546 | Nature | Vol 582 | 25 June 2020
Article
These results show that L2/3 excitatory neurons are excited by
both classical and inverse stimuli centred on their ffRF. Classical and
inverse stimuli were mutually antagonistic—the responses to both
stimuli together (for example, full-field gratings) were smaller than the
responses to either of the stimuli alone (Fig. 1c). We defined a neuron as
inverse-tuned if its response to at least one inverse stimulus of any size
centred on its ffRF was larger than its response to a full-field stimulus
(Methods). Of the visually responsive excitatory neurons in L2/3, 79%
were inverse-tuned (943 of 1,190 neurons in 9 mice) (Fig. 1a–c, Extended
Data Figs. 2, 3a). We then computed the inverse-tuning index (ITI),
in which 0.5 denotes an equal response to both classical and inverse
stimuli of the preferred size; 0 indicates a response to classical stimuli
only and 1 indicates a response to inverse stimuli only (Methods). The
ITI distribution of L2/3 excitatory neurons was unimodal, with a mean
of 0.52 ± 0.01 (mean ± s.e.m.) (Fig. 1c).
Surrounding the ffRF of inverse-tuned neurons is a region that is
either suppressive or excitatory depending on whether it is stimulated
in the presence or in the absence of a stimulus in the ffRF, respectively.
We used classical and inverse stimuli to compare the tuning properties
of the surrounding excitatory region with those of the ffRF. On average,
the orientation tuning to inverse stimuli was sharper than that to clas-
sical stimuli (Extended Data Fig. 3b–f ). Moreover, individual neurons
were not necessarily tuned to the same orientation when stimulated by
classical or inverse stimuli (Extended Data Fig. 3g). We determined the
interaction between the surrounding region and the ffRF in neurons
with a similar orientation preference by independently varying the
contrast of simultaneously presented classical and inverse stimuli. At
matching contrasts greater than 13%, the interaction between the sur-
rounding region and the ffRF was antagonistic (Extended Data Fig. 3h).
To address whether inverse tuning in L2/3 is inherited from earlier
stages of cortical processing, we measured the responses of excita-
tory neurons in layer 4 (L4) to classical and inverse stimuli (Fig. 1d–f,
Extended Data Fig. 3i). In contrast to L2/3 neurons, the suppressive
regions of L4 neurons surrounded their ffRF, creating a ring around
the centre (Fig. 1e; the absence of a suppressive ring around the ffRF
of L2/3 neurons was not due to insufficient spatial resolution of the
mapping stimuli; Extended Data Fig. 4). Further, the responses of L4
neurons to inverse stimuli placed on the centre of their ffRF decreased
monotonically with stimulus size, which is consistent with the progres-
sive reduction of feedforward drive (Extended Data Fig. 2) and is again
different from L2/3 neurons (compare Fig. 1c and Fig. 1f). Overall, the
spatial organization of suppressive regions of L4 neurons is consistent
with previous models and observations^7 ,^13 and is distinct from that of
L2/3 neurons. Inverse tuning in L2/3 excitatory neurons is therefore
not inherited from L4 neurons.
Sources of input to L2/3 neurons are specific to the neuron type^11 ,^14.
To address whether inverse tuning is also present in L2/3 inhibi-
tory neurons, we characterized the responses of the three major
classes of cortical inhibitory neurons—parvalbumin-expressing
(PV), vasoactive-intestinal-peptide-expressing (VIP) and
somatostatin-expressing (SOM) neurons—to classical and inverse
stimuli (Fig. 2 , Extended Data Fig. 3j–l). Both PV and VIP neuronsL2/3Inverse stimulusClassical stimulusClassical InverseClassical
Inverse...
......
...L2/3
L4
L5/6
GCaMP6f | 7f
GADcre x tdTomatoResponse (norm.ΔF/F)Stimulus size (°)L2/30.4ΔF/F
4 s 525456585...
......
...(^00309600)
0.8
0.4
ab
L4
0
0.6
0.4
0.2
L2/3L4
L5/6
L2/3
L4
f
15°
Response(norm.
ΔF
/F)
Response(norm.
ΔF
/F)
2 Δ
F/F
2 s
15°
15°
Classical Inverse
15°
4 Δ
F/F
4 s
15°
15°
6 Δ
F/F
2 s
L4
c
d e
Flex-GCaMP6fScnn1a-Cre
0.9
0.3
0.6
0
0.2
0.4
0.6
0
V1
V1
Response (norm.
ΔF
/F)
Stimulus size (°)
(^03) Stimulus size (°) 0960 0
(^5254) Stimulus size (°) 56585
ClassicalInverse
ClassicalInverse
ClassicalInverse
ClassicalInverse
(^001)
0.6
FractionITI
(^001)
0.18
FractionITI
"
Fig. 1 | Layer-specif ic responses to inverse stimuli. a, Experimental
configuration: Two-photon calcium imaging in excitatory L2/3 neurons
(green triangles) of awake head-fixed mice while presenting classical and
inverse stimuli. b, Top, example trial-averaged responses of an excitatory
L2/3 neuron for each stimulus location. Here and in all figures, shaded
areas represent stimulus presentation periods. Bottom, population-averaged
receptive fields aligned to the centre of the classical ff RF (2,601 excitatory
L2/3 neurons in 9 mice). c, Top, example trial-averaged responses of a
L2/3 neuron for each stimulus diameter. Stimuli are centred on the ff RF.
Bottom, population-averaged size-tuning functions, normalized to the
maximum response to classical stimuli. In all figures, solid lines show fits to the
data and triangles indicate the median preferred size. The inset shows the ITI
distribution of L2/3 excitatory neurons, with a median value of 0.54 as
indicated by the triangle (the dashed line indicates 0.5; 1,190 excitatory L2/3
neurons in 9 mice). d, Schematic of results and experimental configuration.
Imaging in excitatory L4 neurons reveals that inverse-tuning in excitatory L2/3
neurons is not simply inherited from L4 neurons. e, f, Same as b, c, but for L4
excitatory neurons. The results in e are obtained from 24 neurons in 4 mice. In f,
the ITI for L4 neurons is shown in black (median 0.053; 35 neurons in 6 mice),
with the results for L2/3 neurons (c) shown in grey. Comparison was performed
using a two-sided Wilcoxon rank-sum test; ***P = 1. 5 × 10−18. In all plots, traces or
data points represent the mean and shading or error bars show the s.e.m.
PV
Stimulus size (°)
ClassicalInverse
0
030960 0
0.8
0.4
a Classical Inverse
15°
0.3
0.9
0.6
0
VIP
Stimulus size (°)
0
030960 0
0.8
0.4
Classical Inverse c
15°
0.2
0.4
0.6
0
SOM
Stimulus size (°)
(^00309600)
0.6
0.3
Classical Inverse
15°
0.4
0.8
0
1.2
b
0.3
Fraction (^00) ITI 1
0.25
Fraction (^00) ITI 1
0.45
Fraction (^00) ITI 1
NS
Response (norm.
ΔF
/F)
Response(norm.
ΔF/
F)
Fig. 2 | Neuron-type specif ic response to inverse stimuli. a, Top,
population-averaged receptive field for PV neurons aligned to the centre of
ff RF (82 neurons in 6 mice). Bottom, population-averaged size-tuning
functions, normalized to the maximum response to classical stimuli. The inset
shows the ITI distribution, with a median of 0.53 as indicated by the triangle
(60 neurons in 7 mice). The results for L2/3 neurons are shown in grey (Fig. 1c),
and comparison was performed using a two-sided Wilcoxon rank-sum test; NS,
P = 0.79. b, Same as in a but for VIP neurons. Top, results from 126 neurons in
4 mice. The median ITI (bottom, inset) is 0.42 (74 neurons in 8 mice). The results
for L2/3 neurons are shown in grey (Fig. 1c), and comparison was performed
using a two-sided Wilcoxon rank-sum test; P = 1 .1 × 10−5. c, Same as in a but for
SOM neurons. Top, results from 315 neurons in 5 mice. The median ITI (bottom,
inset) is 0.12 (179 neurons in 9 mice). The results for L2/3 neurons are shown in
grey (Fig. 1c), and comparison was performed using a two-sided Wilcoxon
rank-sum test; ***P = 1 .1 × 10−42.