Source-linked AI summary
A Reconfigurable Nanophotonics Platform for Sub-Millisecond, Deep Brain Neural Stimulation
Aseema Mohanty, Qian Li, Mohammad Amin Tadayon, Gaurang Bhatt, Euijae Shim, Xingchen Ji, Jaime Cardenas, Steven A. Miller, Adam Kepecs, Michal Lipson
TL;DR
The study combines a reconfigurable nanophotonic probe with viral targeting and electrophysiological analysis for deep-brain optical stimulation. In vivo experiments reached approximately 1300 µm, while waveform-based analysis identified directly light-activated neurons and addressed light-induced recording artifacts.
Problem
Reliable identification of optically activated neurons requires distinguishing light-induced neural spikes from light-induced artifacts during electrophysiological recording.
Method
The paper uses a silicon nitride Mach-Zehnder switching network with thermally controlled phase, an 8-beam probe, viral targeting, and waveform-based spike analysis.
Results
Approximately 1300 µm depth was reached in vivo, and light-induced and spontaneous spikes in a ChETA-expressing Gad2 interneuron showed waveform similarity r = 0.95.
Takeaways & Limitations
The platform supports patterned optical stimulation across visual-cortex layers and hippocampus alongside multi-channel neural recording.
Takeaways & Limitations
Light-induced artifacts can occur when exposed tungsten electrode tips are directly illuminated, requiring spike-sorting discrimination.
Abstract
from arXiv · showhide
Nanophotonics provides the ability to rapidly and precisely reconfigure light beams on a compact platform. Infrared nanophotonic devices are widely used in data communications to overcome traditional bandwidth limitations of electrical interconnects. Nanophotonic devices also hold promise for use in biological applications that require visible light, but this has remained technically elusive due to the challenges of reconfiguring and guiding light at these smaller dimensions. In neuroscience, for example, there is a need for implantable optical devices to optogenetically stimulate neurons across deep brain regions with the speed and precision matching state-of-the-art recording probes. Here we demonstrate the first platform for reconfigurable nanophotonic devices in the visible wavelength range and show its application in vivo in the brain. We demonstrate an implantable probe endowed with the ability to rapidly switch and route multiple optical beams using a nanoscale switching network. Each switch consists of a silicon nitride waveguide structure that can be reconfigured by electrically tuning the phase of light and is designed for robustness to fabrication variation, enabling scalable devices. By implanting our probe in mouse visual cortex, we demonstrate in vivo the ability to stimulate identified sets of neurons across layers to produce multi-neuron spike patterns and record them simultaneously with sub-millisecond temporal precision. This nanophotonic platform can be scaled up and integrated with high-density neural recording technologies, opening the door to implantable probe technologies that are able to simultaneously record and stimulate the activity of large neural populations at distant regions of the brain with sub-millisecond precision. We expect this platform will enable researchers to gain a deeper understanding into the spatio-temporal precision of the neural code.
METHODS
The probe uses a silicon nitride nanophotonic switching network designed for fabrication, wavelength, and polarization robustness. Thermally tuned Mach-Zehnder interferometers provide phase control for routing light.
- METHODS: Seven thermally tuned Mach-Zehnder interferometer switches form the 1x8 nanophotonic switching network.Each switch uses a 300 µm platinum microheater for phase control.
- METHODS: EigenMode Expansion simulations optimize the Mach-Zehnder switch for fabrication, wavelength, and polarization insensitivity.
- METHODS: Adult Gad2-IRES-Cre knock-in mice were used under an approved institutional animal-care protocol.The mice were maintained on a reverse 12-hour light/dark cycle with food and water available ad libitum.
Virus injection
The study used Cre-dependent viral delivery to express a ChR2 variant in targeted visual-cortex and hippocampal regions of Gad2-IRES-Cre mice.
- Virus injection: A Cre-dependent AAV9-EF1a-DIO-ChETA-eYFP vector was injected into visual cortex and hippocampus at four depths.The injections used 0.25 µl per site at depths of 0.2, 0.4, 0.8, and 1.3 mm.
- Virus injection: The viral vector delivered ChETA-eYFP, a ChR2 variant designed for high-frequency stimulation.
- Virus injection: Gad2-IRES-Cre knock-in mice enabled Cre-dependent targeting of the injected construct.
In vivo experiments
Acute experiments inserted an 8-beam probe across visual-cortex layers 2–6 to stimulate infected areas while recording extracellular spikes.
- In vivo experiments: The probe reached activation sites as deep as approximately 1300 µm from the brain surface.It was inserted through a 2 x 2 mm^2 craniotomy to target viral-infected areas across visual-cortex layers 2–6.
- In vivo experiments: Three tungsten stereotrodes recorded extracellular spike signals simultaneously at 30 kHz.Signals were preamplified 20X and band-pass filtered from 600–6 KHz.
- In vivo experiments: Patterned stimulation used 1 ms laser pulses delivered through the nanophotonic probe.
Electrophysiological data analysis
Neural data were spike-sorted and analyzed using waveform similarity, spike probability, latency, and jitter to identify directly light-activated neurons.
- Electrophysiological data analysis: Spikes were manually sorted offline into clusters using peak amplitude and waveform energy.The analysis used MClust software and custom Matlab tools.
- Electrophysiological data analysis: Directly light-activated neurons showed high waveform similarity, high spike probability, short latency, and low jitter.Waveform similarity was defined as the correlation between averaged light-induced and spontaneous spike waveforms.
- Electrophysiological data analysis: Histology verified the electrode insertion track after acute electrophysiological recordings.Brains were sectioned into 90-micron coronal slices including V1 and hippocampus.
FIGURES
The figures show a visible-wavelength implantable probe that routes light through a switching network to independently controllable grating emitters, demonstrating configurable optical patterns and multi-neuron stimulation.
- Figure 1: The probe routes input light through a switching network to grating emitters that direct independently switchable beams toward neurons.The architecture places active switching photonics outside the brain and uses long routing waveguides to reach the emitters.
- Figure 1: The visible-wavelength switch uses electrical power to tune light continuously between two output ports.The inset depicts a 1x2 switch whose output allocation changes as applied power increases.
- Figure 2: The switch operates at 473 nm with 30 mW full-conversion power, a 50:1 ON:OFF contrast ratio, and a 20 μs switching time.The figure reports 17 dB extinction and derives switching time by fitting an exponential response.
- Figure 3: Different switch configurations generate distinct spatial patterns from highly collimated beams emitted by the probe.The patterns are visualized in fluorescent dye, with full and half-power settings represented in the configurations.
- Figure 4: The 8-beam probe drives three identified interneurons across cortical layers with regular, random, and 200 Hz stimulation patterns while recording spikes.The figure pairs optical stimulation with nearby electrical recordings and reports latency and jitter across repeated trials.
SUPPLEMENTARY FIGURES
The supplementary figures explain the thermo-optic switching mechanism, fabrication and in vivo setup, electrode arrangement, neuron identification, and artifact discrimination.
- Supp 1 Thermo-optic Switch Design: The thermo-optic switch uses a silicon nitride Mach–Zehnder interferometer with a platinum heater to change phase and route light between outputs.Light is split into two arms, one arm is heated, and recombination changes the output distribution.
- Supp 3 Experimental setup for In-vivo demonstration: The in vivo setup synchronizes seven switch-control voltage patterns and laser pulses while acquiring eight-channel neural activity.A pulse generator provides timed control and a synchronized trigger for neural recording.
- Supp 4 Electrode arrangement and switching configuration: Three tungsten stereotrodes are aligned with beams 1–3, and histology documents probe insertion across visual-cortex layers 2–6.The arrangement supports simultaneous recording near selected optical emitters during the in vivo demonstration.
- Supp 5 Identification of ChETA-expressing Gad2 interneuron: Light tagging identifies ChETA-expressing Gad2 interneurons through spike sorting, waveform similarity, and light-induced spike probability.The supplementary analysis distinguishes light-responsive Gad2 neurons from non-Gad2 putative pyramidal neurons.
- Supp 6 In-vivo characterization of single switch performance: The supplementary single-switch experiment examines optogenetic control of neurons near two beams separated by 875 μm.The setup places tungsten electrodes next to beam 1 and beam 2 to assess switching-dependent activation.
- Supp 7 Characterization of light-induced artifact: Light-induced artifacts are distinguished from neural spikes using waveform and temporal-response differences during optical stimulation.The comparison includes latency and jitter for spikes and artifacts, with electrodes positioned near the emitter.