ArticlePDF Available

Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex

Authors:

Abstract and Figures

The potential for neuronal representations of external stimuli to be modified by previous experience is critical for efficient sensory processing and improved behavioral outcomes. To investigate how repeated exposure to a visual stimulus affects its representation in mouse primary visual cortex (V1), we performed two-photon calcium imaging of layer 2/3 neurons and assessed responses before, during, and after the presentation of a repetitive stimulus over 5 consecutive days. We found a stimulus-specific enhancement of the neuronal representation of the repetitively presented stimulus when it was associated with a reward. This was observed both after mice actively learned a rewarded task and when the reward was randomly received. Stimulus-specific enhanced representation resulted both from neurons gaining selectivity and from increased response reliability in previously selective neurons. In the absence of reward, there was either no change in stimulus representation or a decreased representation when the stimulus was viewed at a fixed temporal frequency. Pairing a second stimulus with a reward led to a similar enhanced representation and increased discriminability between the equally rewarded stimuli. Single-neuron responses showed that separate subpopulations discriminated between the two rewarded stimuli depending on whether the stimuli were displayed in a virtual environment or viewed on a single screen. We suggest that reward-associated responses enable the generalization of enhanced stimulus representation across these V1 subpopulations. We propose that this dynamic regulation of visual processing based on the behavioral relevance of sensory input ultimately enhances and stabilizes the representation of task-relevant features while suppressing responses to non-relevant stimuli.
Enhanced Representations of Two Repetitively Presented and Equally Rewarded Stimuli (A) Experimental timeline. Orientation selectivity was assessed on the day before (pre) phase 1 (P1) training with a single virtual corridor (vertical gratings) followed by phase 2 (P2), where two virtual corridors were repetitively presented with a water reward, and a final assessment of orientation selectivity (post). (B) Responses (DF/F 0 ) of 3 example neurons; individual trials (gray) and average response (black). Neuron 1 and 2: corridor-selective responses are shown. Neuron 3: corridor responsive neuron is shown, not selective for a single corridor grating but responsive to both corridors. (C) Correlation between the percentage of corridor-responsive neurons and the behavioral performance (spatial modulation index [SMI]; R = 0.82; p < 0.001; n = 15 [novice, mid-training, and expert sessions from 5 mice]; Pearson's coefficient). (D) Proportion of corridor-specific (green) and corridor-responsive (gray) neurons on the first day of phase 2 (novice), the midpoint training day (i.e., after first half of training days; training), and the last day of training (expert). Increase in the percentage of corridor-selective neurons from novice to expert day is shown (p = 0.047; n = 5; Wilcoxon signed rank). (E) Accuracy of a template-matching decoder to determine which corridor grating (vertical or angled) was presented per trial for the novice and expert day (p = 0.031; n = 5; Wilcoxon signed rank). Dashed line represents chance level. (F) Change in the proportion of orientation selective neurons from pre to post day for both vertical gratings (p = 0.905; Mann-Whitney U test) and angled gratings (p = 0.032; Mann-Whitney U test) for animals that only had phase 1 (P1; 4 mice) and animals that had both phase 1 and phase 2 (P2; 5 mice). Repetitively presented stimuli are indicated (R). All panels, error bars: SEM.
… 
Different Subpopulations of Grating-Selective Neurons across Viewing Contexts (A) Percentage of corridor-selective neurons in the virtual environment (green; for novice [top] or expert [bottom]) and orientation-selective neurons during single-screen viewing (gray; selective for any orientation [either 0 , 45 , 90 , or 135 ] on pre [top] or post [bottom] testing day); percentage of overlapping indicated in center. (B) Among the population of corridor-selective neurons on expert day: the proportion that is orientation selective for the R gratings (green; either vertical or angled) or for the gratings orthogonal to the R gratings (black) is shown on the left and, on the right, the magnitude of their selectivity for both pre and post testing days. *p < 0.05; n = 5 mice; MannWhitney U test. (C) Stimulus discriminability (d') between vertical and angled gratings in the VR on expert day and between the vertical and angled gratings under single-screen viewing conditions on post testing day for each neuron (n = 510 neurons from 5 mice). (D) Change in d' from novice to expert day between the two R gratings in the virtual environment (vertical and angled) for the corridor-selective population (green; n = 158) versus all other neurons (n = 352) and change in d' from the pre to post testing day during single-screen viewing for the population of neurons that are orientation selective for the R grating (R select ; black; n = 166) versus all other neurons (n = 344). ***p < 0.001; Mann-Whitney U test. All panels, error bars: SEM.
… 
Content may be subject to copyright.
Article
Reward Association Enhances Stimulus-Specific
Representations in Primary Visual Cortex
Highlights
dReward association drives the enhanced representation of
repetitive visual stimuli
dIncreased discriminability between two equally rewarded
stimuli with learning
dEnhanced responses to rewarded stimuli generalize across
subpopulations and contexts
dDynamic regulation of V1 responses based on behavioral
relevance of visual input
Authors
Julia U. Henschke, Evelyn Dylda,
Danai Katsanevaki, ...,
Theoklitos Amvrosiadis,
Janelle M.P. Pakan,
Nathalie L. Rochefort
Correspondence
janelle.pakan@med.ovgu.de (J.M.P.P.),
n.rochefort@ed.ac.uk (N.L.R.)
In Brief
Henschke et al. show that reward
association enhances the representation
of repetitive visual stimuli in mouse
primary visual cortex. Two equally
rewarded stimuli led to similar enhanced
representation and increased
discriminability with learning. Responses
to rewarded stimuli generalize across V1
subpopulations and viewing contexts.
Henschke et al., 2020, Current Biology 30, 1–15
May 18, 2020 ª2020 The Author(s). Published by Elsevier Inc.
https://doi.org/10.1016/j.cub.2020.03.018
Current Biology
Article
Reward Association Enhances Stimulus-Specific
Representations in Primary Visual Cortex
Julia U. Henschke,
1,2,5
Evelyn Dylda,
3,5
Danai Katsanevaki,
3,5
Nathalie Dupuy,
3
Stephen P. Currie,
3
Theoklitos Amvrosiadis,
3
Janelle M.P. Pakan,
1,2,5,6,
*and Nathalie L. Rochefort
3,4,5,
*
1
Center for Behavioral Brain Sciences, Institute of Cognitive Neurology and Dementia Research, Otto-von-Guericke University Magdeburg,
Leipziger Str. 44, Magdeburg 39120, Germany
2
German Center for Neurodegenerative Diseases, Leipziger Str. 44, Magdeburg 39120, Germany
3
Centre for Discovery Brain Sciences, Edinburgh Medical School: Biomedical Sciences, University of Edinburgh, 15 George Square,
Edinburgh, EH8 9XD, UK
4
Simons Initiative for the Developing Brain, University of Edinburgh, 15 George Square, Edinburgh EH8 9XD, UK
5
These authors contributed equally
6
Lead Contact
*Correspondence: janelle.pakan@med.ovgu.de (J.M.P.P.), n.rochefort@ed.ac.uk (N.L.R.)
https://doi.org/10.1016/j.cub.2020.03.018
SUMMARY
The potential for neuronal representations of external
stimuli to be modified by previous experience is crit-
ical for efficient sensory processing and improved
behavioral outcomes. To investigate how repeated
exposure to a visual stimulus affects its representa-
tion in mouse primary visual cortex (V1), we per-
formed two-photon calcium imaging of layer 2/3 neu-
rons and assessed responses before, during, and
after the presentation of a repetitive stimulus over 5
consecutive days. We found a stimulus-specific
enhancement of the neuronal representation of the
repetitively presented stimulus when it was associ-
ated with a reward. This was observed both after
mice actively learned a rewarded task and when the
reward was randomly received. Stimulus-specific
enhanced representation resulted both from neurons
gaining selectivity and from increased response reli-
ability in previously selective neurons. In the absence
of reward, there was either no change in stimulus rep-
resentation or a decreased representation when the
stimulus was viewed at a fixed temporal frequency.
Pairing a second stimulus with a reward led to a
similar enhanced representation and increased dis-
criminability between the equally rewarded stimuli.
Single-neuron responses showed that separate sub-
populations discriminated between the two re-
warded stimuli depending on whether the stimuli
were displayed in a virtual environment or viewed
on a single screen. We suggest that reward-associ-
ated responses enable the generalization of
enhanced stimulus representation across these V1
subpopulations. We propose that this dynamic regu-
lation of visual processing based on the behavioral
relevance of sensory input ultimately enhances and
stabilizes the representation of task-relevant features
while suppressing responses to non-relevant stimuli.
INTRODUCTION
Adaptation to the environment is vital for survival and relies on
our ability to selectively integrate relevant sensory information
and ignore irrelevant distractors. This ability depends on the po-
tential for neuronal networks to change through experience, for
example, by learning the association of a specific sensory stim-
ulus with a reward. Neuronal representations of visual stimuli in
adult primary visual cortex (V1) have been shown to be modified
by previous visual experience [1–17]. However, previous studies
in mice have reported inconsistent results regarding the effect of
a repetitively viewed stimulus under various behavioral condi-
tions. Several studies have reported a stimulus-specific
response potentiation to the daily presentation of a given stim-
ulus [7–10] or stimulus sequence [6] without any associated
reward or aversive stimuli. Conversely, other studies have shown
a stimulus-specific adaptation resulting in response suppression
for repetitive stimuli [5]. In addition, studies that involved active
learning tasks, where a specific stimulus was repetitively paired
with an associated behavioral outcome, have shown either an in-
crease in stimulus discriminability of behaviorally relevant stimuli
[3,4,12,16] or a stimulus-specific decrease in the number of
visually responsive neurons and decreased stimulus selectivity
[5]. Therefore, the extent to which the presentation of a visual
stimulus can, by itself, lead to plasticity in V1 and alter subse-
quent neuronal responses to that stimulus remains unclear.
Part of this inconsistency may stem from the behavioral rele-
vance of the stimulus and/or the behavioral state of the animal
when viewing a visual stimulus. A number of factors can
contribute to the salience of a visual stimulus and therefore its
behavioral relevance [13], for example, whether a given stimulus
is associated with a reward [17–19] or an aversive event [5] and
whether this association was learned during a behavioral task
(e.g., goal-directed behavior) [3,4,20]. Additionally, the behav-
ioral state of an animal, such as whether the animal is stationary
or running, modulates the activity of individual neurons in V1 [21–
29] and affects the representation of visual stimuli [30,31] (but
see [32]). While moving through an environment, congruence be-
tween an animal’s self-motion and optic-flow information results
in coupled visuomotor feedback; recent research has revealed
that a subpopulation of neurons in V1 responds to a mismatch
Current Biology 30, 1–15, May 18, 2020 ª2020 The Author(s). Published by Elsevier Inc. 1
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
Figure 1. Enhanced or Attenuated Representation of a Repetitive Stimulus in V1 Layer 2/3 Neurons
(A) Experimental timeline. Layer 2/3 neurons in V1 were imaged in head-fixed mice able to freely run on a cylindrical treadmill, and responses to 4 oriented gratings
were assessed before (pre) and after (post) an intervening 5 consecutive days of repetitive stimulus presentation of one oriented grating.
(legend continued on next page)
2Current Biology 30, 1–15, May 18, 2020
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
between optic-flow and self-motion signals when this visuomo-
tor feedback is uncoupled [33–36]. In order to detect salient stim-
uli and ignore irrelevant input, one may predict that neural cir-
cuits in the visual cortex could selectively reduce responses to
distractors, increase their responses to visual stimuli that are
relevant for a behavioral task, and remain unaffected by the nat-
ural optic flow associated with self-motion. However, how infor-
mation regarding the behavioral relevance of a specific stimulus
is encoded in the visual cortex and modified by experience re-
mains uncertain.
In this study, we performed two-photon calcium imaging of V1
layer 2/3 neurons in awake head-fixed mice to assess neuronal
responses before, during, and after the presentation of a single
repetitive visual stimulus for 5 consecutive days. We systemati-
cally assessed how stimulus-reward association and visuomotor
coupling affected the representation of the repetitive stimulus.
We found a stimulus-specific enhancement of the neuronal rep-
resentation of the repetitive stimulus when it was associated with
a reward. When a second stimulus was subsequently associated
with the same reward, we found a similar stimulus-specific in-
crease, indicating that layer 2/3 neuronal populations can simul-
taneously maintain multiple enhanced representations of re-
warded stimuli. In the absence of reward, there was either no
change or a decrease in repeated stimulus representation over
days. Single-cell responses showed that, although one V1 sub-
population selectively discriminated between rewarded stimuli
in the virtual environment, a largely non-overlapping subpopula-
tion enhanced their selectivity and discriminability for the re-
warded orientations during single-screen viewing. We propose
permissive mechanisms for reward responses to drive general-
ization across V1 populations and enhance stimulus representa-
tion. Altogether, these results support the view of a dynamic
regulation of visual information processing in V1 based on the
behavioral relevance of the visual input.
RESULTS
We characterized the visual responses of layer 2/3 neurons ex-
pressing the genetically encoded calcium indicator GCaMP6
[37] in V1 by using two-photon calcium imaging in head-fixed
mice that were freely running on a cylindrical treadmill (Figure 1).
We first established the baseline level of orientation selectivity in
individual neurons on an initial testing day (referred to
as ‘‘pre’’) by presenting a series of full-field oriented gratings
(four orientations; single-screen, contralateral to imaged V1). An-
imals were then presented with only a single oriented grating for
5 consecutive days (repetitive stimulus presentation; 15-min
session per day), followed by a second assessment of visual re-
sponses to all oriented gratings at the end of the experiments
(referred to as ‘‘post’’; Figure 1A). In a control group, no grating
was presented during the 5 consecutive days between pre and
post imaging (no stimulus group; Figure 1B).
During the presentation of the repetitive stimulus over 5 days,
animals were divided into experimental groups in order to sys-
tematically assess the impact of stimulus-reward association
and visuomotor coupling under conditions ranging from a non-
rewarded, uncoupled stimulus to a stimulus-reward association
learned in a virtual reality (VR) environment (Figure 1B). The first,
passive viewing, group viewed the single oriented grating in the
same experimental conditions as in the pre/post testing days
(single screen, same spatial frequency, and moving at a constant
temporal frequency), and no reward was given. Therefore, this
condition is representative of traditional laboratory settings
used to probe visual responses of neurons in V1 with oriented
gratings on a single screen (Figure 1B). The second, no reward
VR, group represents a more naturalistic condition in a virtual
environment, consisting of a linear corridor with the repetitive
grating presented on the corridor walls and the optic flow of
the stimulus directly linked to the animals’ movements on the
treadmill, creating coupled visuomotor feedback. Animals
passively viewed the repetitive stimulus over consecutive days
with no reward and no requirement to be directly behaviorally
engaged. The third, goal-directed VR, group viewed the repeti-
tive stimulus while performing a rewarded learning task that
required direct behavioral engagement. This later group also
had coupled visuomotor feedback in the virtual environment
and, additionally, a reward associated to this visual context (Fig-
ure 1B). Mice were actively engaged in the task and learned to
associate a water reward with the visual context of the virtual
corridor (Figure S1). For each animal, the same population of
layer 2/3 V1 neurons was imaged across all experimental days
(pre, 5 days of repetitive stimulus presentation, and post; see
(B) During the repetitive visual stimulus over the 5 days, animals were divided into experimental groups: no stimulus presentation; passive viewing condition wit h
one grating at a constant temporal frequency (schematic of setup with single screen); and same grating in a virtual reality (VR) environment where visuomotor
feedback is coupled and with either no reward (no reward VR) or a learning task with water reward (goal-directed VR).
(C) Example two-photon imaging field of view (left) with arrows indicating two example neurons that displayed orientation-selective response for either the
stimulus that was repetitively presented (R) (middle, orange) or for the grating orthogonal to the repetitive grating (O
R
) (right, gray). Responses (DF/F
0
) of the
example neurons to the 4 oriented gratings for each trial are shown in gray with average response across trials in black (A, angled; O
A
orthogonal to A). Tuning
curves show the peak response at the preferred orientation. Scale bar: 50 mm.
(D) Change in the proportion of selective neurons from pre to post day for the R and O
R
gratings (no stimulus: p = 0.918; passive viewing: p = 0.030; no reward VR:
p = 0.199; goal-directed VR: p < 0.001; Student’s t test).
(E) Relative change from pre to post day between the R and O
R
gratings (RO
index
; left; p = 0.001; one-way ANOVA with LSD test) and oth er control orientations (O
A
and O
R
;O
index
; right; p = 0.633; one-way ANOVA).
(F) Shift in the orientation selectivity at the population level from pre to post day. Difference in the average maximal response vector across all neurons for each
animal is shown as the mean across animals for each group. Bars above the red dashed line indicate a shift toward the R grating, and bars below the dashed line
indicate a shift away from the R grating (effect shown in schematic tuning curves on left; p < 0.001; one-way ANOVA with LSD test).
(G) Average change in the decoding accuracy across animals from pre to post day for each orientation (left) using a Bayesian maximum-likelihood decoder to
determine individual neuronal response to each oriented grating, averaged across the population for each animal. Average change in decoding accuracy for the R
grating across animals for each group is shown (right; p < 0.001; one-way ANOVA with LSD test).
All panels: *p < 0.05, **p < 0.01, ***p < 0.001; n = 9 mice for each group; error bars: SEM.
See also Figures S1 and S2 and Table S2.
Current Biology 30, 1–15, May 18, 2020 3
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
Table S1) and neuronal responses were quantified by the change
in mean fluorescence of GCaMP6 (DF/F
0
; for example, see
Figure 1C).
Enhanced or Attenuated Stimulus-Specific
Representation in V1 Layer 2/3 Neurons
To assess the effect of the repetitively presented stimulus, we
first compared visual responses before (pre) and after (post)
the 5 consecutive days of repetitive stimulus presentation. We
quantified the proportion of neurons that were orientation selec-
tive for the repetitively presented grating (R
select
) compared to
neurons that were selective for the grating that was orthogonally
oriented to the repetitive grating (O
select
;Figure 1C; see STAR
Methods). The orthogonal grating was used as an internal control
because the animals only saw this oriented stimulus on the pre
and post testing days, but not during the 5 days in between.
Because animals ran the same amount of time for all gratings
on the pre and post days, we included all collected data, regard-
less of locomotor activity (see also Figure S3).
For the passive viewing group, the proportion of R
select
neu-
rons significantly decreased on the post day compared to pre
day (p = 0.004; n = 9; Wilcoxon signed rank), although there
was no change in this proportion for the no reward VR group
(p = 0.301; n = 9; Wilcoxon signed rank) or the group that was
not exposed to the repetitive stimulus (no stimulus group; p =
0.641; n = 9; Wilcoxon signed rank; Figure S2). In contrast, we
found that, by the post testing day, the proportion of R
select
neu-
rons increased significantly in the goal-directed VR group (p =
0.004; n = 9; Wilcoxon signed rank; Figure S2). These changes
were specific to the repetitive grating, because we found no sig-
nificant change in the proportion of O
select
neurons between pre
and post days for any group (no stimulus, p = 0.426; passive
viewing, p = 0.844; no reward VR, p = 0.742; goal-directed VR,
p = 0.078; n = 9 for each; Wilcoxon signed rank; Figure S2).
This resulted in a statistically significant interaction between
the orientation and experimental group (p < 0.001; two-way anal-
ysis of variance [ANOVA]) and a significant change in the propor-
tion of R
select
compared to O
select
neurons in the goal-directed
VR group (Figure 1D; p < 0.001; n = 9; Student’s t test), as well
as the passive viewing group (Figure 1D; p = 0.030; n = 9; Stu-
dent’s t test). The stimulus specificity was confirmed by quanti-
fying the relative change of R
select
compared to O
select
neurons
(RO
index
=R
select
O
select
/R
select
+O
select
) from pre to post day
(Figure 1E). The change in RO
index
was significantly higher for
the goal-directed VR group compared to all other groups (Fig-
ure 1E; p = 0.001; one-way ANOVA). There was no significant dif-
ference across groups when we calculated a similar index be-
tween two oriented gratings that mice were only exposed to on
pre and post days (Figure 1E; O
index
; p = 0.633; one-way
ANOVA). Finally, for the goal-directed task in the VR environ-
ment, we found a significant correlation between the proportion
of R
select
neurons and behavioral performance (Figure S1C; R =
0.627; p = 0.005; Pearson’s coefficient).
We then examined how these changes in the proportion of se-
lective neurons affected information encoding within the V1
neuronal population, i.e., the ability to decode, at the popula-
tion-level, stimulus-specific information based on individual
neuronal activity [1]. First, we determined the peak angle of the
tuning curve (preferred orientation response vector; see STAR
Methods) across neurons for each animal and examined the shift
in this vector from pre to post testing days across groups.
Although the no reward VR and the no stimulus groups showed
no significant net change in the population orientation vector
(0.32± 3.37and 0.42± 2.59, respectively), we found a gen-
eral shift away from the repetitive stimulus for the passive
viewing group (5.64± 1.90). In contrast, the goal-directed
VR group showed a shift toward the repetitive orientation
(13.81± 2.93), which was significant when compared to all
other experimental groups (Figure 1F; p < 0.001; one-way
ANOVA). In addition, using a Bayesian maximum-likelihood
decoder (see STAR Methods)[1], we found that the average de-
coding accuracy for the repetitive grating in the goal-directed VR
group increased from pre to post testing days (10% ± 3%), lead-
ing to a significant difference in decoding accuracy in compari-
son to all other groups (Figure 1G; p < 0.001; one-way ANOVA)
but no significant difference in decoding accuracy for the orthog-
onal grating (p = 0.823; one-way ANOVA).
Therefore, under passive viewing conditions, exposure to
a repetitive stimulus at a fixed temporal frequency decreased
the proportion of neurons that showed selective responses
for that stimulus. Exposure to the same repetitive stimulus in
a VR environment (no reward VR) resulted in no significant
change although the goal-directed task in a virtual environment
resulted in enhanced representation of the repetitively presented
stimulus.
Stimulus-Specific Enhanced Representation following
Reward Association
To determine the specific impact of reward and visuomotor
coupling on the experience-dependent enhanced representa-
tion observed during the goal-directed VR task, we first de-
coupled the stimulus presentation from the motor activity of
the mouse (i.e., a visuomotor uncoupled VR playback condition;
see STAR Methods); however, the behavioural task remained the
same: to lick at the reward zone in order to receive the reward
(uncoupled-rewarded group; Figure 2A). One consequence of
decoupling the visuomotor feedback in this group was that,
although animals still performed the task (D5 success rate =
74% ± 11%), their licking was not confined to the immediate
reward zone region; hence, the spatial modulation index (SMI)
(see STAR Methods) did not increase from day 1 to day 5 (p =
0.652; n = 9; Wilcoxon signed rank). Under these conditions,
there was a significant increase in the proportion of R
select
compared to O
select
neurons from pre to post testing days (Fig-
ure 2B; p = 0.013; n = 9; Student’s t test) and difference in the
proportion of R
select
neurons across groups, such that there
was a significant increase compared to the no stimulus control
but no difference compared to the previous goal-directed VR
group with coupled visuomotor feedback (Figure 2B; p =
0.023; one-way ANOVA with least significant difference [LSD]
test; p = 0.025 uncoupled-rewarded versus no stimulus;p=
0.440 uncoupled-rewarded versus goal-directed VR;Table S2).
Likewise, there was a significant increase in both the RO
index
and shift of the response vector toward the repetitive grating
when compared to the no stimulus control but no significant dif-
ference between this group and the goal-directed VR group
(RO
index
:Figure 2C; p = 0.032, one-way ANOVA with LSD test;
p = 0.019 uncoupled-rewarded versus no stimulus; p = 0.748
4Current Biology 30, 1–15, May 18, 2020
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
Figure 2. Stimulus-Specific Enhanced Representation following Reward Association
(A) Rewarded experimental groups. Over the 5 days of repeated exposure to the oriented grating, mice received a water reward either randomly (random reward)
or during a learning task within the specified reward zone (uncoupled-rewarded and goal-directed VR). The same visual stimulus was displayed in a virtual
environment either uncoupled (random reward and uncoupled-rewarded) or coupled to the mice locomotion (goal-directed VR). The group with no stimulus
presentation over 5 days was used as a control.
(legend continued on next page)
Current Biology 30, 1–15, May 18, 2020 5
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
uncoupled-rewarded versus goal-directed VR; orientation shift:
Figure 2D; p = 0.039, one-way ANOVA with LSD test; p =
0.049 uncoupled-rewarded versus no stimulus; p = 0.350 un-
coupled-rewarded versus goal-directed VR). This increased
response to the repetitive stimulus led to an increase in the
decoding accuracy from pre to post testing days for the un-
coupled-rewarded group (p = 0.048; n = 9; paired t test); how-
ever, the magnitude of this increase was smaller than the goal-
directed VR group but still significant when compared to the
no stimulus control (Figure 2E; p = 0.003, one-way ANOVA
with LSD test; p = 0.021 uncoupled-rewarded versus no stim-
ulus; p = 0.043 uncoupled-rewarded versus goal-directed VR).
To further confirm the importance of the reward separately
from engagement in learning the task, we tested an additional
experimental group, random reward, in which mice did not
have to lick in a designated reward zone but received rewards
randomly anywhere along the VR corridor with the same repeti-
tive stimulus presentation as for the uncoupled-rewarded group.
Here, we also found an increase in the proportion of R
select
compared to O
select
neurons from pre to post testing days (Fig-
ure 2B) and no significant difference between the random reward
and the other rewarded groups for the RO
index
, orientation shift
toward the repetitive grating, and increase in decoder accuracy
for the repetitive grating (Figures 2C–2E). Therefore, the associ-
ation of the repetitive stimulus with a reward resulted in a stim-
ulus-specific enhanced representation in V1, both after animals
had been engaged in learning a rewarded task and when the an-
imals passively received a reward given randomly during stim-
ulus presentation; the rewarded learning task was associated
with the largest enhancement on average.
Finally, we removed the reward entirely from the task so that
mice were exposed to the same repetitive stimulus in the VR
environment with uncoupled visuomotor feedback but received
no reward during the 5 consecutive days (uncoupled-unre-
warded group; Figure 2F). We found no increase in the propor-
tion of R
select
compared to O
select
neurons from pre to post
testing days (Figure 2G) and no difference across other non-re-
warded groups (no stimulus control as well as the no reward
VR group) with regard to the proportion of R
select
neurons (Fig-
ure 2G), RO
index
(Figure 2H), orientation shift (Figure 2I), or
decoding accuracy (Figure 2J). Therefore, in the absence of a
reward-associated stimulus, we found either no change of stim-
ulus representation in groups with a dynamic optic flow (either
visuomotor coupled [no reward VR group] or uncoupled [un-
coupled-rewarded group]) or a decrease in stimulus representa-
tion with a fixed temporal frequency repetitive stimulus (passive
viewing group; see Figure 1).
Impact of Locomotion on Experience-Dependent
Changes of Stimulus Representation in V1
Because locomotion is known to modulate the gain of visual re-
sponses in V1 [21,24,27,33], we assessed whether changes in
the proportion of responsive neurons correlated with running
behavior. For the rewarded groups, mice were motivated by us-
ing water restriction and we found that they spent more time
running (total time running, averaged across all days: rewarded
80% ± 2% versus non-rewarded 34% ± 1%; see Figure S3A).
However, the difference in the amount of running from pre to
post days was highly variable across animals, even within groups
(Figure S3B). We found no significant correlation between the
change in the proportion of R
select
neurons and the proportion
of time spent running either during pre and post days (Fig-
ure S3C; R = 0.206; p = 0.103; Pearson’s coefficient) or across
all experimental days (Figure S3D) within any of the experimental
groups, i.e., for any given experimental condition, the animals
that ran more did not have larger changes in the proportion of
R
select
neurons. Although there was a positive correlation in
this regard across all conditions (Figure S3D; R = 0.445; p <
0.001; n = 63; Pearson’s coefficient), this became non-significant
when we analyzed, separately, the groups that were rewarded
(R = 0.045; p = 0.823; n = 27; Pearson’s coefficient) versus
non-rewarded (including control; R = 0.068; p = 0.695; n = 36;
Pearson’s coefficient), again, suggesting that the presence of
the reward was the driving factor.
Importantly, on the pre and post testing days, there was no
bias for the animals to run more during the repetitive grating
compared to other orientations and no change in the modulation
of responses by locomotion from pre to post days (Figures S3E–
S3G). Together, this suggests that locomotion did not directly
affect the change in visual response properties on the pre and
(B) Change in the proportion of selective neurons from pre to post day compared between the R and O
R
gratings within groups (#p < 0.05; ###p < 0.001; goal-
directed VR: p < 0.001; uncoupled-rewarded: p = 0.013; random reward: p = 0.025; Student’s t test) and the change in proportion of R selective neurons
compared across groups (p = 0.023; one-way ANOVA with LSD test; *p < 0.05; **p < 0.01).
(C) Relative change from pre to post day between the R and O
R
gratings (RO
index
; p = 0.032; one-way ANOVA with LSD test; *p < 0.05).
(D) Shift in the orientation selectivity at the population level from pre to post day. Difference in the average maximal response vector across all neurons for each
animal is shown as the mean across animals for each group (n = 9 mice for each). Bars above the red dashed line indicate a shift toward the R grating, and bars
below the dashed line indicate a shift away from the R grating (p = 0.039; one-way ANOVA with LSD test; *p < 0.05; **p < 0.01).
(E) Change in the decoding accuracy across groups for the R grating from pre to post day using a Bayesian maximum-likelihood decoder to determine individual
neuronal response and averaged across the population for each animal (p = 0.003; one-way ANOVA with LSD test; *p < 0.05; ***p < 0.001).
(F) Non-rewarded experimental groups. Over the 5 days of repeated exposure to the oriented grating, visual stimulus was di splayed in a virtual environment either
uncoupled (uncoupled-unrewarded) or coupled (no reward VR), with all groups receiving no reward. The group with no stimulus presentation over 5 days was
used as a control.
(G) Change in the proportion of selective neurons from pre to post day compared between the R and O
R
gratings within groups was non-significant for all groups
(no stimulus: p = 0.641; uncoupled-unrewarded: p = 0.322; no reward VR: p = 0.301; Student’s t test) and the change in proportion of R selective neurons not
significant across groups (p = 0.310; one-way ANOVA).
(H) Relative change from pre to post day between the R and O
R
gratings (RO
index
) was non-significant across non-rewarded groups (p = 0.895; one-way ANOVA).
(I) Shift in the orientation selectivity at the population level from pre to post day was non-significant across non-rewarded groups (p = 0.921; one-way ANOVA).
(J) Change in the decoding accuracy for the R grating from pre to post day was non-significant across non-rewarded groups (p = 0.123; one-way ANOVA).
All panels: n = 9 mice for each group; error bars: SEM.
See also Figure S3 and Table S2.
6Current Biology 30, 1–15, May 18, 2020
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
Figure 3. Stimulus-Reward Association Increases Selectivity and Decreases Trial-by-Trial Variability
(A) Responses (DF/F
0
) to the four oriented gratings for three example neurons before (pre) and after (post) 5 days of repetitive stimulus presentation are shown for
neurons that either remain selective (left), gain selectivity (middle), or lose selectivity (right) for the stimulus that was repetitively presented (R). Individual trials are
shown in gray and average response across trials in black.
(B) Experimental groups depending on reward association and visuomotor coupling during the repetitive visual stimulus over 5 days. Rewarded and non-re-
warded groups are indicated. Visuomotor feedback was either coupled (C) or uncoupled (UC).
(C) Proportion of the total number of neurons that are selective for the R grating or the total number of neurons that are selective for the O
R
grating on the pre
testing day for the rewarded (black) and non-rewarded (white) groups that then either remain selective (left; R, p = 0.016; O
R
, p = 0.0421) or gain selectivity (middle;
(legend continued on next page)
Current Biology 30, 1–15, May 18, 2020 7
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
post imaging days within each experimental group. However,
because the presence of the reward itself increased engagement
and, subsequently, running time was higher for all rewarded
groups within the 5 days of stimulus exposure, it is possible
that running induced or enhanced plasticity mechanisms associ-
ated with the stimulus-reward association.
Stimulus-Reward Associations Enhance Selectivity and
Response Reliability
We next investigated the single-cell dynamics underlying expe-
rience-dependent changes by following the selective properties
of each individual neuron across pre and post days (Figure 3). We
grouped neurons into four categories: a neuron selective for the
repetitive grating on the initial pre day can either ‘‘remain selec-
tive’’ or ‘‘lose selectivity’’ on the post day; alternatively, neurons
that are not initially selective for the repetitive grating on the pre
day can either ‘‘gain selectivity’’ or remain ‘‘non-selective’’ by the
post day (Figure 3A). The increased proportion of R
select
neurons
in the rewarded groups (Figure 3B) resulted from both more neu-
rons that remained selective (Figure 3C; p = 0.016; n = 27; Stu-
dent’s t test) as well as more initially non-selective neurons
that gained selectivity (Figure 3C; p < 0.001; n = 27; Student’s t
test). For the population of stable R
select
neurons (i.e., remain se-
lective neurons), the selectivity magnitude increased for the
goal-directed VR group in comparison to the no stimulus and
all non-rewarded conditions, and the variability across trials
(coefficient of variation) was significantly lower for the rewarded
groups compared to the no stimulus control (Figure 3D;
Table S2).
Lastly, there was a significant shift in the distribution of the
maximum response vector toward the repetitive stimulus within
the population of neurons that remained non-selective from pre
to post testing days for the rewarded groups and no significant
change for the non-rewarded groups (Figure 3E; rewarded: p <
0.001; non-rewarded: p = 0.914; Kolmogorov-Smirnov test),
similar to that observed at the level of the whole population
(see Figure 2). Therefore, reward association with the repetitive
stimulus resulted in more neurons showing stable, more reliable,
and more selective responses to this stimulus, leading to a glob-
ally enhanced stimulus representation in both selective and non-
selective neurons.
Enhanced Representation of Two Equally Rewarded
Stimuli and Increased Discriminability
Because we found that the representation of a reward-associ-
ated stimulus was enhanced in V1, we next tested how an addi-
tional rewarded stimulus would be represented: whether the
second stimulus would be comparatively enhanced and whether
the discriminability between these two equally rewarded stimuli
would be maintained or increased. To this end, after the first
phase of goal-directed VR learning with one repetitive stimulus,
5 mice were used for a second phase, where they were pre-
sented with an additional grating along the virtual corridor (Fig-
ure 4A). Here, the length of the corridors was different between
the two gratings (vertical and angled) to encourage the separa-
tion of these virtual environments as different contexts, but all
other task parameters remained the same. We found that a pop-
ulation of layer 2/3 neurons responded to the visual stimulus pre-
sentation along the virtual corridors (corridor responsive neu-
rons; Figure 4B): these neurons were defined by having a
significantly larger response across trials when the corridor walls
contain a visual stimulus compared to black corridor walls (p <
0.001; paired t test). We found a positive correlation between
the percentage of corridor responsive neurons and the behav-
ioral performance (SMI; Figure 4C; R = 0.82; p < 0.001; n = 15
[novice, mid-training, and expert sessions from 5 mice]; Pear-
son’s coefficient), indicating that the proportion of corridor
responsive neurons in V1 is related to task engagement.
The corridor responsive V1 population consisted of neurons
that were either responsive to both corridors (e.g., neuron 3 in
Figure 4B) or selectively responsive to only one of the corridor
gratings (corridor-selective neurons; e.g., neurons 1 and 2 in Fig-
ure 4B). Only the proportion of the corridor-selective neurons
significantly increased from the novice to the expert days (Fig-
ure 4D; novice: 13% ± 4%; expert: 29% ± 5%; p = 0.047; n =
5; Wilcoxon signed rank). Because there were significantly
more neurons that responded to only one corridor, decoder ac-
curacy (i.e., population-level stimulus-discriminability using a
template-matching decoder) between these two corridors signif-
icantly increased from the novice to expert day (Figure 4E; pre:
71% ± 4%; post: 84% ± 5%; p = 0.031; n = 5; Wilcoxon signed
rank). Therefore, the increase in responses to rewarded stimuli in
V1 is also context specific and results in, not only a general in-
crease in the proportion of responsive neurons, but also an in-
crease in the level of stimulus discriminability between stimuli
of similar relevance in V1 neuronal population.
Finally, we compared the orientation selectivity of neurons for
mice from phase 1 and phase 2. For animals that received phase
2 training with both vertical and angled repetitive gratings, we
found an increased proportion of neurons responded to the
angled grating on the post testing day, although the mice from
phase 1 showed no change in the representation of angled grat-
ings (Figure 4F; phase 1: 1% ± 2%; phase 2: 8% ± 4%; p =
0.032; Mann-Whitney U test). Interestingly, we found that there
was no additive increase in the response to vertical gratings
from phase 1 to phase 2 (Figure 4F; phase 1: 10% ± 5%; phase
2: 10% ± 4%; p = 0.905; Mann-Whitney U test), indicating that
the experience-dependent changes that occur after 5 days of
R, p < 0.001; O
R
, p = 0.144) by the post testing day. Proportion of the total number of all neurons in the rewarded (black) and non-rewarded (white) groups that then
lose (right; R, p = 0.093; O
R
, p = 0.816) selectivity for the R or the O
R
grating is shown. For all, n = 27 mice; Student’s t test.
(D) Factor change in the magnitude of selectivity (preferre d orientation response vector; left; p = 0.006, one-way ANOVA with LSD test; **p < 0.01; ***p < 0.001) and
in the variability of the selectivity across trials (as measured by the coefficient of variation; right; p = 0.004, one-way ANOVA with LSD test; *p < 0.05; **p < 0.01;
***p < 0.001) on post day compared to pre day (post/pre) for the population of neurons that remain selective for the R grating for each experimental group in (B).
Red dashed line represents no change from pre to post.
(E) Distribution of the preferred orientation (maximum response vector in degrees) for the individual neurons that remain non-selective from pre (black) to post
(gray) testing day for rewarded (***p < 0.001; Kolmogorov-Smirnov test) and non-rewarded (p = 0.914; Kolmogorov-Smirnov test) groups. For each condition,
both distributions are overlaid to show the changes between pre and post days. Dashed line represents the orientation of the R grating.
All panels, error bars: SEM. See also Table S2.
8Current Biology 30, 1–15, May 18, 2020
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
exposure to a repetitive stimulus are maintained, but not ampli-
fied further by additional days of exposure to the same stimulus.
Different Subpopulations of Grating-Selective Neurons
across Viewing Contexts
We then investigated the overlap between V1 neurons that
showed grating-selective properties in different environmental
contexts (i.e., different visual display): either in the virtual envi-
ronment (corridor-selective neurons) or when viewing gratings
of the same orientation with monocular stimulation on a single
screen (orientation selective neurons on pre and post testing
days). Before training, we found that largely distinct populations
were selective under the two conditions, with only a minority of
overlap (Figure 5A; 2% of the total population); after training in
the virtual environment, the proportion of overlap increased but
remained in the minority (Figure 5A; 13% of the total population).
Specifically, among the corridor-selective population on expert
day, we found that an equal proportion of neurons were orienta-
tion selective for the repetitive gratings (i.e., vertical/angled;
17% ± 7%) and orthogonal gratings (18% ± 6%) on the
pre day (p = 0.955; n = 5; Mann-Whitney U test) and that the
magnitude of their selectivity was also equal (vertical/angled:
0.44 ± 0.04; orthogonal: 0.42 ± 0.05; p = 0.328; n = 5; Mann-
Whitney U test; Figure 5B). By the post day, the proportion
of the corridor-selective population that was orientation selec-
tive for vertical/angled gratings (33% ± 7%) increased in com-
parison to the orthogonal gratings (12% ± 4%; p = 0.022;
n = 5; Mann-Whitney U test), as did the selectivity magnitude
Figure 4. Enhanced Representations of Two Repetitively Presented and Equally Rewarded Stimuli
(A) Experimental timeline. Orientation selectivity was assessed on the day before (pre) phase 1 (P1) training with a single virtual corridor (vertical gratings) followed
by phase 2 (P2), where two virtual corridors were repetitively presented with a water reward, and a final assessment of orientation selectivity (post).
(B) Responses (DF/F
0
) of 3 example neurons; individual trials (gray) and average response (black). Neuron 1 and 2: corridor-selective responses are shown.
Neuron 3: corridor responsive neuron is shown, not selective for a single corridor grating but responsive to both corridors.
(C) Correlation between the percentage of corridor-responsive neurons and the behavioral performance (spatial modulation index [SMI]; R = 0.82; p < 0.001; n =
15 [novice, mid-training, and expert sessions from 5 mice]; Pearson’s coefficient).
(D) Proportion of corridor-specific (green) and corridor-responsive (gray) neurons on the first day of phase 2 (novice), the midpoint training day (i.e., after first half
of training days; training), and the last day of training (expert). Increase in the percentage of corridor-selective neurons from novice to expert day is shown (p =
0.047; n = 5; Wilcoxon signed rank).
(E) Accuracy of a template-matching decoder to determine which corridor grating (vertical or angled) was presented per trial for the novice and expert day (p =
0.031; n = 5; Wilcoxon signed rank). Dashed line represents chance level.
(F) Change in the proportion of orientation selective neurons from pre to post day for both vertical gratings (p = 0.905; Mann-Whitney U test) and angled gratings
(p = 0.032; Mann-Whitney U test) for animals that only had phase 1 (P1; 4 mice) and animals that had both phase 1 and phase 2 (P2; 5 mice). Repetitively presented
stimuli are indicated (R). All panels, error bars: SEM.
Current Biology 30, 1–15, May 18, 2020 9
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
(vertical/angled: 0.62 ± 0.04; orthogonal: 0.46 ± 0.03; p = 0.006,
respectively; n = 5; Mann-Whitney U test; Figure 5B). However,
even after learning, only a minority of the population that was se-
lective for a specific corridor in the VR environment was selective
for that same grating during single-screen viewing.
To examine the ability for an individual neuron in V1 to distin-
guish between the two behaviorally relevant gratings (vertical/
angled) under these different viewing contexts, we used a mea-
sure of stimulus discriminability (d’; see STAR Methods) between
vertical and angled gratings. We found that stimulus discrimina-
bility across contexts was not correlated (R = 0.079; p = 0.077;
n = 510 neurons; Pearson’s coefficient): most neurons showed
discriminability in either the VR environment or during single-
screen viewing, but not under both conditions (Figure 5C). How-
ever, the discriminability between the two repetitive gratings
increased with training for both the corridor-selective population
(change in d’ from novice to expert day in the VR environment;
p < 0.001; Mann-Whitney U test) and the R
select
population of
neurons (change in d’ from pre to post testing day during single
screen presentation; p < 0.001; Mann-Whitney U test; Figure 5D).
This suggests that there is a mechanism for the transfer, or
generalization, of stimulus-specific information during training
in the VR environment to the population of R
select
neurons in a
different environmental context on the post testing day.
Generalization of Reward-Associated Responses
We next investigated the potential mechanisms for the general-
ization of stimulus-specific reward association across environ-
ments. We examined reward influences on neuronal responses
in all experimental animals that received rewards. We first
defined reward-responsive neurons during D1–D5 as having
significantly higher activity in a 1-s window following the reward
onset compared to a 1-s window preceding the reward onset
(Figure 6A; see also STAR Methods). We then examined the
properties of orientation-selective neurons that were also reward
responsive during D1–D5 of the repetitive presentation. We
found that the proportion of reward-responsive R
select
neurons
compared to O
select
neurons was higher on the post day (i.e., af-
ter exposure to the reward-associated repetitive stimulus; 80%
versus 20%, respectively; Figure 6A). Additionally, although the
peak response to the reward onset for the R
select
neurons tended
to increase from the pre to post selective population, reward-
responsive post-O
select
neurons had a significantly lower peak
response to the reward onset (Figure 6A; p = 0.023; Student’s t
test). These results suggest that the population of post-R
select
neurons had stronger reward influences during the repetitive
grating sessions on D1–D5 compared to the population of
post-O
select
neurons.
Indeed, when we then examined the response properties of
neurons that were reward-responsive during D1–D5 and either
gained or lost selectivity from the pre to post days, we found
that a higher proportion of neurons that became R
select
had pre-
viously been reward responsive during D1–D5 and that these
neurons that gained selectivity also had significantly higher
peak responses to the reward onset (Figure 6B; p = 0.010; Stu-
dent’s t test). Importantly, the converse was true for O
select
pop-
ulations; neurons that lost their selectivity from the pre to post
day showed significantly higher peak reward responses during
Figure 5. Different Subpopulations of
Grating-Selective Neurons across Viewing
Contexts
(A) Percentage of corridor-selective neurons in the
virtual environment (green; for novice [top] or expert
[bottom]) and orientation-selective neurons during
single-screen viewing (gray; selective for any
orientation [either 0,45
,90
,or135
] on pre [top]
or post [bottom] testing day); percentage of over-
lapping indicated in center.
(B) Among the population of corridor-selective
neurons on expert day: the proportion that is
orientation selective for the R gratings (green; either
vertical or angled) or for the gratings orthogonal to
the R gratings (black) is shown on the left and, on the
right, the magnitude of their selectivity for both pre
and post testing days. *p < 0.05; n = 5 mice; Mann-
Whitney U test.
(C) Stimulus discriminability (d’) between vertical
and angled gratings in the VR on expert day and
between the vertical and angled gratings under
single-screen viewing conditions on post testing
day for each neuron (n = 510 neurons from 5 mice).
(D) Change in d’ from novice to expert day between
the two R gratings in the virtual environment (vertical
and angled) for the corridor-selective population
(green; n = 158) versus all other neurons (n = 352)
and change in d’ from the pre to post testing day
during single-screen viewing for the population of
neurons that are orientation selective for the R
grating (R
select
; black; n = 166) versus all other
neurons (n = 344). ***p < 0.001; Mann-Whitney U
test. All panels, error bars: SEM.
10 Current Biology 30, 1–15, May 18, 2020
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
Figure 6. Generalization of Reward-Associated Activity across V1 Subpopulations and Viewing Contexts
(A) Average activity (DF/F
0
) across trials for all animals that received rewards; activity associated with reward onset (time = 0; ±2-s window) is shown as traces,
averaged over neurons in selected populations: reward-responsive and selective for the R grating (orange) on either pre (n = 108) or post (n = 150) testing days
(top) or reward responsive and selective for the orthogonal to the R grating (orthogonal
R
; black) on either pre (n = 74) or post (n = 37) testing days (bottom).
Proportion of total for these populations is shown as pie charts for pre and post days. Average peak response (max DF/F
0
within window) for these populations is
shown on right (repetitive selective: p = 0.058; orthogonal selective: p = 0.023; Student’s t test).
(B) Average activity (DF/F
0
) across trials for all animals that received rewards; activity associated with reward onset (time = 0; ±2-s window) is shown as traces,
averaged over neurons in selected populations: reward-responsive and gained selectivity (n = 113; solid orange) or lost selectivity (n = 71; open orange) for the R
grating on post testing day (top) or reward responsive and gained selectivity (n = 38; solid black) or lost selectivity (n = 74; open black) for the orthogonal
R
on post
testing day (bottom). Proportion of total for these populations is shown as pie chart. Average peak response (max DF/F
0
within window) for these populations is
shown on right (repetitive: p = 0.010; orthogonal selective: p = 0.001; Student’s t test).
(C) Summary schematic of conditions underlying generalization of reward-associated activity across environmental contexts and days. Populations selective for
the R grating (orange), orthogonal
R
(black), and non-selective (white) are indicated. Stimulus-associated reward influences shift V1 populations responses toward
the R grating on the post testing day.
All panels: *p < 0.05; **p < 0.01; ***p < 0.001; error bars: SEM.
Current Biology 30, 1–15, May 18, 2020 11
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
D1–D5 (Figure 6B; p = 0.001; Student’s t test). Ultimately, strong
reward responses during D1–D5 were associated with a gain in
selectivity for the repetitive stimulus, resulting in more R
select
neurons and a loss of selectivity in O
select
neurons. As a result,
enhanced reward-related responses in the VR context trans-
ferred to stimulus-specific responses in the pre/post context.
This led to an overall enhancement of the repetitive stimulus rep-
resentation with a shift of population responses toward the re-
petitive grating, driven by reward associations (Figure 6C; see
also Figures 1F and 2D).
DISCUSSION
In this study, we demonstrated that the association of a reward
with a repetitive visual stimulus over 5 consecutive days resulted
in an enhanced representation of the reward-associated stim-
ulus in mouse V1 layer 2/3 neurons, while responses to other
stimuli were unaffected. This experience-dependent plasticity
was observed both after animals had been engaged in learning
a rewarded task and when the reward was given randomly dur-
ing the stimulus presentation. By following changes in single-cell
activity over time, we found that both a gain in selectivity for the
rewarded stimulus and an increase in response reliability of pre-
viously responsive neurons underlie the net enhancement of the
repetitive grating representation with reward. When a second
stimulus was paired with a reward, we found a similar enhanced
representation of this stimulus as well as increased discriminabil-
ity between both stimuli of similar reward value. Enhanced re-
sponses to rewarded stimuli generalized across V1 subpopula-
tions and viewing contexts. In the absence of reward, there
was either no change of stimulus representation over days or a
decrease in stimulus representation when the stimulus was
consistently viewed at a fixed temporal frequency. Taken
together, our findings highlight the importance of the behavioral
relevance of sensory stimuli in experience-dependent plasticity
in primary sensory cortices to enhance and stabilize the repre-
sentation of rewarded features while suppressing responses to
non-relevant stimuli.
Previous studies investigating changes in V1 after learned as-
sociations between a visual stimulus and a reward have also
found an enhanced representation of task-relevant stimuli in
layer 2/3 neurons [3,4,12,15,16] and a difference in adaptation
between non-rewarded (passive) and rewarded conditions [12].
Our results extend these conclusions by systematically testing
the impact of visuomotor coupling, stimulus-reward association,
and goal-directed task engagement on this plasticity. We
conclude that the main factor leading to enhanced stimulus-spe-
cific representation in V1 neurons is the association of the repet-
itive stimulus with a reward; this enhancement was observed af-
ter both coupled and uncoupled visuomotor experience and
after both expected (active learning) and unexpected (randomly
given) rewards. Beyond an increase in the proportion of neurons
responsive to the repetitive stimulus, we also found that stim-
ulus-reward associations led to network-wide plasticity: with
shifts in population selectivity and increased information encod-
ing, discriminability, and reliability in responses over time. The
rewarded task was a simple detection task, but our results are
consistent with previous studies using a more challenging
learning discrimination task [3,15,38], which also demonstrated
that enhanced selectivity of task-relevant stimuli was largely due
to the stabilization of existing responses and the recruitment of
previously non-responsive neurons. However, in these studies,
selectivity is generally defined by the difference in the responses
to the rewarded and the non-rewarded stimuli, both of which are
modified by experience. This suggests that changes in re-
sponses to individual gratings differ for discrimination tasks,
where a non-rewarded stimulus can gain relevance by providing
a ‘‘no-go’’ signal, compared to a detection task, where non-re-
warded stimuli act as irrelevant signals.
In addition to the enhanced encoding of a single rewarded
stimulus, we also found similar increased representations of
two equally rewarded stimuli and increased discriminability be-
tween these rewarded stimuli with learning. This demonstrates
that, even if two separate contexts hold the same behavioral
relevance (each led to the same probability of a reward), expe-
rience-dependent changes in V1 promote not only increased
efficiency in stimulus encoding but also discriminability be-
tween equally relevant contexts. This highlights the capacity
of V1 networks to encode both the value (relevance) of a stim-
ulus as well as its specificity compared to equally valuable
stimuli. Furthermore, in line with previous studies, we found
that experience-dependent plasticity occurring within the first
few days of exposure to a repetitive stimulus is maintained,
but not amplified further with additional exposure to the same
stimulus, and that plasticity in response to a new repetitive
stimulus can be induced concurrently [7]. The representation
of the new repetitive stimulus was enhanced to the same
extent as the first rewarded stimulus, highlighting specific
network constraints for this type of experience-dependent
plasticity. Further experiments are required to determine the
capacity of the network to maintain more than two enhanced
representations of rewarded stimuli simultaneously and to
establish the conditions of the potential extinction of these rep-
resentations [39].
In contrast to our results with stimulus-reward associations,
we observed stimulus habituation during passive viewing of a re-
petitive stimulus at a fixed temporal frequency. These findings
potentially contrast with experiments showing stimulus-specific
response potentiation to a repetitive stimulus under similar con-
ditions, using chronic electrophysiological recordings of visually
evoked potentials in layer 4 [7–9] (but see [40]). Our results are in
agreement with a previous two-photon calcium imaging study
that observed a stimulus-specific decrease in the proportion
and selectivity of layer 2/3 neurons after passive viewing of a
grating [5]. Although we found stimulus-specific habituation
when the stimulus was displayed at a fixed temporal frequency
and uncoupled to the animals’ movements, no change was
observed when the same stimulus was not predictable (un-
coupled playback) or when it was coupled to the animal’s loco-
motion, mimicking natural optic flow. This is consistent with the
idea that habituation occurs for stimuli that are learned to be irrel-
evant for the animal’s behavior [11].
Our results revealed that the same oriented gratings elicited
responses in largely separate neuronal populations whether dis-
played in the virtual environment or viewed under visual stimula-
tion conditions with a single screen [41]. Although pairing of the
reward with an oriented grating occurred only when animals
were in the virtual environment, the enhanced representation
12 Current Biology 30, 1–15, May 18, 2020
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
occurred in both subpopulations of neurons, i.e., generalized
across neurons that were not observed to respond selectively
during pairing with the reward. One hypothesis is that stim-
ulus-related reward associations may strengthen existing con-
nections between excitatory neurons of different orientation
preference and shift orientation selectivity within individual
neurons [42]. Additionally, some neurons may display a bias in
their subthreshold responses during pairing with the reward,
which could not be detected with GCaMP6 imaging. Further in-
vestigations are needed to characterize how these changes
generalize to neurons responding to the same orientation in
different contexts and to determine the specific connectivity be-
tween these neuronal ensembles. A recent study using a compu-
tational model of V1 layer 2/3 neuronal circuits has shown that
specialized interneuron circuits could store information about re-
warded stimuli and instruct changes in two excitatory networks
with, for instance, different visual receptive field locations [17].
This transfer of reward-associated enhanced responses across
subpopulations is consistent with the observation of invariance
of learned representations, such that stimulus representations
are generalized across variable viewing contexts and receptive
fields.
As a growing number of studies demonstrate altered levels of
neuronal activity according to behavioral-state-dependent
changes, in subcortical regions, such as the superior colliculus
[43], thalamus [24,44,45], and cerebellum [46–48], as well as
contextual and experience-dependent changes in higher cortical
areas with reciprocally connections to V1 [49], such as the retro-
splenial cortex [50–54], the anterior cingulate cortex [35,55], pa-
rietal cortex [56], and visual association cortex [41,57,58], it
seems likely that these experience-dependent changes in V1
involve modulation of activity in circuits at multiple processing
levels. Additionally, neuromodulatory inputs likely play an impor-
tant role in behavioral-state-dependent plasticity, as cholinergic
and noradrenergic influence have already been shown to modu-
late cortical plasticity, attention, and learning [59,60] in V1, as
well as responses to locomotion and behavioral-state changes
[22,25,61,62]. Indeed, one potential confounding factor in
this study is the time that mice spent running during the presen-
tation of the repetitive grating—because locomotion has been
shown to modulate both visual responses and plasticity of V1
layer 2/3 neurons [10,13,27,31,63,64]. However, locomotion
itself has been shown to increase the gain of visual responses
[21], but not directly affect the orientation tuning of individual
neurons [21,30] or general visual acuity [32]. For instance, previ-
ous work has found that viewing an oriented grating for 60 min
per day while running and without being rewarded led to a shift
in orientation toward a repetitive stimulus and sharpening of
orientation tuning [10], as well as a general increase in stim-
ulus-specific information and response reliability [31]. These
studies suggest that locomotion is required for enhanced stim-
ulus encoding, an idea consistent with literature showing
increased experience-dependent plasticity in V1 with running
[64]. In the current study, we show that, for shorter daily expo-
sure to a stimulus (15 min per day), the absence of a reward pre-
vented the stimulus response enhancement. However, mice in
the unrewarded groups also spent less total time running than
mice in the rewarded groups, but there was no significant corre-
lation between the enhanced stimulus-specific representation
and the proportion of time spent running within any of the exper-
imental groups, i.e., the animals that ran more did not have larger
response changes. Therefore, in the current study, running time
did not appear to be the main factor correlated with the magni-
tude of the observed enhanced stimulus representation in V1.
However, because the presence of the reward itself increased
levels of motivation and engagement, resulting in increasing total
running time, it is possible that running time higher than a certain
threshold induces or enhances plasticity of stimulus-evoked
responses.
Altogether, our results show that repeated exposure to a visual
stimulus leads to enhanced responses when the stimulus was
associated with a reward and to either no change or attenuated
responses when the stimulus was non-rewarded. This mecha-
nism could prove to be highly adaptive for behavior by suppress-
ing responses to irrelevant stimuli that may act as distractors
while optimizing stimulus encoding and information processing
for behaviorally relevant stimuli, such as those associated with
a reward.
STAR+METHODS
Detailed methods are provided in the online version of this paper
and include the following:
dKEY RESOURCES TABLE
dLEAD CONTACT AND MATERIALS AVAILABILITY
dEXPERIMENTAL MODEL AND SUBJECT DETAILS
BSubjects
dMETHOD DETAILS
BSurgery
BTwo-photon imaging
BStimulus presentation: pre and post testing days
BStimulus presentation: repetitive stimulus
dQUANTIFICATION AND STATISTICAL ANALYSIS
BImage analysis
BLicking behavior
BOrientation selectivity
BStimulus decoding and discriminability
BReward-responsive neurons
BAnalysis of locomotion
BStatistics
dDATA AND CODE AVAILABILITY
SUPPLEMENTAL INFORMATION
Supplemental Information can be found online at https://doi.org/10.1016/j.
cub.2020.03.018.
ACKNOWLEDGMENTS
We thank the GENIE Program and the Janelia Research Campus, specifically
V. Jayaraman, R. Kerr, D. Kim, L. Looger, and K. Svoboda, for making
GCaMP6 available. This work was funded by the Wellcome Trust and the Royal
Society (Sir Henry Dale fellowship to N.L.R.); the Marie Curie Actions of the Eu-
ropean Union’s FP7 program (MC-CIG 631770 to N.L.R. and IEF 624461 to
J.M.P.P.); the Shirley Foundation, the Patrick Wild Center, and the RS Mac-
Donald Charitable Trust Seedcorn Grant (to N.L.R.); the Simons Initiative for
the Developing Brain (to N.L.R.); the University of Edinburgh (Graduate School
of Life Sciences to E.D.); funds from the German Center for Neurodegenerative
Diseases (J.M.P.P.); and the European Regional Development Fund (ERDF)
Current Biology 30, 1–15, May 18, 2020 13
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
(Center for Behavioral Brain Sciences; ZS/2016/04/78113 to J.U.H. and
J.M.P.P.).
AUTHOR CONTRIBUTIONS
E.D., S.P.C., J.M.P.P., and N.L.R. designed the experiments. E.D., J.U.H.,
D.K., S.P.C., and T.A. performed the experiments. E.D., J.U.H., N.D., and
J.M.P.P. analyzed the data. J.M.P.P. and N.L.R. wrote the manuscript with
input from all authors.
DECLARATION OF INTERESTS
The authors declare no competing interests.
Received: July 16, 2019
Revised: February 7, 2020
Accepted: March 9, 2020
Published: April 2, 2020
REFERENCES
1. Montijn, J.S., Vinck, M., and Pennartz, C.M.A. (2014). Population coding in
mouse visual cortex: response reliability and dissociability of stimulus tun-
ing and noise correlation. Front. Comput. Neurosci. 8, 58.
2. Tong, F. (2003). Primary visual cortex and visual awareness. Nat. Rev.
Neurosci. 4, 219–229.
3. Poort, J., Khan, A.G., Pachitariu, M., Nemri, A., Orsolic, I., Krupic, J.,
Bauza, M., Sahani, M., Keller, G.B., Mrsic-Flogel, T.D., and Hofer, S.B.
(2015). Learning enhances sensory and multiple non-sensory representa-
tions in primary visual cortex. Neuron 86, 1478–1490.
4. Jurjut, O., Georgieva, P., Busse, L., and Katzner, S. (2017). Learning en-
hances sensory processing in mouse V1 before improving behavior.
J. Neurosci. 37, 6460–6474.
5. Makino, H., and Komiyama, T. (2015). Learning enhances the relative
impact of top-down processing in the visual cortex. Nat. Neurosci. 18,
1116–1122.
6. Gavornik, J.P., and Bear, M.F. (2014). Learned spatiotemporal sequence
recognition and prediction in primary visual cortex. Nat. Neurosci. 17,
732–737.
7. Frenkel, M.Y., Sawtell, N.B., Diogo, A.C.M., Yoon, B., Neve, R.L., and
Bear, M.F. (2006). Instructive effect of visual experience in mouse visual
cortex. Neuron 51, 339–349.
8. Cooke, S.F., and Bear, M.F. (2010). Visual experience induces long-term
potentiation in the primary visual cortex. J. Neurosci. 30, 16304–16313.
9. Cooke, S.F., Komorowski, R.W., Kaplan, E.S., Gavornik, J.P., and Bear,
M.F. (2015). Visual recognition memory, manifested as long-term habitua-
tion, requires synaptic plasticity in V1. Nat. Neurosci. 18, 262–271.
10. Kaneko, M., Fu, Y., and Stryker, M.P. (2017). Locomotion induces stim-
ulus-specific response enhancement in adult visual cortex. J. Neurosci.
37, 3532–3543.
11. Clopath, C., Bonhoeffer, T., Hu
¨bener, M., and Rose, T. (2017). Variance
and invariance of neuronal long-term representations. Philos. Trans. R.
Soc. Lond. B Biol. Sci. 372, 20160161.
12. Keller, A.J., Houlton, R., Kampa, B.M., Lesica, N.A., Mrsic-Flogel, T.D.,
Keller, G.B., and Helmchen, F. (2017). Stimulus relevance modulates
contrast adaptation in visual cortex. eLife 6, e21589.
13. Pakan, J.M.P., Francioni, V., and Rochefort, N.L. (2018). Action and
learning shape the activity of neuronal circuits in the visual cortex. Curr.
Opin. Neurobiol. 52, 88–97.
14. Gilbert, C.D.D., and Li, W. (2012). Adult visual cortical plasticity. Neuron
75, 250–264.
15. Goltstein, P.M., Meijer, G.T., and Pennartz, C.M. (2018). Conditioning
sharpens the spatial representation of rewarded stimuli in mouse primary
visual cortex. eLife 7, e37683.
16. Goltstein, P.M., Coffey, E.B.J., Roelfsema, P.R., and Pennartz, C.M.A.
(2013). In vivo two-photon Ca2+ imaging reveals selective reward effects
on stimulus-specific assemblies in mouse visual cortex. J. Neurosci. 33,
11540–11555.
17. Wilmes, K.A., and Clopath, C. (2019). Inhibitory microcircuits for top-down
plasticity of sensory representations. Nat. Commun. 10, 5055.
18. Schuler, M.G., and Bear, M.F. (2006). Reward timing in the primary visual
cortex. Science 311, 1606–1609.
19. Levy, J.M., Zold, C.L., Namboodiri, V.M.K., and Hussain Shuler, M.G.
(2017). The timing of reward-seeking action tracks visually cued theta os-
cillations in primary visual cortex. J. Neurosci. 37, 10408–10420.
20. Pessoa, L. (2015). Multiple influences of reward on perception and atten-
tion. Vis. Cogn. 23, 272–290.
21. Niell, C.M., and Stryker, M.P. (2010). Modulation of visual responses by
behavioral state in mouse visual cortex. Neuron 65, 472–479.
22. Polack, P.-O., Friedman, J., and Golshani, P. (2013). Cellular mechanisms
of brain state-dependent gain modulation in visual cortex. Nat. Neurosci.
16, 1331–1339.
23. Saleem, A.B., Ayaz, A., Jeffery, K.J., Harris, K.D., and Carandini, M. (2013).
Integration of visual motion and locomotion in mouse visual cortex. Nat.
Neurosci. 16, 1864–1869.
24. Erisken, S., Vaiceliunaite, A., Jurjut, O., Fiorini, M., Katzner, S., and Busse,
L. (2014). Effects of locomotion extend throughout the mouse early visual
system. Curr. Biol. 24, 2899–2907.
25. Fu, Y., Tucciarone, J.M., Espinosa, J.S., Sheng, N., Darcy, D.P., Nicoll,
R.A., Huang, Z.J., and Stryker, M.P. (2014). A cortical circuit for gain con-
trol by behavioral state. Cell 156, 1139–1152.
26. Reimer, J., Froudarakis, E., Cadwell, C.R., Yatsenko, D., Denfield, G.H.,
and Tolias, A.S. (2014). Pupil fluctuations track fast switching of cortical
states during quiet wakefulness. Neuron 84, 355–362.
27. Pakan, J.M.P., Lowe, S.C., Dylda, E., Keemink, S.W., Currie, S.P., Coutts,
C.A., and Rochefort, N.L. (2016). Behavioral-state modulation of inhibition
is context-dependent and cell type specific in mouse visual cortex. eLife 5,
1–18.
28. Vinck, M., Batista-Brito, R., Knoblich, U., and Cardin, J.A. (2015). Arousal
and locomotion make distinct contributions to cortical activity patterns
and visual encoding. Neuron 86, 740–754.
29. Francioni, V., Padamsey, Z., and Rochefort, N.L. (2019). High and asym-
metric somato-dendritic coupling of V1 layer 5 neurons independent of be-
havioural state and visual stimulation. eLife 8:e49145, e49145.
30. Mineault, P.J., Tring, E., Trachtenberg, J.T., and Ringach, D.L. (2016).
Enhanced spatial resolution during locomotion and heightened attention
in mouse primary visual cortex. J. Neurosci. 36, 6382–6392.
31. Dadarlat, M.C., and Stryker, M.P. (2017). Locomotion enhances neural en-
coding of visual stimuli in mouse V1. J. Neurosci. 37, 3764–3775.
32. Swain, A.D., Park, E., Cheng, Z.Y., Kowalewski, N., Sun, A., Allen, T., and
Kuhlman, S.J. (2019). Visual acuity performance level is independent of
locomotion. bioRxiv. https://doi.org/10.1101/750844.
33. Keller, G.B.B., Bonhoeffer, T., and Hu
¨bener, M. (2012). Sensorimotor
mismatch signals in primary visual cortex of the behaving mouse.
Neuron 74, 809–815.
34. Attinger, A., Wang, B., and Keller, G.B. (2017). Visuomotor coupling
shapes the functional development of mouse visual cortex. Cell 169,
1291–1302.e14.
35. Leinweber, M., Ward, D.R., Sobczak, J.M., Attinger, A., and Keller, G.B.
(2017). A sensorimotor circuit in mouse cortex for visual flow predictions.
Neuron 95, 1420–1432.e5.
36. Zmarz, P., and Keller, G.B. (2016). Mismatch receptive fields in mouse vi-
sual cortex. Neuron 92, 766–772.
37. Chen, T.-W., Wardill, T.J., Sun, Y., Pulver, S.R., Renninger, S.L., Baohan,
A., Schreiter, E.R., Kerr, R.A., Orger, M.B., Jayaraman, V., et al. (2013).
Ultrasensitive fluorescent proteins for imaging neuronal activity. Nature
499, 295–300.
14 Current Biology 30, 1–15, May 18, 2020
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
38. Khan, A.G., Poort, J., Chadwick, A., Blot, A., Sahani, M., Mrsic-Flogel,
T.D., and Hofer, S.B. (2018). Distinct learning-induced changes in stim ulus
selectivity and interactions of GABAergic interneuron classes in visual cor-
tex. Nat. Neurosci. 21, 851–859.
39. Bieszczad, K.M., and Weinberger, N.M. (2012). Extinction reveals that pri-
mary sensory cortex predicts reinforcement outcome. Eur. J. Neurosci.
35, 598–613.
40. Kim, T., Chaloner, F.A., Cooke, S.F., Harnett, M.T., and Bear, M.F. (2020).
Opposing somatic and dendritic expression of stimulus-selective
response plasticity in mouse primary visual cortex. Front. Cell. Neurosci.
13, 555.
41. Diamanti, E.M., Reddy, C., Schro
¨der, S., Harris, K.D., Saleem, A.B., and
Carandini, M. (2019). Spatial encoding in the visual pathway arises in cor-
tex and depends on active navigation. bioRxiv 832915,https://doi.org/10.
1101/832915.
42. Jia, H., Rochefort, N.L., Chen, X., and Konnerth, A. (2010). Dendritic orga-
nization of sensory input to cortical neurons in vivo. Nature 464, 1307–
1312.
43. Ito, S., Feldheim, D.A., and Litke, A.M. (2017). Segregation of visual
response properties in the mouse superior colliculus and their modulation
during locomotion. J. Neurosci. 37, 8428–8443.
44. Roth, M.M., Dahmen, J.C., Muir, D.R., Imhof, F., Martini, F.J., and Hofer,
S.B. (2016). Thalamic nuclei convey diverse contextual information to layer
1 of visual cortex. Nat. Neurosci. 19, 299–307.
45. Aydın, C¸ ., Couto, J., Giugliano, M., Farrow, K., and Bonin, V. (2018).
Locomotion modulates specific functional cell types in the mouse visual
thalamus. Nat. Commun. 9, 4882.
46. Powell, K., Mathy, A., Duguid, I., and H
ausser, M. (2015). Synaptic repre-
sentation of locomotion in single cerebellar granule cells. eLife 4, 1–18.
47. Muzzu, T., Mitolo, S., Gava, G.P., and Schultz, S.R. (2018). Encoding of
locomotion kinematics in the mouse cerebellum. PLoS ONE 13, e0203900.
48. Albergaria, C., Silva, N.T., Pritchett, D.L., and Carey, M.R. (2018).
Locomotor activity modulates associative learning in mouse cerebellum.
Nat. Neurosci. 21, 725–735.
49. Gilbert, C.D., and Li, W. (2013). Top-down influences on visual processing.
Nat. Rev. Neurosci. 14, 350–363.
50. Milczarek, M.M., Vann, S.D., and Sengpiel, F. (2018). Spatial memory
engram in the mouse retrosplenial cortex. Curr. Biol. 28, 1975–1980.e6.
51. Mao, D., Kandler, S., McNaughton, B.L., and Bonin, V. (2017). Sparse
orthogonal population representation of spatial context in the retrosplenial
cortex. Nat. Commun. 8, 243.
52. Mao, D., Neumann, A.R., Sun, J., Bonin, V., Mohajerani, M.H., and
McNaughton, B.L. (2018). Hippocampus-dependent emergence of spatial
sequence coding in retrosplenial cortex. Proc. Natl. Acad. Sci. USA 115,
8015–8018.
53. V
elez-Fort, M., Bracey, E.F., Keshavarzi, S., Rousseau, C.V., Cossell, L.,
Lenzi, S.C., Strom, M., and Margrie, T.W. (2018). A circuit for integration
of head- and visual-motion signals in layer 6 of mouse primary visual cor-
tex. Neuron 98, 179–191.e6.
54. Fischer, L.F., Mojica Soto-Albors, R., Buck, F., and Harnett, M.T. (2020).
Representation of visual landmarks in retrosplenial cortex. eLife 9, e51458.
55. Zhang, S., Xu, M., Kamigaki, T., Hoang Do, J.P., Chang, W.C., Jenvay, S.,
Miyamichi, K., Luo, L., and Dan, Y. (2014). Selective attention. Long-range
and local circuits for top-down modulation of visual cortex processing.
Science 345, 660–665.
56. Pho, G.N., Goard, M.J., Woodson, J., Crawford, B., and Sur, M. (2018).
Task-dependent representations of stimulus and choice in mouse parietal
cortex. Nat. Commun. 9, 2596.
57. Ramesh, R.N., Burgess, C.R., Sugden, A.U., Gyetvan, M., and
Andermann, M.L. (2018). Intermingled ensembles in visual association
cortex encode stimulus identity or predicted outcome. Neuron 100, 900–
915.e9.
58. Pafundo, D.E., Nicholas, M.A., Zhang, R., and Kuhlman, S.J. (2016). Top-
down-mediated facilitation in the visual cortex is gated by subcortical neu-
romodulation. J. Neurosci. 36, 2904–2914.
59. Kang, J.I., Hupp
e-Gourgues, F., and Vaucher, E. (2014). Boosting visual
cortex function and plasticity with acetylcholine to enhance visual percep-
tion. Front. Syst. Neurosci. 8, 172.
60. Chubykin, A.A., Roach, E.B., Bear, M.F., and Shuler, M.G.H. (2013). A
cholinergic mechanism for reward timing within primary visual cortex.
Neuron 77, 723–735.
61. Pinto, L., Goard, M.J., Estandian, D., Xu, M., Kwan, A.C., Lee, S.-H.,
Harrison, T.C., Feng, G., and Dan, Y. (2013). Fast modulation of visual
perception by basal forebrain cholinergic neurons. Nat. Neurosci. 16,
1857–1863.
62. Lee, A.M., Hoy, J.L., Bonci, A., Wilbrecht, L., Stryker, M.P., and Niell, C.M.
(2014). Identification of a brainstem circuit regulating visual cortical state in
parallel with locomotion. Neuron 83, 455–466.
63. Fu, Y., Kaneko, M., Tang, Y., Alvarez-Buylla, A., and Stryker, M.P. (2015). A
cortical disinhibitory circuit for enhancing adult plasticity. eLife 4, e05558.
64. Kalogeraki, E., Greifzu, F., Haack, F., and Lo
¨wel, S. (2014). Voluntary phys-
ical exercise promotes ocular dominance plasticity in adult mouse primary
visual cortex. J. Neurosci. 34, 15476–15481.
65. Aronov, D., and Tank, D.W. (2014). Engagement of neural circuits underly-
ing 2D spatial navigation in a rodent virtual reality system. Neuron 84,
442–456.
66. Kaifosh, P., Zaremba, J.D., Danielson, N.B., and Losonczy, A. (2014).
SIMA: Python software for analysis of dynamic fluorescence imaging
data. Front. Neuroinform. 8, 80.
67. Keemink, S.W., Lowe, S.C., Pakan, J.M.P., Dylda, E., van Rossum,
M.C.W., and Rochefort, N.L. (2018). FISSA: A neuropil decontamination
toolbox for calcium imaging signals. Sci. Rep. 8, 3493.
68. Dylda, E., Pakan, J.M.P., and Rochefort, N.L. (2018). Chronic Two-Photon
Calcium Imaging in the Visual Cortex of Awake Behaving Mice. In
Handbook of Behavioral Neuroscience, 28, D. Manahan-Vaughan, ed.
(Elsevier), pp. 235–251.
69. Ringach, D.L., Shapley, R.M., and Hawken, M.J. (2002). Orientation selec-
tivity in macaque V1: diversity and laminar dependence. J. Neurosci. 22,
5639–5651.
70. Brainard, D.H. (1997). The psychophysics toolbox. Spat. Vis. 10, 433–436.
71. Pakan, J.M.P., Currie, S.P., Fischer, L., and Rochefort, N.L. (2018). The
impact of visual cues, reward, and motor feedback on the representation
of behaviorally relevant spatial locations in primary visual cortex. Cell Rep.
24, 2521–2528.
72. Mazurek, M., Kager, M., and Van Hooser, S.D. (2014). Robust quantifica-
tion of orientation selectivity and direction selectivity. Front. Neural
Circuits 8, 92.
73. Cohen, M.R., and Maunsell, J.H.R. (2009). Attention improves perfor-
mance primarily by reducing interneuronal correlations. Nat. Neurosci.
12, 1594–1600.
74. Ruxton, G.D., and Beauchamp, G. (2008). Time for some a priori thinking
about post hoc testing. Behav. Ecol. 19, 690–693.
Current Biology 30, 1–15, May 18, 2020 15
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
STAR+METHODS
KEY RESOURCES TABLE
LEAD CONTACT AND MATERIALS AVAILABILITY
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Janelle
MP Pakan (janelle.pakan@med.ovgu.de). This study did not generate new unique reagents.
EXPERIMENTAL MODEL AND SUBJECT DETAILS
Subjects
All animal experiments were approved by either the Animal Welfare and Ethical Review Board (AWERB) of the Universityof Edinburgh
(and experiments performed under a project license granted by the UK Home Office) or by the animal care committee of Sachsen-
Anhalt, Germany, and conformed with the UK Animals (Scientific Procedures) Act 1986 and the European Directive 86/609/EEC and
2010/63/EU on the protection of animals used for experimental purposes.
Animals were group housed (typically 2–4 mice) and a total of seven experimental groups were included in the study, each with 9
mice per group (see Table S1). Both male and female mice, aged 6-12 weeks, with a C57BL/6J background were used (RRID:
IMSR_JAX:000664; Jackson Laboratory). Mice were housed in standard cages at a 12h/12h light dark cycle, food and water were
provided ad libitum (except for during behavioral testing involving reward, see below).
METHOD DETAILS
Surgery
For cranial window implantation and virus injection, mice were anaesthetized with isoflurane (4% for induction and 1%–2% mainte-
nance during surgery) and mounted on a stereotaxic frame (David Kopf Instruments). Eye cream was applied to protect the eyes (Be-
panthen, Bayer), analgesics and anti-inflammatory drugs were injected subcutaneously (buprenorphine, 0.1 mg/kg of body weight,
carprofen, 0.15mg, and dexamethasone, 2mg). A section of scalp was removed, and the underlying bone was cleaned before a
REAGENT or RESOURCE SOURCE IDENTIFIER
Bacterial and Virus Strains
AAV1.Syn.GCaMP6s.WPRE.SV40 Addgene RRID:Addgene_100843
AAV1.Syn.GCaMP6f.WPRE.SV40 Addgene RRID:Addgene_100837
Experimental Models: Organisms/Strains
Mouse: C57BL/6J The Jackson Laboratory RRID:IMSR_JAX:000664
Software and Algorithms
MATLAB 2013/2017a Mathworks RRID:SCR_001622
Psychophysics Toolbox package for MATLAB http://psychtoolbox.org RRID:SCR_002881
LabVIEW version 8.2 National Instruments RRID:SCR_014325
Jet Ball system PhenoSys RRID:SCR_004072
ThorImage LS Microscopy Software Thorlabs version 2.4
ViRMEn [65]https://pni.princeton.edu/pni-software-tools/virmen
SIMA 1.3.2 (sequential image analysis) [66]https://pypi.org/project/sima/
ScanImage Vidrio Technologies RRID:SCR_014307
FISSA [67]https://github.com/rochefort-lab/fissa
ImageJ (Fiji) NIH – public domain http://fiji.sc; RRID:SCR_002285
Analyses were performed using custom-written
MATLAB scripts
This paper https://github.com/rochefort-lab
Other
Optical encoder Pewatron E7P, 250cpr
Reward spout Harvard Apparatus Cat#59-8636
Capacitive touch sensor Sparkfun Cat#SEN-12041
e1 Current Biology 30, 1–15.e1–e5, May 18, 2020
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
craniotomy (approximately 2x2 mm) was made over the left V1 (centered around 2.5 mm lateral and 0.5 mm anterior to lambda). Then
an adeno-associated virus (AAV) was injected using a pipette with ~20 mm tip diameter (Nanoject, Drummond Scientific) at a speed of
10 nL min
-1
at depths throughout the cortex (to label a cortical column, 2-3 injection sites were made ranging 250-600 um deep; 50-
100 nL per injection site for a total volume of 200 nl). Either AAV1.Syn.GCaMP6s.WPRE.SV40 (RRID:Addgene_100843) or AAV1.
Syn.GCaMP6f.WPRE.-SV40 (RRID:Addgene_100837) was injected in V1; GCaMP6s was used for all experimental groups except
for the 5 mice trained for both phase 1 and phase 2 in goal-directed VR group, for which we used GCaMP6f [see Table S1]; note,
there were no significant differences in responses to the repetitive grating between the GCaMP6s and GCaMP6f expressing mice
for the goal-directed VR group in phase 1, so results were pooled, see Figure 4F). After each injection, pipettes were left in situ
for an additional 5 min to prevent backflow. The craniotomy was then sealed with a glass coverslip and fixed with cyanoacrylate
glue (gel control, Loctite). A custom-built head-post was implanted on the exposed skull with glue and cemented with dental acrylic
(Paladur, Heraeus Kulzer). After recovery from anesthesia, animals were returned to their home cage for 2-3 weeks to allow for virus
expression and clearing of the cranial window [68] before imaging.
Two-photon imaging
Two-photon calcium imaging was performed using one of three resonant scanning two-photon microscopes (see Table S1). The first,
a custom built 12 kHz resonant scanning system with a Ti:Sapphire pulsing laser (Chameleon Vision-S, Coherent; < 70 fs pulse width,
80 MHz repetition rate) tuned to 920 nm. Images were acquired at 40 Hz (using a 40X 0.8 NA or a 25X 1.05 NA, Nikon objective, see
Table S1) with a custom-programmed LabVIEW based software (version 8.2; National Instruments). The second, a 8kHz resonant
scanning microscope (B-scope, Thorlabs) with a Ti:Sapphire pulsing laser (Chameleon Ultra II, Coherent; 140 fs pulse width,
80 MHz repetition rate) tuned to 920 nm. Images were acquired at 30 Hz (using a 20X objective 1.0 NA, Olympus) with ThorImageLS
software (version 2.4., Thorlabs). The third, using an 8kHz resonant scanning microscope (HyperScope, Scientifica) with an Ultasfast
laser (InSight X3 Dual output laser; Spectra-Physics; < 120 fs pulse width, 80 MHz repetition rate) tuned to 940 nm. Images were ac-
quired at 30 Hz (using a 16X objective 0.8 NA, Nikon, zoom factor 2x) with ScanImage software (Vidrio Technologies). With all two-
photon systems, chronic imaging of the same L2/3 field-of-view (at cortical depths between 180–280 mm) was performed across
consecutive days.
For all groups and all imaging sessions, mice were awake, head restrained, and placed on a cylindrical treadmill (either a 20 cm
polystyrene cylinder mounted on a ball-bearing axis with the custom two-photon system [69], or an air-suspended polystyrene 20
cm ball with B-Scope and HyperScope systems, which was fixed on both sides so that mice could run freely only in a linear direction).
Movements were monitored using an optical encoder (E7P, 250cpr, Pewatron, with custom two-photon system, sampling frequency
12 kHz; or Jet Ball system, PhenoSys GmbH, with B-Scope and HyperScope systems, sampling frequency 60 Hz;). In all cases, the
sampling frequency of the optical encoders were down-sampled to meet the sampling frequency of the imaging.
Stimulus presentation: pre and post testing days
For each experimental group, the orientation selectivity of neurons within the selected field-of-view for each animal was assessed
before (pre) and after (post) the presentation of a repetitive grating while animals were awake and head-fixed, but free to move at
will. Visual stimuli were generated using the Psychophysics Toolbox package [70] for MATLAB (Mathworks) and displayed on a single
LCD monitor (51 329 cm, Dell) placed 20 cm from the eye contralateral to the cranial window, covering 104x72
of the visual field.
Visual stimulation (10-20 trials) consisted of stationary full-field gratings for 2-4 s and corresponding moving stimulus for 2-3 s (fixed
spatial frequency [0.03-0.05 cpd] and constant temporal frequency [1-1.5 Hz], 4 equally spaced orientations in randomized order,
contrast 80%, mean luminance 37 cd/m
2
). Each oriented grating was separated by a gray period (isoluminant, 5 s) and each trial
started and ended with a gray screen for 2 s. Spontaneous activity in the dark was assessed by randomly interleaving trials where
no visual stimulation was presented (~10 trials, 60-90 s each).
Stimulus presentation: repetitive stimulus
For the repetitive stimulation days, a single oriented grating at a fixed spatial frequency was displayed in a 15 minute daily session for
5 consecutive days. The display and reward-association of the repetitive stimulus varied across groups based on: (1) either a con-
stant or dynamic temporal frequency, (2) coupling or uncoupling of the visual stimulus to locomotor behavior (i.e., coupled or un-
coupled optic-flow), and (3) the presence or absence of a reward associated with the repetitive stimulus (rewarded or unrewarded;
see also Table S1). Spontaneous activity in the dark was collected before the repetitive stimulus presentation on each day. The con-
trol no stimulus group did not receive any stimulus presentation in the 5 consecutive days between the pre and post testing days.
Unrewarded experimental groups
For the passive viewing group, a full-field grating was displayed on a single screen as in the pre and post testing days (see above). For
5 consecutive days, a single oriented grating was displayed in a 15 minute daily session with grating presentations of 4 s (at same
fixed spatial and constant temporal frequency as pre and post testing days), interleaved by isoluminant gray periods of randomized
duration between 5 to 15 s, to simulate the trial by trial nature of the goal-directed VR task, see below. The total time of grating stim-
ulation per session was 5 min. Mice were able to voluntarily run on the treadmill, however, running speed and visual stimulus presen-
tation were not coupled as the temporal frequency and spatial frequency of the repetitive grating were fixed and constant throughout.
For the uncoupled-unrewarded group, mice were placed in a virtual environment (custom two-photon system) consisting of a linear
corridor with the repetitive stimulus (vertical oriented grating) on the corridor walls. Animals ran freely, but visuomotor feedback was
Current Biology 30, 1–15.e1–e5, May 18, 2020 e2
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
uncoupled to optic-flow and varied randomly throughout each session independent of the animals movements (i.e., uncoupled VR
playback conditions); therefore, the stimulus temporal frequency was dynamic and randomly varied according to the same range of
parameters as in the goal-directed VR group (see below; range 0.5-3.5 Hz; average 2 Hz across trials).
For the no reward VR group, mice were placed in a virtual environment (PhenoSys GmbH; B-Scope two-photon system) consisting
of a linear corridor with the repetitive stimulus (vertical oriented grating) on the corridor walls. Animals ran freely, and visuomotor feed-
back was coupled with optic-flow throughout each session; therefore, the stimulus temporal frequency was dynamic and controlled
by the locomotion of each mouse.
Rewarded experimental groups
Mice were put on a 1 ml/day water restriction regime 3 days before training commenced to increase motivation during behavioral
training. This regime maintained bodyweight at 85%–90% of their free feeding weight, calculated as the mean of the last 3 days
before water restriction. Mice were positioned in front of two angled computer screens forming the VR environment with a reward
spout within reach. The reward spout (59-8636; Harvard Apparatus, UK) was fitted with a capacitive touch sensor (SEN-12041;
Sparkfun, CO, USA) to detect animal licking behavior. Either PhenoSys software (PhenoSys GmbH; for the HyperScope two-photon
system) or the MATLAB based package ViRMEn [65] (for the custom two-photon system) combined with custom written code was
used to design and run the presentation of the virtual environment and collect related data. The virtual reality system was updated at a
rate of 60 Hz. Imaging and behavioral datafiles were aligned post hoc, where the behavioral datafile was matched to the imaging
datafile by down-sampling and interpolating such that the aligned dataset had the same number of frames. Training consisted of daily
sessions where a single droplet of water (~5 ml/reward) was dispensed per trial (average trials per session for rewarded groups: D1,
129 ± 9; D2, 144 ± 12; D3, 166 ± 12; D4, 154 ± 10; D5, 168 ± 16). Following each session, the volume of water consumed during the
task was supplemented to 1ml if necessary.
For the goal-directed VR group, two-photon imaging was performed in combination with a virtual reality system as previously
described [71]. Briefly, the task required the mice to lick a spout for a water reward at a specific location along a virtual corridor
(120 cm from the beginning of the corridor), which was indicated by a change in visual stimulus from the oriented grating pattern
to black walls, referred to as the reward zone. Once the animal entered the reward zone (40 cm total length), within the first
20 cm (120-140cm) it could lick for a water droplet; this was considered a successful trial. When a reward was not triggered by
the mouse (unsuccessful trial), animals were given a water droplet at a default location 20 cm beyond the reward zone onset. In phase
1, all animals were presented with a single repetitive grating (vertical oriented bars) on the virtual corridor walls and in phase 2 a sub-
group of these animals (5 of 9 mice, see Table S1) were presented with the initial repetitive grating as well as an additional repetitive
grating (120 cm total length, 40 cm reward zone) angled at 45(presented in alternating blocks of 5 trials each). In phase 2, sessions
were 30 minutes long with an equal number of trials with each grating.
For the uncoupled-rewarded group, animals ran freely and the same grating and black zone were presented on the corridor walls
as in the goal-directed VR group but visuomotor feedback was uncoupled to optic-flow and varied randomly throughout each session
independent of the animals movements (i.e., uncoupled VR playback conditions); therefore, the stimulus temporal frequency was
dynamic and randomly varied according to the same range of parameters as in the goal-directed VR group (range 0.5-3.5 Hz; average
2 Hz across trials). Mice received rewards by self-initiated licking during the first 1.5 s (successful trial) of the presentation of the
reward zone (marked by black corridor walls) or, if this trigger was missed, dispensed by default (unsuccessful trial) after this time.
For the random reward group, animals ran freely and the same grating and black zone were presented on the corridor walls as in the
goal-directed VR group but visuomotor feedback was uncoupled to optic-flow and varied randomly throughout each session inde-
pendent of the animals movements (i.e., uncoupled VR playback conditions); therefore, the stimulus temporal frequency was dy-
namic and randomly varied according to the same range of parameters as in the goal-directed VR group (range 0.5-3.5 Hz; average
2 Hz across trials). Mice received rewards randomly throughout the session, either during the repetitive stimulus presentation (vertical
oriented grating on corridor walls) or during the presentation of the reward zone (marked by black corridor walls).
QUANTIFICATION AND STATISTICAL ANALYSIS
Image analysis
Images resulting from two-photon imaging were analyzed as previously described [27,71]. Briefly, we used discrete Fourier 2D-
based image alignment for motion correction of image frames (SIMA 1.3.2, sequential image analysis [66]). Regions of interest
(ROIs) corresponding to neuronal cell bodies were selected manually and aligned across days. Pixel intensity within each ROI
was averaged to create a raw fluorescence time series F(t). Baseline fluorescence F
0
was computed for each neuron by taking
the 5
th
percentile of the smoothed F(t) (1 Hz lowpass, zero-phase, 60
th
-order FIR filter) and the change in fluorescence relative to
baseline (DF/F
0
) was calculated (F(t)-F
0
/F
0
). In order to remove neuropil contamination, we used nonnegative matrix factorization
(NMF), as implemented in FISSA [67](https://github.com/rochefort-lab/fissa). All further analyses were performed using custom-writ-
ten scripts in MATLAB (MathWorks, MA, USA), which are freely available via GitHub (https://github.com/rochefort-lab/
Henschke_et_al_CurrBiology2020).
Licking behavior
To assess the behavioral performance of the mice during the active VR task, a spatial modulation index (SMI) of licking was calculated
[71]. The licks of each trial were randomly permuted, and we determined the proportion of trials in which at least one lick event was
e3 Current Biology 30, 1–15.e1–e5, May 18, 2020
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
inside the reward zone. This was repeated 1000 times and the mean success rate of the shuffled distribution was calculated. The SMI
value was calculated by dividing the original success rate (successful trials/total number of trials) by the mean of the shuffled distri-
bution. If the animal licks few times but in the right spot, this number will be high (> 1). In contrast, if the animal licks in a spatially
indiscriminate pattern, the number will approach 1. If the animal licks often, but keeps missing the reward zone, the SMI will be < 1.
Orientation selectivity
The response rtðqkÞof a neuron to an oriented grating qkduring trial twas calculated by averaging the DF/F
0
over the stimulation
period. This response was then normalized by subtracting the local baseline activity (rtðqkÞ=mean DF/F
0
- minimum DF/F
0
during
the 2 s window preceding stimulation). The main response RðqkÞto orientation qkwas obtained by averaging the responses rtðqkÞ
across trials. The preferred stimulus of a neuron was the orientation that elicited the maximal response RðqkÞ. The orientation selec-
tivity was characterized by the circular variance (CirVar) [69]:
1CirVar =
PkRðqkÞexpð2iqkÞ
PkRðqkÞ
The peak angle and peak magnitude of the tuning curve were estimated by taking the argument and modulus of the response vector
PkRðqkÞexpð2iqkÞ=PkRðqkÞ[72]. The peak angle was used to compute the orientation shift between pre and post days. A neuron then
qualified as orientation selective if it passed two criteria: (i) it was significantly tuned (peak magnitude > 25th percentile of pre testing
day for each animal), and (ii) the response to its preferred orientation was significantly higher than to the orthogonal-to-preferred
orientation across trials (responses rtðqkÞ, preferred versus orthogonal, across all trials; p < 0.05, Wilcoxon signed rank test).
The reliability of the orientation selectivity was assessed by calculating the coefficient of variation (CV; the ratio of the standard
deviation to the mean) of the peak magnitude of the tuning curve across trials.
Stimulus decoding and discriminability
To quantify the specific increase in the proportion of neurons selective for the repetitive grating (R
select
neurons) relative to the change
in the proportion of neurons that were selective for the orthogonal grating (O
select
neurons; which was only presented to the mice on
pre and post days but not during the 5 consecutive day) we calculating an index RO
index
=R
select
-O
select
/R
select
+O
select
and quantified
the change in this ratio from pre to post (post RO
index
– pre RO
index
). As an additional control, we calculated a similar index (O
index
)
between two different oriented gratings (orthogonal to the repetitive, OR
select,
and orthogonal to the angled grating, OA
select
) both of
which, mice were only exposed to on pre and post days (O
index
=OA
select
-OR
select
/OA
select
+OR
select
).
To quantify the accuracy by which V1 activity could be classified based on the population activity, we used a template-matching
decoder [1], which compares the population activity per trial to response templates of the different oriented gratings. These templates
are generated by taking the mean DF/F
0
during each oriented stimulus period, or corridor, (q), for each neuron in a single field-of-view;
resulting in a template of population activity (R
q
) per trial. The similarity of this template to the actual population activity (R
P
) for all
other trials per stimulus orientation is given by:
Iq=P
N
i=1
RP
i
$Rq
i
jRqj$jRpj;
where iindexes the N elements (neurons) of R. The similarity index Iis calculated for all presented stimulus orientations and the de-
coded output is determined by taking the condition with the highest similarity to the template population activity. Decoder accuracy is
given by the percentage of correctly decoded trials.
To quantify how individual neurons encode information regarding the separate orientated stimuli we used a Bayesian maximum-
likelihood decoder for each neuron separately. For each trial, the response of a neuron to a specific orientation was calculated by
taking the average DF/F
0
over the 2 s stimulation period. For each orientation qwe approximated the corresponding response dis-
tribution of a neuron p(R|q) with a Gaussian. Leaving one trial out (test trial), we determined the orientation–specific likelihood distri-
bution by computing the mean and standard deviation of the responses to that orientation across the remaining trials (training trials).
We then decoded the responses of the test trial; we assumed a uniform prior across orientations and hence the posterior p(q|R) is
directly proportional to the likelihood [1]. Thus, for each response R_t of the test trial we selected the orientation qthat maximized
the likelihood p(R_t|q). We repeated the leave-one-out procedure by looping over trials. The performance for a given neuron was eval-
uated for each orientation by calculated the percentage of correct inferences. We then averaged the performance for each orientation
across all neurons per animal. Finally, we estimated the changes of decoder performance per animal and per orientation between the
post and pre testing days.
For Phase 2 of the goal-directed VR task, to determine grating responsive neurons (corridor responsive and corridor-selective neu-
rons, i.e neurons that were responsive to the oriented grating stimulus pattern along the virtual corridor), we compared the activity
within 25 cm blocks before and after the reward-zone onset for each trial (R
pre
versus R
post
; p < 0.001, Wilcoxon signed rank test).
Corridor responsive neurons were categorized as those where R
pre
(35 to 10 cm before the reward-zone onset) was significantly
greater than R
post
(0 to 25 cm after the reward-zone onset), hence, the neuron decreased its activity at the region of the virtual
corridor that lacked a visual stimulus (the reward zone), for either, or both, of the presented virtual corridors (vertical and/or angled).
Current Biology 30, 1–15.e1–e5, May 18, 2020 e4
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
Corridor-selective neurons were then further defined as those significantly responsive only to a single virtual corridor (e.g., vertical
corridor pattern) and not to the other (e.g., angled corridor pattern).
Stimulus discriminability (d’) was calculated following previously described methods [31,73] based on the average responses (DF/
F
0
) across trials for the two behaviorally relevant stimuli (vertical and angled gratings): taking the 2 s stimulation period during the
single screen condition and a 2 s period at the start of each trial when each grating was presented on the walls of the virtual corridor
for the conditions in the VR environment. The d’ was then calculated as follows:
d0=m1m2
1
2s2
1+s2
21
2
;
where m
1
and m
2
were the means of the responses for each stimulus, and s
2
their respective variances and the absolute value taken
for analysis.
Reward-responsive neurons
To define if a neuron was reward-responsive, we first aligned responses (DF/F
0
) to the reward event (i.e., time of reward onset) for
each trial. We then examined the average responses within a 1 s window before the onset of the reward event (2 s from onset
to 1 s from onset) compared to the average responses within a 1 s window after the onset of the reward (reward onset to 1 s after
reward onset) for each individual trial. If the neuron had significantly higher responses after the reward onset compared to the window
before the reward onset it was considered reward-responsive (p < 0.05; paired t test).
Analysis of locomotion
For all two-photon systems, locomotion and stationary periods were calculated as previously described [27]. Briefly, stationary pe-
riods were defined as time points when the instantaneous speed was < 0.1 cm/s. Locomotion corresponded to periods meeting three
criteria: instantaneous speed R0.1 cm/s, 0.25 Hz lowpass filtered speed R0.1 cm/s, and an average speed R0.1 cm/s over a 2 s
window centered at the time point. Any inter-locomotion interval shorter than 500 ms was also labeled as locomotion. Periods less
than 3 s after or 0.2 s before a period of locomotion were not considered as stationary. We quantified the effect of locomotion on
neuronal activity by using a locomotion modulation index (LMI), which is the difference between the DF/F
0
during locomotion (R
L
)
and stationary (R
s
) periods, normalized by the activity during both periods: LMI=(R
L
-R
s
)(R
L
+R
s
).
Statistics
Unless otherwise stated, error bars in all graphs indicate standard error of the mean (s.e.m.) and all statistical tests were two-tailed.
Unless otherwise stated, mean, error values and statistics were calculated across animals; for analysis where we isolated specific
populations of neurons with multiple response parameters across experimental groups (e.g., for Figure 6), mean and error values
and statistics were calculated across neurons. For planned comparisons across different experimental groups, we used one-way
ANOVA with Fisher’s least significant difference (lsd) test with no correction for multiple comparisons. Since, in our study design,
we have formulated a specific hypothesis for the tests we perform across specific experimental groups, we are testing these planned
comparison predictions using the lsd test to increase our statistical power and avoid the increased probability of Type II errors that
may occur with other post hoc tests [74]; this choice is at the cost of potentially increasing the likelihood of Type 1 errors. We report
exact p values for all the comparisons that were made in Table S2. For paired comparisons across conditions or days for the same
population of neurons, or when assessing the proportion of neurons within each group, we used the Wilcoxon signed rank test (non-
parametric paired difference test) or paired t test. For unpaired comparisons across days involving different underlying population of
neurons within the same group, we used Mann-Whitney U-test or unpaired Student’s t test. For all correlations we report Pearson’s
correlation coefficient.
DATA AND CODE AVAILABILITY
The code used for analysis in this study are freely available via GitHub repository (https://github.com/rochefort-lab/
Henschke_et_al_CurrBiology2020).
e5 Current Biology 30, 1–15.e1–e5, May 18, 2020
Please cite this article in press as: Henschke et al., Reward Association Enhances Stimulus-Specific Representations in Primary Visual Cortex, Current
Biology (2020), https://doi.org/10.1016/j.cub.2020.03.018
... Over the course of 5 days, mice learned to lick within the reward zone to receive the water reward. This visual detection task was used to investigate V1 neuronal activity before, during and after learning (Henschke et al., 2020). The goal of the experiments was to elucidate how repeated exposure to a stimulus modulates neural population responses, particularly in the presence of a stimulus-associated reward. ...
... Our analysis was based on deconvolved spike trains instead of the calcium transients. Spiking activity had been reconstructed using the MLspike algorithm (Deneux et al., 2016) (see Henschke et al., 2020 for more details). The data we used in this study were limited to one mouse on day 4 of the experiment when the animal was an expert at the task. ...
... Having validated the vine flow copula framework with artificial data, we subsequently focused on spiking activity from a subset of 5 V1 layer 2/3 neurons while the animal was navigating a virtual reality corridor ( Figure 6A) (Henschke et al., 2020). A parametric vine copula framework with Gaussian processes (Kudryashova et al., 2022) was recently applied to calcium transients from the same V1 layer 2/3 neurons (Henschke et al., 2020). ...
Article
Full-text available
Recordings of complex neural population responses provide a unique opportunity for advancing our understanding of neural information processing at multiple scales and improving performance of brain computer interfaces. However, most existing analytical techniques fall short of capturing the complexity of interactions within the concerted population activity. Vine copula-based approaches have shown to be successful at addressing complex high-order dependencies within the population, disentangled from the single-neuron statistics. However, most applications have focused on parametric copulas which bear the risk of misspecifying dependence structures. In order to avoid this risk, we adopted a fully non-parametric approach for the single-neuron margins and copulas by using Neural Spline Flows (NSF). We validated the NSF framework on simulated data of continuous and discrete types with various forms of dependency structures and with different dimensionality. Overall, NSFs performed similarly to existing non-parametric estimators, while allowing for considerably faster and more flexible sampling which also enables faster Monte Carlo estimation of copula entropy. Moreover, our framework was able to capture low and higher order heavy tail dependencies in neuronal responses recorded in the mouse primary visual cortex during a visual learning task while the animal was navigating a virtual reality environment. These findings highlight an often ignored aspect of complexity in coordinated neuronal activity which can be important for understanding and deciphering collective neural dynamics for neurotechnological applications.
... In order to assess our framework's applicability to neuronal activity we used 2-photon calcium imaging data of neuronal activity in the mouse primary visual cortex, that were was collected at the Rochefort lab (see [50] for more details). Briefly, V1 layer 2/3 neurons labeled with the calcium indicator GCamP6s were imaged while the animal was headfixed, freely running on a cylindrical treadmill and navigating through a virtual reality environment (160 cm). ...
... Over the course of 5 days, mice learned to lick within the reward zone to receive the water reward. This visual detection task was used to investigate V1 neuronal activity before, during and after learning [50]. The goal of the experiments was to elucidate how repeated exposure to a stimulus modulates neural population responses, particularly in the presence of a stimulus-associated reward. ...
... Our analysis was based on deconvolved spike trains instead of the calcium transients. Spiking activity had been reconstructed using the MLspike algorithm [51] (see [50] for more details). The data we used in this study were limited to one mouse on day 4 of the experiment when the animal was an expert at the task. ...
Preprint
Full-text available
Recordings of complex neural population responses provide a unique opportunity for advancing our understanding of neural information processing at multiple scales and improving performance of brain computer interfaces. However, most existing analytical techniques fall short of capturing the complexity of interactions within the concerted population activity. Vine copula-based approaches have shown to be successful at addressing complex high-order dependencies within the population, disentangled from the single-neuron statistics. However, most applications have focused on parametric copulas which bear the risk of misspecifying dependence structures. In order to avoid this risk, we adopted a fully non-parametric approach for the single-neuron margins and copulas by using Neural Spline Flows (NSF). We validated the NSF framework on simulated data of continuous and discrete type with various forms of dependency structures and with different dimensionality. Overall, NSFs performed similarly to existing non-parametric estimators, while allowing for considerably faster and more flexible sampling which also enables faster Monte Carlo estimation of copula entropy. Moreover, our framework was able to capture low and higher order heavy tail dependencies in neuronal responses recorded in the mouse primary visual cortex during a visual learning task while the animal was navigating a virtual reality environment. These findings highlight an often ignored aspect of complexity in coordinated neuronal activity which can be important for understanding and deciphering collective neural dynamics for neurotechnological applications.
... Modification is not limited to deprivation paradigms. Sensory representations are persistently altered by reinforced behavioral training under conditions in which specific visual cues are associated with salience [16][17][18][19][20] , and learning impacts the representation of nontrained stimuli as well 21 . These studies raise the possibility that every time new associations are made, existing function is altered. ...
... To determine whether our results generalize to a single-modality task, we trained mice in a visual discrimination task in which it was previously established that V1 activity is required for improved behavioral performance 16,50,51 . Converging evidence indicates visual discrimination training enhances the neural representation of rewarded stimuli by increasing selectivity for the stimuli experienced during training 16,21 and in some cases improving response reliability 16 , and at the same time suppresses responses to non-relevant stimuli 16 . Enhanced responses to rewarded stimuli are known to generalize across task variations experienced in the training environment. ...
... As such, reward-induced changes in selectivity are considered to be context-specific. Furthermore, in many instances long-lasting changes observed in the training environment are restricted to stimulus-specific assemblies, and enhancement to more than one rewarded stimulus is possible due to assembly-specific plasticity 20,21 . Based on these previous observations, we hypothesized that similar to our non-matched sensory task, reward training-induced changes are context specific, and tuning to visual stimuli remains largely undisturbed outside of the training environment. ...
Article
Full-text available
Acquisition of new skills has the potential to disturb existing network function. To directly assess whether previously acquired cortical function is altered during learning, mice were trained in an abstract task in which selected activity patterns were rewarded using an optical brain-computer interface device coupled to primary visual cortex (V1) neurons. Excitatory neurons were longitudinally recorded using 2-photon calcium imaging. Despite significant changes in local neural activity during task performance, tuning properties and stimulus encoding assessed outside of the trained context were not perturbed. Similarly, stimulus tuning was stable in neurons that remained responsive following a different, visual discrimination training task. However, visual discrimination training increased the rate of representational drift. Our results indicate that while some forms of perceptual learning may modify the contribution of individual neurons to stimulus encoding, new skill learning is not inherently disruptive to the quality of stimulus representation in adult V1.
... The presentation of feedback elicits a cascade of neural processes that ranges from feedback valence processing in subcortical areas of the brain (Delgado et al., 2005) to the monitoring and updating of stimulus-reward contingencies and subsequent decision making with regard to whether or not future behavior needs to be adjusted in the frontal cortices (Rushworth et al., 2011). In addition, the sensory cortices show changes in neural activity dependent on the sensory features of the reward-associated stimuli (van den Berg et al., 2019;Henschke et al., 2020;Schiffer et al., 2014;Folstein et al., 2013). The result of these neurocognitive processes might ensure that if the rewarded stimulus is encountered again in the future, attention will be directed to this information, resulting in faster and/or more accurate performance (e.g., Hickey et al., 2010). ...
... As a result of feedback processing and learning of stimulus-reward associations, the sensitivity of the sensory cortex for reward-related information might increase. Evidence supporting this account stems from neuroimaging studies that indeed report enhanced neural discriminability of relevant stimulus features in the sensory cortex after learning (Henschke et al., 2020;Schiffer et al., 2014;Folstein et al., 2013). Moreover, the post-feedback re-activation of the sensory cortex was found to be localized to those areas of the visual cortex that were involved in initial processing of the reward-related stimuli. ...
Article
Full-text available
Learned stimulus-reward-associations can modulate behavior and the underlying neural processing of information. We investigated the cascade of these neurocognitive mechanisms involved in the learning of spatial stimulus-reward-associations. Using electroencephalogram recordings (EEG) while participants performed a probabilistic spatial reward learning task, we observed that the feedback-related negativity (FRN) component was more negative in response to loss feedback compared to gain feedback, but showed no modulation by learning. The late positive component (LPC) became larger in response to losses as the learning set progressed, but smaller in response to gains. Additionally, feedback-locked alpha frequency oscillations measured over occipital sites were predictive of N2pc amplitudes, a marker of spatial attention orienting, observed on the next trial. This relationship was found to become stronger with learning set progression. Taken together, we elucidated neurocognitive dynamics underlying feedback processing during spatial reward learning, and the subsequent effects of these learned spatial stimulus-reward associations on spatial attention. (150 words)
... Sample sizes were chosen to be comparable to sample sizes of other studies in the field, and were not chosen based on statistical methods 20,22,26,48,56 . No data were excluded from the analyses. ...
Article
Full-text available
Primary sensory cortex has long been believed to play a straightforward role in the initial processing of sensory information. Yet, the superficial layers of cortex overall are sparsely active, even during sensory stimulation; additionally, cortical activity is influenced by other modalities, task context, reward, and behavioral state. Our study demonstrates that reinforcement learning dramatically alters representations among longitudinally imaged neurons in superficial layers of mouse primary somatosensory cortex. Learning an object detection task recruits previously unresponsive neurons, enlarging the neuronal population sensitive to touch and behavioral choice. Cortical responses decrease upon repeated stimulus presentation outside of the behavioral task. Moreover, training improves population encoding of the passage of time, and unexpected deviations in trial timing elicit even stronger responses than touches do. In conclusion, the superficial layers of sensory cortex exhibit a high degree of learning-dependent plasticity and are strongly modulated by non-sensory but behaviorally-relevant features, such as timing and surprise.
... Finally, we would like to note that the demonstration of reinforcement-like learning via the transformation of eTraces in V1, somewhat unexpected in a primary sensory cortex, suggests the universality of the rules governing cortical plasticity. Indeed, the primary visual cortex is increasingly considered not merely an early stage of visual processing but also a primary locus of perceptual learning driven by behavioral relevance variables including reward, fear, attention, saliency, novelty [44][45][46][47][48][49][50][51] in which neuromodulatory systems likely play a central roles 2,52,53 . ...
Article
Full-text available
Reinforcement allows organisms to learn which stimuli predict subsequent biological relevance. Hebbian mechanisms of synaptic plasticity are insufficient to account for reinforced learning because neuromodulators signaling biological relevance are delayed with respect to the neural activity associated with the stimulus. A theoretical solution is the concept of eligibility traces (eTraces), silent synaptic processes elicited by activity which upon arrival of a neuromodulator are converted into a lasting change in synaptic strength. Previously we demonstrated in visual cortical slices the Hebbian induction of eTraces and their conversion into LTP and LTD by the retroactive action of norepinephrine and serotonin Here we show in vivo in mouse V1 that the induction of eTraces and their conversion to LTP/D by norepinephrine and serotonin respectively potentiates and depresses visual responses. We also show that the integrity of this process is crucial for ocular dominance plasticity, a canonical model of experience-dependent plasticity. Previous work has identified synaptic eligibility traces in slice preparations. Here the authors provide demonstration of eligibility traces in vivo during reinforcement learning.
... The increase in spontaneous activity that we found after habituation differs from changes in neuronal activity that occur after instrumental learning, where learned sequences of activity replay at rest and may form a kind of internal practice (Eagleman and Dragoi, 2012;Guidotti et al., 2015;Henschke et al., 2020;Ji and Wilson, 2007). In contrast, after habituation, we see a widespread increase in V1 activity across the entire spectrum of orientation selectivity. ...
Article
Full-text available
The cerebral cortex is spontaneously active, but the function of this ongoing activity remains unclear. To test whether spontaneous activity encodes learned experiences, we measured the response of neuronal populations in mouse primary visual cortex with chronic two-photon calcium imaging during visual habituation to a specific oriented stimulus. We find that, during habituation, spontaneous activity increases in neurons across the full range of orientation selectivity, eventually matching that of evoked levels. This increase in spontaneous activity robustly correlates with the degree of habituation. Moreover, boosting spontaneous activity with two-photon optogenetic stimulation to the levels of visually evoked activity accelerates habituation. Our study shows that cortical spontaneous activity is linked to habituation, and we propose that habituation unfolds by minimizing the difference between spontaneous and stimulus-evoked activity levels. We conclude that baseline spontaneous activity could gate incoming sensory information to the cortex based on the learned experience of the animal.
Article
Sensory areas are spontaneously active in the absence of sensory stimuli. This spontaneous activity has long been studied; however, its functional role remains largely unknown. Recent advances in technology, allowing large-scale neural recordings in the awake and behaving animal, have transformed our understanding of spontaneous activity. Studies using these recordings have discovered high-dimensional spontaneous activity patterns, correlation between spontaneous activity and behavior, and dissimilarity between spontaneous and sensory-driven activity patterns. These findings are supported by evidence from developing animals, where a transition toward these characteristics is observed as the circuit matures, as well as by evidence from mature animals across species. These newly revealed characteristics call for the formulation of a new role for spontaneous activity in neural sensory computation.
Article
Visual discrimination improves with training, a phenomenon that is thought to reflect plastic changes in the responses of neurons in primary visual cortex (V1). However, the identity of the neurons that undergo change, the nature of the changes, and the consequences of these changes for other visual behaviors remain unclear. We used chronic in vivo 2-photon calcium imaging to monitor the responses of neurons in the V1 of tree shrews learning a Go/No-Go fine orientation discrimination task. We observed increases in neural population measures of discriminability for task-relevant stimuli that correlate with performance and depend on a select subset of neurons with preferred orientations that include the rewarded stimulus and nearby orientations biased away from the non-rewarded stimulus. Learning is accompanied by selective enhancement in the response of these neurons to the rewarded stimulus that further increases their ability to discriminate the task stimuli. These changes persist outside of the trained task and predict observed enhancement and impairment in performance of other discriminations, providing evidence for selective and persistent learning-induced plasticity in the V1, with significant consequences for perception.
Article
Learning is an essential cognitive mechanism allowing behavioral adaptation through adjustments in neuronal processing. It is associated with changes in the activity of sensory cortical neurons evoked by task-relevant stimuli. However, the exact nature of those modifications and the computational advantages they may confer are still debated. Here, we investigated how learning an orientation discrimination task alters the neuronal representations of the cues orientations in the primary visual cortex (V1) of male and female mice. When comparing the activity evoked by the task stimuli in naïve mice and mice performing the task, we found that the representations of the orientation of the rewarded and non-rewarded cues were more accurate and stable in trained mice. This better cue representation in trained mice was associated with a distortion of the orientation representation space such that stimuli flanking the task-relevant orientations were represented as the task stimuli themselves, suggesting that those stimuli were generalized as the task cues. This distortion was context dependent as it was absent in trained mice passively viewing the task cues and enhanced in the behavioral sessions where mice performed best. Those modifications of the V1 population orientation representation in performing mice were supported by a suppression of the activity of neurons tuned for orientations neighboring the orientations of the task cues. Thus, visual processing in V1 is dynamically adapted to enhance the reliability of the representation of the learned cues and favor generalization in the task-relevant computational space.Significance statementPerformance improvement in a task often requires facilitating the extraction of the information necessary to its execution. Here, we demonstrate the existence of a suppression mechanism that improves the representation of the orientations of the task stimuli in the V1 of mice performing an orientation discrimination task. We also show that this mechanism distorts the V1 orientation representation space, leading stimuli flanking the task stimuli orientations to be generalized as the task stimuli themselves.
Article
Full-text available
The process by which visual information is incorporated into the brain’s spatial framework to represent landmarks is poorly understood. Studies in humans and rodents suggest that retrosplenial cortex (RSC) plays a key role in these computations. We developed an RSC-dependent behavioral task in which head-fixed mice learned the spatial relationship between visual landmark cues and hidden reward locations. Two-photon imaging revealed that these cues served as dominant reference points for most task-active neurons and anchored the spatial code in RSC. This encoding was more robust after task acquisition. Decoupling the virtual environment from mouse behavior degraded spatial representations and provided evidence that supralinear integration of visual and motor inputs contributes to landmark encoding. V1 axons recorded in RSC were less modulated by task engagement but showed surprisingly similar spatial tuning. Our data indicate that landmark representations in RSC are the result of local integration of visual, motor, and spatial information.
Article
Full-text available
Active dendrites impact sensory processing and behaviour. However, it remains unclear how active dendritic integration relates to somatic output in vivo. We imaged semi-simultaneously GCaMP6s signals in the soma, trunk and distal tuft dendrites of layer 5 pyramidal neurons in the awake mouse primary visual cortex. We found that apical tuft signals were dominated by widespread, highly correlated calcium transients throughout the tuft. While these signals were highly coupled to trunk and somatic transients, the frequency of calcium transients was found to decrease in a distance-dependent manner from soma to tuft. Ex vivo recordings suggest that low-frequency back-propagating action potentials underlie the distance-dependent loss of signals, while coupled somato-dendritic signals can be triggered by high-frequency somatic bursts or strong apical tuft depolarization. Visual stimulation and locomotion increased neuronal activity without affecting somato-dendritic coupling. High, asymmetric somato-dendritic coupling is therefore a widespread feature of layer 5 neurons activity in vivo.
Article
Full-text available
Daily exposure of awake mice to a phase-reversing visual grating stimulus leads to enhancement of the visual-evoked potential (VEP) in layer 4 of the primary visual cortex (V1). This stimulus-selective response potentiation (SRP) resembles and shares mechanistic requirements with canonical long-term synaptic potentiation (LTP). However, it remains to be determined how this augmentation of a population response translates into altered neuronal activity of individual V1 neurons. To address this question, we performed longitudinal calcium imaging of layer 4 excitatory neurons in V1 and tracked changes associated with the induction and expression of SRP. We found no evidence for a net change in the fraction of visually responsive neurons as the stimulus became familiar. However, endoscopic calcium imaging of layer 4 principal neurons revealed that somatic calcium transients in response to phase-reversals of the familiar visual stimulus are reduced and undergo strong within-session adaptation. Conversely, neuropil calcium responses and VEPs are enhanced during familiar stimulus viewing, and the VEPs show reduced within-session adaptation. Consistent with the exquisite selectivity of SRP, the plasticity of cellular responses to phase-reversing gratings did not translate into altered orientation selectivity to drifting gratings. Our findings suggest a model in which augmentation of fast, short-latency synaptic (dendritic) responses, manifested as enhanced layer 4 VEPs, recruits inhibition to suppress cellular activity. Reduced cellular activity to the familiar stimulus may account for the behavioral correlate of SRP, orientation-selective long-term habituation.
Article
Full-text available
Rewards influence plasticity of early sensory representations, but the underlying changes in circuitry are unclear. Recent experimental findings suggest that inhibitory circuits regulate learning. In addition, inhibitory neurons are highly modulated by diverse long-range inputs, including reward signals. We, therefore, hypothesise that inhibitory plasticity plays a major role in adjusting stimulus representations. We investigate how top-down modulation by rewards interacts with local plasticity to induce long-lasting changes in circuitry. Using a computational model of layer 2/3 primary visual cortex, we demonstrate how interneuron circuits can store information about rewarded stimuli to instruct long-term changes in excitatory connectivity in the absence of further reward. In our model, stimulus-tuned somatostatin-positive interneurons develop strong connections to parvalbumin-positive interneurons during reward such that they selectively disinhibit the pyramidal layer henceforth. This triggers excitatory plasticity, leading to increased stimulus representation. We make specific testable predictions and show that this two-stage model allows for translation invariance of the learned representation. Rewards can improve stimulus processing in early sensory areas but the underlying neural circuit mechanisms are unknown. Here, the authors build a computational model of layer 2/3 primary visual cortex and suggest that plastic inhibitory circuits change first and then increase excitatory representations beyond the presence of rewards.
Preprint
Full-text available
During navigation, the visual responses of neurons in primary visual cortex (V1) are modulated by the animal’s spatial position. Here we show that this spatial modulation is similarly present across multiple higher visual areas but largely absent in the main thalamic pathway into V1. Similar to hippocampus, spatial modulation in visual cortex strengthens with experience and requires engagement in active behavior. Active navigation in a familiar environment, therefore, determines spatial modulation of visual signals starting in the cortex.
Preprint
Full-text available
Locomotion has a global impact on circuit function throughout the cortex, including regulation of spatiotemporal dynamics in primary visual cortex (V1). The mechanisms driving state-changes in V1 result in a 2-3 fold gain of responsiveness to visual stimuli. To determine whether locomotion-mediated increases in response gain improve the perception of spatial acuity we developed a head-fixed task in which mice were free to run or sit still during acuity testing. Spatial acuity, ranging from 0.1 to 0.7 cycles/deg, was assessed before and after 3-4 weeks of reward-based training in adult mice. Training on vertical orientations once a day improved the average performance across mice by 22.5 +/- 0.05%. Improvement transferred to non-trained orientations presented at 45 deg, indicating that the improvement in acuity generalized. Furthermore we designed a second closed-loop task in which acuity threshold could be directly assessed in a single session. Using this design, we established that acuity threshold matched the upper limit of the trained spatial frequency; in two mice spatial acuity threshold reached as high as 1.5 cycles/deg. During the 3-4 weeks of training we collected a sufficient number of stimulus trials in which mice performed above chance but below 100% accuracy. Using this subset of stimulus trials, we found that perceptual acuity was not enhanced on trials in which mice were running compared to trials in which mice were still. Our results demonstrate that perception of spatial acuity is not improved by locomotion.
Chapter
Full-text available
Our ability to learn relies on the potential of neuronal circuits to change through experience. Recent advances in genetic tools and in vivo imaging have enabled novel investigations into the mechanisms of plasticity at the level of neuronal circuits. This chapter describes an experimental method for long-term activity monitoring of large neuronal populations, activity, with single-cell resolution, in the cortex of awake behaving mice. By combining genetically encoded calcium indicators with two-photon imaging, this technique allows researchers to relate the activity of subtypes of neurons directly to the animal's behavior over temporal scales from hundreds of milliseconds to several weeks. We detail the key experimental aspects of long-term, repeated two-photon calcium imaging: surgical procedures for chronic craniotomies, repeated image acquisition, and critical analysis parameters for two-photon calcium imaging data. Finally, we review recent applications of this technique to study the plasticity of neuronal circuits in the primary visual cortex.