Diode probes for spatiotemporal optical control of multiple neurons in freely moving animals

Abstract

Neuronal control with high temporal precision is possible with optogenetics, yet currently available methods do not enable to control independently multiple locations in the brains of freely moving animals. Here, we describe a diode-probe system that allows real-time and location-specific control of neuronal activity at multiple sites. Manipulation of neuronal activity in arbitrary spatiotemporal patterns is achieved by means of an optoelectronic array, manufactured by attaching multiple diode-fiber assemblies to high-density silicon probes or wire tetrodes and implanted into the brains of animals that are expressing light-responsive opsins. Each diode can be controlled separately, allowing localized light stimulation of neuronal activators and silencers in any temporal configuration and concurrent recording of the stimulated neurons. Because the only connections to the animals are via a highly flexible wire cable, unimpeded behavior is allowed for circuit monitoring and multisite perturbations in the intact brain. The capacity of the system to generate unique neural activity patterns facilitates multisite manipulation of neural circuits in a closed-loop manner and opens the door to addressing novel questions.

Fig. 1.

Multidiode arrays integrated with multielectrode arrays. A: structure of a single diode-fiber assembly. The diode is coupled to a 50-μm multimode optical fiber, etched to a point at the distal end. Top: magnified view of the fiber tip; calibration, 1 mm. B: light intensity in brain tissue. Light intensity as a function of axial and transverse distance from the fiber tip (left) and as a function of axial distance alone (right) shown for a blue (red) light source of 40 (300) μW. Light at intensity > 1 (7) mW/mm² (for blue/red light; see blue/red lines on left) covers a volume of ∼0.001 mm³, yielding highly localized photostimulation. C: diode-probe example. Six assemblies (4 blue and 2 red) were attached to separate shanks of a 6-shank silicon probe. Left: probe on a movable drive; arrow, metal rod used for connecting the diode ends of the fibers. Top: magnified frontal view of all 6 shanks (top middle) and an oblique view of 2 of the shanks (top right). Bottom middle: 4 shanks illuminated with blue light. Bottom right: 2 adjacent shanks illuminated with blue and red light. D: long-term optical stimulation and neuronal recordings. Top: explant of a multidiode array after 2 mo of recordings from the right dorsal hippocampus. The diode probe, equipped with 6 light-emitting diode (LED)-fiber assemblies, was left in situ, and 3 nonadjacent shanks were illuminated while photographing the skull from the inside out. Bottom: fluorescence image of a coronal section through the hippocampus of the same animal [tdTomato fluorescence, viewed via a rhodamin (RH) filter] and Nissl staining of an adjacent 80-μm section (inset) showing the track of a probe shank ending in CA3c pyramidal layer (arrow; calibration, 500 μm).

Fig. 2.

Diode-fiber assembly construction and driving. A: example LEDs. Top: wired LED (APA1606QBC/G; Kingbright). Bottom: nonwired diodes (LB P4SG, OSRAM; APA1606PBC/Z, Kingbright). Arrows show light emitters. B: alignment and coupling of diode and fiber. Left: 3 micromanipulators are used: I holds the wired diode; II holds the etched fiber; and III holds a light sensor (S130A; Thorlabs) set to the 100-μW range. Two pieces of black paper with pinholes are slipped over the fiber and cover the sensor aperture and diode such that the sensor reading is 0 when the diode is not lit. Right: following optimization of the relative location of diode and fiber, a drop of UV-curable glue is placed on the interface and cured using a UV light gun (arrow). C, left: tetrode-based diode probe. Each tetrode is coupled to a diode-coupled fiber (arrow) and mounted on a separate movable drive. Right: tetrode attached to a diode-coupled fiber. The tip of a 100-μm core optical fiber (AFS105/125; Thorlabs) was etched to a cone, and the fiber was coupled to a blue LED. The assembly was then attached, using UV-curable glue, to a tetrode such that the fiber tip was ∼100 μm above the recording end of the wires. D: multichannel precision current source. Top: schematic of a single channel. The input stage consists of an NPN transistor in an emitter-follower configuration, a digital buffer, a source selector, and an op-amp follower to ensure high input impedance; this enables 3 different operating modes: manual (push button), digital input [transistor-transistor logic (TTL)], and analog input. The current control stage consists of a differential amplifier and a PNP transistor with an op-amp negative-feedback; this yields linear compliance for up to 10-V voltage (V) drop at the load while enabling programming relative to ground. Bottom: front panel of the rack-mounted device, showing for each channel (from top to bottom) the push buttons, potentiometers, analog input selectors, and analog/digital source selectors.

Fig. 3.

Estimation of light intensity in brain tissue. A: theoretical calculation for an idealized cleaved fiber. Top left: scheme of the geometric arrangement; gray, cleaved fiber; yellow, idealized light cone [0.22-numerical aperture (NA) fiber, resulting in divergence angle (α) = 9.3° in brain tissue]. Bottom left: light attenuation through brain as a function of wavelength (470, 589, and 639 nm) and the thickness of brain tissue between the fiber tip (inserted into a piece of brain) and the measurement sensor (placed just below the brain). Middle: geometric dispersion of light (the inverse of the cone cross-section) as a function of the distance (d) and the radius (r) from the fiber tip (top) and along the optical axis (fiber center; bottom). Right: light-intensity map (source: blue light, 40 μW at tip of fiber) in the same coordinates as the geometric dispersion map; vertical dashed red line shows the distance for which intensity falls below threshold (1 mW/mm²). B: empirical calculation for a pointed fiber. Top left: fluorescence image of the cone of blue light emitted by a 0.22-NA fiber etched to a point (viewed via an FITC filter). The measured divergence in PBS (0.1 M) is α = 16.3°, resulting in an effective NA (NAeff) of ∼0.37. Middle: geometric dispersion map, estimated from the fluorescence image. Note that the distribution of light is not uniform at locations equidistant from the source and that the distance attenuation is stronger than for a cleaved fiber. Right: blue light intensity map for this fiber. Bottom left: number of neurons in illuminated brain tissue as a function of blue light intensity at the fiber tip. The estimation, shown for 2 different intensity thresholds for ChR2 activation, assumes neuronal density of 20,000 cells/mm³ and ignores shadowing by the probe shank.

Fig. 4.

Diode probe induction of neuronal activity in the freely moving wild-type rat. A: example of the effect of low intensity light on the activity of a single unit in the rat hippocampus (dorsal CA1). Raster plot (top) and peristimulus time histogram (PSTH; bottom) of spiking activity during 100 light stimuli, each 10 ms long (blue bar; 4 μW at fiber tip, corresponding to ∼0.12 mW/mm² at the shank center). Inset shows autocorrelation histogram. Right: wideband (1–5,000 Hz) spike waveforms. B: effect of localized light on spiking rate. Left: firing rates during light vs. spontaneous activity. Red circle marks unit shown in A. Middle: light modulation index (difference between firing rates during photostimulation and baseline divided by the sum). The distribution is bimodal. Right: fraction of light-modulated units is above chance level (dashed red line) and does not differ between CA1, dentate gyrus, and CA3 regions of the hippocampus. C: effect of light on spiking probability. Left: spiking probability during light vs. during spontaneous activity. Conventions are the same as in B. Right: number of spikes per light stimulus. Stimuli were rectangular light pulses, 2–50 ms long; units are partitioned according to the median pulse duration used (30 ms). Dashed red line shows median number of spikes per stimulus (1.92). D: effect of light on extracellular waveform. Waveform consistency was defined as the Pearson correlation between the mean waveforms during light and during spontaneous activity.

Fig. 5.

Spatiotemporal extent of diode-probe light stimulation. A: temporal precision of light-induced spiking. Left: spiking center-of-mass relative to stimulus onset (across-trial average of trial-specific centers-of-mass). Dashed red line, median. Middle: spiking latency (time to 1st significantly high PSTH bin). Right: spiking precision (across-trial SD of the trial-specific centers-of-mass). B: example of spatially localized light-induced spiking. Wideband (1–5,000 Hz) traces during light stimulation. Red/blue/green traces show recordings from channels 11 through 40 (shanks 2–4) of a 6-shank silicon probe implanted in the dorsal CA3 pyramidal layer of the rat hippocampus. Probe geometry is shown at right; blue light was applied locally at shank 3 (blue bars; 20-ms pulses, 5 Hz, 0.15 mW/mm² at shank center). Top right: expanded view during spontaneous activity, showing spiking on 3 adjacent shanks. Bottom right: expanded view during shank-specific light stimulation, showing light-driven spiking specifically on the illuminated shank. Note lack of electromagnetic or light artifacts. C: 2 light-responsive single units recorded on shank 3. Conventions are the same as in Fig. 4A. D: spatial extent of the localized light stimulus (CA1 pyramidal layer; 10 shanks in 3 rats). In contrast to the illuminated shank, the fraction of light-modulated units on other shanks does not differ significantly from chance level (exact binomial test, P > 0.05; dashed red line corresponds to 0.05, the fraction of units expected to be defined as modulated assuming that the null hypothesis of “no light modulation” is correct). Error bars, SE. E: cell type-specific stimulation in freely moving mice (CA1). Bottom: each row corresponds to the 4-ms binned PSTH of 1 unit, averaged over 60 light pulses, and scaled to 0–1. Top: before scaling, all PSTHs were averaged; horizontal calibration: 50 ms; vertical calibration: pyramidal cells, 1 spike/s; interneurons: 20 spikes/s. During stimulation (∼0.9 mW/mm²), putative pyramidal cells are activated in CaMKII::ChR2 and silenced in PV::ChR2 mice.

Fig. 6.

Concurrent silencing and activation of the same neurons in the freely moving rat. A: multiple-unit example. Wideband (1–5,000 Hz) traces are from shank 1 of a 6-shank multicolor diode probe implanted in layers 4 and 5 of the somatosensory cortex of a wild-type rat expressing CAG-ChR2 and CaMKII-Halo locally. Bottom traces show the occurrence of blue (10 ms, 0.34 mW/mm² on the center of the illuminated shank) and red (200 ms, 9.2 mW/mm²) light pulses, delivered on shanks 1 and 2, respectively. Each row in the raster plot represents the spiking activity of 1 unit isolated from shank 1. Multiple units are activated (silenced) during the blue (red) light pulses. B: example statistics for 1 unit. Each PSTH-raster pair shows spiking activity of the unit marked in blue in A during blue light pulses (left), red light pulses (right), and concurrent red and blue light (middle). The unit is activated by blue light and silenced by red light; induced spiking is delayed during concurrent activation and silencing. C: example of unit silencing at multiple intensities. Effect of 200-ms pulses (1 Hz for 10 s) at varying intensities (2.1–16.7 mW/mm², measured at the shank center). D, top left: fraction of light-modulated units recorded on the illuminated shanks (dashed red line, P = 0.05). Top right: light-modulation index for the same units: 0, no effect; −1, complete silencing. Error bands, SE. Bottom: the probability of observing suppressed units and the light-modulation index depend on light intensity and on the transverse distance from the illuminated shank.

Fig. 7.

Spike-triggered closed-loop spatiotemporal photostimulation in the freely moving animal. A: the plot shows a single 50-ms sweep through the recording. A spike was detected (arrow) and isolated in real-time from the wideband signal (1–5,000 Hz) recorded from shank 4 of a 6-shank silicon probe implanted in dorsal CA1 of a wild-type rat injected with CAG-ChR2. One millisecond later, an arbitrary sequence of 3 blue light pulses (at the times indicated by the colored bars below the wideband traces) was given on shanks 3, 5, and 6. B: this resulted in consistent unit spiking, as evident by the multiunit PSTH (the spiking of each well-isolated unit was binned at 1 ms and normalized to the 0–1 range; top; same time scale as in A) and the cross-correlation histograms (CCHs) between the triggering pyramidal unit and 3 units on the other shanks (3 bottom left panels). CCHs were estimated in 0.5-ms bins and scaled to multiples of chance coincidence counts. Calibration for CCHs: horizontal, 50 ms; vertical, 10. Red-colored bins indicate significantly high counts (convolution method, window size, 10.5 ms). C: the same display as B but during a duration-matched (16 min) baseline period. Note that the precise temporal correlations in B were generated by the photostimuli, as CCHs between the same units were nearly flat (with an exception of a 1-ms lag between the units recorded on shanks 4 and 3) before the stimulation.

Fig. 8.

Closed-loop spatiotemporal photostimulation by diode probes on the linear track (rat CA3c). A: on alternating left-to-right runs, 2 shanks were illuminated sequentially for 1 s, starting at a predetermined position (magenta lines in all panels). For each shank (4, black; 1, red), traces show light intensity (top) and wideband signal on 1 recording site (1–5,000 Hz; bottom); tick marks below indicate spike times for all simultaneously recorded units (shank 4, 6 units; shank 1, 10 units). Ranges: speed: 0–51 cm/s; position: 0–150 cm; light intensity, 0–0.1 mW/mm². B: examples of light-induced position-related spiking (left: shank 4, red unit from A; peak firing rate, 23 spikes/s; right: shank 1, black unit, peak firing rate, 28 spikes/s). Here and in C, horizontal bars indicate the average position of the animal during photostimuli. Note consistent position-related firing in both cases and induction of a novel place field for the shank 1 unit (arrow). C: interaction between light-driven spiking and native θ-oscillations. Top: autocorrelation histograms for the pyramidal cells in B; side lobes indicate θ (∼8 Hz) modulation. Bottom: θ-phase (0: peak) of spikes shifts as a function of position during stimulation trials, illustrating an interaction between light-induced and network-controlled effects.

Fig. 9.

Spatiotemporal photostimulation evoked by brain patterns in the behaving rat. A: as the animal (wild-type rat injected with CAG-ChR2 in the dentate gyrus) was running back and forth on the linear track, θ-phase was determined in real time (dashed black lines here and in B: 0-crossing events). Starting from a predetermined position (magenta), on alternating right-to-left laps, a train of 6 brief light pulses was delivered at 6 separate sites (shifted horizontal bars, bottom) during θ-troughs. Each row in the raster plot represents the spiking activity of 1 unit, colored according to the shank by which it was recorded. B: closed-loop stimulation resulted in precise spatiotemporal spiking within each individual θ-cycle. Each row corresponds to the 1-ms binned PSTH of 1 unit recorded by 1 of the 6 shanks, averaged over 104 θ-cycles, smoothed (Gaussian, SD = 5 ms), and scaled to 0–1. Inset: relative activity times of units recorded at the different shanks (medians). C: on a broader space and time scale, cells also exhibited position-related spiking. Each row corresponds to the scaled difference between stimulation/control position-related firing of 1 unit.