what are the majority of brain cell good for? Clment Lna - - PowerPoint PPT Presentation

The cerebellum: what are the majority of brain cell good for?

Clément Léna

(lena@biologie.ens.fr) Institut de Biologie de l‘École Normale Supérieure, Paris Jussieu 1/12/2015

The cerebellum

10.3% of brain mass

80.2% of brain neurons

81.8% of brain mass

19.0% of brain neurons

7.8% of brain mass 0.8% of brain neurons

cerebellar nuclei cerebellar cortex

The cerebellum

Herculano-Houzel Front Neuroanat 2010

cerebellar nuclei cerebellar cortex

Learning with the cerebellum

• The cerebellum as an associative device
• Plasticity in the cerebellum.
• Prediction of sensory inputs with anti-Hebbian learning
• Generating movement from the cerebellum : eyeblink conditioning.
• Modulating movement with the cerebellum : gain control.
• Programming movement?

cerebellar nuclei

Pre-cerebellar nuclei Post-cerebellar nuclei (mostly pre-motor)

cerebellar cortex cerebellar nuclei

granule cell Purkinje cell Pre-cerebellar nuclei Post-cerebellar nuclei (mostly pre-motor)

cerebellar cortex cerebellar nuclei inferior

• live

granule cell Purkinje cell Pre-cerebellar nuclei Post-cerebellar nuclei (mostly pre-motor)

cerebellar cortex cerebellar nuclei inferior

• live

granule cell Golgi cell stellate/ basket cell Purkinje cell Pre-cerebellar nuclei Post-cerebellar nuclei (mostly pre-motor)

4 mossy fiber/granule cell >150.000 parallel fiber/PC 1 climbing fiber/PC

nb PC ~ 300 x nb gc

Topology of the cerebellar cortex

Coronal view Sagittal view

Pre- and post-cerebellar nuclei

Vestibular information

• Inner ear
• Vestibular nuclei

Sensory-motor information:

• spinal proprioceptive & sensory
• Nucleus dorsalis (dorsal spino-cerebellar tract),
• Cuneate nucleus
• brainstem nuclei : trigeminal & Reticular nuclei (lateral,

paramedian,reticulo-tegmental)

Neocortical inputs:

• Pontine nuclei

Mostly premotor regions in :

• Vestibular nuclei
• Reticular formation
• Red nucleus
• Thalamus (mostly to motor and premotor cortex)

Input: Mossy fibers afferents Output: Projections from cerebellar nuclei Input: Climbing fibers afferents Inferior olive

Medio-lateral segmentation of the cerebellar cortex

Differential expression of molecular markers in Purkinje cells:

CAUDAL ASPECT DORSAL ASPECT ANTERIOR ASPECT

Apps & Hawkes Nature Rev Neurosci 2009

Topography of PC afferents: mossy-fiber input types

Topography of PC afferents: mossy fiber receptive fields

Fractured somatotopy of mossy fibers, conserved across individuals Voogd and Glickstein (1998) TINS 21(9):370-375

Parallel fibers extend over the whole lobule. Thus Purkinje cells receive multiple sensory input types.

Topography of PC afferents: climbing fibers

Inferior olive Cerebellar cortex

Topography of convergent Purkinje cells

Cerebellar nuclei Cerebellar cortex

Summary of the cerebellar circuitry: a powerful associative device

Cerebellar nuclei Cerebellar cortex Inferior olive

Summary

• The topology of cerebellar connectivity maximizes the associative power of

the cerebellum.

Plasticity sites in the cerebellum

cerebellar cortex cerebellar nuclei

granule cell Golgi cell stellate/ basket cell Purkinje cell

Hansel et al. (2001) Nature Neurosci 4, 467 - 475

Learning with the cerebellum: 2 inputs of Purkinje cells

Kreitzer et al. Neuron 2000

20 mm

Philippe Isope & Boris Barbour

Contact fibre parallèle-cellule de Purkinje

Anti-Hebbian learning

Before pairing After pairing Pairing

Learning with the cerebellum: inferior olive induce LTD in Purkinje cell

Sakurai, J. Physiol. 1987 Intracellular recordings in the Purkinje cells

Temps (min)

Vestibular inputs Inferior

• live

granule cells Purkinje cells

mossy fibers climbing fibers parallel fibers

LTD

Vestibular stim 3 min. pairing Inferior olive + Vestibular stim Vestibular stim

5 cells

Temps (min)

Spike counts

Ito et al. (1982) J Physiol. 324:113-34 Extracellular recordings in the Purkinje cells

Learning with the cerebellum: bidirectional plasticity at the synapse between parallel fiber and Purkinje cell

Microcystin: PP inhibitor Chelerythrine: PKC inhibitor

Jörntell & Hansel (2006) Neuron. 52(2):227-38

“Anti-Hebbian” rule: PF-CF coincidence leads to reduced excitation

Learning with the cerebellum: hint of timing window from Ca2+ imaging

Nature Neuroscience 3, 1266 - 1273 (2000)

Wang et al. (2000) Nature Neurosci 3, 1266 - 1273

PF: parallel fibers CF: climbing fibers

Summary

• The topology of cerebellar connectivity maximizes the associative power of

the cerebellum.

• The cerebellum hosts anti-Hebbian learning rule(s) between the parallel fiber

(encoding context) and the climbing fiber (encoding a learning signal).

Sensory prediction: lessons from cerebellum-like structures

Bell et al. Annu Rev Neurosci 31:1-24

Modification of motor command to reduce the error (the climbing fiber)

Adaptive filtering: the decorrelation algorithms

Adaptive interference cancellation Decorrelation control

Subtractions of expected sensory signal the output neuron should discharge only on unexpected inputs

(decorrelation by using anti-Hebbian rule)

Dean et al. (2002) Proc Biol Sci. 269(1503):1895-904.

Cancellation of predictable sensory inputs by the cerebellum

Vestibular neurons Inner ear Motor command

• f head mvt

Cerebellum

=> The cerebellum predicts the sensory inputs and cancels the expected inputs

Summary

• The topology of cerebellar connectivity maximizes the associative power of

the cerebellum.

• The cerebellum hosts anti-Hebbian learning rule(s) between the parallel fiber

(encoding context) and the climbing fiber (encoding a learning signal).

• The anti-Hebbian rule allows the cerebellum to implement adaptive filters

(cancellation of expected input)

Learning with the cerebellum: eyeblink conditioning

Krupa, Thompson and Thompson (1993 ) Science. 260(5110):989-91.

CS+US, interpositus inactivated CS+US, saline in the interpositus CS+US, no infusion

Local infusion of GABA agonist

Regions in the rabbit cerebellum involved in eyeblink control

Hesslow J Physiol 1994

Eyelid Sensory neurons Inferior

• live

Eyelid muscles Cerebellar nuclei

Purkinje cell firing changes in a longitudinal study of aversive conditioning

Jirenhed, D.-A. et al. (2007) J. Neurosci. 27:2493-2502

=> climbing fiber => mossy fiber

Principle of eyeblink conditioning

LTD

Cerebellum microzones: sensory receptive fields of climbing fibers

Martin Garwicz (2002) Brain Research Reviews 40:152–165

Climbing Fiber/mossy fibers sensory receptive field

Apps&Garwicz (2005) Nature Reviews Neuroscience 6, 297-311

Correspondence between single muscle nociceptive receptive fields and climbing fiber receptive fields Muscle EMG following noxious mecanichal stim. Quantitative comparison of the receptive fields of withdrawal reflex and climbing fibers Categories of CF receptive fields Receptive fields for single muscle withdrawal reflex

Cerebellum microzones: relation between olivary inputs and controlled motor units

Muscle group controlled by the target

• f

Purkinje cells Climbing Fiber sensory receptive field Apps&Garwicz (2005) Nature Reviews Neuroscience 6, 297-311

Cerebellum microzones: sensory receptive fields of neurons

Apps&Garwicz (2005) Nature Reviews Neuroscience 6, 297-311

Climbing fiber receptive fields Parallel fiber receptive fields Molecular layer interneuron receptive fields

Purkinje cells receptive fields are defined by plasticity

Jörntell & Hansel (2006) Neuron. 52(2):227-38

Plasticity in interneurons: complementarity with LTD in Purkinje cells

Hebbian learning in interneurons: Parallel fibers and climbing fibers coincidence induce LTP in interneuron (and anti-coincidence produce LTD). Few parallel fibers activate the interneurons Many parallel fibers activate the interneurons

Jorntell & Ekerot J Neurosci (2003) 23(29):9620-9631

NB: The climbing fiber does not make direct synaptic contacts with the interneuron

PF + CF stim.

Learning with the cerebellum: a synthesis with microzonal organization

Summary

• The topology of cerebellar connectivity maximizes the associative power of

the cerebellum.

• The cerebellum hosts anti-Hebbian learning rule(s) between the parallel fiber

(encoding context) and the climbing fiber (encoding a learning signal).

• The anti-Hebbian rule allows the cerebellum to implement adaptive filters

(cancellation of expected input).

• The peripheral control of the climbing fiber is organized as for reflex loops.
• The cerebellar learning leads to climbing fiber cancellation.

Sensory neuron Motoneuron Purkinje cell Context Learning

Two main types of “internal models”

“instructor” (what you want to do) “controler” (how to do it) “effector” (do it!) “instructor” “controler” “effector” Knowing the context, you should do it a bit more this way “inverse” internal model “instructor” “controler” “effector” Knowing the context, that's what you are about to get... “forward” internal model

Forward model in cerebello-cerebral loops

Premotor areas Motor ctx Motoneurons Cerebellum a possible implementation: “instructor” “controler” “effector” If you do it this way, that's what you will get... “forward” internal model

A cerebro-cerebellar loop?

thalamus pontine nucleus cerebellum cortex

granule cells receiving (indirect) inputs from M1

Kelly and Strick, J. Neurosci., 2003

Trans-synaptic virus micro-injection

→ a given site in the motor cortex projects to specific areas of the cerebellar cortex and receives inputs from these specific areas

Anterograde trans-synaptic (HSV) Retrograde trans- synaptic (rabies)

Purkinje cells granule cells

Anatomical evidence for reciprocal connections between the motor cerebral cortex and the cerebellar cortex

Purkinje cells projecting (indirectly) to M1

Sensory-motor convergence in the Cb hemisphere

500um

Crus I

Urethane-anesthetized mice

→ the sensory cortex (provides ‘context’) and the motor cortex (provides ‘current plans’) converge in the cerebellar hemispheres.

Rémi Proville Maria Spolidoro

Auto-fluorescence (metabolic imaging) Electrophysiology

Topography of the ascending pathway

Région activée par M1v

Nicolas Guyon

• Coll. Fekrije Selimi (ICBI, Paris),

Philippe Isope (INCI, Strasbourg)

Souris L7-ChR2 Chaumont et al. PNAS 2013

→ existence of functional cerebello-cerebral loops

Role of cerebro-cerebral loops in motor control

whisking

Rémi Proville Maria Spolidoro

Teaching a forward model

“instructor” “controler” “effector” If you do it this way, that's what you will get... “forward” internal model Sensory processing “sensor” Mmh, wrong prediction

Learning with a forward model

“instructor” Sensorimotor cortex “effector” Cerebellum Sensory processing “sensor” Inferior Olive “forward” internal model

?

The inhibitory nucleo-olivary pathway

Inhibition of climbing fiber response by the nucleo-olivary pathway NO stim (1 to 5 pulses, 200Hz)

Svensson et al. (2006) Exp Brain Res 168: 241–253

Slow because the NO pathway is dominated by asynchronous GABA release (Best and Regehr, 2009)

Purkinje cell illumination induce a delayed increase in olivary firing

Chaumont et al. PNAS 2013

The extended micro-circuit

“instructor” Sensorimotor cortex “effector” Cerebellum Sensory processing “sensor” Inferior Olive

Data from viral tracing : Cortical areas with reciprocal connections with cerebellum

adapted from Middleton and Strick Brain Res Rev (2000)

→ Potential cerebro-cerebellar loops for many anterior brain areas → Include areas involved in cognitive functions

Perspective: non motor functions for the cerebellum?

Co-evolution of the # of neurons in the Cb and Cx

Herculano-Houzel, S. (2010) Front Neuroanat, 4(12):1-8

Human Chimp. Capuchin % of cb volume

Summary

• The topology of cerebellar connectivity maximizes the associative power of

the cerebellum.

• The cerebellum hosts anti-Hebbian learning rule(s) between the parallel fiber

(encoding context) and the climbing fiber (encoding a learning signal).

• The anti-Hebbian rule allows the cerebellum to implement adaptive filters

(cancellation of expected input).

• The peripheral control of the climbing fiber is organized as for reflex loops.
• The cerebellar learning leads to climbing fiber cancellation.
• The cerebellum is involved in loops with the cerebral cortex which may

provide forward models for many cortical areas.

Thank you for your attention!