Three-dimensional scanless holographic optogenetics with temporal focusing (3D-SHOT)

Abstract

Optical methods capable of manipulating neural activity with cellular resolution and millisecond precision in three dimensions will accelerate the pace of neuroscience research. Existing approaches for targeting individual neurons, however, fall short of these requirements. Here we present a new multiphoton photo-excitation method, termed three-dimensional scanless holographic optogenetics with temporal focusing (3D-SHOT), which allows precise, simultaneous photo-activation of arbitrary sets of neurons anywhere within the addressable volume of a microscope. This technique uses point-cloud holography to place multiple copies of a temporally focused disc matching the dimensions of a neuron's cell body. Experiments in cultured cells, brain slices, and in living mice demonstrate single-neuron spatial resolution even when optically targeting randomly distributed groups of neurons in 3D. This approach opens new avenues for mapping and manipulating neural circuits, allowing a real-time, cellular resolution interface to the brain.

Fig. 1

Simplified experimental set-up for 3D-SHOT. a A spherical phase mask is applied to a Gaussian femtosecond laser pulse incident on a blazed diffraction grating. After the grating, the direction of propagation of the first-order beam is wavelength-specific, decomposing the pulse in the temporal domain. Temporal focusing and the associated nonlinear response thus only happens across a disc-shaped area at depth planes corresponding to images of the diffraction grating (dashed, green). We call this 3D intensity distribution a Custom Temporally Focused Pattern (CTFP). b We then copied this temporally focused disc pattern to different areas in 3D by computer generated holography (CGH). To do so, a spatial light modulator (SLM) in Fourier space (dashed, red) generates a point-cloud hologram that replicates the CTFP at each target point in the volume image. The resulting field is demagnified again before impinging on the sample to create custom 3D positioning of the temporally focused disk, each targeted at a particular neuron

Fig. 2

Optical characterization of the spatial resolution of CGH vs. 3D-SHOT. a We used a fluorescent calibration slide and an inverted microscope to compare two-photon absorption patterns in 3D. b For conventional holography, we consider a 10 µm diameter disk target, and show (from top to bottom) example projection views of two-photon absorption in the (x,y), (z,y), and (z,x) planes. With 3D-SHOT, the CTFP was adjusted to a 10 µm diameter target and the same projection views were recorded c experimentally, and d rendered with simulation. e Simulation additionally provides space-time projections (Supplementary Notes 1−3). The primary focus (green arrow) at z = 0 displays the characteristic depth-sectioned properties of temporal focusing, with a perceived femtosecond pulse line object rapidly swept along the x-axis. The secondary focus (pink arrow), a static line along the x-axis in y = 0, is a geometric projection of the spherical phase pattern induced by lens L _c. The pulse duration at the secondary focus is stretched in time (~1000 fs) which minimizes nonlinear response

Fig. 3

3D-SHOT generates axially confined photo-activation. a A photostimulation pattern generated with CGH was mechanically stepped along the optical z-axis and passed through a cell expressing opsin. Photocurrents were recorded in the whole-cell voltage clamp configuration. b−f The response profile for CGH with a 10 µm disk target and different power levels on CHO cells for CGH b or 3D-SHOT f (n = 12 cells, data points are a representative example cell). d The FWHM of the characteristic response profile for both methods at various power levels on CHO cells (n = 12 cells, data represent mean ± s.e.m.). e As in a, but in mouse brain slices. c, g as in b, f, but in mouse brain slices (n = 11 neuron, data points are a representative example neuron). h The FWHM of the characteristic response profile for both methods at various power levels in mouse brain slices (n = 11 neurons, data represent mean ± s.e.m.)

Fig. 4

3D-SHOT provides high spatial resolution photo-activation of neurons in vitro and in vivo. a Spatial profile of two-photon evoked spiking of a L2/3 pyramidal neuron in a mouse brain slice in the radial dimension. (CGH: black, n = 12 neurons; 3D-SHOT: red, n = 14 neurons; p = 0.56 Mann−Whitney U-test. Data represent mean ± s.e.m). b As in a) but along the axial dimension (CGH n = 11 neurons, 3D-SHOT n = 5 neurons, p = 0.006). c Quantification of the FWHM comparing CGH and 3D-SHOT (data represent mean and s.e.m.). d Full volumetric assessment of photostimulation resolution using 3D-SHOT. Points throughout the volume were tested, but only points that elicited spike probability greater than zero are shown (data represent mean of n = 7 neurons). e−g As in a−c but for neurons recorded via in vivo 2P guided patch: radial: p = 0.46; axial: p = 0.004, CGH (n = 6 neurons), and 3D-SHOT (n = 6 neurons). h as in d, but in vivo cell-attached patch using 3D-SHOT (data represent mean of n = 5 neurons)

Fig. 5

3D-SHOT provides cellular resolution photostimulation in a large volume through digital focusing. a To quantify the spatial resolution of 3D-SHOT as a function of hologram target depth, we recorded photocurrents in CHO cells while digitally targeting varying positions along the optical axis (z), and measuring resolution by mechanically sweeping the objective over the entire (z) range and measuring the response at each point. b Normalized photocurrent in CHO cells while targeting the same cell from different axial displacements (n = 5 cells; p = 0.39, Kruskal−Wallis test with multiple comparisons correction, data are mean and s.e.m.). c Axial photocurrent resolution as a function of digital displacement—shaded green colors denote mechanical sweeps across the optical axis for different digital displacements. d Quantification of the FWHM for the axial fit of photocurrents in CHO cells as a function of digital defocus from the focal plane (n = 5 cells, p = 0.07, data are mean and s.e.m.). e−h As in a−d, but spike probability recorded in mouse brain slices via current clamp instead of photocurrent in CHO cells recorded in voltage clamp (f: p = 0.2; h: p = 0.17, n = 3 neurons)

Fig. 6

Single-neuron resolution with simultaneous photostimulation of two targets along the axial dimension. a Using CGH or 3D-SHOT we simultaneously photostimulated two targets separated by 80 µm along the optical z-axis. The photocurrent was recorded as a function of the respective displacement Δz between one patched cell and the volume hologram in an example CHO cell (b, n = 11 cells) or in an example cortical neuron (c, n = 5 neurons), with conventional holography (black) and with 3D-SHOT (red). d As in c, but recording spike probability instead of photocurrent for 3D-SHOT

Fig. 7

Power corrected 3D-SHOT simultaneously illuminates 50 spots. a Uncorrected simultaneous 3D-SHOT with a hologram targeting 50 spots in a spiral pattern, occupying 50 individual z-planes. Spots are color coded by normalized 2P absorption in each spot (measured with a sub-stage camera). b Correction factor computed for each spot via 3D power interpolant. c Measurements of 2P-induced fluorescence with 3D-SHOT, and a power-corrected hologram targeting the same locations as in a. d Mean intensity images of power-corrected 50 spot spiral hologram showing (z,y) (top left), (z,x) (bottom left), or (x,y) (right) projection views. e Box plots showing the variance in normalized 2P absorption from each spot within the corresponding hologram before and after power correction (p < 1 × 10⁻¹², F-test of equality of variances, error bars correspond to range, red line is median, n = 50 spots)

Fig. 8

Spatial resolution with simultaneous targets throughout a large volume. a−b Example 3D renderings from tomographic images of two-photon absorption from single-shot holograms targeting increasing numbers of randomly distributed spots. The FWHMs of the two-photon response was computed for each target, c radially, and d axially. Results show that spatial resolution and axial confinement are not significantly degraded by increasing the number of simultaneous targets in any given hologram. e The contrast ratio of the holograms was quantified as a function of the number of targets, and shows that contrast, rather than resolution determines the total number of targets that can be photostimulated in a single hologram. f 3D recordings of various scaling of the same point cloud were made to evaluate density as a possible factor affecting spatial resolution (error bars correspond to range, red line is median)

Fig. 9

Spatially precise volumetric optogenetics with 3D-SHOT. a Schematic demonstrating 3D-SHOT stimulation of a CHO cell with a single spot in a hologram targeting multiple spots in the volume. The objective was moved to measure the physiological point spread function via voltage clamp recordings. b Normalized photocurrent and quantification of FWHM as a function of radial displacement for holograms targeting 1–50 spots (data represent mean ± s.e.m. of n = 3 cells, p = 0.39, Kruskal−Wallis test with multiple comparison correction). c As in b but axially (n = 3 cells, p = 0.64). d Schematic showing whole cell recordings in brain slices while stimulating multiple targets. To measure spatial resolution of 3D-SHOT, a series of new holograms was computed that displaced only the center target while holding all other spots stationary. e Photocurrent measurements from brain slices for radial digital offset of a center spot at several stimulation powers (colors, 0.5–2 W), quantification of FWHM as a function of stimulation power through the objective (data are mean ± s.e.m. of n = 5 neurons, p = 0.69, Kruskal−Wallis test with multiple comparison). f As in e but axially (p = 0.11). g As in d, but for in vivo loose patch recordings; example image of a 2P guided patch of a L2/3 pyramidal neuron expressing ST-ChrimsonR-mRuby2 (scale bar = 50 µm). h Radial and axial physiological point spread functions showing resolution of 3D-SHOT while targeting an ensemble of 21 neurons while only the center spot was moved (data represent mean and s.e.m. from an example neuron). i Bar graph showing mean ± s.e.m. of the radial and axial PPSFs from targeted in vivo patch while stimulating 21 cells (n = 17 radial and n = 18 axial). j Scatter plot showing axial FWHM vs. the depth below the cortical surface of individual neurons (axial PPSF is from experiments with 1 or 21 holograms, n = 24). Red dot indicates the mean values, and the black line is the line of best fit—within this range, cortical depth can explain none of the variance in axial FWHM (R ² = −0.03)