Measuring Absolute Membrane Potential Across Space and Time

Abstract

Membrane potential (V_mem) is a fundamental biophysical signal present in all cells. V_mem signals range in time from milliseconds to days, and they span lengths from microns to centimeters. V_mem affects many cellular processes, ranging from neurotransmitter release to cell cycle control to tissue patterning. However, existing tools are not suitable for V_mem quantification in many of these areas. In this review, we outline the diverse biology of V_mem, drafting a wish list of features for a V_mem sensing platform. We then use these guidelines to discuss electrode-based and optical platforms for interrogating V_mem. On the one hand, electrode-based strategies exhibit excellent quantification but are most effective in short-term, cellular recordings. On the other hand, optical strategies provide easier access to diverse samples but generally only detect relative changes in V_mem. By combining the respective strengths of these technologies, recent advances in optical quantification of absolute V_mem enable new inquiries into V_mem biology.

1. INTRODUCTION

Membrane potential (V_mem), or voltage across a lipid bilayer, is ubiquitous in biology. In electrically excitable cells such as neurons and cardiomyocytes, millisecond V_mem fluctuations cause release of neurotransmitters and contraction. Both excitable and nonexcitable cells exhibit V_mem signals, which vary over timespans of seconds to days. Furthermore, an estimated 10–50% of the cellular energy budget goes to maintain V_mem, even in nonexcitable cells (96). Given this large energy expenditure, the functions of V_mem span far beyond excitable electrical activity in the brain or heart (2); recent work has linked V_mem to cell proliferation (26, 42), differentiation (25, 104), and tissue patterning (64).

Membrane potential (V_mem):

a voltage across a membrane, arising from differences in ion concentration on either side

V_mem results from differences in ion concentration across a semipermeable membrane. Although this phenomenon is best studied with respect to the plasma membrane of mammalian cells, any selectively permeable membrane can maintain a voltage. For example, voltages in bacterial communities (58, 60, 89) and across organelle membranes (20, 124) also have signaling roles. Unless otherwise indicated, we use the term V_mem to indicate the plasma membrane potential between the cytosol and the extracellular space. In addition, we use absolute V_mem to indicate the value of the membrane potential in millivolts, rather than a relative measure of changes in V_mem.

Absolute V_mem:

the value of V_mem in mV units; this term is used to differentiate between V_mem recording techniques that can quantify V_mem and those that can only track relative changes

V_mem is inherently a system-level property, determined by an array of ion channels and pumps and sensed by diverse cellular components. In mammalian cells, the primary ions involved in generating V_mem are Na⁺ and K⁺ (Figure 1a), but many other species (including Cl⁻, Ca²⁺, H⁺, and organic anions) play a role. Within the cellular V_mem response, voltage-gated ion channels are the best-documented class of voltage-sensitive proteins; these channels transduce V_mem changes into additional ionic currents or fluctuations in the second messenger Ca²⁺. Nevertheless, all membrane-localized proteins experience V_mem, and more of these proteins may be sensitive to V_mem than was previously thought. Evidence of V_mem sensitivity exists for various membrane components (121), including phosphatases (80), G protein–coupled receptors (92, 93), and the membrane organization itself (123). For most systems, we have an incomplete understanding of both the determinants of V_mem and the factors that respond to V_mem.

Figure 1.

Membrane potential (V_mem) signals and measurement techniques across time and space. (a) A schematic of cellular V_mem, with examples of seven biological processes where V_mem signaling plays a role. The resistance to ionic flow between compartments sets the length scale of V_mem signals; V_mem can scale across tissues or be compartmentalized within organelles or dendritic spines. The properties of ion channels and pumps that determine V_mem, as well as ion diffusion between compartments, dictate the time duration of a V_mem response. (b) Biological space accessible to the three most common strategies for absolute V_mem measurement (shaded regions), overlaid with example biological process of interest (outlined boxes, numbered as in panel a). While many techniques can report cellular V_mem on the scale of milliseconds or seconds, longer recordings or recordings across large areas are difficult to access. Shaded areas indicate regions where single trial recordings retain voltage resolution on the scale of biological V_mem changes (≤20 mV) and minimally alter the biological sample. Recordings with fluorescence electrochromic dyes require recalibration with an electrode on each cell, whereas fluorescence lifetime and patch-clamp electrophysiology can be applied as stand-alone measurements that are comparable between cells. Shaded areas are restricted to demonstrated application space for each technique and do not indicate the full extent of its possible use.

These gaps in our understanding are largely attributable to limitations in the existing V_mem recording toolkit, as most V_mem recording strategies are only tractable in a small subset of biological samples. In particular, strategies to record absolute V_mem and absolute V_mem changes (i.e., to document that V_mem depolarized by 20 mV) are unavailable in many systems (Figure 1b). To better understand necessary features of V_mem recording strategies, we first summarize known V_mem signaling pathways and build a wish list of features for an optimized V_mem reporter. We then turn our focus to the two primary approaches for recording V_mem: electrodes and fluorescent sensors. We review recent progress in V_mem sensing platforms and highlight areas where development of new tools would facilitate discovery.

2. MEMBRANE POTENTIAL SIGNALING ACROSS TIME AND SPACE

V_mem signaling is most extensively studied in excitable tissue, where rapid changes in cellular V_mem transmit information. However, important V_mem signals also occur across minutes, hours, and days(2). Similarly, as a spatial signal, V_mem is canonically treated as cell autonomous and uniform in spherical model cells. Nevertheless, V_mem can be compartmentalized subcellularly (20, 30, 120, 124) and delocalized across tissues (21, 69, 76, 95). In this section, we break down V_mem signals into three length scales: cellular, subcellular, and tissue. From there, we enumerate the unique challenges that each scale brings for V_mem measurement.

2.1. Cellular Membrane Potentials: Action Potentials, Cell Cycle Control, Etc.

The many ion gradients across the plasma membrane create a cellular V_mem, which is generally assumed to be uniform across a cell due to free diffusion of ions within the cytosol. Many cellular V_mem events are short-lived; their quantification requires both high temporal resolution and high voltage resolution. An important fast V_mem signal is the action potential (AP) in excitable cells. APs in mammalian neurons typically last a few milliseconds, involving a rapid depolarization of approximately 100 mV followed by rapid hyperpolarization back to the resting V_mem (8). Many aspects of APs have important cellular effects, including waveform, frequency, and V_mem (AP amplitude, as well as initial and final absolute V_mem) (8). V_mem recording platforms have been engineered with the goal of recapitulating AP waveform and frequency, and modern versions of both electrode-based and optical strategies reflect excellent progress toward this end (14, 57, 77, 115, 116). However, options for recording AP amplitude or absolute V_mem are more limited (see below).

Action potential (AP):

a rapid, all-or-nothing V_mem signal in excitable cells that results from the activity of V_mem-sensitive ion channels

To extend recordings from milliseconds to minutes, hours, or days, two new requirements emerge. First, the V_mem recording technique must be stable over the relevant timeframe, displaying no artefactual, V_mem-independent changes in signal. Second, it must be noninvasive, with cellular processes continuing normally despite chronic observation. With existing tools, V_mem recordings on timescales longer than minutes are challenging and are infrequently attempted. As a result, we have an incomplete picture of what V_mem changes occur and little clarity regarding their effects. Nevertheless, it is evident that slower changes in V_mem are pleiotropic signals with important cellular roles. In neuroscience, for example, the resting V_mem of neurons influences the propensity to fire APs. Furthermore, neuronal resting V_mem changes in association with—and perhaps also regulates—diverse processes, including cellular metabolism (97) and circadian rhythm (63). In addition, V_mem appears to hyperpolarize during neuronal development and differentiation, although the magnitude of the observed changes varies drastically with the measurement technique used (48, 75, 105).

Furthermore, non-excitable cells also show V_mem changes on long timescales, often broadly related to growth. For example, V_mem changes associated with progression from the G1 to the S phase of the cell cycle have been documented in various cell lines (6, 108, 111), and acute epidermal growth factor receptor activation can be accompanied by a V_mem signal (59, 71, 86). In a disease context, cancer cells may be generally more depolarized than their nontransformed counterparts, implying that V_mem modulation may offer selective growth advantages (117). Nevertheless, because current techniques cannot monitor absolute V_mem on the same cell over hours to days, we lack a road map of V_mem throughout the cell cycle or in growth signaling.

2.2. Subcellular Membrane Potentials: Organelles and Membrane Compartmentalization

Because V_mem originates from ion concentration gradients, subcellular V_mem differences can only exist in areas of electrical compartmentalization, where diffusion is regionally restricted. Partial compartmentalization produces transient or metastable local differences in V_mem. To study subcellular V_mem differences, a technique requires exquisite spatial resolution, as well as access to subcellular structures without damaging them. Isolating the V_mem of a subcellular compartment from the cellular V_mem (which may be larger in magnitude and certainly occupies a larger region of space) is particularly challenging. However, direct evidence for subcellular V_mem differences exists, along with a preponderance of indirect evidence that V_mem is locally regulated in membrane-bound compartments.

Electrical compartmentalization:

a restriction on ionic flow that allows for local differences in ion concentration and V_mem

Electrical compartmentalization occurs where the plasma membrane forms compartmentalized processes, and the longevity of localized V_mem signals is determined by the resistance to ion flow in and out of the compartment. Transient subcellular V_mem signals occur in neurons, as voltage waveforms are transmitted and processed by dendrites (102) and the axon initial segment (53). Because these electrical compartments have nearly free ionic flow with other parts of the cell, the signals are fleeting. Visualization of dendritic computation and AP propagation requires exceptional temporal resolution, with recording rates often exceeding 10 kHz (9). Longer-lived subcellular differences likely exist in the micron-scale dendritic spines, insulated by high-resistance bottlenecks between the spine and the dendrite. Mounting evidence suggests that spine V_mem can be electrically distinct from the rest of the neuron, and its role in learning and memory is an area of active study (37, 120). In non-excitable cells, V_mem compartmentalization at the plasma membrane has yet to be conclusively shown. Nevertheless, the subcellular enrichment of particular ion channels in primary cilia (30, 99) and filopodia (44) suggests that nonuniform plasma membrane V_mem is likely.

Organellar membranes also offer an opportunity for ionic, and therefore V_mem, compartmentalization. Organellar electrode-based recordings are both invasive and challenging, as they often involve rupturing the plasma membrane. Nevertheless, through labor-intensive experimental efforts, we have some direct evidence for V_mem signals in organelles. For example, the mitochondrial V_mem (and H⁺ gradient) is required for successful oxidative phosphorylation (124), and depolarized mitochondrial V_mem, as well as defective oxidative phosphorylation, is associated with neurodegenerative diseases (47). The lysosome contains excitable channels (20), and the endoplasmic reticulum V_mem has been reported to respond to cellular V_mem (90). Furthermore, where direct recordings have not been made, proteomic and transcriptomic studies of organellar channels and transporters imply complex physiology (114), and improved methods tailored to the unique constraints of these subcellular systems are of great interest.

2.3. Membrane Potentials in Tissue: Long-Range Interactions and Delocalization

At the other extreme of the length scale, V_mem carries information via long-range interactions in tissues. Direct experimental interrogation of V_mem in tissue requires throughput and multiplexed measurement across many cells simultaneously, as well as the ability to measure in situ. Electrical communication over distances is facilitated by synaptic transmission, as well as by gap junctions. In the brain, neurons do not act in isolation, but instead form circuits (5, 13); understanding behavior requires simultaneous V_mem mapping across many cells. Beyond neural circuits, V_mem can be delocalized across tissues via gap junctions, which allow free flow of ions and many small molecules between cells. Early studies of electrical coupling and gap junctions were performed in the Drosophila salivary gland, where resistance to current flow between cells is only slightly higher than the resistance to flow within cells (69). Since then, gap junctions have been found to be ubiquitous, with dysfunction connected to various diseases (109). Gap junctions also enable rapid exchange of V_mem information across tissues, as in the synchronous electrical signaling in the heart (95).

Electrical coupling via gap junctions not only enables transmission of V_mem signals across a population of cells, but also inextricably links V_mem between coupled cells. Despite this, many observations about cellular V_mem signaling are made on isolated cells in culture. For epithelial cells that exist in vivo as a tightly coupled tissue, this is an artificial electrical state. What happens to the surrounding tissue when an individual cell changes its V_mem throughout the cell cycle or as a response to growth factor signaling? It remains unclear how and to what extent cell-autonomous V_mem signals discovered in isolated cells will translate into tissues. Theoretical studies suggest that, in some conditions, stable and metastable electrical patterning is possible, but it is highly dependent on the electrical coupling in the tissue (22, 23). These results imply that V_mem signals are accompanied—or at least modulated—by dynamic regulation of cell–cell coupling. Although some progress has been made toward direct visualization of V_mem distributions in tissue (21, 76), such measurements are largely unexplored.

Taken together, diverse biological questions necessitate a noninvasive, multiplexed, in situ V_mem platform with excellent spatiotemporal and voltage resolution. Throughput and technical ease of use of the platform are also important, as they enable top-down profiling experiments that can help elucidate V_mem signaling networks. Of the array of methods available today, no single approach possesses all of these characteristics (Tables 1 and 2). Indeed, more realistically, a toolkit of recording strategies will be required to fully interrogate V_mem. Below, we discuss the performance of existing V_mem measurement techniques, with an eye toward their ability to report V_mem across space and time.

Table 1.

Comparison of electrode-based V_mem recording strategies

Performance attribute	Whole-cell patch clampa	Cell attachedb	Perforated patchc	Nanopipettesd	Nanopillarse	Planar patch clampf
Spatial resolution	− (single point)	− (single point)	− (single point)	− (single point)	++ (grid of points)	− (single point)
Temporal resolution	+++ (μs–ms)	+++ (μs–ms)	+++ (μs–ms)	+++ (μs–ms)	+++ (μs–ms)	+++ (μs–ms)
Voltage resolution	+++ (1 mV)	+ (variable)	+++ (1 mV)	+++ (1 mV)	+++ (1 mV)	+++ (1 mV)
Stability	+ (mins–1 h)	+ (mins–1 h)	++ (~1 h)	+ (mins–1 h)	+++ (days)	+ (mins–1 h)
Noninvasiveness	−	++	+	+	+	−
In situ capabilities	++	++	++	++	−	−
Access to subcellular structures	+	+	+	+++	++	−
Throughput and multiplexing	−	−	−	−	++	+++
Ease of use	−	−	−	−	+	++

^aDescribed in References 7, 14, 38, 49, 52, 73, 100, 103, 105, and 110.

^bDescribed in References 87, 105, and 106.

^cDescribed in References 40, 54, and 66.

^dDescribed in References 45 and 46.

^eDescribed in References 94 and 113.

^fDescribed in References 32 and 82.

Technique performance in each category is ranked as follows: −, poor; +, fair; ++, good; +++, excellent. Abbreviation: V_mem, membrane potential.

Table 2.

Comparison of optical V_mem recording strategies

Performance attribute	Single-color fluorescence intensitya	Single-color fluorescence lifetimeb	FRET-basedc	Electrochromic (ratio-based)d	GEVI temporal dynamicse	Stimulated Raman scatteringf
Spatial resolutiong	+++ (<1 μm)	++ (~1 μm)h	+++ (<1 μm)	+++ (<1 μm)	++ (~1 μm)h	+ (>1 μm)h
Temporal resolution	++ (1 ms)	+ (5 ms)h	+ (1 ms or >10 ms)i	++ (1 ms)	− (s)h	++ (1 ms)h
Voltage resolutionj	− (≫00 mV)	++ (VF: 20 mV, GEVI: >100 mV)	ND	+ (~100 mV)	++ (10 mV)	ND
Stability	+ (minutes)	++ (hours)	++ (hours)	+ (minutes)	ND	ND
Noninvasiveness	+++	+++	+++	+++	+++	+++
In situ capabilities	+++	+++	+++	+++	+++	++
Access to subcellular structuresk	++	++	++	++	++	++
Throughput and multiplexing	+++	++	+++	+++	+	++
Ease of use	+++	+	++	++	−	−

^aMany single-color fluorescence intensity V_mem sensors have been designed, although they are seldom used for absolute V_mem quantification. Data are compiled from References 57, 77, 115, and 116.

^bSingle-color fluorescence lifetime recordings have been shown for genetically encoded voltage indicators (16) and for small-molecule VFs (59).

^cFRET-based systems may use conformational changes of the indicator (4), changes in the absorption spectrum of a rhodopsin derivative (3, 125), or translocation of charged groups in the membrane (15, 35, 72) to report V_mem. Many properties are similar across these architectures; metrics in this case represent all types of FRET-based strategies unless otherwise noted.

^dPerformance metrics are aggregated from References 17, 33, 34, 68, 79, and 122.

^eDemonstrated in Reference 41.

^fDemonstrated in References 61 and 62.

^gSpatial resolution is determined by the diffraction limit, as well as optics used in the microscope. Techniques rated ++ rather than +++ generally require moderate spatial binning to obtain adequate signal.

^hFor fluorescence lifetime, GEVI temporal dynamics, and stimulated Raman scattering measurements, there is a trade-off between spatial and temporal resolution, and one may be sacrificed to improve the other. We list the best values of either spatial or temporal resolution shown in the literature or that we have demonstrated in our laboratory.

ⁱFRET sensors based on protein conformational changes or dye translocation exhibit temporal resolution of >10 ms, whereas electrochromic FRET-based sensors can resolve millisecond-time V_mem events.

^jVoltage resolution listed is for applying a calibration determined on one cell onto another individual cell; more precise V_mem measurements can be obtained if measurements from multiple cells are averaged. Most techniques have substantially better resolution for quantifying the magnitude of a V_mem change on an individual cell (e.g., VF-FLIM has 5-mV resolution for quantifying V_mem changes on a single cell versus 20-mV resolution for quantifying absolute V_mem alone).

^kLimited by delivery of sensor to intracellular structures and/or the ability to isolate signal from only the subcellular membrane structure of interest when all membrane structures are stained.

Technique performance in each category is ranked as follows: −, poor; +, fair; ++, good; +++, excellent. Abbreviations: FLIM, fluorescence lifetime imaging; FRET, Förster resonance energy transfer; GEVI, genetically encoded voltage indicator; ND, not determined; VF, VoltageFluor; V_mem, membrane potential.

3. ELECTRODE-BASED APPROACHES FOR RECORDING ABSOLUTE MEMBRANE POTENTIAL

Electrode-based techniques are the gold standard for recording V_mem, ion channel properties, and electrical currents in cells. Three primary configurations are used in cellular V_mem recordings: whole-cell patch clamp, cell attached, and perforated patch (Figure 2). All three can quantify absolute V_mem with excellent temporal resolution, but they suffer from high invasiveness, instability over time, low throughput, and poor spatial resolution (Table 1).

Figure 2.

Electrophysiological configurations for V_mem recording. (a) General schematic for electrode-based V_mem recordings, in which a cell comes into direct contact with an electrode. Voltage is measured as the difference between the recording electrode and a reference electrode in the bath solution. (b–d) Close-up of the interface between the membrane and the electrode in different electrophysiology configurations. (b) In whole-cell patch-clamp electrophysiology, the plasma membrane is ruptured, and the cytosol mixes with the recording electrode solution. (c) In the cell-attached configuration, the plasma membrane is left intact. (d) In perforated patch, the membrane is not ruptured, but ionophores introduced into the recording solution allow ionic exchange across the membrane.

Whole-cell patch clamp:

a direct, electrode-based technique for recording V_mem in which physical contact is established between an electrode and the inside of a cell

3.1. Intracellular Recordings of Absolute Membrane Potential

In intracellular recording, V_mem is determined from the difference between a recording electrode in the cytosol and a reference electrode in the extracellular solution. More detailed treatment of the many capabilities of intracellular recording can be found elsewhere (14); we focus only on V_mem measurement. Two main types of intracellular recordings exist: sharp electrode and whole-cell recordings. Sharp microelectrodes possess very fine tips and high tip resistances, which produce a variable tip potential and a V_mem artifact on the order of tens of millivolts (14, 65). Furthermore, they create a leak in the cell of interest, producing artificially depolarized apparent V_mem and altered input resistance (65, 101). Although sharp electrode recordings are still in use in certain in vivo preparations, they have largely been replaced by intracellular recordings in the whole-cell configuration. Whole-cell intracellular recordings use much larger patch electrodes with lower tip potentials; such electrodes can be sealed onto the cell with less damage than sharp electrodes (100). Whole-cell recordings can also provide higher-resolution information about the particular ions and types of channels involved in setting V_mem.

Intracellular recording provides a direct, in situ measurement of V_mem with sub-mV precision and excellent temporal resolution, but it suffers from a variety of drawbacks. First, the whole-cell patch is highly invasive and disrupts the very processes under investigation. By dialyzing the cytosol with electrode internal solution, whole-cell patch clamp washes out soluble signaling factors (14, 49, 73). In addition, physical contact with the electrode can activate mechanosensitive channels and signal transduction pathways (70, 112), introducing further artifacts. With respect to the measurement itself, leaks in the seal between the recording electrode and the membrane can produce errors in recorded V_mem (105). More subtly, an electrode has poor spatial resolution, reporting V_mem at a point of contact that may or may not reflect other parts of the cell or tissue. For complex membrane structures or electrically coupled tissues, this space clamp error renders the effective recording area unknown (7, 110). For subcellular recordings, electrode-based strategies struggle to physically access subcellular compartments, as the size of the electrode tip provides a lower bound on the size of structures that can be interrogated (30, 37). Furthermore, intracellular recordings are challenging to execute. Attempting each whole-cell recording takes an expert researcher or specialized robot 5–10 minutes, with a 10–30% success rate for these efforts (38, 52, 103). While in situ automated patch systems have reduced the need for human expertise and labor, they still do not achieve high throughput or multiplexing. Therefore, recording across many cells in the same tissue is nearly impossible, as is cataloging V_mem subpopulations over many cells.

3.2. Reducing Invasiveness: Cell-Attached and Perforated Patch Configurations

To mitigate the washout of soluble factors associated with whole-cell recording, V_mem must be measured without disrupting the plasma membrane. This is possible in the cell-attached configuration, in which an extracellular patch electrode is sealed onto the cell without rupturing the membrane. If the membrane resistance is approximately 100-fold lower than the seal resistance, then the voltage across the small membrane region under the pipette will reflect cellular V_mem with reasonable accuracy (87). Alternatively, if the channel composition and electrical properties are known in the cells of interest, then the reversal potential of these channels (determined in the cell-attached configuration) can also be used to infer V_mem. Such techniques have been demonstrated with the K⁺ reversal potential in hippocampal interneurons (106), as well as the NMDA reversal potential in hippocampal CA3 neurons (105). However, this strategy is difficult to extend, as it requires detailed knowledge of the inventory and behavior of channels in the specific system of interest.

Cell-attached configuration:

an electrophysiological configuration where an electrode is sealed onto the membrane of a cell but the membrane is not ruptured

Perforated patch recordings offer a more general solution to reduce invasiveness of whole-cell recordings (40, 66). In perforated patch, an ionophore placed in the pipette internal solution allows ions to cross the membrane without completely disrupting it, thereby electrically connecting the cytosol with the pipette solution. Because the membrane remains impermeable to the diffusion of larger molecules, perforated patch recording can last over an hour without run-down of relevant cytosolic factors (40, 54).

Ionophore:

a molecule that can bind specific ions and diffuse across membranes, equalizing the electrochemical gradient for that ion

3.3. Engineering Electrodes for Improved Performance

Efforts to improve electrode-based recordings have also turned to the electrodes themselves, where recent developments have reduced invasiveness and improved throughput of electrode-based recordings. Below, we highlight two advances in electrode design: planar patch clamp and nanoscale electrodes.

3.3.1. Planar patch clamp.

With the development of low-noise arrays of patch electrodes (32), commercial planar patch clamp systems can now perform simultaneous, automated electrophysiological experiments in 384- or 768-well arrayed formats (82). These systems are increasingly used to screen compound libraries against ion channel targets, expediting the drug discovery process. However, the system requires a suspension of dissociated cells, which precludes recording in situ and over extended time courses. As a result, planar patch clamp is more suited for high-throughput screening of ion channel pharmacology than for analysis of cellular or tissue-level voltage signaling.

3.3.2. Nanopipettes and nanopillar electrodes.

The development of nanopipettes and nanopillar electrodes has enabled electrode-based access to broader length scales. Nanopipettes are a smaller version of sharp microelectrodes; their reduced size brings flexibility and access to small cellular compartments. Nanopipettes have been used to record from micron-scale dendritic spines, providing some of the first direct evidence for functional voltage compartmentalization in these structures (45). In addition, the flexibility of the nanopipette can stabilize in vivo recordings during animal behavior (46). As with sharp electrodes, the high tip resistances of the nanopipettes filter the signal and create a shunt that changes the observed V_mem, so the output signal must be processed to obtain accurate results. Nanopillar electrode arrays enable simultaneous intracellular recordings from many cells in a culture via a grid of nanoscale electrodes (94, 113). These platforms facilitate mapping of electrical activity across neuronal circuits or beating cardiomyocytes. However, the cells of interest must be cultured directly onto the electrode grid; nanopillar electrode arrays are not yet capable of in situ mapping of tissue V_mem. Overall, electrode engineering offers a promising avenue for increasing the spatial reach of electrophysiology.

4. CALIBRATED OPTICAL SIGNALS FOR RECORDING ABSOLUTE MEMBRANE POTENTIAL

A variety of optical V_mem measurement strategies use fluorescence as a proxy for V_mem, employing small-molecule or protein-based sensors engineered to display V_mem-sensitive fluorescence (Table 2). Optical platforms generally enjoy excellent spatial resolution, low invasiveness, and medium throughput. However, achieving the requisite temporal resolution, voltage resolution, and signal-to-noise ratio has required more engineering. Optical strategies fall into four categories (single-color intensity, ratio based, pump probe, and single-color lifetime; Figure 3). In all cases, rigorous calibration must be performed to relate the optical signal to the underlying absolute V_mem, which is most effectively achieved by performing the optical recording while V_mem is tuned across a range of known values. As we discuss in this section, a key difference among the various fluorescence-based strategies is voltage resolution, which is a result of the reproducibility of this calibration over time and between cells.

Figure 3.

Optical strategies for reporting absolute V_mem. (a) Single-color fluorescence intensity recordings detect changes in V_mem by changes in a sensor’s fluorescence quantum yield (shown), extinction coefficient, or cellular concentration. This technique generally cannot report absolute V_mem optically, although estimates can be made if calibrations are performed for every cell of interest. (b) Single-color fluorescence lifetime can report absolute V_mem, particularly if the V_mem sensor operates via photoinduced electron transfer.(c) FRET-based sensors show differences in the FRET ratios of two fluorophores, often resulting from V_mem-dependent changes in the distance between the two. The diagram shows a FRET-oxonol system. (d) The excitation and/or emission spectra of electrochromic dyes depend on the electric field in the plasma membrane. (e) Certain GEVIs show V_mem-related differences in the temporal dynamics of the absorption of their excited state. (f) SRS imaging reports changes in V_mem-dependent vibrational frequencies. Abbreviations:λ, wavelength; depol, depolarization; FRET, Förster resonance energy transfer; GEVI, genetically encoded voltage indicator; hyperpol, hyperpolarization; SRS, stimulated Raman scattering; V_mem, membrane potential.

4.1. Strategies for Calibration of Optical Membrane Potential Reporters

To calibrate an optical V_mem recording platform, one must be able to measure the optical response across a range of V_mem values. By fitting this measured relationship to a function, one can then convert the optically determined parameter into V_mem values and, equally importantly, estimate the accuracy of the optical V_mem determination. For example, for a linear relationship, one would determine the slope and the y-intercept and use those values to relate the optically recorded parameter to V_mem. To this end, there are three common strategies for setting cellular V_mem: electrode-based, pharmacological, and optogenetic.

4.1.1. Electrode-based calibration.

The most accurate method for calibrating fluorescence with respect to V_mem is whole-cell voltage clamp electrophysiology, which uses an intracellular recording configuration to inject current until the cell reaches the desired V_mem. The response across a range of precisely set V_mem can then be evaluated, determining the relationship between the optical readout and the underlying voltage. However, this approach is best when performed in isolated, single cells, and it suffers from the limitations of electrophysiology discussed above.

4.1.2. Pharmacological calibration.

If the system under study is inaccessible to electrode-based strategies, either pharmacology or intrinsic V_mem signals can be used to approximate the relationship between the optical signal and absolute V_mem. Reference values of V_mem have been established using ionophores for the primary ion establishing V_mem (39, 51, 74, 81), high extracellular K⁺ (120–150 mM) (55, 74), and cell death or fixation with paraformaldehyde (PFA) (51). PFA and ionophores produce a more reliable 0-mV point than do high K⁺ treatments, but they interact with the plasma membrane itself, which can change the fluorescence properties of the voltage sensor (79). Furthermore, there is no reliable way to access the sensitivity of the sensor (i.e., slope of signal versus V_mem) from a single pharmacological set point. However, in samples with well-described intrinsic electrical responses, a V_mem event of known magnitude can provide a standard for approximate determination of absolute V_mem sensitivity. Demonstrated reference V_mem signals include sustained hyperpolarization in Purkinje cells (19), depolarization resulting from glutamate uncaging (107), and back-propagating APs in dendrites (85). Such strategies are noninvasive but are limited to the small subset of samples where a useful and consistent V_mem signal can be identified.

4.1.3. Efforts toward optogenetic calibration.

In many cases, this limited set of calibration options is inadequate. Recently developed optical strategies for V_mem control may eventually provide an alternative to the above approaches, although they cannot yet tune V_mem to defined voltages. Engineering efforts for optical V_mem actuators have focused on channelrhodopsins, proteins that pass current in response to light. For setting V_mem optically, step-function opsins may be particularly useful, as they continue to pass current for minutes after initial activation (12, 119). Recently, light sensitivity was engineered into other ion channels via domain insertion, allowing diverse channels to serve as optogenetic actuators (27). These strategies could eventually allow precise setting of V_mem without an electrode.

4.2. Single-Color Fluorescence Intensity Recordings Are Difficult to Relate to Absolute Membrane Potential

In single-color fluorescence intensity–based recordings, V_mem affects the quantum yield or absorption coefficient of a sensor, producing changes in its fluorescence emission. Fluorescence intensity of a voltage sensor in a membrane is recorded over seconds or minutes, enabling detection of V_mem changes and characterization of V_mem waveforms. A variety of sensor architectures have been documented (57, 77, 115, 116), including both small-molecule dyes and genetically encoded voltage indicators (GEVIs). Many sensors display sufficiently fast temporal resolution to follow millisecond-time APs. State-of-the-art systems can achieve cell-resolved recording of single APs in vivo from superficial brain regions, often with subcellular detail (1, 3, 88). However, single-color fluorescence intensity is best suited for V_mem event detection and generally cannot report absolute V_mem. Methods that rely solely on fluorescence intensity to monitor V_mem display large differences in baseline signal between cells and over minutes to hours. These variations are V_mem independent, arising from variable sensor loading or trafficking, cell morphology, sensor photobleaching, fluorescence quenching, and illumination intensity. As a result, calibrations made on one cell at a given time cannot be extended to other cells or even to the same cell hours later.

Fluorescence intensity:

the total amount of fluorescence signal from a sample; for fluorescent sensors, intensity depends on the amount of sensor present, the amount of light absorbed, and the quantum yield

Voltage indicator:

a small molecule or protein fluorophore that exhibits V_mem dependence in some or all of its fluorescent properties (e.g., intensity, wavelength, lifetime)

Some attempts have been made to calibrate fluorescence intensity to known V_mem at each cell or structure of interest, but caution must be employed in interpreting these results. Electrophysiological calibration of each individual cell is often challenging or impractical, so fluorescence intensity calibrations are generally performed with less accurate pharmacological strategies. In some cases, this approach is the only viable way to interrogate V_mem—for example, gramicidin-based calibrations and the V_mem indicator Archaerhodopsin (Arch) revealed subcellular differences in the AP waveform (39). However, pharmacological calibration of a single-color V_mem sensor is prone to artifacts. When calibrating from a single pharmacological set point, rather than a range of V_mem set by an electrode, the researcher must assume that the fractional change in fluorescence (%ΔF/F) per mV of the indicator is the same for all samples. In reality, the ΔF/F response to a given ΔV_mem will depend on the ratio of productively engaged sensor to background sensor, which may vary in space with trafficking or loading of the sensor. Furthermore, even with electrode-based calibration, many voltage sensors display sensitivity to membrane composition (36), which varies not only between cells but also within cells with complex membrane structures (10).

4.3. Ratio-Based Fluorescent Sensors Improve Reproducibility

Ratio-based voltage sensors show changes in the relative intensities of two fluorescence channels. The primary strategies for ratio-based fluorescent V_mem sensors rely on either Förster resonance energy transfer (FRET)-based indicators or electrochromic dyes. If the stoichiometry between the two signals is known, then the second channel can be used to correct for variability arising from cell morphology and dye loading. Ideally, the corrected signal would then be a stable proxy for absolute V_mem over time and across many cells without recalibration. In practice, ratio-based sensors achieve better absolute V_mem resolution than single-color intensity-based techniques, but they generally do not have sufficient resolution to quantify the biologically relevant V_mem differences between cells (~10–20 mV).

Ratio-based:

expressed as a ratio between two components; in ratio-based fluorescent imaging, the ratio of signal at two different wavelengths changes in response to the phenomenon of interest

4.3.1. Förster resonance energy transfer–based membrane potential sensors.

FRET-based V_mem sensors exhibit V_mem-dependent energy transfer from a donor to an acceptor fluorophore. Because the fluorescence intensities of the donor and acceptor are voltage sensitive in opposite directions, the ratio of the two can provide better fractional responses and a higher signal-to-noise ratio than a single color alone. Although a variety of FRET-based V_mem sensors exist, we are not aware of their use to document absolute V_mem, largely due to variable stoichiometry between the donor and acceptor. For genetically encoded systems (3, 4, 125), different rates of photobleaching, folding, and productive trafficking to the plasma membrane lead to variability in FRET ratios between cells. For small-molecule FRET-oxonol systems (15, 35), differences in loading between the two lipophilic indicators lead to variability in the %ΔF/F (72). As a result, FRET ratio measurements are not extensible between cells. Furthermore, the FRET-based reporters that undergo conformational changes to sense V_mem suffer from reduced temporal resolution and toxicity from capacitive load. In practice, these probes are primarily used to detect APs, with the second color serving to improve the signal-to-noise ratio or reduce motion artifacts.

4.3.2. Electrochromic ratio-based membrane potential sensors.

Electrochromic dyes, such as the ANEPPS (68) and ANNINE (34) classes, show V_mem-dependent excitation and/or emission spectra. A ratio signal is obtained by comparing the emission in a fixed band produced by excitation at two different wavelengths (or vice versa with a fixed excitation and two emission bands). Because electrochromic dyes are based on a charge shift mechanism, they have submillisecond temporal responses but display low sensitivity (~10% change per 100 mV) (33, 79). Normalized ANEPPS fluorescence ratios are accurate to approximately 5 mV for quantifying V_mem changes on an individual cell (17, 122). In combination with random access microscopy, di-8-ANEPPS ratios can report the AP waveform in absolute V_mem optically with excellent spatiotemporal resolution (exposure times of 0.5 ms and spatial resolution of 2 μm) (17). However, the V_mem calibration does not translate between cells and must be performed with an electrode on each cell of interest to achieve this accuracy. The origin of the variability of the absolute ratio between cells is unclear; it likely depends on a complex combination of temperature, membrane composition, and probe loading.

4.3.3. Molecular beacons in ratio-based membrane potential recording.

A recent, promising architecture for ratio-based V_mem sensors uses a molecular beacon, where the sensor contains one fluorophore that is voltage sensitive and a second fluorophore that is a voltage-independent reference. This strategy was tested with the GEVI Arch(D95N) fused to green fluorescent protein (GFP), but absolute V_mem quantification was not possible because of differential photobleaching of GFP and Arch(D95N) (41). More recently, the molecular beacon system was implemented with small-molecule V_mem sensors in Voltair (98), which comprises the VoltageFluor (VF) RVF5 (56) and the reference fluorophore ATTO647N on a DNA chassis. By normalizing the RVF5 signal to ATTO647N, Voltair was able to estimate organellar V_mem optically based on a pharmacological calibration, but absolute V_mem resolution of this technique has not yet been determined.

4.4. Pump-Probe Recordings: Accurate, with Complex Instrumentation

The temporal dynamics and vibrational signature of a probe do not depend on probe concentration; therefore, they can be used as a reproducible proxy for absolute V_mem. The temporal dynamics of the GEVI Arch(D95H)-eGFP can report absolute V_mem between cells with 10-mV resolution (41), the best yet reported for an optical system. However, access to and interpretation of such data is challenging for most laboratories, limiting this optical system’s utility as a general V_mem sensing platform. Stimulated Raman scattering (SRS) has also been investigated as a strategy for absolute V_mem recordings, but its V_mem resolution has yet to be determined. Recently, the SRS signal of near-infrared opsins was shown to be sensitive to bulk depolarization in Escherichia coli membranes (61). Moreover, label-free detection of neuronal APs with SRS has been reported, although the mechanism of such V_mem sensitivity remains unclear (62). Further investigations of temporal dynamics and SRS as V_mem reporting platforms are of great interest, especially with the aim of translating them into more widely available systems.

4.5. Fluorescence Lifetime Offers Accessible Absolute Optical Membrane Potential Recordings

Fluorescence lifetime (τ_fl), or the amount of time that probe molecules remain in the fluorescent excited state, has recently garnered attention for its ability to report cellular properties (e.g., Ca²⁺ concentration or V_mem) absolutely using a single fluorescence channel (118). τ_fl is largely independent of fluorophore concentration, cellular morphology, and photobleaching, although it displays sensitivity to parameters such as temperature and viscosity. The primary downside to fluorescence lifetime imaging (FLIM) is its poor temporal resolution, with most acquisition times on the order of seconds.

Fluorescence lifetime (τ_fl):

the amount of time that a fluorophore remains in the fluorescent excited state before returning to the ground state, via either fluorescence ora nonradiative pathway

Fluorescence lifetime imaging (FLIM):

a microscopy technique in which both fluorescence intensity and τ_fl are recorded at each pixel of an image

4.5.1. Fluorescence lifetime of genetically encoded voltage indicators.

τ_fl-based V_mem recordings can be performed if τ_fl of a sensor changes reproducibly with V_mem. Many GEVIs sense V_mem through complex photocycles or multiple conformational states; this leads to low sensitivity in τ_fl or nonlinear relationships between τ_fl and V_mem. For example, the rhodopsin derivative Arch(D95H) was first evaluated in FLIM for its absolute V_mem-reporting capabilities, but its τ_fl–V_mem relationship was difficult to interpret (41). Comparison of the V_mem sensitivity of three GEVIs in two-photon illuminated FLIM (16) revealed that only CAESR (125) was suitable for absolute V_mem recording. CAESR exhibits 20-mV accuracy for quantifying V_mem changes on a given cell in a one-second bandwidth, but it displays too much variability to extend optical V_mem calibrations between cells (16). This between-cell variability may arise from the complex dynamics of rhodopsin fluorescence, coupled with intracellular fluorescence signal from incorrectly trafficked protein. V_mem-sensitive FLIM has also been demonstrated for FRET-oxonol V_mem sensors(31) and for accumulation-based sensors of mitochondrial potential (83), but no estimation of the absolute V_mem resolution was made in these systems.

4.5.2. Fluorescence lifetime of VoltageFluor small-molecule dyes.

τ_fl-based absolute V_mem imaging was developed further in our laboratory, achieving stable calibration between cells through the use of small-molecule VFs as voltage sensors. The voltage-sensing trigger for VFs is photoinduced electron transfer (78), which translates to a rapid, linear V_mem response in both fluorescence intensity (11, 78) and τ_fl (59). The τ_fl of VF2.1.Cl (VF-FLIM) can be calibrated to report absolute V_mem with a resolution of 5 mV for quantifying voltage changes on individual cells. These calibrations can be extended from one cell to another while retaining a single-trial V_mem resolution of 20 mV, which is sufficient to detect biologically relevant variation in resting V_mem, especially once multiple measurements are averaged. Because VF-FLIM calibrations are consistent for a given cell line, we were able to extend V_mem calibrations across thousands of cells after an initial electrode-based calibration on only a few cells. We applied VF-FLIM to study epidermal growth factor signaling and found a 15-mV hyperpolarizing response in carcinoma cells, showcasing its potential for elucidating diverse signaling roles of V_mem.

In addition to its improved V_mem resolution, VF-FLIM shows many promising attributes as a V_mem sensing platform. The increased stability of the calibration enhances throughput; an individual researcher using VF-FLIM could perform V_mem recordings at the speed of optical imaging (thousands of cells in an afternoon) rather than electrophysiology (~10 cells per day). Importantly, the temporal and spatial resolution of FLIM-based strategies are related: Acquisition rates as fast as 200 Hz (exposures of 5 ms) can be obtained with reduced spatial resolution, and spatial resolution can be improved from 5 microns (the width of a binned pixel) to approximately 1 micron at the expense of imaging speed (Miller lab, unpublished data). The VF-FLIM technology can be further improved through combination with red-shifted V_mem sensors (28, 43) and targeting of sensors to specific cell populations in thick tissue (29, 67, 84), both of which are ongoing efforts in our lab. Overall, FLIM-based absolute V_mem imaging extends the timescale over which V_mem can be recorded and displays some of the best V_mem resolution to date for optical systems.

5. OUTLOOK

V_mem signals run the gamut from milliseconds to days and from the subcellular to the tissue level (Figure 1). The existing V_mem toolkit can record most cellular V_mem signals and some subcellular ones, with time resolutions ranging from milliseconds to minutes. V_mem signals delocalized across tissues are still challenging to interrogate, especially those occurring over hours to days.

Electrode-based absolute V_mem recordings have defined the field for years and will remain pivotal as engineering efforts continue to improve throughput and reduce invasiveness. However, progress in optical approaches for absolute V_mem recording has already expanded the range of space and time over which V_mem can be recorded. One of the key limitations to the scope of optical, absolute V_mem recording is the need for calibration in the cell or tissue of interest; we excitedly await improved optical actuators to facilitate V_mem calibrations in diverse systems. Further improvements in voltage-sensitive dyes and proteins stand to increase the absolute V_mem resolution of existing optical approaches. Improvements to photon-counting FLIM hardware, as well as fast frequency-domain FLIM strategies (91), are promising avenues for faster lifetime recordings. Optical V_mem recording in vivo can be multiplexed across more cells and larger areas by combining it with lattice light-sheet imaging (24) and other fast volumetric imaging strategies (50).

With these technological advances, we are constructing a picture of V_mem across biological time and space scales. Nevertheless, our atlas remains far from complete. In many cases, the information carried by V_mem is unknown, and the molecular mechanisms of V_mem response remain obfuscated. Each newly observed V_mem signal raises additional, exciting questions: How does the cell bring about these V_mem changes? Which molecules respond to V_mem? How are these signals transduced to downstream cellular processes? Uncovering the signaling networks surrounding V_mem will require platforms that interface with other cell profiling and omics techniques, such as the recently pioneered Patch-Seq system (18). With such efforts, we move ever closer to a systems-level understanding of V_mem in all of its diversity.

SUMMARY POINTS.

1. V_mem influences cellular processes on timescales ranging from seconds to days and across length scales from microns to centimeters. Beyond excitable cell APs, V_mem also contributes to cell growth, differentiation, and organellar function.

2. V_mem recording strategies generally fall into one of two categories: electrode-based or optical recordings.

3. Electrode-based V_mem recordings have high temporal resolution and voltage resolution, but they are invasive, have low throughput, and only record V_mem at a single point.

4. Recent advances in electrodes and electrode arrays have allowed for electrophysiology of smaller cellular structures, as well as improved throughput and multiplexing.

5. Optical V_mem recordings can track V_mem noninvasively across many cells at once, but most optical strategies cannot quantify biologically relevant V_mem signals (i.e., measure absolute V_mem).

6. Recent improvements in certain optical V_mem recording platforms have improved V_mem resolution, thereby enabling quantification of absolute V_mem. In particular, lifetime or temporal dynamics of a sensor’s fluorescence can quantify V_mem across longer space and timescales.

7. Optical recordings are indirect measurements of V_mem, so they always require calibration to report absolute V_mem. Various optical V_mem recording architectures show differences in the reproducibility of this calibration; this is usually the limiting factor in absolute V_mem resolution.

FUTURE ISSUES.

1. Many areas of V_mem biology are understudied, especially V_mem changes occurring over long periods of time and in very small or very large structures. Continued efforts toward developing platforms for absolute V_mem recordings are necessary before we can fully document or understand V_mem signaling in these areas.

2. Electrical signaling in vivo is likely delocalized across multiple cells by gap junctions. As tools tailored to measure V_mem in tissue are developed, further investigations into cell–cell coupling and long-range V_mem signals are needed.

3. Absolute V_mem recording platforms with signals that are stable over days are needed. Although nanopillar electrode arrays are a major step forward, electrode-based platforms are generally too invasive for chronic recordings. A key challenge for optical platforms is maintaining stable levels of V_mem sensor in the sample during extended recordings.

4. Standard metrics for—and routine determination of—absolute V_mem resolution should be incorporated into the evaluation and presentation of optical platforms for quantitative V_mem recordings (ratio-based recordings, lifetime-based recordings, SRS, etc.).

5. More generalizable strategies for calibration of V_mem are needed. In systems that are accessible optically but not with electrodes, it is difficult to determine the accuracy of optical absolute V_mem measurements.

6. Although FLIM and temporal dynamics–based strategies have accessed resolutions in the range of 10–20 mV, many interesting V_mem changes are on the scale of 5–10 mV. Further engineering of these optical sensors and platforms could enable detection of subtle V_mem changes.