Abstract
Our eyes are never still. Even when we attempt to fixate, the visual gaze is never motionless, as we continuously perform miniature oculomotor movements termed as fixational eye movements. The fastest eye movements during the fixation epochs are termed microsaccades (MSs) that are leading to continual motion of the visual input, affecting mainly neurons in the fovea. Yet our vision appears to be stable. To explain this gap, previous studies suggested the existence of an extraretinal input (ERI) into the visual cortex that can account for the motion and produce visual stability. Here, we investigated the existence of an ERI to V1 fovea in macaque monkeys (male) while they performed spontaneous MSs, during fixation. We used voltage-sensitive dye imaging (VSDI) to measure and characterize at high spatiotemporal resolution the influence of MSs on neural population activity, in the foveal region of the primary visual cortex (V1). Microsaccades, performed over a blank screen, induced a two-phase response modulation: an early suppression followed by an enhancement. A correlation analysis revealed a widespread foveal increase in neural synchronization, peaking around ∼100 ms after MS onset. Next, we investigated the MS effects in the presence of a small visual stimulus and found that this modulation was different from the blank condition yet both modulations coexisted in the fovea. Finally, the VSD response to an external motion of the fixation point could not explain the MS modulation. These results support an ERI that may be involved in visual stabilization already at the level of V1.
Significance Statement
Microsaccades are tiny fixational saccades, leading to the continual motion of the visual input on the fovea, during visual fixation. Yet our vision appears to be stable. To explain this gap, we investigated the existence of an extraretinal input into the fovea of the primary visual cortex (V1) in behaving monkeys while they performed microsaccades over a blank screen with a tiny fixation point. The population response aligned on microsaccades showed a widespread, transient increased neural synchronization along with a two-phase response modulation. Microsaccades in the presence of a visual stimulus induced distinct spatiotemporal response from that in the blank condition. Our results support the existence of an extraretinal input that may be involved in visual stabilization at V1 area.
Introduction
Our eyes are never still. We scan the surroundings with large and rapid ballistic eye movements termed as saccades (size >1°) which enable us to gather visual information from points of interest in the visual scene. Even when we attempt to fixate, the visual gaze is never entirely motionless, because we continuously perform miniature oculomotor movements termed as fixational eye movements. These tiny eye movements are classified into two main categories: microsaccades (MSs) and drift. Microsaccades are fast and small (size <1°) fixational saccades and are considered to be the smaller version of saccades, while drift is the slow eye motion during the intersaccadic interval (Martinez-Conde et al., 2004, 2009; Rolfs, 2009; Rucci and Victor, 2015; Ahissar et al., 2016; Krauzlis et al., 2017).
In primates, MSs and saccades enhance visual processing by allowing to flexibly allocate the fovea, a specialized high-acuity zone at the center of the primate's retina, toward points of interest in the visual scene (Green, 1970; Wurtz, 2008; Poletti et al., 2013; Intoy et al., 2021). Another line of studies have suggested that MSs enhance processing by heightening information acquisition during visual exploration and by reversing visual fading (Martinez-conde et al., 2006; Otero-millan et al., 2008; McCamy et al., 2014). However, these fast and ballistic eye movements pose also a substantial challenge for the visual system, because each MS or saccade produces an abrupt and rapid motion of the visual input over the retina. Yet, despite the continual displacement of the visual input on the retina, we do not perceive these shifts and our visual perception remains stable. Saccadic suppression, which is a decrease in sensitivity to a visual stimulus, was suggested as a mechanism that can mediate stable vision (Volkmann, 1986; Diamond et al., 2000; Burr, 2004; Binda and Morrone, 2018). Recently, an increase in visual thresholds was described also for MSs (MSs suppression; Scholes et al., 2018; Intoy et al., 2021). Moreover, the decrease in sensitivity starts before the MS or saccade onset, suggesting the existence of an extraretinal input (ERI; e.g., an efference copy of the motor command to the oculomotor muscles).
Past neurophysiological studies investigated the effects of saccades and MSs on neuronal activity in the visual cortex of humans and monkeys (for reviews, see Ibbotson and Krekelberg, 2011; Krauzlis et al., 2017; Martinez-Conde et al., 2013; Rolfs, 2009; Ross et al., 2001; Wurtz, 2008). Studies investigating the effects of saccades on neural responses in animals, reported on neural modulation in various visual cortical areas, comprised neural suppression, neural enhancement, or both (Royal et al., 2006; Bremmer et al., 2009; Idrees et al., 2020; Parker et al., 2023). Several studies investigated the existence of a saccade-related ERI into the visual cortex (Royal et al., 2006; Rajkai et al., 2008; Miura and Scanziani, 2022; Niemeyer et al., 2022; Denagamage et al., 2023).
The effects of MSs on neuronal responses in monkeys and humans were also studied in various visual areas and reported to be early suppression or late enhancement or a sequence of early suppression followed by an enhancement (Kagan et al., 2008; Meirovithz et al., 2012; Gilad et al., 2014; Troncoso et al., 2015; Wu et al., 2022). The effects of MSs in early visual cortex, during visual stimulus presentation, showed that at least part of the modulation can be explained by the stimulus displacement on the retina (Meirovithz et al., 2012). Only few studies focused on the involvement of an ERI in MSs and investigated the effects of MSs without a visual stimulus in V1 or V2 (Martinez-Conde et al., 2000, 2002; Snodderly et al., 2001; Troncoso et al., 2015; Wu et al., 2022).
The possible existence of an ERI into V1 suggests also a modulation in neural synchronization. Yet, only very few studies investigated the effects of neural synchronization around saccades onset (Maldonado et al., 2008; Niemeyer et al., 2022; Denagamage et al., 2023) and MSs onset (Bosman et al., 2009; Meirovithz et al., 2012; Lowet et al., 2018; discussed in Leopold and Logothetis, 1998; Martinez-Conde et al., 2004). In particular, the relation between the changes in synchronization and MSs over homogenous blank screen in early visual cortex was not studied.
Here we used voltage-sensitive dye imaging (VSDI) in monkeys to investigate the effects of MSs on the neural population responses in the foveal region of V1, when the animals fixate over a blank screen (a homogenous gray screen with a tiny fixation point). Past neurophysiological studies did not focus on V1 fovea and foveola, where neurons are expected to be most influenced by the tiny fixational saccades. Moreover, most of these studies focused mainly on MSs effects in the presence of a visual stimulus. Here we applied a different approach: using VSDI, we characterized the influence of MSs on V1 activity at the fovea, in blank trials, where only a tiny fixation point appears. Our results show that MSs induce a biphasic modulation in V1, composed of suppression followed by enhancement. This biphasic modulation was associated with a transient increase in neural synchronization. Additionally, the MS modulation driven by a small visual stimulus was different from that in the blank trials. Finally, we compared the neural response evoked by an artificial movement of the fixation point (mimicking the fixation point movement on the retina; similarly to Troncoso et al., 2015) to that generated by MSs. Overall, our results reveal robust evidence for ERI related to MSs in V1.
Materials and Methods
Animals and dataset
Three adult male monkeys (Macaca fascicularis, males, 8–12 kg; Monkeys L, G, B) were used in this study. Nine recording sessions (three from each monkey) were used for retinotopic mapping of the imaged chamber and defining the V1/V2 border (see additional details below, in Defining the border between V1 and V2 and retinotopic mapping of V1). Data analysis on MSs effect was performed on a total of 21 recording sessions from three hemispheres of the three monkeys (Figs. 1–5). The analysis of MSs modulation in blank trials was done on a total of six recording sessions (∼29 trials per session) from the left hemisphere of Monkey G, six recording sessions (∼60 trials per session) from the right hemisphere of Monkey L, and one recording session (68 trials) from the right hemisphere of Monkey B. In each session we analyzed only correct trials (see below, Behavioral paradigm and data acquisition) that were carefully inspected for eye movements. A total of 521 MSs were analyzed from the three monkeys [Monkey L: 219 MSs, amplitude 0.45 ± 0.04° (mean ± SEM); duration 24 ± 5.87 ms (median ± MAD); Monkey G: 243 MSs, amplitude: 0.37 ± 0.02°; duration 24 ± 4.86 ms; Monkey B: 59 MSs, 0.36 ± 0.02°; duration 30 ± 6.67 ms; MSs duration are within the previously reported range; Hafed and Krauzlis, 2012; Troncoso et al., 2015].
The analysis on trials with a small visual stimulus (Fig. 4) was done on 50 MSs from five sessions (Monkey G, amplitude 0.53 ± 0.02°). Another set of control experiments of simulated MSs were done in Monkey B and included: 56 trials of artificial FP movement from three sessions (30 trials with inward movement of the fixation point; 26 trials with FP outward movement; Fig. 5). Additional details on the behavioral paradigms are below (see in Behavioral paradigm and data acquisition).
Surgical procedure and voltage-sensitive dye imaging
The surgical procedure and voltage-sensitive dye (VSD) staining have been reported with details, previously (Shoham et al., 1999; Shtoyerman et al., 2000; Arieli et al., 2002; Slovin et al., 2002). All experimental procedures were carried out according to the NIH guidelines, approved by the Animal Care and Use Guidelines Committee of Bar-Ilan University, and supervised by the Israeli authorities for animal experiments. Briefly, the monkeys were anesthetized and ventilated, and an intravenous catheter was inserted. A head holder and two cranial windows (25 mm, i.d.) were bilaterally placed over the primary visual cortices and cemented to the cranium with dental acrylic cement. After craniotomy, the dura mater was removed, exposing the visual cortex. A thin, transparent artificial dura of silicone was implanted over the visual cortex. Appropriate analgesics and antibiotics were given during surgery and postoperatively. The center of the imaged V1 area for the three animals laid 0.75–1.5° below the horizontal meridian and 1–2° from the vertical meridian. We stained the cortex with RH-1691 or RH-1838 VSD supplied by Optical Imaging.
VSDI was performed using the Micam Ultima system based on a sensitive fast camera with up to 10 kHz sampling rate. We used a sampling rate of 10 ms per frame with a spatial resolution of 10,000 pixels where each pixel measures the activity from an area of 1702 µm2. Each pixel sums the VSD signal from the population activity of a few hundreds of neurons, mainly from the upper layers (2–3) of the cortex (Petersen et al., 2003). The fluorescence dye signal from each pixel reflects the sum of the membrane potential from all the neuronal elements of that pixel, i.e., a population signal (rather than a single neuron). Thus, the VSD signal reflects both subthreshold and suprathreshold membrane potentials from all neurons. The exposed cortex was illuminated by an epi-illumination stage with appropriate excitation filter (peak transmission 630 nm, width at half-height 10 nm) and a dichroic mirror (DRLP 650), both from Omega Optical. To collect the fluorescence and reject stray excitation light, a barrier postfilter was placed above the dichroic mirror (RG 665, Schott).
Behavioral paradigm and data acquisition
Two adult male monkeys (Monkeys G, L) were trained on a fixation paradigm. Each trial started when a small fixation point (FP, 0.1°) appeared in the center of a screen (coordinates: 0,0) on a gray background (Fig. 1A; note that the FP appears 8–11° away from the screen edges; gray background luminance, 40–50 cd/m2). The animals were required to maintain fixation throughout the entire trial duration that lasted for 4–5 s, during which VSD data acquisition (DAQ) was performed. In blank trials, only a FP appeared over the gray background. In stimulated trials, after a random fixation interval of 3–4 s, a small visual stimulus was turned on for a variable duration (small black or white dot; radius, 0.05–0.15°; position, 0.75° below the horizontal meridian; 0.6° from the vertical meridian). During all trials, the animals were required to maintain tight fixation within a small fixation window. The animals were rewarded only if the trial was successfully completed, i.e., the animal maintained fixation within the fixation window, throughout the entire trial. These trials were termed as correct trials and were used for analysis.
A schematic illustration of the fixation screen and microsaccades (MSs) characteristics. A, Left, The monkeys fixated on a small fixation point (0.1–0.2°), positioned at the middle of a gray screen (coordinates: 0,0; fixation point shown in scale). The red contour represents the fixation window (radius, 1°; not visible to the animals) where the monkeys have performed MSs while maintaining fixation. Dashed vertical and horizontal lines indicate the center position (zero) on each axis. Right, An enlargement of the screen display around the fixation point, with an illustration of two example MSs, one from each monkey. B, Distribution histograms of MSs amplitude for each monkey (n = 219 Monkey L; n = 243 Monkey G). C, Distribution histograms of MSs duration for each monkey. D, The main sequence computed for MSs from each animal: max velocity versus amplitude. The continuous line is a linear regression of the data. The correlation between maximal velocity and amplitude: r = 0.91, p < 0.001, Monkey L; r = 0.96, p < 0.001, Monkey G.
Another set of control experiments were performed on a third monkey (Monkey B). The animal performed a fixation task (as described for Monkeys G and L) on a small FP (0.2°) that appeared at the center of a gray screen (coordinates: 0,0°). In part of the experiments the FP moved to one out of two possible locations: (1) 0.25° below and 0.25° left to the initial position of the FP (0,0) and (2) 0.2° above and 0.2 right to the initial position of the FP (0,0) (Fig. 5A). Following a period of 200–300 ms, the FP returned to its initial position. In another set of experiments, a small green or white point (0.1°) shifted between the above two locations, for a period of 200–300 ms. The spatial shift range of the FP and the small point was 0.3–0.6°, thus mimicking typical amplitude size of MSs. Therefore, both control experiments allowed us to simulate an artificial movement of FP, similar to the FP shift on the retina occurring when a MS is generated. Trials were divided into two groups based on the motion direction of the FP (or the point stimulus): (1) the FP (or the point stimulus) moved into the imaged cortical region (inward movement). This simulated a MS away from the imaging chamber, which leads to a spatial shift of the FP toward the chamber, (2) the FP (or the point stimulus) moved outside the imaging chamber. This simulated an inward MS that is directed toward the chamber, which leads to a spatial shift of the FP away from the chamber. In both experiments, the animal was required to maintain fixation and was rewarded only if the trial was successfully completed.
The monkeys performed the tasks in dark rooms. The visual stimuli were presented on a CRT SGI screen (85 Hz) or LCD screen (100 Hz; Display +++, CRS) that were viewed at a distance of 1 meter from the animal's eye. The luminance of the gray background was 40 or 50 cd/m2. Stimuli appeared at high contrast (100%, Weber contrast; black luminance, 0.1 cd/m2; white luminance was 80 for Monkeys G and L and 100 cd/m2 for Monkey B). Data acquisition was done using one of the following setups: (1) two linked computers controlled the behavioral task, visual stimulation, data acquisition, and the recording of the monkey's behavior (CORTEX software package). We used a combination of imaging system (Micam Ultima) and the NIMH-Cortex software package. This system was equipped with a PCI-DAS 1602/12 card to control the behavioral task and data acquisition. (2) A computer installed with MonkeyLogic (ML) software and equipped with a NI card to control the behavioral task and the data acquisition. We used a combination of ML with the imaging system (Micam Ultima). For both setups, behavioral and neuronal data from single trials was saved in separate files to enable single trial analysis. Additional details on the protocol for VSDI DAQ has been described in detail previously (Slovin et al., 2002).
Eye position recording and detection of MSs
Eye position was monitored by an infrared eye tracker (Dr. Bouis; Bach et al., 1983), sampled at 1 kHz, recorded at 250–1,000 Hz. During the behavioral task, the animals were required to maintain fixation on a small fixation point and typically made a few MSs per second (Fig. 1A). To detect the MSs on each trial, we implemented an algorithm for MSs detection on the monkeys’ eye position data (Engbert and Mergenthaler, 2006). MSs were detected in 2D velocity space of the eye position, using thresholds for peak velocity (in units of median-based standard deviations, SDs) and a minimal MS duration. This produced an elliptical threshold in the 2D velocity space (3.5–6 MAD from median velocity). Additionally, the threshold for minimal MS duration was set to 7–12 ms and a minimal 50 ms interval was set between successive MSs. MSs were also defined according to their kinematic properties (amplitude <1° and maximal velocity <100°/s; Engbert and Mergenthaler, 2006; Meirovithz et al., 2012). To minimize noise contamination in the dataset, we set a lower limit for MS amplitude (0.1°). The reliability of the algorithm for MSs detection in our data was already confirmed in previous VSD studies (Meirovithz et al., 2012; Gilad et al., 2014).
To compute the MS start point (MS onset) and end point (MS landing), we used a modified algorithm based on previous studies (Hafed et al., 2009; Chen and Hafed, 2013). The eye position traces were filtered using a low-pass filter (third-order Savitzky–Golay filter, 25 ms window) and the velocity was recomputed (Cherici et al., 2012). Then, a narrow time window (±30 ms) was defined around each detected MS, and all velocity points above a threshold (exceeding 3–8°/s; set for each animal based on the data SNR) were used for acceleration analysis. Next, a threshold of acceleration (1–2 MAD from the median acceleration) was used to determine the MS onset and MS landing. MS direction was computed from MS onset to MS landing and the MS amplitude was defined as max amplitude.
In addition, we plotted the MSs amplitude and duration histograms as well as computed the main sequence for all MSs (i.e., the relation between peak velocity and amplitude; Fig. 1B–D). MSs that were 3 SD away from the mean distribution of the main sequence were removed. Finally, we visualized all the eye position traces and verified each detected MS.
Defining the border between V1 and V2 and retinotopic mapping of V1
A total of nine recording sessions were used for retinotopic mapping and defining the V1/V2 border. In each monkey, we identified V1 and V2 regions in the imaged cortical area using optical imaging of intrinsic signals (Shtoyerman et al., 2000). Ocular dominance and orientation maps were used to identify the border between V1 and V2 areas in each animal. To obtain the retinotopic maps in V1, we used optical imaging of intrinsic signals combined with visual stimulation of horizontal or vertical bars and VSDI combined with visual stimulation of small points (Shtoyerman et al., 2000; Slovin et al., 2002). Based on the empirical retinotopic maps and using a 2D analytical retinotopic model (Ayzenshtat et al., 2012), we could approximately define the visual field eccentricities in the imaged chamber: 0.25–3° and 0.2–2.5° for Monkeys L and G, respectively. According to previous studies, the fovea region is defined from 0 to 2.5° eccentricity and parafovea from 2.5 to 5° eccentricity (Green, 1970; Moiseenko et al., 2018; Chen et al., 2019). We then defined the foveal and parafoveal regions in the monkeys imaged V1, for each animal (Fig. 2A,E maps, −60 ms). Finally, we also marked the 0.75° eccentricity on the imaged V1 and termed it as the central fovea.
The effects of microsaccades on population responses in blank trials. A, A sequence of VSD maps in V1 aligned on MS onset (t = 0; mean across MSs, n = 34; Monkey L), an example session. Each VSD map is averaged over 20 ms and color coded (blue and red denote suppressed and enhanced activity relative to the baseline). The baseline response (150–50 ms before MS onset) was subtracted for each pixel as well as the shuffled response (see Materials and Methods). Maps were low-pass filtered for visualization purposes only. The map at t = −60 ms shows the approximate eccentricity lines of 0.75 and 2.5°, denoting the regions of the fovea (F), parafovea (P), and central fovea (CF) in the imaged V1 (see Materials and Methods). B, Time course of the VSD signal in V1 (mean over all pixels) for the example session in A. The effect of MSs appears in a continuous curve and the shuffle data in a dashed curve (see Materials and Methods). The shaded area represents ±1 SEM across trials or shuffle data. C, Grand analysis of MSs modulation in Monkey L (n = 6 sessions): TC of the VSD signal in V1. Shaded area represents ±1 SEM across sessions. The top bars in B and C mark the time points with significant difference between MS modulation and shuffled data (p value range: <0.001 to <0.05). D, Left, Suppression and enhancement peak amplitude (mean across all sessions), grand analysis. Right, Time to peak for the MS suppression (supp.) and enhancement (enha.), mean across all sessions, grand analysis. Error bars represent ±1 SEM across sessions. *p> 0.05; ***p> 0.001 Wilcoxon rank sum. E–H, As in A–D, but for Monkey G. E–F, n = 60 MSs; G–H, grand analysis, n = 6 sessions.
Basic VSDI analysis and aligning the VSD signal on MS onset
Basic VSD data processing was described in details previously (Slovin et al., 2002; Ayzenshtat et al., 2010). Briefly, this consisted of choosing pixels with minimal threshold fluorescence, then normalizing each pixel to its baseline fluorescence level, and, finally, subtracting the average blank condition to remove the heartbeat artifact. This processing removes, in an unbiased manner, most of the slow fluctuations originating from heartbeat artifact or dye bleaching within a trial (Shoham et al., 1999). These steps were schematically illustrated and explained in Figure S12 in Ayzenshtat et al. (2010). In addition, we removed pixels in the vicinity of large blood vessels and pixels outside the margins of the imaged cortical area. Single trials that deviated >2.5 SDs from the mean (across trials) response were further removed from analysis to avoid noise contamination (e.g., animal motion artifact). VSD maps were low-pass filtered with a 2D Gaussian filter (sigma, 2 pixels) for visualization purposes only.
To analyze the effect of MSs on the VSD signal, we analyzed MSs occurring at 200–1,300 ms from onset of VSD DAQ in the blank condition, and for the stimulated condition, MSs were analyzed in a time window of 100 ms after stimulus onset until 50 ms after stimulus offset. Next, the VSD signal was aligned on MS onset, using a 500 ms window: 150 ms before MS and 340 ms after MS onset. Next, we subtracted from each pixel the baseline activity (mean response at 150–50 ms before MS onset) and compared the VSD response to the MS shuffled condition (see below, Statistical analysis and computing the MS shuffled data). The peak response in MS modulation (negative or positive peak) was computed by averaging the VSD signal within a 30 ms window around the peak response time. Matlab software was used for all calculations and statistical analyses.
Correlation analysis
To investigate the effect of MSs on synchronization between neural population, we computed the correlation between V1 pixels, for each MS (Eq. 1; Meirovithz et al., 2010, 2012). First, we subtracted the mean blank response from each MS trial, for each time point and each pixel. Next, using a sliding window of 100 ms, a Pearson’s correlation coefficient (r) was computed between the VSD response during a MS, for each V1 pixel (Pi) and any other pixel (Pj) in the imaged V1 area. This computation resulted in a vector of correlations for pixel (Pi): corr(Pi,Pj). Next, we computed the mean correlation over this vector, synchrony(Pi) and the value was assigned to pixel (Pi). This procedure was repeated for each pixel in V1, which resulted in a correlation map for each single time window. The computation of correlation maps over sequential time windows resulted in a sequence of correlation maps for each MS. Finally, the correlation maps were averaged across MSs within a single session (Fig. 3A). In summary, the correlation value within each pixel in the map reflects the synchronization between pixel (Pi) and all other pixels in V1 (Eq. 1). Positive, zero, and negative values indicate increased synchronization, no synchronization, and desynchronization between V1 pixels, accordingly. To evaluate the statistical significance, the correlation maps were computed for shuffled data (see below, Statistical analysis and computing the MS shuffled data) and compared with the original MS data.
Microsaccades induce a transient increase of synchronization. A, A sequence of mean correlation maps for all V1 pixels in Monkey L, with a 100 ms sliding window, example session (n = 42 MSs). The value of each pixel in each map represents the correlation (r) between the pixel and all the other pixels in V1 at that time window (the numbers above the maps denote the middle time point of the window). The correlations were averaged across the MSs of the same session and the shuffle correlation was subtracted. The gray pixels denote blood vessels that were removed from analysis. B, Time courses of correlation averaged over all V1 pixels in example sessions for Monkey L (left; same data as in A; n = 42 MSs) and Monkey G (right; n = 66 MSs). The continuous line represents MSs data and dashed line the shuffle data. Shaded area represent ±1 SEM and the top bar with asterisk depicts the time window with a significant difference between the correlation for MSs and shuffle data (Wilcoxon rank sum, *p> 0.05; **p> 0.01; ***p> 0.001). Shaded blue and orange bars denote the time window used for the single pairwise correlation matrices in D. C, Grand analysis of correlation TCs for Monkey L (left; n = 6 sessions) and Monkey G (right; n = 6 sessions). D, Top, Pairwise correlation matrices (same example session as in A). The matrices were computed for baseline period (correlations were averaged over 30 ms before MS onset) and peak correlation period (correlations were averaged over 100–120 s after MS onset). The shuffle correlation was subtracted. Bottom, Histograms of the pairwise correlation (r) from each matrix (***p < 0.001).
Analysis of MSs effects in trials with a small visual stimulus
To investigate the MS effects in trials with a small visual stimulus (Fig. 4, a black or white circle; radius, 0.05–0.15°; position, 0.75° below the horizontal meridian and 0.6° from the vertical meridian), we selected trials with the following criteria: (1) The VSD maps showed a clear patch of activation evoked by the stimulus. (2) The properties of the MSs (typically amplitude of 0.2–0.7°) enables to visualize the neural activation displacement over the imaged V1 area. Next, we fitted an elliptical ROI to the stimulus evoked activity in each trial with MS (as previously done in Meirovithz et al., 2012). We then computed the VSD signal using the ROIs and TCs were smoothed (with a window size of 3 time points). We defined three ROIs: (1) an ROI fitted to the stimulus evoked response before MS onset, pre-MS ROI (Fig. 4C, left top); (2) an ROI fitted to the stimulus evoked response after MS landing, post-MS ROI (Fig. 4C, left middle); and (3) an ROI of V1 area excluding the stimulus evoked response, as defined in (1) and (2). This ROI was termed as the nonstimulated ROI (Fig. 4C, left bottom). We verified no overlap between the different ROIs.
The effects of microsaccades in trials with a small visual stimulus and the comparison to blank trials. A, A sequence of VSD maps aligned on MS onset in the presence of a small visual stimulus (black dot; position, 0.75° below the horizontal meridian; 0.6° from the vertical meridian; size, r = 0.1°), an example trial. The contour lines depict an ellipse fit for two activation patches: prior to the MS (pre-MS; continuous line) and after MS landing (post-MS; dashed line). The map at t = 150 ms shows the approximate eccentricity lines of 0.75 and 2.5°, denoting the regions of the fovea (F), parafovea (P), and central fovea (CF) in the imaged V1 (see Materials and Methods). B, Eye position traces of the example trial in A and the detected MS (in red). Top, The horizontal and vertical eye position in time. Bottom, The eye position in 2D space. C, Left, Maps of the three analyzed ROIs in V1 of the example trial in A: pre-MS ROI, post-MS ROI, and the rest of the imaged region in V1, i.e., the nonstimulated (NS) ROI. Middle, Mean VSD signal in each ROI (n = 50 MSs). Baseline activity was subtracted from the VSD signal in the NS-ROI. Shaded area represents ±1 SEM across MSs. Top bars with asterisk depict the time points with significant difference between the NS-ROI modulation to the post-MS ROI and pre-MS ROI modulations (Wilcoxon rank sum test; ***p < 0.001). Right, Same data as in C middle but rescaled for comparison of the TC dynamics. D, Time courses of the VSD signal in the NS-ROI averaged across all stimulated trials in comparison with the MS modulation of the grand analysis, in blank trials (Fig. 2G). The correlation coefficient (r) between the two responses is 0.96 (p < 0.001).
Statistical analysis and computing the MS shuffled data
To evaluate statistical significance, we used nonparametric statistical tests, signed rank test to compare a population's median to zero, and Wilcoxon rank sum test to compare between medians of two populations. To assess the statistical significance of the MSs modulation in the VSD signal, we created MSs shuffled data. In each recording session, data was pooled across all trials, and the original MSs onset times were dissociated from the matched VSD trials and then randomly shuffled over the VSD trials (Meirovithz et al., 2012). This approach preserved the statistical distribution of MSs onset times relative to the VSD signal. Next, we computed the VSD response aligned on the shuffled MS data, for each session and computed the VSD response over ∼1,000 shuffled MSs. Similar results were obtained when computing random times of MSs in the VSD trials or when restricting the shuffle to close neighboring trials within the same third of the session (thus taking into account possible gradual response drift over time). To compute the shuffled condition for the correlation analysis, we used a random time for MS onsets relative to the VSD trials (the random times for shuffled data were at least 100 ms shifted from the real data). For each session, we computed ∼100 shuffled MSs. Statistical significance of the correlation in the real data was computed relative to the shuffle data.
Results
Population responses were recorded from V1 area in the right and left hemispheres of two monkeys while they performed a fixation task. The animals were maintaining fixation, i.e., holding their eye gaze on a small fixation point that appeared at the center of a visual display with a gray background (blank trials, see Materials and Methods; Fig. 1A, left). VSDI was performed in the foveal and parafoveal regions of V1 (lower visual field, approximate eccentricity: 0.25–3° and 0.2–2.5° for Monkeys L and G, respectively; see Materials and Methods). While the animals were maintaining fixation, they performed spontaneous miniature fixational saccades, i.e., microsaccades (MSs; Fig. 1A, right). The MSs amplitude was 0.45 ± 0.04 (mean ± SEM) and 0.37 ± 0.02 for Monkeys L and G, respectively (Fig. 1B), and the MSs duration was 24 ± 5.87 ms (median ± MAD) and 24 ± 4.86 for Monkeys L and G, respectively (Fig. 1C). The relation between MSs amplitude and maximal velocity confirmed the main sequence (Fig. 1D; r = 0.91, p < 0.001 for Monkey L; r = 0.96, p < 0.001 for Monkey G; see Materials and Methods for MSs detection and analysis).
To measure the neural activity evoked by MSs, we used VSDI at high spatial (mesoscale, 1702 µm2/pixel) and temporal resolution (100 Hz) from V1 area, spanning an area of 14–16 mm in diameter. The fluorescence dye signal from each pixel reflects the sum of the membrane potential from all the neuronal elements of that pixel, i.e., a population signal (rather than the response recorded from a single neuron). Thus, the VSD signal reflects the membrane potential from subthreshold activity (i.e., synaptic potentials) to suprathreshold activity (i.e., spiking activity; Ayzenshtat et al., 2010; Grinvald and Hildesheim, 2004; Jancke et al., 2004). The sensitivity to the subthreshold activity means that even subtle changes in membrane potential can be detected using VSDI. A main advantage of VSDI in this work is the ability to measure the activation patterns evoked by MSs over large parts of the foveal representation in V1, the main region to be influenced by these small fixational saccades (size <1°). We utilized the VSDI to investigate the effects of MSs generated over a homogenous blank screen, when only a tiny fixation point is present.
Microsaccades induce a biphasic modulation in V1 activity
Figure 2 shows the VSD maps, i.e., population response maps in V1 of the two monkeys, aligned on MSs onset in blank trials (no visual stimulus except for the small fixation point). Figure 2, A and E, depicts a time sequence of V1 VSD maps (20 ms apart) aligned on MSs onset, from two example sessions in Monkeys L and G. The first map of each example session illustrates the approximate regions of the fovea for each animal. Negative and positive times denote the neural responses before and after the MS onset. The VSD maps show a biphasic response that appears over most of V1 area: following MS onset (t = 0), there is an early suppression transient (dark blue pixels denote negative response; ∼0–80 ms and 40–80 ms after MS onset, for Monkeys L and G, respectively) that is followed by an enhancement transient (∼140–200 ms after MS onset; pink-white pixels denote positive response). To quantify this, we computed the time course (TC) of the VSD signal by averaging over all V1 pixels in the example sessions. Figure 2, B and F, shows the VSD TCs in each of the example sessions which reveal a biphasic modulation composed of an early suppression (negative peak at t = 70 ms after MS onset for both monkeys) followed by a late enhancement (positive peak at t = 160 and 180 ms after MS onset, for Monkeys L and G, respectively). The negative peak response amplitude of suppression is −1.68 ± 0.35 × 10−4 and −0.51 ± 0.2 × 10−4 ΔF/F (mean ± SEM over MSs) for Monkeys L and G, respectively. The peak response amplitude of enhancement is 2.81 ± 0.48 × 10−4 and 1.0 ± 0.27 × 10−4 ΔF/F for Monkeys L and G, respectively. Both the negative and positive peak responses were significantly different from the shuffled data (p value range: < 0.001 to < 0.05; Wilcoxon rank sum test; see different shuffling approaches in Material and Methods).
Next, we computed the grand analysis, i.e., across all imaging sessions, for each animal. The MSs modulation was averaged over all sessions (n = 6 for each monkey). Figure 2, C and G, depict the TCs in V1, confirming a biphasic modulation for both animals, and Figure 2, D, and H, show the quantification of this modulation. Across sessions, the suppression transient reached a negative peak at 60 ± 2.58 and 73.3 ± 3.33 ms (mean ± SEM across sessions) after MS onset for Monkeys L and G, respectively. The enhancement transient reached its peak value at 165.7 ± 6.71 ms and 206.7 ± 10.22 for Monkeys L and G, respectively (Fig. 2D,H, right). The negative peak response of the suppression had an amplitude of −1.08 ± 0.24 × 10−4 and −0.46 ± 0.11 × 10−4 ΔF/F for Monkeys L and G, respectively, and peak response amplitude of the enhancement was 1.68 ± 0.31 × 10−4 and 0.82 ± 0.15 × 10−4 ΔF/F for Monkeys L and G, respectively (Fig. 2D,H, left; p < 0.05 Wilcoxon signed rank, for significant difference from zero; p < 0.005, Wilcoxon rank sum for significant difference between negative and positive peaks).
In summary, the results suggest a clear biphasic modulation in V1 response following MSs in blank trials (68–73% of V1 pixels across all recording sessions showed a significant biphasic modulation). Notably, the first phase of the modulation is suppression, which is different from the typical increase in V1 population response to visual stimulus onset (or following a spatial shift of a small stimulus; Slovin et al., 2002; Meirovithz et al., 2012). In addition, the very small fixation point was located most of the time at the center of the foveola that is outside the imaging chamber. Thus, we anticipated only minimal effects due to possible visual stimulation of the fixation point itself on the imaged population responses within the optical chamber.
Microsaccades induce a widespread, transient increase of synchronization in V1
The analyzed MSs were performed over a homogeneous blank screen. This suggests that the MS neural modulation is not caused by a visual stimulation but may reflect a common neuronal input into V1 that can induce changes in neuronal correlation and drives the modulation. To investigate this, we computed the zero time-lag correlation (Pearson’s correlation coefficient, r) between V1 pixels for each MS (i.e., at the single trial level), using a sliding window of 100 ms (see Materials and Methods). The results of this analysis are correlation maps (i.e., synchronization maps; averaged across MSs), where the value in each pixel indicates the mean correlation between this pixel and all other V1 pixels. Thus, the map sequence shows the temporal evolution of correlation between all V1 pixels. An example session with a time sequence of correlation maps is shown in Figure 3A: the dark blue colors denote lower correlation values and warmer colors denote higher correlation values. The maps show that, at approximately t = 80–130 ms (time denotes mid correlation windows) after MS onset, there is an increase in correlation, i.e., increase in synchronization, that is widespread over V1. The time window of increased synchronization overlaps with the suppression phase of the MSs modulation and the initial part of the enhancement phase (similar results were obtained using shorter or longer time windows; see Materials and Methods).
Next, we plotted the time course of synchronization by computing the mean correlation value for all V1 pixels, in the correlation maps. Figure 3B shows the TCs of correlation from two example sessions of the two monkeys. The correlation for Monkey L peaked at t = 110 ms after MS onset with a correlation value of 0.22 ± 0.02 and for Monkey G at t = 130 ms after MS onset with a value of 0.06 ± 0.01, both are significantly different from the shuffled correlation data (see Materials and Methods; Wilcoxon rank sum p < 0.001 and p < 0.01 for Monkeys L and G, respectively). The correlation analysis in Figure 3C shows the grand average (mean across all sessions) with similar results to the example sessions: peak correlation amplitude of 0.19 ± 0.04 at t = 110 ms from MS onset for Monkey L and peak correlation amplitude of 0.08 ± 0.01 at t = 130 ms for Monkey G (peak correlations are significantly different from shuffled correlation data, p < 0.05, Wilcoxon rank sum).
Next, we wanted to test whether the single pairwise correlation matrix (rather than the mean correlation between a pixel and all V1 pixels; see Materials and Methods) also shows increased synchronization after MS onset. Figure 3D shows the pairwise correlation matrix for the same example session as in Figure 3A. Figure 3D, top left, shows the pairwise correlation matrix at the baseline period (30 ms before MS onset), and Figure 3D, top right, shows the pairwise correlation matrix at the peak TC of correlation (100 to 120 ms after MS onset; see Fig. 3B for highlighted time windows). Figure 3D, bottom, shows the distribution histograms of the correlation values in the two time windows, and there is a significant difference between the baseline correlation and peak correlation (Wilcoxon rank sum, p < 0.001). In summary, the correlation analysis (correlation maps or matrix of single pairwise correlations) revealed a transient increase in synchronization that was widespread in V1 following MSs. This result further supports the existence of a driving input into V1 that can generate a transient synchronization, aligned on MSs onset.
MS modulation in visually stimulated trials and the relation to MS modulation in blank trials
The MS modulation we found appears in blank trials, i.e., only a small fixation point is presented on the screen. Therefore, it is unclear whether this modulation exists in the presence of a visual stimulus and how it relates to the MS modulation reported in the presence of a visual stimulus (Meirovithz et al., 2012). To investigate this, we analyzed trials where the animal was fixating and a small visual stimulus that was retinotopically mapped to the imaged V1 area, appeared over the screen (Fig. 4; see Material and Methods). The small visual stimulus (position, 0.75° below the horizontal meridian; 0.6° from the vertical meridian; size, 0.05–0.15°) activated only a small part of the imaged V1 area, which enabled us to measure the MS modulation in V1 regions that were directly activated by the visual stimulus and V1 regions that were not activated by it, as in the blank trials.
Figure 4A depicts a sequence of VSD maps aligned on MS onset (t = 0), from an example trial where a small visual stimulus was turned on (prior to the MS; the eye position traces of the MS are depicted in Fig. 4B, top). The maps show a clear patch of activation before MS onset (t = 0) that is corresponding to the stimulus evoked response (VSD maps at −20 to 0). We defined two elliptical ROIs: the neural activation region evoked by the stimulus before the MS onset (pre-MS ROI; continuous line; Fig. 4C, left top), and the activation region following the MS landing (post-MS ROI; dashed line; Fig. 4C, left, middle). The eye position traces indicated that the MS was directed toward the visual stimulus, and therefore the stimulus shifted toward the center of the fovea (Fig. 4B, bottom). Thus, the VSD maps show that following a MS, there is a gradual shift of the stimulus evoked response in the fovea, toward more lateral regions in V1, i.e., toward the center of the fovea (Fig. 4A; VSD maps from 50–150 after MS onset). The last map in Figure 4A illustrates this shift toward the center of fovea, where the post-MS ROI is shifted closer to the eccentricity line of 0.75°.
Next, we wanted to investigate whether the MS modulation in the blank trials exists also in trials with visual stimulation. We therefore defined the V1 region that was not activated by the visual stimulus as the nonstimulated (NS) ROI (Fig. 4C, left, bottom). The NS-ROI thus reflects the “blank” state in the stimulated trials. Figure 4C, middle, shows the mean VSD TC of response over all stimulated trials (n = 50 MSs, 6 sessions) in the pre-MS ROI, post-MS ROI, and NS-ROI. As we previously reported (Meirovithz et al., 2012), the population response aligned on MS onset in the pre-MS and post-MS ROIs reflects the stimulus shift over the retinotopic map in V1, as dictated by the shift of the stimulus over the retina. The population response in the pre-MS shows a decrease in activity (continuous curve), while at the post-ROI (dashed curve), there is an increase of response following the MS. The VSD response in the pre-MS ROI decreased from 9.74 ± 0.6 × 10−4 ΔF/F (mean activity over −70 to 0 ms prior to MS onset) to 4.55 ± 0.7 × 10−4 ΔF/F (mean activity over 200–270 ms after MS onset; Wilcoxon rank sum p < 0.001, for difference between pre- and post-MS onset). The VSD response in the post-MS ROI increased from 3.73 ± 0.55 × 10−4 ΔF/F (mean activity over −70 to 0 ms prior to MS onset) to 10 ± 0.76 × 10−4 ΔF/F at peak response (mean activity over 120–190 ms after MS onset; Wilcoxon rank sum p < 0.001, for difference between response at pre- and post-MS onset).
Interestingly, the VSD signal in the NS-ROI shows a biphasic modulation aligned on MS onset, and notably, this modulation is ∼10 times smaller than that induced by the visual stimulation (Fig. 4C, middle, gray curve). There is a significant difference between the VSD response in the NS-ROI and the pre-MS and post-MS ROIs responses in time windows that are corresponding to the suppression phase (0–70 ms after MS onset, Wilcoxon rank sum p < 0.001), enhancement phase (120–190 ms, Wilcoxon rank sum p < 0.001), and later times 250–320 ms (Wilcoxon rank sum p < 0.001). To compare the response dynamics of the MS modulation in all three ROIs, we normalized the VSD response (to overcome the large response amplitude difference). Figure 4C, right, shows the comparison of the normalized VSD response in all three ROIs, and there is a clear temporal difference between the dynamics of the TC modulation in the NS-ROI and the VSD response in the pre- and post-ROIs.
Next, we wanted to compare the MS modulation in the blank trials with the MS modulation in the NS-ROI of the stimulated trials (Fig. 4D). The amplitude and dynamics of the MS modulation in both conditions is highly similar as indicated by the value of Pearson’s correlation (r = 0.96, p < 0.001). In summary, the MS modulation in blank trials appears also in the nonstimulated V1 region, during visual stimulation.
The MS modulation cannot be explained by an external motion of the fixation point
Thus far, we have studied V1 population responses to MSs generated by the animals, spontaneously, while the monkeys were maintaining fixation around a fixation point (FP). To test whether the MSs modulation might be explained by the FP movement following a MS, we performed a set of control experiments on a third monkey (Monkey B). Here we artificially moved the FP or a tiny point located at the perimeter of the FP (i.e., at the surrounding border of the FP; see Materials and Methods). Figure 5A shows the experimental design of the control experiment: the FP (or a small dot located at the perimeter of the FP), moved inward, into the imaging chamber or outward, i.e., away from the imaging chamber. The inward motion of the FP simulated a MS away from the imaging chamber and the outward motion of the FP stimulated a MS toward the imaging chamber (see Materials and Methods). To mimic real MSs properties, we used similar MSs amplitude and MSs frequency: the points shifted with an amplitude of 0.3–0.6°, inward or outward of the imaging chamber, for 200–300 ms, and then returned back to the initial position.
Artificial movement of the fixation point in comparison with the microsaccade modulation. A, A schematic illustration of the control paradigm with an artificial movement of the fixation point (FP). Left, While the animal is fixating, the FP is shifted over the screen (in this example by ∼0.3°), mimicking a typical MS amplitude. Following 200–300 ms, the FP returned to its initial point. Right, Movement direction of the FP: inward or outward of the imaging chamber. B, Left, A comparison between the V1 VSD response to a MS (black curve, right y-axis; t = 0 is the MS landing; n = 59 MSs) and the response generated by an artificial inward movement of the FP (cyan curve, left y-axis; t = 0 is movement onset; n = 30 trials). Top bars with asterisk depict the time windows with significant difference between the two responses. Right, Same as for left, but when the FP moves outward (n = 26 trials). Baseline activity was subtracted from each condition (Wilcoxon rank sum; *p > 0.05; ***p > 0.001). C, The MS modulation is divided into two classes: MSs directed inward (leading to FP out of the chamber; gray curve) or MS outward (leading to FP into the chamber; black curve) of the chamber for each monkey. Left, Monkey L, MSs with inward direction (n = 52), MSs with outward direction (n = 167). Right, Monkey G, MSs with inward direction (n = 146), MSs with outward direction (n = 107). The correlation coefficient (r) between the MS modulation for the two MS directions: r = 1, p < 0.001, for both monkeys. Shaded area represents ±1 SEM across MSs.
Figure 5B shows the VSD TC of response in V1 for the two control conditions (t = 0 for FP motion, cyan curves, left y-axes): inward motion (Fig. 5B, left) and outward motion (Fig. 5B, right). For comparison purposes, in the same graphs, we plotted the MS modulation in the blank trials of the same animal, aligned on MS landing (t = 0, black curve; right y-axes; the two y-axes have different scales because the V1 response to the FP movement was much larger than the MS modulation in the blank trials). Figure 5B, left, shows that following an inward movement of the FP (the FP moved into the imaging chamber), a very sluggish and slow response gradually develops in V1. This response may reflect low subthreshold spread of activation due to the FP stimulation in the new location (note that this response is much lower and slower than a typical V1 response to a visual stimulus). Figure 5B, right, shows that following an outward motion of the FP (the FP moved away from the imaging chamber), there is very little change in V1 response. To quantify the response differences generated by the artificial FP movement versus the MS modulation (in blank trials), we compared the response amplitude during the MS suppression (30–100 ms after MS landing) and at peak enhancement (170–240 ms after MS landing) in the inward movement (Fig. 5B, left). The response amplitude in the FP control stimulus during the time window of MS suppression (30–100 ms) was 3.52 ± 1.22 × 10−5 ΔF/F (n = 30 trials), which was significantly higher than the negative peak response in the MS modulation: −0.77 ± 0.69 × 10−5 ΔF/F (n = 59 MSs; Wilcoxon rank sum test p < 0.05). The response amplitude for the FP control stimulus at the peak time of MS enhancement (170–240 ms) was 21.66 ± 1.93 × 10−5 ΔF/F (n = 30 trials), which was significantly higher than the response amplitude in the MS modulation 5.74 ± 0.83 × 10−5 ΔF/F (n = 59 MSs; Wilcoxon rank sum test p < 0.001). Finally, the response amplitude for the FP outward movement, at the MS suppression phase (30–100 ms), was 4.5 ± 1.42 × 10−5 ΔF/F (n = 26 trials) that was significantly higher than the response amplitude in the MS modulation −0.77 ± 0.69 × 10−5 ΔF/F (n = 59 trials; Wilcoxon rank sum test p < 0.05). These results suggest that the influences of artificial movement of the FP and the MS modulation are different.
The VSD responses in the FP control experiments showed small changes between the inward and outward FP movement. We therefore decided to test whether the direction of real MSs can induce different V1 response. The VSD response for inward MSs (shifting the FP outward the chamber, gray curve) and for outward MSs (shifting the FP inward, toward the chamber, black curve) is plotted in Figure 5C for both monkeys. The VSD responses to both directions of real MSs had similar amplitude and dynamics and showed a high correlation between the TCs for the two MSs directions (r = 1, p < 0.001 for Monkeys L and G). In summary, the MS modulation induced by real MSs is different from that generated by an artificial movement of the FP (in agreement with Troncoso et al., 2015). Moreover, MSs with opposite directions, which shift the FP toward or outward of the imaging chamber, generate a highly similar VSD response, in accordance with a recent study in V1 of monkeys (Wu et al., 2022). Finally, a correlation analysis aligned on the FP shift did not show increased correlation relative to the shuffle data.
Discussion
Our vision appears as stable despite the incessant eye movements that are leading to continual movement of the visual input over the retina. This can suggest the existence of an extraretinal neuronal signal that can be used for correcting the stimulus shift and lead to the generation of visual stability. While numerous studies suggested several neural mechanisms that can mediate visual stability for saccades (Duhamel et al., 1992; Khayat et al., 2004; Merriam et al., 2007; Wurtz, 2008; Zirnsak et al., 2014; Sun and Goldberg, 2016; Niemeyer et al., 2022), only a few addressed this notion in MSs (Leopold and Logothetis, 1998; Martinez-Conde et al., 2004; Kagan et al., 2008; Hafed and Krauzlis, 2010; Watamaniuk et al., 2023). Furthermore, the few neurophysiological studies that investigated this topic in MSs did not focus on V1 fovea, the area that is expected to be most influenced by the tiny fixational saccades. In our work we implemented a different approach: using VSDI we characterized the influence of MSs on V1 activity at the fovea, in the blank condition where only a small fixation point is present over a gray homogenous screen. Following MSs, we found a biphasic modulation in V1 population response. Additionally, there is an increase in synchronization following a MS, supporting the existence of a common input into V1 that drives the neuronal population upon the MS onset. The biphasic modulation could not be explained by an artificial movement of the FP and coexisted in V1 next to the MS modulation in the presence of a visual stimulus.
MSs effect in V1 reveal a biphasic modulation
Some of the studies that investigated the influence of fast eye movement in the presence of a visual stimulus suggested that the modulation in V1 can be explained by the stimulus shift over the retina: for saccades (Idrees et al., 2020; Parker et al., 2023) and for MSs (Gur and Snodderly, 1997; Leopold and Logothetis, 1998; Martinez-Conde et al., 2000, 2002; Meirovithz et al., 2012; Gilad et al., 2014). However, other studies raised argumentations beyond the stimulus-related component and reported on evidence that can support the existence of an ERI to V1 (Kagan et al., 2008; Bremmer et al., 2009; Troncoso et al., 2015; Niemeyer et al., 2022; Wu et al., 2022).
Following MS onset, we found a biphasic modulation in V1 population response, composed of a suppression followed by enhancement. This biphasic modulation could not be explained by an artificial movement of the FP, thus further supporting the existence of an ERI to V1. Previous studies on saccades in the presence or absence of visual stimuli reported on neural modulation that can support ERI in various visual cortical areas. This neural modulation showed typically suppression and/or enhancement following the saccade onset (Royal et al., 2006; Rajkai et al., 2008; Bremmer et al., 2009; Miura and Scanziani, 2022; Niemeyer et al., 2022; Denagamage et al., 2023). An evidence for ERI in MSs was reported by just a few studies (Snodderly et al., 2001; Troncoso et al., 2015; Wu et al., 2022), and some of them reported on a biphasic modulation composed of suppression (70–100 ms from MS onset) followed by enhancement (160–200 ms from MS onset). These findings are in accordance with our results. Finally, we note that we found similar biphasic modulation in V1 for large saccades that were performed over a blank screen (unpublished results).
The suppression phase of the neural modulation in large saccades and fixational saccades was suggested to play a role in “saccadic suppression,” a behavioral phenomenon characterized by a transient reduction in visual sensitivity around the saccadic onset (Volkmann, 1986; Diamond et al., 2000; Burr, 2004; Binda and Morrone, 2018) and recently also around MSs onset (Scholes et al., 2018; Intoy et al., 2021; but see also Martinez-Conde et al., 2013). This behavioral phenomenon along with the neural correlates was suggested to be involved in perceptual stability also in MSs (Martinez-Conde et al., 2004, 2013; Hafed and Krauzlis, 2010). Postsaccadic neural enhancement was suggested as a mechanism that can underlie behavioral enhancement of visual processing after saccades, such as increase in visual sensitivity (Burr and Ross, 1982), increased contrast sensitivity (Diamond et al., 2000), and reduction in reaction times (Johns et al., 2009). Other neurophysiological studies reported that the postsaccadic neural enhancement is correlated with increased visual sensitivity to high spatial frequencies (Niemeyer et al., 2022), sharpening of orientation tuning curves (Ibbotson and Krekelberg, 2011), and faster evolution of figure-ground segregation (Gilad et al., 2014). In recent years, behavioral enhancement of visual processing was also reported in relation to MSs, for example, improved performance in visual acuity demanding tasks (Poletti et al., 2013; Ko et al., 2016; Intoy and Rucci, 2020), enhanced sensitivity for a specific range of spatial frequencies 100–200 ms after MS onset (Bellet et al., 2017; Scholes et al., 2018; Intoy et al., 2021; but see Mostofi et al., 2016), enhanced visual processing during visual exploration and search (Otero-Millan et al., 2008; McCamy et al., 2014), and restoration of visibility after perceptual fading (Martinez-Conde et al., 2004, 2006). Finally, saccades and MSs were suggested to serve as a clock which parses vision between fixation epochs (Paradiso et al., 2012). According to this notion, there is a link between the two behavioral phenomena of saccadic suppression and the followed increased visual enhancement: their temporal sequence allows to actively mask the presaccadic stimulus and boost the response to the postsaccadic stimulus (Wurtz, 2008; Ibbotson and Krekelberg, 2011; Paradiso et al., 2012; Binda and Morrone, 2018). While this mechanism was suggested to play a role in perceptual stability in relation to saccades (Niemeyer et al., 2022), it may play a similar role for MSs, which exhibit also a biphasic neuronal modulation.
Microsaccades induce increased synchronization
The neuronal population in V1 showed an increase in synchronization shortly after MS onset, the peak amplitude of the synchronization occurred ∼100 ms after MS onset. Previous studies of neuronal synchronization in relation to saccades and MSs in the visual cortex were reported in animals that were presented with various visual stimuli (Leopold and Logothetis, 1998; Martinez-Conde et al., 2004; Maldonado et al., 2008; Bosman et al., 2009; Meirovithz et al., 2012; Lowet et al., 2016; Niemeyer et al., 2022). A previous VSD study in fixating monkeys reported that MSs performed during the presence of small and large visual stimuli induced increased synchronization in V1 ∼100 ms after MS onset (Meirovithz et al., 2012). Bosman et al. (2009) reported on increased coherence in V1 and V4 following MSs, and Lowet et al. (2018) reported on gamma band synchronization in V1 and V2. These observations were suggested to be linked with enhanced information processing of the stimulus, in its new landing position, following the spatial shift over the retina (Meirovithz et al., 2012; Niemeyer et al., 2022).
The increase in synchronization following a MS onset, which we found in the blank trials, can be linked to the above reports and may reflect the neural preparation in V1 for enhanced processing of an upcoming stimulus. In addition, Martinez-Conde et al. (2000, 2002) suggested that the synchronization after each MS allows the MS signal to propagate and synchronize the neural activity across the visual areas. Additionally, Melcher (2011) suggested the activation of a “visual stability network” and the widespread synchronization signal we found could support this notion.
Possible neural sources for the MS modulation
The biphasic MS modulation in our work could not be explained by an artificial movement of the FP, which further supports the existence of an ERI to V1. The few neurophysiological studies which reported on evidence for ERI to V1 in MSs suggested several possible neuronal sources: proprioceptive signals, corollary discharge (CD, efference copy of the motor command to the oculomotor muscles), global motion signals, and attentional signals (Kagan et al., 2008; Troncoso et al., 2015; Wu et al., 2022). Studies which focused on saccades offered similar sources of ERI signals to the visual system (Duffy and Burchfiel, 1975; Corbetta et al., 1998; Royal et al., 2006; Rajkai et al., 2008; Ibbotson and Krekelberg, 2011; Miura and Scanziani, 2022; Niemeyer et al., 2022).
The superior colliculus (SC) plays a key role in motor execution of saccades and MSs (Hafed et al., 2009). A CD signal, starting tens of milliseconds before MSs onset, can emerge from the intermediate layers of the SC and arrive to the visual cortex via different pathways (Wurtz, 2008; Krauzlis et al., 2017). A recent study by Denagamage et al. (2023) reported on neural suppression following saccades that was initiated by the activation of inhibitory V4 neurons. The authors suggested the pulvinar as a possible neural source for this response, which was previously suggested to carry CD signal from the SC to various visual areas such as MT in primates and V1 in rodents and primates (Stepniewska et al., 2000; Shipp, 2004; Berman and Wurtz, 2010; Kuang et al., 2012; Schneider et al., 2020, 2023; Miura and Scanziani, 2022). Therefore, the initial suppression phase we report can fit to a similar possible pathway from the SC through the pulvinar into the inhibitory neurons in V1. Another possible pathway of the CD travels through the mediodorsal nucleus of the thalamus to the frontal eye field (FEF; Sommer and Wurtz, 2002) and from there, through V4, to V1. However, this track might be too slow to account for the early neural suppression immediately after MS onset that may start even before the MS onset (Fig. 2C).
Proprioceptive signals from the oculomotor muscles can carry information about the eye position and movement kinematics. While such signals were previously shown to affect the neural activity in V1 (Trotter et al., 1993; Buisseret, 1995), most of the previous studies indicated that the oculomotor proprioceptive signal is too slow to account for the fast suppression in MSs (Balslev et al., 2012; Sun and Goldberg, 2016). One may speculate that a proprioceptive signal can be a relevant candidate for the late enhancement phase present in our results, yet the anatomical pathways that carry such information to the visual cortex were not reported yet.
Attention has been extensively linked with saccades and MSs (Corbetta et al., 1998; Hafed and Clark, 2002; Engbert and Kliegl, 2003; Martinez-Conde et al., 2004; Shipp, 2004). A well-established pathway initiating from FEF to V4 may mediate an attention-related input to V1 (Corbetta et al., 1998; Fries et al., 2001; Martinez-Conde et al., 2013; Denagamage et al., 2023). The SC and the pulvinar might be also involved in a parallel pathway mediating attentional signal from the FEF through these subcortical centers (Shipp, 2004). The attentional effects in V1 activity appear ∼200–250 ms after stimulus onset (Roelfsema et al., 1998; Lamme and Roelfsema, 2000). This timing fits well with the temporal dynamics of the enhancement in the MS modulation, as we reported.
Limitations of the blank condition and the MS modulation in the presence of a visual stimulus
The blank, i.e., the “no stimulus,” condition in this study contains several limitations. Although the animals were fixating on a homogeneous gray screen (which was fixed for the entire trial and intertrial interval), it is still possible that some V1 neurons were driven by this weak visual stimulation. In addition, although the animals performed the task in a dark room, we cannot completely rule out some reflections from the gray screen. Previous studies, investigating MSs in similar “no stimulus” or “blank” conditions, faced similar challenges (Martinez-Conde et al., 2000, 2002; Snodderly et al., 2001; Kagan et al., 2008; Troncoso et al., 2015; Wu et al., 2022). Nevertheless, MSs modulation similar to those found in this study were reported when using a blank gray screen or when the screen was completely dark (Snodderly et al., 2001; Kagan et al., 2008). In our work we validated that the MSs modulation in the blank condition was not due to an artificial movement of the small FP, when this movement was applied over the same gray screen. In addition, the RFs of the imaged neuronal population were far from the screen margins and surroundings (Fig. 1A), and thus the MS modulation cannot emerge from the screen margins. Finally, we note that the effects of any noncontrol visual stimulation (e.g., screen reflection) were not observed in baseline activity or in the shuffle data or during external motion of the FP—thus minimizing the influence of such factors on our results.
Our study showed the coexistence of two different MS modulations (Fig. 4): one that is induced by the visual stimulus shift over the retina (visually driven modulation) and another that is evident in the nonstimulated region. The stimulus-related modulation is 10 times larger than the MS modulation in the nonstimulated region and the latter is highly similar to the MS modulation in the blank trials. Interestingly, the temporal evolution of the MS modulation in the two conditions is different. The suppression phase in nonstimulated region (and MS modulation in the blank trials) occurs before the activation decrease for the visually driven modulation (pre-MS ROI). In contrast, the enhancement phase in the nonstimulated region (and MS modulation in the blank trials) is delayed relative to the response increase in the post-MS ROI. This may further support the notion that sources for the two modulations are, at least partially, independent and carry different information. It could be also that at some point downstream the two modulations converge into one signal which mediates visual stability, for example, by remapping (shifts in visual neurons RFs; Duhamel et al., 1992; Colby and Goldberg, 1999; Wurtz, 2008; Melcher, 2011). Additional future studies are required to uncover and dissect the anatomical pathways underlying the MS modulation in the absence of visual stimulation and how these interact with visual stimulation and possibly leads to the generation of visual stabilization.
Footnotes
We thank Roy Oz, Shany Nivinsky Margalit, Ofir Korch, and Ossnat Bar-Shira. This work was supported by the Israel Science Foundation 955/16 and 2443/18.
The authors declare no competing financial interests.
- Correspondence should be addressed to Hamutal Slovin at Hamutal.Slovin{at}biu.ac.il.
