Whisker Twitching - Espci

Paris, 2008
COMPUTING
WITH
NEURAL ENSEMBLES
Miguel Nicolelis
1 Center for Neuroengineering,
Duke University, Durham, USA
2 Edmond and Lily Safra International
Instituteof Neuroscience of Natal,
Natal, Brazil
nicoleli@neuro.duke.edu
1

Donald O. Hebb
1949: The Organization of Behavior
“…the cell assembly .[is] a diffus
structure comprising [brain] cells
In the cortex and diencephalon,
capable of acting briefly as a
closed system, delivering
facilitation to other such
systems..”

DEFINING THE PRINCIPLES OF
NEURAL ENSEMBLE PHYSIOLOGY
n What is the minimal neural ensemble size
required to sustain a given behavior?
n How does its membership vary from
moment to moment?
n What factors influence its dynamics?
n Can different ensembles produce the
same behavior?
n Can the same ensemble yield multiple
outputs?

John Lillie: an early attempt
625 macroelectrodes
Implanted in a Rhesus monkey
“Carving Brain Activity”

HIGH DENSITY MICROWIRE ARRAYS

Implanting mice

Movable Microwire Array for
Cortical and Subcortical Recodings

PROBING A NEURAL SYSTEM:
THE NEW WAY
CHRONIC IMPLANTS
MULTIPLE
ELECTRODES
MULTIPLE SITES

NEURAL ENSEMBLE DATA ANALYSIS STRATEGY

� SOMATOSENSORY
� MOTOR
� GUSTATORY
� LIMBIC
Mouse = 30-70 neurons
Rat = 100-200 neurons
NEURAL SYSTEMS INVESTIGATED
IN DIFFERENT SPECIES
� SOMATOSENSORY
� MOTOR
Owl Monkey = 100-200 neurons
Rhesus Monkey = 300-500 neurons
� MOTOR
(acute recordings)
Human = 50 neurons

EXPERIMENTAL MODEL:
THE RAT TRIGEMINAL SOMATOSENSORY SYSTEM

Single Trial, Aperture Width
Discrimination
Narrow Aperture Wide Aperture

THE MAP
layer IV

THE RAT TRIGEMINAL
SOMATOSENSORY SYSTEM AND ITS
MAPS
C2
The Classical Model

Simultaneous, Multi-site, Neural Ensemble Recordings

SIMULTANEOUS, MULTI-SITE
NEURAL ENSEMBLE RECORDINGS
Nicolelis et. al., Science 1995

DISTRIBUTED PROCESSING OF TACTILE INFORMATION
IN THE RAT SI CORTEX

SPATIOTEMPORAL RECEPTIVE FIELDS OF SI NEURONS
Ghazanfar and Nicolelis, Cerebral Cortex, 1999.

DYNAMIC RECEPTIVE FIELDS:
MAIN FEATURES
� CORTICAL AND SUBCORTICAL (SI and VPM).
�OBSERVED IN THE SOMATOSENSORY, VISUAL,
AUDITORY, AND TASTE SYSTEMS.
�FORMED BY THE ASYNCHRONOUS CONVERGENCE
OF BOTH FEEDFOWARD AND FEEDBACK PATHWAYS.
�BOTH SPATIAL AND TEMPORAL DOMAINS ARE
SHAPED BY EARLY SENSORY EXPERIENCE.
�PROVIDE THE POTENTIAL FOR IMMEDIATE PLASTIC
REORGANIZATION FOLLOWING PERIPHERAL INJURY.
Nicolelis et. al. Science, 1995.
Ghazanfar and Nicolelis. Cerebral Cortex, 2001.

Active
Exploration
Quiet
Waking
LFPs and behavioral states
Whisker Twitching
(Mu rhythm)
Slow wave
sleep
Paradoxical
/REM sleep

Power
Spectrum
Density
Ratio1:
0-20 Hz
————
0-55 Hz
Ratio2:
0-4 Hz
————
0-10 Hz
Frequency (0-55Hz)
Power
ratio
Frequency (0-55Hz)
Power
ratio
Frequency (0-10Hz)
0.9
0.8
0.7
0.6
0.5
0.6
0.5
0.4
4
sec
1 sec step
Cortical LFP (100 sec, 500Hz sampling)
Blue: raw power ratio
Red: smoothed with Hanning(20) window

PC1 from
Ratio1 of
all LFPs
from Cx,
Th, Hipp,
CPu
PC1 calculated
from Ratio2 of
all LFPs
THE “GREAT ATTRACTOR”

Whisker Twitching
During REM
Whisker Twitching
During μ Rhythm
THE GREAT ATTRACTOR
AND ITS BEHAVIORS
Slow Wave
Sleep
Whisking

THE GREAT ATTRACTOR DEPICTS ALL MAJOR
RAT BEHAVIORAL STATES
Mu
Intermediate
Sleep
REM
Active
SWS
Quiet
Intermediate Sleep
REM Sleep
Slow Wave Sleep
Quiet Awake
Active Awake
Mu State

2DMaps-3

THE GREAT ATTRACTOR CAN BE USED TO
QUANTIFY TRAJECTORIES BETWEEN THE
MAJOR BEHAVIORAL STATES
Mu
Intermediate
Sleep
REM
Active
SWS
Quiet
Intermediate Sleep
REM Sleep
Slow Wave Sleep
Quiet Awake
Active Awake
Mu State

S1
Figure S1.
Diagrams of spontaneous transitions between behavioral states for 120 hours of a rat’s life, determined by the classical classification approach
combining the observation of the overt behavior and analysis of LFP spectral features (Timo-Iaria et al., 1970; Winson, 1974; Fanselow and
Nicolelis, 1999). Each state is represented by a circle, with area proportional to the amount of time spent in that state (% indicated in italic).
Arrows indicate transitions from one state to another. Non-italicized numbers represent the relative occurrence probability of specific transitions.
In agreement with previous studies, we found that rats spent ~60% of the day (light period) sleeping, and ~60% of the night (dark period) awake.
Some state transitions are highly prevalent (e.g. AE↔QW, QW↔SWS, QW↔WT, SWS→REM and REM→QW), while others are either very
rare (SWS→AE, REM→AE and WT→AE) or absent (SWS→WT and AE→REM). REM episodes are nearly always terminated by QW.

WT
IS
The Hypno-map
REM
SWS
QW

TransitionsLength

Transitions

Figure 3. The topography of the two-dimensional state-spaces is consistent across animals. (A) Scatter plots of the 2-D state-space
(conventions as in Fig. 2A). For all animals, 48 hours of recording are displayed; to avoid graphic saturation, only 20% of the data points were
evenly sampled and plotted. (B) Density plots, calculated from the scatter plots, show the conserved cluster topography and the relative
abundance of various states (see also Fig. 1D). (C) Speed plots representing the average velocity of spontaneous trajectories within the 2-D state-
space. Stationarity (low speed) can be observed within the three main clusters, while a maximum speed is reached during transitions from one
cluster to another, i.e. between brain states.

Pooled Coherence (S1, VPM, Hippocampus, Striatum)

6
Figure 6. Pooled coherence, a single measure to address the dynamic of global brain states. 3-D state-space in four rats derived from to
the two amplitude ratios (X and Y axes), and the additional average pooled coherence between 7-55 Hz (Z- axis). The use of pooled
coherence as a single measure of the coupling between forebrain areas captured state-dependent patterns and further improved the
separation between states. Notice that the WT cluster can be easily identified.

CONCLUSIONS – PART I
� A large scale dynamic structure can be identified by combining simultaneously
recorded LFP activity in the SI cortex, VPM thalamus, hippocampus, and striatum.
� Different regions of this dynamic structure correspond to distinct behavioral
states and transitions between states.
� Transitions between states are characterized by the occurrence of transient
periods of high coherence between cortical, thalamic, hippocampal, and striatal
LPF activity. By combining the duration, level of pooled coherence, and phase of
these brief periods of high coherence one may identify the behavioral state the
animal was departured from and the likely state the animal will attain.

Behavioral States Investigated
Quiet:
no movement
Active:
Body movement not involving
whiskers
Whisking:
large-amplitude exploratory
whisker movements
Whisker Twitching:
small-amplitude, rhythmic
whisker movements

STIMULATION AND
RECORDING
METHODS
MICROWIRES
VPM
SI
SURGICAL
SUTURE
CUFF ELECTRODE
TEFLON-COATED
WIRES
PLATINUM
BANDS
SYLGARD
1 mm
Fanselow and Nicolelis. J. Neurosci. 1999

RESPONSES TO ELECTRICAL NERVE CUFF STMULATION
VS. MANUAL WHISKER DEFLECTION
SINGLE UNIT
MULTI-UNIT
CLUSTER
80
40
0
80
40
NERVE CUFF
STIMULATION
MANUAL WHISKER
DEFLECTION
0
-50 0 50 -50 0 50 ms

SPIKES/SEC
200
100
0
200
100
0
BEHAVIORAL MODIFICATION OF TACTILE RESPONSES
IN SINGLE UNITS
VPM THALAMUS
SI CORTEX
QUIET ACTIVE WHISKING WHISKER
TWITCHING

BEHAVIORAL MODIFICATION OF TACTILE RESPONSES
IN SINGLE UNITS
100
INTEGRATED
MAXIMUM
RESPONSE MAGNITUDE
50
0
100
50
0
VPM THALAMUS SI CORTEX
*
*
Q A W WT Q A W WT
*
*
*
*

QUIET
ACTIVE
WHISKING
WHISKER
TWITCHING
MODULATION OF POST-STIMULUS INHIBITION
BY BEHAVIORAL STATE
40
20
0
40
20
0
40
20
0
40
20
0
VPM THALAMUS SI CORTEX
0 50 100 150 200 0 50 100 150 200
MSEC
SPIKES/SEC

MAXIMUM
MAGNITUDE
INTEGRATED
RESPONSE
150
100
50
0
150
100
50
0
BEHAVIORAL MODULATION OF RESPONSES TO
PAIRED STIMULI IN RAT SI CORTEX
QUIET ACTIVE WHISKING
1st 50 100 150 200 1st 50 100 150 200 1st 50 100 150 200
2nd stimulus
ISI IN MSEC

BURST FIRING IN VPM THALAMUS
quiet whisker twitching
Fanselow et. al. PNAS, 2001

TONIC AND BURSTING MODES
(Jahansen and Llinas 1984, McCormick 1992, Sherman and Guillery 1996)

SI
SI
VPM
VPM
Field potential recordings
during the whisker twitching
behavior
SI
SI
VPM
VPM

High Sensitivity Correlated with Whisker Movements
Fanselow
et al. 2001
Nicolelis
et al. 1995
Facilitated detection of novel or faint stimuli during bursting mode: Guido and Weyand 1995

7-12Hz Oscillations (μ rhythm) tend to spread
from the SI cortex to the VPM thalamus

SI
VPM
SI Inactivation Abolishes VPM bursting during Whisker Twitching
SI INTACT
SI BLOCKED
SI INTACT
SI BLOCKED

CONCLUSIONS PART II
� THE PHYSIOLOGICAL PROPERTIES OF BOTH THALAMIC
AND CORTICAL NEURONS ARE DETERMINED BY THE
ANIMAL’S BEHAVIORAL STATE.
� WE PROPOSE THAT BOTH CELLULAR PROPERTIES AND CIRCUIT
INTERACTIONS IN THE RAT THALAMOCORTICAL LOOP SHIFT IN PARALLEL
WITH CHANGES IN THE ANIMAL’S EXPLORATORY STRATEGY.
� IN THIS MODEL, DURING QUIET AND WHISKER TWITCHING STATES, THE
PHYSIOLOGICAL PROPERTIES OF THE THALAMOCORTICAL ARE
OPTIMIZED FOR DETECTION OF NOVEL OR SMALL TACTILE STIMULI.
� CONVERSELY, DURING WHISKING, THALAMIC AND CORTICAL NEURONS
BECOME OPTIMIZED TO PROCESS INFORMATION DERIVED FROM FAST
SEQUENCES OF MULTI-WHISKER STIMULI, WHICH UNDERLIE THE ANIMAL’S
ABILITY TO DISCRIMINATE MULTIPLE TACTILE ATTRIBUTES OF AN OBJECT.

TACTILE DISCRIMINATION TASK

TACTILE DISCRIMINATION OF THE APERTURE WIDTH
REQUIRES MULTIPLE WHISKERS

TASK COMPLETION REQUIRES THE SI CORTEX

BEHAVIORAL PERFORMANCE OVER TIME

Chronic Recordings in Multiple Cortical Layers
RAT
“BARREL”
SI
CORTEX
Layer II/III
Layer IV
Layer V

NEURAL ENSEMBLE DATA ANALYSIS STRATEGY

Active Vs. Passive Discrimination

SI NEURONS HAVE BILATERAL
RECEPTIVE FIELDS
Shuler, Krupa, and Nicolelis. J. Neurosci. 2001.

CONCLUSIONS – PART III
• Tactile cortical responses during active exploration of objects are very
distinct from those observed in anesthetized or even immobilized rats.
• Inhibition may play an important role in the representation of bilateral
tactile stimuli during active whisker discrimination.
• Different types of inhibition patterns are seen in the rat S1 cortex. Different
layers exhibit a predominance of distinct patterns of inhibition.
• Callosal interactions may account for some of the inhibition observed
during integration of bilateral whisker stimuli.
• Other afferents may contribute to other types of inhibitory effects (long-
inhibition) that account for sensory gating of tactile signals during active
whisker exploration

ROBUST TACTILE
RESPONSES
IN THE RAT
DORSAL
HIPPOCAMPUS
Pereira at. al., PNAS 2007

MUSCIMOL INJECTION
S1
MUSCIMOL INJECTION
VPM THALAMUS
MUSCIMOL INJECTION
S1
TACTILE REPONSES IN THE RAT
DORSAL HIPPOCAMPUS
DEPEND ON SOMATOSENSORY
LEMNISCAL PATHWAY
MUSCIMOL INJECTIONS
S1 AND VPM

Pereira at. al., PNAS 2007
Perforant

Pereira at. al., PNAS 2007
active

HIPPOCAMPAL ENSEMBLE ACTIVITY
PREDICTS OUTCOME OF
TACTILE DISCRIMINATION TASK
Pereira at. al., PNAS 2007
• Differences on neuronal
ensemble activity as a
function of task
performance
• Learning Vector
Quantization (LVQ) –
non-parametric statistical
pattern recognition.
• Implemented on an
Artificial Neural Network
• M. Laubach, J. Wessberg, M. A. Nicolelis,
Nature 405, 567 (2000).

COLLABORATORS
DUKE UNIVERSITY
DEPT. NEUROBIOLOGY
• DAVID KRUPA, Ph.D.
• MICHAEL WIEST, Ph.D.
• SIDARTA RIBEIRO, Ph.D.
• DAMIEN GERVASONI, Ph.D.
• SHIH-CHIEH LIN
• ANTONIO PEREIRA
• GARY LEHEW

COLLABORATORS
DUKE UNIVERSITY
DEPT. NEUROBIOLOGY
• David Krupa, Ph.D.
• Erika Fanselow, Ph.D.
• Marshall Shuler, Ph.D.
• Sidarta Ribeiro, Ph.D.
• Damien Gervasoni, Ph.D.
• Janaina Pantoja
• Gary Lehew
• Laura Oliveira, M.D.