A dark quencher genetically encodable voltage indicator (dqGEVI) exhibits high fidelity and speed

Abstract

Voltage sensing with genetically expressed optical probes is highly desirable for large-scale recordings of neuronal activity and detection of localized voltage signals in single neurons. Most genetically encodable voltage indicators (GEVI) have drawbacks including slow response, low fluorescence, or excessive bleaching. Here we present a dark quencher GEVI approach (dqGEVI) using a Förster resonance energy transfer pair between a fluorophore glycosylphosphatidylinositol-enhanced green fluorescent protein (GPI-eGFP) on the outer surface of the neuronal membrane and an azo-benzene dye quencher (D3) that rapidly moves in the membrane driven by voltage. In contrast to previous probes, the sensor has a single photon bleaching time constant of ∼40 min, has a high temporal resolution and fidelity for detecting action potential firing at 100 Hz, resolves membrane de- and hyperpolarizations of a few millivolts, and has negligible effects on passive membrane properties or synaptic events. The dqGEVI approach should be a valuable tool for optical recordings of subcellular or population membrane potential changes in nerve cells.

Keywords: GEVI; cultured neurons; fluorescent membrane potential measurement; genetically encoded voltage indicator.

Fig. 1.

Simultaneous optical and electrical recordings of membrane potential changes in a cultured neuron using the dqGEVI approach with 10 μM D3. (A) Single continuous traces of optical recordings (red) without image processing or filtering, as sampled at 1.08 kHz with an EM-CCD camera. The patch-clamp recordings in the I-clamp configuration (black) were sampled at 50 kHz. Various current pulses of 300-ms duration were injected into the neuron to produce hyper- and depolarizations of the membrane and AP firing. The average ΔF/F (± SEM) for the 27 APs depicted in this trace was 5.01 ± 0.05%. Horizontal dashed line indicates the RMP of −65 mV. (B–E) Parts of the traces labeled with the respective letters on panel A are shown in enlarged snapshots as superimposed traces of optical (red) and electrical (black) recordings. Note the highly accurate temporal overlap between the fluorescence and membrane potential changes. (F) Overlay of fluorescence (red) and membrane voltage (black) during an AP at higher temporal resolution. Note the superimposition of the two traces during both the pre-AP voltage rising to threshold and during the post-AP hyperpolarization. (G) Fluorescence and membrane potential are shown normalized to the peak of the AP. (H) The sampling data points are shown for the two recording modalities to illustrate the rapid change in fluorescence despite the relatively low sampling frequency.

Fig. 2.

Calculation of the correspondence between ΔF/F and membrane potential. (A) A current injection protocol similar to that used in Fig. 1 in another cultured neuron expressing GPI-eGFP recorded in the presence of 10 μM D3. Superimposed traces of nonimage processed fluorescence sampled at 1.08 kHz (red) and the I-clamped membrane voltage (black) downsampled to the same sampling interval from the original 50 kHz. The horizontal dashed line indicates the RMP of −60 mV. (B) A portion of the traces is shown from the part enclosed in the dashed lined box of A. Note the excellent overlay between the optical and electrical recordings as indicated by the accurate reporting of the decreasing AP amplitudes during the train, and the subthreshold depolarizing events (∼5 mV) prior to the current injection (probably spontaneous excitatory postsynaptic potentials; arrowhead) of <10 mV amplitudes. (C) The relationship between ΔF/F and membrane potential was calculated by plotting the two traces point-by-point against each other. The slope of the linear regression (black line) yields the relationship for this cell as indicated in the inset.

Fig. 3.

Speed of the dqGEVI approach as measured during the decay of APs. (A) Rapidly decaying APs (usually those early on during a current pulse injection) were normalized to their peaks (gray: electrophysiology; black: fluorescence) and single exponentials were fitted to their decay phases (dotted red lines). (B) The same was done for slower APs (usually recorded during the late phases of the depolarizing current injections). (C) Comparison of the fast AP decays show a remarkable one-to-one correspondence between the optical and the electrical measurements (dotted line with a slope = 1 for comparison; red line is the linear fit to the data; slope = 1.0334 ± 0.0366; R² = 0.746). (D) Similar to C but for the APs with slower decays (slope = 1.0495 ± 0.026; R² = 0.939). Values expressed as mean ± SD. (E) AP waveforms recorded in HEK293T cells transfected with ASAP2s and GPI-eGFP D3 in voltage clamp mode. Optical traces (black) show the average of 12 traces from 12 cells. The AP waveform voltage trace is indicated in gray. (Inset) The lower sampling rate (1 kHz) of the fluorescent traces makes it look as if the upswing on the AP voltage trace is delayed, which is not the case. (F) Comparison of AP width of the optical traces measured at 50% amplitude. Significance was assessed by Wilcoxon rank-sum test.

Fig. 4.

Amplitude and phase correspondence between ΔF/F and V_m during subthreshold stimuli of increasing frequencies. (A) Superimposed plots of the ΔF/F (red) and V_m (black) during a 4-s chirp pulse (10 to 100 Hz). Morlet wavelet transforms of the ΔF/F trace (Top) and of the V_m trace (Bottom) showing the linearly increasing frequency responses. Note how the ΔF/F response faithfully follows the linear change in frequency. (B and C) Zoomed images of the ΔF/F and V_m traces from the shaded boxes labeled in A. (Top) The smoothed ΔF/F (dark red) and raw ΔF/F (light red) traces, superimposed with the V_m trace (black). (Bottom) The phases of the ΔF/F (red) and V_m (black) with the smoothed ΔF/F and V_m traces in the background shown in fainter colors. (D) Plot of ΔF/F phase vs. V_m phase. Individual points are represented in red, and the greyscale represents the binned histogram values with a bin width of 0.04π rad. Pearson’s R = 0.61; P = 0. (E) Normalized cross-correlation between ΔF/F phase and V_m phase. The RMS normalization of the cross-correlogram was done as described in Materials and Methods. (F) Histogram of point-by-point differences between ΔF/F phase and V_m phase. The mean of the Gaussian fit (red) is at −0.011 rad, that is, −311 μs. The mean ± SEM lag betwen ΔF/F phase and V_m phase determined in this manner in eight experiments was −276 ± 240 μs.

Fig. 5.

High photostability of GPI-eGFP D3. (A) Normalized fluorescence intensity of GPI-eGFP D3 over 15-min continuous illumination with an LED (illumination intensity 4.3 mW × mm⁻², the standard intensity used in all of our experiments). Points are average fluorescence values sampled every 60 s in three cells. The fitted exponential decay (green line) has a time constant of 2,268 s (∼38 min) with a ±SD (shaded green) of 177 s (∼3 min). Graph is superimposed with the photostability curves for Voltron (black) and Positron (red) (17) (illuminated with an LED, light intensity 18 mW × mm⁻²). (B) Normalized SNRs of GPI-eGFP D3 over 15-min continuous illumination (n = 5 neurons). Error bars indicate ± SEM. (C) Example raw traces from a single cultured neuron showing spontaneous APs at baseline (Upper) and after 15 min of continuous illumination (Lower).

Fig. 6.

Accuracy of AP detection at 50 Hz and 100 Hz and its ROC analysis. (A) Detection of APs elicited with 4-ms current pulses at 50 Hz. (Upper) Raw (light red) and smoothed (red) ΔF/F signal of example recording. Threshold (dashed line) for detection of fluorescent APs (fAPs) was set at 75% peak amplitude of the first fAP, determined as the peak ΔF/F in a ±3-ms time window of the first electrophysiological (V_m) AP relative to a 180 ms baseline period. Crosses indicate threshold crossings, peaks of detected fAPs are indicated in blue. (Lower) Corresponding electrophysiological trace. Threshold for detection of V_m APs was set at 0 mV. (B) Superimposed raw fluorescent and electrophysiological traces from (A) (R = 0.98). (C) Same as A at 100 Hz. (D) Superimposed raw fluorescent and electrophysiological traces from C (R = 0.95). (E) ROC analysis of the indicated number of cells and traces with 579 APs and 102 failures elicited at 50 Hz and 396 APs and 296 failures elicited at 100 Hz.

Fig. 7.

Optical recordings of synchronous activity. (A) Two fluorescent cultured neurons on the same coverslip. Cell 1 was recorded in the I-clamp whole-cell mode. Cell 2 was only optically monitored. White circles indicate the ROI from which the imaging trace was taken. (B) When treated with 4-AP (50 μM) neurons developed epileptiform bursting. The simultaneous recordings show the fluorescent signals in the two cells (red and green traces) and the electrical recording in Cell 1 (black trace). (C) The area shaded in gray on the left of B magnified to show suprathreshold activity in Cell 2 and subthreshold activity in Cell 1. (D) The area shaded in gray on the right of B magnified to show suprathreshold activity in both cells. Note the alignment of the start of the synchronous discharges in both C and D. (E) A 128- × 128-pixel image of a GPI-eGFP–expressing cultured cortical primary mouse neuron imaged at 10 kHz. Spontaneous activity was elicited by incubation with 4-AP. (F) Superimposed single traces of the dendritic signals at a proximal location (orange) and at a more distal location (green). The somatic signal (black, ROI indicated at soma with white circle) precedes the dendritic signals. (G) Quantification of the soma–dendritic delay in a single neuron measured on average at a distance of 12.60 ± 0.09 µm (orange) and 40.62 ± 0.17 µm (green) from the soma (n = 13 depolarizations). Error bars indicate ± SEM.

Fig. 8.

Long-lasting optical measurements of membrane voltage following removal of the acceptor/quencher D3 from the extracellular space. (A) Individual traces of raw unfiltered and unprocessed fluorescent signals (sampled at 1.08 kHz) of membrane potential changes following the indicated times after the washout of 10 μM D3 from the recording chamber. The current injection protocol is the same as that shown in Figs. 2 and 3. Each protocol required a 6-s continuous illumination and was repeated four times every 10 min. Despite the multiple exposures to light, the SNR (z-score) of the first AP in the train was remarkably constant over time (at 0 min: 13.8 and at 60 min: 12.2). (B) Simultaneous optical (red) and electrical (black) recording from a mouse cultured neuron recorded 24 h after 10 min of incubation in 2 µM D3. The patch-clamp recordings in the I-clamp configuration (black) were sampled at 50 kHz. (Upper) Whole trace of current pulses of 300-ms duration injected into the neuron to produce hyper- and depolarizations of the membrane and AP firing. (Lower) Enlarged snapshots of hyerpolarizing (Lower Left) and depolarizing (Lower Right) current injections. (C) Superimposed average trace of five AP waveforms induced in a HEK293T cell expressing GPI-eGFP measured 24 and 48 h after 10 min of incubation with 2 µM D3. (D) Bar graph showing SNRs 24 and 48 h after incubation compared to Ctrl. Error bars indicate ± SEM.