The Role of Feedback Connections from Higher Visual Areas in Orientation-Tuned Surround Suppression in Mice V1 Single Unit Activity in Anesthetized and Awake State

(1)

MSc Brain and Cognitive Sciences

Cognitive Neuroscience

Research Project I

The Role of Feedback Connections from Higher Visual Areas

in Orientation-Tuned Surround Suppression in Mice V1

Single Unit Activity in Anesthetized and Awake State

By

Enny H. van Beest

UvA ID: 10002255

September, 2014

32 ECTS

January – August

Supervisor:

Examiner:

Dr. Matthew Self

Dr. Umberto Olcese

(2)

Abstract

Cells in the primary visual cortex (V1) are known to respond to stimuli inside their classical receptive field (cRF) in an orientation dependent fashion. Thereby, their responses can be modulated by the orientation of the context, which is thought to enhance the ability to recognise objects in the visual scene. When the orientation of the surround is equal to that of the centre, larger suppression of the response occurs compared to when the surround has an opposite orientation, a phenomenon called Orientation Tuned Surround Suppression (OTSS). Mechanisms behind OTSS are unknown, but with recent findings that this effect also exists in mouse V1, the role of feedback in relation to OTSS can be investigated. Optogenetics can be used to decrease feedback from higher visual areas to V1 in both anesthetized and awake mice. Here we show, based on single unit data, that feedback is necessary for OTSS in deep layers of V1 in the awake state, but not in other layers or in anesthetized state. A possible explanation could be the asymmetric targeting of layers by feedback projections that has been found in vitro before. In addition, both OTSS in general and the effect of reducing feedback on it, are higher in awake compared to anesthetized mice V1, which could be due to the altered balance between excitation and inhibition. However no strong conclusions can be drawn about the role of feedback in OTSS in general, it is clear from these results that different layers of an area are functionally and anatomically different. Future research should focus on their distinct roles in figure-ground modulation and should thereby distinguish between awake and anesthetized recordings.

Introduction

The visual system is a complex system with many different cells, pathways and brain areas involved. Despite the large amount of research already performed on it, a lot about the visual system is still unknown. However, pioneers in the fifties already showed that, on different levels of the visual pathway, such as ganglion cells in the retina (Barlow, Fitzhugh, & Kuffler, 1957) and excitatory and inhibitory cells in the primary visual cortex (V1) (Hubel & Wiesel, 1959), cells respond to visual stimulation of a certain area within the visual field, which is termed classical receptive field (cRF). Since then, there has been a lot of progress in finding the mechanisms and brain areas involved in vision. It became clear that cells in V1 respond more strongly to stimuli with certain properties, such as specific orientations and sizes (Hubel & Wiesel, 1998; Niell & Stryker, 2008; Ringach, 2004). Not only stimuli that fall within the cRF are important for a neuron its response, but it is believed that context can modulate it (Blakemore & Tobin, 1972; Lamme, 1995). It is unclear whether this contextual modulation happens via lateral (Adesnik, Bruns, & Taniguchi, 2012; Blakemore & Tobin,

(4)

Figure 1: Orientation-Tuned Surround Suppression, adapted from Self et al., (2014). This figure

shows the effect of orientation context on perceived contrast. In all three images, centre gratings are identical, but the addition of an iso-oriented grating in the surround (left image) reduces the perception of contrast of the centre grating more than a cross-orientated surround (rotated by 90◦, right image).

1972), feedforward (Supèr, Romeo, & Keil, 2010) or feedback (Bullier, Hupé, James, & Girard, 2001) connections. Angelucci and Bressloff (2006) suggested all these pathways to contribute in a different way in a recurrent network model. The fast feed-forward connections are thought to mainly shape the cRF, whereas recurrent networks modulate responses to stimuli in this cRF depending on factors such as visual context and spatial attention (Lamme & Roelfsema, 2000). One example of contextual modulation is figure-ground (FG)-modulation, in which the response of a neuron is enhanced when the cRF is on a figure, compared to when it is on a background, even though the visual input is exactly the same. This is thought to be important for grouping parts of an image together during perception (Roelfsema, 2006). Another example is surround suppression, which describes the decreased response to a stimulus that extends beyond the cRF (Cavanaugh, Bair, & Movshon, 2002; Knierim & van Essen, 1992). Remarkably the suppression from the surround is orientation tuned. Strong suppression occurs when an iso-oriented surround (i.e. the surround of a cRF has the same orientation as a grating within the cRF) is presented, but the response is less suppressed (relative to iso-suppression, therefore sometimes called cross-facilitation) when stimuli with cross-orientated surroundings are presented (Cannon & Fullenkamp, 1991; Self et al., 2014, Figure 1). This phenomenon is described as Orientation-Tuned Surround Suppression (OTSS) and is thought to be important for increasing the neural representation of possible objects in the visual scene.

(5)

What the mechanisms behind OTSS exactly are, is yet unknown. Feedforward connections from LGN to V1 cannot explain the effect on their own, since the size of V1 cell surrounds extend far beyond those of LGN cells. Therefore they could prefer differently sized stimuli, which leads to possible cases where surround suppression in a LGN cell does not lead to surround suppression in the V1 cell it is projecting to (Angelucci & Bressloff, 2006). The lack of orientation-columns in mice (Van Hooser, Heimel, Chung, Nelson, & Toth, 2005) rules out the possibility that OTSS originates from neighbouring cells with the same preferred orientation inhibiting each other. A recent study of Adesnik et al. (2012) suggested that somatostatin (SOM)-interneurons in superficial layers of V1 could play a role, since these interneurons are found to be involved in general surround suppression as well. However, blocking activity in superficial layers in V1, and thereby indirectly SOM activity, did not change OTSS significantly in the deeper layers and layer 4 (Self et al., 2014). The larger latencies of orientation-tuned modulation compared to latencies of general surround suppression observed in the same study, indicate a role for feedback from higher visual areas. This idea is in line with the finding that inactivating higher visual areas by cooling in macaques, results in decreased surround suppression (Nassi, Lomber, & Born, 2013).

A recently developed technique that could be used to find out the functional role of feedback connections from higher visual areas in OTSS in V1 is optogenetics, in which virus-infected cells express a light-sensitive protein allowing them to be activated or inhibited by light (Yizhar, Fenno, & Davidson, 2011). Here this technique is used to inject Adeno Associated Virus (AAV) with a cre-dependent gene for Channelrhodospin2 (Chr2) into higher visual areas of Gad2Cre mice, leading to expression of Chr2 in all transfected interneurons at the injection site. In this way feedback provided to V1 by higher visual areas can be reduced temporally, by activating interneurons locally with blue light. This can help answering open questions about OTSS in V1, such as 1) are feedback connections from higher visual areas involved in OTSS?

Apart from whether OTSS in V1 depends on feedback connections from higher visual areas, another question is what the role of OTSS would be in perception. It is likely that OTSS contributes to FG-segregation, as it acts to enhance the responses to regions of orientation contrast (possible figures) and suppress the responses to uniform regions of the visual scene (possible backgrounds). It is unknown whether recurrent FG-segregation mechanisms, such as boundary detection and grouping of figure regions, as described by Roelfsema, Lamme, Spekreijse, & Bosch (2002), indeed play a role. In support of OTSS being dependent on these mechanisms, a computational model by Sakai & Nishimura (2006) shows that iso-suppression and cross-facilitation are involved in assigning border ownership (BO), which is assumed to be important for constructing figures. Following that reasoning, it would be interesting to measure OTSS in awake mice, since until now it has only been

(6)

studied in anesthetized mice, while in monkeys it has been found that perceptual segregation of figure and ground only takes place in awake, but not in anesthetized monkeys (Lamme, Zipser, & Spekreijse, 1998). Thereby, inhibitory currents are found to dominate excitatory currents in awake, but not in anesthetized rodents (Andermann, Kerlin, & Roumis, 2011; Haider, Häusser, & Carandini, 2013). Therefore another important question is 2) whether OTSS is different in awake compared to anesthetized mice. In the current study various forms of neural signals (i.e. SUA, MUAe and LFP) are recorded from all layers of mouse V1, while stimuli with either an iso-surround in-phase or out-of-phase with the centre, or a cross-surround are shown. For half of the trials, blue light will be turned on, causing reduced feedback from higher visual areas to V1. In this report SUA analysis will be discussed to show that 1) OTSS in V1 is likely to be dependent on feedback from higher visual areas in a layer-dependent fashion in awake, but not anesthetized animals and 2) that both OTSS in general and the effect of reducing feedback on this, is smaller in anesthetized compared to awake mice.

Methods

Surgery and electrophysiology

All experimental procedures complied with the National Institutes of Health Guide for Care and Use of Laboratory Animals, and were approved by the institutional animal care and use committee of the Royal Netherlands Academy of Arts and Sciences. Data was recorded from 27 penetrations in 16 hemispheres of 14 anesthetized animals with a maximum of 2 penetrations per hemisphere, and 9 penetrations in 3 hemispheres of 2 awake GAD2-cre mice. All anesthetized animals were male and were aged 2 to 4 months; one animal for the awake recordings was aged over 1 year and the other 3 months. They were both female. The awake animals were handled for two weeks, where after they were treated in the same way as the anesthetized animals. Anesthesia was induced before surgery using 5% isoflurane in oxygen, and maintained with 1.5-2.5%. The mouse was placed in a stereotactic frame and Cavasan eye ointment was applied to prevent eyes from drying. Body temperature was maintained between 36.5 and 37.5 ˚C. Depth of anesthesia was determined by testing foot reflexes when carefully pinching between the toes and monitoring breathing rates. If necessary the level of isoflurane was adapted. Hair on the head, at the location of the pedestal and chamber, was shaved off, where after a small amount of lidocaïne was applied as local analgesia. After 5 minutes an incision was made above the frontal and parietal bone and the skin was gently pulled aside. The bone underneath it was cleaned and etched with a phosphoric acid for approximately 10 seconds. For awake recordings, two screws were inserted in the frontal bone. A primer and adhesive layer were applied. A small metal plate (pedestal) was attached to the frontal

(7)

and parietal bones using the screws and a composite that was hardened by UV light, originating from dental surgery and adapted for use in laboratory animals. For both anesthetized and awake recordings, a hole, just large enough for insertion of a sharp glass pipette, was drilled in the skull at the coordinates of a posterior and/or anterior higher visual region (LM or AL respectively). Using this pipette a maximum of 250 nanoliter of viral fluid (i.e. AAV with a cre-dependent expression vector) was injected by pressure at multiple depths. A small craniotomy (1-2 mm in diameter) was made in the occipital bone at 0.4 mm anterior to lambda and 2.9 mm lateral to the midline with an acrylic chamber around it. A reference wire (Ag/Cl) was inserted between the skull and the dura of the frontal cortex while a ground electrode was inserted under the skin. A craniotomy with radius < 0.5 mm was performed. Sterile NaCl with an antibiotic mixture of gramicidin was applied to the chamber, where after it was sealed with silicon and sterile bone wax. 2 mg/kg metacam was provided subcutaneously as analgesic. Animals for awake recordings received recovery for two weeks in a filter top cage, before acclimatization to the set-up began. Anesthetized animals were immediately put in the set-up (figure 2A). They were head-fixed with ear- and mouth bars in a stereotactic frame, whereas awake mice were head-fixed using a head-bar and their pedestal. During recording sessions, a 16 contact laminar electrode (Neuronexus A1x16-10mm-100-413, contacts spaced 100 m apart) was lowered into the V1 craniotomy. In awake recordings, the shallowest electrode (channel 16), which was not in the brain, was used as a reference to subtract movements out of the signal. An optic fibre was brought in place over higher visual area LM (anesthetized and awake), AL or both (anesthetized only). Electrical signals from each of the contacts were amplified and digitized at 24.4 KHz (Tucker-Davis Technologies). LFP was measured using a low-pass filter (corner frequency 200 Hz) and sampled at 763 Hz. Current-source density (CSD) (1) was calculated as the second spatial derivative of LFP (Mitzdorf, 1985).

( ) ( ) 2 (₂ ) ( ) h h x x h x x CSD 







 







 , ₍₁₎

where ϕ is the voltage, x is the point at which the CSD (in A.mm-3) is calculated, h is the spacing of recording sites for the computation (here 0.2 mm) and σ is the tissue conductivity (400 S.mm-1) (Logothetis, Kayser, & Oeltermann, 2007). CSD response was calculated to the onset of a full-screen, full-contrast checkerboard stimulus (check-size = 20°) to place the electrode at approximately the same depth for each penetration. We examined the CSD traces for the earliest current sinks induced by the checkerboard in combination with a reversal from current sources to current sinks as these features mark the location of the boundary between layer 5 and layer 4c (Mitzdorf, 1985) (Figure

2B). This reversal agrees well with histological assignments of this laminar boundary in mice (Niell &

(8)

to this boundary. Sites from -600 to -100 m were assigned to deep layers, 0-100 m to layer 4 and 200-500 m to superficial layers.

SUA was recorded by setting a spike-amplitude threshold for each recording site. To isolate single units, we clustered spike shapes using WaveClus (Quiroga, 2009) and only included well separated clusters with a refractory period in our single unit analysis. To record MUAe, the signal from the electrode was bandpass filtered (500 Hz - 5 kHz) to extract high-frequency (spiking) activity, rectified, and then low-pass filtered at 200 Hz to measure the envelope of this signal. MUAe provides an instantaneous measure of the number and amplitude of spikes in the vicinity of the electrode without the setting of a spike-detection threshold. MUAe responses are similar to thresholded multi-unit data and to the average single-multi-unit response (Self, Kooijmans, Supèr, Lamme, & Roelfsema, 2012; Supèr & Roelfsema, 2005).

Visual Stimuli

Visual stimuli were presented on a back-projection screen 15 cm in front of the mouse, using a gamma-corrected PLUS U2-X1130 DLP projector (mean luminance = 40.6 cd.m-2 ). Stimuli were created with MATLAB (Mathworks, inc.) package Cogent. The MUAe RF of every recording site was mapped using one (anesthetized) or four (awake) briefly (200 ms) bright square(s) of 10° presented at each point of a grid covering the entire screen (136° x 102°) (Figure 2C). To assess the preferred tuning properties of multi- and single-units, we used drifting gratings with a varying orientation (0-360° in 12 steps of 30°) in blocks of at least 20 repeats of 1 sec duration (mean luminance 40.66 cd.m -2_{, contrast 80%) with a spatial frequency of 0.08 cyc.deg}-1 _{and drift speed of 28 deg.s}-1_{. MUAe activity} was averaged at each recording site to construct tuning curves and predominantly preferred values were chosen. Size-tuning profiles were established using drifting sine-wave gratings with the previously determined predominantly preferred orientation, and 10 aperture sizes varying from 5-200°. Gratings were presented for 0.5 sec and alternated with a grey screen of mean luminance for 0.5 sec as inter-trial interval. Based on the size-tuning curve optimum two centre grating sizes were chosen; a ‘small’ grating (mode = 20°, range 15-30°), and a ‘big’ grating which was 20° bigger than the small centre size, since RF-sizes are thought to increase with depth of cells. A surround grating with outer diameter of 120° was added in anesthetized experiments, and a full-screen surround in awake experiments. Both gratings drifted at 28 deg.s-1 .

The centre of the visual stimulus was either small or large, and had a preferred or non-preferred (+90°) orientation. The surround could 1) be equal to the centre (iso condition) or 2) have an opposite orientation (+90°, cross condition) (figure 2D). This resulted in a 2 (small or large centre) x 2 (preferred or orthogonal centre orientation) x 2 (cross- or iso-oriented surround) x 2 (light on or

(9)

Figure 2: A) Schematic overview of set-up. A mouse is either anesthetized and stabilized with ear

bars, or awake and head-fixed in the set-up. A Neuronexus recording electrode with 16 channels 100 µm space in between is lowered into V1. An optic fibre with blue light is positioned over higher visual area LM (awake and anesthetized), AL or both (anesthetized only). A projection screen is placed 15 cm in front of the mouse, on which stimuli are presented using Matlab package Cogent.

B) Current Source Density (CSD) as the second spatial derivative of local field potential. This can be

used to determine the sink (red area in right plot) and the reversal potential (where the signal changes from negative to positive), which is equal to layer 4C in mice V1. C) Receptive field

mapping using 1 (anesthetized) or 4 (awake) white squares to map where the highest visual

response is evoked (left). Right side of this figure shows normalized MUA that is evoked when a white square appears on that specific spatial location. Circles show determined RF (highest visually evoked MUA signal) for selected channels that are thought to be located in V1. D) Visual stimuli

used for the task. Conditions were; cross-oriented surround (cross), iso-oriented surround (iso),

Surround only (surround), Centre only (centre) and grey screen (Spont). To determine the small centre size, size tuning was performed. Big centre size was always 20 visual degrees bigger than the small centre size.

light off) design. To assess activation by the surround of the RF, a stimulus without a centre grating (grey centre) was presented. For spontaneous activity a grey screen was used. All stimuli were on the screen

for 1 sec.

(10)

Figure 3: Virus Localization. A) Stimuli used for optic intrinsic imaging. Stimuli (drifting

gratings) were presented at four different locations in the visual field (responding to the different colours). The signal that it evoked in anesthetized mice V1 was imaged to localize the responding area. B) Average maps of activity for single conditions. From top left to bottom right; red, green, blue and yellow area of the visual field (as in A). Axes are distance from lambda (red cross) in mm in x and y direction. Colours show normalized intrinsic signal values. Higher values (reddish) mean higher visually evoked signal to the corresponding stimulus described in (A). C) Winner-Takes-All map (WTA) of V1. This map is created with the use of the single condition maps in B. 147 sessions were used. Axes again give distance from lambda, which is the red cross here. Overestimated border of V1 would be around the yellow circle, where relative high signal changes to lower signal. Areas neighbouring V1 with slightly higher intensity are probably higher visual areas.

Viral expression

A fluorescence microscope was used to visualize viral expression in the cortex. This expression was overlaid on an average winner-takes-all (WTA) map of V1 to control for direct light effects on cells in this area (supplementary figure 1). First 147 retinotopic maps, coming from a database for intrinsic imaging by J.A. Heimel, were averaged for each spatial location in the visual field separately (Figure 3A). These average maps per location were overlaid relative to lambda (Figure 3B) and from these an average WTA map relative to lambda was created (figure 3C).

(11)

Data analysis

All evoked firing rates were taken as the response during 0-1 sec of stimulus duration. Single units with a mean evoked firing rate during the small and/or big centre size smaller than 0.5 Hz above the pre-trial baseline activity (PBA) (-0.3 to 0 sec relative to stimulus onset) were excluded from analysis (anesthetized 36.21%, awake 33.85%), leaving a total of n = 148 from anesthetized, and n = 43 cells from awake recordings. As a specific modulation index, the Orientation Specific Suppression Index (OSSI) (2) was calculated for each cell, to quantify the difference between cross- and iso- orientated conditions.

, (2)

where Cz is the mean response in Hz to the cross-orientated stimulus for size z (i.e. small or big centre size), Iz the mean response in Hz to the iso-orientated stimulus for size z, and PBA the pre-trial baseline activity. Means are calculated over both preferred and non-preferred centre orientation. Outliers in OSSI values were excluded based on a z-score threshold of 3. Since data did not differ much from a normal distribution as determined by eye inspection of histograms and normality plots, a mixed factorial ANOVA was performed using SPSS (IBM Statistics) with a repeated measures factor per cell (light on or light off) and a between factor being layer (deep layers, layer 4 (L4) or superficial layers) with OSSI per size as outcome variables.

To compare conditions directly without using indexes, evoked firing rates (Hz) of all conditions were normalized to the evoked firing rates (Hz) of the small centre no light condition. This resulted in a value of 1 for this condition for every cell, and for all other conditions, the evoked firing rates in both small and large size stimuli were scaled to this value. A mixed factorial ANOVA was then performed on the normalized firing rates of the cells in a 2 (light on or light off) x 2 (cross or ISO) x 3 (deep, L4 or superficial) design.

In order to test the difference between awake and anesthetized, OSSI values were compared in a mixed ANOVA with a 2 (light on/light off) x 2 (awake or anesthetized) x 3 (deep layer, L4 or superficial layer) design.

(12)

Results

Role of feedback - Anesthetized data

First the role of feedback in OTSS was investigated in anesthetized animals. Because the virus was injected in two separate higher visual areas, it was tested whether the role of feedback differed between AL and LM, which was not the case: for both small and big centre size stimuli, injection location of the virus did not influence the modulation between cross and iso stimuli, which was determined by comparing OSSI values (Supplementary Figure 2). Therefore data was pooled over all injection locations for the rest of the analysis.

Subsequently we were able to replicate the results of Self et al. (2014), who found orientation tuned surround suppression in anesthetized mouse V1. For both centre sizes SUA was higher for oriented than for iso-oriented surrounds. Small size: main effect of normalized cross-oriented (0.907 ± 0.054Hz) being larger than iso-cross-oriented responses (0.644 ± 0.051Hz), difference is 0.263 ± 0.031Hz, F(1,125) = 74.347, p < 0.001, η = 0.375, (Figure 4A), large size: F(1,125) = 8.300, p = 0.005, η = 0.062, (Figure 4B), with responses to cross-oriented (0.691 ± 0.057Hz) being on average 0.074 ± 0.026Hz bigger than responses to iso-oriented (0.617 ± 0.054Hz) stimuli. For small centre sizes only, this difference was not constant, but dependent on the layer that was recorded in. Small: F(2,125) = 8.693, p < 0.001, η = 0.122 (Figure 4C). There was no main effect of light found in both small and big centre size stimuli, suggesting that inhibiting feedback to V1 has no effect on general evoked activity in V1. Instead a trend for a three-way interaction was found between condition, light and compartment in the small centre size condition (F(2,125) = 2.768, p = 0.067, η = 0.042). More specifically, the difference in normalized response between cross- and iso-oriented stimuli, that was larger in superficial layers (cross: 1.029 ± 0.131Hz, iso: 0.567 ± 0.124Hz) compared to L4 (cross: 0.804 ± 0.081Hz, iso: 0.609 ± 0.077Hz) and deep layers (cross: 0.888 ± 0.049Hz, iso: 0.755 ± 0.047Hz), was even larger when the light was turned on compared to when it was turned off (Figure 4C left side,

Supplementary Table 1). These interaction effects were not observed when using big size centre

stimuli.

Next, we tried to capture the modulation between cross- and iso-oriented responses in the OSSI value, which was calculated using raw spike rates of cells in response to cross- and iso- oriented stimuli. When doing analysis on these values, similar results were observed as described for the normalized response rates. In Figure 5A and Figure 5C data points are OSSI values in the light off condition (x-axis) plotted against those of light on condition (y-axis) for every cell in the small and big centre size condition, respectively. In the small centre size condition, responses to cross-oriented

(13)

Figure 4: A) Relative firing rates (Hz) small centre size. For both light off and light on (x-axis, light

on has faded colours), relative spike rates are plotted in Hz (y-axis, same for all subfigures). ‘Relative’ means relative to the spike rate of the small centre light off condition, which is set to be 1 (dotted line in all figures). Therefore this bar is missing in (A) light off. Other bars are for iso-, cross- and surround condition as shown in the legend. Normalized response to cross-oriented stimuli was significantly bigger than to iso-oriented stimuli. B) Relative firing rates (Hz) big centre

size. Bars represent same information as in figure (A). Firing rates are relative to small centre size

light off (dotted lines). Again normalized response to cross- is significantly bigger than iso-oriented stimuli. C) Relative firing rate per layer. For every subfigure relative firing rates when a cross (red) or iso (orange) stimulus is presented, are given for different layers (x-axis). Top left gives information for small centre size, light off. Bottom left for small centre size, light on (faded colours). For small centre size the difference in response to cross- and iso-oriented stimuli depended on the layer signals were recorded from. In addition turning the light on could influence the difference between cross- and iso-oriented stimuli in a layer-dependent way. Top right figure is for big centre size, light off and bottom right for big centre light on.

(14)

Figure 5: A) OSSI values light off vs. Light on for small centre size. For every cell

(deep layer = dark blue, L4 = light blue, superficial layer = yellow) the OSSI value of the no light condition (x-axis) is plotted against the light on condition (y-axis). The dark green area is the area were OSSI values are negative, meaning that there are higher responses for iso than cross. B) Average OSSI value per layer small centre size for both light off (black) and light on (blue). Asterisk (*) means OSSI values are significantly higher than 0 (all *, p < 0.001). C) OSSI values light off vs. light on for big

centre size. Same labels as in (A). D) Average OSSI value per layer big centre size.

Same labels as in (B), accept now * is p < 0.05.

stimuli were generally bigger than to iso-oriented. Consequently, OSSI values were significantly higher than 0 in every compartment, meaning that there was significant surround suppression in all layers (Figure 5B, Bonferroni corrected Wilcoxon Signed Rank tests, all but superficial light off, p < 0.001). However, the amount of modulation differed significantly per layer (F(2,119) = 4.364, p = 0.015, η = 0.0168). Post-hoc tests showed that this effect was driven by OSSI values in superficial layers, which were 0.146 ± 0.053 bigger than values in deep layers (p = 0.021, Bonferroni corrected). No significant effect of blocking activity in higher visual areas was observed. For big centre size stimuli, OSSI values were generally smaller, more spread and more often negative (Figure 5C), resulting in less orientation-tuned surround suppression. In Figure 5D it is shown that only in L4 OSSI values were significantly larger than 0 (both light off and light on L4: p < 0.05, Bonferroni corrected, Wilcoxon signed rank test). No significant effects of light or compartment or an interaction between them were found.

(15)

Role of feedback - Awake data

Since processes in the anesthetized brain are known to differ from the awake brain, and more importantly, figure-segregation processes are thought to be absent in the anesthetized brain, we did the same analyses on data recorded from chronic recordings in awake mice. Because there was no significant difference in response measured between the two mice we used for the awake recordings, and in addition only a small number of awake recordings were available to include in this analysis, data from both mice was pooled. When looking at the normalized responses to iso-oriented and cross-oriented stimuli, we found a main effect of stimulus type on responses in the small centre size condition (F(1,36) = 7.293, p = 0.010, η = 0.168, Figure 6A). Similar to results from anesthetized recordings, this effect came from responses to cross- (1.214 ± 0.162Hz) being larger than responses to iso-oriented stimuli (0.863 ± 0.112Hz, difference between two stimuli: 0.351 ± 0.130Hz). There was no general main effect of light on reducing responses, however in interaction with compartment (F(2,36) = 3.655, p = 0.036, η = 0.169) there was an effect of light. This is mainly seen in Figure 6C (left side), that shows that the average response of both conditions decreased or did not change in deep and superficial layers when the light was turned on, while it increased in L4. This effect observed was not limited to the response to iso-oriented and cross-oriented stimuli, but was a general phenomenon that we observed for all visually evoked responses (Supplementary Figure 3). Thus without light (i.e. under normal circumstances), L4 responses to centre stimuli seemed suppressed compared to deep- and superficial layer responses. No significant interaction was found between light and condition, but Supplementary Table 2 and Figure 6C show that light decreased the difference between cross and iso slightly for deep layers only. For big centre stimuli no main effect of condition was found (Figure 6B). However, the interaction between light on/off and compartment observed in small centre size was preserved as a trend (F(2,35) = 2.954, p = 0.065, η = 0.144, Figure

6C, right side). In addition there was a trend for a main effect of light (F(1,35) = 3.436, p = 0.072, η =

0.089), but no relative changes between responses to cross- and iso-oriented stimuli due to the light were found.

To quantify what the effects of reducing feedback on the modulation between cross- and iso oriented stimuli were, analysis on the OSSI values was performed as well. In awake animals we observed that in deep layers most of the cells had reduced modulation (i.e. lower OSSI values) when the light was on, compared to when the light was off (Figure 7A). However, in superficial layers and L4, this was not the case. In Figure 7B this is expressed in the average OSSI value per compartment. In both the deep and superficial layers, OSSI values were significantly larger than 0 under normal circumstances (both p < 0.05, Bonferroni corrected Wilcoxon Signed Rank). While this effect remained in superficial layers when feedback was prevented (p < 0.05, Bonferroni corrected

(16)

Figure 6: Normalized responses in awake recordings. A) Normalized firing rates (Hz) for small centre size (y-axis, same for every subplot) for every condition (centre, iso, cross, surround and out of phase,

see legend for corresponding bars) during both light off and light on (x-axis, light on in faded colours). The evoked response depends on the type of stimulus that is given. More specifically, responses to cross- are significantly higher than to iso-oriented stimuli. B) Normalized firing rates (Hz) big centre

size. Same axes as in (A). No significant effect of condition was found. C) Normalized firing rates (Hz) per compartment. Left column is for the small centre size, right column for the big centre size. Top

row (bright colours) is for light off condition and bottom row (faded colours) for light on. Normalized responses differed between light off and light on, in a layer-dependent fashion. This is also illustrated by the black line, which marks the average of all conditions for every layer.

Wilcoxon Signed Rank), the OSSI value did not differ significantly from 0 in deep layers with the light on. This layer dependent change when the light was turned on was also indicated by a trend interaction between light and layer (F(2,33) = 2.574, p = 0.091, η = 0.135). Apart from this interaction with light, there was a significant main effect of compartment on OSSI values, meaning that the amount of orientation specific surround suppression depended on the layer that was recorded from (F(2,33) = 5.455, p = 0.009, η = 0.248). Post-hoc tests showed that this effect was mainly driven by superficial layers having higher OSSI values (0.300 ± 0.053) compared to L4 (0.042 ± 0.064, p = 0.012). There was only a trend for the deep layers (0.104 ± 0.071) having higher OSSI values than L4 (p = 0.100), which was probably due to the reduced values with light on. For big centre size stimuli, replicating anesthetized results, OSSI values were generally smaller and no OSSI values significantly

(17)

Figure 7: Awake OSSI values A) Small size OSSI values per cell for light off (x-axis) compared to light on

(y-axis). Cells come from different compartments, which is indicated with different colours (dark blue for deep layers, light blue for L4 and yellow for superficial layers). The dark-green area marks the area where OSSI values become negative, and the black middle line shows the area where OSSI values between light off and light on are equal. B) Small size average OSSI values. Asterisk (*) marks conditions in which the OSSI is significantly bigger than 0 (Wilcoxon signed rank test). On the x-axis different groups compartments are given (deep layers, L4 and superficial layers) and values are for light off (black) and light on (blue). OSSI values differed per layer (F(2,33) = 5.455, p = 0.009, η = 0.248) and the change in OSSI values when the light was on compared to light off was also modulated by in which layer the cell was (F(2,33) = 2.574, p = 0.091, η = 0.135). C) Big size OSSI values per cell. Same axes and legends used as in (A). D) Big size average OSSI values. Same axes values as in (B). No significant OSSI values or interaction effects were found.

higher than 0 were found, neither with light off or light on (Figure 7C and D). Although it seemed that OSSI values did decrease when the light was on compared to when it was off in L4 and superficial layers, there were no significant effects found.

(18)

Figure 8: OSSI values Anesthetized vs. Awake under normal circumstances. A) All layers averaged. OSSI

values (y-axis) are significantly higher in awake compared to anesthetized animals, indicated with an asterisk (*) (Wilcoxon rank sum, p = 0.028). B) Deep layers. In this figure it becomes clear that this effect is mainly driven by OSSI values in the deep layers, where they are significantly higher in awake compared to anesthetized state (Wilcoxon rank sum, Bonferroni corrected p = 0.031). C) Layer 4. In layer 4 the difference between awake and anesthetized is opposite to that of what is generally observed. However, this is not significant. D) Superficial layers. Superficial layers show the effect as was observed in deep layers, however this was not significant.

Awake vs. Anesthetized

As we already observed in the analysis of both anesthetized and awake data, OSSI values were lower and - except for L4 in anesthetized animals - non-significant in big centre size stimuli, even without decreasing feedback. Therefore, for this analysis we only used small centre size data. When we directly compared awake and anesthetized OSSI values under normal circumstances (i.e. no light), OSSI values were on average higher in awake compared to anesthetized mice (Wilcoxon rank sum test, Bonferroni corrected p = 0.028, Figure 8A). However, this difference was mainly driven by deep layers (Wilcoxon rank sum test, Bonferroni corrected p = 0.031, Figure 8B,C,D), as might be expected from earlier described analysis. Superficial layers also had higher OSSI values in awake compared to anesthetized, but this was not significant. L4 showed exactly the opposite of what was seen on average and in both superficial and deep layers, but this was also not significant. Thus under normal circumstances there is a difference in the modulation between awake and anesthetized animals, question remains what the effects are of reducing feedback with light on this difference. Averaged over compartments, no significant interaction between light and state was found (Figure

9A). But as you might expect from results described above, there was an interaction between light,

state and compartment, which means that whether state influences the change in modulation with light depended on in which layer of the cortex you record (F(2,152) = 4.178, p = 0.017, η = 0.052). Deep layers seemed to be the driving force for this effect (Figure 9B). L4 and superficial layers showed no explicit different change in OSSI values with light on, in awake compared to anesthetized animals (Figure 9C,D).

(19)

Figure 9: Interaction between optogenetics and state. A) Averaged OSSI values over all layers (y-axis).

The difference between light off and light on (x-axis) OSSI values seem bigger in awake (purple line) compared to anesthetized (green line) state, but this is not significant. B) In deep layers, the difference in modulation between awake and anesthetized recordings is most clear. Effects were smaller in C) Layer 4 and D) superficial layers, which is consistent with the finding that the difference in OSSI values for light on compared to light off is not only dependent on state, but also on the layer that was recorded from.

(20)

Discussion

Main findings

First of all, we were able to replicate the findings of Self et al. (2014), namely that there was orientation tuned surround suppression in V1 of anesthetized mice. In a first attempt to see whether feedback projections play a role, we did not find evidence that blocking feedback activity from higher visual areas leads to decreased differences between responses to cross- and iso-oriented stimuli in anesthetized animals. Nevertheless, when recording in awake mice, which also showed these OTSS effects, decreasing feedback by turning the light on did seem to decrease modulation in deep layers specifically, as measured by OSSI values. Even though these effects are not very robust yet, finding an effect at all with such a small amount of cells is promising for further recordings.

Assuming these findings to be correct and persistent in future analyses, why do we find the effect of reducing feedback from higher order areas on OTSS only in awake, and not in anesthetized mice? Therefore one has to keep in mind that the domination of inhibitory currents over excitatory currents found in awake mice V1, is known to be absent under anesthesia (Haider et al., 2013). This probably results in a diminished effect of excitatory feedback from extrastriate areas to V1 under anesthesia, which would be a possible explanation why reducing this already decreased role of feedback, has little effect in our anesthetized recordings. To see whether this theory finds any support in our data, we directly compared OSSI values from cells in anesthetized to cells in awake animals. In line with this theory, under normal circumstances (i.e. no light) there was a bigger difference in awake compared to anesthetized mice between responses to cross- and iso-orientated stimuli, as observed in the OSSI value differences. In addition, light on decreased OSSI values more in awake compared to anesthetized mice.

Layer Specific Modulation

An important observation in all analyses performed, was that both the modulation between responses to cross- and iso-orientated stimuli and the influence of decreased feedback by light on them had different effects in different layers. More specifically, deep layers had a significant modulation under normal circumstances similar to superficial layers. But where modulation in deep layers seemed to be affected by light, superficial layers were not so much affected by decreasing feedback from higher order areas. Layer 4 cells acted in an opposite fashion to both the deep and superficial layers and showed reduced responses to all stimuli and no significant OTTS. When feedback was decreased by the light, deep- and superficial layer responses to all (but surround) stimuli did not change or decrease, while L4 responses increased. Layer-specific modulation has been

(21)

observed by others, for example in mice somatosensory cortex when stimulating horizontal connections, and it was also shown that L4 had less FG-modulation in macaques compared to deep and superficial layers as well (Hillel Adesnik & Scanziani, 2010; Self, van Kerkoerle, Supèr, & Roelfsema, 2013). Nevertheless, a layer-specific effect of OTSS as we have observed in our awake-, but not anesthetized recordings, was not expected from the previous study of Self et al. (2014) in anesthetized animals as well, suggesting that this layer-specific modulation might be dependent on the dominance of inhibitory currents in awake animals.

Possible Mechanisms

One possible mechanism for the layer-specific modulation we observed, is that feedback projections target layers asymmetrically in V1. It was already shown that feedback projections from higher visual areas tend to avoid layer 4 in primates (Callaway, 2004; Felleman & Van Essen, 1991). In addition, Yang, Carrasquillo, Hooks, Nerbonne, & Burkhalter (2013) in slices of mouse visual cortex found that, while feedback connections from LM to V1 cause a balanced depolarization between the primary excitatory (pyramidal) cells and parvalbumin-positive (PV+) interneurons in superficial layer 2/3, excitatory feedback was biased to targeting pyramidal cells in deep layer 5. This asymmetric feedback results in more net excitation from feedback connections in deep layers compared to superficial layers. When feedback is decreased, it is likely to affect deep layers more because of this net excitation being reduced, whereas in superficial layers the feedback probably has a more balancing role and no net change occurs. This idea fits with our findings, that reducing feedback from LM to V1 in awake mice only has an effect on contextual modulation in deep layers, and not in L4 or superficial layers. But if feedback projections only contribute to OTSS in deep, and not in other layers, what are the mechanisms behind OTSS in these layers? Perhaps SOM inhibitory neurons in superficial layers after all (Adesnik et al., 2012)? Another option is that, because of the different RF-size for every layer, centre sizes were too big for superficial layers and L4 to benefit from a cross-oriented surround, then the near surround of their RF could actually be the iso-oriented stimulus of the centre stimulus. It might be worth using even smaller centre stimuli in awake recordings, since most studies about RF sizes were performed in anesthetized animals. Nevertheless, it is unlikely, since centre-sizes are determined during recordings based on signals in all channels. In addition superficial layer cells show OTSS as well, and those cells are known to have smaller RF sizes than deep layer cells. A last option that we could think of is that our virus was mainly expressed in deep layers. If organization would be maintained across areas, it could be that we therefore only decreased feedback in deep layers with light, but not in superficial layers. This is something that needs to be investigated in histology after recordings. However, even if expression was only found in deep layers, there is no literature that we know of that shows a maintenance of layer-division across areas (i.e. feedback

(22)

from LM to layer k in V1 is always coming from layer k in LM as well). Therefore the first option, which is supported by the findings of Yang et al. (2013), is most likely.

Future Directions

For further research, the first goal would be to have more recordings in awake mice to find more robust effects. Also analyses on MUAe and LFP can be added to the single unit data, to provide an insight in the collaborative action of multiple cells. Additionally, latency analysis and using different timings of turning on the light could provide insight in the specific timing of feedback that is needed to contribute to contextual modulation. However, to completely understand and rightfully interpret results of the functional role of feedback, more details about the anatomy of feedback connections in mice are needed. For example, which specific layers in LM target different layers in V1. It would be nice to replicate the findings of Yang et al. (2013) in vivo, to know whether the differences between layers apply to awake animals as well.

On a higher level, it would be nice to know whether OTSS depends on FG-segregation mechanisms such as edge-detection and labelling by attention. Therefore a new stimulus could be introduced which has an iso-oriented surround with a shifted phase, visual responses to this stimulus could then be compared to responses to stimuli we already used here. Even better would be to be able to couple these different neural signals to certain behaviours of the mice, to see whether the animal can consciously discriminate between different stimuli. That would make it easier to compare results of mouse electrophysiology to findings about FG-segregation in macaque electrophysiology, or even human psychophysics studies, as well.

Conclusions

We conclude that feedback does play a role in orientation tuned surround suppression in deep layers of V1 in awake mice. What the mechanisms exactly are for other layers is unknown. Future research should focus on functional and anatomical specificity of feedback connections, in which different layers of a target- or source area should be treated as different structures with their own preferences and functions.

(23)

Acknowledgements

I would like to thank all the contributors to this research project, and members and principle investigators of the groups “vision and cognition” and “cortical structure and function”, both accommodated at the Netherlands Institute for Neuroscience, Amsterdam. In particular: M.W. Self, J. Vangeneugden, J.A.M. Lorteije, J.A. Heimel and P.R. Roelfsema. Thanks to U. Olcese from the “Cognitive and Systems Neuroscience”, located at University of Amsterdam, as a co-assessor and UvA representative. A special thanks to M.W. Self for guidance and supervision.

Supplementary Figures and Tables

B) Individual session maps.

Same applies as in (A), but now for every individual session. Created by J. Vangeneugden. For most sessions, virus expression was completely outside the borders of V1. When borders of expression and V1 overlapped, it was only little. Since the border of V1 is always an over-estimation due to neighbouring higher visual areas that have an opposite retinotopy, there was no concern of V1 contamination in our analysis.

Supplementary Figure 1: A) Winner-Takes-All map (WTA) of V1 with virus overlay. This map is created with the use of the WTA

map in Figure 3C. 147 sessions were used. Axes give distance from lambda, which is the white cross here. Black circles in transparent figures overlaying the WTA-map are two examples of borders of fluorescent microscopy images of virus expression (transparent green colour). Virus was not expressed within the borders of V1 (white circle), but always outside.

(24)

Supplementary Figure 2: OSSI values per virus location for deep (A), L4 (B), and Superficial (C) for small centre size. OSSI values per virus location for deep (D), L4 (E) and Superficial (F) for big centre size. In all

figures you can see the OSSI values (y-axis) as calculated by the second function per virus location (x- axis) and compartment (different subplots) for light off (black) and light on (blue). Virus location did not significantly change OSSI values, neither as a main effect, neither as an interaction effect with

(25)

Table 1: Anesthetized Mean ± std normalized firing rates (Hz) small centre size condition. Values are

given for cross-oriented and iso-oriented conditions per layer, for both light off and light on.

Table 2: Awake Mean ± std normalized firing rates (Hz) small centre size condition. Values are given for

iso-oriented, cross-oriented and OOP conditions per layer, for both light off and light on.

Layer Light CROSS ISO

Deep Off 0.900 ± 0.048 0.748 ± 0.053 On 0.875 ± 0.055 0.761 ± 0.046 L4 Off 0.816 ± 0.079 0.593 ± 0.088 On 0.792 ± 0.090 0.625 ± 0.076 Superficial Off 0.964 ± 0.127 0.584 ± 0.141 On 1.094 ± 0.145 0.550 ± 0.123

Layer Light ISO CROSS

Deep Off 1.004 ± 0.225 1.368 ± 0.304 On 0.688 ± 0.185 0.719 ± 0.305 L4 Off 0.917 ± 0.224 0.974 ± 0.331 On 1.040 ± 0.202 1.234 ± 0.332 Superficial Off 0.786 ± 0.209 1.463 ± 0.283 On 0743 ± 0.173 1.524 ± 0.284

(26)

Supplementary Figure 3: Normalized Firing rates (Hz) per compartment, awake. Same graphs and data

is shown as in figure 6C, with in addition centre and surround to show that suppression without light and release of inhibition with light in L4 accounts for visual stimuli inside the classical RF in general, but not for stimuli in the surround. OOP stands for Out Of Phase, which is a stimulus that could function as a control, since it has an iso-oriented surround with a shifted phase of 90 degrees in relation to the centre.

References

 Adesnik, H., Bruns, W., & Taniguchi, H. (2012). A neural circuit for spatial summation in visual cortex.

Nature, 490(7419), 226–231. doi:10.1038/nature11526

 Adesnik, H., & Scanziani, M. (2010). Lateral competition for cortical space by layer-specific horizontal circuits. Nature, 464(7292), 1155–60. doi:10.1038/nature08935

 Andermann, M., Kerlin, A., & Roumis, D. (2011). Functional specialization of mouse higher visual cortical areas. Neuron, 72(6), 1–32. doi:10.1016/j.neuron.2011.11.013.Andermann

 Angelucci, A., & Bressloff, P. (2006). Contribution of feedforward, lateral and feedback connections to the classical receptive field center and extra-classical receptive field surround of primate V1 neurons.

Progress in Brain Research, 154(SUPPL. A), 93–120. doi:10.1016/S0079-6123(06)54005-1

 Barlow, H., Fitzhugh, R., & Kuffler, S. (1957). Change of organization in the receptive fields of the cat’s retina during dark adaptation. The Journal of Physiology, 338–354. Retrieved from

http://jp.physoc.org/content/137/3/338.full.pdf

 Blakemore, C., & Tobin, E. (1972). Lateral inhibition between orientation detectors in the cat’s visual cortex. Experimental Brain Research, 440, 439–440. Retrieved from

http://www.springerlink.com/index/J342Q1645P754459.pdf

 Bullier, J., Hupé, J.-M., James, a C., & Girard, P. (2001). The role of feedback connectionss in shaping rhe responses of visual cortical neurons. Prog.Brain Res., 134, 193–204.

(27)

 Callaway, E. M. (2004). Feedforward, feedback and inhibitory connections in primate visual cortex.

Neural Networks : The Official Journal of the International Neural Network Society, 17(5-6), 625–32.

doi:10.1016/j.neunet.2004.04.004

 Cannon, M., & Fullenkamp, S. (1991). Spatial interactions in apparent contrast: inhibitory effects among grating patterns of different spatial frequencies, spatial positions and orientations. Vision

Research, 31(Ii), 1985–1998. Retrieved from

http://www.sciencedirect.com/science/article/pii/0042698991901939

 Cavanaugh, J. R., Bair, W., & Movshon, J. A. (2002). Nature and interaction of signals from the receptive field center and surround in macaque V1 neurons. Journal of Neurophysiology, 88(5), 2530– 46. doi:10.1152/jn.00692.2001

 Felleman, D. J., & Van Essen, D. C. (1991). Distributed hierarchical processing in the primate cerebral cortex. Cerebral Cortex (New York, N.Y. : 1991), 1(1), 1–47. Retrieved from

http://www.ncbi.nlm.nih.gov/pubmed/1822724

 Haider, B., Häusser, M., & Carandini, M. (2013). Inhibition dominates sensory responses in the awake cortex. Nature, 493(7430), 97–100. doi:10.1038/nature11665.Inhibition

 Hubel, D. H., & Wiesel, T. N. (1998). Early exploration of the visual cortex. Neuron. doi:10.1016/S0896-6273(00)80984-8

 Hubel, D., & Wiesel, T. (1959). Receptive fields of single neurones in the cat’s striate cortex. The

Journal of Physiology, 574–591. Retrieved from http://jp.physoc.org/content/148/3/574.full.pdf

 Knierim, J. J., & van Essen, D. C. (1992). Neuronal responses to static texture patterns in area V1 of the alert macaque monkey. Journal of Neurophysiology, 67(4), 961–80. Retrieved from

http://www.ncbi.nlm.nih.gov/pubmed/1588394

 Lamme, V. (1995). The neurophysiology of figure-ground segregation in primary visual cortex. The

Journal of Neuroscience, (February). Retrieved from

http://www.jneurosci.org/content/15/2/1605.short

 Lamme, V. a, Zipser, K., & Spekreijse, H. (1998). Figure-ground activity in primary visual cortex is suppressed by anesthesia. Proceedings of the National Academy of Sciences of the United States of

America, 95(6), 3263–8. Retrieved from

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=19730&tool=pmcentrez&rendertype=abs tract

 Lamme, V., & Roelfsema, P. (2000). The distinct modes of vision offered by feedforward and recurrent processing. Trends in Neurosciences, 23(11), 571–579. Retrieved from

http://www.sciencedirect.com/science/article/pii/S016622360001657X

 Logothetis, N. K., Kayser, C., & Oeltermann, A. (2007). In vivo measurement of cortical impedance spectrum in monkeys: implications for signal propagation. Neuron, 55(5), 809–23.

doi:10.1016/j.neuron.2007.07.027

 Mitzdorf, U. (1985). Current source-density method and application in cat cerebral cortex: investigation of evoked potentials and EEG phenomena. Physiological Reviews, 65(1), 37–100. Retrieved from http://europepmc.org/abstract/MED/3880898

 Nassi, J. J., Lomber, S. G., & Born, R. T. (2013). Corticocortical feedback contributes to surround suppression in V1 of the alert primate. The Journal of Neuroscience : The Official Journal of the Society

for Neuroscience, 33(19), 8504–17. doi:10.1523/JNEUROSCI.5124-12.2013

 Niell, C., & Stryker, M. (2008). Highly selective receptive fields in mouse visual cortex. The Journal of

Neuroscience, 28(30), 7520–7536. doi:10.1523/JNEUROSCI.0623-08.2008

 Quiroga, R. (2009). “Wave clus, 5, 1–6. Retrieved from

http://scholar.google.com/scholar?hl=en&btnG=Search&q=intitle:Wave+clus#1

 Ringach, D. (2004). Mapping receptive fields in primary visual cortex. The Journal of Physiology, 558(Pt 3), 717–28. doi:10.1113/jphysiol.2004.065771

 Roelfsema, P. R. (2006). Cortical algorithms for perceptual grouping. Annual Review of Neuroscience,

29, 203–27. doi:10.1146/annurev.neuro.29.051605.112939

 Roelfsema, P. R., Lamme, V. a F., Spekreijse, H., & Bosch, H. (2002). Figure-ground segregation in a recurrent network architecture. Journal of Cognitive Neuroscience, 14(4), 525–37.

doi:10.1162/08989290260045756

 Sakai, K., & Nishimura, H. (2006). Surrounding suppression and facilitation in the determination of border ownership. Journal of Cognitive Neuroscience, 18(4), 562–79. doi:10.1162/jocn.2006.18.4.562

(28)

 Self, M. W., Kooijmans, R. N., Supèr, H., Lamme, V. a, & Roelfsema, P. R. (2012). Different glutamate receptors convey feedforward and recurrent processing in macaque V1. Proceedings of the National

Academy of Sciences of the United States of America, 109(27), 11031–6.

doi:10.1073/pnas.1119527109

 Self, M. W., Lorteije, J. a. M., Vangeneugden, J., van Beest, E. H., Grigore, M. E., Levelt, C. N., … Roelfsema, P. R. (2014). Orientation-Tuned Surround Suppression in Mouse Visual Cortex. Journal of

Neuroscience, 34(28), 9290–9304. doi:10.1523/JNEUROSCI.5051-13.2014

 Self, M. W., van Kerkoerle, T., Supèr, H., & Roelfsema, P. R. (2013). Distinct roles of the cortical layers of area V1 in figure-ground segregation. Current Biology : CB, 23(21), 2121–9.

doi:10.1016/j.cub.2013.09.013

 Supèr, H., & Roelfsema, P. R. (2005). Chronic multiunit recordings in behaving animals: advantages and limitations. Progress in Brain Research, 147, 263–82. doi:10.1016/S0079-6123(04)47020-4

 Supèr, H., Romeo, A., & Keil, M. (2010). Feed-forward segmentation of figure-ground and assignment of border-ownership. PloS One, 5(5), e10705. doi:10.1371/journal.pone.0010705

 Van Hooser, S. D., Heimel, J. A. F., Chung, S., Nelson, S. B., & Toth, L. J. (2005). Orientation selectivity without orientation maps in visual cortex of a highly visual mammal. The Journal of Neuroscience : The

Official Journal of the Society for Neuroscience, 25(1), 19–28. doi:10.1523/JNEUROSCI.4042-04.2005

 Yang, W., Carrasquillo, Y., Hooks, B. M., Nerbonne, J. M., & Burkhalter, A. (2013). Distinct balance of excitation and inhibition in an interareal feedforward and feedback circuit of mouse visual cortex. The

Journal of Neuroscience : The Official Journal of the Society for Neuroscience, 33(44), 17373–84.

doi:10.1523/JNEUROSCI.2515-13.2013

 Yizhar, O., Fenno, L., & Davidson, T. (2011). Optogenetics in neural systems. Neuron, 71(1), 9–34. doi:10.1016/j.neuron.2011.06.004