Fluorescent Protein-Based Redox Probes

1. rxYFP

The introduction of the T203Y mutation converted EGFP to YFP, a red-shifted GFP variant with an excitation maximum at 513 nm and emission at 527 nm (124). Based on the assumption that dislocation of Y203 and accompanying changes in π-electron stacking would affect fluorescence significantly, Ostergaard and colleagues used YFP as a template for redox-sensitive variants (rxYFP) (92). Dislocation of Y203 was expected to occur when the β-strand 10 (aa 199–208) of the β-barrel of the YFP protein was forced into a slightly different position. To achieve such conformational change, pairs of cysteines were introduced in different positions on β-strands 7 and 10 such that the two residues are capable of forming an intramolecular disulfide bridge. Of four different combinations of cysteine pairs only the variant N149C/S202C gave a strong redox-dependent change in fluorescence (92). Formation of the disulfide bond between these two cysteines resulted in a 2.2-fold reduction of the emission peak without a significant shift in the excitation wavelength. YFP exhibits two absorption peaks for the neutral A-band (λ_max = 392 nm) and the anionic B-band (λ_max = 514 nm), with a clean isosbestic point at 432 nm (125). However, due to fluorescence quenching, the neutral form of YFP is nonfluorescent and thus recombinant rxYFP exhibits only a single excitation peak with a maximum at 512 nm similar to wtYFP (92). The midpoint redox potential E°′ of rxYFP was determined as −261 mV through redox titration of recombinant protein with 2GSH/GSSG buffers (92). Since the reactivity of rxYFP with glutathione is rather low, protein samples had to equilibrate for 21 hours and thus the adjusted 2GSH/GSSG mixture had to be carefully controlled for air oxidation of GSH.

Although fluorescence quenching of the neutral form of the chromophore in rxYFP would argue for exclusive usability of rxYFP as a single excitation wavelength probe only, Maulucci and colleagues report the use of rxYFP as a ratiometric probe (76). When expressed in human cells, the redox-dependent change in emission after excitation at 458 nm was only ∼50% of the change observed after excitation at 488 nm. This difference was used for ratiometric calculations (60, 76, 77). However, it is not entirely clear if this kind of ratiometry can separate variations in probe concentration from changes in redox state. The exact excitation spectra for reduced and oxidized rxYFP inside cells are not known. Because the reference wavelength (458 nm) is close to the isosbestic point of YFP, the fluorescence excited is very low and thus may be significantly obscured by background noise, which needs to be subtracted very carefully. This already implies that such analysis of rxYFP might be very sensitive to autofluorescence. Furthermore, the fact that the redox-dependent change in fluorescence points in the same direction for both excitation wavelengths inevitably lowers the dynamic range of the probe. Another issue that may be relevant for the interpretation of measurements is the sensitivity of YFP to changes in pH at near neutral conditions with a chromophore pK _a of 7.0 (30) and the significant sensitivity of YFP to chloride (62,125).

2. roGFPs

The limitations due to lack of ratiometric properties were overcome with the development of redox-sensitive GFPs (roGFPs) (49). Starting from either wtGFP or EGFP, respectively, two cysteines were engineered into the positions S147 and Q204, which are located on β-strands 7 and 10 in close proximity to the positions mutated in rxYFP. The resulting redox probes were called roGFP1, which is based on wtGFP, and roGFP2, which is derived from EGFP. The two excitation maxima characteristic of GFP (∼400 nm for the A-band and 475–490 nm for the B-band) are maintained in the respective roGFP variants with band A being the main excitation peak in roGFP1 and band B dominating in roGFP2. Oxidation results in an increase in the A-band and a decrease in the B-band and an inverse behavior under reducing conditions, as shown exemplarily for roGFP2 in Figure 2. For both probes a clean isosbestic point at ∼425 nm was identified. With these spectral characteristics both roGFPs are bona fide ratiometric sensors.

OPEN IN VIEWER

FIG. 2.

Redox-dependent changes in the excitation spectrum of roGFP2. Because of the S65T mutation, the chromophore of reduced roGFP2 is preferentially deprotonated, thus exhibiting a predominant excitation maximum around 488 nm (red curve). Upon oxidation, the chromophore of roGFP2 is preferentially protonated, thus gaining excitability at 405 nm and loosing excitability at 488 nm (blue curve).

RoGFP2 has been crystallized in both the reduced and oxidized states (PDB entries 1JC0 and 1JC1) (49). A comparison between the two structures highlights structural changes that likely explain the disulfide-induced shift in chromophore protonation. Formation of the disulfide bridge shifts one β-strand relative to the other and this causes several small structural rearrangements. Significant movements involve H148 (neighbor of C147 on strand 7) and S205 (neighbor of C204 on strand 10) (Fig. 3). H148 is located in H-bonding distance to the phenolic oxygen of Y66 and moves slightly away from the chromophore upon disulfide formation. The OH group of S205 (part of the ESPT hydrogen bonding network) moves relative to the water molecule and the E222 carboxylate. It appears likely that these changes influence the Y66 protonation state and thus explain the ratiometric shift in the excitation spectrum.

OPEN IN VIEWER

FIG. 3.

Disulfide bond formation influences roGFP2 chromophore protonation. Upon formation of the C147–C204 disulfide bond, neighboring residues H148 and S205 are slightly shifted relative to the chromophore. H148 is in H-bonding distance to the phenolic oxygen of the chromophore. S205 is part of the hydrogen bonding network involved in proton transfer between the chromophore and E222. The figure is based on PDB entries 1JC0 (reduced roGFP2) and 1JC1 (oxidized roGFP2). Important atoms are shown as Van-der-Waals spheres: Sγ(C147), Sγ(C204), Nδ1(H148), O(Y66), Oγ(S205), Oɛ1(E222), and O(H₂O).

Midpoint redox potentials for roGFPs have been determined through titration with DTT, lipoic acid, or bis(2-mercaptoethyl)sulfone (BMES) (27, 49). Generally, the determined midpoint potentials of −288 to −294 mV for roGFP1 and −272 to −287 mV for roGFP2 are more negative than the midpoint potential of rxYFP. Although the use of different redox buffers for titration led to slightly different apparent midpoint potentials, the ranking of roGFP1 as more reducing than roGFP2 was found to be preserved (27). For future applications, Dooley and colleagues suggested consensus midpoint potentials of −291 mV for roGFP1 and −280 mV for roGFP2 (27). A more precise determination of E°′ for redox-sensitive FPs requires exact knowledge of the midpoint potential of the calibrating redox couple and avoidance of impurities in reduced and oxidized compounds.

In addition to the difference in midpoint potential, roGFP1 and roGFP2 also show some differences in their maximum achievable dynamic range for the change in fluorescence between fully reduced and oxidized probes (Table 1). If roGFPs are used as a single wavelength probe, they exhibit approximately a 3-fold change in fluorescence, which is only slightly better than the 2.2-fold change for rxYFP. However, because roGFPs are ratiometric, their dynamic range is significantly increased. With fully flexible filter-based detection systems such as epifluorescence microscopes or fluorescence plate readers, the probes can be excited exactly on the two excitation peaks, which would provide the largest achievable dynamic range. With excitation at 390 ± 5 nm and 480 ± 5 nm, roGFP2 has a dynamic range of about 12 (81). On appliances that use fixed wavelength lasers for fluorescence excitation such as confocal microscopes or fluorescence-activated cell sorting (FACS), the dynamic range is significantly lower due to off-peak excitation. The canonical laser wavelengths used for excitation of GFP are 405 and 488 nm. While the S65T mutation in roGFP2 shifts the excitation peak B for the anionic form of the chromophore to 490 nm, the anionic form of roGFP1 is excited with a maximum at 475 nm. When excited at 488 nm the dynamic range of roGFP1 is thus below 3, while roGFP2 would provide a theoretical dynamic range of ∼9 (107).

3. Variants of rxYFP and roGFPs with altered properties

To match different needs and applications, it would be desirable to obtain a collection of probes with different midpoint potentials. For instance, to detect deflections from the steady state in both directions, probes with a midpoint potential matching the steady state redox potential of the environment under study are most appropriate. Along these lines, in addition to roGFP1 and roGFP2, four other variants, roGFP3 to roGFP6, have been engineered (Fig. 4; Table 1) (49).

OPEN IN VIEWER

FIG. 4.

Genealogy of genetically engineered fluorescent proteins with redox-sensitive disulfides. rxYFP^3R refers to the mutant rxYFP^{200R/204R/227R}. For further details and references, see text.

RoGFP3 and roGFP4 differ from their siblings roGFP1 and roGFP2 in that they contain the disulfide bridge in position C149/C204. RoGFP5 and roGFP6 contain all four cysteines and thus can potentially form two internal disulfide bridges. However, none of these variants have been investigated in much detail so far and the presence of two disulfide bridges may render quantitative interpretation of the measured fluorescence ratio rather difficult. The variants roGFP3 and roGFP4, however, may have some potential for future applications. Although the observed dynamic range of theses variants is lower than that observed for roGFP1 and roGFP2, the midpoint potentials are slightly more negative. This is particularly true for roGFP3, which is derived from wtGFP, with the most negative midpoint potential of all roGFPs at −299 mV (49). Thus, in highly reducing compartments roGFP3 may have some potential for achieving more accurate measurements compared to other roGFPs.

The first generation of redox-sensitive FPs (rxYFP, roGFPs1-6) all had midpoint redox potentials of −260 mV or below, which made them most useful in reducing compartments, such as the cytosol, the mitochondrial matrix and chloroplast stroma (Table 1) (49, 92).

In the lumen of the endoplasmic reticulum (ER), rxYFP, roGFP1, and roGFP2 were fully oxidized, suggesting a redox potential higher than −240 mV (2, 49, 93). Thus, in oxidizing compartments the use of these probes is limited by their rather low midpoint potential. Recently, Lohman and Remington developed a new subfamily of roGFP variants derived from roGFP1, called roGFP1-iX, where ‘X’ denominates the insertion of different single amino acids into β-strand 7, adjacent to C147 (Fig. 4) (67). In conjunction with the exchange H148S the insertion resulted in an increased geometric strain in the disulfide, which renders the midpoint potential for the disulfide bond less negative. The least negative values were achieved with the insertion of leucine (−229 mV) and glutamate (−236 mV) (Table 1). Despite the very significant shift in E°′, it is not yet clear whether an E°′ of −230 mV is already sufficient for dynamic measurements in the ER.

To be reactive, thiols need to be de-protonated. The pK _a values of the two cysteines in rxYFP and roGFPs are all located between 8.9 and 9.5 (47, 49), which results in mostly protonated thiols and hence low reactivity. Introduction of basic amino acids next to cysteines can lead to stabilization of the thiolate form and thus was expected to increase the reactivity of the thiols in a series of 2^nd generation redox-sensitive FPs (Fig. 4). Insertion of two lysines close to the two cysteines plus an arginine in position 223 led to a 6-fold increased rate constant in roGFP1-R12 (13). The insertion of up to three positive charges near the disulfide in rxYFP (rxYFP^{200R/204R/227R}) led to a 13-fold rate enhancement for the oxidation of rxYFP by GSSG (47). However, the achieved rate enhancements appear negligible for reactions in vivo, where the kinetic properties of redox-sensitive FPs are mainly determined through interaction with glutaredoxins (Grxs) (see Section III.A).

4. Interaction of rxYFP and roGFP with cellular redox pairs

When expressed in HeLa cells, the effective concentration for H₂O₂-induced oxidation of roGFPs was much lower than the concentration required in vitro (27). This observation suggested that oxidation of roGFPs inside cells is a catalyzed process. A link between roGFP oxidation and the cellular glutathione pool was indicated by a slight oxidation of roGFP in HeLa cells after a 2-hour incubation with 100 μM L-buthionine (S,R)-sulfoximine (BSO), a highly specific inhibitor of GSH biosynthesis. Furthermore, BSO-treatment also led to a highly increased sensitivity of roGFP to exogenously applied H₂O₂ (27).

At the same time, Jakob Winther and co-workers reported that rxYFP in yeast cells senses the glutathione redox potential (E_GSH ) through interaction with Grxs (93). Thioredoxins, in contrast, were not capable of reducing or oxidizing rxYFP. A sensitive response of roGFP2 to depletion of cytosolic glutathione was shown in neuronal cells (121) and detailed studies in Arabidopsis thaliana confirmed that roGFP2, like rxYFP, is a quantitative biosensor for E_GSH (81). Expression of roGFP2 in the partially GSH-deficient Arabidopsis mutant cad2, which is restricted in the activity of the first and rate-limiting GSH biosynthetic enzyme glutamate-cysteine ligase (GSH1) (21), resulted in an increased 405/488 nm fluorescence ratio reflecting a shift in E_GSH towards less reducing values (Figs. 5B and 5C) (81). The detected shift in E_GSH closely matched expectations based on only 30% remaining glutathione and the assumption of constant glutathione reductase (GR) activity. In a more severe GSH1 mutant, rml1, which contains less than 5% of wild-type GSH (12,120), roGFP2 was almost completely oxidized (Fig. 5D). Taken together, studies on rxYFP and roGFP strongly support the notion that within cells these proteins communicate with the glutathione redox couple through mediation by endogenous Grxs.

OPEN IN VIEWER

FIG. 5.

Expression and ratiometric analysis of roGFP2 in Arabidopsis . (A) Biosynthetic pathway for GSH consisting of two enzymatic steps catalyzed by glutamate-cysteine ligase (GSH1) and glutathione synthase (GSH2). cad2 and rml1 are two partially GSH-deficient GSH1 mutants. (B–D) Ratiometric analysis of roGFP2 in root tips of wild-type (Col-0), cad2, and rml1 mutants indicates increasing oxidation of roGFP2 with deceasing GSH content. Propidiumiodide (PI) in intact cells exclusively labels cell walls and was used as a viability indicator. Scale bars = 50 μm.

Several other redox-active compounds like NADPH and ascorbate, as well as oxidoreductases, such as Trxs and protein disulfide isomerases (PDIs), have been tested for interaction with roGFP in vitro, but none of these significantly affected the redox status of roGFP (40, 81) (Meyer et al., unpublished results). However, given the broad spectrum of oxidoreductases in the Trx superfamily and other thiol-reactive proteins in cells, it cannot be excluded that some of these proteins may be capable of interacting with roGFP. As long as the Grx-catalyzed equilibration of roGFP with 2GSH/GSSG is kinetically favored, all such theoretically possible side reactions of roGFP should be overruled and high specificity of roGFP for E_GSH guaranteed. Translational fusion of an appropriate Grx to roGFP is a potent strategy to ensure constant kinetic coupling and fully reproducible probe specificity (see Section III.A).

5. The Nernst equation and its implications for using roGFPs

Based on the very well supported notion that roGFPs predominantly, if not exclusively, equilibrate with the 2GSH/GSSG redox couple, the redox behavior of roGFPs can be described and predicted by the Nernst equation, with various implications for the practical use of the probe. How fast the equilibrium is reached, or if it is reached at all, depends on the availability of Grx, which may differ between experimental systems. As discussed later on, the translational fusion of Grx to roGFP is a suitable strategy to ensure real-time equilibration (see Section III). The following calculations exemplarily apply to roGFP2.

b. Measuring range and calibration of roGFP2

The relationship between OxD_roGFP2 and the redox potential E can be derived from the right half of the preceding equation: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6} \begin{document} \begin{align*} OxD_{roGFP2} = \left( e^{\frac {2F \left( - E + {E^{ \circ \prime}_{roGFP2}}\right)}{RT}} + 1 \right)^{ -1} \end{align*} \end{document}

This relationship reveals that the response of the probe becomes increasingly nonlinear when approaching 5% and 95% oxidation, respectively, and thus renders the detected signal sensitive to background noise (Fig. 6). For practical purposes, changes below 5% and above 95% sensor oxidation cannot be sufficiently resolved in fluorescence measurements. At best, roGFP2 can be expected to measure changes in E_GSH in the 80 mV-wide window between −320 and −240 mV (Fig. 6).

OPEN IN VIEWER

FIG. 6.

Relationship between the redox potential ( E ) and the degree of sensor oxidation ( OxD_roGFP2 ). Changes in probe fluorescence can be resolved in the window between 5% and 95% probe oxidation, which translates to an effective measuring range from − 320 to − 240 mV.

Suitably, the cytosolic E_GSH of unstressed cells is typically found in the range between −320 to −300 mV, close to the lower end of the practical measuring range. The measuring range of roGFP2 is well suited for measuring E_GSH in the cytosol and several other compartments, including the nucleus, mitochondria, and plastids. It is, however, not suitable for the ER, where it is oxidized at all times (81).

Given the available measuring range, is it feasible to derive absolute E_GSH values from fluorescence data? Of course, it would be highly desirable to calibrate intracellular measurements of E_GSH against independently confirmed standards. This, however, would require an independent compartment-specific approach to determine intracellular E_GSH . Unfortunately, no such alternative method is available. Since absolute intracellular calibration is not feasible, a more indirect way of calibration has to be performed. In a practical approach, to estimate intracellular E_GSH from the measured fluorescence ratio (R), externally applied cell-permeable reagents (typically dithiothreitol and diamide or H₂O₂) are used to deflect the probe in situ to the fully reduced and oxidized states. The resulting fluorescence ratios (R_red and R_ox ) define the total range of the sensor inside intact cultured cells or transgenic tissues. Thus, the full set of measurements delivers six intensity values from which three ratios can be calculated (Table 2).

OPEN IN VIEWER

Table 2.

Fluorescence Intensity Measurements Required for the Calculation of Redox Potentials

	Complete reduction (DTT)	Complete oxidation (Diamide or H2O2)	Actual measurement
Intensity at 405 nm	I405 red	I405 ox	I405
Intensity at 488 nm	I488 red	I488 ox	I488
Corresponding ratio	\documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document}\begin{align*}\displaystyle R_ { red } = \frac { I 405_ { red } } { I 488_ { red } }\end{align*}\end{document}	\documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document}\begin{align*}\displaystyle R_ { ox } = \frac { I 405_ { ox } } { I 488_ { ox } }\end{align*}\end{document}	\documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document}\begin{align*}\displaystyle R = \frac { I 405 } { I 488 }\end{align*}\end{document}

To establish a relationship between the macroscopic observables and OxD_roGFP2 , several molecular quantities need to be defined: i405 _red , i405 _ox , i488 _red , and i488 _ox are the fluorescence intensities contributed by a single roGFP2 molecule at the indicated wavelength and redox state, N_total is the total number of roGFP2 molecules, N_red is the number of reduced roGFP2 molecules, and N_ox is the number of oxidized roGFP2 molecules. Based on the assumption that the overall fluorescence intensity corresponds to the summation of fluorescence intensities from individual roGFP2 molecules, the relationships given in Table 3 apply.

OPEN IN VIEWER

Table 3.

Relationships between Fluorescence Intensities and Probe Molecules

Limiting condition 1:	Ntotal = Nred	I405 red = Ntot · i405 red
Complete reduction	Nox = 0	I488 red = Ntot · i488 red
Limiting condition 2:	Ntotal = Nox	I405 ox = Ntot · i405 ox
Complete oxidation	Nred = 0	I488 ox = Ntot · i488 ox
Actual measurement	Ntotal = Nred + Nox	I405 = Nox · i405 ox + Nred · i405 red
		I488 = Nox · i488 ox + Nred · i488 red

By inserting the resulting equations (right column of Table 3) into OxD_roGFP2 = N_ox/N_total, the relationship between OxD_roGFP2 and the macroscopic observables can be deduced. As a result, OxD_roGFP2 can be calculated directly from the six measured intensity values: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6} \begin{document} \begin{align*} OxD_ { roGFP2 } \\&= \frac { I 405 \cdot I 488_ { red } - I 405_ { red } \cdot I 488 } { I 405 \cdot I 488_ { red } - I 405 \cdot I 488_ { ox } + I 405_ { ox } \cdot I 488 - I 405_ { red } \cdot I 488 } \end{align*} \end{document}

Or, alternatively, after transforming the same equation, using ratios: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6} \begin{document} \begin{align*} OxD_ { roGFP2 } = \frac { R - R_{ red }} {\frac { I 488_ { ox } } {I 488_ {red}} ( R_ { ox } - R ) + ( R - R_ { red } ) } \end{align*} \end{document}

In this equation, the quotient (I488 _ox /I488 _red ) is most appropriately considered the ‘instrument factor’. As the intensity at both wavelengths is linearly dependent on OxD_roGFP2 , the ratio R is not (Fig. 7). Only in the special case that the isosbestic point (425 nm in the case of roGFP2) is used as the reference wavelength for ratiometry (instead of 405 nm) the instrument factor is canceling out from the equation, which then describes a linear relationship.

OPEN IN VIEWER

FIG. 7.

The nonlinear relationship between OxD_roGFP2 and the observed fluorescence intensity ratio R . R is the experimentally derived ratio of fluorescence intensities (I405/I488). In this graph the fully reduced probe corresponds to R = 0.1 and the fully oxidized probe to R = 1.

Having obtained OxD_roGFP2 , the corresponding intracellular sensor redox potential E_roGFP2 can be calculated from the Nernst equation: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6} \begin{document} \begin{align*} E_ { roGFP2 } = E_ { roGFP2 } ^ { \circ \prime } - \frac { RT } { 2F } \ln \left( \frac { 1 - OxD_ { roGFP2 } } { OxD_ { roGFP2 } } \right) \end{align*} \end{document}

Finally, assuming complete equilibration, E_GSH should equal E_roGFP2 . Of course, the accuracy of the estimate is limited by measurement errors and differences in pH and temperature relative to the standard conditions. For a more detailed discussion of pH influence, see Section II.B.5.d below.

c. Sensitivity of roGFP2 towards changes in OxD_GSH and [GSSG]

As the midpoint redox potential of roGFP2 (−280 mV) is much more negative than that of glutathione at physiological concentrations (ranging from −151 mV at 1 mM to −181 mV at 10 mM), even a small increase in the degree of glutathione oxidation will give rise to a large and disproportionate degree of sensor oxidation (Fig. 8). However, since E_GSH depends on GSH_total, so does the sensitivity of the probe towards changes in OxD_GSH and [GSSG].

OPEN IN VIEWER

FIG. 8.

Nernst relationship between the roGFP2 and glutathione redox couples. Comparison between roGFP2 (blue curve), glutathione at GSH_total = 10 mM (left red curve), and at GSH_total = 1 mM (right red curve). Within the measuring range of roGFP2 (highlighted box), small increases in the oxidation of glutathione lead to large increases in sensor oxidation. The difference in midpoint potentials \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document}\begin{align*}(E_{roGFP2}^{ \circ \prime} -\ E_{GSH}^{ \circ \prime})\end{align*}\end{document} is inversely correlated to GSH_total.

It follows that roGFP2 should be a more sensitive indicator of OxD_GSH and [GSSG] in cells with low GSH_total as compared to cells with high GSH_total. Comparing overall cytosolic glutathione concentrations of 1 and 10 mM, broadly reflecting the variation between mammalian cell types, the difference in sensor oxidizablity is quite significant: At GSH_total = 1 mM 50% sensor oxidation is achieved with a 10-fold lower OxD_GSH (0.000088 vs. 0.00088) (Fig. 9A) and a 100-fold lower [GSSG] (44 nM vs. 4.4 μM) (Fig. 9B).

OPEN IN VIEWER

FIG. 9.

Sensitivity of the roGFP2 response towards changes in OxD_GSH and [GSSG]. Comparison between 1 mM and 10 mM total glutathione concentration. (A) Relationship between OxD_roGFP2 and OxD_GSH . (B) Relationship between OxD_roGFP2 and [GSSG]. The measuring range of roGFP2 is highlighted for the curves representing GSH_total = 1 mM.

To quantify the ‘transmission’ between glutathione oxidation and sensor oxidation—as afforded by the difference in midpoint potential between the two redox pairs—the OxD_roGFP2 /OxD_GSH ratio at the sensor midpoint potential (−280 mV) can be considered. The ‘transmission ratio’ increases steeply with decreasing GSH_total and is 10-fold higher at 1 mM than at 10 mM (Fig. 10).

OPEN IN VIEWER

FIG. 10.

The ‘transmission ratio’ OxD_roGFP2/OxD_GSH depends on the total concentration of glutathione. The ratio OxD_roGFP2/OxD_GSH is 560 at GSH_total = 10 mM and reaches 5700 at GSH_total = 1 mM.

Although the ‘transmission ratio’ becomes very high with decreasing GSH, at GSH_total = 10 mM the ‘transmission’ is just sufficient to reliably observe changes in the lowest part of the physiological range between −320 and −300 mV. Specifically, at GSH_total = 10 mM, the shift of E_GSH from −320 to −300 mV corresponds to an absolute increase of [GSSG] from 200 to 930 nM, and a shift in OxD_GSH from 0.00004 to 0.00018. As a consequence, the oxidation of roGFP2 increases from 4% to 17%. If, however, a redox-sensitive FP with a midpoint potential of −260 mV had been used instead of roGFP2, it could not reliably ‘perceive’ this change at GSH_total = 10 mM, as its oxidation would just increase from 1 to 4%.

d. The influence of pH

How does pH influence the roGFP2 response? In principle, there are two ways by which pH can influence probes based on redox-sensitive FPs. Firstly, by influencing the fluorophore itself, and, secondly, by influencing the redox-sensing dithiol/disulfide pair. Direct influence on the fluorophore has been a concern, because the S65T mutation is expected to render roGFPs pH-sensitive. In fact, roGFP2 loses fluorescence at low pH, like other GFP variants with the S65T mutation (30). Fortunately, it turned out that pH-dependent fluorescence quenching affects both excitation wavelengths equally and that the calculation of excitation ratios effectively cancels out any pH effect in the range between 5.5 and 8.0 (107). The pH-insensitivity of the ratio was also confirmed for the fusion protein Grx1-roGFP2 (40), to be discussed later (see Section III.A.2).

What about the general influence of pH on thiol–disulfide equilibria? The reactivity of thiols and thus the redox potential is directly linked to the ambient pH because two protons are liberated during the formation of a disulfide bridge. As long as the pH does not approach the pK _a of the thiols forming the disulfide bond, the redox potential will change by approximately 59 mV per pH unit (103, 127). Consequently, a change in pH leads to a simultaneous change in E_GSH and E_roGFP2 , into the same direction and by a very similar amount. Given a pK _a of 8.92 for glutathione (103) and assuming a pK _a of 9.0 for the roGFP2 cysteines (14), it can be expected that the difference in redox potential between the two redox pairs is not significantly influenced by pH changes in the physiological range (Fig. 11A). The predicted deviation from co-linearity (ΔΔE) amounts to just 2 mV at pH 8, but becomes more significant at higher pH (Fig. 11B).

OPEN IN VIEWER

FIG. 11.

Collinear pH-dependency of E_roGFP2 and E_GSH . (A) The expected relationship between E and pH demonstrates collinearity for GSH (calculated for pK _a = 8.92 at GSH_total = 5 mM, blue curve) and roGFP2 (for pK _a = 9.0, red curve). (B) Deviation from collinearity (ΔΔE) becomes significant when the pK _a is approached, especially above pH 8.

This inference is supported by the in vitro observation that recombinant roGFP2 or Grx1-roGFP2 show the same response to GSH solutions adjusted to different pH in the range between 5.5 and 8.0 (Meyer et al., unpublished data; Gutscher et al., unpublished data). Thus, roGFP2 does not respond to pH-induced changes in E_GSH and should be regarded an indicator for pH-independent changes of E_GSH . It should be noted that pH nevertheless influences the kinetics of roGFP2-glutathione equilibration, which is faster at higher pH, attesting to the deprotonation of thiol groups involved in the exchange (Meyer et al., unpublished data; Gutscher et al., unpublished data). In conclusion, expressing roGFP2 in a compartment with different pH (e.g., mitochondria) should not influence its sensitivity to oxidation by GSSG (compare Section II.B.5.c). However, if absolute mV values are to be estimated from fluorescence data, a correction for the assumed pH difference is required.

6. Applications of rxYFP and roGFPs

The applicability of conventional redox-sensitive FPs is limited by the kinetics of equilibration with the glutathione system (which may differ between species and cell types). As complete intracellular probe equilibration can take tens of minutes, conventional redox-sensitive FPs are best suited to determine steady state redox conditions and to document redox changes that are sufficiently long-lived to allow the probe to equilibrate.

c. roGFPs as topology reporters for membrane proteins

Current estimates indicate that 20%–30% of all predicted open reading frames in a typical genome code for transmembrane (TM) proteins (122), but the number, location and orientation of transmembrane domains (TMDs), as predicted by bioinformatics algorithms are often highly contradictory (29). Despite the general biological importance of TM proteins, tools for experimental analysis of the protein topology are limited. In yeast, the topology of TM proteins can be investigated in vivo using fusions with essential metabolic enzymes as topology reporters and expression of these fusions in a mutant background (58). This approach, however, is not feasible for most other eukaryotic cell types because suitable mutants are often lacking. An advanced method to analyze the topology of TM proteins in vivo is the redox-based topology analysis (ReTA) in which roGFP2 is exploited as a topology reporter (8) (Fig. 12). The only requirement for ReTA to be applicable is a distinct gradient in E_GSH across the membrane the TM protein of interest is targeted to. Such a gradient exists across the ER membrane and thus all TM proteins passing through the ER are prime candidates for ReTA. The straightforward ratiometric analysis without the need of any calibration provides a binary readout for the orientation of the protein within the membrane (8). For further analysis of the entire protein topology, roGFP can be fused to truncated versions of the protein or the probe can be inserted into (putative) extra-membrane loops. The latter approach has been shown to be feasible by inserting conventional GFP into PIN5, an atypical member of the PIN auxin-efflux-carrier family in plants (88). Oxidized roGFP2 has a pK _a of 6 (49) and thus could be quenched to a large extent in acidic compartments such as lysosomes, plant vacuoles, and the apoplast in plants. Loss of fluorescence on one side of the membrane would impede binary ratiometric readouts. Since most proteins in these membranes pass through the ER and their topology is assumed not to change during passage through the secretory pathway, it may still be possible to analyze their topology by ReTA within the ER (8). Another possibility to overcome fluorescence quenching in acidic compartments may be to use the pH-insensitive roGFP1, which is assumed to have a much lower pK _a in the order of the apparent pK _a near 4.5 determined for wtGFP (117).

OPEN IN VIEWER

FIG. 12.

The redox-based topology assay (ReTA) enables direct visualization of membrane protein topology. Ratiometric imaging of roGFP fused to membrane proteins either at their N- or C-termini provides a binary readout for the orientation of the roGFP tag relative to the membrane.

ReTA can be expanded to the analysis of protein orientation in nonsecretory membranes provided a sufficient E_GSH gradient is present. For mitochondria it has been shown that rxYFP is partially oxidized in the intermembrane space (IMS), whereas the probe was considerably more reduced in the cytosol and the mitochondrial matrix (56). Based on this finding, ReTA should also be applicable to mitochondrial membrane proteins. Applicability of ReTA to TM proteins of the chloroplast envelope in plants would also depend on the ability to measure E_GSH in the IMS of the chloroplast envelope. So far, however, no successful expression of any fluorescent protein has been reported for the plastidic IMS (e.g., (5)). Nevertheless, it can be envisaged that the topology of TM proteins in the inner membrane can be investigated in a mutant background specifically affected in plastidic GSH levels and thus plastidic E_GSH . With these amendments ReTA is considered to be broadly applicable and the most simple-to-use method currently available for the analysis of TM protein topology in higher eukaryotic cells (8).

1. rxYFP-Grx1p

Following the observation that rxYFP in yeast cells equilibrates with glutathione (93), the first investigation into the direction of fusion-based redox relays was done by the group of Jakob Winther. The fusion between rxYFP and yeast Grx1p (rxYFP-Grx1p) was primarily used as a biochemical tool to study the reaction mechanism of Grx. The fused Grx was found to operate through the monothiol mechanism. RxYFP oxidation was improved by a factor of 3300 and proceeded with almost absolute glutathione specificity (7). The crystal structure of rxYFP-Grx1p did not reveal direct contacts between the two protein domains, thus suggesting that it is the enhanced collision rate between the two domains that is responsible for the much improved thiol–disulfide exchange efficiency (44).

2. Grx1-roGFP2

As described in Section II.B.4, it had become clear that roGFPs, like rxYFP, interact with the glutathione system (81). This led to the question if Grx-based chimeric fusion proteins are suitable as redox probes for live imaging of redox changes in mammalian cells. To create a probe for E_GSH , human Grx1 and roGFP2 were chosen as fusion partners (40). RoGFP2 was considered to be the most suitable redox-sensitive FP for the purpose of developing a biosensor. As discussed before, a general advantage of roGFPs over rxYFP is that they are ratiometric by excitation, thus making measurements insensitive to differences in sensor concentration and photobleaching (49). In addition, the midpoint potential of roGFP2 (−280 mV) is more negative than that of rxYFP (−261 mV), thus making the roGFP2 response more sensitive to oxidation by small increases in glutathione oxidation (compare Section II.B.5.c). roGFP2 was preferred over roGFP1 (49), mainly for three reasons: (i) roGFP2 is brighter, (ii) its two excitation maxima better match standard laser lines (405 and 488 nm), and (iii) the S65T mutation makes roGFP2 resistant to photoswitching, which may create artifacts by shifting fluorescence ratios (107). Due to their limited dynamic range, roGFP3 and roGFP4 (49) were not considered for the purpose of biosensor development.

a. The general concept behind Grx1-roGFP2

The intracellular response of nonfused roGFP2 is limited by the availability of endogenous Grx, which appears to be insufficient for rapid equilibration in many systems and situations. This also means that differences in Grx abundance and activity between species, cell types, subcellular compartments and developmental or environmental situations will influence the kinetics of equilibration between roGFP2 and glutathione, making observations much less comparable and reliable. To solve the problem of slow and variable equilibration, roGFP2-fused Grx is exploited to enforce continuous equilibration between the two redox pairs roGFP2_red/roGFP2_ox and 2GSH/GSSG. By covalently attaching the redox catalyst to roGFP2, its local concentration is dramatically increased. For approximation, assuming that Grx1 is restricted by the linker to an operating range of ∼75 Å, the local Grx1 concentration experienced by each roGFP2 molecule is ∼1000 times higher (∼1 mM) than the estimated absolute concentration of Grx1-roGFP2 in the cell (∼1 μM). The local Grx1 concentration experienced by roGFP2 is thus also at least three orders of magnitude higher than the concentration of endogenous Grx1 distributed in the mammalian cytosol (37). By making Grx an integral part of the probe, the biosensor becomes catalytically self-sufficient in securing rapid and efficient equilibration under all circumstances (Fig. 13).

OPEN IN VIEWER

FIG. 13.

Principle of the Grx1-roGFP2 redox biosensor. (A) RoGFP2-linked Grx efficiently catalyzes equilibration between glutathione and roGFP2. (B) Structural model showing glutathione disulfide and the Grx1–roGFP2 fusion protein in the reduced state. The redox state of roGFP2 is obtained by ratiometric fluorescence intensity (I) measurements at two different excitation wavelengths (λ₁, λ₂).

b. The molecular mechanism of Grx1-roGFP2

The response properties of Grx1-roGFP2 are based on the well established monothiol mechanism of Grxs (32, 98). This conclusion is supported by the finding that removal of the second active site thiol (C26) does not compromise the redox-sensing behavior of Grx1-roGFP2 in vitro or in live cell experiments (Gutscher et al., unpublished results). Likewise, the monothiol mechanism has also been established for rxYFP-Grx1p (7). In the oxidative response, the nucleophilic cysteine C23 of Grx1 specifically reacts with GSSG to form a mixed Grx1-glutathione disulfide intermediate. The latter reacts with one of the two thiols on roGFP2, which, as a consequence, becomes S-glutathionylated. The existence of S-glutathionylated intermediates has been confirmed by mass spectrometry (Meyer et al., unpublished data), but it remains unclear if one of the two roGFP2 cysteines is preferentially S-glutathionylated by Grx. In the last step, glutathionylated roGFP2 re-arranges to form the internal disulfide bridge (C147–C204). Importantly, the three consecutive steps of thiol–disulfide exchange are fully reversible. When an oxidative event fades away, [GSSG] and thus E_GSH rapidly normalize, and the whole three-step thiol–disulfide cascade re-equilibrates (Fig. 14A). Grx1-roGFP2 is a truly dynamic probe because it is based on a completely symmetrical ping-pong mechanism (Fig. 14B).

OPEN IN VIEWER

FIG. 14.

Reversibility of the Grx1-roGFP2 response. (A) Live cell imaging of the intracellular Grx1–roGFP2 response to an oxidative challenge from outside the cell. H₂O₂ entering the cell is rapidly converted to GSSG (presumably by glutathione peroxidases), which then oxidizes the probe. Once H₂O₂ is depleted, GSSG levels drop (due to glutathione reductase) and the sensor re-equilibrates. (B) Molecular mechanism of the Grx1–roGFP2 biosensor. Each individual step of the three-step thiol-disulfide exchange cascade is fully reversible.

f. Implications for redox signaling

Interestingly, the dynamic behavior of Grx1-roGFP2 inside living cells may have implications beyond its use as a biosensor. If it is assumed that natural Grx1 target proteins behave similarly to roGFP2, which seems reasonable, they should also be dynamically oxidized (S-glutathionylated) and reduced (de-glutathionylated) in response to physiological changes in E_GSH (Fig. 15). Grx1-roGFP2 clearly demonstrates that glutathionylation can drive the formation of stable disulfide bonds, suggesting that Grx1 target proteins frequently convert S-glutathionylation into intra- or intermolecular disulfide bonds (if a second thiol is available nearby). Naturally, the extent of Grx1-catalyzed disulfide formation would depend on the respective dithiol/disulfide midpoint potential, which is certain to differ among target proteins. Although Grx1 is predominantly known for its reductive functions, protein S-glutathionylation is increasingly discussed as a possible pro-oxidative function of Grx1 (82). The dynamic behavior of Grx1-roGFP2 in living cells now clearly demonstrates that Grx1 catalyzes dynamically in both directions and that this happens readily within the physiological range of E_GSH . Accordingly, a large set of thiol proteins may be redox-regulated by Grx1-mediated real-time equilibration with the 2GSH/GSSG redox couple. How-ever, Grx1-roGFP2 also demonstrates that co-localization between Grx1 and target protein is very important, and it may be that only those Grx1 target proteins in close proximity to endogenous Grx1 are efficiently coupled to E_GSH . ASK1 and Procaspase-3 may be a case in point as these proteins seem to associate directly with Grx1 (96, 112).

OPEN IN VIEWER

FIG. 15.

Glutaredoxin as a general mediator of reversible protein thiol oxidation in H₂O₂ signaling. Grx may play a general role in reversible protein thiol oxidation. (A) Through the action of glutathione peroxidases, H₂O₂ transiently increases E_GSH . In response, oxidation-sensitive proteins co-localized with Grx become S-glutathionylated. Glutathionylation may in turn lead to disulfide bond formation. (B) Upon renormalization of E_GSH , GSH attacks the same disulfide bonds and Grx catalyzes de-glutathionylation of the same proteins.

Grx-mediated protein oxidation may also help to understand the phenomenon of H₂O₂-based signaling. An important question is how H₂O₂ can oxidize redox-regulated proteins efficiently and specifically, despite competition by highly efficient and abundant scavenger systems. It has been suggested that regulatory protein thiol oxidation by H₂O₂ may sometimes, if not frequently, be an indirect process (130). The Grx1-roGFP2 probe potentially exemplifies such an indirect pathway (Fig. 15) in that H₂O₂ facilitates protein thiol oxidation through the GSSG intermediate. In other words, ultrasensitive H₂O₂-scavenging enzymes (e.g., glutathione peroxidases in mammalian cells or ascorbate peroxidase together with dehydroascorbate reductase in plant cells) may act as highly efficient initiators of oxidative signaling by generating transient bursts of GSSG. The resulting pro-oxidative shift in E_GSH then triggers Grx-mediated protein oxidation in those oxidation-sensitive proteins that are in close physical association with appropriate Grxs, but not in others (Fig. 15A). Once E_GSH normalizes by the action of GRs, Grx catalyzes the reduction of the same proteins (Fig. 15B). It may be speculated that Grx1-roGFP2 dynamically responds to endogenous H₂O₂ signals precisely because it mimics a natural pathway of regulatory protein thiol oxidation.

1. The concept behind Trx1-roGFP2

The successful redox coupling of Grx1 and roGFP2 and the considerations regarding the increased local concentration and therefore probability of interaction (see Section III.A.2.a) suggested the possibility that a Trx-roGFP2 fusion protein may likewise serve as reversible probe for the Trx redox state. This possibility is conceivable for two reasons. First, Trx has been observed to become transiently oxidized during oxidative stress and redox signaling (45), thus suggesting that Trx is not kept in the reduced state at all times. Second, disulfide exchange between Trx and substrates is reversible in principle, and prokaryotic Trxs were previously shown to be able to act as oxidases (23, 113). Depending on the position of the equilibrium, Trx may thus oxidize roGFP by dithiol/disulfide exchange. In principle, this may allow visualization of transient Trx oxidation, potentially associated with oxidative signaling events. Assuming that specific and efficient equilibration between Trx1 and roGFP2 occurs in a fusion protein, the following Nernst relationship should apply: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6} \begin{document} \begin{align*} E_ { Trx1 } & =E_ { Trx1 } ^ { \circ \prime} - \frac { RT } { 2F } \ln \left( \frac { 1 - OxD_ { Trx1 } } { OxD_ { Trx1 } } \right) \\&=E_ { roGFP2 } ^ { \circ \prime} - \frac { RT } { 2F } \ln \left( \frac { 1 - OxD_ { roGFP2 } } { OxD_ { roGFP2 } } \right) = E_ { roGFP2 } \end{align*} \end{document}

The midpoint potential of the active site dithiol–disulfide pair of human Trx1 has been determined as −230 mV (128). The midpoint potential of roGFP2 is yet more negative (Fig. 16A), leading to the prediction that only a highly reduced Trx1 would keep roGFP2 in the reduced state. Any weak increase in Trx1 oxidation would translate into substantial roGFP2 oxidation. Specifically, a shift towards 2% Trx1 oxidation should entail 50% roGFP2 oxidation (Fig. 16B).

OPEN IN VIEWER

FIG. 16.

Theoretical thermodynamic relationship between the roGFP2 and Trx1 redox couples. (A) Relationship between OxD and E, shown for roGFP2 (blue curve), Trx1 (red curve), and, for comparison, GSH at 5 mM (green curve). (B) Minor increases in Trx1 oxidation are expected to entail major increases in roGFP2 oxidation.

2. The geometrical limits of thiol–disulfide exchange

All these theoretical predictions assume effective equilibration by thiol–disulfide exchange. However, it turned out that human Trx1 does not interact with roGFP2. Neither does fully reduced Trx (coupled to the complete Trx system) target oxidized roGFP, nor does the reverse reaction take place (40). Likewise, the same observation applies to rxYFP and Trx (93). Why does roGFP2 readily interact with Grx, but not with Trx? It is not well understood by which criteria Trx1 targets particular disulfide bonds (while ignoring others, even if well exposed). The stereochemical conformation of the roGFP2 disulfide bridge is rather unusual. Bonds of this type are known as ‘right-handed staples’ or ‘cross-stranded’ disulfides (55, 105, 132). Interestingly, this atypical geometry is shared by several disulfides known to be readily reduced by Trx [e.g., a disulfide in the CD4 cell surface receptor (73) and another in BASI (barley α-amylase/subtilisin inhibitor)] (69). On these grounds, the disulfide on roGFP2 would appear to be a likely target for Trx. It is well exposed on the barrel surface and the ‘staple’ geometry is compatible with Trx-mediated reduction in other proteins. The most likely explanation for the resistance of the roGFP2 disulfide bond against Trx is a structural constraint imposed by the reaction mechanism. In general, thiol–disulfide exchange reactions proceed by bimolecular nucleophilic substitution (S_N2). This implies that the three participating sulfur atoms (nucleophile, electrophile, and leaving group) have to be arranged in a straight line (Fig. 17). It is somewhat surprising that the implications of S_N2 geometry for protein disulfide exchange interactions are rarely discussed in the literature, with only few exceptions (28, 52, 129).

OPEN IN VIEWER

FIG. 17.

Bimolecular nucleophilic substitution (S_N2) requires an in-line arrangement between nucleophile, electrophile, and leaving group. The nucleophilic sulfur of Trx (C32γS) needs to be precisely positioned for the 180° in-line attack on the target disulfide (cross hairs). The geometry of this mechanism imposes restrictions for enzyme substrate interactions.

The S_N2 constraint likely explains why right-handed ‘staple’ disulfides are preferentially targeted by Trx in certain other proteins, but not in roGFP2 (or rxYFP). When a ‘staple’ disulfide bond is positioned on the edge of a β-sheet (as in BASI), the interface encountering the active site surface of Trx is relatively flat. The point in space that needs to be reached by the nucleophilic sulfur of C32 in order to mount an in-line attack is unobstructed by side-chains (Fig. 18). More specifically, it appears to be essential that the two residues flanking the target cysteine do not project their side chains into the direction of Trx. In fact, the β-sheet configuration effectively prevents steric interference of this kind. Any side chain (even alanine) projecting towards Trx would be predicted to get into steric conflict with W31 and P34 on Trx1, two highly conserved residues flanking the nucleophilic C32.

OPEN IN VIEWER

FIG. 18.

Molecular model for the encounter between Trx1 and a ‘staple’ disulfide bond located on a β-sheet edge. The geometry allows the electrophilic sulfur atom to reach the S_N2 attack point (cross hairs). The flat edge of the β-sheet allows positioning of the electrophilic sulfur atom between conserved residues W31 and P34 without steric interference by side chains. (A) Front view and (B) side view rotated by 90°. The figure is based on PDB entries 1ERT (thioredoxin) and 1AVA (staple disulfide bond with surrounding β-sheet).

These considerations may also explain why the situation is fundamentally different in the case of the roGFP2-Trx interaction: The roGFP2 disulfide is not located at the edge of the β-sheet and the sulfur–sulfur axis of the roGFP disulfide bond lies flat and parallel to the barrel surface. Therefore, most likely, the nucleophilic sulfur of Trx, located in a narrow groove between W31, P34, and other residues, is sterically prevented from positioning itself for the S_N2 reaction (Figs. 19 and 20A).

OPEN IN VIEWER

FIG. 19.

A likely explanation for the lack of disulfide exchange between roGFP2 and Trx1. Proper in-line alignment for S_N2 is prevented by massive steric conflict (A), while the angled approach is not compatible with the geometric necessities of a S_N2 reaction (B).

OPEN IN VIEWER

FIG. 20.

Differing sterical constraints for Trx and Grx. Most likely, thioredoxin ignores roGFP2 because its nucleophilic cysteine cannot take the S_N2 attack position (cross hairs) (A). In contrast, glutathione is small enough to position itself and engage in disulfide exchange with roGFP2. The disulfide bridge of the resulting glutathione–roGFP2 conjugate is more accessible and thus engages successfully with Grx (B).

The same considerations may also explain why the roGFP2 disulfide bond is a very good substrate for the Grx/GSH system (Fig. 20B). Most likely, the small glutathione molecule can easily reach the position required for in line attack on the roGFP2 disulfide bond. The resulting roGFP2–S-SG mixed disulfide bond is much more accessible, because it is not as close to the barrel surface. In addition, the sulfur-containing bonds of roGFP2–S-SG are rotatable, thus offering many alternative positions for subsequent nucleophilic attack. This may explain why there is no steric problem for Grx to attack roGFP2–S-SG. It may be asked if Trx would also be able to attack and reduce the roGFP2-S-SG conjugate with its more accessible disulfide bond. In vitro experiments indicate that this is not the case (Gutscher et al., unpublished data), suggesting that either roGFP–S-SG is still not accessible enough for Trx, or that Trx somehow discriminates against protein–S-SG mixed disulfides as substrates.

In conclusion, it seems unlikely that a working Trx-roGFP redox relay can be created. On the positive side, the fact that Trx refuses to interact with the roGFP2 intramolecular disulfide bond (and also with the roGFP2-S-SG disulfide bond) strengthens the concept that roGFP2 (and especially Grx1-roGFP2) really are highly specific probes for 2GSH/GSSG. Of course, it remains possible that certain Trxs from other organisms or organelles, possibly those with a shallower active site groove (i.e., with a more exposed nucleophile), can engage in thiol–disulfide exchange with roGFPs. However, even if a disulfide-exchanging Trx/roGFP combination can be found, the usefulness of the corresponding fusion protein to visualize the native endogenous Trx redox state would be quite uncertain. The bulky roGFP domain may sterically hinder the interaction of Trx with its native redox partners TrxR and the peroxiredoxins. In that case the probe may not properly reflect the native redox state of Trx. Moreover, the interpretation of the probe response may be further complicated as slow interactions with the endogenous GSH/Grx system may take place in parallel. It seems that fresh ideas are needed to finally develop a genetically encoded redox probe specific to the Trx redox state (see also Section V.A.2).

1. The concept behind roGFP2-Orp1

Is it possible to convert roGFPs into specific probes for H₂O₂ or other peroxides? Peroxiredoxins and glutathione peroxidases are highly efficient and selective H₂O₂ receptors. Intriguingly, some peroxidases have been found to transmit oxidation to other proteins. The best studied example is the regulatory oxidation of the transcription factor Yap1 by the peroxidase Orp1 (Gpx3) in yeast (24). This finding led to the question if Orp1 would accept roGFP in place of its natural target protein, Yap1. To test the idea, roGFP2 was fused to Orp1 to create a redox relay in which roGFP2 replaces Yap1 and the linker enforces proximity (41) (Fig. 21). In fact, Orp1 was found to form a highly efficient redox relay with roGFP2. In vitro, Orp1 mediated near-stoichiometric oxidation of roGFP2 by H₂O₂, converting almost every H₂O₂ molecule into a roGFP2 disulfide bridge. Moreover, the Orp1-roGFP2 redox couple effectively converts physiological H₂O₂ signals into measurable fluorescent signals in living cells (41).

OPEN IN VIEWER

FIG. 21.

Principle of the roGFP2–Orp1 redox relay. (A) RoGFP2-linked Orp1 efficiently catalyzes equilibration between H₂O₂ and roGFP2. (B) Structural model showing H₂O₂ and the roGFP2-Orp1 fusion protein in the reduced state. The redox state of roGFP2 is obtained by ratiometric fluorescence intensity (I) measurements at two different excitation wavelengths (λ₁, λ₂).

The intracellular mechanism of the roGFP2-Orp1 redox relay was found to be based on dithiol-disulfide exchange between the two domains (Fig. 22). Oxidation of Orp1 by H₂O₂ first generates an intramolecular disulfide bond between C36 and C82. In a second step, the disulfide bond is transferred to roGFP2 by thiol–disulfide exchange (41). The Orp1 C36/C82 dithiol/disulfide midpoint potential was measured as −255 (± 8) mV at pH 6.0 (72), suggesting a value of −314 mV at pH 7.0 by extrapolation (based on the −59 mV per pH unit dependence). Considering thiol–disulfide exchange between Orp1 and roGFP2, Orp1 should be rather inefficient in oxidizing roGFP2 (50% sensor oxidation would require approx. 90% Orp1 oxidation). However, in situations where Orp1 encounters H₂O₂ frequently, it will be kept at 100% oxidation by H₂O₂ and thus transfer oxidative equivalents to roGFP2 very efficiently. Thus, Orp1 can be seen as a redox catalyst mediating electron flow between the H₂O₂/H₂O pair (approx. + 1350 mV at pH 7) and the roGFP2_ox/roGFP2_red redox pair.

OPEN IN VIEWER

FIG. 22.

Molecular mechanism of the roGFP2–Orp1 biosensor. Upon encountering H₂O₂, the Orp1 domain forms a sulfenic acid residue which is rapidly converted into a disulfide bridge within the same domain. Subsequent thiol–disulfide exchange between the Orp1 and roGFP2 domains leads to formation of the roGFP2 disulfide bridge.

The roGFP2-Orp1 relay demonstrates dynamic behavior in living cells. Thus, it is important to ask how the probe is reduced inside cells. In principle, there are two possibilities. Similar to conventional roGFP2, the roGFP2 disulfide within the fusion protein may be reduced by GSH. This would depend on endogenous Grx and be rather slow. Based on the thermodynamic relationship between roGFP2 and Orp1, it appears more likely that roGFP2 passes the disulfide bond back to Orp1. This leads to the question of how Orp1 is reduced inside living cells. In yeast, Trx is the natural reductant of Orp1 and this may also be the case in mammalian cells. The −314 mV midpoint potential of Orp1 is compatible with its reduction by human Trx1 \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document}\begin{align*}(E_{Trx1}^{ \circ \prime} = - 230\ { \rm mV})\end{align*}\end{document} , if the Trx1 pool is highly reduced (OxD_Trx < 1%). At present, it is not known if roGFP2–Orp1 is exclusively reduced by Trx1, or, potentially, by both Trx1 and GSH/Grx1. In any case, as roGFP2–Orp1 reports on the thiol redox state of Orp1, its response is best seen as a reflection of the balance between H₂O₂-mediated thiol oxidation and (Trx- and/or Grx-mediated) thiol reduction.

2. Implications for H₂O₂ signaling

Similar to the potential role of Grx1 in regulatory protein oxidation (Section III.A.2.f.), the fact that roGFP2-Orp1 converts physiological concentrations of H₂O₂ into protein disulfide bonds may as well have implications beyond its use as a biosensor, namely for the mechanism of H₂O₂ signaling. Apparently, Orp1 is not mechanistically restricted to use Yap1 as the recipient of oxidative equivalents and thus harbors a more general ability to act as a thiol oxidase for other proteins (41). In turn, this may suggest that certain other peroxidases, peroxiredoxins, and/or members of the glutathione peroxidase family have a similar intrinsic capacity to promote the oxidation of other proteins. In fact, the human selenocysteine-based glutathione peroxidase Gpx4 was already implicated in the disulfide cross-linking of protamines and is also capable of mediating the oxidation of roGFP2 in living cells (41). Close pre-positioning between peroxidases and target proteins likely is critical for the transfer of oxidizing equivalents. Only protein thiols close to the peroxidase active site are likely to intercept the peroxidatic cycle. If roGFP2-Orp1 is indicative of a more general principle of H₂O₂ signaling, a major task for the future is to identify and study natural peroxidase-target relationships in higher eukaryotic cells.

1. The concept behind HyPer

HyPer has been designed to serve as a specific indicator for intracellular H₂O₂ (6). It is built on the H₂O₂-sensing regulatory domain (RD) of the prokaryotic transcription factor OxyR. The active site cysteine of OxyR-RD (C199) readily reacts with H₂O₂ to form a cysteine sulfenic acid (S-OH) which subsequently condenses with a second cysteine in the same domain (C208) to form an intramolecular disulfide bridge. In the reduced state, C199 and C208 are separated by 17 Å and a substantial conformational change, predominantly affecting peptide segment 205-222, takes place upon oxidation (20). To create HyPer, Belousov and colleagues inserted cpYFP between residues 205 and 206 of OxyR-RD. The cpYFP moiety was further optimized for folding and chromophore maturation, and the Y203F mutation was introduced to visualize the protonated form of the cpYFP chromophore. The resulting cpYFP–OxyR-RD chimera, named HyPer, displays a single emission maximum at 516 nm and two excitation maxima at 420 (protonated chromophore) and 500 nm (deprotonated chromophore). Importantly, the conformational influence of disulfide bond formation leads to a ratiometric shift with the 420 nm peak decreasing and the 500 nm peak increasing. HyPer was shown to detect nanomolar H₂O₂ in vitro and, when expressed in cells, to respond to micromolar H₂O₂ added externally. HyPer also allowed detection of oxidative changes induced by growth factors (6, 70). The maximal ratiometric dynamic range in vitro was reported as 3- to 4-fold. However, ratiometric changes in live cell imaging after physiological stimuli were more moderate (i.e., a 15% ratio increase after NGF stimulation), indicating that accurate measurements of physiologically relevant events can be technically demanding.

2. How specific is HyPer as a probe for H₂O₂?

How accurately does the HyPer signal reflect dynamic variations in endogenous H₂O₂ levels? On the one hand, there is little doubt that the OxyR regulatory domain is highly selective in its reactivity towards H₂O₂. On the other hand, the cpYFP moiety of HyPer does not sense H₂O₂ directly. Instead, the fluorescence response reflects the thiol redox state of OxyR-RD. The thiol redox state of OxyR-RD is the result of the dynamic interplay between disulfide formation and disulfide reduction. This leads to the question of how the HyPer disulfide bond is reduced within eukaryotic cells. In E. coli the disulfide form of OxyR is reduced by the glutathione system, specifically by GSH under catalysis of Grx1 (136). It is likely that reduction of HyPer in eukaryotic cells also depends on the glutathione system. Therefore, in principle, events curtailing the disulfide reducing function of the glutathione system (e.g., limiting NADPH levels or loss of GSH) may well shift the balance towards increased HyPer oxidation, even if H₂O₂ levels remain the same.

Regarding the role of the glutathione system, there is an additional concern: If Grx mediates reduction of the HyPer disulfide it may also catalyze the reverse reaction, namely disulfide formation, based on the Nernst relationship between the glutathione redox couple and the dithiol/disulfide couple of HyPer. Of note, in vitro, when purified OxyR is incubated with a combination of GSSG and Grx1 (rather than GSSG alone), it is, on a molar basis, as efficiently oxidized as with H₂O₂ (136). Is it conceivable that Grx-catalyzed oxidation of HyPer by GSSG occurs in intact cells? It is unlikely to be a major concern, at least for the cytosolic compartment. First, oxidation of OxyR-RD by GSSG becomes thermodynamically favorable only at relatively high GSSG concentrations. The midpoint potential of the OxyR dithiol/disulfide couple has been reported as −185 mV (136). Thus, to oxidize HyPer (OxyR) by 10%, an E_GSH of −214 mV is required (at GSH_total = 5 mM this would correspond to [GSSG] = 165 μM). As discussed before, measurements with rxYFP, roGFPs, and Grx1-roGFP2 strongly indicate that such a high E_GSH is unlikely to occur in the cytosol of living cells under physiological conditions. Second, it is not clear if intracellular Grx levels would be sufficient to support equilibration. Most likely, GSSG-mediated oxidation of HyPer is limited by the local availability of Grx and therefore a slow process. It appears safe to assume that under most circumstances H₂O₂ is the predominant oxidant for HyPer. It should be noted, however, that the midpoint potential of the 2GSH/GSSG redox couple depends on GSH_total. The lower the concentration of the total GSH pool, the more favored is HyPer (OxyR) oxidation by GSSG.

With regard to reversibility, HyPer differs fundamentally from Grx1-roGFP2. Grx1-roGFP2 measures the redox potential of a single redox pair by equilibrating with the same. In contrast, HyPer (i.e., OxyR-RD) communicates with two redox pairs simultaneously: 2H₂O/H₂O₂ and 2GSH/GSSG, thus displaying a fundamental asymmetry between the oxidation and reduction reactions. Obviously, roGFP2-Orp1 is subject to the same mechanistic asymmetry as HyPer. Thus, it is probably more appropriate to consider HyPer a probe for the balance between H₂O₂-mediated disulfide formation and glutathione-mediated disulfide reduction and, correspondingly, roGFP2-Orp1 a probe for the balance between H₂O₂-mediated disulfide formation and (probably Trx-mediated) disulfide reduction. This kind of probe behavior should not be considered a shortcoming, as measuring the thiol oxidation-reduction balance is highly relevant for the investigation of cellular redox signaling.

Finally, another concern that deserves attention is the influence of pH on HyPer fluorescence. From the pH titration curve (6) it appears that the 500/420 excitation ratio is significantly shifted by pH and that a shift of 0.2 pH units can be sufficient to change the ratio as much as full reduction or oxidation. Thus, for practical purposes it seems important to monitor pH changes in the same experiment.

3. Applications of HyPer

HyPer measurements have been successfully applied to study the wound response of zebra fish larvae (90). Cytoplasmic sensor expression was achieved in whole animals by injecting HyPer mRNA into embryos. Local injury of the tail fin produced a localized, graded, and transient HyPer signal. The interpretation of the observed HyPer response as a H₂O₂ signal was supported by the fact that pH changes could not be detected. A fluorogenic dye also indicated oxidant formation under the same conditions, and the signal was dependent on the expression of the NADPH oxidase Duox. Thus, it appears very likely that the observed HyPer response was directly caused by H₂O₂.

1. Improvements and new modalities

2. Novel probes for key redox couples

Redox-sensitive FPs provide a probe scaffold that is amenable to improvement, reconstruction, and expansion. Numerous opportunities for the creation of novel probes can be envisaged. In the following paragraphs we give just a few examples for the possible deployment of redox-sensitive FPs in the design of novel probes.

First of all, peroxidase-roGFP redox relays clearly offer opportunities for further probe development. The finding that the thiol-based peroxidase Orp1 efficiently mediates electron flow between H₂O₂ and roGFP2 (41) suggests that the same principle may also be applicable to other peroxidases. In fact, the selenocysteine-based glutathione peroxidase Gpx4 (also known as phospholipid hydroperoxidase, PHGPx) also mediates roGFP oxidation in a live cell context (41). It is now conceivable that a number of roGFP-peroxidase probes can be created to visualize the intracellular oxidative response of individual peroxidases. In this regard, a potentially attractive goal may be the design of a genetically encoded probe for lipid peroxidation.

Another important goal for the future is the development of the hitherto unavailable Trx-specific redox biosensor. As discussed before, existing roGFPs and rxYFP do not interact with Trxs (see Section III.B). Thus, a different strategy is needed to generate a Trx-sensitive probe. It may be possible to introduce novel disulfide switches into GFP, designed from the outset to be accessible to the Trx active site. It may also be possible to create a chimeric redox relay between a Trx target protein and a roGFP. Yet another possibility may be to insert a Trx-sensitive protein domain directly into the barrel of a redox-insensitive FP. A precedent for the latter strategy is the Ca^2 +-sensor Camgaroo, in which Ca^2 +-binding calmodulin was inserted into YFP at position 145 (3, 39).

A probe for the redox pair AA/DHA is also high on the wish list. After disproportionation of two molecules monodehydroascorbate (MDHA) to AA and DHA the reduction of DHA to AA is catalyzed by dehydroascorbate reductase (DHAR) and leads to a mixed glutathione-enzyme disulfide (26, 110). Normally this mixed disulfide would be reduced by GSH to form GSSG, but in analogy to the catalytic mechanism of Grx1–roGFP2 it may also be transferable to roGFP2 where it may trigger disulfide formation and a corresponding change in fluorescence. It remains to be seen if a specific probe for AA/DHA can be build upon these or similar principles.

Another highly desired probe is one for the real-time observation of NO^•. NO^• can undergo nitrosylation reactions with protein and low molecular weight thiols. It is conceivable that a trans-nitrosylation reaction targeted to roGFP could trigger formation of the roGFP disulfide bridge. A fusion between roGFP and a primary NO^• acceptor protein may thus allow NO^• imaging. Recently, proteomic profiling of S-nitrosylation identified proteins with unexpected stability of the nitrosothiol group (95). The authors suggest that stable S-nitrosylation reflects a protein conformational change that shields the nitrosothiol group. Whether such NO^•-dependent conformational changes can be exploited for the generation of future NO^• probes remains to be tested.

3. Redox biosensor transgenic model organisms

Arguably, one of the ultimate methodological goals of redox biology is the ability to study clearly defined redox processes as they occur in the intact physiological context of the living organism. Genetically encoded redox probes now offer an opportunity to make progress into this direction. Much initial work on the characterization and application of roGFP-based probes has been done in transgenic Arabidopsis plants (57, 71, 81, 107). For the near future, it can be expected that the yeast S. cerevisiae will also play a major role in this context. Although rxYFP has been initially characterized in yeast (92, 93), the ratiometric behavior of roGFP provides a principle advantage over nonratiometric rxYFP and future work in yeast will likely build on roGFP variants. For example, yeast cells expressing roGFP-based probes may be genetically screened for deviations from steady state redox state and thus help identifying novel regulators of redox homeostasis. Similarly, a variety of genetically tractable model organisms expressing roGFP-based probes or HyPer are currently being created and studied to address diverse biological questions. To our knowledge, the list of transgenic species currently generated in various laboratories includes mice, zebra fish, C. elegans, D. melanogaster, barley, tobacco, trypanosomes, Chlamydomonas rheinhardtii, and Synechocystis. It remains to be seen what can be achieved with these transgenic models. In this regard, the expression and application of HyPer in zebra fish larvae is a promising first step (90). To unlock the potential of redox imaging in non-transparent organisms, various technical challenges need to be overcome. For example, the potential use of roGFP-based probes in two-photon imaging will have to be investigated. Another potential direction of technical investigation is the in situ chemical conservation of the biosensor redox state and its analysis in tissue sections. An attractive long-term perspective for redox biosensor transgenic mouse models is the comparison between healthy and diseased states and the testing of drugs for their actual in vivo influence on defined redox states in defined target tissues.

1. Exemplary applications in cell biology: Glutathione compartmentalization and transport

Glutathione is required in most subcellular compartments of eukaryotic cells. Despite the ubiquitous need for GSH, its biosynthesis is restricted to either the cytosol in mammalian cells or the plastids and the cytosol in plant cells, which highlights the need for membrane transport of GSH into other organelles (Fig. 23). In Arabidopsis, the rate-limiting enzyme GSH1 is exclusively targeted to the plastids while the second enzyme, glutathione synthase (GSH2), is targeted to both plastids and cytosol. This subcellular distribution of the two GSH biosynthetic enzymes implies the need for export of γ-glutamylcysteine (γ-EC) from plastids to supply cytosolic GSH2 with its substrate (97, 123) (Fig. 23A). First candidate proteins for mediating this transport of γ-EC and/or GSH from plastids to the cytosol were identified recently in Arabidopsis. Based on high sequence homology to the chloroquine-resistance transporter from Plasmodium falciparum, PfCRT, these proteins were named CRT-like transporters (CLTs) (75) (Fig. 23A). Lack of all three CLT isoforms, CLT1, CLT2, and CLT3, caused a slight decrease in total GSH content in shoot tissue, but a more pronounced effect in subcellular compartmentation of GSH was resolved with roGFP2 in situ. While plastid-targeted roGFP2 in both wild-type and the triple knockout mutant was almost fully reduced, roGFP2 expressed in the cytosol showed a distinct oxidation in the triple mutant. In contrast to a cytosolic E_GSH of about −315 mV in wild-type plants, the cytosolic E_GSH in the triple mutant was about 20 mV less reducing. This result is similar to the change in E_GSH observed for the partially GSH-deficient cad2 mutant (Fig. 5). In general, targeting of roGFP to organelles like mitochondria and peroxisomes may provide an excellent opportunity to test candidate proteins for their ability to transport GSH in situ and to better understand the dynamics of glutathione compartmentation within eukaryotic cells.

OPEN IN VIEWER

FIG. 23.

Biosynthesis and distribution of GSH within eukaryotic cells. (A) Subcellular distribution of GSH in Arabidopsis thaliana. (B) GSH compartmentation in mammalian cells.

2. High throughput screening

Biosensors on the basis of fluorescent proteins have a huge potential for high throughput screening (HTS) to identify novel drugs affecting specific physiological processes. The use of genetically encoded probes avoids dye loading and washing steps and leads to more robust assays. An example of successful HTS based on GFP-derived sensors is the use of the Cl^--sensitive YFP-H148Q to screen for new classes of potent CFTR activators of the cystic fibrosis transmembrane conductance regulator (CFTR) (68). Rapid quantitative screening of a 60,000 compound library on cells grown on 96-well plates led to the identification of a large number of strong and weak activators. Recently, similar HTS to identify proteins that correct the F508del-CFTR defect was done with YFP–H148Q/I152L (116).

Similar screening approaches can be envisaged with roGFP-based probes to identify novel genes involved in cellular redox homeostasis and drugs affecting defined redox processes. Provided that roGFP variants with appropriate midpoint potentials are available, such screens can be conducted for both reductive and pro-oxidative processes. Although a ratiometric analysis for roGFP is technically more demanding than a single wavelength screen like the one described for Cl^--sensitive YFP, it seems feasible to do such screens on a plate reader, automated microscopes or by FACS. For identifying fast and transient effects on E_GSH in a time frame where correction for protein concentration is less critical, it may also be possible to use only a single wavelength to monitor changes in roGFP2 oxidation (121).