548 | Nature | Vol 582 | 25 June 2020
Article
suggesting that the response to inverse stimuli may be driven by feed-
back projections from HVAs.
To directly test the involvement of feedback projections in inverse
tuning, we silenced HVAs by scanning them with a laser to optoge-
netically activate inhibitory neurons while recording extracellular
electrophysiological activity in V1 (Fig. 5a–e, Extended Data Fig. 7a–
c). In inverse-tuned units, silencing HVAs reduced both spontane-
ous activity and responses to small-diameter classical stimuli. In
surround-suppressed units, silencing HVAs increased responses to
large-diameter classical stimuli (Fig. 5c, d, Extended Data Fig. 8), as has
previously been shown^19 –^21. The response to inverse stimuli, however,
was strongly suppressed (Fig. 5c–e). Upon silencing HVAs, inverse
stimuli evoked responses that decreased with increasing size of the
grey patch, reminiscent of the responses of L4 neurons. These effects
could not be explained by a direct effect of scattered laser light on V1
(Extended Data Fig. 7d–g), nor by the activation of putative inhibitory
neurons in the HVA with long-range axonal projections targeting V1, as
these projections—consistent with the findings of a previous study^22 —
were rare (Extended Data Fig. 7h–l).
To test whether distinct HVAs contribute equally to inverse responses
in V1, we silenced individual HVAs while recording single-unit responses
in V1 to classical and inverse stimuli (Extended Data Fig. 9). Although
the silencing of several visual areas reduced the response to inverse
stimuli, the strongest stimulus-specific effect on inverse responses
occurred when silencing the lateromedial visual area (LM).
To address whether inverse tuning is directly inherited from
inverse-tuned neurons in the LM, we determined the response prop-
erties of LM axonal boutons in layer 1 of V1 while mapping the retino-
topic coordinates of the V1 site (Fig. 5f–i, Extended Data Fig. 10a–h). LM
boutons for which the receptive fields were centred on the retinotopic
coordinates of the imaged V1 site showed surround suppression to
classical stimuli and were not inverse-tuned (Fig. 5g). When present-
ing inverse stimuli centred on the receptive field of the LM boutons,
the response decreased with increasing diameter of the grey patch,
as seen in L4 neurons (Fig. 1f). This was not a general property of LM
neurons, because directly imaging cell bodies in the LM showed inverse
tuning in some neurons and in the population average (Extended Data
Fig. 10i–k). We therefore established that V1 neurons do not directly
inherit inverse tuning from the LM.
Inverse tuning of L2/3 neurons could result from LM inputs that,
although not inverse-tuned, have spatially offset receptive fields rela-
tive to those of L2/3 neurons. These LM inputs would respond to an
inverse stimulus centred on the V1 retinotopic coordinates because
their receptive field, being offset relative to the grey patch, would be
stimulated by the grating. We mapped the spatial offset of the recep-
tive field of LM boutons relative to the retinotopic coordinates of the
V1 sites. The centres of the receptive fields of LM boutons showed a
wide scatter relative to the retinotopic coordinates of the V1 site, larger
than the scatter of ffRF centres of L2/3 neurons at the V1 site (Fig. 5h,
Extended Data Fig. 10f, g), consistent with the findings of a previousHVA
ControlClassical 15° Inverse 15°473 nm –1 021 –1 021
Time (s)
00.8
0.6
0.4
0.2ab c ClassicalTime (s)V1Example
unitOpto.Response (normalized ring rate)Classical Inverse0>1.1 **** *** *d*** (^51515) Stimulus size (°) 25 35 45
Bsl (Hz)^04
P
LMLI
AL
AM RL
PM
M
125 Hz
scanning
V1
VGAT-ChR2
Δ(Control,Opto.)
(normalized ring rate)
NS
0
–0.5
–1.0
0.5
–1.0–0.50 0 .5
Classical modulation index
Inverse modulation index<–2.0
e
L2/3L5/6
Opto.Control
0
0.8
0.6
1515
0.4
5 25
0.2
35 45
Inverse
Stimulus size (°)
**
Opto.Control
HVA
V1 HVA
V1
fhg i
?
L2/3L4
L5/6
Flex-GCaMP6fFlex-RGECO1a
V1 LM
Emx1-Cre
0
0.6
0.9
0.3
Response (norm.
ΔF
/F)
Stimulus size (°)
030960 0
ClassicalInverse
0
7
ΔF
/F 4
0.5***
Fraction (^00) ITI 1
10°
V1
0.016
Fraction 0 LM Response (norm.
ΔF
/F)
Stimulus size (°)
ClassicalInverse
(^00309600)
0.6
0
7
4
0.3
ΔF
/F
7 <0.7
0.5
0
Response (Hz) 2
3
1
0
2
3
1
Opto.Control
Classical
Inverse
5 1515 25 35 45
Stimulus size (°)
Opto.Control
Fig. 5 | Higher visual areas contribute to inverse tuning in V1. a, Schematic of
results and experimental configuration for the optogenetic silencing of HVAs.
b, Top, raster plot of an example L5/6 unit (30 trials each). Black horizontal lines
represent the period of stimulus presentation; blue horizontal lines represent
period of HVA silencing. ‘Control’ and ‘opto.’ indicate trials without and with
optogenetic silencing of HVAs, respectively. Bottom, example size-tuning
function with or without HVA silencing. c, Population-averaged size-tuning
function to classical (top) or inverse (bottom) stimuli, with or without HVA
silencing, normalized to the maximum control response to classical stimuli.
Responses under control and silenced conditions were compared using
two-sided Wilcoxon signed-rank tests; classical, *P = 3. 2 × 10−4; inverse,
*P = 3.0 × 10−3. The inset shows the mean baseline firing rate (bsl) with or
without HVA silencing, compared using a two-sided Wilcoxon rank-sum test;
P = 5.0 × 10−10; 44 units in 12 mice. d, Difference in firing rates under control
conditions and after HVA silencing. Comparisons were performed using
two-sided Wilcoxon signed-rank tests. At 5°, P = 8.0 × 10−3; at 15°, P = 1 .4 × 10−3;
at 25°, P = 4.0 × 10−5; at 35°, P = 3. 5 × 10−3; at 45°, NS, P = 0.70; 44 units in 12
mice. e, Optogenetic modulation indices (Methods). The green symbol
indicates the example unit from b. Comparisons were performed using a
two-sided Wilcoxon signed-rank test; P = 7.4 × 10−4; 44 units in 12 mice.
f, Schematic of results and experimental configuration. LM boutons in
V1 are not inverse-tuned as revealed by two-photon calcium imaging.
g, Population-averaged size-tuning function for LM boutons retinotopically
aligned to their putative V1 targets. The left inset shows maximum responses,
with horizontal lines denoting the median values. Responses to classical and
inverse stimuli were compared using two-sided Wilcoxon signed-rank tests;
**P = 6.1 × 10−16. The right inset shows the ITI distribution with a median value of
0.11. Black represents results for LM boutons (87 boutons in 5 mice), grey
represents the results for L2/3 neurons (Fig. 1c), compared using a two-sided
Wilcoxon rank-sum test, P = 2. 8 × 10−38. h, Top, LM boutons respond to inverse
stimuli that are centred on their putative V1 targets. Bottom, retinotopic
spread of the ff RF of V1 neurons and LM boutons (2,352 neurons and 311
boutons in the same 5 mice). i, Population-averaged size-tuning functions of
LM boutons not retinotopically aligned with their putative V1 targets. The inset
shows maximum responses, with horizontal lines denoting the median values.
Responses to classical and inverse stimuli were compared using a two-sided
Wilcoxon signed-rank test; ***P = 4.8 × 10−36; 362 boutons in 5 mice.