Optical estimation of absolute membrane potential using fluorescence lifetime imaging

^aMeasurements vary too much to be converted to absolute voltage or interpreted across populations of cells. This variability is attributable to numerous confounding factors, including dye loading, photobleaching, and sample movement (Peterka et al., 2011). ^bWhile in principle less variable than a single-color fluorescence intensity measurement, in practice, the signal depends strongly on the loading of two independent lipophilic indicators (Adams and Levin, 2012; Maher et al., 2007), which can vary substantially. ^cANEPPS excitation ratios depend on a variety of non-voltage factors, in particular the membrane composition, leading to substantial artifacts in optical V_mem determinations (Zhang et al., 1998; Gross et al., 1994). ^dWith the GEVI CAESR in our hands, apparently poor protein trafficking produces large amounts of non-voltage-sensitive signal, which contaminates the FLIM recording and contributes to high cell to cell variability (Figure 1—figure supplement 4, Materials and methods). ^ePatch-clamp electrophysiology requires physical contact with the cell of interest, which causes damage to the cell and, in whole cell configurations, washout of intracellular factors. Slight movement of the cell or sample generally result in loss of the patch. ^fMovement of the cell and photobleaching of the dye both cause large changes to the signal over seconds to minutes. ^gRatio-calibrated imaging approaches use a second signal (usually another color of fluorescence) to correct for differences in dye concentration or changes in the region of interest that contaminate single-color intensity signals. If the rate of photobleaching is the same for both components, photobleaching artifacts can also be avoided. ^hLimited by photon count rates. ⁱLimited by probe movement in the membrane, which depends mostly on lipophilicity (Briggman et al., 2010). ^jPhoton counting based lifetime imaging, like epifluorescence intensity imaging, is limited by photon count rates. Large numbers of photons per pixel must be collected to fit TCSPC FLIM data, often using a line scanning confocal approach, leading to slower acquisition speeds than epifluorescence-based intensity imaging. ^kToxicity from capacitive load of the sensor (Briggman et al., 2010). ^lThe spatial resolution of electrophysiology is compromised by space clamp error, preventing interpretation of V_mem in regions far from the electrode (e.g. many neuronal processes) (Williams and Mitchell, 2008; 35,36). ^mAs demonstrated by Cohen and co-workers (Brinks et al., 2015); in our hands with CAESR, we also experienced significant improvements in voltage resolution by fitting a single curve per FLIM image instead of processing the images pixel-wise (see Materials and methods) ⁿIn this work, we calibrated VF-FLIM for V_mem measurements with single cell resolution. In principle, subcellular spatial resolution could be achieved with the VF-FLIM technique.

Fluorescein and erythrosin B standards were measured in drops of solution placed on a coverslip. For VF dyes, voltage sensitivities from intensity-based fluorescence imaging in HEK293T cells (%ΔF/F, percent change in fluorescence intensity for a voltage step from −60 mV to +40 mV) are from previously published work (Woodford et al., 2015). Lifetime data were obtained from voltage-clamp electrophysiology of HEK293T cells loaded with 100 nM VF. Lifetime listed here is the average 0 mV lifetime from the electrophysiology calibration. % Δτ/τ is the percent change in lifetime corresponding to a 100 mV step from −60 mV to +40 mV. Lifetime sample sizes: fluorescein 25, erythrosin B 25, VF2.1.Cl 17, VF2.0.Cl 17. For lifetime standards, each measurement was taken on a separate day. VF2.1.Cl data in HEK293T is duplicated in Figure 2—source data 1. Values are tabulated as mean ± SEM.

Data are compiled from Figure 1 (VF-FLIM, this work), Figure 1—figure supplement 4 (CAESR; Brinks et al., 2015), and Figure 1—figure supplement 5 (Di-8-ANEPPS; Zhang et al., 1998). All data were obtained by simultaneous whole cell voltage clamp electrophysiology and optical recording in HEK293T (VF-FLIM n = 17 cells, CAESR n = 9, di-8-ANEPPS n = 16). Calculation of intra and inter cell accuracies are performed via root-mean-square deviation (RMSD) and discussed in detail in the Methods (see Resolution of VF-FLIM…). Regions of interest were chosen at the plasma membrane in all cases. Di-8-ANEPPS data are presented as the ratio of signal obtained with blue excitation to signal obtained with green excitation (R, see Materials and methods) and are not normalized to the 0 mV R.

Whole-cell voltage-clamp electrophysiology was used to determine the relationship between VF2.1.Cl lifetime and membrane potential in five different cell lines. Parameters of this linear model are listed above. The %Δτ/τ is the percent change in the lifetime observed for a voltage step from −60 mV to +40 mV. The intra-cell RMSD represents the accuracy for quantifying voltage changes in a particular cell (see Materials and methods). The inter-cell RMSD represents the expected variability in single-trial absolute V_mem determinations. Sample sizes: A431 12, CHO 8, HEK293T 17, MCF-7 24, MDA-MB-231 11. All values are tabulated as mean ± SEM.

Comparison of optically-determined resting membrane potential values (in millivolts) and previously reported values. This table summarizes data presented in Figure 3 and Figure 3—figure supplement 1. Optically determined membrane potentials were calculated from lifetime-V_mem standard curves (Figure 2—source data 1). For tabulated literature values, measures of error and central tendency were used from the original publication. In some cases, none were given or only ranges were discussed. The mean of the reported ephys values is the mean of the values listed here. Sample sizes for resting and elevated K⁺, respectively: A431 1056, 368; CHO 2410, 1310; HEK293T 1613, 520; MCF-7 1259, 681; MDA-MB-231 1840, 558.