Dynamic Patterns of Spontaneous Ongoing Activity in the Visual Cortex of Anesthetized and Awake Monkeys are Different

Abstract

Ongoing internal cortical activity plays a major role in perception and behavior both in animals and humans. Previously we have shown that spontaneous patterns resembling orientation-maps appear over large cortical areas in the primary visual-cortex of anesthetized cats. However, it remains unknown 1) whether spontaneous-activity in the primate also displays similar patterns and 2) whether a significant difference exists between cortical ongoing-activity in the anesthetized and awake primate. We explored these questions by combining voltage-sensitive-dye imaging with multiunit and local-field-potential recordings. Spontaneously emerging orientation and ocular-dominance maps, spanning up to 6 × 6 mm², were readily observed in anesthetized but not in awake monkeys. Nevertheless, spontaneous correlated-activity involving orientation-domains was observed in awake monkeys. Under both anesthetized and awake conditions, spontaneous correlated-activity coincided with traveling waves. We found that spontaneous activity resembling orientation-maps in awake animals spans smaller cortical areas in each instance, but over time it appears across all of V1. Furthermore, in the awake monkey, our results suggest that the synaptic strength had been completely reorganized including connections between dissimilar elements of the functional architecture. These findings lend support to the notion that ongoing-activity has many more fast switching representations playing an important role in cortical function and behavior.

awake primate, cortical column, spontaneous activity, visual cortex, voltage-sensitive dye

Introduction

The brain at rest, without external sensory input, continuously produces substantial internal activity (James 1890) that consumes one order of magnitude more metabolic energy than does evoked activity (Sokoloff et al. 1955; Raichle and Mintun 2006), and therefore, it most likely plays an important role in brain function. This nonevoked activity, which is observed in the absence of any overt input, is referred to as “internal,” “spontaneous,” or “ongoing.” The relationship between the dynamic patterns of internal ongoing activity and the cortical functional architecture at the cortical column level is currently unknown in awake animals.

Previous studies in anesthetized cats demonstrated that ongoing activity in the primary visual cortex (V1) is spatially structured, that is, it exhibits distinct patterns of cortical activity similar to the orientation columns. These findings suggest that ongoing activity is not “noise” but may contribute to information processing (Arieli et al. 1995, 1996; Abbott and Dayan 1999; Tsodyks et al. 1999; Kenet et al. 2003; Gutnisky et al. 2017). Following this line of thought, multiple theoretical and experimental studies performed on awake human and animal subjects found substantial evidence for a major role played by ongoing activity in shaping perception and behavior (Wagner et al. 1998; Pessoa et al. 2002; Ress and Heeger 2003; Dehaene and Changeux 2005; Pessoa and Padmala 2005; Fox et al. 2006, 2007; Hesselmann, Kell, Eger et al. 2008; Hesselmann, Kell, Kleinschmidt et al. 2008; Northoff et al. 2010; Raichle 2010; Berkes et al. 2011; Civillico and Contreras 2012; Ponce-Alvarez et al. 2013; Sachidhanandam et al. 2013; Singer 2013; Zagha et al. 2013; Zou et al. 2013; Netser et al. 2014; Zhang et al. 2014; Carcea et al. 2017; Wilf et al. 2017). However, the relationship to the functional architecture in awake animals has not yet been explored.

In particular, spontaneous activity in primary visual cortex (both areas 17 and 18) of anesthetized cats displays patterns that span large cortical areas—up to 4 × 7 mm²—and resemble the columnar functional architecture (functional maps) associated with the processing of orientation (Arieli et al. 1995, 1996; Tsodyks et al. 1999; Kenet et al. 2003; O’Hashi et al. 2017). These studies in anesthetized animals raise the question of the existence of spontaneous map-like cortical states in awake animals.

To address this question, we combined simultaneous voltage-sensitive dye imaging (VSDI) together with recordings of local field potentials (LFP) and multiunit activity (MUA) in the primary visual cortex of awake and anesthetized monkeys (Macaca fascicularis). Combining these methodologies allowed us to study the ongoing activity at 2 fundamentally different resolutions: 1) the mesoscopic, multicolumnar resolution—using VSDI and 2) a single-column resolution—using MUA for studying the local cortical circuits.

The data presented here, to the best of our knowledge, provide the first images of the mesoscopic cortical dynamics of ongoing activity in an anesthetized monkey, but most importantly—also in an awake monkey. In total we recorded 529 110 movie frames of spontaneous ongoing cortical activity during wakefulness from 3 awake monkeys and 370 140 movie frames from 3 additional anesthetized monkeys. These extensive data from 6 monkeys revealed many instances of the spontaneous emergence of cortical states resembling the classical functional architecture of orientation and ocular dominance columns in the anesthetized monkey, but not in the awake monkey, despite the fact that the response size and imaging quality for the 2 conditions were similar. Nevertheless, we found direct evidence for the existence of spontaneous iso-orientation correlated activity in the awake monkey. Under both conditions—anesthesia and wakefulness—columnar correlated activity displayed similar spatiotemporal dynamics and coincided with traveling waves over the primary visual cortex. These findings suggest that in awake animals the number of spontaneous activity patterns is much larger and that their dynamics is faster. Moreover, these results indicate that lateral cortical connectivity in the awake monkey is different and that spontaneous activity patterns in the awake animal are not directly related to classical functional connectivity of mostly like-to-like long range horizontal connection (Ts’o et al. 1986; Gilbert and Wiesel 1989) but instead to a more complex connectivity rules deviating from a like-to like organization which was recently published by 3 different groups using different approaches (Chavane et al. 2011; Huang et al. 2014; Martin et al. 2014).

Materials and Methods

Animals and Surgery

We collected data from the primary visual cortex of 3 behaving monkeys during wakefulness and from 3 additional monkeys under anesthesia. All experiments were carried out under a protocol approved by the Institutional Animal Care and Use Committee of the Weizmann Institute of Science, whose guidelines are based on the NIH guidelines. Long-term recordings and VSD imaging were conducted chronically from 3 anesthetized and 3 awake adult male Macaca fascicularis monkeys. The surgical procedure has been reported in detail previously (Shtoyerman et al. 2000; Slovin et al. 2002) and is outlined briefly here. The experimental paradigms for the visual stimulation experiments were also reported in detail previously in Omer et al. (2013).

Implant surgeries were performed under sterile conditions and under anesthesia. A head holder and 2 cranial windows for optical recordings (25 mm internal diameter) were implanted over the right and left V1. Several months after this procedure, the monkeys underwent craniotomy and the dura inside the chamber was resected to expose the visual cortex. The anterior border of the exposed cortex was always 3–6 mm anterior to the lunate sulcus. Typically, the center of the chamber was 0°–4° below the representation of the vertical meridian in V1 and 2°–4° lateral to the horizontal meridian. A thin, transparent silicone artificial dura was implanted over the exposed cortex (Arieli et al. 2002). Local and systemic antibiotics were applied according to microbiological examinations of the chamber. We used local antibiotics above the artificial dura over long periods (sometimes several months), but we did not observe any effects of the antibiotics on the quality of the collected data.

Voltage-Sensitive Dye Imaging

We used here an Oxonol voltage-sensitive dye (VSD), RH-1916 (a close analog of RH −1691 (Shoham et al. 1999), in which the para methoxy phenyl substituent on the rhodanine moiety was replaced by para methyl phenyl N-rhodanine (Fig. S1).

To minimize vibration noise in awake experiments, we fixed the head to a heavy stand with an implanted head holder. The anesthetized experiments were carried out on a vibration isolation table. For real-time optical imaging we used a custom-written program to control the sensitive MICAM Ultima high-speed camera (SciMedia, Japan), with a resolution of 100 × 100 pixels. The exposed cortex was illuminated using epi-illumination with an excitation filter (peak transmission 630 nm, width-at-half-height 30 nm) and a dichroic mirror (650 DRLP), both from Omega Optical (Bratlleboro, VT, USA). To collect the fluorescence light and reject stray excitation light, a barrier high-pass filter (RG 665; Schott, Mainz, Germany) was placed above the dichroic mirror. Data were recorded at 200 Hz and the images covered a cortical area of 6 × 6 mm².

Electrophysiological Recordings

Multiunit spiking activity (MUA) and LFP were recorded simultaneously with the VSDI data, using tungsten microelectrodes (FHC, USA; impedance ~300 KΩ). In each experiment, a single microelectrode was lowered into the center of the imaging chamber by penetrating the artificial dura. In order to avoid bleeding, special care was made to target the electrode to a cortical area without large blood vessels. This was done before each experiment and the position of the penetration site, relative to the functional domains of orientation and ocular dominance was not known. The electrophysiological signals were recorded using an in-house software based on a National Instruments data acquisition card (National Instruments, USA). The analog signal from the microelectrode was amplified and sampled at 50 kHz. For the LFP, the analog signal was bandpass filtered (forth order Butterworth filter, with cutoff 0.5–400 Hz) and decimated. MUA spike times were detected by applying a 3 s.d. (standard deviation) threshold on the bandpass analog signal (forth order Butterworth filter, with cutoff 0.5–30 kHz). MUA stability was assessed by comparing the average MUA firing rate on the first and the last trial (trial duration of 10 s) of each experiment day. Electrophysiological signals were recorded in all the experiments.

MUA Events

MUA-events were defined as events in the MUA instantaneous firing rate, which exceeded a threshold of 3 s.d. above the mean MUA firing rate (note that 2 thresholds were used. One threshold to detect spike times in the analog signal, and one threshold to detect MUA-events in the MUA firing rate). MUA events with inter-event intervals of less than 200 ms were discarded to avoid any overlap between events (a conservative 200 ms was used since the average MUA event duration was <200 ms).

To estimate the MUA event time courses presented in Figure 7f, l, we utilized the fact that the optical signals during MUA events exhibited a significant increase in power at low frequency bands (1–5 Hz; see Fig. 7e,k). The instantaneous power of the filtered optical signals (1–5 Hz) was estimated using the Hilbert transform; this calculation was repeated independently for all the pixels, and was summed across pixels to produce the MUA event time course.

Experimental Design

To test the spatial relationship between spontaneously occurring activity patterns and the functional architecture, we thoroughly mapped the underlying functional architecture of the orientation and ocular dominance columns, before each experimental session, using visual stimulation (see Omer et al. (2013) for a detailed description of the methodology for high-quality imaging of functional maps, which is described only briefly below).

Visual stimulation and mapping the functional architecture of orientation columns: We used 4 drifting-grating stimuli (0°, 90°, 180°, and 270° orientation; contrast, 90%; size, 13 × 13°; spatial frequency, 2 cycles/deg; temporal frequency, 2°/s; mean screen luminance 23 cd/m²). Each stimulus lasted 50 ms, and the 4 stimuli were presented pseudo-randomly with an inter-stimulus interval of ~150–250 ms, for a total period of ~10 s. To map the functional architecture of the ocular dominance columns, we used full-field isoconcentric rings of flickering checkers as stimuli; the stimulus was presented to one eye at a time, alternating between eyes every 200 ms by the use of a pair of ferroelectric shutters (DisplayTech, USA). The precise time course of the stimulus was monitored by a photodiode and was saved as an analog channel that was synchronized with the optical imaging. The screen background was kept isoluminant for the entire trial period, including the period of fixation prior to stimulus onset. Stimuli were presented on a 21-inch CRT monitor (Sony GDM-F500) at 120 Hz, with the monitor placed 100 cm in front of the monkey’s eyes.

To image functional maps in the awake monkey experiments, the monkeys were trained to perform a long fixation of 11–15 s, either with each eye separately or with both eyes. The trial started when the monkey fixated within a 2° × 2° window on a small spot of light (fixation point, 0.1°) throughout the entire 10-s recording trial, until the spot disappeared. At 200 ms following the onset of the fixation point, a stimulus appeared on the screen. Eye positions were monitored by an infrared eye tracker and recorded at 1 kHz (Dr. Bouis Devices, Karlsruhe, Germany). The eye-tracking device was mounted at a 90° angle to the animal, using a half-silvered mirror, and therefore, it did not interfere with the eye shutters. Whenever the monkey broke fixation, the trial was immediately aborted. Otherwise, the monkey was rewarded with 0.2–0.3 mL of water or juice.

Imaging the ongoing activity: For the ongoing activity sessions in awake monkeys, the monkeys were trained to sit quietly for periods of 10 s while their eyes were covered with eye shutters to ensure a uniform illumination pattern. The eye shutters did not interfere with the eye-tracking device. At the end of each ongoing session, the monkey was rewarded with 0.2–0.4 mL of water or juice. For the ongoing activity sessions in anesthetized monkeys, the animals were fully paralyzed using Pancuronium bromide, and their eyes were covered with eye shutters.

Both for the visually evoked and the ongoing recorded optical signals, the images were binned offline by a factor of 2 (100 × 100 pixels → 50 × 50 pixels; each binned pixel imaged cortical activity emanating from a cortical area of 120 × 120 μm²). The bleaching artifact was removed by subtracting a 2-exponential fit from each pixel (Arieli et al. 1995, 1996; Tsodyks et al. 1999; Kenet et al. 2003). Breathing artifacts were removed by using a high-pass filter with a cutoff frequency of 0.2 Hz. Heartbeat artifacts were removed by subtracting the ECG-triggered averaged optical signal from the raw data at each heartbeat (Grinvald et al. 1984; Arieli et al. 1996; Tsodyks et al. 1999). If time series analysis was carried out, data were transformed into fractional units (∆F/F) by dividing each optical movie by the average frame, and subtracting one. Before carrying out pattern analysis, the data were denoised using a PCA-based approach and a spatial 2D filter described in detail in O’Hashi et al. (2017).

Optical frames of ongoing activity were compared with visually evoked maps using the Pearson correlation coefficient as a measure of similarity. For the correlations we used a region of interest (ROI) which included only area V1—based on the clear V1–V2 border in the evoked ocular-dominance maps. Only experiments which yielded reliable evoked functional maps of both orientation columns and ocular-dominance columns were included in the analysis.

We performed awake monkey experiments in 3 animals: 17 experiments in the first animals, 2 experiments in the second animal, and 8 experiments in the third monkey.

We performed anesthetized monkey experiments in 3 animals: 17 experiments in the first animal, 12 experiments in the second animal, and 4 experiments in the third animal.

IODC (“iso-orientation domain correlations”) (Figs 3c, 4c): To detect spontaneous correlated activity related to orientation domains, we correlated 2 variables: 1) the correlation coefficient between 10 s-long (single trial) ongoing activities recorded optically from 2 different cortical loci, and 2) the difference in the preferred orientation angle (Δθ) between the 2 cortical loci (based on the visually evoked maps). Cortical loci pairs less than 1000 μm apart were excluded from the analysis to avoid correlations due to light scattering. IODC were contrasted with both spatial shuffling and temporal shuffling. Spatial shuffling was performed by shuffling the orientation preference of each cortical loci (pixel). Temporal shuffling was performed by calculating the correlation coefficient between time courses of ongoing activities which were recorded during different trials.

The IODC time courses used in Figure 7f,l were calculated by repeating the above calculation on sliding overlapping windows of 200 ms (50 ms overlap).

MUA event absolute latency maps (Fig. 7b, h): One map was calculated for each MUA event. For each pixel we calculated the response latency (time-to-peak relative to MUA) of the optical signal (filtered: 1–5 Hz response latency map). For normalization purposes, the shortest response latency in each map was subtracted from the response latency of the reset of the pixels in the map. These maps were subsequently used to calculate the Mean MUA event relative latency maps (below).

Mean MUA event relative latency maps (Fig. 7c, i): These maps were used to estimate the average spatial spread pattern across the imaged area and the spatial propagation speeds of the activity waves. The maps were calculated from the MUA event absolute latency maps (above). First, each MUA event absolute latency map was zero-padded and rereferenced relative to the onset time of one of the image pixels. This calculation was repeated for each pixel in the map independently, resulting in 2500 maps (one for each pixel) × the total number of MUA event absolute latency maps. Next, each rereferenced map was spatially aligned by centering it at the reference pixel position (maps were zero-padded to avoid edge effects). Finally, we averaged all the rereferenced, re-centered maps together to obtain the Mean MUA event relative latency map (Fig. 7c, i). Trajectory slopes across this map represent the propagation speeds along that trajectory (the derivative of the response latencies). The mean and maximum propagation speeds of the traveling waves in Figure 7d, j were estimated by calculating the average and the minimum slope of the linear trajectories starting from the center of the map outward.

Estimating SNR for Evoked Functional Maps

The SNR for evoked functional maps was defined as the ratio between the average peak response (dF/F) at preferred domains and the standard deviation of the responses at the orthogonal domains.

Results

Spontaneous Cortical Activity Patterns Resembling Orientation and Ocular Dominance Maps in the Anesthetized Monkey

Spontaneously emerging cortical activity patterns resembling visually evoked orientation and ocular dominance maps were observed directly in the anesthetized monkey (Figs 1b and 2b; Fig. S2; Movies S1–S2). To quantify the emergence of such patterns, we calculated the frame-by-frame Pearson correlation between spontaneous ongoing activity, and visually evoked orientation and ocular dominance maps (Figs 1a and 2a). We used several tests to determine whether cortical activity patterns that resemble the functional maps occurred more frequently than expected by chance. First, we computed the distributions of the frame-by-frame correlations of the spontaneous activity to “template maps,” that is, the visually evoked differential maps for orientation and ocular dominance (differential maps were obtained by subtracting the average single condition map from the orthogonal map. The 0°/90°, 45°/135° differential orientation maps and the left-eye/right-eye differential ocular dominance map were used here); we then compared these correlations with the distributions of the frame-by-frame correlations to horizontally flipped template maps—the control maps (Kenet et al. 2003; O’Hashi et al. 2017). Results are shown in Figure 3a,b (see also Fig. S3). The distributions of correlation values to real template maps were significantly wider and shifted towards higher positive and negative values, as compared with the null distributions of correlation values obtained using control template maps (Fig. 3a,b: compare red histograms [data] to blue histograms [control]; F-test; orientation maps: F = 2.643, df = 49, P = 9 × 10⁻⁴; ocular-dominance maps: F = 3.828, df = 49, P = 6.3 × 10^–6; Kolmogorov–Smirnov test: orientation maps P < 10⁻⁶; ocular-dominance maps P < 10⁻¹⁰; See Figure S3 for examples from individual animals). Spontaneous cortical patterns resembling evoked maps appeared at average frequency of 1.75 Hz ± 0.79 Hz (mean ± s.d.), and 2.41 Hz ± 0.98 (mean ± s.d.) for orientation and ocular dominance respectively. Furthermore, the correlation to orientation and ocular dominance template maps exceeded a threshold (P < 0.01) derived from the control distributions far above what is expected by chance (only 1% of the time to exceed this threshold). This demonstrates that spontaneously emerging activity patterns indeed often resemble the visually evoked functional orientation and the ocular dominance maps.

Figure 1.

Example of a spontaneously emerging cortical state resembling the ocular dominance map in the anesthetized monkey. (a) Time course of correlation values between the visually evoked ocular dominance map (the first frame is denoted by a green rectangle in panel (b)) and consecutive images of ongoing activity recorded over 10 s. (b) Short movie sequence imaged over area V1 in an anesthetized monkey depicting a spontaneously emerging cortical state resembling an ocular dominance map. The first frame (the green rectangle) depicts the average visually evoked map (template map) for comparison. The correlation value of each frame to the visually evoked map is color-coded by dots in the upper-right corner of each frame; see the color-bar on the right. Frames are sequentially ordered from top-left (second frame) to bottom-right. The imaged area covers 6 × 6 mm² of the cortex; frames are shown every 80 ms.

Figure 2.

Example of a spontaneously emerging cortical state resembling an orientation map in the anesthetized monkey. (a) As in Figure 1: the time course of correlation values between a visually evoked orientation map for 45° stimulus (the first frame is denoted by a red rectangle in panel (b) and the consecutive images of ongoing activity recorded over a 10 s epoch. (b) A short movie sequence imaged over area V1 in an anesthetized monkey, depicting a spontaneously emerging cortical state resembling an orientation map. The first frame (the red rectangle) depicts the average visually evoked map for comparison. The correlation value of each frame to the visually evoked map is color-coded by dots in the upper-right corner of each frame; See the color bar on the right. Frames are sequentially ordered from top-left (second frame) to bottom-right. The imaged area covers 6 × 6 mm of cortex; frames are shown every 40 ms.

Figure 3.

Spontaneously emerging orientation and ocular dominance activity patterns in the anesthetized primate. (a, b) Distribution of the frame-by-frame correlations to visually evoked maps (a—orientation, b—ocular dominance). Red: the distribution obtained using visually evoked functional maps. Blue: the shuffle distribution obtained using control (flipped) maps. Blue line: the fitted Gaussian function. Data were pooled across all 3 animals. (c) Red: in the absence of visual stimulation; pairwise correlations were computed between different cortical loci—plotted here as a function of the difference in their preferred orientation angle (⁠${∆\theta }^{°})$⁠. We termed the correlation between these 2 parameters “Iso-orientation domains correlations” IODC. Black: pairwise correlations were obtained using the randomly shuffled polar map for estimating each pixel’s preferred orientation. Inset gray: pairwise correlations in anesthetized monkeys obtained during visual stimulation (response to full-field-oriented stimuli). Correlations were restricted to pairs of cortical loci that were at least 1000-μm apart. Results are plotted as mean ± s.e.m. Data were pooled across all 3 animals. (d) Distribution of the relative occurrence of activity patterns corresponding to visually evoked maps of ocular dominance and orientation. Activity patterns corresponding to ocular dominance maps were found to be significantly more prevalent than were activity patterns corresponding to orientation maps. Data were pooled across all 3 animals. (e) Average correlation values for spontaneous activity patterns and their corresponding visually evoked maps. The analysis includes spontaneous activity patterns whose correlation to visually evoked maps exceeded the 99th percentile threshold of the shuffle distribution (obtained using control maps—see (a) and (b)). Data were pooled across all 3 animals.

Next, we investigated how the correlation between the ongoing activities in 2 different cortical loci depends on the difference between their preferred orientation angles (Δθ; preferred orientation angle computed based on the evoked orientation maps). Each pixel in the imaged area represented a single cortical loci covering a cortical area of 120 × 120 μm². We repeated this calculation over all possible pairs of cortical loci (only V1 pixels were used for the analysis; number of pairs < ~3 000 000). Cortical loci pairs less than 1000-μm apart (within the total imaged area of 6 × 6 mm²) were excluded from this analysis to avoid correlations due to light scattering. We termed the correlation between these 2 measured parameters—the difference in preferred angle (Δθ), and the signal correlation between 2 cortical loci—the iso-orientation domain correlations (IODC) (Materials and Methods). In the visual stimulation data, as expected, we found strong negative IODC—meaning that the correlation between the activities in 2 separate cortical loci was strongly negatively correlated with the difference in their preferred orientation angle (Fig. 3c, inset, in gray; r = −0.99 ± 0.003 (mean ± s.d.)). However, a strong negative IODC also existed for the ongoing spontaneous data (Fig. 3c, red; r = −0.92 ± 0.3 (mean ± s.d.). Taken together, these results provide the first direct evidence that spontaneous activity patterns resembling functional maps exist in the primary visual cortex of an anesthetized primate, similar to those reported in the auditory cortex by Fukushima et al. (2012), thus, confirming that the existence of such cortical states in the primary visual cortex is not unique to cats but instead is most likely a more general phenomenon in the anesthetized brains of mammals with columnar organization.

States Corresponding to Ocular Dominance Maps were More Prevalent Than States Corresponding to Orientation Maps

The pattern-correlation analysis, using visually evoked maps as templates, uncovered substantial differences between the activity patterns corresponding to the orientations and ocular dominance maps in the anesthetized monkey. First, there was nearly a 2-fold increase in the duration of ocular dominance states versus the orientation states (in 17.8% of the total imaging time the correlation to the orientation template exceeded the P < 0.01 threshold derived from the control distribution, as compared with 30.02% for the ocular dominance states; Fig. 3a,b).

Second, the average correlation values for each of the states, pooled over the total collected frames, also indicated a significant difference between states corresponding to the ocular dominance and orientation maps (Fig. 3d,e). These findings are consistent with previous studies that showed that in primate V1, long-range horizontal connections are biased toward ocular dominance domains (Malach et al. 1993) and that the amplitude of ocular dominance maps is significantly larger than that of orientation maps (Shtoyerman et al. 2000; Slovin et al. 2002).

In Contrast with the Anesthetized Monkey, the Patterns of Orientation and Ocular Dominance Maps were Elusive in the Awake Monkey

Next, we wanted to determine whether cortical states resembling functional maps also appear spontaneously in the awake monkey. To address this issue, we first repeated the correlation analysis that we performed on the anesthetized monkey’s data. The visually evoked maps from awake and anesthetized monkeys had similar spatial signal-to-noise-ratios (SNR; Materials and Methods); awake monkey single-condition orientation maps: 1.8 ± 0.005 (mean ± s.d.); anesthetized monkey single-condition orientation maps: 1.7 ± 0.15 (mean ± s.d.); unpaired t-test, P = 0.39; awake monkey differential ocular-dominance maps: 1.6 ± 0.06 (mean ± s.d.); anesthetized monkey differential ocular-dominance maps: 1.4 ± 0.09; unpaired t-test, P = 0.13). Nevertheless, using correlation analysis, there were no significant occurrences of map-like spontaneous activity patterns in the awake monkey data, beyond what would be expected by chance. Figure 4a,b depicts distributions from awake monkeys, pooled over all 3 animals (see Fig. S4 for individual animals). No significant difference between the real distributions and the null distributions for both types of maps was evident (F-test; orientation maps: F = 1.052, df = 49, P = 0.86; Ocular dominance maps: F = 1.009, df = 49, P = 0.98; Kolmogorov–Smirnov test: orientation maps P = 1; ocular dominance maps P = 1; See Fig. S4 for individual animals).

Figure 4.

Spontaneous activity in V1 of awake monkeys is correlated between iso-orientation domains. (a,b) As in Figure 3a,b. Red: the distribution of frame-by-frame correlations to visually evoked maps (a—orientation, b—ocular dominance). Blue: the shuffle distribution of frame-by-frame correlations obtained using control (flipped) maps. The distributions of the correlation values for both the orientation and ocular dominance maps were not significantly different from what is expected by chance (the red and blue histograms overlap). (c) Red: in the absence of visual stimulation: pairwise correlations between different cortical loci plotted as a function of the difference in their preferred orientation angle (⁠${∆\mathit{\theta }}^{°})$⁠. Black: pairwise correlations obtained using the randomly shuffled polar map for estimating each pixel’s preferred orientation. Green: pairwise correlations for ongoing activity from anesthetized monkeys (identical to Fig. 3c). Note the similarity between the curves for the awake and anesthetized monkeys. Inset, Gray: pairwise correlations in the awake monkey obtained during visual stimulation (response to full-field-oriented stimuli). Correlations were restricted to pairs of cortical loci that were at least 1000-μm apart. Results are shown as mean ± s.e.m. Data were pooled across all 3 animals.

Evidence for the Existence of Spontaneous Orientation Cortical Activity in the Awake Monkey

Nevertheless, similar to the anesthetized monkey, we found significant IODC (Fig. 4c; r = −0.96 ± 0.01 [mean ± s.d.]; for evoked data r = −0.99 ± 0.002 [mean ± s.d.]; see Fig. S5 for individual animals, and Fig. S6 for temporal shuffling). Surprisingly, IODC were as strong as we found in the anesthetized monkey. Taken together, the failure to detect cortical states by using the correlation analysis, on the one hand, and the strong IODC correlations, on the other hand, suggests that spontaneous cortical states resembling orientation maps do exist in the awake monkey, but in each instance they appear partially over the imaged area.

IODC is a sensitive measure for detecting temporal correlations, but this comes at the expense of detailed spatial information. To further test the relationship between ongoing activity and the columnar architecture, while maintaining the columnar spatial resolution, we constructed spatial correlation maps in which the gray level intensity at each pixel represents the average correlation between the time course of the ongoing activity recorded over that particular pixel, and all pixels that were tuned to 1 of the 4 orientations we tested (provided the pixels were located at least ~500-μm away from each other). Employing this procedure revealed spatial correlation maps that showed a strong resemblance to the visually evoked orientation maps (Fig. 5a,b depicts 2 typical examples: the top row in each panel shows a set of 4 visually evoked single-condition orientation maps; red pluses denote peaks of preferred iso-orientation domains). This effect was also shown for the population (Fig. 5c; population result; n = 3 animals). Map-correlations between spatial correlation maps and evoked maps were significantly higher than map-correlations between horizontally flipped evoked maps (Fig. 5c; unpaired t-test, P < 3.2 × 10⁻³¹; Mean correlation between evoked and spatial correlation maps: 0.46 ± 0.02 s.e.m. (standard error of the mean). Mean correlation between horizontally flipped evoked and spatial correlation maps: 0.02 ± 0.009 s.e.m. In this case a correlation of 0.46 is considered high. Note that the absolute correlation values also depend on the high frequency noise in the data and the pixel spatial resolution. Therefore, correlation values should be evaluated with those obtained from independent data that have similar noise level). Interestingly, and in strict contrast to the anesthetized monkey, we found no spontaneous correlated activity between ocular dominance domains. These results further support our finding that spontaneous activity patterns related to orientation domains also exist in the awake monkey over smaller cortical areas, and also suggest that over time, these small orientation patterns appear across all V1. Next, we explored the dynamics of the activity patterns we observed.

Figure 5.

Spontaneous iso-orientation correlated activity in awake monkeys spans smaller cortical areas, but over time (on average), appears across all of V1. (a) Example of data from one experiment. Top row: 4 single-condition visually evoked orientation maps (0°, 45°, 90°, and 135°). The gray levels represents z-scored dF/F. Bottom row: spontaneous spatial correlation-maps. The gray levels represents the mean pairwise-correlation between the time course recorded over that pixel and all pixels with a preferred orientation angle equal to either 0°, 45°, 90°, or 135° (from left to right). Peaks of iso-orientation domains are denoted on each map by red plus signs. (b) Same as in (a). An additional example from an independent experiment. (c) Population: Each curve represents the correlation between a spatial correlation-map ((a) bottom row) and one of the visually evoked orientation maps ((a) top row). Black, control: correlations between each spatial correlation-map and the control (flipped) maps. Data were pooled across all 3 awake animals. Results are shown as mean ± s.e.m.

Similar Spatiotemporal Dynamics of Spontaneous Activity Patterns in the Anesthetized and Awake Monkey

To characterize and compare the spatiotemporal dynamics of spontaneous activity patterns between the anesthetized and awake monkey, we combined the simultaneously recorded MUA, which has a local intracolumnar resolution, with the VSDI, which has mesoscopic intercolumnar resolution. Already from the local MUA the similarity between the anesthetized and awake monkey was evident. In both conditions, the MUA revealed synchronized epochs of high-rate spiking activity lasting up to a few hundreds of milliseconds (“MUA events”), interleaved by long periods of low-rate activity lasting up to a few seconds (Fig. 6a,b, top panels). MUA events were also detected optically by the VSDI (Fig. 6a,b, bottom panels). In the awake data, there was often low similarity between the mean optically detected activity over the entire imaged area and the local MUA events. This was due to the spatial nonhomogeneous activation pattern over the imaged cortical area, whereas the electrical recording was local.

Figure 6.

The MUA and population activity detected optically exhibits similar timing. (a) A 10-s epoch of simultaneous recordings of optical signals and MUA in an anesthetized monkey. Top: the firing rate of the recorded MUA over a 10 s epoch. Bottom: the simultaneously recorded optical time course averaged over all image pixels. Note that MUA events were also clearly observed in the optical signal (red asterisks denote the start of a MUA event). (b) Similar analyses for awake monkeys. The often low similarity between local MUA events and the mean global activity detected by the optical signals (red asterisks) and averaged over a large area are due to the nonhomogeneous activation patterns over the imaged cortical surface.

To further analyze the activity over a spatial scale of several columns, we used the MUA events to trigger temporally the optical signals (Materials and Methods). The average optical signals triggered by MUA events revealed large amplitude depolarization events (Fig. 7a, g, bottom panel; displayed over a ± 400 ms window around the MUA events, for both anesthetized and awake monkeys, taken from one experiment; top panel, trigger-averaged optical signals from each pixel of the entire imaged area, aligned with the MUA events’ onsets; Materials and Methods)—demonstrating the co-occurrence of local MUA events with large-scale depolarization optical events under both conditions.

Figure 7.

The spatiotemporal dynamics of spontaneous correlated activity in anesthetized and awake monkeys is similar. (a) Top: optical line scan plots (x-axis: time, y-axis: pixels), averaged after triggering on MUA events onsets in an anesthetized monkey. The color code represents the fractional change in florescence (dF/F). All image pixels are shown. Bottom: time courses of MUA firing rates triggered by the MUA events onsets and stacked along the y-axis. Firing rates were normalized to maximum firing rate. Example taken from one session. (b) Example of a time-to-peak latency map around a single MUA event, showing isolatency lines across the cortex. The latency value taken from the pixel with the shortest latency time was subtracted from all pixels across the imaged area. (c) The mean relative latency map (n = 780; MUA events × recording days × 2500 pixels × 3 animals). (d) Radiating linear trajectories on the latency map in c, starting from the map-center outwards. The trajectories’ slope represents the propagation velocity of activity over the cortex. The mean propagating speed was 0.12 m/s (the yellow dashed line). The maximum propagating speed was 0.33 m/s (green dashed line). (e) Average time-frequency plot of the optical signals triggered by MUA events onsets for an anesthetized monkey showing a distinct increase in power at low frequencies (1–5 Hz). The color code represents the z-score units (n = 780; MUA events over all recording days × 3 animals). (f) Crosscorrelogram between 2 variables: 1) the frame-by-frame correlation to visually evoked maps (absolute values) and 2) the time course of MUA events (instantaneous power of the filtered optical signal across the image at 1–5 Hz). Blue and green: crosscorrelograms for differential orientation maps 0°/90° and 45°/135°, respectively. Red: crosscorrelogram for differential ocular dominance maps. Data were pooled over 3 animals. Results are shown as mean ± s.e.m. (g–k) Similar analyses for awake monkeys (except for panel (l)). (l) Crosscorrelogram between 2 variables: 1) the time course of MUA events (instantaneous power of the filtered optical signal across the image at 1–5 Hz) and 2) the time course of the instantaneous IODC (sliding window 200 ms; Materials and Methods). Note that IODC is inversely correlated with the difference in the preferred orientation angle, Δθ (Fig. 4c); Data were pooled over 3 animals. Results are shown as mean ± s.e.m.

Propagating Traveling Waves in the Anesthetized and Awake Monkey

The large-scale depolarization optical events which co-occurred with MUA-events displayed nearly similar dynamics in the awake and anesthetized monkey. First, under both conditions, the optical events propagated as traveling wave over the cortex as revealed by MUA event absolute latency maps (Fig. 7b,h showing an example MUA event absolute latency maps in anesthetized and awake monkeys, respectively). These maps display the time-to-peak latency of the optical signals in each pixel around each MUA-event onset (Materials and Methods). We than used all the MUA event absolute latency maps to calculate a Mean MUA event relative latency maps (Materials and Methods). These average maps revealed a centered round convexity, indicating that the optical events propagated across the imaged area as traveling waves under both conditions (Fig. 7c—anesthetized money, 7i—awake monkey), but with different latency relative to MUA events (note the response latency difference between the optical signals and the MUA events between the 2 conditions—Fig. 7a, g).

In each map the mean slope of the lines going from the center location outwards represents the mean propagating speeds of the waves across the cortex (Fig. 7d—anesthetized money, 7j—awake monkey; maximal propagation speeds 0.33 m/s for both anesthetized and awake monkeys). The spectrograms of optical signals around the occurrences of MUA-events showed similar characteristic power at 1–5 Hz (Fig. 7e, k for awake and anesthetized monkeys, respectively). These findings of large-scale coordinated depolarization events in the anesthetized and awake monkey resemble the previously reported up-down states in anesthetized rodents using intracellular recordings (Steriade, McCormick et al. 1993; Steriade, Nunez et al. 1993).

Next, we investigated whether a relationship exists between the MUA events (and the larger-scale optical events associated with them) and the spontaneous emergence of cortical map-like patterns under anesthesia and during wakefulness. First, in the anesthetized monkey, where spontaneous activity patterns were readily observed, we constructed 2 time series: 1) the frame-by-frame correlation of the spontaneous activity to the visually evoked maps, one for each map (differential orientation maps 90°/0°, 45°/135°, and differential ocular dominance maps), and 2) the time course of MUA events (Materials and Methods). We then cross-correlated the 2 time series (Fig. 7f; see also Fig. S7 for data from independent experiments). We found significant correlations, with a zero time-lag, between the occurrences of spontaneous activity patterns corresponding to the orientation and ocular dominance maps, and MUA events—showing that the emergence of spontaneous mesoscopic activity patterns is time locked to the MUA events in the anesthetized monkey. Second, using the awake monkey data, we cross-correlated 2 time series: 1) the instantaneous IODC (sliding window 200 ms; Materials and Methods), and 2) the time course of MUA events (Fig. 7l; Materials and Methods; see also Fig. S8 for data from independent experiments). Importantly, we found a significant negative correlation, with a zero time-lag, between the IODC time course and MUA events, that is, the IODC was more negative around MUA events onsets (Note that IODC is inversely correlated with the difference in the preferred orientation angle, Δθ, as shown in Fig. 4c). This finding provides evidence that the emergence of spontaneous mesoscopic activity patterns in awake monkeys is also time-locked to MUA events and large-scale depolarization events. Furthermore, it shows that both in anesthetized and awake monkeys, spontaneous map-like cortical activity shares similar spatiotemporal dynamics—both emerge over propagating traveling waves in V1 (Grinvald et al. 1994; Jancke et al. 2004; Ferezou et al. 2007; Benucci et al. 2007 (but see Sit et al 2009, Suppl. Fig. 4), Nauhaus et al. 2009; Mohajerani et al. 2010; Nauhaus et al. 2012; Sato et al. 2012; Muller et al. 2014; Xu et al. 2007; Huang et al. 2010; Muller et al. 2018).

The Spontaneous Emergence of Orientation Activity Patterns in Awake Monkeys Cannot be Explained by Eye Movements or by Sleep Periods

To test for a possible confounding effect of eye movements, we conducted 6 independent experiments on 2 awake monkeys, where we recorded the eye movements during ongoing activity sessions while the monkey’s eyes were covered with eye shutters (Materials and Methods). Specifically, we wanted to determine whether the monkey’s eye movements were correlated with the MUA events. To this end, we averaged the eye movement velocities, triggered by the occurrence of MUA events, which indicates the occurrence of spontaneous activity patterns. This analysis did not reveal any correlation between the 2 (Fig. 8a). We then averaged the LFP and optical signals, triggered by the occurrence of the same MUA events that we used to trigger the eye movement velocities. In contrast with the eye movements, the LFP and the optical signals revealed large population events (Fig. 8b, c) that could not be explained by the eye movements (Fig. 8a). Importantly, the response amplitudes of the MUA event triggered LFP (Fig. 6b, black curve) was of a magnitude similar to the visually evoked spike-triggered average LFP in response to visual stimuli having full-field drifting grating (green curve). These results, from the LFP and optics, stand out in sharp contrast to the eye movement results above. Taken together, we can conclude that spontaneous MUA events—along with their concurrent activity patterns—are internally generated, and are not driven by eye movements.

Figure 8.

Spontaneous correlated activity in an awake monkey cannot be explained by eye movements or by short episodes of sleep. (a) Eye movement velocity averaged after triggering by MUA events in an awake monkey. No significant eye movements were correlated with MUA events. (b) Top panel: time courses of the LFP triggered by MUA events onsets and stacked along the y-axis (x-axis: time, y-axis: n = 120 MUA events). The same MUA events were used as in a. Bottom panel: the black line shows the LFP averaged after triggering by MUA events (the same MUA events as in a). Green: LFP averaged after triggering by spikes in response to a visual stimulation with full-field drifting grating stimuli—shown here for comparison. Note the high resemblance between the evoked and spontaneous responses. (c) Top panel: optical line-scan plots (x-axis: time, y-axis: pixels), averaged after triggering by MUA events onsets. The same MUA events were used as in (a) and (b). The color code represents the fractional change in fluorescence (dF/F). Bottom panel: Optical signals averaged after triggering by the same MUA events as in (a) and (b). (d) Blue trace: the power spectrum of LFPs from awake monkey experiments displayed the characteristic power distribution of wakefulness with a significant increase of power in the alpha and gamma frequency bands, indicating that the animals were awake during the recording sessions of spontaneous activity. Red trace: the power spectrum of the LFP recorded from anesthetized monkeys. (e) Blue line: the mean power spectra of ± 2 s segments of LFP centered on detected MUA events. The orange line shows the mean power spectra of ± 2 s segments of LFP centered at random times, ruling out the possibility that MUA events represent short episodes of sleep. In panels (a–e) results are shown as mean ± s.e.m.; data in (a–c) were pooled over 2 animals; data in (d) and (e) pooled over 3 awake and 3 anesthetized animals.

Another potential concern is the behavioral state of the monkeys during the awake experiments—specifically, the animals may have been potentially drowsy. However, this seems unlikely because the power spectrum of the LFP in the awake experiments displayed the characteristic power distribution of the wakefulness state—with a significant increase of power in the alpha (10–15 Hz) and gamma frequency range (25–40 Hz), and a decrease in delta (1–4 Hz) frequencies (Fig. 8d, blue curve; data shown as mean ± s.e.m.). For comparison, Fig. 8d also depicts the power spectrum of ongoing epochs recorded in anesthetized animal experiments (red curve).

Next, we also wanted to rule out the possibility that MUA events might represent short episodes of sleep. To this end, we compared the mean power spectra of ±2 s segments of LFP centered on MUA events to the mean power spectra of ±2 s of random segments (Fig. 8e; data shown as mean ± s.e.m.). The power spectra were statistically indistinguishable—which rules out the possibility that during the awake monkey experiments, MUA events occurred during short episodes of sleep. Taken together, these results confirmed that in the awake experiments the animals were in a wakefulness state and that the neural dynamics reported here indeed represent the internally generated cortical dynamics in awake monkeys.

Discussion

In this study we combined VSD imaging, LFP, and MUA recordings in the primary visual cortex (V1) of anesthetized and awake monkeys. The data reported here compared for the first time the spatiotemporal dynamics of ongoing activity in awake versus anesthetized primates, at a high subcolumnar spatial resolution and the fast millisecond time domain. We showed here that ongoing activity in the anesthetized monkey comprises spontaneously emerging activity patterns, some of which resemble the functional maps of orientation and ocular dominance domains, which span large cortical areas (at least up to the 6 × 6 mm² area imaged here). In the awake monkey, such spatially extended activity patterns resembling the monkey’s functional architecture were not detected. Instead, temporal correlation analysis (IDOC) and spatial correlation map analysis revealed that spontaneous cortical activity involving the underlying columnar organization also exists in the awake monkey. However, it is most likely that such activity covered smaller cortical areas. Spontaneous correlated cortical activity under both conditions displayed similar spatiotemporal dynamics (MUA events, and optical large-scale hyperpolarization events) and were triggered by propagating traveling waves (Grinvald et al. 1994; Jancke et al. 2004; Ferezou et al. 2007; Benucci et al. 2007 (but see Sit et al. 2009, Suppl. Fig. S4); Nauhaus et al. 2009; Mohajerani et al. 2010; Nauhaus et al. 2012; Sato et al. 2012; Muller et al. 2014; Xu et al. 2007; Huang et al. 2010; Muller et al. 2018). Taken together, these results provide strong evidence that coherent local columnar correlated activity exists also in awake animals.

Spontaneous Activity Patterns Resembling Orientation and Ocular Dominance Maps in the Anesthetized Monkey

The first finding of this study is the observation of spontaneously emerging activity patterns resembling functional maps in the anesthetized monkey (Figs1–3). This is the first such finding in primates, and it extends previous reports in anesthetized cats (Arieli et al. 1995, 1996; ; Tsodyks et al. 1999; Kenet et al. 2003; Chan et al. 2015; Reyes-Puerta et al. 2016; O’Hashi et al. 2017) also to primates. This previous finding in cats can thus no longer be considered as a species-specific observation. Spontaneously occurring activity patterns in the anesthetized monkey were found to comprise orientation and ocular dominance maps covering large cortical areas (up to the 6 × 6 mm²—the imaged area limit in this study). These results are in agreement with previous anatomical findings that reported evidence for large bias in local cortical circuit connectivity between cortical columns with similar tuning properties, referred to as like-to-like connectivity (Rockland and Lund 1983; Ts’o et al. 1986; Gilbert et al. 1989; Malach et al. 1993; Yoshioka et al. 1996; Bosking et al. 1997; Lund et al. 2003). The bias in the occurrence of ocular dominance states compared with orientation states is also in agreement with the well-known strength of evoked maps in the primate—where ocular dominance maps are known to be stronger than orientation maps (Shtoyerman et al. 2000; Slovin et al. 2002) (the opposite relationship occurs in cats). Taken together, these observations suggest that in the anesthetized monkey the ongoing activity is partly shaped by activity patterns that reflect the underlying columnar organization. Furthermore, this finding in primates suggests that this is a general phenomenon across mammals with columnar cortical organization (Mountcastle 1957; Hubel and Wiesel 1963; Humphrey and Norton 1980; Yacoub et al. 2008; Wang et al. 2015).

Previous studies in anesthetized animals suggested that spontaneous correlated activity in V1 can be fully explained by state-dependent global uniform fluctuations (standing waves) regardless of orientation preference (Greenberg et al. 2008; Ecker et al. 2010; Ecker Alexander et al. 2014; Tan et al. 2014; Schölvinck et al. 2015). Interestingly, our analysis revealed nearly the opposite: First, the optically detected large-scale coordinated depolarization waves revealed by MUA event triggering were not globally uniform at all. In fact they propagated over the cortical surface as traveling waves at speeds that are in agreement with previously reported propagation speeds over non myelinated long-range cortical connections in V1 (Hirsch and Gilbert 1991; Grinvald et al. 1994). This may suggest that the waves propagate over corticocortical connections. Nevertheless, we cannot exclude propagation over thalamocortical loops. Second, we found spontaneous correlated activity that was strongly correlated with the underlying columnar organization (orientation and ocularity preference). The reason for the discrepancy is most likely due to the different methodologies. In particular, optical imaging data emphasize both subthreshold and suprathreshold activity, and has higher spatial resolution at the columnar level.

Evidence for the Existence of Spontaneous Correlated Orientation Activity in the Awake Monkey

The second key finding of this study was very different from the first one. Using the full-field visually evoked maps as templates for pattern correlation analysis, we found no direct evidence for the spontaneous emergence of widespread activity patterns that resemble orientation maps or ocular dominance maps during wakefulness (Fig. 4a,b). At first glance, these observations seem to be in full agreement with previous studies reporting that spontaneous cortical activity in awake animals is decorrelated (Poulet and Petersen 2008; Ecker et al. 2010) and are in line with other studies pointing to the possible detrimental effects of internal correlated activity on sensory neural population coding (Averbeck et al. 2006). However, the template correlation analysis we employed here is inherently biased towards detecting patterns spanning large cortical areas. Employing a more sensitive analysis revealed a different picture. The IODC in the awake monkey was as strong as in the anesthetized monkey (Figs 4c, 3c), and the spatial correlation maps showed high similarity to the evoked orientation maps (Fig. 5). Both observations strongly suggest that spontaneous correlated internally generated activity exists in awake monkeys, but that it is faster and spans smaller cortical areas in comparison with the activity in the anesthetized state. If correlated cortical activity limits information capacity in cortical circuits, then the correlations we found in the awake state may seem nonbeneficial. A recent study by Moreno-Bote et al. (2014) suggests that the amount of information that can be stored in population activity is not limited by correlations induced by shared connections, but rather by particular type of correlations which are proportional to the product of the derivatives of the tuning curves. One may speculate that the correlations we observe in the anesthetized state are dominated by the information limiting correlations as opposed to the awake state.

In contrast to the evidence for spontaneous orientation correlated activity, we found no evidence for spontaneous ocular dominance correlated activity for reasons that remain to be explored. It seems that the lack of ocular dominance patterns during awake state is related to the different inputs which are associated with each columnar system. Orientation maps reflecting more recurrent cortical activity and ocular dominance more feedforward inputs. We can only speculate here that, in contrast to the anesthetized state, spontaneous activity in awake state reflects more recurrent cortical activity which masks the feed forward activity. Other alternative explanations remain to be explored.

The third key finding in this study concerns the spatiotemporal dynamics of spontaneous activity patterns in both the anesthetized and the awake monkeys. Specifically, under both conditions spontaneous correlated activity emerged, overriding spatial traveling waves. The similar traveling speeds and the involvement of MUA burst activity under both conditions may suggest that the waves propagate over cortical long-range horizontal connections. Interestingly, our findings here confirm and extend a previous finding by Tan et al. (2014), using whole-cell recordings in alert fixating monkeys, which reported that ongoing activity in V1 neurons results from infrequent correlated internally generated events that elicit large fluctuations in the neurons’ membrane potential—which strongly resemble the MUA events we observed.

In summary, to the best of our knowledge, our experiments show, for the first time, images of spontaneous cortical dynamics at high spatiotemporal resolution integrated with MUA and LFP recordings, in V1 of anesthetized and awake monkeys. Importantly, we found that spontaneous correlated columnar activity exists in both anesthetized and awake monkeys. In the anesthetized monkey the patterns of ongoing activity were dominated by the classical functional columnar connectivity, whereas in the awake monkey the correlated activity was more local and faster and may also reflect other connections. The finding of dynamic patterns of spontaneous activity in the awake monkey supports the notion that ongoing activity plays an important role in cortical functions and behavior.

Finally, our data indicate that V1 lateral connectivity in the awake monkey is richer and far more heterogeneous than that found in the anesthetized monkey. This difference is in agreement with recent anatomical and physiological studies reporting that the lateral connectivity in the superficial layers of V1 is more heterogeneous than what was known before: pyramidal cells at iso-orientation domains receive input originating from a heterogeneous set of orientation domains (Chavane et al. 2011; Huang et al. 2014; Martin et al. 2014). The functional role of this heterogeneous lateral cortical connectivity in the awake behaving animal remains to be uncovered and elucidated.

Funding

Research grants to A.G. from ISF, BMBF, and the EU (FP6-2004-IST-FETPI-015803, DAISY) and by the Grodesky Center.

Notes

We thank N. Ulanovsky, D. Derdikman, L. Las, M. Geva-Sagiv, A. Finkelstein, and E. Tamir for comments; and L. Rom, S. Kaufman, and E. Tsabari for assistance in surgeries and primate training. Conflict of interest: The authors declare no competing financial interests.