Non-genetic neuromodulation with graphene optoelectronic actuators for disease models, stem cell maturation, and biohybrid robotics

Abstract

Light can serve as a tunable trigger for neurobioengineering technologies, enabling probing, control, and enhancement of brain function with unmatched spatiotemporal precision. Yet, these technologies often require genetic or structural alterations of neurons, disrupting their natural activity. Here, we introduce the Graphene-Mediated Optical Stimulation (GraMOS) platform, which leverages graphene's optoelectronic properties and its ability to efficiently convert light into electricity. Using GraMOS in longitudinal studies, we found that repeated optical stimulation enhances the maturation of hiPSC-derived neurons and brain organoids, underscoring GraMOS's potential for regenerative medicine and neurodevelopmental studies. To explore its potential for disease modeling, we applied short-term GraMOS to Alzheimer's stem cell models, uncovering disease-associated alterations in neuronal activity. Finally, we demonstrated a proof-of-concept for neuroengineering applications by directing robotic movements with GraMOS-triggered signals from graphene-interfaced brain organoids. By enabling precise, non-invasive neural control across timescales from milliseconds to months, GraMOS opens new avenues in neurodevelopment, disease treatment, and robotics.

Fig. 1. Mechanism and neuronal biocompatibility of GraMOS.

a Mechanism of action of GraMOS. Left: Dirac cones for multilayer graphene materials exhibiting a small bandgap. K = momentum; E = energy; E_f  = Fermi energy (red dotted line). Light generates high-energy “hot” electrons (blue spheres) which, in turn, can produce additional electrons via impact ionization, leading to carrier multiplication. Holes are depicted as red spheres. Center: cartoon of a light-activated neuron on a graphene interface. Graphene lattice (one hexagonal unit cell is ~0.0524 nm²) and neuron (10,000 nm) are shown at different scales for illustration. Right: zoom-in of neuronal membrane at rest and under light illumination. Photogenerated electrons from the electron cloud outside the graphene lattice can dynamically induce capacitive effects at the cell-electrolyte-graphene interface, leading to membrane depolarization. b Representative SEM image of rGO flakes on the ITO substrate selected from 8 images of 4 samples. c Histogram of rGO flake lateral dimensions (n = 1000). d Histogram of rGO flake areas (n = 250). e Representative image of rGO flakes spray-coated on glass coverslips, producing G-substrates with ~90%, 70%, and 50% optical transmittance (left to right). f Representative SEM image of hiPSC-derived neurons on G-coverslips (false-colored blue), selected from 97 images of 15 samples. g Representative fluorescent images showing hiPSC-derived neurons on control coverslips (left) and G-coverslips (right). Neurons were labeled with EthD-1 (dead cells, red fluorescence) and calcein-AM (live cells, green fluorescence). h Summary of viability experiments using hiPSC-derived neurons on control coverslips and G-coverslips (three replicates, five random fields of view each). Data are presented as violin plots: center lines = median, red squares = mean, and individual data points are shown as circles (**P = 0.00924, one-way ANOVA). i Representative action potentials from hiPSC-derived neurons on control coverslips and G-coverslips. j Selected electrophysiological properties of hiPSC-derived neurons on control (n = 9) and G-coated (n = 12) coverslips. AHP = Afterhyperpolarization. Data are presented as box plots: center lines = median, red squares = mean, box = upper/lower quartiles, whiskers = 5th–95th percentiles, and individual points are shown as circles.

Fig. 2. Photophysical characterization of G-interfaces.

a Visible-range absorption spectra of rGO, showing broadband absorption with decreasing efficiency at longer wavelengths. b Photocurrents acquired from G-coverslips (the optical transmittance (OT) of 90% (squares), 70% (circles), and 50% (diamonds)) when illuminated by 480 nm light (2.1 mW/mm²), 535 nm light (3.7 mW/mm²), and 575 nm light (8.3 mW/mm²). The holding potential = 0 mV. Insert: an example of a photocurrent trace (bars: 20 pA/500 ms). c Photocurrents generated under different light intensities. Measurements from the same G-coverslips are connected by lines. d Photocurrents acquired at 0 mV and +5 mV holding potentials. The 0-mV and +5 mV values from the same G-coverslips are connected by lines. e Temperature measurements on the surface of G-coverslips as a function of photocurrent amplitudes (n = 50). The average surface temperature of G-coverslips when illuminated by light was 23.6 ± 0.03 °C vs 23.7 ± 0.04 °C in the dark as detected using a non-contact digital IR thermometer Mestek 800 C. f Local temperature measurements near G-coverslips during 15-minute light exposure (535 nm, 3.7 mW/mm²; n = 3). The beige box indicates the light exposure period. Data are presented as violin plots: center line = median, red square = mean, and individual data points are shown as circles. g pH values of an electrolyte solution containing submerged 70-OT G-coverslips during 15 min light exposure (535 nm, 3.7 mW/mm²; n = 4). The beige box indicates the light exposure period. Data are presented as violin plots: center line = median, red square = mean, and individual data points are shown as circles. h CM-H₂DCFD-based quantification of reactive oxygen species (ROS) levels in control and G-interfaced 2-month-old WT83 brain cortical organoids in the dark and after light stimulation (2 Hz, 1 h, 1.9 mW/mm²) for 7 consecutive days. n = 100 cells per condition. Data are presented as violin plots: center line = median, red square = mean, and individual data points are shown as circles (**P = 0.0091, ****P = 4.3648 × 10^-5; Two-way ANOVA).

Fig. 3. GraMOS-enabled activation of neurons on G-interfaces.

a Experimental scheme (left) and a representative brightfield image of a neuron on a G-substrate during patch-clamp experiments (right) (n > 50 experiments). b Currents triggered by 5-s light illumination (452 nm; 1.9 mW/mm², 4.1 mW/mm², and 8 mW/mm² for 1x, 2x, and 4x light intensities, respectively) in voltage-clamped G-interfaced neurons (V_h = -70 mV). The insert on the left (yellow box) shows a zoom-in of the raising phase of light-triggered currents in neurons. c Light-triggered action potentials in G-interfaced neurons (452 nm; 4.1 mW/mm²; 1-s duration for the left trace, and 2-ms duration for the center and right traces). d Experimental scheme for GraMOS-empowered all-optical calcium imaging under wide-field illumination. e Representative images of Fluo-4-labeled neurons on G-substrates (n > 50 experiments). f Light parameters for GraMOS-based assays minimize optical crosstalk: wavelengths of stimulation light (L_S) lie outside the fluorophore absorption range, while the intensity of fluorophore excitation light (Lₑ) remains below the GraMOS activation threshold. g Representative calcium transients from several regions of interest in the same field of view of G-interfaced cells (neurons – N; astrocyte – A) and Fluo-4 signal outside cells (background – the bottom trace) in response to pulsed wide-field light illumination (638 nm; 3.9 mW/mm², 5 ms duration). h Representative bursting calcium transients triggered by prolonged wide-field light illumination (638 nm; 3.9 mW/mm²; 2-s duration) in G-interfaced primary cortical neurons.

Fig. 4. GraMOS enables functional phenotyping and promotes maturation of G-interfaced hiPSC-derived neurons.

a Scheme for GraMOS-empowered all-optical studies using single-cell single-pulse light. b Videoframes of G-interfaced Fluo-4-labeled neurons at rest and following activation of a central neuron by a single light pulse (561 nm, 5.2 mW/mm²). c Calcium transients across WTAD and M233L networks triggered by a single light pulse. d Ratio of GraMOS-activated neurons in WTAD and M233L models (n = 9 and 18 G-coverslips, respectively; **P = 0.0098, one-way ANOVA). e Calcium transient amplitudes in G-interfaced WTAD and M233L neurons (n = 251 and 678, respectively; ***P = 0.0004, one-way ANOVA) activated by a single light pulse. Box plots in (d) and (e): center line = median, red squares = mean, box = upper/lower quartiles, whiskers = 5th–95th percentiles, and individual points = circles. f Spearman correlation matrix for GraMOS-initiated neuronal activity in G-interfaced WTAD and M233L neurons. g Spatial maps showing the number of spikes in GraMOS-activated WTAD and M233L neuronal networks. h Spatio-temporal maps depicting the activation times (in seconds) after a light pulse in WTAD and M233L neurons (****P = 1.0063 × 10^-235, two-sided Wilcoxon rank sum). i Histogram of activation times for WTAD and M233L models. j Scheme for GraMOS-driven maturation of G-interfaced hiPSC-neurons. k Ratio of GraMOS-activated WTAD neurons at different timepoints during light training (n = 47, 16, 9, 20, 15 for Weeks 0, 2, 3, 4, 6) (*P = 0.0422 (Weeks 0 vs 3); ***P = 0.0006 (Weeks 2 vs 4); ****P = 8.119 × 10⁻⁷ (Weeks 0 vs 4); P for Week 6 vs Week 0–6.6059 × 10⁻¹⁸, Week 2–8.2329 × 10⁻⁷, Week 3–1.3661 × 10⁻⁷, Week 4–0.0011; Two-way ANOVA). l Amplitudes of GraMOS-triggered calcium transients in WTAD neurons at different timepoints during light training (n = 542, 202, 245, 364, 675 neurons for Weeks 0, 2, 3, 4, 6) (****P = 7.0376 × 10⁻⁷, (Weeks 0 vs 3); ***P = 3.1209 × 10⁻⁴ (Weeks 0 vs 2); ****P = 5.154 × 10⁻⁵ (Weeks 0 vs 4); P = 2.5673 × 10⁻⁸ (Weeks 0 vs 6); Two-way ANOVA). Box plots in (k) and (l): center line = median, red squares = mean, box = upper/lower quartiles, whiskers = 5th–95th percentiles, and individual points = circles. m Spearman correlation matrix for GraMOS-initiated neuronal activity in WTAD models at different timepoints during light training. A.U. arbitrary units.

Fig. 5. GraMOS in hiPSC-derived brain cortical organoids.

a Schematic of a hiPSC-derived brain organoid interfaced with internal G-flakes. b Bright-field image of brain organoids with G-flakes inside. c Control and G-interfaced brain organoids labeled with EthD-1 (dead cells, red fluorescence) and calcein-AM (live cells, green fluorescence). d Summary of cell viability experiments performed in control and G-interfaced brain organoids (n = 5 per group). Data are presented as violin plots: center line = median, red square = mean, and individual data points are shown as symbols. Data were analyzed using one-way ANOVA; no statistically significant differences were observed. e Representative immunostaining images of 10-week-old control and G-interfaced brain organoids labeled with DAPI and antibodies for characterization of neuronal morphology and composition of neuronal networks. f Population analysis of specific markers in control and G-interfaced brain organoids. Data are presented as violin plots: center line = median, red square = mean, and individual data points are shown as symbols (n = 3 per group; *P = 0.0416, one-way ANOVA). g Experimental scheme of GraMOS-empowered all-optical calcium imaging on a confocal microscope using pulsed spatially-limited light illumination (top and bottom) with a representative confocal plane inside a Fluo-4-labeled G-interfaced brain organoid (center). h Calcium transients triggered by spatially confined 561 nm light pulses (5.2 mW/mm², 1 ms duration), with excitation propagating across the neuronal network and appearing in selected neurons within the field of view. The white dashed line shows the timing of the pulsed light signal. A.U. - arbitrary units.

Fig. 6. GraMOS-evoked electrical activity in G-interfaced hiPSC-derived brain cortical organoids on microelectrode arrays.

a Schematic of GraMOS-empowered MEA experiments from G-interfaced brain cortical organoids (BCOs) (left), and representative examples of spontaneous and GraMOS-evoked electrical activity (right). b GraMOS during external (c−g) and internal (h–l) interfacing of rGO flakes with brain organoids. c Raster plots showing GraMOS-evoked electrical activity in brain organoids with external G-interfaces (top and central panels) and control brain organoids (bottom panel). A representative zoom-in raster plot shows electrical activity after a single light pulse (c, center). d Normalized mean firing rates (MFR) in externally G-interfaced and control organoids before, during, and after GraMOS (approximately 10 G-interfaced and 10 control organoids on 64 electrodes each). Data are shown as violin plots: center lines = median, red squares = mean, and individual data points = circles (P = 1.1052 × 10^-85 for “before” vs “GraMOS”; P = 5.3556 × 10^-82 for “GraMOS” vs “after”; two-way ANOVA). e Lag times of GraMOS-evoked neuronal spikes in externally G-interfaced brain organoids. f MFR fold increase per active electrode in externally G-interfaced and control organoids before, during, and after GraMOS (n > 100 per group). g Population histogram of MFR fold increase in externally G-interfaced brain organoids (n > 100 per group). h Raster plots showing GraMOS-evoked electrical activity in brain organoids with internal G-interfaces (top and central panels) and control brain organoids (bottom panel). i Normalized MFR in internally G-interfaced and control organoids before, during, and after GraMOS (approximately 50 G-interfaced organoids on 253, 153, 135 electrodes, respectively; approximately 20 control organoids on 54, 54, 34 electrodes). Data are shown as violin plots: center line = median, red squares = mean, and individual data points = circles (P = 1.0183 × 10^-7 for “before” vs “GraMOS”; P = 3.2932 × 10^-5 for “GraMOS” vs “after”; two-way ANOVA). j Lag times of GraMOS-evoked neuronal spikes in internally G-interfaced brain organoids. k MFR fold increase per active electrode in internally G-interfaced and control brain organoids before, during, and after GraMOS (n > 100 per group). l Population histogram of MFR fold increase in internally G-interfaced brain organoids.

Fig. 7. GraMOS-driven maturation of G-interfaced brain cortical organoids via long-term optical stimulation.

a Schematic for GraMOS-enabled activity-dependent enhancement of maturation of G-interfaced brain cortical organoids. b Mean firing rates (left) and bursting rates (right) per electrode in G-interfaced (blue bars) and control (gray bars) brain organoids at different time points after long-term GraMOS (50 G-interfaced and 20 control organoids). Data presented as box plots: center line = median, red squares = mean, box = upper/lower quartiles, whiskers = 5th–95th percentiles, and individual points = circles. c Representative STTC matrices for MEA recordings from G-interfaced and control brain organoids on Day 1 and Day 21 after light training. d Volcano plot of DEGs between GraMOS-trained G-interfaced and control brain organoids. Cutoffs are at +/− 0.6 log₂ fold change and P-values < 0.05. Labeled DEGs are linked to brain developmental processes. P-values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR) correction method. e Top 30 GO analysis results by FDR. f Top 30 GO analysis Neuronal metrics by FDR. Enrichment score was calculated as the -log₁₀(FDR).