1.1.

Foundation of Optical Recording of Membrane Potential Transients from Individual Nerve Cells

The feasibility of multiple site optical recording from subregions of individual nerve cells was initially demonstrated using monolayer neuronal culture and extracellular application of the voltage-sensitive dye. Grinvald et al.⁵ showed that voltage-sensitive dyes and multisite optical measurements can be employed successfully to determine conduction velocity, space constants, and regional variations in the electrical properties of neuronal processes in measurements from dissociated neurons in monolayer culture. Both absorption and fluorescence measurements were used in these experiments, and the individual dissociated neurons in a culture dish were stained from the extracellular side by bath application of the dye.⁶ It has also been possible, using the same approach, to study synaptic interactions among several interconnected neurons in culture^7,8 and the cell-to-cell propagation of the AP in patterned growth cardiac myocytes forming two-dimensional (2-D) hearts in culture.⁹ The monolayer neuronal culture is a low opacity system especially convenient for extracellular staining of the outside cellular membrane. However, primary cultures are networks of neurons that grow under artificial conditions. Therefore, a number of important questions in cellular neurophysiology can only be studied in intact or semi-intact preparations. However, extending the same approach to in situ conditions has proven difficult and has been slow to develop. In in situ conditions, the extracellular application of fluorescent dyes cannot provide subcellular resolution because of the large background fluorescence from the dye bound indiscriminately to all membranes in the preparation.¹⁰

Based on these arguments, Grinvald et al. designed an approach to optical analysis of electrical events in processes of individual nerve cells. The key conceptual advance was the idea to stain particular neurons in situ selectively by intracellular application of an impermeant fluorescent voltage-sensitive dye. This approach was based on pioneering measurements in Larry Cohen’s and Brian Salzberg’s laboratories at the Marine Biological Laboratory in Woods Hole, performed on the giant axon of the squid, which demonstrated that optical signals may be obtained when the dye is applied from the inside.^11–14 In these experiments, optical signals were obtained from the large membrane surface area of the giant axon. Substantial additional efforts were required to demonstrate that the same approach is feasible at the spatial scale of average size neurons and their processes. The initial experiments in this direction, using intracellular labeling with voltage-sensitive dyes, were carried out by Obaid et al.¹⁵ and by Grinvald et al.¹⁶ on leech neurons. The results demonstrated the essential advantages of using intracellular application of fluorescent potentiometric probes. Fluorescence measurements are more effective than absorption measurements when measuring from a small membrane area,¹⁷ particularly in situations where the image of the object (e.g., thin process) is much smaller than the size of the photodetector picture element (pixel).^18,19 When transmitted light is used, only a small fraction of the total light captured by individual pixels will be modulated by the signal from the neuronal process (dendrite or axon) projected onto that pixel ($\mathrm{\Delta }I$). Most of the light will be projected directly and will only contribute to the resting light intensity ($I$). Thus, the fractional signal ($\mathrm{\Delta }I/I$) will be very small. On the other hand, in fluorescence measurements, practically all of the light captured by individual pixels will come from the object (if autofluorescence is negligible) regardless of the fraction of the pixel surface area covered by the image of the object. In now classical experiments on leech neurons,¹⁶ fluorescence measurements were used to record APs and synaptic potential signals from processes of selectively stained single neurons. As predicted, the $S/N$ was substantially improved relative to prior absorption measurements, but the sensitivity was still too low to be of practical value—the available $S/N$ was insufficient for multiple site optical recording of complex electrical interactions at the level of thin neuronal processes. The limits to the sensitivity of these measurements were determined primarily by (a) the relatively low voltage sensitivity of the available dyes (amino-phenyl styryl dyes, e.g., RH437 and RH461; fractional change in fluorescence intensity per AP, in intracellular application, of the order of 0.01% to 0.1%); (b) the choice of the excitation light bandwidth; and (c) the relatively low intensity of the incident light that could be obtained from a 100-W mercury arc lamp. Despite low sensitivity, these experiments clearly showed the advantages of selective staining of individual neurons by intracellular application of the dye and paved the way for further improvements in sensitivity. At the time, the reported sensitivity was the result of a modest screening effort, suggesting that better signals might be obtained by (a) synthesizing and screening new molecules for higher sensitivity, (b) increasing the concentration of the dye to increase the fluorescence intensity, (c) using an excitation light source capable of providing higher light intensity and better stability, and (d) using detector devices with lower dark-noise and adequate spatial and temporal resolution. Following this rationale, the first substantial improvement in the $S/N$ was obtained by finding an intracellular voltage-sensitive dye with sensitivity in intracellular application two-orders of magnitude higher than what was previously available. The amino-naphthalene styryl dye JPW1114 synthesized by Wuskell and Loew at the University of Connecticut Health Center had the sensitivity expressed as $\mathrm{\Delta }F/F$ per AP, in intracellular application, of the order of 10%.²⁰ The higher sensitivity of this dye translates directly into higher $S/N$.

The second significant improvement in sensitivity was the introduction of cooled, backilluminated CCD cameras. The CCD camera, because it is cooled and can have about 1000 times smaller pixel surface area and proportionally smaller dark currents compared to available diode arrays, exhibits substantially lower dark noise. Following the introduction of CCD cameras, a number of studies demonstrated that the $V_{\mathrm{m}}$-imaging is quite efficient at the spatial scale of dendritic branches.^20–22 The available $S/N$, however, was still not adequate to allow multisite recordings from thin axons and axon collaterals at high frame rates (5 to 10 kHz) without extensive averaging ($>100$ trials). In addition, it was still not possible to monitor $V_{\mathrm{m}}$ signals at the higher optical magnification and finer spatial scale necessary for resolving electrical events at the level of individual dendritic spines. Clearly, further improvements in sensitivity (expressed as the $S/N$) were needed.

One way to increase the $S/N$ with a given voltage-sensitive dye is to increase the photon flux ($\mathrm{\varphi }$) by increasing the excitation light intensity. Another possibility is to increase the fractional fluorescence change per unit change in $V_{\mathrm{m}}$ (sensitivity of the dye) by choosing the optimal excitation wavelength. Following this rationale, Holthoff et al. used one of the most sensitive voltage-sensitive dyes in terms of $S/N$ (JPW3028^21,23,24) and improved both the excitation (and, hence, emission) light intensity and the relative fluorescence change in response to $V_{\mathrm{m}}$ change by utilizing a laser as an illumination source in wide-field epifluorescence microscopy mode.²⁰ In measurements from layer 5 pyramidal neurons in rat visual cortex slices, the light from a frequency-doubled 200-mW diode-pumped $\mathrm{Nd}:{\mathrm{YVO}}_4$ continuous wave laser emitting at 532 nm was directed to a quartz optical fiber coupled to the microscope via a single-port epifluorescence condenser (TILL Photonics) designed to overfill the back aperture of the objective. In this way, approximately uniform illumination of the object plane was attained. In the initial series of experiments with this apparatus, the extent of propagation and the time course of backpropagating action potentials (bAPs) signals were monitored from dendrites at relatively low magnification [the full CCD frame ($80\times 80\text{pixels}$) corresponded to a $300\times 300\mu \mathrm{m}$ area in the object plane]. In this configuration, bAPs could be monitored with unprecedented sensitivity in single-trial measurements at all dendritic sites, including the terminal apical tuft as well as the oblique and the basal dendrites.²⁰ The combined effect of an increase in light intensity and the use of a near-optimal excitation wavelength was a dramatic improvement in the sensitivity of voltage imaging by a factor of about 50 in recording from different sites on neuronal processes.

2.1.

Slices, Patch-Clamp Recording and Intracellular Application of Dyes

All procedures were performed in accordance with Public Health Service Policy on Humane Care and Use of Laboratory Animals and approved by Yale University Institutional Animal Care, as well as by Paris Descartes Ethics Committee for Animal Research (registered number CEEA34.EV.118.12) in accordance with European Union institutional guidelines. The experiments were carried out on somatosensory cortex slices from 15- to 21-day-old rats of either sex. The rats were decapitated following deep halothane anesthesia, the brain was quickly removed, and $300\text{-}\mu \mathrm{m}$-thick coronal cortical slices were cut in ice-cold solution. Slices were incubated at 37°C for about 30 min and then maintained at room temperature (23°C to 25°C). The standard extracellular solution used during recording contained (in mM): 125 NaCl, 25 $\mathrm{NaHC}{\mathrm{O}}_3$, 20 glucose, 2.5 KCl, 1.25 ${\mathrm{NaH}}_2{\mathrm{PO}}_4$, 2 ${\mathrm{CaCl}}_2$, and 1 ${\mathrm{MgCl}}_2$, pH 7.4 when bubbled with a 5% ${\mathrm{CO}}_2$ gas mixture balanced with 95% ${\mathrm{O}}_2$. Somatic whole-cell recordings in current clamp or voltage-clamp mode were made with 4 to $6\mathrm{M}\mathrm{\Omega }$ patch pipettes using a Multiclamp 700B amplifier (Axon Instruments Inc., Union City, California). Voltage-clamp recordings were made with series resistance compensation set at 70%. The pipette solution contained (in mM): 120 K-gluconate, 3 KCl, 7 NaCl, 4 Mg-ATP, 0.3 Na-GTP, 20 HEPES, and 14 tris-phosphocreatin (pH 7.3, adjusted with KOH) and 0.8 mM of the voltage-sensitive dye JPW3028.⁴³ In all experiments, an attempt was made to identify layer 5 pyramidal cells with intact dendrites in one plane of focus close to the surface of the slice (to minimize light scattering) using infrared differential interference contrast (DIC) video microscopy. The recordings were carried out from spines on superficial basal dendrites at different distances from the soma (range 30 to $120\mu \mathrm{m}$). Labeling of pyramidal neurons with the voltage-sensitive dye was carried out by free diffusion from a somatic patch electrode in the whole-cell configuration. We used the most successful voltage probe for intracellular application, JPW3028, a close analog of JPW1114^20,45,51 with similar voltage sensitivity available from Thermo Fisher Scientific as D6923. The electrode tips were first filled with dye-free solution by applying negative pressure and then back filled with the solution containing the voltage probe (0.8 mM). The patch electrode was detached from the neuron by forming an outside-out patch after staining was accomplished, as determined by measuring resting fluorescence intensity from the soma. Following the staining period, the preparation was incubated for an additional 1.5 to 2 h at room temperature to allow the voltage-sensitive dye to spread into dendritic processes and spines. Before optical recording, the cell was repatched to obtain electrical recording using an electrode filled with dye-free intracellular solution.

2.2.

Optical Setup

The recording setup was built around a stationary upright microscope (Olympus BX51; Olympus Inc.) equipped with a high-spatial resolution CCD camera for infrared DIC video microscopy and for detailed morphological reconstruction of dye-loaded neurons (Photometrics, CoolSnap ES2) and a fast data acquisition camera with relatively low-spatial resolution ($80\times 80\text{pixels}$) and exceptionally low read noise (NeuroCCD-SM, RedShirtImaging LLC, Decatur, Georgia) used for voltage imaging.

The microscope was coupled with two holographic-patterned illumination paths, one for voltage imaging and one for uncaging. The simplified schematic of the setup is shown in Fig. 1(a). The voltage imaging part was based on a 532-nm diode-pumped laser source (MLL-FN-532-450-5-LAB-TTL, Changchun New Industries Optoelectronics Tech. Co. Ltd., Changchun, China). The laser beam was attenuated by neutral density filters, enlarged by a telescope, and cleaned by a $50\text{-}\mu \mathrm{m}$ pinhole before it was directed to the sensitive area of a liquid crystal of an SLM (Hamamatsu X10468-01). The reflected beam was projected onto the back focal plane of $100\times$ microscope objective ($\mathrm{NA}=1.0$ water dipping, Olympus) by an afocal telescope formed by L1 ($f=500\mathrm{mm}$) and L2 ($f=250\mathrm{mm}$) and a dichroic mirror D1 (FF560-FDi01-25x36, Semrock). The zero-order nonmodulated components reflected from the SLM were suppressed by an intermediate point block. A defocus of the beam was explicitly introduced before the SLM such that the zero-order focus was displaced by around 5 cm. For the diffracted first-order component, the defocus was compensated with a spherical Fresnel lens at the SLM. In this way, the point block for the zero-order component did not significantly perturb the propagation of the holographic pattern.⁴⁵ For voltage imaging, the illumination power delivered to the slice depended on several factors, including the level of voltage-sensitive dye staining, the depth of the dendritic branch with spines, and the exact pattern of stimulation. Thus, the illumination power delivered to the preparation varied in the range of 0.05 to 0.6 mW.

Fig. 1

Two-color CGH illumination: (a) diagram of the optical setup for holographic illumination patterns at 532 and 405 nm. The 532-nm (green) laser beam from a solid-state laser is attenuated by neutral density filters (ND), enlarged by a telescope, and directed to the liquid crystal SLM. The plane of the SLM is projected with an afocal telescope (L1 and L2) to the back aperture of a microscope objective (OBJ). The 405-nm patterned beam (violet) is generated by a commercial module (phasor) and coupled to the 532-nm beam path thorough a dichroic mirror D1. The patterned beams are reflected toward the objective by the dichroic mirror D2. The collected fluorescence is filtered by the long pass filter (LPF) and directed either to the high-resolution camera or to the high-speed NeuroCCD camera using a movable mirror. (b) Upper panels: Fluorescence pattern generated by a 532-nm (right) and 405-nm (left) digital holography on a thin layer of rhodamine 6G as detected with the NeuroCCD. These patterns were used to calibrate the submicrometric positioning of the holographic spots. Scale bar $1\mu \mathrm{m}$. Lower panels: spatial profile of the rhodamine emission generated by a 532-nm (green) or 405-nm (blue) CGH spot recorded with the high-resolution camera.

The uncaging illumination was obtained using a commercial module for holographic light patterning (Phasor–3i Intelligent Imaging Innovation, Inc. Denver, Colorado), modulating a 405-nm laser source (Obis Coherent, 100 mW) controlled by LaserStack and SlideBook software (3i Intelligent Imaging Innovation). The illumination power delivered to each uncaging spot varied between 0.1 to 0.5 mW. The imaging and uncaging beam were combined using a dichroic mirror D2 (Di03-R405/488/561/635-25x36, Semrock). The induced fluorescence from a voltage-sensitive probe was long-pass filtered LPF (FF01-593/LP-25, Semrock) and imaged using NeuroCCD-SM camera.

2.3.

Computer-Generated Holography

Multipoint CGH patterns for both imaging and uncaging were generated using a modified weighted Gerchberg and Saxton algorithm (GS).^47,52 The relative intensity between spots was controlled by adjusting their weight in the input pattern of the algorithm.^53,54 The pattern and the corresponding wGS-generated phase profiles were defined, computed, and addressed to the SLM using “Wavefront Designer IV,” in house software written in C++ with Qt 4.4.0 and fftw 3.1.2. Calibration and characterization of the generated patterns were performed by detecting induced fluorescence on a thin film of rhodamine 6G spin-coated on a coverslip. The multipoint patterns shown on Fig. 1(b) acquired on the NeuroCCD camera were used to calibrate the exact positioning of the holographic spots. By introducing a correcting stretch, translation, and rotation transformation to the input pattern that was provided to the algorithm, we could achieve submicrometric precision in spot positioning. The experimentally recorded beam size [Fig. 1(b)] results from the convolution of excitation (405 or 532 nm) and the detection point spread function (PSF) at the wavelength of the collected rhodamine emission (filtered by the LPF filter). We estimated the effective excitation beam size based on the nominal value of the detection PSF and the measured experimental profile [acquired by high-resolution camera and shown in Fig. 3(b)] to be $\sim 0.3\mu \mathrm{m}$ (FWHM) for the 405-nm spots and $\sim 0.7\mu \mathrm{m}$ for the 532-nm ones. A slight under-filling of the objective back aperture was used to increase the size of holographic spots of the 532-nm imaging pattern to better match the size of a typical spine. A similar GS algorithm^54,55 was also applied to generate a large 532-nm holographic spot (diameter $\sim 25\mu \mathrm{m}$) that was used to mimic wide-field illumination. The large illumination spot was used to obtain an image of dye-loaded dendrites and identify structures of interest on the NeuroCCD for voltage imaging. Because extended holographic shapes are characterized by randomly distributed intensity inhomogeneities (speckles), wide-field images (as shown in Fig. 3) were obtained by averaging five different speckle patterns. The obtained wide-field fluorescence images of selected dendritic segments were used to define the pattern of holographic illumination for both imaging and uncaging (Fig. 3). We adjusted different illumination intensities for different illumination spots to enhance signals from specific regions of interest. Typically, the spots covering spine heads were adjusted to have 2 to 10 times stronger illumination intensities relative to those on the dendritic shaft. For voltage imaging, laser light was gated by a high-speed shutter (Uniblitz LS6, driver D880C) for 10- to 30-ms recording trails. Recordings were carried out at a frame rate of 2 or 5 kHz. Acquisition and analysis of data were carried out with NeuroPlex software (RedShirtImaging). For glutamate uncaging, we used bath-applied DNI-glutamate (5 mM) provided by Femtonics KFT (Budapest, Hungary). Illumination pattern for uncaging consisted of single or multiple spots placed near individual spine heads, approximately at a distance of $0.5\mu \mathrm{m}$. The actual spatial relationship between the uncaging spots and spine heads was uncertain at the submicrometer spatial scale because of light scattering in the brain tissue. Short pulses (0.2 to 0.3 ms) of patterned illumination were applied to induce local glutamate release near spine heads.

2.4.

Data Analysis

Subthreshold EPSP signals and bAP signals were recorded typically for 30 ms at a frame rate of 2 or 5 kHz at room temperature kept at 25°C to 27°C or at near physiological temperature of 32°C to 34°C. Repetition of typically nine trials was acquired and averaged. Analysis and display of data were carried out using the NeuroPlex software (RedShirtImaging) written in IDL (Exelis Visual Information Solutions, Boulder, Colorado) and custom visual basic routines. No correction for background fluorescence was carried out because it was found to be negligible under holographic illumination restricted to several spots on labeled dendrites. The signal alignment software was used to correct for temporal jitter in AP initiation, as well as for possible small movements of the preparation during averaging.⁴³ In the temporal domain, AP signals were aligned by cross-correlation of the electrically recorded APs in each trial to the reference signal acquired at the start of averaging. In the spatial domain, camera images were aligned in 2-D offline by image cross-correlation to compensate for possible small lateral movements of the preparation. Correct focus of the image in the $z$-dimension was verified frequently; small adjustments were often necessary. The spatially and temporally aligned signals were averaged and slow changes in light intensity due to bleaching of the dye were corrected by dividing the data by an appropriate dual exponential function derived from the recording trials with no stimulation. Dual exponential fitting requires an algorithm for least-squares estimation of nonlinear parameters. We used the Levenberg–Marquardt algorithm. Subthreshold optical signals were calibrated on an absolute scale (in mV) by normalizing to an optical signal from a bAP, which has a known declining amplitude along basal dendrites, as it has been determined by patch-pipette recordings.^56,57 In an earlier study,⁴³ we experimentally confirmed the previously reported result⁵⁸ showing that this method of calibration produces the same results as normalizing signals to optical recordings corresponding to long hyperpolarizing pulses delivered to the soma that attenuate relatively little as they propagate along dendrites.