Daniel J. Thengone, Yunguo Yu, Jonathan D. Victor
0
90
250
180
270
0
1
0
0
0
200
400
600 >800
Trough-peak interval (µs)
this unit
40
0.6 0
20
-0.6 −200 0
Time (µs)
800
0
200
400
600 >800
Trough-peak interval (µs)
25
0 repulsive
Untuned Inhibition
Neural Response
90
50
Pre-Synaptic
45
25
0
90 repulsive Relative peak position post-adaptation(deg)
el ls fc ro be Nu
m
Tuned Inhibition
Site of Adaptation
ro fc be m
Number of cells
be m
45
Tuned Inhibition
el ls Half-Bandwidth post-adaptation (deg)
el ls fc 800
Number of cells
m Nu
ro
−200 0
Nu
ro fc be
Number of cells Number of cells
Half-Bandwidth post-adaptation (deg)
el ls fc ro be
Number of cells
Nu m
Half-Bandwidth post-adaptation (deg)
ls fc el
el ls of c be r
be ro
ls el fc ro be Nu m
Nu m
Nu m
Number of cells
0
14 0
45
45
repulsive
20
40
40 broadening
45
Half-Bandwidth post-adaptation (deg)
20
Nu
ls el fc ro be N um
Number of cells Number of cells
-0.6
Relative peak position post-adaptation(deg)
20
Circular Variance
Half-Bandwidth
el ls
ls el ro fc be m Nu
Number of cells Number of cells Relative peak position post-adaptation(deg) Relative peak position post-adaptation(deg)
Normalized Voltage Normalized Voltage
Normalized Voltage
0
Normalized Voltage
S-1
Half-Bandwidth post-adaptation (deg)
Normalized Voltage
deg deg
Normalized Amplitude
Normalized Amplitude
Number of cells
Spike Rate (S-1) Spike Rate (S-1) Spike Rate (S-1)
10
100
40
0.6
Time (µs)
Amplitude
0
this unit
270
adapted
12
Number of cells
W HB
180
14
attractive
0
45
Relative peak position pre-adaptation(deg)
0 10
20
in
90
0
0
Half-Bandwidth pre-adaptation (deg)
0
ge
0
0
W
S-1
1
0 20
an
30
30
Circular Variance
40
V2
Trough-peak interval (µs)
0
HB
600 >800
10 Half-Bandwidth
35
20
Ch
400
in
200
ge
Time (µs)
Amplitude
0
0
800
adapted
30
0
−200 0
0
0
The bimodality in the distribu on of trough to peak widths (p < 0.01 by the Har gan dip test) was used to classify extracellular waveforms as narrow-spiking (< 405μs) puta ve inhibitory interneurons and broad-spiking (> 430μs) puta ve excitatory neurons (consistent with Mitchell et al., 2007). Neurons within 10% of the notch were labeled unclassified and not used in further analyses.
-0.6
0
narrowing
0
Trough-peak width (µs)
>800
20
Number of cells
an
600
0
50
40
broadening
45
50
25
attractive
0
0
Number of cells
10
40
20
20
narrowing 0
20
40
Half-Bandwidth pre-adaptation (deg)
0 15
0
40 broadening
0
45 90 Relative peak position pre-adaptation(deg)
narrowing
25 50 Half-Bandwidth pre-adaptation (deg)
40
20
Common to all Models
20
narrowing 20
40
0 15
0
Number of cells
15
Number of cells
15
• • •
In the granular layer of V1, brief adaptation of inhibitory celltypes mostly broadens tuning bandwidth while prolonged adaptation typically narrows it.
20
-2
400
0
0.6
0
20
14
200
35
Ch
0
Relative peak position pre-adaptation(deg)
40
-3
Trough-peak width (µs)
>800
0
40
0
600
45
20
0
400
270
1
this unit
40
90
0
Excitatory and inhibitory neurons both show adaptive shifts of their orientation tuning curves, for brief and prolonged presentations of the adapting stimulus.
25
200
180
0
attractive
0
0
0
0
90
0
0
mean and 95% CI
Circular Variance
Half-Bandwidth
30
S-1 0
V2
0
0
Conclusions
14
30
45
30
Unclassified
Amplitude
45
) eg
20
adapted
30
20
(d
Broad-spiking (Excitatory neuron)
repulsive
10 baseline adapted
0
40 broadening
0
Narrow-spiking (Inhibitory interneuron)
800
45
15
Time (µs)
20
0
0
20
Half-Bandwidth pre-adaptation (deg)
20
0
35
35
0
0
10
V1 Layers Combined
10
20
−200
(+/- 1 SD)
0
Number of cells
ift
Time (µs)
800
Trough-peak width
0 10
0
0
0 -0.6
Trough-peak interval (µs)
20
Half-Bandwidth pre-adaptation (deg)
Sh
−200
(+/- 1 SD)
40
600 >800
ak Pe
-0.6
400
narrowing 40
0
0
0.6
Time (µs)
270
200
Number of cells
0
10
40
0.6
•
•
V2
0
800
10
20 15 -
V1
−200 0
Adapta on can induce changes in halfbandwidth, and the tuning curve peak shi can be either towards the adapter (a rac ve) or away from it (repulsive). Layer 4 shows a dis nc ve behavior: brief adapta on typically broadens the tuning bandwidth, while prolonged adapta on typically narrows it. In area V2 (shown below), adapta on both broadens and narrows half-bandwidth in the inhibitory units, while the excitatory units mostly narrow their half-bandwidths. Orienta on tuning in the two cell types mostly shi s away from the adapter with brief or prolonged s mula on.
•
Relative peak position pre-adaptation(deg)
-0.6
0
)
Extracellular wave shape analysis
180
20
0
0 12
eg
360
90
0
0
20
attractive 45
(d
0
0
0
0
ift
360
-1
40
0.6
0
0
S
1
this unit
20
-3
360
•
40
15
Circular Variance
Half-Bandwidth
20
0
0
30
Amplitude
40
Sh
30
F.
16
40 broadening
ak
35
E.
600 >800
40
Pe
D.
0
400
0
0
Time (µs)
200
Trough-peak interval (µs)
45
45 repulsive
30
0
50 C.
60 B.
0
800
45
35
Spike Rate (s -1)
A.
−200 0
20
35
10
-0.6
adapted
0
Based on online analysis of tuning measured at six tetrodes, the adap ng orienta ons are chosen. In this example, neural ac vity at three of six tetrodes have peaks within 22.5 deg of a common direc on. Orienta ons of 90 and 112.5 deg (indicated by ver cal bars) are therefore selected for the adapta on experiments. S muli are presented at the spa al frequency that is op mal for at least one of the six tetrodes.
270
0
deg
Selection of Adaptation Parameters
180
Spike Rate (S-1)
For each recording site, two adapta on experiments are performed, one at 0.4 sec and another at 40 sec dura on. These mes are chosen to demonstrate adapta on-induced effects in V1 (Pa erson et al, 2013).
90
0
0
0
Normalized Voltage
0
deg
Time (sec)
S
-1
0
Half-Bandwidth post-adaptation (deg)
Normalized Voltage
deg deg deg
Normalized Voltage
an 180.0
deg
1.0
Spike Rate (S-1)
5.0
Relative peak position post-adaptation(deg)
deg
Spike Rate (S-1) Spike Rate (S-1) Spike Rate (S-1) Spike Rate (S-1)
k
t
Bl
-u To p
Te s
p
t
p To p
-u
t
Te s
t Ad
an
ap
k
t
Bl
Bl
Te s
k an
t
Te s
t
Te s 1.0
0
W
5.0
HB
Time (sec)
1.0
in
40.0
20
25
20
1.0
0
45
0
1.0
1
50
20
1.0
10
90
0
0.4
0.6
20
Change in Half-Bandwidth
Peak Shifts
0
14 0
0
0.4
40
12
this unit
40
6
-3
1.2
...
Circular Variance
ge
Trough-peak interval (µs)
0
an
600 >800
Number of cells
Ch
400
15
14
200
0
0
0
800
0
15
OFF AXIS
14 -20
−200 0
W
-0.6
HB
20
40
14
0
0
20
0
Half-Bandwidth pre-adaptation (deg)
in
0.6
adapted
Half-Bandwidth
10
8
narrowing
ge
this unit
40
Time (µs)
Amplitude
Number of cells
20
0
6
-2
270
0
6
180
0
0
0
0 8
Half-Bandwidth pre-adaptation (deg)
)
90
0
11
30
S-1 0
Half-Bandwidth
20
40
20
an
Amplitude
0
narrowing 40
40 broadening
Ch
mean and 95% CI Circular Variance
20
40
7
baseline adapted
10
Number of cells
12
0
0.4
Relative peak position pre-adaptation(deg)
0
20
0
adapted
20
45
eg
ap
Trough-peak interval (µs)
Time (µs)
12
0.4
0
600 >800
(d
Prolonged Adaptation
Ad
400
ift
k
200
Sh
an
0
800
ak
Bl
−200 0
attractive
Pe
t
-0.6
8
Te s
270
0
0
20
0
0
k
0
40
0.6
10 30
an
S-1
1
V1 Granular Layer
Adaptation Paradigms
Bl
20
40
W HB
180
10
40 broadening
in
90
0
Half-Bandwidth
this unit
45
45
e
adapted
repulsive
ng
600 >800
40
a Ch
400
Trough-peak interval (µs)
Circular Variance
12
200
20
20
20
0
800
0
0
0
−200 0
45
0
-0.6
Time (µs)
Amplitude
0
0
20
8
10 -20
0
0
10
20
0
20 0
0
0
270
-1
0.6
0 -2
180
1
40
0 -3
90
0
40
this unit
8
10
S
Circular Variance
Half-Bandwidth
0
Amplitude
0
0
W HB
Trough-peak interval (µs)
Time (µs)
in
600 >800
e
400
12
ng
200
a Ch
270
0
800
6
−200 0
0
-0.6
12
6
g)
0
20
e (d
0
0
ift
0
40
0.6
Sh
12
12
Brief Adaptation
S-1
1
adapted
12
6-tetrode array, each independently movable Spike sor ng (KlustaKwik and Klusters) Lesions and histology post experiment
30
ak
180
20
this unit
Pe
90
0
Circular Variance
Half-Bandwidth
0
0
mean and 95% CI
12
Amplitude
0
Macaque V1 and V2 Anesthesia: sufentanil and propofol Neuromuscular blockade: rocuronium
We performed mul -tetrode single-unit recordings to measure neural responses to dri ing sinusoidal gra ngs before and a er adapta on to preferred and non-preferred orienta ons.
22
Bandwidth Changes
Peak Shifts
0 20 3
Physiological Methods and Recordings
baseline adapted
Normalized Voltage
adapted
Post-Synaptic
V1 Supragranular Layer
Sensory adapta on u lizes recent s mulus history to modify neuronal response proper es to adjust to the ever-changing features of the visual world. To delineate the influence of adapta on on neural and network proper es, we inves gated its effect on orienta on tuning in func onally dis nct cell categories (excitatory vs. inhibitory) in V1 and V2 of the macaque.
Methods
Population Responses
Modeled Adaptation Effects
Untuned Inhibition
Weill Cornell Medical College, New York, NY 10065
Single-unit Responses
Introduction and Motivation
ON AXIS
Unadapted tuning
contact:
[email protected]
33.4100 VSS 2016
Predicted Adaptation Effects
Model
Half-Bandwidth post-adaptation (deg)
Effects of adaptation on orientation tuning in excitatory and inhibitory neurons in macaque V1 and V2
In V2, adaptation mostly induces narrowing of orientation tuning in excitatory cells, but can increase or decrease the bandwidth of inhibitory cells. Both repulsive and attractive shifts in tuning are observed in V2. Both pre-synaptic adaptation and tuned inhibition are required to account for both repulsive and attractive shifts in tuning, as observed in the data.
• •
Dis nguishing between Models
• •
Feedforward architecture 80% excitatory neurons 20% inhibitory neurons Input neuron orienta on tuning modeled with von Mises func ons Tuning func on parameters drawn from our neural data Output neuron ac vity modeled as Poisson spike trains
Tuned vs. untuned inhibi on Post-synap c vs. pre-synap c adapta on
Adapta on Rules
• •
Post-synap c adapta on influences connec on strengths of all inputs by a common factor, determined by post-synap c neuron Pre-synap c adapta on influences connec on strengths propor onally to the pre-synap c unadapted response
percen les th
Symbol Mean 25
Excitatory Input Neurons Half-Bandwidth (deg) n/a Concentra on β j Amplitude (s-1) M j Baseline Firing Rate (s-1) B j Inhibitory Input Neurons Half-Bandwidth (deg) n/a Concentra on β j -1 Amplitude (s ) M j Baseline Firing Rate (s-1) B j
50th 75th
26.3 20.5 24.5 50.2 1.9 2.6 2.4 0.5 11.5
6.2
8.2
29.2
9.5
3.2
9.5
12.2
Rinput (
stim , j ) = B j + M j e
1 (µ PE ( j , k ) = e j K 1 PI ( j , k ) = (e tuned K
µ k )2 / 2
j
cos(2(
stim
µ j )) 1)
excitatory
2
/ 2)) 2 / 2
( µ j (µk
2
+e
( µ j ( µ k + / 2)) 2 / 2
2
)
untuned inhibitory
PIuntuned ( j , k ) = constant
32.3 18.6 35.5 60.2 1.1 2.8 0.8 0.2 12.2
8.4
14.2 35.8
12.4
7.6
8.5
Range Explored
Typical Min
Max
Network parameters Average number of inputs 1000 100 1000 Frac on of inhibitory inputs 0.2 0.05 0.5 5 60 Connec vity Rule width (deg)( σ) 20 Adapta on strength (α) 0.5 0.1 0.75
12.5
f(
adapt
, j) = 1
(
R(
tuned inhibitory
adapt
, j ) Rmin ( j )
Rmax ( j ) Rmin ( j ) Rmin ( j ) = min ( R( stim , j )) Rmax ( j ) = max ( R (
stim
)
, j ))
Output Neuron Routput (
stim
, k) = j input
C ( j, k ) A( j, k ) Rinput (
stim
, j)
C ( j , k ) = 1 with probability P ( j , k ) and 0 otherwise A( j , k ) = 1 before adaptation A( j , k ) = f ( adapt , j ) for pre-synaptic adaptation A( j , k ) = f (
adapt
, k ) for pre-synaptic adaptation
References
1. Abbo , L.F., Varela, J.A., Sen, K., & Nelson, S.B. (1997). Synap c depression and cor cal gain control. Science, 275(5297), 221-224. 2. Dragoi, V., & Sur, M. (2004). Plas city of orienta on processing in adult visual cortex. Visual Neurosciences, 1654-1664. 3. Kohn, A. (2007). Visual adapta on: physiology, mechanisms, and func onal benefits. Journal of Neurophysiology, 97(5), 3155-3164. 4. Mitchell, J.F., Sundberg, K.A., & Reynolds, J.H. (2007). Differen al a en on-dependent response modula on across cell classes in macaque visual area V4. Neuron, 55(1), 131-141. 5. Pa erson, C.A., Wissig, S.C., & Kohn, A. (2013). Dis nct effects of brief and prolonged adapta on on orienta on tuning in primary visual cortex. The Journal of Neuroscience, 33(2), 532-543. 6. Shapley, R., Hawken, M., & Ringach, D. (2003). Dynamics of orienta on selec vity in the primary visual cortex and the importance of cor cal inhibi on. Neuron, 38(5), 689-699.
Support: EY9314 The Fred Plum Fellowship
We thank Mary M. Conte for assistance with poster design and creation.
