Methods). In control trials, activity trajectories from
“lick left” and “lick right” trials diverged from each other along the
cd
during the sample epoch and remained separated throughout the delay epoch, providing
a stable neural substrate for a short-term memory of the future movement (Fig 4i)
20
. Fastigial photo-activation resulted in a collapse of activity trajectories. Additional
experiments showed that fastigial photo-activation destroyed the relationship of activity
trajectories to specific future movements (Supplementary Information, Extended Data
Fig 10). In contrast, dentate photo-activation minimally impacted activity trajectories
along the
cd
, despite having an impact on ALM neuron responses (Fig 4i, Extended Data Fig 9c).
Thus fastigial photo-activation destroyed the coding of future movement, whereas dentate
photo-activation did not.
Our results show that the cerebellum is critical for the coding of future movement
in frontal cortex. Given its roles in movement timing
26
, the cerebellum may provide an urgency drive that is necessary for the emergence
of selectivity in ALM thalamo-cortical loop
17
in preparation of movement, consistent with attractor models of motor planning
30
.
Methods
Mice
This study is based on data from 70 mice (age > P60, both male and female mice, Supplemental
Table 1). 13 C57B1/6 mice were used for CN ChR2 photo-activation experiments. A subset
of these mice (10 mice) were used for ALM recordings during photo-activation of the
CN. Another subset (2 mice) and 6 additional C57B1/6 mice were used for CN lesion
experiments (8 mice). 7 L7-cre
31
crossed to Ai32 (Rosa26-LSL-ChR2-EYFP, JAX Stock#012569)
32
mice were used for CN photo-inhibition experiments during behavior. A subset of these
mice (4 mice) were used for ALM recordings during CN photo-inhibition. 2 additional
L7-cre × Ai32 mice were used to characterize CN photo-inhibition using electrophysiology.
6 additional L7-cre × Ai32 mice (not listed in Supplemental Table 1) were used to
quantify general effects of CN photo-inhibition on movements. 10 C57B1/6 mice were
used for CN recording experiments. 8 VGAT-ChR2-EYFP mice (Jackson laboratory, JAX
Stock#014548)
33
were used for CN recordings during ALM photo-inhibition. 4 VGAT-ChR2-EYFP mice were
used for behavioral experiments in which both ALM and CN activity were manipulated
to test the causal role of ALM activity in behavior following a fastigial perturbation.
12 C57B1/6 mice (not listed in Supplemental Table 1) were used for anatomical tracing.
All procedures were in accordance with protocols approved by the Institutional Animal
Care and Use Committees at Baylor College of Medicine, Janelia Research Campus, and
Erasmus Medical Center. Mice were housed in a 12:12 reverse light:dark cycle and tested
during the dark phase. On days not tested, mice received 0.5–1 mL of water. On other
days, mice were tested in experimental sessions lasting 1 to 2 hours where they received
all their water (0.3 to 2 mL). If mice did not maintain a stable body weight, they
received supplementary water
34
. All surgical procedures were carried out aseptically under 1–2 % isoflurane anesthesia.
Buprenorphine HCl (0.1 mg/kg) or Sustained Release Meloxicam (4mg/kg) was used for
pre- and post-operative analgesia. Ketoprofen (5 mg/kg) was used at the time of surgery
and post-operatively to reduce inflammation. After the surgery, mice were allowed
to recover for at least three days with free access to water before water restriction.
Surgery
Mice were prepared for photostimulation and electrophysiology with a clear-skull cap
19
and a headpost
34
. The scalp and periosteum over the dorsal surface of the skull were removed. A layer
of cyanoacrylate adhesive (Krazy glue, Elmer’s Products Inc) was directly applied
to the intact skull. A custom made headpost was placed on the skull (approximately
over visual cortex) and cemented in place with clear dental acrylic (Lang Dental Jet
Repair Acrylic; Part# 1223-clear). A thin layer of clear dental acrylic was applied
over the cyanoacrylate adhesive covering the entire exposed skull, followed by a thin
layer of clear nail polish (Electron Microscopy Sciences, Part# 72180).
In wild-type mice prepared for CN ChR2 photo-activation experiments, 150 nL of AAV2-hSyn1-(h134R)ChR2-EYFP
virus (UNC viral core, titer 5.712 vg/ml) was injected in the fastigial nucleus (posterior
2.7 mm from lambda, lateral 0.5 mm, depth 2.3 mm) or dentate nucleus (posterior 2.5
mm lambda, lateral 2.7 mm, depth 2.3 mm), followed by implantation of a 2 mm long
optical fiber (Thorlabs, Part#CFML12L02) over the injection site. In VGAT-ChR2-EYFP
mice prepared for fastigial perturbation experiments, an optical fiber was implanted
at the same stereotaxic coordinate to target the fastigial nucleus.
Behavior
The behavioral task and training have been described
19,34
. The stimulus was a metal pin (0.9 mm in diameter), presented at one of two possible
positions (Fig. 1a). The two pole positions were 5 mm apart along the anterior-posterior
axis and were constant across sessions. The posterior pole position was 5 mm from
the whisker pad. A two-spout lickport (4.5 mm between spouts) was used to deliver
water rewards and record answer licks. Behavioral data were acquired using commercial
hardware and software (Bpod, Sanworks).
At the beginning of each trial, the vertical pole moved into reach of the whiskers
(0.2 s travel time), where it remained for 1 second, after which it was retracted
(retraction time 0.2 s). The sample epoch was defined as the time between the pole
movement onset to 0.1 s after the pole retraction onset (sample epoch, 1.3 s, Fig.
1a). Mice touched the object at both pole positions, typically with a different set
of whiskers. The delay epoch (durations, 1.3 s) followed the sample epoch. An auditory
‘go’ cue indicated the end of the delay epoch (pure tone, 3.4 kHz, 0.1 s duration).
Licking early during the trial was punished by a loud alarm sound (siren buzzer, 0.05
s duration), followed by a brief timeout (1–1.2 s). Licking the correct lickport after
the ‘go’ cue led to a liquid reward (2–3 µL). Licking the incorrect lickport triggered
a timeout (2–6 s). Trials in which mice did not lick within a 1.5 second window after
the ‘go’ cue (‘no lick’) were rare and typically occurred at the end of a session.
Reaction time was from the ‘go’ cue onset to the first lickport contact.
CN Lesion
In 8 wild-type mice, we performed localized lesions in the CN. Intense 473nm light
(>10mW) was delivered to either the fastigial or dentate nucleus through the implanted
optical fiber for 5–10 min while the mice were under 2 % isoflurane anesthesia. Light-induced
heating produced localized lesion around the fiber tip. The lesion areas were estimated
post hoc, by measuring the size of areas in which neuron labelling were absent (anti-NeuN
staining, Millipore ABN78, 1:1000, Fig 1c, Extended Data Fig 1a). Lesion sizes were
between 602–878 um, close to the diameters of the fastigial nucleus or the dentate
nucleus. The mice were tested on subsequent days. See Supplemental Table 1 for the
list of mice and lesioned hemispheres.
CN ChR2 photo-activation
For ChR2 photo-activation of the CN, 13 wild-type mice injected with AAV2-hSyn1-(h134R)ChR2-EYFP
virus were used. Light was delivered to the CN through an optical fiber. In 6 mice,
both the fastigial nucleus and the dentate nucleus on opposite hemispheres were tested:
left fastigial/right dentate nucleus in 3 mice and right fastigial/left dentate in
the other 3 mice. In those mice, the fastigial and dentate nucleus were tested on
separate days. In another 3 mice, only the right fastigial nucleus was tested. In
another 4 mice, only the right dentate nucleus was tested. In 1 of the fastigial-stimulation
mice, the injection of ChR2 virus missed the CN and its data were excluded. See Supplemental
Table 1 for the list of mice and manipulated hemispheres in individual experiments.
Light from a 473 nm laser (Laser Quantum, Part# Gem 473) was controlled by an acousto-optical
modulator (AOM; Quanta Tech) and a shutter (Vincent Associates). To prevent the mice
from distinguishing photostimulation trials from control trials using visual cues,
a ‘masking flash’ was delivered using 470 nm LEDs (Luxeon Star) near the eyes of the
mice. The masking flash began as the pole started to move and continued through the
end of the epoch in which photostimulation could occur. Photostimulation was deployed
on 25% of the behavioral trials. During ALM recordings, photostimulation was deployed
on 33% of the trials. The photostimulus was pulses of light (5 ms pulse duration)
delivered at 20 Hz and a range of peak powers (5, 10, 15mW). The power values reported
in the paper indicate average powers (0.5, 1, 1.5 mW). The powers were measured at
the fiber tip. The photostimulus started at the beginning of a task epoch and continued
for either 0.455 s (10 pulses) or 1.255 s (26 pulses).
We performed photostimulation both during early parts of the sample and delay epochs
(0.455 s; Fig 1g–h), as well as the entire epochs (1.255 s; Extended Data Fig 1e).
Our primary experiments were centered around sub-epoch manipulations (early sample
and early delay). This experimental design was chosen for two reasons. First, sub-epoch
manipulations sought to better resolve specific involvements of the CN in time. Manipulations
during the entire sample epoch (1.3s) may affect the beginning of motor planning process,
and manipulations during the entire delay epoch up to the ‘go’ cue may interfere with
movement initiation
20
. Second, we sought to use the early delay epoch manipulation to examine any persistent
effect on preparatory activity (Fig 4). If a brain region is involved in maintaining
preparatory activity, transiently disrupting it during the early delay should disrupt
preparatory activity subsequently. We later tested manipulations that spanned the
entire behavioral epochs and included the response epoch to examine the involvement
of the CN in movement initiation and execution (Fig 1h–i, Extended Data Fig 1e).
CN photo-inhibition
In L7-cre × Ai32 mice, ChR2 was expressed in cerebellar Purkinje cells. We photostimulated
Purkinje cells to inhibit neurons in the CN (Fig 1j). To characterize the effect of
CN photo-inhibition (Extended Data Fig 2), we performed silicon probe recordings in
the fastigial nucleus (details described in Electrophysiology). Photostimulation of
Purkinje cells was carried out by directing a blue laser (beam diameter: 400 µm at
4σ) to the brain surface through a cranial window. The photostimulus was a 40 Hz sinusoid
(average power, 4.5 mW) lasting for 1.3 sec, including a 100–200ms linear ramp during
the laser offset to reduce rebound neuronal activity. Photostimulation was deployed
on 25% of the behavioral trials. During ALM recordings, photostimulation was deployed
on 33% of the trials. Masking flash was delivered on all trials. Photostimulation
was performed at two different locations: 3 mm posterior and 2 mm lateral from lambda
(the cerebellar cortex), 1 mm posterior and 2 mm lateral to lambda (the inferior colliculus).
Photostimulation of the cerebellar cortex silenced fastigial activity whereas photostimulation
of the inferior colliculus did not (Extended Data Fig 2).
For photo-inhibition of the CN during behavior, we tested 7 L7-cre × Ai32 mice. Photostimulation
of Purkinje cells was carried out by directing the blue laser to the brain surface
through a cranial window. We manipulated the left hemisphere in 3 mice and the right
hemisphere in 4 mice. See Supplemental Table 1 for the list of mice and manipulated
hemispheres in individual experiments. The photostimulus (average power, 0.5, 1.5,
or 4.5 mW) started at the beginning of a task epoch and continued for either 0.5 or
1.3 sec, including the 100–200ms linear ramp.
We performed CN photo-inhibition both during parts of the behavioral epochs (0.5 s;
Fig 1k), as well as the entire epochs (1.3 s; Extended Data Fig 2g), for the same
reasons described for CN ChR2 photo-activation. In addition to early sample and early
delay photo-inhibitions, we also included late sample and late delay photo-inhibitions
to resolve CN involvements during specific sub-epochs of the delayed response task
(Extended Data Fig 2h). 7 mice were initially tested for sub-epoch photo-inhibitions.
Later, 3 of the 7 mice were tested with photo-inhibitions during the entire behavior
epochs to confirm the directional bias induced by response epoch photo-inhibition
(Extended Data Fig 2g, Supplemental Table 1).
In 4 L7-cre × Ai32 mice (Supplemental Table 1), we recorded from ALM during contralateral
CN photo-inhibition. Photo-inhibition (average power, 1.5 or 4.5 mW) started at the
beginning of the delay epoch and continued for 1.3 sec, including the 100–200ms linear
ramp.
CN photo-inhibition and effects on movement
In 6 naïve L7-cre × Ai32 mice, we examined the general effects of CN photo-inhibition
on movements. Two optical fibers were implanted in the fastigial and dentate nucleus
on opposite hemispheres to photostimulate Purkinje cell axons in the CN. We tested
a range of laser powers (average power, 1.5, 5, 10, 20 mW), including the photostimulus
intensities used to photo-inhibit CN activity in the delayed response task (1.5 mW).
The photostimulus was pulses of light (5–10 ms pulse duration, 20–40 Hz, 15–50mW peak
powers). Videos (Point Grey, Part# CM3-US-13Y3M-CS or Basler, Part# acA640–300gm)
were recorded while mice underwent photo-inhibition either in head-fixed configuration
or unrestrained. In head-fixed mice (n=6), jaw movement and eyeblink were tracked
using custom software. In unrestrained mice (n=2), the effect of photo-inhibition
on movements and postures were manually scored on a scale of 0 to 4 that rated dystonia-like
movements
35
: 0 - no motor abnormalities; 1 - slightly slowed or abnormal motor behavior, no dystonia;
2 - mild impairment, mild and transient dystonic postures, weak tremor; 3 - moderate
impairment, dystonic postures, cannot balance the body, major tremor; 4 - severe impairment,
sustained dystonic postures and limited movements. Two individuals scored the videos
independently and the scores were averaged. The scoring was blinded to the experimental
conditions.
In head-fixed mice, high-intensity photo-inhibition (20 mW) of the dentate nucleus
produced jaw movements (Extended Data Fig 2f). No tongue protrusion was observed in
all photo-inhibition conditions. No consistent eyeblink was evoked. At the photo-inhibition
intensities used in the delay response task (1.5–5mW), no movement was consistent
evoked.
In unrestrained mice, high-intensity photo-inhibition (20 mW) of both the fastigial
and dentate nucleus produced dystonia-like movements and posture changes. This included
loss of balance and extension of the contralateral limbs (Extended Data Fig 2e). However,
at the photo-inhibition intensities used in the delayed response task (1.5–5 mW),
no consistent movement was detected (Extended Data Fig 2e).
ALM photo-inhibition
ALM is centered on bregma anterior 2.5 mm, lateral 1.5 mm
19–21
. For photo-inhibition of ALM, we photostimulated cortical GABAergic neurons in VGAT-ChR2-EYFP
mice (8 mice). Photostimulation was performed through the clear-skull cap implant
by directing the blue laser over the skull (beam diameter: 400 µm at 4σ, bregma anterior
2.5 mm, lateral 1.5 mm). The light transmission through the intact skull was 50%
19
. Photo-inhibition was deployed on 25% of the behavioral trials. During CN recordings,
photo-inhibition was deployed on 33% of the trials. Masking flash was delivered on
all trials.
We photo-inhibited ALM for 1.3 s at the beginning of the delay epoch, including a
100 ms linear ramp at the laser offset to minimize rebound excitation. This photostimulus
was empirically determined to produce robust photo-inhibition in ALM
19,20
. The photo-inhibition silenced 90% of spikes in a cortical area of 1mm radius (at
half-max) through all cortical layers. For unilateral ALM photo-inhibition (Extended
Data Figs 6, 10), we used a 40 Hz sinusoidal photostimulus (1.5mW average power at
the skull surface) at 2.5 mm anterior and 1.5 mm lateral from bregma. For bilateral
ALM photo-inhibition (Fig 3, Extended Data Figs 5, 6, 10), we used a constant photostimulus
and a scanning galvo (GVSM002, Thorlabs), which stepped the laser beam sequentially
through the photo-inhibition sites at the rate of 1 step per 5 ms (step time: 0.2
ms; dwell time: 4.8 ms; measured using a photodiode). 8 photo-inhibition sites were
spaced in 1 mm at anterior 2–3 mm and lateral 1–2 mm from bregma, covering ALM. Peak
power was adjusted based on the number of photo-inhibition sites to achieve 1.5 mW
average power per site.
ALM photo-inhibition after a fastigial perturbation
To test the causal role of ALM activity in driving directional licking after a fastigial
perturbation, we tested 4 VGAT-ChR2-EYFP mice (Extended Data Fig 10). We confirmed
that ChR2 was expressed in Purkinje cells as well as other GABAergic neurons in the
cerebellum
33
. Photostimulation (40 Hz sinusoid; average power, 1.5 mW) was delievered to the fastigial
nucleus through an optical fiber to perturb CN activity. Photostimulation started
at the onset of the delay epoch and always lasted for 500 ms. We targeted the left
fastigial nucleus in 3 mice and the right fastigial nucleus in 1 mouse. See Supplemental
Table 1 for the list of mice and manipulated hemispheres. CN perturbations in VGAT-ChR2-EYFP
mice might be different from CN photo-activation using ChR2 virus and photo-inhibition
in L7-cre × Ai32 mice. We thus first confirmed that in VGAT-ChR2-EYFP mice, fastigial
perturbations during the early delay epoch reduced behavioral performance for both
trial types (control: “contra lick” trials, 74.5% correct, “ipsi lick” trials, 67.8%;
photostimulation: 58.5% and 57.0%; p<0.05 for both trial types, bootstrap), similar
to other CN optogenetic manipulations during the early delay epoch (c.f. Fig 1h, k).
Photo-inhibition of ALM was performed as described above. For unilateral ALM photo-inhibition,
photostimulation started 800 ms after the onset of the delay epoch and ended at the
onset of the ‘go’ cue (500ms in duration). Unilateral ALM photo-inhibition before
the ‘go’ cue was previously found to bias the upcoming licking to the ipsilateral
direction
20
. We confirmed the ipsilateral bias here (Extended Data Fig 6b, 10e). In 3 VGAT-ChR2-EYFP
mice, we also tested bilateral ALM photo-inhibition. Bilateral photostimulation started
at the onset of the ‘go’ cue and lasted for 1.3 s. On randomly interleaved trials,
we either perturbed the fastigial nucleus, or photo-inhibited ALM (unilateral or bilateral),
or perturbed the fastigial nucleus followed by ALM photo-inhibition (Extended Data
Fig 10f–g). Photostimulation was deployed on 25% of the behavioral trials. Masking
flash was delivered on all trials.
Behavioral data analysis
Performance was computed as the fraction of correct choices, excluding lick early
trials and no lick trials (Fig 1g, Extended Data Figs 1e, 2g, 10b). Chance performance
was 50%. We also separately computed the performance for “lick right” and “lick left”
trials (Fig 1; Extended Data Figs 1, 2, 6, 10). Significance of the performance change
in each photostimulation condition was determined using bootstrap to account for variability
across mice, sessions, and trials. We tested against the null hypothesis that the
performance change caused by photostimulation was due to normal behavioral variability.
In each round of bootstrap, we replaced the original behavioral dataset with a re-sampled
dataset in which we re-sampled with replacement from: 1) mice, 2) sessions performed
by each mice, 3) the trials within each session. We then computed the performance
change on the re-sampled dataset. Repeating this procedure 10,000 times produced a
distribution of performance changes that reflected the behavioral variability. The
p-value of the observed performance change was computed as the fraction of times the
bootstrap produced an inconsistent performance change (e.g. if a performance decrease
was observed during photostimulation, the p-value was the fraction of times a performance
increase was observed during bootstrap, one-tailed test).
Electrophysiology
Extracellular spikes were recorded using 32-channel NeuroNexus silicon probes (Part#
A4×8–5mm-100–200-177) or 64-channel Cambridge NeuroTech silicon probes (H2 acute probe,
25 µm spacing, 2 shanks). The 32-channel voltage signals were multiplexed, digitized
by a PCI6133 board at 400 kHz (National Instruments) at 14 bit, demultiplexed (sampling
at 25,000 Hz) and stored for offline analysis. The 64-channel voltage signals were
amplified and digitized on an Intan RHD2164 64-Channel Amplifier Board (Intan Technology)
at 16 bit, recorded on an Intan RHD2000-Series Amplifier Evaluation System (sampling
at 20,000 Hz) using Open-Source RHD2000 Interface Software from Intan Technology (version
1.5.2), and stored for offline analysis.
For ALM recordings, a small craniotomy (diameter, <1 mm) was made one day prior to
the recording session (2.5 mm anterior, 1.5 mm lateral from bregma)
19
. A silicon probe was acutely inserted 0.9–1.11 mm below the brain surface prior to
the start of each recording session. To minimize brain movement, a drop of silicone
gel (3–4680, Dow Corning, Midland, MI) was applied over the craniotomy after the electrode
was in the tissue. The tissue was allowed to settle for several minutes before the
recording started. 2 to 6 recordings were made from each craniotomy. In 10 mice, recordings
were from the left ALM. In 2 mice, recordings were from the right ALM. ALM recordings
were always contralateral to the side of CN manipulations (Supplemental Table 1).
For CN recordings, two craniotomies (diameter, 1.5 mm) were made successively over
the left cerebellum to target the CN (first at 2.5 mm posterior, 1.5mm lateral from
lambda; then at 2 mm posterior, 3 mm lateral from lambda). After a craniotomy was
opened, recordings were made at different locations with the craniotomy over the next
5–7 days. A silicon probe was acutely inserted 2.1–2.4 mm below the brain surface
prior to the start of each session. On the last recording session, the silicon probe
was painted with DiI to label the recording track. All the recording locations within
a craniotomy were reconstructed post-hoc relative to the labeled track (Extended Data
Fig 3). Reconstructions were based on the insertion locations on the brain surface.
Recording depths were inferred from manipulator readings and verified based on histology.
The second craniotomy was opened after the first craniotomy has been sampled.
Electrophysiology data analysis
The extracellular recording traces were band-pass filtered (300–6 kHz). Events that
exceeded an amplitude threshold (4 standard deviations of the background) were subjected
to manual spike sorting to extract single-units
19
.
For ALM recordings during behavior (Fig 2), 1309 single-units were recorded in 10
mice across 39 sessions. Spike widths were computed as the trough-to-peak interval
in the mean spike waveform. Units with spike width < 0.35 ms were defined as fast-spiking
neurons (82/1309) and units with spike widths > 0.45 ms as putative pyramidal neurons
(1194/1309). Units with intermediate values (0.35 – 0.45 ms, 33/1309) were excluded
from analyses. This classification was previously verified by optogenetic tagging
of GABAergic neurons
19
. We concentrated our analyses on the putative pyramidal neurons (n=1194). 9 of the
10 mice were tested with CN ChR2 photo-activation during ALM recordings (Fig 4, Extended
Data Fig 10; n=912 putative pyramidal neurons). 1 mouse was excluded because the injection
of ChR2 virus missed the CN. For ALM recordings during CN photo-inhibition (Fig 3,
Extended Data Fig 5), 490 single-units were recorded in 4 mice across 19 sessions.
454/490 units were putative pyramidal neurons.
For CN recordings during behavior, 878 single-units were recorded in 18 mice across
169 sessions (Fig 2). CN single-units were not classified based on spike shapes. We
estimated unit locations based on recording track labeling, recording depth, and the
lamination of activity patterns across the recording shanks (Extended Data Fig 3).
CN boundaries were visible in DAPI staining (Extended Data Fig 3). 314 units were
from outside of the CN (mostly from the cerebellar cortex). 268 units were from the
fastigial nucleus, 139 units were from the interposed nucleus, and 157 units were
from the dentate nucleus. 8 of the 18 mice were tested with ALM photo-inhibition during
CN recordings (Fig 3, Extended Data Figs 5, 6; n=389).
To characterize CN photo-inhibition in L7-cre × Ai32 mice, 60 single-units were recorded
in 2 mice across 11 recording sessions (Fig 1j, Extended Data Fig 2). The recordings
were targeted to the left fastigial nucleus. 34 units were from outside of CN (mostly
from the cerebellar cortex). 26 units were from the fastigial nucleus. The mice were
awake but non-behaving during the recordings.
Neurons were tested for significant trial-type selectivity during the sample, delay,
or response epochs, using the spike counts from the “lick left” and “lick right” trials
(two-tailed t-test, p<0.05). Neurons that significantly differentiated trial types
during any one of the epochs were deemed “selective”. To compute selectivity in Figures
2, 3 and Extended Data Figures 5, 6, we first determined each neuron’s preferred trial
type using spike counts from a subset of the trials (10 trials for ALM, 20 for CN),
selectivity was calculated as the spike rate difference between “lick left” and “lick
right” trials on the remaining data. Selecting preferred trial type using either sample,
delay, or response epoch data yielded qualitatively similar preparatory activity.
To quantify the effect of photostimulation on individual neuron spike rates (Figs
3, 4, Extended Data Figs 5, 9), spike counts from “lick left” and “lick right” trials
were pooled. We used two-tailed t-tests on spike counts binned in various windows
(e.g. Fig 3, and Extended Data Fig 5, the whole delay epoch; Fig 4f, 400 ms time bins
in steps of 50ms) to test for significant spike rate change: control versus photostimulation
trials.
Quantification of the effect of photostimulation on selectivity was complicated by
the fact that selectivity was coupled to upcoming movements (Extended Data Fig 4e).
For example, using only correct trials to compute selectivity would miss the trials
in which photostimulation caused the mice to switch future movements, thus underestimating
the effect of photostimulation on selectivity. We therefore used all trials (correct
and incorrect, grouped by instructed trial types) when quantifying selectivity changes
caused by photostimulation (Fig 3c, f, Extended Data Figs 5, 6, 9).
Analysis of ALM population dynamics in the activity space
To analyze the relationship between ALM population activity and upcoming movements,
we restricted the analysis to the recording sessions with > 5 neurons recorded simultaneously
for > 5 trials per condition (28/39 sessions, 8–46 neurons recorded simultaneously;
Fig 4, Extended Data Fig 10). For a population of n neurons, we found a n × 1 vector,
in the n dimensional activity space that maximally separated the response vectors
in “lick right” and “lick left” trials, we term this vector the “coding direction”
(
cd
).
Average spike counts were computed in 400 ms windows in 10 ms steps. For each movement
direction ( “lick right” and “lick left”) we computed the average spike counts
x
¯
l
i
c
k
r
i
g
h
t
and
x
¯
l
i
c
k
l
e
f
t
,
n
×
1
response vectors that described the population response at each time point, t. We
computed the difference in the mean response vectors,
c
d
t
=
x
¯
l
i
c
k
r
i
h
g
t
−
x
¯
l
i
c
k
l
e
f
t
cd
t
was stable throughout the delay epoch. We averaged the
cd
t
’s from the delay epoch to obtain a single
cd
. Only correct control trials were used to estimate the
cd
. The projection along the
cd
captured 64.1±3.4% of the population selectivity for “lick left” and “lick right”
trials over the sample and delay epochs (root mean square, RMS, of the spike rate
difference between “lick right” and “lick left” trials), and 27.0±4.2% of the total
variance in ALM task-related activity. Activity variance was quantified as the RMS
of the baseline subtracted activity over the sample and delay epochs.
To visualize the effect of CN photo-activation on ALM activity trajectories, we projected
the activity trajectories from the control trials, fastigial photo-activation trials,
or dentate photo-activation trials on the
cd
. We used independent control and photo-activation trials from the trials used to
compute the
cd
. For each trial we computed the spike counts for each neuron,
x
(n × 1 response vector), at each time point. The projected trajectories in Figure
4i were obtained as
cd
T
x
. Both correct and incorrect trials were used to compute the projected trajectries,
grouped by the instructed trial types. To average trajectories across multiple behavioral
sessions, we first offset the trajectories for a particularly session by subtracting
the mean
cd
T
x
across all trials and time points in that session. This removed fluctuations in mean
activity from session to session. The offsets were computed using the independent
control trials that were used to calculate the
cd
. Standard errors of the mean was obtained by bootstrapping individual sessions.
To predict upcoming movements using ALM responses projected onto the
cd
(Extended Data Fig. 10c–d), we used the response vector
x
from the last time bin before the ‘go’ cue (last 400 ms of the delay epoch). For each
session, we computed the projected responses,
cd
T
x
, from “lick right” and “lick left” trials. The
cd
was computed using independent control trials (correct trials only). Separate photostimulation
trials were used to predict performance (both correct and incorrect trials). Probability
of the mouse licking right in photostimulation trials was calculated as a function
of projected responses,
cd
T
x
(Extended Data Fig. 10c–d). Data from multiple sessions were pooled.
Anatomy, tracing
We mapped connectivity between the CN, thalamus, ALM, substantia nigra reticulata
(SNr), and the lateral superior colliculus previously implicated to play roles in
controlling licking
36
(Fig 4a–c, Extended Data Figs 7, 8). Mice were anesthetized using isofluorane and
fixated on a high precision stereotaxic frame (David Kopf or Stoelting). Coordinates
used to target brain areas of interest used either bregma or lambda as the reference
points. The coordinates were (in mm): ALM, bregma anterior 2.5, lateral 1.5, depth
0.8; SNr, lambda anterior 1.1, lateral 1.4, depth 4.5; superior colliculus, lambda
anterior 0, lateral 1.3, depth 1.8; fastigial nucleus, lambda posterior 2.7, lateral
0.7, depth 2.3; dentate nucleus, lambda posterior 2.5, lateral 2.5, depth 2.4. To
label ALM-projecting thalamus, 10% biotin dextran amine BDA-10k (Sigma), or 1% Cholera
toxin b subunit Ctb-Alexa647 (Invitrogen) disolved in PBS were iontophoretically injected.
Glass pipette with a tip size of 20 um were filled with either BDA or Ctb, 4 mA of
positive currents were applied with 7s duty cycle for 15 mins. Viral vector assisted
tracing were applied to the other brain regions. 100–150 nl of AAV2-CAG-eGFP or AAV2-CAG-tRFP
(UNC viral core, titers 1.0–1.2 × 1013 vg/ml) were pressure injected in the target
areas. After each injection, the pipette was left in place for >10 minutes before
being slowly withdrawn. Brains were processed for immunohistochemistry 10 days after
the injection. Mice were deeply anesthetized with an overdose of Nembutal (0.2 ml,
i.p.) and transcardially perfused with 20 ml saline followed by 50 ml 4% paraformaldehyde
(PFA) in PBS. Brains were extracted and post-fixed in 4% PFA for 2 hours and incubated
in 10% sucrose overnight at 4 °C. Brains were then embedded in gelatine and cryoprotected
in 30% sucrose in PB, frozen on dry ice, and sectioned using a freezing microtome
(50 μm thick). For immunohistochemistry, sections were blocked for 1 hour at room
temperature in PBS with 0.4% Triton X-100 and 10% NHS solution and incubation for
48 hrs at 4°C in a mixture of primary antibodies diluted in PBS with 2% NHS and 0.4%
Triton X-100. Sections were then washed and incubated for 2 hours at room temperature
in the fluorescent secondary antibodies. Axonal projections and neurons were visulized
using: mouse anti-NeuN (1:1000, Millipore), chicken anti-GFP (1:2000, Aves), rabbit
anti-RFP (1:2000, Rockland), mouse anti-vGlut2 (1:1000, Millipore) and Streptavidin-Alexa647
(1:400, Jackson). Slices were counterstained with DAPI (1:100,000, Invitrogen) and
mounted with mounting medium for fluorescence (Mowiol, Sigma). Stacks of images were
acquired on an upright LSM 700 confocal microscope (Carl Zeiss, Jena, Germany) operated
with Zeiss Zen Software (v2.2) and post-hoc adjusted and processed in FIJI software
(v.1.46) with appropriate plugins (https://imagej.net/Fiji).
Anatomy, analysis
We aligned each coronal section to the Allen Mouse Common Coordinate Framework (CCF)
37,38
using landmark-based image registration. The registration target was the 10µm/voxel
CCF anatomical template. To align a coronal section, we first manually selected the
coronal plane in the anatomical template that best corresponded to the section. Next,
we manually placed control points at corresponding local landmarks in each image (Extended
Data Fig 7a). 30–50 control points were placed in a single image. All control points
were outside of the thalamus. Next, the image was warped to the CCF using an affine
transformation followed by a nonrigid transformation using b-splines
39
. Images were warped using the B-spline Grid, Image and Point based Registration package
available on the Matlab FileExchange (https://www.mathworks.com/matlabcentral/fileexchange/20057-b-spline-grid--image-and-point-based-registration).
We performed this procedure independently for each brain section. 2–6 sections at
the rostro-caudal level of the thalamus were analyzed (bregma posterior 1.1–2.1 mm
in the CCF).
We quantified the connectivity of the fastigial nucleus, the dentate nucleus, SNr,
and superior colliculus with ALM thalamo-cortical loop based on the overlaps in the
anterograde fluorescence from a source area with ALM-projecting thalamus. In each
case, quantifications of the anterograde fluorescence from a source area were always
made in brains that contained co-injections of tracers in ALM to label ALM-projecting
thalamus. Injections of retrograde (Ctb) and anterograde (BDA) tracers in ALM produced
similar patterns of labeling in the thalamus (Extended Data Fig 8)
17
. We therefore used both types of tracers to label ALM-projecting thalamus. Quantifications
of fluorescence were performed on images post alignment to the CCF anatomical template
(Extended Data Fig 7b). We found consistent labeling patterns in the thalamus across
different injection cases (Extended Data Fig 7c). Alignment to the CCF allowed us
to combine quantifications of fluorescence overlaps across different injection cases
(Fig 4a–c, Extended Data Fig 7d–f).
To quantify fluorescence overlaps with ALM-projecting thalamus, we first limited the
analysis to a subregion of the coronal section that contained the thalamus ipsilateral
to the ALM injection (shown in Extended Data Fig 7b at different rostro-caudal levels).
We then thresholded the fluorescence intensities independently for each channel at
approximately 89th percentile (85th-94th across different cases). The labeled area
was defined as all the pixels that exceeded this threshold (Extended Data Fig 7e).
Finally, the overlap in fluorescence was computed as the number of pixels co-labeled
by ALM injections and a projection (e.g. the fastigial nucleus), divided by the total
number of pixels labeled by that projection (“area fraction” overlap, Extended Data
Fig 7f). This procedure was carried out separately for each coronal section, which
consituted one data point in Fig 4c and Extended Data Figs 7e–f, 8b.
Because we always performed triple injections: a retrograde/anterograde tracer in
ALM, and anterograde tracers in two source areas (i.e. the fastigial and dentate nucleus,
or the fastigial nucleus and SNr, or the fastigial nucleus and superior colliculus,
or other combinations). Quantifications of fluorescence overlaps were further validated
by visualizing all three channels in individual injection cases (Extended Data Fig
8).
Statistics
The sample sizes are similar to sample sizes used in the field: for behavior, 3 mice
or more per condition; for electrophysiology, more than 100 units per brain region.
No statistical methods were used to determine sample size. In 1 of the fastigial photo-activation
mice, the injection of ChR2 virus missed the CN and its data were excluded. All key
results were replicated in multiple mice. Mice were randomly pre-allocated into experimental
groups. Unless stated otherwise, the investigators were not blinded to allocation
during experiments and outcome assessment. Trial types were randomly determined by
a computer program. During spike sorting, experimenters cannot tell the trial type,
so experimenters were blind to conditions. Statistical comparisons using t-tests and
bootstrap are described in detail in sections above.
Code availability
Software used for data analysis is available from the corresponding author upon reasonable
request.
Data availability
Raw and processed data are available from the corresponding author upon reasonable
request.
Extended Data
Extended Data Figure 1.
Histology and behavioral data.
a. Coronal sections showing CN lesion sites for all mice. Fastigial lesion, n=4; dentate
lesion, n=4.
b. Licking movement is not affected by fastigial lesion. Individual licks are shown
as dots (blue, lick right; red, lick left). Licks are aligned to the onset of the
‘go’ cue (vertical line).
c. Average lick rate within a lick bout and the duration of lick bouts are not affected
by fastigial lesion. A lick bout is defined as a sequence of licks with inter-lick-interval
less than 500 ms. Data from the fastigial-lesion mice only (n=4). Data are grouped
by licking direction relative to the lesioned hemisphere: contralesional licking (blue)
or ipsilesional licking (red).
d. Coronal sections showing ChR2-EYFP expression in the fastigial and dentate nucleus.
9 mice were used for various fastigial ChR2 photo-activation experiments, and 10 mice
were used for dentate ChR2 photo-activation experiments. In 1 of the fastigial photo-activation
mice, the injection of ChR2 virus missed the CN (not shown) and its data were excluded.
2 of the dentate photo-activation mice were subsequently used for lesion experiments
(not shown). See Supplemental Table 1 for the list of mice and manipulated hemispheres
in individual experiments.
e. CN photo-activation during specific behavioral epochs. Top, experiment timeline.
Photostimulation was for the entire task epoch (1.3 s). Bottom, behavioral performance.
Left, fastigial photo-activation (n=6); right, dentate photo-activation (n=8). For
fastigial photo-activation, average behavioral performance for control and photo-activation
trials (S, sample epoch photo-activation; D, delay epoch; R, response epoch) is also
shown. Gray lines show individual mice. Chance is 50%. For behavioral performance
grouped by trial type, “lick left” and “lick right” trials are grouped by instructed
licking direction relative to the manipulated hemisphere. Blue, contralateral; red,
ipsilateral. Both hemispheres were tested. See Supplemental Table 1 for the list of
mice and manipulated hemispheres. Thick lines, mean; thin lines, individual mice (n=6).
*, p<0.05, **, p<0.01, ***, p<0.001, one-sided test, bootstrap (Methods).
f. Early lick rate and no lick rate with and without CN photo-activation during the
delay epoch. Control trials (gray) and photo-activation trials (cyan). Lines, individual
mice. FN, fastigial photo-activation (n=6); DN, dentate photo-activation (n=8).
Extended Data Figure 2.
Characterization of CN photo-inhibition and behavioral data.
a. Strategy to silence the CN. Left, ChR2 is expressed in cerebellar Purkinje cells.
Photostimulation of Purkinje cells inhibits CN neurons (‘photo-inhibition’).
b.
Top left, an image of the cerebellar cortex showing ChR2-EYFP expression in Purkinje
cells in L7-cre × Ai32 mouse. Top right, an example silicon probe recording in a L7-cre
× Ai32 mouse. The coronal section shows ChR2-EYFP expression in the cerebellar cortex
and DiI labeled electrode tracks. Bottom, electrode tracks within the fastigial nucleus
(dashed contour). Photostimulation of Purkinje cells was carried out by directing
a blue laser (0.5–4.5 mW average power, beam diameter: 400 µm at 4σ) to the brain
surface through a cranial window.
c. Example voltage traces from a fastigial nucleus recording during photo-inhibition.
Multi-unit activity in the fastigial nucleus was silenced. Multiple traces show repeats
of photostimulation.
d. Photo-inhibition reduced activity specifically in the CN. Average spike rates before
(500 ms) versus during photo-inhibition (1.2 s). Neurons from the fastigial nucleus
(left, n=26) and from the surrounding cerebellar cortex (right, n=34). Photostimulation
of the cerebellar cortex (lambda posterior 3 mm, lateral 2 mm) silenced fastigial
activity whereas photostimulation of the inferior colliculus (posterior 1mm, lateral
2mm) did not. Solid dots, neurons with significant spike rate change (p<0.01, two-tailed
t-test).
e. Dose dependent effect of CN photo-inhibition on posture and movement. Left, video
frames showing an unrestrained mouse undergoing CN photo-inhibition. During low-intensity
photo-inhibition (1.5mW, the intensity typically used in the delayed response task),
no consistent movement was detected. High-intensity photo-inhibition (20 mW) of both
the fastigial and dentate nucleus produced dystonia-like movements and posture changes,
such as loss of balance and extensions of the contralateral limbs (arrows). Right,
scores that rated dystonia-like movements on a scale of 0–4 (Methods): 0 - no motor
abnormalities; 1 - slightly slowed or abnormal motor behavior, no dystonia; 2 - mild
impairment, mild and transient dystonic postures, weak tremor; 3 - moderate impairment,
dystonic postures, cannot balance the body, major tremor; 4 - severe impairment, sustained
dystonic postures and limited movements. The scoring was blinded to the experimental
conditions. Mean ± s.e.m. across trials. N=2 mice, 10 trials per condition.
f. Dose dependent effect of CN photo-inhibition on jaw movement and eyeblink. Top,
video frames showing a head-fixed mouse undergoing CN photo-inhibition. Bottom left,
eyeblink responses during CN photo-inhibition. No consistent eyeblink was evoked.
Bottom right, jaw movements during CN photo-inhibition. Only high-intensity photo-inhibition
(20 mW) of the dentate nucleus produced jaw movements. Mean ± s.e.m. across trials.
N=4 mice, 10 trials per condition for eyeblink, 20 trials per condition for jaw movement.
g. Behavioral performance in the delayed response task with and without fastigial
photo-inhibition. Photo-inhibition during the entire task epochs (n=3). Left, average
behavioral performance for control and photo-inhibition trials (S, sample epoch photo-inhibition;
D, delay epoch; R, response epoch). Gray lines show individual mice. Chance is 50%.
Right, behavioral performance for each trial type. “Lick left” and “lick right” trials
are grouped by instructed licking direction relative to the manipulated hemisphere.
Blue, contralateral; red, ipsilateral. Both hemispheres were tested. See Supplemental
Table 1 for the list of mice and manipulated hemispheres. Thick lines, mean; thin
lines, individual mice. *, p<0.05, **, p<0.01, ***, p<0.001, one-sided test, bootstrap
(Methods).
h. Behavioral performance. Photo-inhibition during the early sample, late sample,
early delay, and late delay epochs (n=7 mice). *, p=0.04, ***, p=0.0002, one-sided
test, bootstrap.
i. Early lick rate and no lick rate during delay epoch fastigial photo-inhibition.
Control trials (gray) and photo-inhibition trials (cyan). Lines, individual mice (n=7).
Extended Data Figure 3.
Silicon probe recordings in the CN.
a. DiI labeled recording tracks in coronal sections stained by DAPI. Borders of the
CN were visible in DAPI staining. Electrode location was estimated based on DiI labeling
and manipulator depth. The lamination of activity patterns across the electrodes corresponded
well to the anatomical structures: i.e. high activity level in the CN and cerebellar
cortex, low activity level in the white matter. Single-unit locations were estimated
based on electrode location. Top, a coronal section showing a recording in the fastigial
nucleus. Bottom, the fastigial nucleus (yellow dashed line), estimated electrode location,
and the locations of single-units (units 1–6) from an example recording.
b.
Left, voltage trace from the recording site of unit 1 during a single behavioral trial.
Dashed lines, behavioral epochs. Right, waveforms of unit 1 after spike sorting.
c. PSTHs of all the units from the example fastigial recording. Dashed box indicates
that units 1–5 were within the fastigial nucleus. Unit 6 was outside of the fastigial
nucleus. Blue, “lick right” trials; red, “lick left” trials.
d. Same as a and c, but for an example recording in the interposed nucleus.
e. Same as a and c, but for an example recording in the dentate nucleus.
Extended Data Figure 4.
ALM and CN activity related to motor planning and movement.
a. Criteria used to identify neuronal activity modulated by licking. For each lick,
spike counts were calculated in two adjacent 50ms time windows, starting 20ms from
the detection time of the lick. Across all licks and all trials, neurons with significant
spike count difference between the two time windows were deemed to be modulated by
licking (p<0.01, two-tailed t-test). Licking modulation is calculated for each neuron:
mean spike rate difference between the two time windows.
b. Example neurons significantly modulated by licking. PSTH from the correct “lick
right” and “lick left” trials during the response epoch. Spike times are aligned to
the first lick. Trials are grouped by licking direction relative to the recorded hemisphere:
contralateral (blue) or ipsilateral (red). Averaging window, 10 ms.
c. Fraction of neurons significantly modulated by licking (p<0.01, two-tailed t-test)
from each brain region. ALM, the anterior lateral motor cortex; FN, the fastigial
nucleus; IP, the interposed nucleus; DN, the dentate nucleus. Error bars, s.e.m. across
mice, bootstrap (Methods). Dots, individual mice (ALM, n=10; FN, n=15; IP, n=9; DN,
n=11).
d. No correlation between whether a neuron was modulated by licking and whether it
exhibited preparatory activity. Top, ALM; bottom, the cerebellar nuclei. Selectivity
during the delay epoch is the spike rate difference between the “lick right” and “lick
left” trials. Solid dots, neurons with significant licking modulation (p<0.01, two-tailed
t-test).
e. Preparatory activity in the CN. Left, 2 example CN neurons. PSTHs for correct and
error trials. Trial types are based on sensory instruction (blue, “lick right”; red,
“lick left”). The same trial-type preference in correct and error trials indicates
selectivity for object location. Opposite trial-type preference indicates selectivity
for upcoming movement directions. A negative ratio of trial-type selectivity between
error and correct trials means a neuron switches trial-type preference to predict
upcoming movement directions on error trials. Right, selectivity ratios for CN neurons.
Neurons with significant delay epoch selectivity and tested for >3 error trials (n=73).
f. Population selectivity for each cerebellar nucleus (mean ± s.e.m. across neurons,
bootstrap; fastigial, n=87; interposed, n=50; dentate, n=60). Selectivity is the spike
rate difference between the preferred and non-preferred trial type (Methods). Dashed
lines separate sample, delay and response epochs.
g. Proportion of contra- versus ipsi-preferring neurons based on spike counts during
the first 500ms (left) and the last 500ms (right) of the delay epoch. Error bars,
s.e.m. across mice. Dots, individual mice (ALM, n=10; FN, n=15; IP, n=9; DN, n=11).
***, p=0.0001, one-sided test, bootstrap.
Extended Data Figure 5.
ALM drives CN preparatory activity and vice versa.
a.
Top, schematic showing CN recording during bilateral ALM photo-inhibition. Laser power,
1.5 mW per location (Methods). Bottom, PSTHs of 4 example CN neurons with and without
ALM photo-inhibition. Cyan, photo-inhibition. Blue, “lick right” trials; red, “lick
left”.
b.
Left, average fastigial selectivity in control and photo-inhibition trials (mean ±
s.e.m. across neurons, bootstrap). Only selective neurons tested for >3 trials in
all conditions are included (n=54 neurons, 8 mice). Orange dashed line, mean from
control trials. Right, relationship between delay epoch selectivity of individual
fastigial neurons and changes in firing rate due to ALM photo-inhibition. Filled circles,
neurons significantly modulated by ALM photo-inhibition (p<0.01, two-tailed t-test).
c. Same as b, but for the interposed nucleus (n=44 neurons, 5 mice).
d. Same as b, but for the dentate nucleus (n=59 neurons, 6 mice).
e. Same as a, but for ALM recording during contralateral CN photo-inhibition. Laser
power, 1.5–4.5 mW.
Extended Data Figure 6.
CN preparatory activity is driven by both hemispheres of ALM.
a. CN recording during contralateral or ipsilateral ALM photo-inhibition.
b. Behavioral performance during unilateral and bilateral ALM photo-inhibition.
c. CN population selectivity from control and photo-inhibition trials (mean ± s.e.m.
across neurons, bootstrap). Only selective neurons tested for >3 trials in all conditions
are included (n=157). Orange dashed line, mean from control trials. Cyan, photo-inhibition.
CN selectivity was reduced by photo-inhibiting either side of ALM.
d. Example CN neurons during ipsilateral or contralateral ALM photo-inhibition. The
effect was heterogeneous across individual neurons. Some neurons were affected by
ipsilateral ALM photo-inhibition (row 1–3), other neurons were affected by contralateral
ALM photo-inhibition (row 4), others were affected by photo-inhibiting either side
(row 5, 6).
e. Example CN neurons during bilateral ALM or M1 photo-inhibition. ALM, bregma anterior
2–3 mm, lateral 1–2 mm; M1 anterior 0–1 mm, lateral 1–2 mm.
Extended Data Figure 7.
Alignment of anatomical data and comparing fastigial and dentate projections to ALM-projecting
thalamus.
a.
Top, an example coronal section before and after alignment to the corresponding section
of the anatomical template in the Allen Mouse Common Coordinate Framework (CCF). Fluorescence
is anterograde labeling (BDA) from an ALM injection, labeling the ALM-projecting thalamus
(Methods). Yellow dots, control points for b-spline transformation. Dashed line, a
region of interest containing the thalamus ipsilateral to the ALM injection. Bottom,
overlay of the aligned image (red) and the anatomical template (green). ALM axons
in the thalamus obey structural boundaries in the anatomical template. Critically,
the alignment procedure did not use any structures inside the thalamus as landmarks.
b. Coronal sections showing an example fastigial injection case with anterograde labeling
in the thalamus. The sections are aligned to and superimposed on the anatomical template.
The yellow outlines show labeled areas after thresholding (Methods).
c. Coronal sections showing 3 different fastigial injection cases. Different injections
show similar patterns of anterograde labeling in the thalamus.
d.
Left, outlines of the labeled areas from all fastigial injection cases (n=6). Borders
of the thalamic nuclei (VM and VAL) are based on annotations in the Allen Reference
Brain. Right, outlines of the labeled areas from all fastigial (n=6), dentate (n=4),
and ALM (n=8) injection cases. There was a greater overlap between fastigial and ALM
labeling.
e. Total areas labeled (number of pixels) in the thalamus by fastigial, dentate, and
ALM injections. Dots, individual coronal sections. Lines, individual injections.
f. Overlaps between fastigial/dentate projections and ALM-projecting thalamus. The
co-labeled area is normalized to the total thalamic area labeled by fastigial/dentate
projections (‘area fraction overlap’). Dots, individual coronal sections. Lines, individual
injections.
Extended Data Figure 8.
ALM-projecting thalamus receives converging inputs from the fastigial nucleus, basal
ganglia substantia nigra reticulata (SNr), and superior colliculus.
a. Example triple injections showing co-labeling of fastigial and dentate projections
(left), fastigial and SNr projections (middle), and fastigial and superior colliculus
projections (right) in the thalamus. Injections were repeated multiple times in different
mice with similar results (ALM-fastigial-dentate, n=4; ALM-fastigial-SNr, n=3; ALM-fastigial-superior
colliculus, n=2).
b. Overlaps between ALM-projecting thalamus and projections of different areas (Methods).
Fastigial, SNr, and superior colliculus labeling show comparable amount of overlaps
with ALM-projecting thalamus. Dentate labeling show less overlap. The thalamic area
co-labeled by ALM injection and a projection (fastigial, dentate, SNr, or superior
colliculus) is normalized to the total thalamic area labeled by that projection (‘area
fraction’). Dots, individual coronal sections. Lines, individual injections.
c. A confocal image showing vGlut2 staining in ALM-projecting thalamus. Anterograde
tracer injections in ALM (blue), the fastigial nucleus (green), and SNr (red). Fastigial
axons form glutamatergic synapses (vGlut2 positive) within ALM-projecting thalamus.
vGlut2 staining was performed in one ALM-fastigial-SNr injection case.
Extended Data Figure 9.
CN ChR2 photo-activation drives rapid ALM activity change.
a. ALM neurons firing rate change during fastigial or dentate ChR2 photo-activation.
Firing rate change is the difference in spike rate between control and photo-activation
trials. For neurons that were suppressed by CN photo-activation, the firing rate differences
are multiplied by −1 so that the firing rate changes are always positive for latency
quantifications. Mean ± s.e.m. across neurons. Only neurons that were significantly
modulated by CN photo-activation are included (p<0.01, two-tailed t-test; fastigial
photo-activation, n=227; dentate photo-activation, n=163). Top, ALM firing rate change
time course. Cyan, CN photo-activation (500ms). Bottom, ALM firing rate change onset
(arrows). Firing rate change is quantified in 10ms time bins in 1ms steps. The onset
time is the first time bin in which population firing rate change significantly deviated
from 0 (p<0.01, two-tailed t-test). We repeated the onset time estimation 10,000 times.
In each round, we re-sampled with replacement from the neurons in the dataset and
re-estimated the onset time. The arrows indicate the mean estimated onset time.
b. Relationship between selectivity of individual ALM neurons and changes in firing
rate due to CN photo-activation. Filled circles, ALM neurons that were significantly
modulated by CN photo-activation (p<0.01, two-tailed t-test). Firing rate change is
the difference in spike rate between control and photo-activation trials during photostimulation
(top) or last 500ms of the delay epoch (bottom). Selectivity is the spike rate difference
between “lick right” and “lick left” trials during the delay epoch.
c. ALM population selectivity from control and photo-activation trials (mean ± s.e.m.
across neurons, bootstrap; selective neurons tested for >3 trials in all conditions).
Orange dashed line, mean from control trials. Top, fastigial photo-activation (n=328);
bottom, dentate photo-activation (n=377). Cyan, photo-activation.
Extended Data Figure 10.
Fastigial photo-activation destroys the relationship between ALM activity and upcoming
movements
a. ALM recording during bilateral ALM photo-inhibition (top) or contralateral fastigial
photo-activation (bottom).
b. Performance in control (grey) and photostimulation trials (cyan). “Lick left” and
“lick right” trials are pooled. Chance is 50% (dashed line). Lines, individual mice
(ALM photo-inhibition, n=7; fastigial photo-activation, n=6). ***, p<0.001, one-sided
test, bootstrap (Methods).
c.
Left, schematic. Dashed lines, control trials, ALM activity trajectories converge
to discrete endpoints at the end of the delay epoch that predict upcoming movement
directions (“lick right”, blue; “lick left”, red). Solid black lines, bilateral ALM
photo-inhibition trials, the distance of the perturbed trajectories to the endpoints
predicts future movement directions even though behavioral performance is near chance.
Right, experimental data. Probability of the mouse licking right as a function of
trajectory distance to the endpoints along the
cd
(Methods). Bilateral ALM photo-inhibition trials only (n=690). The
cd
and endpoints were estimated from independent control trials. Data are binned along
the
cd
projection. s.e.m. is obtained by bootstrapping the trials in each bin.
d. Same as c, but for fastigial photo-activation trials (n=337). ALM activity no longer
predicts future movement directions after a fastigial perturbation.
e. It is possible that fastigial perturbation activated downstream motor circuits
that could maintain the motor plan and generate movements independent of ALM. We examined
the necessity of ALM activity in driving directional licking after a fastigial perturbation.
In VGAT-ChR2-EYFP mice expressing ChR2 in both Purkinje neurons and cortical GABAergic
neurons, we independently manipulated activity in the fastigial nucleus and ALM (Methods).
Left, unilateral ALM photo-inhibition during the late delay epoch. Right, behavioral
performance for each trial type in control and photo-inhibition trials. Left ALM photo-inhibition
biased upcoming licking to the left, resulting in lower performance in the “lick right”
trials. The opposite pattern of bias was induced by right ALM photo-inhibition. Thick
lines, mean; thin lines, individual mice (n=4). ***, p=0.0002, one-sided test, bootstrap
(Methods).
f. Same as e, but for unilateral ALM photo-inhibition after fastigial perturbation.
Note the same pattern of behavioral bias as e. *, p=0.04, **, p=0.008, one-sided test,
bootstrap.
g.
Left, after fastigial perturbation, ALM activity was bilaterally silenced during movement
initiation. Right, fraction of trials in which mice did not lick after the ‘go’ cue.
FN, fastigial perturbation only; ALM, bilateral ALM photo-inhibition; FN+ALM, fastigial
perturbation followed by bilateral ALM photo-inhibition. Thick lines, mean; thin lines,
individual mice (n=3). **, p=0.002, ***, p=0.0002, one-sided test, bootstrap.
Supplementary Material
1
2