This site needs JavaScript to work properly. Please enable it to take advantage of the complete set of features!
Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

NIH NLM Logo
Log in
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2017 May;20(5):727-734.
doi: 10.1038/nn.4531. Epub 2017 Mar 20.

Cerebellar granule cells acquire a widespread predictive feedback signal during motor learning

Affiliations

Cerebellar granule cells acquire a widespread predictive feedback signal during motor learning

Andrea Giovannucci et al. Nat Neurosci. 2017 May.

Abstract

Cerebellar granule cells, which constitute half the brain's neurons, supply Purkinje cells with contextual information necessary for motor learning, but how they encode this information is unknown. Here we show, using two-photon microscopy to track neural activity over multiple days of cerebellum-dependent eyeblink conditioning in mice, that granule cell populations acquire a dense representation of the anticipatory eyelid movement. Initially, granule cells responded to neutral visual and somatosensory stimuli as well as periorbital airpuffs used for training. As learning progressed, two-thirds of monitored granule cells acquired a conditional response whose timing matched or preceded the learned eyelid movements. Granule cell activity covaried trial by trial to form a redundant code. Many granule cells were also active during movements of nearby body structures. Thus, a predictive signal about the upcoming movement is widely available at the input stage of the cerebellar cortex, as required by forward models of cerebellar control.

PubMed Disclaimer

Conflict of interest statement

COMPETING FINANCIAL INTERESTS

The authors declare no competing financial interests.

Figures

Figure 1
Figure 1
GCaMP expression and signals in cerebellar GrCs. (a) Dorsal view of the cerebellum with imaged area of lobule VI indicated in green. (b) GABAA α6 staining (red) highlights dense virally driven expression of GCaMP6f (green) in GrCs (GrCs; each FOV of 100 ×ばつ 100 μm contained an average of 187 GrCs; 20 FOVs and 2 mice per expression strategy, NeuroD1-cre and hSyn). Note the absence of GCaMP signal in the mossy fiber (mf) bundle and Purkinje cells (asterisks). Arrows indicate putative Golgi or Lugaro cells. (c) Top: three panels show (from top to bottom) two-photon imaging FOV in the granular layer, overlaid with a subset of spatial components identified by NMF and classified as GrCs (red) or a putative Golgi cell (green, G) and the same GrCs on a dark background. Bottom: GrC fluorescence traces and corresponding spatial mask. (d) Two-photon imaging of parallel fiber activity. Top: an example FOV without and with manually selected regions of interest. Bottom left: fluorescence traces. Bottom right: cross-correlations reveal high correlation between mediolaterally aligned boutons. (e) Coronal cerebellar sections of the mouse shown in d, counterstained for GABAA α6. Arrows point to a parallel fiber expressing GCaMP6f. (f) Fluorescent traces from parallel fiber boutons recorded from a trained mouse, aligned to corneal airpuffs and light flashes.
Figure 2
Figure 2
Role of imaged regions in eyeblink conditioning. (a) Task schematic. CS (ultraviolet LED flash to the contralateral eye or weak puff to the ipsilateral vibrissa) and US (periorbital airpuff) were delivered to a head-fixed mouse on a freely moving treadmill while blinks, snout movement, body movement and treadmill rotation were monitored by high-speed infrared camera (100 frames/s) and (in some experiments) a magnet attached to the lower eyelid. (b) CRs, quantified as eyelid closure as a fraction of US response, during CS–US paired trials in a single animal. Green (CS, light) and red (US, airpuff) shaded zones indicate stimulus presentations. Blue shaded zone indicates time window for computing e*, eyelid movement as a measure of CR amplitude. (c) Evolution of learning in 6 animals trained for up to 12 consecutive d (2 mice trained with whisker CS, open squares; 4 mice trained with light CS, closed gray diamonds; average indicated by open circles). Error bars for the averaged plot indicate ± s.e.m. (d) Left: focal injection of muscimol, but not saline vehicle, led to a reduction in the percentage of CRs (muscimol (drug) versus saline CR probability, P = 0.005, paired t-test; each line represents one animal). Filled symbols depict data for the example mouse shown in e. Right: focal injection of muscimol, but not saline vehicle, led to a reduction in the amplitude of CRs (muscimol versus saline CR, P = 0.026, paired t-test; n = 5). Blue and red dots depict averages. Error bars indicate ± s.e.m. (e) Example of the inactivation experiment in one mouse (filled diamonds in d, left). Left: a coronal cerebellar section counterstained with DAPI (blue) and aldolase C (green) reveals the injection position (red, muscimol + Evans Blue). Right: individual (light gray) and averaged (black) eyelid responses during baseline trials before injections and after drug or saline injections.
Figure 3
Figure 3
Calcium signals of a single GrC during eyeblink conditioning. (a) Eyelid movement (left), locomotion activity (center) and single-cell calcium signal (right) from one neuron followed over 9 d of training. Each horizontal line represents a single trial. Date labels indicate the start of a day’s training, and unlabeled ticks indicate the start of a new session within the same day. Vertical solid lines indicate onset and offset of CS stimulus. Dashed vertical lines indicate delivery time of the US stimulus. (b) Top: data from the last 3 d of training, resorted according to whether the CS did (CR+, middle) or did not (CR−, top) evoke an anticipatory eyelid closure before the US. Trials with significant locomotion (> 2 cm/s) during the CS are excluded. Bottom left: overlaid average eyelid responses in US trials (cyan) and at the final stage of training when the animal did (CR+, black) or did not (CR−, red) produce a CR. Bottom middle: overlaid average locomotion speeds during US, CR+ and CR− trials. Bottom right: overlaid average calcium signals during US, CR+ and CR− trials. f* is the integral of the neural response in an 85-ms window before the UR.
Figure 4
Figure 4
Cerebellar GrCs acquire a neural correlate of the learned conditional response. (a) Top: in a mouse trained with light CS, average CR amplitude (first column) and fluorescence responses (second column) for populations of GrCs grouped by trial type (n = 29 neurons). US, cyan; CR+ trials, black; CR− trials, red. Bottom: same results for a mouse trained with whisker CS (N = 60 neurons). Behavioral and neural traces averaged over an 85-ms time window (brackets) preceding the UR for eyelid responses (e*) and fluorescence response (f*). Shaded regions in the right column represent the amount of GrC response in CR− trials (red) and the excess GrC response in CR+ vs. CR− trials (black). (b) Top: evolution of the calcium response during 9 d of conditioning to a CS light stimulus, shown separately for CR+ (left) and CR− (right) trials in 29 individual GrCs from a single animal. Gray indicates that not enough trials (5) were available to compute a value. Middle: average calcium responses for the 29 GrCs in the population as a function of training day. Responses before and after training were significantly different (paired t-test, P = 5.7 ×ばつ 10−7). Bottom: eyelid response and locomotion for CR+ and CR− trials. Error bars indicate ± s.e.m. (c) Evolution of GrC calcium responses across all animals (n = 6). Plotted values show average ± s.e.m. calculated by bootstrap-resampling (n = 480 neurons, eyelid-amplitude CR+ trials; n = 1,120 cells, eyelid-amplitude CR− trials; n = 459 neurons, percentage responses CR+ trials; n = 1,109 neurons, percentage responses CR− trials) to obtain equal weighting among FOVs across 6 mice. Left: 0–20%, 20–40%, 40–60%, 60–80% and 80–100% of CRs. Right: blocks of trials sorted by e*. For the significant bins, from left to right, t-tests: CR+ vs. CR−, P = 2.5 ×ばつ 10−4, 7.8 ×ばつ 10−3 and 6.1 ×ばつ 10−4 (left), and P = 0.041 and 1.6 ×ばつ 10−4, 08.1 ×ばつ 10−6 (right). Error bars indicate ± s.e.m. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. (d) Histograms of GrC responses in 2 fully trained mice (one trained with whisker CS and one with light CS; for other mice see Supplementary Fig. 8d). Each histogram bar represents the response of a single GrC to a CS stimulus without wheel movement. Red histograms, ΔF/F for trials with no conditional response (CR−, same time bin as f*); black histograms, difference in ΔF/F between CR+ and CR− trial responses for the same cells, i.e., the CR+ excess (same time bin as f*); n = 60 cells, light CS; n = 60 cells, whisker CS.
Figure 5
Figure 5
GrCs simultaneously express correlates of eyeblink, locomotion and snout movement. (a) Simultaneous recording of three GrCs whose calcium signals (black) were correlated with eyelid (magenta), wheel (orange) or snout (cyan) movements. Pearson correlations over one day’s recording session are given in the left column. Dashed traces superimposed on the calcium signals are the same movement-related traces shown in the panels above. (b) In a single GrC, eyelid responses to CS presentation (e*) as a function of fluorescence signal (f*). Error bars indicate ± s.e.m. (c) Left: in a single GrC, US-only presentation caused a range of eyelid response amplitudes (URs) concurrent with changes in f*. Right: eyelid closure amplitude as a function of neural activity in a single GrC, averaged across US-only trials. (d) Examples of GrCs with high correlations to two behavioral parameters. (e) In three mice with the strongest learning and a high CR rate, tuning of all imaged GrCs (gray lines and open circles) to eyelid CR, snout movement, wheel speed and eyelid UR amplitude. Tunings corresponding to the GrCs in a (colored lines and circles) and d (black lines and circles) are highlighted. Bottom: the distribution of GrCs with strongest correlation to each individual behavior. Asterisks indicate US-alone trials.
Figure 6
Figure 6
Redundancy in the GrC representation of learned responses. (a) For one example mouse, a histogram of Pearson correlations for all neurons (red, values larger than the shuffled range shown). The black dashed line indicates the Pearson correlation between the output of the population multivariate regressor and the actual eyelid response (e*) for multiple runs. Gray shading indicates the ± 1 s.d. range of regressor performance. The cyan band indicates the correlations for single cells when the eyelid responses in individual trials are randomly shuffled. (b) Information content of the population regressor (black) as a function of the number of neurons included. The red curve indicates the maximum possible information in the case of no redundancy.

Comment in

References

    1. Ito M. Cerebellar circuitry as a neuronal machine. Prog Neurobiol. 2006;78:272–303. - PubMed
    1. Arenz A, Bracey EF, Margrie TW. Sensory representations in cerebellar granule cells. Curr Opin Neurobiol. 2009;19:445–451. - PubMed
    1. Marr D. A theory of cerebellar cortex. J Physiol (Lond) 1969;202:437–470. - PMC - PubMed
    1. Albus JS. A theory of cerebellar function. Math Biosci. 1971;10:25–61.
    1. Spanne A, Jörntell H. Questioning the role of sparse coding in the brain. Trends Neurosci. 2015;38:417–427. - PubMed
Full text links
Nature Publishing Group full text link Nature Publishing Group Free PMC article
Cite

AltStyle によって変換されたページ (->オリジナル) /