Palette of fluorinated voltage-sensitive hemicyanine dyes

Abstract

Optical recording of membrane potential permits spatially resolved measurement of electrical activity in subcellular regions of single cells, which would be inaccessible to electrodes, and imaging of spatiotemporal patterns of action potential propagation in excitable tissues, such as the brain or heart. However, the available voltage-sensitive dyes (VSDs) are not always spectrally compatible with newly available optical technologies for sensing or manipulating the physiological state of a system. Here, we describe a series of 19 fluorinated VSDs based on the hemicyanine class of chromophores. Strategic placement of the fluorine atoms on the chromophores can result in either blue or red shifts in the absorbance and emission spectra. The range of one-photon excitation wavelengths afforded by these new VSDs spans 440–670 nm; the two-photon excitation range is 900–1,340 nm. The emission of each VSD is shifted by at least 100 nm to the red of its one-photon excitation spectrum. The set of VSDs, thus, affords an extended toolkit for optical recording to match a broad range of experimental requirements. We show the sensitivity to voltage and the photostability of the new VSDs in a series of experimental preparations ranging in scale from single dendritic spines to whole heart. Among the advances shown in these applications are simultaneous recording of voltage and calcium in single dendritic spines and optical electrophysiology recordings using two-photon excitation above 1,100 nm.

Optical recording techniques provide powerful tools for neurobiologists (1) and cardiac physiologists (2) to study detailed patterns of electrical activity over time and space in cells, tissues, and organs. Rational design methods, based on molecular orbital calculations of the dye chromophores and characterization of their binding and orientations in membranes (3–5), were used to engineer dye structures. The general class of dye chromophores called hemicyanine (also referred to as styryl dyes) has emerged from this effort as a good foundation for voltage-sensitive dyes (VSDs), because they exhibit electrochromism. This mechanism, also referred to as the molecular Stark effect, involves the differential interaction of the electric field in the membrane with the ground and excited states of the dye chromophore. Several important hemicyanine dyes were produced over the years, including di-4-ANEPPS (6, 7), di-8-ANEPPS (8), di-2-ANEPEQ (also known as JPW-1114) (9, 10), RH-421 and RH-795 (11), ANNINE-6 and ANNINE-6+ (12, 13), di-3-ANEPPDHQ (14, 15), di-4-ANBDQBS, and di-4-ANBDQPQ (16, 17). Because the electrochromic mechanism is a direct interaction of the electric field with the chromophore and does not require any movement of the dye molecule, all of these dyes provide rapid absorbance and fluorescence responses to membrane potential (V_m); they are, therefore, capable of recording action potentials (APs). Other mechanisms can give more sensitive voltage responses in specialized applications (18–22). Additionally, new fluorescent protein-based voltage sensors are being developed (23–26), with the promise of being able to genetically target specific cells in an organism. However, to date, hemicyanine dyes are the most universally used and applicable VSDs available. Indeed, among the most recent new advances enabled by these dyes, there has been recording of deep aberrant activity patterns in human hearts from transplant patients (27, 28) and recording V_m spikes (∼1 ms) from individual dendritic spines (29–31).

In this work, we combine instrumentation and experimental protocols to characterize fluorinated hemicyanine VSDs. They are shown in voltage-sensing applications in cardiac and neuronal systems. These systems include random access two-photon imaging of AP propagation in cerebellum; two-photon recording of back-propagating APs (bAPs) from individual dendritic spines in mouse cortical brain slices; tests of long-wavelength two-photon voltage sensing in cultured hippocampal neurons; single-sweep one-photon imaging of dendritic bAPs in cortical neurons; and multiwavelength imaging of V_m and intracellular free calcium ([Ca²⁺]_i) in whole perfused guinea pig heart. These applications are enabled by a series of fluorinated hemicyanine dyes that collectively sample a broad spectral range. Appropriate placement of the fluorine substituents on the dye chromophore allows the excitation and emission spectra of the dyes to be finely tuned. This tuning permits one to choose dyes for optimal sensitivity in relation to the linear or nonlinear excitation sources appropriate to the experimental system as well as minimization of spectral overlap with other fluorescent sensors or optical manipulations. Furthermore, fluorination is known to enhance the photostability of dye chromophores (32, 33). For measurements that may be limited by low-dye copy numbers in small volumes and/or the need to record rapid transients, better photostability permits the collection of more photons per VSD molecule, thus improving the signal to noise ratio (S:N) of the optically recorded voltage activity.

Results

Fluorinated Hemicyanine VSDs.

We have synthesized a series of 19 VSDs with various fluorine substitution patterns on three hemicyanine backbone chromophores. The resulting VSDs are shown in Table 1 along with their spectral data when bound to the lipid membranes of sonicated soybean phosphatidylcholine vesicles. Also, Table 1 shows the corresponding unsubstituted dyes for comparison. The hemicyanine dye chromophores are composed of an electron-donating aromatic amine, a π-linker region, and a heterocyclic electron acceptor (17). The central portions of the names in Table 1 comprise a short code for identifying the structure of the chromophore. The donor moiety for all of these VSDs is 6-amino-naphth-2-yl (AN in the name of the VSD); the linker group is either ethene (E) or a butadiene (BD), and the acceptor is either pyridinium (P) or quinolinium (Q). Thus, the first chromophore in Table 1 has an ANEP chromophore (aminonaphthyl ethene pyridinium). A pair of alkyl groups on the amino end anchors the VSD to the membrane, each having two, three, or four carbons that are identified by the prefixes as di-2-, di-3-, or di-4-, respectively. The positions of fluorine substituents are identified by (F), (F2), or (CF3) placed immediately after the moiety on which the substituent has been placed. Finally, all of these VSDs have a 1-prop-3-yl (triethylammonium) head group to impart water solubility and inhibit dye flipping across the membrane; this group is designated by the suffix PTEA.

Table 1.

Fluorinated VSDs

As can be seen, substitution of fluorine at the aminonaphthyl end of the chromophore results in blue shifts of the emission and absorbance spectra; fluorine substitution within the linker region or the heterocyclic rings (i.e., pyridinium or quinolinium) results in significant red shifts of the spectra. In general, we have prepared both N,N-diethylamino- and N,N-dibutylamino- VSDs. VSDs with n = 2 are more water-soluble, which improves tissue penetration or intracellular spread from a pipette. VSDs with n = 4 are more tightly bound to cell membranes and therefore, more persistent in long-term experiments. All of these VSDs are essentially nonfluorescent in aqueous solution but highly fluorescent when bound to lipid membranes. They have Stokes shifts (difference between the absorbance and emission wavelength maxima) of >150 nm, facilitating their use combined with other fluorescent probes. Also, as detailed in the example applications below, fluorination imparts more photostability than previous generations of hemicyanine dyes. Most important, of course, is their sensitivity to fast voltage changes, which will also be shown by diverse sample applications below.

Recording bAP from Individual Dendrites and Spines in a Cortical Brain Slice.

To optimize the voltage sensitivity of the fluorescent signal (ΔF/F), it is important to excite the hemicyanine dyes at the edge of their spectra, where the change in fluorescence (ΔF) is large but baseline (F) is low (13, 34). The mode-locked Ti-Sapphire lasers most commonly used for two-photon excitation can be readily tuned up to ca. 1,060 nm, but tuning to higher wavelengths, which would be optimal for the nonfluorinated ANEP dyes, results in a rapid decline in laser power. Mode-locked single-wavelength fiber lasers are less costly than tunable Ti-Sapphire lasers and available for excitation at 1,064 nm (35). Therefore, our first syntheses of fluorinated dyes aimed to tune the dye absorbance spectra to place 1,060 nm at the red edge. This tuning produced di-2-AN(F)EPPTEA and di-4-AN(F)EPPTEA, VSDs with fluorine substitution on the naphthalene to elicit a blue shift; these dyes are the only dyes in Table 1 that have been previously reported (29, 36). The initial report using di-2-AN(F)EPPTEA was able to show high-quality recordings of bAP from single dendritic spines on a dye-filled pyramidal neuron in a mouse brain slice (29). In Fig. S1, we extend this demonstration with a two-photon excited fluorescence record of 5 bAPs evoked by repeated current injection at the soma. Using 1,060-nm two-photon excitation, we observe optical bAP waveforms with amplitudes in excess of 17% change in fluorescence. To capture the entire spike train, an extended recording duration of 160 ms was used. Fig. S1 shows optical recording can follow rapid, repeated AP firing in spines. Additionally, V_m responses to hyperpolarizing current injections can be recorded in a single spine.

It is important to emphasize, however, that the wavelength dependence of ΔF/F is shallow, which allows some flexibility in the choice of excitation wavelength in response to special experimental demands. We sought to simultaneously record voltage and [Ca²⁺]_i in single spines by using di-2-AN(F)EPPTEA combined with a fluorescent calcium indicator. This combination was achieved with Calcium Green-1 (conjugated to 3-kDa dextran), which has a one-photon absorbance maximum at 506 nm and emission at 531 nm. The results are shown in Fig. 1. We chose 1,020-nm excitation to excite both fluorescent indicators simultaneously. Xanthene dyes such as Calcium Green have their best two-photon absorbance cross-section into the second electronic excited state at about 800 nm; we reasoned, however, that we could achieve sufficient two-photon excitation directly into the lowest excited state at 1,020 nm, which would allow for simultaneous voltage-sensitive excitation of di-4-AN(F)EPPTEA. The large Stokes shift of the VSD permitted us to readily separate the [Ca²⁺]_i and voltage signals with appropriate emission filters. Fig. 1 shows a characteristically slower time course for onset and recovery of the calcium response to bAPs compared with the directly measure optical bAP. Although the ΔF/F for the VSD is somewhat diminished at 1,020 nm (15% compared with 17% at 1,060 nm in Fig. S1), the S:N of the measurement is still sufficient to readily detect a bAP in a single sweep. Thus we achieved a simultaneous recording of voltage and [Ca²⁺]_i in dendritic spines.

Fig. 1.

Simultaneous voltage and calcium imaging in a single dendritic spine. (A) Reconstruction of the segment of proximal basal dendrite studied from a mouse acute cortical brain slice indicating the targeted spine. (B) Calcium Green 1 Dextran fluorescence in response to two somatically initiated bAPs. (C) Simultaneously recorded di-2-AN(F)EPPTEA VSD fluorescence transients. Both dyes are excited with the same two-photon laser tuned to 1,020 nm. Waveforms are averages of 10 trials, with bleaching subtracted. (D) Single-sweep VSD fluorescence. The vertical scale bars in B and C correspond to ΔF/F.

In Fig. S2, we use one-photon excitation with a sensitive fast camera to show the ability of two longer-wavelength fluorinated VSDs to record bAPs in dendritic segments of cortical pyramidal neurons; di-2-ANEP(F)PTEA and di-2-ANEP(F2)PTEA have single and double substitutions, respectively, of fluorine at the pyridinium heterocycle to produce a red shift of the hemicyanine chromophore. As above, these dyes were applied intracellularly through a patch pipette, and therefore, they required high solubility in aqueous solution. For both VSDs, the results in Fig. S2 were obtained with excitation centered at 510 nm. As can be seen, the ΔF/F for both of these experiments is significantly lower than the spine recording in Fig. 1 or Fig. S1. This finding is not likely to be caused by significantly lower sensitivity of these dyes. Rather, lower ΔF/F can be attributed to a higher total fluorescence caused by internal membrane staining in dendrites compared with spines as previously reported (29); ΔF/F would also be greater if the excitation wavelength was shifted farther to the red edge of each dye spectrum. However, a range of wavelengths can be used for a given VSD without too much impact on ΔF/F; because these dyes have greater fluorescence at 510 nm and a larger area is illuminated, the S:N in Fig. S2 is improved compared with the two-photon records in Fig. S1. This result occurs, because at these low-light levels, S:N is largely determined by the Poisson statistics of the emitted photons: sampling over an ∼100-μm² segment of dendrite membrane will deliver more photons than the ∼3-μm² surface of a typical single spine. Additional S:N considerations are described in Long-Wavelength Two-Photon Measurement of V_m in Cultured Hippocampal Neurons.

Spatiotemporal Mapping of Cardiac Tissue Physiology.

We have already shown the efficacy of di-4-AN(F)EPPTEA to record APs from multiple subcellular sites on single cardiomyocytes (36); in this section, we show the efficacy of other fluorinated VSDs in a whole-heart preparation. For illustrative purposes, three spectrally distinct dyes (blue, green, and red excitation wavelengths) were used in three common isolated whole-heart optical mapping applications: (i) sinus rhythm (di-4-AN(CF3)E(F)PPTEA), (ii) ventricular fibrillation (di-4-AN(F)EP(F)PTEA), and (iii) simultaneous voltage and [Ca²⁺]_i imaging (di-4-ANEQ(F)PTEA). Fig. S3A shows a schematic of the imaging system used for all three applications (2, 37). Fig. S3 B–D show AP recordings from the ventricles in initial experiments characterizing dyes di-4-AN(CF3)E(F)PPTEA, di-4-AN(F)EP(F)PTEA, and di-4-ANEQ(F)PTEA, respectively, in guinea pig hearts in sinus rhythm. Washout kinetics and photobleaching rate were also compared between spectrally similar di-4-ANEPPS and di-4-AN(F)EP(F)PTEA (Fig. S4). This finding shows the improved properties of these VSDs. Their ΔF/F and S:N are as good as standard VSDs used today (2, 38). For di-4-ANEQ(F)PTEA (Fig. S3D), the dye was also exposed to an excitation source for the low-affinity [Ca²⁺]_I reporter fura-4F (AM). A multiband emission filter collecting emission fluorescence for both fura-4F (AM) and di-4-ANEQ(F)PTEA was used (F3) (Fig. S3A), showing that no cross-talk will occur in tissue additionally loaded with fura-4F (AM).

Fig. 2 shows activation waves on whole-heart ventricles (mostly left ventricle view) for the three VSDs. Fig. 2A shows AP progression during sinus rhythm. Fig. 2B shows sequential snapshots of chaotic electrical activity during ventricular fibrillation caused by bursts of rapid electrical pacing. Fig. 2C shows simultaneous voltage and [Ca²⁺]_i imaging with the heart paced at the apex and coloaded with di-4-ANEQ(F)PTEA and fura-4F (AM), revealing the well-established delay between V_m and [Ca²⁺]_i peaks.

Fig. 2.

Guinea pig heart optical mapping. (A) Normalized fluorescence intensity maps of V_m at six progressive time points during sinus rhythm. (B) Normalized fluorescence intensity maps of V_m at seven progressive time points during ventricular fibrillation. (C) Normalized fluorescence intensity maps of V_m and [Ca²⁺]_i at four progressive time points during local 4-Hz electrical stimulation of the apex (at the site of the white circle). Raw (unfiltered) V_m and [Ca²⁺]_i signals, on a 16-bit scale, are shown from a 4 × 4-pixel region on the ventricle. The color bar is shown at the bottom right for all normalized fluorescence intensity maps. Fluorescent indicators, excitation wavelengths, and time interval between frames are indicated below each image series. (Scale bar: 5 mm.)

Random Access Two-Photon Recording of Electrical Activity in Cerebellum.

We bathed a slice of cerebellum in di-4-AN(F)EPPTEA and were able to record spontaneous AP spikes from multiple Purkinje cells using two-photon excitation with a 1,064-nm mode-locked fiber laser (Fig. 3). This recording is achieved by rapidly positioning the laser excitation with an acousto-optic modulator to sample a patch of membrane from each cell in <100 μs; for recording from five cells as in Fig. 3A, multiplexing permits a temporal resolution of ∼400 μs—sufficient to capture every spike. Fig. 3B shows spontaneous activity recorded in this manner over 800 ms from the five neighboring cells, showing that spiking is not temporally correlated. Simultaneous optical and electrical recording from PC₁ shows the high fidelity of the optical measurement. The expanded trace for PC₅ in Fig. 3B reveals the characteristic after hyperpolarization and shows the high temporal resolution of the random access microscope. Fig. 3 C–E shows that the dye and the optical recording protocol do not produce any significant photodamage to the preparation.

Fig. 3.

Real-time multicellular AP recording by random access multiphoton microscopy. (A) Two-photon fluorescence image (taken at a depth of 70 μm) of a parasagittal acute cerebellar slice stained with di-4-AN(F)EPPTEA. The molecular (ML) and granular (GL) layers are clearly distinguishable. The multiunit optical recording was carried out from the lines drawn (red) on the five Purkinje cells (PCs). The electrical activity (cell-attached recording) of PC1 was also monitored. (B) Real-time multiplexed optical recording of spontaneous activities from the five PCs (black traces) with a temporal resolution ∼400 μs. PC1 electrical activity measured by the electrode (blue trace) shows the reliability of the optical recording in spike detection. A simple spike in PC5 trace (red box) is temporally magnified, revealing the undershoot phase. (C) Electrical recording of spontaneous activity after the multiline scans session. A central region of the trace is temporally magnified (6×) in the dashed red box. (D and E) The interspike interval (ISI) and the coefficient of variation (CV) measured before (CTRL) and after (AR) the optical recording session are not statistically different, indicating negligible photodamage induced by di-4-AN(F)EPPTEA.

Long-Wavelength Two-Photon Measurement of V_m in Cultured Hippocampal Neurons.

The common near-infrared femtosecond laser sources used for two-photon microscopy have wavelength ranges from ca. 680 to 1,080 nm. This range restricts the range of fluorescent probes for two-photon microscopy to chromophores with absorbance spectra below ∼540 nm. For hemicyanine VSDs, where red-edge excitation is optimal, the best dyes would have absorbance maxima at or below ∼470 nm, which is the case for di-3-ANEPPDHQ (15), di-2-AN(F)EPPTEA (Fig. 1 and Fig. S1), and di-4-AN(F)EPPTEA (Fig. 3). However, this finding excludes most of the chromophores in Table 1 from two-photon applications. In particular, two-photon applications of a VSD combined with optogenetic manipulation or many [Ca²⁺]_i indicators would be facilitated if higher wavelength could be used for the VSD. We, therefore, decided to test several of the long-wavelength fluorinated VSDs by exciting them with light in the range of 1,100–1,300 nm using a microscope in which the Ti-sapphire laser was coupled to an optical parametric oscillator.

We applied the VSDs externally to cultured dissociated hippocampal neurons to obtain a closer estimate of the intrinsic V_m sensitivity of the VSDs (Fig. 4A), and in a first set of experiments, we acquired the fluorescence by rapid line scans along a short plasma membrane segment of a patch-clamped cell. The fluorescence responses to a 50-Hz train of square voltage-clamp hyperpolarizations were recorded. The excitation wavelength for each dye was chosen by finding the region of maximum slope on the red wing of its absorbance spectrum in dimethyl sulfoxide solution (all of the dyes were highly soluble in this solvent); no additional attempt was made to find the wavelength for optimal ΔF/F. Fig. 4B shows the response of the five tested dyes. Although di-2-ANEP(F)PTEA and di-4-ANEQ(F)PTEA seem to be more sensitive than the long-wavelength nonfluorinated dye, di-2-ANEQPTEA, the differences were not statistically significant (P = 0.23, Kruskal–Wallis). However, under two-photon excitation, the effect of fluorination significantly improved photostability (Fig. 4C) of di-2-ANEP(F)PTEA, di-2-ANEP(F2)PTEA, and di-4-ANEQ(F)PTEA (with P < 0.0125, Wilcoxon rank sum and Bonferroni criterion); it showed, on average, bleaching up to 11% compared with di-2-ANEQPTEA, which showed 32% after exposure to 2,150 scans. Based on these results, we chose to test more di-2-ANEP(F)PTEA for single-sweep AP detection in a fixed spot illumination mode. With 20 mW at 1,160 nm during brief 30-ms exposures, we could clearly resolve single APs (Fig. 4D) without observing any increase in leak current, even after 20 repetitions (the fluorescence was corrected for ∼25% bleaching). By averaging the filtered optical signal from 20-well timed APs, we found a close match between the optical and patch-clamp recorded AP waveforms (Fig. 4E), showing the linearity of the dye response. The results summarized in Fig. 4F show that di-2-ANEP(F)PTEA had a V_m sensitivity of 14.6 ± 4.6%/100 mV (median ± SD); the S:N of the total fluorescence was 176 ± 82 (median ± SD). This finding enabled single-trial 90-mV AP-induced changes in the V_m to be detected with a S:N of 23.2 ± 5.0 (median ± SD) from a micrometer-sized spot under two-photon excitation. Although the wavelength dependence of the S:N for F has a generally negative slope, which would be expected for shot noise limited detection, S:N for ΔF/F has no obvious wavelength dependence; this result occurs, because as the wavelength increases, the ΔF/F becomes larger as the noise becomes relatively larger (of course, if there is enough laser power, it can be used to improve the S:N at the red edge of the VSD spectra). After pooling both the voltage step and AP datasets, we found that di-2-ANEP(F)PTEA was not only more photostable than di-2-ANEQPTEA but also, significantly more sensitive to the V_m (P = 0.003, Wilcoxon rank sum) with ΔF/F =12.4 ± 4.2%/100 mV (median ± SD) compared with 7.9 ± 1.0%/100 mV (median ± SD), respectively. The maximum ΔF/F observed for di-2-ANEP(F)PTEA was 21%/100 mV; the lower average ΔF/F and its variability can be attributed to fluorescence coming from internalized dye bound to intracellular membranes.

Fig. 4.

Characterization of long-wavelength VSDs with two-photon excited fluorescence on cultured hippocampal neurons. (A) Example of a patch-clamped hippocampal cultured neuron stained with a VSD. (Scale bar: 10 μm.) Fluorescence measurements were done in both line scan mode at 3.5 KHz along the somatic membrane and fixed spot illumination mode with 20 KHz sampling (red marks). Screening of VSDs was done using line scans along the soma membrane during the multiple voltage-clamp stepping cycles. The optical trace (green) shows the average of 60 trials with 20 50-mV steps per trial using di-4-ANEQ(F)PTEA with 1,290-nm excitation. (B) Sensitivity of the dyes (notched box plot; red, median; dark blue circles, individual data points). The ordinate shows the percentage change in two-photon fluorescence (ΔF/F) per 100 mV voltage-clamp step and the abscissa common to C. (C) Photostability of the VSDs after a 614-ms exposure to 2,150 line scans at 3.5 KHz (notched box plot; red, median; dark blue circles, individual data points; red circles, outliers). In both B and C, each point in the two charts corresponds to a different cell. Laser powers used for each experiment were adjusted to produce approximately the same S:N for each VSD. Together with the wavelength and dye used, they are also listed along the abscissa. In another set of experiments, di-2-ANEP(F)PTEA was tested for single AP detection using a fixed spot illumination mode and 20-mW excitation at 1,160 nm. (D) Bleaching-corrected fluorescence signal sampled at 20 KHz and filtered with a five-point moving average (for presentation, trace has been inverted). (E) Average of 20 AP fluorescent traces corrected for bleaching and filtered with a five-point moving average (blue) and average of 20 patch-clamp recorded APs (red). (F) Summary of fluorescence S:N (blue filled circles) and 90-mV AP detection S:N (green triangles) of the 20 KHz sampled and five-point moving average-filtered optical trace and their medians (red). The S:N of detecting a 90-mV AP is the product of the ordinate S:N F and the abscissa ΔF/F × 0.9. Note that fluorescence decreased with depolarization, because the outer membrane leaflet was stained.

Discussion

The VSDs listed in Table 1 cover a broad range of wavelengths, with absorbance maxima of the membrane-bound dyes ranging from 400 to 553 nm and emission maxima ranging from 589 to 743 nm. For each VSD, red-edge excitation at about 60 nm above the absorbance maximum offers optimal ΔF/F, but there is a broad-wavelength window over which the response is close to the maximum (7, 17, 29). This window stretches the usable excitation range over which good sensitivity can be obtained with the new palette of VSDs from 440 to 670 nm (Fig. 2). Additionally, the hemicyanines have the distinct advantage of large Stokes shifts. Although common cyanine and xanthene fluorescent dyes and typical fluorescent protein probes have Stokes shifts of 20–40 nm, hemicyanines have emission spectra that are shifted by ∼170 nm from their absorbance/excitation maxima (Table 1). Thus, a single excitation source can be used to excite both a hemicyanine VSD and a second fluorescent probe or sensor, and the responses readily separated with appropriate emission filters (Fig. 1). Additionally, hemicyanine dyes have good two-photon cross-sections for excitation into their lowest-energy excited states, and therefore, the two-photon absorbance spectra are approximately at two times the wavelength of the one-photon spectra (39). Although the range for two-photon excitation is shown from 1,020 to 1,290 nm in this work (Figs. 1, 3, and 4), the repertoire of VSDs in Table 1 should allow for effective two-photon voltage sensing at any wavelength between 900 and 1,340 nm. Importantly, commercial Ti-sapphire laser systems are becoming available to cover the entire range from 680 to 1,300 nm. Thus, the availability of the VSDs in Table 1 should allow superior flexibility in the design of experiments with various light sources and cell and tissue preparations and combined with other optical tools for manipulation and sensing of cell physiology.

The photostability of fluorescent probes can be a critical factor in their practical application when the duration of the event of interest or the copy number of fluorescent molecules is small, which would occur in subcellular microscopic measurements. This finding is because the noise in such measurements is dominated by the shot noise caused by the limited number of photons being detected. A dye with high photostability will produce more photons before it is bleached than a dye with low photostability, and therefore, the former can produce fluorescence with intrinsically better S:N. Although we do not present a systematic study of the effect of fluorination on photostability, which is beyond the scope of this work, there is ample precedent that fluorination is beneficial (32, 33). We do show, in several practical tests for our fluorinated VSDs, that they show improved photostability and correspondingly, display little evidence for photodynamic damage to the specimen. Fig. S1 shows that 40 160-ms exposure trials cause no deterioration in the optical signal obtained from a single dendritic spine. Fig. S4 shows that di-4-ANEPPS bleaches almost two times as fast in a guinea pig heart as di-4-AN(F)EP(F)PTEA, which has nearly identical spectral properties. Similarly, for long-wavelength two-photon excitation, Fig. 4C shows that the three fluorinated VSDs tested are three to six times more photostable than a long-wavelength nonfluorinated VSD. Additionally, Figs. 1 and 3 provide evidence that, under conditions of prolonged optical recording, these VSDs do not seem to change the electrophysiology properties of the biological substrate.

The application vignettes in Figs. 1, 2, 3, and 4 show the efficacy of the dyes for single cells in culture, externally and internally stained neurons in brain slices, and whole-heart preparations. Additionally, we have recently reported the ability of di-4-AN(F)EPPTEA to map subcellular AP properties in single cardiomyocytes (36). These examples show the high sensitivity of the fluorinated VSDs in subcellular, single-cell, and tissue-level applications. They also show the efficacy of the dyes with both one- and two-photon excitation. Figs. 1, 2, 3, and 4 indicate that the voltage sensitivities of the fluorinated VSDs are comparable with and often exceed the sensitivities of the best conventional hemicyanines. The ability to choose from the large palette of VSDs listed in Table 1 assures that sensitivity and other properties (solubility, photostability, toxicity, excitation wavelength, emission wavelength, etc.) can be optimized to meet the varying demands of a large range of optical electrophysiology experiments.

Materials and Methods

The detailed preparation and characterization of di-4-AN(F)EPPTEA and di-2-AN(F)EPPTEA have been reported (29, 36). The synthetic procedures for the other VSDs in Table 1 all involve the same aldol coupling chemistry used for the published syntheses. Details on the dye spectral measurements, biological preparations, electrophysiological recording methods, and fluorescence imaging setups (Fig. S5) can be found in SI Materials and Methods.