Sensitivity optimization of a rhodopsin-based fluorescent voltage indicator

Abstract

The ability to optically image cellular transmembrane voltages at millisecond-timescale resolutions can offer unprecedented insight into the function of living brains in behaving animals. Here, we present a point mutation that increases the sensitivity of Ace2 opsin-based voltage indicators. We use the mutation to develop Voltron2, an improved chemigeneic voltage indicator that has a 65% higher sensitivity to single APs and 3-fold higher sensitivity to subthreshold potentials than Voltron. Voltron2 retained the sub-millisecond kinetics and photostability of its predecessor, although with lower baseline fluorescence. In multiple in vitro and in vivo comparisons with its predecessor across multiple species, we found Voltron2 to be more sensitive to APs and subthreshold fluctuations. Finally, we used Voltron2 to study and evaluate the possible mechanisms of interneuron synchronization in the mouse hippocampus. Overall, we have discovered a generalizable mutation that significantly increases the sensitivity of Ace2 rhodopsin-based sensors, improving their voltage reporting capability.

Keywords: biosensors; fluorescence imaging; fluorescent proteins; genetically encoded indicators; voltage imaging.

Figure 1:. Mutagenesis and screening of Voltron in cultured neurons.

a. Residues targeted for SSM in the Ace2N rhodopsin domain of Voltron, colored by the rationale for targeting them. b. Mutagenesis and screening workflow. c. Diagram of field stimulation assay performed in 96-well plates. d. Representative image of a neuron from the screen expressing Voltron2₅₂₅. Inset shows representative frame during fast (1.497 kHz) stream acquisition. Scale bar: 10 μm. e. Field stimulation parameters (top, black) and acquired fluorescence response of the neuron shown in d (bottom, red). All imaging in the screen was performed at a light density of 1.14 mW/mm² measured in the image plane. f. Field stimulation assay results of the SSM Voltron₅₂₅ screen, ranked by maximum $\left|\text{ΔF}/{\text{F}}_0\right|$ for each variant, normalized to in-plate Voltron₅₂₅ controls. g. Mutated residues colored by the maximum increase in $\left|\text{ΔF}/{\text{F}}_0\right|$ achieved in that position. Top three mutations are labeled. h. Representative traces (mean ± s.e.m.) from a single plate containing Voltron₅₂₅ (8 wells) and Voltron2₅₂₅ (8 wells). i. Single AP $\text{ΔF}/{\text{F}}_0$ (mean ± s.e.m.; Voltron₅₂₅, −.059 ± 0.001, n=338 wells; Voltron2₅₂₅: −0.090 ± 0.002, n=130 wells; p<0.0001, Mann-Whitney U test) and SNR of Voltron2₅₂₅ and SNR (Voltron₅₂₅: 1.80 ± 0.013; Voltron2₅₂₅: 2.24 ± 0.040; p<0.0001, Mann-Whitney $U$ test). See also Fig. S1.

Figure 2:. Automated patch-clamp screening and characterization of Voltron2 in cultured neurons.

a. Fully automated uM workstation screening platform, based on PatcherBot. The pipette cleaning procedure is shown where a used pipette is dipped into a reservoir of cleaning solution (step 1, “c”) and back to the neuronal culture for a subsequent patch-clamp attempt without the need for replacing the pipette (step 2). b. Peak fluorescence responses to voltage steps (−70 to +30 mV) of Voltron₅₂₅, Voltron2₅₂₅ and the top two variants from the field stimulation assay (mean ± s.e.m.; Voltron2₅₂₅ vs. Voltron₅₂₅: p=0.012; Voltron2₅₂₅ vs. Voltron₅₂₅.V74G: p=0.015; Voltron2₅₂₅ vs. Voltron₅₂₅.V74W: p=0.0003, one-way ANOVA followed by Dunnett’s post-hoc test). Inset: Voltron₅₂₅ and Voltron2₅₂₅ fluorescence traces (solid line: mean, shading: s.e.m.) in response to −70 to +30 mV voltage steps. N values (cells) indicated in figure. c. Mutated residues from 1^st screening round (single sites) colored by the maximum $\text{ΔF}/{\text{F}}_0$ response to 100 mV (−70 to +30 mV) voltage steps, measured with the uM workstation. Top mutations at each position are labeled. d. Onset (top) and decay (bottom) fluorescence kinetics of Voltron₅₂₅ and Voltron2₅₂₅ in response to a +100 mV voltage step from −70 mV. Vertical axis scaled to match $\text{ΔF}/{\text{F}}_0$ between the sensors. e. Onset and decay kinetics (mean ± s.e.m.) of the traces in (d). Onset kinetics: *p=0.03, Mann-Whitney U test. Decay kinetics: *p=0.03, Mann-Whitney U test; Voltron₅₂₅: n=4 cells, Voltron2₅₂₅: n=4 cells. f. Representative fluorescence responses to single evoked APs in current clamp. Scale bar: 10 μm. g. $\text{ΔF}/{\text{F}}_0$ in response to single AP stimulation in current clamp mode (mean ± s.e.m.; *p=0.03, Student’s t test, Voltron₅₂₅: n=5 cells, Voltron2₅₂₅: n=7 cells). h. Normalized resting fluorescence relative to mTagBFP2 fused to the C terminus (mean ± s.e.m.; ****p<0.0001; Voltron₅₂₅: n=105 cells, Voltron2₅₂₅: n=115 cells, Student’s t test). i. Photobleaching comparison of Voltron₅₂₅ and Voltron2₅₂₅ over 10 mins (solid line: mean; shading: s.e.m.). All experiments were performed at room temperature. See also Figs. S2-S5.

Figure 3:. Synthetic PSP (synPSP) detection using Voltron2 in mouse brain slices.

a. Experimental setup in acute mouse brain slice. b,c. Percent change in fluorescence over time for Voltron₅₈₅ (b; n=6 cells; 2 mice) or Voltron2₅₈₅ (c, n=4 cells; 2 mice) in response to changes from resting membrane potential of −15mV to +15mV in 5mV increments (lower panels), intended to mimic typical inhibitory or excitatory synaptic transmission. Solid lines: mean; shading: s.d.. A representative cell for each construct is shown in the inset (scale bar = 10μm). d. Top: percent change in fluorescence as a function of the peak amplitude of the synthetic postsynaptic potential (synPSP) applied to the cell (solid line: mean, shading: s.d.). Bottom: sensitivity index (d’/mV) of Voltron2₅₈₅ is significantly higher than that of Voltron₅₈₅ (p=0.025, Welch’s t-test). See also Fig. S6.

Figure 4:. Simultaneous voltage imaging and optogenetic stimulation.

a. (Top) Average intensity projection of 457 Hz confocal images showing Voltron2-expressing cells labeled with JF₅₈₅ in an acute slice of motor cortex (n=1 mouse). Pipette used for whole-cell recordings illustrated in red. (Bottom) Post-hoc confocal image showing pan-neuronal expression of ChR2-GFP in the same field of view (FOV) shown in top panel, with patched cell #1 indicated by white arrow. b. Whole-cell membrane voltage (black traces) and corresponding Voltron2 fluorescent signal (red traces) from patched cell #1 shown in a, showing responses to 400 ms stimulation with 10 (top), 30 (middle), and 50 μW/mm² (bottom) blue light. c. Voltron2₅₈₅ signals (red and gray traces) recorded across 10 distinct cells in the FOV shown in (a) in response to 400 ms stimulation with 30 μW/mm² blue light. Corresponding membrane voltage is shown for patched cell # 1 (upper black trace). d. Raster plots show trial-aligned APs detected in fluorescent signals from cells #1–10 shown in (a) and (c), across 10 repeated 400 ms blue stimulus trials. See also Fig. S7.

Figure 5:. In vivo comparison of Voltron-ST and Voltron2-ST in zebrafish olfactory sensory neurons.

a. Experimental setup. Left: Olfactory sensory neurons expressing Voltron-ST or Voltron2-ST, labeled with JF₅₅₂ and imaged at 400 Hz using a lattice-lightsheet microscope. ex: excitation objective lens, em: imaging objective lens. Right: Volumetric rendering of olfactory sensory neurons in the nasal cavity. r, rostral; c, cadual; m, medial; l, lateral b. Representative FOVs and recordings. Spatial weights optimized for individual spiking neurons are shown in distinct colors over the structural image (left). The activity trace of corresponding neurons is shown in the same color (right). c. Performance comparisons of Voltron₅₅₂-ST and Voltron2₅₅₂-ST. Left: Distribution of spike-related fluorescence change of Voltron₅₅₂-ST and Voltron2₅₅₂-ST. Right: Distribution of SNR of Voltron₅₅₂-ST and Voltron2₅₅₂-ST. P values: Wilcoxon rank-sum test.

Figure 6:. Imaging of voltage activity in vivo in flies.

a. Experimental setup. A head-fixed fly is imaged using an sCMOS camera at 800 Hz. Voltron is loaded with JF₅₅₂-Halotag ligand via a one-hour incubation/one-hour wash-out protocol. b,d. Voltage recordings in MBON-$\text{γ1pedc}$>$\text{α}/\text{β}$ (MB112C-Gal4) and PPL1-$\text{γ1pedc}$ (MB320C-Gal4). Neuron schematics are shown for the left hemisphere with the MB in shaded gray (arrowheads indicate axonal outputs). Fluorescence images were acquired from the $\text{γ1}$ compartment (inset, 50 μm × 50 μm), which contains dendrites of MBON-$\text{γ1pedc}$>$\text{α}/\text{β}$ and axon terminals of PPL1-$\text{γ1pedc}$. Single-trial recordings of $\text{ΔF}/{\text{F}}_0$ traces are shown (8.4 and 6.0 mW/mm² for b and d respectively). c. Spike amplitude with Voltron2₅₅₂ and Voltron₅₅₂ in MBON-$\text{γ1pedc}$>$\text{α}/\text{β}$. p=6.4×10⁻¹⁴, Wilcoxon rank sum test. For Voltron2₅₅₂, the data set was from 15 hemispheres (8 flies) at three levels of illumination for a total of 45 experiments, for Voltron₅₅₂, 13 hemispheres (7 flies) with 39 experiments. Box represents interquartile range (IQR), center represents median, notch represents 95% CI, and whiskers indicate 1.5xIQR. e. Spike amplitude in PPL1-$\text{γ1pedc}$. p=5.0×10⁻¹¹, Wilcoxon rank sum test. For both Voltron2₅₅₂ and Voltron₅₅₂, the dataset was from 10 flies at three levels of illumination for 30 total experiments. f,h. SNR calculated as spike amplitude over standard deviation of the spike-free zones of the trace. p=0.07, 0.006, 0.003 between Voltron2₅₅₂ and Voltron₅₅₂ in MBON-$\text{γ1pedc}$>$\text{α}/\text{β}$, p=0.28, 0.16, 0.17 in PPL1-$\text{γ1pedc}$, Student’s t-test. g,i. Lower basal fluorescence levels with Voltron2₅₅₂. p=0.0027, 0.0048, 0.0046 in MBON-$\text{γ1pedc}$>$\text{α}/\text{β}$, p< 0.05 in PPL1-$\text{γ1pedc}$, Student’s t-test. p=0.0132, 0.0148, 0.02. Values in (f-i) shown as mean ± s.e.m. *p<0.05, no correction for multiple comparisons was performed.

Figure 7:. Imaging of voltage activity in vivo in mouse hippocampus and cortex with Voltron and Voltron2.

a. Example image of hippocampal PV neurons expressing Voltron2-ST labeled with JF₅₅₂. Scale bar: 200 μm. b. Sample fluorescence traces of cells 1–4 in (a). c. Spike waveforms of cells expressing Voltron₅₅₂-ST or Voltron2₅₅₂-ST. Solid line: mean; shading: s.e.m. d-g. Comparison of Voltron₅₅₂-ST and Voltron2₅₅₂-ST spike amplitude (d), baseline fluorescence (e), noise standard deviation (f), and SNR (g) in hippocampal PV neurons (mean ± s.e.m.). N values indicated in (c). h. Example images of cortical pyramidal neurons expressing Voltron₅₂₅-ST (top) or Voltron2₅₂₅-ST (bottom) labeled with JF₅₂₅. Scale bar: 10 μm. i. Example fluorescence traces from individual neurons recorded using Voltron₅₂₅-ST (blue) and Voltron2₅₂₅-ST (red) detrended using a 5s median filter. Grey dashed boxes indicate detection of 3–5Hz oscillations shown in (j) and quantified in (o). j. Zoomed portions of the fluorescence traces in (i) showing spikes and 3–5Hz oscillations. k-o. Comparison of Voltron₅₂₅-ST and Voltron2₅₂₅-ST spike amplitude at both 500 and 1000 Hz imaging rates (k), baseline fluorescence (l), noise standard deviation (m), SNR (n), and 3–5Hz oscillation amplitude (o) in cortical pyramidal neurons (mean ± s.e.m.; 107 neurons expressing Voltron₅₅₂-ST in 4 mice, 102 expressing Voltron2₅₅₂-ST in 4 mice). p. NDNF interneurons in mouse ALM expressing Voltron2₅₅₂-ST. Scale bar: 100 μm. q. $\text{ΔF}/{\text{F}}_0$ traces during 3 min of recording at 400 Hz from neurons shown in (p), in decreasing order of SNR. Scale bars are 10 s and −5% $\text{ΔF}/{\text{F}}_0$. r. $\text{ΔF}/{\text{F}}_0$ traces from color-coded regions of (q) with action potentials represented as black dots. s-v. Comparison of Voltron₅₅₂-ST and Voltron2₅₅₂-ST spike amplitude (s), subthreshold $\text{ΔF}/{\text{F}}_0$ (t), % cells passing quality control (u), and expression density (v) (245 neurons expressing Voltron₅₅₂-ST in 5 mice, 181 expressing Voltron2₅₅₂-ST in 2 mice). For all plots: Statistically significant differences between groups were determined by two-sided Wilcoxon rank-sum test. *p < 0.05, **p < 0.01, ***p < 0.001. See also Fig. S8.

Figure 8:. Co-depolarization of hippocampal PV interneurons in mice.

a. Example fluorescence traces of a pair of PV neurons showing their synchronous action potentials (asterisks). b. Spike cross-correlogram (CCG) of the same pair of neurons in (a). c. Synchrony strength plotted against the distance between cells (n=4,376 pairs, 7 mice). d. Spatial locations (left) and CCGs (right) of three representative PV cell pairs showing varying strength of spike synchrony. e. Example fluorescence traces (the same cells as in a) showing solitary spikes (asterisks) and co-depolarization in the other cell (arrowheads). f. Averaged spike (top) and co-depolarization (bottom) of the two neurons in (e). The double arrow indicates the size of co-depolarization. g. Co-depolarization averaged over all cell pairs. The dashed vertical line indicates the time of the reference spikes. Solid line: mean, shading: s.d. h. Maps of spike synchrony (left) and co-depolarization (right) of a target neuron (shown by a white circle) relative to each of the reference cells. Strength of synchrony and size of co-depolarization was color coded (from blue to red, minimum to maximum) and shown on each reference cell. i. Co-depolarization (top) and spike synchrony (bottom) of the target cell in (h) relative to each of the reference cells in (h). j. Correlation between synchrony and co-depolarization for the target cell shown in (h). Each dot indicates a different reference neuron. Colors correspond to the representative pairs in (d) and (h). k. Correlation between synchrony and co-depolarization measured relative to the same or different reference neurons. For the “Different pairs” condition, synchrony to a given reference neuron was correlated with the co-depolarization triggered by a different reference neuron, located at a similar distance from the target. A correlation value was computed for each target cell (n = 204) that had more than 10 simultaneously imaged reference neurons. The bars show the mean correlation values of all target cells (n=204). Error bars indicate s.e.m. *p<10⁻⁴⁷, Student’s paired t test.