To determine the minimal duration of activity in visual cortex necessary for accurate visual discrimination by the animal, we needed to develop a perceptual task that requires visual cortex. We developed a visual discrimination task in which mice are head-fixed yet free to run on a treadmill (Figure 1A and Video 1) Visual stimuli (circular patches of gratings,~30 degrees, oriented at different angles) shown on a monitor placed on the right side of the animal moved horizontally from the anterior to the posterior end of the monitor at a speed that was proportional to the running speed of the animal. One of the stimuli (a grating oriented at 90 degrees) was the target stimulus, while the other stimulus (a grating oriented at 45 degrees) was the distractor. Mice were rewarded with water for bringing the target stimulus to the center of the monitor, the reward zone, and holding it there for a minimum time set by the experimenter (~1 s; a trial in which the stimulus is held in the reward zone for at least the minimum time is called a ‘stop trial’; see Materials and methods). To start the next trial mice had to bring the stimulus out of the posterior end of the monitor and continue running for some distance. To be most efficient in this task, mice had to continue running when the distractors appeared (Figure 1B). To ensure that mice did not solve the task by using local differences in contrast between the two gratings, we varied the position of the stripes in the circular patch, i.e. the spatial phase of the grating, randomly. At the beginning of each trial the stimulus appeared at the anterior end of the monitor and was frozen (i.e. insensitive to the rotation of the treadmill) for 350 ms, after which time the stimulus could be moved by the locomotion of the mouse. This ensured reproducibility of stimulus position across trials in the initial 350 ms. Further, during the task the position of the eye varied little trial to trial during the initial 350 ms (the standard deviation of the position of the right eye over this interval across trials was 2.4 ± 0.6 degrees; mean ±std across five mice; Figure 1—figure supplement 1). Mice learned to perform the task with accuracy above 85% correct in 23 ± 7 days (mean ±std; n = 15 wild type mice; Figure 1—figure supplement 2; accuracy is defined as the average of the percentage of stop trials upon target presentation and the percentage of non-stop trials upon distractor presentation; chance level is 50%) completing on average 200 ± 30 trials each day (transgenic mice, VGat-ChR2-EYFP, learned the task in 50 ± 20 days, n = 8 mice; difference in learning rates was significant: p=0.0096, Wilcoxon ranksum test, Figure 1—figure supplement 2).

Figure 1 with 3 supplements see all

Download asset Open asset

To determine whether visual cortex (VC) is required for this visual discrimination task, we used two approaches: optogenetic silencing to determine the impact of an acute and reversible perturbation and surgical lesions to establish the effect of an irreversible ablation. We silenced cortical activity by optogenetically activating cortical inhibitory neurons (Atallah et al., 2012; Lien and Scanziani, 2013; Olsen et al., 2012) with a 1 mm optic fiber placed over the left primary visual cortex (V1) (i.e. contralateral to the visual stimulus) in transgenic mice (VGAT-mhChR2-YFP) that selectively express the microbial light activated cation channel Channelrhodopsin2 (ChR2) in inhibitory neurons (Zhao et al., 2011) (Figure 1C). In these mice, V1 activity could be completely, rapidly and reversibly silenced (see Figure 1—figure supplement 3) with a delay of 8 ms after the onset of illumination by a blue LED (450–490 nm, Lien and Scanziani, 2013; Olsen et al., 2012). Cortical silencing started 76 ± 6 ms before the stimulus appeared (mean ±std across mice) and ended just after the stimulus had exited the monitor, and was performed on a third of the trials interleaved randomly. During silencing trials, the behavioral performance of mice was severely disrupted (51 ± 3% accuracy; n = 3 mice; Figure 1D). On these trials, mice either kept on running no matter whether the target or distractor was presented (e.g. Figure 1D) or, on a fraction of trials, they sufficiently slowed down to center the grating (i.e. stop trial) but did so indiscriminately for both stimuli (p>0.16, Wilcoxon rank sum test on choice data; e.g. Figure 1D). Because the distinction between stop and non-stop trials is binary, i.e. based on a threshold duration that the stimulus spends in the reward zone, it is conceivable that while performing at or close to chance when silencing cortex, mice may still hold the target for a longer time than the distractor in the reward zone. For example, targets and distractors may both spend less than the threshold time in the reward zone and hence be categorized as non-stop trials yet the targets may spend a longer time than the distractor in the reward zone. This would imply the ability of the mouse to discriminate despite performing at chance according to the criteria of the task. An advantage of our task is that it can reveal differences in the animal’s behavior for target versus distractor that are not captured by the binary classification of stop versus non-stop trials. We thus verified that an ideal observer could not disambiguate the target from the distractor based on times spent by each of the two stimuli in the reward zone using receiver operating characteristic (ROC) analysis (see methods). The discrimination accuracy of the ideal observer was 55 ± 6% (mean ±std, n = 3 mice), hence very close to the actual performance of the task. (Figure 1D). To exclude the possibility that the optogenetic silencing simply distracted the mice from performing the task, we silenced the right visual cortex (i.e. ipsilateral to the visual stimulus) (Figure 1E). This manipulation resulted in no substantial impairment in the behavioral performance (88 ± 6% accuracy for LED trials versus 89 ± 6% for no LED trials; Figure 1E), thus showing that the impairment was specific to the visual cortex processing visual information in the contralateral hemifield.

We verified that silencing visual cortex did not affect the ability of mice to express the decision, that is, to place the stimulus at the center of the monitor. Mice were trained as above but with the target stimulus only. In other words, mice where trained to perform a simple detection rather than a discrimination task. The distance that mice had to run to start the next trial was randomly varied. On trials where the contralateral V1 was silenced, mice centered the target image almost as frequently as in control trials (Figure 1F) demonstrating that contralateral V1 is not required to express the decision.

Behavioral deficits resulting from acute perturbations of the activity of a given brain area may lead to incorrect interpretations relative to the actual role of that area for behavior (Otchy et al., 2015), since following permanent lesions of said area animal’s behavior can recover without additional training (Kawai et al., 2015). To further assess the necessity of VC in visual discrimination we trained mice to perform the visual discrimination task and, after they had reached proficiency (accuracy of 93 ± 7%, n = 4 mice), we surgically removed VC contralateral to the side of stimulus presentation (e.g. Figure 2A) and allowed the animals to recover for ten days post-surgery before behavioral testing. Lesioned animals performed at chance (p>0.3, Wilcoxon rank sum test on choice data, n = 4 mice, Figure 2B). The impairment in behavioral performance was not due to the ten day interval from the last behavioral session because trained control animals experiencing even longer intervals between behavioral sessions remained proficient (accuracy of 80 ± 10%, Figure 2C). Furthermore the behavioral impairment was not due to either anesthesia or to some unspecific impact of surgery because proficiency was preserved after removing the ipsilateral VC (accuracy of 85%, n = 1 mouse, Figure 2B) or following anesthesia to perform craniotomy for physiological recordings (80 ± 10%, see results below). Taken together, these results show that this visual discrimination task requires visual cortex.

Figure 2

Download asset Open asset

To determine over what time interval stimulus evoked spiking activity in individual V1 neurons can be used to disambiguate the target from the distractor stimulus we recorded extracellular action potentials while the animals performed the task (Figure 3A). We inserted a multichannel probe in V1 at the beginning of a behavioral session in trained mice (performance accuracy during recordings: 80 ± 10%, mean ± std; n = 9 mice). To ensure that the units were maximally excited by the stimulus, we placed the monitor so that the position of the stimulus in the initial 350 ms, when the stimulus is stationary, was superimposed on the multiunit spatial receptive field (center of stimulus was 2 ± 1 degrees from center of receptive field, mean ± std, n = 8 mice). Further, we compared the eye position during receptive field mapping (performed outside of the task) with the eye position during the task. While the animals moved their eyes more outside of the task than during the task (Figure 1—figure supplement 1), the median eye position during receptive field mapping and during the task was very similar (difference of 3 ± 1 degrees; mean ±std across five mice; Figure 1—figure supplement 1). For comparison, the size of the receptive field of an individual cortical neuron is 12–20 degrees (Niell and Stryker, 2008). Thus, the position of the stimulus was well aligned with respect to the center of the receptive field. The cortical response to the visual stimulus began 40 ± 5 ms after stimulus onset (mean ±std across mice, Figure 3C) consistent with previous reports (Niell and Stryker, 2008). The onset of cortical response was quantified as the earliest deflection in the local field potential that exceeded three standard deviations from baseline. We verified that the earliest deflection corresponded to layer 4 of V1, the major thalamo-recipient layer, based on current source density analysis (Niell and Stryker, 2008) (Figure 3—figure supplement 1B).

Figure 3 with 2 supplements see all

Download asset Open asset

To determine whether the spiking of an individual neuron allows an ideal observer to discriminate the target from the distractor stimulus we performed ROC analysis (Tolhurst et al., 1983) on 72 well isolated units in nine behaving animals (Figure 3B,D–E; see Materials and methods for cell type and layer distribution). About half of the neurons (46%) discriminated the target from the distractor when their activity was integrated over a time window of 300 ms, starting at the onset the cortical response and ending just before the stimulus could be moved by the animal (Figure 3E) (p<0.012, Wilcoxon ranksum test on spike counts across trials, Benjamini-Hochberg correction for multiple comparisons). Below we refer to these units as ‘discriminating units’. How early do discriminating units start discriminating? We performed ROC analysis at various intervals from the onset of the cortical response (Figure 3E). The fraction of discriminating units increased rapidly between 40 and 80 ms (Figure 3F). While at 40 ms after the onset of the cortical response only ~20% of the discriminating units discriminated the target from the distractor above chance, by 80 ms already ~50% of units did so with a median discrimination accuracy of 66% (range: 58–79%). The fraction of discriminating units discriminating increased more slowly following these initial 80 ms. By 300 ms (when, per definition, 100% of discriminating units are discriminating) they reached a median discrimination accuracy of 74% (range: 58–96%). Thus, already by 80 ms following the onset of the cortical response ~50% of discriminating units discriminate the target from the distractor.

To determine how well the orientation tuning curve of a neuron predicts its ability to discriminate we measured the tuning properties of discriminating neurons after the end of the behavioral session. We presented drifting gratings of twelve different orientations that had the same size and spatial frequency and were presented at the same location as the stimuli used during the task, yet they were not rewarded and their location was insensitive to the movement of the wheel (passive viewing; stimulus properties: 20°/s; 0.5 s; 15° steps; chosen in a random order; drifting in either of the two directions perpendicular to the grating’s orientation). Most discriminating units that preferred the target during the task showed a peak response to orientations larger than 90 degrees (109 ± 8 degrees, median ±SEM; n = 9, four mice; Figure 3G). Furthermore most discriminating units that preferred the distractor during the task, showed a peak response to orientations less than 45 degrees (30 ± 20 degrees, median ±SEM; Figure 3G; n = 8, four mice). Non-discriminating neurons had either very sharp tuning curves peaking far away from target and distractor orientations, or flat tuning curves, or tuning curves peaking in between the target and distractor orientation (Figure 3—figure supplement 2). We compared the difference in spike number in response to grating presented at 45 and 90 degrees during passive viewing with how well discriminating units distinguish the target from the distractor during the task. The difference in spike number during passive viewing correlated with the value obtained from ROC analysis over the initial 80 ms following the onset of cortical response during the task (R² based on linear fit: 0.35; Figure 3H).

What is the minimal duration of activity in visual cortex necessary for accurate visual discrimination? And how many action potentials are fired by individual neurons during this time? If by 80 ms from the onset of visually evoked cortical activity information about stimulus identity is available to an independent observer, it may also be available to the mouse. Thus, the minimal duration of visual cortical activity enabling discrimination may be around 80 ms.

To control the duration of the visually evoked cortical response we optogenetically silenced visual cortex, as described above, at varying intervals after the onset of the response (Figure 4A). In each experiment we ensured that the LED intensity was sufficiently high such that performance accuracy was at chance when the illumination started before the stimulus appeared (p>0.05, Wilcoxon ranksum test on choice data, n = 8 mice). Furthermore, as above, for each animal we verified that despite chance performance the hold times of the stimulus in the reward zone of the monitor did not differ between target and distractor stimulus (p>0.05, Wilcoxon ranksum test on stimulus centering times in the reward zone). We verified this again at the very end after testing all LED onset intervals.

Figure 4

Download asset Open asset

The accuracy of the behavior increased with increasing interval between the onset of the cortical response to the stimulus and the onset of cortical silencing (Figure 4B,C). When cortical silencing followed the onset of cortical response by 44 ± 6 ms the performance was close to chance (54 ± 5%; mean ±std across mice, Figure 4D), similar to when the LED onset preceded stimulus presentation (51 ± 3%; mean ±std across mice). Strikingly, however, when cortical silencing was delayed by a further 40 ms, hence with a latency of 80 ms after the onset of the cortical response, performance accuracy of the animals sharply increased to 76 ± 7% (mean ±std across mice). Performance accuracy continued to increase, yet less sharply, over the longer intervals tested reaching 92 ± 5% when the LED onset followed the onset of the cortical response by 300 ms (mean ±std across mice). With this interval the animals performed similarly to control conditions, in the absence of LED illumination (94 ± 2%; mean ±std across mice). Thus, there is a sharp increase in performance when visual cortex is allowed to function between 44 and 80 ms after the onset of the cortical response. As above, we used ROC analysis to compare behavioral performance with the ability of an ideal observer to disambiguate the target from the distractor based on times spent by each stimulus in the reward zone when silencing cortex at 44 ms following the onset of the cortical response. The discrimination accuracy of the ideal observer was 54 ± 7%, hence very close to the actual performance of the task at 44 ms (54 ± 5%). These results show that the minimal duration of visually evoked activity in V1 for an animal to perform the present task above chance lies between 40 and 80 ms.

If the estimated time window indeed approximates the threshold duration of V1 activity for perceptual discrimination, performance accuracy in trials when V1 is active for only 80 ms should be very sensitive to the difficulty of the task. We thus trained mice to discriminate a narrower angle difference between target and distractor, namely 15 degrees. Mice were first trained to perform the standard 45 degrees discrimination task and their behavioral performance measured across various intervals of cortical silencing, as above. We then re-trained those same animals to discriminate a target from a distractor separated by 15 degrees until they reached a similar level of proficiency as for the 45 degrees task (accuracy of 90 ± 4% for 15 degrees versus accuracy of 93 ± 2% for 45 degrees, mean ±std, n = 3 mice, Figure 5B). We silenced the cortex of these animals at various intervals following the onset of the cortical response and compared the decrease in performance between the 45 and the 15 degrees discrimination tasks. Silencing cortex at 80 ms after the onset of the cortical response reduced performance significantly more for the 15 degrees as compared to the 45 degrees discrimination task in all animals (p<0.05, Wilcoxon ranksum test on choice data, n = 3 mice Figure 5B,C). While silencing V1 80 ms following the onset of the cortical response still enabled the 15 degrees discrimination to occur above chance (p<0.02, Wilcoxon ranksum test on choice data, n = 3 mice), the accuracy was significantly lower than for 45 degrees discrimination (p<0.02, Wilcoxon ranksum test on choice data, n = 3 mice, Figure 5B,C). This difference cannot be accounted for simply by a difference in motivation or in control performance because in two out of three mice, non-LED trials during the 15 degrees discrimination task were as accurate as non-LED trials during the 45 degrees discrimination task (Wilcoxon ranksum test on choice data, p=0.65 and 0.67, Figure 5B). Thus, these experiments demonstrate that the time between 40–80 ms following the onset of the cortical response indeed captures the threshold duration of V1 activity for a simple perceptual discrimination.

Figure 5

Download asset Open asset

Over these initial 80 ms from the onset of the cortical response discriminating units in primary visual cortex fired only 25 ± 4% of all the spikes fired above baseline during the 300 ms window (mean ±sem across units, Figure 3—figure supplement 1D). During this 80 ms interval discriminating units fired 0.6 ± 0.1 (median ±SEM across units) action potentials in response to their preferred stimulus (Figure 6F, response not different for the wild type and the transgenic mice as shown in Figure 3—figure supplement 1C), corresponding to a firing rate of 7.5 Hz, and 0.22 ± 0.06 action potentials for their non-preferred stimulus. Furthermore, in response to their preferred stimulus, discriminating units fired only one or no action potential in 80% of the trials and two action potentials in only 11% of the trials (Figure 6E), similar to what is expected by Poisson statistics (% of variance explained across units: R² = 97 ± 5%, median ±SEM; median time until first action potential: 70 ± 10 ms and 80 ± 20 ms from the onset of the cortical response for the preferred and non-preferred stimulus, respectively (median ±SEM across units; analysis performed over the initial 300 ms from the onset of the cortical response, Figure 6C); mean latency difference: 12 ± 6 ms (mean ±sem across units; p=0.03; t-test; Figure 6D)). Thus, over the initial 80 ms from the onset of the cortical response the vast majority of discriminating units in primary visual cortex get to fire either one or no action potentials.

Figure 6 with 1 supplement see all

Download asset Open asset

To determine whether indeed the first action potential in response to a stimulus is sufficient to discriminate the target from the distractor we performed ROC analysis (Figure 6G) after removing from each unit all but the first action potential after the onset of the cortical response. As above we performed this analysis for various intervals from the onset of the cortical response. The first action potential was sufficient for ~33% of units to discriminate by 300 ms (compared to 46% if all the action potentials were available), and more than half of those units (54%) could discriminate above chance at 80 ms (Figure 6H). Thus for most units the first action potential substantially contributes to their ability to discriminate within the initial 80 ms after the onset of the cortical response.

Finally, the accuracy of the behavioral response during the initial 80 ms can be explained by pooling the activity of ~5 discriminating neurons on average (Figure 6—figure supplement 1B), or ~20 neurons if non-discriminating neurons are also included in the pool (Figure 6—figure supplement 1C).

Taken together, these results show that the threshold duration of visually evoked cortical activity for a simple visual discrimination lies between 40 and 80 ms, a time window during which most individual cortical neurons get to fire one or no spike.

All experimental procedures were approved by the University of California San Diego Animal Care and Use Committee (protocol number S02160M). Mice were on a 12 hr light/dark cycle, lights on at eight pm. Training and experiments were performed during the dark cycle. Mice were single-housed. Data were collected from C57BL6 mice (Charles River Laboratories) or for optogenetic silencing, VGat-ChR2-EYFP mice, which have ChR2 targeted to the Slc32a1 locus (Jackson Laboratories; stock number: 014548). All mice were male and adults (2–5 months old) at the start of experiments.

Headbar implantation: Each animal was implanted with a custom made headbar for head-fixation. Briefly, animals were anesthetized with 2.5% isoflurane. Body temperature was controlled by a thermal blanket connected to a rectal thermometer (FHC; DC Temperature Controller). To expose the skull, the skin and periosteum were removed. The gap between the edge of the skull and skin was sealed with Vetbond (Fisher Scientific). The headbar was affixed to the skull with Krazy glue. Dental cement (Lang Dental; Ortho-Jet BCA) was mixed with black ink and applied to reinforce the affixation of the headbar. Animals were allowed to recover for at least 3 days before the start of water restriction (1 ml/day). Craniotomy: On the day before the extracellular recording, animals were anesthetized as above and a craniotomy was made over V1 (size: 400 µm x 1 mm, anterioposterior x mediolateral, center approximately 2.3 mm from midline and 1.3 mm from lambdoid suture). At the end of the craniotomy, to protect the brain the craniotomy was covered with a drop of artificial cerebrospinal fluid (ACSF; 142 mM NaCl, 5 mM KCl, 10 mM D-glucose, 10 mM Hepes, 3.1 mM CaCl2, 1.3 mM MgCl2) and Kwik-Cast (WPI). Cortical Ablation: The animals were anesthetized as above. Using stereotaxic coordinates the outline of visual cortex (from Paxinos and Franklin mouse brain atlas (Paxinos and Franklin, 2007)) was marked at the surface of the skull. Using a dental drill (700 µm) the area of visual cortex was thinned and removed. Sterile PBS was used to hydrate the exposed brain area. A cut of 1 mm depth was performed around the outline of VC using a microsurgical blade (FST). The cortical tissue was removed using a spoon shaped microsurgical blade (FST 10317–14). The area was washed with PBS to remove blood and consequently covered with Silicon Kwik-Cast (WPI). Upon polymerization a layer of cyanoacrylate glue was applied to cover the lesioned area. An additional layer of dental cement was applied to permanently cover the lesioned site.

A schematic of the setup is shown in Figure 1A. Mice ran on a custom made flat transparent disc, or wheel (diameter: 15 cm). The wheel was mounted on the shaft of a rotary encoder (MA3-A10-125-B, US Digital), which provided an analog output voltage proportional to the absolute shaft position. The encoder was mounted via an adaptor to a small Noga arm (MSC; part number: 09560459). Data were acquired with a National Instruments data acquisition board (NI USB-6009). To deliver water (~10 µl/reward) we used gravitational flow under the control of a solenoid valve (NResearch; Model 161K011; valve driver: CoolDrive). The valve was connected to a lickspout (hypodermic tubing; gauge 14) via Tygon tubing (1/16 inch ID).

Visual stimulation. Visual stimuli were presented on an LCD monitor (20.5 × 11.5 inches, 1920 × 1080 pixels, 60 Hz refresh rate, gamma corrected mean background luminance: 47 cd/m² for optogenetic silencing for 6 mice and 120 cd/m² for two mice, and 110 cd/m² for electrophysiology). The anterior edge of the monitor was positioned 25 cm from the right eye and the monitor subtended 50 degrees to 150 degrees of the visual field. The monitor was placed on the right side of the animal such that the antero-posterior body axis had a 15 degrees angle relative to the horizontal axis of the monitor with the axes converging rostrally. During the recording, the monitor was moved slightly so that the stimulus when stationary, i.e. in the first 350 ms (Figure 1A), overlapped with the multiunit spatial receptive field (see ‘Extracellular Electrophysiology’). To quantify how well the stimulus was centered relative to the center of the receptive field, at the end of the recording session we presented black and white squares of 3.6° in a grid of 9 × 9 locations (one stimulus at a time) covering the whole stimulus area for four mice and squares of 6° in a 5 × 5 grid for four mice. The stimuli were generated using PsychToolbox (Brainard, 1997) and custom written software in Matlab (Mathworks; code written by Shawn Olsen and available at https://github.com/aresulaj/ResRueOlsSca18; Behavior visual stimulation code and Passive visual stimulation code; Resulaj, 2018. Copy archived at https://github.com/elifesciences-publications/ResRueOlsSca18).

For the behavioral task, stimuli were circular patches of static sinusoidal gratings (spatial frequency: 0.146 cycles/degree, diameter: ~30°, contrast: 50%). On each trial, the spatial phase of the grating was chosen randomly out of 7 evenly spaced phases. We monitored the timing of stimulus onset by placing a photodiode (response time 15 ns; PDB-C156-ND; Digikey) at the bottom anterior part of the monitor, where a white square appeared concurrently with the stimulus after a 5 ms delay (accounted for in our analysis). The horizontal motion of the stimulus was controlled by the running of the animal, and updated at ~20 Hz (monitor refresh rate: 60 Hz). The gain, defined as stimulus displacement on the monitor (cm)/running distance (cm) varied from 0.3 to 0.6 across animals depending on how fast each animal ran (Supplementary file 1). The distance between two consecutive stimuli in the track was 1.25 times the width of the monitor.

For the recording sessions, the spatial phase of the grating was constant and did not vary trial to trial. During passive viewing (Figure 3G,H and Figure 3—figure supplement 2), the stimuli had the same size and spatial frequency and were presented at the same location as the stimuli used during the task during the initial 350 ms, yet the stimuli were not rewarded and their location was insensitive to the movement of the wheel (passive viewing). We presented drifting gratings of twelve different orientations (3 Hz thus 20 degrees/s; 0.5 s; 15^o steps; chosen in a random order; drifting in either of the two directions perpendicular to the grating’s orientation). The duration of the inter trial grey screen was 0.75 s. We presented ~30 repetitions/direction. The size and spatial frequency of the stimulus as projected on the retina will vary depending on its position on the monitor. However data on electrophysiological recordings only report activity for a specific position of the stimulus on the monitor, when the stimulus is fixed. This position is the same for stimuli presented during the task and during passive viewing.

Eye movements were monitored as previously described (Liu et al., 2016). Briefly, we used a high speed infrared (IR) camera (Imperx IPX-VGA 210; 200 Hz) to capture the reflection of the right eye on an IR mirror (Edmund Optics #64–471). Images were acquired using National Instruments PCI-6036E and custom written software in LabVIEW (National Instruments). Requests for eye tracking code should be addressed to Satoru Miura (Satoru.Miura@ucsf.edu), who wrote the code and kindly shared it with us. The pupil was identified by thresholding the pixel values. The eye position was quantified as the distance between the center of the pupil and the center of the corneal reflection of a reference IR LED placed above the camera along the optical axis of the camera. The position was calibrated by swinging the IR LED and the camera by ±10 degrees along a circumference centered on the image of the eye. Because eye movements impair the calibration and our mice move their eyes less during the task than outside of the task, (Figure 1—figure supplement 1) the calibration was performed for each mouse during the behavioral task the day before the measurement were taken. Following calibration, the camera was locked in place.

An optical fiber (1 mm) coupled to a blue LED (470 nm; Doric Lenses) was placed over V1 above the intact skull covered with a thin layer of Krazy glue. The fiber was placed at approximately the retinotopic location corresponding to the stimulus during the initial 350 ms in the task: ~2.3 mm from midline and ~1.3 mm from lambdoid suture. To find these coordinates, we recorded multiunit activity in V1 with the monitor in the same position as during optogenetic silencing (monitor was moved <15 degrees to center the spatial receptive field of the multiunit activity). We used these same approximate coordinates for all of our recordings.

For each animal, the total power was increased until the performance was at chance level when the LED illumination started before the stimulus appeared (3.3–20 mW across animals; p>0.05; Wilcoxon ranksum test on stimulus centering times in the reward zone). To turn the LED on at specific delays after stimulus onset, the photodiode signal detecting the onset of the stimulus was sent to an amplifier (Newark; TWLUX - TW-MF2CAB) and then to an external microprocessor (Mega 1280; Arduino). The microprocessor waited for the amplified photodiode signal to cross a threshold before sending out a digital trigger to the LED driver (Thorlabs). The jitter (s.d.) of the LED onset was 4 ms.

Training began after animals had been on water restriction for at least 7 days (~1 ml water/day). During training, mice were kept at 80% or above of their initial body weight. Additional water was provided if the body weight fell below 80% of the initial weight. The initial behavioral parameters were: 100% contrast, gain (gain = stimulus displacement in the monitor (cm)/running distance (cm)): 0.6, hold time (minimal time in reward zone for a reward): 0.2 s. With these parameters, mice would get a water reward every time a new target stimulus would pass the reward zone, that is, as long as they kept running. After mice began to run consistently (one to a few sessions), the gain was decreased to 0.45 and the hold time was initially increased to 0.4 s to get the first ~10 rewards each session (i.e. during the warm up period) and then increased to 0.9 s. Mice learned to perform the task, that is to hold the target in the reward zone for at least the minimal hold time for a reward, but not the distractor, with accuracy >85% in 23 ± 7 days (mean ±std; n = 15 wild type mice) completing on average 200 ± 30 trials each day (transgenic mice learned the task in 50 ± 20 days, Figure 1—figure supplement 2). Over this period, if mice were running too fast and not stopping on a substantial fraction of targets, the gain was decreased (lowest value: 0.3). Conversely, if mice were stopping on a substantial fraction of distractors, the hold time was increased (up to a value of 1.5 s). After mice achieved ~85% accuracy, stimulus contrast was decreased to 50%. Mice easily generalized to stimuli of 50% contrast.

The animal’s discrimination accuracy based on the binary classification of stop versus non-stop trials could be lower than that of an ideal observer monitoring the time that the stimulus spends in the reward zone. This could be the case if, for example, on some target trials the mouse slows down more than it would for a distractor trial but not sufficiently so for the target to spend the minimal amount of time in the reward zone and hence be classified as a stop trial. This scenario would lead to an underestimate of the animal’s ability to discriminate. For cortical silencing, when the LED illumination started before the stimulus appeared, the animal’s discrimination accuracy was similar to that of an ideal observer based on ROC analysis of the stimulus centering times in the reward zone. However, when the LED illumination started after the stimulus appeared, particularly for intervals longer than 80 ms from onset of the cortical response, the animal’s discrimination accuracy was usually noticeably lower than that of an ideal observer (>10% difference). This difference would often occur because on some target trials mice would not slow down sufficiently for the trial to be a ‘stop trial’ but they would slow down more than they would for distractor trials. Thus, an advantage of our task is that it revealed differences in the animal’s behavior for target versus distractor that were not captured by the binary classification of stop versus non-stop trials.

To motivate mice to make choices with similar accuracy to that of an ideal observer monitoring the time that the stimulus spends in the reward zone, we adjusted the probability that an image would be a target and the minimum time that the target had to be centered for a reward (hold time). Decreasing target probability increases the gap between rewards. With decreasing probability of the image being a target, any miss is accompanied by a longer average time interval until the next target arrives. Similarly, as the probability of the image being a distractor increases, a given false alarm rate increases the average time until the next target is available. The parameters were adjusted until for LED illuminations starting 301 ms after cortical onset (Figure 3) mice made choices that had an accuracy similar to that of an ideal observer (difference did not exceed 7%). Adjusting parameters for the longest interval between the onset of the cortical response and the LED illumination (301 ms) was usually sufficient for the discrimination accuracy by the animal to be similar to that of an ideal observer for shorter intervals too. For data collection, parameters were kept constant across different intervals. For a list of final parameters for all mice included in the main experiments see Supplementary file 1. The probability that a stimulus would be a target varied from 25–50% across mice, and the hold time varied from 0.6 s – 1.0 s across mice. Each interval was tested for 1–3 sessions totaling 130 ± 70 trials (range: 42–372 trials per interval), and data were pooled together for analysis.

On the day of the recording the Kwik-Cast was removed, ACSF was added and, before the recording electrode penetrated the brain, a drop of 1% agarose (Type IIIA; Sigma-Aldrich) was added to reduce movement artifacts. The recording electrode was a NeuroNexus 32 channel linear probe (A1 × 32-Edge-5mm-20–177) that span 620 µm in depth across cortical layers. The probe was inserted approximately perpendicular to pia and lowered to a depth of ~850–900 µm (the curvature of the V1 surface was estimated using the Franklin and Paxinos brain atlas (Paxinos and Franklin, 2007)). The probe was connected with a Plexon adaptor to two 16-channel A-M Systems headstages (gain 20x) and then connected to two 16 channel A-M Systems amplifiers (Model 3600; gain: 500x, high pass filter: 0.3 Hz, low pass filter: 5 kHz). The voltage signals were acquired with a National Instruments data acquisition board and extracted with custom written software in Matlab (code written by Shawn Olsen and included in https://github.com/aresulaj/ResRueOlsSca18; Electrophysiology Acquisition; Resulaj, 2018. Copy archived at https://github.com/elifesciences-publications/ResRueOlsSca18).

Data collection began at least 30 min after insertion of the probe. We first presented black or white squares of ~10° to map the location of the receptive field across all channels of the probe. To ensure that all receptive fields overlapped with the stimulus position during the behavior in the first 350 ms of a trial (Figure 1A), we either moved the monitor slightly if the movement was approximately <15°, or reinserted the probe at a different location mediolaterally. We mapped the receptive field at a higher spatial resolution at the end of the recording for eight mice (for same units and same stimulus and monitor position, see ‘Visual Stimulation’).

Mice were transcardially perfused with 4% paraformaldehyde (PFA) in PBS. The brain was post-fixed in 4% PFA overnight in 4°C and then cut in 100 μm thick coronal sections using a vibratome. To estimate the extent of the lesion, consecutive sections were used. All mice had lesions in V1 and surrounding V2 areas. The lesion extended slightly into the following areas (as outlined in the Paxinos and Franklin mouse brain atlas (Paxinos and Franklin, 2007)): hippocampus (2/5 mice), retrosplenial cortex (2/5 mice), primary somatosensory cortex (2/5 mice), primary and secondary auditory cortex (1/5 mice), temporal association cortex (1/5 mice), parietal association cortex (2/5 mice), dorsal subiculum (1/5 mice), postsubiculum (2/5 mice).

We visualized the positioning of the stimulus on the monitor by plotting the position of the leading (caudal) edge of the stimulus (Figures 1 and 3). A stop trial is defined as the stimulus spending ≥the minimal time for reward in the reward zone (500 pixels wide). Error bars in stop probability (Figures 1, 2, 4 and 5) indicate 95% confidence intervals assuming a binomial distribution of stops and non-stops at each orientation (data pooled from all sessions that each condition was tested). Accuracy for the target stimulus is defined as the percentage of stop trials upon target presentation; accuracy for the distractor stimulus is defined as the percentage of non-stop trials upon distractor presentation. Overall accuracy is taken as the average of these two choice accuracies; chance level is 50% correct.

We used UltraMegaSort (Fee et al., 1996; Hill et al., 2011) to sort spike waveforms into clusters. We then manually sorted the clusters into putative units. Units were accepted as well isolated according to the following criteria. First, the spike waveform across four channels had to be different than that of neighboring units in the cluster. If there were similarities, the orientation tuning curves had to be different otherwise the units were merged. Second, the refractory period violations for each unit did not exceed 0.1% (except for Figure 1—figure supplement 3 where threshold was 0.2%). To reach this criteria, we occasionally removed outliers (Hill et al., 2011) identified using the distribution of the Mahalanbois distance of spike waveforms from the cluster center. Third, for each unit, the fraction of spikes with amplitudes below detection threshold (4 s.d. of high frequency noise) did not exceed 15% (including any removed outliers) as estimated by a Gaussian fit to the distribution of spike amplitudes. Finally, we ensured that the spike waveform was stable for the duration of trials in our analysis.

Out of 98 well isolated units (n = 9 mice, one recording/animal), 12 units (12%) were putative inhibitory units based on spike waveform (Figure 1—figure supplement 3). Isolated units spanned all cortical layers.

To estimate the onset of the cortical response (Figure 3C), we first averaged for each channel the raw voltage traces across trials. We then filtered the averaged trace (fourth order Butterworth filter, low pass frequency cutoff: 300 Hz, applied bidirectionally) to get the local field potential. The filtered trace was almost indistinguishable from the raw trace (blue versus black, Figure 3C). The onset of the cortical response was defined as the earliest deflection in the filtered traces exceeding 3 s.d. from baseline. The baseline was computed for each channel as the average over the interval −80 ms to 20 ms from stimulus onset.

The current source density analysis based on the averaged traces was computed as described before (Niell and Stryker, 2008; Reinhold et al., 2015). For the average CSD across mice in Figure 3—figure supplements 1B, three mice were excluded because either the recording electrode did not span the same depth of cortex as other recordings (n = 1 mouse) or there was noise in a few channels (n = 2 mice); both cases would lead to discontinuities in the average CSD if included.

To quantify for each recording how far the center of the receptive field was from the center of the stimulus, at each stimulus location (black or white squares in a 9 × 9 grid in 4 mice and 5 × 5 grid in four mice covering the whole stimulus area; see 'Extracellular Electrophysiology') and for each channel we calculated the baseline subtracted response over a window of 170 ms (stimulus duration: 120 ms), and then normalized the responses to the peak response. After, for each location we averaged the responses across all channels. We then computed for each location the average of the average response to the white squares and the average response to the black squares. Lastly, we estimated the center of the receptive field by calculating the center of mass from the average responses at different locations. The distance from the center of the receptive field to the center of the stimulus was 2 ± 1 degrees (mean ±std, n = 8 mice).

We computed the area under the receiver operating characteristic (ROC) using function perfcurve in Matlab. To exclude the possibility that this analysis was not sensitive enough for low firing units, out of 98 well isolated units, 13 (all regular spiking) were excluded because they fired <1 spike every six trials over the initial 300 ms. This threshold was chosen because units firing at rates just above this threshold could discriminate (p<0.012; Wilcoxon ranksum test comparing the distributions of the number of action potentials for target versus distractor). We confirmed that the distribution of running speeds for the two stimuli was not significantly different in the initial 350 ms and thus did not affect our ROC analysis (p>0.02, Wilcoxon ranksum test using the Benjamini-Hochberg correction for multiple comparisons, n = 9 mice).

To determine how many neurons are needed to explain behavior, we first artificially increased the number of units by randomly shuffling the trials of each unit to get six new units. We increased the total number of units to 231 units for the pool containing discriminating units only and 504 units for the pool containing both discriminating and non-discriminating units. We then randomly picked N units, where N was 2, 5, 10, 20, 50, 100, or 200. For all units preferring the distractor, we switched the target responses with the distractor responses. Next we created a 'pooling neuron' that had for each trial all the spikes of all N neurons. We then performed ROC analysis on the spikes of this pooling neuron. We repeated the random sampling step 1000 times and averaged the resulting areas under the ROC curve. We repeated the whole procedure 10 times. Error bars in Figure 6—figure supplement 1 are sem from these 10 repetitions.

To compute the time of the first spike for the preferred versus the non-preferred stimulus (Figure 6B–D) we quantified for each unit the time of the first spike for each trial over the initial 300 ms following the onset of the cortical response. For each unit we then normalized the distribution of these spike times (total number of spikes = 1) and then averaged across units the fraction of spikes in each time bin (20 ms bins, Figure 6C). We also show the mean of the spike times for each unit and the distribution of these means across units (Figure 6D). To assess whether the first spike occurred earlier for the preferred versus the non-preferred stimulus, for each unit we computed the difference in the mean time of the first spike for the preferred versus the non-preferred stimulus and tested whether the mean of the differences from all units was different than zero (Student’s t-test).

To compute orientation tuning curves, the firing rate for each orientation was calculated over the initial 330 ms following cortical onset (i.e. the first cycle of presentation), averaged across repetitions, and normalized by the maximal firing rate across orientations.

To compute the preferred orientation for each unit, we used the following equation (Lien and Scanziani, 2013):

We computed orientation selectivity index (OSI) using the equation (Lien and Scanziani, 2013):

To evaluate the dependence of the area under the ROC curve over the initial 80 ms after cortical onset during the task versus the difference in the number of action potentials in response to passively viewed stimuli of 45° and 90°, we fitted a linear function using least squares estimation (Dobson and Barnett, 2008). The standard errors of the slope and offset parameters of the fit were based on the inverse of the information matrix (Dobson and Barnett, 2008). The slope was significantly larger than 0 (p<0.05; t-test).

The stated p-values are the results of the Wilcoxon ranksum test unless otherwise noted. For medians, we report standard errors calculated using bootstrapping.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We thank J Evora for help with mouse husbandry, Jeffery Isaacson, Lindsey Glickfeld and Michael Shadlen for comments on the manuscript, Maria Dadarlat and members of the Scanziani lab and Isaacson lab for helpful discussions, Satoru Miura, Guy Bouvier and Baohua Li for help with headbar implantations and eye tracking, and Yi Li for help with behavioral training and histology. We are grateful to the undergraduate students that helped with training of the mice. This project was supported by the Gatsby Charitable Foundation and the Howard Hughes Medical Institute.

Animal experimentation: All experimental procedures were approved by the University of California San Diego Animal Care and Use Committee. (protocol number S02160M).

© 2018, Resulaj et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.