Layers of the monkey visual cortex are selectively modulated during electrical stimulation

The study investigates how external electrical stimulation modulates visually evoked local field potentials across cortical layers of primate primary visual cortex (V1). Although electrical stimulation can alter neuronal membrane potentials and entrain single-neuron activity, its layer-specific impact on population-level signals like LFPs is not well understood. Understanding this relationship is important for linking laminar microcircuit dynamics to human EEG and for optimizing neuromodulation therapies. Given the conserved laminar architecture and known feedforward/feedback pathways in V1, the authors test whether transcranial alternating current stimulation (tACS) induces layer- and phase-specific modulation of sensory-evoked LFPs in nonhuman primates, hypothesizing stronger, phase-dependent effects in deeper layers due to their anatomical and physiological properties.

Prior work has characterized laminar connectivity in neocortex, with thalamic inputs to layer 4 projecting to layers 2/3 and then to layers 5/6. Laminar LFP recordings show early sinks in layer 4 followed by superficial and deep activity, with deeper layers often exhibiting larger evoked LFPs, attributed to stronger dipoles from large pyramidal neurons and higher firing rates. Electrical brain stimulation (including tACS) can modulate neuronal excitability and entrain spikes, yet translation to layer-specific LFP effects remains unclear. Nonhuman primates provide a suitable model for TES biophysics matching humans. The literature also notes that electric field strength and tissue conductivity vary across layers, and prior modeling suggests deeper layers can contribute strongly to LFPs. These studies motivate testing whether tACS modulates sensory-evoked LFPs in a layer- and phase-dependent manner and whether biophysical field distributions alone explain observed effects.

Subjects: Two female capuchin monkeys (Cebus apella), 15 years old, 1.5–3 kg, lightly anesthetized with 1–1.5% isoflurane. Approvals by IACUC (AP2014-510). Implants: headpost and occipital recording chamber. Laminar recordings: A 23-contact linear laminar probe (U-Probe, Plexon) with 0.1 mm spacing, impedance 300–750 kΩ, inserted perpendicular to V1 to span all layers and beyond. Signals amplified (×10), analog-filtered (0.1–500 Hz for LFP; MUA 300–5,000 Hz), digitized and downsampled to 2 kHz (then to 1 kHz for analysis). Reference on dura. Visual stimulation: Full-field high-intensity flash at 2.3 Hz (one every 435 ms), 200–400 trials per session in an electrically shielded chamber under dim light. Two sessions per monkey: Flash (no stimulation) then Flash + AC (with stimulation), with rest between sessions. Electrical stimulation (tACS): Starstim (Neuroelectrics) with 10 mm radius Ag/AgCl electrodes. One electrode over occipital scalp near probe, the other over right temporal area. Sinusoidal 1.5 Hz. Current intensity peak-to-zero 0.1 mA (monkey 1) and 0.2 mA (monkey 2). 30 s ramp-up, continuous stimulation through trials, 30 s ramp-down. Montage aimed to orient fields parallel to probe. Artifact rejection: For Flash + AC data, applied ICA using SOBI to isolate 1.5 Hz stimulation components. Identified sinusoid at 1.5 Hz by FFT (no harmonics detected). Removed using discrete Fourier transform subtraction of the 1.5 Hz component. Back-projected cleaned signals. Verified removal via power spectral density showing suppression at 1.5 Hz. Preprocessing: Digital bandpass 0.5–100 Hz (4th-order Butterworth), notch at 60 Hz, downsample to 1 kHz. Epochs from −50 to 250 ms relative to flash onset, baseline correction on −50 to 0 ms. Identified P1 (55–90 ms) and N1 (100–140 ms) peaks per trial. Trials with amplitude beyond mean ± 5 SD excluded. Trials per condition: Monkey 1, Flash 320, Flash + AC 216; Monkey 2, Flash 295, Flash + AC 389. Layer assignment: Combined current source density (from Flash condition), electric field profile, and anatomical references to delineate layers: layer 1 boundary where electric field began to rise; layers 2/3 between layer 1 and layer 4AB; four contacts assigned to layer 4C and 5/6, two to 4AB, five or six to 2/3, and one or two to layer 1. White matter control obtained by lowering probe 1.3 mm to place five tip contacts in white matter. Electric field measurement: Bandpass 0.5–2 Hz on raw data, divide by 10 to correct amplifier gain, compute spatial gradient along contacts to estimate radial electric field per contact, extract 1.5 Hz amplitude and phase via FFT. Phase dependency analysis (LFP): For Flash + AC, determined AC phase at flash onset per trial and contact; verified uniform phase distribution due to non-harmonic 2.3 vs 1.5 Hz. Sorted layer-averaged trials into 20 phase bins (18°). Computed phase-binned mean amplitudes for P1 and N1, then mean direction and vector length per layer. Addressed bimodality (noted in monkey 1) using axial mean methods (CircStat). Control: applied same analysis in Flash condition using a virtual 1.5 Hz reference. MUA analysis (monkey 2): Zero-phase digital bandpass 300–5,000 Hz, threshold-based spike detection using median absolute deviation (α=3.5). Constructed peristimulus time histograms (10 ms bins), averaged across trials and channels within each layer. Compared firing rates between layers 2/3 vs 5/6 and across conditions. For phase effects, analyzed subsets aligned to AC peak (96 trials) and trough (90 trials). Computational modeling: Used biophysically detailed V1 cortical column model (230,924 neurons; 17 classes; excitatory and inhibitory; LGN input via FilterNet; background Poisson). Visual stimulus: 50 ms flash following 500 ms gray, then 350 ms gray, 10 trials. LFP computed via filtered extracellular potentials (≤100 Hz), baseline corrected. Integrated tACS by assigning extracellular potentials from a uniform vertical electric field (16 mV/mm) with 1.5 Hz sinusoid; started at 250 ms. Shifted waveform so flash coincided with AC peak or trough. Recorded membrane currents in passive basal dendrites of 500 randomly selected excitatory neurons per layer; subtracted tACS-only currents to isolate LGN-driven components under peak vs trough. Statistics: Cluster-based nonparametric permutation tests (2,000 permutations, alpha 0.01) comparing Flash vs Flash + AC across contacts and time points. Phase dependency significance via permutation on phase labels (5,000 shuffles), computing vector lengths and one-tailed p-values. MUA comparisons via paired t tests (layers within condition) and unpaired t tests (conditions within layer).

• Deeper layers show enhanced evoked LFPs with AC: LFP amplitudes after ~100 ms post-flash increased under Flash + AC compared to Flash, especially in layers 4–6. Cluster-based permutation tests found significant clusters from ~100–250 ms in layer 4AB to 5/6 for both monkeys (monkey 1: p=4.99×10⁻4; monkey 2: p=9.99×10⁻4). No significant differences 0–100 ms.
• Quantitative component changes (deep layers 5/6): Monkey 1 N1 0.59±0.10 (Flash + AC) vs 0.47±0.12 (Flash); P1 0.13±0.08 vs 0.12±0.09. Monkey 2 N1 0.38±0.12 vs 0.28±0.06; P1 0.10±0.09 vs 0.06±0.06.
• Phase dependence is layer-specific: P1 and N1 amplitudes showed significant phase preference only in deeper layers (layers 4–6) in both monkeys. Monkey 1 exhibited bimodal circular distributions with strong directionality for P1 in layers 4–6; mean directions (bimodal) for P1 ranged approximately −145° (L1) to −113° (L5/6). N1 bimodality was weaker. Monkey 2 showed unimodal distributions with strong directionality (P1 mean direction ~−159° in L1 to ~−110° in L5/6; N1 mean directions varied across layers). Control analysis using virtual AC in the Flash condition showed no significant phase preference in any layer.
• Electric field distribution vs physiological effects: Measured electric fields were largest in layers 2/3 and decreased with depth. Monkey 1 average fields (V/m): L1 0.62, L2/3 2.53, L4AB 1.18, L4C 0.62, L5/6 0.37. Monkey 2: L1 0.97, L2/3 2.76, L4AB 1.60, L4C 1.59, L5/6 1.42. Despite stronger fields superficially, significant LFP modulation and phase dependence occurred only in deeper layers, indicating dominance of neurophysiological layer properties over field amplitude.
• MUA increases in deep layers with AC (monkey 2): Layers 5/6 had higher firing than layers 2/3 in both conditions (Flash: 6.83±5.35 vs 5.32±2.22 spikes/s, p=2.03×10⁻6; Flash + AC: 9.85±6.09 vs 5.03±2.03 spikes/s, p=6.09×10⁻42). AC increased firing in layers 5/6 vs Flash (p=2.54×10⁻12); no significant change in layers 2/3. Peak vs trough phase alignment showed no significant firing rate differences (layers 2/3 p=0.36; layers 5/6 p=0.51).
• Modeling explains phase dependence: The model reproduced higher firing in deeper layers and showed that P1 (~50 ms) precedes spiking, implicating synaptic currents rather than spikes. Basal dendritic membrane currents (I_mem) in deeper layers differed between AC peak and trough conditions during the P1 window, with peak phase yielding more positive (less negative) I_mem and more positive LFP deflection relative to trough. This supports a mechanism of AC phase modulating electrochemical driving force and EPSCs in deep-layer pyramidal neurons, yielding opposite phase preferences for P1 and N1.
• White matter control: No modulation or phase dependence of LFPs in white matter, supporting gray matter layer specificity.

The findings demonstrate that transcranial alternating current stimulation modulates visually evoked LFPs in a layer- and phase-specific manner, with modulation localized to deeper cortical layers (4–6). This selectivity cannot be explained solely by electric field magnitude, which was stronger in superficial layers, implying that anatomical and physiological properties of deeper layers (e.g., large pyramidal cells, higher neuronal density, greater synchrony and firing rates) render them more susceptible to AC modulation. The phase dependence of P1 and N1 components indicates that AC phase alters the electrochemical driving force and postsynaptic currents in basal dendrites of deep-layer pyramidal neurons, producing more positive LFPs during depolarizing phases and stronger negative deflections during hyperpolarizing phases. The temporal dissociation of P1 (preceding spikes) and N1 (more aligned with spiking) suggests distinct mechanisms: P1 reflects subthreshold synaptic currents, while N1 relates more closely to evoked firing. MUA data confirmed increased firing in deep layers with AC but did not show clear phase dependence, likely due to the higher sensitivity of LFPs to subthreshold modulation compared to spike generation. Differences between the two monkeys (bimodal vs unimodal phase preference) may reflect differences in field strength and interactions with spontaneous low-frequency activity. Overall, results bridge single-neuron level effects of TES with laminar network dynamics and support targeting deeper layers for more effective neuromodulation.

Sensory-evoked LFPs in primate V1 are selectively and phase-dependently modulated by transcranial alternating current stimulation, predominantly in deeper layers. A cortical column model attributes this to phase-dependent changes in driving force and synaptic currents in deep-layer pyramidal neurons, explaining opposite phase preferences of P1 and N1 and the stronger deep-layer effects. These insights clarify how electric fields interact with cortical microcircuits at the laminar level and can guide neuromodulation strategies to target specific layers, potentially enhancing therapeutic efficacy. Future work should examine frequency- and dose-dependence of phase effects, incorporate spontaneous activity into models to capture inter-animal variability, validate in awake conditions, explore multi-channel TES and temporal interference for focal deep-layer targeting, and extend models from rodents to primates and humans.

• Small sample size (n=2 monkeys) and lightly anesthetized state; awake-state dynamics may differ.
• Residual stimulation artifacts cannot be entirely excluded despite ICA/DFT removal, although controls suggest minimal impact.
• Layer assignment relies on CSD profiles, electric field gradients, and anatomical references; exact boundaries carry uncertainty.
• Electric field estimates are along probe axis and assume quasi-static, uniform fields; local conductivity variations (e.g., vasculature, astrocytes) may affect measurements.
• MUA phase dependence was not detected, possibly due to limited sensitivity or strong sensory drive overshadowing subtle phase effects.
• Computational model adapted from rodent V1; species differences and lack of spontaneous activity modeling may limit capture of bimodality and inter-animal variability.
• Only one stimulation frequency (1.5 Hz) and limited current amplitudes were tested; generalization across parameters remains to be established.