Abstract

Binocular vision requires that the brain integrate input from both eyes to form a unified percept. Small interocular differences support depth perception (stereopsis), while larger disparities can cause diplopia or binocular rivalry. The neural mechanisms by which early visual circuits process concordant versus conflicting binocular signals remain incompletely understood. Here, we used visually evoked potentials (VEPs), unit recordings, and 2-photon calcium imaging in the binocular region of mouse primary visual cortex (bV1) to examine how distinct forms of binocular disparity engage local circuits. We found that interocular phase disparities reduced VEP magnitude through decreased neuronal firing early in the response (40–80 ms after stimulus onset). Orientation disparities also decreased VEP magnitude, but via increased firing later in the response (100–200 ms). This late activity was enhanced in both regular-spiking (putative excitatory) and fast-spiking (putative parvalbumin-positive inhibitory) units. In contrast, calcium imaging revealed that somatostatin-positive interneurons were suppressed during orientation conflict. These findings suggest that phase differences suppress bV1 responses via feedforward mechanisms, while orientation disparities prolong activity through a process associated with suppression of somatostatin-positive interneurons. Our results reveal cell-type specific circuit mechanisms engaged by different forms of binocular conflict and provide a foundation for mechanistic investigations of perceptual suppression and rivalry.

Introduction

Visual processing of information coming from the two eyes requires the integration of conflicting binocular signals. Under physiologic conditions, disparities between the images projected onto each retina tend to be modest, and the brain can effectively fuse the two images into a single percept. Processing of spatial image disparities, within a range, enables the perception of depth (stereopsis) (Parker, 2007). How the brain processes conflicting visual signals has been studied since the foundational work of Hubel and Wiesel (Hubel and Wiesel, 1962). In mammals, signals from the two eyes largely first converge in primary visual cortex (V1) where monocularly responsive neurons with retinotopically aligned receptive fields synapse onto binocular neurons. Many binocular neurons are modulated by modest spatial differences between the inputs from each eye, a phenomenon known as disparity-tuning. Disparity-tuned neurons are thought to serve stereopsis (Cumming and DeAngelis, 2001; Hubel and Livingstone, 1987; Hubel and Wiesel, 1970). In mice, even relatively large spatial disparities can be integrated by neurons in V1 (La Chioma et al., 2020; Scholl et al., 2013; Scholl et al., 2017), corresponding to the short viewing distance at which mice tend to interact with objects (Samonds et al., 2019).

Although disparity-tuned neurons in V1 have been extensively studied under conditions of modest spatial offset, less is known about how cortical circuits respond when binocular inputs are strongly conflicting. In strabismus, a clinical condition defined by eye misalignment, large interocular disparities exceed the capacity for binocular integration and preclude stereopsis (Harrad et al., 1996). When disparities between the images viewed by each eye are too great, fusion breaks down into diplopia (double vision) and binocular rivalry (an alternating, bistable percept). These outcomes can arise from different types of interocular conflict, including spatial disparities (differences in the relative phase of images between the eyes) and feature disparities (such as differences in stimulus orientation), which may engage distinct cortical mechanisms. However, how early visual circuits respond to qualitatively different forms of interocular conflict remains poorly understood.

We sought to study how mammalian V1 processes large interocular image disparities using the tractability of the mouse visual system to identify the populations of neurons involved. Our goal was to identify how distinct forms of interocular conflict are represented at the level of local circuits in primary visual cortex, providing a critical step toward understanding the neural basis of perceptual suppression. Using a combination of visually evoked potentials (VEPs), unit recordings, and 2-photon (2P) calcium imaging, we characterized cortical responses to two types of interocular stimulus disparities in mice. The first—orientation-matched grating stimuli with an interocular phase difference—mimics the magnitude of spatial disparity that tends to give rise to diplopia in humans. The second—orthogonally oriented grating stimuli—employs stimulus combinations often used to study binocular rivalry in humans (Blake and Logothetis, 2002) and non-human primates (Leopold and Logothetis, 1996). Although both types of disparity led to reductions in VEP magnitude, they did so via markedly different mechanisms: phase offset stimuli suppressed the early negative component of the VEP through decreased early evoked firing, while orthogonal stimuli suppressed the late positive component through prolonged spiking activity. These stimulus-specific patterns were evident across cortical layers and in the responses of both regular-spiking (RS, putative excitatory) and fast-spiking (FS, putative parvalbumin-positive inhibitory) neurons. Furthermore, 2P calcium imaging revealed that a separate class of interneurons expressing somatostatin (SOM+) exhibited decreased activity in response to orthogonal stimuli. These findings demonstrate that V1 circuits in mice exhibit distinct patterns of activity in response to different forms of interocular conflict. This divergence is reflected in anatomically and temporally specific activity patterns across excitatory and inhibitory cell types, providing mechanistic insight into how early visual cortex processes discordant binocular input.

Materials and methodsResources availabilityLead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Eric D. Gaier.1 Key Resources are outlined in Supplementary Table 1.

Materials availability

This study did not generate new unique reagents.

Data and code availability

The code used to analyze visually evoked potentials are available on Github.2 The datasets generated during this study have not been deposited in a public repository but are available from the corresponding author on request.

Experimental design and subject details

All procedures adhered to the guidelines of the National Institutes of Health and were approved by the Committee on Animal Care at Massachusetts Institute of Technology (MIT). For local field potential and laminar probe experiments, we used male and female mice on a C57BL/6 N background (Charles River Laboratories). For principal cell calcium imaging experiments, EMX1. Cre mice (B6.129S2-Emx1tm1(cre)Krj/J, The Jackson Laboratory, RRID: IMSR_JAX:005628) and Ai93(TITL-GCaMP6f)-D; CaMK2a-tTA mice (Igs7tm93.1(tetO−GCaMP6f)Hze Tg(Camk2a- tTA)1Mmay/J, The Jackson Laboratory, RRID: IMSR_JAX:024108) were crossbred in the animal facility at MIT to produce EMX1. GCaMP6f.tTA triple transgenic mice. All EMX1. GCaMP6f.tTA mice were fed with a sterile doxycycline food pellet diet (200 mg/kg, Bio-Serv, Flemington, NJ, United States) until weaned and changed to a normal diet. For SOM + cell calcium imaging experiments, SOM-Cre mice (B6N. Cg-Ssttm2.1(cre)Zjh/J; catalog #018973, The Jackson Laboratory; RRID: IMSR_JAX:018973). The effects reported in this study did not differ qualitatively by sex, so both were combined. Animals were housed in groups of 2–5 same-sex littermates after weaning at postnatal day 21 (P21). They had access to food and water ad libitum and were maintained on a 12 h light–dark cycle.

Method detailsSurgeries

For local field potential experiments, young adult C57BL/6 mice (P31-32) were injected with 0.1 mg/kg buprenex HCl and subcutaneously (s.c.) to provide analgesia. Induction of anesthesia was achieved via inhalation of isoflurane (3% in oxygen) and thereafter maintained via inhalant isoflurane (1–2% in oxygen). Before surgical incision, the head was shaved and the scalp cleaned with povidone–iodine (10% w/v) and ethanol (70% v/v). The scalp was resected, and the skull surface was scored with a scalpel. A steel head post was affixed to the skull (anterior to bregma) with cyanoacrylate glue. Small burr holes were drilled above both hemispheres of binocular V1 (3.1 mm lateral of lambda). Tapered 300–500 kΩ tungsten recording electrodes (FHC), 75 μm in diameter at their widest point, were implanted in each hemisphere, 470 mm below the cortical surface. Silver wire (A-M Systems) reference electrodes were placed over the left frontal cortex. Electrodes were secured using cyanoacrylate, and the skull was covered with dental cement. Nonsteroidal anti-inflammatory drugs were administered on return to the home cage (meloxicam, 1.0 mg/kg s.c.). Buprenex was administered on post-operative day 1, and meloxicam on post-operative days 1 and 2. Signs of infection and discomfort were carefully monitored. Mice were allowed to recover for at least 48 h before head fixation.

For headplate surgeries for acute electrophysiology experiments, adult mice (P60-P80) were anesthetized and prepared as described above. Following scalp incision, a lidocaine (1%) solution was applied onto the periosteum, and the exposed area of skull gently scraped with a scalpel blade. A mark on the skull was made above bV1 (AP -3.6, ML + 3.1) with marker. A custom stainless-steel head plate was attached to the skull and dental acrylic (C&B Metabond Quick Adhesive Cement System) was used to form a 5 mm x 5 mm well above bV1. The well was filled with Kwik-Sil (World Precision Instruments) which was held in place using a thin bridge of Ortho-Jet dental cement (Lang Dental). Mice were returned to their home cage and monitored for at least 48 h prior to head fixation. On the morning of the acute recording, mice were placed again under isoflurane anesthesia. The Ortho-Jet bridge and Kwik-Sil cover were removed, and a ~ 2 mm craniotomy was made around the mark previously made above bV1. Kwik-Sil was once again applied, and the mice were removed from anesthesia and single-house for 2 h. The mice were head fixed, the Kwik- Sil removed, and a 64-channel laminar probe (H3, Cambridge NeuroTech) was inserted slowly (~100 μm/min) into bV1 perpendicular to the cortical surface. Recordings were obtained and the mice were euthanized by pentobarbital overdose immediately thereafter.

For cranial window implantations for two-photon calcium imaging, adult EMX1. GCaMP6f.tTA and SOM-Cre mice (P60-P80) were anesthetized and prepared as described above. A 3 mm craniotomy was made over bV1. For SOM-Cre mice, an adeno-associated virus containing the GCaMP7f gene (pGP-AAV9-syn-FLEX-jGCaMP7f-WPRE; catalog #104488- AAV9, Addgene) was loaded into a glass micropipette with a tip diameter of 40–50 μm attached to a Nanoject II injection system (Drummond Scientific). The micropipette was then inserted into bV1 (AP -3.6, ML + 3.1) layer 2/3 at depths of 300 and 200 mm below the pial surface, and 50 nL of virus was delivered at each depth. For EMX1. GCaMP6f.tTA mice with genetic expression of GCaMP, an image of the brain was taken during the surgery with a pin at the coordinates for bV1 (AP -3.6, ML + 3.1). On subsequent imaging days, the imaging field of view was matched to the vasculature at the coordinates indicated by the pin to ensure that imaging was centered on bV1. Next, a sterile 3-mm-round glass coverslip (CS-3R-0; Warner Instruments) was gently laid on top of the exposed dura mater. The coverslip was secured with cyanoacrylate glue, and a custom stainless steel headplate was attached to the skull. Once the glue had set, dental acrylic (C&B Metabond Quick Adhesive Cement System) was mixed and applied throughout the exposed skull surface. Nonsteroidal anti-inflammatory drugs were administered upon return to the home cage (meloxicam, 1 mg/kg s.c.). Signs of infection and discomfort were carefully monitored. Mice were allowed to recover for at least 3 weeks prior to head-fixation to allow time for the window to clear and for the virus to be fully expressed.

Visual stimulus delivery

Prior to stimulus delivery, mice were acclimated to head restraint in front of a gray screen for a 30-min session on each of two consecutive days. For electrophysiological experiments, mice viewed stimuli on a passive 3D monitor (Vizeo E3D420VX, 120 Hz refresh rate) with reciprocally polarized lenses placed in front of each eye. For calcium imaging experiments, mice viewed stimuli on an active 3D monitor (Acer UMFG6AAB01, 144 Hz refresh rate) with shutter lenses (modified from NVIDIA 3D Vision 2 glasses) placed in front of each eye. Visual stimuli were generated using custom software written in either C++ for interaction with a VSG2/2 card (Cambridge Research Systems) or MATLAB (MathWorks) using the Psychtoolbox extension3 to control stimulus drawing and timing and to present different stimuli to each eye. Grating stimuli spanned the full range of monitor display values between black and white, with gamma correction to ensure constant total luminance in both gray-screen and patterned stimulus conditions.

During recordings, mice were head fixed 25 cm from the screen and viewed grating stimuli with a Gabor filter to limit the size of the stimuli to a 40° field of view directly in front of each mouse. Except where noted otherwise, phase-reversing grating stimuli were presented at 1 Hz, 0.05 cycles per degree, and an orientation of 45° or 135°. Stimuli were always presented to the left eye (contralateral to electrode or window) at 100% contrast.

We observed ~10% interocular bleed-through on both 3D display systems, such that each eye received approximately 10% of the stimulus presented to the other eye. For VEP and 2-photon calcium imaging experiments, this bleed-through was not corrected. Because the contralateral eye stimulus was held constant across conditions (100% contrast), bleed-through from the contralateral eye contributed a fixed ~10% to the ipsilateral eye and therefore cannot account for differences between stimulus conditions. Bleed-through from the ipsilateral eye to the contralateral eye was limited to 1–5% effective contrast given the low ipsilateral stimulus contrasts used (10–50%). The principal findings were replicated using the corrected laminar probe display, suggesting that bleed-through does not account for the observed effects. For acute laminar recordings, we implemented a correction for this crosstalk by reducing the stimulus presented to each eye by 10% of the contrast intended for the opposite eye. This reduced interocular bleed-through to below detectable levels by eye but resulted in a slightly reduced maximum effective contrast range (0–95% instead of 0–100%).

For VEP recordings, stimuli were presented with the contrast of the right (ipsilateral) eye stimulus increasing from 0 to 50% or decreasing from 50 to 0% in increments of 10% contrast between blocks. Each block lasted 75 s with a 15 s interblock interval for 4 blocks per condition. For acute laminar recordings and calcium imaging experiments the right (ipsilateral) eye stimulus was presented at 50% contrast (for Concordant, Phase Offset, and Orthogonal conditions) or 0% contrast (for the Monocular condition). Stimuli for acute laminar recordings were presented in pseudorandomly interleaved 100 s blocks with 30 s interblock intervals for 5 blocks per condition, and stimuli for calcium imaging experiments were presented in pseudorandomly interleaved 10 s blocks with 10 s interblock intervals for 24 blocks per condition.

Electrophysiology recordings and analysis

Electrophysiological recordings were conducted in awake, head-restrained mice. Recordings were amplified and digitized using the Recorder-64 system (Plexon Inc.) or the RHD Recording system (Intan Technologies). Local field potentials were recorded from V1 with 1-kHz sampling using a 500-Hz low-pass filter. For the laminar recordings on the Intan system, we sampled at 25 kHz and used a 0.1-Hz high-pass and a 7.5-kHz low-pass filter.

All analyses were conducted using custom MATLAB code and the Chronux toolbox (Bokil et al., 2010). For laminar recordings, the raw 25-kHz data from each channel were extracted and converted to mV. Then they were down sampled to 1,000 Hz and a third order 1- to 300-Hz Butterworth filter was applied to isolate the LFP frequency band. For all data, the mean of the entire channel’s data was subtracted from each time point to account for any DC offset in the system. Next, the data were locally detrended using the locdetrend function in the Chronux toolbox using a 0.5 s window sliding in chunks of 0.1 s to remove slow non-stationary drifts not fully eliminated by the high-pass filter. Finally, a third-order Butterworth filter was used to notch frequencies between 58 and 62 Hz. For the multiunit activity of laminar recordings, the raw 25-kHz data were extracted for each channel. A 60 Hz, 10-dB bandwidth IIR notch filter was applied to each channel and the median value of each channel was subtracted from that channel. The median value across channels for each time point was subtracted from all channel’s timepoints. Visually evoked potentials were normalized by subtracting the average of the first 10 ms of each trial from that trial to set each trial’s pre-stimulus baseline to zero. A 10-ms moving Gaussian was applied to smooth the visually evoked potential waveform.

Current-source density analysis and laminar identification

Current-source density (CSD) analysis on laminar local field potential (LFP) data in V1 was used to identify layer 4 (L4) for each mouse and align our recordings (Aizenman et al., 1996; Mitzdorf, 1985, 1987). The LFP data were temporally smoothed with a 20 ms Gaussian window, then a five-point hamming window was used to compute the CSD (Ulbert et al., 2001). We define 0 mm as the site of the earliest and deepest sink immediately below the superficial source. All other channels were referenced according to that landmark. Thus, superficial channels were above 0 mm and deep channels were below 0 mm. For the sake of reporting in Results, we broke the laminar data into four segments roughly corresponding to each cortical layer (Senzai et al., 2019): L2/3 between +300 and +60 μm, L4 between +60 and −80 μm, L5 between −80 and −260 μm, and L6 between −260 and −460 μm.

Multiunit activity analysis

To measure multiunit activity, we calculated the multiunit activity envelope (MUAe), which provides an instantaneous measure of the number and size of action potentials of neurons in the vicinity of the electrode tip without the requirement of setting an arbitrary threshold level (Brosch et al., 1995; Legatt et al., 1980; Super and Roelfsema, 2005). MUAe was calculated by applying a third order bandpass (500-5000 Hz) Butterworth filter to the common median referenced data to isolate spiking unit activity. Next, the absolute values of the data were taken (units of mV) to remove contamination from far-field signals. Finally, a third order low pass (<250 Hz) Butterworth filter was applied to smooth the signal and the data were down sampled to 1,000 Hz to align with local field potential data. MUAe was z-scored based on the average and standard deviation across all interblock intervals.

Single unit activity analysis

Single unit activity was calculated for the common median referenced data using Kilosort 2.5 (Pachitariu et al., 2024; Kilosort parameters: final projection threshold = 4; spike threshold = 6; channel count = 64; clusters per channel = 4; min spike rate = 0.02; high pass filter frequency = 300; optimization projection threshold = 10; number of blocks = 5; sigma mask = 30; batch size = 1,024; whitening range = 32). Spikes identified by Kilosort were manually curated using Phy to isolate single units. Curation was performed by a single experimenter applying consistent qualitative criteria across all recordings: (1) a clear refractory period in the autocorrelogram, with near-absent ISIs in the 0–2 ms range and refractory-period contamination below 10%, (2) stable waveform amplitude and shape across the recording session, (3) continuous firing throughout the recording, without prolonged silent periods suggestive of drift, and (4) visible separation from neighboring clusters in principal component feature space. Numerical thresholds for these metrics were not pre-specified. Single units with a trough-to-peak latency of less than 0.4 ms were classified as fast-spiking, and units with a latency of greater than 0.4 ms were classified as regular-spiking. The spike rate (in spikes/s) for each unit was calculated by binning spikes into 1 ms bins and averaging across all 500 trials per condition to obtain the mean spike rate in spikes/s. Average RS and FS unit PSTHs were calculated by averaging the spike rates of each class of unit. Firing rates during early and late time bins were calculated by averaging firing rates between 40 and 80 ms after stimulus onset, and 100–200 ms after stimulus onset, respectively. Z-scored SUA was calculated based on the average and standard deviation of the firing rate of each unit across all interblock intervals.

In vivo two-photon calcium imaging

Three to 4 weeks following craniotomy surgery, mice were habituated to the behavior restraint apparatus in front of a gray screen with the objective lens of the two-photon microscope positioned on the head plate for 30 min for two consecutive days before beginning their visual stimulus delivery. A Ti:sapphire laser (Coherent) was used for imaging at a wave length of 930 nm. Photomultiplier tubes (Hamamatsu) and the objective lens (20×, 0.95 numerical aperture, XLUMPLFLN, Olympus) were used to detect fluorescence images. Calcium image recordings were triggered by time-locked transistor-transistor logic pulses generated from the USB-1208 fs data acquisition device (Measurement Computing) using PrairieView and TriggerSync software (Bruker) and imaged at a frequency of ~14.9 Hz at a depth of ~200 mm in bV1. The size of the imaging field of view was ~600 × 600 mm2 at 512 × 512 pixels.

Calcium imaging analysis

Acquired time series of calcium imaging files were processed using Suite2p (Pachitariu et al., 2017). All recorded files were registered to stabilize shifts due to animal movement. Regions of interest (ROIs) were automatically detected, then manually curated based on the maximum projection of all frames. Mice were excluded from all further analysis if they had fewer than 100 ROIs within the field of view, as this indicated excessive noise during the recording. We extracted the fluorescence of each ROI for all time points. In line with previous work, for each ROI we calculated the estimated true fluorescence of the ROI. This is the measured fluorescence of the ROI minus 7/10ths of the average measured fluorescence of the surrounding neuropil (Chen et al., 2013). We used the average interblock gray period to compute the response relative to gray (Fstim–Favg_gray)/Favg_gray. The average population response for each mouse was calculated by taking the average dF/F of all ROIs across all blocks.

Quantification and statistical analysisStatistics

Throughout the results section, all data is expressed as group mean ± standard error of the mean (SEM). Each dataset was assessed for normality and homogeneity of variance prior to choosing a statistical approach, using Levene’s, D’Agostino and Pearson, and, in the case of small sample sizes, Shapiro–Wilk tests. One-way repeated-measures (RM) analysis-of-variance (ANOVA) were used to compare group responses across conditions. Two-way RM ANOVA were used to compare group responses across conditions, stimulus contrasts, or cortical layers. Greenhouse–Geisser was used to modify the degrees of freedom of repeated-measures tests to correct for violations of sphericity. Interactions were followed by tests of simple main effects. Tukey or Sidak’s methods were applied to compensate for multiple comparisons. Uncorrected alpha was set to 0.05. Statistical analyses were performed with Prism 10 (GraphPad; RRID: SCR_002798).

ResultsInterocular phase or orientation disparities differentially reduce VEP magnitude

We first sought to determine the effect of presenting disparate stimuli to each eye on the cortical responses of mice. We implanted tungsten microelectrodes into layer 4 (L4) of bV1 (Figure 1A) and measured VEPs in awake, head-fixed mice using phase-reversing sinusoidal grating stimuli presented on a passive 3D display (Figure 1B). Reciprocally polarized lenses placed in front of each eye enabled the independent presentation of binocularly discordant stimuli. For all recordings, the eye contralateral to the recording electrode viewed full contrast stimuli at a 45° orientation. The ipsilateral eye viewed stimuli ranging from 0% contrast (i.e., grey screen) to 50% contrast. Ipsilateral eye stimuli were either in phase (concordant) or 180° out of phase (phase offset) relative to the contralateral eye stimulus (Figure 1C). We found that the VEP magnitudes (peak negative to peak positive voltage) elicited by concordant stimuli were larger than those elicited by monocular (0% ipsilateral eye contrast) or phase offset stimuli (Figure 1D). Notably, a statistically significant difference between concordant and phase offset stimuli emerged at an ipsilateral eye contrast as low as 20%.

Binocular stimuli with an interocular phase or orientation disparity elicit smaller visually evoked potentials. (A) Schematic of mouse brain showing placement of a chronic recording electrode in binocular V1 (V1b). Adapted with permission from Fong et al. (2021), licensed under CC BY 4.0. (B) Mice viewed phase reversing sinusoidal grating stimuli on a dichoptic display. (C) Phase offset experimental design. The contralateral (Contra) eye viewed full-contrast stimuli, while the ipsilateral (Ipsi) eye viewed stimuli ranging from 0 to 50% contrast, presented either in phase (Concordant) or out of phase (Phase Offset) with the contra eye stimulus. (D) For ispi eye contrasts ≥20% contrast, concordant stimuli elicited significantly larger VEPs than phase offset stimuli. Average VEP waveforms for each condition shown above bars. Two-way repeated measures ANOVA (main effect of condition, F (1,18) = 16.4, p < 0.001; condition × contrast, F (5,90) = 2.37, p = 0.045) followed by Šídák’s multiple comparisons test (concordant vs. phase offset at 0, 10, 20, 30, 40, and 50% ipsi eye contrast: p = 0.522, 0.126, <0.001, <0.001, <0.001, <0.001; Concordant 0% vs. 10, 20, 30, 40, and 50% contrast: p = 0.999, 0.088, 0.999, 0.110, <0.001; Phase Offset 0% vs. 10, 20, 30, 40, and 50% contrast: p = 0.955, 0.999, 0.678, 0.999, 0.999). (E) Orthogonal orientation experimental design. Similar to C, but the ipsi eye viewed stimuli at either the same orientation as the contra eye (Concordant) or rotated 90° (Orthogonal). (F) For ispi eye contrasts ≥10% contrast, concordant stimuli elicited larger VEPs than Orthogonal stimuli. Two-way repeated measures ANOVA (main effect of condition, F (1,23) = 32.6, p < 0.001; condition × contrast, F (5,115) = 13.8, p < 0.001) followed by Šídák’s multiple comparisons test (Concordant vs. Orthogonal at 0, 10, 20, 30, 40, and 50% ipsi eye contrast: p = 0.437, 0.037, <0.001, <0.001, <0.001, <0.001; Concordant 0% vs. 10, 20, 30, 40, and 50% contrast: p = 0.968, 0.282, 0.017, 0.003, <0.001; Orthogonal 0% vs. 10, 20, 30, 40, and 50% contrast: p = 0.137, 0.003, <0.001, <0.001, <0.001). Scale bars: 200 ms and 100 μV.

Next, we looked at how orientation differences between the two eyes affects bV1 responses. To test this, we used the same dichoptic display to present stimuli that were either at the same orientation for each eye (concordant) or rotated 90° (orthogonal). As before, the contralateral eye viewed a full contrast stimulus while the ipsilateral eye viewed stimuli ranging from 0 to 50% contrast (Figure 1E). Again, a substantial contrast-dependent difference in VEP magnitude between concordant and orthogonal ipsilateral eye stimuli emerged, with a statistically significant difference between concordant and orthogonal stimuli at just 10% ipsilateral eye contrast (Figure 1F). Relative to monocular stimuli, we also observed a robust increase in VEP magnitude for concordant stimuli, and decrease in VEP magnitude for orthogonal stimuli. These findings suggest that even low contrast engagement of the inherently weaker ipsilateral eye is sufficient to modulate cortical responses in the context of interocular stimulus disparities, particularly in the case of orthogonal stimuli.

To better understand which features of the VEP were modulated by these disparities, we separately analyzed the negative peak (most negative voltage) and positive peak (most positive voltage following the negative peak) of the waveform. For phase offset stimuli, the divergence from concordant in VEP magnitude was primarily reflected in the negative peak (Figure 2A), where concordant stimuli showed a progressive enhancement with increasing ipsilateral contrast that was not observed for phase offset stimuli, while changes in the positive peak were smaller and more variable (Figure 2B). In contrast, for orthogonal stimuli the divergence from concordant was evident in both the negative and positive components of the VEP (Figures 2C,D), with the positive component showing the larger effect. These results suggest that distinct features of the VEP are differentially modulated by interocular phase and orientation disparities.

Phase offset and orthogonal stimuli affect different components of the VEP. (A) VEP negativities (most negative voltage peak) for concordant (grey) and phase offset (purple) stimuli at different ipsilateral eye contrasts. Analyses performed using a two-way repeated measures ANOVA (main effect of condition, F (1,18) = 14.0, p = 0.002; condition × contrast, F (5,90) = 2.88, p = 0.019) followed by Šídák’s multiple comparisons test (concordant vs. phase offset at 0, 10, 20, 30, 40, and 50% ipsi eye contrast: p = 0.393, 0.800, 0.009, <0.001, 0.001, <0.001). (B) VEP positivities (most positive voltage following the negative peak) for concordant and phase offset stimuli at different ipsilateral eye contrasts. Two-way repeated measures ANOVA (main effect of condition, F (1,18) = 9.42, p = 0.007; condition × contrast, F (5,90) = 1.03, p = 0.407) followed by Šídák’s multiple comparisons test (concordant vs. phase offset at 0, 10, 20, 30, 40, and 50% ipsi eye contrast: p = > 0.999, 0.029, 0.015, 0.304, 0.349, 0.176). (C) VEP negativities for concordant (grey) and orthogonal (red) stimuli. Two-way repeated measures ANOVA (main effect of condition, F (1,23) = 11.8, p = 0.002; condition × contrast, F (5,115) = 6.09, p < 0.001) followed by Šídák’s multiple comparisons test (concordant vs. discordant at 0, 10, 20, 30, 40, and 50% ipsi eye contrast: p = 0.847, 0.586, 0.210, 0.002, 0.002, <0.001). (D) VEP positivities for oncordant and rthogonal stimuli. Two-way repeated measures ANOVA (main effect of condition, F (1,23) = 48.73, p < 0.001; condition × contrast, F(5,115) = 15.4, p < 0.001) followed by Šídák’s multiple comparisons test (concordant vs. discordant at 0, 10, 20, 30, 40, and 50% ipsi eye contrast: p = 0.437, 0.009, <0.001, <0.001, <0.001, <0.001).

Because the divergence between concordant and discordant conditions was more pronounced for orthogonal stimulus orientations than for phase offset disparities, we next sought to further characterize the features of VEP responses elicited by this stimulus combination. First, we examined how sensitive this effect was to graded orientation disparities between the eyes. Fixing the contralateral eye grating stimulus at 100% contrast at a 45° orientation and the ipsilateral eye at 50% contrast, we varied the grating orientation of the ipsilateral eye stimulus in 5° increments (from 0° to 30° differences, plus a fully orthogonal 90° difference relative to the contralateral eye) (Figure 3A). VEPs elicited by stimuli with interocular orientation differences of as little at 10° showed a statistically significant decrease relative to the 0° (concordant) condition (Figure 3B), indicating that the VEP is highly sensitive to interocular orientation disparity.

Reductions in VEP magnitude for discordant stimuli are sensitive to interocular orientation differences, spatial frequency, and binocularity. (A) Mice viewed dichoptic grating stimuli (100% contra and 50% ipsi eye contrast) with varying interocular orientation differences (0° to 30° in 5° increments, plus an orthogonal orientation). (B) VEP magnitudes were normalized to the baseline (0°) condition. Orientation differences of 10° or larger elicited significantly smaller VEPs. One-way repeated measures ANOVA with Geisser–Greenhouse correction (F(3.32, 29.8) = 12.7, p < 0.001) followed by Dunnett’s multiple comparisons test (0° vs. 5°, 10°, 15°, 20°, 25°, 30°, and 90°: p = 0.086, 0.025, 0.001, 0.005, <0.001, <0.001, 0.003). (C) Mice viewed discordant stimuli at higher spatial frequencies (0.2 and 0.4 cpd) with orthogonal ipsi eye stimuli (100% contra eye contrast, 0 to 50% ipsi eye contrast). (D) VEP magnitude was reduced for 0.2 cpd, but not for 0.4 cpd. Two-way repeated measures ANOVA with Geisser–Greenhouse correction (main effect of SF, F (1,20) = 22.66, p < 0.001; SF × contrast, F (2.92,58.4) = 4.69, p = 0.006) followed by Dunnett’s multiple comparisons test (0% vs. 10, 20, 30, 40, and 50% ipsi eye contrast for 0.02 cpd: p = 0.003, <0.001, <0.001, <0.001, <0.001; for 0.04 cpd: p = 0.701, >0.999, >0.999, >0.999, >0.999). (E) Mice viewed dichoptic grating stimuli (100% contra eye contrast) with a conflicting orthogonal grating. The orthogonal grating (0 to 50% contrast) was either presented to the ipsi eye (Binocular) or overlayed on top of the initial stimulus orientation for the contra eye (Monocular). (F) VEP magnitude for binocular and monocular conditions decreased as orthogonal grating contrast increased, but this decrease was sensitive to lower stimulus contrast in the binocular condition. Two-way repeated measures ANOVA with Geisser–Greenhouse correction (main effect of condition, F (1,23) = 4.38, p = 0.048; condition × contrast, F (3.48, 80.0) = 1.44, p = 0.235) followed by Dunnett’s multiple comparisons test (0% vs. 10, 20, 30, 40, and 50% contrast for Monocular: p = 0.817, 0.011, <0.001, <0.001, <0.001; for Binocular: p = 0.006, <0.001, <0.001, <0.001, <0.001). (B,D,F) Error bars indicate SEM. Scale bars: 200 ms and 100 μV.

Next, we examined the orientation disparity effect at spatial frequencies greater than 0.05 cycles per degree (cpd). We again presented orthogonal stimulus orientations to the eyes, with the contralateral eye fixed at 100% contrast, and the ipsilateral eye ranging from 0 to 50% contrast. Stimuli were presented at either 0.2 or 0.4 cpd (Figure 3C). At 0.2 cpd we again observed a strong reduction in VEP magnitude relative to the 0% contrast baseline, but no change was observed for 0.4 cpd (Figure 3D). This suggests that rivalrous stimulus effects on the VEP are spatial frequency-dependent, consistent with the known spatial frequency dependence of interocular suppression and binocular rivalry in humans (Fahle, 1982; Lei and Schor, 1994), though further work will be needed to determine whether the same mechanisms are involved.

Finally, we asked whether the reduction in VEP magnitude observed for orthogonal binocular stimuli could be accounted for by interocular interactions alone, or whether similar effects might also be elicited by overlapping stimuli within a single eye. To test this, we presented a full contrast grating to the contralateral eye, and added an overlapping orthogonal grating to either the same eye (monocular condition) or to the opposite eye (binocular condition), with contrast ranging from 0 to 50% (Figure 3E). In both cases, the superimposed gratings created a plaid-like pattern, differing only in whether the component gratings were presented to one or both eyes. While VEP magnitude decreased with increasing contrast in both conditions, the reduction was more sensitive to contrast in the binocular condition (Figure 3F). These results suggest that orthogonal stimuli presented binocularly produce modestly greater VEP suppression than monocular cross-oriented stimuli (main effect of condition, p = 0.048), though this difference did not depend significantly on contrast (interaction, p > 0.23).

Changes in the VEP are driven by an early decrease in V1 firing with phase disparity, and prolongation of firing with orientation disparity

While VEPs are a useful method for measuring summed neuronal responses within V1, it is difficult to infer how the underlying cortical circuitry gives rise to differential, stimulus-dependent VEP waveform changes. To better answer this question, we used 64-channel laminar probes to measure neuronal activity across all layers of V1 (Figures 4A,B). Mice viewed dichoptic phase reversing grating stimuli as described previously (see Figure 1B). The eye contralateral to the recording electrode viewed full contrast stimuli at a fixed orientation, while the ipsilateral eye viewed grey screen (monocular condition), an in-phase grating stimulus at the same angle (concordant condition), an out-of-phase grating stimulus at the same angle (phase offset condition), or an orthogonal grating (orthogonal condition). In order to test all conditions within each animal, ipsilateral eye stimuli were presented only at 50% contrast for each dichoptic condition.

L4 VEPs measured using laminar probes reveal reductions in different components of the VEP for phase offset and discordant stimuli. (A) Recording setup. LFPs were recorded from bV1 using a 64-channel silicon probe. The probe spanned all cortical layers with channel spacing of 20 μm. Adapted with permission from Fong et al. (2021), licensed under CC BY 4.0. (B,C) Monocular VEPs recorded from each electrode channel (B) were used to calculate current source densities (CSDs) shown in (C). The earliest current sink in the CSD for each mouse was used to identify the electrode corresponding to layer 4 (shown in red in B). Approximate layer boundaries shown as white dashed lines in C. (D) Stimulus conditions. Mice viewed dichoptic phase reversing gratings (2 Hz, 0.05 cpd) as described previously. Contra eye stimuli (45°, 100% contrast) were kept constant across conditions. Ipsi eye stimuli varied across 4 conditions: monocular (0% contrast), concordant (50% contrast, 45°), phase offset (50% contrast, 45° with 180° phase offset), and orthogonal (50% contrast, 135°). L4 VEPs for each condition shown below schematic. (E) VEP magnitude for the phase offset (ΔPhase) and orthogonal (ΔAngle) conditions were decreased relative to the monocular (Monoc) and concordant (Conc) conditions. One-way repeated measures ANOVA with Geisser–Greenhouse correction [F (2.20, 22.0) = 20.0, p < 0.001] followed by Tukey’s multiple comparison’s test. (F) Relative to monocular, the negative component of the L4 VEP was larger (more negative) for concordant, smaller for phase offset, and no different for orthogonal stimuli. One-way repeated measures ANOVA with Geisser–Greenhouse correction [F (2.15, 21.5) = 24.9, p < 0.001] followed by Tukey’s multiple comparison’s test. (G) Relative to monocular, the positive peak of the layer 4 VEP was slightly decreased for phase offset stimuli, and substantially diminished for orthogonal stimuli. One-way repeated measures ANOVA with Geisser–Greenhouse correction [F (2.16, 21.6) = 19.7, p < 0.001] followed by Tukey’s multiple comparison’s test. (E–G) *p < 0.05, **p < 0.01, ***p < 0.001.

Recording VEPs across all channels enabled us to calculate a current source density plot for each mouse. The electrode contact corresponding to L4 was identified for each mouse based on the earliest current sink in the current source density (CSD) plots (Figure 4C). Consistent with our results using unipolar tungsten electrodes (Figure 1), we observed a reduction in L4 VEP magnitude for the phase offset and orthogonal conditions relative to concordant stimuli (Figures 4D,E). We also found a statistically significant decrease relative to monocular stimuli, which was previously only observed for orthogonal stimuli. Again, the reduction in VEP magnitude in the phase offset and discordant conditions were caused by distinct changes in VEP waveform; phase offset stimuli primarily decreased the VEP negativity, whereas discordant stimuli primarily decreased the VEP positivity (Figures 4F,G).

To determine how changes in different components of the VEP relate to changes in neuronal spiking in bV1, we next analyzed unit activity across cortical layers. Recordings from individual mice were aligned based on the channel that showed the earliest L4 current sink, and the multiunit activity envelope (MUAe) was calculated for each electrode contact. Approximate layer boundaries were assigned based on the distance from the L4 sink and the L5 peak in MUAe (Senzai et al., 2019; Figure 5A). To account for the variability in spiking between layers, we z-scored the MUAe for each channel based on its average across the entire recording session. We observed a strong visually evoked response across all layers of V1 when mice viewed the grating monocularly through the contralateral eye (Figure 5B). This response peaked around 60–80 ms after stimulus presentation and MUAe returned to near-baseline levels by ~100 ms. To compare how activity across layers was affected by the different stimulus conditions, we then assessed MUAe when the ipsilateral eye viewed a concordant, phase offset, or orthogonal grating stimulus (Figure 5C). MUAe changed across most layers of V1 for all three conditions (Figure 5D). Relative to monocular stimuli, concordant and orthogonal stimuli increased the response in the early period following the phase reversal (roughly corresponding in time with the negative component of the VEP), whereas phase offset stimuli elicited a reduced response in this period (Figures 5E,F). Unexpectedly, the orthogonal condition prolonged firing well beyond the visually evoked response of the other conditions (Figures 5E,G). This prolonged period of firing persisted through 200 ms after the phase reversal, corresponding to the timing of the VEP positivity in response monocular and concordant stimuli (Figure 4). Thus, the reduction in the positive component of the VEP waveform by the orthogonal condition may reflect prolongation of activity.