Fig 1.
Modeling MS V1 neural dynamics.
(A) Conceptual overview over the model. A (small) region in V1 was modeled as a PING network of spiking excitatory (E cells, regular-spiking) and inhibitory neurons (I cells, fast-spiking). They were connected through AMPA- and GABA-A-type synapses (see Materials and methods and Fig 2A for details). On the PING network, we imposed currents mimicking MS-modulated input from LGN and/or corollary discharges to V1 (see Materials and methods). We used the spikes and approximate LFPs in the PING network for subsequent analysis. (B) A representative experimental MS-triggered TFR of V1 LFP power. MSs occurred at t = 0 s. Note the broadband activity just after the saccade onset (0–100 ms) followed by a narrow-band gamma signal (100 ms onwards). (C) An MS-triggered TFR of the simulated LFP. Conventions as in panel B. The Vm is able to produce both a broadband signal directly following the saccade onset (transient) as well as the narrow-band gamma response afterwards (sustained). AMPA, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid; GABA, gamma-aminobutyric acid; LFP, local field potential; LGN, lateral geniculate nucleus; MS, microsaccade; PING, Pyramidal-InterNeuron Gamma; TFR, time-frequency representation; Vm, model visual cortex.
Fig 2.
Synchrony across the network depends on connection strength and input difference more strongly during sustained gamma-band activity than during MS-induced transients.
Synchrony was measured by PLV. (A) A scaled-down schematic representation of the model network. Inhibitory neurons are omitted for clarity. The full network consists of excitatory neurons placed on a square 40 × 40 grid together with a square grid of the same diameter containing 20 × 20 inhibitory neurons (not depicted). LFP electrodes are placed on a 10 × 10 grid spread equally across the 40 × 40 neuronal grid. The boundaries of the grid are periodic, i.e., the neurons are placed atop a toroidal surface. The 2 red arrows indicate 2 electrode pairs: a neighbor pair, at minimal distance, and a pair at maximal distance. (B) Network was driven by input consisting of retinotopically smoothed Gaussian white noise. This input current was modulated over time by multiplying it by the MS modulation kernel (see Fig 1A, below right) for a total of 50 MSs. The SNR factor of the temporal Gaussian noise was equal to 2 (see Materials and methods). (C) Mean synchrony strength expressed as the mean PLV (see Materials and methods) across MSs between neighboring LFP electrodes (see smallest of the red arrows in panel A). (D) Same as panel C, but now between LFP electrodes with maximal distance to each other (see longest of the red arrows in panel A). Note the lack of synchrony in the gamma band during the sustained period (t = 150–350 ms). (E) The correlation coefficients between electrode distance and PLV for a broad range of frequencies and all intersaccade time points. Only the synchrony in the narrow-band gamma activity (25–50 Hz) during the sustained period depends on the electrode distance. (F) The same as panel E but for stimulus input differences. Only the neuron pairs that are within 4 interneuron distances from each other were used for generating this figure. Neurons further apart have little to no synchrony during the sustained period (i.e., the correlation in panel E is negative) and were therefore not included. LFP, local field potential; MS, microsaccade; PLV, phase-locking value; SNR, signal-to-noise ratio.
Fig 3.
MS-dependent spatiotemporal organization of network activity.
Network was a lattice with excitatory–inhibitory neurons locally connected with periodic boundary conditions (similar to Fig 2A). (A) Oriented stimulus, isotropic connectivity. Left: firing rates of excitatory RS neurons in the network that was driven by a horizontal (top) or vertical (bottom) bar-shaped stimulus. Network synchronization (PLV with the center neuron, denoted by a black cross, as reference) is shown in the transient (middle) and in the sustained (right) period with 50-ms time windows. Middle: in the transient period, PLV with the center neuron was high over the whole network due to the MS-induced reset. Right: in contrast, in the sustained period, the PLV was high only close to the reference neuron in the network center, and the PLV profile was shaped by the orientation of the bar-shaped stimulus. (B) Isotropic stimulus, anisotropic connectivity. Left: the center neuron (indicated by a black cross) was preferentially connected to neurons left and right from it (top), or to neurons below and above it (bottom). Middle: in the transient period throughout the network, the PLV with the center neuron (indicated by a black cross) was high. Right: during the sustained period, PLV was only high close to the reference neuron, and the shape of the PLV profile reflected the preferred connectivity orientation. MS, microsaccade; PLV, phase-locking value; RS, regular-spiking neuron.
Fig 4.
Monkey LFP measurements across V1 cortical locations replicate model results.
(A–B) Single contact pair examples. (C–E) Population-level analysis (12 recording sessions, each with 3 laminar probes from 2 monkeys). (A) Temporally resolved PLVs for multiple frequencies between 2 LFP electrodes in V1 relatively far apart (4–6 mm,) as a function of time, aligned to the onset of MSs. Middle subplots represents PLV values for stimulus grating condition including large contrast variation. Right subplot represents PLV values with stimulus grating containing low contrast variation. Black lines represent MS-triggered averaged eye speed. (B) The same as panel A, but for relatively close probes (2–3 mm). (C) The population-averaged V1-V1 MS-triggered PLV spectrum. (D) Represents the explained variance of MS-triggered PLV values as a function of RF distance between probes. (E) Same as panel D, but as a function of stimulus contrast difference. LFP, local field potential; MS, microsaccade; PLV, phase-locking value; RF, receptive field.
Fig 5.
Monkey V1 and V2 cortical locations replicating model results.
(A) Schematic illustration of how the laminar probes were inserted in V1 and reaching, in many cases, V2 lying beneath. Panel is taken from [29]. (B) Single contact pair examples. (C–E) Population-level analysis (12 recording sessions, each with 3 laminar probes from 2 monkeys). (B) Temporally resolved PLVs for multiple frequencies a contact pair situated in V1 and in V2, aligned to the onset of MSs. Upper panel shows the RFs of the corresponding. Lower subplots represent PLVs for stimulus grating condition including large (left) and large (right) contrast variation. (C) The population-averaged V1-V1 MS-triggered PLV spectrum. (D) Represents the explained variance of MS-triggered PLVs as a function of RF distance between probes. (E) Same as panel D, but as a function of stimulus contrast difference. MS, microsaccade; PLV, phase-locking value; RF, receptive field.
Fig 6.
Illustration of differential synchronization of 2 V1 locations to 1 V2 location (from 1 session, Monkey M1).
(A) The RF locations of the 2 V1 locations (V1a, V1b) and V2 location. (B–C) Stimulus grating had uniform contrast. (B) V1a-V2 MS-triggered TFR PLV. Squared dashed box indicates the sustained period. (C) Same as panel B, but for V1b-V2 MS-triggered TFR PLV. (D–E) Stimulus had varying spatial contrast. (D) V1a-V2 MS-triggered TFR PLV. (E) Same as panel D, but for V1b-V2 MS-triggered TFR PLV. MS, microsaccade; PLV, phase-locking value; RF, receptive field; TFR, time-frequency representation.
Fig 7.
MS-dependent synchronization shapes information transfer to downstream neurons.
(A) Excitatory–inhibitory visual model network, receiving MS-modulated spatially structured input derived from natural image luminance (bottom). The V1 network was unidirectionally connected to a downstream V2 neuron. (B) The synchronization profile (measured by PLV) of the center V1 neuron to all other neurons in V1 is shown for the input in panel A. During the transient period, synchrony was high across the V1 network, whereas synchrony was local and anisotropic in the sustained period. Synchrony was high for neurons receiving similar input. (C) Information transfer of V1 neurons to V2 (quantified as STA(V1→V2), see Materials and methods). (Left) In the transient period, synSTA was high for neurons with strong input (and therefore high firing rate). (Right) In contrast, in the sustained period, the synSTA map showed high values for neurons with similar input. (D) Contour lines of PLV in V1 during the sustained period (see panel B) as well as for the synSTA from V1 to V2 (see panel C) overlaid on of the anatomical connectivity. The synSTA maps (green and red contour lines) did not strictly follow the synaptic connectivity profile (heat map). During the transient, they were biased towards higher input. In the sustained period, they were biased towards higher synchrony (PLV, blue contours). (E) The same simulation as in panel A, but now using a different stimulus image, where the center neuron in V1 is situated over a bright patch. (F) Similar to panel C, but for the simulation using the stimulus in panel E. This second stimulus led to synSTA maps that were similar in both the transient (left) and sustained (right) periods. (G) Similar to panel D, but using the stimulus in panel E. With this stimulus, the synSTA maps showed large amount of overlap in both the transient and sustained periods. This is due to the fact that the neurons were synchronizing to the center neuron, which were also the neurons with high input. MS, microsaccade; PLV, phase-locking value; synSTA, synaptically confined spike-triggered average.
Fig 8.
Information transmission (quantified by the STA, see Materials and methods) from V1 neurons to V2 neurons.
(A–C) Schematic illustration of factors that could modulate the influence of a particular V1 neuron on a V2 neuron. We compare the information transfer in the transient and sustained period if the activation of V2 is mostly determined by higher spike rates (A), local synchronization (B), or optimal phase differences (C) in V1. (D) Scatter plots from simulations using 100 different natural images. synSTA is plotted as a function of (left) spike rate, (middle) synchrony, (right) gamma phase difference. Note: Because the transient is relatively short, “phase” is equivalent to spike latency relative to the saccade onset. (E) We used (normalized) MI to quantify how much of the variance in synSTA between V1 and V2 neuron is explained by the V1 neuron’s spike rate, synchrony, and phase. In the transient period, the transfer entropy was best explained by spike rate and phase. (F) Similar to panel D, but for the sustained period instead of the transient. (G). Similar to panel E, but the sustained period was analyzed instead of the transient. Here, the synSTA is mainly explained by synchrony and phase. This shows that synchrony-dependent organization in the sustained period has strong effects on feedforward information transmission. MI, mutual information; synSTA,.
Table 1.
Parameters for the RSs and FSs.
Table 2.
Parameter values for the MS modulation kernel (see Eq 10).