Mark B Cannon and S James Remington

Abstract

The quantification of transient redox events within subcellular compartments, such as those involved in certain signal transduction pathways, requires specific probes with high spatial and temporal resolution. Redox-sensitive variants of the green fluorescent protein (roGFP) have recently been developed that allow the noninvasive monitoring of intracellular thiol-disulfide equilibria. In this chapter, the biophysical properties of these probes are discussed, including recent efforts to enhance their response times. Several recent applications of roGFPs are highlighted, including roGFP expression within Arabidopsis to monitor redox status during root elongation, expression in neurons to measure oxidative stress during ischemia, and targeting of roGFPs to endosomal compartments demonstrating unexpectedly oxidizing potentials within these compartments. Possible future directions for the optimization of roGFPs or new classes of redox-sensitive fluorescent probes are also discussed.

Keywords: Biosensor, disulfide, GFP, midpoint potential, protein engineering, redox.

1. Introduction

Cysteine residues in proteins are reactive and play many key roles in enzymatic activity as well as in protein folding, for example, by stabilizing folded states through intramolecular disulfide bond formation. More recently, it has come to light that cysteine reactivity, in particular toward reactive oxygen and nitrogen species (ROS and RNS), is also critical for dynamic processes such as gene regulation, oxidative stress response, and cell signaling (1). Despite the increasingly clear significance and apparent complexity of thiol chemistry in living systems, the conventional method

John T. Hancock (ed.), Methods in Molecular Biology, Redox-Mediated Signal Transduction, vol. 476 © 2008 Humana Press, a part of Springer Science + Business Media, Totowa, NJ DOI: 10.1007/978-1-59745-129-1_4

for determining intracellular thiol redox status still requires assays of whole-cell extracts to determine ratios of reduced/oxidized redox buffering components. Such an invasive approach has low accuracy and little or no temporal and spatial resolution. The drawbacks of such methods are especially significant, given that individual subcellular compartments often have very different redox environments (2-4). In addition, transient oxidative signaling bursts, such as the localized production of hydrogen peroxide or other highly reactive species, are known to occur but are necessarily restricted to subcellular microdomains (5, 6). Clearly, the study of these complex systems would benefit greatly from use of noninvasive techniques and tools that allow local monitoring of redox status in real time. This chapter focuses on recent developments that allow intracellular redox measurements with high spatial and temporal resolution using oxidation-reduction sensitive variants of the green fluorescent protein (GFP; also see Chap. 5 by Mullineaux and Lawson, this volume).

2. The Use of GFP and Its Derivatives

2.1. The Green Fluorescent Protein

The Aequorea victoria GFP is a small protein of 238 amino acids and molecular weight of approx 26 kDa. Its structure is a nearly perfect barrel-shape composed of 11 antiparallel P-strands surrounding a rather distorted coaxial P-helix (7, 8). The "P-can"

(8) comprising GFP is capped on both ends by short helices and loops that restrict solvent accessibility to the interior of the structure. Three sequential residues on the interior coaxial helix, Ser65, Tyr66, and Gly67, spontaneously form the green light-emitting 4-(^-hydroxybenzylidene)imidazolidin-5-one chromo-phore (Fig. 4.1). The nucleophilic attack of the amide nitrogen of Gly67 on the carbonyl carbon of Ser65 forms a five-membered imidazolone ring, followed by dehydration of the carbonyl oxygen of Ser65 and the much-slower step of oxidation of the Tyr66 Ca-CP bond to complete the conjugation of the ring systems

(9). This process requires no accessory proteins or external cofac-tors other than molecular oxygen.

GFP is extremely resistant to protease and remains stable under very harsh conditions. In fact, the stability of the protein structure is a requirement for efficient fluorescence, because neither unfolded GFP, nor the naked chromophore, are fluorescent. However, the protein backbone tolerates insertions of entire proteins at several locations, as well as cyclic permutation and/or addition of N- and C-terminal fusion proteins (10). These characteristics make GFP exceptionally suitable as a fusion-tag in both prokaryotic and eukaryotic organisms (11).

Fig. 4.1. Reaction mechanism for the autocatalytic formation of the GFP chromophore from three amino acid residues in the interior of the p-barrel. The mature chromophore is in equilibrium between the neutral and anionic forms, but excitation of the neutral form leads almost immediately to ESPT and the chromophore, which is then anionic, emits light at 508 nm.

Fig. 4.1. Reaction mechanism for the autocatalytic formation of the GFP chromophore from three amino acid residues in the interior of the p-barrel. The mature chromophore is in equilibrium between the neutral and anionic forms, but excitation of the neutral form leads almost immediately to ESPT and the chromophore, which is then anionic, emits light at 508 nm.

Although several GFP homologs have been isolated from marine organisms, with a rainbow of emission colors ranging from cyan to yellow to red, the original GFP isolated from Aequorea victoria remains unique in its complex excitation/emission spectra. Wild-type GFP has two excitation peaks: a major peak at about 400 nm and a minor peak at about 480 nm with approximately one-third the amplitude of the major peak. Excitation at either peak leads to green emission at 508 nm (Fig. 4.2). The presence of two separate excitation peaks suggests that within the protein, equilibrium exists between two distinct chromophore states with similar emission wavelengths.

This observation presents a puzzle, which was partially resolved when it was discovered that the two excitation peaks arise from the protonation state of the chromophore. The ani-onic form of the chromophore is maximally excited at 470-480 nm and emits at about 510 nm. The neutral form of the chromophore, which has a pKa near neutrality in aqueous solution, is efficiently excited by light with wavelength in the range of 370-400 nm and would normally emit blue light at around 450 nm; however, excited state proton transfer (ESPT) converts the neutral,

300 400 500 600 700

Fig. 4.2. Fluorescence excitation (dashed line, emission recorded at 510 nm) and emission (solid line, excitation at 400 nm) spectra for wild-type GFP. Intensities have been scaled to line up excitation and emission peaks. Excitation at either peak (~400 and 480 nm) leads to the characteristic green emission at 510 nm.

300 400 500 600 700

Fig. 4.2. Fluorescence excitation (dashed line, emission recorded at 510 nm) and emission (solid line, excitation at 400 nm) spectra for wild-type GFP. Intensities have been scaled to line up excitation and emission peaks. Excitation at either peak (~400 and 480 nm) leads to the characteristic green emission at 510 nm.

protonated chromophore into an anionic, green emitting species (see Fig.1 in ref. 12). Upon excitation, the pKa of the neutral chromophore is dramatically reduced and proton transfer to an acceptor leads to generation of the excited state anion.

Two groups proposed a structural basis for this dual-excitation behavior (13, 14). Upon excitation, the neutral chromophore, which is the dominant ground-state form in wild-type GFP, undergoes rapid ESPT from the Tyr66 phenol via a hydrogen bonding network to internal Glu222. The anionic form of the chromophore, which can also be directly excited at 480 nm, then emits a 510-nm light.

Presumably, the protein structure determines the equilibrium between the neutral and anionic chromophore populations and thus is subject to modification by external influences. Indeed, in the early 1980s it was determined that the equilibrium between protonation states could be perturbed to some extent by changes in pH, salt concentration, or protein concentration (presumably leading to formation of multimers of GFP; see refs. 15,16). These early results presaged the development of GFP-based indicators of environmental conditions within cells.

The dual excitation behavior of GFP is particularly well suited for the design of active biosensors of various cellular phenomena. Through mutagenesis and protein fusions, novel fluorescent protein-based indicators have been developed that respond to a wide variety of compounds and biological events (see reviews in refs. 17, 18). Here, we review the development and application of redox-sensitive GFP indicators.

2.2. Redox-Sensitive The two widely spaced excitation maxima of wild-type GFP

GFP Indicators depend on the protonation state of the chromophore, which in turn depends on the structure of the protein. In principle, structural alterations that induce a change in the protonation state form an excellent basis for the creation of ratiometric sensors of external conditions. Ratiometric sensors are particularly desirable, because they reduce or eliminate measurement errors due to changes in illumination intensity, cell thickness or indicator concentration.

The technique of ratiometry depends on the presence of two excitation maxima (as in wild-type GFP) or two emission maxima (e.g., red and green), the relative intensities of which are altered in opposite ways by external factors. Assuming two-state behavior, one can factor out the effects of variations in the concentration of the fluorescent indicator and/or the light source intensity by forming a fluorescence intensity ratio. This permits direct quantification of the stimulus (19). On the other hand, an indicator that responds to an environmental stimulus with only an increase or decrease in overall fluorescence may be subject to considerable measurement error because the emission intensity will also depend on variable factors such as illumination intensity, cell thickness, and indicator concentration. In addition, such indicators may be difficult or impossible to calibrate and would give indication only of a relative change in the observed parameter.

Ratiometric redox-sensitive versions of GFP have been developed S20) to take advantage of these principles. The probes, termed redox-sensitive green fluorescent proteins (roGFPs), were constructed by placing pairs of cysteine residues on neighboring strands on the surface of the GFP P-barrel in positions favorable for formation of disulfide linkages. Two sites were selected: positions 149/202, and positions 147/204. Six versions of roGFP were initially developed: with cysteine substitutions at the 147/204 site (roGFPl and roGFP2), at the 149/202 site (roGFP3 and roGFP4), and with cysteine substitutions at both locations (roGFP5 and roGFP6). Half of the roGFPs were developed from wtGFP (roGFP1, roGFP3, and roGFP5) and half from the S65T GFP background (roGFP2, roGFP4, and roGFP6).

The cysteine locations straddle a bulge in the GFP barrel structure around His148, the side chain of which is oriented inside the P-barrel very near the phenolic end of the chromophore (Fig. 4.3). Transition between neutral and anionic forms of the chromophore occurs by protonation/deprotonation of this phenolic hydroxyl group, and therefore formation of a strand-bridging disulfide bond at this location was thought to have a n

Fig. 4.3. Illustration of roGFPI (the "R7" variant) in the oxidized form. The chromophore, His148, and the two engineered cysteine residues (Cys147 and Cys204) are shown in ball-and-stick format to illustrate their positions relative to each other. The positions (149, 202, and 223) used for positively charged substitutions in the rate-enhanced variants of roGFPI are indicated. Figure produced using PyMOL (43) and PDB file 2AH8 (30).

Fig. 4.3. Illustration of roGFPI (the "R7" variant) in the oxidized form. The chromophore, His148, and the two engineered cysteine residues (Cys147 and Cys204) are shown in ball-and-stick format to illustrate their positions relative to each other. The positions (149, 202, and 223) used for positively charged substitutions in the rate-enhanced variants of roGFPI are indicated. Figure produced using PyMOL (43) and PDB file 2AH8 (30).

high probability of affecting fluorescent excitation ratios. A crystal structure of roGFP2 shows that disulfide formation between the pair of engineered surface cysteine residues results in a shift of one P-strand relative to the other (20), which causes subtle internal structural rearrangements, including repositioning of side chains contacting the chromophore (i.e., His148 and Ser205), such that the neutral chromophore is favored over the anionic. Therefore, as a population of roGFP is oxidized, disulfide formation leads to an increase in the excitation peak at 400 nm at the expense of the 480 nm peak (20) (see Fig. 4.4).

RoGFPs were expressed in mammalian cells and were shown to be effective indicators of the ambient cellular redox potential, as perturbed by exogenous oxidants and reductants, as well as by physiological redox changes (21, 22). RoGFPs were expressed in the cytosol as well as in the mitochondrial matrix of HeLa cells. Calibration of the probe was accomplished in situ by measuring the 400/480-nm excitation ratio (with emission measured at 508 nm) in the presence of excess exogenous membrane-permeable oxidants (H2O2) or reductants (dithiothreitol [DTT]).

Fig. 4.4. Plot showing a typical redox titration of roGFP1. The solid line represents fluorescence excitation from the completely oxidized protein (emission measured at 510 nm). Excitation diminishes at the 400 nm peak while increasing at the 480-nm peak as the protein is reduced. Measurement of 480 to 400 nm excitation peak intensities allows determination of fractional oxidation of the probe, and thus ambient midpoint potential.

Fig. 4.4. Plot showing a typical redox titration of roGFP1. The solid line represents fluorescence excitation from the completely oxidized protein (emission measured at 510 nm). Excitation diminishes at the 400 nm peak while increasing at the 480-nm peak as the protein is reduced. Measurement of 480 to 400 nm excitation peak intensities allows determination of fractional oxidation of the probe, and thus ambient midpoint potential.

These peak ratios correspond to the 100% or 0% oxidized forms of the probe, respectively. In vitro titration of the probe with a range of redox buffers (e.g., oxidized/reduced DTT or GSH/ GSSH) and application of the Nernst equation allows alignment of peak ratios, and thus % fractional oxidation of the probe, with redox midpoint potentials (see Fig. 1, and Materials and Methods in ref. 20).

The midpoint potentials of oxidation/reduction reactions involving H+ (such as the oxidation/reduction of roGFP) are intrinsically dependent on the pH of the solution. Midpoints calculated using the Nernst equation must therefore by adjusted according to the solution pH. Reduction of a disulfide to the dithiol form requires the input of two H+:

ox red

At the midpoint, [roGFPox] = [roGFPred] at equilibrium, so the K = [H+]-2. The Nernst equation can therefore be used to calculate theoretical midpoints at any pH:

Here, T is the absolute temperature in Kelvin and the result is expressed in millivolts. It should be noted that the above reaction assumes both cysteines to be protonated in the reduced form. The pKas of the engineered cysteine residues in the roGFPs have not been precisely determined, but are most likely > ~9 (MBC and SJR, 2003, unpublished observations), and so at physiological pH this assumption should be valid.

2.3!. Redox-Sensitive A similar approach was independently employed by Jakob Win-YFP ther and co-workers to produce redox-sensitive variants of the

GFP-derived yellow fluorescent protein (YFP, [see ref. 7]), termed rxYFP (23). The engineered rxYFP cysteines are located at the same position as in roGFP3 and roGFP4 (positions 149/202) and formation of a disulfide bond at this position was shown to affect fluorescence. However, YFP lacks the dual-excitation behavior of GFP and so these probes respond to redox conditions via (non-ratiometric) changes in the amplitude of a single fluorescence excitation peak.

Subsequent studies on rxYFP by Winther et al. (24, 25) have revealed important information concerning the cellular interaction partners of GFP-based redox sensors. Thiol redox reactions in cells and subcellular compartments are enzymatically catalyzed, complex, and far from perfectly understood. However, it has been established that the thioredoxin and glutaredoxin systems work independently to equilibrate cellular thiols with different reducing pools, including glutathione and NADP/NADPH (24). Enzyme specificities and reaction kinetics determine which systems interact with specific thiol/disulfides. In yeast, rxYFP apparently equilibrates with the glutathione pool through the actions of glutaredoxins whereas interactions with thioredoxins seem to be much less important in determining the probe's in vivo redox status (25). roGFPs appear to behave similarly (SJR, 2006, unpublished observations).

One potential application of roGFP is the study of H2O2 bursts in cell signaling events. Recent evidence implicates Ha O2 as an important second messenger in cell signaling since it is produced in response to various extracellular stimuli, such as cytokines and peptide growth factors, and its intracellular production or exogenous application affects the function of a variety of proteins, including protein kinases, protein phosphatases, ion channels and transcription factors (26, 27). Cell defense mechanisms, the forefront being the high concentration of reduced glutathione in many cellular compartments (2, 3, 28), quickly eliminate HaO2 and other potentially damaging reactive oxygen species and therefore such oxidative bursts are believed to be very transient and highly localized. Understanding the complex relationships involved in this aspect of cell signaling will require specific probes with high temporal and spatial resolution. Dooley et al. (22) investigated the possibility of employing roGFP to detect these oxidative bursts,

2.4. Re-Engineering the roGFP for Improved Response Rate but were unable to measure any significant changes in roGFP excitation peak ratios in response to intracellular H2O2 production (22). It is likely that this inadequacy is due to the relatively slow response time of the roGFP probes, which are on the order of tens of minutes.

Formation of a disulfide bond requires deprotonation of the cysteine thiol to the thiolate form. Once this rate-limiting step has occurred, disulfide formation rapidly follows. Stabilization of the cysteine thiolate by nearby positive charges or dipoles is thought to lower the activation energy of the process and increase disulfide formation rates (29). One approach to improving response time of such a system is to substitute positively charged residues near the reactive thiols, thus lowering their pKs and increasing disulfide formation rates. Using this strategy, several variants of roGFPl (designated roGFP1-R1 through roGFP1-R14) have been constructed with up to three positively charged substitutions (lysine or arginine) at locations near the reactive cysteines, positions 149, 202, and 223 (30) (see Fig. 4.3). Substantial rate increases were observed. Mutants with a single basic substitution exhibited an approximate doubling of the pseudo first-order rate constant, as measured by monitoring excitation peak ratios over time after the addition of excess DTT or H2O2 to buffered solutions of the roGFP1 variants. Each additional basic substitution further increased the rate by approximately twofold, with a maximum rate increase of approximately sevenfold in the variant with three basic substitutions.

Application of nonlinear Poisson-Boltzmann theory (as implemented in the program DelPhi; see refs. 31, 32) verified pKa depression of the reactive cysteine thiols. Other groups have undertaken similar approaches to increase cysteine reactivity, with comparable results (22, 33). Although the observed rate increases from this approach are significant, the solvent-exposed location of these substitutions on the outer surface of a P-barrel imposes natural limits on its effectiveness since these engineered chargecharge interactions are medium-range at best (see Fig. 3 of ref. 30). Nevertheless, the roGFP1 variant with the most significant rate enhancement (termed "roGFP1-R12") is recommended for general use because of its faster response time as well as its more oxidizing disulfide midpoint potential.

2.5. Midpoint Another potential limitation on the use of roGFP probes con-

Potentials of Redox cerns the rather negative midpoint potentials. RoGFPs 1-6

Probes exhibit midpoint potentials that range from -272 mV (roGFP2)

to -299 mV (roGFP3) (20). The utility of the probes in determining midpoint potential drops off rapidly the farther the ambient midpoint is from the probe's own midpoint potential. In practice, roGFPs are most useful in measuring midpoints within ~35 mV of their own midpoint (see Fig. 1 in ref. 20). RoGFP1, for example, has a measurement range of —325 to -255 mV, (i.e., when the probe is between ~ 10% and 90% oxidized). This range makes roGFPI (and its siblings) ideal for determination of thiol redox status in highly reducing compartments, such as the cytosol or mitochondria. However, in more oxidizing (i.e., more positive midpoint potential) compartments such as the endoplas-mic reticulum (ER), the probes are expected to be completely oxidized and thus capable of indicating only this fact.

Re-engineered roGFPs such as roGFP-R12 are somewhat more oxidizing than the first-generation probes. The introduced basic residues act to lower the cysteine thiol pKas, thus stabilizing the thiolate anion, and this, in general, destabilizes the disulfide thermodynamically (34, 35). RoGFP1-R variants with more basic substitutions had slightly more oxidizing midpoints, with roGFP1-R12 and roGFP1-14 (each with three basic mutations near the reactive cysteines) having the most oxidizing midpoint potentials, -265 and -263 mV, respectively (30). However, the cysteine pKas remain high, greater than 9, and it would be desirable to lower these values substantially (see below).

Was this article helpful?

0 0

Post a comment