A general approach to engineer positive-going eFRET voltage indicators

Imaging membrane voltage from genetically defined cells offers the unique ability to report spatial and temporal dynamics of electrical signaling at cellular and circuit levels. Here, we present a general approach to engineer electrochromic fluorescence resonance energy transfer (eFRET) genetically encoded voltage indicators (GEVIs) with positive-going fluorescence response to membrane depolarization through rational manipulation of the native proton transport pathway in microbial rhodopsins. We transform the state-of-the-art eFRET GEVI Voltron into Positron, with kinetics and sensitivity equivalent to Voltron but flipped fluorescence signal polarity. We further apply this general approach to GEVIs containing different voltage sensitive rhodopsin domains and various fluorescent dye and fluorescent protein reporters.

The rhodopsin GEVI optical signal is fast and linear 8 , two desirable features for a voltage indicator.
However, rhodopsin fluorescence is very dim, requiring intense illumination for imaging 8 . Fusions of fluorescent protein (FP) domains or other bright fluorophores to rhodopsin GEVIs were therefore made to facilitate imaging 11,13,14 . These fusions enable voltage-sensitive electrochromic fluorescence resonance energy transfer (eFRET) from a bright fluorophore to the retinal cofactor within the rhodopsin, which acts as a dark quencher. While the absorbance of the retinal cofactor increases with increasing membrane potential, the emission of the eFRET-coupled fluorophore consequently decreases. All reported eFRET GEVIs therefore have negatively sloped fluorescence-voltage relationships; they are brighter at resting membrane potential and become dimmer during an action potential (negative-going). Probes with positively sloped fluorescencevoltage relationships (positive-going) are expected to exhibit higher signal-to-noise ratios due to lower background fluorescence. Although two VSD GEVIs, FlicR 4 and Marina 16 , exhibit positivegoing signals in neurons, they have significantly slower response kinetics than eFRET GEVIs, making detection of action potentials difficult. Here, we present a general approach to engineer eFRET GEVIs with fast, bright, and positive-going fluorescence signals in response to neuronal action potentials by modification of the natural proton transport pathway within microbial rhodopsins.
Previous work fused the Ace2 rhodopsin from Acetabularia acetabulum 17 to the FP mNeonGreen to produce a negative-going eFRET GEVI that allowed in vivo imaging of voltage signals in several model organisms 12 . We recently used the same Ace2 rhodopsin to engineer a negative-going chemigenetic eFRET GEVI called Voltron, which uses a HaloTag protein domain to covalently bind bright and photostable small molecule flurophores 18,19 , extending the duration and number neurons imaged simultaneously in vivo 14 . In both of these GEVIs, photocurrent of Ace2 rhodopsin ( Fig. 1a) is blocked by mutating the residue that normally functions as the proton acceptor (PA) 20 (D81N) (Fig. 1a,b), analogous to the Arch D95N mutation described above. This mutation blocks the primary pathway for exchange of protons from the retinal Schiff base, which links retinal to the rhodopsin protein, to outside the cell 20 . Electrophysiology measurements showing transient inward photocurrents with Ace2 D81N (Fig. 1c) and other rhodopsin based GEVIs 21 suggest that voltage sensitivity in Ace2 D81N and other eFRET GEVIs results from membrane potential changes altering the equilibrium of protonation between the retinal Schiff base, the proton donor (PD) residue 20 , and the cell cytoplasm (Fig 1a,b).
We hypothesized that we could alter the local electrochemical potential of protons on the retinal Schiff base and instead establish a protonation equilibrium with the outside of the cell, which would cause the rhodopsin absorbance and eFRET fluorescence to exhibit the opposite response to membrane potential change. Blocking access of protons to the retinal Schiff base from the cell cytoplasm should enable preferential exchange of protons with the outside of the cell. This hypothesis is supported by a recent report that describes a natural light-driven inward proton pump showing the capacity of microbial rhodopsins to accept protons from outside the cell 22 . To block access of protons from the cytoplasmic side, we substituted the amino acid at the PD position of Voltron for a neutral residue (D92N). As expected, this substitution led to a block of the transient inward photocurrent of Voltron (Fig 1d). Importantly, Voltron D92N (as well as other substitutions to neutral residues at the PD position ( Supplementary Fig. 1)) showed a positivegoing fluorescence signal with membrane depolarization, but with slow kinetics that made it incapable of following neuronal action potentials (Fig 1d).
We reasoned that the proton pathway between the retinal Schiff base of Voltron D92N and the exterior of the cell was inefficient, resulting in the observed slow kinetics. To improve the efficiency of proton movement towards the outside of the cell, we substituted the amino acid at the PA position for a negatively charged aspartate (N81D), as was present in the original Ace2 rhodopsin sequence. This resulted in an indicator (Voltron N81D D92N) that had sufficient response speed to track action potentials in neurons (Fig. 1e). Critically, Voltron N81D D92N exhibited no steady-state photocurrent (Fig. 1e), showing that it can function as a GEVI without pumping protons across the membrane. Voltron N81D D92N had a transient outward photocurrent (Fig. 1e), confirming that the Schiff base proton was now in equilibrium with the outside of the cell. Although Voltron N81D D92N was suitable for monitoring neuronal action potentials with positive-going fluorescence changes, it showed only ~40% of the fluorescence change of the negative-going Voltron (Fig. 1e, g).
To improve the sensitivity of Voltron N81D D92N we focused on the rest of the proton transport pathway of the Ace2 rhodopsin. We reasoned that proton release (PR) residues 20  We previously showed that Voltron was suitable for imaging voltage signals in the brains of live animals such as fruit flies, zebrafish, and mice 14 . To confirm that Positron allows for in vivo imaging, we recorded optical voltage signals from five neurons simultaneously in the forebrain of a larval zebrafish expressing Positron and labeled with JF525 using a widefield fluorescence microscope (Fig. 1j). We observed independent spiking and subthreshold signals in each of the neurons ( Fig. 1j) with fluorescence changes and signal-to-noise comparable to those observed with Voltron imaged similarly in the same preparation (compare Fig. 1j with Fig. S34 of ref. 14).
To demonstrate the generality of our approach to generate eFRET GEVIs with positive-going fluorescence response, we explored different reporter fluorophores and rhodopsin domains.
Positron showed sensitive fluorescence response when labeled with a yellow dye (JF525) or a red dye (JF585) (Fig. 2a). We also showed that the HaloTag could be exchanged for either green or red FP domains (Fig. 2b). We created positive-going versions of the Ace2N-mNeon 12  Positron is the first eFRET GEVI to run in reverse, having lower fluorescence at resting membrane potentials, positive-going signals in response to action potentials, and sensitivity equal to that of state-of-the-art eFRET GEVIs. We achieved this by rational mutation of the PD, PA, and PR sites within the characterized proton transport pathway of microbial rhodopsins and found that the effect of these substitutions can be generalized to other rhodopsins. This work provides further mechanistic insight into the class of rhodopsin eFRET GEVIs and has potential advantages for lower resting fluorescence background and improved signal detectability in densely labeled samples.

Molecular biology
The genes for Ace2 and HaloTag were amplified from a Voltron plasmid 1

Field stimulation in spiking HEK cells or primary neuron culture
A stimulus isolator (A385, World Precision Instruments) with platinum wires was used to deliver field stimuli (50V, 1 ms) to elicit HEK cell spiking or action potentials in cultured neurons as described previously 11 . The stimulation was controlled using Wavesurfer and timing was synchronized with fluorescence acquisition using Wavesurfer and a National Instruments PCIe-

Electrophysiology in primary neuron culture
Filamented glass micropipettes (Sutter Instruments) were pulled to a tip resistance of 4 -6 MW.
Pipettes were positioned with a MPC200 manipulator (Sutter Instruments). Whole cell voltage clamp and current clamp recordings were acquired using an EPC800 amplifier (HEKA), filtered at 10 kHz with the internal Bessel filter, and digitized using a National Instruments PCIe-6353 acquisition board at 20 kHz. Data were acquired from cells with access resistance < 25 MW.
WaveSurfer software was used to generate the various analog and digital waveforms to control the amplifier, camera, light source, and record voltage and current traces. with CsOH, and adjusted to 300 mOsm with sucrose.
For fluorescence voltage curves, cells were held at a potential of -70 mV at the start of each recording and then 1 second voltage steps were applied to step the potential from -110 mV to +50 mV in 20 mV increments. Fluorescence images were acquired at 400 Hz using the same microscope described in the "Microscopy" section above. For determining response speed of indicators, fluorescence images were acquired at 3200 Hz in response to a 100 mV potential step delivered to voltage clamped neurons (from -70 mV to +30 mV). Traces were fit to a double exponential function using MATLAB. All recordings were done at room temperature.

Photocurrent measurements
Photocurrents were recorded at room temperature in voltage-clamp mode with a holding potential of −70 mV in response to 1s light pulses. Photocurrents were recorded using an EPC800 amplifier (HEKA), filtered at 10 kHz with an internal Bessel filter, and digitized using a National Instruments PCIe-6353 acquisition board at 20 kHz controlled using WaveSurfer. Light was delivered to the clamped neurons using the same microscope described above. Irradiance at the imaging plane was set to 70 mW/mm 2 determined with a microscope slide power sensor (S170C, Thorlabs).

Imaging in zebrafish
In vivo wide-field voltage imaging in zebrafish was performed using the procedure described The images were acquired with sCMOS camera (pco.edge 4.2, PCO) at 400 Hz for 1-2 min.