Ultrafast Two-Photon Imaging of a High-Gain Voltage Indicator in Awake Behaving Mice

Abstract

Optical interrogation of voltage in deep brain locations with cellular resolution would be immensely useful for understanding how neuronal circuits process information. Here, we report ASAP3, a genetically encoded voltage indicator with 51% fluorescence modulation by physiological voltages, submillisecond activation kinetics, and full responsivity under two-photon excitation. We also introduce an ultrafast local volume excitation (ULoVE) method for kilohertz-rate two-photon sampling in vivo with increased stability and sensitivity. Combining a soma-targeted ASAP3 variant and ULoVE, we show single-trial tracking of spikes and subthreshold events for minutes in deep locations, with subcellular resolution and with repeated sampling over days. In the visual cortex, we use soma-targeted ASAP3 to illustrate cell-type-dependent subthreshold modulation by locomotion. Thus, ASAP3 and ULoVE enable high-speed optical recording of electrical activity in genetically defined neurons at deep locations during awake behavior.

Keywords: ASAP3; GEVI; RAMP; ULoVE; action potential; electrophysiology; electroporation; theta oscillation; two-photon; voltage imaging.

Figure 1.. Electroporation-based screening of GEVIs.

(A) Left, responses of various GEVIs in HEK293-Kir2.1 cells to a 0.01-ms 150-V pulse recorded at 100 Hz. ASAPIncp is a voltage-insensitive variant of ASAP1 in which the GFP was de-permuted (Chamberland et al., 2017). Solid traces, mean responses. Data are from five or six wells from two separate experiments, with the top five responders in each well analyzed. Right, fluorescence responses were recorded at 200 Hz in HEK293A cells expressing various GEVIs in response to a voltage step from −70 to 0 mV (ASAP1ncp, n = 4; ASAP1, n = 6; Arclight, n = 3; ASAP2s, n = 10). Error bars, standard deviation (SD). (B) ASAP can be directly expressed from these unpurified PCR products. HEK293-Kir2.1 cells were transiently transfected using Lipofectamine 3000 and either pc3-CMV-ASAP2s plasmid or linear ASAP2s product generated by PCR. Intensity scaling is the same across images. (C) Overview of the screening system. HEK293-Kir2.1 cells expressing GEVI variants are plated in 384-well plates on conductive glass slides connected to a square-pulse generator. The medium in each well is sequentially contacted by a motorized platinum electrode during imagine by a high-speed CMOS camera. The entire procedure is automated by MATLAB routines. (D) Left, model of ASAP domain organization. Right, residues targeted for mutagenesis in the S3-cpGFP linker. (E) Evolutionary history of ASAP3. (F) Fluorescence responses from the third round of electrical screening for the parental ASAP2f L146G S147T R414Q (left) and the best performing mutant, ASAP2f L146G S147T S150G H151D R414Q (right). (G) One-photon fluorescence response curves of ArcLight Q239 (n = 3), ASAP2f R414Q (n = 6), and ASAP3 (n = 10) from HEK293 cells stepped for 500ms from a holding potential of −70 mV. See also Figures S1–S4.

Figure 2.. Characterization of ASAP3 in cells.

(A) Fluorescence-voltage (F-V) curves for ASAP variants. Normalized sigmoid traces fit to steady-state fluorescence responses of ASAP1 (blue trace, n = 6 and 4), ASAP2s (purple, n = 5 and 7), or ASAP3 (green, n = 3 and 6). Error bars, standard error of the mean (SEM). (B) ASAP3 fluorescence activation kinetics during a sustained voltage step from −80 to +20 mV (top) and deactivation kinetics following a 1 ms voltage step from −80 to +20 mV (bottom) in an example CHO cell at 33 °C. Activation kinetics (T_fast = 0.7ms, 72%; T_slow = 4.7ms for this cell) were faster than deactivation kinetics (T_fast = 1.0 ms, 50%; T_si_ow = 4.7ms). (C) Top, fast activation time constant and weighted deactivation time constant during various voltage steps from or to a holding potential of −80 mV, respectively. A single weighted deactivation time constant t was calculated for each potential as (a₁T₁ + a₂T₂)/(a₁ + a₂), where a₁ and a₂ are the coefficients for the bi-exponential fit with time constants t₁ and t₂. Circles and error bars represent mean ± standard deviation (SD) for n = 16 cells. Bottom, fraction of steady-state response achieved by the fluorescence transient evoked by a 1-ms depolarizing pulse. (D) Top, one-photon wide-field image of a representative cultured hippocampal neuron showing efficient plasma membrane localization of ASAP3. Below, response of ASAP3 fluorescence in a dendritic region including spines upon a commanded voltage change from −70 to +30 mV. (E) Responses of ASAP2s and ASAP3 to current-evoked APs in cultured hippocampal neurons with similar AP waveforms (peak amplitudes of 63.5 ± 0.8 mV and 63.7 ± 1.0 mV; full width at half-maximum of 5.0 ± 0.1 and 3.9 ± 0.2 ms for ASAP2s and ASAP3 respectively). Grey lines, single-trial responses (n = 25 each). Colored lines, mean responses. (F) Mean peak fluorescence response, optical spike width at half-maximum, and area under the curve of ASAP2s (n = 4) and ASAP3 (n = 5) in cultured hippocampal neurons. Each neuron fired 3–34 APs, whose values were averaged. (G) Electrical and optical responses from a representative cultured hippocampal neuron expressing ASAP3 in current-clamp mode. Asterisks indicate spontaneous spikes not elicited by current injection. (H) Example voltage (top) and ASAP3 fluorescence (bottom) traces showing ASAP3 responses to sEPSPs of various amplitudes. See also Figure S5.

Figure 3.. Spike detection in mammalian brain tissue by ASAP3 and two-photon microscopy.

(A) An overlay image of a two-photon maximum-intensity z projection of an ASAP3-expressing neuron in a cultured hippocampal slice, in the whole-cell configuration with an intracellular solution containing red fluorescent Alexa Fluor 594. The inset shows the expanded somatic region of the neuron, with red dots marking the 20 recorded voxels. (B) Representative examples of ASAP3 optical responses from the neuron in A to current-evoked APs acquired at 20 °C or 32 °C. The ASAP3 signal tracks the rising phase of AP, but exhibits a longer decay time than the AP. About half of the decay kinetics are accounted for by a time constant of 4.3 ± 0.4 ms (n = 12 neurons) at 32 °C or 9.3 ± 1.9 ms (n = 8) at 20 °C. A second component of about 30 ms accounted for the other half of the decay, which may relate to the decay of the underlying after-depolarization. (C) Left, representative examples of ASAP2s and ASAP3 responses to current-evoked APs acquired from cells in cultured hippocampal slices at 32 °C (ave rage of 20 voxels over 10 trials). Right, ASAP3 shows higher peak amplitude responses than ASAP2s to current-evoked APs recorded at 32 °C ***p < 0.0001 (two-tailed t-test). Bars represent mean ± SEM. (D) Left, schematic of whole cell current-clamp and ULoVE recordings of molecular layer interneurons in a parasagittal cerebellar slice expressing ASAP3-Kv. Right, two-photon stack projection of sparse ASAP3 expression in molecular layer interneurons in a parasagittal cerebellar slice. The inset shows the expanded somatic region of the neuron, with red ellipsoids marking the regions of optical acquisition. (E) Left, averaged ASAP3 fluorescence transient (50 trials) for the interneuron in G shows a characteristic interneuron waveform with after-hyperpolarization. Right, ASAP3 shows improved responsivity over ASAP2s for APs in molecular layer interneurons. Bars represent mean ± SEM. *p = 0.026 (Wilcoxon rank sum test). (F) Representative projection confocal images of ASAP3 (left) and ASAP3-Kv (right) expression in cortical slices reveal that the Kv2.1 PRC motif enriches ASAP3 signals in cell soma. (G) Schematic of whole cell current-clamp and optical line-scan recordings of ASAP3-Kv from medium spiny neurons in acute striatal slice. (H) ASAP3-Kv fluorescence reliably tracks 10-Hz (left) or 100-Hz (right) current-evoked AP trains in acute slices. A single unfiltered trace is shown. See also Figure S6.

Figure 4.. Principle and implementation of ultrafast local volume excitation (ULoVE).

(A) Schematic of 3D scanning with AODs. Acoustic frequency is modulated as a sinusoidal function producing both time-varying lateral scanning and axial defocusing. At any time, the optical wavefront at the output of the AOD is the integral of the acoustic frequency function in the AOD window. (B) Schematic description of holographic multiplexing with AODs. A frequency function covering a fraction of the AOD window is imposed to produce a desired homogeneous pattern by holography in the focal plane. The concatenation of this function is equivalent to a convolution with regularly spaced Dirac functions, leading by optical Fourier transform to the generation of fixed regularly spaced diffraction-limited points bounded by the holographic pattern in the focal plane of the objective. (C) Comparison of excitation volumes of a diffraction-limited focal spot (pattern 1) and two ULoVE patterns created by combined multiplexing and scanning (patterns 2 and 3). (D) Left, pattern 3 reduces photobleaching at similar photon flux to pattern 1. Left, mean normalized photon flux in 1-min recordings (n = 24 cells for pattern 1 or 17 cells for pattern 3). Center, fast time constants (6.09 ± 2.78 ms for pattern 1,243 ± 129 ms for pattern 3), obtained by exponential fitting to the initial fast monoexponential decay phase (initial 50 ms in 22 cells for pattern 1 or initial 2 s in 16 cells for pattern 3 after excluding cells with poor fits). Right, quantification of photobleaching for patterns 1 (grey, n = 24 cells) and 3 (red, n = 17 cells). (E) Left, two-photon timelapse projections of an ASAP3-Kv-expressing cortical neuron from a single frame or averaged over multiple frames. ULoVE patterns 2 (blue) and 3 (green) encompass a large fraction of the membrane even with movement blur. Right, brain motion-induced image displacement calculated from the registration data (grey points) and its mean ± SD (black) (n = 29 recordings). Half-widths of the ULoVE patterns are indicated by color-coded lines. (F) Left, quasi-simultaneous interleaved 5-min optical recordings with the three patterns pointing at the same location on a neuron, along with animal running speed. Single trials are shown. Right, mean coefficient of variation for each ULoVE pattern across all neurons (n = 50). (G) ULoVE suppresses depth-dependent motion artifacts. The coefficient of variation of ASAP3-Kv signals increases with depth for patterns 1 and 2 but not for pattern 3. Pearson correlation coefficients are indicated. *** p<0.001; * p<0.05. n = 50 neurons in 6 mice. (H) ULoVE improves photon flux from ASAP3-Kv. Measured photon fluxes were normalized by the squared power at each focal point in the ULoVE patterns and expressed relative to the maximum photon flux (when the spot is aligned with the membrane) from the quasi-simultaneous diffraction-limited spot recording. In all panels, error bars represent SD. *** p < 0.001. * p < 0.05. NS, not significant.

Figure 5.. High-fidelity optical voltage recordings with ULoVE and ASAP3-Kv in the cortex.

(A) Top, schematic of whole-cell patch clamp and ASAP3-Kv recording in head-fixed anesthetized mice. Bottom, representative image of ASAP3-Kv-expressing V1 cortical L2/3 neuron patched in whole-cell mode, with intracellular solution containing Alexa Fluor 594 (AF594). (B) Simultaneous current-clamp (top) and ASAP3 (middle) recording showing faithful optical tracking of both APs and sub-threshold voltage fluctuations. Bottom, ASAP3 optical trace, smoothed for visualization using a 20–150-Hz bandpass filter. Tick marks and cross indicate true-positive and false-negative events, respectively. (C) ASAP3-Kv fluorescence is linearly related to spontaneous voltage fluctuations from −40 to −80 mV (data from B). Probability distribution refers to the time spent at each coordinate out of all time points. (D) Averaged electrical (left) and optical (middle) spike-triggered waveforms (n = 311 spikes, data from B). Right, optical spikes showed decay kinetics of 5.4 ms. (E) Histogram reveals near-millisecond precision of optical spikes. Optical spikes from the brightest in vivo patch-clamped neurons was detected and timed by MLspike, and lag times between optically and electrically recorded spikes were calculated. (F) Left, schematic of ULoVE imaging of awake head-fixed mice running on a wheel. Right, maximal intensity projection of a 25-μm z-stack in V1 cortex L2/3 injected with AAV1-hSyn-Cre and AAV1-eF1α-FLEx-ASAP3-Kv, displaying representative sparse expression. (G) Example ASAP3-Kv recording in awake mouse of a V1 neuron with spontaneous bursty activity and MLspike-detected events (ticks). (H) High-temporal-resolution examples of single APs and bursts (data from H). (I) ΔF/F distribution for detected APs from the 23 sorted neurons recorded in awake mice. (J) ΔF/F vs. F₀×t relationship shows the discriminability (d’) of APs from 23 visual cortex L1–3 neurons. Crosshair indicates mean ± SD.

Figure 6.. Deep multi-unit recordings of spiking and subthreshold activity in cortex and hippocampus.

(A) Left, series of two-photon images at indicated depths in L5, asterisks indicate somata. Right, image of proximal apical dendrites seen at 421 μm. (B) Spontaneous activity of three simultaneously recorded proximal apical dendrites of L5 cortical neurons as indicated in L. Ticks indicate dendritic spikes detected by MLspike. Gray lines indicate up and down states. (C) Average events detected in dendrites #1 (n = 15) and #2 (n = 28). (D) Top, single-trial fluorescence trace illustrating theta-frequency membrane potential oscillations in the hippocampal neuron displayed on the left. Red ticks indicate APs detected by MLspike. Middle, running speed heat map for the corresponding period. Bottom, spectrogram of the optical signal for the same period. (E) Top, average power spectrum reveals a large peak in the theta frequency band during run (black curve) but not rest (dashed black curve). Bottom: scatter plot indicates a positive relationship between mouse speed and theta power. Red line represents best linear fit. (F) Single-trial recordings from the rest (top) and run (bottom) epochs marked in D.Ticks, APs detected by MLspike. Red line, basal sub-threshold fluorescence fluctuation as extracted by MLspike. Note that spikes are nested on hippocampal theta oscillations during run. (G) Left, histogram of the phase probability distribution of optical spikes relative to theta cycles (grey curve). Right, average spike-triggered fluorescent trace (n = 391 spikes) illustrating spike locking to theta oscillation. (H) Large-scale z-stack projection reveals distributed parvalbumin-expressing neurons. Inlet represents a zoom into the somata of a parvalbumin-expressing basket cell. (I) Single-trial optical recording of the neuron in H. (J) The spike-triggered fluorescent trace average (n = 125 spikes) of the neuron in H shows the spike after-hyperpolarization typical of interneurons.

Figure 7.. ASAP3-Kv reveals locomotion modulation of cell bursting in visual cortex.

(A) Left, single-plane two-photon images of 4 neurons recorded in one cortical column at indicated depths. Middle, zoomed z-stack projections of the neurons. Right, single-trial ASAP3-Kv recordings. (B) Average AP shape reveals a variety of waveforms. From top to bottom, n = 1500, 102, 62, and 79 spikes. (C) Left, two representative cumulative interspike interval distributions with tri-exponential fits (dashed lines). Center, scatter plot of the fitted time constant of the first and second components allows segregation of bursting neurons (red) from regular firing neurons (blue). Right, cumulative interspike interval distributions for all neurons, with lines color-coded by burstiness. (D) Classification and spike-aligned traces of the 23 neurons analyzed. Four groups were found, consisting of bursty neurons (11 cells, 2934 spikes), regular-firing neurons with afterhyperpolarization (AHP, 6 cells, 2405 spikes, blue), regular-firing without AHP (3 cells, 803 spikes), and slow-firing cells (3 cells, 77 spikes). (E) Interspike intervals and spike number per burst (inset) display power law distributions, as expected from a sparsely active cortical network. Aggregated data from all 23 analyzed neurons are shown. (F) Example of a recording from a regular-firing neuron with AHP. Detected spikes are indicated by ticks. (G) Stairstep plot of firing rate for a regular-firing neuron with AHP, calculated over 1-s bins. (H) Average firing rates as a function of running speed binned in 3 categories: still (< 2 cm/s), walking (2–5 cm/s), and running (> 5 cm/s), for the neuron in G (Pearson correlation coefficient r = 0.39, p < 0.0001). (l) Mean (dark line) ± SEM (light shaded line) of the spike waveforms for cells during run (orange) and rest (black) periods for regular-firing cells with AHP (4 cells, 827 and 1257 spikes during run and rest, respectively), for bursty cells with speed-modulated firing (3 cells, 601 and 411 spikes) or non-modulated firing (3 cells, 33 and 466 spikes). Dashed lines represent spike threshold chosen as a membrane potential reference. See also Figure S14.