Article | Published:

Coordinated cerebellar climbing fiber activity signals learned sensorimotor predictions

Nature Neurosciencevolume 21pages14311441 (2018) | Download Citation


The prevailing model of cerebellar learning states that climbing fibers (CFs) are both driven by, and serve to correct, erroneous motor output. However, this model is grounded largely in studies of behaviors that utilize hardwired neural pathways to link sensory input to motor output. To test whether this model applies to more flexible learning regimes that require arbitrary sensorimotor associations, we developed a cerebellar-dependent motor learning task that is compatible with both mesoscale and single-dendrite-resolution calcium imaging in mice. We found that CFs were preferentially driven by and more time-locked to correctly executed movements and other task parameters that predict reward outcome, exhibiting widespread correlated activity in parasagittal processing zones that was governed by these predictions. Together, our data suggest that such CF activity patterns are well-suited to drive learning by providing predictive instructional input that is consistent with an unsigned reinforcement learning signal but does not rely exclusively on motor errors.

Access optionsAccess options

Rent or Buy article

Get time limited or full article access on ReadCube.


All prices are NET prices.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Additional information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.


  1. 1.

    Albus, J. S. A theory of cerebellar function. Math. Biosci. 10, 25–61 (1971).

  2. 2.

    Ito, M. Neural design of the cerebellar motor control system. Brain Res. 40, 81–84 (1972).

  3. 3.

    Marr, D. A theory of cerebellar cortex. J. Physiol. (Lond.) 202, 437–470 (1969).

  4. 4.

    Ito, M., Sakurai, M. & Tongroach, P. Climbing fibre induced depression of both mossy fibre responsiveness and glutamate sensitivity of cerebellar Purkinje cells. J. Physiol. (Lond.) 324, 113–134 (1982).

  5. 5.

    Ekerot, C. F. & Kano, M. Long-term depression of parallel fibre synapses following stimulation of climbing fibres. Brain Res. 342, 357–360 (1985).

  6. 6.

    Tank, D. W., Sugimori, M., Connor, J. A. & Llinás, R. R. Spatially resolved calcium dynamics of mammalian Purkinje cells in cerebellar slice. Science 242, 773–777 (1988).

  7. 7.

    Llinás, R. & Sugimori, M. Electrophysiological properties of in vitro Purkinje cell somata in mammalian cerebellar slices. J. Physiol. (Lond.) 305, 171–195 (1980).

  8. 8.

    Medina, J. F., Nores, W. L., Ohyama, T. & Mauk, M. D. Mechanisms of cerebellar learning suggested by eyelid conditioning. Curr. Opin. Neurobiol. 10, 717–724 (2000).

  9. 9.

    Ohyama, T., Nores, W. L., Murphy, M. & Mauk, M. D. What the cerebellum computes. Trends Neurosci. 26, 222–227 (2003).

  10. 10.

    Raymond, J. L., Lisberger, S. G. & Mauk, M. D. The cerebellum: a neuronal learning machine? Science 272, 1126–1131 (1996).

  11. 11.

    Schultz, S. R., Kitamura, K., Post-Uiterweer, A., Krupic, J. & Häusser, M. Spatial pattern coding of sensory information by climbing fiber-evoked calcium signals in networks of neighboring cerebellar Purkinje cells. J. Neurosci. 29, 8005–8015 (2009).

  12. 12.

    Tsutsumi, S. et al. Structure-function relationships between aldolase C/zebrin II expression and complex spike synchrony in the cerebellum. J. Neurosci. 35, 843–852 (2015).

  13. 13.

    Ozden, I., Dombeck, D. A., Hoogland, T. M., Tank, D. W. & Wang, S. S. Widespread state-dependent shifts in cerebellar activity in locomoting mice. PLoS One 7, e42650 (2012).

  14. 14.

    De Gruijl, J. R., Hoogland, T. M. & De Zeeuw, C. I. Behavioral correlates of complex spike synchrony in cerebellar microzones. J. Neurosci. 34, 8937–8947 (2014).

  15. 15.

    Tang, T., Suh, C. Y., Blenkinsop, T. A. & Lang, E. J. Synchrony is key: complex spike inhibition of the deep cerebellar nuclei. Cerebellum 15, 10–13 (2016).

  16. 16.

    Lee, K. H. et al. Circuit mechanisms underlying motor memory formation in the cerebellum. Neuron 86, 529–540 (2015).

  17. 17.

    Histed, M. H., Carvalho, L. A. & Maunsell, J. H. Psychophysical measurement of contrast sensitivity in the behaving mouse. J. Neurophysiol. 107, 758–765 (2012).

  18. 18.

    Chettih, S. N., McDougle, S. D., Ruffolo, L. I. & Medina, J. F. Adaptive timing of motor output in the mouse: the role of movement oscillations in eyelid conditioning. Front. Integr. Neurosci. 5, 72 (2011).

  19. 19.

    Gaffield, M. A., Amat, S. B., Bito, H. & Christie, J. M. Chronic imaging of movement-related Purkinje cell calcium activity in awake behaving mice. J. Neurophysiol. 115, 413–422 (2016).

  20. 20.

    Mukamel, E. A., Nimmerjahn, A. & Schnitzer, M. J. Automated analysis of cellular signals from large-scale calcium imaging data. Neuron 63, 747–760 (2009).

  21. 21.

    Najafi, F., Giovannucci, A., Wang, S. S. & Medina, J. F. Coding of stimulus strength via analog calcium signals in Purkinje cell dendrites of awake mice. eLife 3, e03663 (2014).

  22. 22.

    Ozden, I., Lee, H. M., Sullivan, M. R. & Wang, S. S. Identification and clustering of event patterns from in vivo multiphoton optical recordings of neuronal ensembles. J. Neurophysiol. 100, 495–503 (2008).

  23. 23.

    Ozden, I., Sullivan, M. R., Lee, H. M. & Wang, S. S. Reliable coding emerges from coactivation of climbing fibers in microbands of cerebellar Purkinje neurons. J. Neurosci. 29, 10463–10473 (2009).

  24. 24.

    Najafi, F., Giovannucci, A., Wang, S. S. & Medina, J. F. Sensory-driven enhancement of calcium signals in individual Purkinje cell dendrites of awake mice. Cell Rep. 6, 792–798 (2014).

  25. 25.

    Najafi, F. & Medina, J. F. Beyond “all-or-nothing” climbing fibers: graded representation of teaching signals in Purkinje cells. Front. Neural Circuits 7, 115 (2013).

  26. 26.

    Yang, Y. & Lisberger, S. G. Purkinje-cell plasticity and cerebellar motor learning are graded by complex-spike duration. Nature 510, 529–532 (2014).

  27. 27.

    Kitamura, K. & Häusser, M. Dendritic calcium signaling triggered by spontaneous and sensory-evoked climbing fiber input to cerebellar Purkinje cells in vivo. J. Neurosci. 31, 10847–10858 (2011).

  28. 28.

    Konnerth, A., Dreessen, J. & Augustine, G. J. Brief dendritic calcium signals initiate long-lasting synaptic depression in cerebellar Purkinje cells. Proc. Natl. Acad. Sci. USA 89, 7051–7055 (1992).

  29. 29.

    Miyakawa, H., Lev-Ram, V., Lasser-Ross, N. & Ross, W. N. Calcium transients evoked by climbing fiber and parallel fiber synaptic inputs in guinea pig cerebellar Purkinje neurons. J. Neurophysiol. 68, 1178–1189 (1992).

  30. 30.

    Eilers, J., Augustine, G. J. & Konnerth, A. Subthreshold synaptic Ca2+ signalling in fine dendrites and spines of cerebellar Purkinje neurons. Nature 373, 155–158 (1995).

  31. 31.

    Hartell, N. A. Strong activation of parallel fibers produces localized calcium transients and a form of LTD that spreads to distant synapses. Neuron 16, 601–610 (1996).

  32. 32.

    Apps, R. & Garwicz, M. Anatomical and physiological foundations of cerebellar information processing. Nat. Rev. Neurosci. 6, 297–311 (2005).

  33. 33.

    Ohmae, S. & Medina, J. F. Climbing fibers encode a temporal-difference prediction error during cerebellar learning in mice. Nat. Neurosci. 18, 1798–1803 (2015).

  34. 34.

    Ackerley, R., Pardoe, J. & Apps, R. A novel site of synaptic relay for climbing fibre pathways relaying signals from the motor cortex to the cerebellar cortical C1 zone. J. Physiol. (Lond.) 576, 503–518 (2006).

  35. 35.

    Kitazawa, S., Kimura, T. & Yin, P. B. Cerebellar complex spikes encode both destinations and errors in arm movements. Nature 392, 494–497 (1998).

  36. 36.

    Zhou, H. et al. Cerebellar modules operate at different frequencies. eLife 3, e02536 (2014).

  37. 37.

    Paukert, M., Huang, Y. H., Tanaka, K., Rothstein, J. D. & Bergles, D. E. Zones of enhanced glutamate release from climbing fibers in the mammalian cerebellum. J. Neurosci. 30, 7290–7299 (2010).

  38. 38.

    Blenkinsop, T. A. & Lang, E. J. Synaptic action of the olivocerebellar system on cerebellar nuclear spike activity. J. Neurosci. 31, 14708–14720 (2011).

  39. 39.

    Bengtsson, F., Ekerot, C. F. & Jörntell, H. In vivo analysis of inhibitory synaptic inputs and rebounds in deep cerebellar nuclear neurons. PLoS One 6, e18822 (2011).

  40. 40.

    Person, A. L. & Raman, I. M. Purkinje neuron synchrony elicits time-locked spiking in the cerebellar nuclei. Nature 481, 502–505 (2011).

  41. 41.

    McElvain, L. E., Bagnall, M. W., Sakatos, A. & du Lac, S. Bidirectional plasticity gated by hyperpolarization controls the gain of postsynaptic firing responses at central vestibular nerve synapses. Neuron 68, 763–775 (2010).

  42. 42.

    Pugh, J. R. & Raman, I. M. Potentiation of mossy fiber EPSCs in the cerebellar nuclei by NMDA receptor activation followed by postinhibitory rebound current. Neuron 51, 113–123 (2006).

  43. 43.

    Miles, F. A. & Lisberger, S. G. Plasticity in the vestibulo-ocular reflex: a new hypothesis. Annu. Rev. Neurosci. 4, 273–299 (1981).

  44. 44.

    Matias, S., Lottem, E., Dugué, G. P. & Mainen, Z. F. Activity patterns of serotonin neurons underlying cognitive flexibility. eLife 6, e20552 (2017).

  45. 45.

    Belova, M. A., Paton, J. J., Morrison, S. E. & Salzman, C. D. Expectation modulates neural responses to pleasant and aversive stimuli in primate amygdala. Neuron 55, 970–984 (2007).

  46. 46.

    Hayden, B. Y., Heilbronner, S. R., Pearson, J. M. & Platt, M. L. Surprise signals in anterior cingulate cortex: neuronal encoding of unsigned reward prediction errors driving adjustment in behavior. J. Neurosci. 31, 4178–4187 (2011).

  47. 47.

    Pearce, J. M. & Hall, G. A model for Pavlovian learning: variations in the effectiveness of conditioned but not of unconditioned stimuli. Psychol. Rev. 87, 532–552 (1980).

  48. 48.

    Roesch, M. R., Esber, G. R., Li, J., Daw, N. D. & Schoenbaum, G. Surprise! Neural correlates of Pearce-Hall and Rescorla-Wagner coexist within the brain. Eur. J. Neurosci. 35, 1190–1200 (2012).

  49. 49.

    Raymond, J. L. & Medina, J. F. Computational principles of supervised learning in the cerebellum. Annu. Rev. Neurosci. 41, 233–253 (2018).

  50. 50.

    de Solages, C. et al. High-frequency organization and synchrony of activity in the Purkinje cell layer of the cerebellum. Neuron 58, 775–788 (2008).

Download references


We thank L. Glickfeld for helpful discussions and input on calcium-imaging approaches and analyses, S. Lisberger, G. Field, K. Franks, and F. Wang for feedback on early versions of this manuscript, and members of the Hull and Glickfeld labs for input and technical assistance throughout the project. This work was supported by grants from the NIH NINDS (5R01NS096289-02, C.H.) and (F31NS103425, W.H.), the Sloan Foundation (C.H.), and the Whitehall Foundation (C.H.).

Author information

Author notes

    • Andrew McKinney

    Present address: Department of Neuroscience, Baylor College of Medicine, Houston, TX, USA

    • Mati Joshua

    Present address: Edmond and Lily Safra Center for Brain Sciences, The Hebrew University, Jerusalem, Israel


  1. Department of Neurobiology, Duke University Medical School, Durham, NC, USA

    • William Heffley
    • , Eun Young Song
    • , Ziye Xu
    • , Benjamin N. Taylor
    • , Mary Anne Hughes
    • , Andrew McKinney
    • , Mati Joshua
    •  & Court Hull


  1. Search for William Heffley in:

  2. Search for Eun Young Song in:

  3. Search for Ziye Xu in:

  4. Search for Benjamin N. Taylor in:

  5. Search for Mary Anne Hughes in:

  6. Search for Andrew McKinney in:

  7. Search for Mati Joshua in:

  8. Search for Court Hull in:


W.H., E.Y.S., and C.H. designed the experiments. W.H., E.Y.S., A.M., B.N.T., M.A.H., and M.J. conducted the experiments. W.H., E.Y.S., Z.X., B.N.T., A.M., M.J., and C.H. analyzed the data. C.H. wrote the manuscript.

Competing interests

The authors declare no competing interests.

Corresponding author

Correspondence to Court Hull.

Integrated supplementary information

  1. Supplementary Figure 1 Mapping of motor output driven by Arch and electrical stimulation in superficial cerebellar cortex.

    a) Schematics of the results of Arch stimulation over a grid of sites in lobule simplex (Sim) and Crus I for five mice. Colors denote outcome of stimulation. b) Same as A for three mice tested with electrical microstimulation.

  2. Supplementary Figure 2 Retention and extinguishment of learning for cue-prediction regime.

    a) Average reaction times relative to cue for the first (left) and second (right) sessions where animals performed 0.5 s cue prediction on consecutive training days (n = 17 sessions, 5 mice). Error bars are ±SEM across sessions. b) Same as a for 2 s cue prediction sessions (n = 10 sessions, 6 mice). c) Left, example cumulative distributions of reaction times for cue reaction sessions immediately surrounding a cue prediction session (session 0). Trials from the cue prediction session include only the last 1/3 of the session when the animal was anticipating the cue timing. Right, summary of distribution skewness across cue reaction sessions immediately surrounding cue prediction sessions (n = 43 sessions). Error bars are SEM across sessions. d) Same as a for pairs of 0.5 s prediction sessions separated by 3 or more cue reaction sessions (n = 20 sessions, 10 mice).

  3. Supplementary Figure 3 Performance in the cue-prediction condition in control and NBQX sessions.

    a-e) Summary across experiments of mean percent correct lever releases, mean baseline reaction time from the visual cue, mean reaction time variance, mean duration to press, mean number of trials performed, and the mean duration of behavioral sessions. n = 44 sessions, 0.5 s; n = 9, 1.0 s; n = 6, 1.5 s; n = 30, 2.0 s; n = 26, Δt; n = 14, 0.5 s NBQX. Error bars are SEM across sessions. For each cue delay, the minimum and maximum (x,y) number of trials performed in single sessions was: 0.5 s: (309, 700) 1 s: (252, 900) 1.5 s: (202, 775) 2 s: (278,1185) Δt: (313, 772) 0.5 s NBQX: (227, 429). fi) Individual (top) and average (bottom) lever trajectories aligned to lever press (left) and release (right) segregated by duration into quartiles (n = 112 trials/quartile) for a representative 0.5 s cue prediction session. fii) Same as fi for a representative 0.5 s cue prediction session in NBQX. (n = 142 trials/quartile). g) Summary of press (left) and release (right) quartiles across experiments for control (top; n = 15 sessions) and NBQX (bottom; n = 14 sessions) experiments. h) Summary of the ratio between press and release ranges (difference between 1st and 4th quartile) for control (black) and NBQX (blue) experiments.

  4. Supplementary Figure 4 Pharmacological modulation of complex spiking in lobule simplex.

    (a) Single unit recordings during NBQX application to LS. Top, summary of complex spike rates in control and after application of NBQX for individual cells (black; n = 9 cells, 5 mice) and the average of all cells (red). Error bars are ± SEM across cells. Bottom, same as top for simple spike rates. Note that NBQX strongly reduces complex but not simple spike rates. (b) (top) Raw epifluorescence images from parasagittal cerebellar sections showing fluorescein labeling after surface application to lobule simplex in vivo (Methods, 3 mice). (bottom) Pixel masks from the same images above. Pixels were thresholded at 30% of maximum value to visualize and quantify fluorescein labeling. (c) Summary of mediolateral fluorescein spread across sections. (n = 3 animals). Error bars are SEM across animals.

  5. Supplementary Figure 5 Segmentation of PC dendrites from imaging experiments.

    (a) Example single photon imaging field of view (left) is segmented to mask individual dendrites (right) used for subsequent single cell analysis. Scale bar = 100 μm. (b) Example timecourse of raw fluorescence (top) from the circled dendrite in A showing individual calcium transients identified (gray circles) according to peaks in the first derivative of the fluorescence trace (bottom). (c) Same as a for an example two photon field of view. Note that while PC somata are visible (round, top right), image segmentation does not extract activity from these structures. Scale bar = 100 μm. (d) Same as b for the experiment in C. (e) Top, example time course of a single unit PC recording from an awake mouse illustrating detection of complex spikes (gray circles). Bottom, overlay of individual simple (left) and complex (right) spike waveforms (gray) and the average waveform (black). (f) Summary of mean spike rates ± SEM across all two photon dendrite imaging sessions (blue; n = 1146 dendrites) and acute single unit recordings (n = 11 units).

  6. Supplementary Figure 6 Single-unit PC recordings of complex spiking during the cue-prediction condition.

    (a) Top, Schematic of the trial structure schematic, where a constant cue delay of 500 ms was imposed on each trial. Middle top, Average simple (SS, left) and complex (CS, right) spike waveforms from an example single unit. Middle bottom, Histogram of SS firing rate aligned to complex spike time revealing post-CS pause, confirming isolation of single PC. Bottom, raster of single trial complex spikes and session PSTH aligned to lever release. (b) Single trial voltage traces (complex spikes- open blue circle) from the cell in A aligned to lever release (correct- black circle; early- red circle). (c) Expansion of 5 consecutive traces from B at time of lever release. (d) Mean normalized complex spike rate (Methods) from single unit recordings (31 PCs, 8 mice) aligned to lever release for correct (black) and early (red) trials during the cue prediction condition. Shaded error is SEM across PCs. (e) Mean normalized complex spike rates aligned to cue presentation for the first 1/3 of trials (black) and the last 1/3 of trials (orange) across cue prediction sessions for correct (left) and early (right) trials (31 PCs, 8 mice). Shaded error is SEM across PCs. (f) Summary of the S.D. of spike times when aligned to either lever release or visual cue for each PC (n = 31). P-values reflect paired t-tests.

  7. Supplementary Figure 7 Differential complex spiking on correct and early-release trials is not driven by motor signals due to licking.

    (a) Example imaging data from an inter-trial interval (ITI) showing a lick responsive dendrite (top, dark blue) and a lick unresponsive dendrite (bottom, black) in the same field of view. (b) Averaged lick triggered calcium transients for the lick responsive dendrite (dark blue) and a lick unresponsive dendrite (black) in A). n = 28 licks, Error bars are SEM across lick-triggered events. (c) Mean lick triggered calcium transient across all lick responsive (dark blue) and lick unresponsive (black) dendrites. Shaded error is SEM across dendrites. (d) Summary of the Spearman’s correlation between lick rate and amplitude of the calcium transient in each frame across trials for lick responsive (dark blue) and lick unresponsive dendrites (grey). Red points indicate the mean ± SEM across dendrites (447 dendrites, 6 animals and 10 sessions). (e) Summary comparing the amplitude of calcium transients for lick responsive neurons only in response to either a single lick (blue) or a lick bout (3 or more licks with < 300 ms between licks; cyan). Note that additional licking does not produce larger responses. (f) Summary comparing the amplitude of calcium transients on correct (black) and early (red) lever releases for lick unresponsive dendrites only. (g) Same as f, for lick responsive dendrites. P-values reflect paired t-tests.

  8. Supplementary Figure 8 Example single-trial and session average calcium transients from two-photon imaging.

    (a) Example single trial data from an individual dendrite showing the normalized calcium transient across an early release trial from before lever press to after the lever release. (b) Same as a but for a correct trial, including the time of cue presentation. (c) Mean calcium transients across trials (n = 186) for the dendrite in A and B. Error bars are SEM across trials. (d) Single trial example, aligned to lever release, showing all dendrites from the experiment illustrated in Fig. 5c on a correct trial over 1 second surrounding release (n = 115 dendrites). Lever press is off scale. (e) Same as d, but for an early release trial from the same imaging session.

  9. Supplementary Figure 9 Unexpected reward drives complex spiking in lick-unresponsive and -responsive dendrites.

    a-b) Top, Average calcium transient in response to unexpected reward (green, aligned to first lick) and correct lever releases (black, aligned to release). Shaded error is SEM across dendrites. Bottom, Average lick rate for unexpected reward and correct lever releases. Error is SEM across experiments (n = 3). Lick responsiveness was defined according to significant responses in the lick triggered averaged taken from the inter-trial interval (Supp. Figure 7). c) Summary of peak calcium transients in the same neurons for correct lever releases vs unexpected reward trials. Note that responses are proportional, and response amplitude is determined by response probability (Figs 5, 6). P-value reflects paired t-test.

  10. Supplementary Figure 10 Complex spiking scales with lever hold time for early releases in single-photon imaging experiments.

    Summary of peak calcium transients in a 500 ms window at the time of release (Methods) across all single photon experiments for correct (black) and early (red) release trials binned according to hold time (250 ms bins). Linear fits were applied to data from each trial type (n = 10 animals, 17 sessions). Error bars are SEM across sessions.

  11. Supplementary Figure 11 Mice have full control over the lever trajectory.

    (a) Individual (top, n = 217) and average (bottom, 54 per quartile) lever trajectories aligned to lever press (left) and release (right) for manual lever depression followed by unperturbed gravity return. Shaded error is SEM across trials. (b) Summary of press (left) and release (right) quartiles across experiments for manual lever experiments (n = 4). Note the lack of variability across quartiles. (c) Summary scatterplot comparing 20-80% rise times of lever press and release for experiments where the lever was controlled by mice (black, n = 15) and manual press (green, n = 4) (d) Summary comparison measuring the average difference from the mean of the slowest release quartile and the fastest release quartile for experiments where the lever was controlled by mice (black) and manual press (green). Lever trajectories were sorted according to duration in a window from 200 ms before threshold crossing to the time of a threshold crossing half way between the top and bottom of the total lever displacement. The slowest and fastest 25 trials were extracted for comparison, and their average rise time was normalized by subtracting that of the mean trajectory of the whole session. Error bars are SEM across sessions. These results show that mice produce movements both slower and faster than the mean, indicating full control over the lever trajectory.

Supplementary information

  1. Supplementary Text and Figures

    Supplementary Figures 1–11

  2. Reporting Summary

  3. Supplementary Video 1

    Optogenetic silencing of superficial PCs in lobule simplex drives ipsilateral forelimb movements. An external fiber coupled laser was activated at the time indicated by the green circle above lobule simplex, resulting in forelimb movement

  4. Supplementary Video 2

    Example correct and early release trials from the cue prediction condition. Two example trials from the same mouse in the same training session are shown first in real time, and then repeated at 0.25× video rate

About this article

Publication history