High-speed multiplane confocal microscopy for voltage imaging in densely labeled neuronal populations

Abstract

Genetically encoded voltage indicators (GEVIs) hold immense potential for monitoring neuronal population activity. To date, best-in-class GEVIs rely on one-photon excitation. However, GEVI imaging of dense neuronal populations remains difficult because out-of-focus background fluorescence produces low contrast and excess noise when paired with conventional one-photon widefield imaging methods. To address this challenge, we developed an imaging system capable of efficient, high-contrast GEVI imaging at near-kHz rates and demonstrate it for in vivo and ex vivo imaging applications in the mouse neocortex. Our approach uses simultaneous multiplane imaging to monitor activity within contiguous tissue volumes with no penalty in speed or requirement for high excitation power. This approach, multi-Z imaging with confocal detection (MuZIC), permits high signal-to-noise ratio voltage imaging in densely labeled neuronal populations and is compatible with imaging through micro-optics. Moreover, it minimizes artifacts associated with concurrent imaging and optogenetic photostimulation for all-optical electrophysiology.

Extended Data Fig. 1 -. Detailed system schematic.

The light source is a continuous-wave laser diode (“561 nm”). Blue dashed box (“Photo-stim.”) denotes components of the full-field photo-stimulation module. Green dashed box (“Epi-illum.”) denotes components of the epi-illumination widefield fluorescence microscopy unit. M: mirror, Ir: iris, L: lens, DM: dichroic mirror, PS: polygonal scanner, GS: galvanometer scanner, SM: switchable mirror, OL: objective lens, FPs: focal planes, PH: pinhole, D: detector, C: camera. All part numbers are for the manufacturer Thorlabs, except the objective lens (Nikon).

Extended Data Fig. 2 -. System resolution.

Experimental measurements of the combined excitation and emission point spread functions (PSFs)corresponding to the four simultaneously acquired image planes. Confocal pinhole diameters were selected so that PSFs corresponding to adjacent detectors overlap, eliminating gaps between the four imaging planes, which together cover ~40 μm in the axial direction. Lateral resolution was somewhat anisotropic on account of the slower detector fall time. (X: 2.6 μm full-width half-max; Y: 1.9 μm).

Extended Data Fig. 3-. Photobleaching.

a. Left, Example of raw, unprocessed fluorescence from all cells identified from the in vitro imaging experiment depicted in Fig. 6a,b. Right, Exponential fits to raw fluorescence normalized to initial brightness. These data were collected with the highest excitation power of all in vitro experiments (561 nm; 23.6 mW/mm2). b. Raw, unprocessed recorded from all cells identified in the in vivo imaging experiment depicted in Fig. 3b (top) and exponential fits to in vivo fluorescence, as in (a). The same excitation power was used for all in vivo imaging experiments (561 nm; 300 mW/mm2).

Extended Data Fig. 4 -. Multi-Z imaging minimizes axial motion artifacts.

a. During periods of brain motion, fluorescence is redistributed across planes, with cells becoming brighter in some planes and dimmer in others. Top, average change in fluorescence during a period of movement indicated by arrow below in adjacent planes. Bottom, average fluorescence time series corresponding to the cell indicated (yellow arrowhead). b. Top, mean fluorescence (left), difference image prior to (middle) and following (right)lateral motion correction. Sum of 4 image planes shown. During periods of brain motion (indicated by black arrow, bottom), lateral motion correction compensates for brain motion. Bottom, fluorescence time series corresponding to the cell indicated (yellow arrowhead, top). Time series extracted from a single plane (plane 2) before lateral motion correction (light blue, top), after lateral motion correction (dark blue, middle), and from a region of interest (ROI) spanning all four image planes following lateral motion correction (gray, bottom). Artifacts introduced by brain motion are suppressed when extracting time series from multiplane ROIs. Data are representative of axial motion across 34 fields, 5 animals.

Extended Data Fig. 5 -. Confocal collection efficiency.

Simulated confocal collection efficiency for fluorescence emitted from point sources (left) and 15-μm shells (right) located at different axial positions. Axial resolution is degraded as a result of spherical aberration induced by imaging through 0.5 mm cover glass (bottom; RI = 1.52). Lateral resolution is not degraded as much as axial resolution, meaning that the total fluorescence together collected by all detectors is only modestly reduced. See Methods and Supplementary Note 4 for more details.

Extended Data Fig. 6 -. Comparison of MuZIC and widefield fluorescence microscopy (WFM).

a. Signal-to-noise ratio of MuZIC relative to WFM increases with increasing WFM background-to-signal ratio (BSR). SNR of MuZIC is mostly insensitive to out-of-focus background fluorescence, while SNR of WFM decreases with increasing background. When BSR is greater than about 5, SNR of MuZIC is higher than SNR of WFM (assuming contrast of MuZIC is 0.6 and detection efficiency of MuZIC is 25% relative to WFM; see Supplementary Note 2). b. WFM image of Voltron2-labeled neurons in vivo. Representative BSRs associated with different neurons are indicated for reference. Individual neurons were difficult to resolve in regions of higher labeling density (representative of 4 animals). c. Example in-vivo widefield images of neurons without (top) and with (bottom) targeted illumination delimited by dashed circles. Dashed circles span regions of interest encompassing a single neuron (top left: signal + background) and background only (top right). Corresponding images when targeted illumination was applied (bottom) are shown recentered. In this extreme example where only a single neuron is targeted, the BSR is found to be reduced by a factor γ= 0.2 (see Supplementary Note 3, representative of 4 animals). d. Signal-to-noise ratio of MuZIC relative to WFM without (solid) and with (dashed) the addition of targeted illumination of a single neuron (i.e. γ= 0.2). In general, when multiple neurons are targeted, the reduction in BSR is not as significant and the relative SNR gain provided by MuZIC lies within the shaded region. Targeted illumination has the effect of shifting the break-even point where MuZIC becomes advantageous over WFM to a higher BSR value, up to about 25 in the very extreme case of single-neuron targeted illumination.

Figure 1 –. High-contrast multiplane imaging in densely labeled tissue with MuZIC.

a. Schematic of optical system. High speed scanning is achieved with a 128-facet polygonal scanner operating at 54,945 RPM. Excitation light (green) underfills the objective creating an axially extended illumination beam. Fluorescence is collected through a series of axially distributed reflective pinholes which map to different planes in the sample. b. Numerical simulation of illumination beam (left), simulated point spread functions (PSFs) associated with each image plane (middle, colors), and corresponding PSFs measured experimentally (right, representative of 10 measurements). c. Membrane-bound YFP in a fixed section from a VGAT-ChR2-YFP mouse visualized with MuZIC and widefield microscopy to demonstrate optical sectioning. Confocal pinholes reject background fluorescence while preserving in-plane signal (representative of 20 fields). d. Neurons densely labeled with Voltron2-ST_JF585 can be simultaneously visualized across four focal planes (representative of 40 fields).

Figure 2 –. MuZIC enables high-SNR GEVI imaging in vitro.

a. Simultaneous whole-cell patch-clamp recording and voltage imaging in a cortical pyramidal neuron in vitro. Subthreshold membrane potential dynamics and action potentials can be detected optically in single-trial measurements from the region of interest (ROI) outlined in red (representative of 8 cells, 3 animals). b. Averaging across trials (100 repetitions) reduces shot noise to reveal millivolt-scale dynamics, which are reported on a linear scale (c) (representative of 14 cells, 3 animals). d. The relationship between changes in fluorescence and membrane potential was similar across recorded neurons (n=14 cells, 3 animals; box plot displays median and interquartile intervals).

Figure 3 –. MuZIC reports subthreshold and suprathreshold activity in neural ensembles in vivo.

a. Widefield image of neurons labeled with Voltron2-ST_JF552 in vivo (far left). Insets highlight two FOVs imaged with widefield microscopy (left) and MuZIC (right). FOV (i) is centered on an area of sparsely labeled neurons, while FOV (ii) contains densely labeled cells. Individual neurons are difficult to identify with widefield imaging but easily discernable with MuZIC (representative of 5 animals). b. In vivo imaging of subthreshold membrane potential changes and spiking in layer 2/3 pyramidal neurons reveals correlated dynamics. Two example FOVs are shown (n=2 animals). Action potentials ride on top of subthreshold depolarizations, as expected. FOV coordinates represent axial distance from the pial surface to plane 1. c. Spontaneous activity simultaneously measured from cells distributed across four imaging planes (images representative of 34 fields, 5 animals).

Figure 4 –. Depth dependence of imaging with MuZIC.

a. MuZIC can be used to detect spontaneous activity in neocortical layer 2/3 in vivo. Cells labeled with Voltron2-ST_JF552 at 5 axial depths (measured from pia to plane 1) show subthreshold dynamics and spiking in awake animals (yellow arrows, n=5 FOVs, 4 animals). Schematic illustrates depths at which MuZIC data were collected relative to cortical laminae. b. Quantification of fluorescence, contrast, estimated spike SNR, and bleaching rate for all cells identified across all planes in the experiments depicted in Figs. 3b, c and 4a (n = 52 cells, 8 FOVs, 5 animals). Time series illustrated in Figs. 3b, c and 4a correspond to cells indicated in red.

Figure 5 –. MuZIC can be performed through a microprism to visualize cells across cortical layers.

a. Schematic illustration of microprism positioning in the cortex and depths at which MuZIC data were collected through the prism. Images acquired across the horizontal face of the microprism arise from different depths along the vertical face of the prism. Highlighted imaging planes correspond to planes shown in (b). b. Single plane images collected across 4 depths in the cortex reveal neurons with varied morphology, including deep layer pyramidal neurons (bottom right, representative of 15 fields, 2 animals). Yellow arrows highlight cells whose membrane potential profiles are shown in (c). Scale bar, 50 µm. c. Examples of subthreshold and suprathreshold dynamics in vivo recorded from different penetration depths across all cortical layers.

Figure 6 –. Simultaneous imaging and optogenetic stimulation with minimal optical crosstalk.

a. Top, Schematic of experiment. Overlapping subsets of neurons are virally labeled with Voltron2-ST_JF585 and ChR2. Widefield optical stimulation excites ChR2-expressing neurons. Bottom, Expression of Voltron2-ST_JF585 across four imaging planes (images representative of 16 fields, 2 animals). b. Simultaneous electrophysiological (black) and optical recording (red) of membrane potential from cell indicated by red ROI in (a) as well as optical recording from eight additional Chr2+/Voltron2-ST_JF585+ neurons in response to repeating 400-ms blue light stimuli. c. Short (0.5 ms) blue light stimuli induce action potentials (top, gray bars) or subthreshold voltage deflections (bottom, orange bars) in both electrophysiological (left heatmap) and optical (right heatmap) recordings of membrane potential. 256 repetitions shown. Average electrophysiological and optical recordings across all trials in which action potentials (gray box, top) or subthreshold depolarizations (orange box, bottom) were elicited. d. Optical recordings, as in (c) of additional cells from the same field of view that displayed subthreshold responses (left) and action potentials on a minority (middle) and majority (right) of trials.