performed in quadruplicate and medium alone serves as the control. Results are calculated as percent control incorporation of [r4C]leucine. Discussion The diphtheria toxin-based fusion toxin proteins that we have worked with in this laboratory include those targeted by the following ligands: (YMSH, IL-2, mIL-4, IL-6, CD4, sIL-15, SP, GRP, and IL-7. Of these, only the CD4 and sIL-15 receptor-targeted fusion protein constructs could not be purified from inclusion bodies and required purification over anti-diphtheria toxoid immunoaffinity columns. In both instances, these fusion protein toxins did not form inclusion bodies and were found in the first supernatant fraction after lysis of recombinant E. coli. All of the above-mentioned diphtheria toxin-based fusion proteins were constructed as the DABsg9 form and possess wild-type catalytic and transmembrane domains of diphtheria toxin. For constructs in which deletions, exchanges, or even single amino acid substitutions were made in these domains, modification of the purification schemes described above were usually required.
[ 191 Green Fluorescent Protein-Based Sensors for Detecting Signal Transduction and Monitoring Ion Channel Function By MICAHS. SIEGEL and EHUDY.ISACOFF
Introduction Measuring signal transduction is a fundamental problem in studying information processing in the nervous system. To address this problem, we have designed a family of detectors that are chimeras between signal transduction proteins and fluorescent proteins. The prototype sensor is a novel, genetically encoded probe that can be used to measure transmembrane voltage in single cells. In this chapter, we describe a modified green fluorescent protein (GFP) fused to a voltage-sensitive K+ channel so that voltage-dependent rearrangements in the K+ channel induce changes in the fluorescence of GFP. The probe has a maximal fractional fluorescence change that is comparable to some of the best organic voltage-sensitive dyes. Moreover, the fluorescent signal is expanded in time in a way that
Copyright D Moo by Academic Press All rights of reproduction in any form reserved. 0076.6EJ9/00 $3O.CHl
makes the signal 30-fold easier to detect than that of a traditional linear dye. Sensors encoded into DNA have the advantage that they may be introduced into an organism noninvasively and targeted to specific brain regions, cell types, and subcellular compartments. Fluorescent
Dyes as Sensors
Fluorescent indicator dyes have revolutionized our understanding of cellular signaling by providing continuous measurements of physiological events in single cells and cell populations. There are two major impediments to progress with indicator dyes: (1) the lack of direct, noninvasive assays for most cellular communication events and (2) the difficulty in observing selected cell populations. In practice, these dyes must be synthesized chemically and introduced as hydrolyzable esters or by microinjection.1-3 Delivering indicator dyes to specific cell populations has proved to be a difficult problem. In the absence of such localization, optical measurements in neural tissue usually cannot distinguish whether a signal originates from activity in neurons or glia; neither can it distinguish which types of neurons are involved. Chimer% Protein Sensors To increase our understanding of the signaling events that govern development, sensory transduction, and learning and memory, it is necessary to expand the range of signals that can be detected and to develop means of targeting sensors to specific cellular locations. One general approach to this problem is to construct sensors out of proteins, the biological molecules that transduce and transmit cellular signals. This approach would harness the high sensitivity and specificity of biological systems, which can detect an enormous range of signals with exquisite sensitivity. Moreover, it could permit the detection of signaling events throughout a signal transduction cascade: from the earliest stage of membrane transduction, through the network of signaling relays and amplification steps, and finally to downstream events that occur anywhere from the nucleus back to the plasma membrane. Because protein-based sensors are encoded in DNA, they can be placed under the control of cell-specific promoters introduced in vivo or in vitro ’ L. Cohen and S. Lesher, Sot. Gen. Physiol. Ser. 40,71 (1986). ’ R. Y. Tsien, Annu. Rev. Neurosci. 12,227 (1989). 3 D. Gross and L. M. Loew, Methods Cell Biol. 30,193 (1989).
by gene transfer techniques, and even targeted to specific subcellular compartments by protein signals recognized by the protein-sorting machinery in the cell. In addition to solving the problem of sensor production and targeting, DNA-encoded sensors have the advantage that, unlike organic dyes, they can be rationally tuned by modification of their functional domains with mutations that are known to adjust their dynamic range of operation. Finally, by creating “biological spies” out of native proteins, it is possible to avoid the introduction of foreign substances that could interfere with cellular physiology. We have designed a family of protein-based detectors for measuring signal transduction events in cells (Fig. 1A). These sensors are chimeras between a signal transduction protein fragment (detector) and a fluorescent protein (reporter). This chapter describes a prototypical example of this class of chimeric proteins: a GFP sensor that we have engineered to measure fast membrane potential changes in single cells embedded within a population of cells.
Detector Signal Transduction
FIG. 1. GFP-based sensors of cell signaling. (A) GFP is fused in-frame into the middle of a signal transduction protein (detector) so that conformational rearrangements in the detector perturb the fluorescence of GFE’. The GFl’ has been sensitized to the rearrangements of the detector protein by deletion of the last eight residues in the C terminal, which are disordered in the crystal structure. 2o ( B ) Flash is a chimeric protein in which GFP has been fused into the Shaker potassium channel.
Fluorescent Shaker K’ Channel: Measure Voltage
In outline, we have engineered a chimeric protein that is a modified GI? fused in-frame at a site just after the sixth transmembrane segment (S6; Fig. 1B) of the voltage-activated Shaker K+ channel.5-7 The detailed description and characterization of this sensor protein have been described previously.8 (Henceforth this chimeric protein is called Flash, for Fluorescent Shaker.) To prevent Flash from loading down target cells with an additional potassium current, we engineered a point mutation into the pore of the channel. This mutation prevents ion conduction but preserves the gating rearrangements of the channel in response to voltage changes. Our idea was that the voltage-dependent rearrangements in the channel could be transmitted to GFP, resulting in a measurable change in its spectral properties. The Shaker-GFP chimeric protein reports changes in membrane potential by a change in its fluorescence emission. In addition, the fluorescence response is amplified in time over the electrical event, drastically increasing the optical signal power per event. Temporal amplification in Flash is due to the response kinetics of the Shaker channel. Taken together, the properties of genetic encoding and temporal amplification allow the sensor to be delivered to specific cells in which action potentials may be detected with standard imaging equipment. Blueprint
Shaker K+ Channel
We have characterized the behavior of Flash in single Xenopus laevis oocytes by cRNA injection and voltage-clamp fluorimetry. Voltage steps from a holding potential of -80 mV evoke fluorescent emission changes and gating currents, but no ionic currents (Fig. 2). The relation of the steady state fluorescence change to voltage is sigmoidal and correlates closely with the steady state gating charge-to-voltage relation (Fig. 3), indicating that in Flash, the fluorescent emission of GFP is coupled to the voltage-dependent rearrangements of the Shaker channel. The dynamic range of Flash is steep, from approximately -50 to -30 mV.
4 M. Chalfie, Y. Tu, G. Euskirchen, W. W. Ward, and D. C. Prasher, Science 263, 802 (1994). 5 3. L. Tempel, D. M. Papazian, T. L. Schwaa, Y. N. Jan, and L. Y. Jan, Science 237,770 (1987). 6A. Baumann, A. Grupe, A. Ackermann, and 0. Pongs, EMBO J. 7,2457 (1988). ‘A. Kamb, J. Tseng-Crank, and M. A. Tanouye, Neuron 1,421 (1988). ’ M. S. Siegel and E. Y. Isa&f, Neuron 19,735 (1997).
Time (ms) FIG. 2. Cell membrane potential modulates fluorescence output in Flash. Simultaneous two-electrode voltage-clamp recording and photometry show current and fluorescence changes in response to voltage steps (V) between -60 and 10 mV, in lo-mV increments. The holding potential was -80 mV. Flash exhibits on-and-off gating currents (Zp) but no ionic current because of the W434F mutation of Shaker. Integrating the gating current gives the total gating charge (Q) moved during the pulse. Flash fluorescence (F) decreases reversibly in response to membrane depolarizations. Traces are the average of 20 sweeps. Fluorescence scale, 5% AFIF.
Impact on Target Cell
To prevent Flash from altering the physiology of cells in which it is expressed, we made the W434F point mutation in the Shaker pore to prevent ion conduction. This mutation blocks conduction by locking a gate in the pore into a closed conformation. Normally this gate closes slowly during sustained depolarization, producing slow inactivation. Other gating processes and rearrangements remain normal in the mutant channel, including activation in response to depolarization, opening of the activation gate, ball-and-chain (N-type) inactivation, and the rearrangement that consolidates slow inactivation and changes the fluorescence of GFP.8-‘2 The use of the W434F mutation works to prevent ion conduction in 9 E. Perozo, R. MacKinnon, F. Bezanilla, and E. Stefani, Neuron l&353 (1993). lo F. Bezanilla, E. Peroza, D. M. Papazian, and E. Stefani, Science 254,679 (1991). I1 Y. Yang, Y. Yan, and F. J. S&worth, J. Gem Physiol. 109,779 (1997). i2 E. Loots and E. Y. Isacoff, .I. Gen. Physiol. 112,377 (1998).
FIG. 3. Flash fluorescence is correlated with Shaker activation gating. The voltage dependence of the normalized steady state gating charge displacement (Q) overlaps with the normalized steady state fluorescence change (F). Both relations were fit by a single Boltzmann equation (solid lines). Values are means + SEM from five oocytes.
nonexcitable cells, such as Xenopus oocytes, where Flash subunits are the only subunits from the Shaker K+ channel subfamily that are expressed, so that Flash channels form as nonconducting (permanently inactivated) homotetramers. However, in excitable cells such as neurons and muscle, where native subunits from the Shaker subfamily (which carry the wildtype W at position 434) are expressed, Flash subunits may co-assemble with those native subunits to form heterotetramers. In such heterotetramers, the slow inactivation gate will shut more quickly than in wild-type (W434) homotetramers, meaning that the properties of the K+ conductances in the cell will be altered. Although the effect of altering one class of voltagegated K+ channels pharmacologically is often subtle, such heterotetrameric channels may nevertheless affect the functional properties of the cells, a side effect of sensor expression that would be better avoided. Our approach to circumventing coassembly between Flash and native channels in excitable cells is to link four Flash cDNAs in tandem in such a way that the four subunits of the channel are covalently attached.13 This approach has been used earlier to force subunits to assemble in a known I3 E.
Isacoff, Y. N. Jan, and L. Y. Jan, Nature
stoichiometry.‘4-16 The expectation is that linked Flash constructs should assemble into Flash homotetramers even in excitable cells because of the higher likelihood of intramolecular assembly between linked Flash subunits than intermolecular assembly with native channel subunits.
Modifying Dynamic Range of Fluorescent through Mutagenesis
Shaker K+ Channel
Given the narrow range of voltage over which Shaker channels gate, and over which Flash modulates its brightness (Fig. 3) it is clear that some voltage signals will be reported more efficiently, while others may be missed altogether. Because mammalian neurons tend to rest at about -70 mV, small excitatory and inhibitory postsynaptic potentials will not fall within the dynamic range of Flash (-50 to -30 mV). However, suprathreshold excitatory postsynaptic potentials and action potentials should be reported. We examined the response of Flash to physiologically realistic voltage traces. Voltage transients measured in response to light in a variety of salamander retinal cell types17 were applied via a voltage clamp to oocytes expressing Flash. Flash fluorescence reflected the dynamics of on-bipolar cell transients quite well, because the response of the on-bipolar cell is within the dynamic range of Flash (-50 to -30 mV). However, Flash did not capture the response of wide-field amacrine cells, because the main response of these cell types occurs from -70 to -50 mV, outside of its dynamic range. One advantage of using the Shaker channel is that many mutations have been described that produce unique alterations in its voltage dependence and kinetics. This provides flexibility in tuning Flash to an operating range that best suits the signals of interest. For example, we have made versions of Flash with a more negative operating range based on mutations identified by Lopez et al. I8 These provide a good optical sensor for measuring the voltage waves from wide-field amacrine cells.r9 Similarly, mutations
I4 R S. Hurst, M. P. Kavanaugh, J. Yakel, J. P. Adehnan, and R. A. North, J. Biol. C&n. 267,23742 (1992). I5 E. R. Liman, J. Tytgat, and P. Hess, Neuron, 9, 861 (1992). l6 D. T. Liu, G. R. Tibbs, and S. A. Sigelbaum, Neuron 16,983 (1996). I7 B. Roska, E. Nemeth, and F. S. Werblin, J. Neurosci. l&34.51 (1998). l8 G. A. Lopez, Y. N. Jan, and L. Y. Jan, Neuron 7,327 (1991). I9 G. Guerrero, M. S. Siegel, B. Roska, C. Dean, and E. Y. Isacoff, in preparation (2000).
that alter the kinetics of the rearrangement that consolidates channel inactivation predictably alter the kinetics of the fluorescence output of FlaSh.l’ Designing
In general, we have found that the polymerase chain reaction (PCR) works well for inserting GFP into signal transduction proteins and that channel proteins, in particular, are surprisingly tolerant to GFP insertions. In the case of Shaker, we find that GFP can be inserted at the N terminal, at the C terminal, and also at a variety of internal sites. In our experience, these chimeric proteins are usually fluorescent, and the signal transduction protein usually functions. The tolerance of many proteins to GFP insertion is probably due to the structure of GFP, in which the N and C termini emerge in close proximity to one another on the same side of the barrel.*’ Sometimes we find that the insertion of GFP can alter the kinetics of the signal transduction protein (Siegel and Isacoff, unpublished data, 1999). We have found that it can be helpful to examine sequences from a wide variety of homologous proteins (e.g., Shaker, Shab, Shaw). Often homology can give insight into those regions in the protein sequence that are highly conserved through evolution, and that therefore might be intolerant to GFP insertions. For example, we never were able to make functional proteins in which GFP was fused into regions near the fourth transmembrane segment (S4) in Shaker. We had hoped that this region would be interesting because of the high homology between S4 transmembrane segments in a variety of channel proteins, and because the Shaker S4 segment is known to undergo conformational rearrangements in response to transmembrane voltage. Unfortunately, the Shaker channel was intolerant to GFP insertions in three regions before (external side of the membrane) and after (internal side of the membrane) the S4 region. These fusion proteins were not fluorescent, nor did the channel function, implying that they interfered with protein folding, assembly, or stability. Designing
We have found it useful to consider endogenous targeting sequences that are present in the signal transduction protein of interest. For example, the Shaker channel contains a PDZ interaction domain at its C terminus that targets Shaker preferentially to postsynaptic specializations.21-23 This PDZ “M
Ormo A. B. Cubitt, K. Kallio, L. A. Gross, R. Y. Tsien, and S. J. Remington, Science 273, 1392’(1996). ‘IF J. Tejedor, A. Bokhari, 0. Rogero, M. Gorczyca, J. Zhang, E. Kim, M. Sheng, and v. Budnik, .I. Neurosci. 17, 152 (1997). ” K. Zito, R. D. Fetter, C. S. Goodman, and E. Y. Isacoff, Neuron 19,1007 (1997). 23 K. Zito, D. Pamas, R. D. Fetter, E. Y. Isacoff, and C. S. Goodman, Neuron 22,719 (1999).
interaction domain can target heterologous proteins to the synapse22,23 and therefore can be used to target other GFP-based sensors. In contrast, some mutations in the PDZ interaction domain prevent targeting22 and can be used to ensure that Flash is distributed more uniformly throughout the cell. Practical Issues in Screening Protein-Based Sensors
for Green Fluorescent
One strategy for producing GFP-based sensors is first to find a fusion protein that exhibits a small fluorescence change in response to a signal transduction event. The next step is to modify the fusion protein in order to increase the size of its fluorescence change. In our experience, when looking for the presence of a small fluorescence change, it is extremely important to have a well-controlled, reproducible method of activating the signal transduction cascade. For example, in the case of Flash, it was important to test the system with rapid voltage jumps. In our experience with ligand-gated receptors, it has been useful to be able to wash ligands into the solution quickly and reproducibly, and then to remove them quickly and reproducibly. Reproducibility in the stimuli is important because it is often desirable to average a number of stimulus presentations to find a weak fluorescence change in a noisy background. When the onset of the stimuli is ambiguous (e.g., with a ligand wash), it is possible to confuse bleaching, arc-lamp jitter, and other optical artifacts for a genuine fluorescence change. It is worthwhile to test a positive control in parallel with the fusion protein of interest. In our experience, most GFP fusion proteins do not exhibit a fluorescence change. It is obviously important to know whether the fusion protein is a failure or the test apparatus is broken. We have found that organic calcium-sensitive dyes can serve as a useful control in this context. Optical Issues in Screening Protein-Based Sensors
We have sometimes found that different GFP mutants can behave differently when fused into the same signal transduction protein. For example, wild-type GFP inserted into the Shaker potassium channel (e.g., Flash, as described above) exhibits a fluorescence change on membrane depolarization. However, the mutant S65TGFP= inserted into Shaker at the same 24 R. Y. Tsien,
67, 509 (1998).
location produces a fusion protein that is brighter than Flash but gives a fluorescence change that is approximately lo-fold smaller. It is important to be aware of different excitation and emission spectra of different GFP mutants, and to choose appropriate filters for looking at the fluorescent sensor. For Flash, we use an HQ-GFP filter (Chroma Technology, Brattleboro, VT) with the following filters: excitation, 425-475 nm, dichroic, 480-nm long-pass; emission, 485-535 nm. We have found that Flash is also visible with a standard fluorescein filter set (e.g., Chroma Technology HQ-FITC filter set). The relative fluorescence change with an HQ-FITC set is similar to that with an HQ-GFP set; however, the total fluorescence emission is smaller because HQ-GFP is better matched to the excitation spectra of wild-type GFP. Finally, we have found that some GFP fusion proteins contain multiple optical excitation peaks, as would be expected from the excitation spectra of wild-type GFP. It would be desirable to detect changes in the relative excitation levels of these peaks. It is useful in these cases to test the resulting fusion proteins with a variety of filters, each tuned to a specific peak. Future Improvements
Shaker K+ Channel
One advantage to using GFP is that many mutations have been discovered that alter the fluorescence absorption or emission properties of GFP.” Several groups have used variants of GFP to monitor the spatial location of different proteins within the same cell. The same approach could be used to monitor different signal transduction pathways within the same cell, by creating chimeric sensor proteins with GFP variants. This would require that the GFP variants be distinguishable in the same way, e.g., fluorescence absorption, emission, or lifetime (see filter discussion above). For example, sensors with cyan-fluorescent protein could potentially be distinguished from sensors with yellow-fluorescent protein. Discussion
Flash is not a typical fluorescent voltage probe. Traditional “fast” voltage-sensitive dyes have been designed to respond quickly and linearly to membrane potential. 1,3,25In contrast, Flash provides a different solution to the underlying problem of detecting fast voltage transients: Flash gives long, stereotypical fluorescent pulses in response to brief voltage spikes. The temporally expanded response from Flash provides advantages for detecting individual electrical events, as the area under the response to 25 J. E. Gonzalez
and R. Y. Tsien,
single spikes is approximately 30-fold larger than the area under the input spike (converting units appropriately). In general, it is best to consider ways to visualize fast, discrete cellular events (e.g., action potentials) by exploiting the slower kinetics of a chimeric protein. The sensor protein can often be constructed so that fast physiological events give rise to a slowed fluorescence signature that is matched to standard imaging hardware (e.g., video-rate cameras). The success of Flash suggests that a modular approach could be used to produce optical sensors to detect other signaling events. We are using this approach to engineer signaling proteins so that changes in their biological activity are converted into changes in fluorescence emission. These chimeric sensor proteins have two domains: a “detector” that undergoes a conformational rearrangement during a cell signaling event (e.g., a channel, receptor, enzyme, G protein) and a “reporter” fluorophore (e.g., GFP). Ideally, GFP is fused to the detector protein near a domain that undergoes a conformational rearrangement when the detector protein is activated. Movement in the detector domain alters the environment of GFP or places stress on the structure of GFP, thus altering its spectral properties. These constructs typically include a variant of GFP as a reporter, a signal transduction protein as a detector, and a subcellular targeting peptide. A thorough characterization of the sensor should make it possible to show definitively that the fluorescence reports on detector protein activity, rather than on other physiological signals, such as changes in ion concentration, pH, or oxidative-reductive state. We showed that the fluorescence output depends on the site of GFP insertion, perfectly follows the voltage dependence of channel gating, and is modified in parallel with channel gating by mutations that alter the dynamic range or kinetics of the channel. Protein-based sensor proteins analogous to Flash may enable the noninvasive detection of activity in a variety of proteins, including receptors, G proteins, enzymes, and motor proteins. The developmental timing and cellular specificity of expression can be directed by placing the construct under the transcriptional control of a specific promoter. The combined ability to tune the sensor module via mutagenesis and to target the sensor to specific locations affords powerful advantages for the study of signal transduction events in intact tissues. Acknowledgments We thank Botond Roska for preparing voltage traces from the salamander retina; and Scott Fraser, Henry Lester, Carver Mead, Gilles Laurent, Norman Davidson, Sanjoy Mahajan, John Ngai, and members of the Isacoff laboratory for helpful discussions. Research was supported by the McKnight and Klingenstein, and Whitehall (#J9529) Foundations. M. S. S. is a Howard Hughes predoctoral fellow in the biological sciences.