Optical Estimation of Absolute Membrane Potential Using One- and Two-Photon Fluorescence Lifetime Imaging Microscopy

Abstract

Background: Membrane potential (V_mem) exerts physiological influence across a wide range of time and space scales. To study V_mem in these diverse contexts, it is essential to accurately record absolute values of V_mem, rather than solely relative measurements.

Materials and Methods: We use fluorescence lifetime imaging of a small molecule voltage sensitive dye (VF2.1.Cl) to estimate mV values of absolute membrane potential.

Results: We test the consistency of VF2.1.Cl lifetime measurements performed on different single-photon counting instruments and find that they are in striking agreement (differences of <0.5 ps/mV in the slope and <50 ps in the y-intercept). We also demonstrate that VF2.1.Cl lifetime reports absolute V_mem under two-photon (2P) illumination with better than 20 mV of V_mem resolution, a nearly 10-fold improvement over other lifetime-based methods.

Conclusions: We demonstrate that VF-FLIM is a robust and portable metric for V_mem across imaging platforms and under both one-photon and 2P illumination. This work is a critical foundation for application of VF-FLIM to record absolute membrane potential signals in thick tissue.

Introduction

Membrane potential (V_mem), or a voltage arising from ionic concentration gradients across semipermeable membranes, plays diverse and important roles in biological systems. Of particular note is the range of time and length scales over which V_mem can influence cellular and organismal physiology.^1–3 On short timescales, V_mem changes control neuronal communication and cardiomyocyte contraction. However, all cells, even nonelectrically excitable cells, maintain a transmembrane potential, spending an estimated 10–50% of their cellular adenosine triphosphate (ATP) budget to pump ions in opposing directions.⁴ On longer timescales, V_mem displays diverse patterns, including responding to growth factor signals,^5,6 oscillating throughout the cell cycle,⁷ and marking developmental boundaries in tissue.^3,8 Furthermore, these voltage signals can be compartmentalized into small areas such as dendritic spines⁹ or delocalized over larger tissues.¹⁰

Accurate and noninvasive V_mem recording techniques are required to document and understand the many roles of V_mem. On the one hand, patch-clamp electrophysiology, although highly accurate, is damaging to cells and low throughput. On the other hand, recently developed optical V_mem recording techniques using voltage sensitive proteins or small molecules have enabled higher throughput and less invasive recordings, at the cost of voltage resolution. These optical strategies have been particularly successful in reporting on fast changes in V_mem in electrically excitable systems, such as neuron or cardiomyocyte action potentials, but they cannot generally report on an absolute V_mem (in, e.g., millivolts). In addition, the ability to monitor changes in V_mem through two-photon (2P) excitation¹¹ would aid in monitoring V_mem changes and measuring absolute V_mem in model systems of increasing complexity and size.

To address this need, we recently reported VF-FLIM,⁶ a technique for monitoring absolute V_mem across the plasma membrane with the fluorescence lifetime (τ_fl) of the VoltageFluor dye VF2.1.Cl.¹² Fluorescence lifetime is a measure of how long a fluorophore stays in the excited state before emitting a photon. Because it is an intrinsic property of a dye in its environment, fluorescence lifetime can be used to quantitatively make measurements in cells.¹³ In our previously published work, we demonstrated that the photo-induced electron transfer (PeT)-based mechanism of VF2.1.Cl^12,14 leads to a linear response of lifetime with respect to V_mem.

In this article, we expand upon VF-FLIM, showing that our measurements are robust across instruments (two additional time-correlated single-photon counting [TCSPC]-FLIM systems) and that VF-FLIM can be used under one-photon (1P) or 2P illumination. We also describe some considerations for image analysis and provide a code package that we use for analyzing FLIM data of fluorescence associated with membranes (FLIM fluorescence lifetime analysis module, or FLIM-FLAM). To our knowledge, this is the first demonstration of 2P illuminated absolute V_mem recordings with voltage resolution better than 20 mV.

Materials and Methods

Materials

VF2.1.Cl was synthesized in-house according to the published synthesis.¹² VF2.1.Cl was stored as a 1000 × (100 μM) stock in dimethyl sulfoxide at −20°C, and stocks were checked by liquid chromatography-mass spectrometry (LC-MS) every few months to confirm no decomposition had occurred. All chemicals were obtained from either Sigma-Aldrich or Thermo Fisher Scientific.

Cell culture

HEK293T cells were obtained from the UC Berkeley Cell Culture Facility and were verified by short tandem repeat profiling. The experiments in this study did not require IRB approval. Cells were maintained in a humidified 37°C incubator with 5% CO₂ and were discarded after 30 passages. Cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 4.5 g/L glucose, 2 mM GlutaMAX, and 10% fetal bovine serum (FBS). Cells were dissociated with 0.05% trypsin–ethylenediaminetetraacetic acid for passaging and preparation of microscopy samples. FBS was purchased from Seradigm; all other media and supplements were purchased from Gibco (Thermo Fisher Scientific).

Fourteen to 24 hours before electrophysiology recordings, HEK293T cells were dissociated and plated at a density of 21,000 cells/cm² in low-glucose DMEM (1 g/L glucose, 2 mM GlutaMAX, 10% FBS, 1 mM sodium pyruvate) on poly-d-lysine coated 25 mm coverslips (Electron Microscopy Sciences) in a six-well tissue culture plate (Corning). Coverslips were prepared by acid wash in 1 M HCl for 2–5 h, followed by three overnight washes in 100% ethanol and three overnight washes in MilliQ purified water. Coverslips were sterilized by heating to 150°C for 2–5 h. Before addition of cells, coverslips were incubated with poly-d-lysine (made as a 0.1 mg/mL solution in phosphate-buffered saline with 10 mM Na₃BO₄; Sigma-Aldrich) for 1–24 h in a humidified 37°C incubator and washed twice with sterile MilliQ purified water and twice with dPBS.

Microscopy sample preparation

Cells were loaded with 100 nM VF2.1.Cl for 20 min in a humidified 37°C incubator with 5% CO₂ in imaging buffer (IB; pH 7.25; 290 mOsmol/L; composition in mM: NaCl 139.5, KCl 5.33, CaCl₂ 1.26, MgCl₂ 0.49, KH₂PO₄ 0.44, MgSO₄ 0.41, Na₂HPO₄ 0.34, HEPES 10, d-glucose 5.56). Cells were washed once in IB and transferred to fresh IB for electrophysiology.

Whole-cell patch-clamp electrophysiology

Electrodes were pulled from glass capillaries with filament (Sutter Instruments) with a P-97 pipette puller (Sutter Instruments) to resistances of 4–7 MΩ. Electrodes were filled with a K-gluconate internal solution (pH 7.25; 285 mOsmol/L, composition in mM: 125 potassium gluconate, 10 KCl, 5 NaCl, 1 EGTA, 10 HEPES, 2 ATP sodium salt, 0.3 GTP sodium salt). EGTA (tetraacid form) was prepared as a stock solution in 1 M KOH before addition to the internal solution. Voltage steps were corrected for the calculated liquid junction (pClamp software package; Molecular Devices) between IB and the K-gluconate internal.¹⁵

Electrodes were position with an MP-225 micromanipulator (Sutter Instruments) to obtain a gigaseal before breaking into the whole-cell configuration. Recordings were sampled at a rate of >10 kHz using an Axopatch 200B amplifier, filtered with a 5 kHz low-pass Bessel filter, and digitized with a DigiData 1440A (Molecular Devices). Only recordings that maintained a 30:1 ratio of membrane resistance R_m to access resistance R_a were used for analysis. Pipette capacitance was corrected with the fast magnitude knob only; series resistance compensation was not performed. Voltage steps of −80, −40, 0, and +40 mV were applied in random order, followed by a voltage step to +80 mV.

Nikon A1R-HD25 on Ti2E with Becker and Hickl FLIM

Measurements were performed on a Nikon A1R-HD25 confocal on a Ti2E base. Excitation was provided by a 488 nm pulsed diode laser (repetition rate 50 MHz) and directed to the sample with a line pass dichroic. Emission was collected through a 40 × oil immersion objective immersed in Immersol 518F (Zeiss) and was filtered through an additional emission filter (488 nm long pass) before it reached the detector. Individual photons were detected with a hybrid detector (HPM-100-40; Becker and Hickl) and converted to photon arrival times with an SPC-150 photon counting card (Becker and Hickl). Acquisition times of 30 s were used at each V_mem step. A confocal pinhole of ∼2–3 Airy units was used to increase photon counts at the expense of optical sectioning. Fluorescence lifetime images were acquired in Nikon Elements and transferred to SPCImage for fitting of exponential decays. VF2.1.Cl fluorescence lifetime was modeled as a biexponential decay at each pixel of the resulting FLIM image. Data were processed with the following fit parameters fixed: shift 0, offset 0. Images were binned by a factor of 1 before analysis (as supported by SPCImage software, moving average of adjacent pixels). A measured instrument response function (IRF) was used in the fit to a biexponential decay model (see Fluorescence lifetime biexponential fitting below). Regions of interest were identified in FIJI;¹⁶ lifetime is reported as the average of the weighted average τ_m over a region.

Zeiss LSM 880 with PicoQuant FLIM

Data were acquired on an inverted LSM 880 confocal microscope (Carl Zeiss AG, Oberkochen, Germany) equipped with a FLIM upgrade kit (PicoQuant GmbH, Berlin, Germany). The LSM 880 was controlled with Zen Black software (Zeiss); the TCSPC unit was controlled with SymPhoTime 64 software (PicoQuant). A 485 nm diode laser operating at a repetition rate of 40 MHz was used to provide pulsed excitation for FLIM. Emission was collected through a 40 × oil immersion objective immersed in Immersol 518F (Zeiss) and was filtered through an additional bandpass emission filter (550/49 nm; Semrock) before it reached the detector. Single photons were detected with a PMA-Hybrid 40 detector (PicoQuant) and processed with a TH260 Pico Dual TCSPC Unit (T3 TCSPC mode). Data were acquired at an approximate frame rate of 1 Hz; successive frames were binned before analysis to obtain a total acquisition of 5 s (5–6 frames per V_mem step). A confocal pinhole of ∼2–3 Airy units was used to increase photon counts at the expense of optical sectioning.

Fluorescence lifetime analysis was performed with global analysis in SymPhoTime (PicoQuant). All photons from a defined ROI per frame were combined for a global fit (see Supplementary Appendix SA1 for further explanation). An experimentally measured IRF was used for fitting to a biexponential model (see Fluorescence lifetime biexponential fitting below). Shift was fixed to 0, and offset was determined by the software from the baseline. Lifetimes of successive frames at the same V_mem were similar. Data presented here were averaged across all frames at a given potential for each cell before lines of best fit were determined.

2P fluorescence lifetime imaging

2P-FLIM was performed on an inverted Zeiss LSM 510 equipped with a Becker and Hickl SPC-150N photon counting card. Excitation was provided by a MaiTai HP Ti:Sapphire laser tuned to 820 nm with a repetition rate of 80 MHz. Excitation power was controlled by a half-wave plate followed by a polarizing beamsplitter, as well as an acousto-optic modulator controlled by the Zen software (Zeiss). Light was directed into the microscope from a series of silver mirrors (Thorlabs or Newport Corp.).

Emitted photons were collected with a 40 × /1.3 NA EC Plan-Neofluar oil immersion objective (Zeiss) and detected with an HPM-100-40 single-photon counting detector (Becker and Hickl). Photons were detected from the nondescanned side port of the LSM 510 after being reflected off of a 680 nm long pass dichroic mirror and passing through an IR-blocking filter. Emission light was further filtered through a bandpass emission filter (550/49 nm; Semrock). Reference pulses for time-correlated single-photon timing were sampled from the excitation beam and detected with PHD-400-N high speed photodiode (Becker and Hickl).

2P-FLIM data for VF2.1.Cl were fit to an incomplete biexponential decay model in SPCImage (Becker and Hickl) as described hereunder.⁶ Using the standard solutions described hereunder, the optimal value for the color shift between the IRF and the measured decay was determined to be 0.7 (of 256 time channels on the analog-to-digital converter). The photon count threshold for fitting was set to 200 counts in the brightest time channel. Other fitting parameters and binning were as described for the published 1P signal.^6,24 Lifetime images in lifetime-intensity overlays are scaled as indicated; underlying photon count images are scaled to maximize contrast.

Measurement of IRF

For all instruments, the IRF was determined experimentally from a solution of 0.5 mM fluorescein and saturating (12.2 M) NaI in 0.1 N NaOH. The IRF was measured frequently (at a minimum daily) during imaging. The IRF was cropped to only include the time bins with the major pulse (∼10 bins in the analog-to-digital converter space). Standard solutions of 2 μM fluorescein in 0.1 NaOH and 1 mg/mL erythrosin B²⁵ were used to assess quality of lifetime determinations and to refine fit parameters.

Fluorescence lifetime biexponential fitting

The fluorescence lifetime of VF2.1.Cl was modeled as a biexponential decay [Eq. (1)] before convolution with the experimentally measured IRF. Lifetimes are reported as the weighted average τ_m [Eq. (2)], consistent with the reported analysis for VF-FLIM.⁶

where I is the time resolved fluorescence intensity as a function of time t, a₁ and a₂ are the amplitudes of each exponential component, and τ₁ and τ₂ are the decay constants (lifetime) of each exponential component. We used both commercial software and custom Matlab code to perform these analyses. The custom Matlab (MathWorks) code is available on GitHub and documented further in Supplementary Appendices SA1 and SA2.

We use τ_m as a proxy for voltage because it can be more precisely determined with the relatively low photon counts that are typical of biological samples. Further commentary on this is provided in Supplementary Appendix A1. For calibration of τ_m versus V_mem, we performed a linear fit using Microsoft Excel or Matlab. We quantified the variability measured between calibration samples using intra- and intercell variability as described previously.⁶ In brief, we calculated the root-mean-square deviation (RMSD) between our electrophysiologically determined V_mem and the calculated V_mem from the lifetime. The “intra” cell variability reports on the error one might expect in successive VF-FLIM measurements on the same cell (e.g., during a time course measurement), and the “inter” cell variability reports on the error one might expect in comparing V_mem of two different cells using VF-FLIM (e.g., comparing an experimental and control cell populations).

Results

1P VF-FLIM is consistent across different photon counting instruments

To evaluate the portability of the VF-FLIM technique to other laboratories, we developed a standardized measurement scheme for calibrating lifetime with respect to V_mem. Whole-cell voltage clamp electrophysiology allows us to set V_mem of a cell of interest to a defined value. If fluorescence lifetime is measured during these electrophysiological steps, a lifetime-V_mem curve can be constructed, as described previously.⁶ We selected this experiment as a comparison point across systems because it is both precise and relatively easy to perform on accessible samples (HEK293T cells).

We measured the lifetime-V_mem relationship for VF2.1.Cl on two independent TCSPC instruments (Fig. 1 and Table 1). These two TCSPC systems were both distinct from the instrument used in the previously published work.⁶ All instruments in this comparison were both point-scanning confocal instruments with excitation at ∼488 nm. However, there were considerable differences among the instruments: each was equipped with a distinct microscope bodies, light sources, detectors, and photon counting modules. Furthermore, the analysis was performed in two different commercial software packages and with different binning schemes. Additional instrumentation details are provided in the Materials and Methods section.

FIG. 1.

Lifetime-V_mem calibrations for VF2.1.Cl on different TCSPC systems. HEK293T cells were held at specified V_mem with whole-cell voltage clamp electrophysiology while lifetime was measured on (a) a Nikon confocal equipped with a Becker and Hickl FLIM system and (b) a Zeiss LSM 880 confocal equipped with a PicoQuant FLIM system. Gray lines indicate lines of best fit for individual cells; black line indicates the average line of best fit. Data are shown as mean ± SEM. TCSPC, time-correlated single-photon counting.

Table 1.

Summarized Lifetime-V_mem Calibration for VF2.1.Cl in HEK293T on Different Time-Correlated Single-Photon Counting Systems

System	Slope (ps/mV)	y-Intercept (ps)	Intercell RMSD (mV)	Intracell RMSD (mV)
VF-FLIM6	3.50 ± 0.08	1.77 ± 0.02	19	3.5 ± 0.4
Zeiss/PQ	3.43 ± 0.08	1.75 ± 0.03	19	3.1 ± 0.6
Nikon/BH	3.18 ± 0.03	1.80 ± 0.02	14	3.4 ± 0.5

Summary of data presented in Figure 1. Slope, y-intercept, and intracell RMSD values were averaged across all cells measured on a particular instrument and are presented as mean ± SEM. Data represent the following numbers of cells: VF-FLIM 17 (data from previously published study),⁶ Zeiss/PicoQuant (PQ) 6, Nikon/Becker-Hickl (BH) 7.

RMSD, root-mean-square deviation.

We observed strikingly similar lifetime results across all three systems in this comparison. The slopes, y-intercepts, and spreads among the different calibrations are nearly identical (Table 1). Furthermore, the voltage resolution obtained on all three systems is similar. To quantify this, we calculated “intracell” and “intercell” accuracy in the calibration, as described previously.⁶ On the one hand, intracell variability captures the expected error in successive measurements on the same cell (e.g., during a longitudinal time course measurement). On the other hand, intercell variability captures the expected error in measurements of resting membrane potential differences between different cells (e.g., experimental and control cell populations). Variability inherent in the VF-FLIM measurement imposes a lower limit on the size of V_mem differences that can be reliably quantified. Lower values of inter- and intracell error allow detection of smaller V_mem changes; if the inter-/intracell error values are too high, the technique will suffer from false negatives, failing to identify voltage differences when they are present. Upon calculating the inter and intracell error, we find that all three instruments provide similar accuracies in lifetime-based V_mem determination. Of note, the intracell resolution achieved with VF-FLIM (3.1–3.5 mV) is ∼6-fold better than what we could achieve using a ratio-based approach with di-8-ANEPPS (18 mV resolution), and ∼10-fold better than what we could achieve using lifetime measurements with the genetically encoded citrine-Arch electrochromic Förster resonance energy transfer sensor (CAESR,^11,17 33 mV resolution). The intercell resolution of VF-FLIM (14-19 mV) also compares favorably (10- to 20-fold improvement) to ratio-based techniques (150 mV resolution) or lifetime with CAESR (370 mV resolution).⁶

VF2.1.Cl fluorescence lifetime reports voltage under 2P excitation

We next sought to determine whether the fluorescence lifetime of VF2.1.Cl under 2P excitation could be used to make determinations of absolute membrane potential. For these measurements, we used an FLIM instrument similar to the one described previously,⁶ but with an altered beampath to enable 2P excitation (see Materials and Methods section). We selected an excitation wavelength of 820 nm, near the previously reported 2P absorption maximum of fluorescein.¹⁸ Excellent membrane staining and optical sectioning was observed (Fig. 2a–d). Because of the relatively poor 2P cross-section of VF2.1.Cl,¹⁹ high light power was required to obtain sufficient photon output. As such, photobleaching appeared more rapid under these conditions than under 1P excitation with similar photon output (data not shown). Nevertheless, it was still possible to obtain high-quality patches, and nonpatched cells were relatively unperturbed by the light exposure (Fig. 2c, d).

FIG. 2.

VF2.1.Cl fluorescence lifetime reports absolute V_mem under 2P excitation. Simultaneous whole-cell patch-clamp electrophysiology and 2P FLIM (820 nm excitation) of VF2.1.Cl in HEK293T reveals a V_mem-sensitive τ_fl. (a, c) Photon count images of 100 nM VF2.1.Cl in two sample images of HEK293T cells held at the indicated V_mem. White arrowheads indicates the patched cell; scale bar is 20 μm. (b, d) Lifetime-intensity overlay of the corresponding cells from (a) or (c). (e) Quantification of the lifetime images in the above panels, revealing a clear linear relationship between τ_fl and V_mem. Points indicate average lifetime for the patched cell in (b) or (d) within a membrane localized region of interest at each potential. Black line indicates the line of best fit. (f) Aggregated lifetime-V_mem calibration for VF2.1.Cl, including the data in (e), where individual cells are represented by gray lines. Black line represents the average slope and y-intercept for n = 16 HEK293T cells. Data are shown as mean ± SEM. 2P, two photon.

Using simultaneous fluorescence lifetime imaging and whole-cell patch-clamp electrophysiology as described previously, we determined that the τ_fl of VF2.1.Cl under these excitation conditions is sensitive to V_mem. With 820 nm 2P excitation, we observe a sensitivity of 3.86 ± 0.04 ps/mV with a 0 mV lifetime (y-intercept) of 1.95 ± 0.01 ns (n = 16 cells; Fig. 2). These results are qualitatively similar to the 1P calibrations (Table 1), but we do observe consistently larger slopes and y-intercept values under 2P illumination. Using this 2P calibration, we observed similar or slightly improved V_mem resolution to the 1P calibration (14 mV intercell RMSD and 4.4 ± 0.6 mV intracell RMSD for the 2P set-up, calculated as in Table 1).⁶

Discussion

To broaden the reach of the VF-FLIM⁶ technique and standardize its use, we compared the fluorescence lifetime of VF2.1.Cl in HEK293T cells under whole-cell voltage clamp across entirely different TCSPC systems, as well as under 1P and 2P illumination. We find that the results are similar across different lifetime instruments and configurations, consistent with the intrinsic nature of fluorescence lifetime. This observation highlights a key advantage of fluorescence lifetime as opposed to fluorescence intensity, namely its relative independence from illumination intensity and dye concentration-related artifacts. In addition to demonstrating the robustness of VF-FLIM, the expansion of the technique to 2P illumination opens up the possibility of measuring V_mem in thick tissue, enabling studies of V_mem during more complex developmental processes.

VF-FLIM is reproducible across lifetime instruments

We compared the results of a standardized electrophysiology experiment (five voltage steps in HEK293T cells) across three TCSPC FLIM microscopes (one from published data and two additional microscopes). We observe excellent agreement between these systems, suggesting that a calibration obtained on one TCSPC FLIM system could be used on a different system with minimal recalibration. Taken together, our results demonstrate that VF2.1.Cl lifetime is a robust proxy for membrane potential across a variety of TCSPC configurations.

Nevertheless, there were slight differences in calibration measured between the instruments, so the most accurate V_mem determinations would require electrode-based calibration on the system to be used for data collection. These slight discrepancies could result from a variety of factors, including differences in temperature, stray light, IRF stability, or simply random variation between cells. We note that the slightly reduced intercell RMSD on the Nikon instrument is difficult to interpret. These data were collected over a much shorter timeframe than the previously published results (∼2 weeks vs. 1.5 years), so the effects of any instrument-related drift would be diminished.

Along with these data, we have also provided extensive commentary on the processing of VF-FLIM data (Supplementary Appendices SA1 and SA2), together with a custom Matlab codebase for data analysis. We have observed that incorrect selection of the fitting model and parameters can introduce sizable artifacts into the lifetime results, often much larger than the differences we observe between microscopes (Table 1). We hope that our added notes will facilitate the broader use of VF-FLIM for absolute membrane potential determinations.

Because of limitations in our access to diverse FLIM instrumentation, we have not evaluated the performance of VF-FLIM in a frequency domain (FD) configuration. Exploration of FD FLIM would offer the opportunity to make considerably faster lifetime recordings in a widefield imaging configuration. Because many interesting V_mem dynamics occur on shorter timescales, investigation of VF-FLIM in combination with fast FD techniques such as siFLIM²⁰ are of considerable future interest.

2P absolute voltage determinations with VF2.1.Cl

We determined that the fluorescence lifetime of VF2.1.Cl under 2P illumination is sensitive to the transmembrane potential. This observation is a critical step forward in 2P voltage imaging, as translation between 1P and 2P illumination is not always straightforward for V_mem indicators. For example, a comparative study of genetically encoded voltage indicators revealed that many of them did not display sufficient brightness or sensitivity to record V_mem under 2P illumination.¹¹ However, we were optimistic that the PeT mechanism of the VoltageFluor dye VF2.1.Cl would be translatable to 2P lifetime imaging, based on previous work using VoltageFluor in 2P intensity-based imaging.^19,21

Although it is not strictly required, fluorophore emission spectra and lifetime are generally similar under 1P and 2P excitation.²² Consistent with this, we observed good agreement between the properties of VF2.1.Cl under 1P and 2P illumination. The slight increase in the y-intercept (0 mV lifetime) is most likely attributable to reduced contribution of cellular autofluorescence when exciting at 820 nm (2P) versus 479 nm (1P). This hypothesis is consistent with the slightly increased sensitivity under 2P illumination, as autofluorescence will contribute a V_mem-insensitive background to the lifetime recording. This reduction in autofluorescence could also produce the observed slight resolution improvement for V_mem recordings (intercell variability of 14 mV in 2P vs. 19 mV in 1P). Nevertheless, these data were collected over the course of a much smaller time window than the 1P data in our previous study (1 month vs. 1.5 years). Slight differences in instrument performance over time may also contribute to this 1P versus 2P resolution difference.

Our work demonstrates that the fluorescence lifetime of VF2.1.Cl under 2P excitation could be a useful method for measuring absolute V_mem, as it displays excellent V_mem sensitivity and resolution. However, a key limitation of the work presented here is the poor 2P cross-section of the dichlorofluorescein chromophore in VF2.1.Cl. Expansion of VF-FLIM to additional VoltageFluors and colors,²³ including those with improved 2P cross-sections,^19,21 will likely lead to considerable improvement in performance. Together, we hope these technologies can begin to address important questions of V_mem signaling over long timescales and in thick tissue.

Conclusion

We demonstrate that the fluorescence lifetime of the VoltageFluor dye VF2.1.Cl (VF-FLIM) is a reproducible and robust measure of absolute membrane potential across different instruments and illumination modes (1P and 2P). We also offer extensive documentation of our data processing routines in the hopes that it will facilitate more widespread adoption of the VF-FLIM technique. We are optimistic that the flexibility and robustness of VF-FLIM will enable its application to study the diverse biological roles of membrane potential across space and time.

Authors' Contributions

J.R.L.-D. performed experiments, analyzed data, wrote, and edited the article. E.W.M. conceived the study, analyzed data, wrote, and edited the article. All authors have approved the final version of this article. This article is not under consideration at any other peer-reviewed journal.

Author Disclosure Statement

E.W.M. is listed as an inventor on a patent describing voltage-sensitive fluorophores. This patent (us20170315059) is owned by the Regents of the University of California. E.W.M. and J.R.L.-D. declare that no other competing interests exist.

Funding Information

A subset of FLIM experiments were performed on “Deckard” at the UC Berkeley CRL Molecular Imaging Center, supported in part by NSF DBI-0116016.

Supplementary Material

Supplementary Appendix SA1

Supplementary Appendix SA2