You are on page 1of 12

European Journal of Neuroscience, pp. 1–12, 2018 doi:10.1111/ejn.

13830

Adaptation to stimulus orientation in mouse primary visual


cortex

Jillian L. King and Nathan A. Crowder


Department of Psychology and Neuroscience, Dalhousie University, 1355 Oxford Street, PO Box 15000, Halifax, NS, B3H 4R2,
Canada

Keywords: animal model, comparative neuroscience, electrophysiology, tilt-after-effect, vision

Abstract
Information processing in the visual system is shaped by recent stimulus history, such that prolonged viewing of an adapting stim-
ulus can alter the perception of subsequently presented test stimuli. In the tilt-after-effect, the perceived orientation of a grating is
often repelled away from the orientation of a previously viewed adapting grating. A possible neural correlate for the tilt-after-effect
has been described in cat and macaque primary visual cortex (V1), where adaptation produces repulsive shifts in the orientation
tuning curves of V1 neurons. We investigated adaptation to stimulus orientation in mouse V1 to determine whether known species
differences in orientation processing, notably V1 functional architecture and proportion of tightly tuned cells, are important for
these repulsive shifts. Unlike the consistent repulsion reported in other species, we found that repulsion was only about twice as
common as attraction in our mouse data. Furthermore, adapted responses were attenuated across all orientations. A simple
model that captured key physiological findings reported in cats and mice indicated that the greater proportion of broadly tuned
neurons in mice may explain the observed species differences in adaptation.

Introduction
Studies of sensory adaptation and/or after effects suggest that the 2012; Patterson et al., 2013). Several related studies have made pro-
visual system possesses a number of mechanisms to calibrate itself gress in establishing how aspects of the stimulus protocol can affect
based on what has been viewed in the recent past (for reviews see the nature of orientation adaptation (duration of adaptation: Dragoi
Carandini, 2000; Kohn, 2007). Adaptation to the orientation of an & Sur, 2000; Ghisovan et al., 2009; adapting angle: Patterson et al.,
edge or texture is especially interesting because unlike brightness, 2013; and stimulus size: Wissig & Kohn, 2012), but there are many
colour or contrast, this process is thought to emphasize the primary remaining details to be revealed about the neural mechanisms under-
visual cortex (V1) and extrastriate areas as antecedent stages of anal- lying this form of short-term plasticity.
ysis show little selectivity for orientation (but see Smith et al., We examined orientation adaptation in the mouse primary visual
1990; Xu et al., 2002; Kuhkmann & Vidyasagar, 2011; Naito et al., cortex to gain a comparative perspective on this phenomenon. Despite
2013). In the tilt-after-effect and similar visual illusions, prolonged mice having spatial acuity up to two orders of magnitude lower than
viewing of an adaptor grating causes the orientation of the subse- classical animal models of vision (Wong & Brown, 2006), mouse V1
quently presented test grating to appear shifted away from (or ‘re- neurons share many features with those of more visual species includ-
pelled’ by) the adaptor’s orientation (Gibson & Radner, 1937; ing: tuning for spatial and temporal frequencies (Gao et al., 2010); the
Levinson & Sekuler, 1976; Patterson & Becker, 1996; Schrater & presence of simple and complex cells (Liu et al., 2009; Van den
Simoncelli, 1998; Clifford, 2002). A potential neural correlate for Bergh et al., 2010); binocular disparity selectivity (Scholl et al.,
these perceptual effects has been described in cat and macaque, 2013); and most relevant for the current work, selectivity for orienta-
where adaptation within the classical receptive field of V1 neurons tion and direction (Niell & Stryker, 2008; Liu et al., 2011; Tan et al.,
produces repulsive shifts in the orientation tuning curves of these 2011). Critically, however, there are also at least three differences in
cells (Dragoi & Sur, 2000; Dragoi et al., 2001; Wissig & Kohn, the way orientation information is dealt with in mouse V1 compared
to classical models. First, there is a larger proportion of neurons that
are not selective or poorly tuned for orientation in mouse V1 (Niell &
Correspondence: Nathan A. Crowder, as above. E-mail: nathan.crowder@dal.ca Stryker, 2008; Gao et al., 2010; Liu et al., 2011; Tan et al., 2011;
Stroud et al., 2012). Second, there is evidence that orientation selec-
Received 25 August 2017, revised 15 December 2017, accepted 8 January 2018
tivity emerges prior to V1 in the mouse, and the sharpening of orienta-
Edited by Panayiota Poirazi. Reviewed by Athanassia Papoutsi, Institute of Molecular tion tuning as signals pass from the dorsal lateral geniculate nucleus
Biology and Biotechnology, Greece; and Nicholas Price, Monash University, Australia (dLGN) to V1 is not as dramatic as in cats or primates (Li et al.,
All peer review communications can be found with the online version of the article. 2013; Piscopo et al., 2013; Scholl et al., 2013; Zhao et al., 2013; but

© 2018 Federation of European Neuroscience Societies and John Wiley & Sons Ltd
2 J. L. King and N. A. Crowder

see Lien & Scanziani, 2013). Finally, whereas cats and primates have generated in Spike2 from triggered transistor-transistor logic (TTL)
columns of orientation selective V1 neurons organized into an orienta- pulses from a window discriminator (Cornerstone by Dagan). Spike
tion map with iso-orientation domains that converge in pinwheel cen- sorting was performed offline with Spike2 software, which first
tres, these orientation maps are either absent or poorly developed in searched for and sorted spikes using a supervised template-matching
mice (Ohki et al., 2005; Ringach et al., 2016). This last difference algorithm, and then displayed candidate spikes. We used a principle
may be especially important because there is evidence that orientation components analysis to check the clustering of spike waveforms. Data
adaptation in cats depends on neuronal location within the orientation were exported to MATLAB (Math Works, Natick, MA, USA) and neu-
map (Dragoi et al., 2001). Any differences in the pattern of orienta- ronal responses were represented as spike density functions with
tion adaptation between species could indicate that the features absent 1 kHz resolution, generated by convolving a delta function at each
from the mouse are the ones which are critical for the repulsive shifts spike arrival time with a Gaussian window.
seen in the traditional animal models.
We investigated the effects of adaptation on the orientation tuning
Visual stimuli
of mouse V1 neurons using single unit recording and adaptation
paradigms that were comparable with previous studies in cat and Once a visually responsive neuron was isolated, the receptive field
monkey (Dragoi et al., 2001; Wissig & Kohn, 2012; Patterson et al., (RF) was mapped by hand using an ophthalmoscope. Quantitative
2013). We found that unlike in other species, adaptation induced stimuli, which were programmed in MATLAB using the Psychophysics
repulsive shifts in the tuning curves of orientation selective neurons Toolbox extension (Brainard, 1997; Pelli, 1997), were presented on a
only about twice as frequently as it induced attractive shifts. We also calibrated CRT monitor (LG Flatron 915FT Plus 19 inch display,
found that adaptation attenuated responses across all orientations, and 100 Hz refresh, 1024 9 768 pixels, mean luminance = 30 cd/m2) at
this effect appears to be more pronounced in mice than in other spe- a viewing distance of 20–30 cm. All stimuli were presented for 8–12
cies. We used a simplified model to show that these differences in repetitions. Each unit’s preferred orientation, RF size, spatial fre-
adaptation found in mice could arise from the greater proportion of quency (SF), and temporal frequency (TF) were analyzed online to
neurons that are broadly tuned or poorly modulated by orientation in ensure robust responses during the orientation adaptation protocol.
this species. Preferred size was tested with two stimuli: (1) sine-wave gratings in
different sized circular apertures, and (2) a full field grating with dif-
ferent sized grey discs at its centre. These two stimuli helped centre
Materials and methods the monitor, and ensured the stimulus was placed within the neuron’s
classical receptive field. The size of the circular aperture used for all
Animals
subsequent stimuli was chosen as either the diameter where the size
Extracellular recordings were performed on 27 male C57BL/6J mice tuning function began to asymptote in neurons lacking surround sup-
obtained from The Jackson Laboratories (JAX stock # 000664) 2– pression or the peak of the size tuning function for neurons that
7 months of age and weighing 22–33 g. All experimental proce- showed surround suppression.
dures were performed in accordance with the guidelines of the
Canadian Council on Animal Care and were approved by the Dal-
Orientation adaptation
housie University Committee on Laboratory Animals.
For the orientation adaptation protocol, orientation tuning was measured
with drifting square-wave gratings (SF = 0.03 cpd; TF = 2 Hz; con-
Physiological preparation
trast = 1) spaced 18° apart and spanning 180° centred on the unit’s pre-
Animals were pre-medicated with an injection of chlorprothixene ferred orientation as determined from the online analysis. Drifting test
(5 mg/kg I.P.; Sigma Aldrich) and then placed in a custom face gratings with randomly selected orientations were presented for 2 s each.
mask and anaesthetized with isoflurane in oxygen for the remainder On adapted trials, this test grating was preceded by a 2-s drifting adaptor
of the experiment (2.5% isoflurane during induction, 1.5% during grating, whereas on nonadapted trials this test grating was preceded only
surgery and 0.5% during recording; Pharmaceutical Partners of by a grey of mean luminance (Fig. 1A). We attempted to select an
Canada). Additional doses of chlorprothixene were given every four adapting orientation that fell midway between the peak and trough of the
hours. Once anesthetized, mice were maintained at a body tempera- orientation tuning curve (flank adaptation), but there was enough vari-
ture of 37.5 °C with a heating pad, and their corneas were protected ability that some units were adapted near their peak, whereas other adap-
by frequent application of optically neutral silicone oil (30 000 cSt, tors were nearly orthogonal to the peak. We referred to adaptation with a
Sigma-Aldrich). In preparation for electrophysiological recordings, stimulus that elicited firing ≤ 10% above the untuned baseline of the
the scalp was removed and a head post was secured using dental control orientation tuning curve as end-of-flank adaptation (Kohn &
epoxy. A small craniotomy (~1 mm2) was then made 0.8 mm ante- Movshon, 2004). Adapted and nonadapted trials were randomized, and
rior and 2.3 mm lateral from lambda (Paxinos & Franklin, 2001). A a grey of mean luminance was shown for the 6-s interstimulus interval
wall of petroleum jelly surrounded the craniotomy and was filled (ISI) between trials. This ISI was based on pilot data from 10 neurons
with saline to prevent dehydration of the cortex. The pupils were where we measured recovery from adaptation by comparing spike rates
not dilated so as to maintain a large depth of focus, and the eyes elicited by adapting gratings that followed either nonadapted or adapted
were not immobilized because eye movements under anaesthesia trials and found no evidence of a difference between these conditions for
have been shown to be negligible in mice (Wang & Burkhalter, the 6-s ISI (P = 0.35, paired t-test).
2007; Niell & Stryker, 2008; Gao et al., 2010).
Extracellular recordings were made with glass micropipettes con-
Data analysis
taining 2 M NaCl, with a tip diameter of 2–5 lm. Signals from indi-
vidual units were isolated, amplified, filtered, and acquired with a The least squares method was used to fit each orientation tuning
CED 1401 interface and Spike2 software (Cambridge Electronic curve to a von Mises function, which is a circular Gaussian often
Designs, Cambridge, UK) sampled at 40 kHz. Online responses were used in modelling orientation selectivity (e.g. Swindale, 1998):

© 2018 Federation of European Neuroscience Societies and John Wiley & Sons Ltd
European Journal of Neuroscience, 1–12
Orientation adaptation in mouse V1 3

of freedom for the simpler function (df1) and von Mises functions (df2)
were calculated as the number of data points minus the number of free
parameters being estimated. A P-value was then calculated to determine
whether the von Mises curve fits the data significantly better than the
simpler function using the MATLAB ‘fcdf’ function.
Each unit’s selectivity for stimulus orientation was also quantified
using a discrimination index for orientation (DIo), which accounts
for both depth of tuning and the variability of responses (Gao et al.,
2010):
RespMax  RespMin
DIo ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð3Þ
RespMax  RespMin þ 2 SSE=ðN  M Þ

where RespMax and RespMin are the maximum and minimum


responses of the neuron to drifting gratings, SSE is the sum of squares
error, N is the total number of stimuli presented (orientations 9
repetitions), and M is the number of different orientations.
The final analysis quantified how many spatially phase sensitive
units were in our sample using the F1/F0 ratio. The F1/F0 ratio is
determined by dividing the first Fourier coefficient of the response
(F1) by the mean time-averaged response (F0). This F1/F0 ratio has
been used in classical animal models of vision to categorize orienta-
tion tuned V1 neurons as simple (F1/F0 > 1) or complex types (F1/
F0 < 1; e.g. Skottun et al., 1991). Here we use the nomenclature of
phase sensitive (F1/F0 > 1) and phase invariant (F1/F0 < 1) because
not all of the units in our sample were orientation selective.

Statistical analysis
We used parametric analyses (specific tests noted in Results)
because the data of interest appeared to be approximately normally
distributed. The Benjamini-Hochberg procedure for controlling false
Fig. 1. Adaptation to orientation in mouse V1 neurons. (A) Diagram depict- discovery rate was applied (33 total comparisons), and adjusted P-
ing the timing of the orientation adaptation protocol, with adapting stimuli
(white blocks), test stimuli (black blocks), and inter-stimulus intervals (grey values are reported (Benjamini & Hochberg, 1995).
blocks) aligned with spiking activity of a sample neuron. (B–E) Orientation
tuning curves of four sample neurons with orientation on the abscissa and
mean response rate on the ordinate. Control and adapted conditions are Modelling
shown as filled and empty circles, respectively. Thin curves represent best
fits of a von Mises function to the control (solid) or adapted (dashed) data.
We implemented a simplified feed-forward model to explore poten-
The spike trace in A and tuning curves in B are from the same unit. Solid tial mechanisms underlying species differences in orientation adapta-
vertical lines indicate the adapting orientation, and dashed horizontal lines tion. The model simulated the responses of individual extra-granular
indicate the spontaneous activity. All error bars indicate SEM. V1 neurons by pooling signals from a population of orientation
tuned excitatory and inhibitory inputs and then generating firing
rp ðhÞ ¼ m þ aebðcosðhhpref Þ1Þ ð1Þ rates with a power-law nonlinearity (Priebe & Ferster, 2008; Tan
et al., 2011). We modelled two sources of adaptation: (1) each input
where the response (rp) at a given angle (h) is determined by the tun- underwent firing rate-dependent divisive scaling; and (2) the simu-
ing bandwidth (b), preferred orientation (hpref), amplitude (a) and lated neuron underwent activity-dependent hyperpolarization (Dhruv
baseline response rate (m). Goodness of fits were calculated with R2. et al., 2011; Wissig & Kohn, 2012). We used a population of 200
Two simpler functions were also fit to the data to further characterize of these simulated neurons to capture the major features and spec-
tuning curve shapes: (1) data points from the control condition were trum of effects observed in our data.
fit to a straight line (2 free parameters) to obtain a statistical measure We defined excitatory input (rexcite) to each simulated neuron as a
of orientation selectivity; and (2) data points from the adapted condi- scaled population of von Mises curves:
tion were fit to a modified von Mises function with hpref fixed at the " !#
nonadapted value (3 free parameters) to obtain a statistical measure X
n   euðcosðhhi Þ1Þ
hðcosðhi hnom Þ1Þ
for peak shift. An F-test was then used to compare the fits of the four rexcite ðhÞ ¼ c þ de 
i¼1 1 þ geuðcosðhadapt hi Þ1Þ
parameter von Mises function to these simpler functions:
ð4Þ

ðSS1  SS2 Þ=ðdf 1  df 2 Þ where hi is the preferred orientation of input i, hadapt is the angle of
F¼ ð2Þ
ðSS2 Þ=ðdf 2 Þ the adapting grating, u is the bandwidth of input i, and g is the gain
of divisive scaling from adaptation (which is set to 0 in the control
where SS1 is the residual sum of squares for the simpler function, and condition). The divisive scaling of each input is firing rate depen-
SS2 is the residual sum of squares for von Mises function. The degrees dent because part of the denominator (euðcosðhadapt hi Þ1Þ ) calculates

© 2018 Federation of European Neuroscience Societies and John Wiley & Sons Ltd
European Journal of Neuroscience, 1–12
4 J. L. King and N. A. Crowder

the firing of input i at the adapting angle. Excitatory synaptic (13), but adaptation affected these two groups similarly so they were
strength among V1 neurons is highest between neurons with similar pooled in all subsequent analyses. Adaptation effects did not corre-
tuning properties (Cossell et al., 2015), so we scaled the amplitude late with recording depth either, so we made no attempt to segregate
of each von Mises curve based on tuning similarity to the simulated units by cortical layer.
neuron. Thus, hnom is the nominal preferred orientation of the simu- The sample neurons in Fig. 1B–E demonstrate not only the spec-
lated neuron as dictated by the strength of its synaptic inputs, h is trum of orientation tuning in our sample, but also the variability of
the bandwidth of this synaptic selectivity, d determines the gain of adaptation effects. The neurons in Fig. 1B and C both had well
the synaptic selectivity scaling factor, and c is an offset that allows tuned nonadapted orientation tuning functions (filled circles). The
some influence from inputs with dissimilar tuning. We used 50 exci- adapted responses (empty circles) of the cell in Fig. 1B showed
tatory and inhibitory inputs (n), but similar results were observed repulsion from the adapting angle (vertical line), whereas the
when n was 20–200. The preferred orientations (hi) for the popula- adapted responses of the cell in Fig. 1C show a generalized
tion of inputs were taken from a normally distributed n-long list of response reduction with little repulsion. The cell in Fig. 1D is an
random numbers. The standard deviation of this list was varied to example of end-of-flank adaptation in a cell with a robust untuned
simulate members of the local cortical network available in cat ori- component in its orientation function, which again produced a gen-
entation tuning columns or mouse salt-and-pepper orientation tuning eralized response reduction with little repulsion. Figure 1E shows
(Dragoi et al., 2001; Ohki et al., 2005; Nauhaus et al., 2008). the effect of adaptation in a poorly tuned neuron. The smooth curves
For simplicity, inhibitory input (rinhibit) did not include a synaptic in Fig. 1B–E show von Mises fits to the nonadapted (solid lines)
strength scaling mechanism, but merely functioned as a broadly and adapted (dotted lines) data. We used parameters from these fits
tuned normalization pool (Carandini & Heeger, 2012): to quantify the effects of adaptation: repulsive or attractive peak
shifts were measured as the change in the preferred orientation
! (hpref); adaptation-induced suppression was measured as changes in
X
n
epðcosðhhj Þ1Þ both peak height (a) and baseline (m); finally, changes in the
rinhibit ðhÞ ¼ ð5Þ
j¼0 1 þ qepðcosðhadapt hj Þ1Þ breadth of orientation selectivity were measured using tuning band-
width (b) and half-width at half-maximum above the untuned base-
where hj is the preferred orientation of input j, p is the bandwidth of line (HWHM; following Niell & Stryker, 2008).
input j, and q is the gain of divisive scaling from adaptation (which
is set to 0 in the control condition). As with excitation, the preferred Segregation of orientation tuned cells
orientations (hj) for the population of inhibitory inputs were taken
from a normally distributed n-long list of random numbers, but the It is apparent from the sample neurons in Fig. 1 that some cells
standard deviation of this distribution was always large to simulate a were much more orientation selective than others. Therefore, reason-
broadly tuned divisive normalization pool. The total input to the cell ing that a neuron’s adaptation to a specific orientation is only mean-
was defined as follows: ingful if it can reliably distinguish between different orientations, we
chose to segregate our data into orientation tuned and nonoriented
cells. Orientation selective units had: (a) an R2 ≥ 0.5 for the von
rexcite ðhÞ Mises curve fits to ensure the data were not too jagged; and (b) a
rinput ðhÞ ¼ ð6Þ
1 þ rinhibit ðhÞ DIo ≥ 0.425, which corresponded to approximately 2 : 1 modula-
tion. We also ensured that using more liberal (DIo ≥ 0.225) or con-
The spiking response of the model neuron was generated by servative (DIo ≥ 0.625) cut-offs did not introduce any noticeable
applying a power-law nonlinearity to the input: biases to our main adaptation effects. Our sample contained 56 ori-
entation tuned and 64 nonoriented units, and this proportion fell
within previously reported ranges (Niell & Stryker, 2008; Gao et al.,
rspike ðhÞ ¼ jrinput ðhÞ  Vthresh jwþ ð7Þ
2010). Furthermore, our median nonadapted HWHM for orientation
tuned cells was 29.7° (interquartile range = 19.6–48°), which is sim-
where Vthresh is the spiking threshold, ||+ indicates half-wave rectifi- ilar to the median of 28° reported by Niell & Stryker (2008). The
cation, and w is the exponent (Priebe & Ferster, 2008; Tan et al., median bandwidth was 4.92 (interquartile range = 1.7–11.4). All
2011). Following Wissig & Kohn (2012), we implemented an activ- units that passed our criteria for orientation selectivity had signifi-
ity-dependent hyperpolarization (simulated as an increase in thresh- cantly better fits to a von Mises function than a straight line
old) to serve as a second adaptation mechanism: (P < 0.05; see Eqn 2 in Materials and methods), whereas no units
that failed our criteria had significantly better fits.
Vthresh adapted ¼ Vthresh þ rspike ðhadapt Þ½k1  k2 ðhnom  hadapt Þ ð8Þ Before focusing exclusively on orientation tuned neurons, it is
important to note that nonoriented cells were strongly visually
where k1 and k2 are arbitrary constants chosen to simulate the range of responsive and consistently showed adaptation. When our discrimi-
effects in our data set. Deeply modulated orientation tuning curves nation index (Eqn 3 in Materials and methods) was modified to
will show more hyperpolarization when the adapt angle is closer to measure how well neurons could distinguish between the preferred
their preferred orientation, whereas poorly modulated tuning curves grating stimulus and a uniform grey field that elicited spontaneous
will show similar amounts of hyperpolarization at all angles. activity, there was no evidence of a difference between nonoriented
and orientation tuned cells (P = 0.21, two-sample t-test). We used
the change in the baseline normalized to the nonadapted maximal
Results
firing rate as a common measure of adaptation across all neurons (as
We analysed the effects of adaptation on orientation tuning in 120 other parameters like tuning width or preferred orientation were
units recorded in mouse V1. Like Gao et al. (2010), our sample invalid for nonoriented cells) and found orientation tuned neurons
contained more phase-invariant (107) than phase-sensitive units showed similar mean decreases in baseline (0.25) to nonoriented

© 2018 Federation of European Neuroscience Societies and John Wiley & Sons Ltd
European Journal of Neuroscience, 1–12
Orientation adaptation in mouse V1 5

cells (0.21) with no evidence of a population difference (P = 0.06,


two-sample t-test).

Population measures of adaptation


When examining adaptation effects across our population of orienta-
tion tuned cells, we first distinguished between flank adaptation
(n = 43) and end-of-flank adaptation (n = 13) because it has been
previously shown in primates that flank adaptation causes mainly
repulsion, whereas the effects of end-of-flank adaptation are more
variable (Wissig & Kohn, 2012; Patterson et al., 2013). We anal-
ysed changes in peak shift, peak height, baseline, bandwidth,
HWHM, and DIo with six mixed repeated-measures ANOVAs with
adaptation (within condition: control vs. adapted) and the flank
effect (between condition: flank vs. end-of-flank adaptation) as fac-
tors. For each ANOVA, we were interested in the main effect of adap-
tation, and the flank interaction that indicated whether there was a
different amount of adaptation in flank or end-of-flank conditions.
Adaptation produced ‘repulsive’ shifts in the preferred orientation
if the adapted peak moved away from the adapt angle, and ‘attrac-
tive’ shifts if the adapted peak moved towards the adapt angle (for
all figures negative peak shifts indicate repulsion). Figure 2A plots
peak shift as a function of the angle between the adaptor and the
nonadapted response peak. Our sample contained 32 cells (57%)
that showed significant repulsive shifts, 18 cells (32%) that showed
significant attractive shifts, and six cells (11%) that did not shift sig-
nificantly (P < 0.05; see Eqn 2 in Materials and methods). The pop-
ulation average peak shift was 2.81°. Previous studies that
Fig. 2. Effects of adaptation on orientation tuning. (A) The peak shift
presented control and adapted stimulus conditions in separate blocks induced by adaptation (ordinate) is plotted against the orientation difference
excluded units with large changes in orientation preference based on between the adapting grating and the control peak (abscissa). Negative peak
the suspicion that these changes reflect drifting isolation over time shift values indicate repulsion. The thick black line shows a 7.5°-wide sliding
(e.g. Wissig & Kohn, 2012). We did not exclude neurons based on window calculating the mean peak shift. Population mean peak shifts for
this criterion because our stimulus had adapted and nonadapted trials flank adapted (empty triangles) and end-of-flank adapted neurons (filled trian-
gles) are shown on the right. (B and C) show the normalized change in peak
randomly shuffled within the same session; therefore, if unit isola- height and baseline with the same format as A, respectively. Negative values
tion varied over time it would affect both types of trials equally. indicate decreased activity following adaptation. (D) Decreases in peak
Although adaptation could induce substantial attractive or repulsive height (ordinate) were larger than decreases in baseline (abscissa) for most
shifts in individual neurons, neither the main effect of adaptation neurons. Proportional changes in HWHM (E), and bandwidth (F) are shown
with the same format as A, but without the lines indicating sliding mean. For
(F1,54, P = 0.3) nor the flank interaction (F1,54, P = 0.96) were sig- all plots, open circles are flank adapted cells and filled circles are end-of-
nificant. Thus, our data from mouse differ from the consistent repul- flank adapted cells.
sion reported in previous cat and primate studies.
In contrast to the variable effects on the preferred orientation,
response amplitude was consistently reduced by adaptation. Changes should be smaller than the decrease in peak firing. The decrease in
in peak and baseline firing rates were both normalized to the non- peak firing was significantly larger than the decrease in baseline
adapted peak firing rate and negative numbers indicate suppression. (P = 2.8 9 104, paired t-test), which is more consistent with
Adaptation decreased peak height in all neurons, regardless of the divisive scaling.
relative angle between the adaptor and the nonadapted peak We used two complementary measures of tuning breadth: HWHM
(Fig. 2B); however, a 7.5°-wide sliding window calculating the depends predominantly on the width near the tuning curve’s peak,
mean normalized change in peak height (i.e. a boxcar filter) indi- whereas bandwidth was affected by the overall shape of the curve.
cates adaptation within ~20° of the nonadapted peak produced Both changes in tuning breadth were calculated as the proportional
slightly larger drops (solid line). Adaptation decreased baseline fir- difference, such that negative values indicated the curve became nar-
ing for all neurons in our sample, and the magnitude of this drop rower (HWHM; Fig. 2E) or less circular (bandwidth; Fig. 2F) fol-
appeared to be independent of the relative angle between the adaptor lowing adaptation. The main effect of adaptation was not significant
and the peak (Fig. 2C). This consistent suppressive effect was sup- for HWHM (F1,54, P = 0.74) or bandwidth (F1,54, P = 0.78). Nei-
ported by a significant main effect of adaptation for both peak ther HWHM (F1,54, P = 0.1) nor bandwidth (F1,54, P = 0.06) had
height (F1,54, P = 3.5 9 1011) and baseline (F1,54, P = 1.7 9 significant flank interactions (Fig. 2E and F). There was also no
1010). The flank interaction was not significant for peak height main effect of adaptation (F1,54, P = 0.48) or flank interaction
(F1,54, P = 0.99) or baseline (F1,54, P = 0.87). We then compared (F1,54, P = 0.42) for DIo.
the magnitude of suppressive effects on peak height and baseline for Orientation tuning curves in mouse V1 are generally broader and
each neuron in Fig. 2D. If the entire adapted curve was shifted less modulated than those in cat and macaque (Dr€ager, 1975; Wagor
downward, as if adaptation had a subtractive influence on firing, et al., 1980; Metin et al., 1988; Sohya et al., 2007; Niell & Stryker,
these two measures should be about equal. However, if adaptation 2008; Liu et al., 2009; Runyan et al., 2010; Tan et al., 2011; Kerlin
induced a divisive scaling of the adapted curve the drop in baseline et al., 2010), so we performed several additional analyses to

© 2018 Federation of European Neuroscience Societies and John Wiley & Sons Ltd
European Journal of Neuroscience, 1–12
6 J. L. King and N. A. Crowder

Fig. 3. Potential influences on the direction and amplitude of peak shifts. (A) Histogram showing proportional differences between the firing rate at the peak
of the adapted curve and the control curve at that same angle (see inset diagram). All differences were negative indicating adapted curves always fell under con-
trol curves. (B) DIo (abscissa) plotted against peak shift (ordinate). Thick line in lower inset shows a 0.04-wide sliding window calculating the range of peak
shifts (window width depicted as a black rectangle with an arrow). (C) Bandwidth (abscissa) plotted against peak shift (ordinate) with the same format as B
(sliding window has a width of 2). (D) Normalized firing rate at the adapt angle (ordinate) plotted against the relative angle between nonadapted peak and
adapting orientation (abscissa). (E, F) Normalized change in peak height (E) or peak shift (F) is plotted against the normalized firing rate at the adapt angle
(now on the abscissa). The thick line in the lower insets shows a 0.04-wide sliding window calculating the range of peak drops (E) or peak shifts (F) for flank
adapted cells. Regression line (solid) and 95% confidence intervals (dashed) are shown in D–F. For all plots, open circles are flank adapted cells and filled cir-
cles are end-of-flank adapted cells.

determine whether this difference could account for the lack of con- insets) indicated that neurons with low DIo (those less modulated by
sistent repulsion in our data. Adapting cat V1 neurons, which have orientation) or low bandwidth values (more circular orientation tun-
narrower orientation tuning than mouse and even primate V1 cells ing) tend to have more variability in the size of their peak shifts.
(Nauhaus et al., 2008), can induce facilitation at some orientations The spike rate elicited by the adapting stimulus is critical for
such that the adapted curve shifts out from under the control curve other forms of adaptation in mouse V1 (LeDue et al., 2012, 2013;
(Dragoi et al., 2001). This effect was never observed in our data, King et al., 2016), so we examined the importance of response
and the histogram in Fig. 3A illustrates that the response at the magnitude on our measures of adaptation. Figure 3D shows that
adapted peak was always smaller than the response at the identical responses to the adapting stimulus (normalized to the peak non-
angle on the nonadapted curve. Given this constraint, large individ- adapted response) were more robust the closer the adapt angle was
ual peak shifts in either direction could only be possible in cells to the peak of the nonadapted curve (Spearman rank correlation =
with broad tuning or large untuned responses. Both DIo (r = 0.02, 0.66, P = 4.24 9 108), as expected from orientation tuned cells.
P = 0.84; Fig. 3B) and bandwidth (r = 0.05, P = 0.71; Fig. 3C) The normalized spike rate elicited at the adapt angle was poorly cor-
were poorly correlated with peak shift, simply because individual related with the decrease in the adapted peak amplitude (Fig. 3E;
cells could show attractive or repulsive shifts. However, sliding win- r = 0.03, P = 0.8) or peak shift (Fig. 3F; r = 0.09, P = 0.53). How-
dows calculating the range of peak shifts (Fig. 3B and C bottom ever, sliding windows calculating the range of both peak drops

Fig. 4. Adaptation over time. (A) Tuning curves from a sample neuron with responses averaged over the entire 2-s test period (format identical to Fig. 1). (B)
Neuron from A with responses divided into eight 250-ms epochs. Inset values indicate DIo of control (top) and adapted curves (bottom). (C) Average control
(filled circles) and adapted (empty circles) values of DIo over eight 250-ms epochs; values were normalized to the mean nonadapted DIo calculated over the
entire 2-s test. (D) Average peak orientation of control (filled circles) and adapted (empty circles) curves over eight 250-ms epochs with control curves aligned
to 0° at the first time bin. (E) Population histogram scoring the consistency of attraction/repulsion across epochs (see Results). (F) The peak shift (ordinate) in
the first 250-ms epoch plotted against the orientation difference between the adaptor and control peak (abscissa). The categorization of flank and end-of-flank
adaptation remained based on tuning curves averaged over the full 2-s stimulus duration. Format as in Fig. 2A. (G) Average peak height over epochs with all
responses normalized to the control peak height in the first epoch. (H) Average baseline over epochs with all responses normalized to the control peak height in
the first epoch. G and H follow format of D. All error bars indicate SEM.

© 2018 Federation of European Neuroscience Societies and John Wiley & Sons Ltd
European Journal of Neuroscience, 1–12
Orientation adaptation in mouse V1 7

© 2018 Federation of European Neuroscience Societies and John Wiley & Sons Ltd
European Journal of Neuroscience, 1–12
8 J. L. King and N. A. Crowder

(Fig. 3E bottom inset) and peak shifts (Fig. 3F bottom inset) of data from each epoch. We ran two-way repeated-measure ANOVAs
flank adapted cells show more variable adaptation effects with (with time epoch and adaptation as factors) to quantify changes in
higher elicited firing rates. Overall these additional analyses sug- DIo, peak location, peak height and baseline firing.
gested that the broader tuning of mouse V1 neurons may lead to lar- Figure 4A shows an example cell’s tuning curves averaged from
ger or more variable peak shifts in either direction. the entire 2-s test window, and Fig. 4B shows the eight 250-msec
epochs from the same cell. Both control and adapted tuning curves
changed in amplitude over time; however, the depth of tuning as
Dynamics of adaptation
measured by DIo (Fig. 4B; noted top left inset of each graph)
Finally, we considered the possibility that subtle transient adaptation remained relatively stable. For population data, control (filled cir-
effects could be masked by averaging spiking activity over the cles) and adapted (empty circles) DIo values in each epoch were
entire 2-s test. Therefore, we examined how control and adapted ori- normalized by the mean nonadapted DIo calculated using the fill 2-s
entation tuning curves evolved over time by parsing responses into interval (Fig. 4C). We found that DIo changed significantly over
250 ms epochs. We then calculated DIo and fit von Mises curves to epochs (Fig. 4C; F1,7, P = 0.0002); however, the effect size was

Fig. 5. Modelling adaptation in cats and mice. (A) Model of how orientation tuned inputs may be combined in a pinwheel centre of cat V1. The flow diagram
on the left shows inputs under control (solid lines) or adapted conditions (dashed lines) that are first depicted as populations of excitatory (black) and inhibitory
(grey) curves (left-most boxes), which are summed to produce pooled inputs (middle boxes). Inhibition acts divisively, and then, the total input is passed
through a power-law nonlinearity (right-most box) to simulate orientation tuning curves (graph on right). (B) Histogram of peak shifts from 200 simulated cat
V1 pinwheel centre neurons generated as in A. (C and D) Simulation of adaptation in an iso-orientation domain of cat V1 with the same format as A and B.
Model parameters for simulations in A–D were: n = 50; hadapt = 12.5; u = 6; g = 0.5; hnom = 0; h = 5; d = 0.7; c = 0.3; p = 2; q = 0.7; Vthresh = 0.9;
w = 1.5; k1 = 0.5; k2 = 0.007. (E and F) Simulation of adaptation in mouse V1 with salt-and-pepper functional architecture in the same format as A and B (see
Results). Model parameters for this simulation were the same as A–D except: u = 3; Vthresh = 0.3. (G) Scatter column graph of individual (grey dots) and popu-
lation mean (horizontal black lines) peak shift values generated when we systematically varied the spiking threshold (Vthresh) while keeping all other model
parameters fixed. (H) Mean tuning breadth (HWHM) decreased when Vthresh was increased, regardless of the bandwidth parameter of excitatory inputs (u; noted
inset). (I) Scatter column graph of peak shift values generated when we systematically varied u while keeping all other parameters fixed (format as in G). Error
bars indicate 95% confidence intervals in G–I.

© 2018 Federation of European Neuroscience Societies and John Wiley & Sons Ltd
European Journal of Neuroscience, 1–12
Orientation adaptation in mouse V1 9

small (partial g2 = 0.07), and the direction of change varied over Modelling results
time. There was no evidence to support an effect of adaptation on
We generated a simplified feed-forward model of orientation adapta-
DIo (F1,1, P = 0.74), or an interaction between time epoch and
tion that first captured several salient features described in cat V1;
adaptation (F1,7, P = 0.12). We were concerned that the periodic fir-
then, in altering its features to align more with mouse V1, we nar-
ing of phase-sensitive neurons to the 2 Hz drifting gratings might
rowed down critical species differences that could explain our data.
have introduced variability into our DIo measures, but excluding
Adjacent neurons within cat V1 can have vastly different orientation
these cells did not affect the outcome of this analysis. Thus, we
preferences depending on their position within the orientation map,
found that the depth of orientation tuning for control and adapted
but anatomical studies show that their local dendritic arbours have a
curves changed inconsistently over time, similar to what was
circularly symmetrical geometry that extends approximately 500–
reported by Tan et al. (2011) with intracellular recordings.
800 lm into surrounding cortex independent of the somatic position
The mean preferred orientations (hpref) of control (filled circles)
(Malach et al., 1993; Das & Gilbert, 1999; Levy et al., 2014). Find-
and adapted (empty circles) curves across epochs are shown in
ings of broader orientation tuning at pinwheel centres (Nauhaus
Fig. 4D. We aligned cells with different preferred orientations by
et al., 2008), and more repulsion following adaptation (Dragoi
setting each nonadapted preferred orientation in the first epoch to
et al., 2001), have been hypothesized to arise from local network
zero. There was no evidence that preferred orientation changed
inputs that include many different preferred orientations. Conversely,
over epochs (F1,7, P = 0.75), that adaptation consistently induced
the more homogenous local network inputs in iso-orientation
repulsive or attractive peak shifts (F1,1, P = 0.87), or of an interac-
domains are thought to give rise to narrower orientation tuning and
tion between epoch and adaptation (F1,7, P = 0.74). We also
little repulsion following adaptation. We simulated adaptation in cat
implemented a scoring system to determine whether individual
V1 cells located near a pinwheel centre by pooling excitatory inputs
neurons consistently showed repulsion or attraction over epochs by
with preferred orientations (hi) that varied widely ( 60–90°) from
assigning a value of 0 to each epoch with an attractive shift and 1
the simulated neuron’s preferred orientation (hnom). Figure 5A
to each epoch with a repulsive shift (Fig. 4E). Here, a hypothetical
shows the simulation of a single neuron that pools multiple inputs
neuron that showed repulsive peak shifts in all epochs would have
(left-most boxes) to generate an orientation tuned spiking response
a score of 8. We performed a Hartigan’s dip test on the scores
(graph on right; solid line) that undergoes a repulsive peak shift fol-
and rejected the null hypothesis of unimodality (P < 0.05; Hartigan
lowing adaptation (dotted line). Figure 5B shows a population peak-
& Hartigan, 1985), which suggests that most cells were fairly con-
shift histogram of 200 simulated neurons indicating repulsion is
sistent in their attraction or repulsion over epochs. Averaging
common under these model conditions. Individual simulated cat V1
across neurons in each epoch may have masked an interaction
neurons located near iso-orientation domains (hi  10–20° from
between the adaptor angle and peak shift, so Fig. 4F plots individ-
hnom) were also orientation tuned (Fig. 5C), but rarely showed
ual peak shifts in the first epoch (when adaptation is expected to
repulsion (Fig. 5D). The modelling above simulated the main results
be strongest) as a function of the angle between the adaptor and
of Dragoi et al. (2001), with repulsion at pinwheel centres but not
the nonadapted response peak in the same format as Fig. 2A. Like
at iso-orientation domains; however, our model also inherently
the data derived from the full 2-s window, both repulsive and
caused the nonadapted curves of pinwheel neurons to have ~18%
attractive shifts are evident, but there was no evidence of a main
broader orientation tuning than iso-orientation domain neurons,
effect of adaptation on peak shift (F1,54, P = 0.81) or a flank inter-
which simulated the results of Nauhaus et al. (2008). Note that in
action (F1,54, P = 0.81). Overall, the pattern of results for peak
Fig. 5A and C adaptation caused peak height to decrease (as in our
shifts evolving over time was similar to our time-averaged analysis
data; Wissig & Kohn, 2012; Patterson et al., 2013), but this could
in that individual cells could show repulsion or attraction but nei-
be prevented (to mimic the data of Dragoi et al., 2001) by changing
ther shift direction was dominant.
the gain of adaptation for inhibitory inputs and activity-dependent
The peak amplitude (a) and baseline (m) parameters of the tuning
hyperpolarization (data not shown).
curves followed the classic evolution of visually evoked responses
To simulate adaptation in mouse V1 neurons, we modified our
over time: they reached a transient peak early in the response before
model of cat V1 cells located at pinwheel centres. The salt-and-pep-
undergoing exponential decay (Albrecht et al., 1984). Figure 4G
per organization of mouse V1 suggests that local cortical networks
shows the mean peak amplitude of control (filled circles) and
here will also contain neurons with diverse orientation tuning like
adapted (empty circles) curves across epochs, and Fig. 4H shows
the pinwheel centres of cat V1 (Malach et al., 1993; Levy et al.,
the mean baseline data across the same time bins. For individual
2014; Fig. 5E, left-most boxes). To simulate the broader orientation
cells, both peak height and baseline were normalized by the peak
tuning and poorer modulation we observed in our mouse data, we
amplitude of the nonadapted curve in the first epoch. Population
made two changes to our model: (1) we lowered the spike threshold
means for both peak amplitude and baseline were maximal in the
(Vthresh) so that simulated spike-rate orientation curves had an
second epoch, which was consistent with the example cell in
untuned component (Fig. 5E, graph on right); and (2) we made the
Fig. 4B. Mean firing in the first epoch was likely lower because it
tuning of the excitatory inputs broader (i.e. smaller values of u).
included the latent period (median latency: 61 ms; interquartile
With these changes, the distribution of peak shifts in our simulated
range = 39–98 ms). Peak amplitude varied significantly with epoch
population became wider, there were more instances of attractive
(F1,7, P = 5.4 9 1011) and adaptation (F1,1, P = 3.1 9 1011);
peak shifts, and the mean peak shift was closer to zero (Fig. 5F).
however, the interaction did not reach significance (F1,7, P = 0.33).
Systematically changing the spike threshold while keeping all other
Baseline also varied significantly with epoch (F1,7, P = 1.6 9
parameters the same did not have any consistent effect on the distri-
1022) and adaptation (F1,1, P = 7.8 9 109), but here the interac-
bution of peak shifts (Fig. 5G), although raising the spike threshold
tion was also significant (F1,7, P = 1.7 9 1010) suggesting that for
did lead to narrower tuning (Fig. 5H). Conversely, systematically
this parameter the effect of adaptation was diminished towards the
increasing the tuning breadth of excitatory inputs while keeping all
end of the 2-s test period. Once again, the consistent effect of adap-
other parameters the same caused more instances of attraction and
tation on peak amplitude and baseline seen in the time-binned
the distribution of peak shifts to approach zero (Fig. 5I). Thus, our
results matched our time-averaged analysis closely.

© 2018 Federation of European Neuroscience Societies and John Wiley & Sons Ltd
European Journal of Neuroscience, 1–12
10 J. L. King and N. A. Crowder

model suggests that it was the broader orientation tuning of mouse attractive shifts were often accompanied by response facilitation
V1 neurons that led adaptation to produce relatively more attractive (Wissig & Kohn, 2012), whereas all of our cells showed suppres-
peak shifts in our data set than observed in other species. sion. Overall, our interleaved stimulus protocol with 2-s adaptors
targeting the classical receptive field could reasonably be predicted
to produce consistent repulsion in cats or primates, so the higher
Discussion
proportion of attraction in our data set appears to be a genuine spe-
The change in the orientation tuning of V1 neurons following pro- cies difference.
longed exposure to a single orientation is a well-known example of There are several differences in the way mouse V1 deals with ori-
adaptation that has been studied in classical animal models of vision entation signals that may underlie our observed differences in adap-
(cats: Priebe et al., 2010; Bachatene et al., 2012; Cattan et al., tation: (1) there is more sharpening of orientation selectivity
2014; monkeys: Wissig & Kohn, 2012; Patterson et al., 2013) and between the dLGN and V1 in cats (Scholl et al., 2013); (2) mice
is a potential neural correlate of the tilt-after-effect and related illu- have a greater proportion of neurons that are nonselective or poorly
sions (Gibson & Radner, 1937; Levinson & Sekuler, 1976; Patterson tuned for orientation (Niell & Stryker, 2008; Gao et al., 2010; Liu
& Becker, 1996; Schrater & Simoncelli, 1998; Clifford, 2002). We et al., 2011; Tan et al., 2011; Stroud et al., 2012); and (3) mice
used a comparative approach by studying orientation adaptation in have salt-and-pepper functional architecture rather than an ordered
mouse V1, which possesses orientation tuned neurons but differs orientation map (Ohki et al., 2005). Our model, which focused on
from classical models in its functional architecture and proportion of the latter two differences, suggested it was the broader orientation
highly orientation selective neurons. Unlike the mainly repulsive tuning of inputs to mouse V1 cells that could lead to our observed
shifts seen in past work, we found adaptation produced approxi- pattern of peak shifts. Additional evidence that tuning breadth can
mately a 2 : 1 proportion of repulsive vs. attractive shifts. Adapta- lead to species differences in adaptation comes from nonoriented V1
tion also consistently attenuated responses across all orientations. neurons. All of our nonoriented mouse cells showed broad response
We first consider whether these differences from past work reflect reduction following adaptation, as if they were fed by essentially
procedural variations or a true species difference and then discuss untuned inputs. Conversely, Wissig & Kohn (2012) showed that
implications for studying this phenomenon in mice. their mean population tuning curve of nonoriented primate V1 neu-
rons developed a ‘dent’ at the adapting angle, as if these cells were
pooling over multiple tightly tuned inputs.
Procedural vs. species differences
As described in the Results section, our sample of mouse V1 neu-
Studying adaptation to orientation in the mouse
rons was similar to previous mouse V1 surveys in terms of the pro-
portion of orientation selective units and tuning breadth. There is currently a poor understanding of what types of changes in
Approximately 11% of the units in our sample were phase sensi- neural responding underlie the perceptual tilt-after-effect illusion.
tive, which is higher than the 6% reported by Gao et al. (2010), Models of V1 adaptation where each orientation tuned cell acts as a
but lower than the ~35–50% reported in other studies (Niell & labelled line predict that repulsive neural peak shifts actually lead to
Stryker, 2008; Van den Bergh et al., 2010). However, both phase- perceptual attraction (e.g. Gilbert & Wiesel, 1990; Kohn & Mov-
sensitive and phase-invariant units showed adaptation in our sample shon, 2004), but decreases in gain, broadening of orientation tuning,
as well as in previous cat and primate work (Dragoi et al., 2001; adapting surround suppression, or processing in subsequent brain
Wissig & Kohn, 2012; Patterson et al., 2013), so the composition areas could counteract this effect (Clifford et al., 2001; Sur et al.,
of our sample seems unlikely to have led to the observed differ- 2002; Kohn & Movshon, 2004; Wissig & Kohn, 2012). There are
ences in adaptation. several potential strategies that could be used to clarify links
Our stimuli were also similar enough to previous work that is between neurophysiological and perceptual aspects of adaptation that
unlikely that the differences we observed were due to our adaptation require developing new methodologies.
protocol. Regarding stimulus timing, past work in cats and primates If the repulsive shifts in orientation tuning observed in past cat
has shown that flank adaptation within the classical receptive field and primate work are indeed the neural basis for the tilt-after-effect
has consistently yielded repulsive peak shifts with adaptation times illusion, our current electrophysiological findings suggest that this
ranging from 50 ms to 10 min (Dragoi & Sur, 2000; Felson et al., illusion should not occur in mice. This would mean that the genetic
2002; Patterson et al., 2013). Furthermore, repulsive shifts were tools available in mice, which are accelerating the visual neuro-
observed whether adaptation was induced with randomly interleaved science community’s ability to monitor large neural populations and
adaptation/control trials or with a top-up protocol (Patterson et al., dissect neural circuits (Baker, 2013), could not be brought to bear
2013). However, several studies in opioid- anaesthetized macaques on this psychophysically well characterized illusion (Gibson & Rad-
have shown that the spatial parameters of the stimulus protocol can ner, 1937; Levinson & Sekuler, 1976; Patterson & Becker, 1996;
affect the type of adaptation observed. Large stimuli that extend into Schrater & Simoncelli, 1998; Clifford, 2002). Thus, it seems the
the extra-classical receptive fields of most V1 neurons produced most straightforward way to determine whether this line of research
attractive rather than repulsive peak shifts (Kohn & Movshon, 2004; should continue in the mouse is to psychophysically test whether
Wissig & Kohn, 2012; Patterson et al., 2013). We are sceptical that mice perceive the tilt-after-effect. This proposal is challenging on
this extra-classical effect was the source of attractive shifts in our several fronts. Whereas other fields of neuroscience have established
data set for three reasons. First, our online analysis helped us limit behavioural tests in rodents that are mediated by vision (e.g. ele-
the adaptation stimuli to the classical receptive field. Second, it has vated plus maze for anxiety: Pellow et al., 1985; Barnes maze for
been reported that isoflurane/urethane anaesthesia decreases surround memory and spatial learning: Barnes, 1979), rigorous psychophysi-
suppression in mice (Adesnik et al., 2012; Vaiceliunaite et al., cal tests to probe perceptual decision making are not yet as devel-
2013), so if the adaptation stimuli did inadvertently extend beyond oped in mice as they are in primates (Carandini & Churchland,
the receptive field any extra-classical modulation would have been 2013; c.f. Busse et al., 2011; Histed et al., 2012). So far, it has
weakened by our isoflurane anaesthesia. Third, in the primate work, been shown that mice are able to coarsely discriminate different

© 2018 Federation of European Neuroscience Societies and John Wiley & Sons Ltd
European Journal of Neuroscience, 1–12
Orientation adaptation in mouse V1 11

orientations (Reuter, 1987; Andermann et al., 2010; Pinto et al., Benjamini, Y. & Hochberg, Y. (1995) Controlling the false discovery rate: a
2013; Long et al., 2015); however, these types of discrimination practical and powerful approach to multiple testing. J. Roy. Stat. Soc. B
Met., 57, 289–300.
task would need to be altered to work in an adaptation paradigm. Brainard, D.H. (1997) The psychophysics tookbox. Spat. Vis., 10, 433–436.
Linking neural activity with perception will be even more challeng- Busse, L., Ayaz, A., Dhruv, N.T., Katzner, S., Saleem, A.B., Sch€ olvinck,
ing. For comparison, there is a much longer history of testing fine- M.L., Zaharia, A.D. & Carandini, M. (2011) The detection of visual con-
grain perceptual discrimination in primates (e.g. Newsome & Pare, trast in the behaving mouse. J. Neurosci., 31, 11351–11361.
1988), but primate studies of adaptation that examine both a change Carandini, M. (2000) Visual cortex: fatigue and adaptation. Curr. Biol., 10,
R605–R607.
in neural activity and a change in perception/behaviour are quite rare Carandini, M. & Churchland, A.K. (2013) Probing perceptual decisions in
(Kohn, 2007; Yang & Lisberger, 2009). In fact, although neural rodents. Nat. Rev. Neurosci., 16, 824–831.
adaptation to orientation has been studied in awake primates (Gut- Carandini, M. & Heeger, D.J. (2012) Normalization as a canonical neural
nisky & Dragoi, 2008; Wang et al., 2011), we are not aware of any computation. Nat. Neurosci., 13, 51–62.
Cattan, S., Bachatene, L., Bharmauria, V., Jeyabalaratnam, J., Milleret, C. &
work that has linked altered perception of orientation with changes Molotchnikoff, S. (2014) Comparative analysis of orientation maps in
in neural encoding. In summary, whereas our electrophysiological areas 17 and 18 of the cat primary visual cortex following adaptation. Eur.
data suggest that neurons across mouse V1 adapt to orientation dif- J. Neurosci., 40, 2554–2563.
ferently than V1 neurons of primates or cats, deciding whether or Clifford, C.W.G. (2002) Perceptual adaptation: motion parallels orientation.
not these differences are actually important will require the develop- Trends. Cogn. Sci., 6, 136–143.
Clifford, C.W.G., Wyatt, A.M., Arnold, D.H., Smith, S.T. & Wenderoth, P.
ment of novel psychophysical tests in multiple species to link neural (2001) Orthogonal adaptation improves orientation discrimination. Vision
adaptation with perception. Res., 41, 151–159.
Cossell, L., Iacaruso, M.F., Muir, D.R., Houlton, R., Sader, E.N., Hofer,
S.B. & Mrsic-Flogel, T.D. (2015) Functional organization of excitatory
Acknowledgements synaptic strength in primary visual cortex. Nature, 518, 399–403.
Das, A. & Gilbert, C.D. (1999) Topography of contextual modulations medi-
The authors thank Matthew P. Lowe for assistance with data collection. This ated by short-range interactions in primary visual cortex. Nature, 399,
work was supported by the Natural Sciences and Engineering Research 655–661.
Council of Canada. Dhruv, N.T., Tailby, C., Sokol, S. & Lennie, P. (2011) Multiple adaptable
mechanisms early in the primate visual pathway. J. Neurosci., 31, 15016–
15025.
Conflicts of interest Dr€ager, U.C. (1975) Receptive fields of single cells and topography in mouse
The authors declare that they have no conflicts of interest in association with visual cortex. J. Comp. Neurol., 16, 269–290.
this work. Dragoi, V. & Sur, M. (2000) Adaptation-induced plasticity of orientation tun-
ing in adult visual cortex. Neuron, 28, 287–298.
Dragoi, V., Rivadulla, C. & Sur, M. (2001) Foci of orientation plasticity in
Author contributions visual cortex. Nature, 411, 80–86.
Felson, F., Shen, Y.S., Yao, H., Spor, G., Li, C. & Dan, Y. (2002) Dynamic
J.L.K. and N.A.C. collected and analysed data; J.L.K. and N.A.C. interpreted modification of cortical orientation tuning mediated by recurrent connec-
results of the experiments; J.L.K. and N.A.C. wrote and revised the manu- tions. Neuron, 36, 945–954.
script; N.A.C. conceived and designed the research; J.L.K. prepared figures; Gao, E., DeAngelis, G.C. & Burkhalter, A. (2010) Parallel input channels to
J.L.K. and N.A.C. approved the final version of the manuscript. mouse primary visual cortex. J. Neurosci., 30, 5912–5926.
Ghisovan, N., Nemri, A., Shumikhina, S. & Molotchinkoff, S. (2009) Long
adaptation reveals mostly attractive shifts of orientation tuning in cat pri-
Data accessibility mary visual cortex. Neuroscience, 164, 1274–1283.
Gibson, J. & Radner, M. (1937) Adaptation, after-effect and contrast in the
Access to supporting data and materials may be arranged by contacting the perception of tilted lines. J. Exp. Psychol., 12, 453–467.
corresponding author. Gilbert, C.D. & Wiesel, T.N. (1990) The influence of contextual stimuli on
the orientation selectivity of cells in primary visual cortex of the cat.
Vision Res., 30, 1689–1701.
Abbreviations Gutnisky, D.A. & Dragoi, V. (2008) Adaptive coding of visual information
DIgrating, discrimination index for gratings; DIo, discrimination index for ori- in neural populations. Nature, 452, 220–224.
entation; dLGN, dorsal lateral geniculate nucleus; HWHM, half-width at half Hartigan, J.A. & Hartigan, P.M. (1985) The dip test of unimodality. Ann.
maximum; ISI, interstimulus interval; RF, receptive field; SF, spatial fre- Statist., 13, 70–84.
quency; TF, temporal frequency; TTL, transistor-transistor logic; V1, primary Histed, M.D., Carvalho, L.A. & Maunsell, J.H.R. (2012) Psychophysical
visual cortex. measurement of contrast sensitivity in the behaving mouse. J. Neurophys-
iol., 107, 758–765.
References Kerlin, A.M., Andermann, M.L., Berezovskii, V.K. & Reid, R.C. (2010)
Broadly tuned response properties of diverse inhibitory neuron subtypes in
Adesnik, H., Bruns, W., Taniguchi, H., Huang, Z.H. & Scanziani, M. (2012) mouse visual cortex. Neuron, 67, 858–871.
A neural circuit for spatial summation in visual cortex. Nature, 490, King, J.L., Lowe, M.P., Stover, K.R., Wong, A.A. & Crowder, N.A. (2016)
226–231. Adaptive processes in thalamus and cortex revealed by silencing of pri-
Albrecht, D.G., Farrar, S.B. & Hamilton, D.B. (1984) Spatial contrast adapta- mary visual cortex during contrast adaptation. Curr. Biol., 26, 1295–1300.
tion characteristics of neurones recorded in the cat’s visual cortex. J. Phys- Kohn, A. (2007) Visual adaptation: physiology, mechanisms, and functional
iol., 347, 713–739. benefits. J. Neurophysiol., 97, 3155–3164.
Andermann, M.L., Kerlin, A.M. & Reid, R.C. (2010) Chronic cellular imag- Kohn, A. & Movshon, J.A. (2004) Adaptation changes the direction tuning
ing of mouse visual cortex during operant behavior and passive viewing. of macaque MT neurons. Nat. Neurosci., 7, 764–772.
Front. Cell. Neurosci., 4, 3. Kuhkmann, L. & Vidyasagar, T.R. (2011) A computational study of how
Bachatene, L., Bharmauria, V., Rouat, J. & Molotchnikoff, S. (2012) Adapta- orientation bias in the lateral geniculate nucleus can give rise to orienta-
tion-induced plasticity and spike waveforms in cat visual cortex. Neurore- tion selectivity in the primary visual cortex. Front. Syst. Neurosci., 5, 81–99.
port, 23, 88–92. LeDue, E.E., Zou, M.Y. & Crowder, N.A. (2012) Spatiotemporal tuning in
Baker, M. (2013) Through the eyes of a mouse. Nature, 502, 156–158. mouse primary visual cortex. Neurosci. Lett., 528, 165–169.
Barnes, C.A. (1979) Memory deficits associated with senescence: a neuro- LeDue, E.E., King, J.L., Stover, K.R. & Crowder, N.A. (2013) Spatiotempo-
physiological and behavioral study in the rat. J. Comp. Physiol. Psych., ral frequency of contrast adaptation in mouse visual cortex. Front. Neural
93, 74–104. Circuit., 7, 154–165.

© 2018 Federation of European Neuroscience Societies and John Wiley & Sons Ltd
European Journal of Neuroscience, 1–12
12 J. L. King and N. A. Crowder

Levinson, E. & Sekuler, R. (1976) Adaptation alters perceived direction of Reuter, J.H. (1987) Tilt discrimination in the mouse. Behav. Brain Res., 24,
motion. Vision Res., 16, 779–781. 81–84.
Levy, M., Lu, Z., Dion, G. & Kara, P. (2014) The shape of dendritic arbors Ringach, D.L., Mineault, P.J., Tring, E., Olivas, N.D., Garcia-Junco-Clem-
in different functional domains of the cortical orientation map. J. Neu- ente, P. & Trachtenberg, J.T. (2016) Spatial clustering of tuning in mouse
rosci., 34, 3231–3236. primary visual cortex. Nat. Commun., 7, 12270–12279.
Li, Y., Ibrahim, L.A., Liu, B., Zhang, L.I. & Tao, W.H. (2013) Linear trans- Runyan, C.A., Schummers, J., Van Wart, A., Kuhlman, S.J., Wilson, N.R.,
formation of thalamocortical input by intracortical excitation. Nat. Neu- Huang, Z.J. & Sur, M. (2010) Response features of parvalbumin-expres-
rosci., 16, 1324–1330. sing interneurons suggest precise roles for subtypes of inhibition in visual
Lien, A. & Scanziani, M. (2013) Tuned thalamic excitation is amplified by cortex. Neuron, 67, 847–857.
visual cortical circuits. Nat. Neurosci., 16, 1315–1323. Scholl, B., Burge, J. & Priebe, N.J. (2013) Binocular integration and dispar-
Liu, B., Li, P., Li, Y., Sun, Y.J., Yanagawa, Y., Obata, K., Zhang, L.I. & ity selectivity in mouse primary visual cortex. J. Neurophysiol., 109,
Tao, H.W. (2009) Visual receptive field structure of cortical inhibitory 3013–3024.
neurons revealed by two-photon imaging guided recording. J. Neurosci., Schrater, P. & Simoncelli, E. (1998) Local velocity representation: evidence
29, 10520–10532. from motion adaptation. Vision Res., 38, 3899–3912.
Liu, B., Li, Y., Ma, W., Pan, C., Zhang, L. & Tao, H.W. (2011) Broad inhi- Skottun, B.C., De Valois, R.L., Grosof, D.H., Movshon, J.A., Albrecht, D.G.
bition sharpens orientation selectivity by expanding input dynamic range & Bonds, A.B. (1991) Classifying simple and complex cells on the basis
in mouse simple cells. Neuron, 71, 542–554. of response modulation. Vision Res., 31, 1079–1086.
Long, M., Jiang, W., Liu, D. & Yao, H. (2015) Contrast-dependent orienta- Smith, E.L., Chino, Y.M., Ridder, W.H., Kitagawa, K. & Langson, A.
tion discrimination in the mouse. Sci. Rep., 5, 15830. (1990) Orientation bias of neurons in the lateral geniculate nucleus of
Malach, R., Amir, Y., Harel, A.M. & Grinvald, A. (1993) Relationship macaque monkeys. Visual Neurosci., 5, 525–545.
between intrinsic connections and functional revealed by optical imaging Sohya, K., Kameyama, K., Yanagawa, Y., Obata, K. & Tsumoto, T.
and in vivo targeted biocytin injections in primate cortex. P. Natl. Acad. (2007) GABAergic neurons are less selective to stimulus orientation
Sci., 90, 10469–10473. than excitatory neurons in layer II/III of visual cortex, as revealed by
Metin, C., Godment, P. & Imbert, M. (1988) The primary visual cortex in in vivo functional Ca2 + imaging in transgenic mice. J. Neurosci., 27,
the mouse: receptive field properties and functional organization. Exp. 2145–2149.
Brain Res., 69, 594–612. Stroud, A.C., LeDue, E.E. & Crowder, N.A. (2012) Orientation specificity of
Naito, T., Okamoto, M., Sadakane, O., Shimegi, S., Osaki, H., Hara, S., contrast adaptation in mouse primary visual cortex. J. Neurophysiol., 108,
Kimura, A., Ishikawa, A. et al. (2013) Effects of stimulus spatial frequency, 1381–1391.
size, and luminance contrast on orientation tuning of neurons in the dorsal Sur, M., Schummers, J. & Dragoi, V. (2002) Cortical plasticity: time for a
lateral geniculate nucleus of the cat. Neurosci. Res., 77, 143–154. change. Curr. Biol., 12, R168–R170.
Nauhaus, I., Benucci, A., Carandini, A. & Ringach, D.L. (2008) Neuronal Swindale, N.V. (1998) Orientation tuning curves: empirical description and
selectivity and local map structure in visual cortex. Neuron, 57, 673–679. estimation of parameters. Biol. Cybern., 78, 45–56.
Newsome, W.T. & Pare, E.B. (1988) A selective impairment of motion per- Tan, A.Y.Y., Brown, B.D., Scholl, B., Mohanty, D. & Priebe, N.J. (2011)
ception following lesions of the middle temporal visual area (MT). J. Neu- Orientation selectivity of synaptic input to neurons in mouse and cat pri-
rosci., 8, 2201–2211. mary visual cortex. J. Neurosci., 31, 12339–12350.
Niell, C. & Stryker, M. (2008) Highly selective receptive fields in mouse Vaiceliunaite, A., Erisken, S., Franze, F., Katzner, S. & Busse, L. (2013)
visual cortex. J. Neurosci., 28, 7520–7536. Spatial integration in mouse primary visual cortex. J. Neurophysiol., 110,
Ohki, K., Chung, S., Ch’ng, Y.H., Kara, P. & Reid, R.C. (2005) Functional 964–972.
imaging with cellular resolution reveals precise micro-architecture in visual Van den Bergh, G., Zhang, B., Arckens, L. & Chino, Y.M. (2010) Recep-
cortex. Nature, 433, 597–603. tive-field properties of V1 and V2 neurons in mice and macaque monkeys.
Patterson, R. & Becker, S. (1996) Direction-selective adaptation and simulta- J. Comp. Neurol., 518, 2051–2070.
neous contrast induced by stereoscopic (cyclopean) motion. Vision Res., Wagor, E., Mangini, N.J. & Pearlman, A.L. (1980) Retinotopic organization
36, 1773–1781. of striate and extrastriate visual cortex in the mouse. J. Comp. Neurol.,
Patterson, C.A., Wissig, S.C. & Kohn, A. (2013) Distinct effects of brief and 193, 187–202.
prolonged adaptation on orientation tuning in primary visual cortex. J. Wang, Q. & Burkhalter, A. (2007) Area map of mouse visual cortex. J.
Neurosci., 33, 532–543. Comp. Neurol., 502, 339–357.
Paxinos, G. & Franklin, K.B. (2001). The Mouse Brain in Stereotaxic Coor- Wang, Y., Iliescu, B.F., Ma, J., Josic, K. & Dragoi, V. (2011) Adaptive
dinates. Academic, New York. changes in neuronal synchronization in macaque V4. J. Neurosci., 37,
Pelli, D.G. (1997) The video toolbox software for visual psychophysics: 13204–13213.
transforming numbers into movies. Spat. Vis., 10, 437–447. Wissig, S.C. & Kohn, A. (2012) The influence of surround suppression on
Pellow, S., Chopin, P., File, S.E. & Briley, M. (1985) Validation of open:- adaptation effects in primary visual cortex. J. Neurophysiol., 107, 3370–
closed arm entries in an elevated plus-maze as a measure of anxiety in the 3384.
rat. J. Neurosci. Meth., 14, 149–167. Wong, A.A. & Brown, R.E. (2006) Visual detection, pattern discrimina-
Pinto, L., Goard, M.J., Estandian, D., Xu, M., Kwan, A.C., Lee, S.H., tion and visual acuity in 14 strains of mice. Genes Brain Behav., 5,
Harrison, T.C., Feng, G. et al. (2013) Fast modulation of visual perception 389–403.
by basal forebrain cholinergic neurons. Nat. Neurosci., 16, 1857–1863. Xu, X., Ichida, J., Shostak, Y., Bonds, A.B. & Casagrande, V.A. (2002) Are
Piscopo, D.M., El-Danaf, R.N., Huberman, A.D. & Niell, C.M. (2013) primate lateral geniculate nucleus (LGN) cells really sensitive to orienta-
Diverse visual features encoded in mouse lateral geniculate nucleus. J. tion or direction? Visual Neurosci., 19, 97–108.
Neurosci., 33, 4642–4656. Yang, J. & Lisberger, S.G. (2009) Relationship between adapted neural
Priebe, N.J. & Ferster, D. (2008) Inhibition, spike threshold, and stimulus population responses in MT and motion adaptation in speed and direc-
selectivity in primary visual cortex. Neuron, 57, 482–497. tion of smooth-pursuit eye movements. J. Neurophysiol., 101, 2693–
Priebe, N.J., Lampl, I. & Ferster, D. (2010) Mechanisms of direction selec- 2707.
tivity in cat primary visual cortex has revealed by visual adaptation. J. Zhao, X., Chen, H., Liu, X. & Cang, J. (2013) Orientation-selective responses
Neurophysiol., 104, 2615–2623. in the mouse lateral geniculate nucleus. J. Neurosci., 33, 12751–12763.

© 2018 Federation of European Neuroscience Societies and John Wiley & Sons Ltd
European Journal of Neuroscience, 1–12

View publication stats

You might also like