Introduction

The brain uses electrical signals to transmit information. These electrical signals are associated with changes in membrane potential in the form of synaptic potentials and nerve impulses, or action potentials. Sharp microelectrode recording, developed in the late 1940s [30], and more recently, the application of the patch-clamp technique [24], have been essential tools for the investigation of neuronal function by allowing the recording of changes in membrane potential in single neurons. Despite their usefulness, these methods suffer from the general problem in that they typically can only be used to record changes in membrane potential at a single site. Even recent advances allowing patch-clamp recordings from multiple locations on the same neuron are restricted to two or three locations, usually on large diameter dendrites [51, 52]. Neurons have complex branching morphologies, with dendritic processes that can extend up to a millimeter from the cell body, and axons that can be significantly longer. These processes are often of small diameter (sub-micrometer) and inaccessible to current recording techniques. These limitations can be addressed through the use of optical techniques, which, in principle, have the capacity to resolve changes in membrane potential at the sub-micrometer level. Here, we review the development and recent advances in the use of optical techniques to investigate fast membrane potential changes in single neurons and their processes.

Development of voltage-sensitive dyes

Work by Hill and Keynes [25] in the late 1940s showed that optical methods could be used to detect changes in light scattering following electrical activation of crab axons. The first optical recordings of fast membrane potential changes, however, were not performed until the late 1960s, in experiments that detected changes in light scattering and birefringence, as well as fluorescence, during single action potentials in crab, lobster, and squid axons [11, 53]. It was subsequently discovered that by staining preparations with “merocyanine” dyes, the signal-to-noise ratio of measurements of membrane potential could be significantly increased [13]. This pioneered the use of voltage-sensitive dyes to measure changes in membrane potential and was first used to record changes in intracellular membrane potential from single neurons of the leech segmental ganglion [48]. These studies paved the way for the search and development of better voltage-sensitive dyes with increased signal to noise [12, 18, 23, 34, 45], with the promise of an optical solution to measuring changes in membrane potential in neurons in both time and space. This goal was realized with the introduction of photodiode arrays allowing the activity of multiple neurons to be monitored simultaneously [17, 20, 49]. Despite these advances, the use of voltage-sensitive dyes to study electrical signaling in single neurons was hampered by their low signal to noise. This situation has changed in recent years due to improvements in technology for the detection of small fluorescent signals at high temporal and spatial resolution, as well as the development of more sensitive hydrophilic voltage-sensitive dyes for intracellular application.

How do voltage-sensitive dyes work?

To sense a change in trans-membrane potential, voltage-sensitive dyes need to contain hydrophobic elements that are embedded in the membrane. While the actual mechanism through which fast (sub-millisecond) voltage-sensitive dyes do this is unclear, work in this area suggests that for some dyes, the change in fluorescence occurs via an “electrochromic” mechanism, whereby the energy of the fluorescent dye between its ground and excited state is modulated by the trans-membrane electric field through a redistribution of electrons within the dye molecule itself [33, 34], or through the use of a monomer–dimer mechanism, whereby changes in the trans-membrane electric field lead to changes in dye aggregation and/or localization in the membrane [54]. The consequence of this electrochromic or monomer–dimer mechanism is that a change in membrane potential causes a shift in the excitation or emission spectrum of the dye, leading to a change in the intensity of the detected fluorescence at a particular wavelength [31].

Imaging single cell responses with voltage-sensitive dyes

Using external application of voltage-sensitive dyes, single cell responses were first achieved by restricting the region from which emitted fluorescence was detected to that arising from a single cell (Fig. 1a) [48]. This method will only work effectively, however, in situations where the neurons examined are in a monolayer or where excitation is restricted to a narrow region in the z-axis around the focal plane. Subsequent studies were able to record responses from neuronal processes using low-density cell culture by improving spatial resolution of the detection system through restricting the area of excitation to a small region of interest or via the use of a photodiode array (Fig. 1b) [19]. Alternatively, it is also possible to detect changes in membrane potential in intact preparations using externally applied voltage-sensitive dyes by signal averaging action potential responses generated following current injection via an intracellular recording pipette (Fig. 1c) [28, 44]. The use of externally applied voltage-sensitive dyes to studying electrical signaling in small neuronal processes in intact preparations is limited, however, as the dye binds to all lipid membranes, including non-neuronal cell membranes, increasing background fluorescence and thereby reducing the signal to noise. Also, spatial resolution is compromised, as activation of multiple neurons or inputs makes the origin of any observed signals unclear. To overcome these limitations, it is necessary to apply the voltage-sensitive dye internally, so that the fluorescent signal is restricted to only the neuron under investigation.

Fig. 1
figure 1

Early single cell voltage-sensitive dye recordings in invertebrate neurons. a Measurement of the fluorescent intensity change during an action potential in a nociceptive neuron from the leech segmental ganglion after bath application of voltage-sensitive dye. Top, measurement of the fluorescent change when the pinhole in the objective image plane is positioned over the stimulated cell. Middle, absence of a fluorescent change when the pinhole is positioned over an adjacent cell. Bottom, membrane potential recorded via a somatic recording pipette in the stimulated cell. Adapted from [48]. b Fluorescent changes at the indicated locations in a N1E-115 neuroblastoma cell in culture stained by bath application of voltage-sensitive dye and recorded by a 10×10 photodetector array (grid). An action potential (bottom left) was generated by current injection via the somatic recording pipette. Adapted from [19]. c Simultaneous fluorescent measurements (right) at the indicated locations from the soma during electrotonic spread of hyperpolarising potential generated by a brief (2 ms) somatic current injection through the recording pipette. Detected using a photodiode array in a neuron from the barnacle supraesophageal ganglion filled with Lucifer Yellow (left) after bath application of a voltage-sensitive dye. Adapted from [44]

Full size image

Voltage imaging using internally applied dyes

The first successful internal application of voltage-sensitive dyes in single neurons was achieved by iontophoresis of dye from a sharp microelectrode recording pipette in an invertebrate neuron [21]. Following on from this important step, the use of intracellular voltage-sensitive dyes to selectively stain individual neurons was aided by the development of new dyes for intracellular application that were more soluble in water and more sensitive to membrane potential changes [4, 32]. These developments significantly increased the sensitivity of the intracellular voltage-sensitive dye technique for monitoring membrane potential transients from multiple sites on neuronal processes, allowing the analysis of the pattern of initiation and propagation of electrical signals in identified snail neurons in the whole ganglion preparation [3, 55]. Initial experiments used pressure ejection from sharp microelectrodes, again in invertebrate neurons [4, 55]. Subsequent experiments applied voltage-sensitive dyes internally via diffusion from the recording pipette during patch-clamp recordings from neurons in brain slices [1, 2]. The use of the patch-clamp technique increased the ability to passively fill neurons with voltage-sensitive dyes due to the larger tip diameter of patch pipettes compared to sharp microelectrodes. These experiments, pioneered by Zecevic, Antic and colleagues [2], showed that internal application of voltage-sensitive dyes could be used to detect changes in membrane potential in the fine dendritic processes of mammalian neurons in intact brain slices. Phototoxicity and non-specific pharmacological effects, however, were found to be much more of a problem than in invertebrate neurons, and procedures needed to be developed to reduce the impact of internal application of voltage-sensitive dyes on neuronal health. It was discovered that the pharmacological effects of the dye could be reduced by withdrawal of the recording pipette after the initial filling period and then waiting 1 to 2 h for recovery (allowing time for the dye to distribute within the neuron). The same neuron was then re-patched with a recording pipette without dye [2]. In addition, phototoxicity was reduced by limiting exposure to high-intensity excitation light to brief intervals (∼100 ms) using low repetition rates. Using these procedures, it was possible to reliably record action potential responses in fine basal and oblique dendrites of cortical pyramidal neurons using voltage-sensitive dyes (Fig. 2) [2]. A follow-up study examined action potential back-propagation into basal dendrites in cortical pyramidal neurons in more detail [5], concluding that action potentials backpropagate up to 200 μm from the soma with little amplitude and shape modulation.

Fig. 2
figure 2

Example of action potential responses recording from a cortical pyramidal neuron after intracellular application of voltage-sensitive dye. a Outline of the 464 element photodiode array superimposed over the fluorescent image of the pyramidal neuron. b Single-trial recording of action potential-related optical signals. Each trace represents the output of one diode. Traces are arranged according to the position of the detectors. Each diode received light from a 14 × 14 μm area. Action potentials were evoked by somatic current injection via the recording pipette. c Comparison of electrical and optical recordings. Top panel, spatial average of optical signals from eight individual diodes from the somatic region (dashed line) is superimposed on the electrical recording from the soma (solid line). Bottom panel, same electrical signal compared with a single trial optical recording. d Action potential signals from detectors located in the basal, oblique and apical dendrites. Traces from different locations are scaled to the same height. The increasing delay between the signal from the somatic region and proximal dendritic segments reflects the time taken for the propagation of the action potential to dendritic locations. Taken from [2]

Full size image

One problem with the interpretation of voltage-sensitive dye signals, however, is that they are difficult to calibrate. Internally applied voltage-sensitive dyes bind not only to the plasma membrane, but also to internal membranes, which do not undergo a change in membrane potential during electrical signaling. In addition, the diameter of dendritic processes is not uniform and is typically smaller at more distal locations. These factors will lead to differences in the contribution of the basal fluorescence signal to the voltage-sensitive signal, leading to differences in the magnitude of the percentage change in fluorescence at different locations, even if the underlying voltage change is the same. One way to get around this problem is to compare relative changes in voltage-sensitive dye signals in response to different stimuli at the same location [14, 26]. Using this method to compare the first and third action potential signal in a burst, a recent study concluded that action potentials are significantly attenuated as they propagate into distal basal dendrites of pyramidal neurons [26], in contrast to the conclusions of Antic [5].

Other applications of internally applied voltage-sensitive dyes include investigation of voltage attenuation of sub-threshold events, with a recent study analyzing the attenuation of synaptically evoked sub-threshold electrical signals in mitral cells in the olfactory bulb [14]. By directly comparing the amplitude of excitatory postsynaptic potentials (EPSPs) with the back-propagating action potential at different dendritic locations, it was concluded that EPSPs undergo only slight attenuation as they propagate from their site of origin in the glomerular tuft to the soma. Internal application of voltage-sensitive dyes has also been used to study regenerative plateau potentials [38, 39] and local dendritic sodium spikes [40] in basal dendrites in cortical pyramidal neurons. A recent study also used internally applied voltage-sensitive dyes to map the site of action potential initiation in cortical pyramidal neurons [43], finding, in contrast to observations in cerebellar Purkinje neurons [10, but see 27], that action potential initiation occurs at the distal end of the axon initial segment. The same study also showed that once initiated, action potentials propagate down the axon in a saltatory manner.

The future of voltage imaging

The above studies clearly show the power of using internally applied voltage-sensitive dyes to investigate processing of both supra-threshold and sub-threshold responses in dendritic and axonal processes of neurons. The strength of voltage imaging lies in its ability to measure fast electrical signals at high spatial resolution. Unlike calcium-sensitive dyes, the response of fast voltage-sensitive dyes is not limited by the speed of the dye, but instead by its sensitivity to changes in voltage. With the development of more sensitive dyes (e.g. [29]), as well as improvements in detection technology, we predict further advances in this area.

In the final part of this review, we look at new technologies that have promise to further increase both the sensitivity and spatial resolution of fast voltage imaging in single neurons and their processes. One of the recent advances in this area is the use of genetically coded voltage sensors. The original idea was to insert green-fluorescent protein (GFP) near the gating domain of a voltage-activated potassium channel, such that voltage-driven rearrangements of the channel lead to changes in GFP fluorescence [22, 50]. Such a genetically coded voltage sensor offers many advantages as it could, in principle, be targeted to specific cell types or cellular compartments of neurons. One problem with these early genetically encoded voltage indicators is that they were relatively slow due to the slow activation kinetics of the voltage-activated potassium channel used, as well as the location of the GFP at the rather distant C-terminus. An additional issue is that the response is not linearly related to the change in membrane potential, but rather follows the sigmoidal voltage dependence of channel activation. Improvements in speed have been obtained by using fluorescence resonance energy transfer (FRET) between a pair of GFP variants linked to the same potassium channel [47] and by inserting GFP at a site closer to the voltage sensor of voltage-activated sodium channels [6]. In a recent development, a hybrid approach, which links GFP to a synthetic voltage-sensing molecule using FRET, has been shown to offer increased signal to noise, as well as a relatively linear change in fluorescence with membrane potential, and fast (sub-millisecond) response time [9].

Another important advance in imaging membrane potential in fine neuronal processes is the use of random-access laser scanning microscopy techniques [7, 8], and a variant of two-photon microscopy using second-harmonic generation (SHG) [15, 16, 35–37, 41, 42, 46]. Laser scanning techniques offer improved spatial resolution, allowing voltage-sensitive imaging from small dendritic processes such as dendritic spines. Although still in the early days of development, SHG has already been used to record action potential propagation along dendrites of invertebrate neurons (Fig. 3a,b) [46] and also to investigate the ability of action potentials to invade dendritic spines in neocortical pyramidal neurons in brain slices (Fig. 3c–f) [42]. One of the main advantages of this method, apart from increased spatial resolution, is that, as the second harmonic signal is obtained exclusively from the cell membrane, it is directly related to the underlying change in membrane potential, allowing calibration of fluorescent signals in terms of the absolute change in membrane potential at different locations. In addition, the use of long wavelength infrared lasers for SHG allows for deeper tissue penetration due to reduced light scattering.

Fig. 3
figure 3

Examples of action potential responses recorded using second harmonic generation (SHG). a SHG projection image of an Aplysia neuron in culture after staining with bath application of FM4-64. The green lines indicate the position for each line scan during the SHG recording of action potentials. b The black traces (left-axis scale), obtained from the averaged line scans, are normalized intensity plots of SHG emission versus time at the positions shown in a. The red traces (right-axis scale) are the membrane potential recorded via an electrode at the soma. a and b from [46]. c Layer 5 pyramidal neuron stained by intracellular application of FM 4-64 and imagined using SHG. d High-resolution image of a dendritic spine on the basal dendrite marked in c. e Calibration of spine SHG signals under voltage-clamp conditions. Normalized SHG signal changes upon voltage steps at spines and their parent dendrites are shown. There are no significant differences between the sensitivity of SHG at spines and dendrites (P = 0.7, n = 5, t test). f SHG measurements during action potentials. A single action potential was initiated by current injection via the somatic recording pipette (top), and the resulting SHG signal changes measured at soma (bottom left) and dendritic spines (bottom right). Average of four and seven recordings, with each recording the result of averaging ∼40 action potentials. c to f from [41]

Full size image

In summary, the strength of voltage imaging lies in its ability to measure the fast electrical signals that underlie neuronal communication both between and within neurons at high spatial and temporal resolution. As technology improves, the application of voltage-sensitive dyes and other voltage imaging methods will continue to be a valuable tool for investigators wishing to explore the electrical events underlying single neuronal computation.