Nature | Vol 582 | 25 June 2020 | 547showed surround suppression to classical stimuli, consistent with pre-
vious reports^11 ,^15. Furthermore, both PV and VIP neurons responded to
inverse stimuli centred on their ffRF (Fig. 2a, b) and showed size-tuning
functions to inverse stimuli that peaked well above their responses to
the largest classical stimuli. By contrast, SOM neurons showed almost
no surround suppression to classical stimuli^11 ,^15 , poor and spatially dif-
fuse responses to inverse stimuli (Fig. 2c), and none of their responses
to inverse stimuli was greater than their response to the largest clas-
sical stimuli. Therefore, similar to L2/3 excitatory neurons, most PV
and VIP neurons were inverse-tuned whereas SOM neurons were not
(Fig. 2 ).
Although L2/3 neurons do not directly inherit inverse tuning from
L4 neurons, the excitatory region surrounding their ffRF could still be
generated via feedforward inputs from L4 neurons with spatially offset
ffRFs (Fig. 3a). Alternatively, the excitatory region surrounding the ffRFs
of L2/3 neurons can be generated via feedback projections^16 ,^17. We there-
fore compared the latency of the responses of inverse-tuned neurons
to classical and inverse stimuli using extracellular electrophysiological
recordings (Fig. 3a) and isolated single units throughout cortical layers,
including infragranular layers (L5/6). A large fraction of infragranular
units (50%, 60 of 119 units) were also inverse-tuned (Extended Data
Fig. 5). In inverse-tuned units recorded in both supra- and infragranu-
lar layers, the time course of the responses to classical and inverse
stimuli were markedly different (Fig. 3b, c). Whereas the response to
classical stimuli showed a fast initial transient followed by a plateau,
the response to inverse stimuli slowly progressed towards steady state
(Fig. 3c–e) and was delayed relative to the classical response (50 ± 20 ms;
mean ± s.e.m.; 15 units) (Fig. 3f). The same biases were observed when
comparing the response dynamics to classical stimuli in all responsive
units with those to inverse stimuli in inverse-tuned units (Fig. 3g). The
difference in latencies and the slower dynamics of responses to inverse
stimuli suggest that the excitatory region surrounding the ffRF of L2/3
neurons is unlikely to emerge from the feedforward pathway.
We next determined whether responses to inverse stimuli depend
on feedback projections from higher visual areas (HVAs). The effect of
anaesthesia on sensory responses has been proposed to be stronger
in HVAs than in V1^18. We therefore compared the effect of isoflurane
on the responses to classical stimuli in V1 and in HVAs that had been
identified beforehand using wide-field intrinsic imaging (Extended
Data Fig. 6a–c). Anaesthesia suppressed responses to a greater extent
in HVAs than in V1 (Extended Data Fig. 6d–f ). If the responses of L2/3
neurons to inverse stimuli rely on feedback projections, they should be
more sensitive to anaesthesia than the responses to classical stimuli.
Consistent with this, anaesthesia preferentially suppressed responses
to inverse rather than classical stimuli in inverse-tuned neurons (Fig. 4 ),>1,6004 Hz
0.1 sClassicalInversebL2/3
L4
L5/6V1L4L2/3L5/6ClassicalInverse01,500**Onset slope (Hz s–1)Rise time (s)
0*Delay (s)
00.25*Classical Inverse00.10.20.3(^00) Classical response delay (s).1 0.20.3
Inverse response delay (s)
g
Δt
L2/3L5/6
0
0.1
0.2
0.3
(^0) Classical rise time (s)0.1 0.2 0.3
Inverse rise time (s)
f
L2/3L5/6
e
0
400
800
1,200
0 400 800 1,200
Classical onset slope (Hz s–1)
Inverse onset slope (Hz s
–1) L2/3L5/6
d
0.4 (norm.)
0.1 s Classical
Inverse
a c
Δt
Δt
Δt
??
0.3
0.25
0.3
Fig. 3 | Slow and delayed responses to inverse stimuli. a, Schematic of results
and experimental configuration for extracellular recordings in V1.
b, Responses of a L5/6 unit to classical and inverse stimuli centred on its ff RF.
Top, raster plot (1,000 trials for each stimulus); bottom, peristimulus time
histogram (PSTH; 10-ms bins). c, Population-averaged PSTHs normalized to
the average response to classical stimuli (15 units in 4 mice). d, Onset slope
of the response to classical and inverse stimuli. The green symbol represents
the example unit from b. The inset shows the PSTH from b to illustrate the
difference in slope between the responses to classical and inverse stimuli.
The dotted lines indicate the lower and upper thresholds used to compute the
slopes. Comparisons were performed using a two-sided Wilcoxon signed-rank
test; P = 1 .8 × 10−4. e, Same as d but for response rise time. Results are for 15 units
in 4 mice. Two-sided Wilcoxon signed-rank test; P = 7. 8 × 10−3. f, Same as d but for
response delay. Comparisons were performed using a two-sided Wilcoxon
signed-rank test; P = 9.8 × 10−4. g, Mean onset slopes (left), rise times (middle)
and delays (right) of independently tuned units. Classical, 51 units in 8 mice;
inverse, 29 units in 8 mice. Two-sided Wilcoxon rank-sum test; onset slope,
*P = 1 .1 × 10−3; rise time, P = 0.017; delay, *P = 0.031.
Iso urane
1 ΔF
4 s/F
0
1.2
0.4
0.8
0
0.4
0 0.2
0.2
0.4 >0. 5
0.5
Inverse peak response (
ΔF
/F)
Awake
Anaesthetized
ab c
0
0.4
0 0.2
0.2
0.4 >0.5
0.5
Response (norm.
ΔF
/F)R
esponse (norm.
ΔF
/F)
525456585
525456585
Stimulus size (°)
Inverse peak response (
ΔF
/F)
Stimulus size (°) Classical peak response (ΔF/F)
Stimulus size (°) Stimulus size (°) Classical peak response (ΔF/F)
030960 0
030960 0
ClassicalInverse
1.6
0
1.2
0.4
0.8
1.6
1 Δ
F/F
4 s
0.2
Fraction (^00) ITI 1
0.2
Fraction (^00) ITI 1
Classical
Inverse
Fig. 4 | Anaesthesia preferentially reduces responses to inverse stimuli.
a, Top, responses of an example neuron to classical and inverse stimuli in an
awake mouse. Bottom, responses of the same example neuron under
isoflurane anaesthesia. b, Population-averaged size-tuning functions in awake
(top) and anaesthetized (bottom) mice, normalized to the maximum awake
response to classical stimuli. The insets show the ITI distributions with median
values of 0.50 (top) and 0.32 (bottom). In the bottom inset, black represents
neurons from anaesthetized mice and grey from awake mice, compared
using a two-sided Wilcoxon signed-rank test; ***P = 1. 5 × 10−5. The same 49
excitatory L2/3 neurons were measured for both, in 5 mice. c, Peak responses
of inverse-tuned neurons in awake (top) and anaesthetized (bottom) mice. The
green symbol represents the example neuron from a. The same 49 excitatory
L2/3 neurons in 5 mice were measured as in b. Peak responses to classical and
inverse stimuli were compared using a two-sided Wilcoxon signed-rank test;
top, P = 0.96; bottom, P = 2 .0 × 10−5.