2.1.

Tissue Preparation

Typically, $30\text{-}\text{to}60\text{-}\text{day}$ -old NIH Swiss female mice were euthanized and their pituitary glands were harvested following Institutional Animal Care and Use Committee approved procedures. The neurohypophysis along with a portion of the infundibular stalk and the surrounding pars intermedia were separated from the anterior pituitary. The intact neurohypophysis was mounted in a custom-designed optical recording chamber and perfused with oxygenated (95% ${\mathrm{O}}_2$ , 5% $\mathrm{C}{\mathrm{O}}_2$ ) physiological saline ( $154\mathrm{mM}$ NaCl, $5.6\mathrm{mM}$ KCl, $2.2\mathrm{mM}$ $\mathrm{Ca}{\mathrm{Cl}}_2$ , $1\mathrm{mM}$ $\mathrm{Mg}{\mathrm{Cl}}_2$ , $20\mathrm{mM}$ $N$ -2-hydroxyethyl perazine- $N$ -2-eteanesulfric acid, $10\mathrm{mM}$ glucose; pH 7.4). The infundibular stalk was placed between two platinum-iridium $(90∕10\%)$ field electrodes, and action potentials were initiated by brief $\left(500\mu \mathrm{s}\right)$ bipolar shocks, supplied by an SIU-102 stimulus-isolation unit (Warner Instruments, LLC; Hamden, CT). The health of the neurohypophysis following dissection was assessed by measuring the intrinsic optical signal associated with excitation-secretion coupling in the unstained preparation.²³ All measurements were performed at room temperature ( $24°\mathrm{C})$ .

2.2.

Staining with VSDs

The action spectra (i.e., the changes in fluorescence dye emission induced by membrane depolarization) of five different styryl-based voltage-sensitive dyes were determined: di-8-ANEPPS, JPW 1114 (di-2-ANEPEQ), RH 414, FM1-43, and the recently developed di-4-ANEPPDHQ.²⁸ All dyes were obtained from Invitrogen (Carlsbad, CA). Stock solutions of most dyes were prepared by dissolving the dye into a $(90∕10\%)$ mixture of dimethyl sulfoxide and Pluoronic F-127 to a final dye concentration of $5–10\mathrm{mM}$ . FM1-43 was dissolved directly in distilled water instead. For staining of the preparation, $3–5\mu \mathrm{L}$ of dye stock was added to $1\mathrm{mL}$ of physiological saline. The preparations were typically stained for $45\min$ , with di-8-ANNEPS requiring $80–90\min$ incubation periods. Prior to fluorescence measurements, the dye solutions were washed out with physiological saline for $15\min$ . Figure 5 compares the relative intensity of resting fluorescence emission (integrated over the total peak area) of all dyes after their respective staining period.

2.3.

Optical Setup and Spectral Recordings of Fluorescence Emission

All fluorescence measurements were performed on an upright BX61WI microscope (Olympus America; Center Valley, PA) using a $20\times 0.5\mathrm{NA}$ water-immersion objective. A $470\mathrm{nm}$ high-power LED (MBLED; Thorlabs, Newton, NJ), driven at $75–600\mathrm{mA}$ with a stabilized constant-current LED controller (Mightex; Pleasanton, CA), was used as excitation light source.²⁹ The emission from the LED was bandpass limited with a $482∕35\mathrm{nm}$ filter (Semrock; Rochester, NY). To record the broadband fluorescence emission from the dye, we replaced the bandpass emission filter of our fluorescein isothiocyanate (FITC) filter cube (Semrock; Rochester, NY) with a $515\mathrm{nm}$ long-pass filter (Schott glass OG 515 lp). To minimize dye bleaching and possible phototoxic damage to the preparation, LED illumination was limited to a total of $0.6\mathrm{s}$ prior to and during electrical stimulation.

To capture the brief and transient changes in fluorescence dye emission during action-potential stimulation, we used an AvaSpec 2048-USB2 CCD-array spectrophotometer (Avantes USA; Broomfield, CO). The spectrophotometer was fitted with a $600\text{lines}∕\mathrm{mm}$ grating and a $2048\text{pixel}$ CCD with a detection range of $267\text{to}819\mathrm{nm}$ , resulting in a nominal resolution of $2.1\mathrm{nm}$ . On the CCD detector, however, the spectral data are sampled every $0.27\mathrm{nm}$ and with $16\text{bit}$ resolution of intensity changes. A built-in order-sorting filter reduces contamination of the spectra from higher-order diffraction peaks. To preserve the small voltage-induced shifts in fluorescence emission, the default smoothing algorithm in the acquisition software had to be disabled. Light output from the camera port on the BX61WI microscope was focused with $10\times 0.25\mathrm{NA}$ objective onto the $600\mu \mathrm{m}$ fiber-optic cable input to the spectrophotometer. Synchronization between the electrical stimulation of the neurohypophysis with spectral recordings of dye emission was provided by an STG1004 stimulus generator (Multichannel Systems, Reutlingen, Germany).

Measurements of changes to the emission spectra of voltage-sensitive dyes during action-potential depolarization are complicated by two factors. First, the spectral shifts in fluorescence dye emission are quite small (see Fig. 1). In addition, membrane depolarization induced by action potentials is short and transient. To address these limitations, we typically summed one hundred individual measurements of emission spectra for resting fluorescence and depolarization-induced spectral changes measured from a single preparation. During action-potential depolarization, spectra of dye fluorescence emission were recorded in a single $8\text{-}\mathrm{ms}$ sweep of the spectrophotometer. Resting spectra of all voltage-sensitive dyes were recorded $120\mathrm{ms}$ prior to generating an action potential. Minimizing the spacing between spectra recorded during rest and during stimulation and reducing illumination intensity avoided distortion of the action spectra due to dye bleaching. Stimulation at frequencies above $1\mathrm{Hz}$ induces plastic changes in the excitable responses of the neurohypophysis.³ Hence, consecutive spectral recordings were spaced at least $4\mathrm{s}$ apart. The total number of stimuli within a given stimulus train was limited to 20, and successive stimulus trains were separated by $15\min$ rest periods.

2.4.

Data Processing

All data were analyzed using Igor Pro 5.02 software (Wavemetrics; Lake Oswego, OR). The emission spectra recorded during rest were subtracted from those taken during stimulation and averaged over 100–200 trials. We used the following procedure to resolve the small voltage-induced wavelength shifts in fluorescence emission. For small spectral shifts, $\Delta F∕\Delta \lambda \approx d{F}_0∕d\lambda$ and, therefore, $\Delta \lambda \approx \Delta F∕(d{F}_0∕d\lambda )$ . Here, $\Delta F$ is the experimentally measured change in fluorescence emission, and $d{F}_0∕d\lambda$ is the local slope of the resting fluorescence emission $F_0$ at a given wavelength $\lambda$ . Therefore, plots of $\Delta F∕(d{F}_0∕d\lambda )$ over the entire emission range can be used to estimate $\Delta \lambda$ . The above considerations also indicate that even very small wavelength shifts $\Delta \lambda$ result in significant changes in fluorescence emission $\Delta F$ wherever the fluorescence emission changes rapidly with $\lambda$ (see insert in Fig. 1). Therefore, high sensitivity to small fluorescence changes permits us to resolve wavelength shifts $\Delta \lambda$ close to the spectral sampling rate of the CCD array $\left(0.27\mathrm{nm}\right)$ and well below the nominal $2.1\mathrm{nm}$ resolution of the spectrophotometer.

The raw difference spectra were still rather noisy due to significant noise from the uncooled CCD array detector and current-to-voltage converters. We used Fourier analysis to filter out the high-frequency components of this white noise. A typical power spectrum obtained via fast Fourier transform (FFT) of the raw difference spectra is shown in Fig. 2 . The first 10–15 components of the power spectrum at low spatial frequencies rise by several orders of magnitude above the uniform background of white noise dominating high spatial frequencies. It is these low-frequency components of the Fourier spectrum that represent the portions of the action spectra slowly varying with wavelength. Hence, we low-pass filtered the spectra by zeroing out all Fourier components below five times the root-mean-square amplitude of the white noise or beyond spatial frequencies of $0.1{\mathrm{nm}}^{-1}$ or higher. Filtered action spectra were obtained via inverse FFT of the filtered Fourier spectra.

Fig. 2

Fourier analysis of the action spectrum for di-4-ANEPPDHQ. Power spectrum for the action spectrum of di-4-ANEPPDHQ shown in Fig. 1. Note the large amplitudes at low spatial frequencies, followed by white noise at higher frequencies. By zeroing out the high-frequency noise components, the S/N ratio of the action spectra could be significantly improved (see Materials and Methods section).

2.5.

Ratiometric Camera Recordings

We tested whether we could obtain ratiometric dye signals using RH414, a dye with a weak action spectrum (i.e., small voltage-induced spectral shifts). Changes in RH414 fluorescence were induced by a single action potential invading the neurohypophysis. Fluorescence changes were recorded using a fast $128\times 128\text{pixel}$ CCD camera cooled to $-60°\mathrm{C}$ (Ixon DV860BV, Andor; Northern Ireland). To resolve fast voltage transients, spatial resolution was reduced by $4\times 4$ binning, thereby increasing the frame rate to $1613\mathrm{FPS}$ . A $470\text{-}\mathrm{nm}$ LED was used for dye excitation. To catch either one of the two peaks in the action spectrum of RH414, fluorescence emission was recorded using a standard FITC filter cube with either a $536∕40\text{-}\mathrm{nm}$ bandpass or a $610\text{-}\mathrm{nm}$ long-pass emission filter. The tissue averaged fluorescence in either emission band was normalized to the fractional change from the resting fluorescence just before stimulation. Typical optical recordings of action potentials obtained in consecutive recordings from either peak in the action spectrum are shown in Fig. 6.

Fig. 6

Ratiometric recordings of neurohypophysial action potential. Fluorescence emission changes recorded with a fast CCD camera (1613 frames/s) during action-potential depolarization of a mouse neurohypophysis stained with RH 414 $(20\mu \mathrm{M}$ for $40\min$ ) using $482∕35\mathrm{nm}$ excitation. The upper trace shows optical voltage signal recorded with a $536∕40\mathrm{nm}$ bandpass filter, while the lower curve shows the signal from the same preparation obtained with a $610\mathrm{nm}$ long-pass filter.

3.1.

In Situ Measurements of Action Spectra

We have measured the resting fluorescence emission spectra and action spectra of four electrochomic dyes: di-8-ANEPPS, di-4-ANEPPDHQ, di-2-ANEPEQ (JPW 1114), and RH 414. We have also included the voltage-sensitive dye FM1-43 for comparison and control. In contrast to the spectral shifts exhibited by the former, voltage changes predominately modulate the amplitude of FM1-43 fluorescence emission.

Figure 1 displays the resting fluorescence-emission spectrum of di-4-ANEPPDHQ in the neurohypophysis averaged over 100 $8\text{-}\mathrm{ms}$ single-sweep recordings of a CCD-array spectrophotometer. The resting emission spectra are somewhat distorted by contributions from higher-order diffraction off the diffraction grating (e.g., the two spikes between 600 and $650\mathrm{nm}$ ). Because we are mostly interested in the voltage-induced changes to the resting emission spectra, we did not attempt to correct for these distortions to the resting spectra. Superimposed on the resting spectrum (solid line) is the emission spectrum recorded during action-potential–induced depolarization of the neurohypophysis, also averaged over 100 recordings. As indicated by the inserts in Fig. 1, the emission spectrum of di-4-ANEPPDHQ is slightly blue-shifted during membrane depolarization. We estimated that the average wavelength shift $\Delta \lambda$ induced during action-potential depolarization was slightly less than $0.4\mathrm{nm}$ (for details, see Materials and Methods). The action spectrum of the dye obtained from one single-sweep recording was too noisy to warrant analysis. To overcome this noise limitation, measurements of the resting and voltage-shifted emission spectra were repeated 100 times and the action spectrum was calculated from their difference. The resulting action spectrum is shown as noisy gray trace in Fig. 1.

3.2.

Low-Pass Filtering of Action Spectra

As shown in Fig. 1, the overall shape of the action spectrum is clearly discernable after the summation of 100 difference spectra. Yet, the signal-to-noise (S/N) ratio is rather disappointing. Closer inspection of the difference spectra reveals that much of the noise seems to arise from broadband white noise, while most of the features of the difference spectrum are slowly varying with wavelength. This is confirmed by calculating the spectral power distribution of the action spectra (Fig. 2). There are several large peaks in the power spectrum at low spatial frequencies, with a background of weaker but uniform white-noise components present at all spatial frequencies. We low-pass filtered the summed action spectra by eliminating all high-frequency Fourier components while preserving the low-frequency components (see Materials and Methods for details). This yielded the filtered action spectrum for di-4-ANEPPDHQ shown as a solid line in Fig. 1.

To ascertain whether low-pass filtering might alter the shape of the difference spectra, we performed the following test. A single emission spectrum of di-4-ANEPPDHQ was recorded at a very low LED intensity (LED current set to $1\mathrm{mA}$ ) and $2\mathrm{ms}$ integration time of the spectrophotometer (Fig. 3a ). This artificially noisy resting spectrum was Fourier-analyzed as described in the previous paragraph (Fig. 3b) and transformed back into real space. The filtered spectrum was plotted together with the rescaled sum of 100 emission spectra for the same dye recorded at normal illumination (Fig. 3c). Although the count rate of the summed resting spectra was over ${10}^4$ times that for the single, low-intensity measurement, low-pass filtering essentially reproduced the emission spectrum recorded at normal illumination intensities.

Fig. 3

Testing the effects of Fourier filtering on fluorescence spectra. (a) A single emission spectrum of di-4-ANEPPDHQ was recorded at very low illumination and short integration time. (b) The power spectrum of this artificially noisy emission spectrum shows strong low-frequency components and a base level of high-frequency noise. (c) After low-pass filtering (for details, see Materials and Methods), the filtered emission spectrum (dashed line) was superimposed onto the sum of 100 emission spectra taken at much higher illumination intensity and with much longer integration times (solid line).

Action spectra for all five VSDs investigated are presented in Fig. 4 . The action spectra of the recently developed dye di-4-ANEPPDHQ yielded larger responses in fluorescence emission with transmembrane voltage than di-8-ANEPPS, JPW 1114, or RH414. More importantly, di-4-ANEPPDHQ much more efficiently stained the preparation (see Fig. 5 ) and, therefore, significantly improved the S/N ratio of the weak fluorescence changes $\Delta F$ . The maximal voltage response for all electrochromic dyes occurs in wavelength bands on either side of the fluorescence emission peaks. Not surprisingly, the action spectra of all amino-naphthylethenylpyridinium-based dyes display similar features, with the largest peak of their action spectra on the short wavelength side of their fluorescence emission. The action spectrum of RH 414, in contrast, develops its maximal response on the long-wavelength shoulder of its resting fluorescence emission. For optical recordings of electrical activity within a single emission band, recording on the favorable emission shoulder clearly improves the optical S/N. The asymmetry in the peak amplitude of the action spectra on either side of the isosbestic point represents an additional challenge to ratiometric optical recordings of voltage transients. Response amplitudes of voltage-sensitive dyes are intrinsically small. In addition, voltage-induced fluorescence signals in complex tissue only arise from a fraction of stained tissue. Hence, the asymmetric peak amplitudes of the action spectra reduces S/N levels during ratiometric measurements of voltage signals, further highlighting the benefit of in situ measurements of action spectra.

Fig. 4

Action spectra for various VSDs. Action spectra $\Delta F∕F_{\max }$ for five VSDs obtained from the mouse neurohypophysis during action-potential depolarization. For each dye, the raw difference spectrum is shown in light gray, with the Fourier filtered spectrum superimposed as a solid line. All spectra are plotted to the same scale. The vertical lines indicate the location of the peak in the resting fluorescence emission. The action spectra of di-4-ANEPPDHQ, JPW 1114, RH 414, and FM1-43 were measured from the average of 100 spectra, while 200 spectra were averaged for di-8-ANEPPS.

Fig. 5

Staining efficay of various VSDs in the mammalian neurohypophysis. Relative intensity of the integrated fluorescence emission of various VSDs after staining of the mammalian neurohypophysis. The preparations were incubated with $20\mu \mathrm{M}$ of dye for $45\min$ , and the dye was washed out prior to measurement. Di-8-ANEPPS was diluted to $40\mu \mathrm{M}$ and was incubated for $90\min$ instead.

3.3.

Ratiometric Measurements of Transmembrane Voltage

Ratiometric measurements of fluorescence changes allow us to discriminate against contributions to the optical transients not related to voltage signals, including dye bleaching or dye internalization. Based on the action spectra measured above, we decided to perform some preliminary ratiometric measurements of voltage transients in the neurohypophysis. With the exception of FM1-43, all of the dyes in this study can be used for ratiometric measurements of transmembrane voltage. For this preparation, di-4-ANEPPDHQ clearly provides the most robust fluorescence signal. However, as shown in Fig. 6 , even with the relatively weak peaks in the action spectrum of RH 414, ratiometric measurements are possible. Optical recordings of the population action potential were obtained with $482∕35\mathrm{nm}$ excitation and fluorescence emission recorded with either a $536∕40\mathrm{nm}$ bandpass filter (upper trace) or a $610\mathrm{nm}$ long-pass filter (lower trace). The emission signals at the two opposite peaks of the action spectrum were recorded in consecutive trials using a fast CCD camera acquiring $1613\text{frames}∕\mathrm{s}$ . The fast transient dominant in either emission band is related to the population action potential depolarizing neurohypophysial axons and their secretory swellings. In addition, a slow after-depolarization following the action-potential transient is present in both recordings. These ratiometric recordings indicate that this slow depolarization is dominated by trans-membrane voltage changes, supporting our previous contention that it arises from glial depolarization following electrical activity.³